Index: user/ngie/more-tests/contrib/tcpdump/interface.h =================================================================== --- user/ngie/more-tests/contrib/tcpdump/interface.h (revision 281974) +++ user/ngie/more-tests/contrib/tcpdump/interface.h (revision 281975) @@ -1,220 +1,218 @@ /* * Copyright (c) 1988-2002 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that: (1) source code distributions * retain the above copyright notice and this paragraph in its entirety, (2) * distributions including binary code include the above copyright notice and * this paragraph in its entirety in the documentation or other materials * provided with the distribution, and (3) all advertising materials mentioning * features or use of this software display the following acknowledgement: * ``This product includes software developed by the University of California, * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of * the University nor the names of its contributors may be used to endorse * or promote products derived from this software without specific prior * written permission. * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. */ #ifndef tcpdump_interface_h #define tcpdump_interface_h #ifdef HAVE_OS_PROTO_H #include "os-proto.h" #endif /* snprintf et al */ #include #if HAVE_STDINT_H #include #endif #if !defined(HAVE_SNPRINTF) int snprintf(char *, size_t, const char *, ...) #ifdef __ATTRIBUTE___FORMAT_OK __attribute__((format(printf, 3, 4))) #endif /* __ATTRIBUTE___FORMAT_OK */ ; #endif /* !defined(HAVE_SNPRINTF) */ #if !defined(HAVE_VSNPRINTF) int vsnprintf(char *, size_t, const char *, va_list) #ifdef __ATTRIBUTE___FORMAT_OK __attribute__((format(printf, 3, 0))) #endif /* __ATTRIBUTE___FORMAT_OK */ ; #endif /* !defined(HAVE_VSNPRINTF) */ #ifndef HAVE_STRLCAT extern size_t strlcat(char *, const char *, size_t); #endif #ifndef HAVE_STRLCPY extern size_t strlcpy(char *, const char *, size_t); #endif #ifndef HAVE_STRDUP extern char *strdup(const char *); #endif #ifndef HAVE_STRSEP extern char *strsep(char **, const char *); #endif #define PT_VAT 1 /* Visual Audio Tool */ #define PT_WB 2 /* distributed White Board */ #define PT_RPC 3 /* Remote Procedure Call */ #define PT_RTP 4 /* Real-Time Applications protocol */ #define PT_RTCP 5 /* Real-Time Applications control protocol */ #define PT_SNMP 6 /* Simple Network Management Protocol */ #define PT_CNFP 7 /* Cisco NetFlow protocol */ #define PT_TFTP 8 /* trivial file transfer protocol */ #define PT_AODV 9 /* Ad-hoc On-demand Distance Vector Protocol */ #define PT_CARP 10 /* Common Address Redundancy Protocol */ #define PT_RADIUS 11 /* RADIUS authentication Protocol */ #define PT_ZMTP1 12 /* ZeroMQ Message Transport Protocol 1.0 */ #define PT_VXLAN 13 /* Virtual eXtensible Local Area Network */ #define PT_PGM 14 /* [UDP-encapsulated] Pragmatic General Multicast */ #define PT_PGM_ZMTP1 15 /* ZMTP/1.0 inside PGM (native or UDP-encapsulated) */ #define PT_LMP 16 /* Link Management Protocol */ #define ESRC(ep) ((ep)->ether_shost) #define EDST(ep) ((ep)->ether_dhost) #ifndef NTOHL #define NTOHL(x) (x) = ntohl(x) #define NTOHS(x) (x) = ntohs(x) #define HTONL(x) (x) = htonl(x) #define HTONS(x) (x) = htons(x) #endif #endif extern char *program_name; /* used to generate self-identifying messages */ extern int32_t thiszone; /* seconds offset from gmt to local time */ /* * True if "l" bytes of "var" were captured. * * The "snapend - (l) <= snapend" checks to make sure "l" isn't so large * that "snapend - (l)" underflows. * * The check is for <= rather than < because "l" might be 0. */ #define TTEST2(var, l) (snapend - (l) <= snapend && \ (const u_char *)&(var) <= snapend - (l)) /* True if "var" was captured */ #define TTEST(var) TTEST2(var, sizeof(var)) /* Bail if "l" bytes of "var" were not captured */ #define TCHECK2(var, l) if (!TTEST2(var, l)) goto trunc /* Bail if "var" was not captured */ #define TCHECK(var) TCHECK2(var, sizeof(var)) extern int mask2plen(uint32_t); extern const char *tok2strary_internal(const char **, int, const char *, int); #define tok2strary(a,f,i) tok2strary_internal(a, sizeof(a)/sizeof(a[0]),f,i) extern void error(const char *, ...) __attribute__((noreturn)) #ifdef __ATTRIBUTE___FORMAT_OK __attribute__((format (printf, 1, 2))) #endif /* __ATTRIBUTE___FORMAT_OK */ ; extern void warning(const char *, ...) #ifdef __ATTRIBUTE___FORMAT_OK __attribute__((format (printf, 1, 2))) #endif /* __ATTRIBUTE___FORMAT_OK */ ; extern char *read_infile(char *); extern char *copy_argv(char **); extern const char *dnname_string(u_short); extern const char *dnnum_string(u_short); /* checksum routines */ extern void init_checksum(void); extern uint16_t verify_crc10_cksum(uint16_t, const u_char *, int); extern uint16_t create_osi_cksum(const uint8_t *, int, int); /* The printer routines. */ #include -extern void pfsync_ip_print(const u_char *, u_int); - extern char *q922_string(const u_char *); extern char *smb_errstr(int, int); extern const char *nt_errstr(uint32_t); #ifdef INET6 extern int mask62plen(const u_char *); #endif /*INET6*/ struct cksum_vec { const uint8_t *ptr; int len; }; extern uint16_t in_cksum(const struct cksum_vec *, int); extern uint16_t in_cksum_shouldbe(uint16_t, uint16_t); #ifndef HAVE_BPF_DUMP struct bpf_program; extern void bpf_dump(const struct bpf_program *, int); #endif #include "netdissect.h" /* forward compatibility */ #ifndef NETDISSECT_REWORKED extern netdissect_options *gndo; #define bflag gndo->ndo_bflag #define eflag gndo->ndo_eflag #define fflag gndo->ndo_fflag #define jflag gndo->ndo_jflag #define Kflag gndo->ndo_Kflag #define nflag gndo->ndo_nflag #define Nflag gndo->ndo_Nflag #define Oflag gndo->ndo_Oflag #define pflag gndo->ndo_pflag #define qflag gndo->ndo_qflag #define Rflag gndo->ndo_Rflag #define sflag gndo->ndo_sflag #define Sflag gndo->ndo_Sflag #define tflag gndo->ndo_tflag #define Uflag gndo->ndo_Uflag #define uflag gndo->ndo_uflag #define vflag gndo->ndo_vflag #define xflag gndo->ndo_xflag #define Xflag gndo->ndo_Xflag #define Cflag gndo->ndo_Cflag #define Gflag gndo->ndo_Gflag #define Aflag gndo->ndo_Aflag #define Bflag gndo->ndo_Bflag #define Iflag gndo->ndo_Iflag #define suppress_default_print gndo->ndo_suppress_default_print #define packettype gndo->ndo_packettype #define sigsecret gndo->ndo_sigsecret #define Wflag gndo->ndo_Wflag #define WflagChars gndo->ndo_WflagChars #define Cflag_count gndo->ndo_Cflag_count #define Gflag_count gndo->ndo_Gflag_count #define Gflag_time gndo->ndo_Gflag_time #define Hflag gndo->ndo_Hflag #define snaplen gndo->ndo_snaplen #define snapend gndo->ndo_snapend extern void default_print(const u_char *, u_int); #endif Index: user/ngie/more-tests/contrib/tcpdump/netdissect.h =================================================================== --- user/ngie/more-tests/contrib/tcpdump/netdissect.h (revision 281974) +++ user/ngie/more-tests/contrib/tcpdump/netdissect.h (revision 281975) @@ -1,599 +1,601 @@ /* * Copyright (c) 1988-1997 * The Regents of the University of California. All rights reserved. * * Copyright (c) 1998-2012 Michael Richardson * The TCPDUMP project * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that: (1) source code distributions * retain the above copyright notice and this paragraph in its entirety, (2) * distributions including binary code include the above copyright notice and * this paragraph in its entirety in the documentation or other materials * provided with the distribution, and (3) all advertising materials mentioning * features or use of this software display the following acknowledgement: * ``This product includes software developed by the University of California, * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of * the University nor the names of its contributors may be used to endorse * or promote products derived from this software without specific prior * written permission. * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. */ #ifndef netdissect_h #define netdissect_h #ifdef HAVE_OS_PROTO_H #include "os-proto.h" #endif #include #ifndef HAVE___ATTRIBUTE__ #define __attribute__(x) #endif /* snprintf et al */ #include #include "ip.h" /* struct ip for nextproto4_cksum() */ #if !defined(HAVE_SNPRINTF) int snprintf (char *str, size_t sz, const char *format, ...) #ifdef __ATTRIBUTE___FORMAT_OK __attribute__((format (printf, 3, 4))) #endif /* __ATTRIBUTE___FORMAT_OK */ ; #endif /* !defined(HAVE_SNPRINTF) */ #if !defined(HAVE_VSNPRINTF) int vsnprintf (char *str, size_t sz, const char *format, va_list ap) #ifdef __ATTRIBUTE___FORMAT_OK __attribute__((format (printf, 3, 0))) #endif /* __ATTRIBUTE___FORMAT_OK */ ; #endif /* !defined(HAVE_SNPRINTF) */ #ifndef HAVE_STRLCAT extern size_t strlcat (char *, const char *, size_t); #endif #ifndef HAVE_STRLCPY extern size_t strlcpy (char *, const char *, size_t); #endif #ifndef HAVE_STRDUP extern char *strdup (const char *str); #endif #ifndef HAVE_STRSEP extern char *strsep(char **, const char *); #endif struct tok { u_int v; /* value */ const char *s; /* string */ }; #define TOKBUFSIZE 128 extern const char *tok2strbuf(const struct tok *, const char *, u_int, char *buf, size_t bufsize); /* tok2str is deprecated */ extern const char *tok2str(const struct tok *, const char *, int); extern char *bittok2str(const struct tok *, const char *, int); extern char *bittok2str_nosep(const struct tok *, const char *, int); typedef struct netdissect_options netdissect_options; struct netdissect_options { int ndo_aflag; /* translate network and broadcast addresses */ int ndo_bflag; /* print 4 byte ASes in ASDOT notation */ int ndo_eflag; /* print ethernet header */ int ndo_fflag; /* don't translate "foreign" IP address */ int ndo_Kflag; /* don't check TCP checksums */ int ndo_nflag; /* leave addresses as numbers */ int ndo_Nflag; /* remove domains from printed host names */ int ndo_qflag; /* quick (shorter) output */ int ndo_Rflag; /* print sequence # field in AH/ESP*/ int ndo_sflag; /* use the libsmi to translate OIDs */ int ndo_Sflag; /* print raw TCP sequence numbers */ int ndo_tflag; /* print packet arrival time */ int ndo_Uflag; /* "unbuffered" output of dump files */ int ndo_uflag; /* Print undecoded NFS handles */ int ndo_vflag; /* verbose */ int ndo_xflag; /* print packet in hex */ int ndo_Xflag; /* print packet in hex/ascii */ int ndo_Aflag; /* print packet only in ascii observing TAB, * LF, CR and SPACE as graphical chars */ int ndo_Bflag; /* buffer size */ int ndo_Iflag; /* rfmon (monitor) mode */ int ndo_Oflag; /* run filter code optimizer */ int ndo_dlt; /* if != -1, ask libpcap for the DLT it names*/ int ndo_jflag; /* packet time stamp source */ int ndo_pflag; /* don't go promiscuous */ int ndo_Cflag; /* rotate dump files after this many bytes */ int ndo_Cflag_count; /* Keep track of which file number we're writing */ int ndo_Gflag; /* rotate dump files after this many seconds */ int ndo_Gflag_count; /* number of files created with Gflag rotation */ time_t ndo_Gflag_time; /* The last time_t the dump file was rotated. */ int ndo_Wflag; /* recycle output files after this number of files */ int ndo_WflagChars; int ndo_Hflag; /* dissect 802.11s draft mesh standard */ int ndo_packet_number; /* print a packet number in the beginning of line */ int ndo_suppress_default_print; /* don't use default_print() for unknown packet types */ int ndo_tstamp_precision; /* requested time stamp precision */ const char *ndo_dltname; char *ndo_espsecret; struct sa_list *ndo_sa_list_head; /* used by print-esp.c */ struct sa_list *ndo_sa_default; char *ndo_sigsecret; /* Signature verification secret key */ struct esp_algorithm *ndo_espsecret_xform; /* cache of decoded */ char *ndo_espsecret_key; int ndo_packettype; /* as specified by -T */ char *ndo_program_name; /*used to generate self-identifying messages */ int32_t ndo_thiszone; /* seconds offset from gmt to local time */ int ndo_snaplen; /*global pointers to beginning and end of current packet (during printing) */ const u_char *ndo_packetp; const u_char *ndo_snapend; /* bookkeeping for ^T output */ int ndo_infodelay; /* pointer to void function to output stuff */ void (*ndo_default_print)(netdissect_options *, register const u_char *bp, register u_int length); void (*ndo_info)(netdissect_options *, int verbose); int (*ndo_printf)(netdissect_options *, const char *fmt, ...) #ifdef __ATTRIBUTE___FORMAT_OK_FOR_FUNCTION_POINTERS __attribute__ ((format (printf, 2, 3))) #endif ; void (*ndo_error)(netdissect_options *, const char *fmt, ...) #ifdef __ATTRIBUTE___NORETURN_OK_FOR_FUNCTION_POINTERS __attribute__ ((noreturn)) #endif /* __ATTRIBUTE___NORETURN_OK_FOR_FUNCTION_POINTERS */ #ifdef __ATTRIBUTE___FORMAT_OK_FOR_FUNCTION_POINTERS __attribute__ ((format (printf, 2, 3))) #endif /* __ATTRIBUTE___FORMAT_OK_FOR_FUNCTION_POINTERS */ ; void (*ndo_warning)(netdissect_options *, const char *fmt, ...) #ifdef __ATTRIBUTE___FORMAT_OK_FOR_FUNCTION_POINTERS __attribute__ ((format (printf, 2, 3))) #endif ; }; #define PT_VAT 1 /* Visual Audio Tool */ #define PT_WB 2 /* distributed White Board */ #define PT_RPC 3 /* Remote Procedure Call */ #define PT_RTP 4 /* Real-Time Applications protocol */ #define PT_RTCP 5 /* Real-Time Applications control protocol */ #define PT_SNMP 6 /* Simple Network Management Protocol */ #define PT_CNFP 7 /* Cisco NetFlow protocol */ #define PT_TFTP 8 /* trivial file transfer protocol */ #define PT_AODV 9 /* Ad-hoc On-demand Distance Vector Protocol */ #define PT_CARP 10 /* Common Address Redundancy Protocol */ #define PT_RADIUS 11 /* RADIUS authentication Protocol */ #define PT_ZMTP1 12 /* ZeroMQ Message Transport Protocol 1.0 */ #define PT_VXLAN 13 /* Virtual eXtensible Local Area Network */ #define PT_PGM 14 /* [UDP-encapsulated] Pragmatic General Multicast */ #define PT_PGM_ZMTP1 15 /* ZMTP/1.0 inside PGM (native or UDP-encapsulated) */ #define PT_LMP 16 /* Link Management Protocol */ #ifndef min #define min(a,b) ((a)>(b)?(b):(a)) #endif #ifndef max #define max(a,b) ((b)>(a)?(b):(a)) #endif /* * Maximum snapshot length. This should be enough to capture the full * packet on most network interfaces. * * * Somewhat arbitrary, but chosen to be: * * 1) big enough for maximum-size Linux loopback packets (65549) * and some USB packets captured with USBPcap: * * http://desowin.org/usbpcap/ * * (> 131072, < 262144) * * and * * 2) small enough not to cause attempts to allocate huge amounts of * memory; some applications might use the snapshot length in a * savefile header to control the size of the buffer they allocate, * so a size of, say, 2^31-1 might not work well. * * XXX - does it need to be bigger still? */ #define MAXIMUM_SNAPLEN 262144 /* * The default snapshot length is the maximum. */ #define DEFAULT_SNAPLEN MAXIMUM_SNAPLEN #define ESRC(ep) ((ep)->ether_shost) #define EDST(ep) ((ep)->ether_dhost) #ifndef NTOHL #define NTOHL(x) (x) = ntohl(x) #define NTOHS(x) (x) = ntohs(x) #define HTONL(x) (x) = htonl(x) #define HTONS(x) (x) = htons(x) #endif /* * True if "l" bytes of "var" were captured. * * The "ndo->ndo_snapend - (l) <= ndo->ndo_snapend" checks to make sure * "l" isn't so large that "ndo->ndo_snapend - (l)" underflows. * * The check is for <= rather than < because "l" might be 0. */ #define ND_TTEST2(var, l) (ndo->ndo_snapend - (l) <= ndo->ndo_snapend && \ (const u_char *)&(var) <= ndo->ndo_snapend - (l)) /* True if "var" was captured */ #define ND_TTEST(var) ND_TTEST2(var, sizeof(var)) /* Bail if "l" bytes of "var" were not captured */ #define ND_TCHECK2(var, l) if (!ND_TTEST2(var, l)) goto trunc /* Bail if "var" was not captured */ #define ND_TCHECK(var) ND_TCHECK2(var, sizeof(var)) #define ND_PRINT(STUFF) (*ndo->ndo_printf)STUFF #define ND_DEFAULTPRINT(ap, length) (*ndo->ndo_default_print)(ndo, ap, length) extern void ts_print(netdissect_options *, const struct timeval *); extern void relts_print(netdissect_options *, int); extern int fn_print(netdissect_options *, const u_char *, const u_char *); extern int fn_printn(netdissect_options *, const u_char *, u_int, const u_char *); extern int fn_printzp(netdissect_options *, const u_char *, u_int, const u_char *); extern const char *tok2str(const struct tok *, const char *, int); #if 0 extern char *read_infile(netdissect_options *, char *); extern char *copy_argv(netdissect_options *, char **); #endif /* * Locale-independent macros for testing character properties and * stripping the 8th bit from characters. Assumed to be handed * a value between 0 and 255, i.e. don't hand them a char, as * those might be in the range -128 to 127. */ #define ND_ISASCII(c) (!((c) & 0x80)) /* value is an ASCII code point */ #define ND_ISPRINT(c) ((c) >= 0x20 && (c) <= 0x7E) #define ND_ISGRAPH(c) ((c) > 0x20 && (c) <= 0x7E) #define ND_TOASCII(c) ((c) & 0x7F) extern void safeputchar(netdissect_options *, const u_char); extern void safeputs(netdissect_options *, const u_char *, const u_int); #ifdef LBL_ALIGN /* * The processor doesn't natively handle unaligned loads, * and the compiler might "helpfully" optimize memcpy() * and memcmp(), when handed pointers that would normally * be properly aligned, into sequences that assume proper * alignment. * * Do copies and compares of possibly-unaligned data by * calling routines that wrap memcpy() and memcmp(), to * prevent that optimization. */ extern void unaligned_memcpy(void *, const void *, size_t); extern int unaligned_memcmp(const void *, const void *, size_t); #define UNALIGNED_MEMCPY(p, q, l) unaligned_memcpy((p), (q), (l)) #define UNALIGNED_MEMCMP(p, q, l) unaligned_memcmp((p), (q), (l)) #else /* * The procesor natively handles unaligned loads, so just use memcpy() * and memcmp(), to enable those optimizations. */ #define UNALIGNED_MEMCPY(p, q, l) memcpy((p), (q), (l)) #define UNALIGNED_MEMCMP(p, q, l) memcmp((p), (q), (l)) #endif #define PLURAL_SUFFIX(n) \ (((n) != 1) ? "s" : "") #if 0 extern const char *dnname_string(netdissect_options *, u_short); extern const char *dnnum_string(netdissect_options *, u_short); #endif /* The printer routines. */ #include typedef u_int (*if_ndo_printer)(struct netdissect_options *ndo, const struct pcap_pkthdr *, const u_char *); typedef u_int (*if_printer)(const struct pcap_pkthdr *, const u_char *); extern if_ndo_printer lookup_ndo_printer(int); extern if_printer lookup_printer(int); extern void eap_print(netdissect_options *,const u_char *, u_int); extern int esp_print(netdissect_options *, const u_char *bp, const int length, const u_char *bp2, int *nhdr, int *padlen); extern void arp_print(netdissect_options *,const u_char *, u_int, u_int); extern void tipc_print(netdissect_options *, const u_char *, u_int, u_int); extern void msnlb_print(netdissect_options *, const u_char *); extern void icmp6_print(netdissect_options *ndo, const u_char *, u_int, const u_char *, int); extern void isakmp_print(netdissect_options *,const u_char *, u_int, const u_char *); extern void isakmp_rfc3948_print(netdissect_options *,const u_char *, u_int, const u_char *); extern void ip_print(netdissect_options *,const u_char *, u_int); extern void ip_print_inner(netdissect_options *ndo, const u_char *bp, u_int length, u_int nh, const u_char *bp2); extern void rrcp_print(netdissect_options *,const u_char *, u_int); extern void loopback_print(netdissect_options *, const u_char *, const u_int); extern void carp_print(netdissect_options *, const u_char *, u_int, int); extern void ether_print(netdissect_options *, const u_char *, u_int, u_int, void (*)(netdissect_options *, const u_char *), const u_char *); extern u_int ether_if_print(netdissect_options *, const struct pcap_pkthdr *,const u_char *); extern u_int netanalyzer_if_print(netdissect_options *, const struct pcap_pkthdr *,const u_char *); extern u_int netanalyzer_transparent_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern int ethertype_print(netdissect_options *,u_short, const u_char *, u_int, u_int); extern int print_unknown_data(netdissect_options *,const u_char *, const char *,int); extern void ascii_print(netdissect_options *, const u_char *, u_int); extern void hex_print_with_offset(netdissect_options *, const char *ident, const u_char *cp, u_int, u_int); extern void hex_print(netdissect_options *,const char *ident, const u_char *cp,u_int); extern void hex_and_ascii_print_with_offset(netdissect_options *, const char *, const u_char *, u_int, u_int); extern void hex_and_ascii_print(netdissect_options *, const char *, const u_char *, u_int); extern int ah_print(netdissect_options *, register const u_char *); extern void beep_print(netdissect_options *, const u_char *, u_int); extern void dtp_print(netdissect_options *, const u_char *, u_int); extern u_int cip_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern int ipcomp_print(netdissect_options *, register const u_char *, int *); extern u_int ipfc_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void udld_print(netdissect_options *, const u_char *, u_int); extern void hsrp_print(netdissect_options *, const u_char *, u_int); extern void igrp_print(netdissect_options *, const u_char *, u_int); extern void msdp_print(netdissect_options *, const u_char *, u_int); extern u_int null_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void mobile_print(netdissect_options *, const u_char *, u_int); extern u_int ap1394_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int bt_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void lane_print(netdissect_options *, const u_char *, u_int, u_int); extern u_int lane_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void otv_print(netdissect_options *, const u_char *, u_int); extern void ahcp_print(netdissect_options *, const u_char *, const u_int); extern void vxlan_print(netdissect_options *, const u_char *, u_int); extern u_int arcnet_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int arcnet_linux_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void bfd_print(netdissect_options *, const u_char *, u_int, u_int); extern void gre_print(netdissect_options *, const u_char *, u_int); extern int vjc_print(netdissect_options *, register const char *, u_short); extern void ipN_print(netdissect_options *, const u_char *, u_int); extern u_int raw_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int usb_linux_48_byte_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int usb_linux_64_byte_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int symantec_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int chdlc_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int chdlc_print(netdissect_options *, register const u_char *, u_int); extern void zmtp1_print(netdissect_options *, const u_char *, u_int); extern void zmtp1_print_datagram(netdissect_options *, const u_char *, const u_int); extern void ipx_print(netdissect_options *, const u_char *, u_int); extern void mpls_print(netdissect_options *, const u_char *, u_int); extern u_int pppoe_print(netdissect_options *, const u_char *, u_int); extern u_int pppoe_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void sunrpcrequest_print(netdissect_options *, const u_char *, u_int, const u_char *); extern u_int pflog_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int token_print(netdissect_options *, const u_char *, u_int, u_int); extern u_int token_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void vqp_print(netdissect_options *, register const u_char *, register u_int); extern void zephyr_print(netdissect_options *, const u_char *, int); extern void fddi_print(netdissect_options *, const u_char *, u_int, u_int); extern u_int fddi_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void mpcp_print(netdissect_options *, const u_char *, u_int); extern void rpki_rtr_print(netdissect_options *, const u_char *, u_int); extern u_int sll_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void dccp_print(netdissect_options *, const u_char *, const u_char *, u_int); extern int llc_print(netdissect_options *, const u_char *, u_int, u_int, const u_char *, const u_char *, u_short *); extern int snap_print(netdissect_options *, const u_char *, u_int, u_int, u_int); extern void eigrp_print(netdissect_options *, const u_char *, u_int); extern void stp_print(netdissect_options *, const u_char *, u_int); extern void l2tp_print(netdissect_options *, const u_char *, u_int); extern void udp_print(netdissect_options *, const u_char *, u_int, const u_char *, int); extern void icmp_print(netdissect_options *, const u_char *, u_int, const u_char *, int); extern void openflow_print(netdissect_options *, const u_char *, const u_int); extern void telnet_print(netdissect_options *, const u_char *, u_int); extern void slow_print(netdissect_options *, const u_char *, u_int); extern void radius_print(netdissect_options *, const u_char *, u_int); extern void lmp_print(netdissect_options *, const u_char *, u_int); extern u_int fr_print(netdissect_options *, register const u_char *, u_int); extern u_int mfr_print(netdissect_options *, register const u_char *, u_int); extern u_int fr_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int mfr_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void q933_print(netdissect_options *, const u_char *, u_int); extern void igmp_print(netdissect_options *, const u_char *, u_int); extern void rip_print(netdissect_options *, const u_char *, u_int); extern void lwapp_control_print(netdissect_options *, const u_char *, u_int, int); extern void lwapp_data_print(netdissect_options *, const u_char *, u_int); extern void pgm_print(netdissect_options *, const u_char *, u_int, const u_char *); extern void pptp_print(netdissect_options *, const u_char *); extern void ldp_print(netdissect_options *, const u_char *, u_int); extern void wb_print(netdissect_options *, const void *, u_int); extern int oam_print(netdissect_options *, const u_char *, u_int, u_int); extern void atm_print(netdissect_options *, u_int, u_int, u_int, const u_char *, u_int, u_int); extern u_int sunatm_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int atm_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void vtp_print(netdissect_options *, const u_char *, u_int); extern int mptcp_print(netdissect_options *, const u_char *, u_int, u_char); extern void ntp_print(netdissect_options *, const u_char *, u_int); extern void cnfp_print(netdissect_options *, const u_char *, const u_char *); extern void dvmrp_print(netdissect_options *, const u_char *, u_int); extern void egp_print(netdissect_options *, const u_char *, u_int); extern u_int enc_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int sl_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int sl_bsdos_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void tftp_print(netdissect_options *, const u_char *, u_int); extern void vrrp_print(netdissect_options *, const u_char *, u_int, const u_char *, int); extern void pimv1_print(netdissect_options *, const u_char *, u_int); extern void cisco_autorp_print(netdissect_options *, const u_char *, u_int); extern void pim_print(netdissect_options *, const u_char *, u_int, u_int); extern const u_char * ns_nprint (netdissect_options *, register const u_char *, register const u_char *); extern void ns_print(netdissect_options *, const u_char *, u_int, int); extern void bootp_print(netdissect_options *, const u_char *, u_int); extern void sflow_print(netdissect_options *, const u_char *, u_int); extern void aodv_print(netdissect_options *, const u_char *, u_int, int); extern void sctp_print(netdissect_options *, const u_char *, const u_char *, u_int); extern char *bgp_vpn_rd_print (netdissect_options *, const u_char *); extern void bgp_print(netdissect_options *, const u_char *, int); extern void olsr_print(netdissect_options *, const u_char *, u_int, int); extern void forces_print(netdissect_options *, const u_char *, u_int); extern void lspping_print(netdissect_options *, const u_char *, u_int); extern void isoclns_print(netdissect_options *, const u_char *, u_int, u_int); extern void krb_print(netdissect_options *, const u_char *); extern void cdp_print(netdissect_options *, const u_char *, u_int, u_int); extern void atalk_print(netdissect_options *, const u_char *, u_int); extern u_int ltalk_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int llap_print(netdissect_options *, const u_char *, u_int); extern void aarp_print(netdissect_options *, const u_char *, u_int); extern u_int juniper_atm1_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_atm2_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_mfr_print(netdissect_options *, const struct pcap_pkthdr *, register const u_char *); extern u_int juniper_mlfr_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_mlppp_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_pppoe_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_pppoe_atm_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_ggsn_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_es_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_monitor_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_services_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_ether_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_ppp_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_frelay_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int juniper_chdlc_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void snmp_print(netdissect_options *, const u_char *, u_int); extern void rx_print(netdissect_options *, register const u_char *, int, int, int, u_char *); extern void nfsreply_print(netdissect_options *, const u_char *, u_int, const u_char *); extern void nfsreply_print_noaddr(netdissect_options *, const u_char *, u_int, const u_char *); extern void nfsreq_print_noaddr(netdissect_options *, const u_char *, u_int, const u_char *); extern void sip_print(netdissect_options *, const u_char *, u_int); extern void syslog_print(netdissect_options *, const u_char *, u_int); extern void lwres_print(netdissect_options *, const u_char *, u_int); extern void cfm_print(netdissect_options *, const u_char *, u_int); extern void nbt_tcp_print(netdissect_options *, const u_char *, int); extern void nbt_udp137_print(netdissect_options *, const u_char *, int); extern void nbt_udp138_print(netdissect_options *, const u_char *, int); extern void smb_tcp_print(netdissect_options *, const u_char *, int); extern void netbeui_print(netdissect_options *, u_short, const u_char *, int); extern void ipx_netbios_print(netdissect_options *, const u_char *, u_int); extern void print_data(netdissect_options *, const unsigned char *, int); extern void decnet_print(netdissect_options *, const u_char *, u_int, u_int); extern void tcp_print(netdissect_options *, const u_char *, u_int, const u_char *, int); extern void ospf_print(netdissect_options *, const u_char *, u_int, const u_char *); extern int ospf_print_te_lsa(netdissect_options *, const uint8_t *, u_int); extern int ospf_print_grace_lsa(netdissect_options *, const uint8_t *, u_int); extern u_int ppp_print(netdissect_options *, register const u_char *, u_int); extern u_int ppp_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int ppp_hdlc_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int ppp_bsdos_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern void lldp_print(netdissect_options *, const u_char *, u_int); extern void rsvp_print(netdissect_options *, const u_char *, u_int); extern void timed_print(netdissect_options *, const u_char *); extern void m3ua_print(netdissect_options *, const u_char *, const u_int); extern void aoe_print(netdissect_options *, const u_char *, const u_int); +extern void pfsync_ip_print(netdissect_options *, const u_char *, u_int); + /* stuff that has not yet been rototiled */ #if 0 extern void ascii_print(netdissect_options *,u_int); extern void default_print(netdissect_options *,const u_char *, u_int); extern char *smb_errstr(netdissect_options *,int, int); extern const char *nt_errstr(netdissect_options *, uint32_t); #endif extern u_int ipnet_if_print(netdissect_options *,const struct pcap_pkthdr *, const u_char *); extern u_int ppi_if_print(netdissect_options *,const struct pcap_pkthdr *, const u_char *); extern u_int nflog_if_print(netdissect_options *,const struct pcap_pkthdr *, const u_char *); extern u_int ieee802_15_4_if_print(netdissect_options *,const struct pcap_pkthdr *, const u_char *); extern u_int pktap_if_print(netdissect_options *,const struct pcap_pkthdr *, const u_char *); extern u_int ieee802_11_radio_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int ieee802_11_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int ieee802_11_radio_avs_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); extern u_int prism_if_print(netdissect_options *, const struct pcap_pkthdr *, const u_char *); #ifdef INET6 extern void ip6_print(netdissect_options *,const u_char *, u_int); extern int frag6_print(netdissect_options *, const u_char *, const u_char *); extern int rt6_print(netdissect_options *, const u_char *, const u_char *); extern int hbhopt_print(netdissect_options *, const u_char *); extern int dstopt_print(netdissect_options *, const u_char *); extern void ripng_print(netdissect_options *, const u_char *, unsigned int); extern int mobility_print(netdissect_options *, const u_char *, const u_char *); extern void dhcp6_print(netdissect_options *, const u_char *, u_int); extern void ospf6_print(netdissect_options *, const u_char *, u_int); extern void babel_print(netdissect_options *, const u_char *, u_int); #endif /*INET6*/ #if 0 struct cksum_vec { const uint8_t *ptr; int len; }; extern uint16_t in_cksum(const struct cksum_vec *, int); extern uint16_t in_cksum_shouldbe(uint16_t, uint16_t); #endif extern int nextproto4_cksum(netdissect_options *ndo, const struct ip *, const uint8_t *, u_int, u_int, u_int); extern int decode_prefix4(netdissect_options *ndo, const u_char *, u_int, char *, u_int); #ifdef INET6 extern int decode_prefix6(netdissect_options *ndo, const u_char *, u_int, char *, u_int); #endif extern void esp_print_decodesecret(netdissect_options *ndo); extern int esp_print_decrypt_buffer_by_ikev2(netdissect_options *ndo, int initiator, u_char spii[8], u_char spir[8], u_char *buf, u_char *end); extern void geonet_print(netdissect_options *ndo,const u_char *eth_hdr,const u_char *geo_pck, u_int len); extern void calm_fast_print(netdissect_options *ndo,const u_char *eth_hdr,const u_char *calm_pck, u_int len); #endif /* netdissect_h */ Index: user/ngie/more-tests/contrib/tcpdump/print-ip.c =================================================================== --- user/ngie/more-tests/contrib/tcpdump/print-ip.c (revision 281974) +++ user/ngie/more-tests/contrib/tcpdump/print-ip.c (revision 281975) @@ -1,717 +1,717 @@ /* * Copyright (c) 1988, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that: (1) source code distributions * retain the above copyright notice and this paragraph in its entirety, (2) * distributions including binary code include the above copyright notice and * this paragraph in its entirety in the documentation or other materials * provided with the distribution, and (3) all advertising materials mentioning * features or use of this software display the following acknowledgement: * ``This product includes software developed by the University of California, * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of * the University nor the names of its contributors may be used to endorse * or promote products derived from this software without specific prior * written permission. * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. * * $FreeBSD$ */ #define NETDISSECT_REWORKED #ifdef HAVE_CONFIG_H #include "config.h" #endif #include #include #include "interface.h" #include "addrtoname.h" #include "extract.h" /* must come after interface.h */ #include "ip.h" #include "ipproto.h" static const char tstr[] = "[|ip]"; static const struct tok ip_option_values[] = { { IPOPT_EOL, "EOL" }, { IPOPT_NOP, "NOP" }, { IPOPT_TS, "timestamp" }, { IPOPT_SECURITY, "security" }, { IPOPT_RR, "RR" }, { IPOPT_SSRR, "SSRR" }, { IPOPT_LSRR, "LSRR" }, { IPOPT_RA, "RA" }, { IPOPT_RFC1393, "traceroute" }, { 0, NULL } }; /* * print the recorded route in an IP RR, LSRR or SSRR option. */ static void ip_printroute(netdissect_options *ndo, register const u_char *cp, u_int length) { register u_int ptr; register u_int len; if (length < 3) { ND_PRINT((ndo, " [bad length %u]", length)); return; } if ((length + 1) & 3) ND_PRINT((ndo, " [bad length %u]", length)); ptr = cp[2] - 1; if (ptr < 3 || ((ptr + 1) & 3) || ptr > length + 1) ND_PRINT((ndo, " [bad ptr %u]", cp[2])); for (len = 3; len < length; len += 4) { ND_PRINT((ndo, " %s", ipaddr_string(ndo, &cp[len]))); if (ptr > len) ND_PRINT((ndo, ",")); } } /* * If source-routing is present and valid, return the final destination. * Otherwise, return IP destination. * * This is used for UDP and TCP pseudo-header in the checksum * calculation. */ static uint32_t ip_finddst(netdissect_options *ndo, const struct ip *ip) { int length; int len; const u_char *cp; uint32_t retval; cp = (const u_char *)(ip + 1); length = (IP_HL(ip) << 2) - sizeof(struct ip); for (; length > 0; cp += len, length -= len) { int tt; ND_TCHECK(*cp); tt = *cp; if (tt == IPOPT_EOL) break; else if (tt == IPOPT_NOP) len = 1; else { ND_TCHECK(cp[1]); len = cp[1]; if (len < 2) break; } ND_TCHECK2(*cp, len); switch (tt) { case IPOPT_SSRR: case IPOPT_LSRR: if (len < 7) break; UNALIGNED_MEMCPY(&retval, cp + len - 4, 4); return retval; } } trunc: UNALIGNED_MEMCPY(&retval, &ip->ip_dst.s_addr, sizeof(uint32_t)); return retval; } /* * Compute a V4-style checksum by building a pseudoheader. */ int nextproto4_cksum(netdissect_options *ndo, const struct ip *ip, const uint8_t *data, u_int len, u_int covlen, u_int next_proto) { struct phdr { uint32_t src; uint32_t dst; u_char mbz; u_char proto; uint16_t len; } ph; struct cksum_vec vec[2]; /* pseudo-header.. */ ph.len = htons((uint16_t)len); ph.mbz = 0; ph.proto = next_proto; UNALIGNED_MEMCPY(&ph.src, &ip->ip_src.s_addr, sizeof(uint32_t)); if (IP_HL(ip) == 5) UNALIGNED_MEMCPY(&ph.dst, &ip->ip_dst.s_addr, sizeof(uint32_t)); else ph.dst = ip_finddst(ndo, ip); vec[0].ptr = (const uint8_t *)(void *)&ph; vec[0].len = sizeof(ph); vec[1].ptr = data; vec[1].len = covlen; return (in_cksum(vec, 2)); } static void ip_printts(netdissect_options *ndo, register const u_char *cp, u_int length) { register u_int ptr; register u_int len; int hoplen; const char *type; if (length < 4) { ND_PRINT((ndo, "[bad length %u]", length)); return; } ND_PRINT((ndo, " TS{")); hoplen = ((cp[3]&0xF) != IPOPT_TS_TSONLY) ? 8 : 4; if ((length - 4) & (hoplen-1)) ND_PRINT((ndo, "[bad length %u]", length)); ptr = cp[2] - 1; len = 0; if (ptr < 4 || ((ptr - 4) & (hoplen-1)) || ptr > length + 1) ND_PRINT((ndo, "[bad ptr %u]", cp[2])); switch (cp[3]&0xF) { case IPOPT_TS_TSONLY: ND_PRINT((ndo, "TSONLY")); break; case IPOPT_TS_TSANDADDR: ND_PRINT((ndo, "TS+ADDR")); break; /* * prespecified should really be 3, but some ones might send 2 * instead, and the IPOPT_TS_PRESPEC constant can apparently * have both values, so we have to hard-code it here. */ case 2: ND_PRINT((ndo, "PRESPEC2.0")); break; case 3: /* IPOPT_TS_PRESPEC */ ND_PRINT((ndo, "PRESPEC")); break; default: ND_PRINT((ndo, "[bad ts type %d]", cp[3]&0xF)); goto done; } type = " "; for (len = 4; len < length; len += hoplen) { if (ptr == len) type = " ^ "; ND_PRINT((ndo, "%s%d@%s", type, EXTRACT_32BITS(&cp[len+hoplen-4]), hoplen!=8 ? "" : ipaddr_string(ndo, &cp[len]))); type = " "; } done: ND_PRINT((ndo, "%s", ptr == len ? " ^ " : "")); if (cp[3]>>4) ND_PRINT((ndo, " [%d hops not recorded]} ", cp[3]>>4)); else ND_PRINT((ndo, "}")); } /* * print IP options. */ static void ip_optprint(netdissect_options *ndo, register const u_char *cp, u_int length) { register u_int option_len; const char *sep = ""; for (; length > 0; cp += option_len, length -= option_len) { u_int option_code; ND_PRINT((ndo, "%s", sep)); sep = ","; ND_TCHECK(*cp); option_code = *cp; ND_PRINT((ndo, "%s", tok2str(ip_option_values,"unknown %u",option_code))); if (option_code == IPOPT_NOP || option_code == IPOPT_EOL) option_len = 1; else { ND_TCHECK(cp[1]); option_len = cp[1]; if (option_len < 2) { ND_PRINT((ndo, " [bad length %u]", option_len)); return; } } if (option_len > length) { ND_PRINT((ndo, " [bad length %u]", option_len)); return; } ND_TCHECK2(*cp, option_len); switch (option_code) { case IPOPT_EOL: return; case IPOPT_TS: ip_printts(ndo, cp, option_len); break; case IPOPT_RR: /* fall through */ case IPOPT_SSRR: case IPOPT_LSRR: ip_printroute(ndo, cp, option_len); break; case IPOPT_RA: if (option_len < 4) { ND_PRINT((ndo, " [bad length %u]", option_len)); break; } ND_TCHECK(cp[3]); if (EXTRACT_16BITS(&cp[2]) != 0) ND_PRINT((ndo, " value %u", EXTRACT_16BITS(&cp[2]))); break; case IPOPT_NOP: /* nothing to print - fall through */ case IPOPT_SECURITY: default: break; } } return; trunc: ND_PRINT((ndo, "%s", tstr)); } #define IP_RES 0x8000 static const struct tok ip_frag_values[] = { { IP_MF, "+" }, { IP_DF, "DF" }, { IP_RES, "rsvd" }, /* The RFC3514 evil ;-) bit */ { 0, NULL } }; struct ip_print_demux_state { const struct ip *ip; const u_char *cp; u_int len, off; u_char nh; int advance; }; static void ip_print_demux(netdissect_options *ndo, struct ip_print_demux_state *ipds) { struct protoent *proto; struct cksum_vec vec[1]; again: switch (ipds->nh) { case IPPROTO_AH: ipds->nh = *ipds->cp; ipds->advance = ah_print(ndo, ipds->cp); if (ipds->advance <= 0) break; ipds->cp += ipds->advance; ipds->len -= ipds->advance; goto again; case IPPROTO_ESP: { int enh, padlen; ipds->advance = esp_print(ndo, ipds->cp, ipds->len, (const u_char *)ipds->ip, &enh, &padlen); if (ipds->advance <= 0) break; ipds->cp += ipds->advance; ipds->len -= ipds->advance + padlen; ipds->nh = enh & 0xff; goto again; } case IPPROTO_IPCOMP: { int enh; ipds->advance = ipcomp_print(ndo, ipds->cp, &enh); if (ipds->advance <= 0) break; ipds->cp += ipds->advance; ipds->len -= ipds->advance; ipds->nh = enh & 0xff; goto again; } case IPPROTO_SCTP: sctp_print(ndo, ipds->cp, (const u_char *)ipds->ip, ipds->len); break; case IPPROTO_DCCP: dccp_print(ndo, ipds->cp, (const u_char *)ipds->ip, ipds->len); break; case IPPROTO_TCP: /* pass on the MF bit plus the offset to detect fragments */ tcp_print(ndo, ipds->cp, ipds->len, (const u_char *)ipds->ip, ipds->off & (IP_MF|IP_OFFMASK)); break; case IPPROTO_UDP: /* pass on the MF bit plus the offset to detect fragments */ udp_print(ndo, ipds->cp, ipds->len, (const u_char *)ipds->ip, ipds->off & (IP_MF|IP_OFFMASK)); break; case IPPROTO_ICMP: /* pass on the MF bit plus the offset to detect fragments */ icmp_print(ndo, ipds->cp, ipds->len, (const u_char *)ipds->ip, ipds->off & (IP_MF|IP_OFFMASK)); break; case IPPROTO_PIGP: /* * XXX - the current IANA protocol number assignments * page lists 9 as "any private interior gateway * (used by Cisco for their IGRP)" and 88 as * "EIGRP" from Cisco. * * Recent BSD headers define * IP_PROTO_PIGP as 9 and IP_PROTO_IGRP as 88. * We define IP_PROTO_PIGP as 9 and * IP_PROTO_EIGRP as 88; those names better * match was the current protocol number * assignments say. */ igrp_print(ndo, ipds->cp, ipds->len); break; case IPPROTO_EIGRP: eigrp_print(ndo, ipds->cp, ipds->len); break; case IPPROTO_ND: ND_PRINT((ndo, " nd %d", ipds->len)); break; case IPPROTO_EGP: egp_print(ndo, ipds->cp, ipds->len); break; case IPPROTO_OSPF: ospf_print(ndo, ipds->cp, ipds->len, (const u_char *)ipds->ip); break; case IPPROTO_IGMP: igmp_print(ndo, ipds->cp, ipds->len); break; case IPPROTO_IPV4: /* DVMRP multicast tunnel (ip-in-ip encapsulation) */ ip_print(ndo, ipds->cp, ipds->len); if (! ndo->ndo_vflag) { ND_PRINT((ndo, " (ipip-proto-4)")); return; } break; #ifdef INET6 case IPPROTO_IPV6: /* ip6-in-ip encapsulation */ ip6_print(ndo, ipds->cp, ipds->len); break; #endif /*INET6*/ case IPPROTO_RSVP: rsvp_print(ndo, ipds->cp, ipds->len); break; case IPPROTO_GRE: /* do it */ gre_print(ndo, ipds->cp, ipds->len); break; case IPPROTO_MOBILE: mobile_print(ndo, ipds->cp, ipds->len); break; case IPPROTO_PIM: vec[0].ptr = ipds->cp; vec[0].len = ipds->len; pim_print(ndo, ipds->cp, ipds->len, in_cksum(vec, 1)); break; case IPPROTO_VRRP: if (ndo->ndo_packettype == PT_CARP) { if (ndo->ndo_vflag) ND_PRINT((ndo, "carp %s > %s: ", ipaddr_string(ndo, &ipds->ip->ip_src), ipaddr_string(ndo, &ipds->ip->ip_dst))); carp_print(ndo, ipds->cp, ipds->len, ipds->ip->ip_ttl); } else { if (ndo->ndo_vflag) ND_PRINT((ndo, "vrrp %s > %s: ", ipaddr_string(ndo, &ipds->ip->ip_src), ipaddr_string(ndo, &ipds->ip->ip_dst))); vrrp_print(ndo, ipds->cp, ipds->len, (const u_char *)ipds->ip, ipds->ip->ip_ttl); } break; case IPPROTO_PGM: pgm_print(ndo, ipds->cp, ipds->len, (const u_char *)ipds->ip); break; #if defined(HAVE_NET_PFVAR_H) case IPPROTO_PFSYNC: - pfsync_ip_print(ipds->cp, ipds->len); + pfsync_ip_print(ndo, ipds->cp, ipds->len); break; #endif default: if (ndo->ndo_nflag==0 && (proto = getprotobynumber(ipds->nh)) != NULL) ND_PRINT((ndo, " %s", proto->p_name)); else ND_PRINT((ndo, " ip-proto-%d", ipds->nh)); ND_PRINT((ndo, " %d", ipds->len)); break; } } void ip_print_inner(netdissect_options *ndo, const u_char *bp, u_int length, u_int nh, const u_char *bp2) { struct ip_print_demux_state ipd; ipd.ip = (const struct ip *)bp2; ipd.cp = bp; ipd.len = length; ipd.off = 0; ipd.nh = nh; ipd.advance = 0; ip_print_demux(ndo, &ipd); } /* * print an IP datagram. */ void ip_print(netdissect_options *ndo, const u_char *bp, u_int length) { struct ip_print_demux_state ipd; struct ip_print_demux_state *ipds=&ipd; const u_char *ipend; u_int hlen; struct cksum_vec vec[1]; uint16_t sum, ip_sum; struct protoent *proto; ipds->ip = (const struct ip *)bp; ND_TCHECK(ipds->ip->ip_vhl); if (IP_V(ipds->ip) != 4) { /* print version if != 4 */ ND_PRINT((ndo, "IP%u ", IP_V(ipds->ip))); if (IP_V(ipds->ip) == 6) ND_PRINT((ndo, ", wrong link-layer encapsulation")); } else if (!ndo->ndo_eflag) ND_PRINT((ndo, "IP ")); ND_TCHECK(*ipds->ip); if (length < sizeof (struct ip)) { ND_PRINT((ndo, "truncated-ip %u", length)); return; } hlen = IP_HL(ipds->ip) * 4; if (hlen < sizeof (struct ip)) { ND_PRINT((ndo, "bad-hlen %u", hlen)); return; } ipds->len = EXTRACT_16BITS(&ipds->ip->ip_len); if (length < ipds->len) ND_PRINT((ndo, "truncated-ip - %u bytes missing! ", ipds->len - length)); if (ipds->len < hlen) { #ifdef GUESS_TSO if (ipds->len) { ND_PRINT((ndo, "bad-len %u", ipds->len)); return; } else { /* we guess that it is a TSO send */ ipds->len = length; } #else ND_PRINT((ndo, "bad-len %u", ipds->len)); return; #endif /* GUESS_TSO */ } /* * Cut off the snapshot length to the end of the IP payload. */ ipend = bp + ipds->len; if (ipend < ndo->ndo_snapend) ndo->ndo_snapend = ipend; ipds->len -= hlen; ipds->off = EXTRACT_16BITS(&ipds->ip->ip_off); if (ndo->ndo_vflag) { ND_PRINT((ndo, "(tos 0x%x", (int)ipds->ip->ip_tos)); /* ECN bits */ if (ipds->ip->ip_tos & 0x03) { switch (ipds->ip->ip_tos & 0x03) { case 1: ND_PRINT((ndo, ",ECT(1)")); break; case 2: ND_PRINT((ndo, ",ECT(0)")); break; case 3: ND_PRINT((ndo, ",CE")); } } if (ipds->ip->ip_ttl >= 1) ND_PRINT((ndo, ", ttl %u", ipds->ip->ip_ttl)); /* * for the firewall guys, print id, offset. * On all but the last stick a "+" in the flags portion. * For unfragmented datagrams, note the don't fragment flag. */ ND_PRINT((ndo, ", id %u, offset %u, flags [%s], proto %s (%u)", EXTRACT_16BITS(&ipds->ip->ip_id), (ipds->off & 0x1fff) * 8, bittok2str(ip_frag_values, "none", ipds->off&0xe000), tok2str(ipproto_values,"unknown",ipds->ip->ip_p), ipds->ip->ip_p)); ND_PRINT((ndo, ", length %u", EXTRACT_16BITS(&ipds->ip->ip_len))); if ((hlen - sizeof(struct ip)) > 0) { ND_PRINT((ndo, ", options (")); ip_optprint(ndo, (u_char *)(ipds->ip + 1), hlen - sizeof(struct ip)); ND_PRINT((ndo, ")")); } if (!ndo->ndo_Kflag && (u_char *)ipds->ip + hlen <= ndo->ndo_snapend) { vec[0].ptr = (const uint8_t *)(void *)ipds->ip; vec[0].len = hlen; sum = in_cksum(vec, 1); if (sum != 0) { ip_sum = EXTRACT_16BITS(&ipds->ip->ip_sum); ND_PRINT((ndo, ", bad cksum %x (->%x)!", ip_sum, in_cksum_shouldbe(ip_sum, sum))); } } ND_PRINT((ndo, ")\n ")); } /* * If this is fragment zero, hand it to the next higher * level protocol. */ if ((ipds->off & 0x1fff) == 0) { ipds->cp = (const u_char *)ipds->ip + hlen; ipds->nh = ipds->ip->ip_p; if (ipds->nh != IPPROTO_TCP && ipds->nh != IPPROTO_UDP && ipds->nh != IPPROTO_SCTP && ipds->nh != IPPROTO_DCCP) { ND_PRINT((ndo, "%s > %s: ", ipaddr_string(ndo, &ipds->ip->ip_src), ipaddr_string(ndo, &ipds->ip->ip_dst))); } ip_print_demux(ndo, ipds); } else { /* Ultra quiet now means that all this stuff should be suppressed */ if (ndo->ndo_qflag > 1) return; /* * if this isn't the first frag, we're missing the * next level protocol header. print the ip addr * and the protocol. */ if (ipds->off & 0x1fff) { ND_PRINT((ndo, "%s > %s:", ipaddr_string(ndo, &ipds->ip->ip_src), ipaddr_string(ndo, &ipds->ip->ip_dst))); if (!ndo->ndo_nflag && (proto = getprotobynumber(ipds->ip->ip_p)) != NULL) ND_PRINT((ndo, " %s", proto->p_name)); else ND_PRINT((ndo, " ip-proto-%d", ipds->ip->ip_p)); } } return; trunc: ND_PRINT((ndo, "%s", tstr)); return; } void ipN_print(netdissect_options *ndo, register const u_char *bp, register u_int length) { struct ip hdr; if (length < 4) { ND_PRINT((ndo, "truncated-ip %d", length)); return; } memcpy (&hdr, bp, 4); switch (IP_V(&hdr)) { case 4: ip_print (ndo, bp, length); return; #ifdef INET6 case 6: ip6_print (ndo, bp, length); return; #endif default: ND_PRINT((ndo, "unknown ip %d", IP_V(&hdr))); return; } } /* * Local Variables: * c-style: whitesmith * c-basic-offset: 8 * End: */ Index: user/ngie/more-tests/contrib/tcpdump/print-pfsync.c =================================================================== --- user/ngie/more-tests/contrib/tcpdump/print-pfsync.c (revision 281974) +++ user/ngie/more-tests/contrib/tcpdump/print-pfsync.c (revision 281975) @@ -1,450 +1,456 @@ /* * Copyright (c) 2012 Gleb Smirnoff * Copyright (c) 2002 Michael Shalayeff * Copyright (c) 2001 Daniel Hartmeier * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. * * $OpenBSD: print-pfsync.c,v 1.38 2012/09/19 13:50:36 mikeb Exp $ * $OpenBSD: pf_print_state.c,v 1.11 2012/07/08 17:48:37 lteo Exp $ */ #ifdef HAVE_CONFIG_H #include "config.h" #endif #include #include #include #include /* XXX */ #include #define TCPSTATES #include #include #include "interface.h" #include "addrtoname.h" -static void pfsync_print(struct pfsync_header *, const u_char *, u_int); -static void print_src_dst(const struct pfsync_state_peer *, +static void pfsync_print(netdissect_options *, struct pfsync_header *, + const u_char *, u_int); +static void print_src_dst(netdissect_options *, + const struct pfsync_state_peer *, const struct pfsync_state_peer *, uint8_t); -static void print_state(struct pfsync_state *); +static void print_state(netdissect_options *, struct pfsync_state *); #ifdef notyet void pfsync_if_print(u_char *user, const struct pcap_pkthdr *h, register const u_char *p) { u_int caplen = h->caplen; ts_print(&h->ts); if (caplen < PFSYNC_HDRLEN) { - printf("[|pfsync]"); + ND_PRINT((ndo, "[|pfsync]")); goto out; } pfsync_print((struct pfsync_header *)p, p + sizeof(struct pfsync_header), caplen - sizeof(struct pfsync_header)); out: if (xflag) { default_print((const u_char *)p, caplen); } - putchar('\n'); + safeputchar(ndo, '\n'); } #endif /* notyet */ void -pfsync_ip_print(const u_char *bp, u_int len) +pfsync_ip_print(netdissect_options *ndo , const u_char *bp, u_int len) { struct pfsync_header *hdr = (struct pfsync_header *)bp; if (len < PFSYNC_HDRLEN) - printf("[|pfsync]"); + ND_PRINT((ndo, "[|pfsync]")); else - pfsync_print(hdr, bp + sizeof(struct pfsync_header), + pfsync_print(ndo, hdr, bp + sizeof(struct pfsync_header), len - sizeof(struct pfsync_header)); } struct pfsync_actions { const char *name; size_t len; - void (*print)(const void *); + void (*print)(netdissect_options *, const void *); }; -static void pfsync_print_clr(const void *); -static void pfsync_print_state(const void *); -static void pfsync_print_ins_ack(const void *); -static void pfsync_print_upd_c(const void *); -static void pfsync_print_upd_req(const void *); -static void pfsync_print_del_c(const void *); -static void pfsync_print_bus(const void *); -static void pfsync_print_tdb(const void *); +static void pfsync_print_clr(netdissect_options *, const void *); +static void pfsync_print_state(netdissect_options *, const void *); +static void pfsync_print_ins_ack(netdissect_options *, const void *); +static void pfsync_print_upd_c(netdissect_options *, const void *); +static void pfsync_print_upd_req(netdissect_options *, const void *); +static void pfsync_print_del_c(netdissect_options *, const void *); +static void pfsync_print_bus(netdissect_options *, const void *); +static void pfsync_print_tdb(netdissect_options *, const void *); struct pfsync_actions actions[] = { { "clear all", sizeof(struct pfsync_clr), pfsync_print_clr }, { "insert", sizeof(struct pfsync_state), pfsync_print_state }, { "insert ack", sizeof(struct pfsync_ins_ack), pfsync_print_ins_ack }, { "update", sizeof(struct pfsync_ins_ack), pfsync_print_state }, { "update compressed", sizeof(struct pfsync_upd_c), pfsync_print_upd_c }, { "request uncompressed", sizeof(struct pfsync_upd_req), pfsync_print_upd_req }, { "delete", sizeof(struct pfsync_state), pfsync_print_state }, { "delete compressed", sizeof(struct pfsync_del_c), pfsync_print_del_c }, { "frag insert", 0, NULL }, { "frag delete", 0, NULL }, { "bulk update status", sizeof(struct pfsync_bus), pfsync_print_bus }, { "tdb", 0, pfsync_print_tdb }, { "eof", 0, NULL }, }; static void -pfsync_print(struct pfsync_header *hdr, const u_char *bp, u_int len) +pfsync_print(netdissect_options *ndo, struct pfsync_header *hdr, + const u_char *bp, u_int len) { struct pfsync_subheader *subh; int count, plen, i; u_int alen; plen = ntohs(hdr->len); - printf("PFSYNCv%d len %d", hdr->version, plen); + ND_PRINT((ndo, "PFSYNCv%d len %d", hdr->version, plen)); if (hdr->version != PFSYNC_VERSION) return; plen -= sizeof(*hdr); while (plen > 0) { if (len < sizeof(*subh)) break; subh = (struct pfsync_subheader *)bp; bp += sizeof(*subh); len -= sizeof(*subh); plen -= sizeof(*subh); if (subh->action >= PFSYNC_ACT_MAX) { - printf("\n act UNKNOWN id %d", subh->action); + ND_PRINT((ndo, "\n act UNKNOWN id %d", + subh->action)); return; } count = ntohs(subh->count); - printf("\n %s count %d", actions[subh->action].name, count); + ND_PRINT((ndo, "\n %s count %d", actions[subh->action].name, + count)); alen = actions[subh->action].len; if (subh->action == PFSYNC_ACT_EOF) return; if (actions[subh->action].print == NULL) { - printf("\n unimplemented action %hhu", subh->action); + ND_PRINT((ndo, "\n unimplemented action %hhu", + subh->action)); return; } for (i = 0; i < count; i++) { if (len < alen) { len = 0; break; } if (vflag) - actions[subh->action].print(bp); + actions[subh->action].print(ndo, bp); bp += alen; len -= alen; plen -= alen; } } if (plen > 0) { - printf("\n ..."); + ND_PRINT((ndo, "\n ...")); return; } if (plen < 0) { - printf("\n invalid header length"); + ND_PRINT((ndo, "\n invalid header length")); return; } if (len > 0) - printf("\n invalid packet length"); + ND_PRINT((ndo, "\n invalid packet length")); } static void -pfsync_print_clr(const void *bp) +pfsync_print_clr(netdissect_options *ndo, const void *bp) { const struct pfsync_clr *clr = bp; - printf("\n\tcreatorid: %08x", htonl(clr->creatorid)); + ND_PRINT((ndo, "\n\tcreatorid: %08x", htonl(clr->creatorid))); if (clr->ifname[0] != '\0') - printf(" interface: %s", clr->ifname); + ND_PRINT((ndo, " interface: %s", clr->ifname)); } static void -pfsync_print_state(const void *bp) +pfsync_print_state(netdissect_options *ndo, const void *bp) { struct pfsync_state *st = (struct pfsync_state *)bp; - putchar('\n'); - print_state(st); + safeputchar(ndo, '\n'); + print_state(ndo, st); } static void -pfsync_print_ins_ack(const void *bp) +pfsync_print_ins_ack(netdissect_options *ndo, const void *bp) { const struct pfsync_ins_ack *iack = bp; - printf("\n\tid: %016jx creatorid: %08x", (uintmax_t )be64toh(iack->id), - ntohl(iack->creatorid)); + ND_PRINT((ndo, "\n\tid: %016jx creatorid: %08x", + (uintmax_t)be64toh(iack->id), ntohl(iack->creatorid))); } static void -pfsync_print_upd_c(const void *bp) +pfsync_print_upd_c(netdissect_options *ndo, const void *bp) { const struct pfsync_upd_c *u = bp; - printf("\n\tid: %016jx creatorid: %08x", (uintmax_t )be64toh(u->id), - ntohl(u->creatorid)); + ND_PRINT((ndo, "\n\tid: %016jx creatorid: %08x", + (uintmax_t)be64toh(u->id), ntohl(u->creatorid))); if (vflag > 2) { - printf("\n\tTCP? :"); - print_src_dst(&u->src, &u->dst, IPPROTO_TCP); + ND_PRINT((ndo, "\n\tTCP? :")); + print_src_dst(ndo, &u->src, &u->dst, IPPROTO_TCP); } } static void -pfsync_print_upd_req(const void *bp) +pfsync_print_upd_req(netdissect_options *ndo, const void *bp) { const struct pfsync_upd_req *ur = bp; - printf("\n\tid: %016jx creatorid: %08x", (uintmax_t )be64toh(ur->id), - ntohl(ur->creatorid)); + ND_PRINT((ndo, "\n\tid: %016jx creatorid: %08x", + (uintmax_t)be64toh(ur->id), ntohl(ur->creatorid))); } static void -pfsync_print_del_c(const void *bp) +pfsync_print_del_c(netdissect_options *ndo, const void *bp) { const struct pfsync_del_c *d = bp; - printf("\n\tid: %016jx creatorid: %08x", (uintmax_t )be64toh(d->id), - ntohl(d->creatorid)); + ND_PRINT((ndo, "\n\tid: %016jx creatorid: %08x", + (uintmax_t)be64toh(d->id), ntohl(d->creatorid))); } static void -pfsync_print_bus(const void *bp) +pfsync_print_bus(netdissect_options *ndo, const void *bp) { const struct pfsync_bus *b = bp; uint32_t endtime; int min, sec; const char *status; endtime = ntohl(b->endtime); sec = endtime % 60; endtime /= 60; min = endtime % 60; endtime /= 60; switch (b->status) { case PFSYNC_BUS_START: status = "start"; break; case PFSYNC_BUS_END: status = "end"; break; default: status = "UNKNOWN"; break; } - printf("\n\tcreatorid: %08x age: %.2u:%.2u:%.2u status: %s", - htonl(b->creatorid), endtime, min, sec, status); + ND_PRINT((ndo, "\n\tcreatorid: %08x age: %.2u:%.2u:%.2u status: %s", + htonl(b->creatorid), endtime, min, sec, status)); } static void -pfsync_print_tdb(const void *bp) +pfsync_print_tdb(netdissect_options *ndo, const void *bp) { const struct pfsync_tdb *t = bp; - printf("\n\tspi: 0x%08x rpl: %ju cur_bytes: %ju", + ND_PRINT((ndo, "\n\tspi: 0x%08x rpl: %ju cur_bytes: %ju", ntohl(t->spi), (uintmax_t )be64toh(t->rpl), - (uintmax_t )be64toh(t->cur_bytes)); + (uintmax_t )be64toh(t->cur_bytes))); } static void -print_host(struct pf_addr *addr, uint16_t port, sa_family_t af, - const char *proto) +print_host(netdissect_options *ndo, struct pf_addr *addr, uint16_t port, + sa_family_t af, const char *proto) { char buf[48]; if (inet_ntop(af, addr, buf, sizeof(buf)) == NULL) - printf("?"); + ND_PRINT((ndo, "?")); else - printf("%s", buf); + ND_PRINT((ndo, "%s", buf)); if (port) - printf(".%hu", ntohs(port)); + ND_PRINT((ndo, ".%hu", ntohs(port))); } static void -print_seq(const struct pfsync_state_peer *p) +print_seq(netdissect_options *ndo, const struct pfsync_state_peer *p) { if (p->seqdiff) - printf("[%u + %u](+%u)", ntohl(p->seqlo), - ntohl(p->seqhi) - ntohl(p->seqlo), ntohl(p->seqdiff)); + ND_PRINT((ndo, "[%u + %u](+%u)", ntohl(p->seqlo), + ntohl(p->seqhi) - ntohl(p->seqlo), ntohl(p->seqdiff))); else - printf("[%u + %u]", ntohl(p->seqlo), - ntohl(p->seqhi) - ntohl(p->seqlo)); + ND_PRINT((ndo, "[%u + %u]", ntohl(p->seqlo), + ntohl(p->seqhi) - ntohl(p->seqlo))); } static void -print_src_dst(const struct pfsync_state_peer *src, +print_src_dst(netdissect_options *ndo, const struct pfsync_state_peer *src, const struct pfsync_state_peer *dst, uint8_t proto) { if (proto == IPPROTO_TCP) { if (src->state <= TCPS_TIME_WAIT && dst->state <= TCPS_TIME_WAIT) - printf(" %s:%s", tcpstates[src->state], - tcpstates[dst->state]); + ND_PRINT((ndo, " %s:%s", tcpstates[src->state], + tcpstates[dst->state])); else if (src->state == PF_TCPS_PROXY_SRC || dst->state == PF_TCPS_PROXY_SRC) - printf(" PROXY:SRC"); + ND_PRINT((ndo, " PROXY:SRC")); else if (src->state == PF_TCPS_PROXY_DST || dst->state == PF_TCPS_PROXY_DST) - printf(" PROXY:DST"); + ND_PRINT((ndo, " PROXY:DST")); else - printf(" ", - src->state, dst->state); + ND_PRINT((ndo, " ", + src->state, dst->state)); if (vflag > 1) { - printf("\n\t"); - print_seq(src); + ND_PRINT((ndo, "\n\t")); + print_seq(ndo, src); if (src->wscale && dst->wscale) - printf(" wscale %u", - src->wscale & PF_WSCALE_MASK); - printf(" "); - print_seq(dst); + ND_PRINT((ndo, " wscale %u", + src->wscale & PF_WSCALE_MASK)); + ND_PRINT((ndo, " ")); + print_seq(ndo, dst); if (src->wscale && dst->wscale) - printf(" wscale %u", - dst->wscale & PF_WSCALE_MASK); + ND_PRINT((ndo, " wscale %u", + dst->wscale & PF_WSCALE_MASK)); } } else if (proto == IPPROTO_UDP && src->state < PFUDPS_NSTATES && dst->state < PFUDPS_NSTATES) { const char *states[] = PFUDPS_NAMES; - printf(" %s:%s", states[src->state], states[dst->state]); + ND_PRINT((ndo, " %s:%s", states[src->state], states[dst->state])); } else if (proto != IPPROTO_ICMP && src->state < PFOTHERS_NSTATES && dst->state < PFOTHERS_NSTATES) { /* XXX ICMP doesn't really have state levels */ const char *states[] = PFOTHERS_NAMES; - printf(" %s:%s", states[src->state], states[dst->state]); + ND_PRINT((ndo, " %s:%s", states[src->state], states[dst->state])); } else { - printf(" %u:%u", src->state, dst->state); + ND_PRINT((ndo, " %u:%u", src->state, dst->state)); } } static void -print_state(struct pfsync_state *s) +print_state(netdissect_options *ndo, struct pfsync_state *s) { struct pfsync_state_peer *src, *dst; struct pfsync_state_key *sk, *nk; int min, sec; if (s->direction == PF_OUT) { src = &s->src; dst = &s->dst; sk = &s->key[PF_SK_STACK]; nk = &s->key[PF_SK_WIRE]; if (s->proto == IPPROTO_ICMP || s->proto == IPPROTO_ICMPV6) sk->port[0] = nk->port[0]; } else { src = &s->dst; dst = &s->src; sk = &s->key[PF_SK_WIRE]; nk = &s->key[PF_SK_STACK]; if (s->proto == IPPROTO_ICMP || s->proto == IPPROTO_ICMPV6) sk->port[1] = nk->port[1]; } - printf("\t%s ", s->ifname); - printf("proto %u ", s->proto); + ND_PRINT((ndo, "\t%s ", s->ifname)); + ND_PRINT((ndo, "proto %u ", s->proto)); - print_host(&nk->addr[1], nk->port[1], s->af, NULL); + print_host(ndo, &nk->addr[1], nk->port[1], s->af, NULL); if (PF_ANEQ(&nk->addr[1], &sk->addr[1], s->af) || nk->port[1] != sk->port[1]) { - printf(" ("); - print_host(&sk->addr[1], sk->port[1], s->af, NULL); - printf(")"); + ND_PRINT((ndo, " (")); + print_host(ndo, &sk->addr[1], sk->port[1], s->af, NULL); + ND_PRINT((ndo, ")")); } if (s->direction == PF_OUT) - printf(" -> "); + ND_PRINT((ndo, " -> ")); else - printf(" <- "); - print_host(&nk->addr[0], nk->port[0], s->af, NULL); + ND_PRINT((ndo, " <- ")); + print_host(ndo, &nk->addr[0], nk->port[0], s->af, NULL); if (PF_ANEQ(&nk->addr[0], &sk->addr[0], s->af) || nk->port[0] != sk->port[0]) { - printf(" ("); - print_host(&sk->addr[0], sk->port[0], s->af, NULL); - printf(")"); + ND_PRINT((ndo, " (")); + print_host(ndo, &sk->addr[0], sk->port[0], s->af, NULL); + ND_PRINT((ndo, ")")); } - print_src_dst(src, dst, s->proto); + print_src_dst(ndo, src, dst, s->proto); if (vflag > 1) { uint64_t packets[2]; uint64_t bytes[2]; uint32_t creation = ntohl(s->creation); uint32_t expire = ntohl(s->expire); sec = creation % 60; creation /= 60; min = creation % 60; creation /= 60; - printf("\n\tage %.2u:%.2u:%.2u", creation, min, sec); + ND_PRINT((ndo, "\n\tage %.2u:%.2u:%.2u", creation, min, sec)); sec = expire % 60; expire /= 60; min = expire % 60; expire /= 60; - printf(", expires in %.2u:%.2u:%.2u", expire, min, sec); + ND_PRINT((ndo, ", expires in %.2u:%.2u:%.2u", expire, min, sec)); bcopy(s->packets[0], &packets[0], sizeof(uint64_t)); bcopy(s->packets[1], &packets[1], sizeof(uint64_t)); bcopy(s->bytes[0], &bytes[0], sizeof(uint64_t)); bcopy(s->bytes[1], &bytes[1], sizeof(uint64_t)); - printf(", %ju:%ju pkts, %ju:%ju bytes", + ND_PRINT((ndo, ", %ju:%ju pkts, %ju:%ju bytes", be64toh(packets[0]), be64toh(packets[1]), - be64toh(bytes[0]), be64toh(bytes[1])); + be64toh(bytes[0]), be64toh(bytes[1]))); if (s->anchor != ntohl(-1)) - printf(", anchor %u", ntohl(s->anchor)); + ND_PRINT((ndo, ", anchor %u", ntohl(s->anchor))); if (s->rule != ntohl(-1)) - printf(", rule %u", ntohl(s->rule)); + ND_PRINT((ndo, ", rule %u", ntohl(s->rule))); } if (vflag > 1) { uint64_t id; bcopy(&s->id, &id, sizeof(uint64_t)); - printf("\n\tid: %016jx creatorid: %08x", - (uintmax_t )be64toh(id), ntohl(s->creatorid)); + ND_PRINT((ndo, "\n\tid: %016jx creatorid: %08x", + (uintmax_t )be64toh(id), ntohl(s->creatorid))); } } Index: user/ngie/more-tests/contrib/tcpdump =================================================================== --- user/ngie/more-tests/contrib/tcpdump (revision 281974) +++ user/ngie/more-tests/contrib/tcpdump (revision 281975) Property changes on: user/ngie/more-tests/contrib/tcpdump ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/contrib/tcpdump:r281414-281974 Index: user/ngie/more-tests/lib/libc/locale/duplocale.3 =================================================================== --- user/ngie/more-tests/lib/libc/locale/duplocale.3 (revision 281974) +++ user/ngie/more-tests/lib/libc/locale/duplocale.3 (revision 281975) @@ -1,79 +1,79 @@ .\" Copyright (c) 2011 The FreeBSD Foundation .\" All rights reserved. .\" .\" This documentation was written by David Chisnall under sponsorship from .\" the FreeBSD Foundation. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" .Dd September 17, 2011 .Dt DUPLOCALE 3 .Os .Sh NAME .Nm duplocale .Nd duplicate an locale .Sh LIBRARY .Lb libc .Sh SYNOPSIS -.In xlocale.h +.In locale.h .Ft locale_t .Fn duplocale "locale_t locale" .Sh DESCRIPTION Duplicates an existing .Fa locale_t returning a new .Fa locale_t that refers to the same locale values but has an independent internal state. Various functions, such as .Xr mblen 3 require a persistent state. These functions formerly used static variables and calls to them from multiple threads had undefined behavior. They now use fields in the .Fa locale_t associated with the current thread by .Xr uselocale 3 . These calls are therefore only thread safe on threads with a unique per-thread locale. The locale returned by this call must be freed with .Xr freelocale 3 . .Sh SEE ALSO .Xr freelocale 3 , .Xr localeconv 3 , .Xr newlocale 3 , .Xr querylocale 3 , .Xr uselocale 3 , .Xr xlocale 3 .Sh STANDARDS This function conforms to .St -p1003.1-2008 . .Sh BUGS Ideally, .Xr uselocale 3 should make a copy of the .Fa locale_t implicitly to ensure thread safety, and a copy of the global locale should be installed lazily on each thread. The FreeBSD implementation does not do this, for compatibility with Darwin. Index: user/ngie/more-tests/lib/libc/locale/freelocale.3 =================================================================== --- user/ngie/more-tests/lib/libc/locale/freelocale.3 (revision 281974) +++ user/ngie/more-tests/lib/libc/locale/freelocale.3 (revision 281975) @@ -1,68 +1,68 @@ .\" Copyright (c) 2011 The FreeBSD Foundation .\" All rights reserved. .\" .\" This documentation was written by David Chisnall under sponsorship from .\" the FreeBSD Foundation. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .Dd September 17, 2011 .Dt FREELOCALE 3 .Os .Sh NAME .Nm freelocale .Nd Frees a locale created with .Xr duplocale 3 or .Xr newlocale 3 .Sh LIBRARY .Lb libc .Sh SYNOPSIS -.In xlocale.h +.In locale.h .Ft int .Fn freelocale "locale_t locale" .Sh DESCRIPTION Frees a .Fa locale_t . This relinquishes any resources held exclusively by this locale. Note that locales share reference-counted components, so a call to this function is not guaranteed to free all of the components. .Sh RETURN VALUES Returns 0 on success or -1 on error. .Sh SEE ALSO .Xr duplocale 3 , .Xr localeconv 3 , .Xr newlocale 3 , .Xr querylocale 3 , .Xr uselocale 3 , .Xr xlocale 3 .Sh STANDARDS The .Fn freelocale function differs from .St -p1003.1-2008 in that its return type is .Vt int rather than .Vt void . Index: user/ngie/more-tests/lib/libc/locale/newlocale.3 =================================================================== --- user/ngie/more-tests/lib/libc/locale/newlocale.3 (revision 281974) +++ user/ngie/more-tests/lib/libc/locale/newlocale.3 (revision 281975) @@ -1,112 +1,112 @@ .\" Copyright (c) 2011 The FreeBSD Foundation .\" All rights reserved. .\" .\" This documentation was written by David Chisnall under sponsorship from .\" the FreeBSD Foundation. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .Dd September 17, 2011 .Dt NEWLOCALE 3 .Os .Sh NAME .Nm newlocale .Nd Creates a new locale .Sh LIBRARY .Lb libc .Sh SYNOPSIS -.In xlocale +.In locale.h .Ft locale_t .Fn newlocale "int mask" "const char * locale" "locale_t base" .Sh DESCRIPTION Creates a new locale, inheriting some properties from an existing locale. The .Fa mask defines the components that the new locale will have set to the locale with the name specified in the .Fa locale parameter. Any other components will be inherited from .Fa base . The .Fa mask is either .Fa LC_ALL_MASK , indicating all possible locale components, or the logical OR of some combination of the following: .Bl -tag -width "LC_MESSAGES_MASK" -offset indent .It LC_COLLATE_MASK The locale for string collation routines. This controls alphabetic ordering in .Xr strcoll 3 and .Xr strxfrm 3 . .It LC_CTYPE_MASK The locale for the .Xr ctype 3 and .Xr multibyte 3 functions. This controls recognition of upper and lower case, alphabetic or non-alphabetic characters, and so on. .It LC_MESSAGES_MASK Set a locale for message catalogs, see .Xr catopen 3 function. .It LC_MONETARY_MASK Set a locale for formatting monetary values; this affects the .Xr localeconv 3 function. .It LC_NUMERIC_MASK Set a locale for formatting numbers. This controls the formatting of decimal points in input and output of floating point numbers in functions such as .Xr printf 3 and .Xr scanf 3 , as well as values returned by .Xr localeconv 3 . .It LC_TIME_MASK Set a locale for formatting dates and times using the .Xr strftime 3 function. .El This function uses the same rules for loading locale components as .Xr setlocale 3 . .Sh RETURN VALUES Returns a new, valid, .Fa locale_t or NULL if an error occurs. You must free the returned locale with .Xr freelocale 3 . .Sh SEE ALSO .Xr duplocale 3 , .Xr freelocale 3 , .Xr localeconv 3 , .Xr querylocale 3 , .Xr uselocale 3 , .Xr xlocale 3 .Sh STANDARDS This function conforms to .St -p1003.1-2008 . Index: user/ngie/more-tests/lib/libc/locale/none.c =================================================================== --- user/ngie/more-tests/lib/libc/locale/none.c (revision 281974) +++ user/ngie/more-tests/lib/libc/locale/none.c (revision 281975) @@ -1,222 +1,222 @@ /*- * Copyright (c) 2002-2004 Tim J. Robbins. All rights reserved. * Copyright (c) 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Paul Borman at Krystal Technologies. * * Copyright (c) 2011 The FreeBSD Foundation * All rights reserved. * Portions of this software were developed by David Chisnall * under sponsorship from the FreeBSD Foundation. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char sccsid[] = "@(#)none.c 8.1 (Berkeley) 6/4/93"; #endif /* LIBC_SCCS and not lint */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include "mblocal.h" static size_t _none_mbrtowc(wchar_t * __restrict, const char * __restrict, size_t, mbstate_t * __restrict); static int _none_mbsinit(const mbstate_t *); static size_t _none_mbsnrtowcs(wchar_t * __restrict dst, const char ** __restrict src, size_t nms, size_t len, mbstate_t * __restrict ps __unused); static size_t _none_wcrtomb(char * __restrict, wchar_t, mbstate_t * __restrict); static size_t _none_wcsnrtombs(char * __restrict, const wchar_t ** __restrict, size_t, size_t, mbstate_t * __restrict); /* setup defaults */ int __mb_cur_max = 1; int __mb_sb_limit = 256; /* Expected to be <= _CACHED_RUNES */ int _none_init(struct xlocale_ctype *l, _RuneLocale *rl) { l->__mbrtowc = _none_mbrtowc; l->__mbsinit = _none_mbsinit; l->__mbsnrtowcs = _none_mbsnrtowcs; l->__wcrtomb = _none_wcrtomb; l->__wcsnrtombs = _none_wcsnrtombs; l->runes = rl; l->__mb_cur_max = 1; l->__mb_sb_limit = 256; return(0); } static int _none_mbsinit(const mbstate_t *ps __unused) { /* * Encoding is not state dependent - we are always in the * initial state. */ return (1); } static size_t _none_mbrtowc(wchar_t * __restrict pwc, const char * __restrict s, size_t n, mbstate_t * __restrict ps __unused) { if (s == NULL) /* Reset to initial shift state (no-op) */ return (0); if (n == 0) /* Incomplete multibyte sequence */ return ((size_t)-2); if (pwc != NULL) *pwc = (unsigned char)*s; return (*s == '\0' ? 0 : 1); } static size_t _none_wcrtomb(char * __restrict s, wchar_t wc, mbstate_t * __restrict ps __unused) { if (s == NULL) /* Reset to initial shift state (no-op) */ return (1); if (wc < 0 || wc > UCHAR_MAX) { errno = EILSEQ; return ((size_t)-1); } *s = (unsigned char)wc; return (1); } static size_t _none_mbsnrtowcs(wchar_t * __restrict dst, const char ** __restrict src, size_t nms, size_t len, mbstate_t * __restrict ps __unused) { const char *s; size_t nchr; if (dst == NULL) { s = memchr(*src, '\0', nms); return (s != NULL ? s - *src : nms); } s = *src; nchr = 0; while (len-- > 0 && nms-- > 0) { if ((*dst++ = (unsigned char)*s++) == L'\0') { *src = NULL; return (nchr); } nchr++; } *src = s; return (nchr); } static size_t _none_wcsnrtombs(char * __restrict dst, const wchar_t ** __restrict src, size_t nwc, size_t len, mbstate_t * __restrict ps __unused) { const wchar_t *s; size_t nchr; if (dst == NULL) { for (s = *src; nwc > 0 && *s != L'\0'; s++, nwc--) { if (*s < 0 || *s > UCHAR_MAX) { errno = EILSEQ; return ((size_t)-1); } } return (s - *src); } s = *src; nchr = 0; while (len-- > 0 && nwc-- > 0) { if (*s < 0 || *s > UCHAR_MAX) { errno = EILSEQ; return ((size_t)-1); } if ((*dst++ = *s++) == '\0') { *src = NULL; return (nchr); } nchr++; } *src = s; return (nchr); } /* setup defaults */ size_t (*__mbrtowc)(wchar_t * __restrict, const char * __restrict, size_t, mbstate_t * __restrict) = _none_mbrtowc; int (*__mbsinit)(const mbstate_t *) = _none_mbsinit; size_t (*__mbsnrtowcs)(wchar_t * __restrict, const char ** __restrict, size_t, size_t, mbstate_t * __restrict) = _none_mbsnrtowcs; size_t (*__wcrtomb)(char * __restrict, wchar_t, mbstate_t * __restrict) = _none_wcrtomb; size_t (*__wcsnrtombs)(char * __restrict, const wchar_t ** __restrict, size_t, size_t, mbstate_t * __restrict) = _none_wcsnrtombs; struct xlocale_ctype __xlocale_global_ctype = { {{0}, "C"}, (_RuneLocale*)&_DefaultRuneLocale, _none_mbrtowc, _none_mbsinit, _none_mbsnrtowcs, _none_wcrtomb, _none_wcsnrtombs, 1, /* __mb_cur_max, */ 256 /* __mb_sb_limit */ }; -const struct xlocale_ctype __xlocale_C_ctype = { +struct xlocale_ctype __xlocale_C_ctype = { {{0}, "C"}, (_RuneLocale*)&_DefaultRuneLocale, _none_mbrtowc, _none_mbsinit, _none_mbsnrtowcs, _none_wcrtomb, _none_wcsnrtombs, 1, /* __mb_cur_max, */ 256 /* __mb_sb_limit */ }; Index: user/ngie/more-tests/lib/libc/locale/querylocale.3 =================================================================== --- user/ngie/more-tests/lib/libc/locale/querylocale.3 (revision 281974) +++ user/ngie/more-tests/lib/libc/locale/querylocale.3 (revision 281975) @@ -1,54 +1,54 @@ .\" Copyright (c) 2011 The FreeBSD Foundation .\" All rights reserved. .\" .\" This documentation was written by David Chisnall under sponsorship from .\" the FreeBSD Foundation. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" .Dd May 3, 2013 .Dt QUERYLOCALE 3 .Os .Sh NAME .Nm querylocale .Nd Look up the locale name for a specified category .Sh LIBRARY .Lb libc .Sh SYNOPSIS -.In xlocale.h +.In locale.h .Ft const char * .Fn querylocale "int mask" "locale_t locale" .Sh DESCRIPTION Returns the name of the locale for the category specified by .Fa mask . This possible values for the mask are the same as those in .Xr newlocale 3 . If more than one bit in the mask is set, the returned value is undefined. .Sh SEE ALSO .Xr duplocale 3 , .Xr freelocale 3 , .Xr localeconv 3 , .Xr newlocale 3 , .Xr uselocale 3 , .Xr xlocale 3 Index: user/ngie/more-tests/lib/libc/locale/uselocale.3 =================================================================== --- user/ngie/more-tests/lib/libc/locale/uselocale.3 (revision 281974) +++ user/ngie/more-tests/lib/libc/locale/uselocale.3 (revision 281975) @@ -1,60 +1,60 @@ .\" Copyright (c) 2011 The FreeBSD Foundation .\" All rights reserved. .\" .\" This documentation was written by David Chisnall under sponsorship from .\" the FreeBSD Foundation. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" .Dd September 17, 2011 .Dt USELOCALE 3 .Os .Sh NAME .Nm uselocale .Nd Sets a thread-local locale .Sh LIBRARY .Lb libc .Sh SYNOPSIS -.In xlocale.h +.In locale.h .Ft locale_t .Fn uselocale "locale_t locale" .Sh DESCRIPTION Specifies the locale for this thread to use. Specifying .Fa LC_GLOBAL_LOCALE disables the per-thread locale, while NULL returns the current locale without setting a new one. .Sh RETURN VALUES Returns the previous locale, or LC_GLOBAL_LOCALE if this thread has no locale associated with it. .Sh SEE ALSO .Xr duplocale 3 , .Xr freelocale 3 , .Xr localeconv 3 , .Xr newlocale 3 , .Xr querylocale 3 , .Xr xlocale 3 .Sh STANDARDS This function conforms to .St -p1003.1-2008 . Index: user/ngie/more-tests/lib/libc =================================================================== --- user/ngie/more-tests/lib/libc (revision 281974) +++ user/ngie/more-tests/lib/libc (revision 281975) Property changes on: user/ngie/more-tests/lib/libc ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/lib/libc:r281925-281974 Index: user/ngie/more-tests/lib/libmd/mdXhl.c =================================================================== --- user/ngie/more-tests/lib/libmd/mdXhl.c (revision 281974) +++ user/ngie/more-tests/lib/libmd/mdXhl.c (revision 281975) @@ -1,98 +1,98 @@ /* mdXhl.c * ---------------------------------------------------------------------------- * "THE BEER-WARE LICENSE" (Revision 42): * wrote this file. As long as you retain this notice you * can do whatever you want with this stuff. If we meet some day, and you think * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp * ---------------------------------------------------------------------------- */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include "mdX.h" char * MDXEnd(MDX_CTX *ctx, char *buf) { int i; unsigned char digest[LENGTH]; static const char hex[]="0123456789abcdef"; if (!buf) buf = malloc(2*LENGTH + 1); if (!buf) return 0; MDXFinal(digest, ctx); for (i = 0; i < LENGTH; i++) { buf[i+i] = hex[digest[i] >> 4]; buf[i+i+1] = hex[digest[i] & 0x0f]; } buf[i+i] = '\0'; return buf; } char * MDXFile(const char *filename, char *buf) { return (MDXFileChunk(filename, buf, 0, 0)); } char * MDXFileChunk(const char *filename, char *buf, off_t ofs, off_t len) { unsigned char buffer[BUFSIZ]; MDX_CTX ctx; struct stat stbuf; int f, i, e; off_t n; MDXInit(&ctx); f = open(filename, O_RDONLY); if (f < 0) return 0; if (fstat(f, &stbuf) < 0) return 0; if (ofs > stbuf.st_size) ofs = stbuf.st_size; if ((len == 0) || (len > stbuf.st_size - ofs)) len = stbuf.st_size - ofs; if (lseek(f, ofs, SEEK_SET) < 0) return 0; n = len; i = 0; while (n > 0) { if (n > sizeof(buffer)) i = read(f, buffer, sizeof(buffer)); else i = read(f, buffer, n); - if (i < 0) + if (i <= 0) break; MDXUpdate(&ctx, buffer, i); n -= i; } e = errno; close(f); errno = e; if (i < 0) return 0; return (MDXEnd(&ctx, buf)); } char * MDXData (const void *data, unsigned int len, char *buf) { MDX_CTX ctx; MDXInit(&ctx); MDXUpdate(&ctx,data,len); return (MDXEnd(&ctx, buf)); } Index: user/ngie/more-tests/sbin/camcontrol/camcontrol.8 =================================================================== --- user/ngie/more-tests/sbin/camcontrol/camcontrol.8 (revision 281974) +++ user/ngie/more-tests/sbin/camcontrol/camcontrol.8 (revision 281975) @@ -1,2103 +1,2102 @@ .\" .\" Copyright (c) 1998, 1999, 2000, 2002, 2005, 2006, 2007 Kenneth D. Merry. .\" All rights reserved. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" 3. The name of the author may not be used to endorse or promote products .\" derived from this software without specific prior written permission. .\" .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" .Dd March 19, 2015 .Dt CAMCONTROL 8 .Os .Sh NAME .Nm camcontrol .Nd CAM control program .Sh SYNOPSIS .Nm .Aq Ar command .Op device id .Op generic args .Op command args .Nm .Ic devlist .Op Fl b .Op Fl v .Nm .Ic periphlist .Op device id .Op Fl n Ar dev_name .Op Fl u Ar unit_number .Nm .Ic tur .Op device id .Op generic args .Nm .Ic inquiry .Op device id .Op generic args .Op Fl D .Op Fl S .Op Fl R .Nm .Ic identify .Op device id .Op generic args .Op Fl v .Nm .Ic reportluns .Op device id .Op generic args .Op Fl c .Op Fl l .Op Fl r Ar reporttype .Nm .Ic readcap .Op device id .Op generic args .Op Fl b .Op Fl h .Op Fl H .Op Fl N .Op Fl q .Op Fl s .Nm .Ic start .Op device id .Op generic args .Nm .Ic stop .Op device id .Op generic args .Nm .Ic load .Op device id .Op generic args .Nm .Ic eject .Op device id .Op generic args .Nm .Ic rescan .Aq all | bus Ns Op :target:lun .Nm .Ic reset .Aq all | bus Ns Op :target:lun .Nm .Ic defects .Op device id .Op generic args .Aq Fl f Ar format .Op Fl P .Op Fl G .Op Fl q .Op Fl s .Op Fl S Ar offset .Op Fl X .Nm .Ic modepage .Op device id .Op generic args .Aq Fl m Ar page | Fl l .Op Fl P Ar pgctl .Op Fl b | Fl e .Op Fl d .Nm .Ic cmd .Op device id .Op generic args .Aq Fl a Ar cmd Op args .Aq Fl c Ar cmd Op args .Op Fl d .Op Fl f .Op Fl i Ar len Ar fmt .Bk -words .Op Fl o Ar len Ar fmt Op args .Op Fl r Ar fmt .Ek .Nm .Ic smpcmd .Op device id .Op generic args .Aq Fl r Ar len Ar fmt Op args .Aq Fl R Ar len Ar fmt Op args .Nm .Ic smprg .Op device id .Op generic args .Op Fl l .Nm .Ic smppc .Op device id .Op generic args .Aq Fl p Ar phy .Op Fl l .Op Fl o Ar operation .Op Fl d Ar name .Op Fl m Ar rate .Op Fl M Ar rate .Op Fl T Ar pp_timeout .Op Fl a Ar enable|disable .Op Fl A Ar enable|disable .Op Fl s Ar enable|disable .Op Fl S Ar enable|disable .Nm .Ic smpphylist .Op device id .Op generic args .Op Fl l .Op Fl q .Nm .Ic smpmaninfo .Op device id .Op generic args .Op Fl l .Nm .Ic debug .Op Fl I .Op Fl P .Op Fl T .Op Fl S .Op Fl X .Op Fl c .Op Fl p .Aq all|off|bus Ns Op :target Ns Op :lun .Nm .Ic tags .Op device id .Op generic args .Op Fl N Ar tags .Op Fl q .Op Fl v .Nm .Ic negotiate .Op device id .Op generic args .Op Fl c .Op Fl D Ar enable|disable .Op Fl M Ar mode .Op Fl O Ar offset .Op Fl q .Op Fl R Ar syncrate .Op Fl T Ar enable|disable .Op Fl U .Op Fl W Ar bus_width .Op Fl v .Nm .Ic format .Op device id .Op generic args .Op Fl q .Op Fl r .Op Fl w .Op Fl y .Nm .Ic sanitize .Op device id .Op generic args .Aq Fl a Ar overwrite | block | crypto | exitfailure .Op Fl c Ar passes .Op Fl I .Op Fl P Ar pattern .Op Fl q .Op Fl U .Op Fl r .Op Fl w .Op Fl y .Nm .Ic idle .Op device id .Op generic args .Op Fl t Ar time .Nm .Ic standby .Op device id .Op generic args .Op Fl t Ar time .Nm .Ic sleep .Op device id .Op generic args .Nm .Ic apm .Op device id .Op generic args .Op Fl l Ar level .Nm .Ic aam .Op device id .Op generic args .Op Fl l Ar level .Nm .Ic fwdownload .Op device id .Op generic args .Aq Fl f Ar fw_image .Op Fl y .Op Fl s .Nm .Ic security .Op device id .Op generic args .Op Fl d Ar pwd .Op Fl e Ar pwd .Op Fl f .Op Fl h Ar pwd .Op Fl k Ar pwd .Op Fl l Ar high|maximum .Op Fl q .Op Fl s Ar pwd .Op Fl T Ar timeout .Op Fl U Ar user|master .Op Fl y .Nm .Ic hpa .Op device id .Op generic args .Op Fl f .Op Fl l .Op Fl P .Op Fl p Ar pwd .Op Fl q .Op Fl s Ar max_sectors .Op Fl U Ar pwd .Op Fl y .Nm .Ic persist .Op device id .Op generic args .Aq Fl i Ar action | Fl o Ar action .Op Fl a .Op Fl I Ar trans_id .Op Fl k Ar key .Op Fl K Ar sa_key .Op Fl p .Op Fl R Ar rel_tgt_port .Op Fl s Ar scope .Op Fl S .Op Fl T Ar res_type .Op Fl U .Nm .Ic help .Sh DESCRIPTION The .Nm utility is designed to provide a way for users to access and control the .Fx CAM subsystem. .Pp The .Nm utility can cause a loss of data and/or system crashes if used improperly. Even expert users are encouraged to exercise caution when using this command. Novice users should stay away from this utility. .Pp The .Nm utility has a number of primary functions, many of which support an optional device identifier. A device identifier can take one of three forms: .Bl -tag -width 14n .It deviceUNIT Specify a device name and unit number combination, like "da5" or "cd3". .It bus:target Specify a bus number and target id. The bus number can be determined from the output of .Dq camcontrol devlist . The lun defaults to 0. .It bus:target:lun Specify the bus, target and lun for a device. (e.g.\& 1:2:0) .El .Pp The device identifier, if it is specified, .Em must come immediately after the function name, and before any generic or function-specific arguments. Note that the .Fl n and .Fl u arguments described below will override any device name or unit number specified beforehand. The .Fl n and .Fl u arguments will .Em not override a specified bus:target or bus:target:lun, however. .Pp Most of the .Nm primary functions support these generic arguments: .Bl -tag -width 14n .It Fl C Ar count SCSI command retry count. In order for this to work, error recovery .Pq Fl E must be turned on. .It Fl E Instruct the kernel to perform generic SCSI error recovery for the given command. This is needed in order for the retry count .Pq Fl C to be honored. Other than retrying commands, the generic error recovery in the code will generally attempt to spin up drives that are not spinning. It may take some other actions, depending upon the sense code returned from the command. .It Fl n Ar dev_name Specify the device type to operate on, e.g.\& "da", "cd". .It Fl t Ar timeout SCSI command timeout in seconds. This overrides the default timeout for any given command. .It Fl u Ar unit_number Specify the device unit number, e.g.\& "1", "5". .It Fl v Be verbose, print out sense information for failed SCSI commands. .El .Pp Primary command functions: .Bl -tag -width periphlist .It Ic devlist List all physical devices (logical units) attached to the CAM subsystem. This also includes a list of peripheral drivers attached to each device. With the .Fl v argument, SCSI bus number, adapter name and unit numbers are printed as well. On the other hand, with the .Fl b argument, only the bus adapter, and unit information will be printed, and device information will be omitted. .It Ic periphlist List all peripheral drivers attached to a given physical device (logical unit). .It Ic tur Send the SCSI test unit ready (0x00) command to the given device. The .Nm utility will report whether the device is ready or not. .It Ic inquiry Send a SCSI inquiry command (0x12) to a device. By default, .Nm will print out the standard inquiry data, device serial number, and transfer rate information. The user can specify that only certain types of inquiry data be printed: .Bl -tag -width 4n .It Fl D Get the standard inquiry data. .It Fl S Print out the serial number. If this flag is the only one specified, .Nm will not print out "Serial Number" before the value returned by the drive. This is to aid in script writing. .It Fl R Print out transfer rate information. .El .It Ic identify Send a ATA identify command (0xec) to a device. .It Ic reportluns Send the SCSI REPORT LUNS (0xA0) command to the given device. By default, .Nm will print out the list of logical units (LUNs) supported by the target device. There are a couple of options to modify the output: .Bl -tag -width 14n .It Fl c Just print out a count of LUNs, not the actual LUN numbers. .It Fl l Just print out the LUNs, and do not print out the count. .It Fl r Ar reporttype Specify the type of report to request from the target: .Bl -tag -width 012345678 .It default Return the default report. This is the .Nm default. Most targets will support this report if they support the REPORT LUNS command. .It wellknown Return only well known LUNs. .It all Return all available LUNs. .El .El .Pp .Nm will try to print out LUN numbers in a reasonable format. It can understand the peripheral, flat, LUN and extended LUN formats. .It Ic readcap Send the SCSI READ CAPACITY command to the given device and display the results. If the device is larger than 2TB, the SCSI READ CAPACITY (16) service action will be sent to obtain the full size of the device. By default, .Nm will print out the last logical block of the device, and the blocksize of the device in bytes. To modify the output format, use the following options: .Bl -tag -width 5n .It Fl b Just print out the blocksize, not the last block or device size. This cannot be used with .Fl N or .Fl s . .It Fl h Print out the device size in human readable (base 2, 1K == 1024) format. This implies .Fl N and cannot be used with .Fl q or .Fl b . .It Fl H Print out the device size in human readable (base 10, 1K == 1000) format. .It Fl N Print out the number of blocks in the device instead of the last logical block. .It Fl q Quiet, print out the numbers only (separated by a comma if .Fl b or .Fl s are not specified). .It Fl s Print out the last logical block or the size of the device only, and omit the blocksize. .El .It Ic start Send the SCSI Start/Stop Unit (0x1B) command to the given device with the start bit set. .It Ic stop Send the SCSI Start/Stop Unit (0x1B) command to the given device with the start bit cleared. .It Ic load Send the SCSI Start/Stop Unit (0x1B) command to the given device with the start bit set and the load/eject bit set. .It Ic eject Send the SCSI Start/Stop Unit (0x1B) command to the given device with the start bit cleared and the load/eject bit set. .It Ic rescan Tell the kernel to scan all busses in the system (with the .Ar all argument), the given bus (XPT_SCAN_BUS), or bus:target:lun (XPT_SCAN_LUN) for new devices or devices that have gone away. The user may specify a scan of all busses, a single bus, or a lun. Scanning all luns on a target is not supported. .It Ic reset Tell the kernel to reset all busses in the system (with the .Ar all argument) or the given bus (XPT_RESET_BUS) by issuing a SCSI bus reset for that bus, or to reset the given bus:target:lun (XPT_RESET_DEV), typically by issuing a BUS DEVICE RESET message after connecting to that device. Note that this can have a destructive impact on the system. .It Ic defects Send the .Tn SCSI READ DEFECT DATA (10) command (0x37) or the .Tn SCSI READ DEFECT DATA (12) command (0xB7) to the given device, and print out any combination of: the total number of defects, the primary defect list (PLIST), and the grown defect list (GLIST). .Bl -tag -width 11n .It Fl f Ar format Specify the requested format of the defect list. The format argument is required. Most drives support the physical sector format. Some drives support the logical block format. Many drives, if they do not support the requested format, return the data in an alternate format, along with sense information indicating that the requested data format is not supported. The .Nm utility attempts to detect this, and print out whatever format the drive returns. If the drive uses a non-standard sense code to report that it does not support the requested format, .Nm will probably see the error as a failure to complete the request. .Pp The format options are: .Bl -tag -width 9n .It block Print out the list as logical blocks. This is limited to 32-bit block sizes, and isn't supported by many modern drives. .It longblock Print out the list as logical blocks. This option uses a 64-bit block size. .It bfi Print out the list in bytes from index format. .It extbfi Print out the list in extended bytes from index format. The extended format allows for ranges of blocks to be printed. .It phys Print out the list in physical sector format. Most drives support this format. .It extphys Print out the list in extended physical sector format. The extended format allows for ranges of blocks to be printed. .El -.Pp .It Fl G Print out the grown defect list. This is a list of bad blocks that have been remapped since the disk left the factory. .It Fl P Print out the primary defect list. This is the list of defects that were present in the factory. .It Fl q When printing status information with .Fl s , only print the number of defects. .It Fl s Just print the number of defects, not the list of defects. .It Fl S Ar offset Specify the starting offset into the defect list. This implies using the .Tn SCSI READ DEFECT DATA (12) command, as the 10 byte version of the command doesn't support the address descriptor index field. Not all drives support the 12 byte command, and some drives that support the 12 byte command don't support the address descriptor index field. .It Fl X Print out defects in hexadecimal (base 16) form instead of base 10 form. .El .Pp If neither .Fl P nor .Fl G is specified, .Nm will print out the number of defects given in the READ DEFECT DATA header returned from the drive. Some drives will report 0 defects if neither the primary or grown defect lists are requested. .It Ic modepage Allows the user to display and optionally edit a SCSI mode page. The mode page formats are located in .Pa /usr/share/misc/scsi_modes . This can be overridden by specifying a different file in the .Ev SCSI_MODES environment variable. The .Ic modepage command takes several arguments: .Bl -tag -width 12n .It Fl d Disable block descriptors for mode sense. .It Fl b Displays mode page data in binary format. .It Fl e This flag allows the user to edit values in the mode page. The user may either edit mode page values with the text editor pointed to by his .Ev EDITOR environment variable, or supply mode page values via standard input, using the same format that .Nm uses to display mode page values. The editor will be invoked if .Nm detects that standard input is terminal. .It Fl l Lists all available mode pages. .It Fl m Ar mode_page This specifies the number of the mode page the user would like to view and/or edit. This argument is mandatory unless .Fl l is specified. .It Fl P Ar pgctl This allows the user to specify the page control field. Possible values are: .Bl -tag -width xxx -compact .It 0 Current values .It 1 Changeable values .It 2 Default values .It 3 Saved values .El .El .It Ic cmd Allows the user to send an arbitrary ATA or SCSI CDB to any device. The .Ic cmd function requires the .Fl c argument to specify SCSI CDB or the .Fl a argument to specify ATA Command Block registers values. Other arguments are optional, depending on the command type. The command and data specification syntax is documented in .Xr cam_cdbparse 3 . NOTE: If the CDB specified causes data to be transferred to or from the SCSI device in question, you MUST specify either .Fl i or .Fl o . .Bl -tag -width 17n .It Fl a Ar cmd Op args This specifies the content of 12 ATA Command Block registers (command, features, lba_low, lba_mid, lba_high, device, lba_low_exp, lba_mid_exp. lba_high_exp, features_exp, sector_count, sector_count_exp). .It Fl c Ar cmd Op args This specifies the SCSI CDB. SCSI CDBs may be 6, 10, 12 or 16 bytes. .It Fl d Specifies DMA protocol to be used for ATA command. .It Fl f Specifies FPDMA (NCQ) protocol to be used for ATA command. .It Fl i Ar len Ar fmt This specifies the amount of data to read, and how it should be displayed. If the format is .Sq - , .Ar len bytes of data will be read from the device and written to standard output. .It Fl o Ar len Ar fmt Op args This specifies the amount of data to be written to a device, and the data that is to be written. If the format is .Sq - , .Ar len bytes of data will be read from standard input and written to the device. .It Fl r Ar fmt This specifies that 11 result ATA Command Block registers should be displayed (status, error, lba_low, lba_mid, lba_high, device, lba_low_exp, lba_mid_exp, lba_high_exp, sector_count, sector_count_exp), and how. If the format is .Sq - , 11 result registers will be written to standard output in hex. .El .It Ic smpcmd Allows the user to send an arbitrary Serial Management Protocol (SMP) command to a device. The .Ic smpcmd function requires the .Fl r argument to specify the SMP request to be sent, and the .Fl R argument to specify the format of the SMP response. The syntax for the SMP request and response arguments is documented in .Xr cam_cdbparse 3 . .Pp Note that SAS adapters that support SMP passthrough (at least the currently known adapters) do not accept CRC bytes from the user in the request and do not pass CRC bytes back to the user in the response. Therefore users should not include the CRC bytes in the length of the request and not expect CRC bytes to be returned in the response. .Bl -tag -width 17n .It Fl r Ar len Ar fmt Op args This specifies the size of the SMP request, without the CRC bytes, and the SMP request format. If the format is .Sq - , .Ar len bytes of data will be read from standard input and written as the SMP request. .It Fl R Ar len Ar fmt Op args This specifies the size of the buffer allocated for the SMP response, and the SMP response format. If the format is .Sq - , .Ar len bytes of data will be allocated for the response and the response will be written to standard output. .El .It Ic smprg Allows the user to send the Serial Management Protocol (SMP) Report General command to a device. .Nm will display the data returned by the Report General command. If the SMP target supports the long response format, the additional data will be requested and displayed automatically. .Bl -tag -width 8n .It Fl l Request the long response format only. Not all SMP targets support the long response format. This option causes .Nm to skip sending the initial report general request without the long bit set and only issue a report general request with the long bit set. .El .It Ic smppc Allows the user to issue the Serial Management Protocol (SMP) PHY Control command to a device. This function should be used with some caution, as it can render devices inaccessible, and could potentially cause data corruption as well. The .Fl p argument is required to specify the PHY to operate on. .Bl -tag -width 17n .It Fl p Ar phy Specify the PHY to operate on. This argument is required. .It Fl l Request the long request/response format. Not all SMP targets support the long response format. For the PHY Control command, this currently only affects whether the request length is set to a value other than 0. .It Fl o Ar operation Specify a PHY control operation. Only one .Fl o operation may be specified. The operation may be specified numerically (in decimal, hexadecimal, or octal) or one of the following operation names may be specified: .Bl -tag -width 16n .It nop No operation. It is not necessary to specify this argument. .It linkreset Send the LINK RESET command to the phy. .It hardreset Send the HARD RESET command to the phy. .It disable Send the DISABLE command to the phy. Note that the LINK RESET or HARD RESET commands should re-enable the phy. .It clearerrlog Send the CLEAR ERROR LOG command. This clears the error log counters for the specified phy. .It clearaffiliation Send the CLEAR AFFILIATION command. This clears the affiliation from the STP initiator port with the same SAS address as the SMP initiator that requests the clear operation. .It sataportsel Send the TRANSMIT SATA PORT SELECTION SIGNAL command to the phy. This will cause a SATA port selector to use the given phy as its active phy and make the other phy inactive. .It clearitnl Send the CLEAR STP I_T NEXUS LOSS command to the PHY. .It setdevname Send the SET ATTACHED DEVICE NAME command to the PHY. This requires the .Fl d argument to specify the device name. .El .It Fl d Ar name Specify the attached device name. This option is needed with the .Fl o Ar setdevname phy operation. The name is a 64-bit number, and can be specified in decimal, hexadecimal or octal format. .It Fl m Ar rate Set the minimum physical link rate for the phy. This is a numeric argument. Currently known link rates are: .Bl -tag -width 5n .It 0x0 Do not change current value. .It 0x8 1.5 Gbps .It 0x9 3 Gbps .It 0xa 6 Gbps .El .Pp Other values may be specified for newer physical link rates. .It Fl M Ar rate Set the maximum physical link rate for the phy. This is a numeric argument. See the .Fl m argument description for known link rate arguments. .It Fl T Ar pp_timeout Set the partial pathway timeout value, in microseconds. See the .Tn ANSI .Tn SAS Protocol Layer (SPL) specification for more information on this field. .It Fl a Ar enable|disable Enable or disable SATA slumber phy power conditions. .It Fl A Ar enable|disable Enable or disable SATA partial power conditions. .It Fl s Ar enable|disable Enable or disable SAS slumber phy power conditions. .It Fl S Ar enable|disable Enable or disable SAS partial phy power conditions. .El .It Ic smpphylist List phys attached to a SAS expander, the address of the end device attached to the phy, and the inquiry data for that device and peripheral devices attached to that device. The inquiry data and peripheral devices are displayed if available. .Bl -tag -width 5n .It Fl l Turn on the long response format for the underlying SMP commands used for this command. .It Fl q Only print out phys that are attached to a device in the CAM EDT (Existing Device Table). .El .It Ic smpmaninfo Send the SMP Report Manufacturer Information command to the device and display the response. .Bl -tag -width 5n .It Fl l Turn on the long response format for the underlying SMP commands used for this command. .El .It Ic debug Turn on CAM debugging printfs in the kernel. This requires options CAMDEBUG in your kernel config file. WARNING: enabling debugging printfs currently causes an EXTREME number of kernel printfs. You may have difficulty turning off the debugging printfs once they start, since the kernel will be busy printing messages and unable to service other requests quickly. The .Ic debug function takes a number of arguments: .Bl -tag -width 18n .It Fl I Enable CAM_DEBUG_INFO printfs. .It Fl P Enable CAM_DEBUG_PERIPH printfs. .It Fl T Enable CAM_DEBUG_TRACE printfs. .It Fl S Enable CAM_DEBUG_SUBTRACE printfs. .It Fl X Enable CAM_DEBUG_XPT printfs. .It Fl c Enable CAM_DEBUG_CDB printfs. This will cause the kernel to print out the SCSI CDBs sent to the specified device(s). .It Fl p Enable CAM_DEBUG_PROBE printfs. .It all Enable debugging for all devices. .It off Turn off debugging for all devices .It bus Ns Op :target Ns Op :lun Turn on debugging for the given bus, target or lun. If the lun or target and lun are not specified, they are wildcarded. (i.e., just specifying a bus turns on debugging printfs for all devices on that bus.) .El .It Ic tags Show or set the number of "tagged openings" or simultaneous transactions we attempt to queue to a particular device. By default, the .Ic tags command, with no command-specific arguments (i.e., only generic arguments) prints out the "soft" maximum number of transactions that can be queued to the device in question. For more detailed information, use the .Fl v argument described below. .Bl -tag -width 7n .It Fl N Ar tags Set the number of tags for the given device. This must be between the minimum and maximum number set in the kernel quirk table. The default for most devices that support tagged queueing is a minimum of 2 and a maximum of 255. The minimum and maximum values for a given device may be determined by using the .Fl v switch. The meaning of the .Fl v switch for this .Nm subcommand is described below. .It Fl q Be quiet, and do not report the number of tags. This is generally used when setting the number of tags. .It Fl v The verbose flag has special functionality for the .Em tags argument. It causes .Nm to print out the tagged queueing related fields of the XPT_GDEV_TYPE CCB: .Bl -tag -width 13n .It dev_openings This is the amount of capacity for transactions queued to a given device. .It dev_active This is the number of transactions currently queued to a device. .It devq_openings This is the kernel queue space for transactions. This count usually mirrors dev_openings except during error recovery operations when the device queue is frozen (device is not allowed to receive commands), the number of dev_openings is reduced, or transaction replay is occurring. .It devq_queued This is the number of transactions waiting in the kernel queue for capacity on the device. This number is usually zero unless error recovery is in progress. .It held The held count is the number of CCBs held by peripheral drivers that have either just been completed or are about to be released to the transport layer for service by a device. Held CCBs reserve capacity on a given device. .It mintags This is the current "hard" minimum number of transactions that can be queued to a device at once. The .Ar dev_openings value above cannot go below this number. The default value for .Ar mintags is 2, although it may be set higher or lower for various devices. .It maxtags This is the "hard" maximum number of transactions that can be queued to a device at one time. The .Ar dev_openings value cannot go above this number. The default value for .Ar maxtags is 255, although it may be set higher or lower for various devices. .El .El .It Ic negotiate Show or negotiate various communication parameters. Some controllers may not support setting or changing some of these values. For instance, the Adaptec 174x controllers do not support changing a device's sync rate or offset. The .Nm utility will not attempt to set the parameter if the controller indicates that it does not support setting the parameter. To find out what the controller supports, use the .Fl v flag. The meaning of the .Fl v flag for the .Ic negotiate command is described below. Also, some controller drivers do not support setting negotiation parameters, even if the underlying controller supports negotiation changes. Some controllers, such as the Advansys wide controllers, support enabling and disabling synchronous negotiation for a device, but do not support setting the synchronous negotiation rate. .Bl -tag -width 17n .It Fl a Attempt to make the negotiation settings take effect immediately by sending a Test Unit Ready command to the device. .It Fl c Show or set current negotiation settings. This is the default. .It Fl D Ar enable|disable Enable or disable disconnection. .It Fl M Ar mode Set ATA mode. .It Fl O Ar offset Set the command delay offset. .It Fl q Be quiet, do not print anything. This is generally useful when you want to set a parameter, but do not want any status information. .It Fl R Ar syncrate Change the synchronization rate for a device. The sync rate is a floating point value specified in MHz. So, for instance, .Sq 20.000 is a legal value, as is .Sq 20 . .It Fl T Ar enable|disable Enable or disable tagged queueing for a device. .It Fl U Show or set user negotiation settings. The default is to show or set current negotiation settings. .It Fl v The verbose switch has special meaning for the .Ic negotiate subcommand. It causes .Nm to print out the contents of a Path Inquiry (XPT_PATH_INQ) CCB sent to the controller driver. .It Fl W Ar bus_width Specify the bus width to negotiate with a device. The bus width is specified in bits. The only useful values to specify are 8, 16, and 32 bits. The controller must support the bus width in question in order for the setting to take effect. .El .Pp In general, sync rate and offset settings will not take effect for a device until a command has been sent to the device. The .Fl a switch above will automatically send a Test Unit Ready to the device so negotiation parameters will take effect. .It Ic format Issue the .Tn SCSI FORMAT UNIT command to the named device. .Pp .Em WARNING! WARNING! WARNING! .Pp Low level formatting a disk will destroy ALL data on the disk. Use extreme caution when issuing this command. Many users low-level format disks that do not really need to be low-level formatted. There are relatively few scenarios that call for low-level formatting a disk. One reason for low-level formatting a disk is to initialize the disk after changing its physical sector size. Another reason for low-level formatting a disk is to revive the disk if you are getting "medium format corrupted" errors from the disk in response to read and write requests. .Pp Some disks take longer than others to format. Users should specify a timeout long enough to allow the format to complete. The default format timeout is 3 hours, which should be long enough for most disks. Some hard disks will complete a format operation in a very short period of time (on the order of 5 minutes or less). This is often because the drive does not really support the FORMAT UNIT command -- it just accepts the command, waits a few minutes and then returns it. .Pp The .Sq format subcommand takes several arguments that modify its default behavior. The .Fl q and .Fl y arguments can be useful for scripts. .Bl -tag -width 6n .It Fl q Be quiet, do not print any status messages. This option will not disable the questions, however. To disable questions, use the .Fl y argument, below. .It Fl r Run in .Dq report only mode. This will report status on a format that is already running on the drive. .It Fl w Issue a non-immediate format command. By default, .Nm issues the FORMAT UNIT command with the immediate bit set. This tells the device to immediately return the format command, before the format has actually completed. Then, .Nm gathers .Tn SCSI sense information from the device every second to determine how far along in the format process it is. If the .Fl w argument is specified, .Nm will issue a non-immediate format command, and will be unable to print any information to let the user know what percentage of the disk has been formatted. .It Fl y Do not ask any questions. By default, .Nm will ask the user if he/she really wants to format the disk in question, and also if the default format command timeout is acceptable. The user will not be asked about the timeout if a timeout is specified on the command line. .El .It Ic sanitize Issue the .Tn SCSI SANITIZE command to the named device. .Pp .Em WARNING! WARNING! WARNING! .Pp ALL data in the cache and on the disk will be destroyed or made inaccessible. Recovery of the data is not possible. Use extreme caution when issuing this command. .Pp The .Sq sanitize subcommand takes several arguments that modify its default behavior. The .Fl q and .Fl y arguments can be useful for scripts. .Bl -tag -width 6n .It Fl a Ar operation Specify the sanitize operation to perform. .Bl -tag -width 16n .It overwrite Perform an overwrite operation by writing a user supplied data pattern to the device one or more times. The pattern is given by the .Fl P argument. The number of times is given by the .Fl c argument. .It block Perform a block erase operation. All the device's blocks are set to a vendor defined value, typically zero. .It crypto Perform a cryptographic erase operation. The encryption keys are changed to prevent the decryption of the data. .It exitfailure Exits a previously failed sanitize operation. A failed sanitize operation can only be exited if it was run in the unrestricted completion mode, as provided by the .Fl U argument. .El .It Fl c Ar passes The number of passes when performing an .Sq overwrite operation. Valid values are between 1 and 31. The default is 1. .It Fl I When performing an .Sq overwrite operation, the pattern is inverted between consecutive passes. .It Fl P Ar pattern Path to the file containing the pattern to use when performing an .Sq overwrite operation. The pattern is repeated as needed to fill each block. .It Fl q Be quiet, do not print any status messages. This option will not disable the questions, however. To disable questions, use the .Fl y argument, below. .It Fl U Perform the sanitize in the unrestricted completion mode. If the operation fails, it can later be exited with the .Sq exitfailure operation. .It Fl r Run in .Dq report only mode. This will report status on a sanitize that is already running on the drive. .It Fl w Issue a non-immediate sanitize command. By default, .Nm issues the SANITIZE command with the immediate bit set. This tells the device to immediately return the sanitize command, before the sanitize has actually completed. Then, .Nm gathers .Tn SCSI sense information from the device every second to determine how far along in the sanitize process it is. If the .Fl w argument is specified, .Nm will issue a non-immediate sanitize command, and will be unable to print any information to let the user know what percentage of the disk has been sanitized. .It Fl y Do not ask any questions. By default, .Nm will ask the user if he/she really wants to sanitize the disk in question, and also if the default sanitize command timeout is acceptable. The user will not be asked about the timeout if a timeout is specified on the command line. .El .It Ic idle Put ATA device into IDLE state. Optional parameter .Pq Fl t specifies automatic standby timer value in seconds. Value 0 disables timer. .It Ic standby Put ATA device into STANDBY state. Optional parameter .Pq Fl t specifies automatic standby timer value in seconds. Value 0 disables timer. .It Ic sleep Put ATA device into SLEEP state. Note that the only way get device out of this state may be reset. .It Ic apm It optional parameter .Pq Fl l specified, enables and sets advanced power management level, where 1 -- minimum power, 127 -- maximum performance with standby, 128 -- minimum power without standby, 254 -- maximum performance. If not specified -- APM is disabled. .It Ic aam It optional parameter .Pq Fl l specified, enables and sets automatic acoustic management level, where 1 -- minimum noise, 254 -- maximum performance. If not specified -- AAM is disabled. .It Ic security Update or report security settings, using an ATA identify command (0xec). By default, .Nm will print out the security support and associated settings of the device. The .Ic security command takes several arguments: .Bl -tag -width 0n .It Fl d Ar pwd .Pp Disable device security using the given password for the selected user according to the devices configured security level. .It Fl e Ar pwd .Pp Erase the device using the given password for the selected user. .Pp .Em WARNING! WARNING! WARNING! .Pp Issuing a secure erase will .Em ERASE ALL user data on the device and may take several hours to complete. .Pp When this command is used against an SSD drive all its cells will be marked as empty, restoring it to factory default write performance. For SSD's this action usually takes just a few seconds. .It Fl f .Pp Freeze the security configuration of the specified device. .Pp After command completion any other commands that update the device lock mode shall be command aborted. Frozen mode is disabled by power-off or hardware reset. .It Fl h Ar pwd .Pp Enhanced erase the device using the given password for the selected user. .Pp .Em WARNING! WARNING! WARNING! .Pp Issuing an enhanced secure erase will .Em ERASE ALL user data on the device and may take several hours to complete. .Pp An enhanced erase writes predetermined data patterns to all user data areas, all previously written user data shall be overwritten, including sectors that are no longer in use due to reallocation. .It Fl k Ar pwd .Pp Unlock the device using the given password for the selected user according to the devices configured security level. .It Fl l Ar high|maximum .Pp Specifies which security level to set when issuing a .Fl s Ar pwd command. The security level determines device behavior when the master password is used to unlock the device. When the security level is set to high the device requires the unlock command and the master password to unlock. When the security level is set to maximum the device requires a secure erase with the master password to unlock. .Pp This option must be used in conjunction with one of the security action commands. .Pp Defaults to .Em high .It Fl q .Pp Be quiet, do not print any status messages. This option will not disable the questions, however. To disable questions, use the .Fl y argument, below. .It Fl s Ar pwd .Pp Password the device (enable security) using the given password for the selected user. This option can be combined with other options such as .Fl e Em pwd .Pp A master password may be set in a addition to the user password. The purpose of the master password is to allow an administrator to establish a password that is kept secret from the user, and which may be used to unlock the device if the user password is lost. .Pp .Em Note: Setting the master password does not enable device security. .Pp If the master password is set and the drive supports a Master Revision Code feature the Master Password Revision Code will be decremented. .It Fl T Ar timeout .Pp Overrides the default timeout, specified in seconds, used for both .Fl e and .Fl h this is useful if your system has problems processing long timeouts correctly. .Pp Usually the timeout is calculated from the information stored on the drive if present, otherwise it defaults to 2 hours. .It Fl U Ar user|master .Pp Specifies which user to set / use for the running action command, valid values are user or master and defaults to master if not set. .Pp This option must be used in conjunction with one of the security action commands. .Pp Defaults to .Em master .It Fl y .Pp Confirm yes to dangerous options such as .Fl e without prompting for confirmation. .El .Pp If the password specified for any action commands does not match the configured password for the specified user the command will fail. .Pp The password in all cases is limited to 32 characters, longer passwords will fail. .It Ic hpa Update or report Host Protected Area details. By default .Nm will print out the HPA support and associated settings of the device. The .Ic hpa command takes several optional arguments: .Bl -tag -width 0n .It Fl f .Pp Freeze the HPA configuration of the specified device. .Pp After command completion any other commands that update the HPA configuration shall be command aborted. Frozen mode is disabled by power-off or hardware reset. .It Fl l .Pp Lock the HPA configuration of the device until a successful call to unlock or the next power-on reset occurs. .It Fl P .Pp Make the HPA max sectors persist across power-on reset or a hardware reset. This must be used in combination with .Fl s Ar max_sectors . .It Fl p Ar pwd .Pp Set the HPA configuration password required for unlock calls. .It Fl q .Pp Be quiet, do not print any status messages. This option will not disable the questions. To disable questions, use the .Fl y argument, below. .It Fl s Ar max_sectors .Pp Configures the maximum user accessible sectors of the device. This will change the number of sectors the device reports. .Pp .Em WARNING! WARNING! WARNING! .Pp Changing the max sectors of a device using this option will make the data on the device beyond the specified value inaccessible. .Pp Only one successful .Fl s Ar max_sectors call can be made without a power-on reset or a hardware reset of the device. .It Fl U Ar pwd .Pp Unlock the HPA configuration of the specified device using the given password. If the password specified does not match the password configured via .Fl p Ar pwd the command will fail. .Pp After 5 failed unlock calls, due to password miss-match, the device will refuse additional unlock calls until after a power-on reset. .It Fl y .Pp Confirm yes to dangerous options such as .Fl e without prompting for confirmation .El .Pp The password for all HPA commands is limited to 32 characters, longer passwords will fail. .It Ic fwdownload Program firmware of the named SCSI device using the image file provided. .Pp Current list of supported vendors: .Bl -bullet -offset indent -compact .It HITACHI .It HP .It IBM .It PLEXTOR .It QUANTUM .It SAMSUNG .It SEAGATE .El .Pp .Em WARNING! WARNING! WARNING! .Pp Little testing has been done to make sure that different device models from each vendor work correctly with the fwdownload command. A vendor name appearing in the supported list means only that firmware of at least one device type from that vendor has successfully been programmed with the fwdownload command. Extra caution should be taken when using this command since there is no guarantee it will not break a device from the listed vendors. Ensure that you have a recent backup of the data on the device before performing a firmware update. .Bl -tag -width 11n .It Fl f Ar fw_image Path to the firmware image file to be downloaded to the specified device. .It Fl y Do not ask for confirmation. .It Fl s Run in simulation mode. Packet sizes that will be sent are shown, but no actual packet is sent to the device. No confirmation is asked in simulation mode. .It Fl v Besides showing sense information in case of a failure, the verbose option causes .Nm to output a line for every firmware segment that is sent to the device by the fwdownload command -- the same as the ones shown in simulation mode. .El .It Ic persist Persistent reservation support. Persistent reservations are a way to reserve a particular .Tn SCSI LUN for use by one or more .Tn SCSI initiators. If the .Fl i option is specified, .Nm will issue the .Tn SCSI PERSISTENT RESERVE IN command using the requested service action. If the .Fl o option is specified, .Nm will issue the .Tn SCSI PERSISTENT RESERVE OUT command using the requested service action. One of those two options is required. .Pp Persistent reservations are complex, and fully explaining them is outside the scope of this manual. Please visit http://www.t10.org and download the latest SPC spec for a full explanation of persistent reservations. .Bl -tag -width 8n .It Fl i Ar mode Specify the service action for the PERSISTENT RESERVE IN command. Supported service actions: .Bl -tag -width 19n .It read_keys Report the current persistent reservation generation (PRgeneration) and any registered keys. .It read_reservation Report the persistent reservation, if any. .It report_capabilities Report the persistent reservation capabilities of the LUN. .It read_full_status Report the full status of persistent reservations on the LUN. .El .It Fl o Ar mode Specify the service action for the PERSISTENT RESERVE OUT command. For service actions like register that are components of other service action names, the entire name must be specified. Otherwise, enough of the service action name must be specified to distinguish it from other possible service actions. Supported service actions: .Bl -tag -width 15n .It register Register a reservation key with the LUN or unregister a reservation key. To register a key, specify the requested key as the Service Action Reservation Key. To unregister a key, specify the previously registered key as the Reservation Key. To change a key, specify the old key as the Reservation Key and the new key as the Service Action Reservation Key. .It register_ignore This is similar to the register subcommand, except that the Reservation Key is ignored. The Service Action Reservation Key will overwrite any previous key registered for the initiator. .It reserve Create a reservation. A key must be registered with the LUN before the LUN can be reserved, and it must be specified as the Reservation Key. The type of reservation must also be specified. The scope defaults to LUN scope (LU_SCOPE), but may be changed. .It release Release a reservation. The Reservation Key must be specified. .It clear Release a reservation and remove all keys from the device. The Reservation Key must be specified. .It preempt Remove a reservation belonging to another initiator. The Reservation Key must be specified. The Service Action Reservation Key may be specified, depending on the operation being performed. .It preempt_abort Remove a reservation belonging to another initiator and abort all outstanding commands from that initiator. The Reservation Key must be specified. The Service Action Reservation Key may be specified, depending on the operation being performed. .It register_move Register another initiator with the LUN, and establish a reservation on the LUN for that initiator. The Reservation Key and Service Action Reservation Key must be specified. .It replace_lost Replace Lost Reservation information. .El .It Fl a Set the All Target Ports (ALL_TG_PT) bit. This requests that the key registration be applied to all target ports and not just the particular target port that receives the command. This only applies to the register and register_ignore actions. .It Fl I Ar tid Specify a Transport ID. This only applies to the Register and Register and Move service actions for Persistent Reserve Out. Multiple Transport IDs may be specified with multiple .Fl I arguments. With the Register service action, specifying one or more Transport IDs implicitly enables the .Fl S option which turns on the SPEC_I_PT bit. Transport IDs generally have the format protocol,id. .Bl -tag -width 5n .It SAS A SAS Transport ID consists of .Dq sas, followed by a 64-bit SAS address. For example: .Pp .Dl sas,0x1234567812345678 .It FC A Fibre Channel Transport ID consists of .Dq fcp, followed by a 64-bit Fibre Channel World Wide Name. For example: .Pp .Dl fcp,0x1234567812345678 .It SPI A Parallel SCSI address consists of .Dq spi, followed by a SCSI target ID and a relative target port identifier. For example: .Pp .Dl spi,4,1 .It 1394 An IEEE 1394 (Firewire) Transport ID consists of .Dq sbp, followed by a 64-bit EUI-64 IEEE 1394 node unique identifier. For example: .Pp .Dl sbp,0x1234567812345678 .It RDMA A SCSI over RDMA Transport ID consists of .Dq srp, followed by a 128-bit RDMA initiator port identifier. The port identifier must be exactly 32 or 34 (if the leading 0x is included) hexadecimal digits. Only hexadecimal (base 16) numbers are supported. For example: .Pp .Dl srp,0x12345678123456781234567812345678 .It iSCSI An iSCSI Transport ID consists an iSCSI name and optionally a separator and iSCSI session ID. For example, if only the iSCSI name is specified: .Pp .Dl iqn.2012-06.com.example:target0 .Pp If the iSCSI separator and initiator session ID are specified: .Pp .Dl iqn.2012-06.com.example:target0,i,0x123 .It PCIe A SCSI over PCIe Transport ID consists of .Dq sop, followed by a PCIe Routing ID. The Routing ID consists of a bus, device and function or in the alternate form, a bus and function. The bus must be in the range of 0 to 255 inclusive and the device must be in the range of 0 to 31 inclusive. The function must be in the range of 0 to 7 inclusive if the standard form is used, and in the range of 0 to 255 inclusive if the alternate form is used. For example, if a bus, device and function are specified for the standard Routing ID form: .Pp .Dl sop,4,5,1 .Pp If the alternate Routing ID form is used: .Pp .Dl sop,4,1 .El .It Fl k Ar key Specify the Reservation Key. This may be in decimal, octal or hexadecimal format. The value is zero by default if not otherwise specified. The value must be between 0 and 2^64 - 1, inclusive. .It Fl K Ar key Specify the Service Action Reservation Key. This may be in decimal, octal or hexadecimal format. The value is zero by default if not otherwise specified. The value must be between 0 and 2^64 - 1, inclusive. .It Fl p Enable the Activate Persist Through Power Loss bit. This is only used for the register and register_ignore actions. This requests that the reservation persist across power loss events. .It Fl s Ar scope Specify the scope of the reservation. The scope may be specified by name or by number. The scope is ignored for register, register_ignore and clear. If the desired scope isn't available by name, you may specify the number. .Bl -tag -width 7n .It lun LUN scope (0x00). This encompasses the entire LUN. .It extent Extent scope (0x01). .It element Element scope (0x02). .El .It Fl R Ar rtp Specify the Relative Target Port. This only applies to the Register and Move service action of the Persistent Reserve Out command. .It Fl S Enable the SPEC_I_PT bit. This only applies to the Register service action of Persistent Reserve Out. You must also specify at least one Transport ID with .Fl I if this option is set. If you specify a Transport ID, this option is automatically set. It is an error to specify this option for any service action other than Register. .It Fl T Ar type Specify the reservation type. The reservation type may be specified by name or by number. If the desired reservation type isn't available by name, you may specify the number. Supported reservation type names: .Bl -tag -width 11n .It read_shared Read Shared mode. .It wr_ex Write Exclusive mode. May also be specified as .Dq write_exclusive . .It rd_ex Read Exclusive mode. May also be specified as .Dq read_exclusive . .It ex_ac Exclusive access mode. May also be specified as .Dq exclusive_access . .It wr_ex_ro Write Exclusive Registrants Only mode. May also be specified as .Dq write_exclusive_reg_only . .It ex_ac_ro Exclusive Access Registrants Only mode. May also be specified as .Dq exclusive_access_reg_only . .It wr_ex_ar Write Exclusive All Registrants mode. May also be specified as .Dq write_exclusive_all_regs . .It ex_ac_ar Exclusive Access All Registrants mode. May also be specified as .Dq exclusive_access_all_regs . .El .It Fl U Specify that the target should unregister the initiator that sent the Register and Move request. By default, the target will not unregister the initiator that sends the Register and Move request. This option only applies to the Register and Move service action of the Persistent Reserve Out command. .El .It Ic help Print out verbose usage information. .El .Sh ENVIRONMENT The .Ev SCSI_MODES variable allows the user to specify an alternate mode page format file. .Pp The .Ev EDITOR variable determines which text editor .Nm starts when editing mode pages. .Sh FILES .Bl -tag -width /usr/share/misc/scsi_modes -compact .It Pa /usr/share/misc/scsi_modes is the SCSI mode format database. .It Pa /dev/xpt0 is the transport layer device. .It Pa /dev/pass* are the CAM application passthrough devices. .El .Sh EXAMPLES .Dl camcontrol eject -n cd -u 1 -v .Pp Eject the CD from cd1, and print SCSI sense information if the command fails. .Pp .Dl camcontrol tur da0 .Pp Send the SCSI test unit ready command to da0. The .Nm utility will report whether the disk is ready, but will not display sense information if the command fails since the .Fl v switch was not specified. .Bd -literal -offset indent camcontrol tur da1 -E -C 4 -t 50 -v .Ed .Pp Send a test unit ready command to da1. Enable kernel error recovery. Specify a retry count of 4, and a timeout of 50 seconds. Enable sense printing (with the .Fl v flag) if the command fails. Since error recovery is turned on, the disk will be spun up if it is not currently spinning. The .Nm utility will report whether the disk is ready. .Bd -literal -offset indent camcontrol cmd -n cd -u 1 -v -c "3C 00 00 00 00 00 00 00 0e 00" \e -i 0xe "s1 i3 i1 i1 i1 i1 i1 i1 i1 i1 i1 i1" .Ed .Pp Issue a READ BUFFER command (0x3C) to cd1. Display the buffer size of cd1, and display the first 10 bytes from the cache on cd1. Display SCSI sense information if the command fails. .Bd -literal -offset indent camcontrol cmd -n cd -u 1 -v -c "3B 00 00 00 00 00 00 00 0e 00" \e -o 14 "00 00 00 00 1 2 3 4 5 6 v v v v" 7 8 9 8 .Ed .Pp Issue a WRITE BUFFER (0x3B) command to cd1. Write out 10 bytes of data, not including the (reserved) 4 byte header. Print out sense information if the command fails. Be very careful with this command, improper use may cause data corruption. .Bd -literal -offset indent camcontrol modepage da3 -m 1 -e -P 3 .Ed .Pp Edit mode page 1 (the Read-Write Error Recover page) for da3, and save the settings on the drive. Mode page 1 contains a disk drive's auto read and write reallocation settings, among other things. .Pp .Dl camcontrol rescan all .Pp Rescan all SCSI busses in the system for devices that have been added, removed or changed. .Pp .Dl camcontrol rescan 0 .Pp Rescan SCSI bus 0 for devices that have been added, removed or changed. .Pp .Dl camcontrol rescan 0:1:0 .Pp Rescan SCSI bus 0, target 1, lun 0 to see if it has been added, removed, or changed. .Pp .Dl camcontrol tags da5 -N 24 .Pp Set the number of concurrent transactions for da5 to 24. .Bd -literal -offset indent camcontrol negotiate -n da -u 4 -T disable .Ed .Pp Disable tagged queueing for da4. .Bd -literal -offset indent camcontrol negotiate -n da -u 3 -R 20.000 -O 15 -a .Ed .Pp Negotiate a sync rate of 20MHz and an offset of 15 with da3. Then send a Test Unit Ready command to make the settings take effect. .Bd -literal -offset indent camcontrol smpcmd ses0 -v -r 4 "40 0 00 0" -R 1020 "s9 i1" .Ed .Pp Send the SMP REPORT GENERAL command to ses0, and display the number of PHYs it contains. Display SMP errors if the command fails. .Bd -literal -offset indent camcontrol security ada0 .Ed .Pp Report security support and settings for ada0 .Bd -literal -offset indent camcontrol security ada0 -U user -s MyPass .Ed .Pp Enable security on device ada0 with the password MyPass .Bd -literal -offset indent camcontrol security ada0 -U user -e MyPass .Ed .Pp Secure erase ada0 which has had security enabled with user password MyPass .Pp .Em WARNING! WARNING! WARNING! .Pp This will .Em ERASE ALL data from the device, so backup your data before using! .Pp This command can be used against an SSD drive to restoring it to factory default write performance. .Bd -literal -offset indent camcontrol hpa ada0 .Ed .Pp Report HPA support and settings for ada0 (also reported via identify). .Bd -literal -offset indent camcontrol hpa ada0 -s 10240 .Ed .Pp Enables HPA on ada0 setting the maximum reported sectors to 10240. .Pp .Em WARNING! WARNING! WARNING! .Pp This will .Em PREVENT ACCESS to all data on the device beyond this limit until HPA is disabled by setting HPA to native max sectors of the device, which can only be done after a power-on or hardware reset! .Pp .Em DO NOT use this on a device which has an active filesystem! .Bd -literal -offset indent camcontrol persist da0 -v -i read_keys .Ed .Pp This will read any persistent reservation keys registered with da0, and display any errors encountered when sending the PERSISTENT RESERVE IN .Tn SCSI command. .Bd -literal -offset indent camcontrol persist da0 -v -o register -a -K 0x12345678 .Ed .Pp This will register the persistent reservation key 0x12345678 with da0, apply that registration to all ports on da0, and display any errors that occur when sending the PERSISTENT RESERVE OUT command. .Bd -literal -offset indent camcontrol persist da0 -v -o reserve -s lun -k 0x12345678 -T ex_ac .Ed .Pp This will reserve da0 for the exlusive use of the initiator issuing the command. The scope of the reservation is the entire LUN. Any errors sending the PERSISTENT RESERVE OUT command will be displayed. .Bd -literal -offset indent camcontrol persist da0 -v -i read_full .Ed .Pp This will display the full status of all reservations on da0 and print out status if there are any errors. .Bd -literal -offset indent camcontrol persist da0 -v -o release -k 0x12345678 -T ex_ac .Ed .Pp This will release a reservation on da0 of the type ex_ac (Exclusive Access). The Reservation Key for this registration is 0x12345678. Any errors that occur will be displayed. .Bd -literal -offset indent camcontrol persist da0 -v -o register -K 0x12345678 -S \e -I sas,0x1234567812345678 -I sas,0x8765432187654321 .Ed .Pp This will register the key 0x12345678 with da0, specifying that it applies to the SAS initiators with SAS addresses 0x1234567812345678 and 0x8765432187654321. .Bd -literal -offset indent camcontrol persist da0 -v -o register_move -k 0x87654321 \e -K 0x12345678 -U -p -R 2 -I fcp,0x1234567812345678 .Ed .Pp This will move the registration from the current initiator, whose Registration Key is 0x87654321, to the Fibre Channel initiator with the Fiber Channel World Wide Node Name 0x1234567812345678. A new registration key, 0x12345678, will be registered for the initiator with the Fibre Channel World Wide Node Name 0x1234567812345678, and the current initiator will be unregistered from the target. The reservation will be moved to relative target port 2 on the target device. The registration will persist across power losses. .Sh SEE ALSO .Xr cam 3 , .Xr cam_cdbparse 3 , .Xr cam 4 , .Xr pass 4 , .Xr xpt 4 .Sh HISTORY The .Nm utility first appeared in .Fx 3.0 . .Pp The mode page editing code and arbitrary SCSI command code are based upon code in the old .Xr scsi 8 utility and .Xr scsi 3 library, written by Julian Elischer and Peter Dufault. The .Xr scsi 8 program first appeared in .Bx 386 0.1.2.4 , and first appeared in .Fx in .Fx 2.0.5 . .Sh AUTHORS .An Kenneth Merry Aq Mt ken@FreeBSD.org .Sh BUGS The code that parses the generic command line arguments does not know that some of the subcommands take multiple arguments. So if, for instance, you tried something like this: .Bd -literal -offset indent camcontrol cmd -n da -u 1 -c "00 00 00 00 00 v" 0x00 -v .Ed .Pp The sense information from the test unit ready command would not get printed out, since the first .Xr getopt 3 call in .Nm bails out when it sees the second argument to .Fl c (0x00), above. Fixing this behavior would take some gross code, or changes to the .Xr getopt 3 interface. The best way to circumvent this problem is to always make sure to specify generic .Nm arguments before any command-specific arguments. Index: user/ngie/more-tests/sbin/growfs/Makefile =================================================================== --- user/ngie/more-tests/sbin/growfs/Makefile (revision 281974) +++ user/ngie/more-tests/sbin/growfs/Makefile (revision 281975) @@ -1,28 +1,28 @@ # @(#)Makefile 8.8 (Berkeley) 6/21/2000 # # $TSHeader: src/sbin/growfs/Makefile,v 1.4 2000/12/05 19:45:24 tomsoft Exp $ # $FreeBSD$ # -#GFSDBG= - .include .PATH: ${.CURDIR}/../mount PROG= growfs SRCS= growfs.c getmntopts.c MAN= growfs.8 CFLAGS+=-I${.CURDIR}/../mount .if defined(GFSDBG) -SRCS+= debug.c +SRCS+= debug.c +CFLAGS+= -DFS_DEBUG +NO_WCAST_ALIGN= yes .endif LIBADD= util .if ${MK_TESTS} != "no" SUBDIR+= tests .endif .include Index: user/ngie/more-tests/sbin/growfs/growfs.c =================================================================== --- user/ngie/more-tests/sbin/growfs/growfs.c (revision 281974) +++ user/ngie/more-tests/sbin/growfs/growfs.c (revision 281975) @@ -1,1742 +1,1742 @@ /* * Copyright (c) 1980, 1989, 1993 The Regents of the University of California. * Copyright (c) 2000 Christoph Herrmann, Thomas-Henning von Kamptz * Copyright (c) 2012 The FreeBSD Foundation * All rights reserved. * * This code is derived from software contributed to Berkeley by * Christoph Herrmann and Thomas-Henning von Kamptz, Munich and Frankfurt. * * Portions of this software were developed by Edward Tomasz Napierala * under sponsorship from the FreeBSD Foundation. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgment: * This product includes software developed by the University of * California, Berkeley and its contributors, as well as Christoph * Herrmann and Thomas-Henning von Kamptz. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $TSHeader: src/sbin/growfs/growfs.c,v 1.5 2000/12/12 19:31:00 tomsoft Exp $ * */ #ifndef lint static const char copyright[] = "@(#) Copyright (c) 2000 Christoph Herrmann, Thomas-Henning von Kamptz\n\ Copyright (c) 1980, 1989, 1993 The Regents of the University of California.\n\ All rights reserved.\n"; #endif /* not lint */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "debug.h" #ifdef FS_DEBUG int _dbg_lvl_ = (DL_INFO); /* DL_TRC */ #endif /* FS_DEBUG */ static union { struct fs fs; char pad[SBLOCKSIZE]; } fsun1, fsun2; #define sblock fsun1.fs /* the new superblock */ #define osblock fsun2.fs /* the old superblock */ /* * Possible superblock locations ordered from most to least likely. */ static int sblock_try[] = SBLOCKSEARCH; static ufs2_daddr_t sblockloc; static union { struct cg cg; char pad[MAXBSIZE]; } cgun1, cgun2; #define acg cgun1.cg /* a cylinder cgroup (new) */ #define aocg cgun2.cg /* an old cylinder group */ static struct csum *fscs; /* cylinder summary */ static void growfs(int, int, unsigned int); static void rdfs(ufs2_daddr_t, size_t, void *, int); static void wtfs(ufs2_daddr_t, size_t, void *, int, unsigned int); static int charsperline(void); static void usage(void); static int isblock(struct fs *, unsigned char *, int); static void clrblock(struct fs *, unsigned char *, int); static void setblock(struct fs *, unsigned char *, int); static void initcg(int, time_t, int, unsigned int); static void updjcg(int, time_t, int, int, unsigned int); static void updcsloc(time_t, int, int, unsigned int); static void frag_adjust(ufs2_daddr_t, int); static void updclst(int); static void mount_reload(const struct statfs *stfs); /* * Here we actually start growing the file system. We basically read the * cylinder summary from the first cylinder group as we want to update * this on the fly during our various operations. First we handle the * changes in the former last cylinder group. Afterwards we create all new * cylinder groups. Now we handle the cylinder group containing the * cylinder summary which might result in a relocation of the whole * structure. In the end we write back the updated cylinder summary, the * new superblock, and slightly patched versions of the super block * copies. */ static void growfs(int fsi, int fso, unsigned int Nflag) { DBG_FUNC("growfs") time_t modtime; uint cylno; int i, j, width; char tmpbuf[100]; DBG_ENTER; time(&modtime); /* * Get the cylinder summary into the memory. */ fscs = (struct csum *)calloc((size_t)1, (size_t)sblock.fs_cssize); if (fscs == NULL) errx(1, "calloc failed"); for (i = 0; i < osblock.fs_cssize; i += osblock.fs_bsize) { rdfs(fsbtodb(&osblock, osblock.fs_csaddr + numfrags(&osblock, i)), (size_t)MIN(osblock.fs_cssize - i, osblock.fs_bsize), (void *)(((char *)fscs) + i), fsi); } #ifdef FS_DEBUG { struct csum *dbg_csp; - int dbg_csc; + u_int32_t dbg_csc; char dbg_line[80]; dbg_csp = fscs; for (dbg_csc = 0; dbg_csc < osblock.fs_ncg; dbg_csc++) { snprintf(dbg_line, sizeof(dbg_line), "%d. old csum in old location", dbg_csc); DBG_DUMP_CSUM(&osblock, dbg_line, dbg_csp++); } } #endif /* FS_DEBUG */ DBG_PRINT0("fscs read\n"); /* * Do all needed changes in the former last cylinder group. */ updjcg(osblock.fs_ncg - 1, modtime, fsi, fso, Nflag); /* * Dump out summary information about file system. */ #ifdef FS_DEBUG #define B2MBFACTOR (1 / (1024.0 * 1024.0)) printf("growfs: %.1fMB (%jd sectors) block size %d, fragment size %d\n", (float)sblock.fs_size * sblock.fs_fsize * B2MBFACTOR, (intmax_t)fsbtodb(&sblock, sblock.fs_size), sblock.fs_bsize, sblock.fs_fsize); printf("\tusing %d cylinder groups of %.2fMB, %d blks, %d inodes.\n", sblock.fs_ncg, (float)sblock.fs_fpg * sblock.fs_fsize * B2MBFACTOR, sblock.fs_fpg / sblock.fs_frag, sblock.fs_ipg); if (sblock.fs_flags & FS_DOSOFTDEP) printf("\twith soft updates\n"); #undef B2MBFACTOR #endif /* FS_DEBUG */ /* * Now build the cylinders group blocks and * then print out indices of cylinder groups. */ printf("super-block backups (for fsck_ffs -b #) at:\n"); i = 0; width = charsperline(); /* * Iterate for only the new cylinder groups. */ for (cylno = osblock.fs_ncg; cylno < sblock.fs_ncg; cylno++) { initcg(cylno, modtime, fso, Nflag); j = sprintf(tmpbuf, " %jd%s", (intmax_t)fsbtodb(&sblock, cgsblock(&sblock, cylno)), cylno < (sblock.fs_ncg - 1) ? "," : "" ); if (i + j >= width) { printf("\n"); i = 0; } i += j; printf("%s", tmpbuf); fflush(stdout); } printf("\n"); /* * Do all needed changes in the first cylinder group. * allocate blocks in new location */ updcsloc(modtime, fsi, fso, Nflag); /* * Now write the cylinder summary back to disk. */ for (i = 0; i < sblock.fs_cssize; i += sblock.fs_bsize) { wtfs(fsbtodb(&sblock, sblock.fs_csaddr + numfrags(&sblock, i)), (size_t)MIN(sblock.fs_cssize - i, sblock.fs_bsize), (void *)(((char *)fscs) + i), fso, Nflag); } DBG_PRINT0("fscs written\n"); #ifdef FS_DEBUG { struct csum *dbg_csp; - int dbg_csc; + u_int32_t dbg_csc; char dbg_line[80]; dbg_csp = fscs; for (dbg_csc = 0; dbg_csc < sblock.fs_ncg; dbg_csc++) { snprintf(dbg_line, sizeof(dbg_line), "%d. new csum in new location", dbg_csc); DBG_DUMP_CSUM(&sblock, dbg_line, dbg_csp++); } } #endif /* FS_DEBUG */ /* * Now write the new superblock back to disk. */ sblock.fs_time = modtime; wtfs(sblockloc, (size_t)SBLOCKSIZE, (void *)&sblock, fso, Nflag); DBG_PRINT0("sblock written\n"); DBG_DUMP_FS(&sblock, "new initial sblock"); /* * Clean up the dynamic fields in our superblock copies. */ sblock.fs_fmod = 0; sblock.fs_clean = 1; sblock.fs_ronly = 0; sblock.fs_cgrotor = 0; sblock.fs_state = 0; memset((void *)&sblock.fs_fsmnt, 0, sizeof(sblock.fs_fsmnt)); sblock.fs_flags &= FS_DOSOFTDEP; /* * XXX * The following fields are currently distributed from the superblock * to the copies: * fs_minfree * fs_rotdelay * fs_maxcontig * fs_maxbpg * fs_minfree, * fs_optim * fs_flags regarding SOFTPDATES * * We probably should rather change the summary for the cylinder group * statistics here to the value of what would be in there, if the file * system were created initially with the new size. Therefor we still * need to find an easy way of calculating that. * Possibly we can try to read the first superblock copy and apply the * "diffed" stats between the old and new superblock by still copying * certain parameters onto that. */ /* * Write out the duplicate super blocks. */ for (cylno = 0; cylno < sblock.fs_ncg; cylno++) { wtfs(fsbtodb(&sblock, cgsblock(&sblock, cylno)), (size_t)SBLOCKSIZE, (void *)&sblock, fso, Nflag); } DBG_PRINT0("sblock copies written\n"); DBG_DUMP_FS(&sblock, "new other sblocks"); DBG_LEAVE; return; } /* * This creates a new cylinder group structure, for more details please see * the source of newfs(8), as this function is taken over almost unchanged. * As this is never called for the first cylinder group, the special * provisions for that case are removed here. */ static void initcg(int cylno, time_t modtime, int fso, unsigned int Nflag) { DBG_FUNC("initcg") static caddr_t iobuf; long blkno, start; ino_t ino; ufs2_daddr_t i, cbase, dmax; struct ufs1_dinode *dp1; struct csum *cs; uint j, d, dupper, dlower; if (iobuf == NULL && (iobuf = malloc(sblock.fs_bsize * 3)) == NULL) errx(37, "panic: cannot allocate I/O buffer"); /* * Determine block bounds for cylinder group. * Allow space for super block summary information in first * cylinder group. */ cbase = cgbase(&sblock, cylno); dmax = cbase + sblock.fs_fpg; if (dmax > sblock.fs_size) dmax = sblock.fs_size; dlower = cgsblock(&sblock, cylno) - cbase; dupper = cgdmin(&sblock, cylno) - cbase; if (cylno == 0) /* XXX fscs may be relocated */ dupper += howmany(sblock.fs_cssize, sblock.fs_fsize); cs = &fscs[cylno]; memset(&acg, 0, sblock.fs_cgsize); acg.cg_time = modtime; acg.cg_magic = CG_MAGIC; acg.cg_cgx = cylno; acg.cg_niblk = sblock.fs_ipg; acg.cg_initediblk = sblock.fs_ipg < 2 * INOPB(&sblock) ? sblock.fs_ipg : 2 * INOPB(&sblock); acg.cg_ndblk = dmax - cbase; if (sblock.fs_contigsumsize > 0) acg.cg_nclusterblks = acg.cg_ndblk / sblock.fs_frag; start = &acg.cg_space[0] - (u_char *)(&acg.cg_firstfield); if (sblock.fs_magic == FS_UFS2_MAGIC) { acg.cg_iusedoff = start; } else { acg.cg_old_ncyl = sblock.fs_old_cpg; acg.cg_old_time = acg.cg_time; acg.cg_time = 0; acg.cg_old_niblk = acg.cg_niblk; acg.cg_niblk = 0; acg.cg_initediblk = 0; acg.cg_old_btotoff = start; acg.cg_old_boff = acg.cg_old_btotoff + sblock.fs_old_cpg * sizeof(int32_t); acg.cg_iusedoff = acg.cg_old_boff + sblock.fs_old_cpg * sizeof(u_int16_t); } acg.cg_freeoff = acg.cg_iusedoff + howmany(sblock.fs_ipg, CHAR_BIT); acg.cg_nextfreeoff = acg.cg_freeoff + howmany(sblock.fs_fpg, CHAR_BIT); if (sblock.fs_contigsumsize > 0) { acg.cg_clustersumoff = roundup(acg.cg_nextfreeoff, sizeof(u_int32_t)); acg.cg_clustersumoff -= sizeof(u_int32_t); acg.cg_clusteroff = acg.cg_clustersumoff + (sblock.fs_contigsumsize + 1) * sizeof(u_int32_t); acg.cg_nextfreeoff = acg.cg_clusteroff + howmany(fragstoblks(&sblock, sblock.fs_fpg), CHAR_BIT); } if (acg.cg_nextfreeoff > (unsigned)sblock.fs_cgsize) { /* * This should never happen as we would have had that panic * already on file system creation */ errx(37, "panic: cylinder group too big"); } acg.cg_cs.cs_nifree += sblock.fs_ipg; if (cylno == 0) for (ino = 0; ino < ROOTINO; ino++) { setbit(cg_inosused(&acg), ino); acg.cg_cs.cs_nifree--; } /* * For the old file system, we have to initialize all the inodes. */ if (sblock.fs_magic == FS_UFS1_MAGIC) { bzero(iobuf, sblock.fs_bsize); for (i = 0; i < sblock.fs_ipg / INOPF(&sblock); i += sblock.fs_frag) { dp1 = (struct ufs1_dinode *)(void *)iobuf; for (j = 0; j < INOPB(&sblock); j++) { dp1->di_gen = arc4random(); dp1++; } wtfs(fsbtodb(&sblock, cgimin(&sblock, cylno) + i), sblock.fs_bsize, iobuf, fso, Nflag); } } if (cylno > 0) { /* * In cylno 0, beginning space is reserved * for boot and super blocks. */ for (d = 0; d < dlower; d += sblock.fs_frag) { blkno = d / sblock.fs_frag; setblock(&sblock, cg_blksfree(&acg), blkno); if (sblock.fs_contigsumsize > 0) setbit(cg_clustersfree(&acg), blkno); acg.cg_cs.cs_nbfree++; } sblock.fs_dsize += dlower; } sblock.fs_dsize += acg.cg_ndblk - dupper; if ((i = dupper % sblock.fs_frag)) { acg.cg_frsum[sblock.fs_frag - i]++; for (d = dupper + sblock.fs_frag - i; dupper < d; dupper++) { setbit(cg_blksfree(&acg), dupper); acg.cg_cs.cs_nffree++; } } for (d = dupper; d + sblock.fs_frag <= acg.cg_ndblk; d += sblock.fs_frag) { blkno = d / sblock.fs_frag; setblock(&sblock, cg_blksfree(&acg), blkno); if (sblock.fs_contigsumsize > 0) setbit(cg_clustersfree(&acg), blkno); acg.cg_cs.cs_nbfree++; } if (d < acg.cg_ndblk) { acg.cg_frsum[acg.cg_ndblk - d]++; for (; d < acg.cg_ndblk; d++) { setbit(cg_blksfree(&acg), d); acg.cg_cs.cs_nffree++; } } if (sblock.fs_contigsumsize > 0) { int32_t *sump = cg_clustersum(&acg); u_char *mapp = cg_clustersfree(&acg); int map = *mapp++; int bit = 1; int run = 0; for (i = 0; i < acg.cg_nclusterblks; i++) { if ((map & bit) != 0) run++; else if (run != 0) { if (run > sblock.fs_contigsumsize) run = sblock.fs_contigsumsize; sump[run]++; run = 0; } if ((i & (CHAR_BIT - 1)) != CHAR_BIT - 1) bit <<= 1; else { map = *mapp++; bit = 1; } } if (run != 0) { if (run > sblock.fs_contigsumsize) run = sblock.fs_contigsumsize; sump[run]++; } } sblock.fs_cstotal.cs_ndir += acg.cg_cs.cs_ndir; sblock.fs_cstotal.cs_nffree += acg.cg_cs.cs_nffree; sblock.fs_cstotal.cs_nbfree += acg.cg_cs.cs_nbfree; sblock.fs_cstotal.cs_nifree += acg.cg_cs.cs_nifree; *cs = acg.cg_cs; memcpy(iobuf, &acg, sblock.fs_cgsize); memset(iobuf + sblock.fs_cgsize, '\0', sblock.fs_bsize * 3 - sblock.fs_cgsize); wtfs(fsbtodb(&sblock, cgtod(&sblock, cylno)), sblock.fs_bsize * 3, iobuf, fso, Nflag); DBG_DUMP_CG(&sblock, "new cg", &acg); DBG_LEAVE; return; } /* * Here we add or subtract (sign +1/-1) the available fragments in a given * block to or from the fragment statistics. By subtracting before and adding * after an operation on the free frag map we can easy update the fragment * statistic, which seems to be otherwise a rather complex operation. */ static void frag_adjust(ufs2_daddr_t frag, int sign) { DBG_FUNC("frag_adjust") int fragsize; int f; DBG_ENTER; fragsize = 0; /* * Here frag only needs to point to any fragment in the block we want * to examine. */ for (f = rounddown(frag, sblock.fs_frag); f < roundup(frag + 1, sblock.fs_frag); f++) { /* * Count contiguous free fragments. */ if (isset(cg_blksfree(&acg), f)) { fragsize++; } else { if (fragsize && fragsize < sblock.fs_frag) { /* * We found something in between. */ acg.cg_frsum[fragsize] += sign; DBG_PRINT2("frag_adjust [%d]+=%d\n", fragsize, sign); } fragsize = 0; } } if (fragsize && fragsize < sblock.fs_frag) { /* * We found something. */ acg.cg_frsum[fragsize] += sign; DBG_PRINT2("frag_adjust [%d]+=%d\n", fragsize, sign); } DBG_PRINT2("frag_adjust [[%d]]+=%d\n", fragsize, sign); DBG_LEAVE; return; } /* * Here we do all needed work for the former last cylinder group. It has to be * changed in any case, even if the file system ended exactly on the end of * this group, as there is some slightly inconsistent handling of the number * of cylinders in the cylinder group. We start again by reading the cylinder * group from disk. If the last block was not fully available, we first handle * the missing fragments, then we handle all new full blocks in that file * system and finally we handle the new last fragmented block in the file * system. We again have to handle the fragment statistics rotational layout * tables and cluster summary during all those operations. */ static void updjcg(int cylno, time_t modtime, int fsi, int fso, unsigned int Nflag) { DBG_FUNC("updjcg") ufs2_daddr_t cbase, dmax, dupper; struct csum *cs; int i, k; int j = 0; DBG_ENTER; /* * Read the former last (joining) cylinder group from disk, and make * a copy. */ rdfs(fsbtodb(&osblock, cgtod(&osblock, cylno)), (size_t)osblock.fs_cgsize, (void *)&aocg, fsi); DBG_PRINT0("jcg read\n"); DBG_DUMP_CG(&sblock, "old joining cg", &aocg); memcpy((void *)&cgun1, (void *)&cgun2, sizeof(cgun2)); /* * If the cylinder group had already its new final size almost * nothing is to be done ... except: * For some reason the value of cg_ncyl in the last cylinder group has * to be zero instead of fs_cpg. As this is now no longer the last * cylinder group we have to change that value now to fs_cpg. */ if (cgbase(&osblock, cylno + 1) == osblock.fs_size) { if (sblock.fs_magic == FS_UFS1_MAGIC) acg.cg_old_ncyl = sblock.fs_old_cpg; wtfs(fsbtodb(&sblock, cgtod(&sblock, cylno)), (size_t)sblock.fs_cgsize, (void *)&acg, fso, Nflag); DBG_PRINT0("jcg written\n"); DBG_DUMP_CG(&sblock, "new joining cg", &acg); DBG_LEAVE; return; } /* * Set up some variables needed later. */ cbase = cgbase(&sblock, cylno); dmax = cbase + sblock.fs_fpg; if (dmax > sblock.fs_size) dmax = sblock.fs_size; dupper = cgdmin(&sblock, cylno) - cbase; if (cylno == 0) /* XXX fscs may be relocated */ dupper += howmany(sblock.fs_cssize, sblock.fs_fsize); /* * Set pointer to the cylinder summary for our cylinder group. */ cs = fscs + cylno; /* * Touch the cylinder group, update all fields in the cylinder group as * needed, update the free space in the superblock. */ acg.cg_time = modtime; if ((unsigned)cylno == sblock.fs_ncg - 1) { /* * This is still the last cylinder group. */ if (sblock.fs_magic == FS_UFS1_MAGIC) acg.cg_old_ncyl = sblock.fs_old_ncyl % sblock.fs_old_cpg; } else { acg.cg_old_ncyl = sblock.fs_old_cpg; } DBG_PRINT2("jcg dbg: %d %u", cylno, sblock.fs_ncg); #ifdef FS_DEBUG if (sblock.fs_magic == FS_UFS1_MAGIC) DBG_PRINT2("%d %u", acg.cg_old_ncyl, sblock.fs_old_cpg); #endif DBG_PRINT0("\n"); acg.cg_ndblk = dmax - cbase; sblock.fs_dsize += acg.cg_ndblk - aocg.cg_ndblk; if (sblock.fs_contigsumsize > 0) acg.cg_nclusterblks = acg.cg_ndblk / sblock.fs_frag; /* * Now we have to update the free fragment bitmap for our new free * space. There again we have to handle the fragmentation and also * the rotational layout tables and the cluster summary. This is * also done per fragment for the first new block if the old file * system end was not on a block boundary, per fragment for the new * last block if the new file system end is not on a block boundary, * and per block for all space in between. * * Handle the first new block here if it was partially available * before. */ if (osblock.fs_size % sblock.fs_frag) { if (roundup(osblock.fs_size, sblock.fs_frag) <= sblock.fs_size) { /* * The new space is enough to fill at least this * block */ j = 0; for (i = roundup(osblock.fs_size - cbase, sblock.fs_frag) - 1; i >= osblock.fs_size - cbase; i--) { setbit(cg_blksfree(&acg), i); acg.cg_cs.cs_nffree++; j++; } /* * Check if the fragment just created could join an * already existing fragment at the former end of the * file system. */ if (isblock(&sblock, cg_blksfree(&acg), ((osblock.fs_size - cgbase(&sblock, cylno)) / sblock.fs_frag))) { /* * The block is now completely available. */ DBG_PRINT0("block was\n"); acg.cg_frsum[osblock.fs_size % sblock.fs_frag]--; acg.cg_cs.cs_nbfree++; acg.cg_cs.cs_nffree -= sblock.fs_frag; k = rounddown(osblock.fs_size - cbase, sblock.fs_frag); updclst((osblock.fs_size - cbase) / sblock.fs_frag); } else { /* * Lets rejoin a possible partially growed * fragment. */ k = 0; while (isset(cg_blksfree(&acg), i) && (i >= rounddown(osblock.fs_size - cbase, sblock.fs_frag))) { i--; k++; } if (k) acg.cg_frsum[k]--; acg.cg_frsum[k + j]++; } } else { /* * We only grow by some fragments within this last * block. */ for (i = sblock.fs_size - cbase - 1; i >= osblock.fs_size - cbase; i--) { setbit(cg_blksfree(&acg), i); acg.cg_cs.cs_nffree++; j++; } /* * Lets rejoin a possible partially growed fragment. */ k = 0; while (isset(cg_blksfree(&acg), i) && (i >= rounddown(osblock.fs_size - cbase, sblock.fs_frag))) { i--; k++; } if (k) acg.cg_frsum[k]--; acg.cg_frsum[k + j]++; } } /* * Handle all new complete blocks here. */ for (i = roundup(osblock.fs_size - cbase, sblock.fs_frag); i + sblock.fs_frag <= dmax - cbase; /* XXX <= or only < ? */ i += sblock.fs_frag) { j = i / sblock.fs_frag; setblock(&sblock, cg_blksfree(&acg), j); updclst(j); acg.cg_cs.cs_nbfree++; } /* * Handle the last new block if there are stll some new fragments left. * Here we don't have to bother about the cluster summary or the even * the rotational layout table. */ if (i < (dmax - cbase)) { acg.cg_frsum[dmax - cbase - i]++; for (; i < dmax - cbase; i++) { setbit(cg_blksfree(&acg), i); acg.cg_cs.cs_nffree++; } } sblock.fs_cstotal.cs_nffree += (acg.cg_cs.cs_nffree - aocg.cg_cs.cs_nffree); sblock.fs_cstotal.cs_nbfree += (acg.cg_cs.cs_nbfree - aocg.cg_cs.cs_nbfree); /* * The following statistics are not changed here: * sblock.fs_cstotal.cs_ndir * sblock.fs_cstotal.cs_nifree * As the statistics for this cylinder group are ready, copy it to * the summary information array. */ *cs = acg.cg_cs; /* * Write the updated "joining" cylinder group back to disk. */ wtfs(fsbtodb(&sblock, cgtod(&sblock, cylno)), (size_t)sblock.fs_cgsize, (void *)&acg, fso, Nflag); DBG_PRINT0("jcg written\n"); DBG_DUMP_CG(&sblock, "new joining cg", &acg); DBG_LEAVE; return; } /* * Here we update the location of the cylinder summary. We have two possible * ways of growing the cylinder summary: * (1) We can try to grow the summary in the current location, and relocate * possibly used blocks within the current cylinder group. * (2) Alternatively we can relocate the whole cylinder summary to the first * new completely empty cylinder group. Once the cylinder summary is no * longer in the beginning of the first cylinder group you should never * use a version of fsck which is not aware of the possibility to have * this structure in a non standard place. * Option (2) is considered to be less intrusive to the structure of the file- * system, so that's the one being used. */ static void updcsloc(time_t modtime, int fsi, int fso, unsigned int Nflag) { DBG_FUNC("updcsloc") struct csum *cs; int ocscg, ncscg; ufs2_daddr_t d; int lcs = 0; int block; DBG_ENTER; if (howmany(sblock.fs_cssize, sblock.fs_fsize) == howmany(osblock.fs_cssize, osblock.fs_fsize)) { /* * No new fragment needed. */ DBG_LEAVE; return; } ocscg = dtog(&osblock, osblock.fs_csaddr); cs = fscs + ocscg; /* * Read original cylinder group from disk, and make a copy. * XXX If Nflag is set in some very rare cases we now miss * some changes done in updjcg by reading the unmodified * block from disk. */ rdfs(fsbtodb(&osblock, cgtod(&osblock, ocscg)), (size_t)osblock.fs_cgsize, (void *)&aocg, fsi); DBG_PRINT0("oscg read\n"); DBG_DUMP_CG(&sblock, "old summary cg", &aocg); memcpy((void *)&cgun1, (void *)&cgun2, sizeof(cgun2)); /* * Touch the cylinder group, set up local variables needed later * and update the superblock. */ acg.cg_time = modtime; /* * XXX In the case of having active snapshots we may need much more * blocks for the copy on write. We need each block twice, and * also up to 8*3 blocks for indirect blocks for all possible * references. */ /* * There is not enough space in the old cylinder group to * relocate all blocks as needed, so we relocate the whole * cylinder group summary to a new group. We try to use the * first complete new cylinder group just created. Within the * cylinder group we align the area immediately after the * cylinder group information location in order to be as * close as possible to the original implementation of ffs. * * First we have to make sure we'll find enough space in the * new cylinder group. If not, then we currently give up. * We start with freeing everything which was used by the * fragments of the old cylinder summary in the current group. * Now we write back the group meta data, read in the needed * meta data from the new cylinder group, and start allocating * within that group. Here we can assume, the group to be * completely empty. Which makes the handling of fragments and * clusters a lot easier. */ DBG_TRC; if (sblock.fs_ncg - osblock.fs_ncg < 2) errx(2, "panic: not enough space"); /* * Point "d" to the first fragment not used by the cylinder * summary. */ d = osblock.fs_csaddr + (osblock.fs_cssize / osblock.fs_fsize); /* * Set up last cluster size ("lcs") already here. Calculate * the size for the trailing cluster just behind where "d" * points to. */ if (sblock.fs_contigsumsize > 0) { for (block = howmany(d % sblock.fs_fpg, sblock.fs_frag), lcs = 0; lcs < sblock.fs_contigsumsize; block++, lcs++) { if (isclr(cg_clustersfree(&acg), block)) break; } } /* * Point "d" to the last frag used by the cylinder summary. */ d--; DBG_PRINT1("d=%jd\n", (intmax_t)d); if ((d + 1) % sblock.fs_frag) { /* * The end of the cylinder summary is not a complete * block. */ DBG_TRC; frag_adjust(d % sblock.fs_fpg, -1); for (; (d + 1) % sblock.fs_frag; d--) { DBG_PRINT1("d=%jd\n", (intmax_t)d); setbit(cg_blksfree(&acg), d % sblock.fs_fpg); acg.cg_cs.cs_nffree++; sblock.fs_cstotal.cs_nffree++; } /* * Point "d" to the last fragment of the last * (incomplete) block of the cylinder summary. */ d++; frag_adjust(d % sblock.fs_fpg, 1); if (isblock(&sblock, cg_blksfree(&acg), (d % sblock.fs_fpg) / sblock.fs_frag)) { DBG_PRINT1("d=%jd\n", (intmax_t)d); acg.cg_cs.cs_nffree -= sblock.fs_frag; acg.cg_cs.cs_nbfree++; sblock.fs_cstotal.cs_nffree -= sblock.fs_frag; sblock.fs_cstotal.cs_nbfree++; if (sblock.fs_contigsumsize > 0) { setbit(cg_clustersfree(&acg), (d % sblock.fs_fpg) / sblock.fs_frag); if (lcs < sblock.fs_contigsumsize) { if (lcs) cg_clustersum(&acg)[lcs]--; lcs++; cg_clustersum(&acg)[lcs]++; } } } /* * Point "d" to the first fragment of the block before * the last incomplete block. */ d--; } DBG_PRINT1("d=%jd\n", (intmax_t)d); for (d = rounddown(d, sblock.fs_frag); d >= osblock.fs_csaddr; d -= sblock.fs_frag) { DBG_TRC; DBG_PRINT1("d=%jd\n", (intmax_t)d); setblock(&sblock, cg_blksfree(&acg), (d % sblock.fs_fpg) / sblock.fs_frag); acg.cg_cs.cs_nbfree++; sblock.fs_cstotal.cs_nbfree++; if (sblock.fs_contigsumsize > 0) { setbit(cg_clustersfree(&acg), (d % sblock.fs_fpg) / sblock.fs_frag); /* * The last cluster size is already set up. */ if (lcs < sblock.fs_contigsumsize) { if (lcs) cg_clustersum(&acg)[lcs]--; lcs++; cg_clustersum(&acg)[lcs]++; } } } *cs = acg.cg_cs; /* * Now write the former cylinder group containing the cylinder * summary back to disk. */ wtfs(fsbtodb(&sblock, cgtod(&sblock, ocscg)), (size_t)sblock.fs_cgsize, (void *)&acg, fso, Nflag); DBG_PRINT0("oscg written\n"); DBG_DUMP_CG(&sblock, "old summary cg", &acg); /* * Find the beginning of the new cylinder group containing the * cylinder summary. */ sblock.fs_csaddr = cgdmin(&sblock, osblock.fs_ncg); ncscg = dtog(&sblock, sblock.fs_csaddr); cs = fscs + ncscg; /* * If Nflag is specified, we would now read random data instead * of an empty cg structure from disk. So we can't simulate that * part for now. */ if (Nflag) { DBG_PRINT0("nscg update skipped\n"); DBG_LEAVE; return; } /* * Read the future cylinder group containing the cylinder * summary from disk, and make a copy. */ rdfs(fsbtodb(&sblock, cgtod(&sblock, ncscg)), (size_t)sblock.fs_cgsize, (void *)&aocg, fsi); DBG_PRINT0("nscg read\n"); DBG_DUMP_CG(&sblock, "new summary cg", &aocg); memcpy((void *)&cgun1, (void *)&cgun2, sizeof(cgun2)); /* * Allocate all complete blocks used by the new cylinder * summary. */ for (d = sblock.fs_csaddr; d + sblock.fs_frag <= sblock.fs_csaddr + (sblock.fs_cssize / sblock.fs_fsize); d += sblock.fs_frag) { clrblock(&sblock, cg_blksfree(&acg), (d % sblock.fs_fpg) / sblock.fs_frag); acg.cg_cs.cs_nbfree--; sblock.fs_cstotal.cs_nbfree--; if (sblock.fs_contigsumsize > 0) { clrbit(cg_clustersfree(&acg), (d % sblock.fs_fpg) / sblock.fs_frag); } } /* * Allocate all fragments used by the cylinder summary in the * last block. */ if (d < sblock.fs_csaddr + (sblock.fs_cssize / sblock.fs_fsize)) { for (; d - sblock.fs_csaddr < sblock.fs_cssize/sblock.fs_fsize; d++) { clrbit(cg_blksfree(&acg), d % sblock.fs_fpg); acg.cg_cs.cs_nffree--; sblock.fs_cstotal.cs_nffree--; } acg.cg_cs.cs_nbfree--; acg.cg_cs.cs_nffree += sblock.fs_frag; sblock.fs_cstotal.cs_nbfree--; sblock.fs_cstotal.cs_nffree += sblock.fs_frag; if (sblock.fs_contigsumsize > 0) clrbit(cg_clustersfree(&acg), (d % sblock.fs_fpg) / sblock.fs_frag); frag_adjust(d % sblock.fs_fpg, 1); } /* * XXX Handle the cluster statistics here in the case this * cylinder group is now almost full, and the remaining * space is less then the maximum cluster size. This is * probably not needed, as you would hardly find a file * system which has only MAXCSBUFS+FS_MAXCONTIG of free * space right behind the cylinder group information in * any new cylinder group. */ /* * Update our statistics in the cylinder summary. */ *cs = acg.cg_cs; /* * Write the new cylinder group containing the cylinder summary * back to disk. */ wtfs(fsbtodb(&sblock, cgtod(&sblock, ncscg)), (size_t)sblock.fs_cgsize, (void *)&acg, fso, Nflag); DBG_PRINT0("nscg written\n"); DBG_DUMP_CG(&sblock, "new summary cg", &acg); DBG_LEAVE; return; } /* * Here we read some block(s) from disk. */ static void rdfs(ufs2_daddr_t bno, size_t size, void *bf, int fsi) { DBG_FUNC("rdfs") ssize_t n; DBG_ENTER; if (bno < 0) err(32, "rdfs: attempting to read negative block number"); if (lseek(fsi, (off_t)bno * DEV_BSIZE, 0) < 0) err(33, "rdfs: seek error: %jd", (intmax_t)bno); n = read(fsi, bf, size); if (n != (ssize_t)size) err(34, "rdfs: read error: %jd", (intmax_t)bno); DBG_LEAVE; return; } /* * Here we write some block(s) to disk. */ static void wtfs(ufs2_daddr_t bno, size_t size, void *bf, int fso, unsigned int Nflag) { DBG_FUNC("wtfs") ssize_t n; DBG_ENTER; if (Nflag) { DBG_LEAVE; return; } if (lseek(fso, (off_t)bno * DEV_BSIZE, SEEK_SET) < 0) err(35, "wtfs: seek error: %ld", (long)bno); n = write(fso, bf, size); if (n != (ssize_t)size) err(36, "wtfs: write error: %ld", (long)bno); DBG_LEAVE; return; } /* * Here we check if all frags of a block are free. For more details again * please see the source of newfs(8), as this function is taken over almost * unchanged. */ static int isblock(struct fs *fs, unsigned char *cp, int h) { DBG_FUNC("isblock") unsigned char mask; DBG_ENTER; switch (fs->fs_frag) { case 8: DBG_LEAVE; return (cp[h] == 0xff); case 4: mask = 0x0f << ((h & 0x1) << 2); DBG_LEAVE; return ((cp[h >> 1] & mask) == mask); case 2: mask = 0x03 << ((h & 0x3) << 1); DBG_LEAVE; return ((cp[h >> 2] & mask) == mask); case 1: mask = 0x01 << (h & 0x7); DBG_LEAVE; return ((cp[h >> 3] & mask) == mask); default: fprintf(stderr, "isblock bad fs_frag %d\n", fs->fs_frag); DBG_LEAVE; return (0); } } /* * Here we allocate a complete block in the block map. For more details again * please see the source of newfs(8), as this function is taken over almost * unchanged. */ static void clrblock(struct fs *fs, unsigned char *cp, int h) { DBG_FUNC("clrblock") DBG_ENTER; switch ((fs)->fs_frag) { case 8: cp[h] = 0; break; case 4: cp[h >> 1] &= ~(0x0f << ((h & 0x1) << 2)); break; case 2: cp[h >> 2] &= ~(0x03 << ((h & 0x3) << 1)); break; case 1: cp[h >> 3] &= ~(0x01 << (h & 0x7)); break; default: warnx("clrblock bad fs_frag %d", fs->fs_frag); break; } DBG_LEAVE; return; } /* * Here we free a complete block in the free block map. For more details again * please see the source of newfs(8), as this function is taken over almost * unchanged. */ static void setblock(struct fs *fs, unsigned char *cp, int h) { DBG_FUNC("setblock") DBG_ENTER; switch (fs->fs_frag) { case 8: cp[h] = 0xff; break; case 4: cp[h >> 1] |= (0x0f << ((h & 0x1) << 2)); break; case 2: cp[h >> 2] |= (0x03 << ((h & 0x3) << 1)); break; case 1: cp[h >> 3] |= (0x01 << (h & 0x7)); break; default: warnx("setblock bad fs_frag %d", fs->fs_frag); break; } DBG_LEAVE; return; } /* * Figure out how many lines our current terminal has. For more details again * please see the source of newfs(8), as this function is taken over almost * unchanged. */ static int charsperline(void) { DBG_FUNC("charsperline") int columns; char *cp; struct winsize ws; DBG_ENTER; columns = 0; if (ioctl(0, TIOCGWINSZ, &ws) != -1) columns = ws.ws_col; if (columns == 0 && (cp = getenv("COLUMNS"))) columns = atoi(cp); if (columns == 0) columns = 80; /* last resort */ DBG_LEAVE; return (columns); } static int is_dev(const char *name) { struct stat devstat; if (stat(name, &devstat) != 0) return (0); if (!S_ISCHR(devstat.st_mode)) return (0); return (1); } /* * Return mountpoint on which the device is currently mounted. */ static const struct statfs * dev_to_statfs(const char *dev) { struct stat devstat, mntdevstat; struct statfs *mntbuf, *statfsp; char device[MAXPATHLEN]; char *mntdevname; int i, mntsize; /* * First check the mounted filesystems. */ if (stat(dev, &devstat) != 0) return (NULL); if (!S_ISCHR(devstat.st_mode) && !S_ISBLK(devstat.st_mode)) return (NULL); mntsize = getmntinfo(&mntbuf, MNT_NOWAIT); for (i = 0; i < mntsize; i++) { statfsp = &mntbuf[i]; mntdevname = statfsp->f_mntfromname; if (*mntdevname != '/') { strcpy(device, _PATH_DEV); strcat(device, mntdevname); mntdevname = device; } if (stat(mntdevname, &mntdevstat) == 0 && mntdevstat.st_rdev == devstat.st_rdev) return (statfsp); } return (NULL); } static const char * mountpoint_to_dev(const char *mountpoint) { struct statfs *mntbuf, *statfsp; struct fstab *fs; int i, mntsize; /* * First check the mounted filesystems. */ mntsize = getmntinfo(&mntbuf, MNT_NOWAIT); for (i = 0; i < mntsize; i++) { statfsp = &mntbuf[i]; if (strcmp(statfsp->f_mntonname, mountpoint) == 0) return (statfsp->f_mntfromname); } /* * Check the fstab. */ fs = getfsfile(mountpoint); if (fs != NULL) return (fs->fs_spec); return (NULL); } static const char * getdev(const char *name) { static char device[MAXPATHLEN]; const char *cp, *dev; if (is_dev(name)) return (name); cp = strrchr(name, '/'); if (cp == 0) { snprintf(device, sizeof(device), "%s%s", _PATH_DEV, name); if (is_dev(device)) return (device); } dev = mountpoint_to_dev(name); if (dev != NULL && is_dev(dev)) return (dev); return (NULL); } /* * growfs(8) is a utility which allows to increase the size of an existing * ufs file system. Currently this can only be done on unmounted file system. * It recognizes some command line options to specify the new desired size, * and it does some basic checkings. The old file system size is determined * and after some more checks like we can really access the new last block * on the disk etc. we calculate the new parameters for the superblock. After * having done this we just call growfs() which will do the work. * We still have to provide support for snapshots. Therefore we first have to * understand what data structures are always replicated in the snapshot on * creation, for all other blocks we touch during our procedure, we have to * keep the old blocks unchanged somewhere available for the snapshots. If we * are lucky, then we only have to handle our blocks to be relocated in that * way. * Also we have to consider in what order we actually update the critical * data structures of the file system to make sure, that in case of a disaster * fsck(8) is still able to restore any lost data. * The foreseen last step then will be to provide for growing even mounted * file systems. There we have to extend the mount() system call to provide * userland access to the file system locking facility. */ int main(int argc, char **argv) { DBG_FUNC("main") const char *device; const struct statfs *statfsp; uint64_t size = 0; off_t mediasize; int error, i, j, fsi, fso, ch, Nflag = 0, yflag = 0; char *p, reply[5], oldsizebuf[6], newsizebuf[6]; void *testbuf; DBG_ENTER; while ((ch = getopt(argc, argv, "Ns:vy")) != -1) { switch(ch) { case 'N': Nflag = 1; break; case 's': size = (off_t)strtoumax(optarg, &p, 0); if (p == NULL || *p == '\0') size *= DEV_BSIZE; else if (*p == 'b' || *p == 'B') ; /* do nothing */ else if (*p == 'k' || *p == 'K') size <<= 10; else if (*p == 'm' || *p == 'M') size <<= 20; else if (*p == 'g' || *p == 'G') size <<= 30; else if (*p == 't' || *p == 'T') { size <<= 30; size <<= 10; } else errx(1, "unknown suffix on -s argument"); break; case 'v': /* for compatibility to newfs */ break; case 'y': yflag = 1; break; case '?': /* FALLTHROUGH */ default: usage(); } } argc -= optind; argv += optind; if (argc != 1) usage(); /* * Now try to guess the device name. */ device = getdev(*argv); if (device == NULL) errx(1, "cannot find special device for %s", *argv); statfsp = dev_to_statfs(device); fsi = open(device, O_RDONLY); if (fsi < 0) err(1, "%s", device); /* * Try to guess the slice size if not specified. */ if (ioctl(fsi, DIOCGMEDIASIZE, &mediasize) == -1) err(1,"DIOCGMEDIASIZE"); /* * Check if that partition is suitable for growing a file system. */ if (mediasize < 1) errx(1, "partition is unavailable"); /* * Read the current superblock, and take a backup. */ for (i = 0; sblock_try[i] != -1; i++) { sblockloc = sblock_try[i] / DEV_BSIZE; rdfs(sblockloc, (size_t)SBLOCKSIZE, (void *)&(osblock), fsi); if ((osblock.fs_magic == FS_UFS1_MAGIC || (osblock.fs_magic == FS_UFS2_MAGIC && osblock.fs_sblockloc == sblock_try[i])) && osblock.fs_bsize <= MAXBSIZE && osblock.fs_bsize >= (int32_t) sizeof(struct fs)) break; } if (sblock_try[i] == -1) errx(1, "superblock not recognized"); memcpy((void *)&fsun1, (void *)&fsun2, sizeof(fsun2)); DBG_OPEN("/tmp/growfs.debug"); /* already here we need a superblock */ DBG_DUMP_FS(&sblock, "old sblock"); /* * Determine size to grow to. Default to the device size. */ if (size == 0) size = mediasize; else { if (size > (uint64_t)mediasize) { humanize_number(oldsizebuf, sizeof(oldsizebuf), size, "B", HN_AUTOSCALE, HN_B | HN_NOSPACE | HN_DECIMAL); humanize_number(newsizebuf, sizeof(newsizebuf), mediasize, "B", HN_AUTOSCALE, HN_B | HN_NOSPACE | HN_DECIMAL); errx(1, "requested size %s is larger " "than the available %s", oldsizebuf, newsizebuf); } } /* * Make sure the new size is a multiple of fs_fsize; /dev/ufssuspend * only supports fragment-aligned IO requests. */ size -= size % osblock.fs_fsize; if (size <= (uint64_t)(osblock.fs_size * osblock.fs_fsize)) { humanize_number(oldsizebuf, sizeof(oldsizebuf), osblock.fs_size * osblock.fs_fsize, "B", HN_AUTOSCALE, HN_B | HN_NOSPACE | HN_DECIMAL); humanize_number(newsizebuf, sizeof(newsizebuf), size, "B", HN_AUTOSCALE, HN_B | HN_NOSPACE | HN_DECIMAL); errx(1, "requested size %s is not larger than the current " "filesystem size %s", newsizebuf, oldsizebuf); } sblock.fs_size = dbtofsb(&osblock, size / DEV_BSIZE); sblock.fs_providersize = dbtofsb(&osblock, mediasize / DEV_BSIZE); /* * Are we really growing? */ if (osblock.fs_size >= sblock.fs_size) { errx(1, "we are not growing (%jd->%jd)", (intmax_t)osblock.fs_size, (intmax_t)sblock.fs_size); } /* * Check if we find an active snapshot. */ if (yflag == 0) { for (j = 0; j < FSMAXSNAP; j++) { if (sblock.fs_snapinum[j]) { errx(1, "active snapshot found in file system; " "please remove all snapshots before " "using growfs"); } if (!sblock.fs_snapinum[j]) /* list is dense */ break; } } if (yflag == 0 && Nflag == 0) { if (statfsp != NULL && (statfsp->f_flags & MNT_RDONLY) == 0) printf("Device is mounted read-write; resizing will " "result in temporary write suspension for %s.\n", statfsp->f_mntonname); printf("It's strongly recommended to make a backup " "before growing the file system.\n" "OK to grow filesystem on %s", device); if (statfsp != NULL) printf(", mounted on %s,", statfsp->f_mntonname); humanize_number(oldsizebuf, sizeof(oldsizebuf), osblock.fs_size * osblock.fs_fsize, "B", HN_AUTOSCALE, HN_B | HN_NOSPACE | HN_DECIMAL); humanize_number(newsizebuf, sizeof(newsizebuf), sblock.fs_size * sblock.fs_fsize, "B", HN_AUTOSCALE, HN_B | HN_NOSPACE | HN_DECIMAL); printf(" from %s to %s? [Yes/No] ", oldsizebuf, newsizebuf); fflush(stdout); fgets(reply, (int)sizeof(reply), stdin); if (strcasecmp(reply, "Yes\n")){ printf("\nNothing done\n"); exit (0); } } /* * Try to access our device for writing. If it's not mounted, * or mounted read-only, simply open it; otherwise, use UFS * suspension mechanism. */ if (Nflag) { fso = -1; } else { if (statfsp != NULL && (statfsp->f_flags & MNT_RDONLY) == 0) { fso = open(_PATH_UFSSUSPEND, O_RDWR); if (fso == -1) err(1, "unable to open %s", _PATH_UFSSUSPEND); error = ioctl(fso, UFSSUSPEND, &statfsp->f_fsid); if (error != 0) err(1, "UFSSUSPEND"); } else { fso = open(device, O_WRONLY); if (fso < 0) err(1, "%s", device); } } /* * Try to access our new last block in the file system. */ testbuf = malloc(sblock.fs_fsize); if (testbuf == NULL) err(1, "malloc"); rdfs((ufs2_daddr_t)((size - sblock.fs_fsize) / DEV_BSIZE), sblock.fs_fsize, testbuf, fsi); wtfs((ufs2_daddr_t)((size - sblock.fs_fsize) / DEV_BSIZE), sblock.fs_fsize, testbuf, fso, Nflag); free(testbuf); /* * Now calculate new superblock values and check for reasonable * bound for new file system size: * fs_size: is derived from user input * fs_dsize: should get updated in the routines creating or * updating the cylinder groups on the fly * fs_cstotal: should get updated in the routines creating or * updating the cylinder groups */ /* * Update the number of cylinders and cylinder groups in the file system. */ if (sblock.fs_magic == FS_UFS1_MAGIC) { sblock.fs_old_ncyl = sblock.fs_size * sblock.fs_old_nspf / sblock.fs_old_spc; if (sblock.fs_size * sblock.fs_old_nspf > sblock.fs_old_ncyl * sblock.fs_old_spc) sblock.fs_old_ncyl++; } sblock.fs_ncg = howmany(sblock.fs_size, sblock.fs_fpg); /* * Allocate last cylinder group only if there is enough room * for at least one data block. */ if (sblock.fs_size % sblock.fs_fpg != 0 && sblock.fs_size <= cgdmin(&sblock, sblock.fs_ncg - 1)) { humanize_number(oldsizebuf, sizeof(oldsizebuf), (sblock.fs_size % sblock.fs_fpg) * sblock.fs_fsize, "B", HN_AUTOSCALE, HN_B | HN_NOSPACE | HN_DECIMAL); warnx("no room to allocate last cylinder group; " "leaving %s unused", oldsizebuf); sblock.fs_ncg--; if (sblock.fs_magic == FS_UFS1_MAGIC) sblock.fs_old_ncyl = sblock.fs_ncg * sblock.fs_old_cpg; sblock.fs_size = sblock.fs_ncg * sblock.fs_fpg; } /* * Update the space for the cylinder group summary information in the * respective cylinder group data area. */ sblock.fs_cssize = fragroundup(&sblock, sblock.fs_ncg * sizeof(struct csum)); if (osblock.fs_size >= sblock.fs_size) errx(1, "not enough new space"); DBG_PRINT0("sblock calculated\n"); /* * Ok, everything prepared, so now let's do the tricks. */ growfs(fsi, fso, Nflag); close(fsi); if (fso > -1) { if (statfsp != NULL && (statfsp->f_flags & MNT_RDONLY) == 0) { error = ioctl(fso, UFSRESUME); if (error != 0) err(1, "UFSRESUME"); } error = close(fso); if (error != 0) err(1, "close"); if (statfsp != NULL && (statfsp->f_flags & MNT_RDONLY) != 0) mount_reload(statfsp); } DBG_CLOSE; DBG_LEAVE; return (0); } /* * Dump a line of usage. */ static void usage(void) { DBG_FUNC("usage") DBG_ENTER; fprintf(stderr, "usage: growfs [-Ny] [-s size] special | filesystem\n"); DBG_LEAVE; exit(1); } /* * This updates most parameters and the bitmap related to cluster. We have to * assume that sblock, osblock, acg are set up. */ static void updclst(int block) { DBG_FUNC("updclst") static int lcs = 0; DBG_ENTER; if (sblock.fs_contigsumsize < 1) /* no clustering */ return; /* * update cluster allocation map */ setbit(cg_clustersfree(&acg), block); /* * update cluster summary table */ if (!lcs) { /* * calculate size for the trailing cluster */ for (block--; lcs < sblock.fs_contigsumsize; block--, lcs++ ) { if (isclr(cg_clustersfree(&acg), block)) break; } } if (lcs < sblock.fs_contigsumsize) { if (lcs) cg_clustersum(&acg)[lcs]--; lcs++; cg_clustersum(&acg)[lcs]++; } DBG_LEAVE; return; } static void mount_reload(const struct statfs *stfs) { char errmsg[255]; struct iovec *iov; int iovlen; iov = NULL; iovlen = 0; *errmsg = '\0'; build_iovec(&iov, &iovlen, "fstype", __DECONST(char *, "ffs"), 4); build_iovec(&iov, &iovlen, "fspath", __DECONST(char *, stfs->f_mntonname), (size_t)-1); build_iovec(&iov, &iovlen, "errmsg", errmsg, sizeof(errmsg)); build_iovec(&iov, &iovlen, "update", NULL, 0); build_iovec(&iov, &iovlen, "reload", NULL, 0); if (nmount(iov, iovlen, stfs->f_flags) < 0) { errmsg[sizeof(errmsg) - 1] = '\0'; err(9, "%s: cannot reload filesystem%s%s", stfs->f_mntonname, *errmsg != '\0' ? ": " : "", errmsg); } } Index: user/ngie/more-tests/sbin =================================================================== --- user/ngie/more-tests/sbin (revision 281974) +++ user/ngie/more-tests/sbin (revision 281975) Property changes on: user/ngie/more-tests/sbin ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sbin:r281925-281974 Index: user/ngie/more-tests/share/mk/src.opts.mk =================================================================== --- user/ngie/more-tests/share/mk/src.opts.mk (revision 281974) +++ user/ngie/more-tests/share/mk/src.opts.mk (revision 281975) @@ -1,394 +1,398 @@ # $FreeBSD$ # # Option file for FreeBSD /usr/src builds. # # Users define WITH_FOO and WITHOUT_FOO on the command line or in /etc/src.conf # and /etc/make.conf files. These translate in the build system to MK_FOO={yes,no} # with sensible (usually) defaults. # # Makefiles must include bsd.opts.mk after defining specific MK_FOO options that # are applicable for that Makefile (typically there are none, but sometimes there # are exceptions). Recursive makes usually add MK_FOO=no for options that they wish # to omit from that make. # # Makefiles must include bsd.mkopt.mk before they test the value of any MK_FOO # variable. # # Makefiles may also assume that this file is included by src.opts.mk should it # need variables defined there prior to the end of the Makefile where # bsd.{subdir,lib.bin}.mk is traditionally included. # # The old-style YES_FOO and NO_FOO are being phased out. No new instances of them # should be added. Old instances should be removed since they were just to # bridge the gap between FreeBSD 4 and FreeBSD 5. # # Makefiles should never test WITH_FOO or WITHOUT_FOO directly (although an # exception is made for _WITHOUT_SRCONF which turns off this mechanism # completely inside bsd.*.mk files). # .if !target(____) ____: .include # # Define MK_* variables (which are either "yes" or "no") for users # to set via WITH_*/WITHOUT_* in /etc/src.conf and override in the # make(1) environment. # These should be tested with `== "no"' or `!= "no"' in makefiles. # The NO_* variables should only be set by makefiles for variables # that haven't been converted over. # # These options are used by src the builds __DEFAULT_YES_OPTIONS = \ ACCT \ ACPI \ AMD \ APM \ AT \ ATM \ AUDIT \ AUTHPF \ AUTOFS \ BHYVE \ BINUTILS \ BINUTILS_BOOTSTRAP \ BLUETOOTH \ BOOT \ BOOTPARAMD \ BOOTPD \ BSD_CPIO \ BSDINSTALL \ BSNMP \ BZIP2 \ CALENDAR \ CAPSICUM \ CASPER \ CCD \ CDDL \ CPP \ CROSS_COMPILER \ CRYPT \ CTM \ CUSE \ CXX \ DICT \ DMAGENT \ DYNAMICROOT \ ED_CRYPTO \ EE \ ELFTOOLCHAIN_TOOLS \ EXAMPLES \ FDT \ FILE \ FINGER \ FLOPPY \ FMTREE \ FORTH \ FP_LIBC \ FREEBSD_UPDATE \ FTP \ GAMES \ GCOV \ GDB \ GNU \ GNU_GREP_COMPAT \ GPIO \ GPL_DTC \ GROFF \ HAST \ HTML \ HYPERV \ ICONV \ INET \ INET6 \ INETD \ IPFILTER \ IPFW \ ISCSI \ JAIL \ KDUMP \ KVM \ LDNS \ LDNS_UTILS \ LEGACY_CONSOLE \ LIB32 \ LIBPTHREAD \ LIBTHR \ LOCALES \ LOCATE \ LPR \ LS_COLORS \ LZMA_SUPPORT \ MAIL \ MAILWRAPPER \ MAKE \ NDIS \ NETCAT \ NETGRAPH \ NLS_CATALOGS \ NS_CACHING \ NTP \ OPENSSL \ PAM \ PC_SYSINSTALL \ PF \ PKGBOOTSTRAP \ PMC \ PORTSNAP \ PPP \ QUOTAS \ RADIUS_SUPPORT \ RCMDS \ RBOOTD \ RCS \ RESCUE \ ROUTED \ SENDMAIL \ SETUID_LOGIN \ SHAREDOCS \ SOURCELESS \ SOURCELESS_HOST \ SOURCELESS_UCODE \ SVNLITE \ SYSCONS \ SYSINSTALL \ TALK \ TCP_WRAPPERS \ TCSH \ TELNET \ TESTS \ TEXTPROC \ TFTP \ TIMED \ UNBOUND \ USB \ UTMPX \ VI \ VT \ WIRELESS \ WPA_SUPPLICANT_EAPOL \ ZFS \ ZONEINFO __DEFAULT_NO_OPTIONS = \ BSD_GREP \ CLANG_EXTRAS \ EISA \ FMAKE \ HESIOD \ LLDB \ NAND \ OFED \ OPENLDAP \ OPENSSH_NONE_CIPHER \ SHARED_TOOLCHAIN \ SORT_THREADS \ SVN # # Default behaviour of some options depends on the architecture. Unfortunately # this means that we have to test TARGET_ARCH (the buildworld case) as well # as MACHINE_ARCH (the non-buildworld case). Normally TARGET_ARCH is not # used at all in bsd.*.mk, but we have to make an exception here if we want # to allow defaults for some things like clang to vary by target architecture. # Additional, per-target behavior should be rarely added only after much # gnashing of teeth and grinding of gears. # .if defined(TARGET_ARCH) __T=${TARGET_ARCH} .else __T=${MACHINE_ARCH} .endif .if defined(TARGET) __TT=${TARGET} .else __TT=${MACHINE} .endif .include .if !${COMPILER_FEATURES:Mc++11} # If the compiler is not C++11 capable, disable clang and use gcc instead. __DEFAULT_YES_OPTIONS+=GCC GCC_BOOTSTRAP GNUCXX __DEFAULT_NO_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_FULL CLANG_IS_CC .elif ${__T} == "aarch64" || ${__T} == "amd64" || ${__T} == "i386" # On x86 and arm64, clang is enabled, and will be installed as the default cc. __DEFAULT_YES_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_FULL CLANG_IS_CC __DEFAULT_NO_OPTIONS+=GCC GCC_BOOTSTRAP GNUCXX .elif ${__TT} == "arm" && ${__T:Marm*eb*} == "" # On little-endian arm, clang is enabled, and it is installed as the default # cc, but since gcc is unable to build the full clang, disable it by default. __DEFAULT_YES_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_IS_CC __DEFAULT_NO_OPTIONS+=CLANG_FULL GCC GCC_BOOTSTRAP GNUCXX .elif ${__T:Mpowerpc*} # On powerpc, clang is enabled, but gcc is installed as the default cc. __DEFAULT_YES_OPTIONS+=CLANG CLANG_FULL GCC GCC_BOOTSTRAP GNUCXX __DEFAULT_NO_OPTIONS+=CLANG_BOOTSTRAP CLANG_IS_CC .else # Everything else disables clang, and uses gcc instead. __DEFAULT_YES_OPTIONS+=GCC GCC_BOOTSTRAP GNUCXX __DEFAULT_NO_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_FULL CLANG_IS_CC .endif .if ${__T} == "aarch64" BROKEN_OPTIONS+=BINUTILS BINUTILS_BOOTSTRAP GDB .endif +# LLVM lacks support for FreeBSD 64-bit atomic operations for ARMv4/ARMv5 +.if ${__T} == "arm" || ${__T} == "armeb" +BROKEN_OPTIONS+=LLDB +.endif .include # # MK_* options that default to "yes" if the compiler is a C++11 compiler. # .for var in \ LIBCPLUSPLUS .if !defined(MK_${var}) .if ${COMPILER_FEATURES:Mc++11} .if defined(WITHOUT_${var}) MK_${var}:= no .else MK_${var}:= yes .endif .else .if defined(WITH_${var}) MK_${var}:= yes .else MK_${var}:= no .endif .endif .endif .endfor # # Force some options off if their dependencies are off. # Order is somewhat important. # .if ${MK_LIBPTHREAD} == "no" MK_LIBTHR:= no .endif .if ${MK_LDNS} == "no" MK_LDNS_UTILS:= no MK_UNBOUND:= no .endif .if ${MK_SOURCELESS} == "no" MK_SOURCELESS_HOST:= no MK_SOURCELESS_UCODE:= no .endif .if ${MK_CDDL} == "no" MK_ZFS:= no MK_CTF:= no .endif .if ${MK_CRYPT} == "no" MK_OPENSSL:= no MK_OPENSSH:= no MK_KERBEROS:= no .endif .if ${MK_CXX} == "no" MK_CLANG:= no MK_GROFF:= no MK_GNUCXX:= no .endif .if ${MK_MAIL} == "no" MK_MAILWRAPPER:= no MK_SENDMAIL:= no MK_DMAGENT:= no .endif .if ${MK_NETGRAPH} == "no" MK_ATM:= no MK_BLUETOOTH:= no .endif .if ${MK_OPENSSL} == "no" MK_OPENSSH:= no MK_KERBEROS:= no .endif .if ${MK_PF} == "no" MK_AUTHPF:= no .endif .if ${MK_TEXTPROC} == "no" MK_GROFF:= no .endif .if ${MK_CROSS_COMPILER} == "no" MK_BINUTILS_BOOTSTRAP:= no MK_CLANG_BOOTSTRAP:= no MK_GCC_BOOTSTRAP:= no .endif .if ${MK_TOOLCHAIN} == "no" MK_BINUTILS:= no MK_CLANG:= no MK_GCC:= no MK_GDB:= no MK_INCLUDES:= no .endif .if ${MK_CLANG} == "no" MK_CLANG_EXTRAS:= no MK_CLANG_FULL:= no .endif # # Set defaults for the MK_*_SUPPORT variables. # # # MK_*_SUPPORT options which default to "yes" unless their corresponding # MK_* variable is set to "no". # .for var in \ BZIP2 \ GNU \ INET \ INET6 \ KERBEROS \ KVM \ NETGRAPH \ PAM \ TESTS \ WIRELESS .if defined(WITHOUT_${var}_SUPPORT) || ${MK_${var}} == "no" MK_${var}_SUPPORT:= no .else MK_${var}_SUPPORT:= yes .endif .endfor # # MK_* options whose default value depends on another option. # .for vv in \ GSSAPI/KERBEROS \ MAN_UTILS/MAN .if defined(WITH_${vv:H}) MK_${vv:H}:= yes .elif defined(WITHOUT_${vv:H}) MK_${vv:H}:= no .else MK_${vv:H}:= ${MK_${vv:T}} .endif .endfor .if !${COMPILER_FEATURES:Mc++11} MK_LLDB:= no .endif # gcc 4.8 and newer supports libc++, so suppress gnuc++ in that case. # while in theory we could build it with that, we don't want to do # that since it creates too much confusion for too little gain. .if ${COMPILER_TYPE} == "gcc" && ${COMPILER_VERSION} >= 40800 MK_GNUCXX:=no MK_GCC:=no .endif .endif # !target(____) Index: user/ngie/more-tests/share =================================================================== --- user/ngie/more-tests/share (revision 281974) +++ user/ngie/more-tests/share (revision 281975) Property changes on: user/ngie/more-tests/share ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/share:r281925-281974 Index: user/ngie/more-tests/sys/amd64/amd64/mp_machdep.c =================================================================== --- user/ngie/more-tests/sys/amd64/amd64/mp_machdep.c (revision 281974) +++ user/ngie/more-tests/sys/amd64/amd64/mp_machdep.c (revision 281975) @@ -1,1740 +1,730 @@ /*- * Copyright (c) 1996, by Steve Passe * Copyright (c) 2003, by Peter Wemm * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. The name of the developer may NOT be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_cpu.h" #include "opt_ddb.h" #include "opt_kstack_pages.h" #include "opt_sched.h" #include "opt_smp.h" #include #include #include #include #ifdef GPROF #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define WARMBOOT_TARGET 0 #define WARMBOOT_OFF (KERNBASE + 0x0467) #define WARMBOOT_SEG (KERNBASE + 0x0469) #define CMOS_REG (0x70) #define CMOS_DATA (0x71) #define BIOS_RESET (0x0f) #define BIOS_WARM (0x0a) -/* lock region used by kernel profiling */ -int mcount_lock; +extern struct pcpu __pcpu[]; -int mp_naps; /* # of Applications processors */ -int boot_cpu_id = -1; /* designated BSP */ - -extern struct pcpu __pcpu[]; - -/* AP uses this during bootstrap. Do not staticize. */ -char *bootSTK; -int bootAP; - -/* Free these after use */ -void *bootstacks[MAXCPU]; - /* Temporary variables for init_secondary() */ char *doublefault_stack; char *nmi_stack; -void *dpcpu; -struct pcb stoppcbs[MAXCPU]; -struct susppcb **susppcbs; - /* Variables needed for SMP tlb shootdown. */ vm_offset_t smp_tlb_addr2; struct invpcid_descr smp_tlb_invpcid; volatile int smp_tlb_wait; uint64_t pcid_cr3; pmap_t smp_tlb_pmap; extern int invpcid_works; -#ifdef COUNT_IPIS -/* Interrupt counts. */ -static u_long *ipi_preempt_counts[MAXCPU]; -static u_long *ipi_ast_counts[MAXCPU]; -u_long *ipi_invltlb_counts[MAXCPU]; -u_long *ipi_invlrng_counts[MAXCPU]; -u_long *ipi_invlpg_counts[MAXCPU]; -u_long *ipi_invlcache_counts[MAXCPU]; -u_long *ipi_rendezvous_counts[MAXCPU]; -static u_long *ipi_hardclock_counts[MAXCPU]; -#endif - -/* Default cpu_ops implementation. */ -struct cpu_ops cpu_ops; - extern inthand_t IDTVEC(fast_syscall), IDTVEC(fast_syscall32); -extern int pmap_pcid_enabled; - /* * Local data and functions. */ -static volatile cpuset_t ipi_nmi_pending; - -/* used to hold the AP's until we are ready to release them */ -struct mtx ap_boot_mtx; - -/* Set to 1 once we're ready to let the APs out of the pen. */ -static volatile int aps_ready = 0; - -/* - * Store data from cpu_add() until later in the boot when we actually setup - * the APs. - */ -struct cpu_info { - int cpu_present:1; - int cpu_bsp:1; - int cpu_disabled:1; - int cpu_hyperthread:1; -} static cpu_info[MAX_APIC_ID + 1]; -int cpu_apic_ids[MAXCPU]; -int apic_cpuids[MAX_APIC_ID + 1]; - -/* Holds pending bitmap based IPIs per CPU */ -volatile u_int cpu_ipi_pending[MAXCPU]; - -static u_int boot_address; -static int cpu_logical; /* logical cpus per core */ -static int cpu_cores; /* cores per package */ - -static void assign_cpu_ids(void); -static void set_interrupt_apic_ids(void); static int start_ap(int apic_id); -static void release_aps(void *dummy); -static u_int hyperthreading_cpus; /* logical cpus sharing L1 cache */ -static int hyperthreading_allowed = 1; static u_int bootMP_size; +static u_int boot_address; -static void -mem_range_AP_init(void) -{ - if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP) - mem_range_softc.mr_op->initAP(&mem_range_softc); -} - -static void -topo_probe_amd(void) -{ - int core_id_bits; - int id; - - /* AMD processors do not support HTT. */ - cpu_logical = 1; - - if ((amd_feature2 & AMDID2_CMP) == 0) { - cpu_cores = 1; - return; - } - - core_id_bits = (cpu_procinfo2 & AMDID_COREID_SIZE) >> - AMDID_COREID_SIZE_SHIFT; - if (core_id_bits == 0) { - cpu_cores = (cpu_procinfo2 & AMDID_CMP_CORES) + 1; - return; - } - - /* Fam 10h and newer should get here. */ - for (id = 0; id <= MAX_APIC_ID; id++) { - /* Check logical CPU availability. */ - if (!cpu_info[id].cpu_present || cpu_info[id].cpu_disabled) - continue; - /* Check if logical CPU has the same package ID. */ - if ((id >> core_id_bits) != (boot_cpu_id >> core_id_bits)) - continue; - cpu_cores++; - } -} - /* - * Round up to the next power of two, if necessary, and then - * take log2. - * Returns -1 if argument is zero. - */ -static __inline int -mask_width(u_int x) -{ - - return (fls(x << (1 - powerof2(x))) - 1); -} - -static void -topo_probe_0x4(void) -{ - u_int p[4]; - int pkg_id_bits; - int core_id_bits; - int max_cores; - int max_logical; - int id; - - /* Both zero and one here mean one logical processor per package. */ - max_logical = (cpu_feature & CPUID_HTT) != 0 ? - (cpu_procinfo & CPUID_HTT_CORES) >> 16 : 1; - if (max_logical <= 1) - return; - - /* - * Because of uniformity assumption we examine only - * those logical processors that belong to the same - * package as BSP. Further, we count number of - * logical processors that belong to the same core - * as BSP thus deducing number of threads per core. - */ - if (cpu_high >= 0x4) { - cpuid_count(0x04, 0, p); - max_cores = ((p[0] >> 26) & 0x3f) + 1; - } else - max_cores = 1; - core_id_bits = mask_width(max_logical/max_cores); - if (core_id_bits < 0) - return; - pkg_id_bits = core_id_bits + mask_width(max_cores); - - for (id = 0; id <= MAX_APIC_ID; id++) { - /* Check logical CPU availability. */ - if (!cpu_info[id].cpu_present || cpu_info[id].cpu_disabled) - continue; - /* Check if logical CPU has the same package ID. */ - if ((id >> pkg_id_bits) != (boot_cpu_id >> pkg_id_bits)) - continue; - cpu_cores++; - /* Check if logical CPU has the same package and core IDs. */ - if ((id >> core_id_bits) == (boot_cpu_id >> core_id_bits)) - cpu_logical++; - } - - KASSERT(cpu_cores >= 1 && cpu_logical >= 1, - ("topo_probe_0x4 couldn't find BSP")); - - cpu_cores /= cpu_logical; - hyperthreading_cpus = cpu_logical; -} - -static void -topo_probe_0xb(void) -{ - u_int p[4]; - int bits; - int cnt; - int i; - int logical; - int type; - int x; - - /* We only support three levels for now. */ - for (i = 0; i < 3; i++) { - cpuid_count(0x0b, i, p); - - /* Fall back if CPU leaf 11 doesn't really exist. */ - if (i == 0 && p[1] == 0) { - topo_probe_0x4(); - return; - } - - bits = p[0] & 0x1f; - logical = p[1] &= 0xffff; - type = (p[2] >> 8) & 0xff; - if (type == 0 || logical == 0) - break; - /* - * Because of uniformity assumption we examine only - * those logical processors that belong to the same - * package as BSP. - */ - for (cnt = 0, x = 0; x <= MAX_APIC_ID; x++) { - if (!cpu_info[x].cpu_present || - cpu_info[x].cpu_disabled) - continue; - if (x >> bits == boot_cpu_id >> bits) - cnt++; - } - if (type == CPUID_TYPE_SMT) - cpu_logical = cnt; - else if (type == CPUID_TYPE_CORE) - cpu_cores = cnt; - } - if (cpu_logical == 0) - cpu_logical = 1; - cpu_cores /= cpu_logical; -} - -/* - * Both topology discovery code and code that consumes topology - * information assume top-down uniformity of the topology. - * That is, all physical packages must be identical and each - * core in a package must have the same number of threads. - * Topology information is queried only on BSP, on which this - * code runs and for which it can query CPUID information. - * Then topology is extrapolated on all packages using the - * uniformity assumption. - */ -static void -topo_probe(void) -{ - static int cpu_topo_probed = 0; - - if (cpu_topo_probed) - return; - - CPU_ZERO(&logical_cpus_mask); - if (mp_ncpus <= 1) - cpu_cores = cpu_logical = 1; - else if (cpu_vendor_id == CPU_VENDOR_AMD) - topo_probe_amd(); - else if (cpu_vendor_id == CPU_VENDOR_INTEL) { - /* - * See Intel(R) 64 Architecture Processor - * Topology Enumeration article for details. - * - * Note that 0x1 <= cpu_high < 4 case should be - * compatible with topo_probe_0x4() logic when - * CPUID.1:EBX[23:16] > 0 (cpu_cores will be 1) - * or it should trigger the fallback otherwise. - */ - if (cpu_high >= 0xb) - topo_probe_0xb(); - else if (cpu_high >= 0x1) - topo_probe_0x4(); - } - - /* - * Fallback: assume each logical CPU is in separate - * physical package. That is, no multi-core, no SMT. - */ - if (cpu_cores == 0 || cpu_logical == 0) - cpu_cores = cpu_logical = 1; - cpu_topo_probed = 1; -} - -struct cpu_group * -cpu_topo(void) -{ - int cg_flags; - - /* - * Determine whether any threading flags are - * necessry. - */ - topo_probe(); - if (cpu_logical > 1 && hyperthreading_cpus) - cg_flags = CG_FLAG_HTT; - else if (cpu_logical > 1) - cg_flags = CG_FLAG_SMT; - else - cg_flags = 0; - if (mp_ncpus % (cpu_cores * cpu_logical) != 0) { - printf("WARNING: Non-uniform processors.\n"); - printf("WARNING: Using suboptimal topology.\n"); - return (smp_topo_none()); - } - /* - * No multi-core or hyper-threaded. - */ - if (cpu_logical * cpu_cores == 1) - return (smp_topo_none()); - /* - * Only HTT no multi-core. - */ - if (cpu_logical > 1 && cpu_cores == 1) - return (smp_topo_1level(CG_SHARE_L1, cpu_logical, cg_flags)); - /* - * Only multi-core no HTT. - */ - if (cpu_cores > 1 && cpu_logical == 1) - return (smp_topo_1level(CG_SHARE_L2, cpu_cores, cg_flags)); - /* - * Both HTT and multi-core. - */ - return (smp_topo_2level(CG_SHARE_L2, cpu_cores, - CG_SHARE_L1, cpu_logical, cg_flags)); -} - -/* * Calculate usable address in base memory for AP trampoline code. */ u_int mp_bootaddress(u_int basemem) { bootMP_size = mptramp_end - mptramp_start; boot_address = trunc_page(basemem * 1024); /* round down to 4k boundary */ if (((basemem * 1024) - boot_address) < bootMP_size) boot_address -= PAGE_SIZE; /* not enough, lower by 4k */ /* 3 levels of page table pages */ mptramp_pagetables = boot_address - (PAGE_SIZE * 3); return mptramp_pagetables; } -void -cpu_add(u_int apic_id, char boot_cpu) -{ - - if (apic_id > MAX_APIC_ID) { - panic("SMP: APIC ID %d too high", apic_id); - return; - } - KASSERT(cpu_info[apic_id].cpu_present == 0, ("CPU %d added twice", - apic_id)); - cpu_info[apic_id].cpu_present = 1; - if (boot_cpu) { - KASSERT(boot_cpu_id == -1, - ("CPU %d claims to be BSP, but CPU %d already is", apic_id, - boot_cpu_id)); - boot_cpu_id = apic_id; - cpu_info[apic_id].cpu_bsp = 1; - } - if (mp_ncpus < MAXCPU) { - mp_ncpus++; - mp_maxid = mp_ncpus - 1; - } - if (bootverbose) - printf("SMP: Added CPU %d (%s)\n", apic_id, boot_cpu ? "BSP" : - "AP"); -} - -void -cpu_mp_setmaxid(void) -{ - - /* - * mp_maxid should be already set by calls to cpu_add(). - * Just sanity check its value here. - */ - if (mp_ncpus == 0) - KASSERT(mp_maxid == 0, - ("%s: mp_ncpus is zero, but mp_maxid is not", __func__)); - else if (mp_ncpus == 1) - mp_maxid = 0; - else - KASSERT(mp_maxid >= mp_ncpus - 1, - ("%s: counters out of sync: max %d, count %d", __func__, - mp_maxid, mp_ncpus)); -} - -int -cpu_mp_probe(void) -{ - - /* - * Always record BSP in CPU map so that the mbuf init code works - * correctly. - */ - CPU_SETOF(0, &all_cpus); - if (mp_ncpus == 0) { - /* - * No CPUs were found, so this must be a UP system. Setup - * the variables to represent a system with a single CPU - * with an id of 0. - */ - mp_ncpus = 1; - return (0); - } - - /* At least one CPU was found. */ - if (mp_ncpus == 1) { - /* - * One CPU was found, so this must be a UP system with - * an I/O APIC. - */ - mp_maxid = 0; - return (0); - } - - /* At least two CPUs were found. */ - return (1); -} - /* * Initialize the IPI handlers and start up the AP's. */ void cpu_mp_start(void) { int i; /* Initialize the logical ID to APIC ID table. */ for (i = 0; i < MAXCPU; i++) { cpu_apic_ids[i] = -1; cpu_ipi_pending[i] = 0; } /* Install an inter-CPU IPI for TLB invalidation */ if (pmap_pcid_enabled) { setidt(IPI_INVLTLB, IDTVEC(invltlb_pcid), SDT_SYSIGT, SEL_KPL, 0); setidt(IPI_INVLPG, IDTVEC(invlpg_pcid), SDT_SYSIGT, SEL_KPL, 0); } else { setidt(IPI_INVLTLB, IDTVEC(invltlb), SDT_SYSIGT, SEL_KPL, 0); setidt(IPI_INVLPG, IDTVEC(invlpg), SDT_SYSIGT, SEL_KPL, 0); } setidt(IPI_INVLRNG, IDTVEC(invlrng), SDT_SYSIGT, SEL_KPL, 0); /* Install an inter-CPU IPI for cache invalidation. */ setidt(IPI_INVLCACHE, IDTVEC(invlcache), SDT_SYSIGT, SEL_KPL, 0); /* Install an inter-CPU IPI for all-CPU rendezvous */ setidt(IPI_RENDEZVOUS, IDTVEC(rendezvous), SDT_SYSIGT, SEL_KPL, 0); /* Install generic inter-CPU IPI handler */ setidt(IPI_BITMAP_VECTOR, IDTVEC(ipi_intr_bitmap_handler), SDT_SYSIGT, SEL_KPL, 0); /* Install an inter-CPU IPI for CPU stop/restart */ setidt(IPI_STOP, IDTVEC(cpustop), SDT_SYSIGT, SEL_KPL, 0); /* Install an inter-CPU IPI for CPU suspend/resume */ setidt(IPI_SUSPEND, IDTVEC(cpususpend), SDT_SYSIGT, SEL_KPL, 0); /* Set boot_cpu_id if needed. */ if (boot_cpu_id == -1) { boot_cpu_id = PCPU_GET(apic_id); cpu_info[boot_cpu_id].cpu_bsp = 1; } else KASSERT(boot_cpu_id == PCPU_GET(apic_id), ("BSP's APIC ID doesn't match boot_cpu_id")); /* Probe logical/physical core configuration. */ topo_probe(); assign_cpu_ids(); /* Start each Application Processor */ init_ops.start_all_aps(); set_interrupt_apic_ids(); } /* - * Print various information about the SMP system hardware and setup. - */ -void -cpu_mp_announce(void) -{ - const char *hyperthread; - int i; - - printf("FreeBSD/SMP: %d package(s) x %d core(s)", - mp_ncpus / (cpu_cores * cpu_logical), cpu_cores); - if (hyperthreading_cpus > 1) - printf(" x %d HTT threads", cpu_logical); - else if (cpu_logical > 1) - printf(" x %d SMT threads", cpu_logical); - printf("\n"); - - /* List active CPUs first. */ - printf(" cpu0 (BSP): APIC ID: %2d\n", boot_cpu_id); - for (i = 1; i < mp_ncpus; i++) { - if (cpu_info[cpu_apic_ids[i]].cpu_hyperthread) - hyperthread = "/HT"; - else - hyperthread = ""; - printf(" cpu%d (AP%s): APIC ID: %2d\n", i, hyperthread, - cpu_apic_ids[i]); - } - - /* List disabled CPUs last. */ - for (i = 0; i <= MAX_APIC_ID; i++) { - if (!cpu_info[i].cpu_present || !cpu_info[i].cpu_disabled) - continue; - if (cpu_info[i].cpu_hyperthread) - hyperthread = "/HT"; - else - hyperthread = ""; - printf(" cpu (AP%s): APIC ID: %2d (disabled)\n", hyperthread, - i); - } -} - -/* * AP CPU's call this to initialize themselves. */ void init_secondary(void) { struct pcpu *pc; struct nmi_pcpu *np; u_int64_t msr, cr0; - u_int cpuid; int cpu, gsel_tss, x; struct region_descriptor ap_gdt; /* Set by the startup code for us to use */ cpu = bootAP; /* Init tss */ common_tss[cpu] = common_tss[0]; common_tss[cpu].tss_rsp0 = 0; /* not used until after switch */ common_tss[cpu].tss_iobase = sizeof(struct amd64tss) + IOPERM_BITMAP_SIZE; common_tss[cpu].tss_ist1 = (long)&doublefault_stack[PAGE_SIZE]; /* The NMI stack runs on IST2. */ np = ((struct nmi_pcpu *) &nmi_stack[PAGE_SIZE]) - 1; common_tss[cpu].tss_ist2 = (long) np; /* Prepare private GDT */ gdt_segs[GPROC0_SEL].ssd_base = (long) &common_tss[cpu]; for (x = 0; x < NGDT; x++) { if (x != GPROC0_SEL && x != (GPROC0_SEL + 1) && x != GUSERLDT_SEL && x != (GUSERLDT_SEL + 1)) ssdtosd(&gdt_segs[x], &gdt[NGDT * cpu + x]); } ssdtosyssd(&gdt_segs[GPROC0_SEL], (struct system_segment_descriptor *)&gdt[NGDT * cpu + GPROC0_SEL]); ap_gdt.rd_limit = NGDT * sizeof(gdt[0]) - 1; ap_gdt.rd_base = (long) &gdt[NGDT * cpu]; lgdt(&ap_gdt); /* does magic intra-segment return */ /* Get per-cpu data */ pc = &__pcpu[cpu]; /* prime data page for it to use */ pcpu_init(pc, cpu, sizeof(struct pcpu)); dpcpu_init(dpcpu, cpu); pc->pc_apic_id = cpu_apic_ids[cpu]; pc->pc_prvspace = pc; pc->pc_curthread = 0; pc->pc_tssp = &common_tss[cpu]; pc->pc_commontssp = &common_tss[cpu]; pc->pc_rsp0 = 0; pc->pc_tss = (struct system_segment_descriptor *)&gdt[NGDT * cpu + GPROC0_SEL]; pc->pc_fs32p = &gdt[NGDT * cpu + GUFS32_SEL]; pc->pc_gs32p = &gdt[NGDT * cpu + GUGS32_SEL]; pc->pc_ldt = (struct system_segment_descriptor *)&gdt[NGDT * cpu + GUSERLDT_SEL]; /* Save the per-cpu pointer for use by the NMI handler. */ np->np_pcpu = (register_t) pc; wrmsr(MSR_FSBASE, 0); /* User value */ wrmsr(MSR_GSBASE, (u_int64_t)pc); wrmsr(MSR_KGSBASE, (u_int64_t)pc); /* XXX User value while we're in the kernel */ lidt(&r_idt); gsel_tss = GSEL(GPROC0_SEL, SEL_KPL); ltr(gsel_tss); /* * Set to a known state: * Set by mpboot.s: CR0_PG, CR0_PE * Set by cpu_setregs: CR0_NE, CR0_MP, CR0_TS, CR0_WP, CR0_AM */ cr0 = rcr0(); cr0 &= ~(CR0_CD | CR0_NW | CR0_EM); load_cr0(cr0); /* Set up the fast syscall stuff */ msr = rdmsr(MSR_EFER) | EFER_SCE; wrmsr(MSR_EFER, msr); wrmsr(MSR_LSTAR, (u_int64_t)IDTVEC(fast_syscall)); wrmsr(MSR_CSTAR, (u_int64_t)IDTVEC(fast_syscall32)); msr = ((u_int64_t)GSEL(GCODE_SEL, SEL_KPL) << 32) | ((u_int64_t)GSEL(GUCODE32_SEL, SEL_UPL) << 48); wrmsr(MSR_STAR, msr); wrmsr(MSR_SF_MASK, PSL_NT|PSL_T|PSL_I|PSL_C|PSL_D); /* signal our startup to the BSP. */ mp_naps++; /* Spin until the BSP releases the AP's. */ while (!aps_ready) ia32_pause(); - /* - * On real hardware, switch to x2apic mode if possible. Do it - * after aps_ready was signalled, to avoid manipulating the - * mode while BSP might still want to send some IPI to us - * (second startup IPI is ignored on modern hardware etc). - */ - lapic_xapic_mode(); - - /* Initialize the PAT MSR. */ - pmap_init_pat(); - - /* set up CPU registers and state */ - cpu_setregs(); - - /* set up SSE/NX */ - initializecpu(); - - /* set up FPU state on the AP */ - fpuinit(); - - if (cpu_ops.cpu_init) - cpu_ops.cpu_init(); - - /* A quick check from sanity claus */ - cpuid = PCPU_GET(cpuid); - if (PCPU_GET(apic_id) != lapic_id()) { - printf("SMP: cpuid = %d\n", cpuid); - printf("SMP: actual apic_id = %d\n", lapic_id()); - printf("SMP: correct apic_id = %d\n", PCPU_GET(apic_id)); - panic("cpuid mismatch! boom!!"); - } - - /* Initialize curthread. */ - KASSERT(PCPU_GET(idlethread) != NULL, ("no idle thread")); - PCPU_SET(curthread, PCPU_GET(idlethread)); - - mca_init(); - - mtx_lock_spin(&ap_boot_mtx); - - /* Init local apic for irq's */ - lapic_setup(1); - - /* Set memory range attributes for this CPU to match the BSP */ - mem_range_AP_init(); - - smp_cpus++; - - CTR1(KTR_SMP, "SMP: AP CPU #%d Launched", cpuid); - printf("SMP: AP CPU #%d Launched!\n", cpuid); - - /* Determine if we are a logical CPU. */ - /* XXX Calculation depends on cpu_logical being a power of 2, e.g. 2 */ - if (cpu_logical > 1 && PCPU_GET(apic_id) % cpu_logical != 0) - CPU_SET(cpuid, &logical_cpus_mask); - - if (bootverbose) - lapic_dump("AP"); - - if (smp_cpus == mp_ncpus) { - /* enable IPI's, tlb shootdown, freezes etc */ - atomic_store_rel_int(&smp_started, 1); - } - - /* - * Enable global pages TLB extension - * This also implicitly flushes the TLB - */ - - load_cr4(rcr4() | CR4_PGE); - if (pmap_pcid_enabled) - load_cr4(rcr4() | CR4_PCIDE); - load_ds(_udatasel); - load_es(_udatasel); - load_fs(_ufssel); - mtx_unlock_spin(&ap_boot_mtx); - - /* Wait until all the AP's are up. */ - while (smp_started == 0) - ia32_pause(); - - /* Start per-CPU event timers. */ - cpu_initclocks_ap(); - - sched_throw(NULL); - - panic("scheduler returned us to %s", __func__); - /* NOTREACHED */ + init_secondary_tail(); } /******************************************************************* * local functions and data */ /* - * We tell the I/O APIC code about all the CPUs we want to receive - * interrupts. If we don't want certain CPUs to receive IRQs we - * can simply not tell the I/O APIC code about them in this function. - * We also do not tell it about the BSP since it tells itself about - * the BSP internally to work with UP kernels and on UP machines. - */ -static void -set_interrupt_apic_ids(void) -{ - u_int i, apic_id; - - for (i = 0; i < MAXCPU; i++) { - apic_id = cpu_apic_ids[i]; - if (apic_id == -1) - continue; - if (cpu_info[apic_id].cpu_bsp) - continue; - if (cpu_info[apic_id].cpu_disabled) - continue; - - /* Don't let hyperthreads service interrupts. */ - if (cpu_logical > 1 && - apic_id % cpu_logical != 0) - continue; - - intr_add_cpu(i); - } -} - -/* - * Assign logical CPU IDs to local APICs. - */ -static void -assign_cpu_ids(void) -{ - u_int i; - - TUNABLE_INT_FETCH("machdep.hyperthreading_allowed", - &hyperthreading_allowed); - - /* Check for explicitly disabled CPUs. */ - for (i = 0; i <= MAX_APIC_ID; i++) { - if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp) - continue; - - if (hyperthreading_cpus > 1 && i % hyperthreading_cpus != 0) { - cpu_info[i].cpu_hyperthread = 1; - - /* - * Don't use HT CPU if it has been disabled by a - * tunable. - */ - if (hyperthreading_allowed == 0) { - cpu_info[i].cpu_disabled = 1; - continue; - } - } - - /* Don't use this CPU if it has been disabled by a tunable. */ - if (resource_disabled("lapic", i)) { - cpu_info[i].cpu_disabled = 1; - continue; - } - } - - if (hyperthreading_allowed == 0 && hyperthreading_cpus > 1) { - hyperthreading_cpus = 0; - cpu_logical = 1; - } - - /* - * Assign CPU IDs to local APIC IDs and disable any CPUs - * beyond MAXCPU. CPU 0 is always assigned to the BSP. - * - * To minimize confusion for userland, we attempt to number - * CPUs such that all threads and cores in a package are - * grouped together. For now we assume that the BSP is always - * the first thread in a package and just start adding APs - * starting with the BSP's APIC ID. - */ - mp_ncpus = 1; - cpu_apic_ids[0] = boot_cpu_id; - apic_cpuids[boot_cpu_id] = 0; - for (i = boot_cpu_id + 1; i != boot_cpu_id; - i == MAX_APIC_ID ? i = 0 : i++) { - if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp || - cpu_info[i].cpu_disabled) - continue; - - if (mp_ncpus < MAXCPU) { - cpu_apic_ids[mp_ncpus] = i; - apic_cpuids[i] = mp_ncpus; - mp_ncpus++; - } else - cpu_info[i].cpu_disabled = 1; - } - KASSERT(mp_maxid >= mp_ncpus - 1, - ("%s: counters out of sync: max %d, count %d", __func__, mp_maxid, - mp_ncpus)); -} - -/* * start each AP in our list */ int native_start_all_aps(void) { vm_offset_t va = boot_address + KERNBASE; u_int64_t *pt4, *pt3, *pt2; u_int32_t mpbioswarmvec; int apic_id, cpu, i; u_char mpbiosreason; mtx_init(&ap_boot_mtx, "ap boot", NULL, MTX_SPIN); /* install the AP 1st level boot code */ pmap_kenter(va, boot_address); pmap_invalidate_page(kernel_pmap, va); bcopy(mptramp_start, (void *)va, bootMP_size); /* Locate the page tables, they'll be below the trampoline */ pt4 = (u_int64_t *)(uintptr_t)(mptramp_pagetables + KERNBASE); pt3 = pt4 + (PAGE_SIZE) / sizeof(u_int64_t); pt2 = pt3 + (PAGE_SIZE) / sizeof(u_int64_t); /* Create the initial 1GB replicated page tables */ for (i = 0; i < 512; i++) { /* Each slot of the level 4 pages points to the same level 3 page */ pt4[i] = (u_int64_t)(uintptr_t)(mptramp_pagetables + PAGE_SIZE); pt4[i] |= PG_V | PG_RW | PG_U; /* Each slot of the level 3 pages points to the same level 2 page */ pt3[i] = (u_int64_t)(uintptr_t)(mptramp_pagetables + (2 * PAGE_SIZE)); pt3[i] |= PG_V | PG_RW | PG_U; /* The level 2 page slots are mapped with 2MB pages for 1GB. */ pt2[i] = i * (2 * 1024 * 1024); pt2[i] |= PG_V | PG_RW | PG_PS | PG_U; } /* save the current value of the warm-start vector */ mpbioswarmvec = *((u_int32_t *) WARMBOOT_OFF); outb(CMOS_REG, BIOS_RESET); mpbiosreason = inb(CMOS_DATA); /* setup a vector to our boot code */ *((volatile u_short *) WARMBOOT_OFF) = WARMBOOT_TARGET; *((volatile u_short *) WARMBOOT_SEG) = (boot_address >> 4); outb(CMOS_REG, BIOS_RESET); outb(CMOS_DATA, BIOS_WARM); /* 'warm-start' */ /* start each AP */ for (cpu = 1; cpu < mp_ncpus; cpu++) { apic_id = cpu_apic_ids[cpu]; /* allocate and set up an idle stack data page */ bootstacks[cpu] = (void *)kmem_malloc(kernel_arena, KSTACK_PAGES * PAGE_SIZE, M_WAITOK | M_ZERO); doublefault_stack = (char *)kmem_malloc(kernel_arena, PAGE_SIZE, M_WAITOK | M_ZERO); nmi_stack = (char *)kmem_malloc(kernel_arena, PAGE_SIZE, M_WAITOK | M_ZERO); dpcpu = (void *)kmem_malloc(kernel_arena, DPCPU_SIZE, M_WAITOK | M_ZERO); bootSTK = (char *)bootstacks[cpu] + KSTACK_PAGES * PAGE_SIZE - 8; bootAP = cpu; /* attempt to start the Application Processor */ if (!start_ap(apic_id)) { /* restore the warmstart vector */ *(u_int32_t *) WARMBOOT_OFF = mpbioswarmvec; panic("AP #%d (PHY# %d) failed!", cpu, apic_id); } CPU_SET(cpu, &all_cpus); /* record AP in CPU map */ } /* restore the warmstart vector */ *(u_int32_t *) WARMBOOT_OFF = mpbioswarmvec; outb(CMOS_REG, BIOS_RESET); outb(CMOS_DATA, mpbiosreason); /* number of APs actually started */ return mp_naps; } /* * This function starts the AP (application processor) identified * by the APIC ID 'physicalCpu'. It does quite a "song and dance" * to accomplish this. This is necessary because of the nuances * of the different hardware we might encounter. It isn't pretty, * but it seems to work. */ static int start_ap(int apic_id) { int vector, ms; int cpus; /* calculate the vector */ vector = (boot_address >> 12) & 0xff; /* used as a watchpoint to signal AP startup */ cpus = mp_naps; ipi_startup(apic_id, vector); /* Wait up to 5 seconds for it to start. */ for (ms = 0; ms < 5000; ms++) { if (mp_naps > cpus) return 1; /* return SUCCESS */ DELAY(1000); } return 0; /* return FAILURE */ } -#ifdef COUNT_XINVLTLB_HITS -u_int xhits_gbl[MAXCPU]; -u_int xhits_pg[MAXCPU]; -u_int xhits_rng[MAXCPU]; -static SYSCTL_NODE(_debug, OID_AUTO, xhits, CTLFLAG_RW, 0, ""); -SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, global, CTLFLAG_RW, &xhits_gbl, - sizeof(xhits_gbl), "IU", ""); -SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, page, CTLFLAG_RW, &xhits_pg, - sizeof(xhits_pg), "IU", ""); -SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, range, CTLFLAG_RW, &xhits_rng, - sizeof(xhits_rng), "IU", ""); - -u_int ipi_global; -u_int ipi_page; -u_int ipi_range; -u_int ipi_range_size; -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_global, CTLFLAG_RW, &ipi_global, 0, ""); -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_page, CTLFLAG_RW, &ipi_page, 0, ""); -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_range, CTLFLAG_RW, &ipi_range, 0, ""); -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_range_size, CTLFLAG_RW, - &ipi_range_size, 0, ""); - -u_int ipi_masked_global; -u_int ipi_masked_page; -u_int ipi_masked_range; -u_int ipi_masked_range_size; -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_masked_global, CTLFLAG_RW, - &ipi_masked_global, 0, ""); -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_masked_page, CTLFLAG_RW, - &ipi_masked_page, 0, ""); -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_masked_range, CTLFLAG_RW, - &ipi_masked_range, 0, ""); -SYSCTL_UINT(_debug_xhits, OID_AUTO, ipi_masked_range_size, CTLFLAG_RW, - &ipi_masked_range_size, 0, ""); -#endif /* COUNT_XINVLTLB_HITS */ - /* - * Init and startup IPI. - */ -void -ipi_startup(int apic_id, int vector) -{ - - /* - * This attempts to follow the algorithm described in the - * Intel Multiprocessor Specification v1.4 in section B.4. - * For each IPI, we allow the local APIC ~20us to deliver the - * IPI. If that times out, we panic. - */ - - /* - * first we do an INIT IPI: this INIT IPI might be run, resetting - * and running the target CPU. OR this INIT IPI might be latched (P5 - * bug), CPU waiting for STARTUP IPI. OR this INIT IPI might be - * ignored. - */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_LEVEL | - APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_INIT, apic_id); - lapic_ipi_wait(100); - - /* Explicitly deassert the INIT IPI. */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_LEVEL | - APIC_LEVEL_DEASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_INIT, - apic_id); - - DELAY(10000); /* wait ~10mS */ - - /* - * next we do a STARTUP IPI: the previous INIT IPI might still be - * latched, (P5 bug) this 1st STARTUP would then terminate - * immediately, and the previously started INIT IPI would continue. OR - * the previous INIT IPI has already run. and this STARTUP IPI will - * run. OR the previous INIT IPI was ignored. and this STARTUP IPI - * will run. - */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_EDGE | - APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_STARTUP | - vector, apic_id); - if (!lapic_ipi_wait(100)) - panic("Failed to deliver first STARTUP IPI to APIC %d", - apic_id); - DELAY(200); /* wait ~200uS */ - - /* - * finally we do a 2nd STARTUP IPI: this 2nd STARTUP IPI should run IF - * the previous STARTUP IPI was cancelled by a latched INIT IPI. OR - * this STARTUP IPI will be ignored, as only ONE STARTUP IPI is - * recognized after hardware RESET or INIT IPI. - */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_EDGE | - APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_STARTUP | - vector, apic_id); - if (!lapic_ipi_wait(100)) - panic("Failed to deliver second STARTUP IPI to APIC %d", - apic_id); - - DELAY(200); /* wait ~200uS */ -} - -/* - * Send an IPI to specified CPU handling the bitmap logic. - */ -static void -ipi_send_cpu(int cpu, u_int ipi) -{ - u_int bitmap, old_pending, new_pending; - - KASSERT(cpu_apic_ids[cpu] != -1, ("IPI to non-existent CPU %d", cpu)); - - if (IPI_IS_BITMAPED(ipi)) { - bitmap = 1 << ipi; - ipi = IPI_BITMAP_VECTOR; - do { - old_pending = cpu_ipi_pending[cpu]; - new_pending = old_pending | bitmap; - } while (!atomic_cmpset_int(&cpu_ipi_pending[cpu], - old_pending, new_pending)); - if (old_pending) - return; - } - lapic_ipi_vectored(ipi, cpu_apic_ids[cpu]); -} - -/* * Flush the TLB on all other CPU's */ static void smp_tlb_shootdown(u_int vector, pmap_t pmap, vm_offset_t addr1, vm_offset_t addr2) { u_int ncpu; ncpu = mp_ncpus - 1; /* does not shootdown self */ if (ncpu < 1) return; /* no other cpus */ if (!(read_rflags() & PSL_I)) panic("%s: interrupts disabled", __func__); mtx_lock_spin(&smp_ipi_mtx); smp_tlb_invpcid.addr = addr1; if (pmap == NULL) { smp_tlb_invpcid.pcid = 0; } else { smp_tlb_invpcid.pcid = pmap->pm_pcid; pcid_cr3 = pmap->pm_cr3; } smp_tlb_addr2 = addr2; smp_tlb_pmap = pmap; atomic_store_rel_int(&smp_tlb_wait, 0); ipi_all_but_self(vector); while (smp_tlb_wait < ncpu) ia32_pause(); mtx_unlock_spin(&smp_ipi_mtx); } static void smp_targeted_tlb_shootdown(cpuset_t mask, u_int vector, pmap_t pmap, vm_offset_t addr1, vm_offset_t addr2) { int cpu, ncpu, othercpus; othercpus = mp_ncpus - 1; if (CPU_ISFULLSET(&mask)) { if (othercpus < 1) return; } else { CPU_CLR(PCPU_GET(cpuid), &mask); if (CPU_EMPTY(&mask)) return; } if (!(read_rflags() & PSL_I)) panic("%s: interrupts disabled", __func__); mtx_lock_spin(&smp_ipi_mtx); smp_tlb_invpcid.addr = addr1; if (pmap == NULL) { smp_tlb_invpcid.pcid = 0; } else { smp_tlb_invpcid.pcid = pmap->pm_pcid; pcid_cr3 = pmap->pm_cr3; } smp_tlb_addr2 = addr2; smp_tlb_pmap = pmap; atomic_store_rel_int(&smp_tlb_wait, 0); if (CPU_ISFULLSET(&mask)) { ncpu = othercpus; ipi_all_but_self(vector); } else { ncpu = 0; while ((cpu = CPU_FFS(&mask)) != 0) { cpu--; CPU_CLR(cpu, &mask); CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, vector); ipi_send_cpu(cpu, vector); ncpu++; } } while (smp_tlb_wait < ncpu) ia32_pause(); mtx_unlock_spin(&smp_ipi_mtx); } void -smp_cache_flush(void) -{ - - if (smp_started) - smp_tlb_shootdown(IPI_INVLCACHE, NULL, 0, 0); -} - -void -smp_invltlb(pmap_t pmap) -{ - - if (smp_started) { - smp_tlb_shootdown(IPI_INVLTLB, pmap, 0, 0); -#ifdef COUNT_XINVLTLB_HITS - ipi_global++; -#endif - } -} - -void smp_invlpg(pmap_t pmap, vm_offset_t addr) { if (smp_started) { smp_tlb_shootdown(IPI_INVLPG, pmap, addr, 0); #ifdef COUNT_XINVLTLB_HITS ipi_page++; #endif } } void smp_invlpg_range(pmap_t pmap, vm_offset_t addr1, vm_offset_t addr2) { if (smp_started) { smp_tlb_shootdown(IPI_INVLRNG, pmap, addr1, addr2); #ifdef COUNT_XINVLTLB_HITS ipi_range++; ipi_range_size += (addr2 - addr1) / PAGE_SIZE; #endif } } void smp_masked_invltlb(cpuset_t mask, pmap_t pmap) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLTLB, pmap, 0, 0); #ifdef COUNT_XINVLTLB_HITS ipi_masked_global++; #endif } } void smp_masked_invlpg(cpuset_t mask, pmap_t pmap, vm_offset_t addr) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLPG, pmap, addr, 0); #ifdef COUNT_XINVLTLB_HITS ipi_masked_page++; #endif } } void smp_masked_invlpg_range(cpuset_t mask, pmap_t pmap, vm_offset_t addr1, vm_offset_t addr2) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLRNG, pmap, addr1, addr2); #ifdef COUNT_XINVLTLB_HITS ipi_masked_range++; ipi_masked_range_size += (addr2 - addr1) / PAGE_SIZE; #endif } } void -ipi_bitmap_handler(struct trapframe frame) +smp_cache_flush(void) { - struct trapframe *oldframe; - struct thread *td; - int cpu = PCPU_GET(cpuid); - u_int ipi_bitmap; - critical_enter(); - td = curthread; - td->td_intr_nesting_level++; - oldframe = td->td_intr_frame; - td->td_intr_frame = &frame; - ipi_bitmap = atomic_readandclear_int(&cpu_ipi_pending[cpu]); - if (ipi_bitmap & (1 << IPI_PREEMPT)) { -#ifdef COUNT_IPIS - (*ipi_preempt_counts[cpu])++; -#endif - sched_preempt(td); - } - if (ipi_bitmap & (1 << IPI_AST)) { -#ifdef COUNT_IPIS - (*ipi_ast_counts[cpu])++; -#endif - /* Nothing to do for AST */ - } - if (ipi_bitmap & (1 << IPI_HARDCLOCK)) { -#ifdef COUNT_IPIS - (*ipi_hardclock_counts[cpu])++; -#endif - hardclockintr(); - } - td->td_intr_frame = oldframe; - td->td_intr_nesting_level--; - critical_exit(); + if (smp_started) + smp_tlb_shootdown(IPI_INVLCACHE, NULL, 0, 0); } -/* - * send an IPI to a set of cpus. - */ void -ipi_selected(cpuset_t cpus, u_int ipi) -{ - int cpu; +smp_invltlb(pmap_t pmap) +{ - /* - * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit - * of help in order to understand what is the source. - * Set the mask of receiving CPUs for this purpose. - */ - if (ipi == IPI_STOP_HARD) - CPU_OR_ATOMIC(&ipi_nmi_pending, &cpus); - - while ((cpu = CPU_FFS(&cpus)) != 0) { - cpu--; - CPU_CLR(cpu, &cpus); - CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); - ipi_send_cpu(cpu, ipi); - } -} - -/* - * send an IPI to a specific CPU. - */ -void -ipi_cpu(int cpu, u_int ipi) -{ - - /* - * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit - * of help in order to understand what is the source. - * Set the mask of receiving CPUs for this purpose. - */ - if (ipi == IPI_STOP_HARD) - CPU_SET_ATOMIC(cpu, &ipi_nmi_pending); - - CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); - ipi_send_cpu(cpu, ipi); -} - -/* - * send an IPI to all CPUs EXCEPT myself - */ -void -ipi_all_but_self(u_int ipi) -{ - cpuset_t other_cpus; - - other_cpus = all_cpus; - CPU_CLR(PCPU_GET(cpuid), &other_cpus); - - if (IPI_IS_BITMAPED(ipi)) { - ipi_selected(other_cpus, ipi); - return; - } - - /* - * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit - * of help in order to understand what is the source. - * Set the mask of receiving CPUs for this purpose. - */ - if (ipi == IPI_STOP_HARD) - CPU_OR_ATOMIC(&ipi_nmi_pending, &other_cpus); - - CTR2(KTR_SMP, "%s: ipi: %x", __func__, ipi); - lapic_ipi_vectored(ipi, APIC_IPI_DEST_OTHERS); -} - -int -ipi_nmi_handler() -{ - u_int cpuid; - - /* - * As long as there is not a simple way to know about a NMI's - * source, if the bitmask for the current CPU is present in - * the global pending bitword an IPI_STOP_HARD has been issued - * and should be handled. - */ - cpuid = PCPU_GET(cpuid); - if (!CPU_ISSET(cpuid, &ipi_nmi_pending)) - return (1); - - CPU_CLR_ATOMIC(cpuid, &ipi_nmi_pending); - cpustop_handler(); - return (0); -} - -/* - * Handle an IPI_STOP by saving our current context and spinning until we - * are resumed. - */ -void -cpustop_handler(void) -{ - u_int cpu; - - cpu = PCPU_GET(cpuid); - - savectx(&stoppcbs[cpu]); - - /* Indicate that we are stopped */ - CPU_SET_ATOMIC(cpu, &stopped_cpus); - - /* Wait for restart */ - while (!CPU_ISSET(cpu, &started_cpus)) - ia32_pause(); - - CPU_CLR_ATOMIC(cpu, &started_cpus); - CPU_CLR_ATOMIC(cpu, &stopped_cpus); - -#ifdef DDB - amd64_db_resume_dbreg(); + if (smp_started) { + smp_tlb_shootdown(IPI_INVLTLB, pmap, 0, 0); +#ifdef COUNT_XINVLTLB_HITS + ipi_global++; #endif - - if (cpu == 0 && cpustop_restartfunc != NULL) { - cpustop_restartfunc(); - cpustop_restartfunc = NULL; } } /* - * Handle an IPI_SUSPEND by saving our current context and spinning until we - * are resumed. - */ -void -cpususpend_handler(void) -{ - u_int cpu; - - mtx_assert(&smp_ipi_mtx, MA_NOTOWNED); - - cpu = PCPU_GET(cpuid); - if (savectx(&susppcbs[cpu]->sp_pcb)) { - fpususpend(susppcbs[cpu]->sp_fpususpend); - wbinvd(); - CPU_SET_ATOMIC(cpu, &suspended_cpus); - } else { - fpuresume(susppcbs[cpu]->sp_fpususpend); - pmap_init_pat(); - initializecpu(); - PCPU_SET(switchtime, 0); - PCPU_SET(switchticks, ticks); - - /* Indicate that we are resumed */ - CPU_CLR_ATOMIC(cpu, &suspended_cpus); - } - - /* Wait for resume */ - while (!CPU_ISSET(cpu, &started_cpus)) - ia32_pause(); - - if (cpu_ops.cpu_resume) - cpu_ops.cpu_resume(); - if (vmm_resume_p) - vmm_resume_p(); - - /* Resume MCA and local APIC */ - lapic_xapic_mode(); - mca_resume(); - lapic_setup(0); - - CPU_CLR_ATOMIC(cpu, &started_cpus); - /* Indicate that we are resumed */ - CPU_CLR_ATOMIC(cpu, &suspended_cpus); -} - -/* * Handlers for TLB related IPIs */ void invltlb_handler(void) { #ifdef COUNT_XINVLTLB_HITS xhits_gbl[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invltlb_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ invltlb(); atomic_add_int(&smp_tlb_wait, 1); } void invltlb_pcid_handler(void) { uint64_t cr3; u_int cpuid; #ifdef COUNT_XINVLTLB_HITS xhits_gbl[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invltlb_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ if (smp_tlb_invpcid.pcid != (uint64_t)-1 && smp_tlb_invpcid.pcid != 0) { if (invpcid_works) { invpcid(&smp_tlb_invpcid, INVPCID_CTX); } else { /* Otherwise reload %cr3 twice. */ cr3 = rcr3(); if (cr3 != pcid_cr3) { load_cr3(pcid_cr3); cr3 |= CR3_PCID_SAVE; } load_cr3(cr3); } } else { invltlb_globpcid(); } if (smp_tlb_pmap != NULL) { cpuid = PCPU_GET(cpuid); if (!CPU_ISSET(cpuid, &smp_tlb_pmap->pm_active)) CPU_CLR_ATOMIC(cpuid, &smp_tlb_pmap->pm_save); } atomic_add_int(&smp_tlb_wait, 1); } void invlpg_handler(void) { #ifdef COUNT_XINVLTLB_HITS xhits_pg[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invlpg_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ invlpg(smp_tlb_invpcid.addr); atomic_add_int(&smp_tlb_wait, 1); } void invlpg_pcid_handler(void) { uint64_t cr3; #ifdef COUNT_XINVLTLB_HITS xhits_pg[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invlpg_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ if (smp_tlb_invpcid.pcid == (uint64_t)-1) { invltlb_globpcid(); } else if (smp_tlb_invpcid.pcid == 0) { invlpg(smp_tlb_invpcid.addr); } else if (invpcid_works) { invpcid(&smp_tlb_invpcid, INVPCID_ADDR); } else { /* * PCID supported, but INVPCID is not. * Temporarily switch to the target address * space and do INVLPG. */ cr3 = rcr3(); if (cr3 != pcid_cr3) load_cr3(pcid_cr3 | CR3_PCID_SAVE); invlpg(smp_tlb_invpcid.addr); load_cr3(cr3 | CR3_PCID_SAVE); } atomic_add_int(&smp_tlb_wait, 1); } static inline void invlpg_range(vm_offset_t start, vm_offset_t end) { do { invlpg(start); start += PAGE_SIZE; } while (start < end); } void invlrng_handler(void) { struct invpcid_descr d; vm_offset_t addr; uint64_t cr3; u_int cpuid; #ifdef COUNT_XINVLTLB_HITS xhits_rng[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invlrng_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ addr = smp_tlb_invpcid.addr; if (pmap_pcid_enabled) { if (smp_tlb_invpcid.pcid == 0) { /* * kernel pmap - use invlpg to invalidate * global mapping. */ invlpg_range(addr, smp_tlb_addr2); } else if (smp_tlb_invpcid.pcid == (uint64_t)-1) { invltlb_globpcid(); if (smp_tlb_pmap != NULL) { cpuid = PCPU_GET(cpuid); if (!CPU_ISSET(cpuid, &smp_tlb_pmap->pm_active)) CPU_CLR_ATOMIC(cpuid, &smp_tlb_pmap->pm_save); } } else if (invpcid_works) { d = smp_tlb_invpcid; do { invpcid(&d, INVPCID_ADDR); d.addr += PAGE_SIZE; } while (d.addr <= smp_tlb_addr2); } else { cr3 = rcr3(); if (cr3 != pcid_cr3) load_cr3(pcid_cr3 | CR3_PCID_SAVE); invlpg_range(addr, smp_tlb_addr2); load_cr3(cr3 | CR3_PCID_SAVE); } } else { invlpg_range(addr, smp_tlb_addr2); } atomic_add_int(&smp_tlb_wait, 1); } - -void -invlcache_handler(void) -{ -#ifdef COUNT_IPIS - (*ipi_invlcache_counts[PCPU_GET(cpuid)])++; -#endif /* COUNT_IPIS */ - - wbinvd(); - atomic_add_int(&smp_tlb_wait, 1); -} - -/* - * This is called once the rest of the system is up and running and we're - * ready to let the AP's out of the pen. - */ -static void -release_aps(void *dummy __unused) -{ - - if (mp_ncpus == 1) - return; - atomic_store_rel_int(&aps_ready, 1); - while (smp_started == 0) - ia32_pause(); -} -SYSINIT(start_aps, SI_SUB_SMP, SI_ORDER_FIRST, release_aps, NULL); - -#ifdef COUNT_IPIS -/* - * Setup interrupt counters for IPI handlers. - */ -static void -mp_ipi_intrcnt(void *dummy) -{ - char buf[64]; - int i; - - CPU_FOREACH(i) { - snprintf(buf, sizeof(buf), "cpu%d:invltlb", i); - intrcnt_add(buf, &ipi_invltlb_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:invlrng", i); - intrcnt_add(buf, &ipi_invlrng_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:invlpg", i); - intrcnt_add(buf, &ipi_invlpg_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:invlcache", i); - intrcnt_add(buf, &ipi_invlcache_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:preempt", i); - intrcnt_add(buf, &ipi_preempt_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:ast", i); - intrcnt_add(buf, &ipi_ast_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:rendezvous", i); - intrcnt_add(buf, &ipi_rendezvous_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:hardclock", i); - intrcnt_add(buf, &ipi_hardclock_counts[i]); - } -} -SYSINIT(mp_ipi_intrcnt, SI_SUB_INTR, SI_ORDER_MIDDLE, mp_ipi_intrcnt, NULL); -#endif - Index: user/ngie/more-tests/sys/amd64/include/smp.h =================================================================== --- user/ngie/more-tests/sys/amd64/include/smp.h (revision 281974) +++ user/ngie/more-tests/sys/amd64/include/smp.h (revision 281975) @@ -1,95 +1,134 @@ /*- * ---------------------------------------------------------------------------- * "THE BEER-WARE LICENSE" (Revision 42): * wrote this file. As long as you retain this notice you * can do whatever you want with this stuff. If we meet some day, and you think * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp * ---------------------------------------------------------------------------- * * $FreeBSD$ * */ #ifndef _MACHINE_SMP_H_ #define _MACHINE_SMP_H_ #ifdef _KERNEL #ifdef SMP #ifndef LOCORE #include #include #include #include #include /* global symbols in mpboot.S */ extern char mptramp_start[]; extern char mptramp_end[]; extern u_int32_t mptramp_pagetables; /* global data in mp_machdep.c */ extern int mp_naps; extern int boot_cpu_id; extern struct pcb stoppcbs[]; extern int cpu_apic_ids[]; +extern int bootAP; +extern void *dpcpu; +extern char *bootSTK; +extern int bootAP; +extern void *bootstacks[]; +extern volatile u_int cpu_ipi_pending[]; +extern volatile int aps_ready; +extern struct mtx ap_boot_mtx; +extern int cpu_logical; +extern int cpu_cores; +extern int pmap_pcid_enabled; +extern u_int xhits_gbl[]; +extern u_int xhits_pg[]; +extern u_int xhits_rng[]; +extern u_int ipi_global; +extern u_int ipi_page; +extern u_int ipi_range; +extern u_int ipi_range_size; +extern u_int ipi_masked_global; +extern u_int ipi_masked_page; +extern u_int ipi_masked_range; +extern u_int ipi_masked_range_size; + +extern volatile int smp_tlb_wait; + +struct cpu_info { + int cpu_present:1; + int cpu_bsp:1; + int cpu_disabled:1; + int cpu_hyperthread:1; +}; +extern struct cpu_info cpu_info[]; + #ifdef COUNT_IPIS extern u_long *ipi_invltlb_counts[MAXCPU]; extern u_long *ipi_invlrng_counts[MAXCPU]; extern u_long *ipi_invlpg_counts[MAXCPU]; extern u_long *ipi_invlcache_counts[MAXCPU]; extern u_long *ipi_rendezvous_counts[MAXCPU]; #endif /* IPI handlers */ inthand_t IDTVEC(invltlb_pcid), /* TLB shootdowns - global, pcid enabled */ IDTVEC(invltlb), /* TLB shootdowns - global */ IDTVEC(invlpg_pcid), /* TLB shootdowns - 1 page, pcid enabled */ IDTVEC(invlpg), /* TLB shootdowns - 1 page */ IDTVEC(invlrng), /* TLB shootdowns - page range */ IDTVEC(invlcache), /* Write back and invalidate cache */ IDTVEC(ipi_intr_bitmap_handler), /* Bitmap based IPIs */ IDTVEC(cpustop), /* CPU stops & waits to be restarted */ IDTVEC(cpususpend), /* CPU suspends & waits to be resumed */ IDTVEC(justreturn), /* interrupt CPU with minimum overhead */ IDTVEC(rendezvous); /* handle CPU rendezvous */ struct pmap; /* functions in mp_machdep.c */ +void assign_cpu_ids(void); void cpu_add(u_int apic_id, char boot_cpu); void cpustop_handler(void); void cpususpend_handler(void); +void init_secondary_tail(void); void invltlb_handler(void); void invltlb_pcid_handler(void); void invlpg_handler(void); void invlpg_pcid_handler(void); void invlrng_handler(void); void invlcache_handler(void); void init_secondary(void); void ipi_startup(int apic_id, int vector); void ipi_all_but_self(u_int ipi); void ipi_bitmap_handler(struct trapframe frame); void ipi_cpu(int cpu, u_int ipi); int ipi_nmi_handler(void); void ipi_selected(cpuset_t cpus, u_int ipi); u_int mp_bootaddress(u_int); +void set_interrupt_apic_ids(void); void smp_cache_flush(void); void smp_invlpg(struct pmap *pmap, vm_offset_t addr); void smp_masked_invlpg(cpuset_t mask, struct pmap *pmap, vm_offset_t addr); void smp_invlpg_range(struct pmap *pmap, vm_offset_t startva, vm_offset_t endva); void smp_masked_invlpg_range(cpuset_t mask, struct pmap *pmap, vm_offset_t startva, vm_offset_t endva); void smp_invltlb(struct pmap *pmap); void smp_masked_invltlb(cpuset_t mask, struct pmap *pmap); int native_start_all_aps(void); +void mem_range_AP_init(void); +void topo_probe(void); +void ipi_send_cpu(int cpu, u_int ipi); #endif /* !LOCORE */ #endif /* SMP */ #endif /* _KERNEL */ #endif /* _MACHINE_SMP_H_ */ Index: user/ngie/more-tests/sys/conf/files.amd64 =================================================================== --- user/ngie/more-tests/sys/conf/files.amd64 (revision 281974) +++ user/ngie/more-tests/sys/conf/files.amd64 (revision 281975) @@ -1,585 +1,586 @@ # This file tells config what files go into building a kernel, # files marked standard are always included. # # $FreeBSD$ # # The long compile-with and dependency lines are required because of # limitations in config: backslash-newline doesn't work in strings, and # dependency lines other than the first are silently ignored. # # linux32_genassym.o optional compat_linux32 \ dependency "$S/amd64/linux32/linux32_genassym.c" \ compile-with "${CC} ${CFLAGS:N-fno-common} -c ${.IMPSRC}" \ no-obj no-implicit-rule \ clean "linux32_genassym.o" # linux32_assym.h optional compat_linux32 \ dependency "$S/kern/genassym.sh linux32_genassym.o" \ compile-with "sh $S/kern/genassym.sh linux32_genassym.o > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "linux32_assym.h" # ia32_genassym.o standard \ dependency "$S/compat/ia32/ia32_genassym.c" \ compile-with "${CC} ${CFLAGS:N-fno-common} -c ${.IMPSRC}" \ no-obj no-implicit-rule \ clean "ia32_genassym.o" # ia32_assym.h standard \ dependency "$S/kern/genassym.sh ia32_genassym.o" \ compile-with "env NM='${NM}' sh $S/kern/genassym.sh ia32_genassym.o > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "ia32_assym.h" # font.h optional sc_dflt_font \ compile-with "uudecode < /usr/share/syscons/fonts/${SC_DFLT_FONT}-8x16.fnt && file2c 'static u_char dflt_font_16[16*256] = {' '};' < ${SC_DFLT_FONT}-8x16 > font.h && uudecode < /usr/share/syscons/fonts/${SC_DFLT_FONT}-8x14.fnt && file2c 'static u_char dflt_font_14[14*256] = {' '};' < ${SC_DFLT_FONT}-8x14 >> font.h && uudecode < /usr/share/syscons/fonts/${SC_DFLT_FONT}-8x8.fnt && file2c 'static u_char dflt_font_8[8*256] = {' '};' < ${SC_DFLT_FONT}-8x8 >> font.h" \ no-obj no-implicit-rule before-depend \ clean "font.h ${SC_DFLT_FONT}-8x14 ${SC_DFLT_FONT}-8x16 ${SC_DFLT_FONT}-8x8" # atkbdmap.h optional atkbd_dflt_keymap \ compile-with "/usr/sbin/kbdcontrol -L ${ATKBD_DFLT_KEYMAP} | sed -e 's/^static keymap_t.* = /static keymap_t key_map = /' -e 's/^static accentmap_t.* = /static accentmap_t accent_map = /' > atkbdmap.h" \ no-obj no-implicit-rule before-depend \ clean "atkbdmap.h" # ukbdmap.h optional ukbd_dflt_keymap \ compile-with "/usr/sbin/kbdcontrol -L ${UKBD_DFLT_KEYMAP} | sed -e 's/^static keymap_t.* = /static keymap_t key_map = /' -e 's/^static accentmap_t.* = /static accentmap_t accent_map = /' > ukbdmap.h" \ no-obj no-implicit-rule before-depend \ clean "ukbdmap.h" # hpt27xx_lib.o optional hpt27xx \ dependency "$S/dev/hpt27xx/amd64-elf.hpt27xx_lib.o.uu" \ compile-with "uudecode < $S/dev/hpt27xx/amd64-elf.hpt27xx_lib.o.uu" \ no-implicit-rule # hptmvraid.o optional hptmv \ dependency "$S/dev/hptmv/amd64-elf.raid.o.uu" \ compile-with "uudecode < $S/dev/hptmv/amd64-elf.raid.o.uu" \ no-implicit-rule # hptnr_lib.o optional hptnr \ dependency "$S/dev/hptnr/amd64-elf.hptnr_lib.o.uu" \ compile-with "uudecode < $S/dev/hptnr/amd64-elf.hptnr_lib.o.uu" \ no-implicit-rule # hptrr_lib.o optional hptrr \ dependency "$S/dev/hptrr/amd64-elf.hptrr_lib.o.uu" \ compile-with "uudecode < $S/dev/hptrr/amd64-elf.hptrr_lib.o.uu" \ no-implicit-rule # amd64/acpica/acpi_machdep.c optional acpi acpi_wakecode.o optional acpi \ dependency "$S/amd64/acpica/acpi_wakecode.S assym.s" \ compile-with "${NORMAL_S}" \ no-obj no-implicit-rule before-depend \ clean "acpi_wakecode.o" acpi_wakecode.bin optional acpi \ dependency "acpi_wakecode.o" \ compile-with "${OBJCOPY} -S -O binary acpi_wakecode.o ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "acpi_wakecode.bin" acpi_wakecode.h optional acpi \ dependency "acpi_wakecode.bin" \ compile-with "file2c -sx 'static char wakecode[] = {' '};' < acpi_wakecode.bin > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "acpi_wakecode.h" acpi_wakedata.h optional acpi \ dependency "acpi_wakecode.o" \ compile-with '${NM} -n --defined-only acpi_wakecode.o | while read offset dummy what; do echo "#define $${what} 0x$${offset}"; done > ${.TARGET}' \ no-obj no-implicit-rule before-depend \ clean "acpi_wakedata.h" # amd64/amd64/amd64_mem.c optional mem #amd64/amd64/apic_vector.S standard amd64/amd64/atomic.c standard amd64/amd64/autoconf.c standard amd64/amd64/bios.c standard amd64/amd64/bpf_jit_machdep.c optional bpf_jitter amd64/amd64/cpu_switch.S standard amd64/amd64/db_disasm.c optional ddb amd64/amd64/db_interface.c optional ddb amd64/amd64/db_trace.c optional ddb amd64/amd64/elf_machdep.c standard amd64/amd64/exception.S standard amd64/amd64/fpu.c standard amd64/amd64/gdb_machdep.c optional gdb amd64/amd64/in_cksum.c optional inet | inet6 amd64/amd64/initcpu.c standard amd64/amd64/io.c optional io amd64/amd64/locore.S standard no-obj amd64/amd64/xen-locore.S optional xenhvm amd64/amd64/machdep.c standard amd64/amd64/mem.c optional mem amd64/amd64/minidump_machdep.c standard amd64/amd64/mp_machdep.c optional smp amd64/amd64/mp_watchdog.c optional mp_watchdog smp amd64/amd64/mpboot.S optional smp amd64/amd64/pmap.c standard amd64/amd64/prof_machdep.c optional profiling-routine amd64/amd64/ptrace_machdep.c standard amd64/amd64/sigtramp.S standard amd64/amd64/stack_machdep.c optional ddb | stack amd64/amd64/support.S standard amd64/amd64/sys_machdep.c standard amd64/amd64/trap.c standard amd64/amd64/uio_machdep.c standard amd64/amd64/uma_machdep.c standard amd64/amd64/vm_machdep.c standard amd64/pci/pci_cfgreg.c optional pci cddl/contrib/opensolaris/common/atomic/amd64/opensolaris_atomic.S optional zfs compile-with "${ZFS_S}" crypto/aesni/aeskeys_amd64.S optional aesni crypto/aesni/aesni.c optional aesni aesni_ghash.o optional aesni \ dependency "$S/crypto/aesni/aesni_ghash.c" \ compile-with "${CC} -c ${CFLAGS:C/^-O2$/-O3/:N-nostdinc} ${WERROR} ${NO_WCAST_QUAL} ${PROF} -mmmx -msse -msse4 -maes -mpclmul ${.IMPSRC}" \ no-implicit-rule \ clean "aesni_ghash.o" aesni_wrap.o optional aesni \ dependency "$S/crypto/aesni/aesni_wrap.c" \ compile-with "${CC} -c ${CFLAGS:C/^-O2$/-O3/:N-nostdinc} ${WERROR} ${NO_WCAST_QUAL} ${PROF} -mmmx -msse -msse4 -maes ${.IMPSRC}" \ no-implicit-rule \ clean "aesni_wrap.o" crypto/blowfish/bf_enc.c optional crypto | ipsec crypto/des/des_enc.c optional crypto | ipsec | netsmb crypto/via/padlock.c optional padlock crypto/via/padlock_cipher.c optional padlock crypto/via/padlock_hash.c optional padlock dev/acpica/acpi_if.m standard dev/acpi_support/acpi_wmi_if.m standard dev/agp/agp_amd64.c optional agp dev/agp/agp_i810.c optional agp dev/agp/agp_via.c optional agp dev/amdsbwd/amdsbwd.c optional amdsbwd dev/amdtemp/amdtemp.c optional amdtemp dev/arcmsr/arcmsr.c optional arcmsr pci dev/asmc/asmc.c optional asmc isa dev/atkbdc/atkbd.c optional atkbd atkbdc dev/atkbdc/atkbd_atkbdc.c optional atkbd atkbdc dev/atkbdc/atkbdc.c optional atkbdc dev/atkbdc/atkbdc_isa.c optional atkbdc isa dev/atkbdc/atkbdc_subr.c optional atkbdc dev/atkbdc/psm.c optional psm atkbdc dev/bxe/bxe.c optional bxe pci dev/bxe/bxe_stats.c optional bxe pci dev/bxe/bxe_debug.c optional bxe pci dev/bxe/ecore_sp.c optional bxe pci dev/bxe/bxe_elink.c optional bxe pci dev/bxe/57710_init_values.c optional bxe pci dev/bxe/57711_init_values.c optional bxe pci dev/bxe/57712_init_values.c optional bxe pci dev/coretemp/coretemp.c optional coretemp dev/cpuctl/cpuctl.c optional cpuctl dev/dpms/dpms.c optional dpms # There are no systems with isa slots, so all ed isa entries should go.. dev/ed/if_ed_3c503.c optional ed isa ed_3c503 dev/ed/if_ed_isa.c optional ed isa dev/ed/if_ed_wd80x3.c optional ed isa dev/ed/if_ed_hpp.c optional ed isa ed_hpp dev/ed/if_ed_sic.c optional ed isa ed_sic dev/fb/fb.c optional fb | vga dev/fb/s3_pci.c optional s3pci dev/fb/vesa.c optional vga vesa dev/fb/vga.c optional vga dev/ichwd/ichwd.c optional ichwd dev/if_ndis/if_ndis.c optional ndis dev/if_ndis/if_ndis_pccard.c optional ndis pccard dev/if_ndis/if_ndis_pci.c optional ndis cardbus | ndis pci dev/if_ndis/if_ndis_usb.c optional ndis usb dev/io/iodev.c optional io dev/ipmi/ipmi.c optional ipmi dev/ipmi/ipmi_acpi.c optional ipmi acpi dev/ipmi/ipmi_isa.c optional ipmi isa dev/ipmi/ipmi_kcs.c optional ipmi dev/ipmi/ipmi_smic.c optional ipmi dev/ipmi/ipmi_smbus.c optional ipmi smbus dev/ipmi/ipmi_smbios.c optional ipmi dev/ipmi/ipmi_ssif.c optional ipmi smbus dev/ipmi/ipmi_pci.c optional ipmi pci dev/ipmi/ipmi_linux.c optional ipmi compat_linux32 dev/ixl/if_ixl.c optional ixl pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/if_ixlv.c optional ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/ixlvc.c optional ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/ixl_txrx.c optional ixl pci | ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/i40e_osdep.c optional ixl pci | ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/i40e_lan_hmc.c optional ixl pci | ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/i40e_hmc.c optional ixl pci | ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/i40e_common.c optional ixl pci | ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/i40e_nvm.c optional ixl pci | ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/ixl/i40e_adminq.c optional ixl pci | ixlv pci \ compile-with "${NORMAL_C} -I$S/dev/ixl" dev/fdc/fdc.c optional fdc dev/fdc/fdc_acpi.c optional fdc dev/fdc/fdc_isa.c optional fdc isa dev/fdc/fdc_pccard.c optional fdc pccard dev/fdt/fdt_x86.c optional fdt dev/hpt27xx/hpt27xx_os_bsd.c optional hpt27xx dev/hpt27xx/hpt27xx_osm_bsd.c optional hpt27xx dev/hpt27xx/hpt27xx_config.c optional hpt27xx dev/hptmv/entry.c optional hptmv dev/hptmv/mv.c optional hptmv dev/hptmv/gui_lib.c optional hptmv dev/hptmv/hptproc.c optional hptmv dev/hptmv/ioctl.c optional hptmv dev/hptnr/hptnr_os_bsd.c optional hptnr dev/hptnr/hptnr_osm_bsd.c optional hptnr dev/hptnr/hptnr_config.c optional hptnr dev/hptrr/hptrr_os_bsd.c optional hptrr dev/hptrr/hptrr_osm_bsd.c optional hptrr dev/hptrr/hptrr_config.c optional hptrr dev/hwpmc/hwpmc_amd.c optional hwpmc dev/hwpmc/hwpmc_intel.c optional hwpmc dev/hwpmc/hwpmc_core.c optional hwpmc dev/hwpmc/hwpmc_uncore.c optional hwpmc dev/hwpmc/hwpmc_piv.c optional hwpmc dev/hwpmc/hwpmc_tsc.c optional hwpmc dev/hwpmc/hwpmc_x86.c optional hwpmc dev/hyperv/netvsc/hv_net_vsc.c optional hyperv dev/hyperv/netvsc/hv_netvsc_drv_freebsd.c optional hyperv dev/hyperv/netvsc/hv_rndis_filter.c optional hyperv dev/hyperv/stordisengage/hv_ata_pci_disengage.c optional hyperv dev/hyperv/storvsc/hv_storvsc_drv_freebsd.c optional hyperv dev/hyperv/utilities/hv_kvp.c optional hyperv dev/hyperv/utilities/hv_util.c optional hyperv dev/hyperv/vmbus/hv_channel.c optional hyperv dev/hyperv/vmbus/hv_channel_mgmt.c optional hyperv dev/hyperv/vmbus/hv_connection.c optional hyperv dev/hyperv/vmbus/hv_hv.c optional hyperv dev/hyperv/vmbus/hv_ring_buffer.c optional hyperv dev/hyperv/vmbus/hv_vmbus_drv_freebsd.c optional hyperv dev/kbd/kbd.c optional atkbd | sc | ukbd | vt dev/nfe/if_nfe.c optional nfe pci dev/ntb/if_ntb/if_ntb.c optional if_ntb dev/ntb/ntb_hw/ntb_hw.c optional if_ntb ntb_hw dev/nvd/nvd.c optional nvd nvme dev/nvme/nvme.c optional nvme dev/nvme/nvme_ctrlr.c optional nvme dev/nvme/nvme_ctrlr_cmd.c optional nvme dev/nvme/nvme_ns.c optional nvme dev/nvme/nvme_ns_cmd.c optional nvme dev/nvme/nvme_qpair.c optional nvme dev/nvme/nvme_sysctl.c optional nvme dev/nvme/nvme_test.c optional nvme dev/nvme/nvme_util.c optional nvme dev/nvram/nvram.c optional nvram isa dev/random/ivy.c optional rdrand_rng dev/random/nehemiah.c optional padlock_rng dev/qlxge/qls_dbg.c optional qlxge pci dev/qlxge/qls_dump.c optional qlxge pci dev/qlxge/qls_hw.c optional qlxge pci dev/qlxge/qls_ioctl.c optional qlxge pci dev/qlxge/qls_isr.c optional qlxge pci dev/qlxge/qls_os.c optional qlxge pci dev/qlxgb/qla_dbg.c optional qlxgb pci dev/qlxgb/qla_hw.c optional qlxgb pci dev/qlxgb/qla_ioctl.c optional qlxgb pci dev/qlxgb/qla_isr.c optional qlxgb pci dev/qlxgb/qla_misc.c optional qlxgb pci dev/qlxgb/qla_os.c optional qlxgb pci dev/qlxgbe/ql_dbg.c optional qlxgbe pci dev/qlxgbe/ql_hw.c optional qlxgbe pci dev/qlxgbe/ql_ioctl.c optional qlxgbe pci dev/qlxgbe/ql_isr.c optional qlxgbe pci dev/qlxgbe/ql_misc.c optional qlxgbe pci dev/qlxgbe/ql_os.c optional qlxgbe pci dev/qlxgbe/ql_reset.c optional qlxgbe pci dev/sfxge/common/efx_bootcfg.c optional sfxge inet pci dev/sfxge/common/efx_ev.c optional sfxge inet pci dev/sfxge/common/efx_filter.c optional sfxge inet pci dev/sfxge/common/efx_intr.c optional sfxge inet pci dev/sfxge/common/efx_mac.c optional sfxge inet pci dev/sfxge/common/efx_mcdi.c optional sfxge inet pci dev/sfxge/common/efx_mon.c optional sfxge inet pci dev/sfxge/common/efx_nic.c optional sfxge inet pci dev/sfxge/common/efx_nvram.c optional sfxge inet pci dev/sfxge/common/efx_phy.c optional sfxge inet pci dev/sfxge/common/efx_port.c optional sfxge inet pci dev/sfxge/common/efx_rx.c optional sfxge inet pci dev/sfxge/common/efx_sram.c optional sfxge inet pci dev/sfxge/common/efx_tx.c optional sfxge inet pci dev/sfxge/common/efx_vpd.c optional sfxge inet pci dev/sfxge/common/efx_wol.c optional sfxge inet pci dev/sfxge/common/siena_mac.c optional sfxge inet pci dev/sfxge/common/siena_mon.c optional sfxge inet pci dev/sfxge/common/siena_nic.c optional sfxge inet pci dev/sfxge/common/siena_nvram.c optional sfxge inet pci dev/sfxge/common/siena_phy.c optional sfxge inet pci dev/sfxge/common/siena_sram.c optional sfxge inet pci dev/sfxge/common/siena_vpd.c optional sfxge inet pci dev/sfxge/sfxge.c optional sfxge inet pci dev/sfxge/sfxge_dma.c optional sfxge inet pci dev/sfxge/sfxge_ev.c optional sfxge inet pci dev/sfxge/sfxge_intr.c optional sfxge inet pci dev/sfxge/sfxge_mcdi.c optional sfxge inet pci dev/sfxge/sfxge_port.c optional sfxge inet pci dev/sfxge/sfxge_rx.c optional sfxge inet pci dev/sfxge/sfxge_tx.c optional sfxge inet pci dev/sio/sio.c optional sio dev/sio/sio_isa.c optional sio isa dev/sio/sio_pccard.c optional sio pccard dev/sio/sio_pci.c optional sio pci dev/sio/sio_puc.c optional sio puc dev/speaker/spkr.c optional speaker dev/syscons/apm/apm_saver.c optional apm_saver apm dev/syscons/scterm-teken.c optional sc dev/syscons/scvesactl.c optional sc vga vesa dev/syscons/scvgarndr.c optional sc vga dev/syscons/scvtb.c optional sc dev/tpm/tpm.c optional tpm dev/tpm/tpm_acpi.c optional tpm acpi dev/tpm/tpm_isa.c optional tpm isa dev/uart/uart_cpu_x86.c optional uart dev/viawd/viawd.c optional viawd dev/vmware/vmxnet3/if_vmx.c optional vmx dev/wbwd/wbwd.c optional wbwd dev/wpi/if_wpi.c optional wpi dev/xen/pci/xen_acpi_pci.c optional xenhvm dev/xen/pci/xen_pci.c optional xenhvm dev/isci/isci.c optional isci dev/isci/isci_controller.c optional isci dev/isci/isci_domain.c optional isci dev/isci/isci_interrupt.c optional isci dev/isci/isci_io_request.c optional isci dev/isci/isci_logger.c optional isci dev/isci/isci_oem_parameters.c optional isci dev/isci/isci_remote_device.c optional isci dev/isci/isci_sysctl.c optional isci dev/isci/isci_task_request.c optional isci dev/isci/isci_timer.c optional isci dev/isci/scil/sati.c optional isci dev/isci/scil/sati_abort_task_set.c optional isci dev/isci/scil/sati_atapi.c optional isci dev/isci/scil/sati_device.c optional isci dev/isci/scil/sati_inquiry.c optional isci dev/isci/scil/sati_log_sense.c optional isci dev/isci/scil/sati_lun_reset.c optional isci dev/isci/scil/sati_mode_pages.c optional isci dev/isci/scil/sati_mode_select.c optional isci dev/isci/scil/sati_mode_sense.c optional isci dev/isci/scil/sati_mode_sense_10.c optional isci dev/isci/scil/sati_mode_sense_6.c optional isci dev/isci/scil/sati_move.c optional isci dev/isci/scil/sati_passthrough.c optional isci dev/isci/scil/sati_read.c optional isci dev/isci/scil/sati_read_buffer.c optional isci dev/isci/scil/sati_read_capacity.c optional isci dev/isci/scil/sati_reassign_blocks.c optional isci dev/isci/scil/sati_report_luns.c optional isci dev/isci/scil/sati_request_sense.c optional isci dev/isci/scil/sati_start_stop_unit.c optional isci dev/isci/scil/sati_synchronize_cache.c optional isci dev/isci/scil/sati_test_unit_ready.c optional isci dev/isci/scil/sati_unmap.c optional isci dev/isci/scil/sati_util.c optional isci dev/isci/scil/sati_verify.c optional isci dev/isci/scil/sati_write.c optional isci dev/isci/scil/sati_write_and_verify.c optional isci dev/isci/scil/sati_write_buffer.c optional isci dev/isci/scil/sati_write_long.c optional isci dev/isci/scil/sci_abstract_list.c optional isci dev/isci/scil/sci_base_controller.c optional isci dev/isci/scil/sci_base_domain.c optional isci dev/isci/scil/sci_base_iterator.c optional isci dev/isci/scil/sci_base_library.c optional isci dev/isci/scil/sci_base_logger.c optional isci dev/isci/scil/sci_base_memory_descriptor_list.c optional isci dev/isci/scil/sci_base_memory_descriptor_list_decorator.c optional isci dev/isci/scil/sci_base_object.c optional isci dev/isci/scil/sci_base_observer.c optional isci dev/isci/scil/sci_base_phy.c optional isci dev/isci/scil/sci_base_port.c optional isci dev/isci/scil/sci_base_remote_device.c optional isci dev/isci/scil/sci_base_request.c optional isci dev/isci/scil/sci_base_state_machine.c optional isci dev/isci/scil/sci_base_state_machine_logger.c optional isci dev/isci/scil/sci_base_state_machine_observer.c optional isci dev/isci/scil/sci_base_subject.c optional isci dev/isci/scil/sci_util.c optional isci dev/isci/scil/scic_sds_controller.c optional isci dev/isci/scil/scic_sds_library.c optional isci dev/isci/scil/scic_sds_pci.c optional isci dev/isci/scil/scic_sds_phy.c optional isci dev/isci/scil/scic_sds_port.c optional isci dev/isci/scil/scic_sds_port_configuration_agent.c optional isci dev/isci/scil/scic_sds_remote_device.c optional isci dev/isci/scil/scic_sds_remote_node_context.c optional isci dev/isci/scil/scic_sds_remote_node_table.c optional isci dev/isci/scil/scic_sds_request.c optional isci dev/isci/scil/scic_sds_sgpio.c optional isci dev/isci/scil/scic_sds_smp_remote_device.c optional isci dev/isci/scil/scic_sds_smp_request.c optional isci dev/isci/scil/scic_sds_ssp_request.c optional isci dev/isci/scil/scic_sds_stp_packet_request.c optional isci dev/isci/scil/scic_sds_stp_remote_device.c optional isci dev/isci/scil/scic_sds_stp_request.c optional isci dev/isci/scil/scic_sds_unsolicited_frame_control.c optional isci dev/isci/scil/scif_sas_controller.c optional isci dev/isci/scil/scif_sas_controller_state_handlers.c optional isci dev/isci/scil/scif_sas_controller_states.c optional isci dev/isci/scil/scif_sas_domain.c optional isci dev/isci/scil/scif_sas_domain_state_handlers.c optional isci dev/isci/scil/scif_sas_domain_states.c optional isci dev/isci/scil/scif_sas_high_priority_request_queue.c optional isci dev/isci/scil/scif_sas_internal_io_request.c optional isci dev/isci/scil/scif_sas_io_request.c optional isci dev/isci/scil/scif_sas_io_request_state_handlers.c optional isci dev/isci/scil/scif_sas_io_request_states.c optional isci dev/isci/scil/scif_sas_library.c optional isci dev/isci/scil/scif_sas_remote_device.c optional isci dev/isci/scil/scif_sas_remote_device_ready_substate_handlers.c optional isci dev/isci/scil/scif_sas_remote_device_ready_substates.c optional isci dev/isci/scil/scif_sas_remote_device_starting_substate_handlers.c optional isci dev/isci/scil/scif_sas_remote_device_starting_substates.c optional isci dev/isci/scil/scif_sas_remote_device_state_handlers.c optional isci dev/isci/scil/scif_sas_remote_device_states.c optional isci dev/isci/scil/scif_sas_request.c optional isci dev/isci/scil/scif_sas_smp_activity_clear_affiliation.c optional isci dev/isci/scil/scif_sas_smp_io_request.c optional isci dev/isci/scil/scif_sas_smp_phy.c optional isci dev/isci/scil/scif_sas_smp_remote_device.c optional isci dev/isci/scil/scif_sas_stp_io_request.c optional isci dev/isci/scil/scif_sas_stp_remote_device.c optional isci dev/isci/scil/scif_sas_stp_task_request.c optional isci dev/isci/scil/scif_sas_task_request.c optional isci dev/isci/scil/scif_sas_task_request_state_handlers.c optional isci dev/isci/scil/scif_sas_task_request_states.c optional isci dev/isci/scil/scif_sas_timer.c optional isci isa/syscons_isa.c optional sc isa/vga_isa.c optional vga kern/kern_clocksource.c standard kern/link_elf_obj.c standard # # IA32 binary support # #amd64/ia32/ia32_exception.S optional compat_freebsd32 amd64/ia32/ia32_reg.c optional compat_freebsd32 amd64/ia32/ia32_signal.c optional compat_freebsd32 amd64/ia32/ia32_sigtramp.S optional compat_freebsd32 amd64/ia32/ia32_syscall.c optional compat_freebsd32 amd64/ia32/ia32_misc.c optional compat_freebsd32 compat/ia32/ia32_sysvec.c optional compat_freebsd32 compat/linprocfs/linprocfs.c optional linprocfs compat/linsysfs/linsysfs.c optional linsysfs # # Linux/i386 binary support # amd64/linux32/linux32_dummy.c optional compat_linux32 amd64/linux32/linux32_locore.s optional compat_linux32 \ dependency "linux32_assym.h" amd64/linux32/linux32_machdep.c optional compat_linux32 amd64/linux32/linux32_support.s optional compat_linux32 \ dependency "linux32_assym.h" amd64/linux32/linux32_sysent.c optional compat_linux32 amd64/linux32/linux32_sysvec.c optional compat_linux32 compat/linux/linux_emul.c optional compat_linux32 compat/linux/linux_file.c optional compat_linux32 compat/linux/linux_fork.c optional compat_linux32 compat/linux/linux_futex.c optional compat_linux32 compat/linux/linux_getcwd.c optional compat_linux32 compat/linux/linux_ioctl.c optional compat_linux32 compat/linux/linux_ipc.c optional compat_linux32 compat/linux/linux_mib.c optional compat_linux32 compat/linux/linux_misc.c optional compat_linux32 compat/linux/linux_signal.c optional compat_linux32 compat/linux/linux_socket.c optional compat_linux32 compat/linux/linux_stats.c optional compat_linux32 compat/linux/linux_sysctl.c optional compat_linux32 compat/linux/linux_time.c optional compat_linux32 compat/linux/linux_timer.c optional compat_linux32 compat/linux/linux_uid16.c optional compat_linux32 compat/linux/linux_util.c optional compat_linux32 dev/amr/amr_linux.c optional compat_linux32 amr dev/mfi/mfi_linux.c optional compat_linux32 mfi # # Windows NDIS driver support # compat/ndis/kern_ndis.c optional ndisapi pci compat/ndis/kern_windrv.c optional ndisapi pci compat/ndis/subr_hal.c optional ndisapi pci compat/ndis/subr_ndis.c optional ndisapi pci compat/ndis/subr_ntoskrnl.c optional ndisapi pci compat/ndis/subr_pe.c optional ndisapi pci compat/ndis/subr_usbd.c optional ndisapi pci compat/ndis/winx64_wrap.S optional ndisapi pci # libkern/memmove.c standard libkern/memset.c standard # # x86 real mode BIOS emulator, required by atkbdc/dpms/pci/vesa # compat/x86bios/x86bios.c optional x86bios | atkbd | dpms | pci | vesa contrib/x86emu/x86emu.c optional x86bios | atkbd | dpms | pci | vesa # # bvm console # dev/bvm/bvm_console.c optional bvmconsole dev/bvm/bvm_dbg.c optional bvmdebug # # x86 shared code between IA32, AMD64 and PC98 architectures # x86/acpica/OsdEnvironment.c optional acpi x86/acpica/acpi_apm.c optional acpi x86/acpica/acpi_wakeup.c optional acpi x86/acpica/madt.c optional acpi x86/acpica/srat.c optional acpi x86/bios/smbios.c optional smbios x86/bios/vpd.c optional vpd x86/cpufreq/powernow.c optional cpufreq x86/cpufreq/est.c optional cpufreq x86/cpufreq/hwpstate.c optional cpufreq x86/cpufreq/p4tcc.c optional cpufreq x86/iommu/busdma_dmar.c optional acpi acpi_dmar pci x86/iommu/intel_ctx.c optional acpi acpi_dmar pci x86/iommu/intel_drv.c optional acpi acpi_dmar pci x86/iommu/intel_fault.c optional acpi acpi_dmar pci x86/iommu/intel_gas.c optional acpi acpi_dmar pci x86/iommu/intel_idpgtbl.c optional acpi acpi_dmar pci x86/iommu/intel_intrmap.c optional acpi acpi_dmar pci x86/iommu/intel_qi.c optional acpi acpi_dmar pci x86/iommu/intel_quirks.c optional acpi acpi_dmar pci x86/iommu/intel_utils.c optional acpi acpi_dmar pci x86/isa/atpic.c optional atpic isa x86/isa/atrtc.c standard x86/isa/clock.c standard x86/isa/elcr.c optional atpic isa | mptable x86/isa/isa.c standard x86/isa/isa_dma.c standard x86/isa/nmi.c standard x86/isa/orm.c optional isa x86/pci/pci_bus.c optional pci x86/pci/qpi.c optional pci x86/x86/busdma_bounce.c standard x86/x86/busdma_machdep.c standard x86/x86/cpu_machdep.c standard x86/x86/dump_machdep.c standard x86/x86/fdt_machdep.c optional fdt x86/x86/identcpu.c standard x86/x86/intr_machdep.c standard x86/x86/io_apic.c standard x86/x86/legacy.c standard x86/x86/local_apic.c standard x86/x86/mca.c standard x86/x86/mptable.c optional mptable x86/x86/mptable_pci.c optional mptable pci +x86/x86/mp_x86.c optional smp x86/x86/msi.c optional pci x86/x86/nexus.c standard x86/x86/pvclock.c standard x86/x86/tsc.c standard x86/x86/delay.c standard x86/xen/hvm.c optional xenhvm x86/xen/xen_intr.c optional xen | xenhvm x86/xen/pv.c optional xenhvm x86/xen/pvcpu_enum.c optional xenhvm x86/xen/xen_apic.c optional xenhvm x86/xen/xenpv.c optional xenhvm x86/xen/xen_nexus.c optional xenhvm x86/xen/xen_msi.c optional xenhvm x86/xen/xen_pci_bus.c optional xenhvm Index: user/ngie/more-tests/sys/conf/files.i386 =================================================================== --- user/ngie/more-tests/sys/conf/files.i386 (revision 281974) +++ user/ngie/more-tests/sys/conf/files.i386 (revision 281975) @@ -1,600 +1,601 @@ # This file tells config what files go into building a kernel, # files marked standard are always included. # # $FreeBSD$ # # The long compile-with and dependency lines are required because of # limitations in config: backslash-newline doesn't work in strings, and # dependency lines other than the first are silently ignored. # linux_genassym.o optional compat_linux \ dependency "$S/i386/linux/linux_genassym.c" \ compile-with "${CC} ${CFLAGS:N-fno-common} -c ${.IMPSRC}" \ no-obj no-implicit-rule \ clean "linux_genassym.o" # linux_assym.h optional compat_linux \ dependency "$S/kern/genassym.sh linux_genassym.o" \ compile-with "sh $S/kern/genassym.sh linux_genassym.o > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "linux_assym.h" # svr4_genassym.o optional compat_svr4 \ dependency "$S/i386/svr4/svr4_genassym.c" \ compile-with "${CC} ${CFLAGS:N-fno-common} -c ${.IMPSRC}" \ no-obj no-implicit-rule \ clean "svr4_genassym.o" # svr4_assym.h optional compat_svr4 \ dependency "$S/kern/genassym.sh svr4_genassym.o" \ compile-with "sh $S/kern/genassym.sh svr4_genassym.o > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "svr4_assym.h" # font.h optional sc_dflt_font \ compile-with "uudecode < /usr/share/syscons/fonts/${SC_DFLT_FONT}-8x16.fnt && file2c 'static u_char dflt_font_16[16*256] = {' '};' < ${SC_DFLT_FONT}-8x16 > font.h && uudecode < /usr/share/syscons/fonts/${SC_DFLT_FONT}-8x14.fnt && file2c 'static u_char dflt_font_14[14*256] = {' '};' < ${SC_DFLT_FONT}-8x14 >> font.h && uudecode < /usr/share/syscons/fonts/${SC_DFLT_FONT}-8x8.fnt && file2c 'static u_char dflt_font_8[8*256] = {' '};' < ${SC_DFLT_FONT}-8x8 >> font.h" \ no-obj no-implicit-rule before-depend \ clean "font.h ${SC_DFLT_FONT}-8x14 ${SC_DFLT_FONT}-8x16 ${SC_DFLT_FONT}-8x8" # atkbdmap.h optional atkbd_dflt_keymap \ compile-with "/usr/sbin/kbdcontrol -L ${ATKBD_DFLT_KEYMAP} | sed -e 's/^static keymap_t.* = /static keymap_t key_map = /' -e 's/^static accentmap_t.* = /static accentmap_t accent_map = /' > atkbdmap.h" \ no-obj no-implicit-rule before-depend \ clean "atkbdmap.h" # ukbdmap.h optional ukbd_dflt_keymap \ compile-with "/usr/sbin/kbdcontrol -L ${UKBD_DFLT_KEYMAP} | sed -e 's/^static keymap_t.* = /static keymap_t key_map = /' -e 's/^static accentmap_t.* = /static accentmap_t accent_map = /' > ukbdmap.h" \ no-obj no-implicit-rule before-depend \ clean "ukbdmap.h" # hpt27xx_lib.o optional hpt27xx \ dependency "$S/dev/hpt27xx/i386-elf.hpt27xx_lib.o.uu" \ compile-with "uudecode < $S/dev/hpt27xx/i386-elf.hpt27xx_lib.o.uu" \ no-implicit-rule # hptmvraid.o optional hptmv \ dependency "$S/dev/hptmv/i386-elf.raid.o.uu" \ compile-with "uudecode < $S/dev/hptmv/i386-elf.raid.o.uu" \ no-implicit-rule # hptnr_lib.o optional hptnr \ dependency "$S/dev/hptnr/i386-elf.hptnr_lib.o.uu" \ compile-with "uudecode < $S/dev/hptnr/i386-elf.hptnr_lib.o.uu" \ no-implicit-rule # hptrr_lib.o optional hptrr \ dependency "$S/dev/hptrr/i386-elf.hptrr_lib.o.uu" \ compile-with "uudecode < $S/dev/hptrr/i386-elf.hptrr_lib.o.uu" \ no-implicit-rule # cddl/contrib/opensolaris/common/atomic/i386/opensolaris_atomic.S optional zfs compile-with "${ZFS_S}" compat/linprocfs/linprocfs.c optional linprocfs compat/linsysfs/linsysfs.c optional linsysfs compat/linux/linux_emul.c optional compat_linux compat/linux/linux_file.c optional compat_linux compat/linux/linux_fork.c optional compat_linux compat/linux/linux_futex.c optional compat_linux compat/linux/linux_getcwd.c optional compat_linux compat/linux/linux_ioctl.c optional compat_linux compat/linux/linux_ipc.c optional compat_linux compat/linux/linux_mib.c optional compat_linux compat/linux/linux_misc.c optional compat_linux compat/linux/linux_signal.c optional compat_linux compat/linux/linux_socket.c optional compat_linux compat/linux/linux_stats.c optional compat_linux compat/linux/linux_sysctl.c optional compat_linux compat/linux/linux_time.c optional compat_linux compat/linux/linux_timer.c optional compat_linux compat/linux/linux_uid16.c optional compat_linux compat/linux/linux_util.c optional compat_linux compat/ndis/kern_ndis.c optional ndisapi pci compat/ndis/kern_windrv.c optional ndisapi pci compat/ndis/subr_hal.c optional ndisapi pci compat/ndis/subr_ndis.c optional ndisapi pci compat/ndis/subr_ntoskrnl.c optional ndisapi pci compat/ndis/subr_pe.c optional ndisapi pci compat/ndis/subr_usbd.c optional ndisapi pci compat/ndis/winx32_wrap.S optional ndisapi pci compat/svr4/imgact_svr4.c optional compat_svr4 compat/svr4/svr4_fcntl.c optional compat_svr4 compat/svr4/svr4_filio.c optional compat_svr4 compat/svr4/svr4_ioctl.c optional compat_svr4 compat/svr4/svr4_ipc.c optional compat_svr4 compat/svr4/svr4_misc.c optional compat_svr4 compat/svr4/svr4_resource.c optional compat_svr4 compat/svr4/svr4_signal.c optional compat_svr4 compat/svr4/svr4_socket.c optional compat_svr4 compat/svr4/svr4_sockio.c optional compat_svr4 compat/svr4/svr4_stat.c optional compat_svr4 compat/svr4/svr4_stream.c optional compat_svr4 compat/svr4/svr4_syscallnames.c optional compat_svr4 compat/svr4/svr4_sysent.c optional compat_svr4 compat/svr4/svr4_sysvec.c optional compat_svr4 compat/svr4/svr4_termios.c optional compat_svr4 bf_enc.o optional crypto | ipsec \ dependency "$S/crypto/blowfish/arch/i386/bf_enc.S $S/crypto/blowfish/arch/i386/bf_enc_586.S $S/crypto/blowfish/arch/i386/bf_enc_686.S" \ compile-with "${CC} -c -I$S/crypto/blowfish/arch/i386 ${ASM_CFLAGS} ${WERROR} ${.IMPSRC}" \ no-implicit-rule crypto/aesni/aeskeys_i386.S optional aesni crypto/aesni/aesni.c optional aesni aesni_ghash.o optional aesni \ dependency "$S/crypto/aesni/aesni_ghash.c" \ compile-with "${CC} -c ${CFLAGS:C/^-O2$/-O3/:N-nostdinc} ${WERROR} ${NO_WCAST_QUAL} ${PROF} -mmmx -msse -msse4 -maes -mpclmul ${.IMPSRC}" \ no-implicit-rule \ clean "aesni_ghash.o" aesni_wrap.o optional aesni \ dependency "$S/crypto/aesni/aesni_wrap.c" \ compile-with "${CC} -c ${CFLAGS:C/^-O2$/-O3/:N-nostdinc} ${WERROR} ${NO_WCAST_QUAL} ${PROF} -mmmx -msse -msse4 -maes ${.IMPSRC}" \ no-implicit-rule \ clean "aesni_wrap.o" crypto/des/arch/i386/des_enc.S optional crypto | ipsec | netsmb crypto/via/padlock.c optional padlock crypto/via/padlock_cipher.c optional padlock crypto/via/padlock_hash.c optional padlock dev/advansys/adv_isa.c optional adv isa dev/agp/agp_ali.c optional agp dev/agp/agp_amd.c optional agp dev/agp/agp_amd64.c optional agp dev/agp/agp_ati.c optional agp dev/agp/agp_i810.c optional agp dev/agp/agp_intel.c optional agp dev/agp/agp_nvidia.c optional agp dev/agp/agp_sis.c optional agp dev/agp/agp_via.c optional agp dev/aic/aic_isa.c optional aic isa dev/amdsbwd/amdsbwd.c optional amdsbwd dev/amdtemp/amdtemp.c optional amdtemp dev/arcmsr/arcmsr.c optional arcmsr pci dev/asmc/asmc.c optional asmc isa dev/atkbdc/atkbd.c optional atkbd atkbdc dev/atkbdc/atkbd_atkbdc.c optional atkbd atkbdc dev/atkbdc/atkbdc.c optional atkbdc dev/atkbdc/atkbdc_isa.c optional atkbdc isa dev/atkbdc/atkbdc_subr.c optional atkbdc dev/atkbdc/psm.c optional psm atkbdc dev/bxe/bxe.c optional bxe pci dev/bxe/bxe_stats.c optional bxe pci dev/bxe/bxe_debug.c optional bxe pci dev/bxe/ecore_sp.c optional bxe pci dev/bxe/bxe_elink.c optional bxe pci dev/bxe/57710_init_values.c optional bxe pci dev/bxe/57711_init_values.c optional bxe pci dev/bxe/57712_init_values.c optional bxe pci dev/ce/ceddk.c optional ce dev/ce/if_ce.c optional ce dev/ce/tau32-ddk.c optional ce \ compile-with "${NORMAL_C} ${NO_WCONSTANT_CONVERSION}" dev/cm/if_cm_isa.c optional cm isa dev/coretemp/coretemp.c optional coretemp dev/cp/cpddk.c optional cp dev/cp/if_cp.c optional cp dev/cpuctl/cpuctl.c optional cpuctl dev/ctau/ctau.c optional ctau dev/ctau/ctddk.c optional ctau dev/ctau/if_ct.c optional ctau dev/cx/csigma.c optional cx dev/cx/cxddk.c optional cx dev/cx/if_cx.c optional cx dev/dpms/dpms.c optional dpms dev/ed/if_ed_3c503.c optional ed isa ed_3c503 dev/ed/if_ed_isa.c optional ed isa dev/ed/if_ed_wd80x3.c optional ed isa dev/ed/if_ed_hpp.c optional ed isa ed_hpp dev/ed/if_ed_sic.c optional ed isa ed_sic dev/fb/fb.c optional fb | vga dev/fb/s3_pci.c optional s3pci dev/fb/vesa.c optional vga vesa dev/fb/vga.c optional vga dev/fdc/fdc.c optional fdc dev/fdc/fdc_acpi.c optional fdc dev/fdc/fdc_isa.c optional fdc isa dev/fdc/fdc_pccard.c optional fdc pccard dev/fdt/fdt_x86.c optional fdt dev/fe/if_fe_isa.c optional fe isa dev/glxiic/glxiic.c optional glxiic dev/glxsb/glxsb.c optional glxsb dev/glxsb/glxsb_hash.c optional glxsb dev/hpt27xx/hpt27xx_os_bsd.c optional hpt27xx dev/hpt27xx/hpt27xx_osm_bsd.c optional hpt27xx dev/hpt27xx/hpt27xx_config.c optional hpt27xx dev/hptmv/entry.c optional hptmv dev/hptmv/mv.c optional hptmv dev/hptmv/gui_lib.c optional hptmv dev/hptmv/hptproc.c optional hptmv dev/hptmv/ioctl.c optional hptmv dev/hptnr/hptnr_os_bsd.c optional hptnr dev/hptnr/hptnr_osm_bsd.c optional hptnr dev/hptnr/hptnr_config.c optional hptnr dev/hptrr/hptrr_os_bsd.c optional hptrr dev/hptrr/hptrr_osm_bsd.c optional hptrr dev/hptrr/hptrr_config.c optional hptrr dev/hwpmc/hwpmc_amd.c optional hwpmc dev/hwpmc/hwpmc_intel.c optional hwpmc dev/hwpmc/hwpmc_core.c optional hwpmc dev/hwpmc/hwpmc_uncore.c optional hwpmc dev/hwpmc/hwpmc_pentium.c optional hwpmc dev/hwpmc/hwpmc_piv.c optional hwpmc dev/hwpmc/hwpmc_ppro.c optional hwpmc dev/hwpmc/hwpmc_tsc.c optional hwpmc dev/hwpmc/hwpmc_x86.c optional hwpmc dev/hyperv/netvsc/hv_net_vsc.c optional hyperv dev/hyperv/netvsc/hv_netvsc_drv_freebsd.c optional hyperv dev/hyperv/netvsc/hv_rndis_filter.c optional hyperv dev/hyperv/stordisengage/hv_ata_pci_disengage.c optional hyperv dev/hyperv/storvsc/hv_storvsc_drv_freebsd.c optional hyperv dev/hyperv/utilities/hv_kvp.c optional hyperv dev/hyperv/utilities/hv_util.c optional hyperv dev/hyperv/vmbus/hv_channel.c optional hyperv dev/hyperv/vmbus/hv_channel_mgmt.c optional hyperv dev/hyperv/vmbus/hv_connection.c optional hyperv dev/hyperv/vmbus/hv_hv.c optional hyperv dev/hyperv/vmbus/hv_ring_buffer.c optional hyperv dev/hyperv/vmbus/hv_vmbus_drv_freebsd.c optional hyperv dev/ichwd/ichwd.c optional ichwd dev/if_ndis/if_ndis.c optional ndis dev/if_ndis/if_ndis_pccard.c optional ndis pccard dev/if_ndis/if_ndis_pci.c optional ndis cardbus | ndis pci dev/if_ndis/if_ndis_usb.c optional ndis usb dev/io/iodev.c optional io dev/ipmi/ipmi.c optional ipmi dev/ipmi/ipmi_acpi.c optional ipmi acpi dev/ipmi/ipmi_isa.c optional ipmi isa dev/ipmi/ipmi_kcs.c optional ipmi dev/ipmi/ipmi_smic.c optional ipmi dev/ipmi/ipmi_smbus.c optional ipmi smbus dev/ipmi/ipmi_smbios.c optional ipmi dev/ipmi/ipmi_ssif.c optional ipmi smbus dev/ipmi/ipmi_pci.c optional ipmi pci dev/ipmi/ipmi_linux.c optional ipmi compat_linux dev/kbd/kbd.c optional atkbd | sc | ukbd | vt dev/le/if_le_isa.c optional le isa dev/mse/mse.c optional mse dev/mse/mse_isa.c optional mse isa dev/nfe/if_nfe.c optional nfe pci dev/nvd/nvd.c optional nvd nvme dev/nvme/nvme.c optional nvme dev/nvme/nvme_ctrlr.c optional nvme dev/nvme/nvme_ctrlr_cmd.c optional nvme dev/nvme/nvme_ns.c optional nvme dev/nvme/nvme_ns_cmd.c optional nvme dev/nvme/nvme_qpair.c optional nvme dev/nvme/nvme_sysctl.c optional nvme dev/nvme/nvme_test.c optional nvme dev/nvme/nvme_util.c optional nvme dev/nvram/nvram.c optional nvram isa dev/pcf/pcf_isa.c optional pcf dev/random/ivy.c optional rdrand_rng dev/random/nehemiah.c optional padlock_rng dev/sbni/if_sbni.c optional sbni dev/sbni/if_sbni_isa.c optional sbni isa dev/sbni/if_sbni_pci.c optional sbni pci dev/sio/sio.c optional sio dev/sio/sio_isa.c optional sio isa dev/sio/sio_pccard.c optional sio pccard dev/sio/sio_pci.c optional sio pci dev/sio/sio_puc.c optional sio puc dev/speaker/spkr.c optional speaker dev/syscons/apm/apm_saver.c optional apm_saver apm dev/syscons/scterm-teken.c optional sc dev/syscons/scvesactl.c optional sc vga vesa dev/syscons/scvgarndr.c optional sc vga dev/syscons/scvtb.c optional sc dev/tpm/tpm.c optional tpm dev/tpm/tpm_acpi.c optional tpm acpi dev/tpm/tpm_isa.c optional tpm isa dev/uart/uart_cpu_x86.c optional uart dev/viawd/viawd.c optional viawd dev/vmware/vmxnet3/if_vmx.c optional vmx dev/acpica/acpi_if.m standard dev/acpi_support/acpi_wmi_if.m standard dev/wbwd/wbwd.c optional wbwd dev/wpi/if_wpi.c optional wpi dev/isci/isci.c optional isci dev/isci/isci_controller.c optional isci dev/isci/isci_domain.c optional isci dev/isci/isci_interrupt.c optional isci dev/isci/isci_io_request.c optional isci dev/isci/isci_logger.c optional isci dev/isci/isci_oem_parameters.c optional isci dev/isci/isci_remote_device.c optional isci dev/isci/isci_sysctl.c optional isci dev/isci/isci_task_request.c optional isci dev/isci/isci_timer.c optional isci dev/isci/scil/sati.c optional isci dev/isci/scil/sati_abort_task_set.c optional isci dev/isci/scil/sati_atapi.c optional isci dev/isci/scil/sati_device.c optional isci dev/isci/scil/sati_inquiry.c optional isci dev/isci/scil/sati_log_sense.c optional isci dev/isci/scil/sati_lun_reset.c optional isci dev/isci/scil/sati_mode_pages.c optional isci dev/isci/scil/sati_mode_select.c optional isci dev/isci/scil/sati_mode_sense.c optional isci dev/isci/scil/sati_mode_sense_10.c optional isci dev/isci/scil/sati_mode_sense_6.c optional isci dev/isci/scil/sati_move.c optional isci dev/isci/scil/sati_passthrough.c optional isci dev/isci/scil/sati_read.c optional isci dev/isci/scil/sati_read_buffer.c optional isci dev/isci/scil/sati_read_capacity.c optional isci dev/isci/scil/sati_reassign_blocks.c optional isci dev/isci/scil/sati_report_luns.c optional isci dev/isci/scil/sati_request_sense.c optional isci dev/isci/scil/sati_start_stop_unit.c optional isci dev/isci/scil/sati_synchronize_cache.c optional isci dev/isci/scil/sati_test_unit_ready.c optional isci dev/isci/scil/sati_unmap.c optional isci dev/isci/scil/sati_util.c optional isci dev/isci/scil/sati_verify.c optional isci dev/isci/scil/sati_write.c optional isci dev/isci/scil/sati_write_and_verify.c optional isci dev/isci/scil/sati_write_buffer.c optional isci dev/isci/scil/sati_write_long.c optional isci dev/isci/scil/sci_abstract_list.c optional isci dev/isci/scil/sci_base_controller.c optional isci dev/isci/scil/sci_base_domain.c optional isci dev/isci/scil/sci_base_iterator.c optional isci dev/isci/scil/sci_base_library.c optional isci dev/isci/scil/sci_base_logger.c optional isci dev/isci/scil/sci_base_memory_descriptor_list.c optional isci dev/isci/scil/sci_base_memory_descriptor_list_decorator.c optional isci dev/isci/scil/sci_base_object.c optional isci dev/isci/scil/sci_base_observer.c optional isci dev/isci/scil/sci_base_phy.c optional isci dev/isci/scil/sci_base_port.c optional isci dev/isci/scil/sci_base_remote_device.c optional isci dev/isci/scil/sci_base_request.c optional isci dev/isci/scil/sci_base_state_machine.c optional isci dev/isci/scil/sci_base_state_machine_logger.c optional isci dev/isci/scil/sci_base_state_machine_observer.c optional isci dev/isci/scil/sci_base_subject.c optional isci dev/isci/scil/sci_util.c optional isci dev/isci/scil/scic_sds_controller.c optional isci dev/isci/scil/scic_sds_library.c optional isci dev/isci/scil/scic_sds_pci.c optional isci dev/isci/scil/scic_sds_phy.c optional isci dev/isci/scil/scic_sds_port.c optional isci dev/isci/scil/scic_sds_port_configuration_agent.c optional isci dev/isci/scil/scic_sds_remote_device.c optional isci dev/isci/scil/scic_sds_remote_node_context.c optional isci dev/isci/scil/scic_sds_remote_node_table.c optional isci dev/isci/scil/scic_sds_request.c optional isci dev/isci/scil/scic_sds_sgpio.c optional isci dev/isci/scil/scic_sds_smp_remote_device.c optional isci dev/isci/scil/scic_sds_smp_request.c optional isci dev/isci/scil/scic_sds_ssp_request.c optional isci dev/isci/scil/scic_sds_stp_packet_request.c optional isci dev/isci/scil/scic_sds_stp_remote_device.c optional isci dev/isci/scil/scic_sds_stp_request.c optional isci dev/isci/scil/scic_sds_unsolicited_frame_control.c optional isci dev/isci/scil/scif_sas_controller.c optional isci dev/isci/scil/scif_sas_controller_state_handlers.c optional isci dev/isci/scil/scif_sas_controller_states.c optional isci dev/isci/scil/scif_sas_domain.c optional isci dev/isci/scil/scif_sas_domain_state_handlers.c optional isci dev/isci/scil/scif_sas_domain_states.c optional isci dev/isci/scil/scif_sas_high_priority_request_queue.c optional isci dev/isci/scil/scif_sas_internal_io_request.c optional isci dev/isci/scil/scif_sas_io_request.c optional isci dev/isci/scil/scif_sas_io_request_state_handlers.c optional isci dev/isci/scil/scif_sas_io_request_states.c optional isci dev/isci/scil/scif_sas_library.c optional isci dev/isci/scil/scif_sas_remote_device.c optional isci dev/isci/scil/scif_sas_remote_device_ready_substate_handlers.c optional isci dev/isci/scil/scif_sas_remote_device_ready_substates.c optional isci dev/isci/scil/scif_sas_remote_device_starting_substate_handlers.c optional isci dev/isci/scil/scif_sas_remote_device_starting_substates.c optional isci dev/isci/scil/scif_sas_remote_device_state_handlers.c optional isci dev/isci/scil/scif_sas_remote_device_states.c optional isci dev/isci/scil/scif_sas_request.c optional isci dev/isci/scil/scif_sas_smp_activity_clear_affiliation.c optional isci dev/isci/scil/scif_sas_smp_io_request.c optional isci dev/isci/scil/scif_sas_smp_phy.c optional isci dev/isci/scil/scif_sas_smp_remote_device.c optional isci dev/isci/scil/scif_sas_stp_io_request.c optional isci dev/isci/scil/scif_sas_stp_remote_device.c optional isci dev/isci/scil/scif_sas_stp_task_request.c optional isci dev/isci/scil/scif_sas_task_request.c optional isci dev/isci/scil/scif_sas_task_request_state_handlers.c optional isci dev/isci/scil/scif_sas_task_request_states.c optional isci dev/isci/scil/scif_sas_timer.c optional isci i386/acpica/acpi_machdep.c optional acpi acpi_wakecode.o optional acpi \ dependency "$S/i386/acpica/acpi_wakecode.S assym.s" \ compile-with "${NORMAL_S}" \ no-obj no-implicit-rule before-depend \ clean "acpi_wakecode.o" acpi_wakecode.bin optional acpi \ dependency "acpi_wakecode.o" \ compile-with "${OBJCOPY} -S -O binary acpi_wakecode.o ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "acpi_wakecode.bin" acpi_wakecode.h optional acpi \ dependency "acpi_wakecode.bin" \ compile-with "file2c -sx 'static char wakecode[] = {' '};' < acpi_wakecode.bin > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "acpi_wakecode.h" acpi_wakedata.h optional acpi \ dependency "acpi_wakecode.o" \ compile-with '${NM} -n --defined-only acpi_wakecode.o | while read offset dummy what; do echo "#define $${what} 0x$${offset}"; done > ${.TARGET}' \ no-obj no-implicit-rule before-depend \ clean "acpi_wakedata.h" # i386/bios/apm.c optional apm i386/bios/mca_machdep.c optional mca i386/bios/smapi.c optional smapi i386/bios/smapi_bios.S optional smapi #i386/i386/apic_vector.s optional apic i386/i386/atomic.c standard \ compile-with "${CC} -c ${CFLAGS} ${DEFINED_PROF:S/^$/-fomit-frame-pointer/} ${.IMPSRC}" i386/i386/autoconf.c standard i386/i386/bios.c optional native i386/i386/bioscall.s optional native i386/i386/bpf_jit_machdep.c optional bpf_jitter i386/i386/db_disasm.c optional ddb i386/i386/db_interface.c optional ddb i386/i386/db_trace.c optional ddb i386/i386/elan-mmcr.c optional cpu_elan | cpu_soekris i386/i386/elf_machdep.c standard i386/i386/exception.s optional native i386/xen/exception.s optional xen i386/i386/gdb_machdep.c optional gdb i386/i386/geode.c optional cpu_geode i386/i386/i686_mem.c optional mem i386/i386/in_cksum.c optional inet | inet6 i386/i386/initcpu.c standard i386/i386/io.c optional io i386/i386/k6_mem.c optional mem i386/i386/locore.s optional native no-obj i386/xen/locore.s optional xen no-obj i386/i386/longrun.c optional cpu_enable_longrun i386/i386/machdep.c standard i386/xen/xen_machdep.c optional xen i386/i386/mem.c optional mem i386/i386/minidump_machdep.c standard i386/i386/mp_clock.c optional smp i386/i386/mp_machdep.c optional native smp i386/xen/mp_machdep.c optional xen smp i386/i386/mp_watchdog.c optional mp_watchdog smp i386/i386/mpboot.s optional smp native i386/xen/mptable.c optional apic xen i386/i386/perfmon.c optional perfmon i386/i386/pmap.c optional native i386/xen/pmap.c optional xen i386/i386/ptrace_machdep.c standard i386/i386/stack_machdep.c optional ddb | stack i386/i386/support.s standard i386/i386/swtch.s standard i386/i386/sys_machdep.c standard i386/i386/trap.c standard i386/i386/uio_machdep.c standard i386/i386/vm86.c standard i386/i386/vm_machdep.c standard i386/ibcs2/ibcs2_errno.c optional ibcs2 i386/ibcs2/ibcs2_fcntl.c optional ibcs2 i386/ibcs2/ibcs2_ioctl.c optional ibcs2 i386/ibcs2/ibcs2_ipc.c optional ibcs2 i386/ibcs2/ibcs2_isc.c optional ibcs2 i386/ibcs2/ibcs2_isc_sysent.c optional ibcs2 i386/ibcs2/ibcs2_misc.c optional ibcs2 i386/ibcs2/ibcs2_msg.c optional ibcs2 i386/ibcs2/ibcs2_other.c optional ibcs2 i386/ibcs2/ibcs2_signal.c optional ibcs2 i386/ibcs2/ibcs2_socksys.c optional ibcs2 i386/ibcs2/ibcs2_stat.c optional ibcs2 i386/ibcs2/ibcs2_sysent.c optional ibcs2 i386/ibcs2/ibcs2_sysi86.c optional ibcs2 i386/ibcs2/ibcs2_sysvec.c optional ibcs2 i386/ibcs2/ibcs2_util.c optional ibcs2 i386/ibcs2/ibcs2_xenix.c optional ibcs2 i386/ibcs2/ibcs2_xenix_sysent.c optional ibcs2 i386/ibcs2/imgact_coff.c optional ibcs2 i386/xen/clock.c optional xen i386/isa/elink.c optional ep | ie i386/isa/npx.c optional npx i386/isa/pmtimer.c optional pmtimer i386/isa/prof_machdep.c optional profiling-routine i386/isa/spic.c optional spic i386/linux/imgact_linux.c optional compat_linux i386/linux/linux_dummy.c optional compat_linux i386/linux/linux_locore.s optional compat_linux \ dependency "linux_assym.h" i386/linux/linux_machdep.c optional compat_linux i386/linux/linux_ptrace.c optional compat_linux i386/linux/linux_support.s optional compat_linux \ dependency "linux_assym.h" i386/linux/linux_sysent.c optional compat_linux i386/linux/linux_sysvec.c optional compat_linux i386/pci/pci_cfgreg.c optional pci i386/pci/pci_pir.c optional pci i386/svr4/svr4_locore.s optional compat_svr4 \ dependency "svr4_assym.h" \ warning "COMPAT_SVR4 is broken and should be avoided" i386/svr4/svr4_machdep.c optional compat_svr4 # isa/syscons_isa.c optional sc isa/vga_isa.c optional vga kern/kern_clocksource.c standard kern/imgact_aout.c optional compat_aout kern/imgact_gzip.c optional gzip kern/subr_sfbuf.c standard libkern/divdi3.c standard libkern/flsll.c standard libkern/memmove.c standard libkern/memset.c standard libkern/moddi3.c standard libkern/qdivrem.c standard libkern/ucmpdi2.c standard libkern/udivdi3.c standard libkern/umoddi3.c standard i386/xbox/xbox.c optional xbox i386/xbox/xboxfb.c optional xboxfb dev/fb/boot_font.c optional xboxfb i386/xbox/pic16l.s optional xbox # # x86 real mode BIOS support, required by atkbdc/dpms/pci/vesa # compat/x86bios/x86bios.c optional x86bios | atkbd | dpms | pci | vesa # # bvm console # dev/bvm/bvm_console.c optional bvmconsole dev/bvm/bvm_dbg.c optional bvmdebug # # x86 shared code between IA32, AMD64 and PC98 architectures # x86/acpica/OsdEnvironment.c optional acpi x86/acpica/acpi_apm.c optional acpi x86/acpica/acpi_wakeup.c optional acpi x86/acpica/madt.c optional acpi apic x86/acpica/srat.c optional acpi x86/bios/smbios.c optional smbios x86/bios/vpd.c optional vpd x86/cpufreq/est.c optional cpufreq x86/cpufreq/hwpstate.c optional cpufreq x86/cpufreq/p4tcc.c optional cpufreq x86/cpufreq/powernow.c optional cpufreq x86/cpufreq/smist.c optional cpufreq x86/iommu/busdma_dmar.c optional acpi acpi_dmar pci x86/iommu/intel_ctx.c optional acpi acpi_dmar pci x86/iommu/intel_drv.c optional acpi acpi_dmar pci x86/iommu/intel_fault.c optional acpi acpi_dmar pci x86/iommu/intel_gas.c optional acpi acpi_dmar pci x86/iommu/intel_idpgtbl.c optional acpi acpi_dmar pci x86/iommu/intel_intrmap.c optional acpi acpi_dmar pci x86/iommu/intel_qi.c optional acpi acpi_dmar pci x86/iommu/intel_quirks.c optional acpi acpi_dmar pci x86/iommu/intel_utils.c optional acpi acpi_dmar pci x86/isa/atpic.c optional atpic x86/isa/atrtc.c optional native x86/isa/clock.c optional native x86/isa/elcr.c optional atpic | apic native x86/isa/isa.c optional isa x86/isa/isa_dma.c optional isa x86/isa/nmi.c standard x86/isa/orm.c optional isa x86/pci/pci_bus.c optional pci x86/pci/qpi.c optional pci x86/x86/busdma_bounce.c standard x86/x86/busdma_machdep.c standard x86/x86/cpu_machdep.c standard x86/x86/dump_machdep.c standard x86/x86/fdt_machdep.c optional fdt x86/x86/identcpu.c standard x86/x86/intr_machdep.c standard x86/x86/io_apic.c optional apic x86/x86/legacy.c optional native x86/x86/local_apic.c optional apic x86/x86/mca.c standard x86/x86/mptable.c optional apic native x86/x86/mptable_pci.c optional apic native pci +x86/x86/mp_x86.c optional native smp x86/x86/msi.c optional apic pci x86/x86/nexus.c standard x86/x86/tsc.c standard x86/x86/pvclock.c standard x86/x86/delay.c standard x86/xen/hvm.c optional xenhvm x86/xen/xen_intr.c optional xen | xenhvm x86/xen/xen_apic.c optional xenhvm x86/xen/xenpv.c optional xen | xenhvm x86/xen/xen_nexus.c optional xen | xenhvm x86/xen/xen_msi.c optional xen | xenhvm Index: user/ngie/more-tests/sys/conf/files.pc98 =================================================================== --- user/ngie/more-tests/sys/conf/files.pc98 (revision 281974) +++ user/ngie/more-tests/sys/conf/files.pc98 (revision 281975) @@ -1,264 +1,265 @@ # This file tells config what files go into building a kernel, # files marked standard are always included. # # modified for PC-9801/PC-9821 # # $FreeBSD$ # # The long compile-with and dependency lines are required because of # limitations in config: backslash-newline doesn't work in strings, and # dependency lines other than the first are silently ignored. # linux_genassym.o optional compat_linux \ dependency "$S/i386/linux/linux_genassym.c" \ compile-with "${CC} ${CFLAGS:N-fno-common} -c ${.IMPSRC}" \ no-obj no-implicit-rule \ clean "linux_genassym.o" # linux_assym.h optional compat_linux \ dependency "$S/kern/genassym.sh linux_genassym.o" \ compile-with "sh $S/kern/genassym.sh linux_genassym.o > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "linux_assym.h" # svr4_genassym.o optional compat_svr4 \ dependency "$S/i386/svr4/svr4_genassym.c" \ compile-with "${CC} ${CFLAGS:N-fno-common} -c ${.IMPSRC}" \ no-obj no-implicit-rule \ clean "svr4_genassym.o" # svr4_assym.h optional compat_svr4 \ dependency "$S/kern/genassym.sh svr4_genassym.o" \ compile-with "sh $S/kern/genassym.sh svr4_genassym.o > ${.TARGET}" \ no-obj no-implicit-rule before-depend \ clean "svr4_assym.h" # ukbdmap.h optional ukbd_dflt_keymap \ compile-with "/usr/sbin/kbdcontrol -L ${UKBD_DFLT_KEYMAP} | sed -e 's/^static keymap_t.* = /static keymap_t key_map = /' -e 's/^static accentmap_t.* = /static accentmap_t accent_map = /' > ukbdmap.h" \ no-obj no-implicit-rule before-depend \ clean "ukbdmap.h" # cddl/contrib/opensolaris/common/atomic/i386/opensolaris_atomic.S optional zfs compile-with "${ZFS_S}" compat/linprocfs/linprocfs.c optional linprocfs compat/linsysfs/linsysfs.c optional linsysfs compat/linux/linux_emul.c optional compat_linux compat/linux/linux_file.c optional compat_linux compat/linux/linux_fork.c optional compat_linux compat/linux/linux_futex.c optional compat_linux compat/linux/linux_getcwd.c optional compat_linux compat/linux/linux_ioctl.c optional compat_linux compat/linux/linux_ipc.c optional compat_linux compat/linux/linux_mib.c optional compat_linux compat/linux/linux_misc.c optional compat_linux compat/linux/linux_signal.c optional compat_linux compat/linux/linux_socket.c optional compat_linux compat/linux/linux_stats.c optional compat_linux compat/linux/linux_sysctl.c optional compat_linux compat/linux/linux_time.c optional compat_linux compat/linux/linux_timer.c optional compat_linux compat/linux/linux_uid16.c optional compat_linux compat/linux/linux_util.c optional compat_linux compat/svr4/imgact_svr4.c optional compat_svr4 compat/svr4/svr4_fcntl.c optional compat_svr4 compat/svr4/svr4_filio.c optional compat_svr4 compat/svr4/svr4_ioctl.c optional compat_svr4 compat/svr4/svr4_ipc.c optional compat_svr4 compat/svr4/svr4_misc.c optional compat_svr4 compat/svr4/svr4_resource.c optional compat_svr4 compat/svr4/svr4_signal.c optional compat_svr4 compat/svr4/svr4_socket.c optional compat_svr4 compat/svr4/svr4_sockio.c optional compat_svr4 compat/svr4/svr4_stat.c optional compat_svr4 compat/svr4/svr4_stream.c optional compat_svr4 compat/svr4/svr4_syscallnames.c optional compat_svr4 compat/svr4/svr4_sysent.c optional compat_svr4 compat/svr4/svr4_sysvec.c optional compat_svr4 compat/svr4/svr4_termios.c optional compat_svr4 bf_enc.o optional crypto | ipsec \ dependency "$S/crypto/blowfish/arch/i386/bf_enc.S $S/crypto/blowfish/arch/i386/bf_enc_586.S $S/crypto/blowfish/arch/i386/bf_enc_686.S" \ compile-with "${CC} -c -I$S/crypto/blowfish/arch/i386 ${ASM_CFLAGS} ${WERROR} ${.IMPSRC}" \ no-implicit-rule crypto/des/arch/i386/des_enc.S optional crypto | ipsec | netsmb dev/agp/agp_ali.c optional agp dev/agp/agp_amd.c optional agp dev/agp/agp_i810.c optional agp dev/agp/agp_intel.c optional agp dev/agp/agp_nvidia.c optional agp dev/agp/agp_sis.c optional agp dev/agp/agp_via.c optional agp dev/aic/aic_cbus.c optional aic isa dev/ce/ceddk.c optional ce dev/ce/if_ce.c optional ce dev/ce/tau32-ddk.c optional ce \ compile-with "${NORMAL_C} ${NO_WCONSTANT_CONVERSION}" dev/cp/cpddk.c optional cp dev/cp/if_cp.c optional cp dev/ct/bshw_machdep.c optional ct dev/ct/ct.c optional ct dev/ct/ct_isa.c optional ct isa dev/ed/if_ed_cbus.c optional ed isa dev/ed/if_ed_wd80x3.c optional ed isa dev/fb/fb.c optional fb | gdc dev/fdc/fdc.c optional fdc dev/fdc/fdc_cbus.c optional fdc isa dev/fe/if_fe_cbus.c optional fe isa dev/hwpmc/hwpmc_amd.c optional hwpmc dev/hwpmc/hwpmc_intel.c optional hwpmc dev/hwpmc/hwpmc_core.c optional hwpmc dev/hwpmc/hwpmc_uncore.c optional hwpmc dev/hwpmc/hwpmc_pentium.c optional hwpmc dev/hwpmc/hwpmc_piv.c optional hwpmc dev/hwpmc/hwpmc_ppro.c optional hwpmc dev/hwpmc/hwpmc_tsc.c optional hwpmc dev/hwpmc/hwpmc_x86.c optional hwpmc dev/io/iodev.c optional io dev/kbd/kbd.c optional pckbd | sc | ukbd dev/le/if_le_cbus.c optional le isa dev/mse/mse.c optional mse dev/mse/mse_cbus.c optional mse isa dev/sbni/if_sbni.c optional sbni dev/sbni/if_sbni_pci.c optional sbni pci dev/sio/sio_pccard.c optional sio pccard dev/sio/sio_pci.c optional sio pci dev/sio/sio_puc.c optional sio puc dev/snc/dp83932.c optional snc dev/snc/dp83932subr.c optional snc dev/snc/if_snc.c optional snc dev/snc/if_snc_cbus.c optional snc isa dev/snc/if_snc_pccard.c optional snc pccard dev/speaker/spkr.c optional speaker dev/syscons/apm/apm_saver.c optional apm_saver apm dev/uart/uart_cpu_pc98.c optional uart i386/bios/apm.c optional apm #i386/i386/apic_vector.s optional apic i386/i386/atomic.c standard \ compile-with "${CC} -c ${CFLAGS} ${DEFINED_PROF:S/^$/-fomit-frame-pointer/} ${.IMPSRC}" i386/i386/autoconf.c standard i386/i386/bios.c standard i386/i386/bioscall.s standard i386/i386/bpf_jit_machdep.c optional bpf_jitter i386/i386/db_disasm.c optional ddb i386/i386/db_interface.c optional ddb i386/i386/db_trace.c optional ddb i386/i386/elf_machdep.c standard i386/i386/exception.s standard i386/i386/gdb_machdep.c optional gdb i386/i386/i686_mem.c optional mem i386/i386/in_cksum.c optional inet | inet6 i386/i386/initcpu.c standard i386/i386/io.c optional io i386/i386/k6_mem.c optional mem i386/i386/locore.s standard no-obj i386/i386/machdep.c standard i386/i386/mem.c optional mem i386/i386/minidump_machdep.c standard i386/i386/mp_clock.c optional smp i386/i386/mp_machdep.c optional smp i386/i386/mp_watchdog.c optional mp_watchdog smp i386/i386/mpboot.s optional smp i386/i386/perfmon.c optional perfmon i386/i386/pmap.c standard i386/i386/ptrace_machdep.c standard i386/i386/stack_machdep.c optional ddb | stack i386/i386/support.s standard i386/i386/swtch.s standard i386/i386/sys_machdep.c standard i386/i386/trap.c standard i386/i386/uio_machdep.c standard i386/i386/vm86.c standard i386/i386/vm_machdep.c standard i386/ibcs2/ibcs2_errno.c optional ibcs2 i386/ibcs2/ibcs2_fcntl.c optional ibcs2 i386/ibcs2/ibcs2_ioctl.c optional ibcs2 i386/ibcs2/ibcs2_ipc.c optional ibcs2 i386/ibcs2/ibcs2_isc.c optional ibcs2 i386/ibcs2/ibcs2_isc_sysent.c optional ibcs2 i386/ibcs2/ibcs2_misc.c optional ibcs2 i386/ibcs2/ibcs2_msg.c optional ibcs2 i386/ibcs2/ibcs2_other.c optional ibcs2 i386/ibcs2/ibcs2_signal.c optional ibcs2 i386/ibcs2/ibcs2_socksys.c optional ibcs2 i386/ibcs2/ibcs2_stat.c optional ibcs2 i386/ibcs2/ibcs2_sysent.c optional ibcs2 i386/ibcs2/ibcs2_sysi86.c optional ibcs2 i386/ibcs2/ibcs2_sysvec.c optional ibcs2 i386/ibcs2/ibcs2_util.c optional ibcs2 i386/ibcs2/ibcs2_xenix.c optional ibcs2 i386/ibcs2/ibcs2_xenix_sysent.c optional ibcs2 i386/ibcs2/imgact_coff.c optional ibcs2 i386/isa/elink.c optional ep | ie i386/isa/npx.c optional npx i386/isa/pmtimer.c optional pmtimer i386/isa/prof_machdep.c optional profiling-routine i386/linux/imgact_linux.c optional compat_linux i386/linux/linux_dummy.c optional compat_linux i386/linux/linux_locore.s optional compat_linux \ dependency "linux_assym.h" i386/linux/linux_machdep.c optional compat_linux i386/linux/linux_ptrace.c optional compat_linux i386/linux/linux_support.s optional compat_linux \ dependency "linux_assym.h" i386/linux/linux_sysent.c optional compat_linux i386/linux/linux_sysvec.c optional compat_linux i386/pci/pci_cfgreg.c optional pci i386/pci/pci_pir.c optional pci i386/svr4/svr4_locore.s optional compat_svr4 \ dependency "svr4_assym.h" \ warning "COMPAT_SVR4 is broken and should be avoided" i386/svr4/svr4_machdep.c optional compat_svr4 kern/kern_clocksource.c standard kern/imgact_aout.c optional compat_aout kern/imgact_gzip.c optional gzip kern/subr_sfbuf.c standard libkern/divdi3.c standard libkern/flsll.c standard libkern/memmove.c standard libkern/memset.c standard libkern/moddi3.c standard libkern/qdivrem.c standard libkern/ucmpdi2.c standard libkern/udivdi3.c standard libkern/umoddi3.c standard pc98/apm/apm_bioscall.S optional apm pc98/cbus/cbus_dma.c optional isa pc98/cbus/gdc.c optional gdc pc98/cbus/nmi.c standard pc98/cbus/olpt.c optional olpt pc98/cbus/pckbd.c optional pckbd pc98/cbus/pcrtc.c standard pc98/cbus/pmc.c optional pmc pc98/cbus/scgdcrndr.c optional sc gdc pc98/cbus/scterm-sck.c optional sc pc98/cbus/scvtb.c optional sc pc98/cbus/sio.c optional sio pc98/cbus/sio_cbus.c optional sio isa pc98/cbus/syscons_cbus.c optional sc pc98/pc98/busio.s standard pc98/pc98/busiosubr.c standard pc98/pc98/canbepm.c optional canbepm pc98/pc98/canbus.c optional canbus pc98/pc98/canbus_if.m optional canbus pc98/pc98/pc98_machdep.c standard # # x86 shared code between IA32, AMD64 and PC98 architectures # x86/isa/atpic.c optional atpic x86/isa/clock.c standard x86/isa/isa.c optional isa x86/pci/pci_bus.c optional pci x86/x86/busdma_bounce.c standard x86/x86/busdma_machdep.c standard x86/x86/cpu_machdep.c standard x86/x86/dump_machdep.c standard x86/x86/identcpu.c standard x86/x86/intr_machdep.c standard x86/x86/io_apic.c optional apic x86/x86/legacy.c standard x86/x86/local_apic.c optional apic x86/x86/mca.c standard +x86/x86/mp_x86.c optional smp x86/x86/mptable.c optional apic x86/x86/mptable_pci.c optional apic pci x86/x86/msi.c optional apic pci x86/x86/nexus.c standard x86/x86/tsc.c standard x86/x86/delay.c standard Index: user/ngie/more-tests/sys/conf =================================================================== --- user/ngie/more-tests/sys/conf (revision 281974) +++ user/ngie/more-tests/sys/conf (revision 281975) Property changes on: user/ngie/more-tests/sys/conf ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys/conf:r281925-281974 Index: user/ngie/more-tests/sys/dev/ipmi/ipmi.c =================================================================== --- user/ngie/more-tests/sys/dev/ipmi/ipmi.c (revision 281974) +++ user/ngie/more-tests/sys/dev/ipmi/ipmi.c (revision 281975) @@ -1,924 +1,929 @@ /*- * Copyright (c) 2006 IronPort Systems Inc. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef LOCAL_MODULE #include #include #else #include #include #endif /* * Driver request structures are allocated on the stack via alloca() to * avoid calling malloc(), especially for the watchdog handler. * To avoid too much stack growth, a previously allocated structure can * be reused via IPMI_INIT_DRIVER_REQUEST(), but the caller should ensure * that there is adequate reply/request space in the original allocation. */ #define IPMI_INIT_DRIVER_REQUEST(req, addr, cmd, reqlen, replylen) \ bzero((req), sizeof(struct ipmi_request)); \ ipmi_init_request((req), NULL, 0, (addr), (cmd), (reqlen), (replylen)) #define IPMI_ALLOC_DRIVER_REQUEST(req, addr, cmd, reqlen, replylen) \ (req) = __builtin_alloca(sizeof(struct ipmi_request) + \ (reqlen) + (replylen)); \ IPMI_INIT_DRIVER_REQUEST((req), (addr), (cmd), (reqlen), \ (replylen)) #ifdef IPMB static int ipmi_ipmb_checksum(u_char, int); static int ipmi_ipmb_send_message(device_t, u_char, u_char, u_char, u_char, u_char, int) #endif static d_ioctl_t ipmi_ioctl; static d_poll_t ipmi_poll; static d_open_t ipmi_open; static void ipmi_dtor(void *arg); int ipmi_attached = 0; static int on = 1; static SYSCTL_NODE(_hw, OID_AUTO, ipmi, CTLFLAG_RD, 0, "IPMI driver parameters"); SYSCTL_INT(_hw_ipmi, OID_AUTO, on, CTLFLAG_RW, &on, 0, ""); static struct cdevsw ipmi_cdevsw = { .d_version = D_VERSION, .d_open = ipmi_open, .d_ioctl = ipmi_ioctl, .d_poll = ipmi_poll, .d_name = "ipmi", }; static MALLOC_DEFINE(M_IPMI, "ipmi", "ipmi"); static int ipmi_open(struct cdev *cdev, int flags, int fmt, struct thread *td) { struct ipmi_device *dev; struct ipmi_softc *sc; int error; if (!on) return (ENOENT); /* Initialize the per file descriptor data. */ dev = malloc(sizeof(struct ipmi_device), M_IPMI, M_WAITOK | M_ZERO); error = devfs_set_cdevpriv(dev, ipmi_dtor); if (error) { free(dev, M_IPMI); return (error); } sc = cdev->si_drv1; TAILQ_INIT(&dev->ipmi_completed_requests); dev->ipmi_address = IPMI_BMC_SLAVE_ADDR; dev->ipmi_lun = IPMI_BMC_SMS_LUN; dev->ipmi_softc = sc; IPMI_LOCK(sc); sc->ipmi_opened++; IPMI_UNLOCK(sc); return (0); } static int ipmi_poll(struct cdev *cdev, int poll_events, struct thread *td) { struct ipmi_device *dev; struct ipmi_softc *sc; int revents = 0; if (devfs_get_cdevpriv((void **)&dev)) return (0); sc = cdev->si_drv1; IPMI_LOCK(sc); if (poll_events & (POLLIN | POLLRDNORM)) { if (!TAILQ_EMPTY(&dev->ipmi_completed_requests)) revents |= poll_events & (POLLIN | POLLRDNORM); if (dev->ipmi_requests == 0) revents |= POLLERR; } if (revents == 0) { if (poll_events & (POLLIN | POLLRDNORM)) selrecord(td, &dev->ipmi_select); } IPMI_UNLOCK(sc); return (revents); } static void ipmi_purge_completed_requests(struct ipmi_device *dev) { struct ipmi_request *req; while (!TAILQ_EMPTY(&dev->ipmi_completed_requests)) { req = TAILQ_FIRST(&dev->ipmi_completed_requests); TAILQ_REMOVE(&dev->ipmi_completed_requests, req, ir_link); dev->ipmi_requests--; ipmi_free_request(req); } } static void ipmi_dtor(void *arg) { struct ipmi_request *req, *nreq; struct ipmi_device *dev; struct ipmi_softc *sc; dev = arg; sc = dev->ipmi_softc; IPMI_LOCK(sc); if (dev->ipmi_requests) { /* Throw away any pending requests for this device. */ TAILQ_FOREACH_SAFE(req, &sc->ipmi_pending_requests, ir_link, nreq) { if (req->ir_owner == dev) { TAILQ_REMOVE(&sc->ipmi_pending_requests, req, ir_link); dev->ipmi_requests--; ipmi_free_request(req); } } /* Throw away any pending completed requests for this device. */ ipmi_purge_completed_requests(dev); /* * If we still have outstanding requests, they must be stuck * in an interface driver, so wait for those to drain. */ dev->ipmi_closing = 1; while (dev->ipmi_requests > 0) { msleep(&dev->ipmi_requests, &sc->ipmi_requests_lock, PWAIT, "ipmidrain", 0); ipmi_purge_completed_requests(dev); } } sc->ipmi_opened--; IPMI_UNLOCK(sc); /* Cleanup. */ free(dev, M_IPMI); } #ifdef IPMB static int ipmi_ipmb_checksum(u_char *data, int len) { u_char sum = 0; for (; len; len--) { sum += *data++; } return (-sum); } /* XXX: Needs work */ static int ipmi_ipmb_send_message(device_t dev, u_char channel, u_char netfn, u_char command, u_char seq, u_char *data, int data_len) { struct ipmi_softc *sc = device_get_softc(dev); struct ipmi_request *req; u_char slave_addr = 0x52; int error; IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_SEND_MSG, data_len + 8, 0); req->ir_request[0] = channel; req->ir_request[1] = slave_addr; req->ir_request[2] = IPMI_ADDR(netfn, 0); req->ir_request[3] = ipmi_ipmb_checksum(&req->ir_request[1], 2); req->ir_request[4] = sc->ipmi_address; req->ir_request[5] = IPMI_ADDR(seq, sc->ipmi_lun); req->ir_request[6] = command; bcopy(data, &req->ir_request[7], data_len); temp[data_len + 7] = ipmi_ipmb_checksum(&req->ir_request[4], data_len + 3); ipmi_submit_driver_request(sc, req); error = req->ir_error; return (error); } static int ipmi_handle_attn(struct ipmi_softc *sc) { struct ipmi_request *req; int error; device_printf(sc->ipmi_dev, "BMC has a message\n"); IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_GET_MSG_FLAGS, 0, 1); ipmi_submit_driver_request(sc, req); if (req->ir_error == 0 && req->ir_compcode == 0) { if (req->ir_reply[0] & IPMI_MSG_BUFFER_FULL) { device_printf(sc->ipmi_dev, "message buffer full"); } if (req->ir_reply[0] & IPMI_WDT_PRE_TIMEOUT) { device_printf(sc->ipmi_dev, "watchdog about to go off"); } if (req->ir_reply[0] & IPMI_MSG_AVAILABLE) { IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_GET_MSG, 0, 16); device_printf(sc->ipmi_dev, "throw out message "); dump_buf(temp, 16); } } error = req->ir_error; return (error); } #endif #ifdef IPMICTL_SEND_COMMAND_32 #define PTRIN(p) ((void *)(uintptr_t)(p)) #define PTROUT(p) ((uintptr_t)(p)) #endif static int ipmi_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int flags, struct thread *td) { struct ipmi_softc *sc; struct ipmi_device *dev; struct ipmi_request *kreq; struct ipmi_req *req = (struct ipmi_req *)data; struct ipmi_recv *recv = (struct ipmi_recv *)data; struct ipmi_addr addr; #ifdef IPMICTL_SEND_COMMAND_32 struct ipmi_req32 *req32 = (struct ipmi_req32 *)data; struct ipmi_recv32 *recv32 = (struct ipmi_recv32 *)data; union { struct ipmi_req req; struct ipmi_recv recv; } thunk32; #endif int error, len; error = devfs_get_cdevpriv((void **)&dev); if (error) return (error); sc = cdev->si_drv1; #ifdef IPMICTL_SEND_COMMAND_32 /* Convert 32-bit structures to native. */ switch (cmd) { case IPMICTL_SEND_COMMAND_32: req = &thunk32.req; req->addr = PTRIN(req32->addr); req->addr_len = req32->addr_len; req->msgid = req32->msgid; req->msg.netfn = req32->msg.netfn; req->msg.cmd = req32->msg.cmd; req->msg.data_len = req32->msg.data_len; req->msg.data = PTRIN(req32->msg.data); break; case IPMICTL_RECEIVE_MSG_TRUNC_32: case IPMICTL_RECEIVE_MSG_32: recv = &thunk32.recv; recv->addr = PTRIN(recv32->addr); recv->addr_len = recv32->addr_len; recv->msg.data_len = recv32->msg.data_len; recv->msg.data = PTRIN(recv32->msg.data); break; } #endif switch (cmd) { #ifdef IPMICTL_SEND_COMMAND_32 case IPMICTL_SEND_COMMAND_32: #endif case IPMICTL_SEND_COMMAND: /* * XXX: Need to add proper handling of this. */ error = copyin(req->addr, &addr, sizeof(addr)); if (error) return (error); IPMI_LOCK(sc); /* clear out old stuff in queue of stuff done */ /* XXX: This seems odd. */ while ((kreq = TAILQ_FIRST(&dev->ipmi_completed_requests))) { TAILQ_REMOVE(&dev->ipmi_completed_requests, kreq, ir_link); dev->ipmi_requests--; ipmi_free_request(kreq); } IPMI_UNLOCK(sc); kreq = ipmi_alloc_request(dev, req->msgid, IPMI_ADDR(req->msg.netfn, 0), req->msg.cmd, req->msg.data_len, IPMI_MAX_RX); error = copyin(req->msg.data, kreq->ir_request, req->msg.data_len); if (error) { ipmi_free_request(kreq); return (error); } IPMI_LOCK(sc); dev->ipmi_requests++; error = sc->ipmi_enqueue_request(sc, kreq); IPMI_UNLOCK(sc); if (error) return (error); break; #ifdef IPMICTL_SEND_COMMAND_32 case IPMICTL_RECEIVE_MSG_TRUNC_32: case IPMICTL_RECEIVE_MSG_32: #endif case IPMICTL_RECEIVE_MSG_TRUNC: case IPMICTL_RECEIVE_MSG: error = copyin(recv->addr, &addr, sizeof(addr)); if (error) return (error); IPMI_LOCK(sc); kreq = TAILQ_FIRST(&dev->ipmi_completed_requests); if (kreq == NULL) { IPMI_UNLOCK(sc); return (EAGAIN); } addr.channel = IPMI_BMC_CHANNEL; /* XXX */ recv->recv_type = IPMI_RESPONSE_RECV_TYPE; recv->msgid = kreq->ir_msgid; recv->msg.netfn = IPMI_REPLY_ADDR(kreq->ir_addr) >> 2; recv->msg.cmd = kreq->ir_command; error = kreq->ir_error; if (error) { TAILQ_REMOVE(&dev->ipmi_completed_requests, kreq, ir_link); dev->ipmi_requests--; IPMI_UNLOCK(sc); ipmi_free_request(kreq); return (error); } len = kreq->ir_replylen + 1; if (recv->msg.data_len < len && (cmd == IPMICTL_RECEIVE_MSG #ifdef IPMICTL_RECEIVE_MSG_32 || cmd == IPMICTL_RECEIVE_MSG_32 #endif )) { IPMI_UNLOCK(sc); return (EMSGSIZE); } TAILQ_REMOVE(&dev->ipmi_completed_requests, kreq, ir_link); dev->ipmi_requests--; IPMI_UNLOCK(sc); len = min(recv->msg.data_len, len); recv->msg.data_len = len; error = copyout(&addr, recv->addr,sizeof(addr)); if (error == 0) error = copyout(&kreq->ir_compcode, recv->msg.data, 1); if (error == 0) error = copyout(kreq->ir_reply, recv->msg.data + 1, len - 1); ipmi_free_request(kreq); if (error) return (error); break; case IPMICTL_SET_MY_ADDRESS_CMD: IPMI_LOCK(sc); dev->ipmi_address = *(int*)data; IPMI_UNLOCK(sc); break; case IPMICTL_GET_MY_ADDRESS_CMD: IPMI_LOCK(sc); *(int*)data = dev->ipmi_address; IPMI_UNLOCK(sc); break; case IPMICTL_SET_MY_LUN_CMD: IPMI_LOCK(sc); dev->ipmi_lun = *(int*)data & 0x3; IPMI_UNLOCK(sc); break; case IPMICTL_GET_MY_LUN_CMD: IPMI_LOCK(sc); *(int*)data = dev->ipmi_lun; IPMI_UNLOCK(sc); break; case IPMICTL_SET_GETS_EVENTS_CMD: /* device_printf(sc->ipmi_dev, "IPMICTL_SET_GETS_EVENTS_CMD NA\n"); */ break; case IPMICTL_REGISTER_FOR_CMD: case IPMICTL_UNREGISTER_FOR_CMD: return (EOPNOTSUPP); default: device_printf(sc->ipmi_dev, "Unknown IOCTL %lX\n", cmd); return (ENOIOCTL); } #ifdef IPMICTL_SEND_COMMAND_32 /* Update changed fields in 32-bit structures. */ switch (cmd) { case IPMICTL_RECEIVE_MSG_TRUNC_32: case IPMICTL_RECEIVE_MSG_32: recv32->recv_type = recv->recv_type; recv32->msgid = recv->msgid; recv32->msg.netfn = recv->msg.netfn; recv32->msg.cmd = recv->msg.cmd; recv32->msg.data_len = recv->msg.data_len; break; } #endif return (0); } /* * Request management. */ static __inline void ipmi_init_request(struct ipmi_request *req, struct ipmi_device *dev, long msgid, uint8_t addr, uint8_t command, size_t requestlen, size_t replylen) { req->ir_owner = dev; req->ir_msgid = msgid; req->ir_addr = addr; req->ir_command = command; if (requestlen) { req->ir_request = (char *)&req[1]; req->ir_requestlen = requestlen; } if (replylen) { req->ir_reply = (char *)&req[1] + requestlen; req->ir_replybuflen = replylen; } } /* Allocate a new request with request and reply buffers. */ struct ipmi_request * ipmi_alloc_request(struct ipmi_device *dev, long msgid, uint8_t addr, uint8_t command, size_t requestlen, size_t replylen) { struct ipmi_request *req; req = malloc(sizeof(struct ipmi_request) + requestlen + replylen, M_IPMI, M_WAITOK | M_ZERO); ipmi_init_request(req, dev, msgid, addr, command, requestlen, replylen); return (req); } /* Free a request no longer in use. */ void ipmi_free_request(struct ipmi_request *req) { free(req, M_IPMI); } /* Store a processed request on the appropriate completion queue. */ void ipmi_complete_request(struct ipmi_softc *sc, struct ipmi_request *req) { struct ipmi_device *dev; IPMI_LOCK_ASSERT(sc); /* * Anonymous requests (from inside the driver) always have a * waiter that we awaken. */ if (req->ir_owner == NULL) wakeup(req); else { dev = req->ir_owner; TAILQ_INSERT_TAIL(&dev->ipmi_completed_requests, req, ir_link); selwakeup(&dev->ipmi_select); if (dev->ipmi_closing) wakeup(&dev->ipmi_requests); } } /* Perform an internal driver request. */ int ipmi_submit_driver_request(struct ipmi_softc *sc, struct ipmi_request *req, int timo) { return (sc->ipmi_driver_request(sc, req, timo)); } /* * Helper routine for polled system interfaces that use * ipmi_polled_enqueue_request() to queue requests. This request * waits until there is a pending request and then returns the first * request. If the driver is shutting down, it returns NULL. */ struct ipmi_request * ipmi_dequeue_request(struct ipmi_softc *sc) { struct ipmi_request *req; IPMI_LOCK_ASSERT(sc); while (!sc->ipmi_detaching && TAILQ_EMPTY(&sc->ipmi_pending_requests)) cv_wait(&sc->ipmi_request_added, &sc->ipmi_requests_lock); if (sc->ipmi_detaching) return (NULL); req = TAILQ_FIRST(&sc->ipmi_pending_requests); TAILQ_REMOVE(&sc->ipmi_pending_requests, req, ir_link); return (req); } /* Default implementation of ipmi_enqueue_request() for polled interfaces. */ int ipmi_polled_enqueue_request(struct ipmi_softc *sc, struct ipmi_request *req) { IPMI_LOCK_ASSERT(sc); TAILQ_INSERT_TAIL(&sc->ipmi_pending_requests, req, ir_link); cv_signal(&sc->ipmi_request_added); return (0); } /* * Watchdog event handler. */ static int ipmi_set_watchdog(struct ipmi_softc *sc, unsigned int sec) { struct ipmi_request *req; int error; if (sec > 0xffff / 10) return (EINVAL); IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_SET_WDOG, 6, 0); if (sec) { req->ir_request[0] = IPMI_SET_WD_TIMER_DONT_STOP | IPMI_SET_WD_TIMER_SMS_OS; req->ir_request[1] = IPMI_SET_WD_ACTION_RESET; req->ir_request[2] = 0; req->ir_request[3] = 0; /* Timer use */ req->ir_request[4] = (sec * 10) & 0xff; req->ir_request[5] = (sec * 10) >> 8; } else { req->ir_request[0] = IPMI_SET_WD_TIMER_SMS_OS; req->ir_request[1] = 0; req->ir_request[2] = 0; req->ir_request[3] = 0; /* Timer use */ req->ir_request[4] = 0; req->ir_request[5] = 0; } error = ipmi_submit_driver_request(sc, req, 0); if (error) device_printf(sc->ipmi_dev, "Failed to set watchdog\n"); else if (sec) { IPMI_INIT_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_RESET_WDOG, 0, 0); error = ipmi_submit_driver_request(sc, req, 0); if (error) device_printf(sc->ipmi_dev, "Failed to reset watchdog\n"); } return (error); /* dump_watchdog(sc); */ } static void ipmi_wd_event(void *arg, unsigned int cmd, int *error) { struct ipmi_softc *sc = arg; unsigned int timeout; int e; if (dumping) return; cmd &= WD_INTERVAL; if (cmd > 0 && cmd <= 63) { timeout = ((uint64_t)1 << cmd) / 1000000000; if (timeout == 0) timeout = 1; e = ipmi_set_watchdog(sc, timeout); if (e == 0) { *error = 0; sc->ipmi_watchdog_active = 1; } else (void)ipmi_set_watchdog(sc, 0); } else if (atomic_readandclear_int(&sc->ipmi_watchdog_active) != 0) { e = ipmi_set_watchdog(sc, 0); if (e != 0 && cmd == 0) *error = EOPNOTSUPP; } } static void ipmi_startup(void *arg) { struct ipmi_softc *sc = arg; struct ipmi_request *req; device_t dev; int error, i; config_intrhook_disestablish(&sc->ipmi_ich); dev = sc->ipmi_dev; /* Initialize interface-independent state. */ mtx_init(&sc->ipmi_requests_lock, "ipmi requests", NULL, MTX_DEF); mtx_init(&sc->ipmi_io_lock, "ipmi io", NULL, MTX_DEF); cv_init(&sc->ipmi_request_added, "ipmireq"); TAILQ_INIT(&sc->ipmi_pending_requests); /* Initialize interface-dependent state. */ error = sc->ipmi_startup(sc); if (error) { device_printf(dev, "Failed to initialize interface: %d\n", error); return; } /* Send a GET_DEVICE_ID request. */ IPMI_ALLOC_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_GET_DEVICE_ID, 0, 15); error = ipmi_submit_driver_request(sc, req, MAX_TIMEOUT); if (error == EWOULDBLOCK) { device_printf(dev, "Timed out waiting for GET_DEVICE_ID\n"); return; } else if (error) { device_printf(dev, "Failed GET_DEVICE_ID: %d\n", error); return; } else if (req->ir_compcode != 0) { device_printf(dev, "Bad completion code for GET_DEVICE_ID: %d\n", req->ir_compcode); return; } else if (req->ir_replylen < 5) { device_printf(dev, "Short reply for GET_DEVICE_ID: %d\n", req->ir_replylen); return; } device_printf(dev, "IPMI device rev. %d, firmware rev. %d.%d%d, " "version %d.%d\n", req->ir_reply[1] & 0x0f, req->ir_reply[2] & 0x7f, req->ir_reply[3] >> 4, req->ir_reply[3] & 0x0f, req->ir_reply[4] & 0x0f, req->ir_reply[4] >> 4); IPMI_INIT_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_CLEAR_FLAGS, 1, 0); ipmi_submit_driver_request(sc, req, 0); /* XXX: Magic numbers */ if (req->ir_compcode == 0xc0) { device_printf(dev, "Clear flags is busy\n"); } if (req->ir_compcode == 0xc1) { device_printf(dev, "Clear flags illegal\n"); } for (i = 0; i < 8; i++) { IPMI_INIT_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), IPMI_GET_CHANNEL_INFO, 1, 0); req->ir_request[0] = i; ipmi_submit_driver_request(sc, req, 0); if (req->ir_compcode != 0) break; } device_printf(dev, "Number of channels %d\n", i); - /* probe for watchdog */ - IPMI_INIT_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), - IPMI_GET_WDOG, 0, 0); + /* + * Probe for watchdog, but only for backends which support + * polled driver requests. + */ + if (sc->ipmi_driver_requests_polled) { + IPMI_INIT_DRIVER_REQUEST(req, IPMI_ADDR(IPMI_APP_REQUEST, 0), + IPMI_GET_WDOG, 0, 0); - ipmi_submit_driver_request(sc, req, 0); + ipmi_submit_driver_request(sc, req, 0); - if (req->ir_compcode == 0x00) { - device_printf(dev, "Attached watchdog\n"); - /* register the watchdog event handler */ - sc->ipmi_watchdog_tag = EVENTHANDLER_REGISTER(watchdog_list, - ipmi_wd_event, sc, 0); + if (req->ir_compcode == 0x00) { + device_printf(dev, "Attached watchdog\n"); + /* register the watchdog event handler */ + sc->ipmi_watchdog_tag = EVENTHANDLER_REGISTER( + watchdog_list, ipmi_wd_event, sc, 0); + } } sc->ipmi_cdev = make_dev(&ipmi_cdevsw, device_get_unit(dev), UID_ROOT, GID_OPERATOR, 0660, "ipmi%d", device_get_unit(dev)); if (sc->ipmi_cdev == NULL) { device_printf(dev, "Failed to create cdev\n"); return; } sc->ipmi_cdev->si_drv1 = sc; } int ipmi_attach(device_t dev) { struct ipmi_softc *sc = device_get_softc(dev); int error; if (sc->ipmi_irq_res != NULL && sc->ipmi_intr != NULL) { error = bus_setup_intr(dev, sc->ipmi_irq_res, INTR_TYPE_MISC, NULL, sc->ipmi_intr, sc, &sc->ipmi_irq); if (error) { device_printf(dev, "can't set up interrupt\n"); return (error); } } bzero(&sc->ipmi_ich, sizeof(struct intr_config_hook)); sc->ipmi_ich.ich_func = ipmi_startup; sc->ipmi_ich.ich_arg = sc; if (config_intrhook_establish(&sc->ipmi_ich) != 0) { device_printf(dev, "can't establish configuration hook\n"); return (ENOMEM); } ipmi_attached = 1; return (0); } int ipmi_detach(device_t dev) { struct ipmi_softc *sc; sc = device_get_softc(dev); /* Fail if there are any open handles. */ IPMI_LOCK(sc); if (sc->ipmi_opened) { IPMI_UNLOCK(sc); return (EBUSY); } IPMI_UNLOCK(sc); if (sc->ipmi_cdev) destroy_dev(sc->ipmi_cdev); /* Detach from watchdog handling and turn off watchdog. */ if (sc->ipmi_watchdog_tag) { EVENTHANDLER_DEREGISTER(watchdog_list, sc->ipmi_watchdog_tag); ipmi_set_watchdog(sc, 0); } /* XXX: should use shutdown callout I think. */ /* If the backend uses a kthread, shut it down. */ IPMI_LOCK(sc); sc->ipmi_detaching = 1; if (sc->ipmi_kthread) { cv_broadcast(&sc->ipmi_request_added); msleep(sc->ipmi_kthread, &sc->ipmi_requests_lock, 0, "ipmi_wait", 0); } IPMI_UNLOCK(sc); if (sc->ipmi_irq) bus_teardown_intr(dev, sc->ipmi_irq_res, sc->ipmi_irq); ipmi_release_resources(dev); mtx_destroy(&sc->ipmi_io_lock); mtx_destroy(&sc->ipmi_requests_lock); return (0); } void ipmi_release_resources(device_t dev) { struct ipmi_softc *sc; int i; sc = device_get_softc(dev); if (sc->ipmi_irq) bus_teardown_intr(dev, sc->ipmi_irq_res, sc->ipmi_irq); if (sc->ipmi_irq_res) bus_release_resource(dev, SYS_RES_IRQ, sc->ipmi_irq_rid, sc->ipmi_irq_res); for (i = 0; i < MAX_RES; i++) if (sc->ipmi_io_res[i]) bus_release_resource(dev, sc->ipmi_io_type, sc->ipmi_io_rid + i, sc->ipmi_io_res[i]); } devclass_t ipmi_devclass; /* XXX: Why? */ static void ipmi_unload(void *arg) { device_t * devs; int count; int i; if (devclass_get_devices(ipmi_devclass, &devs, &count) != 0) return; for (i = 0; i < count; i++) device_delete_child(device_get_parent(devs[i]), devs[i]); free(devs, M_TEMP); } SYSUNINIT(ipmi_unload, SI_SUB_DRIVERS, SI_ORDER_FIRST, ipmi_unload, NULL); #ifdef IMPI_DEBUG static void dump_buf(u_char *data, int len) { char buf[20]; char line[1024]; char temp[30]; int count = 0; int i=0; printf("Address %p len %d\n", data, len); if (len > 256) len = 256; line[0] = '\000'; for (; len > 0; len--, data++) { sprintf(temp, "%02x ", *data); strcat(line, temp); if (*data >= ' ' && *data <= '~') buf[count] = *data; else if (*data >= 'A' && *data <= 'Z') buf[count] = *data; else buf[count] = '.'; if (++count == 16) { buf[count] = '\000'; count = 0; printf(" %3x %s %s\n", i, line, buf); i+=16; line[0] = '\000'; } } buf[count] = '\000'; for (; count != 16; count++) { strcat(line, " "); } printf(" %3x %s %s\n", i, line, buf); } #endif Index: user/ngie/more-tests/sys/dev/ipmi/ipmi_kcs.c =================================================================== --- user/ngie/more-tests/sys/dev/ipmi/ipmi_kcs.c (revision 281974) +++ user/ngie/more-tests/sys/dev/ipmi/ipmi_kcs.c (revision 281975) @@ -1,642 +1,643 @@ /*- * Copyright (c) 2006 IronPort Systems Inc. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #ifdef LOCAL_MODULE #include #include #else #include #include #endif static void kcs_clear_obf(struct ipmi_softc *, int); static void kcs_error(struct ipmi_softc *); static int kcs_wait_for_ibf(struct ipmi_softc *, int); static int kcs_wait_for_obf(struct ipmi_softc *, int); static int kcs_wait_for_ibf(struct ipmi_softc *sc, int state) { int status, start = ticks; status = INB(sc, KCS_CTL_STS); if (state == 0) { /* WAIT FOR IBF = 0 */ while (ticks - start < MAX_TIMEOUT && status & KCS_STATUS_IBF) { DELAY(100); status = INB(sc, KCS_CTL_STS); } } else { /* WAIT FOR IBF = 1 */ while (ticks - start < MAX_TIMEOUT && !(status & KCS_STATUS_IBF)) { DELAY(100); status = INB(sc, KCS_CTL_STS); } } return (status); } static int kcs_wait_for_obf(struct ipmi_softc *sc, int state) { int status, start = ticks; status = INB(sc, KCS_CTL_STS); if (state == 0) { /* WAIT FOR OBF = 0 */ while (ticks - start < MAX_TIMEOUT && status & KCS_STATUS_OBF) { DELAY(100); status = INB(sc, KCS_CTL_STS); } } else { /* WAIT FOR OBF = 1 */ while (ticks - start < MAX_TIMEOUT && !(status & KCS_STATUS_OBF)) { DELAY(100); status = INB(sc, KCS_CTL_STS); } } return (status); } static void kcs_clear_obf(struct ipmi_softc *sc, int status) { int data; /* Clear OBF */ if (status & KCS_STATUS_OBF) { data = INB(sc, KCS_DATA); } } static void kcs_error(struct ipmi_softc *sc) { int retry, status; u_char data; for (retry = 0; retry < 2; retry++) { /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); /* ABORT */ OUTB(sc, KCS_CTL_STS, KCS_CONTROL_GET_STATUS_ABORT); /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); /* Clear OBF */ kcs_clear_obf(sc, status); if (status & KCS_STATUS_OBF) { data = INB(sc, KCS_DATA); if (data != 0) device_printf(sc->ipmi_dev, "KCS Error Data %02x\n", data); } /* 0x00 to DATA_IN */ OUTB(sc, KCS_DATA, 0x00); /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); if (KCS_STATUS_STATE(status) == KCS_STATUS_STATE_READ) { /* Wait for OBF = 1 */ status = kcs_wait_for_obf(sc, 1); /* Read error status */ data = INB(sc, KCS_DATA); if (data != 0) device_printf(sc->ipmi_dev, "KCS error: %02x\n", data); /* Write READ into Data_in */ OUTB(sc, KCS_DATA, KCS_DATA_IN_READ); /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); } /* IDLE STATE */ if (KCS_STATUS_STATE(status) == KCS_STATUS_STATE_IDLE) { /* Wait for OBF = 1 */ status = kcs_wait_for_obf(sc, 1); /* Clear OBF */ kcs_clear_obf(sc, status); return; } } device_printf(sc->ipmi_dev, "KCS: Error retry exhausted\n"); } /* * Start to write a request. Waits for IBF to clear and then sends the * WR_START command. */ static int kcs_start_write(struct ipmi_softc *sc) { int retry, status; for (retry = 0; retry < 10; retry++) { /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); if (status & KCS_STATUS_IBF) return (0); /* Clear OBF */ kcs_clear_obf(sc, status); /* Write start to command */ OUTB(sc, KCS_CTL_STS, KCS_CONTROL_WRITE_START); /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); if (status & KCS_STATUS_IBF) return (0); if (KCS_STATUS_STATE(status) == KCS_STATUS_STATE_WRITE) break; DELAY(1000000); } if (KCS_STATUS_STATE(status) != KCS_STATUS_STATE_WRITE) /* error state */ return (0); /* Clear OBF */ kcs_clear_obf(sc, status); return (1); } /* * Write a byte of the request message, excluding the last byte of the * message which requires special handling. */ static int kcs_write_byte(struct ipmi_softc *sc, u_char data) { int status; /* Data to Data */ OUTB(sc, KCS_DATA, data); /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); if (status & KCS_STATUS_IBF) return (0); if (KCS_STATUS_STATE(status) != KCS_STATUS_STATE_WRITE) return (0); /* Clear OBF */ kcs_clear_obf(sc, status); return (1); } /* * Write the last byte of a request message. */ static int kcs_write_last_byte(struct ipmi_softc *sc, u_char data) { int status; /* Write end to command */ OUTB(sc, KCS_CTL_STS, KCS_CONTROL_WRITE_END); /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); if (status & KCS_STATUS_IBF) return (0); if (KCS_STATUS_STATE(status) != KCS_STATUS_STATE_WRITE) /* error state */ return (0); /* Clear OBF */ kcs_clear_obf(sc, status); /* Send data byte to DATA. */ OUTB(sc, KCS_DATA, data); return (1); } /* * Read one byte of the reply message. */ static int kcs_read_byte(struct ipmi_softc *sc, u_char *data) { int status; u_char dummy; /* Wait for IBF = 0 */ status = kcs_wait_for_ibf(sc, 0); /* Read State */ if (KCS_STATUS_STATE(status) == KCS_STATUS_STATE_READ) { /* Wait for OBF = 1 */ status = kcs_wait_for_obf(sc, 1); if ((status & KCS_STATUS_OBF) == 0) return (0); /* Read Data_out */ *data = INB(sc, KCS_DATA); /* Write READ into Data_in */ OUTB(sc, KCS_DATA, KCS_DATA_IN_READ); return (1); } /* Idle State */ if (KCS_STATUS_STATE(status) == KCS_STATUS_STATE_IDLE) { /* Wait for OBF = 1*/ status = kcs_wait_for_obf(sc, 1); if ((status & KCS_STATUS_OBF) == 0) return (0); /* Read Dummy */ dummy = INB(sc, KCS_DATA); return (2); } /* Error State */ return (0); } /* * Send a request message and collect the reply. Returns true if we * succeed. */ static int kcs_polled_request(struct ipmi_softc *sc, struct ipmi_request *req) { u_char *cp, data; int i, state; IPMI_IO_LOCK(sc); /* Send the request. */ if (!kcs_start_write(sc)) { device_printf(sc->ipmi_dev, "KCS: Failed to start write\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: WRITE_START... ok\n"); #endif if (!kcs_write_byte(sc, req->ir_addr)) { device_printf(sc->ipmi_dev, "KCS: Failed to write address\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Wrote address: %02x\n", req->ir_addr); #endif if (req->ir_requestlen == 0) { if (!kcs_write_last_byte(sc, req->ir_command)) { device_printf(sc->ipmi_dev, "KCS: Failed to write command\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Wrote command: %02x\n", req->ir_command); #endif } else { if (!kcs_write_byte(sc, req->ir_command)) { device_printf(sc->ipmi_dev, "KCS: Failed to write command\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Wrote command: %02x\n", req->ir_command); #endif cp = req->ir_request; for (i = 0; i < req->ir_requestlen - 1; i++) { if (!kcs_write_byte(sc, *cp++)) { device_printf(sc->ipmi_dev, "KCS: Failed to write data byte %d\n", i + 1); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Wrote data: %02x\n", cp[-1]); #endif } if (!kcs_write_last_byte(sc, *cp)) { device_printf(sc->ipmi_dev, "KCS: Failed to write last dta byte\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Wrote last data: %02x\n", *cp); #endif } /* Read the reply. First, read the NetFn/LUN. */ if (kcs_read_byte(sc, &data) != 1) { device_printf(sc->ipmi_dev, "KCS: Failed to read address\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Read address: %02x\n", data); #endif if (data != IPMI_REPLY_ADDR(req->ir_addr)) { device_printf(sc->ipmi_dev, "KCS: Reply address mismatch\n"); goto fail; } /* Next we read the command. */ if (kcs_read_byte(sc, &data) != 1) { device_printf(sc->ipmi_dev, "KCS: Failed to read command\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Read command: %02x\n", data); #endif if (data != req->ir_command) { device_printf(sc->ipmi_dev, "KCS: Command mismatch\n"); goto fail; } /* Next we read the completion code. */ if (kcs_read_byte(sc, &req->ir_compcode) != 1) { device_printf(sc->ipmi_dev, "KCS: Failed to read completion code\n"); goto fail; } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Read completion code: %02x\n", req->ir_compcode); #endif /* Finally, read the reply from the BMC. */ i = 0; for (;;) { state = kcs_read_byte(sc, &data); if (state == 0) { device_printf(sc->ipmi_dev, "KCS: Read failed on byte %d\n", i + 1); goto fail; } if (state == 2) break; if (i < req->ir_replybuflen) { req->ir_reply[i] = data; #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: Read data %02x\n", data); } else { device_printf(sc->ipmi_dev, "KCS: Read short %02x byte %d\n", data, i + 1); #endif } i++; } IPMI_IO_UNLOCK(sc); req->ir_replylen = i; #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: READ finished (%d bytes)\n", i); if (req->ir_replybuflen < i) #else if (req->ir_replybuflen < i && req->ir_replybuflen != 0) #endif device_printf(sc->ipmi_dev, "KCS: Read short: %zd buffer, %d actual\n", req->ir_replybuflen, i); return (1); fail: kcs_error(sc); IPMI_IO_UNLOCK(sc); return (0); } static void kcs_loop(void *arg) { struct ipmi_softc *sc = arg; struct ipmi_request *req; int i, ok; IPMI_LOCK(sc); while ((req = ipmi_dequeue_request(sc)) != NULL) { IPMI_UNLOCK(sc); ok = 0; for (i = 0; i < 3 && !ok; i++) ok = kcs_polled_request(sc, req); if (ok) req->ir_error = 0; else req->ir_error = EIO; IPMI_LOCK(sc); ipmi_complete_request(sc, req); } IPMI_UNLOCK(sc); kproc_exit(0); } static int kcs_startup(struct ipmi_softc *sc) { return (kproc_create(kcs_loop, sc, &sc->ipmi_kthread, 0, 0, "%s: kcs", device_get_nameunit(sc->ipmi_dev))); } static int kcs_driver_request(struct ipmi_softc *sc, struct ipmi_request *req, int timo) { int i, ok; ok = 0; for (i = 0; i < 3 && !ok; i++) ok = kcs_polled_request(sc, req); if (ok) req->ir_error = 0; else req->ir_error = EIO; return (req->ir_error); } int ipmi_kcs_attach(struct ipmi_softc *sc) { int status; /* Setup function pointers. */ sc->ipmi_startup = kcs_startup; sc->ipmi_enqueue_request = ipmi_polled_enqueue_request; sc->ipmi_driver_request = kcs_driver_request; + sc->ipmi_driver_requests_polled = 1; /* See if we can talk to the controller. */ status = INB(sc, KCS_CTL_STS); if (status == 0xff) { device_printf(sc->ipmi_dev, "couldn't find it\n"); return (ENXIO); } #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "KCS: initial state: %02x\n", status); #endif if (status & KCS_STATUS_OBF || KCS_STATUS_STATE(status) != KCS_STATUS_STATE_IDLE) kcs_error(sc); return (0); } /* * Determine the alignment automatically for a PCI attachment. In this case, * any unused bytes will return 0x00 when read. We make use of the C/D bit * in the CTL_STS register to try to start a GET_STATUS transaction. When * we write the command, that bit should be set, so we should get a non-zero * value back when we read CTL_STS if the offset we are testing is the CTL_STS * register. */ int ipmi_kcs_probe_align(struct ipmi_softc *sc) { int data, status; sc->ipmi_io_spacing = 1; retry: #ifdef KCS_DEBUG device_printf(sc->ipmi_dev, "Trying KCS align %d... ", sc->ipmi_io_spacing); #endif /* Wait for IBF = 0 */ status = INB(sc, KCS_CTL_STS); while (status & KCS_STATUS_IBF) { DELAY(100); status = INB(sc, KCS_CTL_STS); } OUTB(sc, KCS_CTL_STS, KCS_CONTROL_GET_STATUS_ABORT); /* Wait for IBF = 0 */ status = INB(sc, KCS_CTL_STS); while (status & KCS_STATUS_IBF) { DELAY(100); status = INB(sc, KCS_CTL_STS); } /* If we got 0x00 back, then this must not be the CTL_STS register. */ if (status == 0) { #ifdef KCS_DEBUG printf("failed\n"); #endif sc->ipmi_io_spacing <<= 1; if (sc->ipmi_io_spacing > 4) return (0); goto retry; } #ifdef KCS_DEBUG printf("ok\n"); #endif /* Finish out the transaction. */ /* Clear OBF */ if (status & KCS_STATUS_OBF) data = INB(sc, KCS_DATA); /* 0x00 to DATA_IN */ OUTB(sc, KCS_DATA, 0); /* Wait for IBF = 0 */ status = INB(sc, KCS_CTL_STS); while (status & KCS_STATUS_IBF) { DELAY(100); status = INB(sc, KCS_CTL_STS); } if (KCS_STATUS_STATE(status) == KCS_STATUS_STATE_READ) { /* Wait for IBF = 1 */ while (!(status & KCS_STATUS_OBF)) { DELAY(100); status = INB(sc, KCS_CTL_STS); } /* Read error status. */ data = INB(sc, KCS_DATA); /* Write dummy READ to DATA_IN. */ OUTB(sc, KCS_DATA, KCS_DATA_IN_READ); /* Wait for IBF = 0 */ status = INB(sc, KCS_CTL_STS); while (status & KCS_STATUS_IBF) { DELAY(100); status = INB(sc, KCS_CTL_STS); } } if (KCS_STATUS_STATE(status) == KCS_STATUS_STATE_IDLE) { /* Wait for IBF = 1 */ while (!(status & KCS_STATUS_OBF)) { DELAY(100); status = INB(sc, KCS_CTL_STS); } /* Clear OBF */ if (status & KCS_STATUS_OBF) data = INB(sc, KCS_DATA); } else device_printf(sc->ipmi_dev, "KCS probe: end state %x\n", KCS_STATUS_STATE(status)); return (1); } Index: user/ngie/more-tests/sys/dev/ipmi/ipmi_smic.c =================================================================== --- user/ngie/more-tests/sys/dev/ipmi/ipmi_smic.c (revision 281974) +++ user/ngie/more-tests/sys/dev/ipmi/ipmi_smic.c (revision 281975) @@ -1,431 +1,432 @@ /*- * Copyright (c) 2006 IronPort Systems Inc. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #ifdef LOCAL_MODULE #include #include #else #include #include #endif static void smic_wait_for_tx_okay(struct ipmi_softc *); static void smic_wait_for_rx_okay(struct ipmi_softc *); static void smic_wait_for_not_busy(struct ipmi_softc *); static void smic_set_busy(struct ipmi_softc *); static void smic_wait_for_tx_okay(struct ipmi_softc *sc) { int flags; do { flags = INB(sc, SMIC_FLAGS); } while (!(flags & SMIC_STATUS_TX_RDY)); } static void smic_wait_for_rx_okay(struct ipmi_softc *sc) { int flags; do { flags = INB(sc, SMIC_FLAGS); } while (!(flags & SMIC_STATUS_RX_RDY)); } static void smic_wait_for_not_busy(struct ipmi_softc *sc) { int flags; do { flags = INB(sc, SMIC_FLAGS); } while (flags & SMIC_STATUS_BUSY); } static void smic_set_busy(struct ipmi_softc *sc) { int flags; flags = INB(sc, SMIC_FLAGS); flags |= SMIC_STATUS_BUSY; flags &= ~SMIC_STATUS_RESERVED; OUTB(sc, SMIC_FLAGS, flags); } /* * Start a transfer with a WR_START transaction that sends the NetFn/LUN * address. */ static int smic_start_write(struct ipmi_softc *sc, u_char data) { u_char error, status; smic_wait_for_not_busy(sc); OUTB(sc, SMIC_CTL_STS, SMIC_CC_SMS_WR_START); OUTB(sc, SMIC_DATA, data); smic_set_busy(sc); smic_wait_for_not_busy(sc); status = INB(sc, SMIC_CTL_STS); if (status != SMIC_SC_SMS_WR_START) { error = INB(sc, SMIC_DATA); device_printf(sc->ipmi_dev, "SMIC: Write did not start %02x\n", error); return (0); } return (1); } /* * Write a byte in the middle of the message (either the command or one of * the data bytes) using a WR_NEXT transaction. */ static int smic_write_next(struct ipmi_softc *sc, u_char data) { u_char error, status; smic_wait_for_tx_okay(sc); OUTB(sc, SMIC_CTL_STS, SMIC_CC_SMS_WR_NEXT); OUTB(sc, SMIC_DATA, data); smic_set_busy(sc); smic_wait_for_not_busy(sc); status = INB(sc, SMIC_CTL_STS); if (status != SMIC_SC_SMS_WR_NEXT) { error = INB(sc, SMIC_DATA); device_printf(sc->ipmi_dev, "SMIC: Write did not next %02x\n", error); return (0); } return (1); } /* * Write the last byte of a transfer to end the write phase via a WR_END * transaction. */ static int smic_write_last(struct ipmi_softc *sc, u_char data) { u_char error, status; smic_wait_for_tx_okay(sc); OUTB(sc, SMIC_CTL_STS, SMIC_CC_SMS_WR_END); OUTB(sc, SMIC_DATA, data); smic_set_busy(sc); smic_wait_for_not_busy(sc); status = INB(sc, SMIC_CTL_STS); if (status != SMIC_SC_SMS_WR_END) { error = INB(sc, SMIC_DATA); device_printf(sc->ipmi_dev, "SMIC: Write did not end %02x\n", error); return (0); } return (1); } /* * Start the read phase of a transfer with a RD_START transaction. */ static int smic_start_read(struct ipmi_softc *sc, u_char *data) { u_char error, status; smic_wait_for_not_busy(sc); smic_wait_for_rx_okay(sc); OUTB(sc, SMIC_CTL_STS, SMIC_CC_SMS_RD_START); smic_set_busy(sc); smic_wait_for_not_busy(sc); status = INB(sc, SMIC_CTL_STS); if (status != SMIC_SC_SMS_RD_START) { error = INB(sc, SMIC_DATA); device_printf(sc->ipmi_dev, "SMIC: Read did not start %02x\n", error); return (0); } *data = INB(sc, SMIC_DATA); return (1); } /* * Read a byte via a RD_NEXT transaction. If this was the last byte, return * 2 rather than 1. */ static int smic_read_byte(struct ipmi_softc *sc, u_char *data) { u_char error, status; smic_wait_for_rx_okay(sc); OUTB(sc, SMIC_CTL_STS, SMIC_SC_SMS_RD_NEXT); smic_set_busy(sc); smic_wait_for_not_busy(sc); status = INB(sc, SMIC_CTL_STS); if (status != SMIC_SC_SMS_RD_NEXT && status != SMIC_SC_SMS_RD_END) { error = INB(sc, SMIC_DATA); device_printf(sc->ipmi_dev, "SMIC: Read did not next %02x\n", error); return (0); } *data = INB(sc, SMIC_DATA); if (status == SMIC_SC_SMS_RD_NEXT) return (1); else return (2); } /* Complete a transfer via a RD_END transaction after reading the last byte. */ static int smic_read_end(struct ipmi_softc *sc) { u_char error, status; OUTB(sc, SMIC_CTL_STS, SMIC_CC_SMS_RD_END); smic_set_busy(sc); smic_wait_for_not_busy(sc); status = INB(sc, SMIC_CTL_STS); if (status != SMIC_SC_SMS_RDY) { error = INB(sc, SMIC_DATA); device_printf(sc->ipmi_dev, "SMIC: Read did not end %02x\n", error); return (0); } return (1); } static int smic_polled_request(struct ipmi_softc *sc, struct ipmi_request *req) { u_char *cp, data; int i, state; /* First, start the message with the address. */ if (!smic_start_write(sc, req->ir_addr)) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: WRITE_START address: %02x\n", req->ir_addr); #endif if (req->ir_requestlen == 0) { /* Send the command as the last byte. */ if (!smic_write_last(sc, req->ir_command)) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Wrote command: %02x\n", req->ir_command); #endif } else { /* Send the command. */ if (!smic_write_next(sc, req->ir_command)) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Wrote command: %02x\n", req->ir_command); #endif /* Send the payload. */ cp = req->ir_request; for (i = 0; i < req->ir_requestlen - 1; i++) { if (!smic_write_next(sc, *cp++)) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Wrote data: %02x\n", cp[-1]); #endif } if (!smic_write_last(sc, *cp)) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Write last data: %02x\n", *cp); #endif } /* Start the read phase by reading the NetFn/LUN. */ if (smic_start_read(sc, &data) != 1) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Read address: %02x\n", data); #endif if (data != IPMI_REPLY_ADDR(req->ir_addr)) { device_printf(sc->ipmi_dev, "SMIC: Reply address mismatch\n"); return (0); } /* Read the command. */ if (smic_read_byte(sc, &data) != 1) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Read command: %02x\n", data); #endif if (data != req->ir_command) { device_printf(sc->ipmi_dev, "SMIC: Command mismatch\n"); return (0); } /* Read the completion code. */ state = smic_read_byte(sc, &req->ir_compcode); if (state == 0) return (0); #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Read completion code: %02x\n", req->ir_compcode); #endif /* Finally, read the reply from the BMC. */ i = 0; while (state == 1) { state = smic_read_byte(sc, &data); if (state == 0) return (0); if (i < req->ir_replybuflen) { req->ir_reply[i] = data; #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: Read data: %02x\n", data); } else { device_printf(sc->ipmi_dev, "SMIC: Read short %02x byte %d\n", data, i + 1); #endif } i++; } /* Terminate the transfer. */ if (!smic_read_end(sc)) return (0); req->ir_replylen = i; #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: READ finished (%d bytes)\n", i); if (req->ir_replybuflen < i) #else if (req->ir_replybuflen < i && req->ir_replybuflen != 0) #endif device_printf(sc->ipmi_dev, "SMIC: Read short: %zd buffer, %d actual\n", req->ir_replybuflen, i); return (1); } static void smic_loop(void *arg) { struct ipmi_softc *sc = arg; struct ipmi_request *req; int i, ok; IPMI_LOCK(sc); while ((req = ipmi_dequeue_request(sc)) != NULL) { IPMI_UNLOCK(sc); ok = 0; for (i = 0; i < 3 && !ok; i++) { IPMI_IO_LOCK(sc); ok = smic_polled_request(sc, req); IPMI_IO_UNLOCK(sc); } if (ok) req->ir_error = 0; else req->ir_error = EIO; IPMI_LOCK(sc); ipmi_complete_request(sc, req); } IPMI_UNLOCK(sc); kproc_exit(0); } static int smic_startup(struct ipmi_softc *sc) { return (kproc_create(smic_loop, sc, &sc->ipmi_kthread, 0, 0, "%s: smic", device_get_nameunit(sc->ipmi_dev))); } static int smic_driver_request(struct ipmi_softc *sc, struct ipmi_request *req, int timo) { int i, ok; ok = 0; for (i = 0; i < 3 && !ok; i++) { IPMI_IO_LOCK(sc); ok = smic_polled_request(sc, req); IPMI_IO_UNLOCK(sc); } if (ok) req->ir_error = 0; else req->ir_error = EIO; return (req->ir_error); } int ipmi_smic_attach(struct ipmi_softc *sc) { int flags; /* Setup function pointers. */ sc->ipmi_startup = smic_startup; sc->ipmi_enqueue_request = ipmi_polled_enqueue_request; sc->ipmi_driver_request = smic_driver_request; + sc->ipmi_driver_requests_polled = 1; /* See if we can talk to the controller. */ flags = INB(sc, SMIC_FLAGS); if (flags == 0xff) { device_printf(sc->ipmi_dev, "couldn't find it\n"); return (ENXIO); } #ifdef SMIC_DEBUG device_printf(sc->ipmi_dev, "SMIC: initial state: %02x\n", flags); #endif return (0); } Index: user/ngie/more-tests/sys/dev/ipmi/ipmivars.h =================================================================== --- user/ngie/more-tests/sys/dev/ipmi/ipmivars.h (revision 281974) +++ user/ngie/more-tests/sys/dev/ipmi/ipmivars.h (revision 281975) @@ -1,260 +1,261 @@ /*- * Copyright (c) 2006 IronPort Systems Inc. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef __IPMIVARS_H__ #define __IPMIVARS_H__ struct ipmi_get_info { int iface_type; uint64_t address; int offset; int io_mode; int irq; }; struct ipmi_device; struct ipmi_request { TAILQ_ENTRY(ipmi_request) ir_link; struct ipmi_device *ir_owner; /* Driver uses NULL. */ u_char *ir_request; /* Request is data to send to BMC. */ size_t ir_requestlen; u_char *ir_reply; /* Reply is data read from BMC. */ size_t ir_replybuflen; /* Length of ir_reply[] buffer. */ int ir_replylen; /* Length of reply from BMC. */ int ir_error; long ir_msgid; uint8_t ir_addr; uint8_t ir_command; uint8_t ir_compcode; }; #define MAX_RES 3 #define KCS_DATA 0 #define KCS_CTL_STS 1 #define SMIC_DATA 0 #define SMIC_CTL_STS 1 #define SMIC_FLAGS 2 struct ipmi_softc; /* Per file descriptor data. */ struct ipmi_device { TAILQ_ENTRY(ipmi_device) ipmi_link; TAILQ_HEAD(,ipmi_request) ipmi_completed_requests; struct selinfo ipmi_select; struct ipmi_softc *ipmi_softc; int ipmi_closing; int ipmi_requests; u_char ipmi_address; /* IPMB address. */ u_char ipmi_lun; }; struct ipmi_kcs { }; struct ipmi_smic { }; struct ipmi_ssif { device_t smbus; int smbus_address; }; struct ipmi_softc { device_t ipmi_dev; union { struct ipmi_kcs kcs; struct ipmi_smic smic; struct ipmi_ssif ssif; } _iface; int ipmi_io_rid; int ipmi_io_type; struct mtx ipmi_io_lock; struct resource *ipmi_io_res[MAX_RES]; int ipmi_io_spacing; int ipmi_irq_rid; struct resource *ipmi_irq_res; void *ipmi_irq; int ipmi_detaching; int ipmi_opened; struct cdev *ipmi_cdev; TAILQ_HEAD(,ipmi_request) ipmi_pending_requests; + int ipmi_driver_requests_polled; eventhandler_tag ipmi_watchdog_tag; int ipmi_watchdog_active; struct intr_config_hook ipmi_ich; struct mtx ipmi_requests_lock; struct cv ipmi_request_added; struct proc *ipmi_kthread; driver_intr_t *ipmi_intr; int (*ipmi_startup)(struct ipmi_softc *); int (*ipmi_enqueue_request)(struct ipmi_softc *, struct ipmi_request *); int (*ipmi_driver_request)(struct ipmi_softc *, struct ipmi_request *, int); }; #define ipmi_ssif_smbus_address _iface.ssif.smbus_address #define ipmi_ssif_smbus _iface.ssif.smbus struct ipmi_ipmb { u_char foo; }; #define KCS_MODE 0x01 #define SMIC_MODE 0x02 #define BT_MODE 0x03 #define SSIF_MODE 0x04 /* KCS status flags */ #define KCS_STATUS_OBF 0x01 /* Data Out ready from BMC */ #define KCS_STATUS_IBF 0x02 /* Data In from System */ #define KCS_STATUS_SMS_ATN 0x04 /* Ready in RX queue */ #define KCS_STATUS_C_D 0x08 /* Command/Data register write*/ #define KCS_STATUS_OEM1 0x10 #define KCS_STATUS_OEM2 0x20 #define KCS_STATUS_S0 0x40 #define KCS_STATUS_S1 0x80 #define KCS_STATUS_STATE(x) ((x)>>6) #define KCS_STATUS_STATE_IDLE 0x0 #define KCS_STATUS_STATE_READ 0x1 #define KCS_STATUS_STATE_WRITE 0x2 #define KCS_STATUS_STATE_ERROR 0x3 #define KCS_IFACE_STATUS_OK 0x00 #define KCS_IFACE_STATUS_ABORT 0x01 #define KCS_IFACE_STATUS_ILLEGAL 0x02 #define KCS_IFACE_STATUS_LENGTH_ERR 0x06 #define KCS_IFACE_STATUS_UNKNOWN_ERR 0xff /* KCS control codes */ #define KCS_CONTROL_GET_STATUS_ABORT 0x60 #define KCS_CONTROL_WRITE_START 0x61 #define KCS_CONTROL_WRITE_END 0x62 #define KCS_DATA_IN_READ 0x68 /* SMIC status flags */ #define SMIC_STATUS_BUSY 0x01 /* System set and BMC clears it */ #define SMIC_STATUS_SMS_ATN 0x04 /* BMC has a message */ #define SMIC_STATUS_EVT_ATN 0x08 /* Event has been RX */ #define SMIC_STATUS_SMI 0x10 /* asserted SMI */ #define SMIC_STATUS_TX_RDY 0x40 /* Ready to accept WRITE */ #define SMIC_STATUS_RX_RDY 0x80 /* Ready to read */ #define SMIC_STATUS_RESERVED 0x22 /* SMIC control codes */ #define SMIC_CC_SMS_GET_STATUS 0x40 #define SMIC_CC_SMS_WR_START 0x41 #define SMIC_CC_SMS_WR_NEXT 0x42 #define SMIC_CC_SMS_WR_END 0x43 #define SMIC_CC_SMS_RD_START 0x44 #define SMIC_CC_SMS_RD_NEXT 0x45 #define SMIC_CC_SMS_RD_END 0x46 /* SMIC status codes */ #define SMIC_SC_SMS_RDY 0xc0 #define SMIC_SC_SMS_WR_START 0xc1 #define SMIC_SC_SMS_WR_NEXT 0xc2 #define SMIC_SC_SMS_WR_END 0xc3 #define SMIC_SC_SMS_RD_START 0xc4 #define SMIC_SC_SMS_RD_NEXT 0xc5 #define SMIC_SC_SMS_RD_END 0xc6 #define IPMI_ADDR(netfn, lun) ((netfn) << 2 | (lun)) #define IPMI_REPLY_ADDR(addr) ((addr) + 0x4) #define IPMI_LOCK(sc) mtx_lock(&(sc)->ipmi_requests_lock) #define IPMI_UNLOCK(sc) mtx_unlock(&(sc)->ipmi_requests_lock) #define IPMI_LOCK_ASSERT(sc) mtx_assert(&(sc)->ipmi_requests_lock, MA_OWNED) #define IPMI_IO_LOCK(sc) mtx_lock(&(sc)->ipmi_io_lock) #define IPMI_IO_UNLOCK(sc) mtx_unlock(&(sc)->ipmi_io_lock) #define IPMI_IO_LOCK_ASSERT(sc) mtx_assert(&(sc)->ipmi_io_lock, MA_OWNED) #if __FreeBSD_version < 601105 #define bus_read_1(r, o) \ bus_space_read_1(rman_get_bustag(r), rman_get_bushandle(r), (o)) #define bus_write_1(r, o, v) \ bus_space_write_1(rman_get_bustag(r), rman_get_bushandle(r), (o), (v)) #endif /* I/O to a single I/O resource. */ #define INB_SINGLE(sc, x) \ bus_read_1((sc)->ipmi_io_res[0], (sc)->ipmi_io_spacing * (x)) #define OUTB_SINGLE(sc, x, value) \ bus_write_1((sc)->ipmi_io_res[0], (sc)->ipmi_io_spacing * (x), value) /* I/O with each register in its in I/O resource. */ #define INB_MULTIPLE(sc, x) \ bus_read_1((sc)->ipmi_io_res[(x)], 0) #define OUTB_MULTIPLE(sc, x, value) \ bus_write_1((sc)->ipmi_io_res[(x)], 0, value) /* * Determine I/O method based on whether or not we have more than one I/O * resource. */ #define INB(sc, x) \ ((sc)->ipmi_io_res[1] != NULL ? INB_MULTIPLE(sc, x) : INB_SINGLE(sc, x)) #define OUTB(sc, x, value) \ ((sc)->ipmi_io_res[1] != NULL ? OUTB_MULTIPLE(sc, x, value) : \ OUTB_SINGLE(sc, x, value)) #define MAX_TIMEOUT 6 * hz int ipmi_attach(device_t); int ipmi_detach(device_t); void ipmi_release_resources(device_t); /* Manage requests. */ struct ipmi_request *ipmi_alloc_request(struct ipmi_device *, long, uint8_t, uint8_t, size_t, size_t); void ipmi_complete_request(struct ipmi_softc *, struct ipmi_request *); struct ipmi_request *ipmi_dequeue_request(struct ipmi_softc *); void ipmi_free_request(struct ipmi_request *); int ipmi_polled_enqueue_request(struct ipmi_softc *, struct ipmi_request *); int ipmi_submit_driver_request(struct ipmi_softc *, struct ipmi_request *, int); /* Identify BMC interface via SMBIOS. */ int ipmi_smbios_identify(struct ipmi_get_info *); /* Match BMC PCI device listed in SMBIOS. */ const char *ipmi_pci_match(uint16_t, uint16_t); /* Interface attach routines. */ int ipmi_kcs_attach(struct ipmi_softc *); int ipmi_kcs_probe_align(struct ipmi_softc *); int ipmi_smic_attach(struct ipmi_softc *); int ipmi_ssif_attach(struct ipmi_softc *, device_t, int); #ifdef IPMB int ipmi_handle_attn(struct ipmi_softc *); #endif extern devclass_t ipmi_devclass; extern int ipmi_attached; #endif /* !__IPMIVARS_H__ */ Index: user/ngie/more-tests/sys/dev/vt/vt_core.c =================================================================== --- user/ngie/more-tests/sys/dev/vt/vt_core.c (revision 281974) +++ user/ngie/more-tests/sys/dev/vt/vt_core.c (revision 281975) @@ -1,2787 +1,2781 @@ /*- * Copyright (c) 2009, 2013 The FreeBSD Foundation * All rights reserved. * * This software was developed by Ed Schouten under sponsorship from the * FreeBSD Foundation. * * Portions of this software were developed by Oleksandr Rybalko * under sponsorship from the FreeBSD Foundation. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(__i386__) || defined(__amd64__) #include #include #endif static tc_bell_t vtterm_bell; static tc_cursor_t vtterm_cursor; static tc_putchar_t vtterm_putchar; static tc_fill_t vtterm_fill; static tc_copy_t vtterm_copy; static tc_param_t vtterm_param; static tc_done_t vtterm_done; static tc_cnprobe_t vtterm_cnprobe; static tc_cngetc_t vtterm_cngetc; static tc_cngrab_t vtterm_cngrab; static tc_cnungrab_t vtterm_cnungrab; static tc_opened_t vtterm_opened; static tc_ioctl_t vtterm_ioctl; static tc_mmap_t vtterm_mmap; const struct terminal_class vt_termclass = { .tc_bell = vtterm_bell, .tc_cursor = vtterm_cursor, .tc_putchar = vtterm_putchar, .tc_fill = vtterm_fill, .tc_copy = vtterm_copy, .tc_param = vtterm_param, .tc_done = vtterm_done, .tc_cnprobe = vtterm_cnprobe, .tc_cngetc = vtterm_cngetc, .tc_cngrab = vtterm_cngrab, .tc_cnungrab = vtterm_cnungrab, .tc_opened = vtterm_opened, .tc_ioctl = vtterm_ioctl, .tc_mmap = vtterm_mmap, }; /* * Use a constant timer of 25 Hz to redraw the screen. * * XXX: In theory we should only fire up the timer when there is really * activity. Unfortunately we cannot always start timers. We really * don't want to process kernel messages synchronously, because it * really slows down the system. */ #define VT_TIMERFREQ 25 /* Bell pitch/duration. */ #define VT_BELLDURATION ((5 * hz + 99) / 100) #define VT_BELLPITCH 800 #define VT_LOCK(vd) mtx_lock(&(vd)->vd_lock) #define VT_UNLOCK(vd) mtx_unlock(&(vd)->vd_lock) #define VT_LOCK_ASSERT(vd, what) mtx_assert(&(vd)->vd_lock, what) #define VT_UNIT(vw) ((vw)->vw_device->vd_unit * VT_MAXWINDOWS + \ (vw)->vw_number) static SYSCTL_NODE(_kern, OID_AUTO, vt, CTLFLAG_RD, 0, "vt(9) parameters"); VT_SYSCTL_INT(enable_altgr, 1, "Enable AltGr key (Do not assume R.Alt as Alt)"); VT_SYSCTL_INT(enable_bell, 1, "Enable bell"); VT_SYSCTL_INT(debug, 0, "vt(9) debug level"); VT_SYSCTL_INT(deadtimer, 15, "Time to wait busy process in VT_PROCESS mode"); VT_SYSCTL_INT(suspendswitch, 1, "Switch to VT0 before suspend"); /* Allow to disable some keyboard combinations. */ VT_SYSCTL_INT(kbd_halt, 1, "Enable halt keyboard combination. " "See kbdmap(5) to configure."); VT_SYSCTL_INT(kbd_poweroff, 1, "Enable Power Off keyboard combination. " "See kbdmap(5) to configure."); VT_SYSCTL_INT(kbd_reboot, 1, "Enable reboot keyboard combination. " "See kbdmap(5) to configure (typically Ctrl-Alt-Delete)."); VT_SYSCTL_INT(kbd_debug, 1, "Enable key combination to enter debugger. " "See kbdmap(5) to configure (typically Ctrl-Alt-Esc)."); VT_SYSCTL_INT(kbd_panic, 0, "Enable request to panic. " "See kbdmap(5) to configure."); static struct vt_device vt_consdev; static unsigned int vt_unit = 0; static MALLOC_DEFINE(M_VT, "vt", "vt device"); struct vt_device *main_vd = &vt_consdev; /* Boot logo. */ extern unsigned int vt_logo_width; extern unsigned int vt_logo_height; extern unsigned int vt_logo_depth; extern unsigned char vt_logo_image[]; /* Font. */ extern struct vt_font vt_font_default; #ifndef SC_NO_CUTPASTE extern struct vt_mouse_cursor vt_default_mouse_pointer; #endif static int signal_vt_rel(struct vt_window *); static int signal_vt_acq(struct vt_window *); static int finish_vt_rel(struct vt_window *, int, int *); static int finish_vt_acq(struct vt_window *); static int vt_window_switch(struct vt_window *); static int vt_late_window_switch(struct vt_window *); static int vt_proc_alive(struct vt_window *); static void vt_resize(struct vt_device *); static void vt_update_static(void *); #ifndef SC_NO_CUTPASTE static void vt_mouse_paste(void); #endif static void vt_suspend_handler(void *priv); static void vt_resume_handler(void *priv); SET_DECLARE(vt_drv_set, struct vt_driver); #define _VTDEFH MAX(100, PIXEL_HEIGHT(VT_FB_DEFAULT_HEIGHT)) #define _VTDEFW MAX(200, PIXEL_WIDTH(VT_FB_DEFAULT_WIDTH)) static struct terminal vt_consterm; static struct vt_window vt_conswindow; static struct vt_device vt_consdev = { .vd_driver = NULL, .vd_softc = NULL, .vd_prev_driver = NULL, .vd_prev_softc = NULL, .vd_flags = VDF_INVALID, .vd_windows = { [VT_CONSWINDOW] = &vt_conswindow, }, .vd_curwindow = &vt_conswindow, .vd_kbstate = 0, #ifndef SC_NO_CUTPASTE .vd_pastebuf = { .vpb_buf = NULL, .vpb_bufsz = 0, .vpb_len = 0 }, .vd_mcursor = &vt_default_mouse_pointer, .vd_mcursor_fg = TC_WHITE, .vd_mcursor_bg = TC_BLACK, #endif }; static term_char_t vt_constextbuf[(_VTDEFW) * (VBF_DEFAULT_HISTORY_SIZE)]; static term_char_t *vt_constextbufrows[VBF_DEFAULT_HISTORY_SIZE]; static struct vt_window vt_conswindow = { .vw_number = VT_CONSWINDOW, .vw_flags = VWF_CONSOLE, .vw_buf = { .vb_buffer = &vt_constextbuf[0], .vb_rows = &vt_constextbufrows[0], .vb_history_size = VBF_DEFAULT_HISTORY_SIZE, .vb_curroffset = 0, .vb_roffset = 0, .vb_flags = VBF_STATIC, .vb_mark_start = {.tp_row = 0, .tp_col = 0,}, .vb_mark_end = {.tp_row = 0, .tp_col = 0,}, .vb_scr_size = { .tp_row = _VTDEFH, .tp_col = _VTDEFW, }, }, .vw_device = &vt_consdev, .vw_terminal = &vt_consterm, .vw_kbdmode = K_XLATE, .vw_grabbed = 0, }; static struct terminal vt_consterm = { .tm_class = &vt_termclass, .tm_softc = &vt_conswindow, .tm_flags = TF_CONS, }; static struct consdev vt_consterm_consdev = { .cn_ops = &termcn_cnops, .cn_arg = &vt_consterm, .cn_name = "ttyv0", }; /* Add to set of consoles. */ DATA_SET(cons_set, vt_consterm_consdev); /* * Right after kmem is done to allow early drivers to use locking and allocate * memory. */ SYSINIT(vt_update_static, SI_SUB_KMEM, SI_ORDER_ANY, vt_update_static, &vt_consdev); /* Delay until all devices attached, to not waste time. */ SYSINIT(vt_early_cons, SI_SUB_INT_CONFIG_HOOKS, SI_ORDER_ANY, vt_upgrade, &vt_consdev); /* Initialize locks/mem depended members. */ static void vt_update_static(void *dummy) { if (!vty_enabled(VTY_VT)) return; if (main_vd->vd_driver != NULL) printf("VT: running with driver \"%s\".\n", main_vd->vd_driver->vd_name); else printf("VT: init without driver.\n"); mtx_init(&main_vd->vd_lock, "vtdev", NULL, MTX_DEF); cv_init(&main_vd->vd_winswitch, "vtwswt"); } static void vt_schedule_flush(struct vt_device *vd, int ms) { if (ms <= 0) /* Default to initial value. */ ms = 1000 / VT_TIMERFREQ; callout_schedule(&vd->vd_timer, hz / (1000 / ms)); } static void vt_resume_flush_timer(struct vt_device *vd, int ms) { if (!(vd->vd_flags & VDF_ASYNC) || !atomic_cmpset_int(&vd->vd_timer_armed, 0, 1)) return; vt_schedule_flush(vd, ms); } static void vt_suspend_flush_timer(struct vt_device *vd) { /* * As long as this function is called locked, callout_stop() * has the same effect like callout_drain() with regard to * preventing the callback function from executing. */ VT_LOCK_ASSERT(vd, MA_OWNED); if (!(vd->vd_flags & VDF_ASYNC) || !atomic_cmpset_int(&vd->vd_timer_armed, 1, 0)) return; callout_stop(&vd->vd_timer); } static void vt_switch_timer(void *arg) { vt_late_window_switch((struct vt_window *)arg); } static int vt_save_kbd_mode(struct vt_window *vw, keyboard_t *kbd) { int mode, ret; mode = 0; ret = kbdd_ioctl(kbd, KDGKBMODE, (caddr_t)&mode); if (ret == ENOIOCTL) ret = ENODEV; if (ret != 0) return (ret); vw->vw_kbdmode = mode; return (0); } static int vt_update_kbd_mode(struct vt_window *vw, keyboard_t *kbd) { int ret; ret = kbdd_ioctl(kbd, KDSKBMODE, (caddr_t)&vw->vw_kbdmode); if (ret == ENOIOCTL) ret = ENODEV; return (ret); } static int vt_save_kbd_state(struct vt_window *vw, keyboard_t *kbd) { int state, ret; state = 0; ret = kbdd_ioctl(kbd, KDGKBSTATE, (caddr_t)&state); if (ret == ENOIOCTL) ret = ENODEV; if (ret != 0) return (ret); vw->vw_kbdstate &= ~LOCK_MASK; vw->vw_kbdstate |= state & LOCK_MASK; return (0); } static int vt_update_kbd_state(struct vt_window *vw, keyboard_t *kbd) { int state, ret; state = vw->vw_kbdstate & LOCK_MASK; ret = kbdd_ioctl(kbd, KDSKBSTATE, (caddr_t)&state); if (ret == ENOIOCTL) ret = ENODEV; return (ret); } static int vt_save_kbd_leds(struct vt_window *vw, keyboard_t *kbd) { int leds, ret; leds = 0; ret = kbdd_ioctl(kbd, KDGETLED, (caddr_t)&leds); if (ret == ENOIOCTL) ret = ENODEV; if (ret != 0) return (ret); vw->vw_kbdstate &= ~LED_MASK; vw->vw_kbdstate |= leds & LED_MASK; return (0); } static int vt_update_kbd_leds(struct vt_window *vw, keyboard_t *kbd) { int leds, ret; leds = vw->vw_kbdstate & LED_MASK; ret = kbdd_ioctl(kbd, KDSETLED, (caddr_t)&leds); if (ret == ENOIOCTL) ret = ENODEV; return (ret); } static int vt_window_preswitch(struct vt_window *vw, struct vt_window *curvw) { DPRINTF(40, "%s\n", __func__); curvw->vw_switch_to = vw; /* Set timer to allow switch in case when process hang. */ callout_reset(&vw->vw_proc_dead_timer, hz * vt_deadtimer, vt_switch_timer, (void *)vw); /* Notify process about vt switch attempt. */ DPRINTF(30, "%s: Notify process.\n", __func__); signal_vt_rel(curvw); return (0); } static int vt_window_postswitch(struct vt_window *vw) { signal_vt_acq(vw); return (0); } /* vt_late_window_switch will done VT switching for regular case. */ static int vt_late_window_switch(struct vt_window *vw) { int ret; callout_stop(&vw->vw_proc_dead_timer); ret = vt_window_switch(vw); if (ret) return (ret); /* Notify owner process about terminal availability. */ if (vw->vw_smode.mode == VT_PROCESS) { ret = vt_window_postswitch(vw); } return (ret); } /* Switch window. */ static int vt_proc_window_switch(struct vt_window *vw) { struct vt_window *curvw; struct vt_device *vd; int ret; vd = vw->vw_device; curvw = vd->vd_curwindow; if (curvw->vw_flags & VWF_VTYLOCK) return (EBUSY); /* Ask current process permission to switch away. */ if (curvw->vw_smode.mode == VT_PROCESS) { DPRINTF(30, "%s: VT_PROCESS ", __func__); if (vt_proc_alive(curvw) == FALSE) { DPRINTF(30, "Dead. Cleaning."); /* Dead */ } else { DPRINTF(30, "%s: Signaling process.\n", __func__); /* Alive, try to ask him. */ ret = vt_window_preswitch(vw, curvw); /* Wait for process answer or timeout. */ return (ret); } DPRINTF(30, "\n"); } ret = vt_late_window_switch(vw); return (ret); } /* Switch window ignoring process locking. */ static int vt_window_switch(struct vt_window *vw) { struct vt_device *vd = vw->vw_device; struct vt_window *curvw = vd->vd_curwindow; keyboard_t *kbd; VT_LOCK(vd); if (curvw == vw) { /* Nothing to do. */ VT_UNLOCK(vd); return (0); } if (!(vw->vw_flags & (VWF_OPENED|VWF_CONSOLE))) { VT_UNLOCK(vd); return (EINVAL); } vt_suspend_flush_timer(vd); vd->vd_curwindow = vw; vd->vd_flags |= VDF_INVALID; cv_broadcast(&vd->vd_winswitch); VT_UNLOCK(vd); if (vd->vd_driver->vd_postswitch) vd->vd_driver->vd_postswitch(vd); vt_resume_flush_timer(vd, 0); /* Restore per-window keyboard mode. */ mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) { if (curvw->vw_kbdmode == K_XLATE) vt_save_kbd_state(curvw, kbd); vt_update_kbd_mode(vw, kbd); vt_update_kbd_state(vw, kbd); } mtx_unlock(&Giant); DPRINTF(10, "%s(ttyv%d) done\n", __func__, vw->vw_number); return (0); } static inline void vt_termsize(struct vt_device *vd, struct vt_font *vf, term_pos_t *size) { size->tp_row = vd->vd_height; size->tp_col = vd->vd_width; if (vf != NULL) { size->tp_row /= vf->vf_height; size->tp_col /= vf->vf_width; } } static inline void vt_winsize(struct vt_device *vd, struct vt_font *vf, struct winsize *size) { size->ws_row = size->ws_ypixel = vd->vd_height; size->ws_col = size->ws_xpixel = vd->vd_width; if (vf != NULL) { size->ws_row /= vf->vf_height; size->ws_col /= vf->vf_width; } } static inline void vt_compute_drawable_area(struct vt_window *vw) { struct vt_device *vd; struct vt_font *vf; vd = vw->vw_device; if (vw->vw_font == NULL) { vw->vw_draw_area.tr_begin.tp_col = 0; vw->vw_draw_area.tr_begin.tp_row = 0; vw->vw_draw_area.tr_end.tp_col = vd->vd_width; vw->vw_draw_area.tr_end.tp_row = vd->vd_height; return; } vf = vw->vw_font; /* * Compute the drawable area, so that the text is centered on * the screen. */ vw->vw_draw_area.tr_begin.tp_col = (vd->vd_width % vf->vf_width) / 2; vw->vw_draw_area.tr_begin.tp_row = (vd->vd_height % vf->vf_height) / 2; vw->vw_draw_area.tr_end.tp_col = vw->vw_draw_area.tr_begin.tp_col + vd->vd_width / vf->vf_width * vf->vf_width; vw->vw_draw_area.tr_end.tp_row = vw->vw_draw_area.tr_begin.tp_row + vd->vd_height / vf->vf_height * vf->vf_height; } static void vt_scroll(struct vt_window *vw, int offset, int whence) { int diff; term_pos_t size; if ((vw->vw_flags & VWF_SCROLL) == 0) return; vt_termsize(vw->vw_device, vw->vw_font, &size); diff = vthistory_seek(&vw->vw_buf, offset, whence); if (diff) vw->vw_device->vd_flags |= VDF_INVALID; vt_resume_flush_timer(vw->vw_device, 0); } static int vt_machine_kbdevent(int c) { switch (c) { case SPCLKEY | DBG: /* kbdmap(5) keyword `debug`. */ if (vt_kbd_debug) kdb_enter(KDB_WHY_BREAK, "manual escape to debugger"); return (1); case SPCLKEY | HALT: /* kbdmap(5) keyword `halt`. */ if (vt_kbd_halt) shutdown_nice(RB_HALT); return (1); case SPCLKEY | PASTE: /* kbdmap(5) keyword `paste`. */ #ifndef SC_NO_CUTPASTE /* Insert text from cut-paste buffer. */ vt_mouse_paste(); #endif break; case SPCLKEY | PDWN: /* kbdmap(5) keyword `pdwn`. */ if (vt_kbd_poweroff) shutdown_nice(RB_HALT|RB_POWEROFF); return (1); case SPCLKEY | PNC: /* kbdmap(5) keyword `panic`. */ /* * Request to immediate panic if sysctl * kern.vt.enable_panic_key allow it. */ if (vt_kbd_panic) panic("Forced by the panic key"); return (1); case SPCLKEY | RBT: /* kbdmap(5) keyword `boot`. */ if (vt_kbd_reboot) shutdown_nice(RB_AUTOBOOT); return (1); case SPCLKEY | SPSC: /* kbdmap(5) keyword `spsc`. */ /* Force activatation/deactivation of the screen saver. */ /* TODO */ return (1); case SPCLKEY | STBY: /* XXX Not present in kbdcontrol parser. */ /* Put machine into Stand-By mode. */ power_pm_suspend(POWER_SLEEP_STATE_STANDBY); return (1); case SPCLKEY | SUSP: /* kbdmap(5) keyword `susp`. */ /* Suspend machine. */ power_pm_suspend(POWER_SLEEP_STATE_SUSPEND); return (1); }; return (0); } static void vt_scrollmode_kbdevent(struct vt_window *vw, int c, int console) { struct vt_device *vd; term_pos_t size; vd = vw->vw_device; /* Only special keys handled in ScrollLock mode */ if ((c & SPCLKEY) == 0) return; c &= ~SPCLKEY; if (console == 0) { if (c >= F_SCR && c <= MIN(L_SCR, F_SCR + VT_MAXWINDOWS - 1)) { vw = vd->vd_windows[c - F_SCR]; if (vw != NULL) vt_proc_window_switch(vw); return; } VT_LOCK(vd); } switch (c) { case SLK: { /* Turn scrolling off. */ vt_scroll(vw, 0, VHS_END); VTBUF_SLCK_DISABLE(&vw->vw_buf); vw->vw_flags &= ~VWF_SCROLL; break; } case FKEY | F(49): /* Home key. */ vt_scroll(vw, 0, VHS_SET); break; case FKEY | F(50): /* Arrow up. */ vt_scroll(vw, -1, VHS_CUR); break; case FKEY | F(51): /* Page up. */ vt_termsize(vd, vw->vw_font, &size); vt_scroll(vw, -size.tp_row, VHS_CUR); break; case FKEY | F(57): /* End key. */ vt_scroll(vw, 0, VHS_END); break; case FKEY | F(58): /* Arrow down. */ vt_scroll(vw, 1, VHS_CUR); break; case FKEY | F(59): /* Page down. */ vt_termsize(vd, vw->vw_font, &size); vt_scroll(vw, size.tp_row, VHS_CUR); break; } if (console == 0) VT_UNLOCK(vd); } static int vt_processkey(keyboard_t *kbd, struct vt_device *vd, int c) { struct vt_window *vw = vd->vd_curwindow; #if VT_ALT_TO_ESC_HACK if (c & RELKEY) { switch (c & ~RELKEY) { case (SPCLKEY | RALT): if (vt_enable_altgr != 0) break; case (SPCLKEY | LALT): vd->vd_kbstate &= ~ALKED; } /* Other keys ignored for RELKEY event. */ return (0); } else { switch (c & ~RELKEY) { case (SPCLKEY | RALT): if (vt_enable_altgr != 0) break; case (SPCLKEY | LALT): vd->vd_kbstate |= ALKED; } } #else if (c & RELKEY) /* Other keys ignored for RELKEY event. */ return (0); #endif if (vt_machine_kbdevent(c)) return (0); if (vw->vw_flags & VWF_SCROLL) { vt_scrollmode_kbdevent(vw, c, 0/* Not a console */); /* Scroll mode keys handled, nothing to do more. */ return (0); } if (c & SPCLKEY) { c &= ~SPCLKEY; if (c >= F_SCR && c <= MIN(L_SCR, F_SCR + VT_MAXWINDOWS - 1)) { vw = vd->vd_windows[c - F_SCR]; if (vw != NULL) vt_proc_window_switch(vw); return (0); } switch (c) { case NEXT: /* Switch to next VT. */ c = (vw->vw_number + 1) % VT_MAXWINDOWS; vw = vd->vd_windows[c]; if (vw != NULL) vt_proc_window_switch(vw); return (0); case PREV: /* Switch to previous VT. */ c = (vw->vw_number - 1) % VT_MAXWINDOWS; vw = vd->vd_windows[c]; if (vw != NULL) vt_proc_window_switch(vw); return (0); case SLK: { vt_save_kbd_state(vw, kbd); VT_LOCK(vd); if (vw->vw_kbdstate & SLKED) { /* Turn scrolling on. */ vw->vw_flags |= VWF_SCROLL; VTBUF_SLCK_ENABLE(&vw->vw_buf); } else { /* Turn scrolling off. */ vw->vw_flags &= ~VWF_SCROLL; VTBUF_SLCK_DISABLE(&vw->vw_buf); vt_scroll(vw, 0, VHS_END); } VT_UNLOCK(vd); break; } case FKEY | F(1): case FKEY | F(2): case FKEY | F(3): case FKEY | F(4): case FKEY | F(5): case FKEY | F(6): case FKEY | F(7): case FKEY | F(8): case FKEY | F(9): case FKEY | F(10): case FKEY | F(11): case FKEY | F(12): /* F1 through F12 keys. */ terminal_input_special(vw->vw_terminal, TKEY_F1 + c - (FKEY | F(1))); break; case FKEY | F(49): /* Home key. */ terminal_input_special(vw->vw_terminal, TKEY_HOME); break; case FKEY | F(50): /* Arrow up. */ terminal_input_special(vw->vw_terminal, TKEY_UP); break; case FKEY | F(51): /* Page up. */ terminal_input_special(vw->vw_terminal, TKEY_PAGE_UP); break; case FKEY | F(53): /* Arrow left. */ terminal_input_special(vw->vw_terminal, TKEY_LEFT); break; case FKEY | F(55): /* Arrow right. */ terminal_input_special(vw->vw_terminal, TKEY_RIGHT); break; case FKEY | F(57): /* End key. */ terminal_input_special(vw->vw_terminal, TKEY_END); break; case FKEY | F(58): /* Arrow down. */ terminal_input_special(vw->vw_terminal, TKEY_DOWN); break; case FKEY | F(59): /* Page down. */ terminal_input_special(vw->vw_terminal, TKEY_PAGE_DOWN); break; case FKEY | F(60): /* Insert key. */ terminal_input_special(vw->vw_terminal, TKEY_INSERT); break; case FKEY | F(61): /* Delete key. */ terminal_input_special(vw->vw_terminal, TKEY_DELETE); break; } } else if (KEYFLAGS(c) == 0) { /* Don't do UTF-8 conversion when doing raw mode. */ if (vw->vw_kbdmode == K_XLATE) { #if VT_ALT_TO_ESC_HACK if (vd->vd_kbstate & ALKED) { /* * Prepend ESC sequence if one of ALT keys down. */ terminal_input_char(vw->vw_terminal, 0x1b); } #endif #if defined(KDB) kdb_alt_break(c, &vd->vd_altbrk); #endif terminal_input_char(vw->vw_terminal, KEYCHAR(c)); } else terminal_input_raw(vw->vw_terminal, c); } return (0); } static int vt_kbdevent(keyboard_t *kbd, int event, void *arg) { struct vt_device *vd = arg; int c; switch (event) { case KBDIO_KEYINPUT: break; case KBDIO_UNLOADING: mtx_lock(&Giant); vd->vd_keyboard = -1; kbd_release(kbd, (void *)vd); mtx_unlock(&Giant); return (0); default: return (EINVAL); } while ((c = kbdd_read_char(kbd, 0)) != NOKEY) vt_processkey(kbd, vd, c); return (0); } static int vt_allocate_keyboard(struct vt_device *vd) { int idx0, idx; keyboard_t *k0, *k; keyboard_info_t ki; idx0 = kbd_allocate("kbdmux", -1, vd, vt_kbdevent, vd); if (idx0 >= 0) { DPRINTF(20, "%s: kbdmux allocated, idx = %d\n", __func__, idx0); k0 = kbd_get_keyboard(idx0); for (idx = kbd_find_keyboard2("*", -1, 0); idx != -1; idx = kbd_find_keyboard2("*", -1, idx + 1)) { k = kbd_get_keyboard(idx); if (idx == idx0 || KBD_IS_BUSY(k)) continue; bzero(&ki, sizeof(ki)); strncpy(ki.kb_name, k->kb_name, sizeof(ki.kb_name)); ki.kb_name[sizeof(ki.kb_name) - 1] = '\0'; ki.kb_unit = k->kb_unit; kbdd_ioctl(k0, KBADDKBD, (caddr_t) &ki); } } else { DPRINTF(20, "%s: no kbdmux allocated\n", __func__); idx0 = kbd_allocate("*", -1, vd, vt_kbdevent, vd); if (idx0 < 0) { DPRINTF(10, "%s: No keyboard found.\n", __func__); return (-1); } } vd->vd_keyboard = idx0; DPRINTF(20, "%s: vd_keyboard = %d\n", __func__, vd->vd_keyboard); return (idx0); } static void vtterm_bell(struct terminal *tm) { struct vt_window *vw = tm->tm_softc; struct vt_device *vd = vw->vw_device; if (!vt_enable_bell) return; if (vd->vd_flags & VDF_QUIET_BELL) return; sysbeep(1193182 / VT_BELLPITCH, VT_BELLDURATION); } static void vtterm_beep(struct terminal *tm, u_int param) { u_int freq, period; if (!vt_enable_bell) return; if ((param == 0) || ((param & 0xffff) == 0)) { vtterm_bell(tm); return; } period = ((param >> 16) & 0xffff) * hz / 1000; freq = 1193182 / (param & 0xffff); sysbeep(freq, period); } static void vtterm_cursor(struct terminal *tm, const term_pos_t *p) { struct vt_window *vw = tm->tm_softc; vtbuf_cursor_position(&vw->vw_buf, p); vt_resume_flush_timer(vw->vw_device, 0); } static void vtterm_putchar(struct terminal *tm, const term_pos_t *p, term_char_t c) { struct vt_window *vw = tm->tm_softc; vtbuf_putchar(&vw->vw_buf, p, c); vt_resume_flush_timer(vw->vw_device, 0); } static void vtterm_fill(struct terminal *tm, const term_rect_t *r, term_char_t c) { struct vt_window *vw = tm->tm_softc; vtbuf_fill_locked(&vw->vw_buf, r, c); vt_resume_flush_timer(vw->vw_device, 0); } static void vtterm_copy(struct terminal *tm, const term_rect_t *r, const term_pos_t *p) { struct vt_window *vw = tm->tm_softc; vtbuf_copy(&vw->vw_buf, r, p); vt_resume_flush_timer(vw->vw_device, 0); } static void vtterm_param(struct terminal *tm, int cmd, unsigned int arg) { struct vt_window *vw = tm->tm_softc; switch (cmd) { case TP_SHOWCURSOR: vtbuf_cursor_visibility(&vw->vw_buf, arg); vt_resume_flush_timer(vw->vw_device, 0); break; case TP_MOUSE: vw->vw_mouse_level = arg; break; } } void vt_determine_colors(term_char_t c, int cursor, term_color_t *fg, term_color_t *bg) { term_color_t tmp; int invert; invert = 0; *fg = TCHAR_FGCOLOR(c); if (TCHAR_FORMAT(c) & TF_BOLD) *fg = TCOLOR_LIGHT(*fg); *bg = TCHAR_BGCOLOR(c); if (TCHAR_FORMAT(c) & TF_REVERSE) invert ^= 1; if (cursor) invert ^= 1; if (invert) { tmp = *fg; *fg = *bg; *bg = tmp; } } #ifndef SC_NO_CUTPASTE int vt_is_cursor_in_area(const struct vt_device *vd, const term_rect_t *area) { - unsigned int mx, my, x1, y1, x2, y2; + unsigned int mx, my; /* * We use the cursor position saved during the current refresh, * in case the cursor moved since. */ mx = vd->vd_mx_drawn + vd->vd_curwindow->vw_draw_area.tr_begin.tp_col; my = vd->vd_my_drawn + vd->vd_curwindow->vw_draw_area.tr_begin.tp_row; - x1 = area->tr_begin.tp_col; - y1 = area->tr_begin.tp_row; - x2 = area->tr_end.tp_col; - y2 = area->tr_end.tp_row; - - if (((mx >= x1 && x2 - 1 >= mx) || - (mx < x1 && mx + vd->vd_mcursor->width >= x1)) && - ((my >= y1 && y2 - 1 >= my) || - (my < y1 && my + vd->vd_mcursor->height >= y1))) - return (1); - - return (0); + if (mx >= area->tr_end.tp_col || + mx + vd->vd_mcursor->width <= area->tr_begin.tp_col || + my >= area->tr_end.tp_row || + my + vd->vd_mcursor->height <= area->tr_begin.tp_row) + return (0); + return (1); } static void vt_mark_mouse_position_as_dirty(struct vt_device *vd) { term_rect_t area; struct vt_window *vw; struct vt_font *vf; int x, y; vw = vd->vd_curwindow; vf = vw->vw_font; x = vd->vd_mx_drawn; y = vd->vd_my_drawn; if (vf != NULL) { area.tr_begin.tp_col = x / vf->vf_width; area.tr_begin.tp_row = y / vf->vf_height; area.tr_end.tp_col = ((x + vd->vd_mcursor->width) / vf->vf_width) + 1; area.tr_end.tp_row = ((y + vd->vd_mcursor->height) / vf->vf_height) + 1; } else { /* * No font loaded (ie. vt_vga operating in textmode). * * FIXME: This fake area needs to be revisited once the * mouse cursor is supported in vt_vga's textmode. */ area.tr_begin.tp_col = x; area.tr_begin.tp_row = y; area.tr_end.tp_col = x + 2; area.tr_end.tp_row = y + 2; } vtbuf_dirty(&vw->vw_buf, &area); } #endif static int vt_flush(struct vt_device *vd) { struct vt_window *vw; struct vt_font *vf; term_rect_t tarea; term_pos_t size; #ifndef SC_NO_CUTPASTE int cursor_was_shown, cursor_moved; #endif vw = vd->vd_curwindow; if (vw == NULL) return (0); if (vd->vd_flags & VDF_SPLASH || vw->vw_flags & VWF_BUSY) return (0); vf = vw->vw_font; if (((vd->vd_flags & VDF_TEXTMODE) == 0) && (vf == NULL)) return (0); #ifndef SC_NO_CUTPASTE cursor_was_shown = vd->vd_mshown; cursor_moved = (vd->vd_mx != vd->vd_mx_drawn || vd->vd_my != vd->vd_my_drawn); /* Check if the cursor should be displayed or not. */ if ((vd->vd_flags & VDF_MOUSECURSOR) && /* Mouse support enabled. */ !(vw->vw_flags & VWF_MOUSE_HIDE) && /* Cursor displayed. */ !kdb_active && panicstr == NULL) { /* DDB inactive. */ vd->vd_mshown = 1; } else { vd->vd_mshown = 0; } /* * If the cursor changed display state or moved, we must mark * the old position as dirty, so that it's erased. */ if (cursor_was_shown != vd->vd_mshown || (vd->vd_mshown && cursor_moved)) vt_mark_mouse_position_as_dirty(vd); /* * Save position of the mouse cursor. It's used by backends to * know where to draw the cursor and during the next refresh to * erase the previous position. */ vd->vd_mx_drawn = vd->vd_mx; vd->vd_my_drawn = vd->vd_my; /* * If the cursor is displayed and has moved since last refresh, * mark the new position as dirty. */ if (vd->vd_mshown && cursor_moved) vt_mark_mouse_position_as_dirty(vd); #endif vtbuf_undirty(&vw->vw_buf, &tarea); vt_termsize(vd, vf, &size); /* Force a full redraw when the screen contents are invalid. */ if (vd->vd_flags & VDF_INVALID) { tarea.tr_begin.tp_row = tarea.tr_begin.tp_col = 0; tarea.tr_end = size; vd->vd_flags &= ~VDF_INVALID; } if (tarea.tr_begin.tp_col < tarea.tr_end.tp_col) { vd->vd_driver->vd_bitblt_text(vd, vw, &tarea); return (1); } return (0); } static void vt_timer(void *arg) { struct vt_device *vd; int changed; vd = arg; /* Update screen if required. */ changed = vt_flush(vd); /* Schedule for next update. */ if (changed) vt_schedule_flush(vd, 0); else vd->vd_timer_armed = 0; } static void vtterm_done(struct terminal *tm) { struct vt_window *vw = tm->tm_softc; struct vt_device *vd = vw->vw_device; if (kdb_active || panicstr != NULL) { /* Switch to the debugger. */ if (vd->vd_curwindow != vw) { vd->vd_curwindow = vw; vd->vd_flags |= VDF_INVALID; if (vd->vd_driver->vd_postswitch) vd->vd_driver->vd_postswitch(vd); } vd->vd_flags &= ~VDF_SPLASH; vt_flush(vd); } else if (!(vd->vd_flags & VDF_ASYNC)) { vt_flush(vd); } } #ifdef DEV_SPLASH static void vtterm_splash(struct vt_device *vd) { vt_axis_t top, left; /* Display a nice boot splash. */ if (!(vd->vd_flags & VDF_TEXTMODE) && (boothowto & RB_MUTE)) { top = (vd->vd_height - vt_logo_height) / 2; left = (vd->vd_width - vt_logo_width) / 2; switch (vt_logo_depth) { case 1: /* XXX: Unhardcode colors! */ vd->vd_driver->vd_bitblt_bmp(vd, vd->vd_curwindow, vt_logo_image, NULL, vt_logo_width, vt_logo_height, left, top, TC_WHITE, TC_BLACK); } vd->vd_flags |= VDF_SPLASH; } } #endif static void vtterm_cnprobe(struct terminal *tm, struct consdev *cp) { struct vt_driver *vtd, **vtdlist, *vtdbest = NULL; struct vt_window *vw = tm->tm_softc; struct vt_device *vd = vw->vw_device; struct winsize wsz; term_attr_t attr; term_char_t c; if (!vty_enabled(VTY_VT)) return; if (vd->vd_flags & VDF_INITIALIZED) /* Initialization already done. */ return; SET_FOREACH(vtdlist, vt_drv_set) { vtd = *vtdlist; if (vtd->vd_probe == NULL) continue; if (vtd->vd_probe(vd) == CN_DEAD) continue; if ((vtdbest == NULL) || (vtd->vd_priority > vtdbest->vd_priority)) vtdbest = vtd; } if (vtdbest == NULL) { cp->cn_pri = CN_DEAD; vd->vd_flags |= VDF_DEAD; } else { vd->vd_driver = vtdbest; cp->cn_pri = vd->vd_driver->vd_init(vd); } /* Check if driver's vt_init return CN_DEAD. */ if (cp->cn_pri == CN_DEAD) { vd->vd_flags |= VDF_DEAD; } /* Initialize any early-boot keyboard drivers */ kbd_configure(KB_CONF_PROBE_ONLY); vd->vd_unit = atomic_fetchadd_int(&vt_unit, 1); vd->vd_windows[VT_CONSWINDOW] = vw; sprintf(cp->cn_name, "ttyv%r", VT_UNIT(vw)); /* Attach default font if not in TEXTMODE. */ if ((vd->vd_flags & VDF_TEXTMODE) == 0) { vw->vw_font = vtfont_ref(&vt_font_default); vt_compute_drawable_area(vw); } /* * The original screen size was faked (_VTDEFW x _VTDEFH). Now * that we have the real viewable size, fix it in the static * buffer. */ if (vd->vd_width != 0 && vd->vd_height != 0) vt_termsize(vd, vw->vw_font, &vw->vw_buf.vb_scr_size); vtbuf_init_early(&vw->vw_buf); vt_winsize(vd, vw->vw_font, &wsz); c = (boothowto & RB_MUTE) == 0 ? TERMINAL_KERN_ATTR : TERMINAL_NORM_ATTR; attr.ta_format = TCHAR_FORMAT(c); attr.ta_fgcolor = TCHAR_FGCOLOR(c); attr.ta_bgcolor = TCHAR_BGCOLOR(c); terminal_set_winsize_blank(tm, &wsz, 1, &attr); if (vtdbest != NULL) { #ifdef DEV_SPLASH vtterm_splash(vd); #endif vd->vd_flags |= VDF_INITIALIZED; } } static int vtterm_cngetc(struct terminal *tm) { struct vt_window *vw = tm->tm_softc; struct vt_device *vd = vw->vw_device; keyboard_t *kbd; u_int c; if (vw->vw_kbdsq && *vw->vw_kbdsq) return (*vw->vw_kbdsq++); /* Make sure the splash screen is not there. */ if (vd->vd_flags & VDF_SPLASH) { /* Remove splash */ vd->vd_flags &= ~VDF_SPLASH; /* Mark screen as invalid to force update */ vd->vd_flags |= VDF_INVALID; vt_flush(vd); } /* Stripped down keyboard handler. */ kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd == NULL) return (-1); /* Force keyboard input mode to K_XLATE */ vw->vw_kbdmode = K_XLATE; vt_update_kbd_mode(vw, kbd); /* Switch the keyboard to polling to make it work here. */ kbdd_poll(kbd, TRUE); c = kbdd_read_char(kbd, 0); kbdd_poll(kbd, FALSE); if (c & RELKEY) return (-1); if (vw->vw_flags & VWF_SCROLL) { vt_scrollmode_kbdevent(vw, c, 1/* Console mode */); vt_flush(vd); return (-1); } /* Stripped down handling of vt_kbdevent(), without locking, etc. */ if (c & SPCLKEY) { switch (c) { case SPCLKEY | SLK: vt_save_kbd_state(vw, kbd); if (vw->vw_kbdstate & SLKED) { /* Turn scrolling on. */ vw->vw_flags |= VWF_SCROLL; VTBUF_SLCK_ENABLE(&vw->vw_buf); } else { /* Turn scrolling off. */ vt_scroll(vw, 0, VHS_END); vw->vw_flags &= ~VWF_SCROLL; VTBUF_SLCK_DISABLE(&vw->vw_buf); } break; /* XXX: KDB can handle history. */ case SPCLKEY | FKEY | F(50): /* Arrow up. */ vw->vw_kbdsq = "\x1b[A"; break; case SPCLKEY | FKEY | F(58): /* Arrow down. */ vw->vw_kbdsq = "\x1b[B"; break; case SPCLKEY | FKEY | F(55): /* Arrow right. */ vw->vw_kbdsq = "\x1b[C"; break; case SPCLKEY | FKEY | F(53): /* Arrow left. */ vw->vw_kbdsq = "\x1b[D"; break; } /* Force refresh to make scrollback work. */ vt_flush(vd); } else if (KEYFLAGS(c) == 0) { return (KEYCHAR(c)); } if (vw->vw_kbdsq && *vw->vw_kbdsq) return (*vw->vw_kbdsq++); return (-1); } static void vtterm_cngrab(struct terminal *tm) { struct vt_device *vd; struct vt_window *vw; keyboard_t *kbd; vw = tm->tm_softc; vd = vw->vw_device; if (!cold) vt_window_switch(vw); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd == NULL) return; if (vw->vw_grabbed++ > 0) return; /* * Make sure the keyboard is accessible even when the kbd device * driver is disabled. */ kbdd_enable(kbd); /* We shall always use the keyboard in the XLATE mode here. */ vw->vw_prev_kbdmode = vw->vw_kbdmode; vw->vw_kbdmode = K_XLATE; vt_update_kbd_mode(vw, kbd); kbdd_poll(kbd, TRUE); } static void vtterm_cnungrab(struct terminal *tm) { struct vt_device *vd; struct vt_window *vw; keyboard_t *kbd; vw = tm->tm_softc; vd = vw->vw_device; kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd == NULL) return; if (--vw->vw_grabbed > 0) return; kbdd_poll(kbd, FALSE); vw->vw_kbdmode = vw->vw_prev_kbdmode; vt_update_kbd_mode(vw, kbd); kbdd_disable(kbd); } static void vtterm_opened(struct terminal *tm, int opened) { struct vt_window *vw = tm->tm_softc; struct vt_device *vd = vw->vw_device; VT_LOCK(vd); vd->vd_flags &= ~VDF_SPLASH; if (opened) vw->vw_flags |= VWF_OPENED; else { vw->vw_flags &= ~VWF_OPENED; /* TODO: finish ACQ/REL */ } VT_UNLOCK(vd); } static int vt_set_border(struct vt_window *vw, term_color_t c) { struct vt_device *vd = vw->vw_device; if (vd->vd_driver->vd_drawrect == NULL) return (ENOTSUP); /* Top bar. */ if (vw->vw_draw_area.tr_begin.tp_row > 0) vd->vd_driver->vd_drawrect(vd, 0, 0, vd->vd_width - 1, vw->vw_draw_area.tr_begin.tp_row - 1, 1, c); /* Left bar. */ if (vw->vw_draw_area.tr_begin.tp_col > 0) vd->vd_driver->vd_drawrect(vd, 0, 0, vw->vw_draw_area.tr_begin.tp_col - 1, vd->vd_height - 1, 1, c); /* Right bar. */ if (vw->vw_draw_area.tr_end.tp_col < vd->vd_width) vd->vd_driver->vd_drawrect(vd, vw->vw_draw_area.tr_end.tp_col - 1, 0, vd->vd_width - 1, vd->vd_height - 1, 1, c); /* Bottom bar. */ if (vw->vw_draw_area.tr_end.tp_row < vd->vd_height) vd->vd_driver->vd_drawrect(vd, 0, vw->vw_draw_area.tr_end.tp_row - 1, vd->vd_width - 1, vd->vd_height - 1, 1, c); return (0); } static int vt_change_font(struct vt_window *vw, struct vt_font *vf) { struct vt_device *vd = vw->vw_device; struct terminal *tm = vw->vw_terminal; term_pos_t size; struct winsize wsz; /* * Changing fonts. * * Changing fonts is a little tricky. We must prevent * simultaneous access to the device, so we must stop * the display timer and the terminal from accessing. * We need to switch fonts and grow our screen buffer. * * XXX: Right now the code uses terminal_mute() to * prevent data from reaching the console driver while * resizing the screen buffer. This isn't elegant... */ VT_LOCK(vd); if (vw->vw_flags & VWF_BUSY) { /* Another process is changing the font. */ VT_UNLOCK(vd); return (EBUSY); } vw->vw_flags |= VWF_BUSY; VT_UNLOCK(vd); vt_termsize(vd, vf, &size); vt_winsize(vd, vf, &wsz); /* Grow the screen buffer and terminal. */ terminal_mute(tm, 1); vtbuf_grow(&vw->vw_buf, &size, vw->vw_buf.vb_history_size); terminal_set_winsize_blank(tm, &wsz, 0, NULL); terminal_set_cursor(tm, &vw->vw_buf.vb_cursor); terminal_mute(tm, 0); /* Actually apply the font to the current window. */ VT_LOCK(vd); if (vw->vw_font != vf && vw->vw_font != NULL && vf != NULL) { /* * In case vt_change_font called to update size we don't need * to update font link. */ vtfont_unref(vw->vw_font); vw->vw_font = vtfont_ref(vf); } /* * Compute the drawable area and move the mouse cursor inside * it, in case the new area is smaller than the previous one. */ vt_compute_drawable_area(vw); vd->vd_mx = min(vd->vd_mx, vw->vw_draw_area.tr_end.tp_col - vw->vw_draw_area.tr_begin.tp_col - 1); vd->vd_my = min(vd->vd_my, vw->vw_draw_area.tr_end.tp_row - vw->vw_draw_area.tr_begin.tp_row - 1); /* Force a full redraw the next timer tick. */ if (vd->vd_curwindow == vw) { vt_set_border(vw, TC_BLACK); vd->vd_flags |= VDF_INVALID; vt_resume_flush_timer(vw->vw_device, 0); } vw->vw_flags &= ~VWF_BUSY; VT_UNLOCK(vd); return (0); } static int vt_proc_alive(struct vt_window *vw) { struct proc *p; if (vw->vw_smode.mode != VT_PROCESS) return (FALSE); if (vw->vw_proc) { if ((p = pfind(vw->vw_pid)) != NULL) PROC_UNLOCK(p); if (vw->vw_proc == p) return (TRUE); vw->vw_proc = NULL; vw->vw_smode.mode = VT_AUTO; DPRINTF(1, "vt controlling process %d died\n", vw->vw_pid); vw->vw_pid = 0; } return (FALSE); } static int signal_vt_rel(struct vt_window *vw) { if (vw->vw_smode.mode != VT_PROCESS) return (FALSE); if (vw->vw_proc == NULL || vt_proc_alive(vw) == FALSE) { vw->vw_proc = NULL; vw->vw_pid = 0; return (TRUE); } vw->vw_flags |= VWF_SWWAIT_REL; PROC_LOCK(vw->vw_proc); kern_psignal(vw->vw_proc, vw->vw_smode.relsig); PROC_UNLOCK(vw->vw_proc); DPRINTF(1, "sending relsig to %d\n", vw->vw_pid); return (TRUE); } static int signal_vt_acq(struct vt_window *vw) { if (vw->vw_smode.mode != VT_PROCESS) return (FALSE); if (vw == vw->vw_device->vd_windows[VT_CONSWINDOW]) cnavailable(vw->vw_terminal->consdev, FALSE); if (vw->vw_proc == NULL || vt_proc_alive(vw) == FALSE) { vw->vw_proc = NULL; vw->vw_pid = 0; return (TRUE); } vw->vw_flags |= VWF_SWWAIT_ACQ; PROC_LOCK(vw->vw_proc); kern_psignal(vw->vw_proc, vw->vw_smode.acqsig); PROC_UNLOCK(vw->vw_proc); DPRINTF(1, "sending acqsig to %d\n", vw->vw_pid); return (TRUE); } static int finish_vt_rel(struct vt_window *vw, int release, int *s) { if (vw->vw_flags & VWF_SWWAIT_REL) { vw->vw_flags &= ~VWF_SWWAIT_REL; if (release) { callout_drain(&vw->vw_proc_dead_timer); vt_late_window_switch(vw->vw_switch_to); } return (0); } return (EINVAL); } static int finish_vt_acq(struct vt_window *vw) { if (vw->vw_flags & VWF_SWWAIT_ACQ) { vw->vw_flags &= ~VWF_SWWAIT_ACQ; return (0); } return (EINVAL); } #ifndef SC_NO_CUTPASTE static void vt_mouse_terminput_button(struct vt_device *vd, int button) { struct vt_window *vw; struct vt_font *vf; char mouseb[6] = "\x1B[M"; int i, x, y; vw = vd->vd_curwindow; vf = vw->vw_font; /* Translate to char position. */ x = vd->vd_mx / vf->vf_width; y = vd->vd_my / vf->vf_height; /* Avoid overflow. */ x = MIN(x, 255 - '!'); y = MIN(y, 255 - '!'); mouseb[3] = ' ' + button; mouseb[4] = '!' + x; mouseb[5] = '!' + y; for (i = 0; i < sizeof(mouseb); i++) terminal_input_char(vw->vw_terminal, mouseb[i]); } static void vt_mouse_terminput(struct vt_device *vd, int type, int x, int y, int event, int cnt) { switch (type) { case MOUSE_BUTTON_EVENT: if (cnt > 0) { /* Mouse button pressed. */ if (event & MOUSE_BUTTON1DOWN) vt_mouse_terminput_button(vd, 0); if (event & MOUSE_BUTTON2DOWN) vt_mouse_terminput_button(vd, 1); if (event & MOUSE_BUTTON3DOWN) vt_mouse_terminput_button(vd, 2); } else { /* Mouse button released. */ vt_mouse_terminput_button(vd, 3); } break; #ifdef notyet case MOUSE_MOTION_EVENT: if (mouse->u.data.z < 0) { /* Scroll up. */ sc_mouse_input_button(vd, 64); } else if (mouse->u.data.z > 0) { /* Scroll down. */ sc_mouse_input_button(vd, 65); } break; #endif } } static void vt_mouse_paste() { term_char_t *buf; int i, len; len = VD_PASTEBUFLEN(main_vd); buf = VD_PASTEBUF(main_vd); len /= sizeof(term_char_t); for (i = 0; i < len; i++) { if (buf[i] == '\0') continue; terminal_input_char(main_vd->vd_curwindow->vw_terminal, buf[i]); } } void vt_mouse_event(int type, int x, int y, int event, int cnt, int mlevel) { struct vt_device *vd; struct vt_window *vw; struct vt_font *vf; term_pos_t size; int len, mark; vd = main_vd; vw = vd->vd_curwindow; vf = vw->vw_font; mark = 0; if (vw->vw_flags & (VWF_MOUSE_HIDE | VWF_GRAPHICS)) /* * Either the mouse is disabled, or the window is in * "graphics mode". The graphics mode is usually set by * an X server, using the KDSETMODE ioctl. */ return; if (vf == NULL) /* Text mode. */ return; /* * TODO: add flag about pointer position changed, to not redraw chars * under mouse pointer when nothing changed. */ if (vw->vw_mouse_level > 0) vt_mouse_terminput(vd, type, x, y, event, cnt); switch (type) { case MOUSE_ACTION: case MOUSE_MOTION_EVENT: /* Movement */ x += vd->vd_mx; y += vd->vd_my; vt_termsize(vd, vf, &size); /* Apply limits. */ x = MAX(x, 0); y = MAX(y, 0); x = MIN(x, (size.tp_col * vf->vf_width) - 1); y = MIN(y, (size.tp_row * vf->vf_height) - 1); vd->vd_mx = x; vd->vd_my = y; if (vd->vd_mstate & MOUSE_BUTTON1DOWN) vtbuf_set_mark(&vw->vw_buf, VTB_MARK_MOVE, vd->vd_mx / vf->vf_width, vd->vd_my / vf->vf_height); vt_resume_flush_timer(vw->vw_device, 0); return; /* Done */ case MOUSE_BUTTON_EVENT: /* Buttons */ break; default: return; /* Done */ } switch (event) { case MOUSE_BUTTON1DOWN: switch (cnt % 4) { case 0: /* up */ mark = VTB_MARK_END; break; case 1: /* single click: start cut operation */ mark = VTB_MARK_START; break; case 2: /* double click: cut a word */ mark = VTB_MARK_WORD; break; case 3: /* triple click: cut a line */ mark = VTB_MARK_ROW; break; } break; case VT_MOUSE_PASTEBUTTON: switch (cnt) { case 0: /* up */ break; default: vt_mouse_paste(); break; } return; /* Done */ case VT_MOUSE_EXTENDBUTTON: switch (cnt) { case 0: /* up */ if (!(vd->vd_mstate & MOUSE_BUTTON1DOWN)) mark = VTB_MARK_EXTEND; else mark = 0; break; default: mark = VTB_MARK_EXTEND; break; } break; default: return; /* Done */ } /* Save buttons state. */ if (cnt > 0) vd->vd_mstate |= event; else vd->vd_mstate &= ~event; if (vtbuf_set_mark(&vw->vw_buf, mark, vd->vd_mx / vf->vf_width, vd->vd_my / vf->vf_height) == 1) { /* * We have something marked to copy, so update pointer to * window with selection. */ vt_resume_flush_timer(vw->vw_device, 0); switch (mark) { case VTB_MARK_END: case VTB_MARK_WORD: case VTB_MARK_ROW: case VTB_MARK_EXTEND: break; default: /* Other types of mark do not require to copy data. */ return; } /* Get current selection size in bytes. */ len = vtbuf_get_marked_len(&vw->vw_buf); if (len <= 0) return; /* Reallocate buffer only if old one is too small. */ if (len > VD_PASTEBUFSZ(vd)) { VD_PASTEBUF(vd) = realloc(VD_PASTEBUF(vd), len, M_VT, M_WAITOK | M_ZERO); /* Update buffer size. */ VD_PASTEBUFSZ(vd) = len; } /* Request copy/paste buffer data, no more than `len' */ vtbuf_extract_marked(&vw->vw_buf, VD_PASTEBUF(vd), VD_PASTEBUFSZ(vd)); VD_PASTEBUFLEN(vd) = len; /* XXX VD_PASTEBUF(vd) have to be freed on shutdown/unload. */ } } void vt_mouse_state(int show) { struct vt_device *vd; struct vt_window *vw; vd = main_vd; vw = vd->vd_curwindow; switch (show) { case VT_MOUSE_HIDE: vw->vw_flags |= VWF_MOUSE_HIDE; break; case VT_MOUSE_SHOW: vw->vw_flags &= ~VWF_MOUSE_HIDE; break; } /* Mark mouse position as dirty. */ vt_mark_mouse_position_as_dirty(vd); vt_resume_flush_timer(vw->vw_device, 0); } #endif static int vtterm_mmap(struct terminal *tm, vm_ooffset_t offset, vm_paddr_t * paddr, int nprot, vm_memattr_t *memattr) { struct vt_window *vw = tm->tm_softc; struct vt_device *vd = vw->vw_device; if (vd->vd_driver->vd_fb_mmap) return (vd->vd_driver->vd_fb_mmap(vd, offset, paddr, nprot, memattr)); return (ENXIO); } static int vtterm_ioctl(struct terminal *tm, u_long cmd, caddr_t data, struct thread *td) { struct vt_window *vw = tm->tm_softc; struct vt_device *vd = vw->vw_device; keyboard_t *kbd; int error, i, s; #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ defined(COMPAT_FREEBSD4) || defined(COMPAT_43) int ival; switch (cmd) { case _IO('v', 4): cmd = VT_RELDISP; break; case _IO('v', 5): cmd = VT_ACTIVATE; break; case _IO('v', 6): cmd = VT_WAITACTIVE; break; case _IO('K', 20): cmd = KDSKBSTATE; break; case _IO('K', 67): cmd = KDSETRAD; break; case _IO('K', 7): cmd = KDSKBMODE; break; case _IO('K', 8): cmd = KDMKTONE; break; case _IO('K', 63): cmd = KIOCSOUND; break; case _IO('K', 66): cmd = KDSETLED; break; case _IO('c', 110): cmd = CONS_SETKBD; break; default: goto skip_thunk; } ival = IOCPARM_IVAL(data); data = (caddr_t)&ival; skip_thunk: #endif switch (cmd) { case KDSETRAD: /* set keyboard repeat & delay rates (old) */ if (*(int *)data & ~0x7f) return (EINVAL); /* FALLTHROUGH */ case GIO_KEYMAP: case PIO_KEYMAP: case GIO_DEADKEYMAP: case PIO_DEADKEYMAP: case GETFKEY: case SETFKEY: case KDGKBINFO: case KDGKBTYPE: case KDGETREPEAT: /* get keyboard repeat & delay rates */ case KDSETREPEAT: /* set keyboard repeat & delay rates (new) */ case KBADDKBD: /* add/remove keyboard to/from mux */ case KBRELKBD: { error = 0; mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) error = kbdd_ioctl(kbd, cmd, data); mtx_unlock(&Giant); if (error == ENOIOCTL) { if (cmd == KDGKBTYPE) { /* always return something? XXX */ *(int *)data = 0; } else { return (ENODEV); } } return (error); } case KDGKBSTATE: { /* get keyboard state (locks) */ error = 0; if (vw == vd->vd_curwindow) { mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) error = vt_save_kbd_state(vw, kbd); mtx_unlock(&Giant); if (error != 0) return (error); } *(int *)data = vw->vw_kbdstate & LOCK_MASK; return (error); } case KDSKBSTATE: { /* set keyboard state (locks) */ int state; state = *(int *)data; if (state & ~LOCK_MASK) return (EINVAL); vw->vw_kbdstate &= ~LOCK_MASK; vw->vw_kbdstate |= state; error = 0; if (vw == vd->vd_curwindow) { mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) error = vt_update_kbd_state(vw, kbd); mtx_unlock(&Giant); } return (error); } case KDGETLED: { /* get keyboard LED status */ error = 0; if (vw == vd->vd_curwindow) { mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) error = vt_save_kbd_leds(vw, kbd); mtx_unlock(&Giant); if (error != 0) return (error); } *(int *)data = vw->vw_kbdstate & LED_MASK; return (error); } case KDSETLED: { /* set keyboard LED status */ int leds; leds = *(int *)data; if (leds & ~LED_MASK) return (EINVAL); vw->vw_kbdstate &= ~LED_MASK; vw->vw_kbdstate |= leds; error = 0; if (vw == vd->vd_curwindow) { mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) error = vt_update_kbd_leds(vw, kbd); mtx_unlock(&Giant); } return (error); } case KDGKBMODE: { error = 0; if (vw == vd->vd_curwindow) { mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) error = vt_save_kbd_mode(vw, kbd); mtx_unlock(&Giant); if (error != 0) return (error); } *(int *)data = vw->vw_kbdmode; return (error); } case KDSKBMODE: { int mode; mode = *(int *)data; switch (mode) { case K_XLATE: case K_RAW: case K_CODE: vw->vw_kbdmode = mode; error = 0; if (vw == vd->vd_curwindow) { mtx_lock(&Giant); kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) error = vt_update_kbd_mode(vw, kbd); mtx_unlock(&Giant); } return (error); default: return (EINVAL); } } case FBIOGTYPE: case FBIO_GETWINORG: /* get frame buffer window origin */ case FBIO_GETDISPSTART: /* get display start address */ case FBIO_GETLINEWIDTH: /* get scan line width in bytes */ case FBIO_BLANK: /* blank display */ if (vd->vd_driver->vd_fb_ioctl) return (vd->vd_driver->vd_fb_ioctl(vd, cmd, data, td)); break; case CONS_BLANKTIME: /* XXX */ return (0); case CONS_GET: /* XXX */ *(int *)data = M_CG640x480; return (0); case CONS_BELLTYPE: /* set bell type sound */ if ((*(int *)data) & CONS_QUIET_BELL) vd->vd_flags |= VDF_QUIET_BELL; else vd->vd_flags &= ~VDF_QUIET_BELL; return (0); case CONS_GETINFO: { vid_info_t *vi = (vid_info_t *)data; if (vi->size != sizeof(struct vid_info)) return (EINVAL); if (vw == vd->vd_curwindow) { kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd != NULL) vt_save_kbd_state(vw, kbd); } vi->m_num = vd->vd_curwindow->vw_number + 1; vi->mk_keylock = vw->vw_kbdstate & LOCK_MASK; /* XXX: other fields! */ return (0); } case CONS_GETVERS: *(int *)data = 0x200; return (0); case CONS_MODEINFO: /* XXX */ return (0); case CONS_MOUSECTL: { mouse_info_t *mouse = (mouse_info_t*)data; /* * All the commands except MOUSE_SHOW nd MOUSE_HIDE * should not be applied to individual TTYs, but only to * consolectl. */ switch (mouse->operation) { case MOUSE_HIDE: if (vd->vd_flags & VDF_MOUSECURSOR) { vd->vd_flags &= ~VDF_MOUSECURSOR; #ifndef SC_NO_CUTPASTE vt_mouse_state(VT_MOUSE_HIDE); #endif } return (0); case MOUSE_SHOW: if (!(vd->vd_flags & VDF_MOUSECURSOR)) { vd->vd_flags |= VDF_MOUSECURSOR; vd->vd_mx = vd->vd_width / 2; vd->vd_my = vd->vd_height / 2; #ifndef SC_NO_CUTPASTE vt_mouse_state(VT_MOUSE_SHOW); #endif } return (0); default: return (EINVAL); } } case PIO_VFONT: { struct vt_font *vf; if (vd->vd_flags & VDF_TEXTMODE) return (ENOTSUP); error = vtfont_load((void *)data, &vf); if (error != 0) return (error); error = vt_change_font(vw, vf); vtfont_unref(vf); return (error); } case PIO_VFONT_DEFAULT: { /* Reset to default font. */ error = vt_change_font(vw, &vt_font_default); return (error); } case GIO_SCRNMAP: { scrmap_t *sm = (scrmap_t *)data; /* We don't have screen maps, so return a handcrafted one. */ for (i = 0; i < 256; i++) sm->scrmap[i] = i; return (0); } case KDSETMODE: /* * FIXME: This implementation is incomplete compared to * syscons. */ switch (*(int *)data) { case KD_TEXT: case KD_TEXT1: case KD_PIXEL: vw->vw_flags &= ~VWF_GRAPHICS; break; case KD_GRAPHICS: vw->vw_flags |= VWF_GRAPHICS; break; } return (0); case KDENABIO: /* allow io operations */ error = priv_check(td, PRIV_IO); if (error != 0) return (error); error = securelevel_gt(td->td_ucred, 0); if (error != 0) return (error); #if defined(__i386__) td->td_frame->tf_eflags |= PSL_IOPL; #elif defined(__amd64__) td->td_frame->tf_rflags |= PSL_IOPL; #endif return (0); case KDDISABIO: /* disallow io operations (default) */ #if defined(__i386__) td->td_frame->tf_eflags &= ~PSL_IOPL; #elif defined(__amd64__) td->td_frame->tf_rflags &= ~PSL_IOPL; #endif return (0); case KDMKTONE: /* sound the bell */ vtterm_beep(tm, *(u_int *)data); return (0); case KIOCSOUND: /* make tone (*data) hz */ /* TODO */ return (0); case CONS_SETKBD: /* set the new keyboard */ mtx_lock(&Giant); error = 0; if (vd->vd_keyboard != *(int *)data) { kbd = kbd_get_keyboard(*(int *)data); if (kbd == NULL) { mtx_unlock(&Giant); return (EINVAL); } i = kbd_allocate(kbd->kb_name, kbd->kb_unit, (void *)vd, vt_kbdevent, vd); if (i >= 0) { if (vd->vd_keyboard != -1) { vt_save_kbd_state(vd->vd_curwindow, kbd); kbd_release(kbd, (void *)vd); } kbd = kbd_get_keyboard(i); vd->vd_keyboard = i; vt_update_kbd_mode(vd->vd_curwindow, kbd); vt_update_kbd_state(vd->vd_curwindow, kbd); } else { error = EPERM; /* XXX */ } } mtx_unlock(&Giant); return (error); case CONS_RELKBD: /* release the current keyboard */ mtx_lock(&Giant); error = 0; if (vd->vd_keyboard != -1) { kbd = kbd_get_keyboard(vd->vd_keyboard); if (kbd == NULL) { mtx_unlock(&Giant); return (EINVAL); } vt_save_kbd_state(vd->vd_curwindow, kbd); error = kbd_release(kbd, (void *)vd); if (error == 0) { vd->vd_keyboard = -1; } } mtx_unlock(&Giant); return (error); case VT_ACTIVATE: { int win; win = *(int *)data - 1; DPRINTF(5, "%s%d: VT_ACTIVATE ttyv%d ", SC_DRIVER_NAME, VT_UNIT(vw), win); if ((win >= VT_MAXWINDOWS) || (win < 0)) return (EINVAL); return (vt_proc_window_switch(vd->vd_windows[win])); } case VT_GETACTIVE: *(int *)data = vd->vd_curwindow->vw_number + 1; return (0); case VT_GETINDEX: *(int *)data = vw->vw_number + 1; return (0); case VT_LOCKSWITCH: /* TODO: Check current state, switching can be in progress. */ if ((*(int *)data) == 0x01) vw->vw_flags |= VWF_VTYLOCK; else if ((*(int *)data) == 0x02) vw->vw_flags &= ~VWF_VTYLOCK; else return (EINVAL); return (0); case VT_OPENQRY: VT_LOCK(vd); for (i = 0; i < VT_MAXWINDOWS; i++) { vw = vd->vd_windows[i]; if (vw == NULL) continue; if (!(vw->vw_flags & VWF_OPENED)) { *(int *)data = vw->vw_number + 1; VT_UNLOCK(vd); return (0); } } VT_UNLOCK(vd); return (EINVAL); case VT_WAITACTIVE: { unsigned int idx; error = 0; idx = *(unsigned int *)data; if (idx > VT_MAXWINDOWS) return (EINVAL); if (idx > 0) vw = vd->vd_windows[idx - 1]; VT_LOCK(vd); while (vd->vd_curwindow != vw && error == 0) error = cv_wait_sig(&vd->vd_winswitch, &vd->vd_lock); VT_UNLOCK(vd); return (error); } case VT_SETMODE: { /* set screen switcher mode */ struct vt_mode *mode; struct proc *p1; mode = (struct vt_mode *)data; DPRINTF(5, "%s%d: VT_SETMODE ", SC_DRIVER_NAME, VT_UNIT(vw)); if (vw->vw_smode.mode == VT_PROCESS) { p1 = pfind(vw->vw_pid); if (vw->vw_proc == p1 && vw->vw_proc != td->td_proc) { if (p1) PROC_UNLOCK(p1); DPRINTF(5, "error EPERM\n"); return (EPERM); } if (p1) PROC_UNLOCK(p1); } if (mode->mode == VT_AUTO) { vw->vw_smode.mode = VT_AUTO; vw->vw_proc = NULL; vw->vw_pid = 0; DPRINTF(5, "VT_AUTO, "); if (vw == vw->vw_device->vd_windows[VT_CONSWINDOW]) cnavailable(vw->vw_terminal->consdev, TRUE); /* were we in the middle of the vty switching process? */ if (finish_vt_rel(vw, TRUE, &s) == 0) DPRINTF(5, "reset WAIT_REL, "); if (finish_vt_acq(vw) == 0) DPRINTF(5, "reset WAIT_ACQ, "); return (0); } else if (mode->mode == VT_PROCESS) { if (!ISSIGVALID(mode->relsig) || !ISSIGVALID(mode->acqsig) || !ISSIGVALID(mode->frsig)) { DPRINTF(5, "error EINVAL\n"); return (EINVAL); } DPRINTF(5, "VT_PROCESS %d, ", td->td_proc->p_pid); bcopy(data, &vw->vw_smode, sizeof(struct vt_mode)); vw->vw_proc = td->td_proc; vw->vw_pid = vw->vw_proc->p_pid; if (vw == vw->vw_device->vd_windows[VT_CONSWINDOW]) cnavailable(vw->vw_terminal->consdev, FALSE); } else { DPRINTF(5, "VT_SETMODE failed, unknown mode %d\n", mode->mode); return (EINVAL); } DPRINTF(5, "\n"); return (0); } case VT_GETMODE: /* get screen switcher mode */ bcopy(&vw->vw_smode, data, sizeof(struct vt_mode)); return (0); case VT_RELDISP: /* screen switcher ioctl */ /* * This must be the current vty which is in the VT_PROCESS * switching mode... */ if ((vw != vd->vd_curwindow) || (vw->vw_smode.mode != VT_PROCESS)) { return (EINVAL); } /* ...and this process is controlling it. */ if (vw->vw_proc != td->td_proc) { return (EPERM); } error = EINVAL; switch(*(int *)data) { case VT_FALSE: /* user refuses to release screen, abort */ if ((error = finish_vt_rel(vw, FALSE, &s)) == 0) DPRINTF(5, "%s%d: VT_RELDISP: VT_FALSE\n", SC_DRIVER_NAME, VT_UNIT(vw)); break; case VT_TRUE: /* user has released screen, go on */ /* finish_vt_rel(..., TRUE, ...) should not be locked */ if (vw->vw_flags & VWF_SWWAIT_REL) { if ((error = finish_vt_rel(vw, TRUE, &s)) == 0) DPRINTF(5, "%s%d: VT_RELDISP: VT_TRUE\n", SC_DRIVER_NAME, VT_UNIT(vw)); } else { error = EINVAL; } return (error); case VT_ACKACQ: /* acquire acknowledged, switch completed */ if ((error = finish_vt_acq(vw)) == 0) DPRINTF(5, "%s%d: VT_RELDISP: VT_ACKACQ\n", SC_DRIVER_NAME, VT_UNIT(vw)); break; default: break; } return (error); } return (ENOIOCTL); } static struct vt_window * vt_allocate_window(struct vt_device *vd, unsigned int window) { struct vt_window *vw; struct terminal *tm; term_pos_t size; struct winsize wsz; vw = malloc(sizeof *vw, M_VT, M_WAITOK|M_ZERO); vw->vw_device = vd; vw->vw_number = window; vw->vw_kbdmode = K_XLATE; if ((vd->vd_flags & VDF_TEXTMODE) == 0) { vw->vw_font = vtfont_ref(&vt_font_default); vt_compute_drawable_area(vw); } vt_termsize(vd, vw->vw_font, &size); vt_winsize(vd, vw->vw_font, &wsz); vtbuf_init(&vw->vw_buf, &size); tm = vw->vw_terminal = terminal_alloc(&vt_termclass, vw); terminal_set_winsize(tm, &wsz); vd->vd_windows[window] = vw; callout_init(&vw->vw_proc_dead_timer, 0); return (vw); } void vt_upgrade(struct vt_device *vd) { struct vt_window *vw; unsigned int i; int register_handlers; if (!vty_enabled(VTY_VT)) return; if (main_vd->vd_driver == NULL) return; for (i = 0; i < VT_MAXWINDOWS; i++) { vw = vd->vd_windows[i]; if (vw == NULL) { /* New window. */ vw = vt_allocate_window(vd, i); } if (!(vw->vw_flags & VWF_READY)) { callout_init(&vw->vw_proc_dead_timer, 0); terminal_maketty(vw->vw_terminal, "v%r", VT_UNIT(vw)); vw->vw_flags |= VWF_READY; if (vw->vw_flags & VWF_CONSOLE) { /* For existing console window. */ EVENTHANDLER_REGISTER(shutdown_pre_sync, vt_window_switch, vw, SHUTDOWN_PRI_DEFAULT); } } } VT_LOCK(vd); if (vd->vd_curwindow == NULL) vd->vd_curwindow = vd->vd_windows[VT_CONSWINDOW]; register_handlers = 0; if (!(vd->vd_flags & VDF_ASYNC)) { /* Attach keyboard. */ vt_allocate_keyboard(vd); /* Init 25 Hz timer. */ callout_init_mtx(&vd->vd_timer, &vd->vd_lock, 0); /* Start timer when everything ready. */ vd->vd_flags |= VDF_ASYNC; callout_reset(&vd->vd_timer, hz / VT_TIMERFREQ, vt_timer, vd); vd->vd_timer_armed = 1; register_handlers = 1; } VT_UNLOCK(vd); /* Refill settings with new sizes. */ vt_resize(vd); if (register_handlers) { /* Register suspend/resume handlers. */ EVENTHANDLER_REGISTER(power_suspend_early, vt_suspend_handler, vd, EVENTHANDLER_PRI_ANY); EVENTHANDLER_REGISTER(power_resume, vt_resume_handler, vd, EVENTHANDLER_PRI_ANY); } } static void vt_resize(struct vt_device *vd) { struct vt_window *vw; int i; for (i = 0; i < VT_MAXWINDOWS; i++) { vw = vd->vd_windows[i]; VT_LOCK(vd); /* Assign default font to window, if not textmode. */ if (!(vd->vd_flags & VDF_TEXTMODE) && vw->vw_font == NULL) vw->vw_font = vtfont_ref(&vt_font_default); VT_UNLOCK(vd); /* Resize terminal windows */ while (vt_change_font(vw, vw->vw_font) == EBUSY) { DPRINTF(100, "%s: vt_change_font() is busy, " "window %d\n", __func__, i); } } } static void vt_replace_backend(const struct vt_driver *drv, void *softc) { struct vt_device *vd; vd = main_vd; if (vd->vd_flags & VDF_ASYNC) { /* Stop vt_flush periodic task. */ VT_LOCK(vd); vt_suspend_flush_timer(vd); VT_UNLOCK(vd); /* * Mute current terminal until we done. vt_change_font (called * from vt_resize) will unmute it. */ terminal_mute(vd->vd_curwindow->vw_terminal, 1); } /* * Reset VDF_TEXTMODE flag, driver who require that flag (vt_vga) will * set it. */ VT_LOCK(vd); vd->vd_flags &= ~VDF_TEXTMODE; if (drv != NULL) { /* * We want to upgrade from the current driver to the * given driver. */ vd->vd_prev_driver = vd->vd_driver; vd->vd_prev_softc = vd->vd_softc; vd->vd_driver = drv; vd->vd_softc = softc; vd->vd_driver->vd_init(vd); } else if (vd->vd_prev_driver != NULL && vd->vd_prev_softc != NULL) { /* * No driver given: we want to downgrade to the previous * driver. */ const struct vt_driver *old_drv; void *old_softc; old_drv = vd->vd_driver; old_softc = vd->vd_softc; vd->vd_driver = vd->vd_prev_driver; vd->vd_softc = vd->vd_prev_softc; vd->vd_prev_driver = NULL; vd->vd_prev_softc = NULL; vd->vd_flags |= VDF_DOWNGRADE; vd->vd_driver->vd_init(vd); if (old_drv->vd_fini) old_drv->vd_fini(vd, old_softc); vd->vd_flags &= ~VDF_DOWNGRADE; } VT_UNLOCK(vd); /* Update windows sizes and initialize last items. */ vt_upgrade(vd); #ifdef DEV_SPLASH if (vd->vd_flags & VDF_SPLASH) vtterm_splash(vd); #endif if (vd->vd_flags & VDF_ASYNC) { /* Allow to put chars now. */ terminal_mute(vd->vd_curwindow->vw_terminal, 0); /* Rerun timer for screen updates. */ vt_resume_flush_timer(vd, 0); } /* * Register as console. If it already registered, cnadd() will ignore * it. */ termcn_cnregister(vd->vd_windows[VT_CONSWINDOW]->vw_terminal); } static void vt_suspend_handler(void *priv) { struct vt_device *vd; vd = priv; if (vd->vd_driver != NULL && vd->vd_driver->vd_suspend != NULL) vd->vd_driver->vd_suspend(vd); } static void vt_resume_handler(void *priv) { struct vt_device *vd; vd = priv; if (vd->vd_driver != NULL && vd->vd_driver->vd_resume != NULL) vd->vd_driver->vd_resume(vd); } void vt_allocate(const struct vt_driver *drv, void *softc) { if (!vty_enabled(VTY_VT)) return; if (main_vd->vd_driver == NULL) { main_vd->vd_driver = drv; printf("VT: initialize with new VT driver \"%s\".\n", drv->vd_name); } else { /* * Check if have rights to replace current driver. For example: * it is bad idea to replace KMS driver with generic VGA one. */ if (drv->vd_priority <= main_vd->vd_driver->vd_priority) { printf("VT: Driver priority %d too low. Current %d\n ", drv->vd_priority, main_vd->vd_driver->vd_priority); return; } printf("VT: Replacing driver \"%s\" with new \"%s\".\n", main_vd->vd_driver->vd_name, drv->vd_name); } vt_replace_backend(drv, softc); } void vt_deallocate(const struct vt_driver *drv, void *softc) { if (!vty_enabled(VTY_VT)) return; if (main_vd->vd_prev_driver == NULL || main_vd->vd_driver != drv || main_vd->vd_softc != softc) return; printf("VT: Switching back from \"%s\" to \"%s\".\n", main_vd->vd_driver->vd_name, main_vd->vd_prev_driver->vd_name); vt_replace_backend(NULL, NULL); } void vt_suspend(struct vt_device *vd) { int error; if (vt_suspendswitch == 0) return; /* Save current window. */ vd->vd_savedwindow = vd->vd_curwindow; /* Ask holding process to free window and switch to console window */ vt_proc_window_switch(vd->vd_windows[VT_CONSWINDOW]); /* Wait for the window switch to complete. */ error = 0; VT_LOCK(vd); while (vd->vd_curwindow != vd->vd_windows[VT_CONSWINDOW] && error == 0) error = cv_wait_sig(&vd->vd_winswitch, &vd->vd_lock); VT_UNLOCK(vd); } void vt_resume(struct vt_device *vd) { if (vt_suspendswitch == 0) return; /* Switch back to saved window */ if (vd->vd_savedwindow != NULL) vt_proc_window_switch(vd->vd_savedwindow); vd->vd_savedwindow = NULL; } Index: user/ngie/more-tests/sys/fs/nfs/nfsport.h =================================================================== --- user/ngie/more-tests/sys/fs/nfs/nfsport.h (revision 281974) +++ user/ngie/more-tests/sys/fs/nfs/nfsport.h (revision 281975) @@ -1,969 +1,969 @@ /*- * Copyright (c) 1989, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Rick Macklem at The University of Guelph. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _NFS_NFSPORT_H_ #define _NFS_NFSPORT_H_ /* * In general, I'm not fond of #includes in .h files, but this seems * to be the cleanest way to handle #include files for the ports. */ #ifdef _KERNEL #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * For Darwin, these functions should be "static" when built in a kext. * (This is always defined as nil otherwise.) */ #define APPLESTATIC #include #include #include #include #include #include #include #include #include #include #include "opt_nfs.h" #include "opt_ufs.h" /* * These types must be defined before the nfs includes. */ #define NFSSOCKADDR_T struct sockaddr * #define NFSPROC_T struct thread #define NFSDEV_T dev_t #define NFSSVCARGS nfssvc_args #define NFSACL_T struct acl /* * These should be defined as the types used for the corresponding VOP's * argument type. */ #define NFS_ACCESS_ARGS struct vop_access_args #define NFS_OPEN_ARGS struct vop_open_args #define NFS_GETATTR_ARGS struct vop_getattr_args #define NFS_LOOKUP_ARGS struct vop_lookup_args #define NFS_READDIR_ARGS struct vop_readdir_args /* * Allocate mbufs. Must succeed and never set the mbuf ptr to NULL. */ #define NFSMGET(m) do { \ MGET((m), M_WAITOK, MT_DATA); \ while ((m) == NULL ) { \ (void) nfs_catnap(PZERO, 0, "nfsmget"); \ MGET((m), M_WAITOK, MT_DATA); \ } \ } while (0) #define NFSMGETHDR(m) do { \ MGETHDR((m), M_WAITOK, MT_DATA); \ while ((m) == NULL ) { \ (void) nfs_catnap(PZERO, 0, "nfsmget"); \ MGETHDR((m), M_WAITOK, MT_DATA); \ } \ } while (0) #define NFSMCLGET(m, w) do { \ MGET((m), M_WAITOK, MT_DATA); \ while ((m) == NULL ) { \ (void) nfs_catnap(PZERO, 0, "nfsmget"); \ MGET((m), M_WAITOK, MT_DATA); \ } \ MCLGET((m), (w)); \ } while (0) #define NFSMCLGETHDR(m, w) do { \ MGETHDR((m), M_WAITOK, MT_DATA); \ while ((m) == NULL ) { \ (void) nfs_catnap(PZERO, 0, "nfsmget"); \ MGETHDR((m), M_WAITOK, MT_DATA); \ } \ } while (0) #define NFSMTOD mtod /* * Client side constant for size of a lockowner name. */ #define NFSV4CL_LOCKNAMELEN 12 /* * Type for a mutex lock. */ #define NFSMUTEX_T struct mtx #endif /* _KERNEL */ /* * NFSv4 Operation numbers. */ #define NFSV4OP_ACCESS 3 #define NFSV4OP_CLOSE 4 #define NFSV4OP_COMMIT 5 #define NFSV4OP_CREATE 6 #define NFSV4OP_DELEGPURGE 7 #define NFSV4OP_DELEGRETURN 8 #define NFSV4OP_GETATTR 9 #define NFSV4OP_GETFH 10 #define NFSV4OP_LINK 11 #define NFSV4OP_LOCK 12 #define NFSV4OP_LOCKT 13 #define NFSV4OP_LOCKU 14 #define NFSV4OP_LOOKUP 15 #define NFSV4OP_LOOKUPP 16 #define NFSV4OP_NVERIFY 17 #define NFSV4OP_OPEN 18 #define NFSV4OP_OPENATTR 19 #define NFSV4OP_OPENCONFIRM 20 #define NFSV4OP_OPENDOWNGRADE 21 #define NFSV4OP_PUTFH 22 #define NFSV4OP_PUTPUBFH 23 #define NFSV4OP_PUTROOTFH 24 #define NFSV4OP_READ 25 #define NFSV4OP_READDIR 26 #define NFSV4OP_READLINK 27 #define NFSV4OP_REMOVE 28 #define NFSV4OP_RENAME 29 #define NFSV4OP_RENEW 30 #define NFSV4OP_RESTOREFH 31 #define NFSV4OP_SAVEFH 32 #define NFSV4OP_SECINFO 33 #define NFSV4OP_SETATTR 34 #define NFSV4OP_SETCLIENTID 35 #define NFSV4OP_SETCLIENTIDCFRM 36 #define NFSV4OP_VERIFY 37 #define NFSV4OP_WRITE 38 #define NFSV4OP_RELEASELCKOWN 39 /* * Must be one greater than the last Operation#. */ #define NFSV4OP_NOPS 40 /* * Additional Ops for NFSv4.1. */ #define NFSV4OP_BACKCHANNELCTL 40 #define NFSV4OP_BINDCONNTOSESS 41 #define NFSV4OP_EXCHANGEID 42 #define NFSV4OP_CREATESESSION 43 #define NFSV4OP_DESTROYSESSION 44 #define NFSV4OP_FREESTATEID 45 #define NFSV4OP_GETDIRDELEG 46 #define NFSV4OP_GETDEVINFO 47 #define NFSV4OP_GETDEVLIST 48 #define NFSV4OP_LAYOUTCOMMIT 49 #define NFSV4OP_LAYOUTGET 50 #define NFSV4OP_LAYOUTRETURN 51 #define NFSV4OP_SECINFONONAME 52 #define NFSV4OP_SEQUENCE 53 #define NFSV4OP_SETSSV 54 #define NFSV4OP_TESTSTATEID 55 #define NFSV4OP_WANTDELEG 56 #define NFSV4OP_DESTROYCLIENTID 57 #define NFSV4OP_RECLAIMCOMPL 58 /* * Must be one more than last op#. */ #define NFSV41_NOPS 59 /* Quirky case if the illegal op code */ #define NFSV4OP_OPILLEGAL 10044 /* * Fake NFSV4OP_xxx used for nfsstat. Start at NFSV4OP_NOPS. */ #define NFSV4OP_SYMLINK (NFSV4OP_NOPS) #define NFSV4OP_MKDIR (NFSV4OP_NOPS + 1) #define NFSV4OP_RMDIR (NFSV4OP_NOPS + 2) #define NFSV4OP_READDIRPLUS (NFSV4OP_NOPS + 3) #define NFSV4OP_MKNOD (NFSV4OP_NOPS + 4) #define NFSV4OP_FSSTAT (NFSV4OP_NOPS + 5) #define NFSV4OP_FSINFO (NFSV4OP_NOPS + 6) #define NFSV4OP_PATHCONF (NFSV4OP_NOPS + 7) #define NFSV4OP_V3CREATE (NFSV4OP_NOPS + 8) /* * This is the count of the fake operations listed above. */ #define NFSV4OP_FAKENOPS 9 /* * and the Callback OPs */ #define NFSV4OP_CBGETATTR 3 #define NFSV4OP_CBRECALL 4 /* * Must be one greater than the last Callback Operation#. */ #define NFSV4OP_CBNOPS 5 /* * Additional Callback Ops for NFSv4.1 only. Not yet in nfsstats. */ #define NFSV4OP_CBLAYOUTRECALL 5 #define NFSV4OP_CBNOTIFY 6 #define NFSV4OP_CBPUSHDELEG 7 #define NFSV4OP_CBRECALLANY 8 #define NFSV4OP_CBRECALLOBJAVAIL 9 #define NFSV4OP_CBRECALLSLOT 10 #define NFSV4OP_CBSEQUENCE 11 #define NFSV4OP_CBWANTCANCELLED 12 #define NFSV4OP_CBNOTIFYLOCK 13 #define NFSV4OP_CBNOTIFYDEVID 14 /* * The lower numbers -> 21 are used by NFSv2 and v3. These define higher * numbers used by NFSv4. * NFS_V3NPROCS is one greater than the last V3 op and NFS_NPROCS is * one greater than the last number. */ #ifndef NFS_V3NPROCS #define NFS_V3NPROCS 22 #define NFSPROC_LOOKUPP 22 #define NFSPROC_SETCLIENTID 23 #define NFSPROC_SETCLIENTIDCFRM 24 #define NFSPROC_LOCK 25 #define NFSPROC_LOCKU 26 #define NFSPROC_OPEN 27 #define NFSPROC_CLOSE 28 #define NFSPROC_OPENCONFIRM 29 #define NFSPROC_LOCKT 30 #define NFSPROC_OPENDOWNGRADE 31 #define NFSPROC_RENEW 32 #define NFSPROC_PUTROOTFH 33 #define NFSPROC_RELEASELCKOWN 34 #define NFSPROC_DELEGRETURN 35 #define NFSPROC_RETDELEGREMOVE 36 #define NFSPROC_RETDELEGRENAME1 37 #define NFSPROC_RETDELEGRENAME2 38 #define NFSPROC_GETACL 39 #define NFSPROC_SETACL 40 /* * Must be defined as one higher than the last Proc# above. */ #define NFSV4_NPROCS 41 /* Additional procedures for NFSv4.1. */ #define NFSPROC_EXCHANGEID 41 #define NFSPROC_CREATESESSION 42 #define NFSPROC_DESTROYSESSION 43 #define NFSPROC_DESTROYCLIENT 44 #define NFSPROC_FREESTATEID 45 #define NFSPROC_LAYOUTGET 46 #define NFSPROC_GETDEVICEINFO 47 #define NFSPROC_LAYOUTCOMMIT 48 #define NFSPROC_LAYOUTRETURN 49 #define NFSPROC_RECLAIMCOMPL 50 #define NFSPROC_WRITEDS 51 #define NFSPROC_READDS 52 #define NFSPROC_COMMITDS 53 /* * Must be defined as one higher than the last NFSv4.1 Proc# above. */ #define NFSV41_NPROCS 54 #endif /* NFS_V3NPROCS */ /* * Stats structure */ struct ext_nfsstats { int attrcache_hits; int attrcache_misses; int lookupcache_hits; int lookupcache_misses; int direofcache_hits; int direofcache_misses; int accesscache_hits; int accesscache_misses; int biocache_reads; int read_bios; int read_physios; int biocache_writes; int write_bios; int write_physios; int biocache_readlinks; int readlink_bios; int biocache_readdirs; int readdir_bios; int rpccnt[NFSV4_NPROCS]; int rpcretries; int srvrpccnt[NFSV4OP_NOPS + NFSV4OP_FAKENOPS]; int srvrpc_errs; int srv_errs; int rpcrequests; int rpctimeouts; int rpcunexpected; int rpcinvalid; int srvcache_inproghits; int srvcache_idemdonehits; int srvcache_nonidemdonehits; int srvcache_misses; int srvcache_tcppeak; int srvcache_size; int srvclients; int srvopenowners; int srvopens; int srvlockowners; int srvlocks; int srvdelegates; int cbrpccnt[NFSV4OP_CBNOPS]; int clopenowners; int clopens; int cllockowners; int cllocks; int cldelegates; int cllocalopenowners; int cllocalopens; int cllocallockowners; int cllocallocks; }; #ifdef _KERNEL /* * Define the ext_nfsstats as nfsstats for the kernel code. */ #define nfsstats ext_nfsstats /* * Define NFS_NPROCS as NFSV4_NPROCS for the experimental kernel code. */ #ifndef NFS_NPROCS #define NFS_NPROCS NFSV4_NPROCS #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * Just to keep nfs_var.h happy. */ struct nfs_vattr { int junk; }; struct nfsvattr { struct vattr na_vattr; nfsattrbit_t na_suppattr; u_int32_t na_mntonfileno; u_int64_t na_filesid[2]; }; #define na_type na_vattr.va_type #define na_mode na_vattr.va_mode #define na_nlink na_vattr.va_nlink #define na_uid na_vattr.va_uid #define na_gid na_vattr.va_gid #define na_fsid na_vattr.va_fsid #define na_fileid na_vattr.va_fileid #define na_size na_vattr.va_size #define na_blocksize na_vattr.va_blocksize #define na_atime na_vattr.va_atime #define na_mtime na_vattr.va_mtime #define na_ctime na_vattr.va_ctime #define na_gen na_vattr.va_gen #define na_flags na_vattr.va_flags #define na_rdev na_vattr.va_rdev #define na_bytes na_vattr.va_bytes #define na_filerev na_vattr.va_filerev #define na_vaflags na_vattr.va_vaflags #include /* * This is the header structure used for the lists, etc. (It has the * above record in it. */ struct nfsrv_stablefirst { LIST_HEAD(, nfsrv_stable) nsf_head; /* Head of nfsrv_stable list */ time_t nsf_eograce; /* Time grace period ends */ time_t *nsf_bootvals; /* Previous boottime values */ struct file *nsf_fp; /* File table pointer */ u_char nsf_flags; /* NFSNSF_ flags */ struct nfsf_rec nsf_rec; /* and above first record */ }; #define nsf_lease nsf_rec.lease #define nsf_numboots nsf_rec.numboots /* NFSNSF_xxx flags */ #define NFSNSF_UPDATEDONE 0x01 #define NFSNSF_GRACEOVER 0x02 #define NFSNSF_NEEDLOCK 0x04 #define NFSNSF_EXPIREDCLIENT 0x08 #define NFSNSF_NOOPENS 0x10 #define NFSNSF_OK 0x20 /* * Maximum number of boot times allowed in record. Although there is * really no need for a fixed upper bound, this serves as a sanity check * for a corrupted file. */ #define NFSNSF_MAXNUMBOOTS 10000 /* * This structure defines the other records in the file. The * nst_client array is actually the size of the client string name. */ struct nfst_rec { u_int16_t len; u_char flag; u_char client[1]; }; /* and the values for flag */ #define NFSNST_NEWSTATE 0x1 #define NFSNST_REVOKE 0x2 #define NFSNST_GOTSTATE 0x4 /* * This structure is linked onto nfsrv_stablefirst for the duration of * reclaim. */ struct nfsrv_stable { LIST_ENTRY(nfsrv_stable) nst_list; struct nfsclient *nst_clp; struct nfst_rec nst_rec; }; #define nst_timestamp nst_rec.timestamp #define nst_len nst_rec.len #define nst_flag nst_rec.flag #define nst_client nst_rec.client /* * At some point the server will run out of kernel storage for * state structures. For FreeBSD5.2, this results in a panic * kmem_map is full. It happens at well over 1000000 opens plus * locks on a PIII-800 with 256Mbytes, so that is where I've set * the limit. If your server panics due to too many opens/locks, * decrease the size of NFSRV_V4STATELIMIT. If you find the server * returning NFS4ERR_RESOURCE a lot and have lots of memory, try * increasing it. */ #define NFSRV_V4STATELIMIT 500000 /* Max # of Opens + Locks */ /* * The type required differs with BSDen (just the second arg). */ void nfsrvd_rcv(struct socket *, void *, int); /* * Macros for handling socket addresses. (Hopefully this makes the code * more portable, since I've noticed some 'BSD don't have sockaddrs in * mbufs any more.) */ #define NFSSOCKADDR(a, t) ((t)(a)) #define NFSSOCKADDRALLOC(a) \ do { \ MALLOC((a), struct sockaddr *, sizeof (struct sockaddr), \ M_SONAME, M_WAITOK); \ NFSBZERO((a), sizeof (struct sockaddr)); \ } while (0) #define NFSSOCKADDRSIZE(a, s) ((a)->sa_len = (s)) #define NFSSOCKADDRFREE(a) \ do { \ if (a) \ FREE((caddr_t)(a), M_SONAME); \ } while (0) /* * These should be defined as a process or thread structure, as required * for signal handling, etc. */ #define NFSNEWCRED(c) (crdup(c)) #define NFSPROCCRED(p) ((p)->td_ucred) #define NFSFREECRED(c) (crfree(c)) #define NFSUIOPROC(u, p) ((u)->uio_td = NULL) #define NFSPROCP(p) ((p)->td_proc) /* * Define these so that cn_hash and its length is ignored. */ #define NFSCNHASHZERO(c) #define NFSCNHASH(c, v) #define NCHNAMLEN 9999999 /* * These macros are defined to initialize and set the timer routine. */ #define NFS_TIMERINIT \ newnfs_timer(NULL) /* * Handle SMP stuff: */ #define NFSSTATESPINLOCK extern struct mtx nfs_state_mutex #define NFSLOCKSTATE() mtx_lock(&nfs_state_mutex) #define NFSUNLOCKSTATE() mtx_unlock(&nfs_state_mutex) #define NFSSTATEMUTEXPTR (&nfs_state_mutex) #define NFSREQSPINLOCK extern struct mtx nfs_req_mutex #define NFSLOCKREQ() mtx_lock(&nfs_req_mutex) #define NFSUNLOCKREQ() mtx_unlock(&nfs_req_mutex) #define NFSSOCKMUTEX extern struct mtx nfs_slock_mutex #define NFSSOCKMUTEXPTR (&nfs_slock_mutex) #define NFSLOCKSOCK() mtx_lock(&nfs_slock_mutex) #define NFSUNLOCKSOCK() mtx_unlock(&nfs_slock_mutex) #define NFSNAMEIDMUTEX extern struct mtx nfs_nameid_mutex #define NFSLOCKNAMEID() mtx_lock(&nfs_nameid_mutex) #define NFSUNLOCKNAMEID() mtx_unlock(&nfs_nameid_mutex) #define NFSNAMEIDREQUIRED() mtx_assert(&nfs_nameid_mutex, MA_OWNED) #define NFSCLSTATEMUTEX extern struct mtx nfs_clstate_mutex #define NFSCLSTATEMUTEXPTR (&nfs_clstate_mutex) #define NFSLOCKCLSTATE() mtx_lock(&nfs_clstate_mutex) #define NFSUNLOCKCLSTATE() mtx_unlock(&nfs_clstate_mutex) #define NFSDLOCKMUTEX extern struct mtx newnfsd_mtx #define NFSDLOCKMUTEXPTR (&newnfsd_mtx) #define NFSD_LOCK() mtx_lock(&newnfsd_mtx) #define NFSD_UNLOCK() mtx_unlock(&newnfsd_mtx) #define NFSD_LOCK_ASSERT() mtx_assert(&newnfsd_mtx, MA_OWNED) #define NFSD_UNLOCK_ASSERT() mtx_assert(&newnfsd_mtx, MA_NOTOWNED) #define NFSV4ROOTLOCKMUTEX extern struct mtx nfs_v4root_mutex #define NFSV4ROOTLOCKMUTEXPTR (&nfs_v4root_mutex) #define NFSLOCKV4ROOTMUTEX() mtx_lock(&nfs_v4root_mutex) #define NFSUNLOCKV4ROOTMUTEX() mtx_unlock(&nfs_v4root_mutex) #define NFSLOCKNODE(n) mtx_lock(&((n)->n_mtx)) #define NFSUNLOCKNODE(n) mtx_unlock(&((n)->n_mtx)) #define NFSLOCKMNT(m) mtx_lock(&((m)->nm_mtx)) #define NFSUNLOCKMNT(m) mtx_unlock(&((m)->nm_mtx)) #define NFSLOCKREQUEST(r) mtx_lock(&((r)->r_mtx)) #define NFSUNLOCKREQUEST(r) mtx_unlock(&((r)->r_mtx)) #define NFSPROCLISTLOCK() sx_slock(&allproc_lock) #define NFSPROCLISTUNLOCK() sx_sunlock(&allproc_lock) #define NFSLOCKSOCKREQ(r) mtx_lock(&((r)->nr_mtx)) #define NFSUNLOCKSOCKREQ(r) mtx_unlock(&((r)->nr_mtx)) #define NFSLOCKDS(d) mtx_lock(&((d)->nfsclds_mtx)) #define NFSUNLOCKDS(d) mtx_unlock(&((d)->nfsclds_mtx)) #define NFSSESSIONMUTEXPTR(s) (&((s)->mtx)) #define NFSLOCKSESSION(s) mtx_lock(&((s)->mtx)) #define NFSUNLOCKSESSION(s) mtx_unlock(&((s)->mtx)) /* * Use these macros to initialize/free a mutex. */ #define NFSINITSOCKMUTEX(m) mtx_init((m), "nfssock", NULL, MTX_DEF) #define NFSFREEMUTEX(m) mtx_destroy((m)) int nfsmsleep(void *, void *, int, const char *, struct timespec *); /* * And weird vm stuff in the nfs server. */ #define PDIRUNLOCK 0x0 #define MAX_COMMIT_COUNT (1024 * 1024) /* * Define these to handle the type of va_rdev. */ #define NFSMAKEDEV(m, n) makedev((m), (n)) #define NFSMAJOR(d) major(d) #define NFSMINOR(d) minor(d) /* * Define this to be the macro that returns the minimum size required * for a directory entry. */ #define DIRENT_SIZE(dp) GENERIC_DIRSIZ(dp) /* * The vnode tag for nfsv4root. */ #define VT_NFSV4ROOT "nfsv4root" /* * Define whatever it takes to do a vn_rdwr(). */ #define NFSD_RDWR(r, v, b, l, o, s, i, c, a, p) \ vn_rdwr((r), (v), (b), (l), (o), (s), (i), (c), NULL, (a), (p)) /* * Macros for handling memory for different BSDen. * NFSBCOPY(src, dst, len) - copies len bytes, non-overlapping * NFSOVBCOPY(src, dst, len) - ditto, but data areas might overlap * NFSBCMP(cp1, cp2, len) - compare len bytes, return 0 if same * NFSBZERO(cp, len) - set len bytes to 0x0 */ #define NFSBCOPY(s, d, l) bcopy((s), (d), (l)) #define NFSOVBCOPY(s, d, l) ovbcopy((s), (d), (l)) #define NFSBCMP(s, d, l) bcmp((s), (d), (l)) #define NFSBZERO(s, l) bzero((s), (l)) /* * Some queue.h files don't have these dfined in them. */ #define LIST_END(head) NULL #define SLIST_END(head) NULL #define TAILQ_END(head) NULL /* * This must be defined to be a global variable that increments once * per second, but never stops or goes backwards, even when a "date" * command changes the TOD clock. It is used for delta times for * leases, etc. */ #define NFSD_MONOSEC time_uptime /* * Declare the malloc types. */ MALLOC_DECLARE(M_NEWNFSRVCACHE); MALLOC_DECLARE(M_NEWNFSDCLIENT); MALLOC_DECLARE(M_NEWNFSDSTATE); MALLOC_DECLARE(M_NEWNFSDLOCK); MALLOC_DECLARE(M_NEWNFSDLOCKFILE); MALLOC_DECLARE(M_NEWNFSSTRING); MALLOC_DECLARE(M_NEWNFSUSERGROUP); MALLOC_DECLARE(M_NEWNFSDREQ); MALLOC_DECLARE(M_NEWNFSFH); MALLOC_DECLARE(M_NEWNFSCLOWNER); MALLOC_DECLARE(M_NEWNFSCLOPEN); MALLOC_DECLARE(M_NEWNFSCLDELEG); MALLOC_DECLARE(M_NEWNFSCLCLIENT); MALLOC_DECLARE(M_NEWNFSCLLOCKOWNER); MALLOC_DECLARE(M_NEWNFSCLLOCK); MALLOC_DECLARE(M_NEWNFSDIROFF); MALLOC_DECLARE(M_NEWNFSV4NODE); MALLOC_DECLARE(M_NEWNFSDIRECTIO); MALLOC_DECLARE(M_NEWNFSMNT); MALLOC_DECLARE(M_NEWNFSDROLLBACK); MALLOC_DECLARE(M_NEWNFSLAYOUT); MALLOC_DECLARE(M_NEWNFSFLAYOUT); MALLOC_DECLARE(M_NEWNFSDEVINFO); MALLOC_DECLARE(M_NEWNFSSOCKREQ); MALLOC_DECLARE(M_NEWNFSCLDS); MALLOC_DECLARE(M_NEWNFSLAYRECALL); MALLOC_DECLARE(M_NEWNFSDSESSION); #define M_NFSRVCACHE M_NEWNFSRVCACHE #define M_NFSDCLIENT M_NEWNFSDCLIENT #define M_NFSDSTATE M_NEWNFSDSTATE #define M_NFSDLOCK M_NEWNFSDLOCK #define M_NFSDLOCKFILE M_NEWNFSDLOCKFILE #define M_NFSSTRING M_NEWNFSSTRING #define M_NFSUSERGROUP M_NEWNFSUSERGROUP #define M_NFSDREQ M_NEWNFSDREQ #define M_NFSFH M_NEWNFSFH #define M_NFSCLOWNER M_NEWNFSCLOWNER #define M_NFSCLOPEN M_NEWNFSCLOPEN #define M_NFSCLDELEG M_NEWNFSCLDELEG #define M_NFSCLCLIENT M_NEWNFSCLCLIENT #define M_NFSCLLOCKOWNER M_NEWNFSCLLOCKOWNER #define M_NFSCLLOCK M_NEWNFSCLLOCK #define M_NFSDIROFF M_NEWNFSDIROFF #define M_NFSV4NODE M_NEWNFSV4NODE #define M_NFSDIRECTIO M_NEWNFSDIRECTIO #define M_NFSDROLLBACK M_NEWNFSDROLLBACK #define M_NFSLAYOUT M_NEWNFSLAYOUT #define M_NFSFLAYOUT M_NEWNFSFLAYOUT #define M_NFSDEVINFO M_NEWNFSDEVINFO #define M_NFSSOCKREQ M_NEWNFSSOCKREQ #define M_NFSCLDS M_NEWNFSCLDS #define M_NFSLAYRECALL M_NEWNFSLAYRECALL #define M_NFSDSESSION M_NEWNFSDSESSION #define NFSINT_SIGMASK(set) \ (SIGISMEMBER(set, SIGINT) || SIGISMEMBER(set, SIGTERM) || \ SIGISMEMBER(set, SIGHUP) || SIGISMEMBER(set, SIGKILL) || \ SIGISMEMBER(set, SIGQUIT)) /* * Convert a quota block count to byte count. */ #define NFSQUOTABLKTOBYTE(q, b) (q) *= (b) /* * Define this as the largest file size supported. (It should probably * be available via a VFS_xxx Op, but it isn't. */ #define NFSRV_MAXFILESIZE ((u_int64_t)0x800000000000) /* * Set this macro to index() or strchr(), whichever is supported. */ #define STRCHR(s, c) strchr((s), (c)) /* * Set the n_time in the client write rpc, as required. */ #define NFSWRITERPC_SETTIME(w, n, v4) \ do { \ if (w) { \ (n)->n_mtime = (n)->n_vattr.na_vattr.va_mtime; \ if (v4) \ (n)->n_change = (n)->n_vattr.na_vattr.va_filerev; \ } \ } while (0) /* * Fake value, just to make the client work. */ #define NFS_LATTR_NOSHRINK 1 /* * Prototypes for functions where the arguments vary for different ports. */ int nfscl_loadattrcache(struct vnode **, struct nfsvattr *, void *, void *, int, int); int newnfs_realign(struct mbuf **, int); /* * If the port runs on an SMP box that can enforce Atomic ops with low * overheads, define these as atomic increments/decrements. If not, * don't worry about it, since these are used for stats that can be * "out by one" without disastrous consequences. */ #define NFSINCRGLOBAL(a) ((a)++) /* * Assorted funky stuff to make things work under Darwin8. */ /* * These macros checks for a field in vattr being set. */ #define NFSATTRISSET(t, v, a) ((v)->a != (t)VNOVAL) #define NFSATTRISSETTIME(v, a) ((v)->a.tv_sec != VNOVAL) /* * Manipulate mount flags. */ #define NFSSTA_HASWRITEVERF 0x00040000 /* Has write verifier */ #define NFSSTA_GOTFSINFO 0x00100000 /* Got the fsinfo */ #define NFSSTA_NOLAYOUTCOMMIT 0x04000000 /* Don't do LayoutCommit */ #define NFSSTA_SESSPERSIST 0x08000000 /* Has a persistent session */ #define NFSSTA_TIMEO 0x10000000 /* Experiencing a timeout */ #define NFSSTA_LOCKTIMEO 0x20000000 /* Experiencing a lockd timeout */ #define NFSSTA_HASSETFSID 0x40000000 /* Has set the fsid */ #define NFSSTA_PNFS 0x80000000 /* pNFS is enabled */ #define NFSHASNFSV3(n) ((n)->nm_flag & NFSMNT_NFSV3) #define NFSHASNFSV4(n) ((n)->nm_flag & NFSMNT_NFSV4) #define NFSHASNFSV4N(n) ((n)->nm_minorvers > 0) #define NFSHASNFSV3OR4(n) ((n)->nm_flag & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) #define NFSHASGOTFSINFO(n) ((n)->nm_state & NFSSTA_GOTFSINFO) #define NFSHASHASSETFSID(n) ((n)->nm_state & NFSSTA_HASSETFSID) #define NFSHASSTRICT3530(n) ((n)->nm_flag & NFSMNT_STRICT3530) #define NFSHASWRITEVERF(n) ((n)->nm_state & NFSSTA_HASWRITEVERF) #define NFSHASINT(n) ((n)->nm_flag & NFSMNT_INT) #define NFSHASSOFT(n) ((n)->nm_flag & NFSMNT_SOFT) #define NFSHASINTORSOFT(n) ((n)->nm_flag & (NFSMNT_INT | NFSMNT_SOFT)) #define NFSHASDUMBTIMR(n) ((n)->nm_flag & NFSMNT_DUMBTIMR) #define NFSHASNOCONN(n) ((n)->nm_flag & NFSMNT_MNTD) #define NFSHASKERB(n) ((n)->nm_flag & NFSMNT_KERB) #define NFSHASALLGSSNAME(n) ((n)->nm_flag & NFSMNT_ALLGSSNAME) #define NFSHASINTEGRITY(n) ((n)->nm_flag & NFSMNT_INTEGRITY) #define NFSHASPRIVACY(n) ((n)->nm_flag & NFSMNT_PRIVACY) #define NFSSETWRITEVERF(n) ((n)->nm_state |= NFSSTA_HASWRITEVERF) #define NFSSETHASSETFSID(n) ((n)->nm_state |= NFSSTA_HASSETFSID) #define NFSHASPNFSOPT(n) ((n)->nm_flag & NFSMNT_PNFS) #define NFSHASNOLAYOUTCOMMIT(n) ((n)->nm_state & NFSSTA_NOLAYOUTCOMMIT) #define NFSHASSESSPERSIST(n) ((n)->nm_state & NFSSTA_SESSPERSIST) #define NFSHASPNFS(n) ((n)->nm_state & NFSSTA_PNFS) /* * Gets the stats field out of the mount structure. */ #define vfs_statfs(m) (&((m)->mnt_stat)) /* * Set boottime. */ #define NFSSETBOOTTIME(b) ((b) = boottime) /* * The size of directory blocks in the buffer cache. * MUST BE in the range of PAGE_SIZE <= NFS_DIRBLKSIZ <= MAXBSIZE!! */ #define NFS_DIRBLKSIZ (16 * DIRBLKSIZ) /* Must be a multiple of DIRBLKSIZ */ /* * Define these macros to access mnt_flag fields. */ #define NFSMNT_RDONLY(m) ((m)->mnt_flag & MNT_RDONLY) #endif /* _KERNEL */ /* * Define a structure similar to ufs_args for use in exporting the V4 root. */ struct nfsex_args { char *fspec; struct export_args export; }; /* * These export flags should be defined, but there are no bits left. * Maybe a separate mnt_exflag field could be added or the mnt_flag * field increased to 64 bits? */ #ifndef MNT_EXSTRICTACCESS #define MNT_EXSTRICTACCESS 0x0 #endif #ifndef MNT_EXV4ONLY #define MNT_EXV4ONLY 0x0 #endif #ifdef _KERNEL /* * Define this to invalidate the attribute cache for the nfs node. */ #define NFSINVALATTRCACHE(n) ((n)->n_attrstamp = 0) /* Used for FreeBSD only */ void nfsd_mntinit(void); /* * Define these for vnode lock/unlock ops. * * These are good abstractions to macro out, so that they can be added to * later, for debugging or stats, etc. */ #define NFSVOPLOCK(v, f) vn_lock((v), (f)) #define NFSVOPUNLOCK(v, f) VOP_UNLOCK((v), (f)) #define NFSVOPISLOCKED(v) VOP_ISLOCKED((v)) /* * Define ncl_hash(). */ #define ncl_hash(f, l) (fnv_32_buf((f), (l), FNV1_32_INIT)) int newnfs_iosize(struct nfsmount *); int newnfs_vncmpf(struct vnode *, void *); #ifndef NFS_MINDIRATTRTIMO #define NFS_MINDIRATTRTIMO 3 /* VDIR attrib cache timeout in sec */ #endif #ifndef NFS_MAXDIRATTRTIMO #define NFS_MAXDIRATTRTIMO 60 #endif /* * Nfs outstanding request list element */ struct nfsreq { TAILQ_ENTRY(nfsreq) r_chain; u_int32_t r_flags; /* flags on request, see below */ struct nfsmount *r_nmp; /* Client mnt ptr */ struct mtx r_mtx; /* Mutex lock for this structure */ }; #ifndef NFS_MAXBSIZE -#define NFS_MAXBSIZE MAXBSIZE +#define NFS_MAXBSIZE MAXBCACHEBUF #endif /* * This macro checks to see if issuing of delegations is allowed for this * vnode. */ #ifdef VV_DISABLEDELEG #define NFSVNO_DELEGOK(v) \ ((v) == NULL || ((v)->v_vflag & VV_DISABLEDELEG) == 0) #else #define NFSVNO_DELEGOK(v) (1) #endif #endif /* _KERNEL */ #endif /* _NFS_NFSPORT_H */ Index: user/ngie/more-tests/sys/fs/nfsclient/nfs_clvfsops.c =================================================================== --- user/ngie/more-tests/sys/fs/nfsclient/nfs_clvfsops.c (revision 281974) +++ user/ngie/more-tests/sys/fs/nfsclient/nfs_clvfsops.c (revision 281975) @@ -1,1880 +1,1880 @@ /*- * Copyright (c) 1989, 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Rick Macklem at The University of Guelph. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from nfs_vfsops.c 8.12 (Berkeley) 5/20/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_bootp.h" #include "opt_nfsroot.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include FEATURE(nfscl, "NFSv4 client"); extern int nfscl_ticks; extern struct timeval nfsboottime; extern struct nfsstats newnfsstats; extern int nfsrv_useacl; extern int nfscl_debuglevel; extern enum nfsiod_state ncl_iodwant[NFS_MAXASYNCDAEMON]; extern struct nfsmount *ncl_iodmount[NFS_MAXASYNCDAEMON]; extern struct mtx ncl_iod_mutex; NFSCLSTATEMUTEX; MALLOC_DEFINE(M_NEWNFSREQ, "newnfsclient_req", "NFS request header"); MALLOC_DEFINE(M_NEWNFSMNT, "newnfsmnt", "NFS mount struct"); SYSCTL_DECL(_vfs_nfs); static int nfs_ip_paranoia = 1; SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_ip_paranoia, CTLFLAG_RW, &nfs_ip_paranoia, 0, ""); static int nfs_tprintf_initial_delay = NFS_TPRINTF_INITIAL_DELAY; SYSCTL_INT(_vfs_nfs, NFS_TPRINTF_INITIAL_DELAY, downdelayinitial, CTLFLAG_RW, &nfs_tprintf_initial_delay, 0, ""); /* how long between console messages "nfs server foo not responding" */ static int nfs_tprintf_delay = NFS_TPRINTF_DELAY; SYSCTL_INT(_vfs_nfs, NFS_TPRINTF_DELAY, downdelayinterval, CTLFLAG_RW, &nfs_tprintf_delay, 0, ""); #ifdef NFS_DEBUG int nfs_debug; SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "Toggle debug flag"); #endif static int nfs_mountroot(struct mount *); static void nfs_sec_name(char *, int *); static void nfs_decode_args(struct mount *mp, struct nfsmount *nmp, struct nfs_args *argp, const char *, struct ucred *, struct thread *); static int mountnfs(struct nfs_args *, struct mount *, struct sockaddr *, char *, u_char *, int, u_char *, int, u_char *, int, struct vnode **, struct ucred *, struct thread *, int, int, int); static void nfs_getnlminfo(struct vnode *, uint8_t *, size_t *, struct sockaddr_storage *, int *, off_t *, struct timeval *); static vfs_mount_t nfs_mount; static vfs_cmount_t nfs_cmount; static vfs_unmount_t nfs_unmount; static vfs_root_t nfs_root; static vfs_statfs_t nfs_statfs; static vfs_sync_t nfs_sync; static vfs_sysctl_t nfs_sysctl; static vfs_purge_t nfs_purge; /* * nfs vfs operations. */ static struct vfsops nfs_vfsops = { .vfs_init = ncl_init, .vfs_mount = nfs_mount, .vfs_cmount = nfs_cmount, .vfs_root = nfs_root, .vfs_statfs = nfs_statfs, .vfs_sync = nfs_sync, .vfs_uninit = ncl_uninit, .vfs_unmount = nfs_unmount, .vfs_sysctl = nfs_sysctl, .vfs_purge = nfs_purge, }; VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK | VFCF_SBDRY); /* So that loader and kldload(2) can find us, wherever we are.. */ MODULE_VERSION(nfs, 1); MODULE_DEPEND(nfs, nfscommon, 1, 1, 1); MODULE_DEPEND(nfs, krpc, 1, 1, 1); MODULE_DEPEND(nfs, nfssvc, 1, 1, 1); MODULE_DEPEND(nfs, nfslock, 1, 1, 1); /* * This structure is now defined in sys/nfs/nfs_diskless.c so that it * can be shared by both NFS clients. It is declared here so that it * will be defined for kernels built without NFS_ROOT, although it * isn't used in that case. */ #if !defined(NFS_ROOT) struct nfs_diskless nfs_diskless = { { { 0 } } }; struct nfsv3_diskless nfsv3_diskless = { { { 0 } } }; int nfs_diskless_valid = 0; #endif SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD, &nfs_diskless_valid, 0, "Has the diskless struct been filled correctly"); SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD, nfsv3_diskless.root_hostnam, 0, "Path to nfs root"); SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD, &nfsv3_diskless.root_saddr, sizeof(nfsv3_diskless.root_saddr), "%Ssockaddr_in", "Diskless root nfs address"); void newnfsargs_ntoh(struct nfs_args *); static int nfs_mountdiskless(char *, struct sockaddr_in *, struct nfs_args *, struct thread *, struct vnode **, struct mount *); static void nfs_convert_diskless(void); static void nfs_convert_oargs(struct nfs_args *args, struct onfs_args *oargs); int newnfs_iosize(struct nfsmount *nmp) { int iosize, maxio; /* First, set the upper limit for iosize */ if (nmp->nm_flag & NFSMNT_NFSV4) { maxio = NFS_MAXBSIZE; } else if (nmp->nm_flag & NFSMNT_NFSV3) { if (nmp->nm_sotype == SOCK_DGRAM) maxio = NFS_MAXDGRAMDATA; else maxio = NFS_MAXBSIZE; } else { maxio = NFS_V2MAXDATA; } if (nmp->nm_rsize > maxio || nmp->nm_rsize == 0) nmp->nm_rsize = maxio; - if (nmp->nm_rsize > MAXBSIZE) - nmp->nm_rsize = MAXBSIZE; + if (nmp->nm_rsize > NFS_MAXBSIZE) + nmp->nm_rsize = NFS_MAXBSIZE; if (nmp->nm_readdirsize > maxio || nmp->nm_readdirsize == 0) nmp->nm_readdirsize = maxio; if (nmp->nm_readdirsize > nmp->nm_rsize) nmp->nm_readdirsize = nmp->nm_rsize; if (nmp->nm_wsize > maxio || nmp->nm_wsize == 0) nmp->nm_wsize = maxio; - if (nmp->nm_wsize > MAXBSIZE) - nmp->nm_wsize = MAXBSIZE; + if (nmp->nm_wsize > NFS_MAXBSIZE) + nmp->nm_wsize = NFS_MAXBSIZE; /* * Calculate the size used for io buffers. Use the larger * of the two sizes to minimise nfs requests but make sure * that it is at least one VM page to avoid wasting buffer * space. */ iosize = imax(nmp->nm_rsize, nmp->nm_wsize); iosize = imax(iosize, PAGE_SIZE); nmp->nm_mountp->mnt_stat.f_iosize = iosize; return (iosize); } static void nfs_convert_oargs(struct nfs_args *args, struct onfs_args *oargs) { args->version = NFS_ARGSVERSION; args->addr = oargs->addr; args->addrlen = oargs->addrlen; args->sotype = oargs->sotype; args->proto = oargs->proto; args->fh = oargs->fh; args->fhsize = oargs->fhsize; args->flags = oargs->flags; args->wsize = oargs->wsize; args->rsize = oargs->rsize; args->readdirsize = oargs->readdirsize; args->timeo = oargs->timeo; args->retrans = oargs->retrans; args->readahead = oargs->readahead; args->hostname = oargs->hostname; } static void nfs_convert_diskless(void) { bcopy(&nfs_diskless.myif, &nfsv3_diskless.myif, sizeof(struct ifaliasreq)); bcopy(&nfs_diskless.mygateway, &nfsv3_diskless.mygateway, sizeof(struct sockaddr_in)); nfs_convert_oargs(&nfsv3_diskless.root_args,&nfs_diskless.root_args); if (nfsv3_diskless.root_args.flags & NFSMNT_NFSV3) { nfsv3_diskless.root_fhsize = NFSX_MYFH; bcopy(nfs_diskless.root_fh, nfsv3_diskless.root_fh, NFSX_MYFH); } else { nfsv3_diskless.root_fhsize = NFSX_V2FH; bcopy(nfs_diskless.root_fh, nfsv3_diskless.root_fh, NFSX_V2FH); } bcopy(&nfs_diskless.root_saddr,&nfsv3_diskless.root_saddr, sizeof(struct sockaddr_in)); bcopy(nfs_diskless.root_hostnam, nfsv3_diskless.root_hostnam, MNAMELEN); nfsv3_diskless.root_time = nfs_diskless.root_time; bcopy(nfs_diskless.my_hostnam, nfsv3_diskless.my_hostnam, MAXHOSTNAMELEN); nfs_diskless_valid = 3; } /* * nfs statfs call */ static int nfs_statfs(struct mount *mp, struct statfs *sbp) { struct vnode *vp; struct thread *td; struct nfsmount *nmp = VFSTONFS(mp); struct nfsvattr nfsva; struct nfsfsinfo fs; struct nfsstatfs sb; int error = 0, attrflag, gotfsinfo = 0, ret; struct nfsnode *np; td = curthread; error = vfs_busy(mp, MBF_NOWAIT); if (error) return (error); error = ncl_nget(mp, nmp->nm_fh, nmp->nm_fhsize, &np, LK_EXCLUSIVE); if (error) { vfs_unbusy(mp); return (error); } vp = NFSTOV(np); mtx_lock(&nmp->nm_mtx); if (NFSHASNFSV3(nmp) && !NFSHASGOTFSINFO(nmp)) { mtx_unlock(&nmp->nm_mtx); error = nfsrpc_fsinfo(vp, &fs, td->td_ucred, td, &nfsva, &attrflag, NULL); if (!error) gotfsinfo = 1; } else mtx_unlock(&nmp->nm_mtx); if (!error) error = nfsrpc_statfs(vp, &sb, &fs, td->td_ucred, td, &nfsva, &attrflag, NULL); if (error != 0) NFSCL_DEBUG(2, "statfs=%d\n", error); if (attrflag == 0) { ret = nfsrpc_getattrnovp(nmp, nmp->nm_fh, nmp->nm_fhsize, 1, td->td_ucred, td, &nfsva, NULL, NULL); if (ret) { /* * Just set default values to get things going. */ NFSBZERO((caddr_t)&nfsva, sizeof (struct nfsvattr)); nfsva.na_vattr.va_type = VDIR; nfsva.na_vattr.va_mode = 0777; nfsva.na_vattr.va_nlink = 100; nfsva.na_vattr.va_uid = (uid_t)0; nfsva.na_vattr.va_gid = (gid_t)0; nfsva.na_vattr.va_fileid = 2; nfsva.na_vattr.va_gen = 1; nfsva.na_vattr.va_blocksize = NFS_FABLKSIZE; nfsva.na_vattr.va_size = 512 * 1024; } } (void) nfscl_loadattrcache(&vp, &nfsva, NULL, NULL, 0, 1); if (!error) { mtx_lock(&nmp->nm_mtx); if (gotfsinfo || (nmp->nm_flag & NFSMNT_NFSV4)) nfscl_loadfsinfo(nmp, &fs); nfscl_loadsbinfo(nmp, &sb, sbp); sbp->f_iosize = newnfs_iosize(nmp); mtx_unlock(&nmp->nm_mtx); if (sbp != &mp->mnt_stat) { bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); } strncpy(&sbp->f_fstypename[0], mp->mnt_vfc->vfc_name, MFSNAMELEN); } else if (NFS_ISV4(vp)) { error = nfscl_maperr(td, error, (uid_t)0, (gid_t)0); } vput(vp); vfs_unbusy(mp); return (error); } /* * nfs version 3 fsinfo rpc call */ int ncl_fsinfo(struct nfsmount *nmp, struct vnode *vp, struct ucred *cred, struct thread *td) { struct nfsfsinfo fs; struct nfsvattr nfsva; int error, attrflag; error = nfsrpc_fsinfo(vp, &fs, cred, td, &nfsva, &attrflag, NULL); if (!error) { if (attrflag) (void) nfscl_loadattrcache(&vp, &nfsva, NULL, NULL, 0, 1); mtx_lock(&nmp->nm_mtx); nfscl_loadfsinfo(nmp, &fs); mtx_unlock(&nmp->nm_mtx); } return (error); } /* * Mount a remote root fs via. nfs. This depends on the info in the * nfs_diskless structure that has been filled in properly by some primary * bootstrap. * It goes something like this: * - do enough of "ifconfig" by calling ifioctl() so that the system * can talk to the server * - If nfs_diskless.mygateway is filled in, use that address as * a default gateway. * - build the rootfs mount point and call mountnfs() to do the rest. * * It is assumed to be safe to read, modify, and write the nfsv3_diskless * structure, as well as other global NFS client variables here, as * nfs_mountroot() will be called once in the boot before any other NFS * client activity occurs. */ static int nfs_mountroot(struct mount *mp) { struct thread *td = curthread; struct nfsv3_diskless *nd = &nfsv3_diskless; struct socket *so; struct vnode *vp; struct ifreq ir; int error; u_long l; char buf[128]; char *cp; #if defined(BOOTP_NFSROOT) && defined(BOOTP) bootpc_init(); /* use bootp to get nfs_diskless filled in */ #elif defined(NFS_ROOT) nfs_setup_diskless(); #endif if (nfs_diskless_valid == 0) return (-1); if (nfs_diskless_valid == 1) nfs_convert_diskless(); /* * XXX splnet, so networks will receive... */ splnet(); /* * Do enough of ifconfig(8) so that the critical net interface can * talk to the server. */ error = socreate(nd->myif.ifra_addr.sa_family, &so, nd->root_args.sotype, 0, td->td_ucred, td); if (error) panic("nfs_mountroot: socreate(%04x): %d", nd->myif.ifra_addr.sa_family, error); #if 0 /* XXX Bad idea */ /* * We might not have been told the right interface, so we pass * over the first ten interfaces of the same kind, until we get * one of them configured. */ for (i = strlen(nd->myif.ifra_name) - 1; nd->myif.ifra_name[i] >= '0' && nd->myif.ifra_name[i] <= '9'; nd->myif.ifra_name[i] ++) { error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td); if(!error) break; } #endif error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td); if (error) panic("nfs_mountroot: SIOCAIFADDR: %d", error); if ((cp = kern_getenv("boot.netif.mtu")) != NULL) { ir.ifr_mtu = strtol(cp, NULL, 10); bcopy(nd->myif.ifra_name, ir.ifr_name, IFNAMSIZ); freeenv(cp); error = ifioctl(so, SIOCSIFMTU, (caddr_t)&ir, td); if (error) printf("nfs_mountroot: SIOCSIFMTU: %d", error); } soclose(so); /* * If the gateway field is filled in, set it as the default route. * Note that pxeboot will set a default route of 0 if the route * is not set by the DHCP server. Check also for a value of 0 * to avoid panicking inappropriately in that situation. */ if (nd->mygateway.sin_len != 0 && nd->mygateway.sin_addr.s_addr != 0) { struct sockaddr_in mask, sin; bzero((caddr_t)&mask, sizeof(mask)); sin = mask; sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); /* XXX MRT use table 0 for this sort of thing */ CURVNET_SET(TD_TO_VNET(td)); error = rtrequest_fib(RTM_ADD, (struct sockaddr *)&sin, (struct sockaddr *)&nd->mygateway, (struct sockaddr *)&mask, RTF_UP | RTF_GATEWAY, NULL, RT_DEFAULT_FIB); CURVNET_RESTORE(); if (error) panic("nfs_mountroot: RTM_ADD: %d", error); } /* * Create the rootfs mount point. */ nd->root_args.fh = nd->root_fh; nd->root_args.fhsize = nd->root_fhsize; l = ntohl(nd->root_saddr.sin_addr.s_addr); snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s", (l >> 24) & 0xff, (l >> 16) & 0xff, (l >> 8) & 0xff, (l >> 0) & 0xff, nd->root_hostnam); printf("NFS ROOT: %s\n", buf); nd->root_args.hostname = buf; if ((error = nfs_mountdiskless(buf, &nd->root_saddr, &nd->root_args, td, &vp, mp)) != 0) { return (error); } /* * This is not really an nfs issue, but it is much easier to * set hostname here and then let the "/etc/rc.xxx" files * mount the right /var based upon its preset value. */ mtx_lock(&prison0.pr_mtx); strlcpy(prison0.pr_hostname, nd->my_hostnam, sizeof(prison0.pr_hostname)); mtx_unlock(&prison0.pr_mtx); inittodr(ntohl(nd->root_time)); return (0); } /* * Internal version of mount system call for diskless setup. */ static int nfs_mountdiskless(char *path, struct sockaddr_in *sin, struct nfs_args *args, struct thread *td, struct vnode **vpp, struct mount *mp) { struct sockaddr *nam; int dirlen, error; char *dirpath; /* * Find the directory path in "path", which also has the server's * name/ip address in it. */ dirpath = strchr(path, ':'); if (dirpath != NULL) dirlen = strlen(++dirpath); else dirlen = 0; nam = sodupsockaddr((struct sockaddr *)sin, M_WAITOK); if ((error = mountnfs(args, mp, nam, path, NULL, 0, dirpath, dirlen, NULL, 0, vpp, td->td_ucred, td, NFS_DEFAULT_NAMETIMEO, NFS_DEFAULT_NEGNAMETIMEO, 0)) != 0) { printf("nfs_mountroot: mount %s on /: %d\n", path, error); return (error); } return (0); } static void nfs_sec_name(char *sec, int *flagsp) { if (!strcmp(sec, "krb5")) *flagsp |= NFSMNT_KERB; else if (!strcmp(sec, "krb5i")) *flagsp |= (NFSMNT_KERB | NFSMNT_INTEGRITY); else if (!strcmp(sec, "krb5p")) *flagsp |= (NFSMNT_KERB | NFSMNT_PRIVACY); } static void nfs_decode_args(struct mount *mp, struct nfsmount *nmp, struct nfs_args *argp, const char *hostname, struct ucred *cred, struct thread *td) { int s; int adjsock; char *p; s = splnet(); /* * Set read-only flag if requested; otherwise, clear it if this is * an update. If this is not an update, then either the read-only * flag is already clear, or this is a root mount and it was set * intentionally at some previous point. */ if (vfs_getopt(mp->mnt_optnew, "ro", NULL, NULL) == 0) { MNT_ILOCK(mp); mp->mnt_flag |= MNT_RDONLY; MNT_IUNLOCK(mp); } else if (mp->mnt_flag & MNT_UPDATE) { MNT_ILOCK(mp); mp->mnt_flag &= ~MNT_RDONLY; MNT_IUNLOCK(mp); } /* * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes * no sense in that context. Also, set up appropriate retransmit * and soft timeout behavior. */ if (argp->sotype == SOCK_STREAM) { nmp->nm_flag &= ~NFSMNT_NOCONN; nmp->nm_timeo = NFS_MAXTIMEO; if ((argp->flags & NFSMNT_NFSV4) != 0) nmp->nm_retry = INT_MAX; else nmp->nm_retry = NFS_RETRANS_TCP; } /* Also clear RDIRPLUS if NFSv2, it crashes some servers */ if ((argp->flags & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) == 0) { argp->flags &= ~NFSMNT_RDIRPLUS; nmp->nm_flag &= ~NFSMNT_RDIRPLUS; } /* Re-bind if rsrvd port requested and wasn't on one */ adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT) && (argp->flags & NFSMNT_RESVPORT); /* Also re-bind if we're switching to/from a connected UDP socket */ adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) != (argp->flags & NFSMNT_NOCONN)); /* Update flags atomically. Don't change the lock bits. */ nmp->nm_flag = argp->flags | nmp->nm_flag; splx(s); if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) { nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10; if (nmp->nm_timeo < NFS_MINTIMEO) nmp->nm_timeo = NFS_MINTIMEO; else if (nmp->nm_timeo > NFS_MAXTIMEO) nmp->nm_timeo = NFS_MAXTIMEO; } if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) { nmp->nm_retry = argp->retrans; if (nmp->nm_retry > NFS_MAXREXMIT) nmp->nm_retry = NFS_MAXREXMIT; } if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) { nmp->nm_wsize = argp->wsize; /* * Clip at the power of 2 below the size. There is an * issue (not isolated) that causes intermittent page * faults if this is not done. */ if (nmp->nm_wsize > NFS_FABLKSIZE) nmp->nm_wsize = 1 << (fls(nmp->nm_wsize) - 1); else nmp->nm_wsize = NFS_FABLKSIZE; } if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) { nmp->nm_rsize = argp->rsize; /* * Clip at the power of 2 below the size. There is an * issue (not isolated) that causes intermittent page * faults if this is not done. */ if (nmp->nm_rsize > NFS_FABLKSIZE) nmp->nm_rsize = 1 << (fls(nmp->nm_rsize) - 1); else nmp->nm_rsize = NFS_FABLKSIZE; } if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) { nmp->nm_readdirsize = argp->readdirsize; } if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0) nmp->nm_acregmin = argp->acregmin; else nmp->nm_acregmin = NFS_MINATTRTIMO; if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0) nmp->nm_acregmax = argp->acregmax; else nmp->nm_acregmax = NFS_MAXATTRTIMO; if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0) nmp->nm_acdirmin = argp->acdirmin; else nmp->nm_acdirmin = NFS_MINDIRATTRTIMO; if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0) nmp->nm_acdirmax = argp->acdirmax; else nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO; if (nmp->nm_acdirmin > nmp->nm_acdirmax) nmp->nm_acdirmin = nmp->nm_acdirmax; if (nmp->nm_acregmin > nmp->nm_acregmax) nmp->nm_acregmin = nmp->nm_acregmax; if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) { if (argp->readahead <= NFS_MAXRAHEAD) nmp->nm_readahead = argp->readahead; else nmp->nm_readahead = NFS_MAXRAHEAD; } if ((argp->flags & NFSMNT_WCOMMITSIZE) && argp->wcommitsize >= 0) { if (argp->wcommitsize < nmp->nm_wsize) nmp->nm_wcommitsize = nmp->nm_wsize; else nmp->nm_wcommitsize = argp->wcommitsize; } adjsock |= ((nmp->nm_sotype != argp->sotype) || (nmp->nm_soproto != argp->proto)); if (nmp->nm_client != NULL && adjsock) { int haslock = 0, error = 0; if (nmp->nm_sotype == SOCK_STREAM) { error = newnfs_sndlock(&nmp->nm_sockreq.nr_lock); if (!error) haslock = 1; } if (!error) { newnfs_disconnect(&nmp->nm_sockreq); if (haslock) newnfs_sndunlock(&nmp->nm_sockreq.nr_lock); nmp->nm_sotype = argp->sotype; nmp->nm_soproto = argp->proto; if (nmp->nm_sotype == SOCK_DGRAM) while (newnfs_connect(nmp, &nmp->nm_sockreq, cred, td, 0)) { printf("newnfs_args: retrying connect\n"); (void) nfs_catnap(PSOCK, 0, "nfscon"); } } } else { nmp->nm_sotype = argp->sotype; nmp->nm_soproto = argp->proto; } if (hostname != NULL) { strlcpy(nmp->nm_hostname, hostname, sizeof(nmp->nm_hostname)); p = strchr(nmp->nm_hostname, ':'); if (p != NULL) *p = '\0'; } } static const char *nfs_opts[] = { "from", "nfs_args", "noac", "noatime", "noexec", "suiddir", "nosuid", "nosymfollow", "union", "noclusterr", "noclusterw", "multilabel", "acls", "force", "update", "async", "noconn", "nolockd", "conn", "lockd", "intr", "rdirplus", "readdirsize", "soft", "hard", "mntudp", "tcp", "udp", "wsize", "rsize", "retrans", "actimeo", "acregmin", "acregmax", "acdirmin", "acdirmax", "resvport", "readahead", "hostname", "timeo", "timeout", "addr", "fh", "nfsv3", "sec", "principal", "nfsv4", "gssname", "allgssname", "dirpath", "minorversion", "nametimeo", "negnametimeo", "nocto", "noncontigwr", "pnfs", "wcommitsize", NULL }; /* * VFS Operations. * * mount system call * It seems a bit dumb to copyinstr() the host and path here and then * bcopy() them in mountnfs(), but I wanted to detect errors before * doing the sockargs() call because sockargs() allocates an mbuf and * an error after that means that I have to release the mbuf. */ /* ARGSUSED */ static int nfs_mount(struct mount *mp) { struct nfs_args args = { .version = NFS_ARGSVERSION, .addr = NULL, .addrlen = sizeof (struct sockaddr_in), .sotype = SOCK_STREAM, .proto = 0, .fh = NULL, .fhsize = 0, .flags = NFSMNT_RESVPORT, .wsize = NFS_WSIZE, .rsize = NFS_RSIZE, .readdirsize = NFS_READDIRSIZE, .timeo = 10, .retrans = NFS_RETRANS, .readahead = NFS_DEFRAHEAD, .wcommitsize = 0, /* was: NQ_DEFLEASE */ .hostname = NULL, .acregmin = NFS_MINATTRTIMO, .acregmax = NFS_MAXATTRTIMO, .acdirmin = NFS_MINDIRATTRTIMO, .acdirmax = NFS_MAXDIRATTRTIMO, }; int error = 0, ret, len; struct sockaddr *nam = NULL; struct vnode *vp; struct thread *td; char hst[MNAMELEN]; u_char nfh[NFSX_FHMAX], krbname[100], dirpath[100], srvkrbname[100]; char *opt, *name, *secname; int nametimeo = NFS_DEFAULT_NAMETIMEO; int negnametimeo = NFS_DEFAULT_NEGNAMETIMEO; int minvers = 0; int dirlen, has_nfs_args_opt, krbnamelen, srvkrbnamelen; size_t hstlen; has_nfs_args_opt = 0; if (vfs_filteropt(mp->mnt_optnew, nfs_opts)) { error = EINVAL; goto out; } td = curthread; if ((mp->mnt_flag & (MNT_ROOTFS | MNT_UPDATE)) == MNT_ROOTFS) { error = nfs_mountroot(mp); goto out; } nfscl_init(); /* * The old mount_nfs program passed the struct nfs_args * from userspace to kernel. The new mount_nfs program * passes string options via nmount() from userspace to kernel * and we populate the struct nfs_args in the kernel. */ if (vfs_getopt(mp->mnt_optnew, "nfs_args", NULL, NULL) == 0) { error = vfs_copyopt(mp->mnt_optnew, "nfs_args", &args, sizeof(args)); if (error != 0) goto out; if (args.version != NFS_ARGSVERSION) { error = EPROGMISMATCH; goto out; } has_nfs_args_opt = 1; } /* Handle the new style options. */ if (vfs_getopt(mp->mnt_optnew, "noac", NULL, NULL) == 0) { args.acdirmin = args.acdirmax = args.acregmin = args.acregmax = 0; args.flags |= NFSMNT_ACDIRMIN | NFSMNT_ACDIRMAX | NFSMNT_ACREGMIN | NFSMNT_ACREGMAX; } if (vfs_getopt(mp->mnt_optnew, "noconn", NULL, NULL) == 0) args.flags |= NFSMNT_NOCONN; if (vfs_getopt(mp->mnt_optnew, "conn", NULL, NULL) == 0) args.flags &= ~NFSMNT_NOCONN; if (vfs_getopt(mp->mnt_optnew, "nolockd", NULL, NULL) == 0) args.flags |= NFSMNT_NOLOCKD; if (vfs_getopt(mp->mnt_optnew, "lockd", NULL, NULL) == 0) args.flags &= ~NFSMNT_NOLOCKD; if (vfs_getopt(mp->mnt_optnew, "intr", NULL, NULL) == 0) args.flags |= NFSMNT_INT; if (vfs_getopt(mp->mnt_optnew, "rdirplus", NULL, NULL) == 0) args.flags |= NFSMNT_RDIRPLUS; if (vfs_getopt(mp->mnt_optnew, "resvport", NULL, NULL) == 0) args.flags |= NFSMNT_RESVPORT; if (vfs_getopt(mp->mnt_optnew, "noresvport", NULL, NULL) == 0) args.flags &= ~NFSMNT_RESVPORT; if (vfs_getopt(mp->mnt_optnew, "soft", NULL, NULL) == 0) args.flags |= NFSMNT_SOFT; if (vfs_getopt(mp->mnt_optnew, "hard", NULL, NULL) == 0) args.flags &= ~NFSMNT_SOFT; if (vfs_getopt(mp->mnt_optnew, "mntudp", NULL, NULL) == 0) args.sotype = SOCK_DGRAM; if (vfs_getopt(mp->mnt_optnew, "udp", NULL, NULL) == 0) args.sotype = SOCK_DGRAM; if (vfs_getopt(mp->mnt_optnew, "tcp", NULL, NULL) == 0) args.sotype = SOCK_STREAM; if (vfs_getopt(mp->mnt_optnew, "nfsv3", NULL, NULL) == 0) args.flags |= NFSMNT_NFSV3; if (vfs_getopt(mp->mnt_optnew, "nfsv4", NULL, NULL) == 0) { args.flags |= NFSMNT_NFSV4; args.sotype = SOCK_STREAM; } if (vfs_getopt(mp->mnt_optnew, "allgssname", NULL, NULL) == 0) args.flags |= NFSMNT_ALLGSSNAME; if (vfs_getopt(mp->mnt_optnew, "nocto", NULL, NULL) == 0) args.flags |= NFSMNT_NOCTO; if (vfs_getopt(mp->mnt_optnew, "noncontigwr", NULL, NULL) == 0) args.flags |= NFSMNT_NONCONTIGWR; if (vfs_getopt(mp->mnt_optnew, "pnfs", NULL, NULL) == 0) args.flags |= NFSMNT_PNFS; if (vfs_getopt(mp->mnt_optnew, "readdirsize", (void **)&opt, NULL) == 0) { if (opt == NULL) { vfs_mount_error(mp, "illegal readdirsize"); error = EINVAL; goto out; } ret = sscanf(opt, "%d", &args.readdirsize); if (ret != 1 || args.readdirsize <= 0) { vfs_mount_error(mp, "illegal readdirsize: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_READDIRSIZE; } if (vfs_getopt(mp->mnt_optnew, "readahead", (void **)&opt, NULL) == 0) { if (opt == NULL) { vfs_mount_error(mp, "illegal readahead"); error = EINVAL; goto out; } ret = sscanf(opt, "%d", &args.readahead); if (ret != 1 || args.readahead <= 0) { vfs_mount_error(mp, "illegal readahead: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_READAHEAD; } if (vfs_getopt(mp->mnt_optnew, "wsize", (void **)&opt, NULL) == 0) { if (opt == NULL) { vfs_mount_error(mp, "illegal wsize"); error = EINVAL; goto out; } ret = sscanf(opt, "%d", &args.wsize); if (ret != 1 || args.wsize <= 0) { vfs_mount_error(mp, "illegal wsize: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_WSIZE; } if (vfs_getopt(mp->mnt_optnew, "rsize", (void **)&opt, NULL) == 0) { if (opt == NULL) { vfs_mount_error(mp, "illegal rsize"); error = EINVAL; goto out; } ret = sscanf(opt, "%d", &args.rsize); if (ret != 1 || args.rsize <= 0) { vfs_mount_error(mp, "illegal wsize: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_RSIZE; } if (vfs_getopt(mp->mnt_optnew, "retrans", (void **)&opt, NULL) == 0) { if (opt == NULL) { vfs_mount_error(mp, "illegal retrans"); error = EINVAL; goto out; } ret = sscanf(opt, "%d", &args.retrans); if (ret != 1 || args.retrans <= 0) { vfs_mount_error(mp, "illegal retrans: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_RETRANS; } if (vfs_getopt(mp->mnt_optnew, "actimeo", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.acregmin); if (ret != 1 || args.acregmin < 0) { vfs_mount_error(mp, "illegal actimeo: %s", opt); error = EINVAL; goto out; } args.acdirmin = args.acdirmax = args.acregmax = args.acregmin; args.flags |= NFSMNT_ACDIRMIN | NFSMNT_ACDIRMAX | NFSMNT_ACREGMIN | NFSMNT_ACREGMAX; } if (vfs_getopt(mp->mnt_optnew, "acregmin", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.acregmin); if (ret != 1 || args.acregmin < 0) { vfs_mount_error(mp, "illegal acregmin: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_ACREGMIN; } if (vfs_getopt(mp->mnt_optnew, "acregmax", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.acregmax); if (ret != 1 || args.acregmax < 0) { vfs_mount_error(mp, "illegal acregmax: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_ACREGMAX; } if (vfs_getopt(mp->mnt_optnew, "acdirmin", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.acdirmin); if (ret != 1 || args.acdirmin < 0) { vfs_mount_error(mp, "illegal acdirmin: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_ACDIRMIN; } if (vfs_getopt(mp->mnt_optnew, "acdirmax", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.acdirmax); if (ret != 1 || args.acdirmax < 0) { vfs_mount_error(mp, "illegal acdirmax: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_ACDIRMAX; } if (vfs_getopt(mp->mnt_optnew, "wcommitsize", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.wcommitsize); if (ret != 1 || args.wcommitsize < 0) { vfs_mount_error(mp, "illegal wcommitsize: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_WCOMMITSIZE; } if (vfs_getopt(mp->mnt_optnew, "timeo", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.timeo); if (ret != 1 || args.timeo <= 0) { vfs_mount_error(mp, "illegal timeo: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_TIMEO; } if (vfs_getopt(mp->mnt_optnew, "timeout", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &args.timeo); if (ret != 1 || args.timeo <= 0) { vfs_mount_error(mp, "illegal timeout: %s", opt); error = EINVAL; goto out; } args.flags |= NFSMNT_TIMEO; } if (vfs_getopt(mp->mnt_optnew, "nametimeo", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &nametimeo); if (ret != 1 || nametimeo < 0) { vfs_mount_error(mp, "illegal nametimeo: %s", opt); error = EINVAL; goto out; } } if (vfs_getopt(mp->mnt_optnew, "negnametimeo", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &negnametimeo); if (ret != 1 || negnametimeo < 0) { vfs_mount_error(mp, "illegal negnametimeo: %s", opt); error = EINVAL; goto out; } } if (vfs_getopt(mp->mnt_optnew, "minorversion", (void **)&opt, NULL) == 0) { ret = sscanf(opt, "%d", &minvers); if (ret != 1 || minvers < 0 || minvers > 1 || (args.flags & NFSMNT_NFSV4) == 0) { vfs_mount_error(mp, "illegal minorversion: %s", opt); error = EINVAL; goto out; } } if (vfs_getopt(mp->mnt_optnew, "sec", (void **) &secname, NULL) == 0) nfs_sec_name(secname, &args.flags); if (mp->mnt_flag & MNT_UPDATE) { struct nfsmount *nmp = VFSTONFS(mp); if (nmp == NULL) { error = EIO; goto out; } /* * If a change from TCP->UDP is done and there are thread(s) * that have I/O RPC(s) in progress with a tranfer size * greater than NFS_MAXDGRAMDATA, those thread(s) will be * hung, retrying the RPC(s) forever. Usually these threads * will be seen doing an uninterruptible sleep on wait channel * "nfsreq". */ if (args.sotype == SOCK_DGRAM && nmp->nm_sotype == SOCK_STREAM) tprintf(td->td_proc, LOG_WARNING, "Warning: mount -u that changes TCP->UDP can result in hung threads\n"); /* * When doing an update, we can't change version, * security, switch lockd strategies or change cookie * translation */ args.flags = (args.flags & ~(NFSMNT_NFSV3 | NFSMNT_NFSV4 | NFSMNT_KERB | NFSMNT_INTEGRITY | NFSMNT_PRIVACY | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/)) | (nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_NFSV4 | NFSMNT_KERB | NFSMNT_INTEGRITY | NFSMNT_PRIVACY | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/)); nfs_decode_args(mp, nmp, &args, NULL, td->td_ucred, td); goto out; } /* * Make the nfs_ip_paranoia sysctl serve as the default connection * or no-connection mode for those protocols that support * no-connection mode (the flag will be cleared later for protocols * that do not support no-connection mode). This will allow a client * to receive replies from a different IP then the request was * sent to. Note: default value for nfs_ip_paranoia is 1 (paranoid), * not 0. */ if (nfs_ip_paranoia == 0) args.flags |= NFSMNT_NOCONN; if (has_nfs_args_opt != 0) { /* * In the 'nfs_args' case, the pointers in the args * structure are in userland - we copy them in here. */ if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX) { vfs_mount_error(mp, "Bad file handle"); error = EINVAL; goto out; } error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize); if (error != 0) goto out; error = copyinstr(args.hostname, hst, MNAMELEN - 1, &hstlen); if (error != 0) goto out; bzero(&hst[hstlen], MNAMELEN - hstlen); args.hostname = hst; /* sockargs() call must be after above copyin() calls */ error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen); if (error != 0) goto out; } else { if (vfs_getopt(mp->mnt_optnew, "fh", (void **)&args.fh, &args.fhsize) == 0) { if (args.fhsize < 0 || args.fhsize > NFSX_FHMAX) { vfs_mount_error(mp, "Bad file handle"); error = EINVAL; goto out; } bcopy(args.fh, nfh, args.fhsize); } else { args.fhsize = 0; } (void) vfs_getopt(mp->mnt_optnew, "hostname", (void **)&args.hostname, &len); if (args.hostname == NULL) { vfs_mount_error(mp, "Invalid hostname"); error = EINVAL; goto out; } bcopy(args.hostname, hst, MNAMELEN); hst[MNAMELEN - 1] = '\0'; } if (vfs_getopt(mp->mnt_optnew, "principal", (void **)&name, NULL) == 0) strlcpy(srvkrbname, name, sizeof (srvkrbname)); else snprintf(srvkrbname, sizeof (srvkrbname), "nfs@%s", hst); srvkrbnamelen = strlen(srvkrbname); if (vfs_getopt(mp->mnt_optnew, "gssname", (void **)&name, NULL) == 0) strlcpy(krbname, name, sizeof (krbname)); else krbname[0] = '\0'; krbnamelen = strlen(krbname); if (vfs_getopt(mp->mnt_optnew, "dirpath", (void **)&name, NULL) == 0) strlcpy(dirpath, name, sizeof (dirpath)); else dirpath[0] = '\0'; dirlen = strlen(dirpath); if (has_nfs_args_opt == 0) { if (vfs_getopt(mp->mnt_optnew, "addr", (void **)&args.addr, &args.addrlen) == 0) { if (args.addrlen > SOCK_MAXADDRLEN) { error = ENAMETOOLONG; goto out; } nam = malloc(args.addrlen, M_SONAME, M_WAITOK); bcopy(args.addr, nam, args.addrlen); nam->sa_len = args.addrlen; } else { vfs_mount_error(mp, "No server address"); error = EINVAL; goto out; } } args.fh = nfh; error = mountnfs(&args, mp, nam, hst, krbname, krbnamelen, dirpath, dirlen, srvkrbname, srvkrbnamelen, &vp, td->td_ucred, td, nametimeo, negnametimeo, minvers); out: if (!error) { MNT_ILOCK(mp); mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_NO_IOPF | MNTK_USES_BCACHE; MNT_IUNLOCK(mp); } return (error); } /* * VFS Operations. * * mount system call * It seems a bit dumb to copyinstr() the host and path here and then * bcopy() them in mountnfs(), but I wanted to detect errors before * doing the sockargs() call because sockargs() allocates an mbuf and * an error after that means that I have to release the mbuf. */ /* ARGSUSED */ static int nfs_cmount(struct mntarg *ma, void *data, uint64_t flags) { int error; struct nfs_args args; error = copyin(data, &args, sizeof (struct nfs_args)); if (error) return error; ma = mount_arg(ma, "nfs_args", &args, sizeof args); error = kernel_mount(ma, flags); return (error); } /* * Common code for mount and mountroot */ static int mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam, char *hst, u_char *krbname, int krbnamelen, u_char *dirpath, int dirlen, u_char *srvkrbname, int srvkrbnamelen, struct vnode **vpp, struct ucred *cred, struct thread *td, int nametimeo, int negnametimeo, int minvers) { struct nfsmount *nmp; struct nfsnode *np; int error, trycnt, ret; struct nfsvattr nfsva; struct nfsclclient *clp; struct nfsclds *dsp, *tdsp; uint32_t lease; static u_int64_t clval = 0; NFSCL_DEBUG(3, "in mnt\n"); clp = NULL; if (mp->mnt_flag & MNT_UPDATE) { nmp = VFSTONFS(mp); printf("%s: MNT_UPDATE is no longer handled here\n", __func__); FREE(nam, M_SONAME); return (0); } else { MALLOC(nmp, struct nfsmount *, sizeof (struct nfsmount) + krbnamelen + dirlen + srvkrbnamelen + 2, M_NEWNFSMNT, M_WAITOK | M_ZERO); TAILQ_INIT(&nmp->nm_bufq); if (clval == 0) clval = (u_int64_t)nfsboottime.tv_sec; nmp->nm_clval = clval++; nmp->nm_krbnamelen = krbnamelen; nmp->nm_dirpathlen = dirlen; nmp->nm_srvkrbnamelen = srvkrbnamelen; if (td->td_ucred->cr_uid != (uid_t)0) { /* * nm_uid is used to get KerberosV credentials for * the nfsv4 state handling operations if there is * no host based principal set. Use the uid of * this user if not root, since they are doing the * mount. I don't think setting this for root will * work, since root normally does not have user * credentials in a credentials cache. */ nmp->nm_uid = td->td_ucred->cr_uid; } else { /* * Just set to -1, so it won't be used. */ nmp->nm_uid = (uid_t)-1; } /* Copy and null terminate all the names */ if (nmp->nm_krbnamelen > 0) { bcopy(krbname, nmp->nm_krbname, nmp->nm_krbnamelen); nmp->nm_name[nmp->nm_krbnamelen] = '\0'; } if (nmp->nm_dirpathlen > 0) { bcopy(dirpath, NFSMNT_DIRPATH(nmp), nmp->nm_dirpathlen); nmp->nm_name[nmp->nm_krbnamelen + nmp->nm_dirpathlen + 1] = '\0'; } if (nmp->nm_srvkrbnamelen > 0) { bcopy(srvkrbname, NFSMNT_SRVKRBNAME(nmp), nmp->nm_srvkrbnamelen); nmp->nm_name[nmp->nm_krbnamelen + nmp->nm_dirpathlen + nmp->nm_srvkrbnamelen + 2] = '\0'; } nmp->nm_sockreq.nr_cred = crhold(cred); mtx_init(&nmp->nm_sockreq.nr_mtx, "nfssock", NULL, MTX_DEF); mp->mnt_data = nmp; nmp->nm_getinfo = nfs_getnlminfo; nmp->nm_vinvalbuf = ncl_vinvalbuf; } vfs_getnewfsid(mp); nmp->nm_mountp = mp; mtx_init(&nmp->nm_mtx, "NFSmount lock", NULL, MTX_DEF | MTX_DUPOK); /* * Since nfs_decode_args() might optionally set them, these * need to be set to defaults before the call, so that the * optional settings aren't overwritten. */ nmp->nm_nametimeo = nametimeo; nmp->nm_negnametimeo = negnametimeo; nmp->nm_timeo = NFS_TIMEO; nmp->nm_retry = NFS_RETRANS; nmp->nm_readahead = NFS_DEFRAHEAD; /* This is empirical approximation of sqrt(hibufspace) * 256. */ nmp->nm_wcommitsize = NFS_MAXBSIZE / 256; while ((long)nmp->nm_wcommitsize * nmp->nm_wcommitsize < hibufspace) nmp->nm_wcommitsize *= 2; nmp->nm_wcommitsize *= 256; if ((argp->flags & NFSMNT_NFSV4) != 0) nmp->nm_minorvers = minvers; else nmp->nm_minorvers = 0; nfs_decode_args(mp, nmp, argp, hst, cred, td); /* * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too * high, depending on whether we end up with negative offsets in * the client or server somewhere. 2GB-1 may be safer. * * For V3, ncl_fsinfo will adjust this as necessary. Assume maximum * that we can handle until we find out otherwise. */ if ((argp->flags & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) == 0) nmp->nm_maxfilesize = 0xffffffffLL; else nmp->nm_maxfilesize = OFF_MAX; if ((argp->flags & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) == 0) { nmp->nm_wsize = NFS_WSIZE; nmp->nm_rsize = NFS_RSIZE; nmp->nm_readdirsize = NFS_READDIRSIZE; } nmp->nm_numgrps = NFS_MAXGRPS; nmp->nm_tprintf_delay = nfs_tprintf_delay; if (nmp->nm_tprintf_delay < 0) nmp->nm_tprintf_delay = 0; nmp->nm_tprintf_initial_delay = nfs_tprintf_initial_delay; if (nmp->nm_tprintf_initial_delay < 0) nmp->nm_tprintf_initial_delay = 0; nmp->nm_fhsize = argp->fhsize; if (nmp->nm_fhsize > 0) bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize); bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN); nmp->nm_nam = nam; /* Set up the sockets and per-host congestion */ nmp->nm_sotype = argp->sotype; nmp->nm_soproto = argp->proto; nmp->nm_sockreq.nr_prog = NFS_PROG; if ((argp->flags & NFSMNT_NFSV4)) nmp->nm_sockreq.nr_vers = NFS_VER4; else if ((argp->flags & NFSMNT_NFSV3)) nmp->nm_sockreq.nr_vers = NFS_VER3; else nmp->nm_sockreq.nr_vers = NFS_VER2; if ((error = newnfs_connect(nmp, &nmp->nm_sockreq, cred, td, 0))) goto bad; /* For NFSv4.1, get the clientid now. */ if (nmp->nm_minorvers > 0) { NFSCL_DEBUG(3, "at getcl\n"); error = nfscl_getcl(mp, cred, td, 0, &clp); NFSCL_DEBUG(3, "aft getcl=%d\n", error); if (error != 0) goto bad; } if (nmp->nm_fhsize == 0 && (nmp->nm_flag & NFSMNT_NFSV4) && nmp->nm_dirpathlen > 0) { NFSCL_DEBUG(3, "in dirp\n"); /* * If the fhsize on the mount point == 0 for V4, the mount * path needs to be looked up. */ trycnt = 3; do { error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp), cred, td); NFSCL_DEBUG(3, "aft dirp=%d\n", error); if (error) (void) nfs_catnap(PZERO, error, "nfsgetdirp"); } while (error && --trycnt > 0); if (error) { error = nfscl_maperr(td, error, (uid_t)0, (gid_t)0); goto bad; } } /* * A reference count is needed on the nfsnode representing the * remote root. If this object is not persistent, then backward * traversals of the mount point (i.e. "..") will not work if * the nfsnode gets flushed out of the cache. Ufs does not have * this problem, because one can identify root inodes by their * number == ROOTINO (2). */ if (nmp->nm_fhsize > 0) { /* * Set f_iosize to NFS_DIRBLKSIZ so that bo_bsize gets set * non-zero for the root vnode. f_iosize will be set correctly * by nfs_statfs() before any I/O occurs. */ mp->mnt_stat.f_iosize = NFS_DIRBLKSIZ; error = ncl_nget(mp, nmp->nm_fh, nmp->nm_fhsize, &np, LK_EXCLUSIVE); if (error) goto bad; *vpp = NFSTOV(np); /* * Get file attributes and transfer parameters for the * mountpoint. This has the side effect of filling in * (*vpp)->v_type with the correct value. */ ret = nfsrpc_getattrnovp(nmp, nmp->nm_fh, nmp->nm_fhsize, 1, cred, td, &nfsva, NULL, &lease); if (ret) { /* * Just set default values to get things going. */ NFSBZERO((caddr_t)&nfsva, sizeof (struct nfsvattr)); nfsva.na_vattr.va_type = VDIR; nfsva.na_vattr.va_mode = 0777; nfsva.na_vattr.va_nlink = 100; nfsva.na_vattr.va_uid = (uid_t)0; nfsva.na_vattr.va_gid = (gid_t)0; nfsva.na_vattr.va_fileid = 2; nfsva.na_vattr.va_gen = 1; nfsva.na_vattr.va_blocksize = NFS_FABLKSIZE; nfsva.na_vattr.va_size = 512 * 1024; lease = 60; } (void) nfscl_loadattrcache(vpp, &nfsva, NULL, NULL, 0, 1); if (nmp->nm_minorvers > 0) { NFSCL_DEBUG(3, "lease=%d\n", (int)lease); NFSLOCKCLSTATE(); clp->nfsc_renew = NFSCL_RENEW(lease); clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew; clp->nfsc_clientidrev++; if (clp->nfsc_clientidrev == 0) clp->nfsc_clientidrev++; NFSUNLOCKCLSTATE(); /* * Mount will succeed, so the renew thread can be * started now. */ nfscl_start_renewthread(clp); nfscl_clientrelease(clp); } if (argp->flags & NFSMNT_NFSV3) ncl_fsinfo(nmp, *vpp, cred, td); /* Mark if the mount point supports NFSv4 ACLs. */ if ((argp->flags & NFSMNT_NFSV4) != 0 && nfsrv_useacl != 0 && ret == 0 && NFSISSET_ATTRBIT(&nfsva.na_suppattr, NFSATTRBIT_ACL)) { MNT_ILOCK(mp); mp->mnt_flag |= MNT_NFS4ACLS; MNT_IUNLOCK(mp); } /* * Lose the lock but keep the ref. */ NFSVOPUNLOCK(*vpp, 0); return (0); } error = EIO; bad: if (clp != NULL) nfscl_clientrelease(clp); newnfs_disconnect(&nmp->nm_sockreq); crfree(nmp->nm_sockreq.nr_cred); if (nmp->nm_sockreq.nr_auth != NULL) AUTH_DESTROY(nmp->nm_sockreq.nr_auth); mtx_destroy(&nmp->nm_sockreq.nr_mtx); mtx_destroy(&nmp->nm_mtx); if (nmp->nm_clp != NULL) { NFSLOCKCLSTATE(); LIST_REMOVE(nmp->nm_clp, nfsc_list); NFSUNLOCKCLSTATE(); free(nmp->nm_clp, M_NFSCLCLIENT); } TAILQ_FOREACH_SAFE(dsp, &nmp->nm_sess, nfsclds_list, tdsp) nfscl_freenfsclds(dsp); FREE(nmp, M_NEWNFSMNT); FREE(nam, M_SONAME); return (error); } /* * unmount system call */ static int nfs_unmount(struct mount *mp, int mntflags) { struct thread *td; struct nfsmount *nmp; int error, flags = 0, i, trycnt = 0; struct nfsclds *dsp, *tdsp; td = curthread; if (mntflags & MNT_FORCE) flags |= FORCECLOSE; nmp = VFSTONFS(mp); /* * Goes something like this.. * - Call vflush() to clear out vnodes for this filesystem * - Close the socket * - Free up the data structures */ /* In the forced case, cancel any outstanding requests. */ if (mntflags & MNT_FORCE) { error = newnfs_nmcancelreqs(nmp); if (error) goto out; /* For a forced close, get rid of the renew thread now */ nfscl_umount(nmp, td); } /* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */ do { error = vflush(mp, 1, flags, td); if ((mntflags & MNT_FORCE) && error != 0 && ++trycnt < 30) (void) nfs_catnap(PSOCK, error, "newndm"); } while ((mntflags & MNT_FORCE) && error != 0 && trycnt < 30); if (error) goto out; /* * We are now committed to the unmount. */ if ((mntflags & MNT_FORCE) == 0) nfscl_umount(nmp, td); /* Make sure no nfsiods are assigned to this mount. */ mtx_lock(&ncl_iod_mutex); for (i = 0; i < NFS_MAXASYNCDAEMON; i++) if (ncl_iodmount[i] == nmp) { ncl_iodwant[i] = NFSIOD_AVAILABLE; ncl_iodmount[i] = NULL; } mtx_unlock(&ncl_iod_mutex); newnfs_disconnect(&nmp->nm_sockreq); crfree(nmp->nm_sockreq.nr_cred); FREE(nmp->nm_nam, M_SONAME); if (nmp->nm_sockreq.nr_auth != NULL) AUTH_DESTROY(nmp->nm_sockreq.nr_auth); mtx_destroy(&nmp->nm_sockreq.nr_mtx); mtx_destroy(&nmp->nm_mtx); TAILQ_FOREACH_SAFE(dsp, &nmp->nm_sess, nfsclds_list, tdsp) nfscl_freenfsclds(dsp); FREE(nmp, M_NEWNFSMNT); out: return (error); } /* * Return root of a filesystem */ static int nfs_root(struct mount *mp, int flags, struct vnode **vpp) { struct vnode *vp; struct nfsmount *nmp; struct nfsnode *np; int error; nmp = VFSTONFS(mp); error = ncl_nget(mp, nmp->nm_fh, nmp->nm_fhsize, &np, flags); if (error) return error; vp = NFSTOV(np); /* * Get transfer parameters and attributes for root vnode once. */ mtx_lock(&nmp->nm_mtx); if (NFSHASNFSV3(nmp) && !NFSHASGOTFSINFO(nmp)) { mtx_unlock(&nmp->nm_mtx); ncl_fsinfo(nmp, vp, curthread->td_ucred, curthread); } else mtx_unlock(&nmp->nm_mtx); if (vp->v_type == VNON) vp->v_type = VDIR; vp->v_vflag |= VV_ROOT; *vpp = vp; return (0); } /* * Flush out the buffer cache */ /* ARGSUSED */ static int nfs_sync(struct mount *mp, int waitfor) { struct vnode *vp, *mvp; struct thread *td; int error, allerror = 0; td = curthread; MNT_ILOCK(mp); /* * If a forced dismount is in progress, return from here so that * the umount(2) syscall doesn't get stuck in VFS_SYNC() before * calling VFS_UNMOUNT(). */ if ((mp->mnt_kern_flag & MNTK_UNMOUNTF) != 0) { MNT_IUNLOCK(mp); return (EBADF); } MNT_IUNLOCK(mp); /* * Force stale buffer cache information to be flushed. */ loop: MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { /* XXX Racy bv_cnt check. */ if (NFSVOPISLOCKED(vp) || vp->v_bufobj.bo_dirty.bv_cnt == 0 || waitfor == MNT_LAZY) { VI_UNLOCK(vp); continue; } if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); goto loop; } error = VOP_FSYNC(vp, waitfor, td); if (error) allerror = error; NFSVOPUNLOCK(vp, 0); vrele(vp); } return (allerror); } static int nfs_sysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req) { struct nfsmount *nmp = VFSTONFS(mp); struct vfsquery vq; int error; bzero(&vq, sizeof(vq)); switch (op) { #if 0 case VFS_CTL_NOLOCKS: val = (nmp->nm_flag & NFSMNT_NOLOCKS) ? 1 : 0; if (req->oldptr != NULL) { error = SYSCTL_OUT(req, &val, sizeof(val)); if (error) return (error); } if (req->newptr != NULL) { error = SYSCTL_IN(req, &val, sizeof(val)); if (error) return (error); if (val) nmp->nm_flag |= NFSMNT_NOLOCKS; else nmp->nm_flag &= ~NFSMNT_NOLOCKS; } break; #endif case VFS_CTL_QUERY: mtx_lock(&nmp->nm_mtx); if (nmp->nm_state & NFSSTA_TIMEO) vq.vq_flags |= VQ_NOTRESP; mtx_unlock(&nmp->nm_mtx); #if 0 if (!(nmp->nm_flag & NFSMNT_NOLOCKS) && (nmp->nm_state & NFSSTA_LOCKTIMEO)) vq.vq_flags |= VQ_NOTRESPLOCK; #endif error = SYSCTL_OUT(req, &vq, sizeof(vq)); break; case VFS_CTL_TIMEO: if (req->oldptr != NULL) { error = SYSCTL_OUT(req, &nmp->nm_tprintf_initial_delay, sizeof(nmp->nm_tprintf_initial_delay)); if (error) return (error); } if (req->newptr != NULL) { error = vfs_suser(mp, req->td); if (error) return (error); error = SYSCTL_IN(req, &nmp->nm_tprintf_initial_delay, sizeof(nmp->nm_tprintf_initial_delay)); if (error) return (error); if (nmp->nm_tprintf_initial_delay < 0) nmp->nm_tprintf_initial_delay = 0; } break; default: return (ENOTSUP); } return (0); } /* * Purge any RPCs in progress, so that they will all return errors. * This allows dounmount() to continue as far as VFS_UNMOUNT() for a * forced dismount. */ static void nfs_purge(struct mount *mp) { struct nfsmount *nmp = VFSTONFS(mp); newnfs_nmcancelreqs(nmp); } /* * Extract the information needed by the nlm from the nfs vnode. */ static void nfs_getnlminfo(struct vnode *vp, uint8_t *fhp, size_t *fhlenp, struct sockaddr_storage *sp, int *is_v3p, off_t *sizep, struct timeval *timeop) { struct nfsmount *nmp; struct nfsnode *np = VTONFS(vp); nmp = VFSTONFS(vp->v_mount); if (fhlenp != NULL) *fhlenp = (size_t)np->n_fhp->nfh_len; if (fhp != NULL) bcopy(np->n_fhp->nfh_fh, fhp, np->n_fhp->nfh_len); if (sp != NULL) bcopy(nmp->nm_nam, sp, min(nmp->nm_nam->sa_len, sizeof(*sp))); if (is_v3p != NULL) *is_v3p = NFS_ISV3(vp); if (sizep != NULL) *sizep = np->n_size; if (timeop != NULL) { timeop->tv_sec = nmp->nm_timeo / NFS_HZ; timeop->tv_usec = (nmp->nm_timeo % NFS_HZ) * (1000000 / NFS_HZ); } } /* * This function prints out an option name, based on the conditional * argument. */ static __inline void nfscl_printopt(struct nfsmount *nmp, int testval, char *opt, char **buf, size_t *blen) { int len; if (testval != 0 && *blen > strlen(opt)) { len = snprintf(*buf, *blen, "%s", opt); if (len != strlen(opt)) printf("EEK!!\n"); *buf += len; *blen -= len; } } /* * This function printf out an options integer value. */ static __inline void nfscl_printoptval(struct nfsmount *nmp, int optval, char *opt, char **buf, size_t *blen) { int len; if (*blen > strlen(opt) + 1) { /* Could result in truncated output string. */ len = snprintf(*buf, *blen, "%s=%d", opt, optval); if (len < *blen) { *buf += len; *blen -= len; } } } /* * Load the option flags and values into the buffer. */ void nfscl_retopts(struct nfsmount *nmp, char *buffer, size_t buflen) { char *buf; size_t blen; buf = buffer; blen = buflen; nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_NFSV4) != 0, "nfsv4", &buf, &blen); if ((nmp->nm_flag & NFSMNT_NFSV4) != 0) { nfscl_printoptval(nmp, nmp->nm_minorvers, ",minorversion", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_PNFS) != 0, ",pnfs", &buf, &blen); } nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_NFSV3) != 0, "nfsv3", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) == 0, "nfsv2", &buf, &blen); nfscl_printopt(nmp, nmp->nm_sotype == SOCK_STREAM, ",tcp", &buf, &blen); nfscl_printopt(nmp, nmp->nm_sotype != SOCK_STREAM, ",udp", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_RESVPORT) != 0, ",resvport", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_NOCONN) != 0, ",noconn", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_SOFT) == 0, ",hard", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_SOFT) != 0, ",soft", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_INT) != 0, ",intr", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_NOCTO) == 0, ",cto", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_NOCTO) != 0, ",nocto", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_NONCONTIGWR) != 0, ",noncontigwr", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & (NFSMNT_NOLOCKD | NFSMNT_NFSV4)) == 0, ",lockd", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & (NFSMNT_NOLOCKD | NFSMNT_NFSV4)) == NFSMNT_NOLOCKD, ",nolockd", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_RDIRPLUS) != 0, ",rdirplus", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & NFSMNT_KERB) == 0, ",sec=sys", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & (NFSMNT_KERB | NFSMNT_INTEGRITY | NFSMNT_PRIVACY)) == NFSMNT_KERB, ",sec=krb5", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & (NFSMNT_KERB | NFSMNT_INTEGRITY | NFSMNT_PRIVACY)) == (NFSMNT_KERB | NFSMNT_INTEGRITY), ",sec=krb5i", &buf, &blen); nfscl_printopt(nmp, (nmp->nm_flag & (NFSMNT_KERB | NFSMNT_INTEGRITY | NFSMNT_PRIVACY)) == (NFSMNT_KERB | NFSMNT_PRIVACY), ",sec=krb5p", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_acdirmin, ",acdirmin", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_acdirmax, ",acdirmax", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_acregmin, ",acregmin", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_acregmax, ",acregmax", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_nametimeo, ",nametimeo", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_negnametimeo, ",negnametimeo", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_rsize, ",rsize", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_wsize, ",wsize", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_readdirsize, ",readdirsize", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_readahead, ",readahead", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_wcommitsize, ",wcommitsize", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_timeo, ",timeout", &buf, &blen); nfscl_printoptval(nmp, nmp->nm_retry, ",retrans", &buf, &blen); } Index: user/ngie/more-tests/sys/fs/nfsserver/nfs_nfsdkrpc.c =================================================================== --- user/ngie/more-tests/sys/fs/nfsserver/nfs_nfsdkrpc.c (revision 281974) +++ user/ngie/more-tests/sys/fs/nfsserver/nfs_nfsdkrpc.c (revision 281975) @@ -1,550 +1,551 @@ /*- * Copyright (c) 1989, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Rick Macklem at The University of Guelph. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * */ #include __FBSDID("$FreeBSD$"); #include "opt_inet6.h" #include "opt_kgssapi.h" #include #include #include #include #include #include NFSDLOCKMUTEX; NFSV4ROOTLOCKMUTEX; struct nfsv4lock nfsd_suspend_lock; /* * Mapping of old NFS Version 2 RPC numbers to generic numbers. */ int newnfs_nfsv3_procid[NFS_V3NPROCS] = { NFSPROC_NULL, NFSPROC_GETATTR, NFSPROC_SETATTR, NFSPROC_NOOP, NFSPROC_LOOKUP, NFSPROC_READLINK, NFSPROC_READ, NFSPROC_NOOP, NFSPROC_WRITE, NFSPROC_CREATE, NFSPROC_REMOVE, NFSPROC_RENAME, NFSPROC_LINK, NFSPROC_SYMLINK, NFSPROC_MKDIR, NFSPROC_RMDIR, NFSPROC_READDIR, NFSPROC_FSSTAT, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP, }; SYSCTL_DECL(_vfs_nfsd); SVCPOOL *nfsrvd_pool; static int nfs_privport = 0; SYSCTL_INT(_vfs_nfsd, OID_AUTO, nfs_privport, CTLFLAG_RWTUN, &nfs_privport, 0, "Only allow clients using a privileged port for NFSv2 and 3"); static int nfs_minvers = NFS_VER2; SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_min_nfsvers, CTLFLAG_RWTUN, &nfs_minvers, 0, "The lowest version of NFS handled by the server"); static int nfs_maxvers = NFS_VER4; SYSCTL_INT(_vfs_nfsd, OID_AUTO, server_max_nfsvers, CTLFLAG_RWTUN, &nfs_maxvers, 0, "The highest version of NFS handled by the server"); static int nfs_proc(struct nfsrv_descript *, u_int32_t, SVCXPRT *xprt, struct nfsrvcache **); extern u_long sb_max_adj; extern int newnfs_numnfsd; extern struct proc *nfsd_master_proc; /* * NFS server system calls */ static void nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt) { struct nfsrv_descript nd; struct nfsrvcache *rp = NULL; int cacherep, credflavor; memset(&nd, 0, sizeof(nd)); if (rqst->rq_vers == NFS_VER2) { - if (rqst->rq_proc > NFSV2PROC_STATFS) { + if (rqst->rq_proc > NFSV2PROC_STATFS || + newnfs_nfsv3_procid[rqst->rq_proc] == NFSPROC_NOOP) { svcerr_noproc(rqst); svc_freereq(rqst); goto out; } nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc]; nd.nd_flag = ND_NFSV2; } else if (rqst->rq_vers == NFS_VER3) { if (rqst->rq_proc >= NFS_V3NPROCS) { svcerr_noproc(rqst); svc_freereq(rqst); goto out; } nd.nd_procnum = rqst->rq_proc; nd.nd_flag = ND_NFSV3; } else { if (rqst->rq_proc != NFSPROC_NULL && rqst->rq_proc != NFSV4PROC_COMPOUND) { svcerr_noproc(rqst); svc_freereq(rqst); goto out; } nd.nd_procnum = rqst->rq_proc; nd.nd_flag = ND_NFSV4; } /* * Note: we want rq_addr, not svc_getrpccaller for nd_nam2 - * NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP * mounts. */ nd.nd_mrep = rqst->rq_args; rqst->rq_args = NULL; newnfs_realign(&nd.nd_mrep, M_WAITOK); nd.nd_md = nd.nd_mrep; nd.nd_dpos = mtod(nd.nd_md, caddr_t); nd.nd_nam = svc_getrpccaller(rqst); nd.nd_nam2 = rqst->rq_addr; nd.nd_mreq = NULL; nd.nd_cred = NULL; if (nfs_privport && (nd.nd_flag & ND_NFSV4) == 0) { /* Check if source port is privileged */ u_short port; struct sockaddr *nam = nd.nd_nam; struct sockaddr_in *sin; sin = (struct sockaddr_in *)nam; /* * INET/INET6 - same code: * sin_port and sin6_port are at same offset */ port = ntohs(sin->sin_port); if (port >= IPPORT_RESERVED && nd.nd_procnum != NFSPROC_NULL) { #ifdef INET6 char b6[INET6_ADDRSTRLEN]; #if defined(KLD_MODULE) /* Do not use ip6_sprintf: the nfs module should work without INET6. */ #define ip6_sprintf(buf, a) \ (sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x", \ (a)->s6_addr16[0], (a)->s6_addr16[1], \ (a)->s6_addr16[2], (a)->s6_addr16[3], \ (a)->s6_addr16[4], (a)->s6_addr16[5], \ (a)->s6_addr16[6], (a)->s6_addr16[7]), \ (buf)) #endif #endif printf("NFS request from unprivileged port (%s:%d)\n", #ifdef INET6 sin->sin_family == AF_INET6 ? ip6_sprintf(b6, &satosin6(sin)->sin6_addr) : #if defined(KLD_MODULE) #undef ip6_sprintf #endif #endif inet_ntoa(sin->sin_addr), port); svcerr_weakauth(rqst); svc_freereq(rqst); m_freem(nd.nd_mrep); goto out; } } if (nd.nd_procnum != NFSPROC_NULL) { if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) { svcerr_weakauth(rqst); svc_freereq(rqst); m_freem(nd.nd_mrep); goto out; } /* Set the flag based on credflavor */ if (credflavor == RPCSEC_GSS_KRB5) { nd.nd_flag |= ND_GSS; } else if (credflavor == RPCSEC_GSS_KRB5I) { nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY); } else if (credflavor == RPCSEC_GSS_KRB5P) { nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY); } else if (credflavor != AUTH_SYS) { svcerr_weakauth(rqst); svc_freereq(rqst); m_freem(nd.nd_mrep); goto out; } #ifdef MAC mac_cred_associate_nfsd(nd.nd_cred); #endif /* * Get a refcnt (shared lock) on nfsd_suspend_lock. * NFSSVC_SUSPENDNFSD will take an exclusive lock on * nfsd_suspend_lock to suspend these threads. * This must be done here, before the check of * nfsv4root exports by nfsvno_v4rootexport(). */ NFSLOCKV4ROOTMUTEX(); nfsv4_getref(&nfsd_suspend_lock, NULL, NFSV4ROOTLOCKMUTEXPTR, NULL); NFSUNLOCKV4ROOTMUTEX(); if ((nd.nd_flag & ND_NFSV4) != 0) { nd.nd_repstat = nfsvno_v4rootexport(&nd); if (nd.nd_repstat != 0) { NFSLOCKV4ROOTMUTEX(); nfsv4_relref(&nfsd_suspend_lock); NFSUNLOCKV4ROOTMUTEX(); svcerr_weakauth(rqst); svc_freereq(rqst); m_freem(nd.nd_mrep); goto out; } } cacherep = nfs_proc(&nd, rqst->rq_xid, xprt, &rp); NFSLOCKV4ROOTMUTEX(); nfsv4_relref(&nfsd_suspend_lock); NFSUNLOCKV4ROOTMUTEX(); } else { NFSMGET(nd.nd_mreq); nd.nd_mreq->m_len = 0; cacherep = RC_REPLY; } if (nd.nd_mrep != NULL) m_freem(nd.nd_mrep); if (nd.nd_cred != NULL) crfree(nd.nd_cred); if (cacherep == RC_DROPIT) { if (nd.nd_mreq != NULL) m_freem(nd.nd_mreq); svc_freereq(rqst); goto out; } if (nd.nd_mreq == NULL) { svcerr_decode(rqst); svc_freereq(rqst); goto out; } if (nd.nd_repstat & NFSERR_AUTHERR) { svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR); if (nd.nd_mreq != NULL) m_freem(nd.nd_mreq); } else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) { svcerr_systemerr(rqst); } if (rp != NULL) { nfsrvd_sentcache(rp, (rqst->rq_reply_seq != 0 || SVC_ACK(xprt, NULL)), rqst->rq_reply_seq); } svc_freereq(rqst); out: NFSEXITCODE(0); } /* * Check the cache and, optionally, do the RPC. * Return the appropriate cache response. */ static int nfs_proc(struct nfsrv_descript *nd, u_int32_t xid, SVCXPRT *xprt, struct nfsrvcache **rpp) { struct thread *td = curthread; int cacherep = RC_DOIT, isdgram, taglen = -1; struct mbuf *m; u_char tag[NFSV4_SMALLSTR + 1], *tagstr = NULL; u_int32_t minorvers = 0; uint32_t ack; *rpp = NULL; if (nd->nd_nam2 == NULL) { nd->nd_flag |= ND_STREAMSOCK; isdgram = 0; } else { isdgram = 1; } /* * Two cases: * 1 - For NFSv2 over UDP, if we are near our malloc/mget * limit, just drop the request. There is no * NFSERR_RESOURCE or NFSERR_DELAY for NFSv2 and the * client will timeout/retry over UDP in a little while. * 2 - nd_repstat == 0 && nd_mreq == NULL, which * means a normal nfs rpc, so check the cache */ if ((nd->nd_flag & ND_NFSV2) && nd->nd_nam2 != NULL && nfsrv_mallocmget_limit()) { cacherep = RC_DROPIT; } else { /* * For NFSv3, play it safe and assume that the client is * doing retries on the same TCP connection. */ if ((nd->nd_flag & (ND_NFSV4 | ND_STREAMSOCK)) == ND_STREAMSOCK) nd->nd_flag |= ND_SAMETCPCONN; nd->nd_retxid = xid; nd->nd_tcpconntime = NFSD_MONOSEC; nd->nd_sockref = xprt->xp_sockref; if ((nd->nd_flag & ND_NFSV4) != 0) nfsd_getminorvers(nd, tag, &tagstr, &taglen, &minorvers); if ((nd->nd_flag & ND_NFSV41) != 0) /* NFSv4.1 caches replies in the session slots. */ cacherep = RC_DOIT; else { cacherep = nfsrvd_getcache(nd); ack = 0; SVC_ACK(xprt, &ack); nfsrc_trimcache(xprt->xp_sockref, ack, 0); } } /* * Handle the request. There are three cases. * RC_DOIT - do the RPC * RC_REPLY - return the reply already created * RC_DROPIT - just throw the request away */ if (cacherep == RC_DOIT) { if ((nd->nd_flag & ND_NFSV41) != 0) nd->nd_xprt = xprt; nfsrvd_dorpc(nd, isdgram, tagstr, taglen, minorvers, td); if ((nd->nd_flag & ND_NFSV41) != 0) { if (nd->nd_repstat != NFSERR_REPLYFROMCACHE && (nd->nd_flag & ND_SAVEREPLY) != 0) { /* Cache a copy of the reply. */ m = m_copym(nd->nd_mreq, 0, M_COPYALL, M_WAITOK); } else m = NULL; if ((nd->nd_flag & ND_HASSEQUENCE) != 0) nfsrv_cache_session(nd->nd_sessionid, nd->nd_slotid, nd->nd_repstat, &m); if (nd->nd_repstat == NFSERR_REPLYFROMCACHE) nd->nd_repstat = 0; cacherep = RC_REPLY; } else { if (nd->nd_repstat == NFSERR_DONTREPLY) cacherep = RC_DROPIT; else cacherep = RC_REPLY; *rpp = nfsrvd_updatecache(nd); } } if (tagstr != NULL && taglen > NFSV4_SMALLSTR) free(tagstr, M_TEMP); NFSEXITCODE2(0, nd); return (cacherep); } static void nfssvc_loss(SVCXPRT *xprt) { uint32_t ack; ack = 0; SVC_ACK(xprt, &ack); nfsrc_trimcache(xprt->xp_sockref, ack, 1); } /* * Adds a socket to the list for servicing by nfsds. */ int nfsrvd_addsock(struct file *fp) { int siz; struct socket *so; int error = 0; SVCXPRT *xprt; static u_int64_t sockref = 0; so = fp->f_data; siz = sb_max_adj; error = soreserve(so, siz, siz); if (error) goto out; /* * Steal the socket from userland so that it doesn't close * unexpectedly. */ if (so->so_type == SOCK_DGRAM) xprt = svc_dg_create(nfsrvd_pool, so, 0, 0); else xprt = svc_vc_create(nfsrvd_pool, so, 0, 0); if (xprt) { fp->f_ops = &badfileops; fp->f_data = NULL; xprt->xp_sockref = ++sockref; if (nfs_minvers == NFS_VER2) svc_reg(xprt, NFS_PROG, NFS_VER2, nfssvc_program, NULL); if (nfs_minvers <= NFS_VER3 && nfs_maxvers >= NFS_VER3) svc_reg(xprt, NFS_PROG, NFS_VER3, nfssvc_program, NULL); if (nfs_maxvers >= NFS_VER4) svc_reg(xprt, NFS_PROG, NFS_VER4, nfssvc_program, NULL); if (so->so_type == SOCK_STREAM) svc_loss_reg(xprt, nfssvc_loss); SVC_RELEASE(xprt); } out: NFSEXITCODE(error); return (error); } /* * Called by nfssvc() for nfsds. Just loops around servicing rpc requests * until it is killed by a signal. */ int nfsrvd_nfsd(struct thread *td, struct nfsd_nfsd_args *args) { char principal[MAXHOSTNAMELEN + 5]; int error = 0; bool_t ret2, ret3, ret4; error = copyinstr(args->principal, principal, sizeof (principal), NULL); if (error) goto out; /* * Only the first nfsd actually does any work. The RPC code * adds threads to it as needed. Any extra processes offered * by nfsd just exit. If nfsd is new enough, it will call us * once with a structure that specifies how many threads to * use. */ NFSD_LOCK(); if (newnfs_numnfsd == 0) { newnfs_numnfsd++; NFSD_UNLOCK(); /* An empty string implies AUTH_SYS only. */ if (principal[0] != '\0') { ret2 = rpc_gss_set_svc_name_call(principal, "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER2); ret3 = rpc_gss_set_svc_name_call(principal, "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER3); ret4 = rpc_gss_set_svc_name_call(principal, "kerberosv5", GSS_C_INDEFINITE, NFS_PROG, NFS_VER4); if (!ret2 || !ret3 || !ret4) printf("nfsd: can't register svc name\n"); } nfsrvd_pool->sp_minthreads = args->minthreads; nfsrvd_pool->sp_maxthreads = args->maxthreads; svc_run(nfsrvd_pool); if (principal[0] != '\0') { rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER2); rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER3); rpc_gss_clear_svc_name_call(NFS_PROG, NFS_VER4); } NFSD_LOCK(); newnfs_numnfsd--; nfsrvd_init(1); } NFSD_UNLOCK(); out: NFSEXITCODE(error); return (error); } /* * Initialize the data structures for the server. * Handshake with any new nfsds starting up to avoid any chance of * corruption. */ void nfsrvd_init(int terminating) { NFSD_LOCK_ASSERT(); if (terminating) { nfsd_master_proc = NULL; NFSD_UNLOCK(); svcpool_destroy(nfsrvd_pool); nfsrvd_pool = NULL; NFSD_LOCK(); } NFSD_UNLOCK(); nfsrvd_pool = svcpool_create("nfsd", SYSCTL_STATIC_CHILDREN(_vfs_nfsd)); nfsrvd_pool->sp_rcache = NULL; nfsrvd_pool->sp_assign = fhanew_assign; nfsrvd_pool->sp_done = fha_nd_complete; NFSD_LOCK(); } Index: user/ngie/more-tests/sys/i386/i386/mp_machdep.c =================================================================== --- user/ngie/more-tests/sys/i386/i386/mp_machdep.c (revision 281974) +++ user/ngie/more-tests/sys/i386/i386/mp_machdep.c (revision 281975) @@ -1,1675 +1,685 @@ /*- * Copyright (c) 1996, by Steve Passe * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. The name of the developer may NOT be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_apic.h" #include "opt_cpu.h" #include "opt_kstack_pages.h" #include "opt_pmap.h" #include "opt_sched.h" #include "opt_smp.h" #if !defined(lint) #if !defined(SMP) #error How did you get here? #endif #ifndef DEV_APIC #error The apic device is required for SMP, add "device apic" to your config file. #endif #if defined(CPU_DISABLE_CMPXCHG) && !defined(COMPILING_LINT) #error SMP not supported with CPU_DISABLE_CMPXCHG #endif #endif /* not lint */ #include #include #include #include /* cngetc() */ #include #ifdef GPROF #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define WARMBOOT_TARGET 0 #define WARMBOOT_OFF (KERNBASE + 0x0467) #define WARMBOOT_SEG (KERNBASE + 0x0469) #define CMOS_REG (0x70) #define CMOS_DATA (0x71) #define BIOS_RESET (0x0f) #define BIOS_WARM (0x0a) /* * this code MUST be enabled here and in mpboot.s. * it follows the very early stages of AP boot by placing values in CMOS ram. * it NORMALLY will never be needed and thus the primitive method for enabling. * #define CHECK_POINTS */ #if defined(CHECK_POINTS) && !defined(PC98) #define CHECK_READ(A) (outb(CMOS_REG, (A)), inb(CMOS_DATA)) #define CHECK_WRITE(A,D) (outb(CMOS_REG, (A)), outb(CMOS_DATA, (D))) #define CHECK_INIT(D); \ CHECK_WRITE(0x34, (D)); \ CHECK_WRITE(0x35, (D)); \ CHECK_WRITE(0x36, (D)); \ CHECK_WRITE(0x37, (D)); \ CHECK_WRITE(0x38, (D)); \ CHECK_WRITE(0x39, (D)); #define CHECK_PRINT(S); \ printf("%s: %d, %d, %d, %d, %d, %d\n", \ (S), \ CHECK_READ(0x34), \ CHECK_READ(0x35), \ CHECK_READ(0x36), \ CHECK_READ(0x37), \ CHECK_READ(0x38), \ CHECK_READ(0x39)); #else /* CHECK_POINTS */ #define CHECK_INIT(D) #define CHECK_PRINT(S) #define CHECK_WRITE(A, D) #endif /* CHECK_POINTS */ -/* lock region used by kernel profiling */ -int mcount_lock; - -int mp_naps; /* # of Applications processors */ -int boot_cpu_id = -1; /* designated BSP */ - extern struct pcpu __pcpu[]; -/* AP uses this during bootstrap. Do not staticize. */ -char *bootSTK; -static int bootAP; - -/* Free these after use */ -void *bootstacks[MAXCPU]; -static void *dpcpu; - -struct pcb stoppcbs[MAXCPU]; -struct susppcb **susppcbs; - /* Variables needed for SMP tlb shootdown. */ vm_offset_t smp_tlb_addr1; vm_offset_t smp_tlb_addr2; volatile int smp_tlb_wait; -#ifdef COUNT_IPIS -/* Interrupt counts. */ -static u_long *ipi_preempt_counts[MAXCPU]; -static u_long *ipi_ast_counts[MAXCPU]; -u_long *ipi_invltlb_counts[MAXCPU]; -u_long *ipi_invlrng_counts[MAXCPU]; -u_long *ipi_invlpg_counts[MAXCPU]; -u_long *ipi_invlcache_counts[MAXCPU]; -u_long *ipi_rendezvous_counts[MAXCPU]; -static u_long *ipi_hardclock_counts[MAXCPU]; -#endif - -/* Default cpu_ops implementation. */ -struct cpu_ops cpu_ops; - /* * Local data and functions. */ -static volatile cpuset_t ipi_nmi_pending; - -/* used to hold the AP's until we are ready to release them */ -static struct mtx ap_boot_mtx; - -/* Set to 1 once we're ready to let the APs out of the pen. */ -static volatile int aps_ready = 0; - -/* - * Store data from cpu_add() until later in the boot when we actually setup - * the APs. - */ -struct cpu_info { - int cpu_present:1; - int cpu_bsp:1; - int cpu_disabled:1; - int cpu_hyperthread:1; -} static cpu_info[MAX_APIC_ID + 1]; -int cpu_apic_ids[MAXCPU]; -int apic_cpuids[MAX_APIC_ID + 1]; - -/* Holds pending bitmap based IPIs per CPU */ -volatile u_int cpu_ipi_pending[MAXCPU]; - -static u_int boot_address; -static int cpu_logical; /* logical cpus per core */ -static int cpu_cores; /* cores per package */ - -static void assign_cpu_ids(void); static void install_ap_tramp(void); -static void set_interrupt_apic_ids(void); static int start_all_aps(void); static int start_ap(int apic_id); static void release_aps(void *dummy); -static u_int hyperthreading_cpus; /* logical cpus sharing L1 cache */ -static int hyperthreading_allowed = 1; +static u_int boot_address; -static void -mem_range_AP_init(void) -{ - if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP) - mem_range_softc.mr_op->initAP(&mem_range_softc); -} - -static void -topo_probe_amd(void) -{ - int core_id_bits; - int id; - - /* AMD processors do not support HTT. */ - cpu_logical = 1; - - if ((amd_feature2 & AMDID2_CMP) == 0) { - cpu_cores = 1; - return; - } - - core_id_bits = (cpu_procinfo2 & AMDID_COREID_SIZE) >> - AMDID_COREID_SIZE_SHIFT; - if (core_id_bits == 0) { - cpu_cores = (cpu_procinfo2 & AMDID_CMP_CORES) + 1; - return; - } - - /* Fam 10h and newer should get here. */ - for (id = 0; id <= MAX_APIC_ID; id++) { - /* Check logical CPU availability. */ - if (!cpu_info[id].cpu_present || cpu_info[id].cpu_disabled) - continue; - /* Check if logical CPU has the same package ID. */ - if ((id >> core_id_bits) != (boot_cpu_id >> core_id_bits)) - continue; - cpu_cores++; - } -} - /* - * Round up to the next power of two, if necessary, and then - * take log2. - * Returns -1 if argument is zero. - */ -static __inline int -mask_width(u_int x) -{ - - return (fls(x << (1 - powerof2(x))) - 1); -} - -static void -topo_probe_0x4(void) -{ - u_int p[4]; - int pkg_id_bits; - int core_id_bits; - int max_cores; - int max_logical; - int id; - - /* Both zero and one here mean one logical processor per package. */ - max_logical = (cpu_feature & CPUID_HTT) != 0 ? - (cpu_procinfo & CPUID_HTT_CORES) >> 16 : 1; - if (max_logical <= 1) - return; - - /* - * Because of uniformity assumption we examine only - * those logical processors that belong to the same - * package as BSP. Further, we count number of - * logical processors that belong to the same core - * as BSP thus deducing number of threads per core. - */ - if (cpu_high >= 0x4) { - cpuid_count(0x04, 0, p); - max_cores = ((p[0] >> 26) & 0x3f) + 1; - } else - max_cores = 1; - core_id_bits = mask_width(max_logical/max_cores); - if (core_id_bits < 0) - return; - pkg_id_bits = core_id_bits + mask_width(max_cores); - - for (id = 0; id <= MAX_APIC_ID; id++) { - /* Check logical CPU availability. */ - if (!cpu_info[id].cpu_present || cpu_info[id].cpu_disabled) - continue; - /* Check if logical CPU has the same package ID. */ - if ((id >> pkg_id_bits) != (boot_cpu_id >> pkg_id_bits)) - continue; - cpu_cores++; - /* Check if logical CPU has the same package and core IDs. */ - if ((id >> core_id_bits) == (boot_cpu_id >> core_id_bits)) - cpu_logical++; - } - - KASSERT(cpu_cores >= 1 && cpu_logical >= 1, - ("topo_probe_0x4 couldn't find BSP")); - - cpu_cores /= cpu_logical; - hyperthreading_cpus = cpu_logical; -} - -static void -topo_probe_0xb(void) -{ - u_int p[4]; - int bits; - int cnt; - int i; - int logical; - int type; - int x; - - /* We only support three levels for now. */ - for (i = 0; i < 3; i++) { - cpuid_count(0x0b, i, p); - - /* Fall back if CPU leaf 11 doesn't really exist. */ - if (i == 0 && p[1] == 0) { - topo_probe_0x4(); - return; - } - - bits = p[0] & 0x1f; - logical = p[1] &= 0xffff; - type = (p[2] >> 8) & 0xff; - if (type == 0 || logical == 0) - break; - /* - * Because of uniformity assumption we examine only - * those logical processors that belong to the same - * package as BSP. - */ - for (cnt = 0, x = 0; x <= MAX_APIC_ID; x++) { - if (!cpu_info[x].cpu_present || - cpu_info[x].cpu_disabled) - continue; - if (x >> bits == boot_cpu_id >> bits) - cnt++; - } - if (type == CPUID_TYPE_SMT) - cpu_logical = cnt; - else if (type == CPUID_TYPE_CORE) - cpu_cores = cnt; - } - if (cpu_logical == 0) - cpu_logical = 1; - cpu_cores /= cpu_logical; -} - -/* - * Both topology discovery code and code that consumes topology - * information assume top-down uniformity of the topology. - * That is, all physical packages must be identical and each - * core in a package must have the same number of threads. - * Topology information is queried only on BSP, on which this - * code runs and for which it can query CPUID information. - * Then topology is extrapolated on all packages using the - * uniformity assumption. - */ -static void -topo_probe(void) -{ - static int cpu_topo_probed = 0; - - if (cpu_topo_probed) - return; - - CPU_ZERO(&logical_cpus_mask); - if (mp_ncpus <= 1) - cpu_cores = cpu_logical = 1; - else if (cpu_vendor_id == CPU_VENDOR_AMD) - topo_probe_amd(); - else if (cpu_vendor_id == CPU_VENDOR_INTEL) { - /* - * See Intel(R) 64 Architecture Processor - * Topology Enumeration article for details. - * - * Note that 0x1 <= cpu_high < 4 case should be - * compatible with topo_probe_0x4() logic when - * CPUID.1:EBX[23:16] > 0 (cpu_cores will be 1) - * or it should trigger the fallback otherwise. - */ - if (cpu_high >= 0xb) - topo_probe_0xb(); - else if (cpu_high >= 0x1) - topo_probe_0x4(); - } - - /* - * Fallback: assume each logical CPU is in separate - * physical package. That is, no multi-core, no SMT. - */ - if (cpu_cores == 0 || cpu_logical == 0) - cpu_cores = cpu_logical = 1; - cpu_topo_probed = 1; -} - -struct cpu_group * -cpu_topo(void) -{ - int cg_flags; - - /* - * Determine whether any threading flags are - * necessry. - */ - topo_probe(); - if (cpu_logical > 1 && hyperthreading_cpus) - cg_flags = CG_FLAG_HTT; - else if (cpu_logical > 1) - cg_flags = CG_FLAG_SMT; - else - cg_flags = 0; - if (mp_ncpus % (cpu_cores * cpu_logical) != 0) { - printf("WARNING: Non-uniform processors.\n"); - printf("WARNING: Using suboptimal topology.\n"); - return (smp_topo_none()); - } - /* - * No multi-core or hyper-threaded. - */ - if (cpu_logical * cpu_cores == 1) - return (smp_topo_none()); - /* - * Only HTT no multi-core. - */ - if (cpu_logical > 1 && cpu_cores == 1) - return (smp_topo_1level(CG_SHARE_L1, cpu_logical, cg_flags)); - /* - * Only multi-core no HTT. - */ - if (cpu_cores > 1 && cpu_logical == 1) - return (smp_topo_1level(CG_SHARE_L2, cpu_cores, cg_flags)); - /* - * Both HTT and multi-core. - */ - return (smp_topo_2level(CG_SHARE_L2, cpu_cores, - CG_SHARE_L1, cpu_logical, cg_flags)); -} - - -/* * Calculate usable address in base memory for AP trampoline code. */ u_int mp_bootaddress(u_int basemem) { boot_address = trunc_page(basemem); /* round down to 4k boundary */ if ((basemem - boot_address) < bootMP_size) boot_address -= PAGE_SIZE; /* not enough, lower by 4k */ return boot_address; } -void -cpu_add(u_int apic_id, char boot_cpu) -{ - - if (apic_id > MAX_APIC_ID) { - panic("SMP: APIC ID %d too high", apic_id); - return; - } - KASSERT(cpu_info[apic_id].cpu_present == 0, ("CPU %d added twice", - apic_id)); - cpu_info[apic_id].cpu_present = 1; - if (boot_cpu) { - KASSERT(boot_cpu_id == -1, - ("CPU %d claims to be BSP, but CPU %d already is", apic_id, - boot_cpu_id)); - boot_cpu_id = apic_id; - cpu_info[apic_id].cpu_bsp = 1; - } - if (mp_ncpus < MAXCPU) { - mp_ncpus++; - mp_maxid = mp_ncpus - 1; - } - if (bootverbose) - printf("SMP: Added CPU %d (%s)\n", apic_id, boot_cpu ? "BSP" : - "AP"); -} - -void -cpu_mp_setmaxid(void) -{ - - /* - * mp_maxid should be already set by calls to cpu_add(). - * Just sanity check its value here. - */ - if (mp_ncpus == 0) - KASSERT(mp_maxid == 0, - ("%s: mp_ncpus is zero, but mp_maxid is not", __func__)); - else if (mp_ncpus == 1) - mp_maxid = 0; - else - KASSERT(mp_maxid >= mp_ncpus - 1, - ("%s: counters out of sync: max %d, count %d", __func__, - mp_maxid, mp_ncpus)); -} - -int -cpu_mp_probe(void) -{ - - /* - * Always record BSP in CPU map so that the mbuf init code works - * correctly. - */ - CPU_SETOF(0, &all_cpus); - if (mp_ncpus == 0) { - /* - * No CPUs were found, so this must be a UP system. Setup - * the variables to represent a system with a single CPU - * with an id of 0. - */ - mp_ncpus = 1; - return (0); - } - - /* At least one CPU was found. */ - if (mp_ncpus == 1) { - /* - * One CPU was found, so this must be a UP system with - * an I/O APIC. - */ - mp_maxid = 0; - return (0); - } - - /* At least two CPUs were found. */ - return (1); -} - /* * Initialize the IPI handlers and start up the AP's. */ void cpu_mp_start(void) { int i; /* Initialize the logical ID to APIC ID table. */ for (i = 0; i < MAXCPU; i++) { cpu_apic_ids[i] = -1; cpu_ipi_pending[i] = 0; } /* Install an inter-CPU IPI for TLB invalidation */ setidt(IPI_INVLTLB, IDTVEC(invltlb), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); setidt(IPI_INVLPG, IDTVEC(invlpg), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); setidt(IPI_INVLRNG, IDTVEC(invlrng), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* Install an inter-CPU IPI for cache invalidation. */ setidt(IPI_INVLCACHE, IDTVEC(invlcache), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* Install an inter-CPU IPI for all-CPU rendezvous */ setidt(IPI_RENDEZVOUS, IDTVEC(rendezvous), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* Install generic inter-CPU IPI handler */ setidt(IPI_BITMAP_VECTOR, IDTVEC(ipi_intr_bitmap_handler), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* Install an inter-CPU IPI for CPU stop/restart */ setidt(IPI_STOP, IDTVEC(cpustop), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* Install an inter-CPU IPI for CPU suspend/resume */ setidt(IPI_SUSPEND, IDTVEC(cpususpend), SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* Set boot_cpu_id if needed. */ if (boot_cpu_id == -1) { boot_cpu_id = PCPU_GET(apic_id); cpu_info[boot_cpu_id].cpu_bsp = 1; } else KASSERT(boot_cpu_id == PCPU_GET(apic_id), ("BSP's APIC ID doesn't match boot_cpu_id")); /* Probe logical/physical core configuration. */ topo_probe(); assign_cpu_ids(); /* Start each Application Processor */ start_all_aps(); set_interrupt_apic_ids(); } - /* - * Print various information about the SMP system hardware and setup. - */ -void -cpu_mp_announce(void) -{ - const char *hyperthread; - int i; - - printf("FreeBSD/SMP: %d package(s) x %d core(s)", - mp_ncpus / (cpu_cores * cpu_logical), cpu_cores); - if (hyperthreading_cpus > 1) - printf(" x %d HTT threads", cpu_logical); - else if (cpu_logical > 1) - printf(" x %d SMT threads", cpu_logical); - printf("\n"); - - /* List active CPUs first. */ - printf(" cpu0 (BSP): APIC ID: %2d\n", boot_cpu_id); - for (i = 1; i < mp_ncpus; i++) { - if (cpu_info[cpu_apic_ids[i]].cpu_hyperthread) - hyperthread = "/HT"; - else - hyperthread = ""; - printf(" cpu%d (AP%s): APIC ID: %2d\n", i, hyperthread, - cpu_apic_ids[i]); - } - - /* List disabled CPUs last. */ - for (i = 0; i <= MAX_APIC_ID; i++) { - if (!cpu_info[i].cpu_present || !cpu_info[i].cpu_disabled) - continue; - if (cpu_info[i].cpu_hyperthread) - hyperthread = "/HT"; - else - hyperthread = ""; - printf(" cpu (AP%s): APIC ID: %2d (disabled)\n", hyperthread, - i); - } -} - -/* * AP CPU's call this to initialize themselves. */ void init_secondary(void) { struct pcpu *pc; vm_offset_t addr; int gsel_tss; int x, myid; - u_int cpuid, cr0; + u_int cr0; /* bootAP is set in start_ap() to our ID. */ myid = bootAP; /* Get per-cpu data */ pc = &__pcpu[myid]; /* prime data page for it to use */ pcpu_init(pc, myid, sizeof(struct pcpu)); dpcpu_init(dpcpu, myid); pc->pc_apic_id = cpu_apic_ids[myid]; pc->pc_prvspace = pc; pc->pc_curthread = 0; gdt_segs[GPRIV_SEL].ssd_base = (int) pc; gdt_segs[GPROC0_SEL].ssd_base = (int) &pc->pc_common_tss; for (x = 0; x < NGDT; x++) { ssdtosd(&gdt_segs[x], &gdt[myid * NGDT + x].sd); } r_gdt.rd_limit = NGDT * sizeof(gdt[0]) - 1; r_gdt.rd_base = (int) &gdt[myid * NGDT]; lgdt(&r_gdt); /* does magic intra-segment return */ lidt(&r_idt); lldt(_default_ldt); PCPU_SET(currentldt, _default_ldt); gsel_tss = GSEL(GPROC0_SEL, SEL_KPL); gdt[myid * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS; PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */ PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL)); PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16); PCPU_SET(tss_gdt, &gdt[myid * NGDT + GPROC0_SEL].sd); PCPU_SET(common_tssd, *PCPU_GET(tss_gdt)); ltr(gsel_tss); PCPU_SET(fsgs_gdt, &gdt[myid * NGDT + GUFS_SEL].sd); /* * Set to a known state: * Set by mpboot.s: CR0_PG, CR0_PE * Set by cpu_setregs: CR0_NE, CR0_MP, CR0_TS, CR0_WP, CR0_AM */ cr0 = rcr0(); cr0 &= ~(CR0_CD | CR0_NW | CR0_EM); load_cr0(cr0); CHECK_WRITE(0x38, 5); /* signal our startup to the BSP. */ mp_naps++; CHECK_WRITE(0x39, 6); /* Spin until the BSP releases the AP's. */ while (!aps_ready) ia32_pause(); /* BSP may have changed PTD while we were waiting */ invltlb(); for (addr = 0; addr < NKPT * NBPDR - 1; addr += PAGE_SIZE) invlpg(addr); #if defined(I586_CPU) && !defined(NO_F00F_HACK) lidt(&r_idt); #endif - /* - * On real hardware, switch to x2apic mode if possible. Do it - * after aps_ready was signalled, to avoid manipulating the - * mode while BSP might still want to send some IPI to us - * (second startup IPI is ignored on modern hardware etc). - */ - lapic_xapic_mode(); - - /* Initialize the PAT MSR if present. */ - pmap_init_pat(); - - /* set up CPU registers and state */ - cpu_setregs(); - - /* set up SSE/NX */ - initializecpu(); - - /* set up FPU state on the AP */ - npxinit(false); - - if (cpu_ops.cpu_init) - cpu_ops.cpu_init(); - - /* A quick check from sanity claus */ - cpuid = PCPU_GET(cpuid); - if (PCPU_GET(apic_id) != lapic_id()) { - printf("SMP: cpuid = %d\n", cpuid); - printf("SMP: actual apic_id = %d\n", lapic_id()); - printf("SMP: correct apic_id = %d\n", PCPU_GET(apic_id)); - panic("cpuid mismatch! boom!!"); - } - - /* Initialize curthread. */ - KASSERT(PCPU_GET(idlethread) != NULL, ("no idle thread")); - PCPU_SET(curthread, PCPU_GET(idlethread)); - - mca_init(); - - mtx_lock_spin(&ap_boot_mtx); - - /* Init local apic for irq's */ - lapic_setup(1); - - /* Set memory range attributes for this CPU to match the BSP */ - mem_range_AP_init(); - - smp_cpus++; - - CTR1(KTR_SMP, "SMP: AP CPU #%d Launched", cpuid); - printf("SMP: AP CPU #%d Launched!\n", cpuid); - - /* Determine if we are a logical CPU. */ - /* XXX Calculation depends on cpu_logical being a power of 2, e.g. 2 */ - if (cpu_logical > 1 && PCPU_GET(apic_id) % cpu_logical != 0) - CPU_SET(cpuid, &logical_cpus_mask); - - if (bootverbose) - lapic_dump("AP"); - - if (smp_cpus == mp_ncpus) { - /* enable IPI's, tlb shootdown, freezes etc */ - atomic_store_rel_int(&smp_started, 1); - } - - mtx_unlock_spin(&ap_boot_mtx); - - /* Wait until all the AP's are up. */ - while (smp_started == 0) - ia32_pause(); - - /* Start per-CPU event timers. */ - cpu_initclocks_ap(); - - /* Enter the scheduler. */ - sched_throw(NULL); - - panic("scheduler returned us to %s", __func__); - /* NOTREACHED */ + init_secondary_tail(); } /******************************************************************* * local functions and data */ /* - * We tell the I/O APIC code about all the CPUs we want to receive - * interrupts. If we don't want certain CPUs to receive IRQs we - * can simply not tell the I/O APIC code about them in this function. - * We also do not tell it about the BSP since it tells itself about - * the BSP internally to work with UP kernels and on UP machines. - */ -static void -set_interrupt_apic_ids(void) -{ - u_int i, apic_id; - - for (i = 0; i < MAXCPU; i++) { - apic_id = cpu_apic_ids[i]; - if (apic_id == -1) - continue; - if (cpu_info[apic_id].cpu_bsp) - continue; - if (cpu_info[apic_id].cpu_disabled) - continue; - - /* Don't let hyperthreads service interrupts. */ - if (cpu_logical > 1 && - apic_id % cpu_logical != 0) - continue; - - intr_add_cpu(i); - } -} - -/* - * Assign logical CPU IDs to local APICs. - */ -static void -assign_cpu_ids(void) -{ - u_int i; - - TUNABLE_INT_FETCH("machdep.hyperthreading_allowed", - &hyperthreading_allowed); - - /* Check for explicitly disabled CPUs. */ - for (i = 0; i <= MAX_APIC_ID; i++) { - if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp) - continue; - - if (hyperthreading_cpus > 1 && i % hyperthreading_cpus != 0) { - cpu_info[i].cpu_hyperthread = 1; - - /* - * Don't use HT CPU if it has been disabled by a - * tunable. - */ - if (hyperthreading_allowed == 0) { - cpu_info[i].cpu_disabled = 1; - continue; - } - } - - /* Don't use this CPU if it has been disabled by a tunable. */ - if (resource_disabled("lapic", i)) { - cpu_info[i].cpu_disabled = 1; - continue; - } - } - - if (hyperthreading_allowed == 0 && hyperthreading_cpus > 1) { - hyperthreading_cpus = 0; - cpu_logical = 1; - } - - /* - * Assign CPU IDs to local APIC IDs and disable any CPUs - * beyond MAXCPU. CPU 0 is always assigned to the BSP. - * - * To minimize confusion for userland, we attempt to number - * CPUs such that all threads and cores in a package are - * grouped together. For now we assume that the BSP is always - * the first thread in a package and just start adding APs - * starting with the BSP's APIC ID. - */ - mp_ncpus = 1; - cpu_apic_ids[0] = boot_cpu_id; - apic_cpuids[boot_cpu_id] = 0; - for (i = boot_cpu_id + 1; i != boot_cpu_id; - i == MAX_APIC_ID ? i = 0 : i++) { - if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp || - cpu_info[i].cpu_disabled) - continue; - - if (mp_ncpus < MAXCPU) { - cpu_apic_ids[mp_ncpus] = i; - apic_cpuids[i] = mp_ncpus; - mp_ncpus++; - } else - cpu_info[i].cpu_disabled = 1; - } - KASSERT(mp_maxid >= mp_ncpus - 1, - ("%s: counters out of sync: max %d, count %d", __func__, mp_maxid, - mp_ncpus)); -} - -/* * start each AP in our list */ /* Lowest 1MB is already mapped: don't touch*/ #define TMPMAP_START 1 static int start_all_aps(void) { #ifndef PC98 u_char mpbiosreason; #endif u_int32_t mpbioswarmvec; int apic_id, cpu, i; mtx_init(&ap_boot_mtx, "ap boot", NULL, MTX_SPIN); /* install the AP 1st level boot code */ install_ap_tramp(); /* save the current value of the warm-start vector */ mpbioswarmvec = *((u_int32_t *) WARMBOOT_OFF); #ifndef PC98 outb(CMOS_REG, BIOS_RESET); mpbiosreason = inb(CMOS_DATA); #endif /* set up temporary P==V mapping for AP boot */ /* XXX this is a hack, we should boot the AP on its own stack/PTD */ for (i = TMPMAP_START; i < NKPT; i++) PTD[i] = PTD[KPTDI + i]; invltlb(); /* start each AP */ for (cpu = 1; cpu < mp_ncpus; cpu++) { apic_id = cpu_apic_ids[cpu]; /* allocate and set up a boot stack data page */ bootstacks[cpu] = (char *)kmem_malloc(kernel_arena, KSTACK_PAGES * PAGE_SIZE, M_WAITOK | M_ZERO); dpcpu = (void *)kmem_malloc(kernel_arena, DPCPU_SIZE, M_WAITOK | M_ZERO); /* setup a vector to our boot code */ *((volatile u_short *) WARMBOOT_OFF) = WARMBOOT_TARGET; *((volatile u_short *) WARMBOOT_SEG) = (boot_address >> 4); #ifndef PC98 outb(CMOS_REG, BIOS_RESET); outb(CMOS_DATA, BIOS_WARM); /* 'warm-start' */ #endif bootSTK = (char *)bootstacks[cpu] + KSTACK_PAGES * PAGE_SIZE - 4; bootAP = cpu; /* attempt to start the Application Processor */ CHECK_INIT(99); /* setup checkpoints */ if (!start_ap(apic_id)) { printf("AP #%d (PHY# %d) failed!\n", cpu, apic_id); CHECK_PRINT("trace"); /* show checkpoints */ /* better panic as the AP may be running loose */ printf("panic y/n? [y] "); if (cngetc() != 'n') panic("bye-bye"); } CHECK_PRINT("trace"); /* show checkpoints */ CPU_SET(cpu, &all_cpus); /* record AP in CPU map */ } /* restore the warmstart vector */ *(u_int32_t *) WARMBOOT_OFF = mpbioswarmvec; #ifndef PC98 outb(CMOS_REG, BIOS_RESET); outb(CMOS_DATA, mpbiosreason); #endif /* Undo V==P hack from above */ for (i = TMPMAP_START; i < NKPT; i++) PTD[i] = 0; pmap_invalidate_range(kernel_pmap, 0, NKPT * NBPDR - 1); /* number of APs actually started */ return mp_naps; } /* * load the 1st level AP boot code into base memory. */ /* targets for relocation */ extern void bigJump(void); extern void bootCodeSeg(void); extern void bootDataSeg(void); extern void MPentry(void); extern u_int MP_GDT; extern u_int mp_gdtbase; static void install_ap_tramp(void) { int x; int size = *(int *) ((u_long) & bootMP_size); vm_offset_t va = boot_address + KERNBASE; u_char *src = (u_char *) ((u_long) bootMP); u_char *dst = (u_char *) va; u_int boot_base = (u_int) bootMP; u_int8_t *dst8; u_int16_t *dst16; u_int32_t *dst32; KASSERT (size <= PAGE_SIZE, ("'size' do not fit into PAGE_SIZE, as expected.")); pmap_kenter(va, boot_address); pmap_invalidate_page (kernel_pmap, va); for (x = 0; x < size; ++x) *dst++ = *src++; /* * modify addresses in code we just moved to basemem. unfortunately we * need fairly detailed info about mpboot.s for this to work. changes * to mpboot.s might require changes here. */ /* boot code is located in KERNEL space */ dst = (u_char *) va; /* modify the lgdt arg */ dst32 = (u_int32_t *) (dst + ((u_int) & mp_gdtbase - boot_base)); *dst32 = boot_address + ((u_int) & MP_GDT - boot_base); /* modify the ljmp target for MPentry() */ dst32 = (u_int32_t *) (dst + ((u_int) bigJump - boot_base) + 1); *dst32 = ((u_int) MPentry - KERNBASE); /* modify the target for boot code segment */ dst16 = (u_int16_t *) (dst + ((u_int) bootCodeSeg - boot_base)); dst8 = (u_int8_t *) (dst16 + 1); *dst16 = (u_int) boot_address & 0xffff; *dst8 = ((u_int) boot_address >> 16) & 0xff; /* modify the target for boot data segment */ dst16 = (u_int16_t *) (dst + ((u_int) bootDataSeg - boot_base)); dst8 = (u_int8_t *) (dst16 + 1); *dst16 = (u_int) boot_address & 0xffff; *dst8 = ((u_int) boot_address >> 16) & 0xff; } /* * This function starts the AP (application processor) identified * by the APIC ID 'physicalCpu'. It does quite a "song and dance" * to accomplish this. This is necessary because of the nuances * of the different hardware we might encounter. It isn't pretty, * but it seems to work. */ static int start_ap(int apic_id) { int vector, ms; int cpus; /* calculate the vector */ vector = (boot_address >> 12) & 0xff; /* used as a watchpoint to signal AP startup */ cpus = mp_naps; ipi_startup(apic_id, vector); /* Wait up to 5 seconds for it to start. */ for (ms = 0; ms < 5000; ms++) { if (mp_naps > cpus) return 1; /* return SUCCESS */ DELAY(1000); } return 0; /* return FAILURE */ } -#ifdef COUNT_XINVLTLB_HITS -u_int xhits_gbl[MAXCPU]; -u_int xhits_pg[MAXCPU]; -u_int xhits_rng[MAXCPU]; -static SYSCTL_NODE(_debug, OID_AUTO, xhits, CTLFLAG_RW, 0, ""); -SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, global, CTLFLAG_RW, &xhits_gbl, - sizeof(xhits_gbl), "IU", ""); -SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, page, CTLFLAG_RW, &xhits_pg, - sizeof(xhits_pg), "IU", ""); -SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, range, CTLFLAG_RW, &xhits_rng, - sizeof(xhits_rng), "IU", ""); - -u_int ipi_global; -u_int ipi_page; -u_int ipi_range; -u_int ipi_range_size; -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_global, CTLFLAG_RW, &ipi_global, 0, ""); -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_page, CTLFLAG_RW, &ipi_page, 0, ""); -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_range, CTLFLAG_RW, &ipi_range, 0, ""); -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_range_size, CTLFLAG_RW, &ipi_range_size, - 0, ""); - -u_int ipi_masked_global; -u_int ipi_masked_page; -u_int ipi_masked_range; -u_int ipi_masked_range_size; -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_global, CTLFLAG_RW, - &ipi_masked_global, 0, ""); -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_page, CTLFLAG_RW, - &ipi_masked_page, 0, ""); -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_range, CTLFLAG_RW, - &ipi_masked_range, 0, ""); -SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_range_size, CTLFLAG_RW, - &ipi_masked_range_size, 0, ""); -#endif /* COUNT_XINVLTLB_HITS */ - /* - * Init and startup IPI. - */ -void -ipi_startup(int apic_id, int vector) -{ - - /* - * This attempts to follow the algorithm described in the - * Intel Multiprocessor Specification v1.4 in section B.4. - * For each IPI, we allow the local APIC ~20us to deliver the - * IPI. If that times out, we panic. - */ - - /* - * first we do an INIT IPI: this INIT IPI might be run, resetting - * and running the target CPU. OR this INIT IPI might be latched (P5 - * bug), CPU waiting for STARTUP IPI. OR this INIT IPI might be - * ignored. - */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_LEVEL | - APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_INIT, apic_id); - lapic_ipi_wait(100); - - /* Explicitly deassert the INIT IPI. */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_LEVEL | - APIC_LEVEL_DEASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_INIT, - apic_id); - - DELAY(10000); /* wait ~10mS */ - - /* - * next we do a STARTUP IPI: the previous INIT IPI might still be - * latched, (P5 bug) this 1st STARTUP would then terminate - * immediately, and the previously started INIT IPI would continue. OR - * the previous INIT IPI has already run. and this STARTUP IPI will - * run. OR the previous INIT IPI was ignored. and this STARTUP IPI - * will run. - */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_EDGE | - APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_STARTUP | - vector, apic_id); - if (!lapic_ipi_wait(100)) - panic("Failed to deliver first STARTUP IPI to APIC %d", - apic_id); - DELAY(200); /* wait ~200uS */ - - /* - * finally we do a 2nd STARTUP IPI: this 2nd STARTUP IPI should run IF - * the previous STARTUP IPI was cancelled by a latched INIT IPI. OR - * this STARTUP IPI will be ignored, as only ONE STARTUP IPI is - * recognized after hardware RESET or INIT IPI. - */ - lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_EDGE | - APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_STARTUP | - vector, apic_id); - if (!lapic_ipi_wait(100)) - panic("Failed to deliver second STARTUP IPI to APIC %d", - apic_id); - - DELAY(200); /* wait ~200uS */ -} - -/* - * Send an IPI to specified CPU handling the bitmap logic. - */ -static void -ipi_send_cpu(int cpu, u_int ipi) -{ - u_int bitmap, old_pending, new_pending; - - KASSERT(cpu_apic_ids[cpu] != -1, ("IPI to non-existent CPU %d", cpu)); - - if (IPI_IS_BITMAPED(ipi)) { - bitmap = 1 << ipi; - ipi = IPI_BITMAP_VECTOR; - do { - old_pending = cpu_ipi_pending[cpu]; - new_pending = old_pending | bitmap; - } while (!atomic_cmpset_int(&cpu_ipi_pending[cpu], - old_pending, new_pending)); - if (old_pending) - return; - } - lapic_ipi_vectored(ipi, cpu_apic_ids[cpu]); -} - -/* * Flush the TLB on all other CPU's */ static void smp_tlb_shootdown(u_int vector, vm_offset_t addr1, vm_offset_t addr2) { u_int ncpu; ncpu = mp_ncpus - 1; /* does not shootdown self */ if (ncpu < 1) return; /* no other cpus */ if (!(read_eflags() & PSL_I)) panic("%s: interrupts disabled", __func__); mtx_lock_spin(&smp_ipi_mtx); smp_tlb_addr1 = addr1; smp_tlb_addr2 = addr2; atomic_store_rel_int(&smp_tlb_wait, 0); ipi_all_but_self(vector); while (smp_tlb_wait < ncpu) ia32_pause(); mtx_unlock_spin(&smp_ipi_mtx); } static void smp_targeted_tlb_shootdown(cpuset_t mask, u_int vector, vm_offset_t addr1, vm_offset_t addr2) { int cpu, ncpu, othercpus; othercpus = mp_ncpus - 1; if (CPU_ISFULLSET(&mask)) { if (othercpus < 1) return; } else { CPU_CLR(PCPU_GET(cpuid), &mask); if (CPU_EMPTY(&mask)) return; } if (!(read_eflags() & PSL_I)) panic("%s: interrupts disabled", __func__); mtx_lock_spin(&smp_ipi_mtx); smp_tlb_addr1 = addr1; smp_tlb_addr2 = addr2; atomic_store_rel_int(&smp_tlb_wait, 0); if (CPU_ISFULLSET(&mask)) { ncpu = othercpus; ipi_all_but_self(vector); } else { ncpu = 0; while ((cpu = CPU_FFS(&mask)) != 0) { cpu--; CPU_CLR(cpu, &mask); CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, vector); ipi_send_cpu(cpu, vector); ncpu++; } } while (smp_tlb_wait < ncpu) ia32_pause(); mtx_unlock_spin(&smp_ipi_mtx); } void -smp_cache_flush(void) -{ - - if (smp_started) - smp_tlb_shootdown(IPI_INVLCACHE, 0, 0); -} - -void smp_invltlb(void) { if (smp_started) { smp_tlb_shootdown(IPI_INVLTLB, 0, 0); #ifdef COUNT_XINVLTLB_HITS ipi_global++; #endif } } void smp_invlpg(vm_offset_t addr) { if (smp_started) { smp_tlb_shootdown(IPI_INVLPG, addr, 0); #ifdef COUNT_XINVLTLB_HITS ipi_page++; #endif } } void smp_invlpg_range(vm_offset_t addr1, vm_offset_t addr2) { if (smp_started) { smp_tlb_shootdown(IPI_INVLRNG, addr1, addr2); #ifdef COUNT_XINVLTLB_HITS ipi_range++; ipi_range_size += (addr2 - addr1) / PAGE_SIZE; #endif } } void smp_masked_invltlb(cpuset_t mask) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLTLB, 0, 0); #ifdef COUNT_XINVLTLB_HITS ipi_masked_global++; #endif } } void smp_masked_invlpg(cpuset_t mask, vm_offset_t addr) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLPG, addr, 0); #ifdef COUNT_XINVLTLB_HITS ipi_masked_page++; #endif } } void smp_masked_invlpg_range(cpuset_t mask, vm_offset_t addr1, vm_offset_t addr2) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLRNG, addr1, addr2); #ifdef COUNT_XINVLTLB_HITS ipi_masked_range++; ipi_masked_range_size += (addr2 - addr1) / PAGE_SIZE; #endif } } void -ipi_bitmap_handler(struct trapframe frame) +smp_cache_flush(void) { - struct trapframe *oldframe; - struct thread *td; - int cpu = PCPU_GET(cpuid); - u_int ipi_bitmap; - critical_enter(); - td = curthread; - td->td_intr_nesting_level++; - oldframe = td->td_intr_frame; - td->td_intr_frame = &frame; - ipi_bitmap = atomic_readandclear_int(&cpu_ipi_pending[cpu]); - if (ipi_bitmap & (1 << IPI_PREEMPT)) { -#ifdef COUNT_IPIS - (*ipi_preempt_counts[cpu])++; -#endif - sched_preempt(td); - } - if (ipi_bitmap & (1 << IPI_AST)) { -#ifdef COUNT_IPIS - (*ipi_ast_counts[cpu])++; -#endif - /* Nothing to do for AST */ - } - if (ipi_bitmap & (1 << IPI_HARDCLOCK)) { -#ifdef COUNT_IPIS - (*ipi_hardclock_counts[cpu])++; -#endif - hardclockintr(); - } - td->td_intr_frame = oldframe; - td->td_intr_nesting_level--; - critical_exit(); + if (smp_started) + smp_tlb_shootdown(IPI_INVLCACHE, 0, 0); } /* - * send an IPI to a set of cpus. - */ -void -ipi_selected(cpuset_t cpus, u_int ipi) -{ - int cpu; - - /* - * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit - * of help in order to understand what is the source. - * Set the mask of receiving CPUs for this purpose. - */ - if (ipi == IPI_STOP_HARD) - CPU_OR_ATOMIC(&ipi_nmi_pending, &cpus); - - while ((cpu = CPU_FFS(&cpus)) != 0) { - cpu--; - CPU_CLR(cpu, &cpus); - CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); - ipi_send_cpu(cpu, ipi); - } -} - -/* - * send an IPI to a specific CPU. - */ -void -ipi_cpu(int cpu, u_int ipi) -{ - - /* - * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit - * of help in order to understand what is the source. - * Set the mask of receiving CPUs for this purpose. - */ - if (ipi == IPI_STOP_HARD) - CPU_SET_ATOMIC(cpu, &ipi_nmi_pending); - - CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); - ipi_send_cpu(cpu, ipi); -} - -/* - * send an IPI to all CPUs EXCEPT myself - */ -void -ipi_all_but_self(u_int ipi) -{ - cpuset_t other_cpus; - - other_cpus = all_cpus; - CPU_CLR(PCPU_GET(cpuid), &other_cpus); - if (IPI_IS_BITMAPED(ipi)) { - ipi_selected(other_cpus, ipi); - return; - } - - /* - * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit - * of help in order to understand what is the source. - * Set the mask of receiving CPUs for this purpose. - */ - if (ipi == IPI_STOP_HARD) - CPU_OR_ATOMIC(&ipi_nmi_pending, &other_cpus); - - CTR2(KTR_SMP, "%s: ipi: %x", __func__, ipi); - lapic_ipi_vectored(ipi, APIC_IPI_DEST_OTHERS); -} - -int -ipi_nmi_handler() -{ - u_int cpuid; - - /* - * As long as there is not a simple way to know about a NMI's - * source, if the bitmask for the current CPU is present in - * the global pending bitword an IPI_STOP_HARD has been issued - * and should be handled. - */ - cpuid = PCPU_GET(cpuid); - if (!CPU_ISSET(cpuid, &ipi_nmi_pending)) - return (1); - - CPU_CLR_ATOMIC(cpuid, &ipi_nmi_pending); - cpustop_handler(); - return (0); -} - -/* - * Handle an IPI_STOP by saving our current context and spinning until we - * are resumed. - */ -void -cpustop_handler(void) -{ - u_int cpu; - - cpu = PCPU_GET(cpuid); - - savectx(&stoppcbs[cpu]); - - /* Indicate that we are stopped */ - CPU_SET_ATOMIC(cpu, &stopped_cpus); - - /* Wait for restart */ - while (!CPU_ISSET(cpu, &started_cpus)) - ia32_pause(); - - CPU_CLR_ATOMIC(cpu, &started_cpus); - CPU_CLR_ATOMIC(cpu, &stopped_cpus); - - if (cpu == 0 && cpustop_restartfunc != NULL) { - cpustop_restartfunc(); - cpustop_restartfunc = NULL; - } -} - -/* - * Handle an IPI_SUSPEND by saving our current context and spinning until we - * are resumed. - */ -void -cpususpend_handler(void) -{ - u_int cpu; - - mtx_assert(&smp_ipi_mtx, MA_NOTOWNED); - - cpu = PCPU_GET(cpuid); - if (savectx(&susppcbs[cpu]->sp_pcb)) { - npxsuspend(susppcbs[cpu]->sp_fpususpend); - wbinvd(); - CPU_SET_ATOMIC(cpu, &suspended_cpus); - } else { - npxresume(susppcbs[cpu]->sp_fpususpend); - pmap_init_pat(); - initializecpu(); - PCPU_SET(switchtime, 0); - PCPU_SET(switchticks, ticks); - - /* Indicate that we are resumed */ - CPU_CLR_ATOMIC(cpu, &suspended_cpus); - } - - /* Wait for resume */ - while (!CPU_ISSET(cpu, &started_cpus)) - ia32_pause(); - - if (cpu_ops.cpu_resume) - cpu_ops.cpu_resume(); - - /* Resume MCA and local APIC */ - lapic_xapic_mode(); - mca_resume(); - lapic_setup(0); - - /* Indicate that we are resumed */ - CPU_CLR_ATOMIC(cpu, &suspended_cpus); - CPU_CLR_ATOMIC(cpu, &started_cpus); -} - -/* * Handlers for TLB related IPIs */ void invltlb_handler(void) { uint64_t cr3; #ifdef COUNT_XINVLTLB_HITS xhits_gbl[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invltlb_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ cr3 = rcr3(); load_cr3(cr3); atomic_add_int(&smp_tlb_wait, 1); } void invlpg_handler(void) { #ifdef COUNT_XINVLTLB_HITS xhits_pg[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invlpg_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ invlpg(smp_tlb_addr1); atomic_add_int(&smp_tlb_wait, 1); } void invlrng_handler(void) { vm_offset_t addr; #ifdef COUNT_XINVLTLB_HITS xhits_rng[PCPU_GET(cpuid)]++; #endif /* COUNT_XINVLTLB_HITS */ #ifdef COUNT_IPIS (*ipi_invlrng_counts[PCPU_GET(cpuid)])++; #endif /* COUNT_IPIS */ addr = smp_tlb_addr1; do { invlpg(addr); addr += PAGE_SIZE; } while (addr < smp_tlb_addr2); atomic_add_int(&smp_tlb_wait, 1); } - -void -invlcache_handler(void) -{ -#ifdef COUNT_IPIS - (*ipi_invlcache_counts[PCPU_GET(cpuid)])++; -#endif /* COUNT_IPIS */ - - wbinvd(); - atomic_add_int(&smp_tlb_wait, 1); -} - -/* - * This is called once the rest of the system is up and running and we're - * ready to let the AP's out of the pen. - */ -static void -release_aps(void *dummy __unused) -{ - - if (mp_ncpus == 1) - return; - atomic_store_rel_int(&aps_ready, 1); - while (smp_started == 0) - ia32_pause(); -} -SYSINIT(start_aps, SI_SUB_SMP, SI_ORDER_FIRST, release_aps, NULL); - -#ifdef COUNT_IPIS -/* - * Setup interrupt counters for IPI handlers. - */ -static void -mp_ipi_intrcnt(void *dummy) -{ - char buf[64]; - int i; - - CPU_FOREACH(i) { - snprintf(buf, sizeof(buf), "cpu%d:invltlb", i); - intrcnt_add(buf, &ipi_invltlb_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:invlrng", i); - intrcnt_add(buf, &ipi_invlrng_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:invlpg", i); - intrcnt_add(buf, &ipi_invlpg_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:invlcache", i); - intrcnt_add(buf, &ipi_invlcache_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:preempt", i); - intrcnt_add(buf, &ipi_preempt_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:ast", i); - intrcnt_add(buf, &ipi_ast_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:rendezvous", i); - intrcnt_add(buf, &ipi_rendezvous_counts[i]); - snprintf(buf, sizeof(buf), "cpu%d:hardclock", i); - intrcnt_add(buf, &ipi_hardclock_counts[i]); - } -} -SYSINIT(mp_ipi_intrcnt, SI_SUB_INTR, SI_ORDER_MIDDLE, mp_ipi_intrcnt, NULL); -#endif Index: user/ngie/more-tests/sys/i386/include/smp.h =================================================================== --- user/ngie/more-tests/sys/i386/include/smp.h (revision 281974) +++ user/ngie/more-tests/sys/i386/include/smp.h (revision 281975) @@ -1,94 +1,131 @@ /*- * ---------------------------------------------------------------------------- * "THE BEER-WARE LICENSE" (Revision 42): * wrote this file. As long as you retain this notice you * can do whatever you want with this stuff. If we meet some day, and you think * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp * ---------------------------------------------------------------------------- * * $FreeBSD$ * */ #ifndef _MACHINE_SMP_H_ #define _MACHINE_SMP_H_ #ifdef _KERNEL #ifdef SMP #ifndef LOCORE #include #include #include #include #include /* global data in mpboot.s */ extern int bootMP_size; /* functions in mpboot.s */ void bootMP(void); /* global data in mp_machdep.c */ extern int mp_naps; extern int boot_cpu_id; extern struct pcb stoppcbs[]; extern int cpu_apic_ids[]; +extern int bootAP; +extern void *dpcpu; +extern char *bootSTK; +extern int bootAP; +extern void *bootstacks[]; +extern volatile u_int cpu_ipi_pending[]; +extern volatile int aps_ready; +extern struct mtx ap_boot_mtx; +extern int cpu_logical; +extern int cpu_cores; +extern volatile int smp_tlb_wait; +extern u_int xhits_gbl[]; +extern u_int xhits_pg[]; +extern u_int xhits_rng[]; +extern u_int ipi_global; +extern u_int ipi_page; +extern u_int ipi_range; +extern u_int ipi_range_size; +extern u_int ipi_masked_global; +extern u_int ipi_masked_page; +extern u_int ipi_masked_range; +extern u_int ipi_masked_range_size; + +struct cpu_info { + int cpu_present:1; + int cpu_bsp:1; + int cpu_disabled:1; + int cpu_hyperthread:1; +}; +extern struct cpu_info cpu_info[]; + #ifdef COUNT_IPIS extern u_long *ipi_invltlb_counts[MAXCPU]; extern u_long *ipi_invlrng_counts[MAXCPU]; extern u_long *ipi_invlpg_counts[MAXCPU]; extern u_long *ipi_invlcache_counts[MAXCPU]; extern u_long *ipi_rendezvous_counts[MAXCPU]; #endif /* IPI handlers */ inthand_t IDTVEC(invltlb), /* TLB shootdowns - global */ IDTVEC(invlpg), /* TLB shootdowns - 1 page */ IDTVEC(invlrng), /* TLB shootdowns - page range */ IDTVEC(invlcache), /* Write back and invalidate cache */ IDTVEC(ipi_intr_bitmap_handler), /* Bitmap based IPIs */ IDTVEC(cpustop), /* CPU stops & waits to be restarted */ IDTVEC(cpususpend), /* CPU suspends & waits to be resumed */ IDTVEC(rendezvous); /* handle CPU rendezvous */ /* functions in mp_machdep.c */ +void assign_cpu_ids(void); void cpu_add(u_int apic_id, char boot_cpu); void cpustop_handler(void); #ifndef XEN void cpususpend_handler(void); #endif +void init_secondary_tail(void); void invltlb_handler(void); void invlpg_handler(void); void invlrng_handler(void); void invlcache_handler(void); void init_secondary(void); void ipi_startup(int apic_id, int vector); void ipi_all_but_self(u_int ipi); #ifndef XEN void ipi_bitmap_handler(struct trapframe frame); #endif void ipi_cpu(int cpu, u_int ipi); int ipi_nmi_handler(void); void ipi_selected(cpuset_t cpus, u_int ipi); u_int mp_bootaddress(u_int); +void set_interrupt_apic_ids(void); void smp_cache_flush(void); void smp_invlpg(vm_offset_t addr); void smp_masked_invlpg(cpuset_t mask, vm_offset_t addr); void smp_invlpg_range(vm_offset_t startva, vm_offset_t endva); void smp_masked_invlpg_range(cpuset_t mask, vm_offset_t startva, vm_offset_t endva); void smp_invltlb(void); void smp_masked_invltlb(cpuset_t mask); +void mem_range_AP_init(void); +void topo_probe(void); +void ipi_send_cpu(int cpu, u_int ipi); #ifdef XEN void ipi_to_irq_init(void); #endif #endif /* !LOCORE */ #endif /* SMP */ #endif /* _KERNEL */ #endif /* _MACHINE_SMP_H_ */ Index: user/ngie/more-tests/sys/i386/xen/mp_machdep.c =================================================================== --- user/ngie/more-tests/sys/i386/xen/mp_machdep.c (revision 281974) +++ user/ngie/more-tests/sys/i386/xen/mp_machdep.c (revision 281975) @@ -1,1300 +1,1292 @@ /*- * Copyright (c) 1996, by Steve Passe * Copyright (c) 2008, by Kip Macy * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. The name of the developer may NOT be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_apic.h" #include "opt_cpu.h" #include "opt_kstack_pages.h" #include "opt_mp_watchdog.h" #include "opt_pmap.h" #include "opt_sched.h" #include "opt_smp.h" #if !defined(lint) #if !defined(SMP) #error How did you get here? #endif #ifndef DEV_APIC #error The apic device is required for SMP, add "device apic" to your config file. #endif #if defined(CPU_DISABLE_CMPXCHG) && !defined(COMPILING_LINT) #error SMP not supported with CPU_DISABLE_CMPXCHG #endif #endif /* not lint */ #include #include #include #include /* cngetc() */ #include #ifdef GPROF #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /*---------------------------- Extern Declarations ---------------------------*/ extern struct pcpu __pcpu[]; extern void Xhypervisor_callback(void); extern void failsafe_callback(void); /*--------------------------- Forward Declarations ---------------------------*/ static driver_filter_t smp_reschedule_interrupt; static driver_filter_t smp_call_function_interrupt; -static void assign_cpu_ids(void); -static void set_interrupt_apic_ids(void); static int start_all_aps(void); static int start_ap(int apic_id); static void release_aps(void *dummy); /*---------------------------------- Macros ----------------------------------*/ #define IPI_TO_IDX(ipi) ((ipi) - APIC_IPI_INTS) /*-------------------------------- Local Types -------------------------------*/ typedef void call_data_func_t(uintptr_t , uintptr_t); -struct cpu_info { - int cpu_present:1; - int cpu_bsp:1; - int cpu_disabled:1; -}; - struct xen_ipi_handler { driver_filter_t *filter; const char *description; }; enum { RESCHEDULE_VECTOR, CALL_FUNCTION_VECTOR, }; /*-------------------------------- Global Data -------------------------------*/ static u_int hyperthreading_cpus; static cpuset_t hyperthreading_cpus_mask; int mp_naps; /* # of Applications processors */ int boot_cpu_id = -1; /* designated BSP */ -static int bootAP; +int bootAP; static union descriptor *bootAPgdt; /* Free these after use */ void *bootstacks[MAXCPU]; struct pcb stoppcbs[MAXCPU]; /* Variables needed for SMP tlb shootdown. */ vm_offset_t smp_tlb_addr1; vm_offset_t smp_tlb_addr2; volatile int smp_tlb_wait; static u_int logical_cpus; static volatile cpuset_t ipi_nmi_pending; /* used to hold the AP's until we are ready to release them */ -static struct mtx ap_boot_mtx; +struct mtx ap_boot_mtx; /* Set to 1 once we're ready to let the APs out of the pen. */ -static volatile int aps_ready = 0; +volatile int aps_ready = 0; /* * Store data from cpu_add() until later in the boot when we actually setup * the APs. */ -static struct cpu_info cpu_info[MAX_APIC_ID + 1]; +struct cpu_info cpu_info[MAX_APIC_ID + 1]; int cpu_apic_ids[MAXCPU]; int apic_cpuids[MAX_APIC_ID + 1]; /* Holds pending bitmap based IPIs per CPU */ -static volatile u_int cpu_ipi_pending[MAXCPU]; +volatile u_int cpu_ipi_pending[MAXCPU]; -static int cpu_logical; -static int cpu_cores; +int cpu_logical; +int cpu_cores; static const struct xen_ipi_handler xen_ipis[] = { [RESCHEDULE_VECTOR] = { smp_reschedule_interrupt, "resched" }, [CALL_FUNCTION_VECTOR] = { smp_call_function_interrupt,"callfunc" } }; /*------------------------------- Per-CPU Data -------------------------------*/ DPCPU_DEFINE(xen_intr_handle_t, ipi_handle[nitems(xen_ipis)]); DPCPU_DEFINE(struct vcpu_info *, vcpu_info); /*------------------------------ Implementation ------------------------------*/ struct cpu_group * cpu_topo(void) { if (cpu_cores == 0) cpu_cores = 1; if (cpu_logical == 0) cpu_logical = 1; if (mp_ncpus % (cpu_cores * cpu_logical) != 0) { printf("WARNING: Non-uniform processors.\n"); printf("WARNING: Using suboptimal topology.\n"); return (smp_topo_none()); } /* * No multi-core or hyper-threaded. */ if (cpu_logical * cpu_cores == 1) return (smp_topo_none()); /* * Only HTT no multi-core. */ if (cpu_logical > 1 && cpu_cores == 1) return (smp_topo_1level(CG_SHARE_L1, cpu_logical, CG_FLAG_HTT)); /* * Only multi-core no HTT. */ if (cpu_cores > 1 && cpu_logical == 1) return (smp_topo_1level(CG_SHARE_NONE, cpu_cores, 0)); /* * Both HTT and multi-core. */ return (smp_topo_2level(CG_SHARE_NONE, cpu_cores, CG_SHARE_L1, cpu_logical, CG_FLAG_HTT)); } /* * Calculate usable address in base memory for AP trampoline code. */ u_int mp_bootaddress(u_int basemem) { return (basemem); } void cpu_add(u_int apic_id, char boot_cpu) { if (apic_id > MAX_APIC_ID) { panic("SMP: APIC ID %d too high", apic_id); return; } KASSERT(cpu_info[apic_id].cpu_present == 0, ("CPU %d added twice", apic_id)); cpu_info[apic_id].cpu_present = 1; if (boot_cpu) { KASSERT(boot_cpu_id == -1, ("CPU %d claims to be BSP, but CPU %d already is", apic_id, boot_cpu_id)); boot_cpu_id = apic_id; cpu_info[apic_id].cpu_bsp = 1; } if (mp_ncpus < MAXCPU) mp_ncpus++; if (bootverbose) printf("SMP: Added CPU %d (%s)\n", apic_id, boot_cpu ? "BSP" : "AP"); } void cpu_mp_setmaxid(void) { mp_maxid = MAXCPU - 1; } int cpu_mp_probe(void) { /* * Always record BSP in CPU map so that the mbuf init code works * correctly. */ CPU_SETOF(0, &all_cpus); if (mp_ncpus == 0) { /* * No CPUs were found, so this must be a UP system. Setup * the variables to represent a system with a single CPU * with an id of 0. */ mp_ncpus = 1; return (0); } /* At least one CPU was found. */ if (mp_ncpus == 1) { /* * One CPU was found, so this must be a UP system with * an I/O APIC. */ return (0); } /* At least two CPUs were found. */ return (1); } /* * Initialize the IPI handlers and start up the AP's. */ void cpu_mp_start(void) { int i; /* Initialize the logical ID to APIC ID table. */ for (i = 0; i < MAXCPU; i++) { cpu_apic_ids[i] = -1; cpu_ipi_pending[i] = 0; } /* Set boot_cpu_id if needed. */ if (boot_cpu_id == -1) { boot_cpu_id = PCPU_GET(apic_id); cpu_info[boot_cpu_id].cpu_bsp = 1; } else KASSERT(boot_cpu_id == PCPU_GET(apic_id), ("BSP's APIC ID doesn't match boot_cpu_id")); cpu_apic_ids[0] = boot_cpu_id; apic_cpuids[boot_cpu_id] = 0; assign_cpu_ids(); /* Start each Application Processor */ start_all_aps(); /* Setup the initial logical CPUs info. */ logical_cpus = 0; CPU_ZERO(&logical_cpus_mask); if (cpu_feature & CPUID_HTT) logical_cpus = (cpu_procinfo & CPUID_HTT_CORES) >> 16; set_interrupt_apic_ids(); } static void iv_rendezvous(uintptr_t a, uintptr_t b) { smp_rendezvous_action(); } static void iv_invltlb(uintptr_t a, uintptr_t b) { xen_tlb_flush(); } static void iv_invlpg(uintptr_t a, uintptr_t b) { xen_invlpg(a); } static void iv_invlrng(uintptr_t a, uintptr_t b) { vm_offset_t start = (vm_offset_t)a; vm_offset_t end = (vm_offset_t)b; while (start < end) { xen_invlpg(start); start += PAGE_SIZE; } } static void iv_invlcache(uintptr_t a, uintptr_t b) { wbinvd(); atomic_add_int(&smp_tlb_wait, 1); } /* * These start from "IPI offset" APIC_IPI_INTS */ static call_data_func_t *ipi_vectors[5] = { iv_rendezvous, iv_invltlb, iv_invlpg, iv_invlrng, iv_invlcache, }; /* * Reschedule call back. Nothing to do, * all the work is done automatically when * we return from the interrupt. */ static int smp_reschedule_interrupt(void *unused) { int cpu = PCPU_GET(cpuid); u_int ipi_bitmap; ipi_bitmap = atomic_readandclear_int(&cpu_ipi_pending[cpu]); if (ipi_bitmap & (1 << IPI_PREEMPT)) { #ifdef COUNT_IPIS (*ipi_preempt_counts[cpu])++; #endif sched_preempt(curthread); } if (ipi_bitmap & (1 << IPI_AST)) { #ifdef COUNT_IPIS (*ipi_ast_counts[cpu])++; #endif /* Nothing to do for AST */ } return (FILTER_HANDLED); } struct _call_data { uint16_t func_id; uint16_t wait; uintptr_t arg1; uintptr_t arg2; atomic_t started; atomic_t finished; }; static struct _call_data *call_data; static int smp_call_function_interrupt(void *unused) { call_data_func_t *func; uintptr_t arg1 = call_data->arg1; uintptr_t arg2 = call_data->arg2; int wait = call_data->wait; atomic_t *started = &call_data->started; atomic_t *finished = &call_data->finished; /* We only handle function IPIs, not bitmap IPIs */ if (call_data->func_id < APIC_IPI_INTS || call_data->func_id > IPI_BITMAP_VECTOR) panic("invalid function id %u", call_data->func_id); func = ipi_vectors[IPI_TO_IDX(call_data->func_id)]; /* * Notify initiating CPU that I've grabbed the data and am * about to execute the function */ mb(); atomic_inc(started); /* * At this point the info structure may be out of scope unless wait==1 */ (*func)(arg1, arg2); if (wait) { mb(); atomic_inc(finished); } atomic_add_int(&smp_tlb_wait, 1); return (FILTER_HANDLED); } /* * Print various information about the SMP system hardware and setup. */ void cpu_mp_announce(void) { int i, x; /* List CPUs */ printf(" cpu0 (BSP): APIC ID: %2d\n", boot_cpu_id); for (i = 1, x = 0; x <= MAX_APIC_ID; x++) { if (!cpu_info[x].cpu_present || cpu_info[x].cpu_bsp) continue; if (cpu_info[x].cpu_disabled) printf(" cpu (AP): APIC ID: %2d (disabled)\n", x); else { KASSERT(i < mp_ncpus, ("mp_ncpus and actual cpus are out of whack")); printf(" cpu%d (AP): APIC ID: %2d\n", i++, x); } } } static int xen_smp_cpu_init(unsigned int cpu) { xen_intr_handle_t *ipi_handle; const struct xen_ipi_handler *ipi; int idx, rc; ipi_handle = DPCPU_ID_GET(cpu, ipi_handle); for (ipi = xen_ipis, idx = 0; idx < nitems(xen_ipis); ipi++, idx++) { /* * The PCPU variable pc_device is not initialized on i386 PV, * so we have to use the root_bus device in order to setup * the IPIs. */ rc = xen_intr_alloc_and_bind_ipi(root_bus, cpu, ipi->filter, INTR_TYPE_TTY, &ipi_handle[idx]); if (rc != 0) { printf("Unable to allocate a XEN IPI port. " "Error %d\n", rc); break; } xen_intr_describe(ipi_handle[idx], "%s", ipi->description); } for (;idx < nitems(xen_ipis); idx++) ipi_handle[idx] = NULL; if (rc == 0) return (0); /* Either all are successfully mapped, or none at all. */ for (idx = 0; idx < nitems(xen_ipis); idx++) { if (ipi_handle[idx] == NULL) continue; xen_intr_unbind(ipi_handle[idx]); ipi_handle[idx] = NULL; } return (rc); } static void xen_smp_intr_init_cpus(void *unused) { int i; for (i = 0; i < mp_ncpus; i++) xen_smp_cpu_init(i); } static void xen_smp_intr_setup_cpus(void *unused) { int i; for (i = 0; i < mp_ncpus; i++) DPCPU_ID_SET(i, vcpu_info, &HYPERVISOR_shared_info->vcpu_info[i]); } #define MTOPSIZE (1<<(14 + PAGE_SHIFT)) /* * AP CPU's call this to initialize themselves. */ void init_secondary(void) { vm_offset_t addr; u_int cpuid; int gsel_tss; /* bootAP is set in start_ap() to our ID. */ PCPU_SET(currentldt, _default_ldt); gsel_tss = GSEL(GPROC0_SEL, SEL_KPL); #if 0 gdt[bootAP * NGDT + GPROC0_SEL].sd.sd_type = SDT_SYS386TSS; #endif PCPU_SET(common_tss.tss_esp0, 0); /* not used until after switch */ PCPU_SET(common_tss.tss_ss0, GSEL(GDATA_SEL, SEL_KPL)); PCPU_SET(common_tss.tss_ioopt, (sizeof (struct i386tss)) << 16); #if 0 PCPU_SET(tss_gdt, &gdt[bootAP * NGDT + GPROC0_SEL].sd); PCPU_SET(common_tssd, *PCPU_GET(tss_gdt)); #endif PCPU_SET(fsgs_gdt, &gdt[GUFS_SEL].sd); /* * Set to a known state: * Set by mpboot.s: CR0_PG, CR0_PE * Set by cpu_setregs: CR0_NE, CR0_MP, CR0_TS, CR0_WP, CR0_AM */ /* * signal our startup to the BSP. */ mp_naps++; /* Spin until the BSP releases the AP's. */ while (!aps_ready) ia32_pause(); /* BSP may have changed PTD while we were waiting */ invltlb(); for (addr = 0; addr < NKPT * NBPDR - 1; addr += PAGE_SIZE) invlpg(addr); #if 0 /* set up SSE/NX */ initializecpu(); #endif /* set up FPU state on the AP */ npxinit(false); #if 0 /* A quick check from sanity claus */ if (PCPU_GET(apic_id) != lapic_id()) { printf("SMP: cpuid = %d\n", PCPU_GET(cpuid)); printf("SMP: actual apic_id = %d\n", lapic_id()); printf("SMP: correct apic_id = %d\n", PCPU_GET(apic_id)); panic("cpuid mismatch! boom!!"); } #endif /* Initialize curthread. */ KASSERT(PCPU_GET(idlethread) != NULL, ("no idle thread")); PCPU_SET(curthread, PCPU_GET(idlethread)); mtx_lock_spin(&ap_boot_mtx); #if 0 /* Init local apic for irq's */ lapic_setup(1); #endif smp_cpus++; cpuid = PCPU_GET(cpuid); CTR1(KTR_SMP, "SMP: AP CPU #%d Launched", cpuid); printf("SMP: AP CPU #%d Launched!\n", cpuid); /* Determine if we are a logical CPU. */ if (logical_cpus > 1 && PCPU_GET(apic_id) % logical_cpus != 0) CPU_SET(cpuid, &logical_cpus_mask); /* Determine if we are a hyperthread. */ if (hyperthreading_cpus > 1 && PCPU_GET(apic_id) % hyperthreading_cpus != 0) CPU_SET(cpuid, &hyperthreading_cpus_mask); #if 0 if (bootverbose) lapic_dump("AP"); #endif if (smp_cpus == mp_ncpus) { /* enable IPI's, tlb shootdown, freezes etc */ atomic_store_rel_int(&smp_started, 1); } mtx_unlock_spin(&ap_boot_mtx); /* wait until all the AP's are up */ while (smp_started == 0) ia32_pause(); PCPU_SET(curthread, PCPU_GET(idlethread)); /* Start per-CPU event timers. */ cpu_initclocks_ap(); /* enter the scheduler */ sched_throw(NULL); panic("scheduler returned us to %s", __func__); /* NOTREACHED */ } /******************************************************************* * local functions and data */ /* * We tell the I/O APIC code about all the CPUs we want to receive * interrupts. If we don't want certain CPUs to receive IRQs we * can simply not tell the I/O APIC code about them in this function. * We also do not tell it about the BSP since it tells itself about * the BSP internally to work with UP kernels and on UP machines. */ -static void +void set_interrupt_apic_ids(void) { u_int i, apic_id; for (i = 0; i < MAXCPU; i++) { apic_id = cpu_apic_ids[i]; if (apic_id == -1) continue; if (cpu_info[apic_id].cpu_bsp) continue; if (cpu_info[apic_id].cpu_disabled) continue; /* Don't let hyperthreads service interrupts. */ if (hyperthreading_cpus > 1 && apic_id % hyperthreading_cpus != 0) continue; intr_add_cpu(i); } } /* * Assign logical CPU IDs to local APICs. */ -static void +void assign_cpu_ids(void) { u_int i; /* Check for explicitly disabled CPUs. */ for (i = 0; i <= MAX_APIC_ID; i++) { if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp) continue; /* Don't use this CPU if it has been disabled by a tunable. */ if (resource_disabled("lapic", i)) { cpu_info[i].cpu_disabled = 1; continue; } } /* * Assign CPU IDs to local APIC IDs and disable any CPUs * beyond MAXCPU. CPU 0 has already been assigned to the BSP, * so we only have to assign IDs for APs. */ mp_ncpus = 1; for (i = 0; i <= MAX_APIC_ID; i++) { if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp || cpu_info[i].cpu_disabled) continue; if (mp_ncpus < MAXCPU) { cpu_apic_ids[mp_ncpus] = i; apic_cpuids[i] = mp_ncpus; mp_ncpus++; } else cpu_info[i].cpu_disabled = 1; } KASSERT(mp_maxid >= mp_ncpus - 1, ("%s: counters out of sync: max %d, count %d", __func__, mp_maxid, mp_ncpus)); } /* * start each AP in our list */ /* Lowest 1MB is already mapped: don't touch*/ #define TMPMAP_START 1 int start_all_aps(void) { int x,apic_id, cpu; struct pcpu *pc; mtx_init(&ap_boot_mtx, "ap boot", NULL, MTX_SPIN); /* set up temporary P==V mapping for AP boot */ /* XXX this is a hack, we should boot the AP on its own stack/PTD */ /* start each AP */ for (cpu = 1; cpu < mp_ncpus; cpu++) { apic_id = cpu_apic_ids[cpu]; bootAP = cpu; bootAPgdt = gdt + (512*cpu); /* Get per-cpu data */ pc = &__pcpu[bootAP]; pcpu_init(pc, bootAP, sizeof(struct pcpu)); dpcpu_init((void *)kmem_malloc(kernel_arena, DPCPU_SIZE, M_WAITOK | M_ZERO), bootAP); pc->pc_apic_id = cpu_apic_ids[bootAP]; pc->pc_vcpu_id = cpu_apic_ids[bootAP]; pc->pc_prvspace = pc; pc->pc_curthread = 0; gdt_segs[GPRIV_SEL].ssd_base = (int) pc; gdt_segs[GPROC0_SEL].ssd_base = (int) &pc->pc_common_tss; PT_SET_MA(bootAPgdt, VTOM(bootAPgdt) | PG_V | PG_RW); bzero(bootAPgdt, PAGE_SIZE); for (x = 0; x < NGDT; x++) ssdtosd(&gdt_segs[x], &bootAPgdt[x].sd); PT_SET_MA(bootAPgdt, vtomach(bootAPgdt) | PG_V); #ifdef notyet if (HYPERVISOR_vcpu_op(VCPUOP_get_physid, cpu, &cpu_id) == 0) { apicid = xen_vcpu_physid_to_x86_apicid(cpu_id.phys_id); acpiid = xen_vcpu_physid_to_x86_acpiid(cpu_id.phys_id); #ifdef CONFIG_ACPI if (acpiid != 0xff) x86_acpiid_to_apicid[acpiid] = apicid; #endif } #endif /* attempt to start the Application Processor */ if (!start_ap(cpu)) { printf("AP #%d (PHY# %d) failed!\n", cpu, apic_id); /* better panic as the AP may be running loose */ printf("panic y/n? [y] "); if (cngetc() != 'n') panic("bye-bye"); } CPU_SET(cpu, &all_cpus); /* record AP in CPU map */ } pmap_invalidate_range(kernel_pmap, 0, NKPT * NBPDR - 1); /* number of APs actually started */ return (mp_naps); } extern uint8_t *pcpu_boot_stack; extern trap_info_t trap_table[]; static void smp_trap_init(trap_info_t *trap_ctxt) { const trap_info_t *t = trap_table; for (t = trap_table; t->address; t++) { trap_ctxt[t->vector].flags = t->flags; trap_ctxt[t->vector].cs = t->cs; trap_ctxt[t->vector].address = t->address; } } extern struct rwlock pvh_global_lock; extern int nkpt; static void cpu_initialize_context(unsigned int cpu) { /* vcpu_guest_context_t is too large to allocate on the stack. * Hence we allocate statically and protect it with a lock */ vm_page_t m[NPGPTD + 2]; static vcpu_guest_context_t ctxt; vm_offset_t boot_stack; vm_offset_t newPTD; vm_paddr_t ma[NPGPTD]; int i; /* * Page 0,[0-3] PTD * Page 1, [4] boot stack * Page [5] PDPT * */ for (i = 0; i < NPGPTD + 2; i++) { m[i] = vm_page_alloc(NULL, 0, VM_ALLOC_NORMAL | VM_ALLOC_NOOBJ | VM_ALLOC_WIRED | VM_ALLOC_ZERO); pmap_zero_page(m[i]); } boot_stack = kva_alloc(PAGE_SIZE); newPTD = kva_alloc(NPGPTD * PAGE_SIZE); ma[0] = VM_PAGE_TO_MACH(m[0])|PG_V; #ifdef PAE pmap_kenter(boot_stack, VM_PAGE_TO_PHYS(m[NPGPTD + 1])); for (i = 0; i < NPGPTD; i++) { ((vm_paddr_t *)boot_stack)[i] = ma[i] = VM_PAGE_TO_MACH(m[i])|PG_V; } #endif /* * Copy cpu0 IdlePTD to new IdlePTD - copying only * kernel mappings */ pmap_qenter(newPTD, m, 4); memcpy((uint8_t *)newPTD + KPTDI*sizeof(vm_paddr_t), (uint8_t *)PTOV(IdlePTD) + KPTDI*sizeof(vm_paddr_t), nkpt*sizeof(vm_paddr_t)); pmap_qremove(newPTD, 4); kva_free(newPTD, 4 * PAGE_SIZE); /* * map actual idle stack to boot_stack */ pmap_kenter(boot_stack, VM_PAGE_TO_PHYS(m[NPGPTD])); xen_pgdpt_pin(VM_PAGE_TO_MACH(m[NPGPTD + 1])); rw_wlock(&pvh_global_lock); for (i = 0; i < 4; i++) { int pdir = (PTDPTDI + i) / NPDEPG; int curoffset = (PTDPTDI + i) % NPDEPG; xen_queue_pt_update((vm_paddr_t) ((ma[pdir] & ~PG_V) + (curoffset*sizeof(vm_paddr_t))), ma[i]); } PT_UPDATES_FLUSH(); rw_wunlock(&pvh_global_lock); memset(&ctxt, 0, sizeof(ctxt)); ctxt.flags = VGCF_IN_KERNEL; ctxt.user_regs.ds = GSEL(GDATA_SEL, SEL_KPL); ctxt.user_regs.es = GSEL(GDATA_SEL, SEL_KPL); ctxt.user_regs.fs = GSEL(GPRIV_SEL, SEL_KPL); ctxt.user_regs.gs = GSEL(GDATA_SEL, SEL_KPL); ctxt.user_regs.cs = GSEL(GCODE_SEL, SEL_KPL); ctxt.user_regs.ss = GSEL(GDATA_SEL, SEL_KPL); ctxt.user_regs.eip = (unsigned long)init_secondary; ctxt.user_regs.eflags = PSL_KERNEL | 0x1000; /* IOPL_RING1 */ memset(&ctxt.fpu_ctxt, 0, sizeof(ctxt.fpu_ctxt)); smp_trap_init(ctxt.trap_ctxt); ctxt.ldt_ents = 0; ctxt.gdt_frames[0] = (uint32_t)((uint64_t)vtomach(bootAPgdt) >> PAGE_SHIFT); ctxt.gdt_ents = 512; #ifdef __i386__ ctxt.user_regs.esp = boot_stack + PAGE_SIZE; ctxt.kernel_ss = GSEL(GDATA_SEL, SEL_KPL); ctxt.kernel_sp = boot_stack + PAGE_SIZE; ctxt.event_callback_cs = GSEL(GCODE_SEL, SEL_KPL); ctxt.event_callback_eip = (unsigned long)Xhypervisor_callback; ctxt.failsafe_callback_cs = GSEL(GCODE_SEL, SEL_KPL); ctxt.failsafe_callback_eip = (unsigned long)failsafe_callback; ctxt.ctrlreg[3] = VM_PAGE_TO_MACH(m[NPGPTD + 1]); #else /* __x86_64__ */ ctxt.user_regs.esp = idle->thread.rsp0 - sizeof(struct pt_regs); ctxt.kernel_ss = GSEL(GDATA_SEL, SEL_KPL); ctxt.kernel_sp = idle->thread.rsp0; ctxt.event_callback_eip = (unsigned long)hypervisor_callback; ctxt.failsafe_callback_eip = (unsigned long)failsafe_callback; ctxt.syscall_callback_eip = (unsigned long)system_call; ctxt.ctrlreg[3] = xen_pfn_to_cr3(virt_to_mfn(init_level4_pgt)); ctxt.gs_base_kernel = (unsigned long)(cpu_pda(cpu)); #endif printf("gdtpfn=%lx pdptpfn=%lx\n", ctxt.gdt_frames[0], ctxt.ctrlreg[3] >> PAGE_SHIFT); PANIC_IF(HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, &ctxt)); DELAY(3000); PANIC_IF(HYPERVISOR_vcpu_op(VCPUOP_up, cpu, NULL)); } /* * This function starts the AP (application processor) identified * by the APIC ID 'physicalCpu'. It does quite a "song and dance" * to accomplish this. This is necessary because of the nuances * of the different hardware we might encounter. It isn't pretty, * but it seems to work. */ int cpus; static int start_ap(int apic_id) { int ms; /* used as a watchpoint to signal AP startup */ cpus = mp_naps; cpu_initialize_context(apic_id); /* Wait up to 5 seconds for it to start. */ for (ms = 0; ms < 5000; ms++) { if (mp_naps > cpus) return (1); /* return SUCCESS */ DELAY(1000); } return (0); /* return FAILURE */ } static void ipi_pcpu(int cpu, u_int ipi) { KASSERT((ipi <= nitems(xen_ipis)), ("invalid IPI")); xen_intr_signal(DPCPU_ID_GET(cpu, ipi_handle[ipi])); } /* * send an IPI to a specific CPU. */ -static void +void ipi_send_cpu(int cpu, u_int ipi) { u_int bitmap, old_pending, new_pending; if (IPI_IS_BITMAPED(ipi)) { bitmap = 1 << ipi; ipi = IPI_BITMAP_VECTOR; do { old_pending = cpu_ipi_pending[cpu]; new_pending = old_pending | bitmap; } while (!atomic_cmpset_int(&cpu_ipi_pending[cpu], old_pending, new_pending)); if (!old_pending) ipi_pcpu(cpu, RESCHEDULE_VECTOR); } else { KASSERT(call_data != NULL, ("call_data not set")); ipi_pcpu(cpu, CALL_FUNCTION_VECTOR); } } /* * Flush the TLB on all other CPU's */ static void smp_tlb_shootdown(u_int vector, vm_offset_t addr1, vm_offset_t addr2) { u_int ncpu; struct _call_data data; ncpu = mp_ncpus - 1; /* does not shootdown self */ if (ncpu < 1) return; /* no other cpus */ if (!(read_eflags() & PSL_I)) panic("%s: interrupts disabled", __func__); mtx_lock_spin(&smp_ipi_mtx); KASSERT(call_data == NULL, ("call_data isn't null?!")); call_data = &data; call_data->func_id = vector; call_data->arg1 = addr1; call_data->arg2 = addr2; atomic_store_rel_int(&smp_tlb_wait, 0); ipi_all_but_self(vector); while (smp_tlb_wait < ncpu) ia32_pause(); call_data = NULL; mtx_unlock_spin(&smp_ipi_mtx); } static void smp_targeted_tlb_shootdown(cpuset_t mask, u_int vector, vm_offset_t addr1, vm_offset_t addr2) { int cpu, ncpu, othercpus; struct _call_data data; othercpus = mp_ncpus - 1; if (CPU_ISFULLSET(&mask)) { if (othercpus < 1) return; } else { CPU_CLR(PCPU_GET(cpuid), &mask); if (CPU_EMPTY(&mask)) return; } if (!(read_eflags() & PSL_I)) panic("%s: interrupts disabled", __func__); mtx_lock_spin(&smp_ipi_mtx); KASSERT(call_data == NULL, ("call_data isn't null?!")); call_data = &data; call_data->func_id = vector; call_data->arg1 = addr1; call_data->arg2 = addr2; atomic_store_rel_int(&smp_tlb_wait, 0); if (CPU_ISFULLSET(&mask)) { ncpu = othercpus; ipi_all_but_self(vector); } else { ncpu = 0; while ((cpu = CPU_FFS(&mask)) != 0) { cpu--; CPU_CLR(cpu, &mask); CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, vector); ipi_send_cpu(cpu, vector); ncpu++; } } while (smp_tlb_wait < ncpu) ia32_pause(); call_data = NULL; mtx_unlock_spin(&smp_ipi_mtx); } void smp_cache_flush(void) { if (smp_started) smp_tlb_shootdown(IPI_INVLCACHE, 0, 0); } void smp_invltlb(void) { if (smp_started) { smp_tlb_shootdown(IPI_INVLTLB, 0, 0); } } void smp_invlpg(vm_offset_t addr) { if (smp_started) { smp_tlb_shootdown(IPI_INVLPG, addr, 0); } } void smp_invlpg_range(vm_offset_t addr1, vm_offset_t addr2) { if (smp_started) { smp_tlb_shootdown(IPI_INVLRNG, addr1, addr2); } } void smp_masked_invltlb(cpuset_t mask) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLTLB, 0, 0); } } void smp_masked_invlpg(cpuset_t mask, vm_offset_t addr) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLPG, addr, 0); } } void smp_masked_invlpg_range(cpuset_t mask, vm_offset_t addr1, vm_offset_t addr2) { if (smp_started) { smp_targeted_tlb_shootdown(mask, IPI_INVLRNG, addr1, addr2); } } /* * send an IPI to a set of cpus. */ void ipi_selected(cpuset_t cpus, u_int ipi) { int cpu; /* * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit * of help in order to understand what is the source. * Set the mask of receiving CPUs for this purpose. */ if (ipi == IPI_STOP_HARD) CPU_OR_ATOMIC(&ipi_nmi_pending, &cpus); while ((cpu = CPU_FFS(&cpus)) != 0) { cpu--; CPU_CLR(cpu, &cpus); CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); ipi_send_cpu(cpu, ipi); } } /* * send an IPI to a specific CPU. */ void ipi_cpu(int cpu, u_int ipi) { /* * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit * of help in order to understand what is the source. * Set the mask of receiving CPUs for this purpose. */ if (ipi == IPI_STOP_HARD) CPU_SET_ATOMIC(cpu, &ipi_nmi_pending); CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); ipi_send_cpu(cpu, ipi); } /* * send an IPI to all CPUs EXCEPT myself */ void ipi_all_but_self(u_int ipi) { cpuset_t other_cpus; /* * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit * of help in order to understand what is the source. * Set the mask of receiving CPUs for this purpose. */ other_cpus = all_cpus; CPU_CLR(PCPU_GET(cpuid), &other_cpus); if (ipi == IPI_STOP_HARD) CPU_OR_ATOMIC(&ipi_nmi_pending, &other_cpus); CTR2(KTR_SMP, "%s: ipi: %x", __func__, ipi); ipi_selected(other_cpus, ipi); } int ipi_nmi_handler() { u_int cpuid; /* * As long as there is not a simple way to know about a NMI's * source, if the bitmask for the current CPU is present in * the global pending bitword an IPI_STOP_HARD has been issued * and should be handled. */ cpuid = PCPU_GET(cpuid); if (!CPU_ISSET(cpuid, &ipi_nmi_pending)) return (1); CPU_CLR_ATOMIC(cpuid, &ipi_nmi_pending); cpustop_handler(); return (0); } /* * Handle an IPI_STOP by saving our current context and spinning until we * are resumed. */ void cpustop_handler(void) { int cpu; cpu = PCPU_GET(cpuid); savectx(&stoppcbs[cpu]); /* Indicate that we are stopped */ CPU_SET_ATOMIC(cpu, &stopped_cpus); /* Wait for restart */ while (!CPU_ISSET(cpu, &started_cpus)) ia32_pause(); CPU_CLR_ATOMIC(cpu, &started_cpus); CPU_CLR_ATOMIC(cpu, &stopped_cpus); if (cpu == 0 && cpustop_restartfunc != NULL) { cpustop_restartfunc(); cpustop_restartfunc = NULL; } } /* * Handlers for TLB related IPIs * * On i386 Xen PV this are no-ops since this port doesn't support SMP. */ void invltlb_handler(void) { } void invlpg_handler(void) { } void invlrng_handler(void) { } void invlcache_handler(void) { } /* * This is called once the rest of the system is up and running and we're * ready to let the AP's out of the pen. */ static void release_aps(void *dummy __unused) { if (mp_ncpus == 1) return; atomic_store_rel_int(&aps_ready, 1); while (smp_started == 0) ia32_pause(); } SYSINIT(start_aps, SI_SUB_SMP, SI_ORDER_FIRST, release_aps, NULL); SYSINIT(start_ipis, SI_SUB_SMP, SI_ORDER_ANY, xen_smp_intr_init_cpus, NULL); SYSINIT(start_cpu, SI_SUB_INTR, SI_ORDER_ANY, xen_smp_intr_setup_cpus, NULL); Index: user/ngie/more-tests/sys/kern/uipc_shm.c =================================================================== --- user/ngie/more-tests/sys/kern/uipc_shm.c (revision 281974) +++ user/ngie/more-tests/sys/kern/uipc_shm.c (revision 281975) @@ -1,1060 +1,1060 @@ /*- * Copyright (c) 2006, 2011 Robert N. M. Watson * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * Support for shared swap-backed anonymous memory objects via * shm_open(2) and shm_unlink(2). While most of the implementation is * here, vm_mmap.c contains mapping logic changes. * * TODO: * * (1) Need to export data to a userland tool via a sysctl. Should ipcs(1) * and ipcrm(1) be expanded or should new tools to manage both POSIX * kernel semaphores and POSIX shared memory be written? * * (2) Add support for this file type to fstat(1). * * (3) Resource limits? Does this need its own resource limits or are the * existing limits in mmap(2) sufficient? */ #include __FBSDID("$FreeBSD$"); #include "opt_capsicum.h" #include "opt_ktrace.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include struct shm_mapping { char *sm_path; Fnv32_t sm_fnv; struct shmfd *sm_shmfd; LIST_ENTRY(shm_mapping) sm_link; }; static MALLOC_DEFINE(M_SHMFD, "shmfd", "shared memory file descriptor"); static LIST_HEAD(, shm_mapping) *shm_dictionary; static struct sx shm_dict_lock; static struct mtx shm_timestamp_lock; static u_long shm_hash; static struct unrhdr *shm_ino_unr; static dev_t shm_dev_ino; #define SHM_HASH(fnv) (&shm_dictionary[(fnv) & shm_hash]) static int shm_access(struct shmfd *shmfd, struct ucred *ucred, int flags); static struct shmfd *shm_alloc(struct ucred *ucred, mode_t mode); static void shm_init(void *arg); static void shm_drop(struct shmfd *shmfd); static struct shmfd *shm_hold(struct shmfd *shmfd); static void shm_insert(char *path, Fnv32_t fnv, struct shmfd *shmfd); static struct shmfd *shm_lookup(char *path, Fnv32_t fnv); static int shm_remove(char *path, Fnv32_t fnv, struct ucred *ucred); static int shm_dotruncate(struct shmfd *shmfd, off_t length); static fo_rdwr_t shm_read; static fo_rdwr_t shm_write; static fo_truncate_t shm_truncate; static fo_stat_t shm_stat; static fo_close_t shm_close; static fo_chmod_t shm_chmod; static fo_chown_t shm_chown; static fo_seek_t shm_seek; static fo_fill_kinfo_t shm_fill_kinfo; /* File descriptor operations. */ static struct fileops shm_ops = { .fo_read = shm_read, .fo_write = shm_write, .fo_truncate = shm_truncate, .fo_ioctl = invfo_ioctl, .fo_poll = invfo_poll, .fo_kqfilter = invfo_kqfilter, .fo_stat = shm_stat, .fo_close = shm_close, .fo_chmod = shm_chmod, .fo_chown = shm_chown, .fo_sendfile = vn_sendfile, .fo_seek = shm_seek, .fo_fill_kinfo = shm_fill_kinfo, .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE }; FEATURE(posix_shm, "POSIX shared memory"); static int uiomove_object_page(vm_object_t obj, size_t len, struct uio *uio) { vm_page_t m; vm_pindex_t idx; size_t tlen; int error, offset, rv; idx = OFF_TO_IDX(uio->uio_offset); offset = uio->uio_offset & PAGE_MASK; tlen = MIN(PAGE_SIZE - offset, len); VM_OBJECT_WLOCK(obj); /* * Read I/O without either a corresponding resident page or swap * page: use zero_region. This is intended to avoid instantiating * pages on read from a sparse region. */ if (uio->uio_rw == UIO_READ && vm_page_lookup(obj, idx) == NULL && !vm_pager_has_page(obj, idx, NULL, NULL)) { VM_OBJECT_WUNLOCK(obj); - return (uiomove(__DECONST(void *, zero_region), len, uio)); + return (uiomove(__DECONST(void *, zero_region), tlen, uio)); } /* * Parallel reads of the page content from disk are prevented * by exclusive busy. * * Although the tmpfs vnode lock is held here, it is * nonetheless safe to sleep waiting for a free page. The * pageout daemon does not need to acquire the tmpfs vnode * lock to page out tobj's pages because tobj is a OBJT_SWAP * type object. */ m = vm_page_grab(obj, idx, VM_ALLOC_NORMAL); if (m->valid != VM_PAGE_BITS_ALL) { if (vm_pager_has_page(obj, idx, NULL, NULL)) { rv = vm_pager_get_pages(obj, &m, 1, 0); m = vm_page_lookup(obj, idx); if (m == NULL) { printf( "uiomove_object: vm_obj %p idx %jd null lookup rv %d\n", obj, idx, rv); VM_OBJECT_WUNLOCK(obj); return (EIO); } if (rv != VM_PAGER_OK) { printf( "uiomove_object: vm_obj %p idx %jd valid %x pager error %d\n", obj, idx, m->valid, rv); vm_page_lock(m); vm_page_free(m); vm_page_unlock(m); VM_OBJECT_WUNLOCK(obj); return (EIO); } } else vm_page_zero_invalid(m, TRUE); } vm_page_xunbusy(m); vm_page_lock(m); vm_page_hold(m); if (m->queue == PQ_NONE) { vm_page_deactivate(m); } else { /* Requeue to maintain LRU ordering. */ vm_page_requeue(m); } vm_page_unlock(m); VM_OBJECT_WUNLOCK(obj); error = uiomove_fromphys(&m, offset, tlen, uio); if (uio->uio_rw == UIO_WRITE && error == 0) { VM_OBJECT_WLOCK(obj); vm_page_dirty(m); vm_pager_page_unswapped(m); VM_OBJECT_WUNLOCK(obj); } vm_page_lock(m); vm_page_unhold(m); vm_page_unlock(m); return (error); } int uiomove_object(vm_object_t obj, off_t obj_size, struct uio *uio) { ssize_t resid; size_t len; int error; error = 0; while ((resid = uio->uio_resid) > 0) { if (obj_size <= uio->uio_offset) break; len = MIN(obj_size - uio->uio_offset, resid); if (len == 0) break; error = uiomove_object_page(obj, len, uio); if (error != 0 || resid == uio->uio_resid) break; } return (error); } static int shm_seek(struct file *fp, off_t offset, int whence, struct thread *td) { struct shmfd *shmfd; off_t foffset; int error; shmfd = fp->f_data; foffset = foffset_lock(fp, 0); error = 0; switch (whence) { case L_INCR: if (foffset < 0 || (offset > 0 && foffset > OFF_MAX - offset)) { error = EOVERFLOW; break; } offset += foffset; break; case L_XTND: if (offset > 0 && shmfd->shm_size > OFF_MAX - offset) { error = EOVERFLOW; break; } offset += shmfd->shm_size; break; case L_SET: break; default: error = EINVAL; } if (error == 0) { if (offset < 0 || offset > shmfd->shm_size) error = EINVAL; else td->td_uretoff.tdu_off = offset; } foffset_unlock(fp, offset, error != 0 ? FOF_NOUPDATE : 0); return (error); } static int shm_read(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct shmfd *shmfd; void *rl_cookie; int error; shmfd = fp->f_data; foffset_lock_uio(fp, uio, flags); rl_cookie = rangelock_rlock(&shmfd->shm_rl, uio->uio_offset, uio->uio_offset + uio->uio_resid, &shmfd->shm_mtx); #ifdef MAC error = mac_posixshm_check_read(active_cred, fp->f_cred, shmfd); if (error) return (error); #endif error = uiomove_object(shmfd->shm_object, shmfd->shm_size, uio); rangelock_unlock(&shmfd->shm_rl, rl_cookie, &shmfd->shm_mtx); foffset_unlock_uio(fp, uio, flags); return (error); } static int shm_write(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct shmfd *shmfd; void *rl_cookie; int error; shmfd = fp->f_data; #ifdef MAC error = mac_posixshm_check_write(active_cred, fp->f_cred, shmfd); if (error) return (error); #endif foffset_lock_uio(fp, uio, flags); if ((flags & FOF_OFFSET) == 0) { rl_cookie = rangelock_wlock(&shmfd->shm_rl, 0, OFF_MAX, &shmfd->shm_mtx); } else { rl_cookie = rangelock_wlock(&shmfd->shm_rl, uio->uio_offset, uio->uio_offset + uio->uio_resid, &shmfd->shm_mtx); } error = uiomove_object(shmfd->shm_object, shmfd->shm_size, uio); rangelock_unlock(&shmfd->shm_rl, rl_cookie, &shmfd->shm_mtx); foffset_unlock_uio(fp, uio, flags); return (error); } static int shm_truncate(struct file *fp, off_t length, struct ucred *active_cred, struct thread *td) { struct shmfd *shmfd; #ifdef MAC int error; #endif shmfd = fp->f_data; #ifdef MAC error = mac_posixshm_check_truncate(active_cred, fp->f_cred, shmfd); if (error) return (error); #endif return (shm_dotruncate(shmfd, length)); } static int shm_stat(struct file *fp, struct stat *sb, struct ucred *active_cred, struct thread *td) { struct shmfd *shmfd; #ifdef MAC int error; #endif shmfd = fp->f_data; #ifdef MAC error = mac_posixshm_check_stat(active_cred, fp->f_cred, shmfd); if (error) return (error); #endif /* * Attempt to return sanish values for fstat() on a memory file * descriptor. */ bzero(sb, sizeof(*sb)); sb->st_blksize = PAGE_SIZE; sb->st_size = shmfd->shm_size; sb->st_blocks = (sb->st_size + sb->st_blksize - 1) / sb->st_blksize; mtx_lock(&shm_timestamp_lock); sb->st_atim = shmfd->shm_atime; sb->st_ctim = shmfd->shm_ctime; sb->st_mtim = shmfd->shm_mtime; sb->st_birthtim = shmfd->shm_birthtime; sb->st_mode = S_IFREG | shmfd->shm_mode; /* XXX */ sb->st_uid = shmfd->shm_uid; sb->st_gid = shmfd->shm_gid; mtx_unlock(&shm_timestamp_lock); sb->st_dev = shm_dev_ino; sb->st_ino = shmfd->shm_ino; return (0); } static int shm_close(struct file *fp, struct thread *td) { struct shmfd *shmfd; shmfd = fp->f_data; fp->f_data = NULL; shm_drop(shmfd); return (0); } static int shm_dotruncate(struct shmfd *shmfd, off_t length) { vm_object_t object; vm_page_t m, ma[1]; vm_pindex_t idx, nobjsize; vm_ooffset_t delta; int base, rv; object = shmfd->shm_object; VM_OBJECT_WLOCK(object); if (length == shmfd->shm_size) { VM_OBJECT_WUNLOCK(object); return (0); } nobjsize = OFF_TO_IDX(length + PAGE_MASK); /* Are we shrinking? If so, trim the end. */ if (length < shmfd->shm_size) { /* * Disallow any requests to shrink the size if this * object is mapped into the kernel. */ if (shmfd->shm_kmappings > 0) { VM_OBJECT_WUNLOCK(object); return (EBUSY); } /* * Zero the truncated part of the last page. */ base = length & PAGE_MASK; if (base != 0) { idx = OFF_TO_IDX(length); retry: m = vm_page_lookup(object, idx); if (m != NULL) { if (vm_page_sleep_if_busy(m, "shmtrc")) goto retry; } else if (vm_pager_has_page(object, idx, NULL, NULL)) { m = vm_page_alloc(object, idx, VM_ALLOC_NORMAL); if (m == NULL) { VM_OBJECT_WUNLOCK(object); VM_WAIT; VM_OBJECT_WLOCK(object); goto retry; } else if (m->valid != VM_PAGE_BITS_ALL) { ma[0] = m; rv = vm_pager_get_pages(object, ma, 1, 0); m = vm_page_lookup(object, idx); } else /* A cached page was reactivated. */ rv = VM_PAGER_OK; vm_page_lock(m); if (rv == VM_PAGER_OK) { vm_page_deactivate(m); vm_page_unlock(m); vm_page_xunbusy(m); } else { vm_page_free(m); vm_page_unlock(m); VM_OBJECT_WUNLOCK(object); return (EIO); } } if (m != NULL) { pmap_zero_page_area(m, base, PAGE_SIZE - base); KASSERT(m->valid == VM_PAGE_BITS_ALL, ("shm_dotruncate: page %p is invalid", m)); vm_page_dirty(m); vm_pager_page_unswapped(m); } } delta = ptoa(object->size - nobjsize); /* Toss in memory pages. */ if (nobjsize < object->size) vm_object_page_remove(object, nobjsize, object->size, 0); /* Toss pages from swap. */ if (object->type == OBJT_SWAP) swap_pager_freespace(object, nobjsize, delta); /* Free the swap accounted for shm */ swap_release_by_cred(delta, object->cred); object->charge -= delta; } else { /* Attempt to reserve the swap */ delta = ptoa(nobjsize - object->size); if (!swap_reserve_by_cred(delta, object->cred)) { VM_OBJECT_WUNLOCK(object); return (ENOMEM); } object->charge += delta; } shmfd->shm_size = length; mtx_lock(&shm_timestamp_lock); vfs_timestamp(&shmfd->shm_ctime); shmfd->shm_mtime = shmfd->shm_ctime; mtx_unlock(&shm_timestamp_lock); object->size = nobjsize; VM_OBJECT_WUNLOCK(object); return (0); } /* * shmfd object management including creation and reference counting * routines. */ static struct shmfd * shm_alloc(struct ucred *ucred, mode_t mode) { struct shmfd *shmfd; int ino; shmfd = malloc(sizeof(*shmfd), M_SHMFD, M_WAITOK | M_ZERO); shmfd->shm_size = 0; shmfd->shm_uid = ucred->cr_uid; shmfd->shm_gid = ucred->cr_gid; shmfd->shm_mode = mode; shmfd->shm_object = vm_pager_allocate(OBJT_DEFAULT, NULL, shmfd->shm_size, VM_PROT_DEFAULT, 0, ucred); KASSERT(shmfd->shm_object != NULL, ("shm_create: vm_pager_allocate")); shmfd->shm_object->pg_color = 0; VM_OBJECT_WLOCK(shmfd->shm_object); vm_object_clear_flag(shmfd->shm_object, OBJ_ONEMAPPING); vm_object_set_flag(shmfd->shm_object, OBJ_COLORED | OBJ_NOSPLIT); VM_OBJECT_WUNLOCK(shmfd->shm_object); vfs_timestamp(&shmfd->shm_birthtime); shmfd->shm_atime = shmfd->shm_mtime = shmfd->shm_ctime = shmfd->shm_birthtime; ino = alloc_unr(shm_ino_unr); if (ino == -1) shmfd->shm_ino = 0; else shmfd->shm_ino = ino; refcount_init(&shmfd->shm_refs, 1); mtx_init(&shmfd->shm_mtx, "shmrl", NULL, MTX_DEF); rangelock_init(&shmfd->shm_rl); #ifdef MAC mac_posixshm_init(shmfd); mac_posixshm_create(ucred, shmfd); #endif return (shmfd); } static struct shmfd * shm_hold(struct shmfd *shmfd) { refcount_acquire(&shmfd->shm_refs); return (shmfd); } static void shm_drop(struct shmfd *shmfd) { if (refcount_release(&shmfd->shm_refs)) { #ifdef MAC mac_posixshm_destroy(shmfd); #endif rangelock_destroy(&shmfd->shm_rl); mtx_destroy(&shmfd->shm_mtx); vm_object_deallocate(shmfd->shm_object); if (shmfd->shm_ino != 0) free_unr(shm_ino_unr, shmfd->shm_ino); free(shmfd, M_SHMFD); } } /* * Determine if the credentials have sufficient permissions for a * specified combination of FREAD and FWRITE. */ static int shm_access(struct shmfd *shmfd, struct ucred *ucred, int flags) { accmode_t accmode; int error; accmode = 0; if (flags & FREAD) accmode |= VREAD; if (flags & FWRITE) accmode |= VWRITE; mtx_lock(&shm_timestamp_lock); error = vaccess(VREG, shmfd->shm_mode, shmfd->shm_uid, shmfd->shm_gid, accmode, ucred, NULL); mtx_unlock(&shm_timestamp_lock); return (error); } /* * Dictionary management. We maintain an in-kernel dictionary to map * paths to shmfd objects. We use the FNV hash on the path to store * the mappings in a hash table. */ static void shm_init(void *arg) { mtx_init(&shm_timestamp_lock, "shm timestamps", NULL, MTX_DEF); sx_init(&shm_dict_lock, "shm dictionary"); shm_dictionary = hashinit(1024, M_SHMFD, &shm_hash); shm_ino_unr = new_unrhdr(1, INT32_MAX, NULL); KASSERT(shm_ino_unr != NULL, ("shm fake inodes not initialized")); shm_dev_ino = devfs_alloc_cdp_inode(); KASSERT(shm_dev_ino > 0, ("shm dev inode not initialized")); } SYSINIT(shm_init, SI_SUB_SYSV_SHM, SI_ORDER_ANY, shm_init, NULL); static struct shmfd * shm_lookup(char *path, Fnv32_t fnv) { struct shm_mapping *map; LIST_FOREACH(map, SHM_HASH(fnv), sm_link) { if (map->sm_fnv != fnv) continue; if (strcmp(map->sm_path, path) == 0) return (map->sm_shmfd); } return (NULL); } static void shm_insert(char *path, Fnv32_t fnv, struct shmfd *shmfd) { struct shm_mapping *map; map = malloc(sizeof(struct shm_mapping), M_SHMFD, M_WAITOK); map->sm_path = path; map->sm_fnv = fnv; map->sm_shmfd = shm_hold(shmfd); shmfd->shm_path = path; LIST_INSERT_HEAD(SHM_HASH(fnv), map, sm_link); } static int shm_remove(char *path, Fnv32_t fnv, struct ucred *ucred) { struct shm_mapping *map; int error; LIST_FOREACH(map, SHM_HASH(fnv), sm_link) { if (map->sm_fnv != fnv) continue; if (strcmp(map->sm_path, path) == 0) { #ifdef MAC error = mac_posixshm_check_unlink(ucred, map->sm_shmfd); if (error) return (error); #endif error = shm_access(map->sm_shmfd, ucred, FREAD | FWRITE); if (error) return (error); map->sm_shmfd->shm_path = NULL; LIST_REMOVE(map, sm_link); shm_drop(map->sm_shmfd); free(map->sm_path, M_SHMFD); free(map, M_SHMFD); return (0); } } return (ENOENT); } /* System calls. */ int sys_shm_open(struct thread *td, struct shm_open_args *uap) { struct filedesc *fdp; struct shmfd *shmfd; struct file *fp; char *path; Fnv32_t fnv; mode_t cmode; int fd, error; #ifdef CAPABILITY_MODE /* * shm_open(2) is only allowed for anonymous objects. */ if (IN_CAPABILITY_MODE(td) && (uap->path != SHM_ANON)) return (ECAPMODE); #endif if ((uap->flags & O_ACCMODE) != O_RDONLY && (uap->flags & O_ACCMODE) != O_RDWR) return (EINVAL); if ((uap->flags & ~(O_ACCMODE | O_CREAT | O_EXCL | O_TRUNC | O_CLOEXEC)) != 0) return (EINVAL); fdp = td->td_proc->p_fd; cmode = (uap->mode & ~fdp->fd_cmask) & ACCESSPERMS; error = falloc(td, &fp, &fd, O_CLOEXEC); if (error) return (error); /* A SHM_ANON path pointer creates an anonymous object. */ if (uap->path == SHM_ANON) { /* A read-only anonymous object is pointless. */ if ((uap->flags & O_ACCMODE) == O_RDONLY) { fdclose(td, fp, fd); fdrop(fp, td); return (EINVAL); } shmfd = shm_alloc(td->td_ucred, cmode); } else { path = malloc(MAXPATHLEN, M_SHMFD, M_WAITOK); error = copyinstr(uap->path, path, MAXPATHLEN, NULL); #ifdef KTRACE if (error == 0 && KTRPOINT(curthread, KTR_NAMEI)) ktrnamei(path); #endif /* Require paths to start with a '/' character. */ if (error == 0 && path[0] != '/') error = EINVAL; if (error) { fdclose(td, fp, fd); fdrop(fp, td); free(path, M_SHMFD); return (error); } fnv = fnv_32_str(path, FNV1_32_INIT); sx_xlock(&shm_dict_lock); shmfd = shm_lookup(path, fnv); if (shmfd == NULL) { /* Object does not yet exist, create it if requested. */ if (uap->flags & O_CREAT) { #ifdef MAC error = mac_posixshm_check_create(td->td_ucred, path); if (error == 0) { #endif shmfd = shm_alloc(td->td_ucred, cmode); shm_insert(path, fnv, shmfd); #ifdef MAC } #endif } else { free(path, M_SHMFD); error = ENOENT; } } else { /* * Object already exists, obtain a new * reference if requested and permitted. */ free(path, M_SHMFD); if ((uap->flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL)) error = EEXIST; else { #ifdef MAC error = mac_posixshm_check_open(td->td_ucred, shmfd, FFLAGS(uap->flags & O_ACCMODE)); if (error == 0) #endif error = shm_access(shmfd, td->td_ucred, FFLAGS(uap->flags & O_ACCMODE)); } /* * Truncate the file back to zero length if * O_TRUNC was specified and the object was * opened with read/write. */ if (error == 0 && (uap->flags & (O_ACCMODE | O_TRUNC)) == (O_RDWR | O_TRUNC)) { #ifdef MAC error = mac_posixshm_check_truncate( td->td_ucred, fp->f_cred, shmfd); if (error == 0) #endif shm_dotruncate(shmfd, 0); } if (error == 0) shm_hold(shmfd); } sx_xunlock(&shm_dict_lock); if (error) { fdclose(td, fp, fd); fdrop(fp, td); return (error); } } finit(fp, FFLAGS(uap->flags & O_ACCMODE), DTYPE_SHM, shmfd, &shm_ops); td->td_retval[0] = fd; fdrop(fp, td); return (0); } int sys_shm_unlink(struct thread *td, struct shm_unlink_args *uap) { char *path; Fnv32_t fnv; int error; path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); error = copyinstr(uap->path, path, MAXPATHLEN, NULL); if (error) { free(path, M_TEMP); return (error); } #ifdef KTRACE if (KTRPOINT(curthread, KTR_NAMEI)) ktrnamei(path); #endif fnv = fnv_32_str(path, FNV1_32_INIT); sx_xlock(&shm_dict_lock); error = shm_remove(path, fnv, td->td_ucred); sx_xunlock(&shm_dict_lock); free(path, M_TEMP); return (error); } /* * mmap() helper to validate mmap() requests against shm object state * and give mmap() the vm_object to use for the mapping. */ int shm_mmap(struct shmfd *shmfd, vm_size_t objsize, vm_ooffset_t foff, vm_object_t *obj) { /* * XXXRW: This validation is probably insufficient, and subject to * sign errors. It should be fixed. */ if (foff >= shmfd->shm_size || foff + objsize > round_page(shmfd->shm_size)) return (EINVAL); mtx_lock(&shm_timestamp_lock); vfs_timestamp(&shmfd->shm_atime); mtx_unlock(&shm_timestamp_lock); vm_object_reference(shmfd->shm_object); *obj = shmfd->shm_object; return (0); } static int shm_chmod(struct file *fp, mode_t mode, struct ucred *active_cred, struct thread *td) { struct shmfd *shmfd; int error; error = 0; shmfd = fp->f_data; mtx_lock(&shm_timestamp_lock); /* * SUSv4 says that x bits of permission need not be affected. * Be consistent with our shm_open there. */ #ifdef MAC error = mac_posixshm_check_setmode(active_cred, shmfd, mode); if (error != 0) goto out; #endif error = vaccess(VREG, shmfd->shm_mode, shmfd->shm_uid, shmfd->shm_gid, VADMIN, active_cred, NULL); if (error != 0) goto out; shmfd->shm_mode = mode & ACCESSPERMS; out: mtx_unlock(&shm_timestamp_lock); return (error); } static int shm_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred, struct thread *td) { struct shmfd *shmfd; int error; error = 0; shmfd = fp->f_data; mtx_lock(&shm_timestamp_lock); #ifdef MAC error = mac_posixshm_check_setowner(active_cred, shmfd, uid, gid); if (error != 0) goto out; #endif if (uid == (uid_t)-1) uid = shmfd->shm_uid; if (gid == (gid_t)-1) gid = shmfd->shm_gid; if (((uid != shmfd->shm_uid && uid != active_cred->cr_uid) || (gid != shmfd->shm_gid && !groupmember(gid, active_cred))) && (error = priv_check_cred(active_cred, PRIV_VFS_CHOWN, 0))) goto out; shmfd->shm_uid = uid; shmfd->shm_gid = gid; out: mtx_unlock(&shm_timestamp_lock); return (error); } /* * Helper routines to allow the backing object of a shared memory file * descriptor to be mapped in the kernel. */ int shm_map(struct file *fp, size_t size, off_t offset, void **memp) { struct shmfd *shmfd; vm_offset_t kva, ofs; vm_object_t obj; int rv; if (fp->f_type != DTYPE_SHM) return (EINVAL); shmfd = fp->f_data; obj = shmfd->shm_object; VM_OBJECT_WLOCK(obj); /* * XXXRW: This validation is probably insufficient, and subject to * sign errors. It should be fixed. */ if (offset >= shmfd->shm_size || offset + size > round_page(shmfd->shm_size)) { VM_OBJECT_WUNLOCK(obj); return (EINVAL); } shmfd->shm_kmappings++; vm_object_reference_locked(obj); VM_OBJECT_WUNLOCK(obj); /* Map the object into the kernel_map and wire it. */ kva = vm_map_min(kernel_map); ofs = offset & PAGE_MASK; offset = trunc_page(offset); size = round_page(size + ofs); rv = vm_map_find(kernel_map, obj, offset, &kva, size, 0, VMFS_OPTIMAL_SPACE, VM_PROT_READ | VM_PROT_WRITE, VM_PROT_READ | VM_PROT_WRITE, 0); if (rv == KERN_SUCCESS) { rv = vm_map_wire(kernel_map, kva, kva + size, VM_MAP_WIRE_SYSTEM | VM_MAP_WIRE_NOHOLES); if (rv == KERN_SUCCESS) { *memp = (void *)(kva + ofs); return (0); } vm_map_remove(kernel_map, kva, kva + size); } else vm_object_deallocate(obj); /* On failure, drop our mapping reference. */ VM_OBJECT_WLOCK(obj); shmfd->shm_kmappings--; VM_OBJECT_WUNLOCK(obj); return (vm_mmap_to_errno(rv)); } /* * We require the caller to unmap the entire entry. This allows us to * safely decrement shm_kmappings when a mapping is removed. */ int shm_unmap(struct file *fp, void *mem, size_t size) { struct shmfd *shmfd; vm_map_entry_t entry; vm_offset_t kva, ofs; vm_object_t obj; vm_pindex_t pindex; vm_prot_t prot; boolean_t wired; vm_map_t map; int rv; if (fp->f_type != DTYPE_SHM) return (EINVAL); shmfd = fp->f_data; kva = (vm_offset_t)mem; ofs = kva & PAGE_MASK; kva = trunc_page(kva); size = round_page(size + ofs); map = kernel_map; rv = vm_map_lookup(&map, kva, VM_PROT_READ | VM_PROT_WRITE, &entry, &obj, &pindex, &prot, &wired); if (rv != KERN_SUCCESS) return (EINVAL); if (entry->start != kva || entry->end != kva + size) { vm_map_lookup_done(map, entry); return (EINVAL); } vm_map_lookup_done(map, entry); if (obj != shmfd->shm_object) return (EINVAL); vm_map_remove(map, kva, kva + size); VM_OBJECT_WLOCK(obj); KASSERT(shmfd->shm_kmappings > 0, ("shm_unmap: object not mapped")); shmfd->shm_kmappings--; VM_OBJECT_WUNLOCK(obj); return (0); } static int shm_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp) { struct shmfd *shmfd; kif->kf_type = KF_TYPE_SHM; shmfd = fp->f_data; mtx_lock(&shm_timestamp_lock); kif->kf_un.kf_file.kf_file_mode = S_IFREG | shmfd->shm_mode; /* XXX */ mtx_unlock(&shm_timestamp_lock); kif->kf_un.kf_file.kf_file_size = shmfd->shm_size; if (shmfd->shm_path != NULL) { sx_slock(&shm_dict_lock); if (shmfd->shm_path != NULL) strlcpy(kif->kf_path, shmfd->shm_path, sizeof(kif->kf_path)); sx_sunlock(&shm_dict_lock); } return (0); } Index: user/ngie/more-tests/sys/kern/vfs_bio.c =================================================================== --- user/ngie/more-tests/sys/kern/vfs_bio.c (revision 281974) +++ user/ngie/more-tests/sys/kern/vfs_bio.c (revision 281975) @@ -1,4635 +1,4637 @@ /*- * Copyright (c) 2004 Poul-Henning Kamp * Copyright (c) 1994,1997 John S. Dyson * Copyright (c) 2013 The FreeBSD Foundation * All rights reserved. * * Portions of this software were developed by Konstantin Belousov * under sponsorship from the FreeBSD Foundation. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * this file contains a new buffer I/O scheme implementing a coherent * VM object and buffer cache scheme. Pains have been taken to make * sure that the performance degradation associated with schemes such * as this is not realized. * * Author: John S. Dyson * Significant help during the development and debugging phases * had been provided by David Greenman, also of the FreeBSD core team. * * see man buf(9) for more info. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "opt_compat.h" #include "opt_swap.h" static MALLOC_DEFINE(M_BIOBUF, "biobuf", "BIO buffer"); struct bio_ops bioops; /* I/O operation notification */ struct buf_ops buf_ops_bio = { .bop_name = "buf_ops_bio", .bop_write = bufwrite, .bop_strategy = bufstrategy, .bop_sync = bufsync, .bop_bdflush = bufbdflush, }; /* * XXX buf is global because kern_shutdown.c and ffs_checkoverlap has * carnal knowledge of buffers. This knowledge should be moved to vfs_bio.c. */ struct buf *buf; /* buffer header pool */ caddr_t unmapped_buf; /* Used below and for softdep flushing threads in ufs/ffs/ffs_softdep.c */ struct proc *bufdaemonproc; static int inmem(struct vnode *vp, daddr_t blkno); static void vm_hold_free_pages(struct buf *bp, int newbsize); static void vm_hold_load_pages(struct buf *bp, vm_offset_t from, vm_offset_t to); static void vfs_page_set_valid(struct buf *bp, vm_ooffset_t off, vm_page_t m); static void vfs_page_set_validclean(struct buf *bp, vm_ooffset_t off, vm_page_t m); static void vfs_clean_pages_dirty_buf(struct buf *bp); static void vfs_setdirty_locked_object(struct buf *bp); static void vfs_vmio_release(struct buf *bp); static int vfs_bio_clcheck(struct vnode *vp, int size, daddr_t lblkno, daddr_t blkno); static int buf_flush(int); static int flushbufqueues(int, int); static void buf_daemon(void); static void bremfreel(struct buf *bp); static __inline void bd_wakeup(void); static int sysctl_runningspace(SYSCTL_HANDLER_ARGS); #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) static int sysctl_bufspace(SYSCTL_HANDLER_ARGS); #endif int vmiodirenable = TRUE; SYSCTL_INT(_vfs, OID_AUTO, vmiodirenable, CTLFLAG_RW, &vmiodirenable, 0, "Use the VM system for directory writes"); long runningbufspace; SYSCTL_LONG(_vfs, OID_AUTO, runningbufspace, CTLFLAG_RD, &runningbufspace, 0, "Amount of presently outstanding async buffer io"); static long bufspace; #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) SYSCTL_PROC(_vfs, OID_AUTO, bufspace, CTLTYPE_LONG|CTLFLAG_MPSAFE|CTLFLAG_RD, &bufspace, 0, sysctl_bufspace, "L", "Virtual memory used for buffers"); #else SYSCTL_LONG(_vfs, OID_AUTO, bufspace, CTLFLAG_RD, &bufspace, 0, "Virtual memory used for buffers"); #endif static long unmapped_bufspace; SYSCTL_LONG(_vfs, OID_AUTO, unmapped_bufspace, CTLFLAG_RD, &unmapped_bufspace, 0, "Amount of unmapped buffers, inclusive in the bufspace"); static long maxbufspace; SYSCTL_LONG(_vfs, OID_AUTO, maxbufspace, CTLFLAG_RD, &maxbufspace, 0, "Maximum allowed value of bufspace (including buf_daemon)"); static long bufmallocspace; SYSCTL_LONG(_vfs, OID_AUTO, bufmallocspace, CTLFLAG_RD, &bufmallocspace, 0, "Amount of malloced memory for buffers"); static long maxbufmallocspace; SYSCTL_LONG(_vfs, OID_AUTO, maxmallocbufspace, CTLFLAG_RW, &maxbufmallocspace, 0, "Maximum amount of malloced memory for buffers"); static long lobufspace; SYSCTL_LONG(_vfs, OID_AUTO, lobufspace, CTLFLAG_RD, &lobufspace, 0, "Minimum amount of buffers we want to have"); long hibufspace; SYSCTL_LONG(_vfs, OID_AUTO, hibufspace, CTLFLAG_RD, &hibufspace, 0, "Maximum allowed value of bufspace (excluding buf_daemon)"); static int bufreusecnt; SYSCTL_INT(_vfs, OID_AUTO, bufreusecnt, CTLFLAG_RW, &bufreusecnt, 0, "Number of times we have reused a buffer"); static int buffreekvacnt; SYSCTL_INT(_vfs, OID_AUTO, buffreekvacnt, CTLFLAG_RW, &buffreekvacnt, 0, "Number of times we have freed the KVA space from some buffer"); static int bufdefragcnt; SYSCTL_INT(_vfs, OID_AUTO, bufdefragcnt, CTLFLAG_RW, &bufdefragcnt, 0, "Number of times we have had to repeat buffer allocation to defragment"); static long lorunningspace; SYSCTL_PROC(_vfs, OID_AUTO, lorunningspace, CTLTYPE_LONG | CTLFLAG_MPSAFE | CTLFLAG_RW, &lorunningspace, 0, sysctl_runningspace, "L", "Minimum preferred space used for in-progress I/O"); static long hirunningspace; SYSCTL_PROC(_vfs, OID_AUTO, hirunningspace, CTLTYPE_LONG | CTLFLAG_MPSAFE | CTLFLAG_RW, &hirunningspace, 0, sysctl_runningspace, "L", "Maximum amount of space to use for in-progress I/O"); int dirtybufferflushes; SYSCTL_INT(_vfs, OID_AUTO, dirtybufferflushes, CTLFLAG_RW, &dirtybufferflushes, 0, "Number of bdwrite to bawrite conversions to limit dirty buffers"); int bdwriteskip; SYSCTL_INT(_vfs, OID_AUTO, bdwriteskip, CTLFLAG_RW, &bdwriteskip, 0, "Number of buffers supplied to bdwrite with snapshot deadlock risk"); int altbufferflushes; SYSCTL_INT(_vfs, OID_AUTO, altbufferflushes, CTLFLAG_RW, &altbufferflushes, 0, "Number of fsync flushes to limit dirty buffers"); static int recursiveflushes; SYSCTL_INT(_vfs, OID_AUTO, recursiveflushes, CTLFLAG_RW, &recursiveflushes, 0, "Number of flushes skipped due to being recursive"); static int numdirtybuffers; SYSCTL_INT(_vfs, OID_AUTO, numdirtybuffers, CTLFLAG_RD, &numdirtybuffers, 0, "Number of buffers that are dirty (has unwritten changes) at the moment"); static int lodirtybuffers; SYSCTL_INT(_vfs, OID_AUTO, lodirtybuffers, CTLFLAG_RW, &lodirtybuffers, 0, "How many buffers we want to have free before bufdaemon can sleep"); static int hidirtybuffers; SYSCTL_INT(_vfs, OID_AUTO, hidirtybuffers, CTLFLAG_RW, &hidirtybuffers, 0, "When the number of dirty buffers is considered severe"); int dirtybufthresh; SYSCTL_INT(_vfs, OID_AUTO, dirtybufthresh, CTLFLAG_RW, &dirtybufthresh, 0, "Number of bdwrite to bawrite conversions to clear dirty buffers"); static int numfreebuffers; SYSCTL_INT(_vfs, OID_AUTO, numfreebuffers, CTLFLAG_RD, &numfreebuffers, 0, "Number of free buffers"); static int lofreebuffers; SYSCTL_INT(_vfs, OID_AUTO, lofreebuffers, CTLFLAG_RW, &lofreebuffers, 0, "XXX Unused"); static int hifreebuffers; SYSCTL_INT(_vfs, OID_AUTO, hifreebuffers, CTLFLAG_RW, &hifreebuffers, 0, "XXX Complicatedly unused"); static int getnewbufcalls; SYSCTL_INT(_vfs, OID_AUTO, getnewbufcalls, CTLFLAG_RW, &getnewbufcalls, 0, "Number of calls to getnewbuf"); static int getnewbufrestarts; SYSCTL_INT(_vfs, OID_AUTO, getnewbufrestarts, CTLFLAG_RW, &getnewbufrestarts, 0, "Number of times getnewbuf has had to restart a buffer aquisition"); static int mappingrestarts; SYSCTL_INT(_vfs, OID_AUTO, mappingrestarts, CTLFLAG_RW, &mappingrestarts, 0, "Number of times getblk has had to restart a buffer mapping for " "unmapped buffer"); static int flushbufqtarget = 100; SYSCTL_INT(_vfs, OID_AUTO, flushbufqtarget, CTLFLAG_RW, &flushbufqtarget, 0, "Amount of work to do in flushbufqueues when helping bufdaemon"); static long notbufdflushes; SYSCTL_LONG(_vfs, OID_AUTO, notbufdflushes, CTLFLAG_RD, ¬bufdflushes, 0, "Number of dirty buffer flushes done by the bufdaemon helpers"); static long barrierwrites; SYSCTL_LONG(_vfs, OID_AUTO, barrierwrites, CTLFLAG_RW, &barrierwrites, 0, "Number of barrier writes"); SYSCTL_INT(_vfs, OID_AUTO, unmapped_buf_allowed, CTLFLAG_RD, &unmapped_buf_allowed, 0, "Permit the use of the unmapped i/o"); /* * Lock for the non-dirty bufqueues */ static struct mtx_padalign bqclean; /* * Lock for the dirty queue. */ static struct mtx_padalign bqdirty; /* * This lock synchronizes access to bd_request. */ static struct mtx_padalign bdlock; /* * This lock protects the runningbufreq and synchronizes runningbufwakeup and * waitrunningbufspace(). */ static struct mtx_padalign rbreqlock; /* * Lock that protects needsbuffer and the sleeps/wakeups surrounding it. */ static struct rwlock_padalign nblock; /* * Lock that protects bdirtywait. */ static struct mtx_padalign bdirtylock; /* * Wakeup point for bufdaemon, as well as indicator of whether it is already * active. Set to 1 when the bufdaemon is already "on" the queue, 0 when it * is idling. */ static int bd_request; /* * Request for the buf daemon to write more buffers than is indicated by * lodirtybuf. This may be necessary to push out excess dependencies or * defragment the address space where a simple count of the number of dirty * buffers is insufficient to characterize the demand for flushing them. */ static int bd_speedupreq; /* * bogus page -- for I/O to/from partially complete buffers * this is a temporary solution to the problem, but it is not * really that bad. it would be better to split the buffer * for input in the case of buffers partially already in memory, * but the code is intricate enough already. */ vm_page_t bogus_page; /* * Synchronization (sleep/wakeup) variable for active buffer space requests. * Set when wait starts, cleared prior to wakeup(). * Used in runningbufwakeup() and waitrunningbufspace(). */ static int runningbufreq; /* * Synchronization (sleep/wakeup) variable for buffer requests. * Can contain the VFS_BIO_NEED flags defined below; setting/clearing is done * by and/or. * Used in numdirtywakeup(), bufspacewakeup(), bufcountadd(), bwillwrite(), * getnewbuf(), and getblk(). */ static volatile int needsbuffer; /* * Synchronization for bwillwrite() waiters. */ static int bdirtywait; /* * Definitions for the buffer free lists. */ #define BUFFER_QUEUES 5 /* number of free buffer queues */ #define QUEUE_NONE 0 /* on no queue */ #define QUEUE_CLEAN 1 /* non-B_DELWRI buffers */ #define QUEUE_DIRTY 2 /* B_DELWRI buffers */ #define QUEUE_EMPTYKVA 3 /* empty buffer headers w/KVA assignment */ #define QUEUE_EMPTY 4 /* empty buffer headers */ #define QUEUE_SENTINEL 1024 /* not an queue index, but mark for sentinel */ /* Queues for free buffers with various properties */ static TAILQ_HEAD(bqueues, buf) bufqueues[BUFFER_QUEUES] = { { 0 } }; #ifdef INVARIANTS static int bq_len[BUFFER_QUEUES]; #endif /* * Single global constant for BUF_WMESG, to avoid getting multiple references. * buf_wmesg is referred from macros. */ const char *buf_wmesg = BUF_WMESG; #define VFS_BIO_NEED_ANY 0x01 /* any freeable buffer */ #define VFS_BIO_NEED_FREE 0x04 /* wait for free bufs, hi hysteresis */ #define VFS_BIO_NEED_BUFSPACE 0x08 /* wait for buf space, lo hysteresis */ static int sysctl_runningspace(SYSCTL_HANDLER_ARGS) { long value; int error; value = *(long *)arg1; error = sysctl_handle_long(oidp, &value, 0, req); if (error != 0 || req->newptr == NULL) return (error); mtx_lock(&rbreqlock); if (arg1 == &hirunningspace) { if (value < lorunningspace) error = EINVAL; else hirunningspace = value; } else { KASSERT(arg1 == &lorunningspace, ("%s: unknown arg1", __func__)); if (value > hirunningspace) error = EINVAL; else lorunningspace = value; } mtx_unlock(&rbreqlock); return (error); } #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) static int sysctl_bufspace(SYSCTL_HANDLER_ARGS) { long lvalue; int ivalue; if (sizeof(int) == sizeof(long) || req->oldlen >= sizeof(long)) return (sysctl_handle_long(oidp, arg1, arg2, req)); lvalue = *(long *)arg1; if (lvalue > INT_MAX) /* On overflow, still write out a long to trigger ENOMEM. */ return (sysctl_handle_long(oidp, &lvalue, 0, req)); ivalue = lvalue; return (sysctl_handle_int(oidp, &ivalue, 0, req)); } #endif /* * bqlock: * * Return the appropriate queue lock based on the index. */ static inline struct mtx * bqlock(int qindex) { if (qindex == QUEUE_DIRTY) return (struct mtx *)(&bqdirty); return (struct mtx *)(&bqclean); } /* * bdirtywakeup: * * Wakeup any bwillwrite() waiters. */ static void bdirtywakeup(void) { mtx_lock(&bdirtylock); if (bdirtywait) { bdirtywait = 0; wakeup(&bdirtywait); } mtx_unlock(&bdirtylock); } /* * bdirtysub: * * Decrement the numdirtybuffers count by one and wakeup any * threads blocked in bwillwrite(). */ static void bdirtysub(void) { if (atomic_fetchadd_int(&numdirtybuffers, -1) == (lodirtybuffers + hidirtybuffers) / 2) bdirtywakeup(); } /* * bdirtyadd: * * Increment the numdirtybuffers count by one and wakeup the buf * daemon if needed. */ static void bdirtyadd(void) { /* * Only do the wakeup once as we cross the boundary. The * buf daemon will keep running until the condition clears. */ if (atomic_fetchadd_int(&numdirtybuffers, 1) == (lodirtybuffers + hidirtybuffers) / 2) bd_wakeup(); } /* * bufspacewakeup: * * Called when buffer space is potentially available for recovery. * getnewbuf() will block on this flag when it is unable to free * sufficient buffer space. Buffer space becomes recoverable when * bp's get placed back in the queues. */ static __inline void bufspacewakeup(void) { int need_wakeup, on; /* * If someone is waiting for BUF space, wake them up. Even * though we haven't freed the kva space yet, the waiting * process will be able to now. */ rw_rlock(&nblock); for (;;) { need_wakeup = 0; on = needsbuffer; if ((on & VFS_BIO_NEED_BUFSPACE) == 0) break; need_wakeup = 1; if (atomic_cmpset_rel_int(&needsbuffer, on, on & ~VFS_BIO_NEED_BUFSPACE)) break; } if (need_wakeup) wakeup(__DEVOLATILE(void *, &needsbuffer)); rw_runlock(&nblock); } /* * runningwakeup: * * Wake up processes that are waiting on asynchronous writes to fall * below lorunningspace. */ static void runningwakeup(void) { mtx_lock(&rbreqlock); if (runningbufreq) { runningbufreq = 0; wakeup(&runningbufreq); } mtx_unlock(&rbreqlock); } /* * runningbufwakeup: * * Decrement the outstanding write count according. */ void runningbufwakeup(struct buf *bp) { long space, bspace; bspace = bp->b_runningbufspace; if (bspace == 0) return; space = atomic_fetchadd_long(&runningbufspace, -bspace); KASSERT(space >= bspace, ("runningbufspace underflow %ld %ld", space, bspace)); bp->b_runningbufspace = 0; /* * Only acquire the lock and wakeup on the transition from exceeding * the threshold to falling below it. */ if (space < lorunningspace) return; if (space - bspace > lorunningspace) return; runningwakeup(); } /* * bufcountadd: * * Called when a buffer has been added to one of the free queues to * account for the buffer and to wakeup anyone waiting for free buffers. * This typically occurs when large amounts of metadata are being handled * by the buffer cache ( else buffer space runs out first, usually ). */ static __inline void bufcountadd(struct buf *bp) { int mask, need_wakeup, old, on; KASSERT((bp->b_flags & B_INFREECNT) == 0, ("buf %p already counted as free", bp)); bp->b_flags |= B_INFREECNT; old = atomic_fetchadd_int(&numfreebuffers, 1); KASSERT(old >= 0 && old < nbuf, ("numfreebuffers climbed to %d", old + 1)); mask = VFS_BIO_NEED_ANY; if (numfreebuffers >= hifreebuffers) mask |= VFS_BIO_NEED_FREE; rw_rlock(&nblock); for (;;) { need_wakeup = 0; on = needsbuffer; if (on == 0) break; need_wakeup = 1; if (atomic_cmpset_rel_int(&needsbuffer, on, on & ~mask)) break; } if (need_wakeup) wakeup(__DEVOLATILE(void *, &needsbuffer)); rw_runlock(&nblock); } /* * bufcountsub: * * Decrement the numfreebuffers count as needed. */ static void bufcountsub(struct buf *bp) { int old; /* * Fixup numfreebuffers count. If the buffer is invalid or not * delayed-write, the buffer was free and we must decrement * numfreebuffers. */ if ((bp->b_flags & B_INVAL) || (bp->b_flags & B_DELWRI) == 0) { KASSERT((bp->b_flags & B_INFREECNT) != 0, ("buf %p not counted in numfreebuffers", bp)); bp->b_flags &= ~B_INFREECNT; old = atomic_fetchadd_int(&numfreebuffers, -1); KASSERT(old > 0, ("numfreebuffers dropped to %d", old - 1)); } } /* * waitrunningbufspace() * * runningbufspace is a measure of the amount of I/O currently * running. This routine is used in async-write situations to * prevent creating huge backups of pending writes to a device. * Only asynchronous writes are governed by this function. * * This does NOT turn an async write into a sync write. It waits * for earlier writes to complete and generally returns before the * caller's write has reached the device. */ void waitrunningbufspace(void) { mtx_lock(&rbreqlock); while (runningbufspace > hirunningspace) { runningbufreq = 1; msleep(&runningbufreq, &rbreqlock, PVM, "wdrain", 0); } mtx_unlock(&rbreqlock); } /* * vfs_buf_test_cache: * * Called when a buffer is extended. This function clears the B_CACHE * bit if the newly extended portion of the buffer does not contain * valid data. */ static __inline void vfs_buf_test_cache(struct buf *bp, vm_ooffset_t foff, vm_offset_t off, vm_offset_t size, vm_page_t m) { VM_OBJECT_ASSERT_LOCKED(m->object); if (bp->b_flags & B_CACHE) { int base = (foff + off) & PAGE_MASK; if (vm_page_is_valid(m, base, size) == 0) bp->b_flags &= ~B_CACHE; } } /* Wake up the buffer daemon if necessary */ static __inline void bd_wakeup(void) { mtx_lock(&bdlock); if (bd_request == 0) { bd_request = 1; wakeup(&bd_request); } mtx_unlock(&bdlock); } /* * bd_speedup - speedup the buffer cache flushing code */ void bd_speedup(void) { int needwake; mtx_lock(&bdlock); needwake = 0; if (bd_speedupreq == 0 || bd_request == 0) needwake = 1; bd_speedupreq = 1; bd_request = 1; if (needwake) wakeup(&bd_request); mtx_unlock(&bdlock); } #ifndef NSWBUF_MIN #define NSWBUF_MIN 16 #endif #ifdef __i386__ #define TRANSIENT_DENOM 5 #else #define TRANSIENT_DENOM 10 #endif /* * Calculating buffer cache scaling values and reserve space for buffer * headers. This is called during low level kernel initialization and * may be called more then once. We CANNOT write to the memory area * being reserved at this time. */ caddr_t kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est) { int tuned_nbuf; long maxbuf, maxbuf_sz, buf_sz, biotmap_sz; /* * physmem_est is in pages. Convert it to kilobytes (assumes * PAGE_SIZE is >= 1K) */ physmem_est = physmem_est * (PAGE_SIZE / 1024); /* * The nominal buffer size (and minimum KVA allocation) is BKVASIZE. * For the first 64MB of ram nominally allocate sufficient buffers to * cover 1/4 of our ram. Beyond the first 64MB allocate additional * buffers to cover 1/10 of our ram over 64MB. When auto-sizing * the buffer cache we limit the eventual kva reservation to * maxbcache bytes. * * factor represents the 1/4 x ram conversion. */ if (nbuf == 0) { int factor = 4 * BKVASIZE / 1024; nbuf = 50; if (physmem_est > 4096) nbuf += min((physmem_est - 4096) / factor, 65536 / factor); if (physmem_est > 65536) nbuf += min((physmem_est - 65536) * 2 / (factor * 5), 32 * 1024 * 1024 / (factor * 5)); if (maxbcache && nbuf > maxbcache / BKVASIZE) nbuf = maxbcache / BKVASIZE; tuned_nbuf = 1; } else tuned_nbuf = 0; /* XXX Avoid unsigned long overflows later on with maxbufspace. */ maxbuf = (LONG_MAX / 3) / BKVASIZE; if (nbuf > maxbuf) { if (!tuned_nbuf) printf("Warning: nbufs lowered from %d to %ld\n", nbuf, maxbuf); nbuf = maxbuf; } /* * Ideal allocation size for the transient bio submap is 10% * of the maximal space buffer map. This roughly corresponds * to the amount of the buffer mapped for typical UFS load. * * Clip the buffer map to reserve space for the transient * BIOs, if its extent is bigger than 90% (80% on i386) of the * maximum buffer map extent on the platform. * * The fall-back to the maxbuf in case of maxbcache unset, * allows to not trim the buffer KVA for the architectures * with ample KVA space. */ if (bio_transient_maxcnt == 0 && unmapped_buf_allowed) { maxbuf_sz = maxbcache != 0 ? maxbcache : maxbuf * BKVASIZE; buf_sz = (long)nbuf * BKVASIZE; if (buf_sz < maxbuf_sz / TRANSIENT_DENOM * (TRANSIENT_DENOM - 1)) { /* * There is more KVA than memory. Do not * adjust buffer map size, and assign the rest * of maxbuf to transient map. */ biotmap_sz = maxbuf_sz - buf_sz; } else { /* * Buffer map spans all KVA we could afford on * this platform. Give 10% (20% on i386) of * the buffer map to the transient bio map. */ biotmap_sz = buf_sz / TRANSIENT_DENOM; buf_sz -= biotmap_sz; } if (biotmap_sz / INT_MAX > MAXPHYS) bio_transient_maxcnt = INT_MAX; else bio_transient_maxcnt = biotmap_sz / MAXPHYS; /* * Artifically limit to 1024 simultaneous in-flight I/Os * using the transient mapping. */ if (bio_transient_maxcnt > 1024) bio_transient_maxcnt = 1024; if (tuned_nbuf) nbuf = buf_sz / BKVASIZE; } /* * swbufs are used as temporary holders for I/O, such as paging I/O. * We have no less then 16 and no more then 256. */ nswbuf = min(nbuf / 4, 256); TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf); if (nswbuf < NSWBUF_MIN) nswbuf = NSWBUF_MIN; /* * Reserve space for the buffer cache buffers */ swbuf = (void *)v; v = (caddr_t)(swbuf + nswbuf); buf = (void *)v; v = (caddr_t)(buf + nbuf); return(v); } /* Initialize the buffer subsystem. Called before use of any buffers. */ void bufinit(void) { struct buf *bp; int i; + CTASSERT(MAXBCACHEBUF >= MAXBSIZE); mtx_init(&bqclean, "bufq clean lock", NULL, MTX_DEF); mtx_init(&bqdirty, "bufq dirty lock", NULL, MTX_DEF); mtx_init(&rbreqlock, "runningbufspace lock", NULL, MTX_DEF); rw_init(&nblock, "needsbuffer lock"); mtx_init(&bdlock, "buffer daemon lock", NULL, MTX_DEF); mtx_init(&bdirtylock, "dirty buf lock", NULL, MTX_DEF); /* next, make a null set of free lists */ for (i = 0; i < BUFFER_QUEUES; i++) TAILQ_INIT(&bufqueues[i]); /* finally, initialize each buffer header and stick on empty q */ for (i = 0; i < nbuf; i++) { bp = &buf[i]; bzero(bp, sizeof *bp); bp->b_flags = B_INVAL | B_INFREECNT; bp->b_rcred = NOCRED; bp->b_wcred = NOCRED; bp->b_qindex = QUEUE_EMPTY; bp->b_xflags = 0; LIST_INIT(&bp->b_dep); BUF_LOCKINIT(bp); TAILQ_INSERT_TAIL(&bufqueues[QUEUE_EMPTY], bp, b_freelist); #ifdef INVARIANTS bq_len[QUEUE_EMPTY]++; #endif } /* * maxbufspace is the absolute maximum amount of buffer space we are * allowed to reserve in KVM and in real terms. The absolute maximum * is nominally used by buf_daemon. hibufspace is the nominal maximum * used by most other processes. The differential is required to * ensure that buf_daemon is able to run when other processes might * be blocked waiting for buffer space. * * maxbufspace is based on BKVASIZE. Allocating buffers larger then * this may result in KVM fragmentation which is not handled optimally * by the system. */ maxbufspace = (long)nbuf * BKVASIZE; - hibufspace = lmax(3 * maxbufspace / 4, maxbufspace - MAXBSIZE * 10); - lobufspace = hibufspace - MAXBSIZE; + hibufspace = lmax(3 * maxbufspace / 4, maxbufspace - MAXBCACHEBUF * 10); + lobufspace = hibufspace - MAXBCACHEBUF; /* * Note: The 16 MiB upper limit for hirunningspace was chosen * arbitrarily and may need further tuning. It corresponds to * 128 outstanding write IO requests (if IO size is 128 KiB), * which fits with many RAID controllers' tagged queuing limits. * The lower 1 MiB limit is the historical upper limit for * hirunningspace. */ - hirunningspace = lmax(lmin(roundup(hibufspace / 64, MAXBSIZE), + hirunningspace = lmax(lmin(roundup(hibufspace / 64, MAXBCACHEBUF), 16 * 1024 * 1024), 1024 * 1024); - lorunningspace = roundup((hirunningspace * 2) / 3, MAXBSIZE); + lorunningspace = roundup((hirunningspace * 2) / 3, MAXBCACHEBUF); /* * Limit the amount of malloc memory since it is wired permanently into * the kernel space. Even though this is accounted for in the buffer * allocation, we don't want the malloced region to grow uncontrolled. * The malloc scheme improves memory utilization significantly on average * (small) directories. */ maxbufmallocspace = hibufspace / 20; /* * Reduce the chance of a deadlock occuring by limiting the number * of delayed-write dirty buffers we allow to stack up. */ hidirtybuffers = nbuf / 4 + 20; dirtybufthresh = hidirtybuffers * 9 / 10; numdirtybuffers = 0; /* * To support extreme low-memory systems, make sure hidirtybuffers cannot * eat up all available buffer space. This occurs when our minimum cannot * be met. We try to size hidirtybuffers to 3/4 our buffer space assuming * BKVASIZE'd buffers. */ while ((long)hidirtybuffers * BKVASIZE > 3 * hibufspace / 4) { hidirtybuffers >>= 1; } lodirtybuffers = hidirtybuffers / 2; /* * Try to keep the number of free buffers in the specified range, * and give special processes (e.g. like buf_daemon) access to an * emergency reserve. */ lofreebuffers = nbuf / 18 + 5; hifreebuffers = 2 * lofreebuffers; numfreebuffers = nbuf; bogus_page = vm_page_alloc(NULL, 0, VM_ALLOC_NOOBJ | VM_ALLOC_NORMAL | VM_ALLOC_WIRED); unmapped_buf = (caddr_t)kva_alloc(MAXPHYS); } #ifdef INVARIANTS static inline void vfs_buf_check_mapped(struct buf *bp) { KASSERT((bp->b_flags & B_UNMAPPED) == 0, ("mapped buf %p %x", bp, bp->b_flags)); KASSERT(bp->b_kvabase != unmapped_buf, ("mapped buf: b_kvabase was not updated %p", bp)); KASSERT(bp->b_data != unmapped_buf, ("mapped buf: b_data was not updated %p", bp)); } static inline void vfs_buf_check_unmapped(struct buf *bp) { KASSERT((bp->b_flags & B_UNMAPPED) == B_UNMAPPED, ("unmapped buf %p %x", bp, bp->b_flags)); KASSERT(bp->b_kvabase == unmapped_buf, ("unmapped buf: corrupted b_kvabase %p", bp)); KASSERT(bp->b_data == unmapped_buf, ("unmapped buf: corrupted b_data %p", bp)); } #define BUF_CHECK_MAPPED(bp) vfs_buf_check_mapped(bp) #define BUF_CHECK_UNMAPPED(bp) vfs_buf_check_unmapped(bp) #else #define BUF_CHECK_MAPPED(bp) do {} while (0) #define BUF_CHECK_UNMAPPED(bp) do {} while (0) #endif static void bpmap_qenter(struct buf *bp) { BUF_CHECK_MAPPED(bp); /* * bp->b_data is relative to bp->b_offset, but * bp->b_offset may be offset into the first page. */ bp->b_data = (caddr_t)trunc_page((vm_offset_t)bp->b_data); pmap_qenter((vm_offset_t)bp->b_data, bp->b_pages, bp->b_npages); bp->b_data = (caddr_t)((vm_offset_t)bp->b_data | (vm_offset_t)(bp->b_offset & PAGE_MASK)); } /* * bfreekva() - free the kva allocation for a buffer. * * Since this call frees up buffer space, we call bufspacewakeup(). */ static void bfreekva(struct buf *bp) { if (bp->b_kvasize == 0) return; atomic_add_int(&buffreekvacnt, 1); atomic_subtract_long(&bufspace, bp->b_kvasize); if ((bp->b_flags & B_UNMAPPED) == 0) { BUF_CHECK_MAPPED(bp); vmem_free(buffer_arena, (vm_offset_t)bp->b_kvabase, bp->b_kvasize); } else { BUF_CHECK_UNMAPPED(bp); if ((bp->b_flags & B_KVAALLOC) != 0) { vmem_free(buffer_arena, (vm_offset_t)bp->b_kvaalloc, bp->b_kvasize); } atomic_subtract_long(&unmapped_bufspace, bp->b_kvasize); bp->b_flags &= ~(B_UNMAPPED | B_KVAALLOC); } bp->b_kvasize = 0; bufspacewakeup(); } /* * binsfree: * * Insert the buffer into the appropriate free list. */ static void binsfree(struct buf *bp, int qindex) { struct mtx *olock, *nlock; BUF_ASSERT_XLOCKED(bp); olock = bqlock(bp->b_qindex); nlock = bqlock(qindex); mtx_lock(olock); /* Handle delayed bremfree() processing. */ if (bp->b_flags & B_REMFREE) bremfreel(bp); if (bp->b_qindex != QUEUE_NONE) panic("binsfree: free buffer onto another queue???"); bp->b_qindex = qindex; if (olock != nlock) { mtx_unlock(olock); mtx_lock(nlock); } if (bp->b_flags & B_AGE) TAILQ_INSERT_HEAD(&bufqueues[bp->b_qindex], bp, b_freelist); else TAILQ_INSERT_TAIL(&bufqueues[bp->b_qindex], bp, b_freelist); #ifdef INVARIANTS bq_len[bp->b_qindex]++; #endif mtx_unlock(nlock); /* * Something we can maybe free or reuse. */ if (bp->b_bufsize && !(bp->b_flags & B_DELWRI)) bufspacewakeup(); if ((bp->b_flags & B_INVAL) || !(bp->b_flags & B_DELWRI)) bufcountadd(bp); } /* * bremfree: * * Mark the buffer for removal from the appropriate free list. * */ void bremfree(struct buf *bp) { CTR3(KTR_BUF, "bremfree(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); KASSERT((bp->b_flags & B_REMFREE) == 0, ("bremfree: buffer %p already marked for delayed removal.", bp)); KASSERT(bp->b_qindex != QUEUE_NONE, ("bremfree: buffer %p not on a queue.", bp)); BUF_ASSERT_XLOCKED(bp); bp->b_flags |= B_REMFREE; bufcountsub(bp); } /* * bremfreef: * * Force an immediate removal from a free list. Used only in nfs when * it abuses the b_freelist pointer. */ void bremfreef(struct buf *bp) { struct mtx *qlock; qlock = bqlock(bp->b_qindex); mtx_lock(qlock); bremfreel(bp); mtx_unlock(qlock); } /* * bremfreel: * * Removes a buffer from the free list, must be called with the * correct qlock held. */ static void bremfreel(struct buf *bp) { CTR3(KTR_BUF, "bremfreel(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); KASSERT(bp->b_qindex != QUEUE_NONE, ("bremfreel: buffer %p not on a queue.", bp)); BUF_ASSERT_XLOCKED(bp); mtx_assert(bqlock(bp->b_qindex), MA_OWNED); TAILQ_REMOVE(&bufqueues[bp->b_qindex], bp, b_freelist); #ifdef INVARIANTS KASSERT(bq_len[bp->b_qindex] >= 1, ("queue %d underflow", bp->b_qindex)); bq_len[bp->b_qindex]--; #endif bp->b_qindex = QUEUE_NONE; /* * If this was a delayed bremfree() we only need to remove the buffer * from the queue and return the stats are already done. */ if (bp->b_flags & B_REMFREE) { bp->b_flags &= ~B_REMFREE; return; } bufcountsub(bp); } /* * Attempt to initiate asynchronous I/O on read-ahead blocks. We must * clear BIO_ERROR and B_INVAL prior to initiating I/O . If B_CACHE is set, * the buffer is valid and we do not have to do anything. */ void breada(struct vnode * vp, daddr_t * rablkno, int * rabsize, int cnt, struct ucred * cred) { struct buf *rabp; int i; for (i = 0; i < cnt; i++, rablkno++, rabsize++) { if (inmem(vp, *rablkno)) continue; rabp = getblk(vp, *rablkno, *rabsize, 0, 0, 0); if ((rabp->b_flags & B_CACHE) == 0) { if (!TD_IS_IDLETHREAD(curthread)) curthread->td_ru.ru_inblock++; rabp->b_flags |= B_ASYNC; rabp->b_flags &= ~B_INVAL; rabp->b_ioflags &= ~BIO_ERROR; rabp->b_iocmd = BIO_READ; if (rabp->b_rcred == NOCRED && cred != NOCRED) rabp->b_rcred = crhold(cred); vfs_busy_pages(rabp, 0); BUF_KERNPROC(rabp); rabp->b_iooffset = dbtob(rabp->b_blkno); bstrategy(rabp); } else { brelse(rabp); } } } /* * Entry point for bread() and breadn() via #defines in sys/buf.h. * * Get a buffer with the specified data. Look in the cache first. We * must clear BIO_ERROR and B_INVAL prior to initiating I/O. If B_CACHE * is set, the buffer is valid and we do not have to do anything, see * getblk(). Also starts asynchronous I/O on read-ahead blocks. */ int breadn_flags(struct vnode *vp, daddr_t blkno, int size, daddr_t *rablkno, int *rabsize, int cnt, struct ucred *cred, int flags, struct buf **bpp) { struct buf *bp; int rv = 0, readwait = 0; CTR3(KTR_BUF, "breadn(%p, %jd, %d)", vp, blkno, size); /* * Can only return NULL if GB_LOCK_NOWAIT flag is specified. */ *bpp = bp = getblk(vp, blkno, size, 0, 0, flags); if (bp == NULL) return (EBUSY); /* if not found in cache, do some I/O */ if ((bp->b_flags & B_CACHE) == 0) { if (!TD_IS_IDLETHREAD(curthread)) curthread->td_ru.ru_inblock++; bp->b_iocmd = BIO_READ; bp->b_flags &= ~B_INVAL; bp->b_ioflags &= ~BIO_ERROR; if (bp->b_rcred == NOCRED && cred != NOCRED) bp->b_rcred = crhold(cred); vfs_busy_pages(bp, 0); bp->b_iooffset = dbtob(bp->b_blkno); bstrategy(bp); ++readwait; } breada(vp, rablkno, rabsize, cnt, cred); if (readwait) { rv = bufwait(bp); } return (rv); } /* * Write, release buffer on completion. (Done by iodone * if async). Do not bother writing anything if the buffer * is invalid. * * Note that we set B_CACHE here, indicating that buffer is * fully valid and thus cacheable. This is true even of NFS * now so we set it generally. This could be set either here * or in biodone() since the I/O is synchronous. We put it * here. */ int bufwrite(struct buf *bp) { int oldflags; struct vnode *vp; long space; int vp_md; CTR3(KTR_BUF, "bufwrite(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); if (bp->b_flags & B_INVAL) { brelse(bp); return (0); } if (bp->b_flags & B_BARRIER) barrierwrites++; oldflags = bp->b_flags; BUF_ASSERT_HELD(bp); if (bp->b_pin_count > 0) bunpin_wait(bp); KASSERT(!(bp->b_vflags & BV_BKGRDINPROG), ("FFS background buffer should not get here %p", bp)); vp = bp->b_vp; if (vp) vp_md = vp->v_vflag & VV_MD; else vp_md = 0; /* * Mark the buffer clean. Increment the bufobj write count * before bundirty() call, to prevent other thread from seeing * empty dirty list and zero counter for writes in progress, * falsely indicating that the bufobj is clean. */ bufobj_wref(bp->b_bufobj); bundirty(bp); bp->b_flags &= ~B_DONE; bp->b_ioflags &= ~BIO_ERROR; bp->b_flags |= B_CACHE; bp->b_iocmd = BIO_WRITE; vfs_busy_pages(bp, 1); /* * Normal bwrites pipeline writes */ bp->b_runningbufspace = bp->b_bufsize; space = atomic_fetchadd_long(&runningbufspace, bp->b_runningbufspace); if (!TD_IS_IDLETHREAD(curthread)) curthread->td_ru.ru_oublock++; if (oldflags & B_ASYNC) BUF_KERNPROC(bp); bp->b_iooffset = dbtob(bp->b_blkno); bstrategy(bp); if ((oldflags & B_ASYNC) == 0) { int rtval = bufwait(bp); brelse(bp); return (rtval); } else if (space > hirunningspace) { /* * don't allow the async write to saturate the I/O * system. We will not deadlock here because * we are blocking waiting for I/O that is already in-progress * to complete. We do not block here if it is the update * or syncer daemon trying to clean up as that can lead * to deadlock. */ if ((curthread->td_pflags & TDP_NORUNNINGBUF) == 0 && !vp_md) waitrunningbufspace(); } return (0); } void bufbdflush(struct bufobj *bo, struct buf *bp) { struct buf *nbp; if (bo->bo_dirty.bv_cnt > dirtybufthresh + 10) { (void) VOP_FSYNC(bp->b_vp, MNT_NOWAIT, curthread); altbufferflushes++; } else if (bo->bo_dirty.bv_cnt > dirtybufthresh) { BO_LOCK(bo); /* * Try to find a buffer to flush. */ TAILQ_FOREACH(nbp, &bo->bo_dirty.bv_hd, b_bobufs) { if ((nbp->b_vflags & BV_BKGRDINPROG) || BUF_LOCK(nbp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) continue; if (bp == nbp) panic("bdwrite: found ourselves"); BO_UNLOCK(bo); /* Don't countdeps with the bo lock held. */ if (buf_countdeps(nbp, 0)) { BO_LOCK(bo); BUF_UNLOCK(nbp); continue; } if (nbp->b_flags & B_CLUSTEROK) { vfs_bio_awrite(nbp); } else { bremfree(nbp); bawrite(nbp); } dirtybufferflushes++; break; } if (nbp == NULL) BO_UNLOCK(bo); } } /* * Delayed write. (Buffer is marked dirty). Do not bother writing * anything if the buffer is marked invalid. * * Note that since the buffer must be completely valid, we can safely * set B_CACHE. In fact, we have to set B_CACHE here rather then in * biodone() in order to prevent getblk from writing the buffer * out synchronously. */ void bdwrite(struct buf *bp) { struct thread *td = curthread; struct vnode *vp; struct bufobj *bo; CTR3(KTR_BUF, "bdwrite(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); KASSERT(bp->b_bufobj != NULL, ("No b_bufobj %p", bp)); KASSERT((bp->b_flags & B_BARRIER) == 0, ("Barrier request in delayed write %p", bp)); BUF_ASSERT_HELD(bp); if (bp->b_flags & B_INVAL) { brelse(bp); return; } /* * If we have too many dirty buffers, don't create any more. * If we are wildly over our limit, then force a complete * cleanup. Otherwise, just keep the situation from getting * out of control. Note that we have to avoid a recursive * disaster and not try to clean up after our own cleanup! */ vp = bp->b_vp; bo = bp->b_bufobj; if ((td->td_pflags & (TDP_COWINPROGRESS|TDP_INBDFLUSH)) == 0) { td->td_pflags |= TDP_INBDFLUSH; BO_BDFLUSH(bo, bp); td->td_pflags &= ~TDP_INBDFLUSH; } else recursiveflushes++; bdirty(bp); /* * Set B_CACHE, indicating that the buffer is fully valid. This is * true even of NFS now. */ bp->b_flags |= B_CACHE; /* * This bmap keeps the system from needing to do the bmap later, * perhaps when the system is attempting to do a sync. Since it * is likely that the indirect block -- or whatever other datastructure * that the filesystem needs is still in memory now, it is a good * thing to do this. Note also, that if the pageout daemon is * requesting a sync -- there might not be enough memory to do * the bmap then... So, this is important to do. */ if (vp->v_type != VCHR && bp->b_lblkno == bp->b_blkno) { VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL); } /* * Set the *dirty* buffer range based upon the VM system dirty * pages. * * Mark the buffer pages as clean. We need to do this here to * satisfy the vnode_pager and the pageout daemon, so that it * thinks that the pages have been "cleaned". Note that since * the pages are in a delayed write buffer -- the VFS layer * "will" see that the pages get written out on the next sync, * or perhaps the cluster will be completed. */ vfs_clean_pages_dirty_buf(bp); bqrelse(bp); /* * note: we cannot initiate I/O from a bdwrite even if we wanted to, * due to the softdep code. */ } /* * bdirty: * * Turn buffer into delayed write request. We must clear BIO_READ and * B_RELBUF, and we must set B_DELWRI. We reassign the buffer to * itself to properly update it in the dirty/clean lists. We mark it * B_DONE to ensure that any asynchronization of the buffer properly * clears B_DONE ( else a panic will occur later ). * * bdirty() is kinda like bdwrite() - we have to clear B_INVAL which * might have been set pre-getblk(). Unlike bwrite/bdwrite, bdirty() * should only be called if the buffer is known-good. * * Since the buffer is not on a queue, we do not update the numfreebuffers * count. * * The buffer must be on QUEUE_NONE. */ void bdirty(struct buf *bp) { CTR3(KTR_BUF, "bdirty(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); KASSERT(bp->b_bufobj != NULL, ("No b_bufobj %p", bp)); KASSERT(bp->b_flags & B_REMFREE || bp->b_qindex == QUEUE_NONE, ("bdirty: buffer %p still on queue %d", bp, bp->b_qindex)); BUF_ASSERT_HELD(bp); bp->b_flags &= ~(B_RELBUF); bp->b_iocmd = BIO_WRITE; if ((bp->b_flags & B_DELWRI) == 0) { bp->b_flags |= /* XXX B_DONE | */ B_DELWRI; reassignbuf(bp); bdirtyadd(); } } /* * bundirty: * * Clear B_DELWRI for buffer. * * Since the buffer is not on a queue, we do not update the numfreebuffers * count. * * The buffer must be on QUEUE_NONE. */ void bundirty(struct buf *bp) { CTR3(KTR_BUF, "bundirty(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); KASSERT(bp->b_bufobj != NULL, ("No b_bufobj %p", bp)); KASSERT(bp->b_flags & B_REMFREE || bp->b_qindex == QUEUE_NONE, ("bundirty: buffer %p still on queue %d", bp, bp->b_qindex)); BUF_ASSERT_HELD(bp); if (bp->b_flags & B_DELWRI) { bp->b_flags &= ~B_DELWRI; reassignbuf(bp); bdirtysub(); } /* * Since it is now being written, we can clear its deferred write flag. */ bp->b_flags &= ~B_DEFERRED; } /* * bawrite: * * Asynchronous write. Start output on a buffer, but do not wait for * it to complete. The buffer is released when the output completes. * * bwrite() ( or the VOP routine anyway ) is responsible for handling * B_INVAL buffers. Not us. */ void bawrite(struct buf *bp) { bp->b_flags |= B_ASYNC; (void) bwrite(bp); } /* * babarrierwrite: * * Asynchronous barrier write. Start output on a buffer, but do not * wait for it to complete. Place a write barrier after this write so * that this buffer and all buffers written before it are committed to * the disk before any buffers written after this write are committed * to the disk. The buffer is released when the output completes. */ void babarrierwrite(struct buf *bp) { bp->b_flags |= B_ASYNC | B_BARRIER; (void) bwrite(bp); } /* * bbarrierwrite: * * Synchronous barrier write. Start output on a buffer and wait for * it to complete. Place a write barrier after this write so that * this buffer and all buffers written before it are committed to * the disk before any buffers written after this write are committed * to the disk. The buffer is released when the output completes. */ int bbarrierwrite(struct buf *bp) { bp->b_flags |= B_BARRIER; return (bwrite(bp)); } /* * bwillwrite: * * Called prior to the locking of any vnodes when we are expecting to * write. We do not want to starve the buffer cache with too many * dirty buffers so we block here. By blocking prior to the locking * of any vnodes we attempt to avoid the situation where a locked vnode * prevents the various system daemons from flushing related buffers. */ void bwillwrite(void) { if (numdirtybuffers >= hidirtybuffers) { mtx_lock(&bdirtylock); while (numdirtybuffers >= hidirtybuffers) { bdirtywait = 1; msleep(&bdirtywait, &bdirtylock, (PRIBIO + 4), "flswai", 0); } mtx_unlock(&bdirtylock); } } /* * Return true if we have too many dirty buffers. */ int buf_dirty_count_severe(void) { return(numdirtybuffers >= hidirtybuffers); } static __noinline int buf_vm_page_count_severe(void) { KFAIL_POINT_CODE(DEBUG_FP, buf_pressure, return 1); return vm_page_count_severe(); } /* * brelse: * * Release a busy buffer and, if requested, free its resources. The * buffer will be stashed in the appropriate bufqueue[] allowing it * to be accessed later as a cache entity or reused for other purposes. */ void brelse(struct buf *bp) { int qindex; CTR3(KTR_BUF, "brelse(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); KASSERT(!(bp->b_flags & (B_CLUSTER|B_PAGING)), ("brelse: inappropriate B_PAGING or B_CLUSTER bp %p", bp)); if (BUF_LOCKRECURSED(bp)) { /* * Do not process, in particular, do not handle the * B_INVAL/B_RELBUF and do not release to free list. */ BUF_UNLOCK(bp); return; } if (bp->b_flags & B_MANAGED) { bqrelse(bp); return; } if (bp->b_iocmd == BIO_WRITE && (bp->b_ioflags & BIO_ERROR) && bp->b_error == EIO && !(bp->b_flags & B_INVAL)) { /* * Failed write, redirty. Must clear BIO_ERROR to prevent * pages from being scrapped. If the error is anything * other than an I/O error (EIO), assume that retrying * is futile. */ bp->b_ioflags &= ~BIO_ERROR; bdirty(bp); } else if ((bp->b_flags & (B_NOCACHE | B_INVAL)) || (bp->b_ioflags & BIO_ERROR) || (bp->b_bufsize <= 0)) { /* * Either a failed I/O or we were asked to free or not * cache the buffer. */ bp->b_flags |= B_INVAL; if (!LIST_EMPTY(&bp->b_dep)) buf_deallocate(bp); if (bp->b_flags & B_DELWRI) bdirtysub(); bp->b_flags &= ~(B_DELWRI | B_CACHE); if ((bp->b_flags & B_VMIO) == 0) { if (bp->b_bufsize) allocbuf(bp, 0); if (bp->b_vp) brelvp(bp); } } /* * We must clear B_RELBUF if B_DELWRI is set. If vfs_vmio_release() * is called with B_DELWRI set, the underlying pages may wind up * getting freed causing a previous write (bdwrite()) to get 'lost' * because pages associated with a B_DELWRI bp are marked clean. * * We still allow the B_INVAL case to call vfs_vmio_release(), even * if B_DELWRI is set. * * If B_DELWRI is not set we may have to set B_RELBUF if we are low * on pages to return pages to the VM page queues. */ if (bp->b_flags & B_DELWRI) bp->b_flags &= ~B_RELBUF; else if (buf_vm_page_count_severe()) { /* * BKGRDINPROG can only be set with the buf and bufobj * locks both held. We tolerate a race to clear it here. */ if (!(bp->b_vflags & BV_BKGRDINPROG)) bp->b_flags |= B_RELBUF; } /* * VMIO buffer rundown. It is not very necessary to keep a VMIO buffer * constituted, not even NFS buffers now. Two flags effect this. If * B_INVAL, the struct buf is invalidated but the VM object is kept * around ( i.e. so it is trivial to reconstitute the buffer later ). * * If BIO_ERROR or B_NOCACHE is set, pages in the VM object will be * invalidated. BIO_ERROR cannot be set for a failed write unless the * buffer is also B_INVAL because it hits the re-dirtying code above. * * Normally we can do this whether a buffer is B_DELWRI or not. If * the buffer is an NFS buffer, it is tracking piecemeal writes or * the commit state and we cannot afford to lose the buffer. If the * buffer has a background write in progress, we need to keep it * around to prevent it from being reconstituted and starting a second * background write. */ if ((bp->b_flags & B_VMIO) && !(bp->b_vp->v_mount != NULL && (bp->b_vp->v_mount->mnt_vfc->vfc_flags & VFCF_NETWORK) != 0 && !vn_isdisk(bp->b_vp, NULL) && (bp->b_flags & B_DELWRI)) ) { int i, j, resid; vm_page_t m; off_t foff; vm_pindex_t poff; vm_object_t obj; obj = bp->b_bufobj->bo_object; /* * Get the base offset and length of the buffer. Note that * in the VMIO case if the buffer block size is not * page-aligned then b_data pointer may not be page-aligned. * But our b_pages[] array *IS* page aligned. * * block sizes less then DEV_BSIZE (usually 512) are not * supported due to the page granularity bits (m->valid, * m->dirty, etc...). * * See man buf(9) for more information */ resid = bp->b_bufsize; foff = bp->b_offset; for (i = 0; i < bp->b_npages; i++) { int had_bogus = 0; m = bp->b_pages[i]; /* * If we hit a bogus page, fixup *all* the bogus pages * now. */ if (m == bogus_page) { poff = OFF_TO_IDX(bp->b_offset); had_bogus = 1; VM_OBJECT_RLOCK(obj); for (j = i; j < bp->b_npages; j++) { vm_page_t mtmp; mtmp = bp->b_pages[j]; if (mtmp == bogus_page) { mtmp = vm_page_lookup(obj, poff + j); if (!mtmp) { panic("brelse: page missing\n"); } bp->b_pages[j] = mtmp; } } VM_OBJECT_RUNLOCK(obj); if ((bp->b_flags & (B_INVAL | B_UNMAPPED)) == 0) { BUF_CHECK_MAPPED(bp); pmap_qenter( trunc_page((vm_offset_t)bp->b_data), bp->b_pages, bp->b_npages); } m = bp->b_pages[i]; } if ((bp->b_flags & B_NOCACHE) || (bp->b_ioflags & BIO_ERROR && bp->b_iocmd == BIO_READ)) { int poffset = foff & PAGE_MASK; int presid = resid > (PAGE_SIZE - poffset) ? (PAGE_SIZE - poffset) : resid; KASSERT(presid >= 0, ("brelse: extra page")); VM_OBJECT_WLOCK(obj); while (vm_page_xbusied(m)) { vm_page_lock(m); VM_OBJECT_WUNLOCK(obj); vm_page_busy_sleep(m, "mbncsh"); VM_OBJECT_WLOCK(obj); } if (pmap_page_wired_mappings(m) == 0) vm_page_set_invalid(m, poffset, presid); VM_OBJECT_WUNLOCK(obj); if (had_bogus) printf("avoided corruption bug in bogus_page/brelse code\n"); } resid -= PAGE_SIZE - (foff & PAGE_MASK); foff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK; } if (bp->b_flags & (B_INVAL | B_RELBUF)) vfs_vmio_release(bp); } else if (bp->b_flags & B_VMIO) { if (bp->b_flags & (B_INVAL | B_RELBUF)) { vfs_vmio_release(bp); } } else if ((bp->b_flags & (B_INVAL | B_RELBUF)) != 0) { if (bp->b_bufsize != 0) allocbuf(bp, 0); if (bp->b_vp != NULL) brelvp(bp); } /* * If the buffer has junk contents signal it and eventually * clean up B_DELWRI and diassociate the vnode so that gbincore() * doesn't find it. */ if (bp->b_bufsize == 0 || (bp->b_ioflags & BIO_ERROR) != 0 || (bp->b_flags & (B_INVAL | B_NOCACHE | B_RELBUF)) != 0) bp->b_flags |= B_INVAL; if (bp->b_flags & B_INVAL) { if (bp->b_flags & B_DELWRI) bundirty(bp); if (bp->b_vp) brelvp(bp); } /* buffers with no memory */ if (bp->b_bufsize == 0) { bp->b_xflags &= ~(BX_BKGRDWRITE | BX_ALTDATA); if (bp->b_vflags & BV_BKGRDINPROG) panic("losing buffer 1"); if (bp->b_kvasize) qindex = QUEUE_EMPTYKVA; else qindex = QUEUE_EMPTY; bp->b_flags |= B_AGE; /* buffers with junk contents */ } else if (bp->b_flags & (B_INVAL | B_NOCACHE | B_RELBUF) || (bp->b_ioflags & BIO_ERROR)) { bp->b_xflags &= ~(BX_BKGRDWRITE | BX_ALTDATA); if (bp->b_vflags & BV_BKGRDINPROG) panic("losing buffer 2"); qindex = QUEUE_CLEAN; bp->b_flags |= B_AGE; /* remaining buffers */ } else if (bp->b_flags & B_DELWRI) qindex = QUEUE_DIRTY; else qindex = QUEUE_CLEAN; binsfree(bp, qindex); bp->b_flags &= ~(B_ASYNC | B_NOCACHE | B_AGE | B_RELBUF | B_DIRECT); if ((bp->b_flags & B_DELWRI) == 0 && (bp->b_xflags & BX_VNDIRTY)) panic("brelse: not dirty"); /* unlock */ BUF_UNLOCK(bp); } /* * Release a buffer back to the appropriate queue but do not try to free * it. The buffer is expected to be used again soon. * * bqrelse() is used by bdwrite() to requeue a delayed write, and used by * biodone() to requeue an async I/O on completion. It is also used when * known good buffers need to be requeued but we think we may need the data * again soon. * * XXX we should be able to leave the B_RELBUF hint set on completion. */ void bqrelse(struct buf *bp) { int qindex; CTR3(KTR_BUF, "bqrelse(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); KASSERT(!(bp->b_flags & (B_CLUSTER|B_PAGING)), ("bqrelse: inappropriate B_PAGING or B_CLUSTER bp %p", bp)); if (BUF_LOCKRECURSED(bp)) { /* do not release to free list */ BUF_UNLOCK(bp); return; } bp->b_flags &= ~(B_ASYNC | B_NOCACHE | B_AGE | B_RELBUF); if (bp->b_flags & B_MANAGED) { if (bp->b_flags & B_REMFREE) bremfreef(bp); goto out; } /* buffers with stale but valid contents */ if (bp->b_flags & B_DELWRI) { qindex = QUEUE_DIRTY; } else { if ((bp->b_flags & B_DELWRI) == 0 && (bp->b_xflags & BX_VNDIRTY)) panic("bqrelse: not dirty"); /* * BKGRDINPROG can only be set with the buf and bufobj * locks both held. We tolerate a race to clear it here. */ if (buf_vm_page_count_severe() && (bp->b_vflags & BV_BKGRDINPROG) == 0) { /* * We are too low on memory, we have to try to free * the buffer (most importantly: the wired pages * making up its backing store) *now*. */ brelse(bp); return; } qindex = QUEUE_CLEAN; } binsfree(bp, qindex); out: /* unlock */ BUF_UNLOCK(bp); } /* Give pages used by the bp back to the VM system (where possible) */ static void vfs_vmio_release(struct buf *bp) { vm_object_t obj; vm_page_t m; int i; if ((bp->b_flags & B_UNMAPPED) == 0) { BUF_CHECK_MAPPED(bp); pmap_qremove(trunc_page((vm_offset_t)bp->b_data), bp->b_npages); } else BUF_CHECK_UNMAPPED(bp); obj = bp->b_bufobj->bo_object; if (obj != NULL) VM_OBJECT_WLOCK(obj); for (i = 0; i < bp->b_npages; i++) { m = bp->b_pages[i]; bp->b_pages[i] = NULL; /* * In order to keep page LRU ordering consistent, put * everything on the inactive queue. */ vm_page_lock(m); vm_page_unwire(m, PQ_INACTIVE); /* * Might as well free the page if we can and it has * no valid data. We also free the page if the * buffer was used for direct I/O */ if ((bp->b_flags & B_ASYNC) == 0 && !m->valid) { if (m->wire_count == 0 && !vm_page_busied(m)) vm_page_free(m); } else if (bp->b_flags & B_DIRECT) vm_page_try_to_free(m); else if (buf_vm_page_count_severe()) vm_page_try_to_cache(m); vm_page_unlock(m); } if (obj != NULL) VM_OBJECT_WUNLOCK(obj); if (bp->b_bufsize) { bufspacewakeup(); bp->b_bufsize = 0; } bp->b_npages = 0; bp->b_flags &= ~B_VMIO; if (bp->b_vp) brelvp(bp); } /* * Check to see if a block at a particular lbn is available for a clustered * write. */ static int vfs_bio_clcheck(struct vnode *vp, int size, daddr_t lblkno, daddr_t blkno) { struct buf *bpa; int match; match = 0; /* If the buf isn't in core skip it */ if ((bpa = gbincore(&vp->v_bufobj, lblkno)) == NULL) return (0); /* If the buf is busy we don't want to wait for it */ if (BUF_LOCK(bpa, LK_EXCLUSIVE | LK_NOWAIT, NULL) != 0) return (0); /* Only cluster with valid clusterable delayed write buffers */ if ((bpa->b_flags & (B_DELWRI | B_CLUSTEROK | B_INVAL)) != (B_DELWRI | B_CLUSTEROK)) goto done; if (bpa->b_bufsize != size) goto done; /* * Check to see if it is in the expected place on disk and that the * block has been mapped. */ if ((bpa->b_blkno != bpa->b_lblkno) && (bpa->b_blkno == blkno)) match = 1; done: BUF_UNLOCK(bpa); return (match); } /* * vfs_bio_awrite: * * Implement clustered async writes for clearing out B_DELWRI buffers. * This is much better then the old way of writing only one buffer at * a time. Note that we may not be presented with the buffers in the * correct order, so we search for the cluster in both directions. */ int vfs_bio_awrite(struct buf *bp) { struct bufobj *bo; int i; int j; daddr_t lblkno = bp->b_lblkno; struct vnode *vp = bp->b_vp; int ncl; int nwritten; int size; int maxcl; int gbflags; bo = &vp->v_bufobj; gbflags = (bp->b_flags & B_UNMAPPED) != 0 ? GB_UNMAPPED : 0; /* * right now we support clustered writing only to regular files. If * we find a clusterable block we could be in the middle of a cluster * rather then at the beginning. */ if ((vp->v_type == VREG) && (vp->v_mount != 0) && /* Only on nodes that have the size info */ (bp->b_flags & (B_CLUSTEROK | B_INVAL)) == B_CLUSTEROK) { size = vp->v_mount->mnt_stat.f_iosize; maxcl = MAXPHYS / size; BO_RLOCK(bo); for (i = 1; i < maxcl; i++) if (vfs_bio_clcheck(vp, size, lblkno + i, bp->b_blkno + ((i * size) >> DEV_BSHIFT)) == 0) break; for (j = 1; i + j <= maxcl && j <= lblkno; j++) if (vfs_bio_clcheck(vp, size, lblkno - j, bp->b_blkno - ((j * size) >> DEV_BSHIFT)) == 0) break; BO_RUNLOCK(bo); --j; ncl = i + j; /* * this is a possible cluster write */ if (ncl != 1) { BUF_UNLOCK(bp); nwritten = cluster_wbuild(vp, size, lblkno - j, ncl, gbflags); return (nwritten); } } bremfree(bp); bp->b_flags |= B_ASYNC; /* * default (old) behavior, writing out only one block * * XXX returns b_bufsize instead of b_bcount for nwritten? */ nwritten = bp->b_bufsize; (void) bwrite(bp); return (nwritten); } static void setbufkva(struct buf *bp, vm_offset_t addr, int maxsize, int gbflags) { KASSERT((bp->b_flags & (B_UNMAPPED | B_KVAALLOC)) == 0 && bp->b_kvasize == 0, ("call bfreekva(%p)", bp)); if ((gbflags & GB_UNMAPPED) == 0) { bp->b_kvabase = (caddr_t)addr; } else if ((gbflags & GB_KVAALLOC) != 0) { KASSERT((gbflags & GB_UNMAPPED) != 0, ("GB_KVAALLOC without GB_UNMAPPED")); bp->b_kvaalloc = (caddr_t)addr; bp->b_flags |= B_UNMAPPED | B_KVAALLOC; atomic_add_long(&unmapped_bufspace, bp->b_kvasize); } bp->b_kvasize = maxsize; } /* * Allocate the buffer KVA and set b_kvasize. Also set b_kvabase if * needed. */ static int allocbufkva(struct buf *bp, int maxsize, int gbflags) { vm_offset_t addr; bfreekva(bp); addr = 0; if (vmem_alloc(buffer_arena, maxsize, M_BESTFIT | M_NOWAIT, &addr)) { /* * Buffer map is too fragmented. Request the caller * to defragment the map. */ atomic_add_int(&bufdefragcnt, 1); return (1); } setbufkva(bp, addr, maxsize, gbflags); atomic_add_long(&bufspace, bp->b_kvasize); return (0); } /* * Ask the bufdaemon for help, or act as bufdaemon itself, when a * locked vnode is supplied. */ static void getnewbuf_bufd_help(struct vnode *vp, int gbflags, int slpflag, int slptimeo, int defrag) { struct thread *td; char *waitmsg; int cnt, error, flags, norunbuf, wait; mtx_assert(&bqclean, MA_OWNED); if (defrag) { flags = VFS_BIO_NEED_BUFSPACE; waitmsg = "nbufkv"; } else if (bufspace >= hibufspace) { waitmsg = "nbufbs"; flags = VFS_BIO_NEED_BUFSPACE; } else { waitmsg = "newbuf"; flags = VFS_BIO_NEED_ANY; } atomic_set_int(&needsbuffer, flags); mtx_unlock(&bqclean); bd_speedup(); /* heeeelp */ if ((gbflags & GB_NOWAIT_BD) != 0) return; td = curthread; cnt = 0; wait = MNT_NOWAIT; rw_wlock(&nblock); while ((needsbuffer & flags) != 0) { if (vp != NULL && vp->v_type != VCHR && (td->td_pflags & TDP_BUFNEED) == 0) { rw_wunlock(&nblock); /* * getblk() is called with a vnode locked, and * some majority of the dirty buffers may as * well belong to the vnode. Flushing the * buffers there would make a progress that * cannot be achieved by the buf_daemon, that * cannot lock the vnode. */ if (cnt++ > 2) wait = MNT_WAIT; ASSERT_VOP_LOCKED(vp, "bufd_helper"); error = VOP_ISLOCKED(vp) == LK_EXCLUSIVE ? 0 : vn_lock(vp, LK_TRYUPGRADE); if (error == 0) { /* play bufdaemon */ norunbuf = curthread_pflags_set(TDP_BUFNEED | TDP_NORUNNINGBUF); VOP_FSYNC(vp, wait, td); atomic_add_long(¬bufdflushes, 1); curthread_pflags_restore(norunbuf); } rw_wlock(&nblock); if ((needsbuffer & flags) == 0) break; } error = rw_sleep(__DEVOLATILE(void *, &needsbuffer), &nblock, (PRIBIO + 4) | slpflag, waitmsg, slptimeo); if (error != 0) break; } rw_wunlock(&nblock); } static void getnewbuf_reuse_bp(struct buf *bp, int qindex) { CTR6(KTR_BUF, "getnewbuf(%p) vp %p flags %X kvasize %d bufsize %d " "queue %d (recycling)", bp, bp->b_vp, bp->b_flags, bp->b_kvasize, bp->b_bufsize, qindex); mtx_assert(&bqclean, MA_NOTOWNED); /* * Note: we no longer distinguish between VMIO and non-VMIO * buffers. */ KASSERT((bp->b_flags & B_DELWRI) == 0, ("delwri buffer %p found in queue %d", bp, qindex)); if (qindex == QUEUE_CLEAN) { if (bp->b_flags & B_VMIO) { bp->b_flags &= ~B_ASYNC; vfs_vmio_release(bp); } if (bp->b_vp != NULL) brelvp(bp); } /* * Get the rest of the buffer freed up. b_kva* is still valid * after this operation. */ if (bp->b_rcred != NOCRED) { crfree(bp->b_rcred); bp->b_rcred = NOCRED; } if (bp->b_wcred != NOCRED) { crfree(bp->b_wcred); bp->b_wcred = NOCRED; } if (!LIST_EMPTY(&bp->b_dep)) buf_deallocate(bp); if (bp->b_vflags & BV_BKGRDINPROG) panic("losing buffer 3"); KASSERT(bp->b_vp == NULL, ("bp: %p still has vnode %p. qindex: %d", bp, bp->b_vp, qindex)); KASSERT((bp->b_xflags & (BX_VNCLEAN|BX_VNDIRTY)) == 0, ("bp: %p still on a buffer list. xflags %X", bp, bp->b_xflags)); if (bp->b_bufsize) allocbuf(bp, 0); bp->b_flags &= B_UNMAPPED | B_KVAALLOC; bp->b_ioflags = 0; bp->b_xflags = 0; KASSERT((bp->b_flags & B_INFREECNT) == 0, ("buf %p still counted as free?", bp)); bp->b_vflags = 0; bp->b_vp = NULL; bp->b_blkno = bp->b_lblkno = 0; bp->b_offset = NOOFFSET; bp->b_iodone = 0; bp->b_error = 0; bp->b_resid = 0; bp->b_bcount = 0; bp->b_npages = 0; bp->b_dirtyoff = bp->b_dirtyend = 0; bp->b_bufobj = NULL; bp->b_pin_count = 0; bp->b_fsprivate1 = NULL; bp->b_fsprivate2 = NULL; bp->b_fsprivate3 = NULL; LIST_INIT(&bp->b_dep); } static int flushingbufs; static struct buf * getnewbuf_scan(int maxsize, int defrag, int unmapped, int metadata) { struct buf *bp, *nbp; int nqindex, qindex, pass; KASSERT(!unmapped || !defrag, ("both unmapped and defrag")); pass = 1; restart: atomic_add_int(&getnewbufrestarts, 1); /* * Setup for scan. If we do not have enough free buffers, * we setup a degenerate case that immediately fails. Note * that if we are specially marked process, we are allowed to * dip into our reserves. * * The scanning sequence is nominally: EMPTY->EMPTYKVA->CLEAN * for the allocation of the mapped buffer. For unmapped, the * easiest is to start with EMPTY outright. * * We start with EMPTYKVA. If the list is empty we backup to EMPTY. * However, there are a number of cases (defragging, reusing, ...) * where we cannot backup. */ nbp = NULL; mtx_lock(&bqclean); if (!defrag && unmapped) { nqindex = QUEUE_EMPTY; nbp = TAILQ_FIRST(&bufqueues[QUEUE_EMPTY]); } if (nbp == NULL) { nqindex = QUEUE_EMPTYKVA; nbp = TAILQ_FIRST(&bufqueues[QUEUE_EMPTYKVA]); } /* * If no EMPTYKVA buffers and we are either defragging or * reusing, locate a CLEAN buffer to free or reuse. If * bufspace useage is low skip this step so we can allocate a * new buffer. */ if (nbp == NULL && (defrag || bufspace >= lobufspace)) { nqindex = QUEUE_CLEAN; nbp = TAILQ_FIRST(&bufqueues[QUEUE_CLEAN]); } /* * If we could not find or were not allowed to reuse a CLEAN * buffer, check to see if it is ok to use an EMPTY buffer. * We can only use an EMPTY buffer if allocating its KVA would * not otherwise run us out of buffer space. No KVA is needed * for the unmapped allocation. */ if (nbp == NULL && defrag == 0 && (bufspace + maxsize < hibufspace || metadata)) { nqindex = QUEUE_EMPTY; nbp = TAILQ_FIRST(&bufqueues[QUEUE_EMPTY]); } /* * All available buffers might be clean, retry ignoring the * lobufspace as the last resort. */ if (nbp == NULL && !TAILQ_EMPTY(&bufqueues[QUEUE_CLEAN])) { nqindex = QUEUE_CLEAN; nbp = TAILQ_FIRST(&bufqueues[QUEUE_CLEAN]); } /* * Run scan, possibly freeing data and/or kva mappings on the fly * depending. */ while ((bp = nbp) != NULL) { qindex = nqindex; /* * Calculate next bp (we can only use it if we do not * block or do other fancy things). */ if ((nbp = TAILQ_NEXT(bp, b_freelist)) == NULL) { switch (qindex) { case QUEUE_EMPTY: nqindex = QUEUE_EMPTYKVA; nbp = TAILQ_FIRST(&bufqueues[QUEUE_EMPTYKVA]); if (nbp != NULL) break; /* FALLTHROUGH */ case QUEUE_EMPTYKVA: nqindex = QUEUE_CLEAN; nbp = TAILQ_FIRST(&bufqueues[QUEUE_CLEAN]); if (nbp != NULL) break; /* FALLTHROUGH */ case QUEUE_CLEAN: if (metadata && pass == 1) { pass = 2; nqindex = QUEUE_EMPTY; nbp = TAILQ_FIRST( &bufqueues[QUEUE_EMPTY]); } /* * nbp is NULL. */ break; } } /* * If we are defragging then we need a buffer with * b_kvasize != 0. XXX this situation should no longer * occur, if defrag is non-zero the buffer's b_kvasize * should also be non-zero at this point. XXX */ if (defrag && bp->b_kvasize == 0) { printf("Warning: defrag empty buffer %p\n", bp); continue; } /* * Start freeing the bp. This is somewhat involved. nbp * remains valid only for QUEUE_EMPTY[KVA] bp's. */ if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL) != 0) continue; /* * BKGRDINPROG can only be set with the buf and bufobj * locks both held. We tolerate a race to clear it here. */ if (bp->b_vflags & BV_BKGRDINPROG) { BUF_UNLOCK(bp); continue; } KASSERT(bp->b_qindex == qindex, ("getnewbuf: inconsistent queue %d bp %p", qindex, bp)); bremfreel(bp); mtx_unlock(&bqclean); /* * NOTE: nbp is now entirely invalid. We can only restart * the scan from this point on. */ getnewbuf_reuse_bp(bp, qindex); mtx_assert(&bqclean, MA_NOTOWNED); /* * If we are defragging then free the buffer. */ if (defrag) { bp->b_flags |= B_INVAL; bfreekva(bp); brelse(bp); defrag = 0; goto restart; } /* * Notify any waiters for the buffer lock about * identity change by freeing the buffer. */ if (qindex == QUEUE_CLEAN && BUF_LOCKWAITERS(bp)) { bp->b_flags |= B_INVAL; bfreekva(bp); brelse(bp); goto restart; } if (metadata) break; /* * If we are overcomitted then recover the buffer and its * KVM space. This occurs in rare situations when multiple * processes are blocked in getnewbuf() or allocbuf(). */ if (bufspace >= hibufspace) flushingbufs = 1; if (flushingbufs && bp->b_kvasize != 0) { bp->b_flags |= B_INVAL; bfreekva(bp); brelse(bp); goto restart; } if (bufspace < lobufspace) flushingbufs = 0; break; } return (bp); } /* * getnewbuf: * * Find and initialize a new buffer header, freeing up existing buffers * in the bufqueues as necessary. The new buffer is returned locked. * * Important: B_INVAL is not set. If the caller wishes to throw the * buffer away, the caller must set B_INVAL prior to calling brelse(). * * We block if: * We have insufficient buffer headers * We have insufficient buffer space * buffer_arena is too fragmented ( space reservation fails ) * If we have to flush dirty buffers ( but we try to avoid this ) */ static struct buf * getnewbuf(struct vnode *vp, int slpflag, int slptimeo, int size, int maxsize, int gbflags) { struct buf *bp; int defrag, metadata; KASSERT((gbflags & (GB_UNMAPPED | GB_KVAALLOC)) != GB_KVAALLOC, ("GB_KVAALLOC only makes sense with GB_UNMAPPED")); if (!unmapped_buf_allowed) gbflags &= ~(GB_UNMAPPED | GB_KVAALLOC); defrag = 0; if (vp == NULL || (vp->v_vflag & (VV_MD | VV_SYSTEM)) != 0 || vp->v_type == VCHR) metadata = 1; else metadata = 0; /* * We can't afford to block since we might be holding a vnode lock, * which may prevent system daemons from running. We deal with * low-memory situations by proactively returning memory and running * async I/O rather then sync I/O. */ atomic_add_int(&getnewbufcalls, 1); atomic_subtract_int(&getnewbufrestarts, 1); restart: bp = getnewbuf_scan(maxsize, defrag, (gbflags & (GB_UNMAPPED | GB_KVAALLOC)) == GB_UNMAPPED, metadata); if (bp != NULL) defrag = 0; /* * If we exhausted our list, sleep as appropriate. We may have to * wakeup various daemons and write out some dirty buffers. * * Generally we are sleeping due to insufficient buffer space. */ if (bp == NULL) { mtx_assert(&bqclean, MA_OWNED); getnewbuf_bufd_help(vp, gbflags, slpflag, slptimeo, defrag); mtx_assert(&bqclean, MA_NOTOWNED); } else if ((gbflags & (GB_UNMAPPED | GB_KVAALLOC)) == GB_UNMAPPED) { mtx_assert(&bqclean, MA_NOTOWNED); bfreekva(bp); bp->b_flags |= B_UNMAPPED; bp->b_kvabase = bp->b_data = unmapped_buf; bp->b_kvasize = maxsize; atomic_add_long(&bufspace, bp->b_kvasize); atomic_add_long(&unmapped_bufspace, bp->b_kvasize); atomic_add_int(&bufreusecnt, 1); } else { mtx_assert(&bqclean, MA_NOTOWNED); /* * We finally have a valid bp. We aren't quite out of the * woods, we still have to reserve kva space. In order * to keep fragmentation sane we only allocate kva in * BKVASIZE chunks. */ maxsize = (maxsize + BKVAMASK) & ~BKVAMASK; if (maxsize != bp->b_kvasize || (bp->b_flags & (B_UNMAPPED | B_KVAALLOC)) == B_UNMAPPED) { if (allocbufkva(bp, maxsize, gbflags)) { defrag = 1; bp->b_flags |= B_INVAL; brelse(bp); goto restart; } atomic_add_int(&bufreusecnt, 1); } else if ((bp->b_flags & B_KVAALLOC) != 0 && (gbflags & (GB_UNMAPPED | GB_KVAALLOC)) == 0) { /* * If the reused buffer has KVA allocated, * reassign b_kvaalloc to b_kvabase. */ bp->b_kvabase = bp->b_kvaalloc; bp->b_flags &= ~B_KVAALLOC; atomic_subtract_long(&unmapped_bufspace, bp->b_kvasize); atomic_add_int(&bufreusecnt, 1); } else if ((bp->b_flags & (B_UNMAPPED | B_KVAALLOC)) == 0 && (gbflags & (GB_UNMAPPED | GB_KVAALLOC)) == (GB_UNMAPPED | GB_KVAALLOC)) { /* * The case of reused buffer already have KVA * mapped, but the request is for unmapped * buffer with KVA allocated. */ bp->b_kvaalloc = bp->b_kvabase; bp->b_data = bp->b_kvabase = unmapped_buf; bp->b_flags |= B_UNMAPPED | B_KVAALLOC; atomic_add_long(&unmapped_bufspace, bp->b_kvasize); atomic_add_int(&bufreusecnt, 1); } if ((gbflags & GB_UNMAPPED) == 0) { bp->b_saveaddr = bp->b_kvabase; bp->b_data = bp->b_saveaddr; bp->b_flags &= ~B_UNMAPPED; BUF_CHECK_MAPPED(bp); } } return (bp); } /* * buf_daemon: * * buffer flushing daemon. Buffers are normally flushed by the * update daemon but if it cannot keep up this process starts to * take the load in an attempt to prevent getnewbuf() from blocking. */ static struct kproc_desc buf_kp = { "bufdaemon", buf_daemon, &bufdaemonproc }; SYSINIT(bufdaemon, SI_SUB_KTHREAD_BUF, SI_ORDER_FIRST, kproc_start, &buf_kp); static int buf_flush(int target) { int flushed; flushed = flushbufqueues(target, 0); if (flushed == 0) { /* * Could not find any buffers without rollback * dependencies, so just write the first one * in the hopes of eventually making progress. */ flushed = flushbufqueues(target, 1); } return (flushed); } static void buf_daemon() { int lodirty; /* * This process needs to be suspended prior to shutdown sync. */ EVENTHANDLER_REGISTER(shutdown_pre_sync, kproc_shutdown, bufdaemonproc, SHUTDOWN_PRI_LAST); /* * This process is allowed to take the buffer cache to the limit */ curthread->td_pflags |= TDP_NORUNNINGBUF | TDP_BUFNEED; mtx_lock(&bdlock); for (;;) { bd_request = 0; mtx_unlock(&bdlock); kproc_suspend_check(bufdaemonproc); lodirty = lodirtybuffers; if (bd_speedupreq) { lodirty = numdirtybuffers / 2; bd_speedupreq = 0; } /* * Do the flush. Limit the amount of in-transit I/O we * allow to build up, otherwise we would completely saturate * the I/O system. */ while (numdirtybuffers > lodirty) { if (buf_flush(numdirtybuffers - lodirty) == 0) break; kern_yield(PRI_USER); } /* * Only clear bd_request if we have reached our low water * mark. The buf_daemon normally waits 1 second and * then incrementally flushes any dirty buffers that have * built up, within reason. * * If we were unable to hit our low water mark and couldn't * find any flushable buffers, we sleep for a short period * to avoid endless loops on unlockable buffers. */ mtx_lock(&bdlock); if (numdirtybuffers <= lodirtybuffers) { /* * We reached our low water mark, reset the * request and sleep until we are needed again. * The sleep is just so the suspend code works. */ bd_request = 0; /* * Do an extra wakeup in case dirty threshold * changed via sysctl and the explicit transition * out of shortfall was missed. */ bdirtywakeup(); if (runningbufspace <= lorunningspace) runningwakeup(); msleep(&bd_request, &bdlock, PVM, "psleep", hz); } else { /* * We couldn't find any flushable dirty buffers but * still have too many dirty buffers, we * have to sleep and try again. (rare) */ msleep(&bd_request, &bdlock, PVM, "qsleep", hz / 10); } } } /* * flushbufqueues: * * Try to flush a buffer in the dirty queue. We must be careful to * free up B_INVAL buffers instead of write them, which NFS is * particularly sensitive to. */ static int flushwithdeps = 0; SYSCTL_INT(_vfs, OID_AUTO, flushwithdeps, CTLFLAG_RW, &flushwithdeps, 0, "Number of buffers flushed with dependecies that require rollbacks"); static int flushbufqueues(int target, int flushdeps) { struct buf *sentinel; struct vnode *vp; struct mount *mp; struct buf *bp; int hasdeps; int flushed; int queue; int error; flushed = 0; queue = QUEUE_DIRTY; bp = NULL; sentinel = malloc(sizeof(struct buf), M_TEMP, M_WAITOK | M_ZERO); sentinel->b_qindex = QUEUE_SENTINEL; mtx_lock(&bqdirty); TAILQ_INSERT_HEAD(&bufqueues[queue], sentinel, b_freelist); mtx_unlock(&bqdirty); while (flushed != target) { maybe_yield(); mtx_lock(&bqdirty); bp = TAILQ_NEXT(sentinel, b_freelist); if (bp != NULL) { TAILQ_REMOVE(&bufqueues[queue], sentinel, b_freelist); TAILQ_INSERT_AFTER(&bufqueues[queue], bp, sentinel, b_freelist); } else { mtx_unlock(&bqdirty); break; } KASSERT(bp->b_qindex != QUEUE_SENTINEL, ("parallel calls to flushbufqueues() bp %p", bp)); error = BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL); mtx_unlock(&bqdirty); if (error != 0) continue; if (bp->b_pin_count > 0) { BUF_UNLOCK(bp); continue; } /* * BKGRDINPROG can only be set with the buf and bufobj * locks both held. We tolerate a race to clear it here. */ if ((bp->b_vflags & BV_BKGRDINPROG) != 0 || (bp->b_flags & B_DELWRI) == 0) { BUF_UNLOCK(bp); continue; } if (bp->b_flags & B_INVAL) { bremfreef(bp); brelse(bp); flushed++; continue; } if (!LIST_EMPTY(&bp->b_dep) && buf_countdeps(bp, 0)) { if (flushdeps == 0) { BUF_UNLOCK(bp); continue; } hasdeps = 1; } else hasdeps = 0; /* * We must hold the lock on a vnode before writing * one of its buffers. Otherwise we may confuse, or * in the case of a snapshot vnode, deadlock the * system. * * The lock order here is the reverse of the normal * of vnode followed by buf lock. This is ok because * the NOWAIT will prevent deadlock. */ vp = bp->b_vp; if (vn_start_write(vp, &mp, V_NOWAIT) != 0) { BUF_UNLOCK(bp); continue; } error = vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT); if (error == 0) { CTR3(KTR_BUF, "flushbufqueue(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); vfs_bio_awrite(bp); vn_finished_write(mp); VOP_UNLOCK(vp, 0); flushwithdeps += hasdeps; flushed++; if (runningbufspace > hirunningspace) waitrunningbufspace(); continue; } vn_finished_write(mp); BUF_UNLOCK(bp); } mtx_lock(&bqdirty); TAILQ_REMOVE(&bufqueues[queue], sentinel, b_freelist); mtx_unlock(&bqdirty); free(sentinel, M_TEMP); return (flushed); } /* * Check to see if a block is currently memory resident. */ struct buf * incore(struct bufobj *bo, daddr_t blkno) { struct buf *bp; BO_RLOCK(bo); bp = gbincore(bo, blkno); BO_RUNLOCK(bo); return (bp); } /* * Returns true if no I/O is needed to access the * associated VM object. This is like incore except * it also hunts around in the VM system for the data. */ static int inmem(struct vnode * vp, daddr_t blkno) { vm_object_t obj; vm_offset_t toff, tinc, size; vm_page_t m; vm_ooffset_t off; ASSERT_VOP_LOCKED(vp, "inmem"); if (incore(&vp->v_bufobj, blkno)) return 1; if (vp->v_mount == NULL) return 0; obj = vp->v_object; if (obj == NULL) return (0); size = PAGE_SIZE; if (size > vp->v_mount->mnt_stat.f_iosize) size = vp->v_mount->mnt_stat.f_iosize; off = (vm_ooffset_t)blkno * (vm_ooffset_t)vp->v_mount->mnt_stat.f_iosize; VM_OBJECT_RLOCK(obj); for (toff = 0; toff < vp->v_mount->mnt_stat.f_iosize; toff += tinc) { m = vm_page_lookup(obj, OFF_TO_IDX(off + toff)); if (!m) goto notinmem; tinc = size; if (tinc > PAGE_SIZE - ((toff + off) & PAGE_MASK)) tinc = PAGE_SIZE - ((toff + off) & PAGE_MASK); if (vm_page_is_valid(m, (vm_offset_t) ((toff + off) & PAGE_MASK), tinc) == 0) goto notinmem; } VM_OBJECT_RUNLOCK(obj); return 1; notinmem: VM_OBJECT_RUNLOCK(obj); return (0); } /* * Set the dirty range for a buffer based on the status of the dirty * bits in the pages comprising the buffer. The range is limited * to the size of the buffer. * * Tell the VM system that the pages associated with this buffer * are clean. This is used for delayed writes where the data is * going to go to disk eventually without additional VM intevention. * * Note that while we only really need to clean through to b_bcount, we * just go ahead and clean through to b_bufsize. */ static void vfs_clean_pages_dirty_buf(struct buf *bp) { vm_ooffset_t foff, noff, eoff; vm_page_t m; int i; if ((bp->b_flags & B_VMIO) == 0 || bp->b_bufsize == 0) return; foff = bp->b_offset; KASSERT(bp->b_offset != NOOFFSET, ("vfs_clean_pages_dirty_buf: no buffer offset")); VM_OBJECT_WLOCK(bp->b_bufobj->bo_object); vfs_drain_busy_pages(bp); vfs_setdirty_locked_object(bp); for (i = 0; i < bp->b_npages; i++) { noff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK; eoff = noff; if (eoff > bp->b_offset + bp->b_bufsize) eoff = bp->b_offset + bp->b_bufsize; m = bp->b_pages[i]; vfs_page_set_validclean(bp, foff, m); /* vm_page_clear_dirty(m, foff & PAGE_MASK, eoff - foff); */ foff = noff; } VM_OBJECT_WUNLOCK(bp->b_bufobj->bo_object); } static void vfs_setdirty_locked_object(struct buf *bp) { vm_object_t object; int i; object = bp->b_bufobj->bo_object; VM_OBJECT_ASSERT_WLOCKED(object); /* * We qualify the scan for modified pages on whether the * object has been flushed yet. */ if ((object->flags & OBJ_MIGHTBEDIRTY) != 0) { vm_offset_t boffset; vm_offset_t eoffset; /* * test the pages to see if they have been modified directly * by users through the VM system. */ for (i = 0; i < bp->b_npages; i++) vm_page_test_dirty(bp->b_pages[i]); /* * Calculate the encompassing dirty range, boffset and eoffset, * (eoffset - boffset) bytes. */ for (i = 0; i < bp->b_npages; i++) { if (bp->b_pages[i]->dirty) break; } boffset = (i << PAGE_SHIFT) - (bp->b_offset & PAGE_MASK); for (i = bp->b_npages - 1; i >= 0; --i) { if (bp->b_pages[i]->dirty) { break; } } eoffset = ((i + 1) << PAGE_SHIFT) - (bp->b_offset & PAGE_MASK); /* * Fit it to the buffer. */ if (eoffset > bp->b_bcount) eoffset = bp->b_bcount; /* * If we have a good dirty range, merge with the existing * dirty range. */ if (boffset < eoffset) { if (bp->b_dirtyoff > boffset) bp->b_dirtyoff = boffset; if (bp->b_dirtyend < eoffset) bp->b_dirtyend = eoffset; } } } /* * Allocate the KVA mapping for an existing buffer. It handles the * cases of both B_UNMAPPED buffer, and buffer with the preallocated * KVA which is not mapped (B_KVAALLOC). */ static void bp_unmapped_get_kva(struct buf *bp, daddr_t blkno, int size, int gbflags) { struct buf *scratch_bp; int bsize, maxsize, need_mapping, need_kva; off_t offset; need_mapping = (bp->b_flags & B_UNMAPPED) != 0 && (gbflags & GB_UNMAPPED) == 0; need_kva = (bp->b_flags & (B_KVAALLOC | B_UNMAPPED)) == B_UNMAPPED && (gbflags & GB_KVAALLOC) != 0; if (!need_mapping && !need_kva) return; BUF_CHECK_UNMAPPED(bp); if (need_mapping && (bp->b_flags & B_KVAALLOC) != 0) { /* * Buffer is not mapped, but the KVA was already * reserved at the time of the instantiation. Use the * allocated space. */ bp->b_flags &= ~B_KVAALLOC; KASSERT(bp->b_kvaalloc != 0, ("kvaalloc == 0")); bp->b_kvabase = bp->b_kvaalloc; atomic_subtract_long(&unmapped_bufspace, bp->b_kvasize); goto has_addr; } /* * Calculate the amount of the address space we would reserve * if the buffer was mapped. */ bsize = vn_isdisk(bp->b_vp, NULL) ? DEV_BSIZE : bp->b_bufobj->bo_bsize; KASSERT(bsize != 0, ("bsize == 0, check bo->bo_bsize")); offset = blkno * bsize; maxsize = size + (offset & PAGE_MASK); maxsize = imax(maxsize, bsize); mapping_loop: if (allocbufkva(bp, maxsize, gbflags)) { /* * Request defragmentation. getnewbuf() returns us the * allocated space by the scratch buffer KVA. */ scratch_bp = getnewbuf(bp->b_vp, 0, 0, size, maxsize, gbflags | (GB_UNMAPPED | GB_KVAALLOC)); if (scratch_bp == NULL) { if ((gbflags & GB_NOWAIT_BD) != 0) { /* * XXXKIB: defragmentation cannot * succeed, not sure what else to do. */ panic("GB_NOWAIT_BD and B_UNMAPPED %p", bp); } atomic_add_int(&mappingrestarts, 1); goto mapping_loop; } KASSERT((scratch_bp->b_flags & B_KVAALLOC) != 0, ("scratch bp !B_KVAALLOC %p", scratch_bp)); setbufkva(bp, (vm_offset_t)scratch_bp->b_kvaalloc, scratch_bp->b_kvasize, gbflags); /* Get rid of the scratch buffer. */ scratch_bp->b_kvasize = 0; scratch_bp->b_flags |= B_INVAL; scratch_bp->b_flags &= ~(B_UNMAPPED | B_KVAALLOC); brelse(scratch_bp); } if (!need_mapping) return; has_addr: bp->b_saveaddr = bp->b_kvabase; bp->b_data = bp->b_saveaddr; /* b_offset is handled by bpmap_qenter */ bp->b_flags &= ~B_UNMAPPED; BUF_CHECK_MAPPED(bp); bpmap_qenter(bp); } /* * getblk: * * Get a block given a specified block and offset into a file/device. * The buffers B_DONE bit will be cleared on return, making it almost * ready for an I/O initiation. B_INVAL may or may not be set on * return. The caller should clear B_INVAL prior to initiating a * READ. * * For a non-VMIO buffer, B_CACHE is set to the opposite of B_INVAL for * an existing buffer. * * For a VMIO buffer, B_CACHE is modified according to the backing VM. * If getblk()ing a previously 0-sized invalid buffer, B_CACHE is set * and then cleared based on the backing VM. If the previous buffer is * non-0-sized but invalid, B_CACHE will be cleared. * * If getblk() must create a new buffer, the new buffer is returned with * both B_INVAL and B_CACHE clear unless it is a VMIO buffer, in which * case it is returned with B_INVAL clear and B_CACHE set based on the * backing VM. * * getblk() also forces a bwrite() for any B_DELWRI buffer whos * B_CACHE bit is clear. * * What this means, basically, is that the caller should use B_CACHE to * determine whether the buffer is fully valid or not and should clear * B_INVAL prior to issuing a read. If the caller intends to validate * the buffer by loading its data area with something, the caller needs * to clear B_INVAL. If the caller does this without issuing an I/O, * the caller should set B_CACHE ( as an optimization ), else the caller * should issue the I/O and biodone() will set B_CACHE if the I/O was * a write attempt or if it was a successfull read. If the caller * intends to issue a READ, the caller must clear B_INVAL and BIO_ERROR * prior to issuing the READ. biodone() will *not* clear B_INVAL. */ struct buf * getblk(struct vnode *vp, daddr_t blkno, int size, int slpflag, int slptimeo, int flags) { struct buf *bp; struct bufobj *bo; int bsize, error, maxsize, vmio; off_t offset; CTR3(KTR_BUF, "getblk(%p, %ld, %d)", vp, (long)blkno, size); KASSERT((flags & (GB_UNMAPPED | GB_KVAALLOC)) != GB_KVAALLOC, ("GB_KVAALLOC only makes sense with GB_UNMAPPED")); ASSERT_VOP_LOCKED(vp, "getblk"); - if (size > MAXBSIZE) - panic("getblk: size(%d) > MAXBSIZE(%d)\n", size, MAXBSIZE); + if (size > MAXBCACHEBUF) + panic("getblk: size(%d) > MAXBCACHEBUF(%d)\n", size, + MAXBCACHEBUF); if (!unmapped_buf_allowed) flags &= ~(GB_UNMAPPED | GB_KVAALLOC); bo = &vp->v_bufobj; loop: BO_RLOCK(bo); bp = gbincore(bo, blkno); if (bp != NULL) { int lockflags; /* * Buffer is in-core. If the buffer is not busy nor managed, * it must be on a queue. */ lockflags = LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK; if (flags & GB_LOCK_NOWAIT) lockflags |= LK_NOWAIT; error = BUF_TIMELOCK(bp, lockflags, BO_LOCKPTR(bo), "getblk", slpflag, slptimeo); /* * If we slept and got the lock we have to restart in case * the buffer changed identities. */ if (error == ENOLCK) goto loop; /* We timed out or were interrupted. */ else if (error) return (NULL); /* If recursed, assume caller knows the rules. */ else if (BUF_LOCKRECURSED(bp)) goto end; /* * The buffer is locked. B_CACHE is cleared if the buffer is * invalid. Otherwise, for a non-VMIO buffer, B_CACHE is set * and for a VMIO buffer B_CACHE is adjusted according to the * backing VM cache. */ if (bp->b_flags & B_INVAL) bp->b_flags &= ~B_CACHE; else if ((bp->b_flags & (B_VMIO | B_INVAL)) == 0) bp->b_flags |= B_CACHE; if (bp->b_flags & B_MANAGED) MPASS(bp->b_qindex == QUEUE_NONE); else bremfree(bp); /* * check for size inconsistencies for non-VMIO case. */ if (bp->b_bcount != size) { if ((bp->b_flags & B_VMIO) == 0 || (size > bp->b_kvasize)) { if (bp->b_flags & B_DELWRI) { /* * If buffer is pinned and caller does * not want sleep waiting for it to be * unpinned, bail out * */ if (bp->b_pin_count > 0) { if (flags & GB_LOCK_NOWAIT) { bqrelse(bp); return (NULL); } else { bunpin_wait(bp); } } bp->b_flags |= B_NOCACHE; bwrite(bp); } else { if (LIST_EMPTY(&bp->b_dep)) { bp->b_flags |= B_RELBUF; brelse(bp); } else { bp->b_flags |= B_NOCACHE; bwrite(bp); } } goto loop; } } /* * Handle the case of unmapped buffer which should * become mapped, or the buffer for which KVA * reservation is requested. */ bp_unmapped_get_kva(bp, blkno, size, flags); /* * If the size is inconsistant in the VMIO case, we can resize * the buffer. This might lead to B_CACHE getting set or * cleared. If the size has not changed, B_CACHE remains * unchanged from its previous state. */ if (bp->b_bcount != size) allocbuf(bp, size); KASSERT(bp->b_offset != NOOFFSET, ("getblk: no buffer offset")); /* * A buffer with B_DELWRI set and B_CACHE clear must * be committed before we can return the buffer in * order to prevent the caller from issuing a read * ( due to B_CACHE not being set ) and overwriting * it. * * Most callers, including NFS and FFS, need this to * operate properly either because they assume they * can issue a read if B_CACHE is not set, or because * ( for example ) an uncached B_DELWRI might loop due * to softupdates re-dirtying the buffer. In the latter * case, B_CACHE is set after the first write completes, * preventing further loops. * NOTE! b*write() sets B_CACHE. If we cleared B_CACHE * above while extending the buffer, we cannot allow the * buffer to remain with B_CACHE set after the write * completes or it will represent a corrupt state. To * deal with this we set B_NOCACHE to scrap the buffer * after the write. * * We might be able to do something fancy, like setting * B_CACHE in bwrite() except if B_DELWRI is already set, * so the below call doesn't set B_CACHE, but that gets real * confusing. This is much easier. */ if ((bp->b_flags & (B_CACHE|B_DELWRI)) == B_DELWRI) { bp->b_flags |= B_NOCACHE; bwrite(bp); goto loop; } bp->b_flags &= ~B_DONE; } else { /* * Buffer is not in-core, create new buffer. The buffer * returned by getnewbuf() is locked. Note that the returned * buffer is also considered valid (not marked B_INVAL). */ BO_RUNLOCK(bo); /* * If the user does not want us to create the buffer, bail out * here. */ if (flags & GB_NOCREAT) return NULL; if (numfreebuffers == 0 && TD_IS_IDLETHREAD(curthread)) return NULL; bsize = vn_isdisk(vp, NULL) ? DEV_BSIZE : bo->bo_bsize; KASSERT(bsize != 0, ("bsize == 0, check bo->bo_bsize")); offset = blkno * bsize; vmio = vp->v_object != NULL; if (vmio) { maxsize = size + (offset & PAGE_MASK); } else { maxsize = size; /* Do not allow non-VMIO notmapped buffers. */ flags &= ~GB_UNMAPPED; } maxsize = imax(maxsize, bsize); bp = getnewbuf(vp, slpflag, slptimeo, size, maxsize, flags); if (bp == NULL) { if (slpflag || slptimeo) return NULL; goto loop; } /* * This code is used to make sure that a buffer is not * created while the getnewbuf routine is blocked. * This can be a problem whether the vnode is locked or not. * If the buffer is created out from under us, we have to * throw away the one we just created. * * Note: this must occur before we associate the buffer * with the vp especially considering limitations in * the splay tree implementation when dealing with duplicate * lblkno's. */ BO_LOCK(bo); if (gbincore(bo, blkno)) { BO_UNLOCK(bo); bp->b_flags |= B_INVAL; brelse(bp); goto loop; } /* * Insert the buffer into the hash, so that it can * be found by incore. */ bp->b_blkno = bp->b_lblkno = blkno; bp->b_offset = offset; bgetvp(vp, bp); BO_UNLOCK(bo); /* * set B_VMIO bit. allocbuf() the buffer bigger. Since the * buffer size starts out as 0, B_CACHE will be set by * allocbuf() for the VMIO case prior to it testing the * backing store for validity. */ if (vmio) { bp->b_flags |= B_VMIO; KASSERT(vp->v_object == bp->b_bufobj->bo_object, ("ARGH! different b_bufobj->bo_object %p %p %p\n", bp, vp->v_object, bp->b_bufobj->bo_object)); } else { bp->b_flags &= ~B_VMIO; KASSERT(bp->b_bufobj->bo_object == NULL, ("ARGH! has b_bufobj->bo_object %p %p\n", bp, bp->b_bufobj->bo_object)); BUF_CHECK_MAPPED(bp); } allocbuf(bp, size); bp->b_flags &= ~B_DONE; } CTR4(KTR_BUF, "getblk(%p, %ld, %d) = %p", vp, (long)blkno, size, bp); BUF_ASSERT_HELD(bp); end: KASSERT(bp->b_bufobj == bo, ("bp %p wrong b_bufobj %p should be %p", bp, bp->b_bufobj, bo)); return (bp); } /* * Get an empty, disassociated buffer of given size. The buffer is initially * set to B_INVAL. */ struct buf * geteblk(int size, int flags) { struct buf *bp; int maxsize; maxsize = (size + BKVAMASK) & ~BKVAMASK; while ((bp = getnewbuf(NULL, 0, 0, size, maxsize, flags)) == NULL) { if ((flags & GB_NOWAIT_BD) && (curthread->td_pflags & TDP_BUFNEED) != 0) return (NULL); } allocbuf(bp, size); bp->b_flags |= B_INVAL; /* b_dep cleared by getnewbuf() */ BUF_ASSERT_HELD(bp); return (bp); } /* * This code constitutes the buffer memory from either anonymous system * memory (in the case of non-VMIO operations) or from an associated * VM object (in the case of VMIO operations). This code is able to * resize a buffer up or down. * * Note that this code is tricky, and has many complications to resolve * deadlock or inconsistant data situations. Tread lightly!!! * There are B_CACHE and B_DELWRI interactions that must be dealt with by * the caller. Calling this code willy nilly can result in the loss of data. * * allocbuf() only adjusts B_CACHE for VMIO buffers. getblk() deals with * B_CACHE for the non-VMIO case. */ int allocbuf(struct buf *bp, int size) { int newbsize, mbsize; int i; BUF_ASSERT_HELD(bp); if (bp->b_kvasize < size) panic("allocbuf: buffer too small"); if ((bp->b_flags & B_VMIO) == 0) { caddr_t origbuf; int origbufsize; /* * Just get anonymous memory from the kernel. Don't * mess with B_CACHE. */ mbsize = (size + DEV_BSIZE - 1) & ~(DEV_BSIZE - 1); if (bp->b_flags & B_MALLOC) newbsize = mbsize; else newbsize = round_page(size); if (newbsize < bp->b_bufsize) { /* * malloced buffers are not shrunk */ if (bp->b_flags & B_MALLOC) { if (newbsize) { bp->b_bcount = size; } else { free(bp->b_data, M_BIOBUF); if (bp->b_bufsize) { atomic_subtract_long( &bufmallocspace, bp->b_bufsize); bufspacewakeup(); bp->b_bufsize = 0; } bp->b_saveaddr = bp->b_kvabase; bp->b_data = bp->b_saveaddr; bp->b_bcount = 0; bp->b_flags &= ~B_MALLOC; } return 1; } vm_hold_free_pages(bp, newbsize); } else if (newbsize > bp->b_bufsize) { /* * We only use malloced memory on the first allocation. * and revert to page-allocated memory when the buffer * grows. */ /* * There is a potential smp race here that could lead * to bufmallocspace slightly passing the max. It * is probably extremely rare and not worth worrying * over. */ if ( (bufmallocspace < maxbufmallocspace) && (bp->b_bufsize == 0) && (mbsize <= PAGE_SIZE/2)) { bp->b_data = malloc(mbsize, M_BIOBUF, M_WAITOK); bp->b_bufsize = mbsize; bp->b_bcount = size; bp->b_flags |= B_MALLOC; atomic_add_long(&bufmallocspace, mbsize); return 1; } origbuf = NULL; origbufsize = 0; /* * If the buffer is growing on its other-than-first allocation, * then we revert to the page-allocation scheme. */ if (bp->b_flags & B_MALLOC) { origbuf = bp->b_data; origbufsize = bp->b_bufsize; bp->b_data = bp->b_kvabase; if (bp->b_bufsize) { atomic_subtract_long(&bufmallocspace, bp->b_bufsize); bufspacewakeup(); bp->b_bufsize = 0; } bp->b_flags &= ~B_MALLOC; newbsize = round_page(newbsize); } vm_hold_load_pages( bp, (vm_offset_t) bp->b_data + bp->b_bufsize, (vm_offset_t) bp->b_data + newbsize); if (origbuf) { bcopy(origbuf, bp->b_data, origbufsize); free(origbuf, M_BIOBUF); } } } else { int desiredpages; newbsize = (size + DEV_BSIZE - 1) & ~(DEV_BSIZE - 1); desiredpages = (size == 0) ? 0 : num_pages((bp->b_offset & PAGE_MASK) + newbsize); if (bp->b_flags & B_MALLOC) panic("allocbuf: VMIO buffer can't be malloced"); /* * Set B_CACHE initially if buffer is 0 length or will become * 0-length. */ if (size == 0 || bp->b_bufsize == 0) bp->b_flags |= B_CACHE; if (newbsize < bp->b_bufsize) { /* * DEV_BSIZE aligned new buffer size is less then the * DEV_BSIZE aligned existing buffer size. Figure out * if we have to remove any pages. */ if (desiredpages < bp->b_npages) { vm_page_t m; if ((bp->b_flags & B_UNMAPPED) == 0) { BUF_CHECK_MAPPED(bp); pmap_qremove((vm_offset_t)trunc_page( (vm_offset_t)bp->b_data) + (desiredpages << PAGE_SHIFT), (bp->b_npages - desiredpages)); } else BUF_CHECK_UNMAPPED(bp); VM_OBJECT_WLOCK(bp->b_bufobj->bo_object); for (i = desiredpages; i < bp->b_npages; i++) { /* * the page is not freed here -- it * is the responsibility of * vnode_pager_setsize */ m = bp->b_pages[i]; KASSERT(m != bogus_page, ("allocbuf: bogus page found")); while (vm_page_sleep_if_busy(m, "biodep")) continue; bp->b_pages[i] = NULL; vm_page_lock(m); vm_page_unwire(m, PQ_INACTIVE); vm_page_unlock(m); } VM_OBJECT_WUNLOCK(bp->b_bufobj->bo_object); bp->b_npages = desiredpages; } } else if (size > bp->b_bcount) { /* * We are growing the buffer, possibly in a * byte-granular fashion. */ vm_object_t obj; vm_offset_t toff; vm_offset_t tinc; /* * Step 1, bring in the VM pages from the object, * allocating them if necessary. We must clear * B_CACHE if these pages are not valid for the * range covered by the buffer. */ obj = bp->b_bufobj->bo_object; VM_OBJECT_WLOCK(obj); while (bp->b_npages < desiredpages) { vm_page_t m; /* * We must allocate system pages since blocking * here could interfere with paging I/O, no * matter which process we are. * * Only exclusive busy can be tested here. * Blocking on shared busy might lead to * deadlocks once allocbuf() is called after * pages are vfs_busy_pages(). */ m = vm_page_grab(obj, OFF_TO_IDX(bp->b_offset) + bp->b_npages, VM_ALLOC_NOBUSY | VM_ALLOC_SYSTEM | VM_ALLOC_WIRED | VM_ALLOC_IGN_SBUSY | VM_ALLOC_COUNT(desiredpages - bp->b_npages)); if (m->valid == 0) bp->b_flags &= ~B_CACHE; bp->b_pages[bp->b_npages] = m; ++bp->b_npages; } /* * Step 2. We've loaded the pages into the buffer, * we have to figure out if we can still have B_CACHE * set. Note that B_CACHE is set according to the * byte-granular range ( bcount and size ), new the * aligned range ( newbsize ). * * The VM test is against m->valid, which is DEV_BSIZE * aligned. Needless to say, the validity of the data * needs to also be DEV_BSIZE aligned. Note that this * fails with NFS if the server or some other client * extends the file's EOF. If our buffer is resized, * B_CACHE may remain set! XXX */ toff = bp->b_bcount; tinc = PAGE_SIZE - ((bp->b_offset + toff) & PAGE_MASK); while ((bp->b_flags & B_CACHE) && toff < size) { vm_pindex_t pi; if (tinc > (size - toff)) tinc = size - toff; pi = ((bp->b_offset & PAGE_MASK) + toff) >> PAGE_SHIFT; vfs_buf_test_cache( bp, bp->b_offset, toff, tinc, bp->b_pages[pi] ); toff += tinc; tinc = PAGE_SIZE; } VM_OBJECT_WUNLOCK(obj); /* * Step 3, fixup the KVM pmap. */ if ((bp->b_flags & B_UNMAPPED) == 0) bpmap_qenter(bp); else BUF_CHECK_UNMAPPED(bp); } } if (newbsize < bp->b_bufsize) bufspacewakeup(); bp->b_bufsize = newbsize; /* actual buffer allocation */ bp->b_bcount = size; /* requested buffer size */ return 1; } extern int inflight_transient_maps; void biodone(struct bio *bp) { struct mtx *mtxp; void (*done)(struct bio *); vm_offset_t start, end; if ((bp->bio_flags & BIO_TRANSIENT_MAPPING) != 0) { bp->bio_flags &= ~BIO_TRANSIENT_MAPPING; bp->bio_flags |= BIO_UNMAPPED; start = trunc_page((vm_offset_t)bp->bio_data); end = round_page((vm_offset_t)bp->bio_data + bp->bio_length); bp->bio_data = unmapped_buf; pmap_qremove(start, OFF_TO_IDX(end - start)); vmem_free(transient_arena, start, end - start); atomic_add_int(&inflight_transient_maps, -1); } done = bp->bio_done; if (done == NULL) { mtxp = mtx_pool_find(mtxpool_sleep, bp); mtx_lock(mtxp); bp->bio_flags |= BIO_DONE; wakeup(bp); mtx_unlock(mtxp); } else { bp->bio_flags |= BIO_DONE; done(bp); } } /* * Wait for a BIO to finish. */ int biowait(struct bio *bp, const char *wchan) { struct mtx *mtxp; mtxp = mtx_pool_find(mtxpool_sleep, bp); mtx_lock(mtxp); while ((bp->bio_flags & BIO_DONE) == 0) msleep(bp, mtxp, PRIBIO, wchan, 0); mtx_unlock(mtxp); if (bp->bio_error != 0) return (bp->bio_error); if (!(bp->bio_flags & BIO_ERROR)) return (0); return (EIO); } void biofinish(struct bio *bp, struct devstat *stat, int error) { if (error) { bp->bio_error = error; bp->bio_flags |= BIO_ERROR; } if (stat != NULL) devstat_end_transaction_bio(stat, bp); biodone(bp); } /* * bufwait: * * Wait for buffer I/O completion, returning error status. The buffer * is left locked and B_DONE on return. B_EINTR is converted into an EINTR * error and cleared. */ int bufwait(struct buf *bp) { if (bp->b_iocmd == BIO_READ) bwait(bp, PRIBIO, "biord"); else bwait(bp, PRIBIO, "biowr"); if (bp->b_flags & B_EINTR) { bp->b_flags &= ~B_EINTR; return (EINTR); } if (bp->b_ioflags & BIO_ERROR) { return (bp->b_error ? bp->b_error : EIO); } else { return (0); } } /* * Call back function from struct bio back up to struct buf. */ static void bufdonebio(struct bio *bip) { struct buf *bp; bp = bip->bio_caller2; bp->b_resid = bip->bio_resid; bp->b_ioflags = bip->bio_flags; bp->b_error = bip->bio_error; if (bp->b_error) bp->b_ioflags |= BIO_ERROR; bufdone(bp); g_destroy_bio(bip); } void dev_strategy(struct cdev *dev, struct buf *bp) { struct cdevsw *csw; int ref; KASSERT(dev->si_refcount > 0, ("dev_strategy on un-referenced struct cdev *(%s) %p", devtoname(dev), dev)); csw = dev_refthread(dev, &ref); dev_strategy_csw(dev, csw, bp); dev_relthread(dev, ref); } void dev_strategy_csw(struct cdev *dev, struct cdevsw *csw, struct buf *bp) { struct bio *bip; KASSERT(bp->b_iocmd == BIO_READ || bp->b_iocmd == BIO_WRITE, ("b_iocmd botch")); KASSERT(((dev->si_flags & SI_ETERNAL) != 0 && csw != NULL) || dev->si_threadcount > 0, ("dev_strategy_csw threadcount cdev *(%s) %p", devtoname(dev), dev)); if (csw == NULL) { bp->b_error = ENXIO; bp->b_ioflags = BIO_ERROR; bufdone(bp); return; } for (;;) { bip = g_new_bio(); if (bip != NULL) break; /* Try again later */ tsleep(&bp, PRIBIO, "dev_strat", hz/10); } bip->bio_cmd = bp->b_iocmd; bip->bio_offset = bp->b_iooffset; bip->bio_length = bp->b_bcount; bip->bio_bcount = bp->b_bcount; /* XXX: remove */ bdata2bio(bp, bip); bip->bio_done = bufdonebio; bip->bio_caller2 = bp; bip->bio_dev = dev; (*csw->d_strategy)(bip); } /* * bufdone: * * Finish I/O on a buffer, optionally calling a completion function. * This is usually called from an interrupt so process blocking is * not allowed. * * biodone is also responsible for setting B_CACHE in a B_VMIO bp. * In a non-VMIO bp, B_CACHE will be set on the next getblk() * assuming B_INVAL is clear. * * For the VMIO case, we set B_CACHE if the op was a read and no * read error occured, or if the op was a write. B_CACHE is never * set if the buffer is invalid or otherwise uncacheable. * * biodone does not mess with B_INVAL, allowing the I/O routine or the * initiator to leave B_INVAL set to brelse the buffer out of existance * in the biodone routine. */ void bufdone(struct buf *bp) { struct bufobj *dropobj; void (*biodone)(struct buf *); CTR3(KTR_BUF, "bufdone(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); dropobj = NULL; KASSERT(!(bp->b_flags & B_DONE), ("biodone: bp %p already done", bp)); BUF_ASSERT_HELD(bp); runningbufwakeup(bp); if (bp->b_iocmd == BIO_WRITE) dropobj = bp->b_bufobj; /* call optional completion function if requested */ if (bp->b_iodone != NULL) { biodone = bp->b_iodone; bp->b_iodone = NULL; (*biodone) (bp); if (dropobj) bufobj_wdrop(dropobj); return; } bufdone_finish(bp); if (dropobj) bufobj_wdrop(dropobj); } void bufdone_finish(struct buf *bp) { BUF_ASSERT_HELD(bp); if (!LIST_EMPTY(&bp->b_dep)) buf_complete(bp); if (bp->b_flags & B_VMIO) { vm_ooffset_t foff; vm_page_t m; vm_object_t obj; struct vnode *vp; int bogus, i, iosize; obj = bp->b_bufobj->bo_object; KASSERT(obj->paging_in_progress >= bp->b_npages, ("biodone_finish: paging in progress(%d) < b_npages(%d)", obj->paging_in_progress, bp->b_npages)); vp = bp->b_vp; KASSERT(vp->v_holdcnt > 0, ("biodone_finish: vnode %p has zero hold count", vp)); KASSERT(vp->v_object != NULL, ("biodone_finish: vnode %p has no vm_object", vp)); foff = bp->b_offset; KASSERT(bp->b_offset != NOOFFSET, ("biodone_finish: bp %p has no buffer offset", bp)); /* * Set B_CACHE if the op was a normal read and no error * occured. B_CACHE is set for writes in the b*write() * routines. */ iosize = bp->b_bcount - bp->b_resid; if (bp->b_iocmd == BIO_READ && !(bp->b_flags & (B_INVAL|B_NOCACHE)) && !(bp->b_ioflags & BIO_ERROR)) { bp->b_flags |= B_CACHE; } bogus = 0; VM_OBJECT_WLOCK(obj); for (i = 0; i < bp->b_npages; i++) { int bogusflag = 0; int resid; resid = ((foff + PAGE_SIZE) & ~(off_t)PAGE_MASK) - foff; if (resid > iosize) resid = iosize; /* * cleanup bogus pages, restoring the originals */ m = bp->b_pages[i]; if (m == bogus_page) { bogus = bogusflag = 1; m = vm_page_lookup(obj, OFF_TO_IDX(foff)); if (m == NULL) panic("biodone: page disappeared!"); bp->b_pages[i] = m; } KASSERT(OFF_TO_IDX(foff) == m->pindex, ("biodone_finish: foff(%jd)/pindex(%ju) mismatch", (intmax_t)foff, (uintmax_t)m->pindex)); /* * In the write case, the valid and clean bits are * already changed correctly ( see bdwrite() ), so we * only need to do this here in the read case. */ if ((bp->b_iocmd == BIO_READ) && !bogusflag && resid > 0) { KASSERT((m->dirty & vm_page_bits(foff & PAGE_MASK, resid)) == 0, ("bufdone_finish:" " page %p has unexpected dirty bits", m)); vfs_page_set_valid(bp, foff, m); } vm_page_sunbusy(m); vm_object_pip_subtract(obj, 1); foff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK; iosize -= resid; } vm_object_pip_wakeupn(obj, 0); VM_OBJECT_WUNLOCK(obj); if (bogus && (bp->b_flags & B_UNMAPPED) == 0) { BUF_CHECK_MAPPED(bp); pmap_qenter(trunc_page((vm_offset_t)bp->b_data), bp->b_pages, bp->b_npages); } } /* * For asynchronous completions, release the buffer now. The brelse * will do a wakeup there if necessary - so no need to do a wakeup * here in the async case. The sync case always needs to do a wakeup. */ if (bp->b_flags & B_ASYNC) { if ((bp->b_flags & (B_NOCACHE | B_INVAL | B_RELBUF)) || (bp->b_ioflags & BIO_ERROR)) brelse(bp); else bqrelse(bp); } else bdone(bp); } /* * This routine is called in lieu of iodone in the case of * incomplete I/O. This keeps the busy status for pages * consistant. */ void vfs_unbusy_pages(struct buf *bp) { int i; vm_object_t obj; vm_page_t m; runningbufwakeup(bp); if (!(bp->b_flags & B_VMIO)) return; obj = bp->b_bufobj->bo_object; VM_OBJECT_WLOCK(obj); for (i = 0; i < bp->b_npages; i++) { m = bp->b_pages[i]; if (m == bogus_page) { m = vm_page_lookup(obj, OFF_TO_IDX(bp->b_offset) + i); if (!m) panic("vfs_unbusy_pages: page missing\n"); bp->b_pages[i] = m; if ((bp->b_flags & B_UNMAPPED) == 0) { BUF_CHECK_MAPPED(bp); pmap_qenter(trunc_page((vm_offset_t)bp->b_data), bp->b_pages, bp->b_npages); } else BUF_CHECK_UNMAPPED(bp); } vm_object_pip_subtract(obj, 1); vm_page_sunbusy(m); } vm_object_pip_wakeupn(obj, 0); VM_OBJECT_WUNLOCK(obj); } /* * vfs_page_set_valid: * * Set the valid bits in a page based on the supplied offset. The * range is restricted to the buffer's size. * * This routine is typically called after a read completes. */ static void vfs_page_set_valid(struct buf *bp, vm_ooffset_t off, vm_page_t m) { vm_ooffset_t eoff; /* * Compute the end offset, eoff, such that [off, eoff) does not span a * page boundary and eoff is not greater than the end of the buffer. * The end of the buffer, in this case, is our file EOF, not the * allocation size of the buffer. */ eoff = (off + PAGE_SIZE) & ~(vm_ooffset_t)PAGE_MASK; if (eoff > bp->b_offset + bp->b_bcount) eoff = bp->b_offset + bp->b_bcount; /* * Set valid range. This is typically the entire buffer and thus the * entire page. */ if (eoff > off) vm_page_set_valid_range(m, off & PAGE_MASK, eoff - off); } /* * vfs_page_set_validclean: * * Set the valid bits and clear the dirty bits in a page based on the * supplied offset. The range is restricted to the buffer's size. */ static void vfs_page_set_validclean(struct buf *bp, vm_ooffset_t off, vm_page_t m) { vm_ooffset_t soff, eoff; /* * Start and end offsets in buffer. eoff - soff may not cross a * page boundry or cross the end of the buffer. The end of the * buffer, in this case, is our file EOF, not the allocation size * of the buffer. */ soff = off; eoff = (off + PAGE_SIZE) & ~(off_t)PAGE_MASK; if (eoff > bp->b_offset + bp->b_bcount) eoff = bp->b_offset + bp->b_bcount; /* * Set valid range. This is typically the entire buffer and thus the * entire page. */ if (eoff > soff) { vm_page_set_validclean( m, (vm_offset_t) (soff & PAGE_MASK), (vm_offset_t) (eoff - soff) ); } } /* * Ensure that all buffer pages are not exclusive busied. If any page is * exclusive busy, drain it. */ void vfs_drain_busy_pages(struct buf *bp) { vm_page_t m; int i, last_busied; VM_OBJECT_ASSERT_WLOCKED(bp->b_bufobj->bo_object); last_busied = 0; for (i = 0; i < bp->b_npages; i++) { m = bp->b_pages[i]; if (vm_page_xbusied(m)) { for (; last_busied < i; last_busied++) vm_page_sbusy(bp->b_pages[last_busied]); while (vm_page_xbusied(m)) { vm_page_lock(m); VM_OBJECT_WUNLOCK(bp->b_bufobj->bo_object); vm_page_busy_sleep(m, "vbpage"); VM_OBJECT_WLOCK(bp->b_bufobj->bo_object); } } } for (i = 0; i < last_busied; i++) vm_page_sunbusy(bp->b_pages[i]); } /* * This routine is called before a device strategy routine. * It is used to tell the VM system that paging I/O is in * progress, and treat the pages associated with the buffer * almost as being exclusive busy. Also the object paging_in_progress * flag is handled to make sure that the object doesn't become * inconsistant. * * Since I/O has not been initiated yet, certain buffer flags * such as BIO_ERROR or B_INVAL may be in an inconsistant state * and should be ignored. */ void vfs_busy_pages(struct buf *bp, int clear_modify) { int i, bogus; vm_object_t obj; vm_ooffset_t foff; vm_page_t m; if (!(bp->b_flags & B_VMIO)) return; obj = bp->b_bufobj->bo_object; foff = bp->b_offset; KASSERT(bp->b_offset != NOOFFSET, ("vfs_busy_pages: no buffer offset")); VM_OBJECT_WLOCK(obj); vfs_drain_busy_pages(bp); if (bp->b_bufsize != 0) vfs_setdirty_locked_object(bp); bogus = 0; for (i = 0; i < bp->b_npages; i++) { m = bp->b_pages[i]; if ((bp->b_flags & B_CLUSTER) == 0) { vm_object_pip_add(obj, 1); vm_page_sbusy(m); } /* * When readying a buffer for a read ( i.e * clear_modify == 0 ), it is important to do * bogus_page replacement for valid pages in * partially instantiated buffers. Partially * instantiated buffers can, in turn, occur when * reconstituting a buffer from its VM backing store * base. We only have to do this if B_CACHE is * clear ( which causes the I/O to occur in the * first place ). The replacement prevents the read * I/O from overwriting potentially dirty VM-backed * pages. XXX bogus page replacement is, uh, bogus. * It may not work properly with small-block devices. * We need to find a better way. */ if (clear_modify) { pmap_remove_write(m); vfs_page_set_validclean(bp, foff, m); } else if (m->valid == VM_PAGE_BITS_ALL && (bp->b_flags & B_CACHE) == 0) { bp->b_pages[i] = bogus_page; bogus++; } foff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK; } VM_OBJECT_WUNLOCK(obj); if (bogus && (bp->b_flags & B_UNMAPPED) == 0) { BUF_CHECK_MAPPED(bp); pmap_qenter(trunc_page((vm_offset_t)bp->b_data), bp->b_pages, bp->b_npages); } } /* * vfs_bio_set_valid: * * Set the range within the buffer to valid. The range is * relative to the beginning of the buffer, b_offset. Note that * b_offset itself may be offset from the beginning of the first * page. */ void vfs_bio_set_valid(struct buf *bp, int base, int size) { int i, n; vm_page_t m; if (!(bp->b_flags & B_VMIO)) return; /* * Fixup base to be relative to beginning of first page. * Set initial n to be the maximum number of bytes in the * first page that can be validated. */ base += (bp->b_offset & PAGE_MASK); n = PAGE_SIZE - (base & PAGE_MASK); VM_OBJECT_WLOCK(bp->b_bufobj->bo_object); for (i = base / PAGE_SIZE; size > 0 && i < bp->b_npages; ++i) { m = bp->b_pages[i]; if (n > size) n = size; vm_page_set_valid_range(m, base & PAGE_MASK, n); base += n; size -= n; n = PAGE_SIZE; } VM_OBJECT_WUNLOCK(bp->b_bufobj->bo_object); } /* * vfs_bio_clrbuf: * * If the specified buffer is a non-VMIO buffer, clear the entire * buffer. If the specified buffer is a VMIO buffer, clear and * validate only the previously invalid portions of the buffer. * This routine essentially fakes an I/O, so we need to clear * BIO_ERROR and B_INVAL. * * Note that while we only theoretically need to clear through b_bcount, * we go ahead and clear through b_bufsize. */ void vfs_bio_clrbuf(struct buf *bp) { int i, j, mask, sa, ea, slide; if ((bp->b_flags & (B_VMIO | B_MALLOC)) != B_VMIO) { clrbuf(bp); return; } bp->b_flags &= ~B_INVAL; bp->b_ioflags &= ~BIO_ERROR; VM_OBJECT_WLOCK(bp->b_bufobj->bo_object); if ((bp->b_npages == 1) && (bp->b_bufsize < PAGE_SIZE) && (bp->b_offset & PAGE_MASK) == 0) { if (bp->b_pages[0] == bogus_page) goto unlock; mask = (1 << (bp->b_bufsize / DEV_BSIZE)) - 1; VM_OBJECT_ASSERT_WLOCKED(bp->b_pages[0]->object); if ((bp->b_pages[0]->valid & mask) == mask) goto unlock; if ((bp->b_pages[0]->valid & mask) == 0) { pmap_zero_page_area(bp->b_pages[0], 0, bp->b_bufsize); bp->b_pages[0]->valid |= mask; goto unlock; } } sa = bp->b_offset & PAGE_MASK; slide = 0; for (i = 0; i < bp->b_npages; i++, sa = 0) { slide = imin(slide + PAGE_SIZE, bp->b_offset + bp->b_bufsize); ea = slide & PAGE_MASK; if (ea == 0) ea = PAGE_SIZE; if (bp->b_pages[i] == bogus_page) continue; j = sa / DEV_BSIZE; mask = ((1 << ((ea - sa) / DEV_BSIZE)) - 1) << j; VM_OBJECT_ASSERT_WLOCKED(bp->b_pages[i]->object); if ((bp->b_pages[i]->valid & mask) == mask) continue; if ((bp->b_pages[i]->valid & mask) == 0) pmap_zero_page_area(bp->b_pages[i], sa, ea - sa); else { for (; sa < ea; sa += DEV_BSIZE, j++) { if ((bp->b_pages[i]->valid & (1 << j)) == 0) { pmap_zero_page_area(bp->b_pages[i], sa, DEV_BSIZE); } } } bp->b_pages[i]->valid |= mask; } unlock: VM_OBJECT_WUNLOCK(bp->b_bufobj->bo_object); bp->b_resid = 0; } void vfs_bio_bzero_buf(struct buf *bp, int base, int size) { vm_page_t m; int i, n; if ((bp->b_flags & B_UNMAPPED) == 0) { BUF_CHECK_MAPPED(bp); bzero(bp->b_data + base, size); } else { BUF_CHECK_UNMAPPED(bp); n = PAGE_SIZE - (base & PAGE_MASK); for (i = base / PAGE_SIZE; size > 0 && i < bp->b_npages; ++i) { m = bp->b_pages[i]; if (n > size) n = size; pmap_zero_page_area(m, base & PAGE_MASK, n); base += n; size -= n; n = PAGE_SIZE; } } } /* * vm_hold_load_pages and vm_hold_free_pages get pages into * a buffers address space. The pages are anonymous and are * not associated with a file object. */ static void vm_hold_load_pages(struct buf *bp, vm_offset_t from, vm_offset_t to) { vm_offset_t pg; vm_page_t p; int index; BUF_CHECK_MAPPED(bp); to = round_page(to); from = round_page(from); index = (from - trunc_page((vm_offset_t)bp->b_data)) >> PAGE_SHIFT; for (pg = from; pg < to; pg += PAGE_SIZE, index++) { tryagain: /* * note: must allocate system pages since blocking here * could interfere with paging I/O, no matter which * process we are. */ p = vm_page_alloc(NULL, 0, VM_ALLOC_SYSTEM | VM_ALLOC_NOOBJ | VM_ALLOC_WIRED | VM_ALLOC_COUNT((to - pg) >> PAGE_SHIFT)); if (p == NULL) { VM_WAIT; goto tryagain; } pmap_qenter(pg, &p, 1); bp->b_pages[index] = p; } bp->b_npages = index; } /* Return pages associated with this buf to the vm system */ static void vm_hold_free_pages(struct buf *bp, int newbsize) { vm_offset_t from; vm_page_t p; int index, newnpages; BUF_CHECK_MAPPED(bp); from = round_page((vm_offset_t)bp->b_data + newbsize); newnpages = (from - trunc_page((vm_offset_t)bp->b_data)) >> PAGE_SHIFT; if (bp->b_npages > newnpages) pmap_qremove(from, bp->b_npages - newnpages); for (index = newnpages; index < bp->b_npages; index++) { p = bp->b_pages[index]; bp->b_pages[index] = NULL; if (vm_page_sbusied(p)) printf("vm_hold_free_pages: blkno: %jd, lblkno: %jd\n", (intmax_t)bp->b_blkno, (intmax_t)bp->b_lblkno); p->wire_count--; vm_page_free(p); atomic_subtract_int(&vm_cnt.v_wire_count, 1); } bp->b_npages = newnpages; } /* * Map an IO request into kernel virtual address space. * * All requests are (re)mapped into kernel VA space. * Notice that we use b_bufsize for the size of the buffer * to be mapped. b_bcount might be modified by the driver. * * Note that even if the caller determines that the address space should * be valid, a race or a smaller-file mapped into a larger space may * actually cause vmapbuf() to fail, so all callers of vmapbuf() MUST * check the return value. */ int vmapbuf(struct buf *bp, int mapbuf) { caddr_t kva; vm_prot_t prot; int pidx; if (bp->b_bufsize < 0) return (-1); prot = VM_PROT_READ; if (bp->b_iocmd == BIO_READ) prot |= VM_PROT_WRITE; /* Less backwards than it looks */ if ((pidx = vm_fault_quick_hold_pages(&curproc->p_vmspace->vm_map, (vm_offset_t)bp->b_data, bp->b_bufsize, prot, bp->b_pages, btoc(MAXPHYS))) < 0) return (-1); bp->b_npages = pidx; if (mapbuf || !unmapped_buf_allowed) { pmap_qenter((vm_offset_t)bp->b_saveaddr, bp->b_pages, pidx); kva = bp->b_saveaddr; bp->b_saveaddr = bp->b_data; bp->b_data = kva + (((vm_offset_t)bp->b_data) & PAGE_MASK); bp->b_flags &= ~B_UNMAPPED; } else { bp->b_flags |= B_UNMAPPED; bp->b_offset = ((vm_offset_t)bp->b_data) & PAGE_MASK; bp->b_saveaddr = bp->b_data; bp->b_data = unmapped_buf; } return(0); } /* * Free the io map PTEs associated with this IO operation. * We also invalidate the TLB entries and restore the original b_addr. */ void vunmapbuf(struct buf *bp) { int npages; npages = bp->b_npages; if (bp->b_flags & B_UNMAPPED) bp->b_flags &= ~B_UNMAPPED; else pmap_qremove(trunc_page((vm_offset_t)bp->b_data), npages); vm_page_unhold_pages(bp->b_pages, npages); bp->b_data = bp->b_saveaddr; } void bdone(struct buf *bp) { struct mtx *mtxp; mtxp = mtx_pool_find(mtxpool_sleep, bp); mtx_lock(mtxp); bp->b_flags |= B_DONE; wakeup(bp); mtx_unlock(mtxp); } void bwait(struct buf *bp, u_char pri, const char *wchan) { struct mtx *mtxp; mtxp = mtx_pool_find(mtxpool_sleep, bp); mtx_lock(mtxp); while ((bp->b_flags & B_DONE) == 0) msleep(bp, mtxp, pri, wchan, 0); mtx_unlock(mtxp); } int bufsync(struct bufobj *bo, int waitfor) { return (VOP_FSYNC(bo->__bo_vnode, waitfor, curthread)); } void bufstrategy(struct bufobj *bo, struct buf *bp) { int i = 0; struct vnode *vp; vp = bp->b_vp; KASSERT(vp == bo->bo_private, ("Inconsistent vnode bufstrategy")); KASSERT(vp->v_type != VCHR && vp->v_type != VBLK, ("Wrong vnode in bufstrategy(bp=%p, vp=%p)", bp, vp)); i = VOP_STRATEGY(vp, bp); KASSERT(i == 0, ("VOP_STRATEGY failed bp=%p vp=%p", bp, bp->b_vp)); } void bufobj_wrefl(struct bufobj *bo) { KASSERT(bo != NULL, ("NULL bo in bufobj_wref")); ASSERT_BO_WLOCKED(bo); bo->bo_numoutput++; } void bufobj_wref(struct bufobj *bo) { KASSERT(bo != NULL, ("NULL bo in bufobj_wref")); BO_LOCK(bo); bo->bo_numoutput++; BO_UNLOCK(bo); } void bufobj_wdrop(struct bufobj *bo) { KASSERT(bo != NULL, ("NULL bo in bufobj_wdrop")); BO_LOCK(bo); KASSERT(bo->bo_numoutput > 0, ("bufobj_wdrop non-positive count")); if ((--bo->bo_numoutput == 0) && (bo->bo_flag & BO_WWAIT)) { bo->bo_flag &= ~BO_WWAIT; wakeup(&bo->bo_numoutput); } BO_UNLOCK(bo); } int bufobj_wwait(struct bufobj *bo, int slpflag, int timeo) { int error; KASSERT(bo != NULL, ("NULL bo in bufobj_wwait")); ASSERT_BO_WLOCKED(bo); error = 0; while (bo->bo_numoutput) { bo->bo_flag |= BO_WWAIT; error = msleep(&bo->bo_numoutput, BO_LOCKPTR(bo), slpflag | (PRIBIO + 1), "bo_wwait", timeo); if (error) break; } return (error); } void bpin(struct buf *bp) { struct mtx *mtxp; mtxp = mtx_pool_find(mtxpool_sleep, bp); mtx_lock(mtxp); bp->b_pin_count++; mtx_unlock(mtxp); } void bunpin(struct buf *bp) { struct mtx *mtxp; mtxp = mtx_pool_find(mtxpool_sleep, bp); mtx_lock(mtxp); if (--bp->b_pin_count == 0) wakeup(bp); mtx_unlock(mtxp); } void bunpin_wait(struct buf *bp) { struct mtx *mtxp; mtxp = mtx_pool_find(mtxpool_sleep, bp); mtx_lock(mtxp); while (bp->b_pin_count > 0) msleep(bp, mtxp, PRIBIO, "bwunpin", 0); mtx_unlock(mtxp); } /* * Set bio_data or bio_ma for struct bio from the struct buf. */ void bdata2bio(struct buf *bp, struct bio *bip) { if ((bp->b_flags & B_UNMAPPED) != 0) { KASSERT(unmapped_buf_allowed, ("unmapped")); bip->bio_ma = bp->b_pages; bip->bio_ma_n = bp->b_npages; bip->bio_data = unmapped_buf; bip->bio_ma_offset = (vm_offset_t)bp->b_offset & PAGE_MASK; bip->bio_flags |= BIO_UNMAPPED; KASSERT(round_page(bip->bio_ma_offset + bip->bio_length) / PAGE_SIZE == bp->b_npages, ("Buffer %p too short: %d %lld %d", bp, bip->bio_ma_offset, (long long)bip->bio_length, bip->bio_ma_n)); } else { bip->bio_data = bp->b_data; bip->bio_ma = NULL; } } #include "opt_ddb.h" #ifdef DDB #include /* DDB command to show buffer data */ DB_SHOW_COMMAND(buffer, db_show_buffer) { /* get args */ struct buf *bp = (struct buf *)addr; if (!have_addr) { db_printf("usage: show buffer \n"); return; } db_printf("buf at %p\n", bp); db_printf("b_flags = 0x%b, b_xflags=0x%b, b_vflags=0x%b\n", (u_int)bp->b_flags, PRINT_BUF_FLAGS, (u_int)bp->b_xflags, PRINT_BUF_XFLAGS, (u_int)bp->b_vflags, PRINT_BUF_VFLAGS); db_printf( "b_error = %d, b_bufsize = %ld, b_bcount = %ld, b_resid = %ld\n" "b_bufobj = (%p), b_data = %p, b_blkno = %jd, b_lblkno = %jd, " "b_dep = %p\n", bp->b_error, bp->b_bufsize, bp->b_bcount, bp->b_resid, bp->b_bufobj, bp->b_data, (intmax_t)bp->b_blkno, (intmax_t)bp->b_lblkno, bp->b_dep.lh_first); if (bp->b_npages) { int i; db_printf("b_npages = %d, pages(OBJ, IDX, PA): ", bp->b_npages); for (i = 0; i < bp->b_npages; i++) { vm_page_t m; m = bp->b_pages[i]; db_printf("(%p, 0x%lx, 0x%lx)", (void *)m->object, (u_long)m->pindex, (u_long)VM_PAGE_TO_PHYS(m)); if ((i + 1) < bp->b_npages) db_printf(","); } db_printf("\n"); } db_printf(" "); BUF_LOCKPRINTINFO(bp); } DB_SHOW_COMMAND(lockedbufs, lockedbufs) { struct buf *bp; int i; for (i = 0; i < nbuf; i++) { bp = &buf[i]; if (BUF_ISLOCKED(bp)) { db_show_buffer((uintptr_t)bp, 1, 0, NULL); db_printf("\n"); } } } DB_SHOW_COMMAND(vnodebufs, db_show_vnodebufs) { struct vnode *vp; struct buf *bp; if (!have_addr) { db_printf("usage: show vnodebufs \n"); return; } vp = (struct vnode *)addr; db_printf("Clean buffers:\n"); TAILQ_FOREACH(bp, &vp->v_bufobj.bo_clean.bv_hd, b_bobufs) { db_show_buffer((uintptr_t)bp, 1, 0, NULL); db_printf("\n"); } db_printf("Dirty buffers:\n"); TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) { db_show_buffer((uintptr_t)bp, 1, 0, NULL); db_printf("\n"); } } DB_COMMAND(countfreebufs, db_coundfreebufs) { struct buf *bp; int i, used = 0, nfree = 0; if (have_addr) { db_printf("usage: countfreebufs\n"); return; } for (i = 0; i < nbuf; i++) { bp = &buf[i]; if ((bp->b_flags & B_INFREECNT) != 0) nfree++; else used++; } db_printf("Counted %d free, %d used (%d tot)\n", nfree, used, nfree + used); db_printf("numfreebuffers is %d\n", numfreebuffers); } #endif /* DDB */ Index: user/ngie/more-tests/sys/modules/oce/Makefile =================================================================== --- user/ngie/more-tests/sys/modules/oce/Makefile (revision 281974) +++ user/ngie/more-tests/sys/modules/oce/Makefile (revision 281975) @@ -1,15 +1,15 @@ # # $FreeBSD$ # .PATH: ${.CURDIR}/../../dev/oce -KMOD = oce +KMOD = if_oce SRCS = oce_if.c oce_hw.c oce_mbox.c oce_util.c oce_queue.c oce_sysctl.c SRCS += bus_if.h device_if.h pci_if.h opt_inet.h opt_inet6.h CFLAGS+= -I${.CURDIR}/../../dev/oce -DSMP # uncomment for lock profiling statistics #CFLAGS += -DLOCK_PROFILING .include Index: user/ngie/more-tests/sys/netinet/tcp_subr.c =================================================================== --- user/ngie/more-tests/sys/netinet/tcp_subr.c (revision 281974) +++ user/ngie/more-tests/sys/netinet/tcp_subr.c (revision 281975) @@ -1,2518 +1,2519 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #endif #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPDEBUG #include #endif #ifdef INET6 #include #endif #ifdef TCP_OFFLOAD #include #endif #ifdef IPSEC #include #include #ifdef INET6 #include #endif #include #include #endif /*IPSEC*/ #include #include #include VNET_DEFINE(int, tcp_mssdflt) = TCP_MSS; #ifdef INET6 VNET_DEFINE(int, tcp_v6mssdflt) = TCP6_MSS; #endif static int sysctl_net_inet_tcp_mss_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_mssdflt = new; } return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(tcp_mssdflt), 0, &sysctl_net_inet_tcp_mss_check, "I", "Default TCP Maximum Segment Size"); #ifdef INET6 static int sysctl_net_inet_tcp_mss_v6_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_v6mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_v6mssdflt = new; } return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(tcp_v6mssdflt), 0, &sysctl_net_inet_tcp_mss_v6_check, "I", "Default TCP Maximum Segment Size for IPv6"); #endif /* INET6 */ /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ VNET_DEFINE(int, tcp_minmss) = TCP_MINMSS; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_minmss), 0, "Minimum TCP Maximum Segment Size"); VNET_DEFINE(int, tcp_do_rfc1323) = 1; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_do_rfc1323), 0, "Enable rfc1323 (high performance TCP) extensions"); static int tcp_log_debug = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_debug, CTLFLAG_RW, &tcp_log_debug, 0, "Log errors caused by incoming TCP segments"); static int tcp_tcbhashsize; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_UINT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(tcbinfo.ipi_count), 0, "Number of active PCBs"); static VNET_DEFINE(int, icmp_may_rst) = 1; #define V_icmp_may_rst VNET(icmp_may_rst) SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp_may_rst), 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static VNET_DEFINE(int, tcp_isn_reseed_interval) = 0; #define V_tcp_isn_reseed_interval VNET(tcp_isn_reseed_interval) SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_isn_reseed_interval), 0, "Seconds between reseeding of ISN secret"); static int tcp_soreceive_stream; SYSCTL_INT(_net_inet_tcp, OID_AUTO, soreceive_stream, CTLFLAG_RDTUN, &tcp_soreceive_stream, 0, "Using soreceive_stream for TCP sockets"); #ifdef TCP_SIGNATURE static int tcp_sig_checksigs = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, signature_verify_input, CTLFLAG_RW, &tcp_sig_checksigs, 0, "Verify RFC2385 digests on inbound traffic"); #endif VNET_DEFINE(uma_zone_t, sack_hole_zone); #define V_sack_hole_zone VNET(sack_hole_zone) VNET_DEFINE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST+1]); static struct inpcb *tcp_notify(struct inpcb *, int); static struct inpcb *tcp_mtudisc_notify(struct inpcb *, int); static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr); static void tcp_timer_discard(struct tcpcb *, uint32_t); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 0 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct tcp_timer tt; struct cc_var ccv; struct osd osd; }; static VNET_DEFINE(uma_zone_t, tcpcb_zone); #define V_tcpcb_zone VNET(tcpcb_zone) MALLOC_DEFINE(M_TCPLOG, "tcplog", "TCP address and flags print buffers"); static struct mtx isn_mtx; #define ISN_LOCK_INIT() mtx_init(&isn_mtx, "isn_mtx", NULL, MTX_DEF) #define ISN_LOCK() mtx_lock(&isn_mtx) #define ISN_UNLOCK() mtx_unlock(&isn_mtx) /* * TCP initialization. */ static void tcp_zone_change(void *tag) { uma_zone_set_max(V_tcbinfo.ipi_zone, maxsockets); uma_zone_set_max(V_tcpcb_zone, maxsockets); tcp_tw_zone_change(); } static int tcp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "tcpinp"); return (0); } /* * Take a value and get the next power of 2 that doesn't overflow. * Used to size the tcp_inpcb hash buckets. */ static int maketcp_hashsize(int size) { int hashsize; /* * auto tune. * get the next power of 2 higher than maxsockets. */ hashsize = 1 << fls(size); /* catch overflow, and just go one power of 2 smaller */ if (hashsize < size) { hashsize = 1 << (fls(size) - 1); } return (hashsize); } void tcp_init(void) { const char *tcbhash_tuneable; int hashsize; tcbhash_tuneable = "net.inet.tcp.tcbhashsize"; if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN, &V_tcp_hhh[HHOOK_TCP_EST_IN], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT, &V_tcp_hhh[HHOOK_TCP_EST_OUT], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); hashsize = TCBHASHSIZE; TUNABLE_INT_FETCH(tcbhash_tuneable, &hashsize); if (hashsize == 0) { /* * Auto tune the hash size based on maxsockets. * A perfect hash would have a 1:1 mapping * (hashsize = maxsockets) however it's been * suggested that O(2) average is better. */ hashsize = maketcp_hashsize(maxsockets / 4); /* * Our historical default is 512, * do not autotune lower than this. */ if (hashsize < 512) hashsize = 512; if (bootverbose) printf("%s: %s auto tuned to %d\n", __func__, tcbhash_tuneable, hashsize); } /* * We require a hashsize to be a power of two. * Previously if it was not a power of two we would just reset it * back to 512, which could be a nasty surprise if you did not notice * the error message. * Instead what we do is clip it to the closest power of two lower * than the specified hash value. */ if (!powerof2(hashsize)) { int oldhashsize = hashsize; hashsize = maketcp_hashsize(hashsize); /* prevent absurdly low value */ if (hashsize < 16) hashsize = 16; printf("%s: WARNING: TCB hash size not a power of 2, " "clipped from %d to %d.\n", __func__, oldhashsize, hashsize); } in_pcbinfo_init(&V_tcbinfo, "tcp", &V_tcb, hashsize, hashsize, "tcp_inpcb", tcp_inpcb_init, NULL, UMA_ZONE_NOFREE, IPI_HASHFIELDS_4TUPLE); /* * These have to be type stable for the benefit of the timers. */ V_tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(V_tcpcb_zone, maxsockets); uma_zone_set_warning(V_tcpcb_zone, "kern.ipc.maxsockets limit reached"); tcp_tw_init(); syncache_init(); tcp_hc_init(); TUNABLE_INT_FETCH("net.inet.tcp.sack.enable", &V_tcp_do_sack); V_sack_hole_zone = uma_zcreate("sackhole", sizeof(struct sackhole), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); /* Skip initialization of globals for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; /* XXX virtualize those bellow? */ tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; if (tcp_rexmit_min < 1) tcp_rexmit_min = 1; tcp_rexmit_slop = TCPTV_CPU_VAR; tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT; tcp_tcbhashsize = hashsize; if (tcp_soreceive_stream) { #ifdef INET tcp_usrreqs.pru_soreceive = soreceive_stream; #endif #ifdef INET6 tcp6_usrreqs.pru_soreceive = soreceive_stream; #endif /* INET6 */ } #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR ISN_LOCK_INIT(); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); EVENTHANDLER_REGISTER(maxsockets_change, tcp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } #ifdef VIMAGE void tcp_destroy(void) { int error; tcp_hc_destroy(); syncache_destroy(); tcp_tw_destroy(); in_pcbinfo_destroy(&V_tcbinfo); uma_zdestroy(V_sack_hole_zone); uma_zdestroy(V_tcpcb_zone); error = hhook_head_deregister(V_tcp_hhh[HHOOK_TCP_EST_IN]); if (error != 0) { printf("%s: WARNING: unable to deregister helper hook " "type=%d, id=%d: error %d returned\n", __func__, HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN, error); } error = hhook_head_deregister(V_tcp_hhh[HHOOK_TCP_EST_OUT]); if (error != 0) { printf("%s: WARNING: unable to deregister helper hook " "type=%d, id=%d: error %d returned\n", __func__, HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT, error); } } #endif void tcp_fini(void *xtp) { } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(struct inpcb *inp, void *ip_ptr, void *tcp_ptr) { struct tcphdr *th = (struct tcphdr *)tcp_ptr; INP_WLOCK_ASSERT(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->inp_flow & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons(sizeof(struct tcphdr)); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } #endif /* INET */ th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(struct inpcb *inp) { struct tcptemp *t; t = malloc(sizeof(*t), M_TEMP, M_NOWAIT); if (t == NULL) return (NULL); tcpip_fillheaders(inp, (void *)&t->tt_ipgen, (void *)&t->tt_t); return (t); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at th and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the segment th, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then th must point to *inside* the mbuf. */ void tcp_respond(struct tcpcb *tp, void *ipgen, struct tcphdr *th, struct mbuf *m, tcp_seq ack, tcp_seq seq, int flags) { int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == (IPV6_VERSION >> 4); ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_WLOCK_ASSERT(inp); } else inp = NULL; if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { /* * reuse the mbuf. * XXX MRT We inherrit the FIB, which is lucky. */ m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, uint32_t); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, uint16_t); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); ip6->ip6_plen = htons(tlen - sizeof(*ip6)); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { tlen += sizeof (struct tcpiphdr); ip->ip_len = htons(tlen); ip->ip_ttl = V_ip_defttl; if (V_path_mtu_discovery) ip->ip_off |= htons(IP_DF); } #endif m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ INP_WLOCK_ASSERT(inp); mac_inpcb_create_mbuf(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_netinet_tcp_reply(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); #ifdef INET6 if (isipv6) { m->m_pkthdr.csum_flags = CSUM_TCP_IPV6; nth->th_sum = in6_cksum_pseudo(ip6, tlen - sizeof(struct ip6_hdr), IPPROTO_TCP, 0); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { m->m_pkthdr.csum_flags = CSUM_TCP; nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); } #endif /* INET */ #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif if (flags & TH_RST) TCP_PROBE5(accept__refused, NULL, NULL, mtod(m, const char *), tp, nth); TCP_PROBE5(send, NULL, tp, mtod(m, const char *), tp, nth); #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); #endif /* INET6 */ #if defined(INET) && defined(INET6) else #endif #ifdef INET (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); #endif } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(struct inpcb *inp) { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(V_tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* Initialise cc_var struct for this tcpcb. */ tp->ccv = &tm->ccv; tp->ccv->type = IPPROTO_TCP; tp->ccv->ccvc.tcp = tp; /* * Use the current system default CC algorithm. */ CC_LIST_RLOCK(); KASSERT(!STAILQ_EMPTY(&cc_list), ("cc_list is empty!")); CC_ALGO(tp) = CC_DEFAULT(); CC_LIST_RUNLOCK(); if (CC_ALGO(tp)->cb_init != NULL) if (CC_ALGO(tp)->cb_init(tp->ccv) > 0) { uma_zfree(V_tcpcb_zone, tm); return (NULL); } tp->osd = &tm->osd; if (khelp_init_osd(HELPER_CLASS_TCP, tp->osd)) { uma_zfree(V_tcpcb_zone, tm); return (NULL); } #ifdef VIMAGE tp->t_vnet = inp->inp_vnet; #endif tp->t_timers = &tm->tt; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? V_tcp_v6mssdflt : #endif /* INET6 */ V_tcp_mssdflt; /* Set up our timeouts. */ callout_init(&tp->t_timers->tt_rexmt, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_persist, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_keep, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_2msl, CALLOUT_MPSAFE); callout_init(&tp->t_timers->tt_delack, CALLOUT_MPSAFE); if (V_tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); if (V_tcp_do_sack) tp->t_flags |= TF_SACK_PERMIT; TAILQ_INIT(&tp->snd_holes); /* * The tcpcb will hold a reference on its inpcb until tcp_discardcb() * is called. */ in_pcbref(inp); /* Reference for tcpcb */ tp->t_inpcb = inp; /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = V_ip_defttl; inp->inp_ppcb = tp; return (tp); /* XXX */ } /* * Switch the congestion control algorithm back to NewReno for any active * control blocks using an algorithm which is about to go away. * This ensures the CC framework can allow the unload to proceed without leaving * any dangling pointers which would trigger a panic. * Returning non-zero would inform the CC framework that something went wrong * and it would be unsafe to allow the unload to proceed. However, there is no * way for this to occur with this implementation so we always return zero. */ int tcp_ccalgounload(struct cc_algo *unload_algo) { struct cc_algo *tmpalgo; struct inpcb *inp; struct tcpcb *tp; VNET_ITERATOR_DECL(vnet_iter); /* * Check all active control blocks across all network stacks and change * any that are using "unload_algo" back to NewReno. If "unload_algo" * requires cleanup code to be run, call it. */ VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); INP_INFO_RLOCK(&V_tcbinfo); /* * New connections already part way through being initialised * with the CC algo we're removing will not race with this code * because the INP_INFO_WLOCK is held during initialisation. We * therefore don't enter the loop below until the connection * list has stabilised. */ LIST_FOREACH(inp, &V_tcb, inp_list) { INP_WLOCK(inp); /* Important to skip tcptw structs. */ if (!(inp->inp_flags & INP_TIMEWAIT) && (tp = intotcpcb(inp)) != NULL) { /* * By holding INP_WLOCK here, we are assured * that the connection is not currently * executing inside the CC module's functions * i.e. it is safe to make the switch back to * NewReno. */ if (CC_ALGO(tp) == unload_algo) { tmpalgo = CC_ALGO(tp); /* NewReno does not require any init. */ CC_ALGO(tp) = &newreno_cc_algo; if (tmpalgo->cb_destroy != NULL) tmpalgo->cb_destroy(tp->ccv); } } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); return (0); } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(struct tcpcb *tp, int errno) { struct socket *so = tp->t_inpcb->inp_socket; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); if (TCPS_HAVERCVDSYN(tp->t_state)) { tcp_state_change(tp, TCPS_CLOSED); (void) tcp_output(tp); TCPSTAT_INC(tcps_drops); } else TCPSTAT_INC(tcps_conndrops); if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } void tcp_discardcb(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ int released; INP_WLOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we delete the * PCB. * * If stopping a timer fails, we schedule a discard function in same * callout, and the last discard function called will take care of * deleting the tcpcb. */ tcp_timer_stop(tp, TT_REXMT); tcp_timer_stop(tp, TT_PERSIST); tcp_timer_stop(tp, TT_KEEP); tcp_timer_stop(tp, TT_2MSL); tcp_timer_stop(tp, TT_DELACK); /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. * * XXXRW: 'so' may be NULL here, and/or socket buffer may be * being torn down. Ideally this code would not use 'so'. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ tcp_reass_flush(tp); #ifdef TCP_OFFLOAD /* Disconnect offload device, if any. */ if (tp->t_flags & TF_TOE) tcp_offload_detach(tp); #endif tcp_free_sackholes(tp); /* Allow the CC algorithm to clean up after itself. */ if (CC_ALGO(tp)->cb_destroy != NULL) CC_ALGO(tp)->cb_destroy(tp->ccv); khelp_destroy_osd(tp->osd); CC_ALGO(tp) = NULL; inp->inp_ppcb = NULL; if ((tp->t_timers->tt_flags & TT_MASK) == 0) { /* We own the last reference on tcpcb, let's free it. */ tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); released = in_pcbrele_wlocked(inp); KASSERT(!released, ("%s: inp %p should not have been released " "here", __func__, inp)); } } void tcp_timer_2msl_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_2MSL); } void tcp_timer_keep_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_KEEP); } void tcp_timer_persist_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_PERSIST); } void tcp_timer_rexmt_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_REXMT); } void tcp_timer_delack_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_DELACK); } void tcp_timer_discard(struct tcpcb *tp, uint32_t timer_type) { struct inpcb *inp; CURVNET_SET(tp->t_vnet); INP_INFO_WLOCK(&V_tcbinfo); inp = tp->t_inpcb; KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp)); INP_WLOCK(inp); KASSERT((tp->t_timers->tt_flags & TT_STOPPED) != 0, ("%s: tcpcb has to be stopped here", __func__)); KASSERT((tp->t_timers->tt_flags & timer_type) != 0, ("%s: discard callout should be running", __func__)); tp->t_timers->tt_flags &= ~timer_type; if ((tp->t_timers->tt_flags & TT_MASK) == 0) { /* We own the last reference on this tcpcb, let's free it. */ tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); if (in_pcbrele_wlocked(inp)) { INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } /* * Attempt to close a TCP control block, marking it as dropped, and freeing * the socket if we hold the only reference. */ struct tcpcb * tcp_close(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); #ifdef TCP_OFFLOAD if (tp->t_state == TCPS_LISTEN) tcp_offload_listen_stop(tp); #endif in_pcbdrop(inp); TCPSTAT_INC(tcps_closed); KASSERT(inp->inp_socket != NULL, ("tcp_close: inp_socket NULL")); so = inp->inp_socket; soisdisconnected(so); if (inp->inp_flags & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_close: !SS_PROTOREF")); inp->inp_flags &= ~INP_SOCKREF; INP_WUNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); return (NULL); } return (tp); } void tcp_drain(void) { VNET_ITERATOR_DECL(vnet_iter); if (!do_tcpdrain) return; VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); struct inpcb *inpb; struct tcpcb *tcpb; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * useful. */ INP_INFO_RLOCK(&V_tcbinfo); LIST_FOREACH(inpb, V_tcbinfo.ipi_listhead, inp_list) { if (inpb->inp_flags & INP_TIMEWAIT) continue; INP_WLOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { tcp_reass_flush(tcpb); tcp_clean_sackreport(tcpb); } INP_WUNLOCK(inpb); } INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(struct inpcb *inp, int error) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_notify: tp == NULL")); /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return (inp); } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tp = tcp_drop(tp, error); if (tp != NULL) return (inp); else return (NULL); } else { tp->t_softerror = error; return (inp); } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, m, n, pcb_count; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = V_tcbinfo.ipi_count + syncache_pcbcount(); n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xtcpcb); return (0); } if (req->newptr != NULL) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_tcbinfo); gencnt = V_tcbinfo.ipi_gencnt; n = V_tcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_tcbinfo); m = syncache_pcbcount(); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + (n + m) * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n + m; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); error = syncache_pcblist(req, m, &pcb_count); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return (ENOMEM); INP_INFO_RLOCK(&V_tcbinfo); for (inp = LIST_FIRST(V_tcbinfo.ipi_listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_flags & INP_TIMEWAIT) { if (intotw(inp) != NULL) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = EINVAL; /* Skip this inp. */ } else error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) { in_pcbref(inp); inp_list[i++] = inp; } } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; void *inp_ppcb; bzero(&xt, sizeof(xt)); xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_flags & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else { bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (xt.xt_tp.t_timers) tcp_timer_to_xtimer(&xt.xt_tp, xt.xt_tp.t_timers, &xt.xt_timer); } if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xt, sizeof xt); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_tcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (!in_pcbrele_rlocked(inp)) INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_tcbinfo); if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&V_tcbinfo); xig.xig_gen = V_tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_tcbinfo.ipi_count + pcb_count; INP_INFO_RUNLOCK(&V_tcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); #ifdef INET static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); inp = in_pcblookup(&V_tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); } else error = ENOENT; if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #endif /* INET */ #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error; #ifdef INET int mapped = 0; #endif error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 || (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) { return (error); } if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { #ifdef INET if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else #endif return (EINVAL); } #ifdef INET if (mapped == 1) inp = in_pcblookup(&V_tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL); else #endif inp = in6_pcblookup(&V_tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); } else error = ENOENT; if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif /* INET6 */ #ifdef INET void tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct icmp *icp; struct in_conninfo inc; tcp_seq icmp_tcp_seq; int mtu; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc_notify; else if (V_icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { icp = (struct icmp *)((caddr_t)ip - offsetof(struct icmp, icmp_ip)); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_WLOCK(&V_tcbinfo); inp = in_pcblookup(&V_tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, INPLOOKUP_WLOCKPCB, NULL); if (inp != NULL) { if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = htonl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) && SEQ_LT(icmp_tcp_seq, tp->snd_max)) { if (cmd == PRC_MSGSIZE) { /* * MTU discovery: * If we got a needfrag set the MTU * in the route to the suggested new * value (if given) and then notify. */ bzero(&inc, sizeof(inc)); inc.inc_faddr = faddr; inc.inc_fibnum = inp->inp_inc.inc_fibnum; mtu = ntohs(icp->icmp_nextmtu); /* * If no alternative MTU was * proposed, try the next smaller * one. */ if (!mtu) mtu = ip_next_mtu( ntohs(ip->ip_len), 1); if (mtu < V_tcp_minmss + sizeof(struct tcpiphdr)) mtu = V_tcp_minmss + sizeof(struct tcpiphdr); /* * Only cache the MTU if it * is smaller than the interface * or route MTU. tcp_mtudisc() * will do right thing by itself. */ if (mtu <= tcp_maxmtu(&inc, NULL)) tcp_hc_updatemtu(&inc, mtu); tcp_mtudisc(inp, mtu); } else inp = (*notify)(inp, inetctlerrmap[cmd]); } } if (inp != NULL) INP_WUNLOCK(inp); } else { bzero(&inc, sizeof(inc)); inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; syncache_unreach(&inc, th); } INP_INFO_WUNLOCK(&V_tcbinfo); } else in_pcbnotifyall(&V_tcbinfo, faddr, inetctlerrmap[cmd], notify); } #endif /* INET */ #ifdef INET6 void tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc_notify; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&V_tcbinfo, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); bzero(&inc, sizeof(inc)); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_flags |= INC_ISIPV6; INP_INFO_WLOCK(&V_tcbinfo); syncache_unreach(&inc, &th); INP_INFO_WUNLOCK(&V_tcbinfo); } else in6_pcbnotify(&V_tcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * * Locking of the global variables isn_secret, isn_last_reseed, isn_offset, * isn_offset_old, and isn_ctx is performed using the TCP pcbinfo lock. In * general, this means holding an exclusive (write) lock. */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) static VNET_DEFINE(u_char, isn_secret[32]); static VNET_DEFINE(int, isn_last); static VNET_DEFINE(int, isn_last_reseed); static VNET_DEFINE(u_int32_t, isn_offset); static VNET_DEFINE(u_int32_t, isn_offset_old); #define V_isn_secret VNET(isn_secret) #define V_isn_last VNET(isn_last) #define V_isn_last_reseed VNET(isn_last_reseed) #define V_isn_offset VNET(isn_offset) #define V_isn_offset_old VNET(isn_offset_old) tcp_seq tcp_new_isn(struct tcpcb *tp) { MD5_CTX isn_ctx; u_int32_t md5_buffer[4]; tcp_seq new_isn; u_int32_t projected_offset; INP_WLOCK_ASSERT(tp->t_inpcb); ISN_LOCK(); /* Seed if this is the first use, reseed if requested. */ if ((V_isn_last_reseed == 0) || ((V_tcp_isn_reseed_interval > 0) && (((u_int)V_isn_last_reseed + (u_int)V_tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&V_isn_secret, sizeof(V_isn_secret)); V_isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &V_isn_secret, sizeof(V_isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; V_isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); if (ticks != V_isn_last) { projected_offset = V_isn_offset_old + ISN_BYTES_PER_SECOND / hz * (ticks - V_isn_last); if (SEQ_GT(projected_offset, V_isn_offset)) V_isn_offset = projected_offset; V_isn_offset_old = V_isn_offset; V_isn_last = ticks; } new_isn += V_isn_offset; ISN_UNLOCK(); return (new_isn); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(struct inpcb *inp, int errno) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); if (tp->t_state != TCPS_SYN_SENT) return (inp); tp = tcp_drop(tp, errno); if (tp != NULL) return (inp); else return (NULL); } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value. Also nudge TCP to send something, since we * know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ static struct inpcb * tcp_mtudisc_notify(struct inpcb *inp, int error) { return (tcp_mtudisc(inp, -1)); } struct inpcb * tcp_mtudisc(struct inpcb *inp, int mtuoffer) { struct tcpcb *tp; struct socket *so; INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_mtudisc: tp == NULL")); tcp_mss_update(tp, -1, mtuoffer, NULL, NULL); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_snd); /* If the mss is larger than the socket buffer, decrease the mss. */ if (so->so_snd.sb_hiwat < tp->t_maxseg) tp->t_maxseg = so->so_snd.sb_hiwat; SOCKBUF_UNLOCK(&so->so_snd); TCPSTAT_INC(tcps_mturesent); tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_free_sackholes(tp); tp->snd_recover = tp->snd_max; if (tp->t_flags & TF_SACK_PERMIT) EXIT_FASTRECOVERY(tp->t_flags); tcp_output(tp); return (inp); } #ifdef INET /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return 0. This routine * is called by TCP routines that access the rmx structure and by * tcp_mss_update to get the peer/interface MTU. */ u_long tcp_maxmtu(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; in_rtalloc_ign(&sro, 0, inc->inc_fibnum); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_mtu, ifp->if_mtu); /* Report additional interface capabilities. */ if (cap != NULL) { if (ifp->if_capenable & IFCAP_TSO4 && ifp->if_hwassist & CSUM_TSO) { cap->ifcap |= CSUM_TSO; cap->tsomax = ifp->if_hw_tsomax; cap->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; cap->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; } } RTFREE(sro.ro_rt); } return (maxmtu); } #endif /* INET */ #ifdef INET6 u_long tcp_maxmtu6(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; in6_rtalloc_ign(&sro6, 0, inc->inc_fibnum); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); /* Report additional interface capabilities. */ if (cap != NULL) { if (ifp->if_capenable & IFCAP_TSO6 && ifp->if_hwassist & CSUM_TSO) { cap->ifcap |= CSUM_TSO; cap->tsomax = ifp->if_hw_tsomax; cap->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; cap->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; } } RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(struct tcpcb *tp) { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) return (0); m = m_gethdr(M_NOWAIT, MT_DATA); if (!m) return (0); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return (hdrsiz); } #endif /* IPSEC */ #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 */ struct secasvar * tcp_get_sav(struct mbuf *m, u_int direction) { union sockaddr_union dst; struct secasvar *sav; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; char ip6buf[INET6_ADDRSTRLEN]; #endif /* Extract the destination from the IP header in the mbuf. */ bzero(&dst, sizeof(union sockaddr_union)); ip = mtod(m, struct ip *); #ifdef INET6 ip6 = NULL; /* Make the compiler happy. */ #endif switch (ip->ip_v) { #ifdef INET case IPVERSION: dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; break; #endif #ifdef INET6 case (IPV6_VERSION >> 4): ip6 = mtod(m, struct ip6_hdr *); dst.sa.sa_len = sizeof(struct sockaddr_in6); dst.sa.sa_family = AF_INET6; dst.sin6.sin6_addr = (direction == IPSEC_DIR_INBOUND) ? ip6->ip6_src : ip6->ip6_dst; break; #endif default: return (NULL); /* NOTREACHED */ break; } /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { ipseclog((LOG_ERR, "%s: SADB lookup failed for %s\n", __func__, (ip->ip_v == IPVERSION) ? inet_ntoa(dst.sin.sin_addr) : #ifdef INET6 (ip->ip_v == (IPV6_VERSION >> 4)) ? ip6_sprintf(ip6buf, &dst.sin6.sin6_addr) : #endif "(unsupported)")); } return (sav); } /* * Compute TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * sav pointer to security assosiation * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * Releases reference to SADB key before return. * * Return 0 if successful, otherwise return -1. * */ int tcp_signature_do_compute(struct mbuf *m, int len, int optlen, u_char *buf, struct secasvar *sav) { #ifdef INET struct ippseudo ippseudo; #endif MD5_CTX ctx; int doff; struct ip *ip; #ifdef INET struct ipovly *ipovly; #endif struct tcphdr *th; #ifdef INET6 struct ip6_hdr *ip6; struct in6_addr in6; uint32_t plen; uint16_t nhdr; #endif u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ ip = mtod(m, struct ip *); #ifdef INET6 ip6 = NULL; /* Make the compiler happy. */ #endif MD5Init(&ctx); /* * Step 1: Update MD5 hash with IP(v6) pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ switch (ip->ip_v) { #ifdef INET case IPVERSION: ipovly = (struct ipovly *)ip; ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); th = (struct tcphdr *)((u_char *)ip + sizeof(struct ip)); doff = sizeof(struct ip) + sizeof(struct tcphdr) + optlen; break; #endif #ifdef INET6 /* * RFC 2385, 2.0 Proposal * For IPv6, the pseudo-header is as described in RFC 2460, namely the * 128-bit source IPv6 address, 128-bit destination IPv6 address, zero- * extended next header value (to form 32 bits), and 32-bit segment * length. * Note: Upper-Layer Packet Length comes before Next Header. */ case (IPV6_VERSION >> 4): in6 = ip6->ip6_src; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); in6 = ip6->ip6_dst; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); plen = htonl(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&plen, sizeof(uint32_t)); nhdr = 0; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); nhdr = IPPROTO_TCP; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); th = (struct tcphdr *)((u_char *)ip6 + sizeof(struct ip6_hdr)); doff = sizeof(struct ip6_hdr) + sizeof(struct tcphdr) + optlen; break; #endif default: + KEY_FREESAV(&sav); return (-1); /* NOTREACHED */ break; } /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, sav->key_auth->key_data, _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } /* * Compute TCP-MD5 hash of a TCP segment. (RFC2385) * * Return 0 if successful, otherwise return -1. */ int tcp_signature_compute(struct mbuf *m, int _unused, int len, int optlen, u_char *buf, u_int direction) { struct secasvar *sav; if ((sav = tcp_get_sav(m, direction)) == NULL) return (-1); return (tcp_signature_do_compute(m, len, optlen, buf, sav)); } /* * Verify the TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * Return 1 if successful, otherwise return 0. */ int tcp_signature_verify(struct mbuf *m, int off0, int tlen, int optlen, struct tcpopt *to, struct tcphdr *th, u_int tcpbflag) { char tmpdigest[TCP_SIGLEN]; if (tcp_sig_checksigs == 0) return (1); if ((tcpbflag & TF_SIGNATURE) == 0) { if ((to->to_flags & TOF_SIGNATURE) != 0) { /* * If this socket is not expecting signature but * the segment contains signature just fail. */ TCPSTAT_INC(tcps_sig_err_sigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } /* Signature is not expected, and not present in segment. */ return (1); } /* * If this socket is expecting signature but the segment does not * contain any just fail. */ if ((to->to_flags & TOF_SIGNATURE) == 0) { TCPSTAT_INC(tcps_sig_err_nosigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (tcp_signature_compute(m, off0, tlen, optlen, &tmpdigest[0], IPSEC_DIR_INBOUND) == -1) { TCPSTAT_INC(tcps_sig_err_buildsig); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (bcmp(to->to_signature, &tmpdigest[0], TCP_SIGLEN) != 0) { TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } TCPSTAT_INC(tcps_sig_rcvgoodsig); return (1); } #endif /* TCP_SIGNATURE */ static int sysctl_drop(SYSCTL_HANDLER_ARGS) { /* addrs[0] is a foreign socket, addrs[1] is a local one. */ struct sockaddr_storage addrs[2]; struct inpcb *inp; struct tcpcb *tp; struct tcptw *tw; struct sockaddr_in *fin, *lin; #ifdef INET6 struct sockaddr_in6 *fin6, *lin6; #endif int error; inp = NULL; fin = lin = NULL; #ifdef INET6 fin6 = lin6 = NULL; #endif error = 0; if (req->oldptr != NULL || req->oldlen != 0) return (EINVAL); if (req->newptr == NULL) return (EPERM); if (req->newlen < sizeof(addrs)) return (ENOMEM); error = SYSCTL_IN(req, &addrs, sizeof(addrs)); if (error) return (error); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: fin6 = (struct sockaddr_in6 *)&addrs[0]; lin6 = (struct sockaddr_in6 *)&addrs[1]; if (fin6->sin6_len != sizeof(struct sockaddr_in6) || lin6->sin6_len != sizeof(struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_V4MAPPED(&fin6->sin6_addr)) { if (!IN6_IS_ADDR_V4MAPPED(&lin6->sin6_addr)) return (EINVAL); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[0]); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[1]); fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; break; } error = sa6_embedscope(fin6, V_ip6_use_defzone); if (error) return (error); error = sa6_embedscope(lin6, V_ip6_use_defzone); if (error) return (error); break; #endif #ifdef INET case AF_INET: fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; if (fin->sin_len != sizeof(struct sockaddr_in) || lin->sin_len != sizeof(struct sockaddr_in)) return (EINVAL); break; #endif default: return (EINVAL); } INP_INFO_WLOCK(&V_tcbinfo); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: inp = in6_pcblookup(&V_tcbinfo, &fin6->sin6_addr, fin6->sin6_port, &lin6->sin6_addr, lin6->sin6_port, INPLOOKUP_WLOCKPCB, NULL); break; #endif #ifdef INET case AF_INET: inp = in_pcblookup(&V_tcbinfo, fin->sin_addr, fin->sin_port, lin->sin_addr, lin->sin_port, INPLOOKUP_WLOCKPCB, NULL); break; #endif } if (inp != NULL) { if (inp->inp_flags & INP_TIMEWAIT) { /* * XXXRW: There currently exists a state where an * inpcb is present, but its timewait state has been * discarded. For now, don't allow dropping of this * type of inpcb. */ tw = intotw(inp); if (tw != NULL) tcp_twclose(tw, 0); else INP_WUNLOCK(inp); } else if (!(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket->so_options & SO_ACCEPTCONN)) { tp = intotcpcb(inp); tp = tcp_drop(tp, ECONNABORTED); if (tp != NULL) INP_WUNLOCK(inp); } else INP_WUNLOCK(inp); } else error = ESRCH; INP_INFO_WUNLOCK(&V_tcbinfo); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_DROP, drop, CTLFLAG_VNET | CTLTYPE_STRUCT | CTLFLAG_WR | CTLFLAG_SKIP, NULL, 0, sysctl_drop, "", "Drop TCP connection"); /* * Generate a standardized TCP log line for use throughout the * tcp subsystem. Memory allocation is done with M_NOWAIT to * allow use in the interrupt context. * * NB: The caller MUST free(s, M_TCPLOG) the returned string. * NB: The function may return NULL if memory allocation failed. * * Due to header inclusion and ordering limitations the struct ip * and ip6_hdr pointers have to be passed as void pointers. */ char * tcp_log_vain(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_in_vain == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } char * tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_debug == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { char *s, *sp; size_t size; struct ip *ip; #ifdef INET6 const struct ip6_hdr *ip6; ip6 = (const struct ip6_hdr *)ip6hdr; #endif /* INET6 */ ip = (struct ip *)ip4hdr; /* * The log line looks like this: * "TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags 0x2" */ size = sizeof("TCP: []:12345 to []:12345 tcpflags 0x2<>") + sizeof(PRINT_TH_FLAGS) + 1 + #ifdef INET6 2 * INET6_ADDRSTRLEN; #else 2 * INET_ADDRSTRLEN; #endif /* INET6 */ s = malloc(size, M_TCPLOG, M_ZERO|M_NOWAIT); if (s == NULL) return (NULL); strcat(s, "TCP: ["); sp = s + strlen(s); if (inc && ((inc->inc_flags & INC_ISIPV6) == 0)) { inet_ntoa_r(inc->inc_faddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); inet_ntoa_r(inc->inc_laddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); #ifdef INET6 } else if (inc) { ip6_sprintf(sp, &inc->inc6_faddr); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); ip6_sprintf(sp, &inc->inc6_laddr); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); } else if (ip6 && th) { ip6_sprintf(sp, &ip6->ip6_src); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); ip6_sprintf(sp, &ip6->ip6_dst); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET6 */ #ifdef INET } else if (ip && th) { inet_ntoa_r(ip->ip_src, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); inet_ntoa_r(ip->ip_dst, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET */ } else { free(s, M_TCPLOG); return (NULL); } sp = s + strlen(s); if (th) sprintf(sp, " tcpflags 0x%b", th->th_flags, PRINT_TH_FLAGS); if (*(s + size - 1) != '\0') panic("%s: string too long", __func__); return (s); } /* * A subroutine which makes it easy to track TCP state changes with DTrace. * This function shouldn't be called for t_state initializations that don't * correspond to actual TCP state transitions. */ void tcp_state_change(struct tcpcb *tp, int newstate) { #if defined(KDTRACE_HOOKS) int pstate = tp->t_state; #endif tp->t_state = newstate; TCP_PROBE6(state__change, NULL, tp, NULL, tp, NULL, pstate); } Index: user/ngie/more-tests/sys/sparc64/pci/apb.c =================================================================== --- user/ngie/more-tests/sys/sparc64/pci/apb.c (revision 281974) +++ user/ngie/more-tests/sys/sparc64/pci/apb.c (revision 281975) @@ -1,312 +1,306 @@ /*- * Copyright (c) 1994,1995 Stefan Esser, Wolfgang StanglMeier * Copyright (c) 2000 Michael Smith * Copyright (c) 2000 BSDi * Copyright (c) 2001, 2003 Thomas Moestl * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: FreeBSD: src/sys/dev/pci/pci_pci.c,v 1.3 2000/12/13 */ #include __FBSDID("$FreeBSD$"); /* * Support for the Sun APB (Advanced PCI Bridge) PCI-PCI bridge. * This bridge does not fully comply to the PCI bridge specification, and is * therefore not supported by the generic driver. * We can use some of the pcib methods anyway. */ #include "opt_ofw_pci.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "pcib_if.h" #include #include /* * Bridge-specific data. */ struct apb_softc { struct ofw_pcib_gen_softc sc_bsc; uint8_t sc_iomap; uint8_t sc_memmap; }; static device_probe_t apb_probe; static device_attach_t apb_attach; static bus_alloc_resource_t apb_alloc_resource; static bus_adjust_resource_t apb_adjust_resource; static device_method_t apb_methods[] = { /* Device interface */ DEVMETHOD(device_probe, apb_probe), DEVMETHOD(device_attach, apb_attach), /* Bus interface */ DEVMETHOD(bus_alloc_resource, apb_alloc_resource), DEVMETHOD(bus_adjust_resource, apb_adjust_resource), DEVMETHOD(bus_release_resource, bus_generic_release_resource), /* pcib interface */ DEVMETHOD(pcib_route_interrupt, ofw_pcib_gen_route_interrupt), /* ofw_bus interface */ DEVMETHOD(ofw_bus_get_node, ofw_pcib_gen_get_node), DEVMETHOD_END }; static devclass_t pcib_devclass; DEFINE_CLASS_1(pcib, apb_driver, apb_methods, sizeof(struct apb_softc), pcib_driver); EARLY_DRIVER_MODULE(apb, pci, apb_driver, pcib_devclass, 0, 0, BUS_PASS_BUS); MODULE_DEPEND(apb, pci, 1, 1, 1); /* APB specific registers */ #define APBR_IOMAP 0xde #define APBR_MEMMAP 0xdf /* Definitions for the mapping registers */ #define APB_IO_SCALE 0x200000 #define APB_MEM_SCALE 0x20000000 /* * Generic device interface */ static int apb_probe(device_t dev) { if (pci_get_vendor(dev) == 0x108e && /* Sun */ pci_get_device(dev) == 0x5000) { /* APB */ device_set_desc(dev, "APB PCI-PCI bridge"); return (0); } return (ENXIO); } static void apb_map_print(uint8_t map, u_long scale) { int i, first; for (first = 1, i = 0; i < 8; i++) { if ((map & (1 << i)) != 0) { printf("%s0x%lx-0x%lx", first ? "" : ", ", i * scale, (i + 1) * scale - 1); first = 0; } } } static int apb_checkrange(uint8_t map, u_long scale, u_long start, u_long end) { int i, ei; i = start / scale; ei = end / scale; if (i > 7 || ei > 7) return (0); for (; i <= ei; i++) if ((map & (1 << i)) == 0) return (0); return (1); } static int apb_attach(device_t dev) { struct apb_softc *sc; struct sysctl_ctx_list *sctx; struct sysctl_oid *soid; sc = device_get_softc(dev); /* * Get current bridge configuration. */ sc->sc_bsc.ops_pcib_sc.domain = pci_get_domain(dev); - sc->sc_bsc.ops_pcib_sc.secstat = - pci_read_config(dev, PCIR_SECSTAT_1, 2); - sc->sc_bsc.ops_pcib_sc.command = - pci_read_config(dev, PCIR_COMMAND, 2); - sc->sc_bsc.ops_pcib_sc.pribus = - pci_read_config(dev, PCIR_PRIBUS_1, 1); + sc->sc_bsc.ops_pcib_sc.pribus = pci_get_bus(dev); + pci_write_config(dev, PCIR_PRIBUS_1, sc->sc_bsc.ops_pcib_sc.pribus, 1); sc->sc_bsc.ops_pcib_sc.bus.sec = pci_read_config(dev, PCIR_SECBUS_1, 1); sc->sc_bsc.ops_pcib_sc.bus.sub = pci_read_config(dev, PCIR_SUBBUS_1, 1); sc->sc_bsc.ops_pcib_sc.bridgectl = pci_read_config(dev, PCIR_BRIDGECTL_1, 2); - sc->sc_bsc.ops_pcib_sc.seclat = - pci_read_config(dev, PCIR_SECLAT_1, 1); sc->sc_iomap = pci_read_config(dev, APBR_IOMAP, 1); sc->sc_memmap = pci_read_config(dev, APBR_MEMMAP, 1); /* * Setup SYSCTL reporting nodes. */ sctx = device_get_sysctl_ctx(dev); soid = device_get_sysctl_tree(dev); SYSCTL_ADD_UINT(sctx, SYSCTL_CHILDREN(soid), OID_AUTO, "domain", CTLFLAG_RD, &sc->sc_bsc.ops_pcib_sc.domain, 0, "Domain number"); SYSCTL_ADD_UINT(sctx, SYSCTL_CHILDREN(soid), OID_AUTO, "pribus", CTLFLAG_RD, &sc->sc_bsc.ops_pcib_sc.pribus, 0, "Primary bus number"); SYSCTL_ADD_UINT(sctx, SYSCTL_CHILDREN(soid), OID_AUTO, "secbus", CTLFLAG_RD, &sc->sc_bsc.ops_pcib_sc.bus.sec, 0, "Secondary bus number"); SYSCTL_ADD_UINT(sctx, SYSCTL_CHILDREN(soid), OID_AUTO, "subbus", CTLFLAG_RD, &sc->sc_bsc.ops_pcib_sc.bus.sub, 0, "Subordinate bus number"); ofw_pcib_gen_setup(dev); if (bootverbose) { device_printf(dev, " domain %d\n", sc->sc_bsc.ops_pcib_sc.domain); device_printf(dev, " secondary bus %d\n", sc->sc_bsc.ops_pcib_sc.bus.sec); device_printf(dev, " subordinate bus %d\n", sc->sc_bsc.ops_pcib_sc.bus.sub); device_printf(dev, " I/O decode "); apb_map_print(sc->sc_iomap, APB_IO_SCALE); printf("\n"); device_printf(dev, " memory decode "); apb_map_print(sc->sc_memmap, APB_MEM_SCALE); printf("\n"); } device_add_child(dev, "pci", -1); return (bus_generic_attach(dev)); } /* * We have to trap resource allocation requests and ensure that the bridge * is set up to, or capable of handling them. */ static struct resource * apb_alloc_resource(device_t dev, device_t child, int type, int *rid, u_long start, u_long end, u_long count, u_int flags) { struct apb_softc *sc; sc = device_get_softc(dev); /* * If this is a "default" allocation against this rid, we can't work * out where it's coming from (we should actually never see these) so * we just have to punt. */ if (start == 0 && end == ~0) { device_printf(dev, "can't decode default resource id %d for " "%s, bypassing\n", *rid, device_get_nameunit(child)); goto passup; } /* * Fail the allocation for this range if it's not supported. * XXX we should probably just fix up the bridge decode and * soldier on. */ switch (type) { case SYS_RES_IOPORT: if (!apb_checkrange(sc->sc_iomap, APB_IO_SCALE, start, end)) { device_printf(dev, "device %s requested unsupported " "I/O range 0x%lx-0x%lx\n", device_get_nameunit(child), start, end); return (NULL); } if (bootverbose) device_printf(sc->sc_bsc.ops_pcib_sc.dev, "device " "%s requested decoded I/O range 0x%lx-0x%lx\n", device_get_nameunit(child), start, end); break; case SYS_RES_MEMORY: if (!apb_checkrange(sc->sc_memmap, APB_MEM_SCALE, start, end)) { device_printf(dev, "device %s requested unsupported " "memory range 0x%lx-0x%lx\n", device_get_nameunit(child), start, end); return (NULL); } if (bootverbose) device_printf(sc->sc_bsc.ops_pcib_sc.dev, "device " "%s requested decoded memory range 0x%lx-0x%lx\n", device_get_nameunit(child), start, end); break; } passup: /* * Bridge is OK decoding this resource, so pass it up. */ return (bus_generic_alloc_resource(dev, child, type, rid, start, end, count, flags)); } static int apb_adjust_resource(device_t dev, device_t child, int type, struct resource *r, u_long start, u_long end) { struct apb_softc *sc; sc = device_get_softc(dev); switch (type) { case SYS_RES_IOPORT: if (!apb_checkrange(sc->sc_iomap, APB_IO_SCALE, start, end)) return (ENXIO); break; case SYS_RES_MEMORY: if (!apb_checkrange(sc->sc_memmap, APB_MEM_SCALE, start, end)) return (ENXIO); break; } return (bus_generic_adjust_resource(dev, child, type, r, start, end)); } Index: user/ngie/more-tests/sys/sys/cdefs.h =================================================================== --- user/ngie/more-tests/sys/sys/cdefs.h (revision 281974) +++ user/ngie/more-tests/sys/sys/cdefs.h (revision 281975) @@ -1,829 +1,835 @@ /*- * Copyright (c) 1991, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Berkeley Software Design, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)cdefs.h 8.8 (Berkeley) 1/9/95 * $FreeBSD$ */ #ifndef _SYS_CDEFS_H_ #define _SYS_CDEFS_H_ /* * Testing against Clang-specific extensions. */ #ifndef __has_attribute #define __has_attribute(x) 0 #endif #ifndef __has_extension #define __has_extension __has_feature #endif #ifndef __has_feature #define __has_feature(x) 0 #endif #ifndef __has_include #define __has_include(x) 0 #endif #ifndef __has_builtin #define __has_builtin(x) 0 #endif #if defined(__cplusplus) #define __BEGIN_DECLS extern "C" { #define __END_DECLS } #else #define __BEGIN_DECLS #define __END_DECLS #endif /* * This code has been put in place to help reduce the addition of * compiler specific defines in FreeBSD code. It helps to aid in * having a compiler-agnostic source tree. */ #if defined(__GNUC__) || defined(__INTEL_COMPILER) #if __GNUC__ >= 3 || defined(__INTEL_COMPILER) #define __GNUCLIKE_ASM 3 #define __GNUCLIKE_MATH_BUILTIN_CONSTANTS #else #define __GNUCLIKE_ASM 2 #endif #define __GNUCLIKE___TYPEOF 1 #define __GNUCLIKE___OFFSETOF 1 #define __GNUCLIKE___SECTION 1 #ifndef __INTEL_COMPILER #define __GNUCLIKE_CTOR_SECTION_HANDLING 1 #endif #define __GNUCLIKE_BUILTIN_CONSTANT_P 1 #if defined(__INTEL_COMPILER) && defined(__cplusplus) && \ __INTEL_COMPILER < 800 #undef __GNUCLIKE_BUILTIN_CONSTANT_P #endif #if (__GNUC_MINOR__ > 95 || __GNUC__ >= 3) && !defined(__INTEL_COMPILER) #define __GNUCLIKE_BUILTIN_VARARGS 1 #define __GNUCLIKE_BUILTIN_STDARG 1 #define __GNUCLIKE_BUILTIN_VAALIST 1 #endif #if defined(__GNUC__) #define __GNUC_VA_LIST_COMPATIBILITY 1 #endif /* * Compiler memory barriers, specific to gcc and clang. */ #if defined(__GNUC__) #define __compiler_membar() __asm __volatile(" " : : : "memory") #endif #ifndef __INTEL_COMPILER #define __GNUCLIKE_BUILTIN_NEXT_ARG 1 #define __GNUCLIKE_MATH_BUILTIN_RELOPS #endif #define __GNUCLIKE_BUILTIN_MEMCPY 1 /* XXX: if __GNUC__ >= 2: not tested everywhere originally, where replaced */ #define __CC_SUPPORTS_INLINE 1 #define __CC_SUPPORTS___INLINE 1 #define __CC_SUPPORTS___INLINE__ 1 #define __CC_SUPPORTS___FUNC__ 1 #define __CC_SUPPORTS_WARNING 1 #define __CC_SUPPORTS_VARADIC_XXX 1 /* see varargs.h */ #define __CC_SUPPORTS_DYNAMIC_ARRAY_INIT 1 #endif /* __GNUC__ || __INTEL_COMPILER */ /* * Macro to test if we're using a specific version of gcc or later. */ #if defined(__GNUC__) && !defined(__INTEL_COMPILER) #define __GNUC_PREREQ__(ma, mi) \ (__GNUC__ > (ma) || __GNUC__ == (ma) && __GNUC_MINOR__ >= (mi)) #else #define __GNUC_PREREQ__(ma, mi) 0 #endif /* * The __CONCAT macro is used to concatenate parts of symbol names, e.g. * with "#define OLD(foo) __CONCAT(old,foo)", OLD(foo) produces oldfoo. * The __CONCAT macro is a bit tricky to use if it must work in non-ANSI * mode -- there must be no spaces between its arguments, and for nested * __CONCAT's, all the __CONCAT's must be at the left. __CONCAT can also * concatenate double-quoted strings produced by the __STRING macro, but * this only works with ANSI C. * * __XSTRING is like __STRING, but it expands any macros in its argument * first. It is only available with ANSI C. */ #if defined(__STDC__) || defined(__cplusplus) #define __P(protos) protos /* full-blown ANSI C */ #define __CONCAT1(x,y) x ## y #define __CONCAT(x,y) __CONCAT1(x,y) #define __STRING(x) #x /* stringify without expanding x */ #define __XSTRING(x) __STRING(x) /* expand x, then stringify */ #define __const const /* define reserved names to standard */ #define __signed signed #define __volatile volatile #if defined(__cplusplus) #define __inline inline /* convert to C++ keyword */ #else #if !(defined(__CC_SUPPORTS___INLINE)) #define __inline /* delete GCC keyword */ #endif /* ! __CC_SUPPORTS___INLINE */ #endif /* !__cplusplus */ #else /* !(__STDC__ || __cplusplus) */ #define __P(protos) () /* traditional C preprocessor */ #define __CONCAT(x,y) x/**/y #define __STRING(x) "x" #if !defined(__CC_SUPPORTS___INLINE) #define __const /* delete pseudo-ANSI C keywords */ #define __inline #define __signed #define __volatile /* * In non-ANSI C environments, new programs will want ANSI-only C keywords * deleted from the program and old programs will want them left alone. * When using a compiler other than gcc, programs using the ANSI C keywords * const, inline etc. as normal identifiers should define -DNO_ANSI_KEYWORDS. * When using "gcc -traditional", we assume that this is the intent; if * __GNUC__ is defined but __STDC__ is not, we leave the new keywords alone. */ #ifndef NO_ANSI_KEYWORDS #define const /* delete ANSI C keywords */ #define inline #define signed #define volatile #endif /* !NO_ANSI_KEYWORDS */ #endif /* !__CC_SUPPORTS___INLINE */ #endif /* !(__STDC__ || __cplusplus) */ /* * Compiler-dependent macros to help declare dead (non-returning) and * pure (no side effects) functions, and unused variables. They are * null except for versions of gcc that are known to support the features * properly (old versions of gcc-2 supported the dead and pure features * in a different (wrong) way). If we do not provide an implementation * for a given compiler, let the compile fail if it is told to use * a feature that we cannot live without. */ #ifdef lint #define __dead2 #define __pure2 #define __unused #define __packed #define __aligned(x) #define __arg_type_tag(arg_kind, arg_idx, type_tag_idx) #define __section(x) #define __weak #else #define __weak __attribute__((__weak__)) #if !__GNUC_PREREQ__(2, 5) && !defined(__INTEL_COMPILER) #define __dead2 #define __pure2 #define __unused #endif #if __GNUC__ == 2 && __GNUC_MINOR__ >= 5 && __GNUC_MINOR__ < 7 && !defined(__INTEL_COMPILER) #define __dead2 __attribute__((__noreturn__)) #define __pure2 __attribute__((__const__)) #define __unused /* XXX Find out what to do for __packed, __aligned and __section */ #endif #if __GNUC_PREREQ__(2, 7) #define __dead2 __attribute__((__noreturn__)) #define __pure2 __attribute__((__const__)) #define __unused __attribute__((__unused__)) #define __used __attribute__((__used__)) #define __packed __attribute__((__packed__)) #define __aligned(x) __attribute__((__aligned__(x))) #define __section(x) __attribute__((__section__(x))) #endif #if __has_attribute(argument_with_type_tag) #define __arg_type_tag(arg_kind, arg_idx, type_tag_idx) \ __attribute__((__argument_with_type_tag__(arg_kind, arg_idx, type_tag_idx))) #else #define __arg_type_tag(arg_kind, arg_idx, type_tag_idx) #endif #if defined(__INTEL_COMPILER) #define __dead2 __attribute__((__noreturn__)) #define __pure2 __attribute__((__const__)) #define __unused __attribute__((__unused__)) #define __used __attribute__((__used__)) #define __packed __attribute__((__packed__)) #define __aligned(x) __attribute__((__aligned__(x))) #define __section(x) __attribute__((__section__(x))) #endif #endif /* lint */ #if !__GNUC_PREREQ__(2, 95) #define __alignof(x) __offsetof(struct { char __a; x __b; }, __b) #endif /* * Keywords added in C11. */ #if !defined(__STDC_VERSION__) || __STDC_VERSION__ < 201112L || defined(lint) #if !__has_extension(c_alignas) #if (defined(__cplusplus) && __cplusplus >= 201103L) || \ __has_extension(cxx_alignas) #define _Alignas(x) alignas(x) #else /* XXX: Only emulates _Alignas(constant-expression); not _Alignas(type-name). */ #define _Alignas(x) __aligned(x) #endif #endif #if defined(__cplusplus) && __cplusplus >= 201103L #define _Alignof(x) alignof(x) #else #define _Alignof(x) __alignof(x) #endif #if !__has_extension(c_atomic) && !__has_extension(cxx_atomic) /* * No native support for _Atomic(). Place object in structure to prevent * most forms of direct non-atomic access. */ #define _Atomic(T) struct { T volatile __val; } #endif #if defined(__cplusplus) && __cplusplus >= 201103L #define _Noreturn [[noreturn]] #else #define _Noreturn __dead2 #endif #if !__has_extension(c_static_assert) #if (defined(__cplusplus) && __cplusplus >= 201103L) || \ __has_extension(cxx_static_assert) #define _Static_assert(x, y) static_assert(x, y) #elif __GNUC_PREREQ__(4,6) /* Nothing, gcc 4.6 and higher has _Static_assert built-in */ #elif defined(__COUNTER__) #define _Static_assert(x, y) __Static_assert(x, __COUNTER__) #define __Static_assert(x, y) ___Static_assert(x, y) #define ___Static_assert(x, y) typedef char __assert_ ## y[(x) ? 1 : -1] \ __unused #else #define _Static_assert(x, y) struct __hack #endif #endif #if !__has_extension(c_thread_local) /* * XXX: Some compilers (Clang 3.3, GCC 4.7) falsely announce C++11 mode * without actually supporting the thread_local keyword. Don't check for * the presence of C++11 when defining _Thread_local. */ #if /* (defined(__cplusplus) && __cplusplus >= 201103L) || */ \ __has_extension(cxx_thread_local) #define _Thread_local thread_local #else #define _Thread_local __thread #endif #endif #endif /* __STDC_VERSION__ || __STDC_VERSION__ < 201112L */ /* * Emulation of C11 _Generic(). Unlike the previously defined C11 * keywords, it is not possible to implement this using exactly the same * syntax. Therefore implement something similar under the name * __generic(). Unlike _Generic(), this macro can only distinguish * between a single type, so it requires nested invocations to * distinguish multiple cases. */ #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L) || \ __has_extension(c_generic_selections) #define __generic(expr, t, yes, no) \ _Generic(expr, t: yes, default: no) #elif __GNUC_PREREQ__(3, 1) && !defined(__cplusplus) #define __generic(expr, t, yes, no) \ __builtin_choose_expr( \ __builtin_types_compatible_p(__typeof(expr), t), yes, no) #endif #if __GNUC_PREREQ__(2, 96) #define __malloc_like __attribute__((__malloc__)) #define __pure __attribute__((__pure__)) #else #define __malloc_like #define __pure #endif #if __GNUC_PREREQ__(3, 1) || (defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 800) #define __always_inline __attribute__((__always_inline__)) #else #define __always_inline #endif #if __GNUC_PREREQ__(3, 1) #define __noinline __attribute__ ((__noinline__)) #else #define __noinline #endif #if __GNUC_PREREQ__(3, 3) #define __nonnull(x) __attribute__((__nonnull__(x))) #define __nonnull_all __attribute__((__nonnull__)) #else #define __nonnull(x) #define __nonnull_all #endif #if __GNUC_PREREQ__(3, 4) #define __fastcall __attribute__((__fastcall__)) #define __result_use_check __attribute__((__warn_unused_result__)) #else #define __fastcall #define __result_use_check #endif #if __GNUC_PREREQ__(4, 1) #define __gnu_inline __attribute__((__gnu_inline__)) #define __returns_twice __attribute__((__returns_twice__)) #else #define __gnu_inline #define __returns_twice #endif #if __has_attribute(alloc_size) || __GNUC_PREREQ__(4, 3) #define __alloc_size(x) __attribute__((__alloc_size__(x))) #else #define __alloc_size(x) #endif +#if __has_attribute(alloc_align) || __GNUC_PREREQ__(4, 9) +#define __alloc_align(x) __attribute__((__alloc_align__(x))) +#else +#define __alloc_align(x) +#endif + /* XXX: should use `#if __STDC_VERSION__ < 199901'. */ #if !__GNUC_PREREQ__(2, 7) && !defined(__INTEL_COMPILER) #define __func__ NULL #endif #if (defined(__INTEL_COMPILER) || (defined(__GNUC__) && __GNUC__ >= 2)) && !defined(__STRICT_ANSI__) || __STDC_VERSION__ >= 199901 #define __LONG_LONG_SUPPORTED #endif /* C++11 exposes a load of C99 stuff */ #if defined(__cplusplus) && __cplusplus >= 201103L #define __LONG_LONG_SUPPORTED #ifndef __STDC_LIMIT_MACROS #define __STDC_LIMIT_MACROS #endif #ifndef __STDC_CONSTANT_MACROS #define __STDC_CONSTANT_MACROS #endif #endif /* * GCC 2.95 provides `__restrict' as an extension to C90 to support the * C99-specific `restrict' type qualifier. We happen to use `__restrict' as * a way to define the `restrict' type qualifier without disturbing older * software that is unaware of C99 keywords. */ #if !(__GNUC__ == 2 && __GNUC_MINOR__ == 95) #if !defined(__STDC_VERSION__) || __STDC_VERSION__ < 199901 || defined(lint) #define __restrict #else #define __restrict restrict #endif #endif /* * GNU C version 2.96 adds explicit branch prediction so that * the CPU back-end can hint the processor and also so that * code blocks can be reordered such that the predicted path * sees a more linear flow, thus improving cache behavior, etc. * * The following two macros provide us with a way to utilize this * compiler feature. Use __predict_true() if you expect the expression * to evaluate to true, and __predict_false() if you expect the * expression to evaluate to false. * * A few notes about usage: * * * Generally, __predict_false() error condition checks (unless * you have some _strong_ reason to do otherwise, in which case * document it), and/or __predict_true() `no-error' condition * checks, assuming you want to optimize for the no-error case. * * * Other than that, if you don't know the likelihood of a test * succeeding from empirical or other `hard' evidence, don't * make predictions. * * * These are meant to be used in places that are run `a lot'. * It is wasteful to make predictions in code that is run * seldomly (e.g. at subsystem initialization time) as the * basic block reordering that this affects can often generate * larger code. */ #if __GNUC_PREREQ__(2, 96) #define __predict_true(exp) __builtin_expect((exp), 1) #define __predict_false(exp) __builtin_expect((exp), 0) #else #define __predict_true(exp) (exp) #define __predict_false(exp) (exp) #endif #if __GNUC_PREREQ__(4, 2) #define __hidden __attribute__((__visibility__("hidden"))) #define __exported __attribute__((__visibility__("default"))) #else #define __hidden #define __exported #endif /* * We define this here since , , and * require it. */ #if __GNUC_PREREQ__(4, 1) #define __offsetof(type, field) __builtin_offsetof(type, field) #else #ifndef __cplusplus #define __offsetof(type, field) \ ((__size_t)(__uintptr_t)((const volatile void *)&((type *)0)->field)) #else #define __offsetof(type, field) \ (__offsetof__ (reinterpret_cast <__size_t> \ (&reinterpret_cast \ (static_cast (0)->field)))) #endif #endif #define __rangeof(type, start, end) \ (__offsetof(type, end) - __offsetof(type, start)) /* * Given the pointer x to the member m of the struct s, return * a pointer to the containing structure. When using GCC, we first * assign pointer x to a local variable, to check that its type is * compatible with member m. */ #if __GNUC_PREREQ__(3, 1) #define __containerof(x, s, m) ({ \ const volatile __typeof(((s *)0)->m) *__x = (x); \ __DEQUALIFY(s *, (const volatile char *)__x - __offsetof(s, m));\ }) #else #define __containerof(x, s, m) \ __DEQUALIFY(s *, (const volatile char *)(x) - __offsetof(s, m)) #endif /* * Compiler-dependent macros to declare that functions take printf-like * or scanf-like arguments. They are null except for versions of gcc * that are known to support the features properly (old versions of gcc-2 * didn't permit keeping the keywords out of the application namespace). */ #if !__GNUC_PREREQ__(2, 7) && !defined(__INTEL_COMPILER) #define __printflike(fmtarg, firstvararg) #define __scanflike(fmtarg, firstvararg) #define __format_arg(fmtarg) #define __strfmonlike(fmtarg, firstvararg) #define __strftimelike(fmtarg, firstvararg) #else #define __printflike(fmtarg, firstvararg) \ __attribute__((__format__ (__printf__, fmtarg, firstvararg))) #define __scanflike(fmtarg, firstvararg) \ __attribute__((__format__ (__scanf__, fmtarg, firstvararg))) #define __format_arg(fmtarg) __attribute__((__format_arg__ (fmtarg))) #define __strfmonlike(fmtarg, firstvararg) \ __attribute__((__format__ (__strfmon__, fmtarg, firstvararg))) #define __strftimelike(fmtarg, firstvararg) \ __attribute__((__format__ (__strftime__, fmtarg, firstvararg))) #endif /* Compiler-dependent macros that rely on FreeBSD-specific extensions. */ #if defined(__FreeBSD_cc_version) && __FreeBSD_cc_version >= 300001 && \ defined(__GNUC__) && !defined(__INTEL_COMPILER) #define __printf0like(fmtarg, firstvararg) \ __attribute__((__format__ (__printf0__, fmtarg, firstvararg))) #else #define __printf0like(fmtarg, firstvararg) #endif #if defined(__GNUC__) || defined(__INTEL_COMPILER) #ifndef __INTEL_COMPILER #define __strong_reference(sym,aliassym) \ extern __typeof (sym) aliassym __attribute__ ((__alias__ (#sym))) #endif #ifdef __STDC__ #define __weak_reference(sym,alias) \ __asm__(".weak " #alias); \ __asm__(".equ " #alias ", " #sym) #define __warn_references(sym,msg) \ __asm__(".section .gnu.warning." #sym); \ __asm__(".asciz \"" msg "\""); \ __asm__(".previous") #define __sym_compat(sym,impl,verid) \ __asm__(".symver " #impl ", " #sym "@" #verid) #define __sym_default(sym,impl,verid) \ __asm__(".symver " #impl ", " #sym "@@" #verid) #else #define __weak_reference(sym,alias) \ __asm__(".weak alias"); \ __asm__(".equ alias, sym") #define __warn_references(sym,msg) \ __asm__(".section .gnu.warning.sym"); \ __asm__(".asciz \"msg\""); \ __asm__(".previous") #define __sym_compat(sym,impl,verid) \ __asm__(".symver impl, sym@verid") #define __sym_default(impl,sym,verid) \ __asm__(".symver impl, sym@@verid") #endif /* __STDC__ */ #endif /* __GNUC__ || __INTEL_COMPILER */ #define __GLOBL1(sym) __asm__(".globl " #sym) #define __GLOBL(sym) __GLOBL1(sym) #if defined(__GNUC__) || defined(__INTEL_COMPILER) #define __IDSTRING(name,string) __asm__(".ident\t\"" string "\"") #else /* * The following definition might not work well if used in header files, * but it should be better than nothing. If you want a "do nothing" * version, then it should generate some harmless declaration, such as: * #define __IDSTRING(name,string) struct __hack */ #define __IDSTRING(name,string) static const char name[] __unused = string #endif /* * Embed the rcs id of a source file in the resulting library. Note that in * more recent ELF binutils, we use .ident allowing the ID to be stripped. * Usage: * __FBSDID("$FreeBSD$"); */ #ifndef __FBSDID #if !defined(lint) && !defined(STRIP_FBSDID) #define __FBSDID(s) __IDSTRING(__CONCAT(__rcsid_,__LINE__),s) #else #define __FBSDID(s) struct __hack #endif #endif #ifndef __RCSID #ifndef NO__RCSID #define __RCSID(s) __IDSTRING(__CONCAT(__rcsid_,__LINE__),s) #else #define __RCSID(s) struct __hack #endif #endif #ifndef __RCSID_SOURCE #ifndef NO__RCSID_SOURCE #define __RCSID_SOURCE(s) __IDSTRING(__CONCAT(__rcsid_source_,__LINE__),s) #else #define __RCSID_SOURCE(s) struct __hack #endif #endif #ifndef __SCCSID #ifndef NO__SCCSID #define __SCCSID(s) __IDSTRING(__CONCAT(__sccsid_,__LINE__),s) #else #define __SCCSID(s) struct __hack #endif #endif #ifndef __COPYRIGHT #ifndef NO__COPYRIGHT #define __COPYRIGHT(s) __IDSTRING(__CONCAT(__copyright_,__LINE__),s) #else #define __COPYRIGHT(s) struct __hack #endif #endif #ifndef __DECONST #define __DECONST(type, var) ((type)(__uintptr_t)(const void *)(var)) #endif #ifndef __DEVOLATILE #define __DEVOLATILE(type, var) ((type)(__uintptr_t)(volatile void *)(var)) #endif #ifndef __DEQUALIFY #define __DEQUALIFY(type, var) ((type)(__uintptr_t)(const volatile void *)(var)) #endif /*- * The following definitions are an extension of the behavior originally * implemented in , but with a different level of granularity. * POSIX.1 requires that the macros we test be defined before any standard * header file is included. * * Here's a quick run-down of the versions: * defined(_POSIX_SOURCE) 1003.1-1988 * _POSIX_C_SOURCE == 1 1003.1-1990 * _POSIX_C_SOURCE == 2 1003.2-1992 C Language Binding Option * _POSIX_C_SOURCE == 199309 1003.1b-1993 * _POSIX_C_SOURCE == 199506 1003.1c-1995, 1003.1i-1995, * and the omnibus ISO/IEC 9945-1: 1996 * _POSIX_C_SOURCE == 200112 1003.1-2001 * _POSIX_C_SOURCE == 200809 1003.1-2008 * * In addition, the X/Open Portability Guide, which is now the Single UNIX * Specification, defines a feature-test macro which indicates the version of * that specification, and which subsumes _POSIX_C_SOURCE. * * Our macros begin with two underscores to avoid namespace screwage. */ /* Deal with IEEE Std. 1003.1-1990, in which _POSIX_C_SOURCE == 1. */ #if defined(_POSIX_C_SOURCE) && _POSIX_C_SOURCE == 1 #undef _POSIX_C_SOURCE /* Probably illegal, but beyond caring now. */ #define _POSIX_C_SOURCE 199009 #endif /* Deal with IEEE Std. 1003.2-1992, in which _POSIX_C_SOURCE == 2. */ #if defined(_POSIX_C_SOURCE) && _POSIX_C_SOURCE == 2 #undef _POSIX_C_SOURCE #define _POSIX_C_SOURCE 199209 #endif /* Deal with various X/Open Portability Guides and Single UNIX Spec. */ #ifdef _XOPEN_SOURCE #if _XOPEN_SOURCE - 0 >= 700 #define __XSI_VISIBLE 700 #undef _POSIX_C_SOURCE #define _POSIX_C_SOURCE 200809 #elif _XOPEN_SOURCE - 0 >= 600 #define __XSI_VISIBLE 600 #undef _POSIX_C_SOURCE #define _POSIX_C_SOURCE 200112 #elif _XOPEN_SOURCE - 0 >= 500 #define __XSI_VISIBLE 500 #undef _POSIX_C_SOURCE #define _POSIX_C_SOURCE 199506 #endif #endif /* * Deal with all versions of POSIX. The ordering relative to the tests above is * important. */ #if defined(_POSIX_SOURCE) && !defined(_POSIX_C_SOURCE) #define _POSIX_C_SOURCE 198808 #endif #ifdef _POSIX_C_SOURCE #if _POSIX_C_SOURCE >= 200809 #define __POSIX_VISIBLE 200809 #define __ISO_C_VISIBLE 1999 #elif _POSIX_C_SOURCE >= 200112 #define __POSIX_VISIBLE 200112 #define __ISO_C_VISIBLE 1999 #elif _POSIX_C_SOURCE >= 199506 #define __POSIX_VISIBLE 199506 #define __ISO_C_VISIBLE 1990 #elif _POSIX_C_SOURCE >= 199309 #define __POSIX_VISIBLE 199309 #define __ISO_C_VISIBLE 1990 #elif _POSIX_C_SOURCE >= 199209 #define __POSIX_VISIBLE 199209 #define __ISO_C_VISIBLE 1990 #elif _POSIX_C_SOURCE >= 199009 #define __POSIX_VISIBLE 199009 #define __ISO_C_VISIBLE 1990 #else #define __POSIX_VISIBLE 198808 #define __ISO_C_VISIBLE 0 #endif /* _POSIX_C_SOURCE */ #else /*- * Deal with _ANSI_SOURCE: * If it is defined, and no other compilation environment is explicitly * requested, then define our internal feature-test macros to zero. This * makes no difference to the preprocessor (undefined symbols in preprocessing * expressions are defined to have value zero), but makes it more convenient for * a test program to print out the values. * * If a program mistakenly defines _ANSI_SOURCE and some other macro such as * _POSIX_C_SOURCE, we will assume that it wants the broader compilation * environment (and in fact we will never get here). */ #if defined(_ANSI_SOURCE) /* Hide almost everything. */ #define __POSIX_VISIBLE 0 #define __XSI_VISIBLE 0 #define __BSD_VISIBLE 0 #define __ISO_C_VISIBLE 1990 #elif defined(_C99_SOURCE) /* Localism to specify strict C99 env. */ #define __POSIX_VISIBLE 0 #define __XSI_VISIBLE 0 #define __BSD_VISIBLE 0 #define __ISO_C_VISIBLE 1999 #elif defined(_C11_SOURCE) /* Localism to specify strict C11 env. */ #define __POSIX_VISIBLE 0 #define __XSI_VISIBLE 0 #define __BSD_VISIBLE 0 #define __ISO_C_VISIBLE 2011 #else /* Default environment: show everything. */ #define __POSIX_VISIBLE 200809 #define __XSI_VISIBLE 700 #define __BSD_VISIBLE 1 #define __ISO_C_VISIBLE 2011 #endif #endif #if defined(__mips) || defined(__powerpc64__) #define __NO_TLS 1 #endif /* * Lock annotations. * * Clang provides support for doing basic thread-safety tests at * compile-time, by marking which locks will/should be held when * entering/leaving a functions. * * Furthermore, it is also possible to annotate variables and structure * members to enforce that they are only accessed when certain locks are * held. */ #if __has_extension(c_thread_safety_attributes) #define __lock_annotate(x) __attribute__((x)) #else #define __lock_annotate(x) #endif /* Structure implements a lock. */ #define __lockable __lock_annotate(lockable) /* Function acquires an exclusive or shared lock. */ #define __locks_exclusive(...) \ __lock_annotate(exclusive_lock_function(__VA_ARGS__)) #define __locks_shared(...) \ __lock_annotate(shared_lock_function(__VA_ARGS__)) /* Function attempts to acquire an exclusive or shared lock. */ #define __trylocks_exclusive(...) \ __lock_annotate(exclusive_trylock_function(__VA_ARGS__)) #define __trylocks_shared(...) \ __lock_annotate(shared_trylock_function(__VA_ARGS__)) /* Function releases a lock. */ #define __unlocks(...) __lock_annotate(unlock_function(__VA_ARGS__)) /* Function asserts that an exclusive or shared lock is held. */ #define __asserts_exclusive(...) \ __lock_annotate(assert_exclusive_lock(__VA_ARGS__)) #define __asserts_shared(...) \ __lock_annotate(assert_shared_lock(__VA_ARGS__)) /* Function requires that an exclusive or shared lock is or is not held. */ #define __requires_exclusive(...) \ __lock_annotate(exclusive_locks_required(__VA_ARGS__)) #define __requires_shared(...) \ __lock_annotate(shared_locks_required(__VA_ARGS__)) #define __requires_unlocked(...) \ __lock_annotate(locks_excluded(__VA_ARGS__)) /* Function should not be analyzed. */ #define __no_lock_analysis __lock_annotate(no_thread_safety_analysis) /* Guard variables and structure members by lock. */ #define __guarded_by(x) __lock_annotate(guarded_by(x)) #define __pt_guarded_by(x) __lock_annotate(pt_guarded_by(x)) #endif /* !_SYS_CDEFS_H_ */ Index: user/ngie/more-tests/sys/sys/param.h =================================================================== --- user/ngie/more-tests/sys/sys/param.h (revision 281974) +++ user/ngie/more-tests/sys/sys/param.h (revision 281975) @@ -1,348 +1,362 @@ /*- * Copyright (c) 1982, 1986, 1989, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)param.h 8.3 (Berkeley) 4/4/95 * $FreeBSD$ */ #ifndef _SYS_PARAM_H_ #define _SYS_PARAM_H_ #include #define BSD 199506 /* System version (year & month). */ #define BSD4_3 1 #define BSD4_4 1 /* * __FreeBSD_version numbers are documented in the Porter's Handbook. * If you bump the version for any reason, you should update the documentation * there. * Currently this lives here in the doc/ repository: * * head/en_US.ISO8859-1/books/porters-handbook/versions/chapter.xml * * scheme is: Rxx * 'R' is in the range 0 to 4 if this is a release branch or * x.0-CURRENT before RELENG_*_0 is created, otherwise 'R' is * in the range 5 to 9. */ #undef __FreeBSD_version #define __FreeBSD_version 1100070 /* Master, propagated to newvers */ /* * __FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD, * which by definition is always true on FreeBSD. This macro is also defined * on other systems that use the kernel of FreeBSD, such as GNU/kFreeBSD. * * It is tempting to use this macro in userland code when we want to enable * kernel-specific routines, and in fact it's fine to do this in code that * is part of FreeBSD itself. However, be aware that as presence of this * macro is still not widespread (e.g. older FreeBSD versions, 3rd party * compilers, etc), it is STRONGLY DISCOURAGED to check for this macro in * external applications without also checking for __FreeBSD__ as an * alternative. */ #undef __FreeBSD_kernel__ #define __FreeBSD_kernel__ #ifdef _KERNEL #define P_OSREL_SIGWAIT 700000 #define P_OSREL_SIGSEGV 700004 #define P_OSREL_MAP_ANON 800104 #define P_OSREL_MAP_FSTRICT 1100036 #define P_OSREL_MAJOR(x) ((x) / 100000) #endif #ifndef LOCORE #include #endif /* * Machine-independent constants (some used in following include files). * Redefined constants are from POSIX 1003.1 limits file. * * MAXCOMLEN should be >= sizeof(ac_comm) (see ) */ #include #define MAXCOMLEN 19 /* max command name remembered */ #define MAXINTERP PATH_MAX /* max interpreter file name length */ #define MAXLOGNAME 33 /* max login name length (incl. NUL) */ #define MAXUPRC CHILD_MAX /* max simultaneous processes */ #define NCARGS ARG_MAX /* max bytes for an exec function */ #define NGROUPS (NGROUPS_MAX+1) /* max number groups */ #define NOFILE OPEN_MAX /* max open files per process */ #define NOGROUP 65535 /* marker for empty group set member */ #define MAXHOSTNAMELEN 256 /* max hostname size */ #define SPECNAMELEN 63 /* max length of devicename */ /* More types and definitions used throughout the kernel. */ #ifdef _KERNEL #include #include #ifndef LOCORE #include #include #endif #ifndef FALSE #define FALSE 0 #endif #ifndef TRUE #define TRUE 1 #endif #endif #ifndef _KERNEL /* Signals. */ #include #endif /* Machine type dependent parameters. */ #include #ifndef _KERNEL #include #endif #ifndef DEV_BSHIFT #define DEV_BSHIFT 9 /* log2(DEV_BSIZE) */ #endif #define DEV_BSIZE (1<>PAGE_SHIFT) #endif /* * btodb() is messy and perhaps slow because `bytes' may be an off_t. We * want to shift an unsigned type to avoid sign extension and we don't * want to widen `bytes' unnecessarily. Assume that the result fits in * a daddr_t. */ #ifndef btodb #define btodb(bytes) /* calculates (bytes / DEV_BSIZE) */ \ (sizeof (bytes) > sizeof(long) \ ? (daddr_t)((unsigned long long)(bytes) >> DEV_BSHIFT) \ : (daddr_t)((unsigned long)(bytes) >> DEV_BSHIFT)) #endif #ifndef dbtob #define dbtob(db) /* calculates (db * DEV_BSIZE) */ \ ((off_t)(db) << DEV_BSHIFT) #endif #define PRIMASK 0x0ff #define PCATCH 0x100 /* OR'd with pri for tsleep to check signals */ #define PDROP 0x200 /* OR'd with pri to stop re-entry of interlock mutex */ #define NZERO 0 /* default "nice" */ #define NBBY 8 /* number of bits in a byte */ #define NBPW sizeof(int) /* number of bytes per word (integer) */ #define CMASK 022 /* default file mask: S_IWGRP|S_IWOTH */ #define NODEV (dev_t)(-1) /* non-existent device */ /* * File system parameters and macros. * * MAXBSIZE - Filesystems are made out of blocks of at most MAXBSIZE bytes * per block. MAXBSIZE may be made larger without effecting * any existing filesystems as long as it does not exceed MAXPHYS, * and may be made smaller at the risk of not being able to use * filesystems which require a block size exceeding MAXBSIZE. * + * MAXBCACHEBUF - Maximum size of a buffer in the buffer cache. This must + * be >= MAXBSIZE and can be set differently for different + * architectures by defining it in . + * Making this larger allows NFS to do larger reads/writes. + * * BKVASIZE - Nominal buffer space per buffer, in bytes. BKVASIZE is the * minimum KVM memory reservation the kernel is willing to make. * Filesystems can of course request smaller chunks. Actual * backing memory uses a chunk size of a page (PAGE_SIZE). + * The default value here can be overridden on a per-architecture + * basis by defining it in . This should + * probably be done to increase its value, when MAXBCACHEBUF is + * defined as a larger value in . * * If you make BKVASIZE too small you risk seriously fragmenting * the buffer KVM map which may slow things down a bit. If you * make it too big the kernel will not be able to optimally use * the KVM memory reserved for the buffer cache and will wind * up with too-few buffers. * * The default is 16384, roughly 2x the block size used by a * normal UFS filesystem. */ #define MAXBSIZE 65536 /* must be power of 2 */ +#ifndef MAXBCACHEBUF +#define MAXBCACHEBUF MAXBSIZE /* must be a power of 2 >= MAXBSIZE */ +#endif +#ifndef BKVASIZE #define BKVASIZE 16384 /* must be power of 2 */ +#endif #define BKVAMASK (BKVASIZE-1) /* * MAXPATHLEN defines the longest permissible path length after expanding * symbolic links. It is used to allocate a temporary buffer from the buffer * pool in which to do the name expansion, hence should be a power of two, * and must be less than or equal to MAXBSIZE. MAXSYMLINKS defines the * maximum number of symbolic links that may be expanded in a path name. * It should be set high enough to allow all legitimate uses, but halt * infinite loops reasonably quickly. */ #define MAXPATHLEN PATH_MAX #define MAXSYMLINKS 32 /* Bit map related macros. */ #define setbit(a,i) (((unsigned char *)(a))[(i)/NBBY] |= 1<<((i)%NBBY)) #define clrbit(a,i) (((unsigned char *)(a))[(i)/NBBY] &= ~(1<<((i)%NBBY))) #define isset(a,i) \ (((const unsigned char *)(a))[(i)/NBBY] & (1<<((i)%NBBY))) #define isclr(a,i) \ ((((const unsigned char *)(a))[(i)/NBBY] & (1<<((i)%NBBY))) == 0) /* Macros for counting and rounding. */ #ifndef howmany #define howmany(x, y) (((x)+((y)-1))/(y)) #endif #define nitems(x) (sizeof((x)) / sizeof((x)[0])) #define rounddown(x, y) (((x)/(y))*(y)) #define rounddown2(x, y) ((x)&(~((y)-1))) /* if y is power of two */ #define roundup(x, y) ((((x)+((y)-1))/(y))*(y)) /* to any y */ #define roundup2(x, y) (((x)+((y)-1))&(~((y)-1))) /* if y is powers of two */ #define powerof2(x) ((((x)-1)&(x))==0) /* Macros for min/max. */ #define MIN(a,b) (((a)<(b))?(a):(b)) #define MAX(a,b) (((a)>(b))?(a):(b)) #ifdef _KERNEL /* * Basic byte order function prototypes for non-inline functions. */ #ifndef LOCORE #ifndef _BYTEORDER_PROTOTYPED #define _BYTEORDER_PROTOTYPED __BEGIN_DECLS __uint32_t htonl(__uint32_t); __uint16_t htons(__uint16_t); __uint32_t ntohl(__uint32_t); __uint16_t ntohs(__uint16_t); __END_DECLS #endif #endif #ifndef lint #ifndef _BYTEORDER_FUNC_DEFINED #define _BYTEORDER_FUNC_DEFINED #define htonl(x) __htonl(x) #define htons(x) __htons(x) #define ntohl(x) __ntohl(x) #define ntohs(x) __ntohs(x) #endif /* !_BYTEORDER_FUNC_DEFINED */ #endif /* lint */ #endif /* _KERNEL */ /* * Scale factor for scaled integers used to count %cpu time and load avgs. * * The number of CPU `tick's that map to a unique `%age' can be expressed * by the formula (1 / (2 ^ (FSHIFT - 11))). The maximum load average that * can be calculated (assuming 32 bits) can be closely approximated using * the formula (2 ^ (2 * (16 - FSHIFT))) for (FSHIFT < 15). * * For the scheduler to maintain a 1:1 mapping of CPU `tick' to `%age', * FSHIFT must be at least 11; this gives us a maximum load avg of ~1024. */ #define FSHIFT 11 /* bits to right of fixed binary point */ #define FSCALE (1<> (PAGE_SHIFT - DEV_BSHIFT)) #define ctodb(db) /* calculates pages to devblks */ \ ((db) << (PAGE_SHIFT - DEV_BSHIFT)) /* * Old spelling of __containerof(). */ #define member2struct(s, m, x) \ ((struct s *)(void *)((char *)(x) - offsetof(struct s, m))) /* * Access a variable length array that has been declared as a fixed * length array. */ #define __PAST_END(array, offset) (((__typeof__(*(array)) *)(array))[offset]) #endif /* _SYS_PARAM_H_ */ Index: user/ngie/more-tests/sys/ufs/ffs/ffs_alloc.c =================================================================== --- user/ngie/more-tests/sys/ufs/ffs/ffs_alloc.c (revision 281974) +++ user/ngie/more-tests/sys/ufs/ffs/ffs_alloc.c (revision 281975) @@ -1,3158 +1,3204 @@ /*- * Copyright (c) 2002 Networks Associates Technology, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Marshall * Kirk McKusick and Network Associates Laboratories, the Security * Research Division of Network Associates, Inc. under DARPA/SPAWAR * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS * research program * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * Copyright (c) 1982, 1986, 1989, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ffs_alloc.c 8.18 (Berkeley) 5/26/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_quota.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include typedef ufs2_daddr_t allocfcn_t(struct inode *ip, u_int cg, ufs2_daddr_t bpref, int size, int rsize); static ufs2_daddr_t ffs_alloccg(struct inode *, u_int, ufs2_daddr_t, int, int); static ufs2_daddr_t ffs_alloccgblk(struct inode *, struct buf *, ufs2_daddr_t, int); static void ffs_blkfree_cg(struct ufsmount *, struct fs *, struct vnode *, ufs2_daddr_t, long, ino_t, struct workhead *); static void ffs_blkfree_trim_completed(struct bio *); static void ffs_blkfree_trim_task(void *ctx, int pending __unused); #ifdef INVARIANTS static int ffs_checkblk(struct inode *, ufs2_daddr_t, long); #endif -static ufs2_daddr_t ffs_clusteralloc(struct inode *, u_int, ufs2_daddr_t, int, - int); +static ufs2_daddr_t ffs_clusteralloc(struct inode *, u_int, ufs2_daddr_t, int); static ino_t ffs_dirpref(struct inode *); static ufs2_daddr_t ffs_fragextend(struct inode *, u_int, ufs2_daddr_t, int, int); static ufs2_daddr_t ffs_hashalloc (struct inode *, u_int, ufs2_daddr_t, int, int, allocfcn_t *); static ufs2_daddr_t ffs_nodealloccg(struct inode *, u_int, ufs2_daddr_t, int, int); static ufs1_daddr_t ffs_mapsearch(struct fs *, struct cg *, ufs2_daddr_t, int); static int ffs_reallocblks_ufs1(struct vop_reallocblks_args *); static int ffs_reallocblks_ufs2(struct vop_reallocblks_args *); /* * Allocate a block in the filesystem. * * The size of the requested block is given, which must be some * multiple of fs_fsize and <= fs_bsize. * A preference may be optionally specified. If a preference is given * the following hierarchy is used to allocate a block: * 1) allocate the requested block. * 2) allocate a rotationally optimal block in the same cylinder. * 3) allocate a block in the same cylinder group. * 4) quadradically rehash into other cylinder groups, until an * available block is located. * If no block preference is given the following hierarchy is used * to allocate a block: * 1) allocate a block in the cylinder group that contains the * inode for the file. * 2) quadradically rehash into other cylinder groups, until an * available block is located. */ int ffs_alloc(ip, lbn, bpref, size, flags, cred, bnp) struct inode *ip; ufs2_daddr_t lbn, bpref; int size, flags; struct ucred *cred; ufs2_daddr_t *bnp; { struct fs *fs; struct ufsmount *ump; ufs2_daddr_t bno; u_int cg, reclaimed; static struct timeval lastfail; static int curfail; int64_t delta; #ifdef QUOTA int error; #endif *bnp = 0; fs = ip->i_fs; ump = ip->i_ump; mtx_assert(UFS_MTX(ump), MA_OWNED); #ifdef INVARIANTS if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) { printf("dev = %s, bsize = %ld, size = %d, fs = %s\n", devtoname(ip->i_dev), (long)fs->fs_bsize, size, fs->fs_fsmnt); panic("ffs_alloc: bad size"); } if (cred == NOCRED) panic("ffs_alloc: missing credential"); #endif /* INVARIANTS */ reclaimed = 0; retry: #ifdef QUOTA UFS_UNLOCK(ump); error = chkdq(ip, btodb(size), cred, 0); if (error) return (error); UFS_LOCK(ump); #endif if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0) goto nospace; if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE, 0) && freespace(fs, fs->fs_minfree) - numfrags(fs, size) < 0) goto nospace; if (bpref >= fs->fs_size) bpref = 0; if (bpref == 0) cg = ino_to_cg(fs, ip->i_number); else cg = dtog(fs, bpref); bno = ffs_hashalloc(ip, cg, bpref, size, size, ffs_alloccg); if (bno > 0) { delta = btodb(size); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta); if (flags & IO_EXT) ip->i_flag |= IN_CHANGE; else ip->i_flag |= IN_CHANGE | IN_UPDATE; *bnp = bno; return (0); } nospace: #ifdef QUOTA UFS_UNLOCK(ump); /* * Restore user's disk quota because allocation failed. */ (void) chkdq(ip, -btodb(size), cred, FORCE); UFS_LOCK(ump); #endif if (reclaimed == 0 && (flags & IO_BUFLOCKED) == 0) { reclaimed = 1; softdep_request_cleanup(fs, ITOV(ip), cred, FLUSH_BLOCKS_WAIT); goto retry; } UFS_UNLOCK(ump); if (reclaimed > 0 && ppsratecheck(&lastfail, &curfail, 1)) { ffs_fserr(fs, ip->i_number, "filesystem full"); uprintf("\n%s: write failed, filesystem is full\n", fs->fs_fsmnt); } return (ENOSPC); } /* * Reallocate a fragment to a bigger size * * The number and size of the old block is given, and a preference * and new size is also specified. The allocator attempts to extend * the original block. Failing that, the regular block allocator is * invoked to get an appropriate block. */ int ffs_realloccg(ip, lbprev, bprev, bpref, osize, nsize, flags, cred, bpp) struct inode *ip; ufs2_daddr_t lbprev; ufs2_daddr_t bprev; ufs2_daddr_t bpref; int osize, nsize, flags; struct ucred *cred; struct buf **bpp; { struct vnode *vp; struct fs *fs; struct buf *bp; struct ufsmount *ump; u_int cg, request, reclaimed; int error, gbflags; ufs2_daddr_t bno; static struct timeval lastfail; static int curfail; int64_t delta; *bpp = 0; vp = ITOV(ip); fs = ip->i_fs; bp = NULL; ump = ip->i_ump; gbflags = (flags & BA_UNMAPPED) != 0 ? GB_UNMAPPED : 0; mtx_assert(UFS_MTX(ump), MA_OWNED); #ifdef INVARIANTS if (vp->v_mount->mnt_kern_flag & MNTK_SUSPENDED) panic("ffs_realloccg: allocation on suspended filesystem"); if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 || (u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) { printf( "dev = %s, bsize = %ld, osize = %d, nsize = %d, fs = %s\n", devtoname(ip->i_dev), (long)fs->fs_bsize, osize, nsize, fs->fs_fsmnt); panic("ffs_realloccg: bad size"); } if (cred == NOCRED) panic("ffs_realloccg: missing credential"); #endif /* INVARIANTS */ reclaimed = 0; retry: if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE, 0) && freespace(fs, fs->fs_minfree) - numfrags(fs, nsize - osize) < 0) { goto nospace; } if (bprev == 0) { printf("dev = %s, bsize = %ld, bprev = %jd, fs = %s\n", devtoname(ip->i_dev), (long)fs->fs_bsize, (intmax_t)bprev, fs->fs_fsmnt); panic("ffs_realloccg: bad bprev"); } UFS_UNLOCK(ump); /* * Allocate the extra space in the buffer. */ error = bread_gb(vp, lbprev, osize, NOCRED, gbflags, &bp); if (error) { brelse(bp); return (error); } if (bp->b_blkno == bp->b_lblkno) { if (lbprev >= NDADDR) panic("ffs_realloccg: lbprev out of range"); bp->b_blkno = fsbtodb(fs, bprev); } #ifdef QUOTA error = chkdq(ip, btodb(nsize - osize), cred, 0); if (error) { brelse(bp); return (error); } #endif /* * Check for extension in the existing location. */ cg = dtog(fs, bprev); UFS_LOCK(ump); bno = ffs_fragextend(ip, cg, bprev, osize, nsize); if (bno) { if (bp->b_blkno != fsbtodb(fs, bno)) panic("ffs_realloccg: bad blockno"); delta = btodb(nsize - osize); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta); if (flags & IO_EXT) ip->i_flag |= IN_CHANGE; else ip->i_flag |= IN_CHANGE | IN_UPDATE; allocbuf(bp, nsize); bp->b_flags |= B_DONE; vfs_bio_bzero_buf(bp, osize, nsize - osize); if ((bp->b_flags & (B_MALLOC | B_VMIO)) == B_VMIO) vfs_bio_set_valid(bp, osize, nsize - osize); *bpp = bp; return (0); } /* * Allocate a new disk location. */ if (bpref >= fs->fs_size) bpref = 0; switch ((int)fs->fs_optim) { case FS_OPTSPACE: /* * Allocate an exact sized fragment. Although this makes * best use of space, we will waste time relocating it if * the file continues to grow. If the fragmentation is * less than half of the minimum free reserve, we choose * to begin optimizing for time. */ request = nsize; if (fs->fs_minfree <= 5 || fs->fs_cstotal.cs_nffree > (off_t)fs->fs_dsize * fs->fs_minfree / (2 * 100)) break; log(LOG_NOTICE, "%s: optimization changed from SPACE to TIME\n", fs->fs_fsmnt); fs->fs_optim = FS_OPTTIME; break; case FS_OPTTIME: /* * At this point we have discovered a file that is trying to * grow a small fragment to a larger fragment. To save time, * we allocate a full sized block, then free the unused portion. * If the file continues to grow, the `ffs_fragextend' call * above will be able to grow it in place without further * copying. If aberrant programs cause disk fragmentation to * grow within 2% of the free reserve, we choose to begin * optimizing for space. */ request = fs->fs_bsize; if (fs->fs_cstotal.cs_nffree < (off_t)fs->fs_dsize * (fs->fs_minfree - 2) / 100) break; log(LOG_NOTICE, "%s: optimization changed from TIME to SPACE\n", fs->fs_fsmnt); fs->fs_optim = FS_OPTSPACE; break; default: printf("dev = %s, optim = %ld, fs = %s\n", devtoname(ip->i_dev), (long)fs->fs_optim, fs->fs_fsmnt); panic("ffs_realloccg: bad optim"); /* NOTREACHED */ } bno = ffs_hashalloc(ip, cg, bpref, request, nsize, ffs_alloccg); if (bno > 0) { bp->b_blkno = fsbtodb(fs, bno); if (!DOINGSOFTDEP(vp)) ffs_blkfree(ump, fs, ip->i_devvp, bprev, (long)osize, ip->i_number, vp->v_type, NULL); delta = btodb(nsize - osize); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + delta); if (flags & IO_EXT) ip->i_flag |= IN_CHANGE; else ip->i_flag |= IN_CHANGE | IN_UPDATE; allocbuf(bp, nsize); bp->b_flags |= B_DONE; vfs_bio_bzero_buf(bp, osize, nsize - osize); if ((bp->b_flags & (B_MALLOC | B_VMIO)) == B_VMIO) vfs_bio_set_valid(bp, osize, nsize - osize); *bpp = bp; return (0); } #ifdef QUOTA UFS_UNLOCK(ump); /* * Restore user's disk quota because allocation failed. */ (void) chkdq(ip, -btodb(nsize - osize), cred, FORCE); UFS_LOCK(ump); #endif nospace: /* * no space available */ if (reclaimed == 0 && (flags & IO_BUFLOCKED) == 0) { reclaimed = 1; UFS_UNLOCK(ump); if (bp) { brelse(bp); bp = NULL; } UFS_LOCK(ump); softdep_request_cleanup(fs, vp, cred, FLUSH_BLOCKS_WAIT); goto retry; } UFS_UNLOCK(ump); if (bp) brelse(bp); if (reclaimed > 0 && ppsratecheck(&lastfail, &curfail, 1)) { ffs_fserr(fs, ip->i_number, "filesystem full"); uprintf("\n%s: write failed, filesystem is full\n", fs->fs_fsmnt); } return (ENOSPC); } /* * Reallocate a sequence of blocks into a contiguous sequence of blocks. * * The vnode and an array of buffer pointers for a range of sequential * logical blocks to be made contiguous is given. The allocator attempts * to find a range of sequential blocks starting as close as possible * from the end of the allocation for the logical block immediately * preceding the current range. If successful, the physical block numbers * in the buffer pointers and in the inode are changed to reflect the new * allocation. If unsuccessful, the allocation is left unchanged. The * success in doing the reallocation is returned. Note that the error * return is not reflected back to the user. Rather the previous block * allocation will be used. */ SYSCTL_NODE(_vfs, OID_AUTO, ffs, CTLFLAG_RW, 0, "FFS filesystem"); static int doasyncfree = 1; -SYSCTL_INT(_vfs_ffs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0, ""); +SYSCTL_INT(_vfs_ffs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0, +"do not force synchronous writes when blocks are reallocated"); static int doreallocblks = 1; -SYSCTL_INT(_vfs_ffs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, ""); +SYSCTL_INT(_vfs_ffs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, +"enable block reallocation"); +static int maxclustersearch = 10; +SYSCTL_INT(_vfs_ffs, OID_AUTO, maxclustersearch, CTLFLAG_RW, &maxclustersearch, +0, "max number of cylinder group to search for contigous blocks"); + #ifdef DEBUG static volatile int prtrealloc = 0; #endif int ffs_reallocblks(ap) struct vop_reallocblks_args /* { struct vnode *a_vp; struct cluster_save *a_buflist; } */ *ap; { if (doreallocblks == 0) return (ENOSPC); /* * We can't wait in softdep prealloc as it may fsync and recurse * here. Instead we simply fail to reallocate blocks if this * rare condition arises. */ if (DOINGSOFTDEP(ap->a_vp)) if (softdep_prealloc(ap->a_vp, MNT_NOWAIT) != 0) return (ENOSPC); if (VTOI(ap->a_vp)->i_ump->um_fstype == UFS1) return (ffs_reallocblks_ufs1(ap)); return (ffs_reallocblks_ufs2(ap)); } static int ffs_reallocblks_ufs1(ap) struct vop_reallocblks_args /* { struct vnode *a_vp; struct cluster_save *a_buflist; } */ *ap; { struct fs *fs; struct inode *ip; struct vnode *vp; struct buf *sbp, *ebp; ufs1_daddr_t *bap, *sbap, *ebap = 0; struct cluster_save *buflist; struct ufsmount *ump; ufs_lbn_t start_lbn, end_lbn; ufs1_daddr_t soff, newblk, blkno; ufs2_daddr_t pref; struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp; - int i, len, start_lvl, end_lvl, ssize; + int i, cg, len, start_lvl, end_lvl, ssize; vp = ap->a_vp; ip = VTOI(vp); fs = ip->i_fs; ump = ip->i_ump; /* * If we are not tracking block clusters or if we have less than 4% * free blocks left, then do not attempt to cluster. Running with * less than 5% free block reserve is not recommended and those that * choose to do so do not expect to have good file layout. */ if (fs->fs_contigsumsize <= 0 || freespace(fs, 4) < 0) return (ENOSPC); buflist = ap->a_buflist; len = buflist->bs_nchildren; start_lbn = buflist->bs_children[0]->b_lblkno; end_lbn = start_lbn + len - 1; #ifdef INVARIANTS for (i = 0; i < len; i++) if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 1"); for (i = 1; i < len; i++) if (buflist->bs_children[i]->b_lblkno != start_lbn + i) panic("ffs_reallocblks: non-logical cluster"); blkno = buflist->bs_children[0]->b_blkno; ssize = fsbtodb(fs, fs->fs_frag); for (i = 1; i < len - 1; i++) if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize)) panic("ffs_reallocblks: non-physical cluster %d", i); #endif /* * If the cluster crosses the boundary for the first indirect * block, leave space for the indirect block. Indirect blocks * are initially laid out in a position after the last direct * block. Block reallocation would usually destroy locality by * moving the indirect block out of the way to make room for * data blocks if we didn't compensate here. We should also do * this for other indirect block boundaries, but it is only * important for the first one. */ if (start_lbn < NDADDR && end_lbn >= NDADDR) return (ENOSPC); /* * If the latest allocation is in a new cylinder group, assume that * the filesystem has decided to move and do not force it back to * the previous cylinder group. */ if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) != dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno))) return (ENOSPC); if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) || ufs_getlbns(vp, end_lbn, end_ap, &end_lvl)) return (ENOSPC); /* * Get the starting offset and block map for the first block. */ if (start_lvl == 0) { sbap = &ip->i_din1->di_db[0]; soff = start_lbn; } else { idp = &start_ap[start_lvl - 1]; if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) { brelse(sbp); return (ENOSPC); } sbap = (ufs1_daddr_t *)sbp->b_data; soff = idp->in_off; } /* * If the block range spans two block maps, get the second map. */ if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) { ssize = len; } else { #ifdef INVARIANTS if (start_lvl > 0 && start_ap[start_lvl - 1].in_lbn == idp->in_lbn) panic("ffs_reallocblk: start == end"); #endif ssize = len - (idp->in_off + 1); if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp)) goto fail; ebap = (ufs1_daddr_t *)ebp->b_data; } /* - * Find the preferred location for the cluster. + * Find the preferred location for the cluster. If we have not + * previously failed at this endeavor, then follow our standard + * preference calculation. If we have failed at it, then pick up + * where we last ended our search. */ UFS_LOCK(ump); - pref = ffs_blkpref_ufs1(ip, start_lbn, soff, sbap); + if (ip->i_nextclustercg == -1) + pref = ffs_blkpref_ufs1(ip, start_lbn, soff, sbap); + else + pref = cgdata(fs, ip->i_nextclustercg); /* * Search the block map looking for an allocation of the desired size. + * To avoid wasting too much time, we limit the number of cylinder + * groups that we will search. */ - if ((newblk = ffs_hashalloc(ip, dtog(fs, pref), pref, - len, len, ffs_clusteralloc)) == 0) { + cg = dtog(fs, pref); + for (i = min(maxclustersearch, fs->fs_ncg); i > 0; i--) { + if ((newblk = ffs_clusteralloc(ip, cg, pref, len)) != 0) + break; + cg += 1; + if (cg >= fs->fs_ncg) + cg = 0; + } + /* + * If we have failed in our search, record where we gave up for + * next time. Otherwise, fall back to our usual search citerion. + */ + if (newblk == 0) { + ip->i_nextclustercg = cg; UFS_UNLOCK(ump); goto fail; } + ip->i_nextclustercg = -1; /* * We have found a new contiguous block. * * First we have to replace the old block pointers with the new * block pointers in the inode and indirect blocks associated * with the file. */ #ifdef DEBUG if (prtrealloc) printf("realloc: ino %ju, lbns %jd-%jd\n\told:", (uintmax_t)ip->i_number, (intmax_t)start_lbn, (intmax_t)end_lbn); #endif blkno = newblk; for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) { if (i == ssize) { bap = ebap; soff = -i; } #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 2"); if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != *bap) panic("ffs_reallocblks: alloc mismatch"); #endif #ifdef DEBUG if (prtrealloc) printf(" %d,", *bap); #endif if (DOINGSOFTDEP(vp)) { if (sbap == &ip->i_din1->di_db[0] && i < ssize) softdep_setup_allocdirect(ip, start_lbn + i, blkno, *bap, fs->fs_bsize, fs->fs_bsize, buflist->bs_children[i]); else softdep_setup_allocindir_page(ip, start_lbn + i, i < ssize ? sbp : ebp, soff + i, blkno, *bap, buflist->bs_children[i]); } *bap++ = blkno; } /* * Next we must write out the modified inode and indirect blocks. * For strict correctness, the writes should be synchronous since * the old block values may have been written to disk. In practise * they are almost never written, but if we are concerned about * strict correctness, the `doasyncfree' flag should be set to zero. * * The test on `doasyncfree' should be changed to test a flag * that shows whether the associated buffers and inodes have * been written. The flag should be set when the cluster is * started and cleared whenever the buffer or inode is flushed. * We can then check below to see if it is set, and do the * synchronous write only when it has been cleared. */ if (sbap != &ip->i_din1->di_db[0]) { if (doasyncfree) bdwrite(sbp); else bwrite(sbp); } else { ip->i_flag |= IN_CHANGE | IN_UPDATE; if (!doasyncfree) ffs_update(vp, 1); } if (ssize < len) { if (doasyncfree) bdwrite(ebp); else bwrite(ebp); } /* * Last, free the old blocks and assign the new blocks to the buffers. */ #ifdef DEBUG if (prtrealloc) printf("\n\tnew:"); #endif for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) { if (!DOINGSOFTDEP(vp)) ffs_blkfree(ump, fs, ip->i_devvp, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize, ip->i_number, vp->v_type, NULL); buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno); #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 3"); #endif #ifdef DEBUG if (prtrealloc) printf(" %d,", blkno); #endif } #ifdef DEBUG if (prtrealloc) { prtrealloc--; printf("\n"); } #endif return (0); fail: if (ssize < len) brelse(ebp); if (sbap != &ip->i_din1->di_db[0]) brelse(sbp); return (ENOSPC); } static int ffs_reallocblks_ufs2(ap) struct vop_reallocblks_args /* { struct vnode *a_vp; struct cluster_save *a_buflist; } */ *ap; { struct fs *fs; struct inode *ip; struct vnode *vp; struct buf *sbp, *ebp; ufs2_daddr_t *bap, *sbap, *ebap = 0; struct cluster_save *buflist; struct ufsmount *ump; ufs_lbn_t start_lbn, end_lbn; ufs2_daddr_t soff, newblk, blkno, pref; struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp; - int i, len, start_lvl, end_lvl, ssize; + int i, cg, len, start_lvl, end_lvl, ssize; vp = ap->a_vp; ip = VTOI(vp); fs = ip->i_fs; ump = ip->i_ump; /* * If we are not tracking block clusters or if we have less than 4% * free blocks left, then do not attempt to cluster. Running with * less than 5% free block reserve is not recommended and those that * choose to do so do not expect to have good file layout. */ if (fs->fs_contigsumsize <= 0 || freespace(fs, 4) < 0) return (ENOSPC); buflist = ap->a_buflist; len = buflist->bs_nchildren; start_lbn = buflist->bs_children[0]->b_lblkno; end_lbn = start_lbn + len - 1; #ifdef INVARIANTS for (i = 0; i < len; i++) if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 1"); for (i = 1; i < len; i++) if (buflist->bs_children[i]->b_lblkno != start_lbn + i) panic("ffs_reallocblks: non-logical cluster"); blkno = buflist->bs_children[0]->b_blkno; ssize = fsbtodb(fs, fs->fs_frag); for (i = 1; i < len - 1; i++) if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize)) panic("ffs_reallocblks: non-physical cluster %d", i); #endif /* * If the cluster crosses the boundary for the first indirect * block, do not move anything in it. Indirect blocks are * usually initially laid out in a position between the data * blocks. Block reallocation would usually destroy locality by * moving the indirect block out of the way to make room for * data blocks if we didn't compensate here. We should also do * this for other indirect block boundaries, but it is only * important for the first one. */ if (start_lbn < NDADDR && end_lbn >= NDADDR) return (ENOSPC); /* * If the latest allocation is in a new cylinder group, assume that * the filesystem has decided to move and do not force it back to * the previous cylinder group. */ if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) != dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno))) return (ENOSPC); if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) || ufs_getlbns(vp, end_lbn, end_ap, &end_lvl)) return (ENOSPC); /* * Get the starting offset and block map for the first block. */ if (start_lvl == 0) { sbap = &ip->i_din2->di_db[0]; soff = start_lbn; } else { idp = &start_ap[start_lvl - 1]; if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) { brelse(sbp); return (ENOSPC); } sbap = (ufs2_daddr_t *)sbp->b_data; soff = idp->in_off; } /* * If the block range spans two block maps, get the second map. */ if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) { ssize = len; } else { #ifdef INVARIANTS if (start_lvl > 0 && start_ap[start_lvl - 1].in_lbn == idp->in_lbn) panic("ffs_reallocblk: start == end"); #endif ssize = len - (idp->in_off + 1); if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp)) goto fail; ebap = (ufs2_daddr_t *)ebp->b_data; } /* - * Find the preferred location for the cluster. + * Find the preferred location for the cluster. If we have not + * previously failed at this endeavor, then follow our standard + * preference calculation. If we have failed at it, then pick up + * where we last ended our search. */ UFS_LOCK(ump); - pref = ffs_blkpref_ufs2(ip, start_lbn, soff, sbap); + if (ip->i_nextclustercg == -1) + pref = ffs_blkpref_ufs2(ip, start_lbn, soff, sbap); + else + pref = cgdata(fs, ip->i_nextclustercg); /* * Search the block map looking for an allocation of the desired size. + * To avoid wasting too much time, we limit the number of cylinder + * groups that we will search. */ - if ((newblk = ffs_hashalloc(ip, dtog(fs, pref), pref, - len, len, ffs_clusteralloc)) == 0) { + cg = dtog(fs, pref); + for (i = min(maxclustersearch, fs->fs_ncg); i > 0; i--) { + if ((newblk = ffs_clusteralloc(ip, cg, pref, len)) != 0) + break; + cg += 1; + if (cg >= fs->fs_ncg) + cg = 0; + } + /* + * If we have failed in our search, record where we gave up for + * next time. Otherwise, fall back to our usual search citerion. + */ + if (newblk == 0) { + ip->i_nextclustercg = cg; UFS_UNLOCK(ump); goto fail; } + ip->i_nextclustercg = -1; /* * We have found a new contiguous block. * * First we have to replace the old block pointers with the new * block pointers in the inode and indirect blocks associated * with the file. */ #ifdef DEBUG if (prtrealloc) printf("realloc: ino %ju, lbns %jd-%jd\n\told:", (uintmax_t)ip->i_number, (intmax_t)start_lbn, (intmax_t)end_lbn); #endif blkno = newblk; for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) { if (i == ssize) { bap = ebap; soff = -i; } #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 2"); if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != *bap) panic("ffs_reallocblks: alloc mismatch"); #endif #ifdef DEBUG if (prtrealloc) printf(" %jd,", (intmax_t)*bap); #endif if (DOINGSOFTDEP(vp)) { if (sbap == &ip->i_din2->di_db[0] && i < ssize) softdep_setup_allocdirect(ip, start_lbn + i, blkno, *bap, fs->fs_bsize, fs->fs_bsize, buflist->bs_children[i]); else softdep_setup_allocindir_page(ip, start_lbn + i, i < ssize ? sbp : ebp, soff + i, blkno, *bap, buflist->bs_children[i]); } *bap++ = blkno; } /* * Next we must write out the modified inode and indirect blocks. * For strict correctness, the writes should be synchronous since * the old block values may have been written to disk. In practise * they are almost never written, but if we are concerned about * strict correctness, the `doasyncfree' flag should be set to zero. * * The test on `doasyncfree' should be changed to test a flag * that shows whether the associated buffers and inodes have * been written. The flag should be set when the cluster is * started and cleared whenever the buffer or inode is flushed. * We can then check below to see if it is set, and do the * synchronous write only when it has been cleared. */ if (sbap != &ip->i_din2->di_db[0]) { if (doasyncfree) bdwrite(sbp); else bwrite(sbp); } else { ip->i_flag |= IN_CHANGE | IN_UPDATE; if (!doasyncfree) ffs_update(vp, 1); } if (ssize < len) { if (doasyncfree) bdwrite(ebp); else bwrite(ebp); } /* * Last, free the old blocks and assign the new blocks to the buffers. */ #ifdef DEBUG if (prtrealloc) printf("\n\tnew:"); #endif for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) { if (!DOINGSOFTDEP(vp)) ffs_blkfree(ump, fs, ip->i_devvp, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize, ip->i_number, vp->v_type, NULL); buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno); #ifdef INVARIANTS if (!ffs_checkblk(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize)) panic("ffs_reallocblks: unallocated block 3"); #endif #ifdef DEBUG if (prtrealloc) printf(" %jd,", (intmax_t)blkno); #endif } #ifdef DEBUG if (prtrealloc) { prtrealloc--; printf("\n"); } #endif return (0); fail: if (ssize < len) brelse(ebp); if (sbap != &ip->i_din2->di_db[0]) brelse(sbp); return (ENOSPC); } /* * Allocate an inode in the filesystem. * * If allocating a directory, use ffs_dirpref to select the inode. * If allocating in a directory, the following hierarchy is followed: * 1) allocate the preferred inode. * 2) allocate an inode in the same cylinder group. * 3) quadradically rehash into other cylinder groups, until an * available inode is located. * If no inode preference is given the following hierarchy is used * to allocate an inode: * 1) allocate an inode in cylinder group 0. * 2) quadradically rehash into other cylinder groups, until an * available inode is located. */ int ffs_valloc(pvp, mode, cred, vpp) struct vnode *pvp; int mode; struct ucred *cred; struct vnode **vpp; { struct inode *pip; struct fs *fs; struct inode *ip; struct timespec ts; struct ufsmount *ump; ino_t ino, ipref; u_int cg; int error, error1, reclaimed; static struct timeval lastfail; static int curfail; *vpp = NULL; pip = VTOI(pvp); fs = pip->i_fs; ump = pip->i_ump; UFS_LOCK(ump); reclaimed = 0; retry: if (fs->fs_cstotal.cs_nifree == 0) goto noinodes; if ((mode & IFMT) == IFDIR) ipref = ffs_dirpref(pip); else ipref = pip->i_number; if (ipref >= fs->fs_ncg * fs->fs_ipg) ipref = 0; cg = ino_to_cg(fs, ipref); /* * Track number of dirs created one after another * in a same cg without intervening by files. */ if ((mode & IFMT) == IFDIR) { if (fs->fs_contigdirs[cg] < 255) fs->fs_contigdirs[cg]++; } else { if (fs->fs_contigdirs[cg] > 0) fs->fs_contigdirs[cg]--; } ino = (ino_t)ffs_hashalloc(pip, cg, ipref, mode, 0, (allocfcn_t *)ffs_nodealloccg); if (ino == 0) goto noinodes; error = ffs_vget(pvp->v_mount, ino, LK_EXCLUSIVE, vpp); if (error) { error1 = ffs_vgetf(pvp->v_mount, ino, LK_EXCLUSIVE, vpp, FFSV_FORCEINSMQ); ffs_vfree(pvp, ino, mode); if (error1 == 0) { ip = VTOI(*vpp); if (ip->i_mode) goto dup_alloc; ip->i_flag |= IN_MODIFIED; vput(*vpp); } return (error); } ip = VTOI(*vpp); if (ip->i_mode) { dup_alloc: printf("mode = 0%o, inum = %ju, fs = %s\n", ip->i_mode, (uintmax_t)ip->i_number, fs->fs_fsmnt); panic("ffs_valloc: dup alloc"); } if (DIP(ip, i_blocks) && (fs->fs_flags & FS_UNCLEAN) == 0) { /* XXX */ printf("free inode %s/%lu had %ld blocks\n", fs->fs_fsmnt, (u_long)ino, (long)DIP(ip, i_blocks)); DIP_SET(ip, i_blocks, 0); } ip->i_flags = 0; DIP_SET(ip, i_flags, 0); /* * Set up a new generation number for this inode. */ if (ip->i_gen == 0 || ++ip->i_gen == 0) ip->i_gen = arc4random() / 2 + 1; DIP_SET(ip, i_gen, ip->i_gen); if (fs->fs_magic == FS_UFS2_MAGIC) { vfs_timestamp(&ts); ip->i_din2->di_birthtime = ts.tv_sec; ip->i_din2->di_birthnsec = ts.tv_nsec; } ufs_prepare_reclaim(*vpp); ip->i_flag = 0; (*vpp)->v_vflag = 0; (*vpp)->v_type = VNON; if (fs->fs_magic == FS_UFS2_MAGIC) (*vpp)->v_op = &ffs_vnodeops2; else (*vpp)->v_op = &ffs_vnodeops1; return (0); noinodes: if (reclaimed == 0) { reclaimed = 1; softdep_request_cleanup(fs, pvp, cred, FLUSH_INODES_WAIT); goto retry; } UFS_UNLOCK(ump); if (ppsratecheck(&lastfail, &curfail, 1)) { ffs_fserr(fs, pip->i_number, "out of inodes"); uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt); } return (ENOSPC); } /* * Find a cylinder group to place a directory. * * The policy implemented by this algorithm is to allocate a * directory inode in the same cylinder group as its parent * directory, but also to reserve space for its files inodes * and data. Restrict the number of directories which may be * allocated one after another in the same cylinder group * without intervening allocation of files. * * If we allocate a first level directory then force allocation * in another cylinder group. */ static ino_t ffs_dirpref(pip) struct inode *pip; { struct fs *fs; int cg, prefcg, dirsize, cgsize; u_int avgifree, avgbfree, avgndir, curdirsize; u_int minifree, minbfree, maxndir; u_int mincg, minndir; u_int maxcontigdirs; mtx_assert(UFS_MTX(pip->i_ump), MA_OWNED); fs = pip->i_fs; avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg; avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; avgndir = fs->fs_cstotal.cs_ndir / fs->fs_ncg; /* * Force allocation in another cg if creating a first level dir. */ ASSERT_VOP_LOCKED(ITOV(pip), "ffs_dirpref"); if (ITOV(pip)->v_vflag & VV_ROOT) { prefcg = arc4random() % fs->fs_ncg; mincg = prefcg; minndir = fs->fs_ipg; for (cg = prefcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_ndir < minndir && fs->fs_cs(fs, cg).cs_nifree >= avgifree && fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { mincg = cg; minndir = fs->fs_cs(fs, cg).cs_ndir; } for (cg = 0; cg < prefcg; cg++) if (fs->fs_cs(fs, cg).cs_ndir < minndir && fs->fs_cs(fs, cg).cs_nifree >= avgifree && fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { mincg = cg; minndir = fs->fs_cs(fs, cg).cs_ndir; } return ((ino_t)(fs->fs_ipg * mincg)); } /* * Count various limits which used for * optimal allocation of a directory inode. */ maxndir = min(avgndir + fs->fs_ipg / 16, fs->fs_ipg); minifree = avgifree - avgifree / 4; if (minifree < 1) minifree = 1; minbfree = avgbfree - avgbfree / 4; if (minbfree < 1) minbfree = 1; cgsize = fs->fs_fsize * fs->fs_fpg; dirsize = fs->fs_avgfilesize * fs->fs_avgfpdir; curdirsize = avgndir ? (cgsize - avgbfree * fs->fs_bsize) / avgndir : 0; if (dirsize < curdirsize) dirsize = curdirsize; if (dirsize <= 0) maxcontigdirs = 0; /* dirsize overflowed */ else maxcontigdirs = min((avgbfree * fs->fs_bsize) / dirsize, 255); if (fs->fs_avgfpdir > 0) maxcontigdirs = min(maxcontigdirs, fs->fs_ipg / fs->fs_avgfpdir); if (maxcontigdirs == 0) maxcontigdirs = 1; /* * Limit number of dirs in one cg and reserve space for * regular files, but only if we have no deficit in * inodes or space. * * We are trying to find a suitable cylinder group nearby * our preferred cylinder group to place a new directory. * We scan from our preferred cylinder group forward looking * for a cylinder group that meets our criterion. If we get * to the final cylinder group and do not find anything, * we start scanning forwards from the beginning of the * filesystem. While it might seem sensible to start scanning * backwards or even to alternate looking forward and backward, * this approach fails badly when the filesystem is nearly full. * Specifically, we first search all the areas that have no space * and finally try the one preceeding that. We repeat this on * every request and in the case of the final block end up * searching the entire filesystem. By jumping to the front * of the filesystem, our future forward searches always look * in new cylinder groups so finds every possible block after * one pass over the filesystem. */ prefcg = ino_to_cg(fs, pip->i_number); for (cg = prefcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_ndir < maxndir && fs->fs_cs(fs, cg).cs_nifree >= minifree && fs->fs_cs(fs, cg).cs_nbfree >= minbfree) { if (fs->fs_contigdirs[cg] < maxcontigdirs) return ((ino_t)(fs->fs_ipg * cg)); } for (cg = 0; cg < prefcg; cg++) if (fs->fs_cs(fs, cg).cs_ndir < maxndir && fs->fs_cs(fs, cg).cs_nifree >= minifree && fs->fs_cs(fs, cg).cs_nbfree >= minbfree) { if (fs->fs_contigdirs[cg] < maxcontigdirs) return ((ino_t)(fs->fs_ipg * cg)); } /* * This is a backstop when we have deficit in space. */ for (cg = prefcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_nifree >= avgifree) return ((ino_t)(fs->fs_ipg * cg)); for (cg = 0; cg < prefcg; cg++) if (fs->fs_cs(fs, cg).cs_nifree >= avgifree) break; return ((ino_t)(fs->fs_ipg * cg)); } /* * Select the desired position for the next block in a file. The file is * logically divided into sections. The first section is composed of the * direct blocks and the next fs_maxbpg blocks. Each additional section * contains fs_maxbpg blocks. * * If no blocks have been allocated in the first section, the policy is to * request a block in the same cylinder group as the inode that describes * the file. The first indirect is allocated immediately following the last * direct block and the data blocks for the first indirect immediately * follow it. * * If no blocks have been allocated in any other section, the indirect * block(s) are allocated in the same cylinder group as its inode in an * area reserved immediately following the inode blocks. The policy for * the data blocks is to place them in a cylinder group with a greater than * average number of free blocks. An appropriate cylinder group is found * by using a rotor that sweeps the cylinder groups. When a new group of * blocks is needed, the sweep begins in the cylinder group following the * cylinder group from which the previous allocation was made. The sweep * continues until a cylinder group with greater than the average number * of free blocks is found. If the allocation is for the first block in an * indirect block or the previous block is a hole, then the information on * the previous allocation is unavailable; here a best guess is made based * on the logical block number being allocated. * * If a section is already partially allocated, the policy is to * allocate blocks contiguously within the section if possible. */ ufs2_daddr_t ffs_blkpref_ufs1(ip, lbn, indx, bap) struct inode *ip; ufs_lbn_t lbn; int indx; ufs1_daddr_t *bap; { struct fs *fs; u_int cg, inocg; u_int avgbfree, startcg; ufs2_daddr_t pref; KASSERT(indx <= 0 || bap != NULL, ("need non-NULL bap")); mtx_assert(UFS_MTX(ip->i_ump), MA_OWNED); fs = ip->i_fs; /* * Allocation of indirect blocks is indicated by passing negative * values in indx: -1 for single indirect, -2 for double indirect, * -3 for triple indirect. As noted below, we attempt to allocate * the first indirect inline with the file data. For all later * indirect blocks, the data is often allocated in other cylinder * groups. However to speed random file access and to speed up * fsck, the filesystem reserves the first fs_metaspace blocks * (typically half of fs_minfree) of the data area of each cylinder * group to hold these later indirect blocks. */ inocg = ino_to_cg(fs, ip->i_number); if (indx < 0) { /* * Our preference for indirect blocks is the zone at the * beginning of the inode's cylinder group data area that * we try to reserve for indirect blocks. */ pref = cgmeta(fs, inocg); /* * If we are allocating the first indirect block, try to * place it immediately following the last direct block. */ if (indx == -1 && lbn < NDADDR + NINDIR(fs) && ip->i_din1->di_db[NDADDR - 1] != 0) pref = ip->i_din1->di_db[NDADDR - 1] + fs->fs_frag; return (pref); } /* * If we are allocating the first data block in the first indirect * block and the indirect has been allocated in the data block area, * try to place it immediately following the indirect block. */ if (lbn == NDADDR) { pref = ip->i_din1->di_ib[0]; if (pref != 0 && pref >= cgdata(fs, inocg) && pref < cgbase(fs, inocg + 1)) return (pref + fs->fs_frag); } /* * If we are at the beginning of a file, or we have already allocated * the maximum number of blocks per cylinder group, or we do not * have a block allocated immediately preceeding us, then we need * to decide where to start allocating new blocks. */ if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) { /* * If we are allocating a directory data block, we want * to place it in the metadata area. */ if ((ip->i_mode & IFMT) == IFDIR) return (cgmeta(fs, inocg)); /* * Until we fill all the direct and all the first indirect's * blocks, we try to allocate in the data area of the inode's * cylinder group. */ if (lbn < NDADDR + NINDIR(fs)) return (cgdata(fs, inocg)); /* * Find a cylinder with greater than average number of * unused data blocks. */ if (indx == 0 || bap[indx - 1] == 0) startcg = inocg + lbn / fs->fs_maxbpg; else startcg = dtog(fs, bap[indx - 1]) + 1; startcg %= fs->fs_ncg; avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; for (cg = startcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } for (cg = 0; cg <= startcg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } return (0); } /* * Otherwise, we just always try to lay things out contiguously. */ return (bap[indx - 1] + fs->fs_frag); } /* * Same as above, but for UFS2 */ ufs2_daddr_t ffs_blkpref_ufs2(ip, lbn, indx, bap) struct inode *ip; ufs_lbn_t lbn; int indx; ufs2_daddr_t *bap; { struct fs *fs; u_int cg, inocg; u_int avgbfree, startcg; ufs2_daddr_t pref; KASSERT(indx <= 0 || bap != NULL, ("need non-NULL bap")); mtx_assert(UFS_MTX(ip->i_ump), MA_OWNED); fs = ip->i_fs; /* * Allocation of indirect blocks is indicated by passing negative * values in indx: -1 for single indirect, -2 for double indirect, * -3 for triple indirect. As noted below, we attempt to allocate * the first indirect inline with the file data. For all later * indirect blocks, the data is often allocated in other cylinder * groups. However to speed random file access and to speed up * fsck, the filesystem reserves the first fs_metaspace blocks * (typically half of fs_minfree) of the data area of each cylinder * group to hold these later indirect blocks. */ inocg = ino_to_cg(fs, ip->i_number); if (indx < 0) { /* * Our preference for indirect blocks is the zone at the * beginning of the inode's cylinder group data area that * we try to reserve for indirect blocks. */ pref = cgmeta(fs, inocg); /* * If we are allocating the first indirect block, try to * place it immediately following the last direct block. */ if (indx == -1 && lbn < NDADDR + NINDIR(fs) && ip->i_din2->di_db[NDADDR - 1] != 0) pref = ip->i_din2->di_db[NDADDR - 1] + fs->fs_frag; return (pref); } /* * If we are allocating the first data block in the first indirect * block and the indirect has been allocated in the data block area, * try to place it immediately following the indirect block. */ if (lbn == NDADDR) { pref = ip->i_din2->di_ib[0]; if (pref != 0 && pref >= cgdata(fs, inocg) && pref < cgbase(fs, inocg + 1)) return (pref + fs->fs_frag); } /* * If we are at the beginning of a file, or we have already allocated * the maximum number of blocks per cylinder group, or we do not * have a block allocated immediately preceeding us, then we need * to decide where to start allocating new blocks. */ if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) { /* * If we are allocating a directory data block, we want * to place it in the metadata area. */ if ((ip->i_mode & IFMT) == IFDIR) return (cgmeta(fs, inocg)); /* * Until we fill all the direct and all the first indirect's * blocks, we try to allocate in the data area of the inode's * cylinder group. */ if (lbn < NDADDR + NINDIR(fs)) return (cgdata(fs, inocg)); /* * Find a cylinder with greater than average number of * unused data blocks. */ if (indx == 0 || bap[indx - 1] == 0) startcg = inocg + lbn / fs->fs_maxbpg; else startcg = dtog(fs, bap[indx - 1]) + 1; startcg %= fs->fs_ncg; avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; for (cg = startcg; cg < fs->fs_ncg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } for (cg = 0; cg <= startcg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (cgdata(fs, cg)); } return (0); } /* * Otherwise, we just always try to lay things out contiguously. */ return (bap[indx - 1] + fs->fs_frag); } /* * Implement the cylinder overflow algorithm. * * The policy implemented by this algorithm is: * 1) allocate the block in its requested cylinder group. * 2) quadradically rehash on the cylinder group number. * 3) brute force search for a free block. * * Must be called with the UFS lock held. Will release the lock on success * and return with it held on failure. */ /*VARARGS5*/ static ufs2_daddr_t ffs_hashalloc(ip, cg, pref, size, rsize, allocator) struct inode *ip; u_int cg; ufs2_daddr_t pref; int size; /* Search size for data blocks, mode for inodes */ int rsize; /* Real allocated size. */ allocfcn_t *allocator; { struct fs *fs; ufs2_daddr_t result; u_int i, icg = cg; mtx_assert(UFS_MTX(ip->i_ump), MA_OWNED); #ifdef INVARIANTS if (ITOV(ip)->v_mount->mnt_kern_flag & MNTK_SUSPENDED) panic("ffs_hashalloc: allocation on suspended filesystem"); #endif fs = ip->i_fs; /* * 1: preferred cylinder group */ result = (*allocator)(ip, cg, pref, size, rsize); if (result) return (result); /* * 2: quadratic rehash */ for (i = 1; i < fs->fs_ncg; i *= 2) { cg += i; if (cg >= fs->fs_ncg) cg -= fs->fs_ncg; result = (*allocator)(ip, cg, 0, size, rsize); if (result) return (result); } /* * 3: brute force search * Note that we start at i == 2, since 0 was checked initially, * and 1 is always checked in the quadratic rehash. */ cg = (icg + 2) % fs->fs_ncg; for (i = 2; i < fs->fs_ncg; i++) { result = (*allocator)(ip, cg, 0, size, rsize); if (result) return (result); cg++; if (cg == fs->fs_ncg) cg = 0; } return (0); } /* * Determine whether a fragment can be extended. * * Check to see if the necessary fragments are available, and * if they are, allocate them. */ static ufs2_daddr_t ffs_fragextend(ip, cg, bprev, osize, nsize) struct inode *ip; u_int cg; ufs2_daddr_t bprev; int osize, nsize; { struct fs *fs; struct cg *cgp; struct buf *bp; struct ufsmount *ump; int nffree; long bno; int frags, bbase; int i, error; u_int8_t *blksfree; ump = ip->i_ump; fs = ip->i_fs; if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize)) return (0); frags = numfrags(fs, nsize); bbase = fragnum(fs, bprev); if (bbase > fragnum(fs, (bprev + frags - 1))) { /* cannot extend across a block boundary */ return (0); } UFS_UNLOCK(ump); error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, NOCRED, &bp); if (error) goto fail; cgp = (struct cg *)bp->b_data; if (!cg_chkmagic(cgp)) goto fail; bp->b_xflags |= BX_BKGRDWRITE; cgp->cg_old_time = cgp->cg_time = time_second; bno = dtogd(fs, bprev); blksfree = cg_blksfree(cgp); for (i = numfrags(fs, osize); i < frags; i++) if (isclr(blksfree, bno + i)) goto fail; /* * the current fragment can be extended * deduct the count on fragment being extended into * increase the count on the remaining fragment (if any) * allocate the extended piece */ for (i = frags; i < fs->fs_frag - bbase; i++) if (isclr(blksfree, bno + i)) break; cgp->cg_frsum[i - numfrags(fs, osize)]--; if (i != frags) cgp->cg_frsum[i - frags]++; for (i = numfrags(fs, osize), nffree = 0; i < frags; i++) { clrbit(blksfree, bno + i); cgp->cg_cs.cs_nffree--; nffree++; } UFS_LOCK(ump); fs->fs_cstotal.cs_nffree -= nffree; fs->fs_cs(fs, cg).cs_nffree -= nffree; fs->fs_fmod = 1; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_blkmapdep(bp, UFSTOVFS(ump), bprev, frags, numfrags(fs, osize)); bdwrite(bp); return (bprev); fail: brelse(bp); UFS_LOCK(ump); return (0); } /* * Determine whether a block can be allocated. * * Check to see if a block of the appropriate size is available, * and if it is, allocate it. */ static ufs2_daddr_t ffs_alloccg(ip, cg, bpref, size, rsize) struct inode *ip; u_int cg; ufs2_daddr_t bpref; int size; int rsize; { struct fs *fs; struct cg *cgp; struct buf *bp; struct ufsmount *ump; ufs1_daddr_t bno; ufs2_daddr_t blkno; int i, allocsiz, error, frags; u_int8_t *blksfree; ump = ip->i_ump; fs = ip->i_fs; if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize) return (0); UFS_UNLOCK(ump); error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, NOCRED, &bp); if (error) goto fail; cgp = (struct cg *)bp->b_data; if (!cg_chkmagic(cgp) || (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) goto fail; bp->b_xflags |= BX_BKGRDWRITE; cgp->cg_old_time = cgp->cg_time = time_second; if (size == fs->fs_bsize) { UFS_LOCK(ump); blkno = ffs_alloccgblk(ip, bp, bpref, rsize); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); bdwrite(bp); return (blkno); } /* * check to see if any fragments are already available * allocsiz is the size which will be allocated, hacking * it down to a smaller size if necessary */ blksfree = cg_blksfree(cgp); frags = numfrags(fs, size); for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++) if (cgp->cg_frsum[allocsiz] != 0) break; if (allocsiz == fs->fs_frag) { /* * no fragments were available, so a block will be * allocated, and hacked up */ if (cgp->cg_cs.cs_nbfree == 0) goto fail; UFS_LOCK(ump); blkno = ffs_alloccgblk(ip, bp, bpref, rsize); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); bdwrite(bp); return (blkno); } KASSERT(size == rsize, ("ffs_alloccg: size(%d) != rsize(%d)", size, rsize)); bno = ffs_mapsearch(fs, cgp, bpref, allocsiz); if (bno < 0) goto fail; for (i = 0; i < frags; i++) clrbit(blksfree, bno + i); cgp->cg_cs.cs_nffree -= frags; cgp->cg_frsum[allocsiz]--; if (frags != allocsiz) cgp->cg_frsum[allocsiz - frags]++; UFS_LOCK(ump); fs->fs_cstotal.cs_nffree -= frags; fs->fs_cs(fs, cg).cs_nffree -= frags; fs->fs_fmod = 1; blkno = cgbase(fs, cg) + bno; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_blkmapdep(bp, UFSTOVFS(ump), blkno, frags, 0); bdwrite(bp); return (blkno); fail: brelse(bp); UFS_LOCK(ump); return (0); } /* * Allocate a block in a cylinder group. * * This algorithm implements the following policy: * 1) allocate the requested block. * 2) allocate a rotationally optimal block in the same cylinder. * 3) allocate the next available block on the block rotor for the * specified cylinder group. * Note that this routine only allocates fs_bsize blocks; these * blocks may be fragmented by the routine that allocates them. */ static ufs2_daddr_t ffs_alloccgblk(ip, bp, bpref, size) struct inode *ip; struct buf *bp; ufs2_daddr_t bpref; int size; { struct fs *fs; struct cg *cgp; struct ufsmount *ump; ufs1_daddr_t bno; ufs2_daddr_t blkno; u_int8_t *blksfree; int i, cgbpref; fs = ip->i_fs; ump = ip->i_ump; mtx_assert(UFS_MTX(ump), MA_OWNED); cgp = (struct cg *)bp->b_data; blksfree = cg_blksfree(cgp); if (bpref == 0) { bpref = cgbase(fs, cgp->cg_cgx) + cgp->cg_rotor + fs->fs_frag; } else if ((cgbpref = dtog(fs, bpref)) != cgp->cg_cgx) { /* map bpref to correct zone in this cg */ if (bpref < cgdata(fs, cgbpref)) bpref = cgmeta(fs, cgp->cg_cgx); else bpref = cgdata(fs, cgp->cg_cgx); } /* * if the requested block is available, use it */ bno = dtogd(fs, blknum(fs, bpref)); if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno))) goto gotit; /* * Take the next available block in this cylinder group. */ bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag); if (bno < 0) return (0); /* Update cg_rotor only if allocated from the data zone */ if (bno >= dtogd(fs, cgdata(fs, cgp->cg_cgx))) cgp->cg_rotor = bno; gotit: blkno = fragstoblks(fs, bno); ffs_clrblock(fs, blksfree, (long)blkno); ffs_clusteracct(fs, cgp, blkno, -1); cgp->cg_cs.cs_nbfree--; fs->fs_cstotal.cs_nbfree--; fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--; fs->fs_fmod = 1; blkno = cgbase(fs, cgp->cg_cgx) + bno; /* * If the caller didn't want the whole block free the frags here. */ size = numfrags(fs, size); if (size != fs->fs_frag) { bno = dtogd(fs, blkno); for (i = size; i < fs->fs_frag; i++) setbit(blksfree, bno + i); i = fs->fs_frag - size; cgp->cg_cs.cs_nffree += i; fs->fs_cstotal.cs_nffree += i; fs->fs_cs(fs, cgp->cg_cgx).cs_nffree += i; fs->fs_fmod = 1; cgp->cg_frsum[i]++; } /* XXX Fixme. */ UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_blkmapdep(bp, UFSTOVFS(ump), blkno, size, 0); UFS_LOCK(ump); return (blkno); } /* * Determine whether a cluster can be allocated. * * We do not currently check for optimal rotational layout if there * are multiple choices in the same cylinder group. Instead we just * take the first one that we find following bpref. */ static ufs2_daddr_t -ffs_clusteralloc(ip, cg, bpref, len, unused) +ffs_clusteralloc(ip, cg, bpref, len) struct inode *ip; u_int cg; ufs2_daddr_t bpref; int len; - int unused; { struct fs *fs; struct cg *cgp; struct buf *bp; struct ufsmount *ump; int i, run, bit, map, got; ufs2_daddr_t bno; u_char *mapp; int32_t *lp; u_int8_t *blksfree; fs = ip->i_fs; ump = ip->i_ump; if (fs->fs_maxcluster[cg] < len) return (0); UFS_UNLOCK(ump); if (bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, NOCRED, &bp)) goto fail_lock; cgp = (struct cg *)bp->b_data; if (!cg_chkmagic(cgp)) goto fail_lock; bp->b_xflags |= BX_BKGRDWRITE; /* * Check to see if a cluster of the needed size (or bigger) is * available in this cylinder group. */ lp = &cg_clustersum(cgp)[len]; for (i = len; i <= fs->fs_contigsumsize; i++) if (*lp++ > 0) break; if (i > fs->fs_contigsumsize) { /* * This is the first time looking for a cluster in this * cylinder group. Update the cluster summary information * to reflect the true maximum sized cluster so that * future cluster allocation requests can avoid reading * the cylinder group map only to find no clusters. */ lp = &cg_clustersum(cgp)[len - 1]; for (i = len - 1; i > 0; i--) if (*lp-- > 0) break; UFS_LOCK(ump); fs->fs_maxcluster[cg] = i; goto fail; } /* * Search the cluster map to find a big enough cluster. * We take the first one that we find, even if it is larger * than we need as we prefer to get one close to the previous * block allocation. We do not search before the current * preference point as we do not want to allocate a block * that is allocated before the previous one (as we will * then have to wait for another pass of the elevator * algorithm before it will be read). We prefer to fail and * be recalled to try an allocation in the next cylinder group. */ if (dtog(fs, bpref) != cg) bpref = cgdata(fs, cg); else bpref = blknum(fs, bpref); bpref = fragstoblks(fs, dtogd(fs, bpref)); mapp = &cg_clustersfree(cgp)[bpref / NBBY]; map = *mapp++; bit = 1 << (bpref % NBBY); for (run = 0, got = bpref; got < cgp->cg_nclusterblks; got++) { if ((map & bit) == 0) { run = 0; } else { run++; if (run == len) break; } if ((got & (NBBY - 1)) != (NBBY - 1)) { bit <<= 1; } else { map = *mapp++; bit = 1; } } if (got >= cgp->cg_nclusterblks) goto fail_lock; /* * Allocate the cluster that we have found. */ blksfree = cg_blksfree(cgp); for (i = 1; i <= len; i++) if (!ffs_isblock(fs, blksfree, got - run + i)) panic("ffs_clusteralloc: map mismatch"); bno = cgbase(fs, cg) + blkstofrags(fs, got - run + 1); if (dtog(fs, bno) != cg) panic("ffs_clusteralloc: allocated out of group"); len = blkstofrags(fs, len); UFS_LOCK(ump); for (i = 0; i < len; i += fs->fs_frag) if (ffs_alloccgblk(ip, bp, bno + i, fs->fs_bsize) != bno + i) panic("ffs_clusteralloc: lost block"); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); bdwrite(bp); return (bno); fail_lock: UFS_LOCK(ump); fail: brelse(bp); return (0); } static inline struct buf * getinobuf(struct inode *ip, u_int cg, u_int32_t cginoblk, int gbflags) { struct fs *fs; fs = ip->i_fs; return (getblk(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, cg * fs->fs_ipg + cginoblk)), (int)fs->fs_bsize, 0, 0, gbflags)); } /* * Determine whether an inode can be allocated. * * Check to see if an inode is available, and if it is, * allocate it using the following policy: * 1) allocate the requested inode. * 2) allocate the next available inode after the requested * inode in the specified cylinder group. */ static ufs2_daddr_t ffs_nodealloccg(ip, cg, ipref, mode, unused) struct inode *ip; u_int cg; ufs2_daddr_t ipref; int mode; int unused; { struct fs *fs; struct cg *cgp; struct buf *bp, *ibp; struct ufsmount *ump; u_int8_t *inosused, *loc; struct ufs2_dinode *dp2; int error, start, len, i; u_int32_t old_initediblk; fs = ip->i_fs; ump = ip->i_ump; check_nifree: if (fs->fs_cs(fs, cg).cs_nifree == 0) return (0); UFS_UNLOCK(ump); error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, NOCRED, &bp); if (error) { brelse(bp); UFS_LOCK(ump); return (0); } cgp = (struct cg *)bp->b_data; restart: if (!cg_chkmagic(cgp) || cgp->cg_cs.cs_nifree == 0) { brelse(bp); UFS_LOCK(ump); return (0); } bp->b_xflags |= BX_BKGRDWRITE; inosused = cg_inosused(cgp); if (ipref) { ipref %= fs->fs_ipg; if (isclr(inosused, ipref)) goto gotit; } start = cgp->cg_irotor / NBBY; len = howmany(fs->fs_ipg - cgp->cg_irotor, NBBY); loc = memcchr(&inosused[start], 0xff, len); if (loc == NULL) { len = start + 1; start = 0; loc = memcchr(&inosused[start], 0xff, len); if (loc == NULL) { printf("cg = %d, irotor = %ld, fs = %s\n", cg, (long)cgp->cg_irotor, fs->fs_fsmnt); panic("ffs_nodealloccg: map corrupted"); /* NOTREACHED */ } } ipref = (loc - inosused) * NBBY + ffs(~*loc) - 1; gotit: /* * Check to see if we need to initialize more inodes. */ if (fs->fs_magic == FS_UFS2_MAGIC && ipref + INOPB(fs) > cgp->cg_initediblk && cgp->cg_initediblk < cgp->cg_niblk) { old_initediblk = cgp->cg_initediblk; /* * Free the cylinder group lock before writing the * initialized inode block. Entering the * babarrierwrite() with the cylinder group lock * causes lock order violation between the lock and * snaplk. * * Another thread can decide to initialize the same * inode block, but whichever thread first gets the * cylinder group lock after writing the newly * allocated inode block will update it and the other * will realize that it has lost and leave the * cylinder group unchanged. */ ibp = getinobuf(ip, cg, old_initediblk, GB_LOCK_NOWAIT); brelse(bp); if (ibp == NULL) { /* * The inode block buffer is already owned by * another thread, which must initialize it. * Wait on the buffer to allow another thread * to finish the updates, with dropped cg * buffer lock, then retry. */ ibp = getinobuf(ip, cg, old_initediblk, 0); brelse(ibp); UFS_LOCK(ump); goto check_nifree; } bzero(ibp->b_data, (int)fs->fs_bsize); dp2 = (struct ufs2_dinode *)(ibp->b_data); for (i = 0; i < INOPB(fs); i++) { dp2->di_gen = arc4random() / 2 + 1; dp2++; } /* * Rather than adding a soft updates dependency to ensure * that the new inode block is written before it is claimed * by the cylinder group map, we just do a barrier write * here. The barrier write will ensure that the inode block * gets written before the updated cylinder group map can be * written. The barrier write should only slow down bulk * loading of newly created filesystems. */ babarrierwrite(ibp); /* * After the inode block is written, try to update the * cg initediblk pointer. If another thread beat us * to it, then leave it unchanged as the other thread * has already set it correctly. */ error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, NOCRED, &bp); UFS_LOCK(ump); ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (error != 0) { brelse(bp); return (error); } cgp = (struct cg *)bp->b_data; if (cgp->cg_initediblk == old_initediblk) cgp->cg_initediblk += INOPB(fs); goto restart; } cgp->cg_old_time = cgp->cg_time = time_second; cgp->cg_irotor = ipref; UFS_LOCK(ump); ACTIVECLEAR(fs, cg); setbit(inosused, ipref); cgp->cg_cs.cs_nifree--; fs->fs_cstotal.cs_nifree--; fs->fs_cs(fs, cg).cs_nifree--; fs->fs_fmod = 1; if ((mode & IFMT) == IFDIR) { cgp->cg_cs.cs_ndir++; fs->fs_cstotal.cs_ndir++; fs->fs_cs(fs, cg).cs_ndir++; } UFS_UNLOCK(ump); if (DOINGSOFTDEP(ITOV(ip))) softdep_setup_inomapdep(bp, ip, cg * fs->fs_ipg + ipref, mode); bdwrite(bp); return ((ino_t)(cg * fs->fs_ipg + ipref)); } /* * Free a block or fragment. * * The specified block or fragment is placed back in the * free map. If a fragment is deallocated, a possible * block reassembly is checked. */ static void ffs_blkfree_cg(ump, fs, devvp, bno, size, inum, dephd) struct ufsmount *ump; struct fs *fs; struct vnode *devvp; ufs2_daddr_t bno; long size; ino_t inum; struct workhead *dephd; { struct mount *mp; struct cg *cgp; struct buf *bp; ufs1_daddr_t fragno, cgbno; ufs2_daddr_t cgblkno; int i, blk, frags, bbase; u_int cg; u_int8_t *blksfree; struct cdev *dev; cg = dtog(fs, bno); if (devvp->v_type == VREG) { /* devvp is a snapshot */ dev = VTOI(devvp)->i_devvp->v_rdev; cgblkno = fragstoblks(fs, cgtod(fs, cg)); } else { /* devvp is a normal disk device */ dev = devvp->v_rdev; cgblkno = fsbtodb(fs, cgtod(fs, cg)); ASSERT_VOP_LOCKED(devvp, "ffs_blkfree_cg"); } #ifdef INVARIANTS if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 || fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) { printf("dev=%s, bno = %jd, bsize = %ld, size = %ld, fs = %s\n", devtoname(dev), (intmax_t)bno, (long)fs->fs_bsize, size, fs->fs_fsmnt); panic("ffs_blkfree_cg: bad size"); } #endif if ((u_int)bno >= fs->fs_size) { printf("bad block %jd, ino %lu\n", (intmax_t)bno, (u_long)inum); ffs_fserr(fs, inum, "bad block"); return; } if (bread(devvp, cgblkno, (int)fs->fs_cgsize, NOCRED, &bp)) { brelse(bp); return; } cgp = (struct cg *)bp->b_data; if (!cg_chkmagic(cgp)) { brelse(bp); return; } bp->b_xflags |= BX_BKGRDWRITE; cgp->cg_old_time = cgp->cg_time = time_second; cgbno = dtogd(fs, bno); blksfree = cg_blksfree(cgp); UFS_LOCK(ump); if (size == fs->fs_bsize) { fragno = fragstoblks(fs, cgbno); if (!ffs_isfreeblock(fs, blksfree, fragno)) { if (devvp->v_type == VREG) { UFS_UNLOCK(ump); /* devvp is a snapshot */ brelse(bp); return; } printf("dev = %s, block = %jd, fs = %s\n", devtoname(dev), (intmax_t)bno, fs->fs_fsmnt); panic("ffs_blkfree_cg: freeing free block"); } ffs_setblock(fs, blksfree, fragno); ffs_clusteracct(fs, cgp, fragno, 1); cgp->cg_cs.cs_nbfree++; fs->fs_cstotal.cs_nbfree++; fs->fs_cs(fs, cg).cs_nbfree++; } else { bbase = cgbno - fragnum(fs, cgbno); /* * decrement the counts associated with the old frags */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, -1); /* * deallocate the fragment */ frags = numfrags(fs, size); for (i = 0; i < frags; i++) { if (isset(blksfree, cgbno + i)) { printf("dev = %s, block = %jd, fs = %s\n", devtoname(dev), (intmax_t)(bno + i), fs->fs_fsmnt); panic("ffs_blkfree_cg: freeing free frag"); } setbit(blksfree, cgbno + i); } cgp->cg_cs.cs_nffree += i; fs->fs_cstotal.cs_nffree += i; fs->fs_cs(fs, cg).cs_nffree += i; /* * add back in counts associated with the new frags */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, 1); /* * if a complete block has been reassembled, account for it */ fragno = fragstoblks(fs, bbase); if (ffs_isblock(fs, blksfree, fragno)) { cgp->cg_cs.cs_nffree -= fs->fs_frag; fs->fs_cstotal.cs_nffree -= fs->fs_frag; fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag; ffs_clusteracct(fs, cgp, fragno, 1); cgp->cg_cs.cs_nbfree++; fs->fs_cstotal.cs_nbfree++; fs->fs_cs(fs, cg).cs_nbfree++; } } fs->fs_fmod = 1; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); mp = UFSTOVFS(ump); if (MOUNTEDSOFTDEP(mp) && devvp->v_type != VREG) softdep_setup_blkfree(UFSTOVFS(ump), bp, bno, numfrags(fs, size), dephd); bdwrite(bp); } TASKQUEUE_DEFINE_THREAD(ffs_trim); struct ffs_blkfree_trim_params { struct task task; struct ufsmount *ump; struct vnode *devvp; ufs2_daddr_t bno; long size; ino_t inum; struct workhead *pdephd; struct workhead dephd; }; static void ffs_blkfree_trim_task(ctx, pending) void *ctx; int pending; { struct ffs_blkfree_trim_params *tp; tp = ctx; ffs_blkfree_cg(tp->ump, tp->ump->um_fs, tp->devvp, tp->bno, tp->size, tp->inum, tp->pdephd); vn_finished_secondary_write(UFSTOVFS(tp->ump)); free(tp, M_TEMP); } static void ffs_blkfree_trim_completed(bip) struct bio *bip; { struct ffs_blkfree_trim_params *tp; tp = bip->bio_caller2; g_destroy_bio(bip); TASK_INIT(&tp->task, 0, ffs_blkfree_trim_task, tp); taskqueue_enqueue(taskqueue_ffs_trim, &tp->task); } void ffs_blkfree(ump, fs, devvp, bno, size, inum, vtype, dephd) struct ufsmount *ump; struct fs *fs; struct vnode *devvp; ufs2_daddr_t bno; long size; ino_t inum; enum vtype vtype; struct workhead *dephd; { struct mount *mp; struct bio *bip; struct ffs_blkfree_trim_params *tp; /* * Check to see if a snapshot wants to claim the block. * Check that devvp is a normal disk device, not a snapshot, * it has a snapshot(s) associated with it, and one of the * snapshots wants to claim the block. */ if (devvp->v_type != VREG && (devvp->v_vflag & VV_COPYONWRITE) && ffs_snapblkfree(fs, devvp, bno, size, inum, vtype, dephd)) { return; } /* * Nothing to delay if TRIM is disabled, or the operation is * performed on the snapshot. */ if (!ump->um_candelete || devvp->v_type == VREG) { ffs_blkfree_cg(ump, fs, devvp, bno, size, inum, dephd); return; } /* * Postpone the set of the free bit in the cg bitmap until the * BIO_DELETE is completed. Otherwise, due to disk queue * reordering, TRIM might be issued after we reuse the block * and write some new data into it. */ tp = malloc(sizeof(struct ffs_blkfree_trim_params), M_TEMP, M_WAITOK); tp->ump = ump; tp->devvp = devvp; tp->bno = bno; tp->size = size; tp->inum = inum; if (dephd != NULL) { LIST_INIT(&tp->dephd); LIST_SWAP(dephd, &tp->dephd, worklist, wk_list); tp->pdephd = &tp->dephd; } else tp->pdephd = NULL; bip = g_alloc_bio(); bip->bio_cmd = BIO_DELETE; bip->bio_offset = dbtob(fsbtodb(fs, bno)); bip->bio_done = ffs_blkfree_trim_completed; bip->bio_length = size; bip->bio_caller2 = tp; mp = UFSTOVFS(ump); vn_start_secondary_write(NULL, &mp, 0); g_io_request(bip, (struct g_consumer *)devvp->v_bufobj.bo_private); } #ifdef INVARIANTS /* * Verify allocation of a block or fragment. Returns true if block or * fragment is allocated, false if it is free. */ static int ffs_checkblk(ip, bno, size) struct inode *ip; ufs2_daddr_t bno; long size; { struct fs *fs; struct cg *cgp; struct buf *bp; ufs1_daddr_t cgbno; int i, error, frags, free; u_int8_t *blksfree; fs = ip->i_fs; if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) { printf("bsize = %ld, size = %ld, fs = %s\n", (long)fs->fs_bsize, size, fs->fs_fsmnt); panic("ffs_checkblk: bad size"); } if ((u_int)bno >= fs->fs_size) panic("ffs_checkblk: bad block %jd", (intmax_t)bno); error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, dtog(fs, bno))), (int)fs->fs_cgsize, NOCRED, &bp); if (error) panic("ffs_checkblk: cg bread failed"); cgp = (struct cg *)bp->b_data; if (!cg_chkmagic(cgp)) panic("ffs_checkblk: cg magic mismatch"); bp->b_xflags |= BX_BKGRDWRITE; blksfree = cg_blksfree(cgp); cgbno = dtogd(fs, bno); if (size == fs->fs_bsize) { free = ffs_isblock(fs, blksfree, fragstoblks(fs, cgbno)); } else { frags = numfrags(fs, size); for (free = 0, i = 0; i < frags; i++) if (isset(blksfree, cgbno + i)) free++; if (free != 0 && free != frags) panic("ffs_checkblk: partially free fragment"); } brelse(bp); return (!free); } #endif /* INVARIANTS */ /* * Free an inode. */ int ffs_vfree(pvp, ino, mode) struct vnode *pvp; ino_t ino; int mode; { struct inode *ip; if (DOINGSOFTDEP(pvp)) { softdep_freefile(pvp, ino, mode); return (0); } ip = VTOI(pvp); return (ffs_freefile(ip->i_ump, ip->i_fs, ip->i_devvp, ino, mode, NULL)); } /* * Do the actual free operation. * The specified inode is placed back in the free map. */ int ffs_freefile(ump, fs, devvp, ino, mode, wkhd) struct ufsmount *ump; struct fs *fs; struct vnode *devvp; ino_t ino; int mode; struct workhead *wkhd; { struct cg *cgp; struct buf *bp; ufs2_daddr_t cgbno; int error; u_int cg; u_int8_t *inosused; struct cdev *dev; cg = ino_to_cg(fs, ino); if (devvp->v_type == VREG) { /* devvp is a snapshot */ dev = VTOI(devvp)->i_devvp->v_rdev; cgbno = fragstoblks(fs, cgtod(fs, cg)); } else { /* devvp is a normal disk device */ dev = devvp->v_rdev; cgbno = fsbtodb(fs, cgtod(fs, cg)); } if (ino >= fs->fs_ipg * fs->fs_ncg) panic("ffs_freefile: range: dev = %s, ino = %ju, fs = %s", devtoname(dev), (uintmax_t)ino, fs->fs_fsmnt); if ((error = bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp))) { brelse(bp); return (error); } cgp = (struct cg *)bp->b_data; if (!cg_chkmagic(cgp)) { brelse(bp); return (0); } bp->b_xflags |= BX_BKGRDWRITE; cgp->cg_old_time = cgp->cg_time = time_second; inosused = cg_inosused(cgp); ino %= fs->fs_ipg; if (isclr(inosused, ino)) { printf("dev = %s, ino = %ju, fs = %s\n", devtoname(dev), (uintmax_t)(ino + cg * fs->fs_ipg), fs->fs_fsmnt); if (fs->fs_ronly == 0) panic("ffs_freefile: freeing free inode"); } clrbit(inosused, ino); if (ino < cgp->cg_irotor) cgp->cg_irotor = ino; cgp->cg_cs.cs_nifree++; UFS_LOCK(ump); fs->fs_cstotal.cs_nifree++; fs->fs_cs(fs, cg).cs_nifree++; if ((mode & IFMT) == IFDIR) { cgp->cg_cs.cs_ndir--; fs->fs_cstotal.cs_ndir--; fs->fs_cs(fs, cg).cs_ndir--; } fs->fs_fmod = 1; ACTIVECLEAR(fs, cg); UFS_UNLOCK(ump); if (MOUNTEDSOFTDEP(UFSTOVFS(ump)) && devvp->v_type != VREG) softdep_setup_inofree(UFSTOVFS(ump), bp, ino + cg * fs->fs_ipg, wkhd); bdwrite(bp); return (0); } /* * Check to see if a file is free. */ int ffs_checkfreefile(fs, devvp, ino) struct fs *fs; struct vnode *devvp; ino_t ino; { struct cg *cgp; struct buf *bp; ufs2_daddr_t cgbno; int ret; u_int cg; u_int8_t *inosused; cg = ino_to_cg(fs, ino); if (devvp->v_type == VREG) { /* devvp is a snapshot */ cgbno = fragstoblks(fs, cgtod(fs, cg)); } else { /* devvp is a normal disk device */ cgbno = fsbtodb(fs, cgtod(fs, cg)); } if (ino >= fs->fs_ipg * fs->fs_ncg) return (1); if (bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp)) { brelse(bp); return (1); } cgp = (struct cg *)bp->b_data; if (!cg_chkmagic(cgp)) { brelse(bp); return (1); } inosused = cg_inosused(cgp); ino %= fs->fs_ipg; ret = isclr(inosused, ino); brelse(bp); return (ret); } /* * Find a block of the specified size in the specified cylinder group. * * It is a panic if a request is made to find a block if none are * available. */ static ufs1_daddr_t ffs_mapsearch(fs, cgp, bpref, allocsiz) struct fs *fs; struct cg *cgp; ufs2_daddr_t bpref; int allocsiz; { ufs1_daddr_t bno; int start, len, loc, i; int blk, field, subfield, pos; u_int8_t *blksfree; /* * find the fragment by searching through the free block * map for an appropriate bit pattern */ if (bpref) start = dtogd(fs, bpref) / NBBY; else start = cgp->cg_frotor / NBBY; blksfree = cg_blksfree(cgp); len = howmany(fs->fs_fpg, NBBY) - start; loc = scanc((u_int)len, (u_char *)&blksfree[start], fragtbl[fs->fs_frag], (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY)))); if (loc == 0) { len = start + 1; start = 0; loc = scanc((u_int)len, (u_char *)&blksfree[0], fragtbl[fs->fs_frag], (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY)))); if (loc == 0) { printf("start = %d, len = %d, fs = %s\n", start, len, fs->fs_fsmnt); panic("ffs_alloccg: map corrupted"); /* NOTREACHED */ } } bno = (start + len - loc) * NBBY; cgp->cg_frotor = bno; /* * found the byte in the map * sift through the bits to find the selected frag */ for (i = bno + NBBY; bno < i; bno += fs->fs_frag) { blk = blkmap(fs, blksfree, bno); blk <<= 1; field = around[allocsiz]; subfield = inside[allocsiz]; for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) { if ((blk & field) == subfield) return (bno + pos); field <<= 1; subfield <<= 1; } } printf("bno = %lu, fs = %s\n", (u_long)bno, fs->fs_fsmnt); panic("ffs_alloccg: block not in map"); return (-1); } /* * Fserr prints the name of a filesystem with an error diagnostic. * * The form of the error message is: * fs: error message */ void ffs_fserr(fs, inum, cp) struct fs *fs; ino_t inum; char *cp; { struct thread *td = curthread; /* XXX */ struct proc *p = td->td_proc; log(LOG_ERR, "pid %d (%s), uid %d inumber %ju on %s: %s\n", p->p_pid, p->p_comm, td->td_ucred->cr_uid, (uintmax_t)inum, fs->fs_fsmnt, cp); } /* * This function provides the capability for the fsck program to * update an active filesystem. Fourteen operations are provided: * * adjrefcnt(inode, amt) - adjusts the reference count on the * specified inode by the specified amount. Under normal * operation the count should always go down. Decrementing * the count to zero will cause the inode to be freed. * adjblkcnt(inode, amt) - adjust the number of blocks used by the * inode by the specified amount. * adjndir, adjbfree, adjifree, adjffree, adjnumclusters(amt) - * adjust the superblock summary. * freedirs(inode, count) - directory inodes [inode..inode + count - 1] * are marked as free. Inodes should never have to be marked * as in use. * freefiles(inode, count) - file inodes [inode..inode + count - 1] * are marked as free. Inodes should never have to be marked * as in use. * freeblks(blockno, size) - blocks [blockno..blockno + size - 1] * are marked as free. Blocks should never have to be marked * as in use. * setflags(flags, set/clear) - the fs_flags field has the specified * flags set (second parameter +1) or cleared (second parameter -1). * setcwd(dirinode) - set the current directory to dirinode in the * filesystem associated with the snapshot. * setdotdot(oldvalue, newvalue) - Verify that the inode number for ".." * in the current directory is oldvalue then change it to newvalue. * unlink(nameptr, oldvalue) - Verify that the inode number associated * with nameptr in the current directory is oldvalue then unlink it. * * The following functions may only be used on a quiescent filesystem * by the soft updates journal. They are not safe to be run on an active * filesystem. * * setinode(inode, dip) - the specified disk inode is replaced with the * contents pointed to by dip. * setbufoutput(fd, flags) - output associated with the specified file * descriptor (which must reference the character device supporting * the filesystem) switches from using physio to running through the * buffer cache when flags is set to 1. The descriptor reverts to * physio for output when flags is set to zero. */ static int sysctl_ffs_fsck(SYSCTL_HANDLER_ARGS); SYSCTL_PROC(_vfs_ffs, FFS_ADJ_REFCNT, adjrefcnt, CTLFLAG_WR|CTLTYPE_STRUCT, 0, 0, sysctl_ffs_fsck, "S,fsck", "Adjust Inode Reference Count"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_BLKCNT, adjblkcnt, CTLFLAG_WR, sysctl_ffs_fsck, "Adjust Inode Used Blocks Count"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NDIR, adjndir, CTLFLAG_WR, sysctl_ffs_fsck, "Adjust number of directories"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NBFREE, adjnbfree, CTLFLAG_WR, sysctl_ffs_fsck, "Adjust number of free blocks"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NIFREE, adjnifree, CTLFLAG_WR, sysctl_ffs_fsck, "Adjust number of free inodes"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NFFREE, adjnffree, CTLFLAG_WR, sysctl_ffs_fsck, "Adjust number of free frags"); static SYSCTL_NODE(_vfs_ffs, FFS_ADJ_NUMCLUSTERS, adjnumclusters, CTLFLAG_WR, sysctl_ffs_fsck, "Adjust number of free clusters"); static SYSCTL_NODE(_vfs_ffs, FFS_DIR_FREE, freedirs, CTLFLAG_WR, sysctl_ffs_fsck, "Free Range of Directory Inodes"); static SYSCTL_NODE(_vfs_ffs, FFS_FILE_FREE, freefiles, CTLFLAG_WR, sysctl_ffs_fsck, "Free Range of File Inodes"); static SYSCTL_NODE(_vfs_ffs, FFS_BLK_FREE, freeblks, CTLFLAG_WR, sysctl_ffs_fsck, "Free Range of Blocks"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_FLAGS, setflags, CTLFLAG_WR, sysctl_ffs_fsck, "Change Filesystem Flags"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_CWD, setcwd, CTLFLAG_WR, sysctl_ffs_fsck, "Set Current Working Directory"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_DOTDOT, setdotdot, CTLFLAG_WR, sysctl_ffs_fsck, "Change Value of .. Entry"); static SYSCTL_NODE(_vfs_ffs, FFS_UNLINK, unlink, CTLFLAG_WR, sysctl_ffs_fsck, "Unlink a Duplicate Name"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_INODE, setinode, CTLFLAG_WR, sysctl_ffs_fsck, "Update an On-Disk Inode"); static SYSCTL_NODE(_vfs_ffs, FFS_SET_BUFOUTPUT, setbufoutput, CTLFLAG_WR, sysctl_ffs_fsck, "Set Buffered Writing for Descriptor"); #define DEBUG 1 #ifdef DEBUG static int fsckcmds = 0; SYSCTL_INT(_debug, OID_AUTO, fsckcmds, CTLFLAG_RW, &fsckcmds, 0, ""); #endif /* DEBUG */ static int buffered_write(struct file *, struct uio *, struct ucred *, int, struct thread *); static int sysctl_ffs_fsck(SYSCTL_HANDLER_ARGS) { struct thread *td = curthread; struct fsck_cmd cmd; struct ufsmount *ump; struct vnode *vp, *vpold, *dvp, *fdvp; struct inode *ip, *dp; struct mount *mp; struct fs *fs; ufs2_daddr_t blkno; long blkcnt, blksize; struct filedesc *fdp; struct file *fp, *vfp; cap_rights_t rights; int filetype, error; static struct fileops *origops, bufferedops; if (req->newlen > sizeof cmd) return (EBADRPC); if ((error = SYSCTL_IN(req, &cmd, sizeof cmd)) != 0) return (error); if (cmd.version != FFS_CMD_VERSION) return (ERPCMISMATCH); if ((error = getvnode(td->td_proc->p_fd, cmd.handle, cap_rights_init(&rights, CAP_FSCK), &fp)) != 0) return (error); vp = fp->f_data; if (vp->v_type != VREG && vp->v_type != VDIR) { fdrop(fp, td); return (EINVAL); } vn_start_write(vp, &mp, V_WAIT); if (mp == 0 || strncmp(mp->mnt_stat.f_fstypename, "ufs", MFSNAMELEN)) { vn_finished_write(mp); fdrop(fp, td); return (EINVAL); } ump = VFSTOUFS(mp); if ((mp->mnt_flag & MNT_RDONLY) && ump->um_fsckpid != td->td_proc->p_pid) { vn_finished_write(mp); fdrop(fp, td); return (EROFS); } fs = ump->um_fs; filetype = IFREG; switch (oidp->oid_number) { case FFS_SET_FLAGS: #ifdef DEBUG if (fsckcmds) printf("%s: %s flags\n", mp->mnt_stat.f_mntonname, cmd.size > 0 ? "set" : "clear"); #endif /* DEBUG */ if (cmd.size > 0) fs->fs_flags |= (long)cmd.value; else fs->fs_flags &= ~(long)cmd.value; break; case FFS_ADJ_REFCNT: #ifdef DEBUG if (fsckcmds) { printf("%s: adjust inode %jd link count by %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.size); } #endif /* DEBUG */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp))) break; ip = VTOI(vp); ip->i_nlink += cmd.size; DIP_SET(ip, i_nlink, ip->i_nlink); ip->i_effnlink += cmd.size; ip->i_flag |= IN_CHANGE | IN_MODIFIED; error = ffs_update(vp, 1); if (DOINGSOFTDEP(vp)) softdep_change_linkcnt(ip); vput(vp); break; case FFS_ADJ_BLKCNT: #ifdef DEBUG if (fsckcmds) { printf("%s: adjust inode %jd block count by %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.size); } #endif /* DEBUG */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp))) break; ip = VTOI(vp); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) + cmd.size); ip->i_flag |= IN_CHANGE | IN_MODIFIED; error = ffs_update(vp, 1); vput(vp); break; case FFS_DIR_FREE: filetype = IFDIR; /* fall through */ case FFS_FILE_FREE: #ifdef DEBUG if (fsckcmds) { if (cmd.size == 1) printf("%s: free %s inode %ju\n", mp->mnt_stat.f_mntonname, filetype == IFDIR ? "directory" : "file", (uintmax_t)cmd.value); else printf("%s: free %s inodes %ju-%ju\n", mp->mnt_stat.f_mntonname, filetype == IFDIR ? "directory" : "file", (uintmax_t)cmd.value, (uintmax_t)(cmd.value + cmd.size - 1)); } #endif /* DEBUG */ while (cmd.size > 0) { if ((error = ffs_freefile(ump, fs, ump->um_devvp, cmd.value, filetype, NULL))) break; cmd.size -= 1; cmd.value += 1; } break; case FFS_BLK_FREE: #ifdef DEBUG if (fsckcmds) { if (cmd.size == 1) printf("%s: free block %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); else printf("%s: free blocks %jd-%jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.value + cmd.size - 1); } #endif /* DEBUG */ blkno = cmd.value; blkcnt = cmd.size; blksize = fs->fs_frag - (blkno % fs->fs_frag); while (blkcnt > 0) { if (blksize > blkcnt) blksize = blkcnt; ffs_blkfree(ump, fs, ump->um_devvp, blkno, blksize * fs->fs_fsize, ROOTINO, VDIR, NULL); blkno += blksize; blkcnt -= blksize; blksize = fs->fs_frag; } break; /* * Adjust superblock summaries. fsck(8) is expected to * submit deltas when necessary. */ case FFS_ADJ_NDIR: #ifdef DEBUG if (fsckcmds) { printf("%s: adjust number of directories by %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DEBUG */ fs->fs_cstotal.cs_ndir += cmd.value; break; case FFS_ADJ_NBFREE: #ifdef DEBUG if (fsckcmds) { printf("%s: adjust number of free blocks by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DEBUG */ fs->fs_cstotal.cs_nbfree += cmd.value; break; case FFS_ADJ_NIFREE: #ifdef DEBUG if (fsckcmds) { printf("%s: adjust number of free inodes by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DEBUG */ fs->fs_cstotal.cs_nifree += cmd.value; break; case FFS_ADJ_NFFREE: #ifdef DEBUG if (fsckcmds) { printf("%s: adjust number of free frags by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DEBUG */ fs->fs_cstotal.cs_nffree += cmd.value; break; case FFS_ADJ_NUMCLUSTERS: #ifdef DEBUG if (fsckcmds) { printf("%s: adjust number of free clusters by %+jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DEBUG */ fs->fs_cstotal.cs_numclusters += cmd.value; break; case FFS_SET_CWD: #ifdef DEBUG if (fsckcmds) { printf("%s: set current directory to inode %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DEBUG */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_SHARED, &vp))) break; AUDIT_ARG_VNODE1(vp); if ((error = change_dir(vp, td)) != 0) { vput(vp); break; } VOP_UNLOCK(vp, 0); fdp = td->td_proc->p_fd; FILEDESC_XLOCK(fdp); vpold = fdp->fd_cdir; fdp->fd_cdir = vp; FILEDESC_XUNLOCK(fdp); vrele(vpold); break; case FFS_SET_DOTDOT: #ifdef DEBUG if (fsckcmds) { printf("%s: change .. in cwd from %jd to %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value, (intmax_t)cmd.size); } #endif /* DEBUG */ /* * First we have to get and lock the parent directory * to which ".." points. */ error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &fdvp); if (error) break; /* * Now we get and lock the child directory containing "..". */ FILEDESC_SLOCK(td->td_proc->p_fd); dvp = td->td_proc->p_fd->fd_cdir; FILEDESC_SUNLOCK(td->td_proc->p_fd); if ((error = vget(dvp, LK_EXCLUSIVE, td)) != 0) { vput(fdvp); break; } dp = VTOI(dvp); dp->i_offset = 12; /* XXX mastertemplate.dot_reclen */ error = ufs_dirrewrite(dp, VTOI(fdvp), (ino_t)cmd.size, DT_DIR, 0); cache_purge(fdvp); cache_purge(dvp); vput(dvp); vput(fdvp); break; case FFS_UNLINK: #ifdef DEBUG if (fsckcmds) { char buf[32]; if (copyinstr((char *)(intptr_t)cmd.value, buf,32,NULL)) strncpy(buf, "Name_too_long", 32); printf("%s: unlink %s (inode %jd)\n", mp->mnt_stat.f_mntonname, buf, (intmax_t)cmd.size); } #endif /* DEBUG */ /* * kern_unlinkat will do its own start/finish writes and * they do not nest, so drop ours here. Setting mp == NULL * indicates that vn_finished_write is not needed down below. */ vn_finished_write(mp); mp = NULL; error = kern_unlinkat(td, AT_FDCWD, (char *)(intptr_t)cmd.value, UIO_USERSPACE, (ino_t)cmd.size); break; case FFS_SET_INODE: if (ump->um_fsckpid != td->td_proc->p_pid) { error = EPERM; break; } #ifdef DEBUG if (fsckcmds) { printf("%s: update inode %jd\n", mp->mnt_stat.f_mntonname, (intmax_t)cmd.value); } #endif /* DEBUG */ if ((error = ffs_vget(mp, (ino_t)cmd.value, LK_EXCLUSIVE, &vp))) break; AUDIT_ARG_VNODE1(vp); ip = VTOI(vp); if (ip->i_ump->um_fstype == UFS1) error = copyin((void *)(intptr_t)cmd.size, ip->i_din1, sizeof(struct ufs1_dinode)); else error = copyin((void *)(intptr_t)cmd.size, ip->i_din2, sizeof(struct ufs2_dinode)); if (error) { vput(vp); break; } ip->i_flag |= IN_CHANGE | IN_MODIFIED; error = ffs_update(vp, 1); vput(vp); break; case FFS_SET_BUFOUTPUT: if (ump->um_fsckpid != td->td_proc->p_pid) { error = EPERM; break; } if (VTOI(vp)->i_ump != ump) { error = EINVAL; break; } #ifdef DEBUG if (fsckcmds) { printf("%s: %s buffered output for descriptor %jd\n", mp->mnt_stat.f_mntonname, cmd.size == 1 ? "enable" : "disable", (intmax_t)cmd.value); } #endif /* DEBUG */ if ((error = getvnode(td->td_proc->p_fd, cmd.value, cap_rights_init(&rights, CAP_FSCK), &vfp)) != 0) break; if (vfp->f_vnode->v_type != VCHR) { fdrop(vfp, td); error = EINVAL; break; } if (origops == NULL) { origops = vfp->f_ops; bcopy((void *)origops, (void *)&bufferedops, sizeof(bufferedops)); bufferedops.fo_write = buffered_write; } if (cmd.size == 1) atomic_store_rel_ptr((volatile uintptr_t *)&vfp->f_ops, (uintptr_t)&bufferedops); else atomic_store_rel_ptr((volatile uintptr_t *)&vfp->f_ops, (uintptr_t)origops); fdrop(vfp, td); break; default: #ifdef DEBUG if (fsckcmds) { printf("Invalid request %d from fsck\n", oidp->oid_number); } #endif /* DEBUG */ error = EINVAL; break; } fdrop(fp, td); vn_finished_write(mp); return (error); } /* * Function to switch a descriptor to use the buffer cache to stage * its I/O. This is needed so that writes to the filesystem device * will give snapshots a chance to copy modified blocks for which it * needs to retain copies. */ static int buffered_write(fp, uio, active_cred, flags, td) struct file *fp; struct uio *uio; struct ucred *active_cred; int flags; struct thread *td; { struct vnode *devvp, *vp; struct inode *ip; struct buf *bp; struct fs *fs; struct filedesc *fdp; int error; daddr_t lbn; /* * The devvp is associated with the /dev filesystem. To discover * the filesystem with which the device is associated, we depend * on the application setting the current directory to a location * within the filesystem being written. Yes, this is an ugly hack. */ devvp = fp->f_vnode; if (!vn_isdisk(devvp, NULL)) return (EINVAL); fdp = td->td_proc->p_fd; FILEDESC_SLOCK(fdp); vp = fdp->fd_cdir; vref(vp); FILEDESC_SUNLOCK(fdp); vn_lock(vp, LK_SHARED | LK_RETRY); /* * Check that the current directory vnode indeed belongs to * UFS before trying to dereference UFS-specific v_data fields. */ if (vp->v_op != &ffs_vnodeops1 && vp->v_op != &ffs_vnodeops2) { vput(vp); return (EINVAL); } ip = VTOI(vp); if (ip->i_devvp != devvp) { vput(vp); return (EINVAL); } fs = ip->i_fs; vput(vp); foffset_lock_uio(fp, uio, flags); vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); #ifdef DEBUG if (fsckcmds) { printf("%s: buffered write for block %jd\n", fs->fs_fsmnt, (intmax_t)btodb(uio->uio_offset)); } #endif /* DEBUG */ /* * All I/O must be contained within a filesystem block, start on * a fragment boundary, and be a multiple of fragments in length. */ if (uio->uio_resid > fs->fs_bsize - (uio->uio_offset % fs->fs_bsize) || fragoff(fs, uio->uio_offset) != 0 || fragoff(fs, uio->uio_resid) != 0) { error = EINVAL; goto out; } lbn = numfrags(fs, uio->uio_offset); bp = getblk(devvp, lbn, uio->uio_resid, 0, 0, 0); bp->b_flags |= B_RELBUF; if ((error = uiomove((char *)bp->b_data, uio->uio_resid, uio)) != 0) { brelse(bp); goto out; } error = bwrite(bp); out: VOP_UNLOCK(devvp, 0); foffset_unlock_uio(fp, uio, flags | FOF_NEXTOFF); return (error); } Index: user/ngie/more-tests/sys/ufs/ffs/ffs_vfsops.c =================================================================== --- user/ngie/more-tests/sys/ufs/ffs/ffs_vfsops.c (revision 281974) +++ user/ngie/more-tests/sys/ufs/ffs/ffs_vfsops.c (revision 281975) @@ -1,2217 +1,2218 @@ /*- * Copyright (c) 1989, 1991, 1993, 1994 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ffs_vfsops.c 8.31 (Berkeley) 5/20/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_quota.h" #include "opt_ufs.h" #include "opt_ffs.h" #include "opt_ddb.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static uma_zone_t uma_inode, uma_ufs1, uma_ufs2; static int ffs_mountfs(struct vnode *, struct mount *, struct thread *); static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, ufs2_daddr_t); static void ffs_ifree(struct ufsmount *ump, struct inode *ip); static int ffs_sync_lazy(struct mount *mp); static vfs_init_t ffs_init; static vfs_uninit_t ffs_uninit; static vfs_extattrctl_t ffs_extattrctl; static vfs_cmount_t ffs_cmount; static vfs_unmount_t ffs_unmount; static vfs_mount_t ffs_mount; static vfs_statfs_t ffs_statfs; static vfs_fhtovp_t ffs_fhtovp; static vfs_sync_t ffs_sync; static struct vfsops ufs_vfsops = { .vfs_extattrctl = ffs_extattrctl, .vfs_fhtovp = ffs_fhtovp, .vfs_init = ffs_init, .vfs_mount = ffs_mount, .vfs_cmount = ffs_cmount, .vfs_quotactl = ufs_quotactl, .vfs_root = ufs_root, .vfs_statfs = ffs_statfs, .vfs_sync = ffs_sync, .vfs_uninit = ffs_uninit, .vfs_unmount = ffs_unmount, .vfs_vget = ffs_vget, .vfs_susp_clean = process_deferred_inactive, }; VFS_SET(ufs_vfsops, ufs, 0); MODULE_VERSION(ufs, 1); static b_strategy_t ffs_geom_strategy; static b_write_t ffs_bufwrite; static struct buf_ops ffs_ops = { .bop_name = "FFS", .bop_write = ffs_bufwrite, .bop_strategy = ffs_geom_strategy, .bop_sync = bufsync, #ifdef NO_FFS_SNAPSHOT .bop_bdflush = bufbdflush, #else .bop_bdflush = ffs_bdflush, #endif }; /* * Note that userquota and groupquota options are not currently used * by UFS/FFS code and generally mount(8) does not pass those options * from userland, but they can be passed by loader(8) via * vfs.root.mountfrom.options. */ static const char *ffs_opts[] = { "acls", "async", "noatime", "noclusterr", "noclusterw", "noexec", "export", "force", "from", "groupquota", "multilabel", "nfsv4acls", "fsckpid", "snapshot", "nosuid", "suiddir", "nosymfollow", "sync", "union", "userquota", NULL }; static int ffs_mount(struct mount *mp) { struct vnode *devvp; struct thread *td; struct ufsmount *ump = NULL; struct fs *fs; pid_t fsckpid = 0; int error, flags; uint64_t mntorflags; accmode_t accmode; struct nameidata ndp; char *fspec; td = curthread; if (vfs_filteropt(mp->mnt_optnew, ffs_opts)) return (EINVAL); if (uma_inode == NULL) { uma_inode = uma_zcreate("FFS inode", sizeof(struct inode), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); uma_ufs1 = uma_zcreate("FFS1 dinode", sizeof(struct ufs1_dinode), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); uma_ufs2 = uma_zcreate("FFS2 dinode", sizeof(struct ufs2_dinode), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); } vfs_deleteopt(mp->mnt_optnew, "groupquota"); vfs_deleteopt(mp->mnt_optnew, "userquota"); fspec = vfs_getopts(mp->mnt_optnew, "from", &error); if (error) return (error); mntorflags = 0; if (vfs_getopt(mp->mnt_optnew, "acls", NULL, NULL) == 0) mntorflags |= MNT_ACLS; if (vfs_getopt(mp->mnt_optnew, "snapshot", NULL, NULL) == 0) { mntorflags |= MNT_SNAPSHOT; /* * Once we have set the MNT_SNAPSHOT flag, do not * persist "snapshot" in the options list. */ vfs_deleteopt(mp->mnt_optnew, "snapshot"); vfs_deleteopt(mp->mnt_opt, "snapshot"); } if (vfs_getopt(mp->mnt_optnew, "fsckpid", NULL, NULL) == 0 && vfs_scanopt(mp->mnt_optnew, "fsckpid", "%d", &fsckpid) == 1) { /* * Once we have set the restricted PID, do not * persist "fsckpid" in the options list. */ vfs_deleteopt(mp->mnt_optnew, "fsckpid"); vfs_deleteopt(mp->mnt_opt, "fsckpid"); if (mp->mnt_flag & MNT_UPDATE) { if (VFSTOUFS(mp)->um_fs->fs_ronly == 0 && vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0) == 0) { vfs_mount_error(mp, "Checker enable: Must be read-only"); return (EINVAL); } } else if (vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0) == 0) { vfs_mount_error(mp, "Checker enable: Must be read-only"); return (EINVAL); } /* Set to -1 if we are done */ if (fsckpid == 0) fsckpid = -1; } if (vfs_getopt(mp->mnt_optnew, "nfsv4acls", NULL, NULL) == 0) { if (mntorflags & MNT_ACLS) { vfs_mount_error(mp, "\"acls\" and \"nfsv4acls\" options " "are mutually exclusive"); return (EINVAL); } mntorflags |= MNT_NFS4ACLS; } MNT_ILOCK(mp); mp->mnt_flag |= mntorflags; MNT_IUNLOCK(mp); /* * If updating, check whether changing from read-only to * read/write; if there is no device name, that's all we do. */ if (mp->mnt_flag & MNT_UPDATE) { ump = VFSTOUFS(mp); fs = ump->um_fs; devvp = ump->um_devvp; if (fsckpid == -1 && ump->um_fsckpid > 0) { if ((error = ffs_flushfiles(mp, WRITECLOSE, td)) != 0 || (error = ffs_sbupdate(ump, MNT_WAIT, 0)) != 0) return (error); DROP_GIANT(); g_topology_lock(); /* * Return to normal read-only mode. */ error = g_access(ump->um_cp, 0, -1, 0); g_topology_unlock(); PICKUP_GIANT(); ump->um_fsckpid = 0; } if (fs->fs_ronly == 0 && vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) { /* * Flush any dirty data and suspend filesystem. */ if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0) return (error); error = vfs_write_suspend_umnt(mp); if (error != 0) return (error); /* * Check for and optionally get rid of files open * for writing. */ flags = WRITECLOSE; if (mp->mnt_flag & MNT_FORCE) flags |= FORCECLOSE; if (MOUNTEDSOFTDEP(mp)) { error = softdep_flushfiles(mp, flags, td); } else { error = ffs_flushfiles(mp, flags, td); } if (error) { vfs_write_resume(mp, 0); return (error); } if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { printf("WARNING: %s Update error: blocks %jd " "files %d\n", fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, fs->fs_pendinginodes); fs->fs_pendingblocks = 0; fs->fs_pendinginodes = 0; } if ((fs->fs_flags & (FS_UNCLEAN | FS_NEEDSFSCK)) == 0) fs->fs_clean = 1; if ((error = ffs_sbupdate(ump, MNT_WAIT, 0)) != 0) { fs->fs_ronly = 0; fs->fs_clean = 0; vfs_write_resume(mp, 0); return (error); } if (MOUNTEDSOFTDEP(mp)) softdep_unmount(mp); DROP_GIANT(); g_topology_lock(); /* * Drop our write and exclusive access. */ g_access(ump->um_cp, 0, -1, -1); g_topology_unlock(); PICKUP_GIANT(); fs->fs_ronly = 1; MNT_ILOCK(mp); mp->mnt_flag |= MNT_RDONLY; MNT_IUNLOCK(mp); /* * Allow the writers to note that filesystem * is ro now. */ vfs_write_resume(mp, 0); } if ((mp->mnt_flag & MNT_RELOAD) && (error = ffs_reload(mp, td, 0)) != 0) return (error); if (fs->fs_ronly && !vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) { /* * If we are running a checker, do not allow upgrade. */ if (ump->um_fsckpid > 0) { vfs_mount_error(mp, "Active checker, cannot upgrade to write"); return (EINVAL); } /* * If upgrade to read-write by non-root, then verify * that user has necessary permissions on the device. */ vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); error = VOP_ACCESS(devvp, VREAD | VWRITE, td->td_ucred, td); if (error) error = priv_check(td, PRIV_VFS_MOUNT_PERM); if (error) { VOP_UNLOCK(devvp, 0); return (error); } VOP_UNLOCK(devvp, 0); fs->fs_flags &= ~FS_UNCLEAN; if (fs->fs_clean == 0) { fs->fs_flags |= FS_UNCLEAN; if ((mp->mnt_flag & MNT_FORCE) || ((fs->fs_flags & (FS_SUJ | FS_NEEDSFSCK)) == 0 && (fs->fs_flags & FS_DOSOFTDEP))) { printf("WARNING: %s was not properly " "dismounted\n", fs->fs_fsmnt); } else { vfs_mount_error(mp, "R/W mount of %s denied. %s.%s", fs->fs_fsmnt, "Filesystem is not clean - run fsck", (fs->fs_flags & FS_SUJ) == 0 ? "" : " Forced mount will invalidate" " journal contents"); return (EPERM); } } DROP_GIANT(); g_topology_lock(); /* * Request exclusive write access. */ error = g_access(ump->um_cp, 0, 1, 1); g_topology_unlock(); PICKUP_GIANT(); if (error) return (error); if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0) return (error); fs->fs_ronly = 0; MNT_ILOCK(mp); mp->mnt_flag &= ~MNT_RDONLY; MNT_IUNLOCK(mp); fs->fs_mtime = time_second; /* check to see if we need to start softdep */ if ((fs->fs_flags & FS_DOSOFTDEP) && (error = softdep_mount(devvp, mp, fs, td->td_ucred))){ vn_finished_write(mp); return (error); } fs->fs_clean = 0; if ((error = ffs_sbupdate(ump, MNT_WAIT, 0)) != 0) { vn_finished_write(mp); return (error); } if (fs->fs_snapinum[0] != 0) ffs_snapshot_mount(mp); vn_finished_write(mp); } /* * Soft updates is incompatible with "async", * so if we are doing softupdates stop the user * from setting the async flag in an update. * Softdep_mount() clears it in an initial mount * or ro->rw remount. */ if (MOUNTEDSOFTDEP(mp)) { /* XXX: Reset too late ? */ MNT_ILOCK(mp); mp->mnt_flag &= ~MNT_ASYNC; MNT_IUNLOCK(mp); } /* * Keep MNT_ACLS flag if it is stored in superblock. */ if ((fs->fs_flags & FS_ACLS) != 0) { /* XXX: Set too late ? */ MNT_ILOCK(mp); mp->mnt_flag |= MNT_ACLS; MNT_IUNLOCK(mp); } if ((fs->fs_flags & FS_NFS4ACLS) != 0) { /* XXX: Set too late ? */ MNT_ILOCK(mp); mp->mnt_flag |= MNT_NFS4ACLS; MNT_IUNLOCK(mp); } /* * If this is a request from fsck to clean up the filesystem, * then allow the specified pid to proceed. */ if (fsckpid > 0) { if (ump->um_fsckpid != 0) { vfs_mount_error(mp, "Active checker already running on %s", fs->fs_fsmnt); return (EINVAL); } KASSERT(MOUNTEDSOFTDEP(mp) == 0, ("soft updates enabled on read-only file system")); DROP_GIANT(); g_topology_lock(); /* * Request write access. */ error = g_access(ump->um_cp, 0, 1, 0); g_topology_unlock(); PICKUP_GIANT(); if (error) { vfs_mount_error(mp, "Checker activation failed on %s", fs->fs_fsmnt); return (error); } ump->um_fsckpid = fsckpid; if (fs->fs_snapinum[0] != 0) ffs_snapshot_mount(mp); fs->fs_mtime = time_second; fs->fs_fmod = 1; fs->fs_clean = 0; (void) ffs_sbupdate(ump, MNT_WAIT, 0); } /* * If this is a snapshot request, take the snapshot. */ if (mp->mnt_flag & MNT_SNAPSHOT) return (ffs_snapshot(mp, fspec)); } /* * Not an update, or updating the name: look up the name * and verify that it refers to a sensible disk device. */ NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td); if ((error = namei(&ndp)) != 0) return (error); NDFREE(&ndp, NDF_ONLY_PNBUF); devvp = ndp.ni_vp; if (!vn_isdisk(devvp, &error)) { vput(devvp); return (error); } /* * If mount by non-root, then verify that user has necessary * permissions on the device. */ accmode = VREAD; if ((mp->mnt_flag & MNT_RDONLY) == 0) accmode |= VWRITE; error = VOP_ACCESS(devvp, accmode, td->td_ucred, td); if (error) error = priv_check(td, PRIV_VFS_MOUNT_PERM); if (error) { vput(devvp); return (error); } if (mp->mnt_flag & MNT_UPDATE) { /* * Update only * * If it's not the same vnode, or at least the same device * then it's not correct. */ if (devvp->v_rdev != ump->um_devvp->v_rdev) error = EINVAL; /* needs translation */ vput(devvp); if (error) return (error); } else { /* * New mount * * We need the name for the mount point (also used for * "last mounted on") copied in. If an error occurs, * the mount point is discarded by the upper level code. * Note that vfs_mount() populates f_mntonname for us. */ if ((error = ffs_mountfs(devvp, mp, td)) != 0) { vrele(devvp); return (error); } if (fsckpid > 0) { KASSERT(MOUNTEDSOFTDEP(mp) == 0, ("soft updates enabled on read-only file system")); ump = VFSTOUFS(mp); fs = ump->um_fs; DROP_GIANT(); g_topology_lock(); /* * Request write access. */ error = g_access(ump->um_cp, 0, 1, 0); g_topology_unlock(); PICKUP_GIANT(); if (error) { printf("WARNING: %s: Checker activation " "failed\n", fs->fs_fsmnt); } else { ump->um_fsckpid = fsckpid; if (fs->fs_snapinum[0] != 0) ffs_snapshot_mount(mp); fs->fs_mtime = time_second; fs->fs_clean = 0; (void) ffs_sbupdate(ump, MNT_WAIT, 0); } } } vfs_mountedfrom(mp, fspec); return (0); } /* * Compatibility with old mount system call. */ static int ffs_cmount(struct mntarg *ma, void *data, uint64_t flags) { struct ufs_args args; struct export_args exp; int error; if (data == NULL) return (EINVAL); error = copyin(data, &args, sizeof args); if (error) return (error); vfs_oexport_conv(&args.export, &exp); ma = mount_argsu(ma, "from", args.fspec, MAXPATHLEN); ma = mount_arg(ma, "export", &exp, sizeof(exp)); error = kernel_mount(ma, flags); return (error); } /* * Reload all incore data for a filesystem (used after running fsck on * the root filesystem and finding things to fix). If the 'force' flag * is 0, the filesystem must be mounted read-only. * * Things to do to update the mount: * 1) invalidate all cached meta-data. * 2) re-read superblock from disk. * 3) re-read summary information from disk. * 4) invalidate all inactive vnodes. * 5) invalidate all cached file data. * 6) re-read inode data for all active vnodes. */ int ffs_reload(struct mount *mp, struct thread *td, int force) { struct vnode *vp, *mvp, *devvp; struct inode *ip; void *space; struct buf *bp; struct fs *fs, *newfs; struct ufsmount *ump; ufs2_daddr_t sblockloc; int i, blks, size, error; int32_t *lp; ump = VFSTOUFS(mp); MNT_ILOCK(mp); if ((mp->mnt_flag & MNT_RDONLY) == 0 && force == 0) { MNT_IUNLOCK(mp); return (EINVAL); } MNT_IUNLOCK(mp); /* * Step 1: invalidate all cached meta-data. */ devvp = VFSTOUFS(mp)->um_devvp; vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); if (vinvalbuf(devvp, 0, 0, 0) != 0) panic("ffs_reload: dirty1"); VOP_UNLOCK(devvp, 0); /* * Step 2: re-read superblock from disk. */ fs = VFSTOUFS(mp)->um_fs; if ((error = bread(devvp, btodb(fs->fs_sblockloc), fs->fs_sbsize, NOCRED, &bp)) != 0) return (error); newfs = (struct fs *)bp->b_data; if ((newfs->fs_magic != FS_UFS1_MAGIC && newfs->fs_magic != FS_UFS2_MAGIC) || newfs->fs_bsize > MAXBSIZE || newfs->fs_bsize < sizeof(struct fs)) { brelse(bp); return (EIO); /* XXX needs translation */ } /* * Copy pointer fields back into superblock before copying in XXX * new superblock. These should really be in the ufsmount. XXX * Note that important parameters (eg fs_ncg) are unchanged. */ newfs->fs_csp = fs->fs_csp; newfs->fs_maxcluster = fs->fs_maxcluster; newfs->fs_contigdirs = fs->fs_contigdirs; newfs->fs_active = fs->fs_active; newfs->fs_ronly = fs->fs_ronly; sblockloc = fs->fs_sblockloc; bcopy(newfs, fs, (u_int)fs->fs_sbsize); brelse(bp); mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; ffs_oldfscompat_read(fs, VFSTOUFS(mp), sblockloc); UFS_LOCK(ump); if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { printf("WARNING: %s: reload pending error: blocks %jd " "files %d\n", fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, fs->fs_pendinginodes); fs->fs_pendingblocks = 0; fs->fs_pendinginodes = 0; } UFS_UNLOCK(ump); /* * Step 3: re-read summary information from disk. */ size = fs->fs_cssize; blks = howmany(size, fs->fs_fsize); if (fs->fs_contigsumsize > 0) size += fs->fs_ncg * sizeof(int32_t); size += fs->fs_ncg * sizeof(u_int8_t); free(fs->fs_csp, M_UFSMNT); space = malloc((u_long)size, M_UFSMNT, M_WAITOK); fs->fs_csp = space; for (i = 0; i < blks; i += fs->fs_frag) { size = fs->fs_bsize; if (i + fs->fs_frag > blks) size = (blks - i) * fs->fs_fsize; error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, NOCRED, &bp); if (error) return (error); bcopy(bp->b_data, space, (u_int)size); space = (char *)space + size; brelse(bp); } /* * We no longer know anything about clusters per cylinder group. */ if (fs->fs_contigsumsize > 0) { fs->fs_maxcluster = lp = space; for (i = 0; i < fs->fs_ncg; i++) *lp++ = fs->fs_contigsumsize; space = lp; } size = fs->fs_ncg * sizeof(u_int8_t); fs->fs_contigdirs = (u_int8_t *)space; bzero(fs->fs_contigdirs, size); loop: MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { /* * Skip syncer vnode. */ if (vp->v_type == VNON) { VI_UNLOCK(vp); continue; } /* * Step 4: invalidate all cached file data. */ if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); goto loop; } if (vinvalbuf(vp, 0, 0, 0)) panic("ffs_reload: dirty2"); /* * Step 5: re-read inode data for all active vnodes. */ ip = VTOI(vp); error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), (int)fs->fs_bsize, NOCRED, &bp); if (error) { VOP_UNLOCK(vp, 0); vrele(vp); MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); return (error); } ffs_load_inode(bp, ip, fs, ip->i_number); ip->i_effnlink = ip->i_nlink; brelse(bp); VOP_UNLOCK(vp, 0); vrele(vp); } return (0); } /* * Possible superblock locations ordered from most to least likely. */ static int sblock_try[] = SBLOCKSEARCH; /* * Common code for mount and mountroot */ static int ffs_mountfs(devvp, mp, td) struct vnode *devvp; struct mount *mp; struct thread *td; { struct ufsmount *ump; struct buf *bp; struct fs *fs; struct cdev *dev; void *space; ufs2_daddr_t sblockloc; int error, i, blks, size, ronly; int32_t *lp; struct ucred *cred; struct g_consumer *cp; struct mount *nmp; bp = NULL; ump = NULL; cred = td ? td->td_ucred : NOCRED; ronly = (mp->mnt_flag & MNT_RDONLY) != 0; dev = devvp->v_rdev; dev_ref(dev); DROP_GIANT(); g_topology_lock(); error = g_vfs_open(devvp, &cp, "ffs", ronly ? 0 : 1); g_topology_unlock(); PICKUP_GIANT(); VOP_UNLOCK(devvp, 0); if (error) goto out; if (devvp->v_rdev->si_iosize_max != 0) mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max; if (mp->mnt_iosize_max > MAXPHYS) mp->mnt_iosize_max = MAXPHYS; devvp->v_bufobj.bo_ops = &ffs_ops; fs = NULL; sblockloc = 0; /* * Try reading the superblock in each of its possible locations. */ for (i = 0; sblock_try[i] != -1; i++) { if ((SBLOCKSIZE % cp->provider->sectorsize) != 0) { error = EINVAL; vfs_mount_error(mp, "Invalid sectorsize %d for superblock size %d", cp->provider->sectorsize, SBLOCKSIZE); goto out; } if ((error = bread(devvp, btodb(sblock_try[i]), SBLOCKSIZE, cred, &bp)) != 0) goto out; fs = (struct fs *)bp->b_data; sblockloc = sblock_try[i]; if ((fs->fs_magic == FS_UFS1_MAGIC || (fs->fs_magic == FS_UFS2_MAGIC && (fs->fs_sblockloc == sblockloc || (fs->fs_old_flags & FS_FLAGS_UPDATED) == 0))) && fs->fs_bsize <= MAXBSIZE && fs->fs_bsize >= sizeof(struct fs)) break; brelse(bp); bp = NULL; } if (sblock_try[i] == -1) { error = EINVAL; /* XXX needs translation */ goto out; } fs->fs_fmod = 0; fs->fs_flags &= ~FS_INDEXDIRS; /* no support for directory indicies */ fs->fs_flags &= ~FS_UNCLEAN; if (fs->fs_clean == 0) { fs->fs_flags |= FS_UNCLEAN; if (ronly || (mp->mnt_flag & MNT_FORCE) || ((fs->fs_flags & (FS_SUJ | FS_NEEDSFSCK)) == 0 && (fs->fs_flags & FS_DOSOFTDEP))) { printf("WARNING: %s was not properly dismounted\n", fs->fs_fsmnt); } else { vfs_mount_error(mp, "R/W mount of %s denied. %s%s", fs->fs_fsmnt, "Filesystem is not clean - run fsck.", (fs->fs_flags & FS_SUJ) == 0 ? "" : " Forced mount will invalidate journal contents"); error = EPERM; goto out; } if ((fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) && (mp->mnt_flag & MNT_FORCE)) { printf("WARNING: %s: lost blocks %jd files %d\n", fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, fs->fs_pendinginodes); fs->fs_pendingblocks = 0; fs->fs_pendinginodes = 0; } } if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { printf("WARNING: %s: mount pending error: blocks %jd " "files %d\n", fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, fs->fs_pendinginodes); fs->fs_pendingblocks = 0; fs->fs_pendinginodes = 0; } if ((fs->fs_flags & FS_GJOURNAL) != 0) { #ifdef UFS_GJOURNAL /* * Get journal provider name. */ size = 1024; mp->mnt_gjprovider = malloc(size, M_UFSMNT, M_WAITOK); if (g_io_getattr("GJOURNAL::provider", cp, &size, mp->mnt_gjprovider) == 0) { mp->mnt_gjprovider = realloc(mp->mnt_gjprovider, size, M_UFSMNT, M_WAITOK); MNT_ILOCK(mp); mp->mnt_flag |= MNT_GJOURNAL; MNT_IUNLOCK(mp); } else { printf("WARNING: %s: GJOURNAL flag on fs " "but no gjournal provider below\n", mp->mnt_stat.f_mntonname); free(mp->mnt_gjprovider, M_UFSMNT); mp->mnt_gjprovider = NULL; } #else printf("WARNING: %s: GJOURNAL flag on fs but no " "UFS_GJOURNAL support\n", mp->mnt_stat.f_mntonname); #endif } else { mp->mnt_gjprovider = NULL; } ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK | M_ZERO); ump->um_cp = cp; ump->um_bo = &devvp->v_bufobj; ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT, M_WAITOK); if (fs->fs_magic == FS_UFS1_MAGIC) { ump->um_fstype = UFS1; ump->um_balloc = ffs_balloc_ufs1; } else { ump->um_fstype = UFS2; ump->um_balloc = ffs_balloc_ufs2; } ump->um_blkatoff = ffs_blkatoff; ump->um_truncate = ffs_truncate; ump->um_update = ffs_update; ump->um_valloc = ffs_valloc; ump->um_vfree = ffs_vfree; ump->um_ifree = ffs_ifree; ump->um_rdonly = ffs_rdonly; ump->um_snapgone = ffs_snapgone; mtx_init(UFS_MTX(ump), "FFS", "FFS Lock", MTX_DEF); bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize); if (fs->fs_sbsize < SBLOCKSIZE) bp->b_flags |= B_INVAL | B_NOCACHE; brelse(bp); bp = NULL; fs = ump->um_fs; ffs_oldfscompat_read(fs, ump, sblockloc); fs->fs_ronly = ronly; size = fs->fs_cssize; blks = howmany(size, fs->fs_fsize); if (fs->fs_contigsumsize > 0) size += fs->fs_ncg * sizeof(int32_t); size += fs->fs_ncg * sizeof(u_int8_t); space = malloc((u_long)size, M_UFSMNT, M_WAITOK); fs->fs_csp = space; for (i = 0; i < blks; i += fs->fs_frag) { size = fs->fs_bsize; if (i + fs->fs_frag > blks) size = (blks - i) * fs->fs_fsize; if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, cred, &bp)) != 0) { free(fs->fs_csp, M_UFSMNT); goto out; } bcopy(bp->b_data, space, (u_int)size); space = (char *)space + size; brelse(bp); bp = NULL; } if (fs->fs_contigsumsize > 0) { fs->fs_maxcluster = lp = space; for (i = 0; i < fs->fs_ncg; i++) *lp++ = fs->fs_contigsumsize; space = lp; } size = fs->fs_ncg * sizeof(u_int8_t); fs->fs_contigdirs = (u_int8_t *)space; bzero(fs->fs_contigdirs, size); fs->fs_active = NULL; mp->mnt_data = ump; mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0]; mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1]; nmp = NULL; if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 || (nmp = vfs_getvfs(&mp->mnt_stat.f_fsid))) { if (nmp) vfs_rel(nmp); vfs_getnewfsid(mp); } mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; MNT_ILOCK(mp); mp->mnt_flag |= MNT_LOCAL; MNT_IUNLOCK(mp); if ((fs->fs_flags & FS_MULTILABEL) != 0) { #ifdef MAC MNT_ILOCK(mp); mp->mnt_flag |= MNT_MULTILABEL; MNT_IUNLOCK(mp); #else printf("WARNING: %s: multilabel flag on fs but " "no MAC support\n", mp->mnt_stat.f_mntonname); #endif } if ((fs->fs_flags & FS_ACLS) != 0) { #ifdef UFS_ACL MNT_ILOCK(mp); if (mp->mnt_flag & MNT_NFS4ACLS) printf("WARNING: %s: ACLs flag on fs conflicts with " "\"nfsv4acls\" mount option; option ignored\n", mp->mnt_stat.f_mntonname); mp->mnt_flag &= ~MNT_NFS4ACLS; mp->mnt_flag |= MNT_ACLS; MNT_IUNLOCK(mp); #else printf("WARNING: %s: ACLs flag on fs but no ACLs support\n", mp->mnt_stat.f_mntonname); #endif } if ((fs->fs_flags & FS_NFS4ACLS) != 0) { #ifdef UFS_ACL MNT_ILOCK(mp); if (mp->mnt_flag & MNT_ACLS) printf("WARNING: %s: NFSv4 ACLs flag on fs conflicts " "with \"acls\" mount option; option ignored\n", mp->mnt_stat.f_mntonname); mp->mnt_flag &= ~MNT_ACLS; mp->mnt_flag |= MNT_NFS4ACLS; MNT_IUNLOCK(mp); #else printf("WARNING: %s: NFSv4 ACLs flag on fs but no " "ACLs support\n", mp->mnt_stat.f_mntonname); #endif } if ((fs->fs_flags & FS_TRIM) != 0) { size = sizeof(int); if (g_io_getattr("GEOM::candelete", cp, &size, &ump->um_candelete) == 0) { if (!ump->um_candelete) printf("WARNING: %s: TRIM flag on fs but disk " "does not support TRIM\n", mp->mnt_stat.f_mntonname); } else { printf("WARNING: %s: TRIM flag on fs but disk does " "not confirm that it supports TRIM\n", mp->mnt_stat.f_mntonname); ump->um_candelete = 0; } } ump->um_mountp = mp; ump->um_dev = dev; ump->um_devvp = devvp; ump->um_nindir = fs->fs_nindir; ump->um_bptrtodb = fs->fs_fsbtodb; ump->um_seqinc = fs->fs_frag; for (i = 0; i < MAXQUOTAS; i++) ump->um_quotas[i] = NULLVP; #ifdef UFS_EXTATTR ufs_extattr_uepm_init(&ump->um_extattr); #endif /* * Set FS local "last mounted on" information (NULL pad) */ bzero(fs->fs_fsmnt, MAXMNTLEN); strlcpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MAXMNTLEN); mp->mnt_stat.f_iosize = fs->fs_bsize; if (mp->mnt_flag & MNT_ROOTFS) { /* * Root mount; update timestamp in mount structure. * this will be used by the common root mount code * to update the system clock. */ mp->mnt_time = fs->fs_time; } if (ronly == 0) { fs->fs_mtime = time_second; if ((fs->fs_flags & FS_DOSOFTDEP) && (error = softdep_mount(devvp, mp, fs, cred)) != 0) { free(fs->fs_csp, M_UFSMNT); ffs_flushfiles(mp, FORCECLOSE, td); goto out; } if (devvp->v_type == VCHR && devvp->v_rdev != NULL) devvp->v_rdev->si_mountpt = mp; if (fs->fs_snapinum[0] != 0) ffs_snapshot_mount(mp); fs->fs_fmod = 1; fs->fs_clean = 0; (void) ffs_sbupdate(ump, MNT_WAIT, 0); } /* * Initialize filesystem stat information in mount struct. */ MNT_ILOCK(mp); mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED | MNTK_NO_IOPF | MNTK_UNMAPPED_BUFS | MNTK_SUSPENDABLE | MNTK_USES_BCACHE; MNT_IUNLOCK(mp); #ifdef UFS_EXTATTR #ifdef UFS_EXTATTR_AUTOSTART /* * * Auto-starting does the following: * - check for /.attribute in the fs, and extattr_start if so * - for each file in .attribute, enable that file with * an attribute of the same name. * Not clear how to report errors -- probably eat them. * This would all happen while the filesystem was busy/not * available, so would effectively be "atomic". */ (void) ufs_extattr_autostart(mp, td); #endif /* !UFS_EXTATTR_AUTOSTART */ #endif /* !UFS_EXTATTR */ return (0); out: if (bp) brelse(bp); if (cp != NULL) { DROP_GIANT(); g_topology_lock(); g_vfs_close(cp); g_topology_unlock(); PICKUP_GIANT(); } if (ump) { mtx_destroy(UFS_MTX(ump)); if (mp->mnt_gjprovider != NULL) { free(mp->mnt_gjprovider, M_UFSMNT); mp->mnt_gjprovider = NULL; } free(ump->um_fs, M_UFSMNT); free(ump, M_UFSMNT); mp->mnt_data = NULL; } dev_rel(dev); return (error); } #include static int bigcgs = 0; SYSCTL_INT(_debug, OID_AUTO, bigcgs, CTLFLAG_RW, &bigcgs, 0, ""); /* * Sanity checks for loading old filesystem superblocks. * See ffs_oldfscompat_write below for unwound actions. * * XXX - Parts get retired eventually. * Unfortunately new bits get added. */ static void ffs_oldfscompat_read(fs, ump, sblockloc) struct fs *fs; struct ufsmount *ump; ufs2_daddr_t sblockloc; { off_t maxfilesize; /* * If not yet done, update fs_flags location and value of fs_sblockloc. */ if ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0) { fs->fs_flags = fs->fs_old_flags; fs->fs_old_flags |= FS_FLAGS_UPDATED; fs->fs_sblockloc = sblockloc; } /* * If not yet done, update UFS1 superblock with new wider fields. */ if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_maxbsize != fs->fs_bsize) { fs->fs_maxbsize = fs->fs_bsize; fs->fs_time = fs->fs_old_time; fs->fs_size = fs->fs_old_size; fs->fs_dsize = fs->fs_old_dsize; fs->fs_csaddr = fs->fs_old_csaddr; fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir; fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree; fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree; fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree; } if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_old_inodefmt < FS_44INODEFMT) { fs->fs_maxfilesize = ((uint64_t)1 << 31) - 1; fs->fs_qbmask = ~fs->fs_bmask; fs->fs_qfmask = ~fs->fs_fmask; } if (fs->fs_magic == FS_UFS1_MAGIC) { ump->um_savedmaxfilesize = fs->fs_maxfilesize; maxfilesize = (uint64_t)0x80000000 * fs->fs_bsize - 1; if (fs->fs_maxfilesize > maxfilesize) fs->fs_maxfilesize = maxfilesize; } /* Compatibility for old filesystems */ if (fs->fs_avgfilesize <= 0) fs->fs_avgfilesize = AVFILESIZ; if (fs->fs_avgfpdir <= 0) fs->fs_avgfpdir = AFPDIR; if (bigcgs) { fs->fs_save_cgsize = fs->fs_cgsize; fs->fs_cgsize = fs->fs_bsize; } } /* * Unwinding superblock updates for old filesystems. * See ffs_oldfscompat_read above for details. * * XXX - Parts get retired eventually. * Unfortunately new bits get added. */ void ffs_oldfscompat_write(fs, ump) struct fs *fs; struct ufsmount *ump; { /* * Copy back UFS2 updated fields that UFS1 inspects. */ if (fs->fs_magic == FS_UFS1_MAGIC) { fs->fs_old_time = fs->fs_time; fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir; fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree; fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree; fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree; fs->fs_maxfilesize = ump->um_savedmaxfilesize; } if (bigcgs) { fs->fs_cgsize = fs->fs_save_cgsize; fs->fs_save_cgsize = 0; } } /* * unmount system call */ static int ffs_unmount(mp, mntflags) struct mount *mp; int mntflags; { struct thread *td; struct ufsmount *ump = VFSTOUFS(mp); struct fs *fs; int error, flags, susp; #ifdef UFS_EXTATTR int e_restart; #endif flags = 0; td = curthread; fs = ump->um_fs; susp = 0; if (mntflags & MNT_FORCE) { flags |= FORCECLOSE; susp = fs->fs_ronly == 0; } #ifdef UFS_EXTATTR if ((error = ufs_extattr_stop(mp, td))) { if (error != EOPNOTSUPP) printf("WARNING: unmount %s: ufs_extattr_stop " "returned errno %d\n", mp->mnt_stat.f_mntonname, error); e_restart = 0; } else { ufs_extattr_uepm_destroy(&ump->um_extattr); e_restart = 1; } #endif if (susp) { error = vfs_write_suspend_umnt(mp); if (error != 0) goto fail1; } if (MOUNTEDSOFTDEP(mp)) error = softdep_flushfiles(mp, flags, td); else error = ffs_flushfiles(mp, flags, td); if (error != 0 && error != ENXIO) goto fail; UFS_LOCK(ump); if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { printf("WARNING: unmount %s: pending error: blocks %jd " "files %d\n", fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, fs->fs_pendinginodes); fs->fs_pendingblocks = 0; fs->fs_pendinginodes = 0; } UFS_UNLOCK(ump); if (MOUNTEDSOFTDEP(mp)) softdep_unmount(mp); if (fs->fs_ronly == 0 || ump->um_fsckpid > 0) { fs->fs_clean = fs->fs_flags & (FS_UNCLEAN|FS_NEEDSFSCK) ? 0 : 1; error = ffs_sbupdate(ump, MNT_WAIT, 0); if (error && error != ENXIO) { fs->fs_clean = 0; goto fail; } } if (susp) vfs_write_resume(mp, VR_START_WRITE); DROP_GIANT(); g_topology_lock(); if (ump->um_fsckpid > 0) { /* * Return to normal read-only mode. */ error = g_access(ump->um_cp, 0, -1, 0); ump->um_fsckpid = 0; } g_vfs_close(ump->um_cp); g_topology_unlock(); PICKUP_GIANT(); if (ump->um_devvp->v_type == VCHR && ump->um_devvp->v_rdev != NULL) ump->um_devvp->v_rdev->si_mountpt = NULL; vrele(ump->um_devvp); dev_rel(ump->um_dev); mtx_destroy(UFS_MTX(ump)); if (mp->mnt_gjprovider != NULL) { free(mp->mnt_gjprovider, M_UFSMNT); mp->mnt_gjprovider = NULL; } free(fs->fs_csp, M_UFSMNT); free(fs, M_UFSMNT); free(ump, M_UFSMNT); mp->mnt_data = NULL; MNT_ILOCK(mp); mp->mnt_flag &= ~MNT_LOCAL; MNT_IUNLOCK(mp); return (error); fail: if (susp) vfs_write_resume(mp, VR_START_WRITE); fail1: #ifdef UFS_EXTATTR if (e_restart) { ufs_extattr_uepm_init(&ump->um_extattr); #ifdef UFS_EXTATTR_AUTOSTART (void) ufs_extattr_autostart(mp, td); #endif } #endif return (error); } /* * Flush out all the files in a filesystem. */ int ffs_flushfiles(mp, flags, td) struct mount *mp; int flags; struct thread *td; { struct ufsmount *ump; int qerror, error; ump = VFSTOUFS(mp); qerror = 0; #ifdef QUOTA if (mp->mnt_flag & MNT_QUOTA) { int i; error = vflush(mp, 0, SKIPSYSTEM|flags, td); if (error) return (error); for (i = 0; i < MAXQUOTAS; i++) { error = quotaoff(td, mp, i); if (error != 0) { if ((flags & EARLYFLUSH) == 0) return (error); else qerror = error; } } /* * Here we fall through to vflush again to ensure that * we have gotten rid of all the system vnodes, unless * quotas must not be closed. */ } #endif ASSERT_VOP_LOCKED(ump->um_devvp, "ffs_flushfiles"); if (ump->um_devvp->v_vflag & VV_COPYONWRITE) { if ((error = vflush(mp, 0, SKIPSYSTEM | flags, td)) != 0) return (error); ffs_snapshot_unmount(mp); flags |= FORCECLOSE; /* * Here we fall through to vflush again to ensure * that we have gotten rid of all the system vnodes. */ } /* * Do not close system files if quotas were not closed, to be * able to sync the remaining dquots. The freeblks softupdate * workitems might hold a reference on a dquot, preventing * quotaoff() from completing. Next round of * softdep_flushworklist() iteration should process the * blockers, allowing the next run of quotaoff() to finally * flush held dquots. * * Otherwise, flush all the files. */ if (qerror == 0 && (error = vflush(mp, 0, flags, td)) != 0) return (error); /* * Flush filesystem metadata. */ vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); error = VOP_FSYNC(ump->um_devvp, MNT_WAIT, td); VOP_UNLOCK(ump->um_devvp, 0); return (error); } /* * Get filesystem statistics. */ static int ffs_statfs(mp, sbp) struct mount *mp; struct statfs *sbp; { struct ufsmount *ump; struct fs *fs; ump = VFSTOUFS(mp); fs = ump->um_fs; if (fs->fs_magic != FS_UFS1_MAGIC && fs->fs_magic != FS_UFS2_MAGIC) panic("ffs_statfs"); sbp->f_version = STATFS_VERSION; sbp->f_bsize = fs->fs_fsize; sbp->f_iosize = fs->fs_bsize; sbp->f_blocks = fs->fs_dsize; UFS_LOCK(ump); sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag + fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks); sbp->f_bavail = freespace(fs, fs->fs_minfree) + dbtofsb(fs, fs->fs_pendingblocks); sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO; sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes; UFS_UNLOCK(ump); sbp->f_namemax = NAME_MAX; return (0); } /* * For a lazy sync, we only care about access times, quotas and the * superblock. Other filesystem changes are already converted to * cylinder group blocks or inode blocks updates and are written to * disk by syncer. */ static int ffs_sync_lazy(mp) struct mount *mp; { struct vnode *mvp, *vp; struct inode *ip; struct thread *td; int allerror, error; allerror = 0; td = curthread; if ((mp->mnt_flag & MNT_NOATIME) != 0) goto qupdate; MNT_VNODE_FOREACH_ACTIVE(vp, mp, mvp) { if (vp->v_type == VNON) { VI_UNLOCK(vp); continue; } ip = VTOI(vp); /* * The IN_ACCESS flag is converted to IN_MODIFIED by * ufs_close() and ufs_getattr() by the calls to * ufs_itimes_locked(), without subsequent UFS_UPDATE(). * Test also all the other timestamp flags too, to pick up * any other cases that could be missed. */ if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0) { VI_UNLOCK(vp); continue; } if ((error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td)) != 0) continue; error = ffs_update(vp, 0); if (error != 0) allerror = error; vput(vp); } qupdate: #ifdef QUOTA qsync(mp); #endif if (VFSTOUFS(mp)->um_fs->fs_fmod != 0 && (error = ffs_sbupdate(VFSTOUFS(mp), MNT_LAZY, 0)) != 0) allerror = error; return (allerror); } /* * Go through the disk queues to initiate sandbagged IO; * go through the inodes to write those that have been modified; * initiate the writing of the super block if it has been modified. * * Note: we are always called with the filesystem marked busy using * vfs_busy(). */ static int ffs_sync(mp, waitfor) struct mount *mp; int waitfor; { struct vnode *mvp, *vp, *devvp; struct thread *td; struct inode *ip; struct ufsmount *ump = VFSTOUFS(mp); struct fs *fs; int error, count, wait, lockreq, allerror = 0; int suspend; int suspended; int secondary_writes; int secondary_accwrites; int softdep_deps; int softdep_accdeps; struct bufobj *bo; wait = 0; suspend = 0; suspended = 0; td = curthread; fs = ump->um_fs; if (fs->fs_fmod != 0 && fs->fs_ronly != 0 && ump->um_fsckpid == 0) panic("%s: ffs_sync: modification on read-only filesystem", fs->fs_fsmnt); if (waitfor == MNT_LAZY) { if (!rebooting) return (ffs_sync_lazy(mp)); waitfor = MNT_NOWAIT; } /* * Write back each (modified) inode. */ lockreq = LK_EXCLUSIVE | LK_NOWAIT; if (waitfor == MNT_SUSPEND) { suspend = 1; waitfor = MNT_WAIT; } if (waitfor == MNT_WAIT) { wait = 1; lockreq = LK_EXCLUSIVE; } lockreq |= LK_INTERLOCK | LK_SLEEPFAIL; loop: /* Grab snapshot of secondary write counts */ MNT_ILOCK(mp); secondary_writes = mp->mnt_secondary_writes; secondary_accwrites = mp->mnt_secondary_accwrites; MNT_IUNLOCK(mp); /* Grab snapshot of softdep dependency counts */ softdep_get_depcounts(mp, &softdep_deps, &softdep_accdeps); MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { /* * Depend on the vnode interlock to keep things stable enough * for a quick test. Since there might be hundreds of * thousands of vnodes, we cannot afford even a subroutine * call unless there's a good chance that we have work to do. */ if (vp->v_type == VNON) { VI_UNLOCK(vp); continue; } ip = VTOI(vp); if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && vp->v_bufobj.bo_dirty.bv_cnt == 0) { VI_UNLOCK(vp); continue; } if ((error = vget(vp, lockreq, td)) != 0) { if (error == ENOENT || error == ENOLCK) { MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); goto loop; } continue; } if ((error = ffs_syncvnode(vp, waitfor, 0)) != 0) allerror = error; vput(vp); } /* * Force stale filesystem control information to be flushed. */ if (waitfor == MNT_WAIT || rebooting) { if ((error = softdep_flushworklist(ump->um_mountp, &count, td))) allerror = error; /* Flushed work items may create new vnodes to clean */ if (allerror == 0 && count) goto loop; } #ifdef QUOTA qsync(mp); #endif devvp = ump->um_devvp; bo = &devvp->v_bufobj; BO_LOCK(bo); if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0) { BO_UNLOCK(bo); vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); error = VOP_FSYNC(devvp, waitfor, td); VOP_UNLOCK(devvp, 0); if (MOUNTEDSOFTDEP(mp) && (error == 0 || error == EAGAIN)) error = ffs_sbupdate(ump, waitfor, 0); if (error != 0) allerror = error; if (allerror == 0 && waitfor == MNT_WAIT) goto loop; } else if (suspend != 0) { if (softdep_check_suspend(mp, devvp, softdep_deps, softdep_accdeps, secondary_writes, secondary_accwrites) != 0) { MNT_IUNLOCK(mp); goto loop; /* More work needed */ } mtx_assert(MNT_MTX(mp), MA_OWNED); mp->mnt_kern_flag |= MNTK_SUSPEND2 | MNTK_SUSPENDED; MNT_IUNLOCK(mp); suspended = 1; } else BO_UNLOCK(bo); /* * Write back modified superblock. */ if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor, suspended)) != 0) allerror = error; return (allerror); } int ffs_vget(mp, ino, flags, vpp) struct mount *mp; ino_t ino; int flags; struct vnode **vpp; { return (ffs_vgetf(mp, ino, flags, vpp, 0)); } int ffs_vgetf(mp, ino, flags, vpp, ffs_flags) struct mount *mp; ino_t ino; int flags; struct vnode **vpp; int ffs_flags; { struct fs *fs; struct inode *ip; struct ufsmount *ump; struct buf *bp; struct vnode *vp; struct cdev *dev; int error; error = vfs_hash_get(mp, ino, flags, curthread, vpp, NULL, NULL); if (error || *vpp != NULL) return (error); /* * We must promote to an exclusive lock for vnode creation. This * can happen if lookup is passed LOCKSHARED. */ if ((flags & LK_TYPE_MASK) == LK_SHARED) { flags &= ~LK_TYPE_MASK; flags |= LK_EXCLUSIVE; } /* * We do not lock vnode creation as it is believed to be too * expensive for such rare case as simultaneous creation of vnode * for same ino by different processes. We just allow them to race * and check later to decide who wins. Let the race begin! */ ump = VFSTOUFS(mp); dev = ump->um_dev; fs = ump->um_fs; ip = uma_zalloc(uma_inode, M_WAITOK | M_ZERO); /* Allocate a new vnode/inode. */ if (fs->fs_magic == FS_UFS1_MAGIC) error = getnewvnode("ufs", mp, &ffs_vnodeops1, &vp); else error = getnewvnode("ufs", mp, &ffs_vnodeops2, &vp); if (error) { *vpp = NULL; uma_zfree(uma_inode, ip); return (error); } /* * FFS supports recursive locking. */ lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL); VN_LOCK_AREC(vp); vp->v_data = ip; vp->v_bufobj.bo_bsize = fs->fs_bsize; ip->i_vnode = vp; ip->i_ump = ump; ip->i_fs = fs; ip->i_dev = dev; ip->i_number = ino; ip->i_ea_refs = 0; + ip->i_nextclustercg = -1; #ifdef QUOTA { int i; for (i = 0; i < MAXQUOTAS; i++) ip->i_dquot[i] = NODQUOT; } #endif if (ffs_flags & FFSV_FORCEINSMQ) vp->v_vflag |= VV_FORCEINSMQ; error = insmntque(vp, mp); if (error != 0) { uma_zfree(uma_inode, ip); *vpp = NULL; return (error); } vp->v_vflag &= ~VV_FORCEINSMQ; error = vfs_hash_insert(vp, ino, flags, curthread, vpp, NULL, NULL); if (error || *vpp != NULL) return (error); /* Read in the disk contents for the inode, copy into the inode. */ error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), (int)fs->fs_bsize, NOCRED, &bp); if (error) { /* * The inode does not contain anything useful, so it would * be misleading to leave it on its hash chain. With mode * still zero, it will be unlinked and returned to the free * list by vput(). */ brelse(bp); vput(vp); *vpp = NULL; return (error); } if (ip->i_ump->um_fstype == UFS1) ip->i_din1 = uma_zalloc(uma_ufs1, M_WAITOK); else ip->i_din2 = uma_zalloc(uma_ufs2, M_WAITOK); ffs_load_inode(bp, ip, fs, ino); if (DOINGSOFTDEP(vp)) softdep_load_inodeblock(ip); else ip->i_effnlink = ip->i_nlink; bqrelse(bp); /* * Initialize the vnode from the inode, check for aliases. * Note that the underlying vnode may have changed. */ if (ip->i_ump->um_fstype == UFS1) error = ufs_vinit(mp, &ffs_fifoops1, &vp); else error = ufs_vinit(mp, &ffs_fifoops2, &vp); if (error) { vput(vp); *vpp = NULL; return (error); } /* * Finish inode initialization. */ if (vp->v_type != VFIFO) { /* FFS supports shared locking for all files except fifos. */ VN_LOCK_ASHARE(vp); } /* * Set up a generation number for this inode if it does not * already have one. This should only happen on old filesystems. */ if (ip->i_gen == 0) { ip->i_gen = arc4random() / 2 + 1; if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { ip->i_flag |= IN_MODIFIED; DIP_SET(ip, i_gen, ip->i_gen); } } #ifdef MAC if ((mp->mnt_flag & MNT_MULTILABEL) && ip->i_mode) { /* * If this vnode is already allocated, and we're running * multi-label, attempt to perform a label association * from the extended attributes on the inode. */ error = mac_vnode_associate_extattr(mp, vp); if (error) { /* ufs_inactive will release ip->i_devvp ref. */ vput(vp); *vpp = NULL; return (error); } } #endif *vpp = vp; return (0); } /* * File handle to vnode * * Have to be really careful about stale file handles: * - check that the inode number is valid * - call ffs_vget() to get the locked inode * - check for an unallocated inode (i_mode == 0) * - check that the given client host has export rights and return * those rights via. exflagsp and credanonp */ static int ffs_fhtovp(mp, fhp, flags, vpp) struct mount *mp; struct fid *fhp; int flags; struct vnode **vpp; { struct ufid *ufhp; struct fs *fs; ufhp = (struct ufid *)fhp; fs = VFSTOUFS(mp)->um_fs; if (ufhp->ufid_ino < ROOTINO || ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg) return (ESTALE); return (ufs_fhtovp(mp, ufhp, flags, vpp)); } /* * Initialize the filesystem. */ static int ffs_init(vfsp) struct vfsconf *vfsp; { ffs_susp_initialize(); softdep_initialize(); return (ufs_init(vfsp)); } /* * Undo the work of ffs_init(). */ static int ffs_uninit(vfsp) struct vfsconf *vfsp; { int ret; ret = ufs_uninit(vfsp); softdep_uninitialize(); ffs_susp_uninitialize(); return (ret); } /* * Write a superblock and associated information back to disk. */ int ffs_sbupdate(ump, waitfor, suspended) struct ufsmount *ump; int waitfor; int suspended; { struct fs *fs = ump->um_fs; struct buf *sbbp; struct buf *bp; int blks; void *space; int i, size, error, allerror = 0; if (fs->fs_ronly == 1 && (ump->um_mountp->mnt_flag & (MNT_RDONLY | MNT_UPDATE)) != (MNT_RDONLY | MNT_UPDATE) && ump->um_fsckpid == 0) panic("ffs_sbupdate: write read-only filesystem"); /* * We use the superblock's buf to serialize calls to ffs_sbupdate(). */ sbbp = getblk(ump->um_devvp, btodb(fs->fs_sblockloc), (int)fs->fs_sbsize, 0, 0, 0); /* * First write back the summary information. */ blks = howmany(fs->fs_cssize, fs->fs_fsize); space = fs->fs_csp; for (i = 0; i < blks; i += fs->fs_frag) { size = fs->fs_bsize; if (i + fs->fs_frag > blks) size = (blks - i) * fs->fs_fsize; bp = getblk(ump->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 0, 0, 0); bcopy(space, bp->b_data, (u_int)size); space = (char *)space + size; if (suspended) bp->b_flags |= B_VALIDSUSPWRT; if (waitfor != MNT_WAIT) bawrite(bp); else if ((error = bwrite(bp)) != 0) allerror = error; } /* * Now write back the superblock itself. If any errors occurred * up to this point, then fail so that the superblock avoids * being written out as clean. */ if (allerror) { brelse(sbbp); return (allerror); } bp = sbbp; if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_sblockloc != SBLOCK_UFS1 && (fs->fs_old_flags & FS_FLAGS_UPDATED) == 0) { printf("WARNING: %s: correcting fs_sblockloc from %jd to %d\n", fs->fs_fsmnt, fs->fs_sblockloc, SBLOCK_UFS1); fs->fs_sblockloc = SBLOCK_UFS1; } if (fs->fs_magic == FS_UFS2_MAGIC && fs->fs_sblockloc != SBLOCK_UFS2 && (fs->fs_old_flags & FS_FLAGS_UPDATED) == 0) { printf("WARNING: %s: correcting fs_sblockloc from %jd to %d\n", fs->fs_fsmnt, fs->fs_sblockloc, SBLOCK_UFS2); fs->fs_sblockloc = SBLOCK_UFS2; } fs->fs_fmod = 0; fs->fs_time = time_second; if (MOUNTEDSOFTDEP(ump->um_mountp)) softdep_setup_sbupdate(ump, (struct fs *)bp->b_data, bp); bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); ffs_oldfscompat_write((struct fs *)bp->b_data, ump); if (suspended) bp->b_flags |= B_VALIDSUSPWRT; if (waitfor != MNT_WAIT) bawrite(bp); else if ((error = bwrite(bp)) != 0) allerror = error; return (allerror); } static int ffs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp, int attrnamespace, const char *attrname) { #ifdef UFS_EXTATTR return (ufs_extattrctl(mp, cmd, filename_vp, attrnamespace, attrname)); #else return (vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname)); #endif } static void ffs_ifree(struct ufsmount *ump, struct inode *ip) { if (ump->um_fstype == UFS1 && ip->i_din1 != NULL) uma_zfree(uma_ufs1, ip->i_din1); else if (ip->i_din2 != NULL) uma_zfree(uma_ufs2, ip->i_din2); uma_zfree(uma_inode, ip); } static int dobkgrdwrite = 1; SYSCTL_INT(_debug, OID_AUTO, dobkgrdwrite, CTLFLAG_RW, &dobkgrdwrite, 0, "Do background writes (honoring the BV_BKGRDWRITE flag)?"); /* * Complete a background write started from bwrite. */ static void ffs_backgroundwritedone(struct buf *bp) { struct bufobj *bufobj; struct buf *origbp; /* * Find the original buffer that we are writing. */ bufobj = bp->b_bufobj; BO_LOCK(bufobj); if ((origbp = gbincore(bp->b_bufobj, bp->b_lblkno)) == NULL) panic("backgroundwritedone: lost buffer"); BO_UNLOCK(bufobj); /* * Process dependencies then return any unfinished ones. */ pbrelvp(bp); if (!LIST_EMPTY(&bp->b_dep)) buf_complete(bp); #ifdef SOFTUPDATES if (!LIST_EMPTY(&bp->b_dep)) softdep_move_dependencies(bp, origbp); #endif /* * This buffer is marked B_NOCACHE so when it is released * by biodone it will be tossed. */ bp->b_flags |= B_NOCACHE; bp->b_flags &= ~B_CACHE; bufdone(bp); BO_LOCK(bufobj); /* * Clear the BV_BKGRDINPROG flag in the original buffer * and awaken it if it is waiting for the write to complete. * If BV_BKGRDINPROG is not set in the original buffer it must * have been released and re-instantiated - which is not legal. */ KASSERT((origbp->b_vflags & BV_BKGRDINPROG), ("backgroundwritedone: lost buffer2")); origbp->b_vflags &= ~BV_BKGRDINPROG; if (origbp->b_vflags & BV_BKGRDWAIT) { origbp->b_vflags &= ~BV_BKGRDWAIT; wakeup(&origbp->b_xflags); } BO_UNLOCK(bufobj); } /* * Write, release buffer on completion. (Done by iodone * if async). Do not bother writing anything if the buffer * is invalid. * * Note that we set B_CACHE here, indicating that buffer is * fully valid and thus cacheable. This is true even of NFS * now so we set it generally. This could be set either here * or in biodone() since the I/O is synchronous. We put it * here. */ static int ffs_bufwrite(struct buf *bp) { struct buf *newbp; int oldflags; CTR3(KTR_BUF, "bufwrite(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags); if (bp->b_flags & B_INVAL) { brelse(bp); return (0); } oldflags = bp->b_flags; if (!BUF_ISLOCKED(bp)) panic("bufwrite: buffer is not busy???"); /* * If a background write is already in progress, delay * writing this block if it is asynchronous. Otherwise * wait for the background write to complete. */ BO_LOCK(bp->b_bufobj); if (bp->b_vflags & BV_BKGRDINPROG) { if (bp->b_flags & B_ASYNC) { BO_UNLOCK(bp->b_bufobj); bdwrite(bp); return (0); } bp->b_vflags |= BV_BKGRDWAIT; msleep(&bp->b_xflags, BO_LOCKPTR(bp->b_bufobj), PRIBIO, "bwrbg", 0); if (bp->b_vflags & BV_BKGRDINPROG) panic("bufwrite: still writing"); } BO_UNLOCK(bp->b_bufobj); /* * If this buffer is marked for background writing and we * do not have to wait for it, make a copy and write the * copy so as to leave this buffer ready for further use. * * This optimization eats a lot of memory. If we have a page * or buffer shortfall we can't do it. */ if (dobkgrdwrite && (bp->b_xflags & BX_BKGRDWRITE) && (bp->b_flags & B_ASYNC) && !vm_page_count_severe() && !buf_dirty_count_severe()) { KASSERT(bp->b_iodone == NULL, ("bufwrite: needs chained iodone (%p)", bp->b_iodone)); /* get a new block */ newbp = geteblk(bp->b_bufsize, GB_NOWAIT_BD); if (newbp == NULL) goto normal_write; KASSERT((bp->b_flags & B_UNMAPPED) == 0, ("Unmapped cg")); memcpy(newbp->b_data, bp->b_data, bp->b_bufsize); BO_LOCK(bp->b_bufobj); bp->b_vflags |= BV_BKGRDINPROG; BO_UNLOCK(bp->b_bufobj); newbp->b_xflags |= BX_BKGRDMARKER; newbp->b_lblkno = bp->b_lblkno; newbp->b_blkno = bp->b_blkno; newbp->b_offset = bp->b_offset; newbp->b_iodone = ffs_backgroundwritedone; newbp->b_flags |= B_ASYNC; newbp->b_flags &= ~B_INVAL; pbgetvp(bp->b_vp, newbp); #ifdef SOFTUPDATES /* * Move over the dependencies. If there are rollbacks, * leave the parent buffer dirtied as it will need to * be written again. */ if (LIST_EMPTY(&bp->b_dep) || softdep_move_dependencies(bp, newbp) == 0) bundirty(bp); #else bundirty(bp); #endif /* * Initiate write on the copy, release the original. The * BKGRDINPROG flag prevents it from going away until * the background write completes. */ bqrelse(bp); bp = newbp; } else /* Mark the buffer clean */ bundirty(bp); /* Let the normal bufwrite do the rest for us */ normal_write: return (bufwrite(bp)); } static void ffs_geom_strategy(struct bufobj *bo, struct buf *bp) { struct vnode *vp; int error; struct buf *tbp; int nocopy; vp = bo->__bo_vnode; if (bp->b_iocmd == BIO_WRITE) { if ((bp->b_flags & B_VALIDSUSPWRT) == 0 && bp->b_vp != NULL && bp->b_vp->v_mount != NULL && (bp->b_vp->v_mount->mnt_kern_flag & MNTK_SUSPENDED) != 0) panic("ffs_geom_strategy: bad I/O"); nocopy = bp->b_flags & B_NOCOPY; bp->b_flags &= ~(B_VALIDSUSPWRT | B_NOCOPY); if ((vp->v_vflag & VV_COPYONWRITE) && nocopy == 0 && vp->v_rdev->si_snapdata != NULL) { if ((bp->b_flags & B_CLUSTER) != 0) { runningbufwakeup(bp); TAILQ_FOREACH(tbp, &bp->b_cluster.cluster_head, b_cluster.cluster_entry) { error = ffs_copyonwrite(vp, tbp); if (error != 0 && error != EOPNOTSUPP) { bp->b_error = error; bp->b_ioflags |= BIO_ERROR; bufdone(bp); return; } } bp->b_runningbufspace = bp->b_bufsize; atomic_add_long(&runningbufspace, bp->b_runningbufspace); } else { error = ffs_copyonwrite(vp, bp); if (error != 0 && error != EOPNOTSUPP) { bp->b_error = error; bp->b_ioflags |= BIO_ERROR; bufdone(bp); return; } } } #ifdef SOFTUPDATES if ((bp->b_flags & B_CLUSTER) != 0) { TAILQ_FOREACH(tbp, &bp->b_cluster.cluster_head, b_cluster.cluster_entry) { if (!LIST_EMPTY(&tbp->b_dep)) buf_start(tbp); } } else { if (!LIST_EMPTY(&bp->b_dep)) buf_start(bp); } #endif } g_vfs_strategy(bo, bp); } int ffs_own_mount(const struct mount *mp) { if (mp->mnt_op == &ufs_vfsops) return (1); return (0); } #ifdef DDB #ifdef SOFTUPDATES /* defined in ffs_softdep.c */ extern void db_print_ffs(struct ufsmount *ump); DB_SHOW_COMMAND(ffs, db_show_ffs) { struct mount *mp; struct ufsmount *ump; if (have_addr) { ump = VFSTOUFS((struct mount *)addr); db_print_ffs(ump); return; } TAILQ_FOREACH(mp, &mountlist, mnt_list) { if (!strcmp(mp->mnt_stat.f_fstypename, ufs_vfsconf.vfc_name)) db_print_ffs(VFSTOUFS(mp)); } } #endif /* SOFTUPDATES */ #endif /* DDB */ Index: user/ngie/more-tests/sys/ufs/ufs/inode.h =================================================================== --- user/ngie/more-tests/sys/ufs/ufs/inode.h (revision 281974) +++ user/ngie/more-tests/sys/ufs/ufs/inode.h (revision 281975) @@ -1,187 +1,189 @@ /*- * Copyright (c) 1982, 1989, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)inode.h 8.9 (Berkeley) 5/14/95 * $FreeBSD$ */ #ifndef _UFS_UFS_INODE_H_ #define _UFS_UFS_INODE_H_ #include #include #include /* * This must agree with the definition in . */ #define doff_t int32_t /* * The inode is used to describe each active (or recently active) file in the * UFS filesystem. It is composed of two types of information. The first part * is the information that is needed only while the file is active (such as * the identity of the file and linkage to speed its lookup). The second part * is the permanent meta-data associated with the file which is read in * from the permanent dinode from long term storage when the file becomes * active, and is put back when the file is no longer being used. * * An inode may only be changed while holding either the exclusive * vnode lock or the shared vnode lock and the vnode interlock. We use * the latter only for "read" and "get" operations that require * changing i_flag, or a timestamp. This locking protocol allows executing * those operations without having to upgrade the vnode lock from shared to * exclusive. */ struct inode { TAILQ_ENTRY(inode) i_nextsnap; /* snapshot file list. */ struct vnode *i_vnode;/* Vnode associated with this inode. */ struct ufsmount *i_ump;/* Ufsmount point associated with this inode. */ u_int32_t i_flag; /* flags, see below */ struct cdev *i_dev; /* Device associated with the inode. */ ino_t i_number; /* The identity of the inode. */ int i_effnlink; /* i_nlink when I/O completes */ struct fs *i_fs; /* Associated filesystem superblock. */ struct dquot *i_dquot[MAXQUOTAS]; /* Dquot structures. */ /* * Side effects; used during directory lookup. */ int32_t i_count; /* Size of free slot in directory. */ doff_t i_endoff; /* End of useful stuff in directory. */ doff_t i_diroff; /* Offset in dir, where we found last entry. */ doff_t i_offset; /* Offset of free space in directory. */ union { struct dirhash *dirhash; /* Hashing for large directories. */ daddr_t *snapblklist; /* Collect expunged snapshot blocks. */ } i_un; + int i_nextclustercg; /* last cg searched for cluster */ + /* * Data for extended attribute modification. */ u_char *i_ea_area; /* Pointer to malloced copy of EA area */ unsigned i_ea_len; /* Length of i_ea_area */ int i_ea_error; /* First errno in transaction */ int i_ea_refs; /* Number of users of EA area */ /* * Copies from the on-disk dinode itself. */ u_int16_t i_mode; /* IFMT, permissions; see below. */ int16_t i_nlink; /* File link count. */ u_int64_t i_size; /* File byte count. */ u_int32_t i_flags; /* Status flags (chflags). */ u_int64_t i_gen; /* Generation number. */ u_int32_t i_uid; /* File owner. */ u_int32_t i_gid; /* File group. */ /* * The real copy of the on-disk inode. */ union { struct ufs1_dinode *din1; /* UFS1 on-disk dinode. */ struct ufs2_dinode *din2; /* UFS2 on-disk dinode. */ } dinode_u; }; /* * These flags are kept in i_flag. */ #define IN_ACCESS 0x0001 /* Access time update request. */ #define IN_CHANGE 0x0002 /* Inode change time update request. */ #define IN_UPDATE 0x0004 /* Modification time update request. */ #define IN_MODIFIED 0x0008 /* Inode has been modified. */ #define IN_NEEDSYNC 0x0010 /* Inode requires fsync. */ #define IN_LAZYMOD 0x0040 /* Modified, but don't write yet. */ #define IN_LAZYACCESS 0x0100 /* Process IN_ACCESS after the suspension finished */ #define IN_EA_LOCKED 0x0200 #define IN_EA_LOCKWAIT 0x0400 #define IN_TRUNCATED 0x0800 /* Journaled truncation pending. */ #define i_devvp i_ump->um_devvp #define i_umbufobj i_ump->um_bo #define i_dirhash i_un.dirhash #define i_snapblklist i_un.snapblklist #define i_din1 dinode_u.din1 #define i_din2 dinode_u.din2 #ifdef _KERNEL /* * The DIP macro is used to access fields in the dinode that are * not cached in the inode itself. */ #define DIP(ip, field) \ (((ip)->i_ump->um_fstype == UFS1) ? \ (ip)->i_din1->d##field : (ip)->i_din2->d##field) #define DIP_SET(ip, field, val) do { \ if ((ip)->i_ump->um_fstype == UFS1) \ (ip)->i_din1->d##field = (val); \ else \ (ip)->i_din2->d##field = (val); \ } while (0) #define SHORTLINK(ip) \ (((ip)->i_ump->um_fstype == UFS1) ? \ (caddr_t)(ip)->i_din1->di_db : (caddr_t)(ip)->i_din2->di_db) #define IS_SNAPSHOT(ip) ((ip)->i_flags & SF_SNAPSHOT) /* * Structure used to pass around logical block paths generated by * ufs_getlbns and used by truncate and bmap code. */ struct indir { ufs2_daddr_t in_lbn; /* Logical block number. */ int in_off; /* Offset in buffer. */ }; /* Convert between inode pointers and vnode pointers. */ #define VTOI(vp) ((struct inode *)(vp)->v_data) #define ITOV(ip) ((ip)->i_vnode) /* Determine if soft dependencies are being done */ #define DOINGSOFTDEP(vp) ((vp)->v_mount->mnt_flag & (MNT_SOFTDEP | MNT_SUJ)) #define MOUNTEDSOFTDEP(mp) ((mp)->mnt_flag & (MNT_SOFTDEP | MNT_SUJ)) #define DOINGSUJ(vp) ((vp)->v_mount->mnt_flag & MNT_SUJ) #define MOUNTEDSUJ(mp) ((mp)->mnt_flag & MNT_SUJ) /* This overlays the fid structure (see mount.h). */ struct ufid { u_int16_t ufid_len; /* Length of structure. */ u_int16_t ufid_pad; /* Force 32-bit alignment. */ uint32_t ufid_ino; /* File number (ino). */ uint32_t ufid_gen; /* Generation number. */ }; #endif /* _KERNEL */ #endif /* !_UFS_UFS_INODE_H_ */ Index: user/ngie/more-tests/sys/vm/uma_int.h =================================================================== --- user/ngie/more-tests/sys/vm/uma_int.h (revision 281974) +++ user/ngie/more-tests/sys/vm/uma_int.h (revision 281975) @@ -1,432 +1,432 @@ /*- * Copyright (c) 2002-2005, 2009, 2013 Jeffrey Roberson * Copyright (c) 2004, 2005 Bosko Milekic * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice unmodified, this list of conditions, and the following * disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * * $FreeBSD$ * */ /* * This file includes definitions, structures, prototypes, and inlines that * should not be used outside of the actual implementation of UMA. */ /* * Here's a quick description of the relationship between the objects: * * Kegs contain lists of slabs which are stored in either the full bin, empty * bin, or partially allocated bin, to reduce fragmentation. They also contain * the user supplied value for size, which is adjusted for alignment purposes * and rsize is the result of that. The Keg also stores information for * managing a hash of page addresses that maps pages to uma_slab_t structures * for pages that don't have embedded uma_slab_t's. * * The uma_slab_t may be embedded in a UMA_SLAB_SIZE chunk of memory or it may * be allocated off the page from a special slab zone. The free list within a * slab is managed with a bitmask. For item sizes that would yield more than * 10% memory waste we potentially allocate a separate uma_slab_t if this will * improve the number of items per slab that will fit. * * The only really gross cases, with regards to memory waste, are for those * items that are just over half the page size. You can get nearly 50% waste, * so you fall back to the memory footprint of the power of two allocator. I * have looked at memory allocation sizes on many of the machines available to * me, and there does not seem to be an abundance of allocations at this range * so at this time it may not make sense to optimize for it. This can, of * course, be solved with dynamic slab sizes. * * Kegs may serve multiple Zones but by far most of the time they only serve * one. When a Zone is created, a Keg is allocated and setup for it. While * the backing Keg stores slabs, the Zone caches Buckets of items allocated * from the slabs. Each Zone is equipped with an init/fini and ctor/dtor * pair, as well as with its own set of small per-CPU caches, layered above * the Zone's general Bucket cache. * * The PCPU caches are protected by critical sections, and may be accessed * safely only from their associated CPU, while the Zones backed by the same * Keg all share a common Keg lock (to coalesce contention on the backing * slabs). The backing Keg typically only serves one Zone but in the case of * multiple Zones, one of the Zones is considered the Master Zone and all * Zone-related stats from the Keg are done in the Master Zone. For an * example of a Multi-Zone setup, refer to the Mbuf allocation code. */ /* * This is the representation for normal (Non OFFPAGE slab) * * i == item * s == slab pointer * * <---------------- Page (UMA_SLAB_SIZE) ------------------> * ___________________________________________________________ * | _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ___________ | * ||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i| |slab header|| * ||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_| |___________|| * |___________________________________________________________| * * * This is an OFFPAGE slab. These can be larger than UMA_SLAB_SIZE. * * ___________________________________________________________ * | _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ | * ||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i| | * ||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_| | * |___________________________________________________________| * ___________ ^ * |slab header| | * |___________|---* * */ #ifndef VM_UMA_INT_H #define VM_UMA_INT_H #define UMA_SLAB_SIZE PAGE_SIZE /* How big are our slabs? */ #define UMA_SLAB_MASK (PAGE_SIZE - 1) /* Mask to get back to the page */ #define UMA_SLAB_SHIFT PAGE_SHIFT /* Number of bits PAGE_MASK */ #define UMA_BOOT_PAGES 64 /* Pages allocated for startup */ /* Max waste percentage before going to off page slab management */ #define UMA_MAX_WASTE 10 /* * I doubt there will be many cases where this is exceeded. This is the initial * size of the hash table for uma_slabs that are managed off page. This hash * does expand by powers of two. Currently it doesn't get smaller. */ #define UMA_HASH_SIZE_INIT 32 /* * I should investigate other hashing algorithms. This should yield a low * number of collisions if the pages are relatively contiguous. */ #define UMA_HASH(h, s) ((((uintptr_t)s) >> UMA_SLAB_SHIFT) & (h)->uh_hashmask) #define UMA_HASH_INSERT(h, s, mem) \ SLIST_INSERT_HEAD(&(h)->uh_slab_hash[UMA_HASH((h), \ (mem))], (s), us_hlink) #define UMA_HASH_REMOVE(h, s, mem) \ SLIST_REMOVE(&(h)->uh_slab_hash[UMA_HASH((h), \ (mem))], (s), uma_slab, us_hlink) /* Hash table for freed address -> slab translation */ SLIST_HEAD(slabhead, uma_slab); struct uma_hash { struct slabhead *uh_slab_hash; /* Hash table for slabs */ int uh_hashsize; /* Current size of the hash table */ int uh_hashmask; /* Mask used during hashing */ }; /* * align field or structure to cache line */ #if defined(__amd64__) #define UMA_ALIGN __aligned(CACHE_LINE_SIZE) #else #define UMA_ALIGN #endif /* * Structures for per cpu queues. */ struct uma_bucket { LIST_ENTRY(uma_bucket) ub_link; /* Link into the zone */ int16_t ub_cnt; /* Count of free items. */ int16_t ub_entries; /* Max items. */ void *ub_bucket[]; /* actual allocation storage */ }; typedef struct uma_bucket * uma_bucket_t; struct uma_cache { uma_bucket_t uc_freebucket; /* Bucket we're freeing to */ uma_bucket_t uc_allocbucket; /* Bucket to allocate from */ uint64_t uc_allocs; /* Count of allocations */ uint64_t uc_frees; /* Count of frees */ } UMA_ALIGN; typedef struct uma_cache * uma_cache_t; /* * Keg management structure * * TODO: Optimize for cache line size * */ struct uma_keg { struct mtx_padalign uk_lock; /* Lock for the keg */ struct uma_hash uk_hash; LIST_HEAD(,uma_zone) uk_zones; /* Keg's zones */ LIST_HEAD(,uma_slab) uk_part_slab; /* partially allocated slabs */ LIST_HEAD(,uma_slab) uk_free_slab; /* empty slab list */ LIST_HEAD(,uma_slab) uk_full_slab; /* full slabs */ uint32_t uk_align; /* Alignment mask */ uint32_t uk_pages; /* Total page count */ uint32_t uk_free; /* Count of items free in slabs */ uint32_t uk_reserve; /* Number of reserved items. */ uint32_t uk_size; /* Requested size of each item */ uint32_t uk_rsize; /* Real size of each item */ uint32_t uk_maxpages; /* Maximum number of pages to alloc */ uma_init uk_init; /* Keg's init routine */ uma_fini uk_fini; /* Keg's fini routine */ uma_alloc uk_allocf; /* Allocation function */ uma_free uk_freef; /* Free routine */ u_long uk_offset; /* Next free offset from base KVA */ vm_offset_t uk_kva; /* Zone base KVA */ uma_zone_t uk_slabzone; /* Slab zone backing us, if OFFPAGE */ uint16_t uk_slabsize; /* Slab size for this keg */ uint16_t uk_pgoff; /* Offset to uma_slab struct */ uint16_t uk_ppera; /* pages per allocation from backend */ uint16_t uk_ipers; /* Items per slab */ uint32_t uk_flags; /* Internal flags */ /* Least used fields go to the last cache line. */ const char *uk_name; /* Name of creating zone. */ LIST_ENTRY(uma_keg) uk_link; /* List of all kegs */ }; typedef struct uma_keg * uma_keg_t; /* * Free bits per-slab. */ #define SLAB_SETSIZE (PAGE_SIZE / UMA_SMALLEST_UNIT) BITSET_DEFINE(slabbits, SLAB_SETSIZE); /* * The slab structure manages a single contiguous allocation from backing * store and subdivides it into individually allocatable items. */ struct uma_slab { uma_keg_t us_keg; /* Keg we live in */ union { LIST_ENTRY(uma_slab) _us_link; /* slabs in zone */ unsigned long _us_size; /* Size of allocation */ } us_type; SLIST_ENTRY(uma_slab) us_hlink; /* Link for hash table */ uint8_t *us_data; /* First item */ struct slabbits us_free; /* Free bitmask. */ #ifdef INVARIANTS struct slabbits us_debugfree; /* Debug bitmask. */ #endif uint16_t us_freecount; /* How many are free? */ uint8_t us_flags; /* Page flags see uma.h */ uint8_t us_pad; /* Pad to 32bits, unused. */ }; #define us_link us_type._us_link #define us_size us_type._us_size /* * The slab structure for UMA_ZONE_REFCNT zones for whose items we * maintain reference counters in the slab for. */ struct uma_slab_refcnt { struct uma_slab us_head; /* slab header data */ uint32_t us_refcnt[0]; /* Actually larger. */ }; typedef struct uma_slab * uma_slab_t; typedef struct uma_slab_refcnt * uma_slabrefcnt_t; typedef uma_slab_t (*uma_slaballoc)(uma_zone_t, uma_keg_t, int); struct uma_klink { LIST_ENTRY(uma_klink) kl_link; uma_keg_t kl_keg; }; typedef struct uma_klink *uma_klink_t; /* * Zone management structure * * TODO: Optimize for cache line size * */ struct uma_zone { struct mtx_padalign uz_lock; /* Lock for the zone */ struct mtx_padalign *uz_lockptr; const char *uz_name; /* Text name of the zone */ LIST_ENTRY(uma_zone) uz_link; /* List of all zones in keg */ LIST_HEAD(,uma_bucket) uz_buckets; /* full buckets */ LIST_HEAD(,uma_klink) uz_kegs; /* List of kegs. */ struct uma_klink uz_klink; /* klink for first keg. */ uma_slaballoc uz_slab; /* Allocate a slab from the backend. */ uma_ctor uz_ctor; /* Constructor for each allocation */ uma_dtor uz_dtor; /* Destructor */ uma_init uz_init; /* Initializer for each item */ uma_fini uz_fini; /* Finalizer for each item. */ uma_import uz_import; /* Import new memory to cache. */ uma_release uz_release; /* Release memory from cache. */ void *uz_arg; /* Import/release argument. */ uint32_t uz_flags; /* Flags inherited from kegs */ uint32_t uz_size; /* Size inherited from kegs */ volatile u_long uz_allocs UMA_ALIGN; /* Total number of allocations */ volatile u_long uz_fails; /* Total number of alloc failures */ volatile u_long uz_frees; /* Total number of frees */ uint64_t uz_sleeps; /* Total number of alloc sleeps */ uint16_t uz_count; /* Amount of items in full bucket */ uint16_t uz_count_min; /* Minimal amount of items there */ /* The next three fields are used to print a rate-limited warnings. */ const char *uz_warning; /* Warning to print on failure */ struct timeval uz_ratecheck; /* Warnings rate-limiting */ /* * This HAS to be the last item because we adjust the zone size * based on NCPU and then allocate the space for the zones. */ - struct uma_cache uz_cpu[]; /* Per cpu caches */ + struct uma_cache uz_cpu[1]; /* Per cpu caches */ }; /* * These flags must not overlap with the UMA_ZONE flags specified in uma.h. */ #define UMA_ZFLAG_MULTI 0x04000000 /* Multiple kegs in the zone. */ #define UMA_ZFLAG_DRAINING 0x08000000 /* Running zone_drain. */ #define UMA_ZFLAG_BUCKET 0x10000000 /* Bucket zone. */ #define UMA_ZFLAG_INTERNAL 0x20000000 /* No offpage no PCPU. */ #define UMA_ZFLAG_FULL 0x40000000 /* Reached uz_maxpages */ #define UMA_ZFLAG_CACHEONLY 0x80000000 /* Don't ask VM for buckets. */ #define UMA_ZFLAG_INHERIT \ (UMA_ZFLAG_INTERNAL | UMA_ZFLAG_CACHEONLY | UMA_ZFLAG_BUCKET) static inline uma_keg_t zone_first_keg(uma_zone_t zone) { uma_klink_t klink; klink = LIST_FIRST(&zone->uz_kegs); return (klink != NULL) ? klink->kl_keg : NULL; } #undef UMA_ALIGN #ifdef _KERNEL /* Internal prototypes */ static __inline uma_slab_t hash_sfind(struct uma_hash *hash, uint8_t *data); void *uma_large_malloc(vm_size_t size, int wait); void uma_large_free(uma_slab_t slab); /* Lock Macros */ #define KEG_LOCK_INIT(k, lc) \ do { \ if ((lc)) \ mtx_init(&(k)->uk_lock, (k)->uk_name, \ (k)->uk_name, MTX_DEF | MTX_DUPOK); \ else \ mtx_init(&(k)->uk_lock, (k)->uk_name, \ "UMA zone", MTX_DEF | MTX_DUPOK); \ } while (0) #define KEG_LOCK_FINI(k) mtx_destroy(&(k)->uk_lock) #define KEG_LOCK(k) mtx_lock(&(k)->uk_lock) #define KEG_UNLOCK(k) mtx_unlock(&(k)->uk_lock) #define ZONE_LOCK_INIT(z, lc) \ do { \ if ((lc)) \ mtx_init(&(z)->uz_lock, (z)->uz_name, \ (z)->uz_name, MTX_DEF | MTX_DUPOK); \ else \ mtx_init(&(z)->uz_lock, (z)->uz_name, \ "UMA zone", MTX_DEF | MTX_DUPOK); \ } while (0) #define ZONE_LOCK(z) mtx_lock((z)->uz_lockptr) #define ZONE_TRYLOCK(z) mtx_trylock((z)->uz_lockptr) #define ZONE_UNLOCK(z) mtx_unlock((z)->uz_lockptr) #define ZONE_LOCK_FINI(z) mtx_destroy(&(z)->uz_lock) /* * Find a slab within a hash table. This is used for OFFPAGE zones to lookup * the slab structure. * * Arguments: * hash The hash table to search. * data The base page of the item. * * Returns: * A pointer to a slab if successful, else NULL. */ static __inline uma_slab_t hash_sfind(struct uma_hash *hash, uint8_t *data) { uma_slab_t slab; int hval; hval = UMA_HASH(hash, data); SLIST_FOREACH(slab, &hash->uh_slab_hash[hval], us_hlink) { if ((uint8_t *)slab->us_data == data) return (slab); } return (NULL); } static __inline uma_slab_t vtoslab(vm_offset_t va) { vm_page_t p; p = PHYS_TO_VM_PAGE(pmap_kextract(va)); return ((uma_slab_t)p->plinks.s.pv); } static __inline void vsetslab(vm_offset_t va, uma_slab_t slab) { vm_page_t p; p = PHYS_TO_VM_PAGE(pmap_kextract(va)); p->plinks.s.pv = slab; } /* * The following two functions may be defined by architecture specific code * if they can provide more effecient allocation functions. This is useful * for using direct mapped addresses. */ void *uma_small_alloc(uma_zone_t zone, vm_size_t bytes, uint8_t *pflag, int wait); void uma_small_free(void *mem, vm_size_t size, uint8_t flags); #endif /* _KERNEL */ #endif /* VM_UMA_INT_H */ Index: user/ngie/more-tests/sys/x86/include/acpica_machdep.h =================================================================== --- user/ngie/more-tests/sys/x86/include/acpica_machdep.h (revision 281974) +++ user/ngie/more-tests/sys/x86/include/acpica_machdep.h (revision 281975) Property changes on: user/ngie/more-tests/sys/x86/include/acpica_machdep.h ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys/x86/include/acpica_machdep.h:r281414-281974 Index: user/ngie/more-tests/sys/x86/x86/mp_x86.c =================================================================== --- user/ngie/more-tests/sys/x86/x86/mp_x86.c (nonexistent) +++ user/ngie/more-tests/sys/x86/x86/mp_x86.c (revision 281975) @@ -0,0 +1,1120 @@ +/*- + * Copyright (c) 1996, by Steve Passe + * Copyright (c) 2003, by Peter Wemm + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. The name of the developer may NOT be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#include +__FBSDID("$FreeBSD$"); + +#ifdef __i386__ +#include "opt_apic.h" +#endif +#include "opt_cpu.h" +#include "opt_kstack_pages.h" +#include "opt_pmap.h" +#include "opt_sched.h" +#include "opt_smp.h" + +#include +#include +#include +#include /* cngetc() */ +#include +#ifdef GPROF +#include +#endif +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#define WARMBOOT_TARGET 0 +#define WARMBOOT_OFF (KERNBASE + 0x0467) +#define WARMBOOT_SEG (KERNBASE + 0x0469) + +#define CMOS_REG (0x70) +#define CMOS_DATA (0x71) +#define BIOS_RESET (0x0f) +#define BIOS_WARM (0x0a) + +/* lock region used by kernel profiling */ +int mcount_lock; + +int mp_naps; /* # of Applications processors */ +int boot_cpu_id = -1; /* designated BSP */ + +extern struct pcpu __pcpu[]; + +/* AP uses this during bootstrap. Do not staticize. */ +char *bootSTK; +int bootAP; + +/* Free these after use */ +void *bootstacks[MAXCPU]; +void *dpcpu; + +struct pcb stoppcbs[MAXCPU]; +struct susppcb **susppcbs; + +#ifdef COUNT_IPIS +/* Interrupt counts. */ +static u_long *ipi_preempt_counts[MAXCPU]; +static u_long *ipi_ast_counts[MAXCPU]; +u_long *ipi_invltlb_counts[MAXCPU]; +u_long *ipi_invlrng_counts[MAXCPU]; +u_long *ipi_invlpg_counts[MAXCPU]; +u_long *ipi_invlcache_counts[MAXCPU]; +u_long *ipi_rendezvous_counts[MAXCPU]; +static u_long *ipi_hardclock_counts[MAXCPU]; +#endif + +/* Default cpu_ops implementation. */ +struct cpu_ops cpu_ops; + +/* + * Local data and functions. + */ + +static volatile cpuset_t ipi_nmi_pending; + +/* used to hold the AP's until we are ready to release them */ +struct mtx ap_boot_mtx; + +/* Set to 1 once we're ready to let the APs out of the pen. */ +volatile int aps_ready = 0; + +/* + * Store data from cpu_add() until later in the boot when we actually setup + * the APs. + */ +struct cpu_info cpu_info[MAX_APIC_ID + 1]; +int cpu_apic_ids[MAXCPU]; +int apic_cpuids[MAX_APIC_ID + 1]; + +/* Holds pending bitmap based IPIs per CPU */ +volatile u_int cpu_ipi_pending[MAXCPU]; + +int cpu_logical; /* logical cpus per core */ +int cpu_cores; /* cores per package */ + +static void release_aps(void *dummy); + +static u_int hyperthreading_cpus; /* logical cpus sharing L1 cache */ +static int hyperthreading_allowed = 1; + +void +mem_range_AP_init(void) +{ + + if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP) + mem_range_softc.mr_op->initAP(&mem_range_softc); +} + +static void +topo_probe_amd(void) +{ + int core_id_bits; + int id; + + /* AMD processors do not support HTT. */ + cpu_logical = 1; + + if ((amd_feature2 & AMDID2_CMP) == 0) { + cpu_cores = 1; + return; + } + + core_id_bits = (cpu_procinfo2 & AMDID_COREID_SIZE) >> + AMDID_COREID_SIZE_SHIFT; + if (core_id_bits == 0) { + cpu_cores = (cpu_procinfo2 & AMDID_CMP_CORES) + 1; + return; + } + + /* Fam 10h and newer should get here. */ + for (id = 0; id <= MAX_APIC_ID; id++) { + /* Check logical CPU availability. */ + if (!cpu_info[id].cpu_present || cpu_info[id].cpu_disabled) + continue; + /* Check if logical CPU has the same package ID. */ + if ((id >> core_id_bits) != (boot_cpu_id >> core_id_bits)) + continue; + cpu_cores++; + } +} + +/* + * Round up to the next power of two, if necessary, and then + * take log2. + * Returns -1 if argument is zero. + */ +static __inline int +mask_width(u_int x) +{ + + return (fls(x << (1 - powerof2(x))) - 1); +} + +static void +topo_probe_0x4(void) +{ + u_int p[4]; + int pkg_id_bits; + int core_id_bits; + int max_cores; + int max_logical; + int id; + + /* Both zero and one here mean one logical processor per package. */ + max_logical = (cpu_feature & CPUID_HTT) != 0 ? + (cpu_procinfo & CPUID_HTT_CORES) >> 16 : 1; + if (max_logical <= 1) + return; + + /* + * Because of uniformity assumption we examine only + * those logical processors that belong to the same + * package as BSP. Further, we count number of + * logical processors that belong to the same core + * as BSP thus deducing number of threads per core. + */ + if (cpu_high >= 0x4) { + cpuid_count(0x04, 0, p); + max_cores = ((p[0] >> 26) & 0x3f) + 1; + } else + max_cores = 1; + core_id_bits = mask_width(max_logical/max_cores); + if (core_id_bits < 0) + return; + pkg_id_bits = core_id_bits + mask_width(max_cores); + + for (id = 0; id <= MAX_APIC_ID; id++) { + /* Check logical CPU availability. */ + if (!cpu_info[id].cpu_present || cpu_info[id].cpu_disabled) + continue; + /* Check if logical CPU has the same package ID. */ + if ((id >> pkg_id_bits) != (boot_cpu_id >> pkg_id_bits)) + continue; + cpu_cores++; + /* Check if logical CPU has the same package and core IDs. */ + if ((id >> core_id_bits) == (boot_cpu_id >> core_id_bits)) + cpu_logical++; + } + + KASSERT(cpu_cores >= 1 && cpu_logical >= 1, + ("topo_probe_0x4 couldn't find BSP")); + + cpu_cores /= cpu_logical; + hyperthreading_cpus = cpu_logical; +} + +static void +topo_probe_0xb(void) +{ + u_int p[4]; + int bits; + int cnt; + int i; + int logical; + int type; + int x; + + /* We only support three levels for now. */ + for (i = 0; i < 3; i++) { + cpuid_count(0x0b, i, p); + + /* Fall back if CPU leaf 11 doesn't really exist. */ + if (i == 0 && p[1] == 0) { + topo_probe_0x4(); + return; + } + + bits = p[0] & 0x1f; + logical = p[1] &= 0xffff; + type = (p[2] >> 8) & 0xff; + if (type == 0 || logical == 0) + break; + /* + * Because of uniformity assumption we examine only + * those logical processors that belong to the same + * package as BSP. + */ + for (cnt = 0, x = 0; x <= MAX_APIC_ID; x++) { + if (!cpu_info[x].cpu_present || + cpu_info[x].cpu_disabled) + continue; + if (x >> bits == boot_cpu_id >> bits) + cnt++; + } + if (type == CPUID_TYPE_SMT) + cpu_logical = cnt; + else if (type == CPUID_TYPE_CORE) + cpu_cores = cnt; + } + if (cpu_logical == 0) + cpu_logical = 1; + cpu_cores /= cpu_logical; +} + +/* + * Both topology discovery code and code that consumes topology + * information assume top-down uniformity of the topology. + * That is, all physical packages must be identical and each + * core in a package must have the same number of threads. + * Topology information is queried only on BSP, on which this + * code runs and for which it can query CPUID information. + * Then topology is extrapolated on all packages using the + * uniformity assumption. + */ +void +topo_probe(void) +{ + static int cpu_topo_probed = 0; + + if (cpu_topo_probed) + return; + + CPU_ZERO(&logical_cpus_mask); + if (mp_ncpus <= 1) + cpu_cores = cpu_logical = 1; + else if (cpu_vendor_id == CPU_VENDOR_AMD) + topo_probe_amd(); + else if (cpu_vendor_id == CPU_VENDOR_INTEL) { + /* + * See Intel(R) 64 Architecture Processor + * Topology Enumeration article for details. + * + * Note that 0x1 <= cpu_high < 4 case should be + * compatible with topo_probe_0x4() logic when + * CPUID.1:EBX[23:16] > 0 (cpu_cores will be 1) + * or it should trigger the fallback otherwise. + */ + if (cpu_high >= 0xb) + topo_probe_0xb(); + else if (cpu_high >= 0x1) + topo_probe_0x4(); + } + + /* + * Fallback: assume each logical CPU is in separate + * physical package. That is, no multi-core, no SMT. + */ + if (cpu_cores == 0 || cpu_logical == 0) + cpu_cores = cpu_logical = 1; + cpu_topo_probed = 1; +} + +struct cpu_group * +cpu_topo(void) +{ + int cg_flags; + + /* + * Determine whether any threading flags are + * necessry. + */ + topo_probe(); + if (cpu_logical > 1 && hyperthreading_cpus) + cg_flags = CG_FLAG_HTT; + else if (cpu_logical > 1) + cg_flags = CG_FLAG_SMT; + else + cg_flags = 0; + if (mp_ncpus % (cpu_cores * cpu_logical) != 0) { + printf("WARNING: Non-uniform processors.\n"); + printf("WARNING: Using suboptimal topology.\n"); + return (smp_topo_none()); + } + /* + * No multi-core or hyper-threaded. + */ + if (cpu_logical * cpu_cores == 1) + return (smp_topo_none()); + /* + * Only HTT no multi-core. + */ + if (cpu_logical > 1 && cpu_cores == 1) + return (smp_topo_1level(CG_SHARE_L1, cpu_logical, cg_flags)); + /* + * Only multi-core no HTT. + */ + if (cpu_cores > 1 && cpu_logical == 1) + return (smp_topo_1level(CG_SHARE_L2, cpu_cores, cg_flags)); + /* + * Both HTT and multi-core. + */ + return (smp_topo_2level(CG_SHARE_L2, cpu_cores, + CG_SHARE_L1, cpu_logical, cg_flags)); +} + + +void +cpu_add(u_int apic_id, char boot_cpu) +{ + + if (apic_id > MAX_APIC_ID) { + panic("SMP: APIC ID %d too high", apic_id); + return; + } + KASSERT(cpu_info[apic_id].cpu_present == 0, ("CPU %d added twice", + apic_id)); + cpu_info[apic_id].cpu_present = 1; + if (boot_cpu) { + KASSERT(boot_cpu_id == -1, + ("CPU %d claims to be BSP, but CPU %d already is", apic_id, + boot_cpu_id)); + boot_cpu_id = apic_id; + cpu_info[apic_id].cpu_bsp = 1; + } + if (mp_ncpus < MAXCPU) { + mp_ncpus++; + mp_maxid = mp_ncpus - 1; + } + if (bootverbose) + printf("SMP: Added CPU %d (%s)\n", apic_id, boot_cpu ? "BSP" : + "AP"); +} + +void +cpu_mp_setmaxid(void) +{ + + /* + * mp_maxid should be already set by calls to cpu_add(). + * Just sanity check its value here. + */ + if (mp_ncpus == 0) + KASSERT(mp_maxid == 0, + ("%s: mp_ncpus is zero, but mp_maxid is not", __func__)); + else if (mp_ncpus == 1) + mp_maxid = 0; + else + KASSERT(mp_maxid >= mp_ncpus - 1, + ("%s: counters out of sync: max %d, count %d", __func__, + mp_maxid, mp_ncpus)); +} + +int +cpu_mp_probe(void) +{ + + /* + * Always record BSP in CPU map so that the mbuf init code works + * correctly. + */ + CPU_SETOF(0, &all_cpus); + if (mp_ncpus == 0) { + /* + * No CPUs were found, so this must be a UP system. Setup + * the variables to represent a system with a single CPU + * with an id of 0. + */ + mp_ncpus = 1; + return (0); + } + + /* At least one CPU was found. */ + if (mp_ncpus == 1) { + /* + * One CPU was found, so this must be a UP system with + * an I/O APIC. + */ + mp_maxid = 0; + return (0); + } + + /* At least two CPUs were found. */ + return (1); +} + +/* + * Print various information about the SMP system hardware and setup. + */ +void +cpu_mp_announce(void) +{ + const char *hyperthread; + int i; + + printf("FreeBSD/SMP: %d package(s) x %d core(s)", + mp_ncpus / (cpu_cores * cpu_logical), cpu_cores); + if (hyperthreading_cpus > 1) + printf(" x %d HTT threads", cpu_logical); + else if (cpu_logical > 1) + printf(" x %d SMT threads", cpu_logical); + printf("\n"); + + /* List active CPUs first. */ + printf(" cpu0 (BSP): APIC ID: %2d\n", boot_cpu_id); + for (i = 1; i < mp_ncpus; i++) { + if (cpu_info[cpu_apic_ids[i]].cpu_hyperthread) + hyperthread = "/HT"; + else + hyperthread = ""; + printf(" cpu%d (AP%s): APIC ID: %2d\n", i, hyperthread, + cpu_apic_ids[i]); + } + + /* List disabled CPUs last. */ + for (i = 0; i <= MAX_APIC_ID; i++) { + if (!cpu_info[i].cpu_present || !cpu_info[i].cpu_disabled) + continue; + if (cpu_info[i].cpu_hyperthread) + hyperthread = "/HT"; + else + hyperthread = ""; + printf(" cpu (AP%s): APIC ID: %2d (disabled)\n", hyperthread, + i); + } +} + +void +init_secondary_tail(void) +{ + u_int cpuid; + + /* + * On real hardware, switch to x2apic mode if possible. Do it + * after aps_ready was signalled, to avoid manipulating the + * mode while BSP might still want to send some IPI to us + * (second startup IPI is ignored on modern hardware etc). + */ + lapic_xapic_mode(); + + /* Initialize the PAT MSR. */ + pmap_init_pat(); + + /* set up CPU registers and state */ + cpu_setregs(); + + /* set up SSE/NX */ + initializecpu(); + + /* set up FPU state on the AP */ +#ifdef __amd64__ + fpuinit(); +#else + npxinit(false); +#endif + + if (cpu_ops.cpu_init) + cpu_ops.cpu_init(); + + /* A quick check from sanity claus */ + cpuid = PCPU_GET(cpuid); + if (PCPU_GET(apic_id) != lapic_id()) { + printf("SMP: cpuid = %d\n", cpuid); + printf("SMP: actual apic_id = %d\n", lapic_id()); + printf("SMP: correct apic_id = %d\n", PCPU_GET(apic_id)); + panic("cpuid mismatch! boom!!"); + } + + /* Initialize curthread. */ + KASSERT(PCPU_GET(idlethread) != NULL, ("no idle thread")); + PCPU_SET(curthread, PCPU_GET(idlethread)); + + mca_init(); + + mtx_lock_spin(&ap_boot_mtx); + + /* Init local apic for irq's */ + lapic_setup(1); + + /* Set memory range attributes for this CPU to match the BSP */ + mem_range_AP_init(); + + smp_cpus++; + + CTR1(KTR_SMP, "SMP: AP CPU #%d Launched", cpuid); + printf("SMP: AP CPU #%d Launched!\n", cpuid); + + /* Determine if we are a logical CPU. */ + /* XXX Calculation depends on cpu_logical being a power of 2, e.g. 2 */ + if (cpu_logical > 1 && PCPU_GET(apic_id) % cpu_logical != 0) + CPU_SET(cpuid, &logical_cpus_mask); + + if (bootverbose) + lapic_dump("AP"); + + if (smp_cpus == mp_ncpus) { + /* enable IPI's, tlb shootdown, freezes etc */ + atomic_store_rel_int(&smp_started, 1); + } + +#ifdef __amd64__ + /* + * Enable global pages TLB extension + * This also implicitly flushes the TLB + */ + load_cr4(rcr4() | CR4_PGE); + if (pmap_pcid_enabled) + load_cr4(rcr4() | CR4_PCIDE); + load_ds(_udatasel); + load_es(_udatasel); + load_fs(_ufssel); +#endif + + mtx_unlock_spin(&ap_boot_mtx); + + /* Wait until all the AP's are up. */ + while (smp_started == 0) + ia32_pause(); + + /* Start per-CPU event timers. */ + cpu_initclocks_ap(); + + sched_throw(NULL); + + panic("scheduler returned us to %s", __func__); + /* NOTREACHED */ +} + +/******************************************************************* + * local functions and data + */ + +/* + * We tell the I/O APIC code about all the CPUs we want to receive + * interrupts. If we don't want certain CPUs to receive IRQs we + * can simply not tell the I/O APIC code about them in this function. + * We also do not tell it about the BSP since it tells itself about + * the BSP internally to work with UP kernels and on UP machines. + */ +void +set_interrupt_apic_ids(void) +{ + u_int i, apic_id; + + for (i = 0; i < MAXCPU; i++) { + apic_id = cpu_apic_ids[i]; + if (apic_id == -1) + continue; + if (cpu_info[apic_id].cpu_bsp) + continue; + if (cpu_info[apic_id].cpu_disabled) + continue; + + /* Don't let hyperthreads service interrupts. */ + if (cpu_logical > 1 && + apic_id % cpu_logical != 0) + continue; + + intr_add_cpu(i); + } +} + +/* + * Assign logical CPU IDs to local APICs. + */ +void +assign_cpu_ids(void) +{ + u_int i; + + TUNABLE_INT_FETCH("machdep.hyperthreading_allowed", + &hyperthreading_allowed); + + /* Check for explicitly disabled CPUs. */ + for (i = 0; i <= MAX_APIC_ID; i++) { + if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp) + continue; + + if (hyperthreading_cpus > 1 && i % hyperthreading_cpus != 0) { + cpu_info[i].cpu_hyperthread = 1; + + /* + * Don't use HT CPU if it has been disabled by a + * tunable. + */ + if (hyperthreading_allowed == 0) { + cpu_info[i].cpu_disabled = 1; + continue; + } + } + + /* Don't use this CPU if it has been disabled by a tunable. */ + if (resource_disabled("lapic", i)) { + cpu_info[i].cpu_disabled = 1; + continue; + } + } + + if (hyperthreading_allowed == 0 && hyperthreading_cpus > 1) { + hyperthreading_cpus = 0; + cpu_logical = 1; + } + + /* + * Assign CPU IDs to local APIC IDs and disable any CPUs + * beyond MAXCPU. CPU 0 is always assigned to the BSP. + * + * To minimize confusion for userland, we attempt to number + * CPUs such that all threads and cores in a package are + * grouped together. For now we assume that the BSP is always + * the first thread in a package and just start adding APs + * starting with the BSP's APIC ID. + */ + mp_ncpus = 1; + cpu_apic_ids[0] = boot_cpu_id; + apic_cpuids[boot_cpu_id] = 0; + for (i = boot_cpu_id + 1; i != boot_cpu_id; + i == MAX_APIC_ID ? i = 0 : i++) { + if (!cpu_info[i].cpu_present || cpu_info[i].cpu_bsp || + cpu_info[i].cpu_disabled) + continue; + + if (mp_ncpus < MAXCPU) { + cpu_apic_ids[mp_ncpus] = i; + apic_cpuids[i] = mp_ncpus; + mp_ncpus++; + } else + cpu_info[i].cpu_disabled = 1; + } + KASSERT(mp_maxid >= mp_ncpus - 1, + ("%s: counters out of sync: max %d, count %d", __func__, mp_maxid, + mp_ncpus)); +} + +#ifdef COUNT_XINVLTLB_HITS +u_int xhits_gbl[MAXCPU]; +u_int xhits_pg[MAXCPU]; +u_int xhits_rng[MAXCPU]; +static SYSCTL_NODE(_debug, OID_AUTO, xhits, CTLFLAG_RW, 0, ""); +SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, global, CTLFLAG_RW, &xhits_gbl, + sizeof(xhits_gbl), "IU", ""); +SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, page, CTLFLAG_RW, &xhits_pg, + sizeof(xhits_pg), "IU", ""); +SYSCTL_OPAQUE(_debug_xhits, OID_AUTO, range, CTLFLAG_RW, &xhits_rng, + sizeof(xhits_rng), "IU", ""); + +u_int ipi_global; +u_int ipi_page; +u_int ipi_range; +u_int ipi_range_size; +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_global, CTLFLAG_RW, &ipi_global, 0, ""); +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_page, CTLFLAG_RW, &ipi_page, 0, ""); +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_range, CTLFLAG_RW, &ipi_range, 0, ""); +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_range_size, CTLFLAG_RW, &ipi_range_size, + 0, ""); + +u_int ipi_masked_global; +u_int ipi_masked_page; +u_int ipi_masked_range; +u_int ipi_masked_range_size; +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_global, CTLFLAG_RW, + &ipi_masked_global, 0, ""); +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_page, CTLFLAG_RW, + &ipi_masked_page, 0, ""); +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_range, CTLFLAG_RW, + &ipi_masked_range, 0, ""); +SYSCTL_INT(_debug_xhits, OID_AUTO, ipi_masked_range_size, CTLFLAG_RW, + &ipi_masked_range_size, 0, ""); +#endif /* COUNT_XINVLTLB_HITS */ + +/* + * Init and startup IPI. + */ +void +ipi_startup(int apic_id, int vector) +{ + + /* + * This attempts to follow the algorithm described in the + * Intel Multiprocessor Specification v1.4 in section B.4. + * For each IPI, we allow the local APIC ~20us to deliver the + * IPI. If that times out, we panic. + */ + + /* + * first we do an INIT IPI: this INIT IPI might be run, resetting + * and running the target CPU. OR this INIT IPI might be latched (P5 + * bug), CPU waiting for STARTUP IPI. OR this INIT IPI might be + * ignored. + */ + lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_LEVEL | + APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_INIT, apic_id); + lapic_ipi_wait(100); + + /* Explicitly deassert the INIT IPI. */ + lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_LEVEL | + APIC_LEVEL_DEASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_INIT, + apic_id); + + DELAY(10000); /* wait ~10mS */ + + /* + * next we do a STARTUP IPI: the previous INIT IPI might still be + * latched, (P5 bug) this 1st STARTUP would then terminate + * immediately, and the previously started INIT IPI would continue. OR + * the previous INIT IPI has already run. and this STARTUP IPI will + * run. OR the previous INIT IPI was ignored. and this STARTUP IPI + * will run. + */ + lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_EDGE | + APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_STARTUP | + vector, apic_id); + if (!lapic_ipi_wait(100)) + panic("Failed to deliver first STARTUP IPI to APIC %d", + apic_id); + DELAY(200); /* wait ~200uS */ + + /* + * finally we do a 2nd STARTUP IPI: this 2nd STARTUP IPI should run IF + * the previous STARTUP IPI was cancelled by a latched INIT IPI. OR + * this STARTUP IPI will be ignored, as only ONE STARTUP IPI is + * recognized after hardware RESET or INIT IPI. + */ + lapic_ipi_raw(APIC_DEST_DESTFLD | APIC_TRIGMOD_EDGE | + APIC_LEVEL_ASSERT | APIC_DESTMODE_PHY | APIC_DELMODE_STARTUP | + vector, apic_id); + if (!lapic_ipi_wait(100)) + panic("Failed to deliver second STARTUP IPI to APIC %d", + apic_id); + + DELAY(200); /* wait ~200uS */ +} + +/* + * Send an IPI to specified CPU handling the bitmap logic. + */ +void +ipi_send_cpu(int cpu, u_int ipi) +{ + u_int bitmap, old_pending, new_pending; + + KASSERT(cpu_apic_ids[cpu] != -1, ("IPI to non-existent CPU %d", cpu)); + + if (IPI_IS_BITMAPED(ipi)) { + bitmap = 1 << ipi; + ipi = IPI_BITMAP_VECTOR; + do { + old_pending = cpu_ipi_pending[cpu]; + new_pending = old_pending | bitmap; + } while (!atomic_cmpset_int(&cpu_ipi_pending[cpu], + old_pending, new_pending)); + if (old_pending) + return; + } + lapic_ipi_vectored(ipi, cpu_apic_ids[cpu]); +} + +void +ipi_bitmap_handler(struct trapframe frame) +{ + struct trapframe *oldframe; + struct thread *td; + int cpu = PCPU_GET(cpuid); + u_int ipi_bitmap; + + critical_enter(); + td = curthread; + td->td_intr_nesting_level++; + oldframe = td->td_intr_frame; + td->td_intr_frame = &frame; + ipi_bitmap = atomic_readandclear_int(&cpu_ipi_pending[cpu]); + if (ipi_bitmap & (1 << IPI_PREEMPT)) { +#ifdef COUNT_IPIS + (*ipi_preempt_counts[cpu])++; +#endif + sched_preempt(td); + } + if (ipi_bitmap & (1 << IPI_AST)) { +#ifdef COUNT_IPIS + (*ipi_ast_counts[cpu])++; +#endif + /* Nothing to do for AST */ + } + if (ipi_bitmap & (1 << IPI_HARDCLOCK)) { +#ifdef COUNT_IPIS + (*ipi_hardclock_counts[cpu])++; +#endif + hardclockintr(); + } + td->td_intr_frame = oldframe; + td->td_intr_nesting_level--; + critical_exit(); +} + +/* + * send an IPI to a set of cpus. + */ +void +ipi_selected(cpuset_t cpus, u_int ipi) +{ + int cpu; + + /* + * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit + * of help in order to understand what is the source. + * Set the mask of receiving CPUs for this purpose. + */ + if (ipi == IPI_STOP_HARD) + CPU_OR_ATOMIC(&ipi_nmi_pending, &cpus); + + while ((cpu = CPU_FFS(&cpus)) != 0) { + cpu--; + CPU_CLR(cpu, &cpus); + CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); + ipi_send_cpu(cpu, ipi); + } +} + +/* + * send an IPI to a specific CPU. + */ +void +ipi_cpu(int cpu, u_int ipi) +{ + + /* + * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit + * of help in order to understand what is the source. + * Set the mask of receiving CPUs for this purpose. + */ + if (ipi == IPI_STOP_HARD) + CPU_SET_ATOMIC(cpu, &ipi_nmi_pending); + + CTR3(KTR_SMP, "%s: cpu: %d ipi: %x", __func__, cpu, ipi); + ipi_send_cpu(cpu, ipi); +} + +/* + * send an IPI to all CPUs EXCEPT myself + */ +void +ipi_all_but_self(u_int ipi) +{ + cpuset_t other_cpus; + + other_cpus = all_cpus; + CPU_CLR(PCPU_GET(cpuid), &other_cpus); + if (IPI_IS_BITMAPED(ipi)) { + ipi_selected(other_cpus, ipi); + return; + } + + /* + * IPI_STOP_HARD maps to a NMI and the trap handler needs a bit + * of help in order to understand what is the source. + * Set the mask of receiving CPUs for this purpose. + */ + if (ipi == IPI_STOP_HARD) + CPU_OR_ATOMIC(&ipi_nmi_pending, &other_cpus); + + CTR2(KTR_SMP, "%s: ipi: %x", __func__, ipi); + lapic_ipi_vectored(ipi, APIC_IPI_DEST_OTHERS); +} + +int +ipi_nmi_handler() +{ + u_int cpuid; + + /* + * As long as there is not a simple way to know about a NMI's + * source, if the bitmask for the current CPU is present in + * the global pending bitword an IPI_STOP_HARD has been issued + * and should be handled. + */ + cpuid = PCPU_GET(cpuid); + if (!CPU_ISSET(cpuid, &ipi_nmi_pending)) + return (1); + + CPU_CLR_ATOMIC(cpuid, &ipi_nmi_pending); + cpustop_handler(); + return (0); +} + +/* + * Handle an IPI_STOP by saving our current context and spinning until we + * are resumed. + */ +void +cpustop_handler(void) +{ + u_int cpu; + + cpu = PCPU_GET(cpuid); + + savectx(&stoppcbs[cpu]); + + /* Indicate that we are stopped */ + CPU_SET_ATOMIC(cpu, &stopped_cpus); + + /* Wait for restart */ + while (!CPU_ISSET(cpu, &started_cpus)) + ia32_pause(); + + CPU_CLR_ATOMIC(cpu, &started_cpus); + CPU_CLR_ATOMIC(cpu, &stopped_cpus); + +#if defined(__amd64__) && defined(DDB) + amd64_db_resume_dbreg(); +#endif + + if (cpu == 0 && cpustop_restartfunc != NULL) { + cpustop_restartfunc(); + cpustop_restartfunc = NULL; + } +} + +/* + * Handle an IPI_SUSPEND by saving our current context and spinning until we + * are resumed. + */ +void +cpususpend_handler(void) +{ + u_int cpu; + + mtx_assert(&smp_ipi_mtx, MA_NOTOWNED); + + cpu = PCPU_GET(cpuid); + if (savectx(&susppcbs[cpu]->sp_pcb)) { +#ifdef __amd64__ + fpususpend(susppcbs[cpu]->sp_fpususpend); +#else + npxsuspend(susppcbs[cpu]->sp_fpususpend); +#endif + wbinvd(); + CPU_SET_ATOMIC(cpu, &suspended_cpus); + } else { +#ifdef __amd64__ + fpuresume(susppcbs[cpu]->sp_fpususpend); +#else + npxresume(susppcbs[cpu]->sp_fpususpend); +#endif + pmap_init_pat(); + initializecpu(); + PCPU_SET(switchtime, 0); + PCPU_SET(switchticks, ticks); + + /* Indicate that we are resumed */ + CPU_CLR_ATOMIC(cpu, &suspended_cpus); + } + + /* Wait for resume */ + while (!CPU_ISSET(cpu, &started_cpus)) + ia32_pause(); + + if (cpu_ops.cpu_resume) + cpu_ops.cpu_resume(); +#ifdef __amd64__ + if (vmm_resume_p) + vmm_resume_p(); +#endif + + /* Resume MCA and local APIC */ + lapic_xapic_mode(); + mca_resume(); + lapic_setup(0); + + /* Indicate that we are resumed */ + CPU_CLR_ATOMIC(cpu, &suspended_cpus); + CPU_CLR_ATOMIC(cpu, &started_cpus); +} + + +void +invlcache_handler(void) +{ +#ifdef COUNT_IPIS + (*ipi_invlcache_counts[PCPU_GET(cpuid)])++; +#endif /* COUNT_IPIS */ + + wbinvd(); + atomic_add_int(&smp_tlb_wait, 1); +} + +/* + * This is called once the rest of the system is up and running and we're + * ready to let the AP's out of the pen. + */ +static void +release_aps(void *dummy __unused) +{ + + if (mp_ncpus == 1) + return; + atomic_store_rel_int(&aps_ready, 1); + while (smp_started == 0) + ia32_pause(); +} +SYSINIT(start_aps, SI_SUB_SMP, SI_ORDER_FIRST, release_aps, NULL); + +#ifdef COUNT_IPIS +/* + * Setup interrupt counters for IPI handlers. + */ +static void +mp_ipi_intrcnt(void *dummy) +{ + char buf[64]; + int i; + + CPU_FOREACH(i) { + snprintf(buf, sizeof(buf), "cpu%d:invltlb", i); + intrcnt_add(buf, &ipi_invltlb_counts[i]); + snprintf(buf, sizeof(buf), "cpu%d:invlrng", i); + intrcnt_add(buf, &ipi_invlrng_counts[i]); + snprintf(buf, sizeof(buf), "cpu%d:invlpg", i); + intrcnt_add(buf, &ipi_invlpg_counts[i]); + snprintf(buf, sizeof(buf), "cpu%d:invlcache", i); + intrcnt_add(buf, &ipi_invlcache_counts[i]); + snprintf(buf, sizeof(buf), "cpu%d:preempt", i); + intrcnt_add(buf, &ipi_preempt_counts[i]); + snprintf(buf, sizeof(buf), "cpu%d:ast", i); + intrcnt_add(buf, &ipi_ast_counts[i]); + snprintf(buf, sizeof(buf), "cpu%d:rendezvous", i); + intrcnt_add(buf, &ipi_rendezvous_counts[i]); + snprintf(buf, sizeof(buf), "cpu%d:hardclock", i); + intrcnt_add(buf, &ipi_hardclock_counts[i]); + } +} +SYSINIT(mp_ipi_intrcnt, SI_SUB_INTR, SI_ORDER_MIDDLE, mp_ipi_intrcnt, NULL); +#endif Property changes on: user/ngie/more-tests/sys/x86/x86/mp_x86.c ___________________________________________________________________ Added: svn:keywords ## -0,0 +1 ## +FreeBSD=%H \ No newline at end of property Index: user/ngie/more-tests/sys =================================================================== --- user/ngie/more-tests/sys (revision 281974) +++ user/ngie/more-tests/sys (revision 281975) Property changes on: user/ngie/more-tests/sys ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys:r281925-281974 Index: user/ngie/more-tests/usr.bin/kdump/kdump.c =================================================================== --- user/ngie/more-tests/usr.bin/kdump/kdump.c (revision 281974) +++ user/ngie/more-tests/usr.bin/kdump/kdump.c (revision 281975) @@ -1,2037 +1,2048 @@ /*- * Copyright (c) 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static const char copyright[] = "@(#) Copyright (c) 1988, 1993\n\ The Regents of the University of California. All rights reserved.\n"; #endif /* not lint */ #ifndef lint #if 0 static char sccsid[] = "@(#)kdump.c 8.1 (Berkeley) 6/6/93"; #endif #endif /* not lint */ #include __FBSDID("$FreeBSD$"); #define _KERNEL extern int errno; #include #undef _KERNEL #include #include #include #define _KERNEL #include #undef _KERNEL #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef HAVE_LIBCAPSICUM #include #include #include #include #endif #include #include #include #ifdef HAVE_LIBCAPSICUM #include #endif #include #include #include #include #include #include #include #include #include "ktrace.h" #include "kdump_subr.h" u_int abidump(struct ktr_header *); int fetchprocinfo(struct ktr_header *, u_int *); int fread_tail(void *, int, int); void dumpheader(struct ktr_header *); void ktrsyscall(struct ktr_syscall *, u_int); void ktrsysret(struct ktr_sysret *, u_int); void ktrnamei(char *, int); void hexdump(char *, int, int); void visdump(char *, int, int); void ktrgenio(struct ktr_genio *, int); void ktrpsig(struct ktr_psig *); void ktrcsw(struct ktr_csw *); void ktrcsw_old(struct ktr_csw_old *); void ktruser_malloc(void *); void ktruser_rtld(int, void *); void ktruser(int, void *); void ktrcaprights(cap_rights_t *); void ktrsockaddr(struct sockaddr *); void ktrstat(struct stat *); void ktrstruct(char *, size_t); void ktrcapfail(struct ktr_cap_fail *); void ktrfault(struct ktr_fault *); void ktrfaultend(struct ktr_faultend *); void limitfd(int fd); void usage(void); void ioctlname(unsigned long, int); #define TIMESTAMP_NONE 0x0 #define TIMESTAMP_ABSOLUTE 0x1 #define TIMESTAMP_ELAPSED 0x2 #define TIMESTAMP_RELATIVE 0x4 extern const char *signames[], *syscallnames[]; extern int nsyscalls; static int timestamp, decimal, fancy = 1, suppressdata, tail, threads, maxdata, resolv = 0, abiflag = 0, syscallno = 0; static const char *tracefile = DEF_TRACEFILE; static struct ktr_header ktr_header; #define TIME_FORMAT "%b %e %T %Y" #define eqs(s1, s2) (strcmp((s1), (s2)) == 0) #define print_number(i,n,c) do { \ if (decimal) \ printf("%c%jd", c, (intmax_t)*i); \ else \ printf("%c%#jx", c, (uintmax_t)(u_register_t)*i); \ i++; \ n--; \ c = ','; \ } while (0) #if defined(__amd64__) || defined(__i386__) void linux_ktrsyscall(struct ktr_syscall *); void linux_ktrsysret(struct ktr_sysret *); extern const char *linux_syscallnames[]; #include static int nlinux_syscalls = sizeof(linux_syscallnames) / \ sizeof(linux_syscallnames[0]); /* * from linux.h * Linux syscalls return negative errno's, we do positive and map them */ static int bsd_to_linux_errno[ELAST + 1] = { -0, -1, -2, -3, -4, -5, -6, -7, -8, -9, -10, -35, -12, -13, -14, -15, -16, -17, -18, -19, -20, -21, -22, -23, -24, -25, -26, -27, -28, -29, -30, -31, -32, -33, -34, -11,-115,-114, -88, -89, -90, -91, -92, -93, -94, -95, -96, -97, -98, -99, -100,-101,-102,-103,-104,-105,-106,-107,-108,-109, -110,-111, -40, -36,-112,-113, -39, -11, -87,-122, -116, -66, -6, -6, -6, -6, -6, -37, -38, -9, -6, -6, -43, -42, -75,-125, -84, -95, -16, -74, -72, -67, -71 }; #endif struct proc_info { TAILQ_ENTRY(proc_info) info; u_int sv_flags; pid_t pid; }; static TAILQ_HEAD(trace_procs, proc_info) trace_procs; #ifdef HAVE_LIBCAPSICUM static cap_channel_t *cappwd, *capgrp; #endif static void strerror_init(void) { /* * Cache NLS data before entering capability mode. * XXXPJD: There should be strerror_init() and strsignal_init() in libc. */ (void)catopen("libc", NL_CAT_LOCALE); } static void localtime_init(void) { time_t ltime; /* * Allow localtime(3) to cache /etc/localtime content before entering * capability mode. * XXXPJD: There should be localtime_init() in libc. */ (void)time(<ime); (void)localtime(<ime); } #ifdef HAVE_LIBCAPSICUM static int cappwdgrp_setup(cap_channel_t **cappwdp, cap_channel_t **capgrpp) { cap_channel_t *capcas, *cappwdloc, *capgrploc; const char *cmds[1], *fields[1]; capcas = cap_init(); if (capcas == NULL) { warn("unable to contact casperd"); return (-1); } cappwdloc = cap_service_open(capcas, "system.pwd"); capgrploc = cap_service_open(capcas, "system.grp"); /* Casper capability no longer needed. */ cap_close(capcas); if (cappwdloc == NULL || capgrploc == NULL) { if (cappwdloc == NULL) warn("unable to open system.pwd service"); if (capgrploc == NULL) warn("unable to open system.grp service"); exit(1); } /* Limit system.pwd to only getpwuid() function and pw_name field. */ cmds[0] = "getpwuid"; if (cap_pwd_limit_cmds(cappwdloc, cmds, 1) < 0) err(1, "unable to limit system.pwd service"); fields[0] = "pw_name"; if (cap_pwd_limit_fields(cappwdloc, fields, 1) < 0) err(1, "unable to limit system.pwd service"); /* Limit system.grp to only getgrgid() function and gr_name field. */ cmds[0] = "getgrgid"; if (cap_grp_limit_cmds(capgrploc, cmds, 1) < 0) err(1, "unable to limit system.grp service"); fields[0] = "gr_name"; if (cap_grp_limit_fields(capgrploc, fields, 1) < 0) err(1, "unable to limit system.grp service"); *cappwdp = cappwdloc; *capgrpp = capgrploc; return (0); } #endif /* HAVE_LIBCAPSICUM */ int main(int argc, char *argv[]) { int ch, ktrlen, size; void *m; int trpoints = ALL_POINTS; int drop_logged; pid_t pid = 0; u_int sv_flags; setlocale(LC_CTYPE, ""); timestamp = TIMESTAMP_NONE; while ((ch = getopt(argc,argv,"f:dElm:np:AHRrSsTt:")) != -1) switch (ch) { case 'A': abiflag = 1; break; case 'f': tracefile = optarg; break; case 'd': decimal = 1; break; case 'l': tail = 1; break; case 'm': maxdata = atoi(optarg); break; case 'n': fancy = 0; break; case 'p': pid = atoi(optarg); break; case 'r': resolv = 1; break; case 'S': syscallno = 1; break; case 's': suppressdata = 1; break; case 'E': timestamp |= TIMESTAMP_ELAPSED; break; case 'H': threads = 1; break; case 'R': timestamp |= TIMESTAMP_RELATIVE; break; case 'T': timestamp |= TIMESTAMP_ABSOLUTE; break; case 't': trpoints = getpoints(optarg); if (trpoints < 0) errx(1, "unknown trace point in %s", optarg); break; default: usage(); } if (argc > optind) usage(); m = malloc(size = 1025); if (m == NULL) errx(1, "%s", strerror(ENOMEM)); if (!freopen(tracefile, "r", stdin)) err(1, "%s", tracefile); strerror_init(); localtime_init(); #ifdef HAVE_LIBCAPSICUM if (resolv != 0) { if (cappwdgrp_setup(&cappwd, &capgrp) < 0) { cappwd = NULL; capgrp = NULL; } } if (resolv == 0 || (cappwd != NULL && capgrp != NULL)) { if (cap_enter() < 0 && errno != ENOSYS) err(1, "unable to enter capability mode"); } #else if (resolv == 0) { if (cap_enter() < 0 && errno != ENOSYS) err(1, "unable to enter capability mode"); } #endif limitfd(STDIN_FILENO); limitfd(STDOUT_FILENO); limitfd(STDERR_FILENO); if (cap_sandboxed()) fprintf(stderr, "capability mode sandbox enabled\n"); TAILQ_INIT(&trace_procs); drop_logged = 0; while (fread_tail(&ktr_header, sizeof(struct ktr_header), 1)) { if (ktr_header.ktr_type & KTR_DROP) { ktr_header.ktr_type &= ~KTR_DROP; if (!drop_logged && threads) { printf( "%6jd %6jd %-8.*s Events dropped.\n", (intmax_t)ktr_header.ktr_pid, ktr_header.ktr_tid > 0 ? (intmax_t)ktr_header.ktr_tid : 0, MAXCOMLEN, ktr_header.ktr_comm); drop_logged = 1; } else if (!drop_logged) { printf("%6jd %-8.*s Events dropped.\n", (intmax_t)ktr_header.ktr_pid, MAXCOMLEN, ktr_header.ktr_comm); drop_logged = 1; } } if (trpoints & (1< size) { m = realloc(m, ktrlen+1); if (m == NULL) errx(1, "%s", strerror(ENOMEM)); size = ktrlen; } if (ktrlen && fread_tail(m, ktrlen, 1) == 0) errx(1, "data too short"); if (fetchprocinfo(&ktr_header, (u_int *)m) != 0) continue; sv_flags = abidump(&ktr_header); if (pid && ktr_header.ktr_pid != pid && ktr_header.ktr_tid != pid) continue; if ((trpoints & (1<ktr_type) { case KTR_PROCCTOR: TAILQ_FOREACH(pi, &trace_procs, info) { if (pi->pid == kth->ktr_pid) { TAILQ_REMOVE(&trace_procs, pi, info); break; } } pi = malloc(sizeof(struct proc_info)); if (pi == NULL) errx(1, "%s", strerror(ENOMEM)); pi->sv_flags = *flags; pi->pid = kth->ktr_pid; TAILQ_INSERT_TAIL(&trace_procs, pi, info); return (1); case KTR_PROCDTOR: TAILQ_FOREACH(pi, &trace_procs, info) { if (pi->pid == kth->ktr_pid) { TAILQ_REMOVE(&trace_procs, pi, info); free(pi); break; } } return (1); } return (0); } u_int abidump(struct ktr_header *kth) { struct proc_info *pi; const char *abi; const char *arch; u_int flags = 0; TAILQ_FOREACH(pi, &trace_procs, info) { if (pi->pid == kth->ktr_pid) { flags = pi->sv_flags; break; } } if (abiflag == 0) return (flags); switch (flags & SV_ABI_MASK) { case SV_ABI_LINUX: abi = "L"; break; case SV_ABI_FREEBSD: abi = "F"; break; default: abi = "U"; break; } if (flags != 0) { if (flags & SV_LP64) arch = "64"; else arch = "32"; } else arch = "00"; printf("%s%s ", abi, arch); return (flags); } void dumpheader(struct ktr_header *kth) { static char unknown[64]; static struct timeval prevtime, prevtime_e, temp; const char *type; + const char *sign; switch (kth->ktr_type) { case KTR_SYSCALL: type = "CALL"; break; case KTR_SYSRET: type = "RET "; break; case KTR_NAMEI: type = "NAMI"; break; case KTR_GENIO: type = "GIO "; break; case KTR_PSIG: type = "PSIG"; break; case KTR_CSW: type = "CSW "; break; case KTR_USER: type = "USER"; break; case KTR_STRUCT: type = "STRU"; break; case KTR_SYSCTL: type = "SCTL"; break; case KTR_PROCCTOR: /* FALLTHROUGH */ case KTR_PROCDTOR: return; case KTR_CAPFAIL: type = "CAP "; break; case KTR_FAULT: type = "PFLT"; break; case KTR_FAULTEND: type = "PRET"; break; default: sprintf(unknown, "UNKNOWN(%d)", kth->ktr_type); type = unknown; } /* * The ktr_tid field was previously the ktr_buffer field, which held * the kernel pointer value for the buffer associated with data * following the record header. It now holds a threadid, but only * for trace files after the change. Older trace files still contain * kernel pointers. Detect this and suppress the results by printing * negative tid's as 0. */ if (threads) printf("%6jd %6jd %-8.*s ", (intmax_t)kth->ktr_pid, kth->ktr_tid > 0 ? (intmax_t)kth->ktr_tid : 0, MAXCOMLEN, kth->ktr_comm); else printf("%6jd %-8.*s ", (intmax_t)kth->ktr_pid, MAXCOMLEN, kth->ktr_comm); if (timestamp) { if (timestamp & TIMESTAMP_ABSOLUTE) { printf("%jd.%06ld ", (intmax_t)kth->ktr_time.tv_sec, kth->ktr_time.tv_usec); } if (timestamp & TIMESTAMP_ELAPSED) { if (prevtime_e.tv_sec == 0) prevtime_e = kth->ktr_time; timevalsub(&kth->ktr_time, &prevtime_e); printf("%jd.%06ld ", (intmax_t)kth->ktr_time.tv_sec, kth->ktr_time.tv_usec); timevaladd(&kth->ktr_time, &prevtime_e); } if (timestamp & TIMESTAMP_RELATIVE) { + if (prevtime.tv_sec == 0) + prevtime = kth->ktr_time; temp = kth->ktr_time; timevalsub(&kth->ktr_time, &prevtime); - prevtime = temp; - printf("%jd.%06ld ", (intmax_t)kth->ktr_time.tv_sec, + if ((intmax_t)kth->ktr_time.tv_sec < 0) { + kth->ktr_time = prevtime; + prevtime = temp; + timevalsub(&kth->ktr_time, &prevtime); + sign = "-"; + } else { + prevtime = temp; + sign = ""; + } + printf("%s%jd.%06ld ", sign, (intmax_t)kth->ktr_time.tv_sec, kth->ktr_time.tv_usec); } } printf("%s ", type); } #include #define KTRACE #include #undef KTRACE int nsyscalls = sizeof (syscallnames) / sizeof (syscallnames[0]); void ktrsyscall(struct ktr_syscall *ktr, u_int flags) { int narg = ktr->ktr_narg; register_t *ip; intmax_t arg; if ((flags != 0 && ((flags & SV_ABI_MASK) != SV_ABI_FREEBSD)) || (ktr->ktr_code >= nsyscalls || ktr->ktr_code < 0)) printf("[%d]", ktr->ktr_code); else { printf("%s", syscallnames[ktr->ktr_code]); if (syscallno) printf("[%d]", ktr->ktr_code); } ip = &ktr->ktr_args[0]; if (narg) { char c = '('; if (fancy && (flags == 0 || (flags & SV_ABI_MASK) == SV_ABI_FREEBSD)) { switch (ktr->ktr_code) { case SYS_bindat: case SYS_connectat: case SYS_faccessat: case SYS_fchmodat: case SYS_fchownat: case SYS_fstatat: case SYS_futimesat: case SYS_linkat: case SYS_mkdirat: case SYS_mkfifoat: case SYS_mknodat: case SYS_openat: case SYS_readlinkat: case SYS_renameat: case SYS_unlinkat: case SYS_utimensat: putchar('('); atfdname(*ip, decimal); c = ','; ip++; narg--; break; } switch (ktr->ktr_code) { case SYS_ioctl: { print_number(ip, narg, c); putchar(c); ioctlname(*ip, decimal); c = ','; ip++; narg--; break; } case SYS_ptrace: putchar('('); ptraceopname(*ip); c = ','; ip++; narg--; break; case SYS_access: case SYS_eaccess: case SYS_faccessat: print_number(ip, narg, c); putchar(','); accessmodename(*ip); ip++; narg--; break; case SYS_open: case SYS_openat: print_number(ip, narg, c); putchar(','); flagsandmodename(ip[0], ip[1], decimal); ip += 2; narg -= 2; break; case SYS_wait4: print_number(ip, narg, c); print_number(ip, narg, c); /* * A flags value of zero is valid for * wait4() but not for wait6(), so * handle zero special here. */ if (*ip == 0) { print_number(ip, narg, c); } else { putchar(','); wait6optname(*ip); ip++; narg--; } break; case SYS_wait6: putchar('('); idtypename(*ip, decimal); c = ','; ip++; narg--; print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); wait6optname(*ip); ip++; narg--; break; case SYS_chmod: case SYS_fchmod: case SYS_lchmod: print_number(ip, narg, c); putchar(','); modename(*ip); ip++; narg--; break; case SYS_mknod: case SYS_mknodat: print_number(ip, narg, c); putchar(','); modename(*ip); ip++; narg--; break; case SYS_getfsstat: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); getfsstatflagsname(*ip); ip++; narg--; break; case SYS_mount: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); mountflagsname(*ip); ip++; narg--; break; case SYS_unmount: print_number(ip, narg, c); putchar(','); mountflagsname(*ip); ip++; narg--; break; case SYS_recvmsg: case SYS_sendmsg: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); sendrecvflagsname(*ip); ip++; narg--; break; case SYS_recvfrom: case SYS_sendto: print_number(ip, narg, c); print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); sendrecvflagsname(*ip); ip++; narg--; break; case SYS_chflags: case SYS_fchflags: case SYS_lchflags: print_number(ip, narg, c); putchar(','); modename(*ip); ip++; narg--; break; case SYS_kill: print_number(ip, narg, c); putchar(','); signame(*ip); ip++; narg--; break; case SYS_reboot: putchar('('); rebootoptname(*ip); ip++; narg--; break; case SYS_umask: putchar('('); modename(*ip); ip++; narg--; break; case SYS_msync: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); msyncflagsname(*ip); ip++; narg--; break; #ifdef SYS_freebsd6_mmap case SYS_freebsd6_mmap: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); mmapprotname(*ip); putchar(','); ip++; narg--; mmapflagsname(*ip); ip++; narg--; break; #endif case SYS_mmap: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); mmapprotname(*ip); putchar(','); ip++; narg--; mmapflagsname(*ip); ip++; narg--; break; case SYS_mprotect: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); mmapprotname(*ip); ip++; narg--; break; case SYS_madvise: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); madvisebehavname(*ip); ip++; narg--; break; case SYS_setpriority: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); prioname(*ip); ip++; narg--; break; case SYS_fcntl: print_number(ip, narg, c); putchar(','); fcntlcmdname(ip[0], ip[1], decimal); ip += 2; narg -= 2; break; case SYS_socket: { int sockdomain; putchar('('); sockdomain = *ip; sockdomainname(sockdomain); ip++; narg--; putchar(','); socktypenamewithflags(*ip); ip++; narg--; if (sockdomain == PF_INET || sockdomain == PF_INET6) { putchar(','); sockipprotoname(*ip); ip++; narg--; } c = ','; break; } case SYS_setsockopt: case SYS_getsockopt: print_number(ip, narg, c); putchar(','); sockoptlevelname(*ip, decimal); if (*ip == SOL_SOCKET) { ip++; narg--; putchar(','); sockoptname(*ip); } ip++; narg--; break; #ifdef SYS_freebsd6_lseek case SYS_freebsd6_lseek: print_number(ip, narg, c); /* Hidden 'pad' argument, not in lseek(2) */ print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); whencename(*ip); ip++; narg--; break; #endif case SYS_lseek: print_number(ip, narg, c); /* Hidden 'pad' argument, not in lseek(2) */ print_number(ip, narg, c); putchar(','); whencename(*ip); ip++; narg--; break; case SYS_flock: print_number(ip, narg, c); putchar(','); flockname(*ip); ip++; narg--; break; case SYS_mkfifo: case SYS_mkfifoat: case SYS_mkdir: case SYS_mkdirat: print_number(ip, narg, c); putchar(','); modename(*ip); ip++; narg--; break; case SYS_shutdown: print_number(ip, narg, c); putchar(','); shutdownhowname(*ip); ip++; narg--; break; case SYS_socketpair: putchar('('); sockdomainname(*ip); ip++; narg--; putchar(','); socktypenamewithflags(*ip); ip++; narg--; c = ','; break; case SYS_getrlimit: case SYS_setrlimit: putchar('('); rlimitname(*ip); ip++; narg--; c = ','; break; case SYS_quotactl: print_number(ip, narg, c); putchar(','); quotactlname(*ip); ip++; narg--; c = ','; break; case SYS_nfssvc: putchar('('); nfssvcname(*ip); ip++; narg--; c = ','; break; case SYS_rtprio: putchar('('); rtprioname(*ip); ip++; narg--; c = ','; break; case SYS___semctl: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); semctlname(*ip); ip++; narg--; break; case SYS_semget: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); semgetname(*ip); ip++; narg--; break; case SYS_msgctl: print_number(ip, narg, c); putchar(','); shmctlname(*ip); ip++; narg--; break; case SYS_shmat: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); shmatname(*ip); ip++; narg--; break; case SYS_shmctl: print_number(ip, narg, c); putchar(','); shmctlname(*ip); ip++; narg--; break; case SYS_shm_open: print_number(ip, narg, c); putchar(','); flagsname(ip[0]); printf(",0%o", (unsigned int)ip[1]); ip += 3; narg -= 3; break; case SYS_minherit: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); minheritname(*ip); ip++; narg--; break; case SYS_rfork: putchar('('); rforkname(*ip); ip++; narg--; c = ','; break; case SYS_lio_listio: putchar('('); lio_listioname(*ip); ip++; narg--; c = ','; break; case SYS_mlockall: putchar('('); mlockallname(*ip); ip++; narg--; break; case SYS_sched_setscheduler: print_number(ip, narg, c); putchar(','); schedpolicyname(*ip); ip++; narg--; break; case SYS_sched_get_priority_max: case SYS_sched_get_priority_min: putchar('('); schedpolicyname(*ip); ip++; narg--; break; case SYS_sendfile: print_number(ip, narg, c); print_number(ip, narg, c); print_number(ip, narg, c); print_number(ip, narg, c); print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); sendfileflagsname(*(int *)ip); ip++; narg--; break; case SYS_kldsym: print_number(ip, narg, c); putchar(','); kldsymcmdname(*ip); ip++; narg--; break; case SYS_sigprocmask: putchar('('); sigprocmaskhowname(*ip); ip++; narg--; c = ','; break; case SYS___acl_get_file: case SYS___acl_set_file: case SYS___acl_get_fd: case SYS___acl_set_fd: case SYS___acl_delete_file: case SYS___acl_delete_fd: case SYS___acl_aclcheck_file: case SYS___acl_aclcheck_fd: case SYS___acl_get_link: case SYS___acl_set_link: case SYS___acl_delete_link: case SYS___acl_aclcheck_link: print_number(ip, narg, c); putchar(','); acltypename(*ip); ip++; narg--; break; case SYS_sigaction: putchar('('); signame(*ip); ip++; narg--; c = ','; break; case SYS_extattrctl: print_number(ip, narg, c); putchar(','); extattrctlname(*ip); ip++; narg--; break; case SYS_nmount: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); mountflagsname(*ip); ip++; narg--; break; case SYS_thr_create: print_number(ip, narg, c); print_number(ip, narg, c); putchar(','); thrcreateflagsname(*ip); ip++; narg--; break; case SYS_thr_kill: print_number(ip, narg, c); putchar(','); signame(*ip); ip++; narg--; break; case SYS_kldunloadf: print_number(ip, narg, c); putchar(','); kldunloadfflagsname(*ip); ip++; narg--; break; case SYS_linkat: case SYS_renameat: case SYS_symlinkat: print_number(ip, narg, c); putchar(','); atfdname(*ip, decimal); ip++; narg--; break; case SYS_cap_fcntls_limit: print_number(ip, narg, c); putchar(','); arg = *ip; ip++; narg--; capfcntlname(arg); break; case SYS_posix_fadvise: print_number(ip, narg, c); print_number(ip, narg, c); print_number(ip, narg, c); (void)putchar(','); fadvisebehavname((int)*ip); ip++; narg--; break; case SYS_procctl: putchar('('); idtypename(*ip, decimal); c = ','; ip++; narg--; print_number(ip, narg, c); putchar(','); procctlcmdname(*ip); ip++; narg--; break; case SYS__umtx_op: print_number(ip, narg, c); putchar(','); umtxopname(*ip); switch (*ip) { case UMTX_OP_CV_WAIT: ip++; narg--; putchar(','); umtxcvwaitflags(*ip); break; case UMTX_OP_RW_RDLOCK: ip++; narg--; putchar(','); umtxrwlockflags(*ip); break; } ip++; narg--; } } while (narg > 0) { print_number(ip, narg, c); } putchar(')'); } putchar('\n'); } void ktrsysret(struct ktr_sysret *ktr, u_int flags) { register_t ret = ktr->ktr_retval; int error = ktr->ktr_error; int code = ktr->ktr_code; if ((flags != 0 && ((flags & SV_ABI_MASK) != SV_ABI_FREEBSD)) || (code >= nsyscalls || code < 0)) printf("[%d] ", code); else { printf("%s", syscallnames[code]); if (syscallno) printf("[%d]", code); printf(" "); } if (error == 0) { if (fancy) { printf("%ld", (long)ret); if (ret < 0 || ret > 9) printf("/%#lx", (unsigned long)ret); } else { if (decimal) printf("%ld", (long)ret); else printf("%#lx", (unsigned long)ret); } } else if (error == ERESTART) printf("RESTART"); else if (error == EJUSTRETURN) printf("JUSTRETURN"); else { printf("-1 errno %d", ktr->ktr_error); if (fancy) printf(" %s", strerror(ktr->ktr_error)); } putchar('\n'); } void ktrnamei(char *cp, int len) { printf("\"%.*s\"\n", len, cp); } void hexdump(char *p, int len, int screenwidth) { int n, i; int width; width = 0; do { width += 2; i = 13; /* base offset */ i += (width / 2) + 1; /* spaces every second byte */ i += (width * 2); /* width of bytes */ i += 3; /* " |" */ i += width; /* each byte */ i += 1; /* "|" */ } while (i < screenwidth); width -= 2; for (n = 0; n < len; n += width) { for (i = n; i < n + width; i++) { if ((i % width) == 0) { /* beginning of line */ printf(" 0x%04x", i); } if ((i % 2) == 0) { printf(" "); } if (i < len) printf("%02x", p[i] & 0xff); else printf(" "); } printf(" |"); for (i = n; i < n + width; i++) { if (i >= len) break; if (p[i] >= ' ' && p[i] <= '~') printf("%c", p[i]); else printf("."); } printf("|\n"); } if ((i % width) != 0) printf("\n"); } void visdump(char *dp, int datalen, int screenwidth) { int col = 0; char *cp; int width; char visbuf[5]; printf(" \""); col = 8; for (;datalen > 0; datalen--, dp++) { vis(visbuf, *dp, VIS_CSTYLE, *(dp+1)); cp = visbuf; /* * Keep track of printables and * space chars (like fold(1)). */ if (col == 0) { putchar('\t'); col = 8; } switch(*cp) { case '\n': col = 0; putchar('\n'); continue; case '\t': width = 8 - (col&07); break; default: width = strlen(cp); } if (col + width > (screenwidth-2)) { printf("\\\n\t"); col = 8; } col += width; do { putchar(*cp++); } while (*cp); } if (col == 0) printf(" "); printf("\"\n"); } void ktrgenio(struct ktr_genio *ktr, int len) { int datalen = len - sizeof (struct ktr_genio); char *dp = (char *)ktr + sizeof (struct ktr_genio); static int screenwidth = 0; int i, binary; printf("fd %d %s %d byte%s\n", ktr->ktr_fd, ktr->ktr_rw == UIO_READ ? "read" : "wrote", datalen, datalen == 1 ? "" : "s"); if (suppressdata) return; if (screenwidth == 0) { struct winsize ws; if (fancy && ioctl(fileno(stderr), TIOCGWINSZ, &ws) != -1 && ws.ws_col > 8) screenwidth = ws.ws_col; else screenwidth = 80; } if (maxdata && datalen > maxdata) datalen = maxdata; for (i = 0, binary = 0; i < datalen && binary == 0; i++) { if (dp[i] >= 32 && dp[i] < 127) continue; if (dp[i] == 10 || dp[i] == 13 || dp[i] == 0 || dp[i] == 9) continue; binary = 1; } if (binary) hexdump(dp, datalen, screenwidth); else visdump(dp, datalen, screenwidth); } const char *signames[] = { "NULL", "HUP", "INT", "QUIT", "ILL", "TRAP", "IOT", /* 1 - 6 */ "EMT", "FPE", "KILL", "BUS", "SEGV", "SYS", /* 7 - 12 */ "PIPE", "ALRM", "TERM", "URG", "STOP", "TSTP", /* 13 - 18 */ "CONT", "CHLD", "TTIN", "TTOU", "IO", "XCPU", /* 19 - 24 */ "XFSZ", "VTALRM", "PROF", "WINCH", "29", "USR1", /* 25 - 30 */ "USR2", NULL, /* 31 - 32 */ }; void ktrpsig(struct ktr_psig *psig) { if (psig->signo > 0 && psig->signo < NSIG) printf("SIG%s ", signames[psig->signo]); else printf("SIG %d ", psig->signo); if (psig->action == SIG_DFL) { printf("SIG_DFL code="); sigcodename(psig->signo, psig->code); putchar('\n'); } else { printf("caught handler=0x%lx mask=0x%x code=", (u_long)psig->action, psig->mask.__bits[0]); sigcodename(psig->signo, psig->code); putchar('\n'); } } void ktrcsw_old(struct ktr_csw_old *cs) { printf("%s %s\n", cs->out ? "stop" : "resume", cs->user ? "user" : "kernel"); } void ktrcsw(struct ktr_csw *cs) { printf("%s %s \"%s\"\n", cs->out ? "stop" : "resume", cs->user ? "user" : "kernel", cs->wmesg); } #define UTRACE_DLOPEN_START 1 #define UTRACE_DLOPEN_STOP 2 #define UTRACE_DLCLOSE_START 3 #define UTRACE_DLCLOSE_STOP 4 #define UTRACE_LOAD_OBJECT 5 #define UTRACE_UNLOAD_OBJECT 6 #define UTRACE_ADD_RUNDEP 7 #define UTRACE_PRELOAD_FINISHED 8 #define UTRACE_INIT_CALL 9 #define UTRACE_FINI_CALL 10 #define UTRACE_DLSYM_START 11 #define UTRACE_DLSYM_STOP 12 struct utrace_rtld { char sig[4]; /* 'RTLD' */ int event; void *handle; void *mapbase; size_t mapsize; int refcnt; char name[MAXPATHLEN]; }; void ktruser_rtld(int len, void *p) { struct utrace_rtld *ut = p; unsigned char *cp; void *parent; int mode; switch (ut->event) { case UTRACE_DLOPEN_START: mode = ut->refcnt; printf("dlopen(%s, ", ut->name); switch (mode & RTLD_MODEMASK) { case RTLD_NOW: printf("RTLD_NOW"); break; case RTLD_LAZY: printf("RTLD_LAZY"); break; default: printf("%#x", mode & RTLD_MODEMASK); } if (mode & RTLD_GLOBAL) printf(" | RTLD_GLOBAL"); if (mode & RTLD_TRACE) printf(" | RTLD_TRACE"); if (mode & ~(RTLD_MODEMASK | RTLD_GLOBAL | RTLD_TRACE)) printf(" | %#x", mode & ~(RTLD_MODEMASK | RTLD_GLOBAL | RTLD_TRACE)); printf(")\n"); break; case UTRACE_DLOPEN_STOP: printf("%p = dlopen(%s) ref %d\n", ut->handle, ut->name, ut->refcnt); break; case UTRACE_DLCLOSE_START: printf("dlclose(%p) (%s, %d)\n", ut->handle, ut->name, ut->refcnt); break; case UTRACE_DLCLOSE_STOP: printf("dlclose(%p) finished\n", ut->handle); break; case UTRACE_LOAD_OBJECT: printf("RTLD: loaded %p @ %p - %p (%s)\n", ut->handle, ut->mapbase, (char *)ut->mapbase + ut->mapsize - 1, ut->name); break; case UTRACE_UNLOAD_OBJECT: printf("RTLD: unloaded %p @ %p - %p (%s)\n", ut->handle, ut->mapbase, (char *)ut->mapbase + ut->mapsize - 1, ut->name); break; case UTRACE_ADD_RUNDEP: parent = ut->mapbase; printf("RTLD: %p now depends on %p (%s, %d)\n", parent, ut->handle, ut->name, ut->refcnt); break; case UTRACE_PRELOAD_FINISHED: printf("RTLD: LD_PRELOAD finished\n"); break; case UTRACE_INIT_CALL: printf("RTLD: init %p for %p (%s)\n", ut->mapbase, ut->handle, ut->name); break; case UTRACE_FINI_CALL: printf("RTLD: fini %p for %p (%s)\n", ut->mapbase, ut->handle, ut->name); break; case UTRACE_DLSYM_START: printf("RTLD: dlsym(%p, %s)\n", ut->handle, ut->name); break; case UTRACE_DLSYM_STOP: printf("RTLD: %p = dlsym(%p, %s)\n", ut->mapbase, ut->handle, ut->name); break; default: cp = p; cp += 4; len -= 4; printf("RTLD: %d ", len); while (len--) if (decimal) printf(" %d", *cp++); else printf(" %02x", *cp++); printf("\n"); } } struct utrace_malloc { void *p; size_t s; void *r; }; void ktruser_malloc(void *p) { struct utrace_malloc *ut = p; if (ut->p == (void *)(intptr_t)(-1)) printf("malloc_init()\n"); else if (ut->s == 0) printf("free(%p)\n", ut->p); else if (ut->p == NULL) printf("%p = malloc(%zu)\n", ut->r, ut->s); else printf("%p = realloc(%p, %zu)\n", ut->r, ut->p, ut->s); } void ktruser(int len, void *p) { unsigned char *cp; if (len >= 8 && bcmp(p, "RTLD", 4) == 0) { ktruser_rtld(len, p); return; } if (len == sizeof(struct utrace_malloc)) { ktruser_malloc(p); return; } printf("%d ", len); cp = p; while (len--) if (decimal) printf(" %d", *cp++); else printf(" %02x", *cp++); printf("\n"); } void ktrcaprights(cap_rights_t *rightsp) { printf("cap_rights_t "); capname(rightsp); printf("\n"); } void ktrsockaddr(struct sockaddr *sa) { /* TODO: Support additional address families #include struct sockaddr_natm *natm; #include struct sockaddr_nb *nb; */ char addr[64]; /* * note: ktrstruct() has already verified that sa points to a * buffer at least sizeof(struct sockaddr) bytes long and exactly * sa->sa_len bytes long. */ printf("struct sockaddr { "); sockfamilyname(sa->sa_family); printf(", "); #define check_sockaddr_len(n) \ if (sa_##n.s##n##_len < sizeof(struct sockaddr_##n)) { \ printf("invalid"); \ break; \ } switch(sa->sa_family) { case AF_INET: { struct sockaddr_in sa_in; memset(&sa_in, 0, sizeof(sa_in)); memcpy(&sa_in, sa, sa->sa_len); check_sockaddr_len(in); inet_ntop(AF_INET, &sa_in.sin_addr, addr, sizeof addr); printf("%s:%u", addr, ntohs(sa_in.sin_port)); break; } case AF_INET6: { struct sockaddr_in6 sa_in6; memset(&sa_in6, 0, sizeof(sa_in6)); memcpy(&sa_in6, sa, sa->sa_len); check_sockaddr_len(in6); getnameinfo((struct sockaddr *)&sa_in6, sizeof(sa_in6), addr, sizeof(addr), NULL, 0, NI_NUMERICHOST); printf("[%s]:%u", addr, htons(sa_in6.sin6_port)); break; } case AF_UNIX: { struct sockaddr_un sa_un; memset(&sa_un, 0, sizeof(sa_un)); memcpy(&sa_un, sa, sa->sa_len); printf("%.*s", (int)sizeof(sa_un.sun_path), sa_un.sun_path); break; } default: printf("unknown address family"); } printf(" }\n"); } void ktrstat(struct stat *statp) { char mode[12], timestr[PATH_MAX + 4]; struct passwd *pwd; struct group *grp; struct tm *tm; /* * note: ktrstruct() has already verified that statp points to a * buffer exactly sizeof(struct stat) bytes long. */ printf("struct stat {"); printf("dev=%ju, ino=%ju, ", (uintmax_t)statp->st_dev, (uintmax_t)statp->st_ino); if (resolv == 0) printf("mode=0%jo, ", (uintmax_t)statp->st_mode); else { strmode(statp->st_mode, mode); printf("mode=%s, ", mode); } printf("nlink=%ju, ", (uintmax_t)statp->st_nlink); if (resolv == 0) { pwd = NULL; } else { #ifdef HAVE_LIBCAPSICUM if (cappwd != NULL) pwd = cap_getpwuid(cappwd, statp->st_uid); else #endif pwd = getpwuid(statp->st_uid); } if (pwd == NULL) printf("uid=%ju, ", (uintmax_t)statp->st_uid); else printf("uid=\"%s\", ", pwd->pw_name); if (resolv == 0) { grp = NULL; } else { #ifdef HAVE_LIBCAPSICUM if (capgrp != NULL) grp = cap_getgrgid(capgrp, statp->st_gid); else #endif grp = getgrgid(statp->st_gid); } if (grp == NULL) printf("gid=%ju, ", (uintmax_t)statp->st_gid); else printf("gid=\"%s\", ", grp->gr_name); printf("rdev=%ju, ", (uintmax_t)statp->st_rdev); printf("atime="); if (resolv == 0) printf("%jd", (intmax_t)statp->st_atim.tv_sec); else { tm = localtime(&statp->st_atim.tv_sec); strftime(timestr, sizeof(timestr), TIME_FORMAT, tm); printf("\"%s\"", timestr); } if (statp->st_atim.tv_nsec != 0) printf(".%09ld, ", statp->st_atim.tv_nsec); else printf(", "); printf("stime="); if (resolv == 0) printf("%jd", (intmax_t)statp->st_mtim.tv_sec); else { tm = localtime(&statp->st_mtim.tv_sec); strftime(timestr, sizeof(timestr), TIME_FORMAT, tm); printf("\"%s\"", timestr); } if (statp->st_mtim.tv_nsec != 0) printf(".%09ld, ", statp->st_mtim.tv_nsec); else printf(", "); printf("ctime="); if (resolv == 0) printf("%jd", (intmax_t)statp->st_ctim.tv_sec); else { tm = localtime(&statp->st_ctim.tv_sec); strftime(timestr, sizeof(timestr), TIME_FORMAT, tm); printf("\"%s\"", timestr); } if (statp->st_ctim.tv_nsec != 0) printf(".%09ld, ", statp->st_ctim.tv_nsec); else printf(", "); printf("birthtime="); if (resolv == 0) printf("%jd", (intmax_t)statp->st_birthtim.tv_sec); else { tm = localtime(&statp->st_birthtim.tv_sec); strftime(timestr, sizeof(timestr), TIME_FORMAT, tm); printf("\"%s\"", timestr); } if (statp->st_birthtim.tv_nsec != 0) printf(".%09ld, ", statp->st_birthtim.tv_nsec); else printf(", "); printf("size=%jd, blksize=%ju, blocks=%jd, flags=0x%x", (uintmax_t)statp->st_size, (uintmax_t)statp->st_blksize, (intmax_t)statp->st_blocks, statp->st_flags); printf(" }\n"); } void ktrstruct(char *buf, size_t buflen) { char *name, *data; size_t namelen, datalen; int i; cap_rights_t rights; struct stat sb; struct sockaddr_storage ss; for (name = buf, namelen = 0; namelen < buflen && name[namelen] != '\0'; ++namelen) /* nothing */; if (namelen == buflen) goto invalid; if (name[namelen] != '\0') goto invalid; data = buf + namelen + 1; datalen = buflen - namelen - 1; if (datalen == 0) goto invalid; /* sanity check */ for (i = 0; i < (int)namelen; ++i) if (!isalpha(name[i])) goto invalid; if (strcmp(name, "caprights") == 0) { if (datalen != sizeof(cap_rights_t)) goto invalid; memcpy(&rights, data, datalen); ktrcaprights(&rights); } else if (strcmp(name, "stat") == 0) { if (datalen != sizeof(struct stat)) goto invalid; memcpy(&sb, data, datalen); ktrstat(&sb); } else if (strcmp(name, "sockaddr") == 0) { if (datalen > sizeof(ss)) goto invalid; memcpy(&ss, data, datalen); if (datalen != ss.ss_len) goto invalid; ktrsockaddr((struct sockaddr *)&ss); } else { printf("unknown structure\n"); } return; invalid: printf("invalid record\n"); } void ktrcapfail(struct ktr_cap_fail *ktr) { switch (ktr->cap_type) { case CAPFAIL_NOTCAPABLE: /* operation on fd with insufficient capabilities */ printf("operation requires "); capname(&ktr->cap_needed); printf(", descriptor holds "); capname(&ktr->cap_held); break; case CAPFAIL_INCREASE: /* requested more capabilities than fd already has */ printf("attempt to increase capabilities from "); capname(&ktr->cap_held); printf(" to "); capname(&ktr->cap_needed); break; case CAPFAIL_SYSCALL: /* called restricted syscall */ printf("disallowed system call"); break; case CAPFAIL_LOOKUP: /* used ".." in strict-relative mode */ printf("restricted VFS lookup"); break; default: printf("unknown capability failure: "); capname(&ktr->cap_needed); printf(" "); capname(&ktr->cap_held); break; } printf("\n"); } void ktrfault(struct ktr_fault *ktr) { printf("0x%jx ", (uintmax_t)ktr->vaddr); vmprotname(ktr->type); printf("\n"); } void ktrfaultend(struct ktr_faultend *ktr) { vmresultname(ktr->result); printf("\n"); } #if defined(__amd64__) || defined(__i386__) void linux_ktrsyscall(struct ktr_syscall *ktr) { int narg = ktr->ktr_narg; register_t *ip; if (ktr->ktr_code >= nlinux_syscalls || ktr->ktr_code < 0) printf("[%d]", ktr->ktr_code); else { printf("%s", linux_syscallnames[ktr->ktr_code]); if (syscallno) printf("[%d]", ktr->ktr_code); } ip = &ktr->ktr_args[0]; if (narg) { char c = '('; while (narg > 0) print_number(ip, narg, c); putchar(')'); } putchar('\n'); } void linux_ktrsysret(struct ktr_sysret *ktr) { register_t ret = ktr->ktr_retval; int error = ktr->ktr_error; int code = ktr->ktr_code; if (code >= nlinux_syscalls || code < 0) printf("[%d] ", code); else { printf("%s", linux_syscallnames[code]); if (syscallno) printf("[%d]", code); printf(" "); } if (error == 0) { if (fancy) { printf("%ld", (long)ret); if (ret < 0 || ret > 9) printf("/%#lx", (unsigned long)ret); } else { if (decimal) printf("%ld", (long)ret); else printf("%#lx", (unsigned long)ret); } } else if (error == ERESTART) printf("RESTART"); else if (error == EJUSTRETURN) printf("JUSTRETURN"); else { if (ktr->ktr_error <= ELAST + 1) error = abs(bsd_to_linux_errno[ktr->ktr_error]); else error = 999; printf("-1 errno %d", error); if (fancy) printf(" %s", strerror(ktr->ktr_error)); } putchar('\n'); } #endif void usage(void) { fprintf(stderr, "usage: kdump [-dEnlHRrSsTA] [-f trfile] " "[-m maxdata] [-p pid] [-t trstr]\n"); exit(1); } Index: user/ngie/more-tests/usr.bin/whois/whois.c =================================================================== --- user/ngie/more-tests/usr.bin/whois/whois.c (revision 281974) +++ user/ngie/more-tests/usr.bin/whois/whois.c (revision 281975) @@ -1,368 +1,486 @@ -/* +/*- * Copyright (c) 1980, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static const char copyright[] = "@(#) Copyright (c) 1980, 1993\n\ The Regents of the University of California. All rights reserved.\n"; #endif /* not lint */ #if 0 #ifndef lint static char sccsid[] = "@(#)whois.c 8.1 (Berkeley) 6/6/93"; #endif /* not lint */ #endif #include __FBSDID("$FreeBSD$"); #include #include +#include #include #include #include #include #include #include #include #include #include #include #include +#include +#include #define ABUSEHOST "whois.abuse.net" #define NICHOST "whois.crsnic.net" #define INICHOST "whois.networksolutions.com" #define GNICHOST "whois.nic.gov" #define ANICHOST "whois.arin.net" #define LNICHOST "whois.lacnic.net" #define KNICHOST "whois.krnic.net" #define RNICHOST "whois.ripe.net" #define PNICHOST "whois.apnic.net" #define MNICHOST "whois.ra.net" #define QNICHOST_TAIL ".whois-servers.net" #define BNICHOST "whois.registro.br" #define NORIDHOST "whois.norid.no" #define IANAHOST "whois.iana.org" #define GERMNICHOST "de.whois-servers.net" #define FNICHOST "whois.afrinic.net" #define DEFAULT_PORT "whois" #define WHOIS_SERVER_ID "Whois Server: " #define WHOIS_ORG_SERVER_ID "Registrant Street1:Whois Server:" #define WHOIS_RECURSE 0x01 #define WHOIS_QUICK 0x02 #define ishost(h) (isalnum((unsigned char)h) || h == '.' || h == '-') static const char *ip_whois[] = { LNICHOST, RNICHOST, PNICHOST, BNICHOST, FNICHOST, NULL }; static const char *port = DEFAULT_PORT; static char *choose_server(char *); static struct addrinfo *gethostinfo(char const *host, int exit_on_error); static void s_asprintf(char **ret, const char *format, ...) __printflike(2, 3); static void usage(void); static void whois(const char *, const char *, int); int main(int argc, char *argv[]) { const char *country, *host; char *qnichost; int ch, flags, use_qnichost; #ifdef SOCKS SOCKSinit(argv[0]); #endif country = host = qnichost = NULL; flags = use_qnichost = 0; while ((ch = getopt(argc, argv, "aAbc:fgh:iIklmp:QrR6")) != -1) { switch (ch) { case 'a': host = ANICHOST; break; case 'A': host = PNICHOST; break; case 'b': host = ABUSEHOST; break; case 'c': country = optarg; break; case 'f': host = FNICHOST; break; case 'g': host = GNICHOST; break; case 'h': host = optarg; break; case 'i': host = INICHOST; break; case 'I': host = IANAHOST; break; case 'k': host = KNICHOST; break; case 'l': host = LNICHOST; break; case 'm': host = MNICHOST; break; case 'p': port = optarg; break; case 'Q': flags |= WHOIS_QUICK; break; case 'r': host = RNICHOST; break; case 'R': warnx("-R is deprecated; use '-c ru' instead"); country = "ru"; break; /* Remove in FreeBSD 10 */ case '6': errx(EX_USAGE, "-6 is deprecated; use -[aAflr] instead"); break; case '?': default: usage(); /* NOTREACHED */ } } argc -= optind; argv += optind; if (!argc || (country != NULL && host != NULL)) usage(); /* * If no host or country is specified determine the top level domain * from the query. If the TLD is a number, query ARIN. Otherwise, use * TLD.whois-server.net. If the domain does not contain '.', fall * back to NICHOST. */ if (host == NULL && country == NULL) { if ((host = getenv("RA_SERVER")) == NULL) { use_qnichost = 1; host = NICHOST; if (!(flags & WHOIS_QUICK)) flags |= WHOIS_RECURSE; } } while (argc-- > 0) { if (country != NULL) { s_asprintf(&qnichost, "%s%s", country, QNICHOST_TAIL); whois(*argv, qnichost, flags); } else if (use_qnichost) if ((qnichost = choose_server(*argv)) != NULL) whois(*argv, qnichost, flags); if (qnichost == NULL) whois(*argv, host, flags); free(qnichost); qnichost = NULL; argv++; } exit(0); } /* * This function will remove any trailing periods from domain, after which it * returns a pointer to newly allocated memory containing the whois server to * be queried, or a NULL if the correct server couldn't be determined. The * caller must remember to free(3) the allocated memory. */ static char * choose_server(char *domain) { char *pos, *retval; if (strchr(domain, ':')) { s_asprintf(&retval, "%s", ANICHOST); return (retval); } for (pos = strchr(domain, '\0'); pos > domain && *--pos == '.';) *pos = '\0'; if (*domain == '\0') errx(EX_USAGE, "can't search for a null string"); if (strlen(domain) > sizeof("-NORID")-1 && strcasecmp(domain + strlen(domain) - sizeof("-NORID") + 1, "-NORID") == 0) { s_asprintf(&retval, "%s", NORIDHOST); return (retval); } while (pos > domain && *pos != '.') --pos; if (pos <= domain) return (NULL); if (isdigit((unsigned char)*++pos)) s_asprintf(&retval, "%s", ANICHOST); else s_asprintf(&retval, "%s%s", pos, QNICHOST_TAIL); return (retval); } static struct addrinfo * gethostinfo(char const *host, int exit_on_error) { struct addrinfo hints, *res; int error; memset(&hints, 0, sizeof(hints)); hints.ai_flags = 0; hints.ai_family = AF_UNSPEC; hints.ai_socktype = SOCK_STREAM; error = getaddrinfo(host, port, &hints, &res); if (error) { warnx("%s: %s", host, gai_strerror(error)); if (exit_on_error) exit(EX_NOHOST); return (NULL); } return (res); } /* * Wrapper for asprintf(3) that exits on error. */ static void s_asprintf(char **ret, const char *format, ...) { va_list ap; va_start(ap, format); if (vasprintf(ret, format, ap) == -1) { va_end(ap); err(EX_OSERR, "vasprintf()"); } va_end(ap); } static void whois(const char *query, const char *hostname, int flags) { FILE *fp; struct addrinfo *hostres, *res; char *buf, *host, *nhost, *p; - int i, s; + int i, j, s = -1, count; size_t c, len; + struct pollfd *fds; + int timeout = 180; - s = -1; hostres = gethostinfo(hostname, 1); - for (res = hostres; res; res = res->ai_next) { - s = socket(res->ai_family, res->ai_socktype, res->ai_protocol); + for (res = hostres, count = 0; res; res = res->ai_next) + count++; + + fds = calloc(count, sizeof(*fds)); + if (fds == NULL) + err(EX_OSERR, "calloc()"); + + /* + * Traverse the result list elements and make non-block + * connection attempts. + */ + count = i = 0; + for (res = hostres; res != NULL; res = res->ai_next) { + s = socket(res->ai_family, res->ai_socktype | SOCK_NONBLOCK, + res->ai_protocol); if (s < 0) continue; - if (connect(s, res->ai_addr, res->ai_addrlen) == 0) - break; - close(s); + if (connect(s, res->ai_addr, res->ai_addrlen) < 0) { + if (errno == EINPROGRESS) { + /* Add the socket to poll list */ + fds[i].fd = s; + fds[i].events = POLLERR | POLLHUP | + POLLIN | POLLOUT; + count++; + i++; + } else { + close(s); + s = -1; + + /* + * Poll only if we have something to poll, + * otherwise just go ahead and try next + * address + */ + if (count == 0) + continue; + } + } else + goto done; + + /* + * If we are at the last address, poll until a connection is + * established or we failed all connection attempts. + */ + if (res->ai_next == NULL) + timeout = INFTIM; + + /* + * Poll the watched descriptors for successful connections: + * if we still have more untried resolved addresses, poll only + * once; otherwise, poll until all descriptors have errors, + * which will be considered as ETIMEDOUT later. + */ + do { + int n; + + n = poll(fds, i, timeout); + if (n == 0) { + /* + * No event reported in time. Try with a + * smaller timeout (but cap at 2-3ms) + * after a new host have been added. + */ + if (timeout >= 3) + timeout <<= 1; + + break; + } else if (n < 0) { + /* + * errno here can only be EINTR which we would want + * to clean up and bail out. + */ + s = -1; + goto done; + } + + /* + * Check for the event(s) we have seen. + */ + for (j = 0; j < i; j++) { + if (fds[j].fd == -1 || fds[j].events == 0 || + fds[j].revents == 0) + continue; + if (fds[j].revents & ~(POLLIN | POLLOUT)) { + close(s); + fds[j].fd = -1; + fds[j].events = 0; + count--; + continue; + } else if (fds[j].revents & (POLLIN | POLLOUT)) { + /* Connect succeeded. */ + s = fds[j].fd; + + goto done; + } + + } + } while (timeout == INFTIM && count != 0); } + + /* All attempts were failed */ + s = -1; + if (count == 0) + errno = ETIMEDOUT; + +done: + /* Close all watched fds except the succeeded one */ + for (j = 0; j < i; j++) + if (fds[j].fd != s && fds[j].fd != -1) + close(fds[j].fd); + + if (s != -1) { + /* Restore default blocking behavior. */ + if ((flags = fcntl(s, F_GETFL)) != -1) { + flags &= ~O_NONBLOCK; + if (fcntl(s, F_SETFL, flags) == -1) + err(EX_OSERR, "fcntl()"); + } else + err(EX_OSERR, "fcntl()"); + } + + free(fds); freeaddrinfo(hostres); - if (res == NULL) + if (s == -1) err(EX_OSERR, "connect()"); fp = fdopen(s, "r+"); if (fp == NULL) err(EX_OSERR, "fdopen()"); if (strcmp(hostname, GERMNICHOST) == 0) { fprintf(fp, "-T dn,ace -C US-ASCII %s\r\n", query); } else if (strcmp(hostname, "dk" QNICHOST_TAIL) == 0) { fprintf(fp, "--show-handles %s\r\n", query); } else { fprintf(fp, "%s\r\n", query); } fflush(fp); nhost = NULL; while ((buf = fgetln(fp, &len)) != NULL) { while (len > 0 && isspace((unsigned char)buf[len - 1])) buf[--len] = '\0'; printf("%.*s\n", (int)len, buf); if ((flags & WHOIS_RECURSE) && nhost == NULL) { host = strnstr(buf, WHOIS_SERVER_ID, len); if (host != NULL) { host += sizeof(WHOIS_SERVER_ID) - 1; for (p = host; p < buf + len; p++) { if (!ishost(*p)) { *p = '\0'; break; } } s_asprintf(&nhost, "%.*s", (int)(buf + len - host), host); } else if ((host = strnstr(buf, WHOIS_ORG_SERVER_ID, len)) != NULL) { host += sizeof(WHOIS_ORG_SERVER_ID) - 1; for (p = host; p < buf + len; p++) { if (!ishost(*p)) { *p = '\0'; break; } } s_asprintf(&nhost, "%.*s", (int)(buf + len - host), host); } else if (strcmp(hostname, ANICHOST) == 0) { for (c = 0; c <= len; c++) buf[c] = tolower((unsigned char)buf[c]); for (i = 0; ip_whois[i] != NULL; i++) { if (strnstr(buf, ip_whois[i], len) != NULL) { s_asprintf(&nhost, "%s", ip_whois[i]); break; } } } } } fclose(fp); if (nhost != NULL) { whois(query, nhost, 0); free(nhost); } } static void usage(void) { fprintf(stderr, "usage: whois [-aAbfgiIklmQrR6] [-c country-code | -h hostname] " "[-p port] name ...\n"); exit(EX_USAGE); } Index: user/ngie/more-tests/usr.sbin/bhyve/pci_emul.c =================================================================== --- user/ngie/more-tests/usr.sbin/bhyve/pci_emul.c (revision 281974) +++ user/ngie/more-tests/usr.sbin/bhyve/pci_emul.c (revision 281975) @@ -1,2089 +1,2100 @@ /*- * Copyright (c) 2011 NetApp, Inc. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include "acpi.h" #include "bhyverun.h" #include "inout.h" #include "ioapic.h" #include "mem.h" #include "pci_emul.h" #include "pci_irq.h" #include "pci_lpc.h" #define CONF1_ADDR_PORT 0x0cf8 #define CONF1_DATA_PORT 0x0cfc #define CONF1_ENABLE 0x80000000ul -#define CFGWRITE(pi,off,val,b) \ -do { \ - if ((b) == 1) { \ - pci_set_cfgdata8((pi),(off),(val)); \ - } else if ((b) == 2) { \ - pci_set_cfgdata16((pi),(off),(val)); \ - } else { \ - pci_set_cfgdata32((pi),(off),(val)); \ - } \ -} while (0) - #define MAXBUSES (PCI_BUSMAX + 1) #define MAXSLOTS (PCI_SLOTMAX + 1) #define MAXFUNCS (PCI_FUNCMAX + 1) struct funcinfo { char *fi_name; char *fi_param; struct pci_devinst *fi_devi; }; struct intxinfo { int ii_count; int ii_pirq_pin; int ii_ioapic_irq; }; struct slotinfo { struct intxinfo si_intpins[4]; struct funcinfo si_funcs[MAXFUNCS]; }; struct businfo { uint16_t iobase, iolimit; /* I/O window */ uint32_t membase32, memlimit32; /* mmio window below 4GB */ uint64_t membase64, memlimit64; /* mmio window above 4GB */ struct slotinfo slotinfo[MAXSLOTS]; }; static struct businfo *pci_businfo[MAXBUSES]; SET_DECLARE(pci_devemu_set, struct pci_devemu); static uint64_t pci_emul_iobase; static uint64_t pci_emul_membase32; static uint64_t pci_emul_membase64; #define PCI_EMUL_IOBASE 0x2000 #define PCI_EMUL_IOLIMIT 0x10000 #define PCI_EMUL_ECFG_BASE 0xE0000000 /* 3.5GB */ #define PCI_EMUL_ECFG_SIZE (MAXBUSES * 1024 * 1024) /* 1MB per bus */ SYSRES_MEM(PCI_EMUL_ECFG_BASE, PCI_EMUL_ECFG_SIZE); #define PCI_EMUL_MEMLIMIT32 PCI_EMUL_ECFG_BASE #define PCI_EMUL_MEMBASE64 0xD000000000UL #define PCI_EMUL_MEMLIMIT64 0xFD00000000UL static struct pci_devemu *pci_emul_finddev(char *name); static void pci_lintr_route(struct pci_devinst *pi); static void pci_lintr_update(struct pci_devinst *pi); static void pci_cfgrw(struct vmctx *ctx, int vcpu, int in, int bus, int slot, int func, int coff, int bytes, uint32_t *val); +static __inline void +CFGWRITE(struct pci_devinst *pi, int coff, uint32_t val, int bytes) +{ + + if (bytes == 1) + pci_set_cfgdata8(pi, coff, val); + else if (bytes == 2) + pci_set_cfgdata16(pi, coff, val); + else + pci_set_cfgdata32(pi, coff, val); +} + +static __inline uint32_t +CFGREAD(struct pci_devinst *pi, int coff, int bytes) +{ + + if (bytes == 1) + return (pci_get_cfgdata8(pi, coff)); + else if (bytes == 2) + return (pci_get_cfgdata16(pi, coff)); + else + return (pci_get_cfgdata32(pi, coff)); +} + /* * I/O access */ /* * Slot options are in the form: * * ::,[,] * [:],[,] * * slot is 0..31 * func is 0..7 * emul is a string describing the type of PCI device e.g. virtio-net * config is an optional string, depending on the device, that can be * used for configuration. * Examples are: * 1,virtio-net,tap0 * 3:0,dummy */ static void pci_parse_slot_usage(char *aopt) { fprintf(stderr, "Invalid PCI slot info field \"%s\"\n", aopt); } int pci_parse_slot(char *opt) { struct businfo *bi; struct slotinfo *si; char *emul, *config, *str, *cp; int error, bnum, snum, fnum; error = -1; str = strdup(opt); emul = config = NULL; if ((cp = strchr(str, ',')) != NULL) { *cp = '\0'; emul = cp + 1; if ((cp = strchr(emul, ',')) != NULL) { *cp = '\0'; config = cp + 1; } } else { pci_parse_slot_usage(opt); goto done; } /* :: */ if (sscanf(str, "%d:%d:%d", &bnum, &snum, &fnum) != 3) { bnum = 0; /* : */ if (sscanf(str, "%d:%d", &snum, &fnum) != 2) { fnum = 0; /* */ if (sscanf(str, "%d", &snum) != 1) { snum = -1; } } } if (bnum < 0 || bnum >= MAXBUSES || snum < 0 || snum >= MAXSLOTS || fnum < 0 || fnum >= MAXFUNCS) { pci_parse_slot_usage(opt); goto done; } if (pci_businfo[bnum] == NULL) pci_businfo[bnum] = calloc(1, sizeof(struct businfo)); bi = pci_businfo[bnum]; si = &bi->slotinfo[snum]; if (si->si_funcs[fnum].fi_name != NULL) { fprintf(stderr, "pci slot %d:%d already occupied!\n", snum, fnum); goto done; } if (pci_emul_finddev(emul) == NULL) { fprintf(stderr, "pci slot %d:%d: unknown device \"%s\"\n", snum, fnum, emul); goto done; } error = 0; si->si_funcs[fnum].fi_name = emul; si->si_funcs[fnum].fi_param = config; done: if (error) free(str); return (error); } static int pci_valid_pba_offset(struct pci_devinst *pi, uint64_t offset) { if (offset < pi->pi_msix.pba_offset) return (0); if (offset >= pi->pi_msix.pba_offset + pi->pi_msix.pba_size) { return (0); } return (1); } int pci_emul_msix_twrite(struct pci_devinst *pi, uint64_t offset, int size, uint64_t value) { int msix_entry_offset; int tab_index; char *dest; /* support only 4 or 8 byte writes */ if (size != 4 && size != 8) return (-1); /* * Return if table index is beyond what device supports */ tab_index = offset / MSIX_TABLE_ENTRY_SIZE; if (tab_index >= pi->pi_msix.table_count) return (-1); msix_entry_offset = offset % MSIX_TABLE_ENTRY_SIZE; /* support only aligned writes */ if ((msix_entry_offset % size) != 0) return (-1); dest = (char *)(pi->pi_msix.table + tab_index); dest += msix_entry_offset; if (size == 4) *((uint32_t *)dest) = value; else *((uint64_t *)dest) = value; return (0); } uint64_t pci_emul_msix_tread(struct pci_devinst *pi, uint64_t offset, int size) { char *dest; int msix_entry_offset; int tab_index; uint64_t retval = ~0; /* * The PCI standard only allows 4 and 8 byte accesses to the MSI-X * table but we also allow 1 byte access to accomodate reads from * ddb. */ if (size != 1 && size != 4 && size != 8) return (retval); msix_entry_offset = offset % MSIX_TABLE_ENTRY_SIZE; /* support only aligned reads */ if ((msix_entry_offset % size) != 0) { return (retval); } tab_index = offset / MSIX_TABLE_ENTRY_SIZE; if (tab_index < pi->pi_msix.table_count) { /* valid MSI-X Table access */ dest = (char *)(pi->pi_msix.table + tab_index); dest += msix_entry_offset; if (size == 1) retval = *((uint8_t *)dest); else if (size == 4) retval = *((uint32_t *)dest); else retval = *((uint64_t *)dest); } else if (pci_valid_pba_offset(pi, offset)) { /* return 0 for PBA access */ retval = 0; } return (retval); } int pci_msix_table_bar(struct pci_devinst *pi) { if (pi->pi_msix.table != NULL) return (pi->pi_msix.table_bar); else return (-1); } int pci_msix_pba_bar(struct pci_devinst *pi) { if (pi->pi_msix.table != NULL) return (pi->pi_msix.pba_bar); else return (-1); } static int pci_emul_io_handler(struct vmctx *ctx, int vcpu, int in, int port, int bytes, uint32_t *eax, void *arg) { struct pci_devinst *pdi = arg; struct pci_devemu *pe = pdi->pi_d; uint64_t offset; int i; for (i = 0; i <= PCI_BARMAX; i++) { if (pdi->pi_bar[i].type == PCIBAR_IO && port >= pdi->pi_bar[i].addr && port + bytes <= pdi->pi_bar[i].addr + pdi->pi_bar[i].size) { offset = port - pdi->pi_bar[i].addr; if (in) *eax = (*pe->pe_barread)(ctx, vcpu, pdi, i, offset, bytes); else (*pe->pe_barwrite)(ctx, vcpu, pdi, i, offset, bytes, *eax); return (0); } } return (-1); } static int pci_emul_mem_handler(struct vmctx *ctx, int vcpu, int dir, uint64_t addr, int size, uint64_t *val, void *arg1, long arg2) { struct pci_devinst *pdi = arg1; struct pci_devemu *pe = pdi->pi_d; uint64_t offset; int bidx = (int) arg2; assert(bidx <= PCI_BARMAX); assert(pdi->pi_bar[bidx].type == PCIBAR_MEM32 || pdi->pi_bar[bidx].type == PCIBAR_MEM64); assert(addr >= pdi->pi_bar[bidx].addr && addr + size <= pdi->pi_bar[bidx].addr + pdi->pi_bar[bidx].size); offset = addr - pdi->pi_bar[bidx].addr; if (dir == MEM_F_WRITE) { if (size == 8) { (*pe->pe_barwrite)(ctx, vcpu, pdi, bidx, offset, 4, *val & 0xffffffff); (*pe->pe_barwrite)(ctx, vcpu, pdi, bidx, offset + 4, 4, *val >> 32); } else { (*pe->pe_barwrite)(ctx, vcpu, pdi, bidx, offset, size, *val); } } else { if (size == 8) { *val = (*pe->pe_barread)(ctx, vcpu, pdi, bidx, offset, 4); *val |= (*pe->pe_barread)(ctx, vcpu, pdi, bidx, offset + 4, 4) << 32; } else { *val = (*pe->pe_barread)(ctx, vcpu, pdi, bidx, offset, size); } } return (0); } static int pci_emul_alloc_resource(uint64_t *baseptr, uint64_t limit, uint64_t size, uint64_t *addr) { uint64_t base; assert((size & (size - 1)) == 0); /* must be a power of 2 */ base = roundup2(*baseptr, size); if (base + size <= limit) { *addr = base; *baseptr = base + size; return (0); } else return (-1); } int pci_emul_alloc_bar(struct pci_devinst *pdi, int idx, enum pcibar_type type, uint64_t size) { return (pci_emul_alloc_pbar(pdi, idx, 0, type, size)); } /* * Register (or unregister) the MMIO or I/O region associated with the BAR * register 'idx' of an emulated pci device. */ static void modify_bar_registration(struct pci_devinst *pi, int idx, int registration) { int error; struct inout_port iop; struct mem_range mr; switch (pi->pi_bar[idx].type) { case PCIBAR_IO: bzero(&iop, sizeof(struct inout_port)); iop.name = pi->pi_name; iop.port = pi->pi_bar[idx].addr; iop.size = pi->pi_bar[idx].size; if (registration) { iop.flags = IOPORT_F_INOUT; iop.handler = pci_emul_io_handler; iop.arg = pi; error = register_inout(&iop); } else error = unregister_inout(&iop); break; case PCIBAR_MEM32: case PCIBAR_MEM64: bzero(&mr, sizeof(struct mem_range)); mr.name = pi->pi_name; mr.base = pi->pi_bar[idx].addr; mr.size = pi->pi_bar[idx].size; if (registration) { mr.flags = MEM_F_RW; mr.handler = pci_emul_mem_handler; mr.arg1 = pi; mr.arg2 = idx; error = register_mem(&mr); } else error = unregister_mem(&mr); break; default: error = EINVAL; break; } assert(error == 0); } static void unregister_bar(struct pci_devinst *pi, int idx) { modify_bar_registration(pi, idx, 0); } static void register_bar(struct pci_devinst *pi, int idx) { modify_bar_registration(pi, idx, 1); } /* Are we decoding i/o port accesses for the emulated pci device? */ static int porten(struct pci_devinst *pi) { uint16_t cmd; cmd = pci_get_cfgdata16(pi, PCIR_COMMAND); return (cmd & PCIM_CMD_PORTEN); } /* Are we decoding memory accesses for the emulated pci device? */ static int memen(struct pci_devinst *pi) { uint16_t cmd; cmd = pci_get_cfgdata16(pi, PCIR_COMMAND); return (cmd & PCIM_CMD_MEMEN); } /* * Update the MMIO or I/O address that is decoded by the BAR register. * * If the pci device has enabled the address space decoding then intercept * the address range decoded by the BAR register. */ static void update_bar_address(struct pci_devinst *pi, uint64_t addr, int idx, int type) { int decode; if (pi->pi_bar[idx].type == PCIBAR_IO) decode = porten(pi); else decode = memen(pi); if (decode) unregister_bar(pi, idx); switch (type) { case PCIBAR_IO: case PCIBAR_MEM32: pi->pi_bar[idx].addr = addr; break; case PCIBAR_MEM64: pi->pi_bar[idx].addr &= ~0xffffffffUL; pi->pi_bar[idx].addr |= addr; break; case PCIBAR_MEMHI64: pi->pi_bar[idx].addr &= 0xffffffff; pi->pi_bar[idx].addr |= addr; break; default: assert(0); } if (decode) register_bar(pi, idx); } int pci_emul_alloc_pbar(struct pci_devinst *pdi, int idx, uint64_t hostbase, enum pcibar_type type, uint64_t size) { int error; uint64_t *baseptr, limit, addr, mask, lobits, bar; assert(idx >= 0 && idx <= PCI_BARMAX); if ((size & (size - 1)) != 0) size = 1UL << flsl(size); /* round up to a power of 2 */ /* Enforce minimum BAR sizes required by the PCI standard */ if (type == PCIBAR_IO) { if (size < 4) size = 4; } else { if (size < 16) size = 16; } switch (type) { case PCIBAR_NONE: baseptr = NULL; addr = mask = lobits = 0; break; case PCIBAR_IO: baseptr = &pci_emul_iobase; limit = PCI_EMUL_IOLIMIT; mask = PCIM_BAR_IO_BASE; lobits = PCIM_BAR_IO_SPACE; break; case PCIBAR_MEM64: /* * XXX * Some drivers do not work well if the 64-bit BAR is allocated * above 4GB. Allow for this by allocating small requests under * 4GB unless then allocation size is larger than some arbitrary * number (32MB currently). */ if (size > 32 * 1024 * 1024) { /* * XXX special case for device requiring peer-peer DMA */ if (size == 0x100000000UL) baseptr = &hostbase; else baseptr = &pci_emul_membase64; limit = PCI_EMUL_MEMLIMIT64; mask = PCIM_BAR_MEM_BASE; lobits = PCIM_BAR_MEM_SPACE | PCIM_BAR_MEM_64 | PCIM_BAR_MEM_PREFETCH; break; } else { baseptr = &pci_emul_membase32; limit = PCI_EMUL_MEMLIMIT32; mask = PCIM_BAR_MEM_BASE; lobits = PCIM_BAR_MEM_SPACE | PCIM_BAR_MEM_64; } break; case PCIBAR_MEM32: baseptr = &pci_emul_membase32; limit = PCI_EMUL_MEMLIMIT32; mask = PCIM_BAR_MEM_BASE; lobits = PCIM_BAR_MEM_SPACE | PCIM_BAR_MEM_32; break; default: printf("pci_emul_alloc_base: invalid bar type %d\n", type); assert(0); } if (baseptr != NULL) { error = pci_emul_alloc_resource(baseptr, limit, size, &addr); if (error != 0) return (error); } pdi->pi_bar[idx].type = type; pdi->pi_bar[idx].addr = addr; pdi->pi_bar[idx].size = size; /* Initialize the BAR register in config space */ bar = (addr & mask) | lobits; pci_set_cfgdata32(pdi, PCIR_BAR(idx), bar); if (type == PCIBAR_MEM64) { assert(idx + 1 <= PCI_BARMAX); pdi->pi_bar[idx + 1].type = PCIBAR_MEMHI64; pci_set_cfgdata32(pdi, PCIR_BAR(idx + 1), bar >> 32); } register_bar(pdi, idx); return (0); } #define CAP_START_OFFSET 0x40 static int pci_emul_add_capability(struct pci_devinst *pi, u_char *capdata, int caplen) { int i, capoff, reallen; uint16_t sts; assert(caplen > 0); reallen = roundup2(caplen, 4); /* dword aligned */ sts = pci_get_cfgdata16(pi, PCIR_STATUS); if ((sts & PCIM_STATUS_CAPPRESENT) == 0) capoff = CAP_START_OFFSET; else capoff = pi->pi_capend + 1; /* Check if we have enough space */ if (capoff + reallen > PCI_REGMAX + 1) return (-1); /* Set the previous capability pointer */ if ((sts & PCIM_STATUS_CAPPRESENT) == 0) { pci_set_cfgdata8(pi, PCIR_CAP_PTR, capoff); pci_set_cfgdata16(pi, PCIR_STATUS, sts|PCIM_STATUS_CAPPRESENT); } else pci_set_cfgdata8(pi, pi->pi_prevcap + 1, capoff); /* Copy the capability */ for (i = 0; i < caplen; i++) pci_set_cfgdata8(pi, capoff + i, capdata[i]); /* Set the next capability pointer */ pci_set_cfgdata8(pi, capoff + 1, 0); pi->pi_prevcap = capoff; pi->pi_capend = capoff + reallen - 1; return (0); } static struct pci_devemu * pci_emul_finddev(char *name) { struct pci_devemu **pdpp, *pdp; SET_FOREACH(pdpp, pci_devemu_set) { pdp = *pdpp; if (!strcmp(pdp->pe_emu, name)) { return (pdp); } } return (NULL); } static int pci_emul_init(struct vmctx *ctx, struct pci_devemu *pde, int bus, int slot, int func, struct funcinfo *fi) { struct pci_devinst *pdi; int err; pdi = calloc(1, sizeof(struct pci_devinst)); pdi->pi_vmctx = ctx; pdi->pi_bus = bus; pdi->pi_slot = slot; pdi->pi_func = func; pthread_mutex_init(&pdi->pi_lintr.lock, NULL); pdi->pi_lintr.pin = 0; pdi->pi_lintr.state = IDLE; pdi->pi_lintr.pirq_pin = 0; pdi->pi_lintr.ioapic_irq = 0; pdi->pi_d = pde; snprintf(pdi->pi_name, PI_NAMESZ, "%s-pci-%d", pde->pe_emu, slot); /* Disable legacy interrupts */ pci_set_cfgdata8(pdi, PCIR_INTLINE, 255); pci_set_cfgdata8(pdi, PCIR_INTPIN, 0); pci_set_cfgdata8(pdi, PCIR_COMMAND, PCIM_CMD_PORTEN | PCIM_CMD_MEMEN | PCIM_CMD_BUSMASTEREN); err = (*pde->pe_init)(ctx, pdi, fi->fi_param); if (err == 0) fi->fi_devi = pdi; else free(pdi); return (err); } void pci_populate_msicap(struct msicap *msicap, int msgnum, int nextptr) { int mmc; CTASSERT(sizeof(struct msicap) == 14); /* Number of msi messages must be a power of 2 between 1 and 32 */ assert((msgnum & (msgnum - 1)) == 0 && msgnum >= 1 && msgnum <= 32); mmc = ffs(msgnum) - 1; bzero(msicap, sizeof(struct msicap)); msicap->capid = PCIY_MSI; msicap->nextptr = nextptr; msicap->msgctrl = PCIM_MSICTRL_64BIT | (mmc << 1); } int pci_emul_add_msicap(struct pci_devinst *pi, int msgnum) { struct msicap msicap; pci_populate_msicap(&msicap, msgnum, 0); return (pci_emul_add_capability(pi, (u_char *)&msicap, sizeof(msicap))); } static void pci_populate_msixcap(struct msixcap *msixcap, int msgnum, int barnum, uint32_t msix_tab_size) { CTASSERT(sizeof(struct msixcap) == 12); assert(msix_tab_size % 4096 == 0); bzero(msixcap, sizeof(struct msixcap)); msixcap->capid = PCIY_MSIX; /* * Message Control Register, all fields set to * zero except for the Table Size. * Note: Table size N is encoded as N-1 */ msixcap->msgctrl = msgnum - 1; /* * MSI-X BAR setup: * - MSI-X table start at offset 0 * - PBA table starts at a 4K aligned offset after the MSI-X table */ msixcap->table_info = barnum & PCIM_MSIX_BIR_MASK; msixcap->pba_info = msix_tab_size | (barnum & PCIM_MSIX_BIR_MASK); } static void pci_msix_table_init(struct pci_devinst *pi, int table_entries) { int i, table_size; assert(table_entries > 0); assert(table_entries <= MAX_MSIX_TABLE_ENTRIES); table_size = table_entries * MSIX_TABLE_ENTRY_SIZE; pi->pi_msix.table = calloc(1, table_size); /* set mask bit of vector control register */ for (i = 0; i < table_entries; i++) pi->pi_msix.table[i].vector_control |= PCIM_MSIX_VCTRL_MASK; } int pci_emul_add_msixcap(struct pci_devinst *pi, int msgnum, int barnum) { uint32_t tab_size; struct msixcap msixcap; assert(msgnum >= 1 && msgnum <= MAX_MSIX_TABLE_ENTRIES); assert(barnum >= 0 && barnum <= PCIR_MAX_BAR_0); tab_size = msgnum * MSIX_TABLE_ENTRY_SIZE; /* Align table size to nearest 4K */ tab_size = roundup2(tab_size, 4096); pi->pi_msix.table_bar = barnum; pi->pi_msix.pba_bar = barnum; pi->pi_msix.table_offset = 0; pi->pi_msix.table_count = msgnum; pi->pi_msix.pba_offset = tab_size; pi->pi_msix.pba_size = PBA_SIZE(msgnum); pci_msix_table_init(pi, msgnum); pci_populate_msixcap(&msixcap, msgnum, barnum, tab_size); /* allocate memory for MSI-X Table and PBA */ pci_emul_alloc_bar(pi, barnum, PCIBAR_MEM32, tab_size + pi->pi_msix.pba_size); return (pci_emul_add_capability(pi, (u_char *)&msixcap, sizeof(msixcap))); } void msixcap_cfgwrite(struct pci_devinst *pi, int capoff, int offset, int bytes, uint32_t val) { uint16_t msgctrl, rwmask; int off, table_bar; off = offset - capoff; table_bar = pi->pi_msix.table_bar; /* Message Control Register */ if (off == 2 && bytes == 2) { rwmask = PCIM_MSIXCTRL_MSIX_ENABLE | PCIM_MSIXCTRL_FUNCTION_MASK; msgctrl = pci_get_cfgdata16(pi, offset); msgctrl &= ~rwmask; msgctrl |= val & rwmask; val = msgctrl; pi->pi_msix.enabled = val & PCIM_MSIXCTRL_MSIX_ENABLE; pi->pi_msix.function_mask = val & PCIM_MSIXCTRL_FUNCTION_MASK; pci_lintr_update(pi); } CFGWRITE(pi, offset, val, bytes); } void msicap_cfgwrite(struct pci_devinst *pi, int capoff, int offset, int bytes, uint32_t val) { uint16_t msgctrl, rwmask, msgdata, mme; uint32_t addrlo; /* * If guest is writing to the message control register make sure * we do not overwrite read-only fields. */ if ((offset - capoff) == 2 && bytes == 2) { rwmask = PCIM_MSICTRL_MME_MASK | PCIM_MSICTRL_MSI_ENABLE; msgctrl = pci_get_cfgdata16(pi, offset); msgctrl &= ~rwmask; msgctrl |= val & rwmask; val = msgctrl; addrlo = pci_get_cfgdata32(pi, capoff + 4); if (msgctrl & PCIM_MSICTRL_64BIT) msgdata = pci_get_cfgdata16(pi, capoff + 12); else msgdata = pci_get_cfgdata16(pi, capoff + 8); mme = msgctrl & PCIM_MSICTRL_MME_MASK; pi->pi_msi.enabled = msgctrl & PCIM_MSICTRL_MSI_ENABLE ? 1 : 0; if (pi->pi_msi.enabled) { pi->pi_msi.addr = addrlo; pi->pi_msi.msg_data = msgdata; pi->pi_msi.maxmsgnum = 1 << (mme >> 4); } else { pi->pi_msi.maxmsgnum = 0; } pci_lintr_update(pi); } CFGWRITE(pi, offset, val, bytes); } void pciecap_cfgwrite(struct pci_devinst *pi, int capoff, int offset, int bytes, uint32_t val) { /* XXX don't write to the readonly parts */ CFGWRITE(pi, offset, val, bytes); } #define PCIECAP_VERSION 0x2 int pci_emul_add_pciecap(struct pci_devinst *pi, int type) { int err; struct pciecap pciecap; CTASSERT(sizeof(struct pciecap) == 60); if (type != PCIEM_TYPE_ROOT_PORT) return (-1); bzero(&pciecap, sizeof(pciecap)); pciecap.capid = PCIY_EXPRESS; pciecap.pcie_capabilities = PCIECAP_VERSION | PCIEM_TYPE_ROOT_PORT; pciecap.link_capabilities = 0x411; /* gen1, x1 */ pciecap.link_status = 0x11; /* gen1, x1 */ err = pci_emul_add_capability(pi, (u_char *)&pciecap, sizeof(pciecap)); return (err); } /* * This function assumes that 'coff' is in the capabilities region of the * config space. */ static void pci_emul_capwrite(struct pci_devinst *pi, int offset, int bytes, uint32_t val) { int capid; uint8_t capoff, nextoff; /* Do not allow un-aligned writes */ if ((offset & (bytes - 1)) != 0) return; /* Find the capability that we want to update */ capoff = CAP_START_OFFSET; while (1) { nextoff = pci_get_cfgdata8(pi, capoff + 1); if (nextoff == 0) break; if (offset >= capoff && offset < nextoff) break; capoff = nextoff; } assert(offset >= capoff); /* * Capability ID and Next Capability Pointer are readonly. * However, some o/s's do 4-byte writes that include these. * For this case, trim the write back to 2 bytes and adjust * the data. */ if (offset == capoff || offset == capoff + 1) { if (offset == capoff && bytes == 4) { bytes = 2; offset += 2; val >>= 16; } else return; } capid = pci_get_cfgdata8(pi, capoff); switch (capid) { case PCIY_MSI: msicap_cfgwrite(pi, capoff, offset, bytes, val); break; case PCIY_MSIX: msixcap_cfgwrite(pi, capoff, offset, bytes, val); break; case PCIY_EXPRESS: pciecap_cfgwrite(pi, capoff, offset, bytes, val); break; default: break; } } static int pci_emul_iscap(struct pci_devinst *pi, int offset) { uint16_t sts; sts = pci_get_cfgdata16(pi, PCIR_STATUS); if ((sts & PCIM_STATUS_CAPPRESENT) != 0) { if (offset >= CAP_START_OFFSET && offset <= pi->pi_capend) return (1); } return (0); } static int pci_emul_fallback_handler(struct vmctx *ctx, int vcpu, int dir, uint64_t addr, int size, uint64_t *val, void *arg1, long arg2) { /* * Ignore writes; return 0xff's for reads. The mem read code * will take care of truncating to the correct size. */ if (dir == MEM_F_READ) { *val = 0xffffffffffffffff; } return (0); } static int pci_emul_ecfg_handler(struct vmctx *ctx, int vcpu, int dir, uint64_t addr, int bytes, uint64_t *val, void *arg1, long arg2) { int bus, slot, func, coff, in; coff = addr & 0xfff; func = (addr >> 12) & 0x7; slot = (addr >> 15) & 0x1f; bus = (addr >> 20) & 0xff; in = (dir == MEM_F_READ); if (in) *val = ~0UL; pci_cfgrw(ctx, vcpu, in, bus, slot, func, coff, bytes, (uint32_t *)val); return (0); } uint64_t pci_ecfg_base(void) { return (PCI_EMUL_ECFG_BASE); } #define BUSIO_ROUNDUP 32 #define BUSMEM_ROUNDUP (1024 * 1024) int init_pci(struct vmctx *ctx) { struct mem_range mr; struct pci_devemu *pde; struct businfo *bi; struct slotinfo *si; struct funcinfo *fi; size_t lowmem; int bus, slot, func; int error; pci_emul_iobase = PCI_EMUL_IOBASE; pci_emul_membase32 = vm_get_lowmem_limit(ctx); pci_emul_membase64 = PCI_EMUL_MEMBASE64; for (bus = 0; bus < MAXBUSES; bus++) { if ((bi = pci_businfo[bus]) == NULL) continue; /* * Keep track of the i/o and memory resources allocated to * this bus. */ bi->iobase = pci_emul_iobase; bi->membase32 = pci_emul_membase32; bi->membase64 = pci_emul_membase64; for (slot = 0; slot < MAXSLOTS; slot++) { si = &bi->slotinfo[slot]; for (func = 0; func < MAXFUNCS; func++) { fi = &si->si_funcs[func]; if (fi->fi_name == NULL) continue; pde = pci_emul_finddev(fi->fi_name); assert(pde != NULL); error = pci_emul_init(ctx, pde, bus, slot, func, fi); if (error) return (error); } } /* * Add some slop to the I/O and memory resources decoded by * this bus to give a guest some flexibility if it wants to * reprogram the BARs. */ pci_emul_iobase += BUSIO_ROUNDUP; pci_emul_iobase = roundup2(pci_emul_iobase, BUSIO_ROUNDUP); bi->iolimit = pci_emul_iobase; pci_emul_membase32 += BUSMEM_ROUNDUP; pci_emul_membase32 = roundup2(pci_emul_membase32, BUSMEM_ROUNDUP); bi->memlimit32 = pci_emul_membase32; pci_emul_membase64 += BUSMEM_ROUNDUP; pci_emul_membase64 = roundup2(pci_emul_membase64, BUSMEM_ROUNDUP); bi->memlimit64 = pci_emul_membase64; } /* * PCI backends are initialized before routing INTx interrupts * so that LPC devices are able to reserve ISA IRQs before * routing PIRQ pins. */ for (bus = 0; bus < MAXBUSES; bus++) { if ((bi = pci_businfo[bus]) == NULL) continue; for (slot = 0; slot < MAXSLOTS; slot++) { si = &bi->slotinfo[slot]; for (func = 0; func < MAXFUNCS; func++) { fi = &si->si_funcs[func]; if (fi->fi_devi == NULL) continue; pci_lintr_route(fi->fi_devi); } } } lpc_pirq_routed(); /* * The guest physical memory map looks like the following: * [0, lowmem) guest system memory * [lowmem, lowmem_limit) memory hole (may be absent) * [lowmem_limit, 0xE0000000) PCI hole (32-bit BAR allocation) * [0xE0000000, 0xF0000000) PCI extended config window * [0xF0000000, 4GB) LAPIC, IOAPIC, HPET, firmware * [4GB, 4GB + highmem) */ /* * Accesses to memory addresses that are not allocated to system * memory or PCI devices return 0xff's. */ lowmem = vm_get_lowmem_size(ctx); bzero(&mr, sizeof(struct mem_range)); mr.name = "PCI hole"; mr.flags = MEM_F_RW | MEM_F_IMMUTABLE; mr.base = lowmem; mr.size = (4ULL * 1024 * 1024 * 1024) - lowmem; mr.handler = pci_emul_fallback_handler; error = register_mem_fallback(&mr); assert(error == 0); /* PCI extended config space */ bzero(&mr, sizeof(struct mem_range)); mr.name = "PCI ECFG"; mr.flags = MEM_F_RW | MEM_F_IMMUTABLE; mr.base = PCI_EMUL_ECFG_BASE; mr.size = PCI_EMUL_ECFG_SIZE; mr.handler = pci_emul_ecfg_handler; error = register_mem(&mr); assert(error == 0); return (0); } static void pci_apic_prt_entry(int bus, int slot, int pin, int pirq_pin, int ioapic_irq, void *arg) { dsdt_line(" Package ()"); dsdt_line(" {"); dsdt_line(" 0x%X,", slot << 16 | 0xffff); dsdt_line(" 0x%02X,", pin - 1); dsdt_line(" Zero,"); dsdt_line(" 0x%X", ioapic_irq); dsdt_line(" },"); } static void pci_pirq_prt_entry(int bus, int slot, int pin, int pirq_pin, int ioapic_irq, void *arg) { char *name; name = lpc_pirq_name(pirq_pin); if (name == NULL) return; dsdt_line(" Package ()"); dsdt_line(" {"); dsdt_line(" 0x%X,", slot << 16 | 0xffff); dsdt_line(" 0x%02X,", pin - 1); dsdt_line(" %s,", name); dsdt_line(" 0x00"); dsdt_line(" },"); free(name); } /* * A bhyve virtual machine has a flat PCI hierarchy with a root port * corresponding to each PCI bus. */ static void pci_bus_write_dsdt(int bus) { struct businfo *bi; struct slotinfo *si; struct pci_devinst *pi; int count, func, slot; /* * If there are no devices on this 'bus' then just return. */ if ((bi = pci_businfo[bus]) == NULL) { /* * Bus 0 is special because it decodes the I/O ports used * for PCI config space access even if there are no devices * on it. */ if (bus != 0) return; } dsdt_line(" Device (PC%02X)", bus); dsdt_line(" {"); dsdt_line(" Name (_HID, EisaId (\"PNP0A03\"))"); dsdt_line(" Name (_ADR, Zero)"); dsdt_line(" Method (_BBN, 0, NotSerialized)"); dsdt_line(" {"); dsdt_line(" Return (0x%08X)", bus); dsdt_line(" }"); dsdt_line(" Name (_CRS, ResourceTemplate ()"); dsdt_line(" {"); dsdt_line(" WordBusNumber (ResourceProducer, MinFixed, " "MaxFixed, PosDecode,"); dsdt_line(" 0x0000, // Granularity"); dsdt_line(" 0x%04X, // Range Minimum", bus); dsdt_line(" 0x%04X, // Range Maximum", bus); dsdt_line(" 0x0000, // Translation Offset"); dsdt_line(" 0x0001, // Length"); dsdt_line(" ,, )"); if (bus == 0) { dsdt_indent(3); dsdt_fixed_ioport(0xCF8, 8); dsdt_unindent(3); dsdt_line(" WordIO (ResourceProducer, MinFixed, MaxFixed, " "PosDecode, EntireRange,"); dsdt_line(" 0x0000, // Granularity"); dsdt_line(" 0x0000, // Range Minimum"); dsdt_line(" 0x0CF7, // Range Maximum"); dsdt_line(" 0x0000, // Translation Offset"); dsdt_line(" 0x0CF8, // Length"); dsdt_line(" ,, , TypeStatic)"); dsdt_line(" WordIO (ResourceProducer, MinFixed, MaxFixed, " "PosDecode, EntireRange,"); dsdt_line(" 0x0000, // Granularity"); dsdt_line(" 0x0D00, // Range Minimum"); dsdt_line(" 0x%04X, // Range Maximum", PCI_EMUL_IOBASE - 1); dsdt_line(" 0x0000, // Translation Offset"); dsdt_line(" 0x%04X, // Length", PCI_EMUL_IOBASE - 0x0D00); dsdt_line(" ,, , TypeStatic)"); if (bi == NULL) { dsdt_line(" })"); goto done; } } assert(bi != NULL); /* i/o window */ dsdt_line(" WordIO (ResourceProducer, MinFixed, MaxFixed, " "PosDecode, EntireRange,"); dsdt_line(" 0x0000, // Granularity"); dsdt_line(" 0x%04X, // Range Minimum", bi->iobase); dsdt_line(" 0x%04X, // Range Maximum", bi->iolimit - 1); dsdt_line(" 0x0000, // Translation Offset"); dsdt_line(" 0x%04X, // Length", bi->iolimit - bi->iobase); dsdt_line(" ,, , TypeStatic)"); /* mmio window (32-bit) */ dsdt_line(" DWordMemory (ResourceProducer, PosDecode, " "MinFixed, MaxFixed, NonCacheable, ReadWrite,"); dsdt_line(" 0x00000000, // Granularity"); dsdt_line(" 0x%08X, // Range Minimum\n", bi->membase32); dsdt_line(" 0x%08X, // Range Maximum\n", bi->memlimit32 - 1); dsdt_line(" 0x00000000, // Translation Offset"); dsdt_line(" 0x%08X, // Length\n", bi->memlimit32 - bi->membase32); dsdt_line(" ,, , AddressRangeMemory, TypeStatic)"); /* mmio window (64-bit) */ dsdt_line(" QWordMemory (ResourceProducer, PosDecode, " "MinFixed, MaxFixed, NonCacheable, ReadWrite,"); dsdt_line(" 0x0000000000000000, // Granularity"); dsdt_line(" 0x%016lX, // Range Minimum\n", bi->membase64); dsdt_line(" 0x%016lX, // Range Maximum\n", bi->memlimit64 - 1); dsdt_line(" 0x0000000000000000, // Translation Offset"); dsdt_line(" 0x%016lX, // Length\n", bi->memlimit64 - bi->membase64); dsdt_line(" ,, , AddressRangeMemory, TypeStatic)"); dsdt_line(" })"); count = pci_count_lintr(bus); if (count != 0) { dsdt_indent(2); dsdt_line("Name (PPRT, Package ()"); dsdt_line("{"); pci_walk_lintr(bus, pci_pirq_prt_entry, NULL); dsdt_line("})"); dsdt_line("Name (APRT, Package ()"); dsdt_line("{"); pci_walk_lintr(bus, pci_apic_prt_entry, NULL); dsdt_line("})"); dsdt_line("Method (_PRT, 0, NotSerialized)"); dsdt_line("{"); dsdt_line(" If (PICM)"); dsdt_line(" {"); dsdt_line(" Return (APRT)"); dsdt_line(" }"); dsdt_line(" Else"); dsdt_line(" {"); dsdt_line(" Return (PPRT)"); dsdt_line(" }"); dsdt_line("}"); dsdt_unindent(2); } dsdt_indent(2); for (slot = 0; slot < MAXSLOTS; slot++) { si = &bi->slotinfo[slot]; for (func = 0; func < MAXFUNCS; func++) { pi = si->si_funcs[func].fi_devi; if (pi != NULL && pi->pi_d->pe_write_dsdt != NULL) pi->pi_d->pe_write_dsdt(pi); } } dsdt_unindent(2); done: dsdt_line(" }"); } void pci_write_dsdt(void) { int bus; dsdt_indent(1); dsdt_line("Name (PICM, 0x00)"); dsdt_line("Method (_PIC, 1, NotSerialized)"); dsdt_line("{"); dsdt_line(" Store (Arg0, PICM)"); dsdt_line("}"); dsdt_line(""); dsdt_line("Scope (_SB)"); dsdt_line("{"); for (bus = 0; bus < MAXBUSES; bus++) pci_bus_write_dsdt(bus); dsdt_line("}"); dsdt_unindent(1); } int pci_bus_configured(int bus) { assert(bus >= 0 && bus < MAXBUSES); return (pci_businfo[bus] != NULL); } int pci_msi_enabled(struct pci_devinst *pi) { return (pi->pi_msi.enabled); } int pci_msi_maxmsgnum(struct pci_devinst *pi) { if (pi->pi_msi.enabled) return (pi->pi_msi.maxmsgnum); else return (0); } int pci_msix_enabled(struct pci_devinst *pi) { return (pi->pi_msix.enabled && !pi->pi_msi.enabled); } void pci_generate_msix(struct pci_devinst *pi, int index) { struct msix_table_entry *mte; if (!pci_msix_enabled(pi)) return; if (pi->pi_msix.function_mask) return; if (index >= pi->pi_msix.table_count) return; mte = &pi->pi_msix.table[index]; if ((mte->vector_control & PCIM_MSIX_VCTRL_MASK) == 0) { /* XXX Set PBA bit if interrupt is disabled */ vm_lapic_msi(pi->pi_vmctx, mte->addr, mte->msg_data); } } void pci_generate_msi(struct pci_devinst *pi, int index) { if (pci_msi_enabled(pi) && index < pci_msi_maxmsgnum(pi)) { vm_lapic_msi(pi->pi_vmctx, pi->pi_msi.addr, pi->pi_msi.msg_data + index); } } static bool pci_lintr_permitted(struct pci_devinst *pi) { uint16_t cmd; cmd = pci_get_cfgdata16(pi, PCIR_COMMAND); return (!(pi->pi_msi.enabled || pi->pi_msix.enabled || (cmd & PCIM_CMD_INTxDIS))); } void pci_lintr_request(struct pci_devinst *pi) { struct businfo *bi; struct slotinfo *si; int bestpin, bestcount, pin; bi = pci_businfo[pi->pi_bus]; assert(bi != NULL); /* * Just allocate a pin from our slot. The pin will be * assigned IRQs later when interrupts are routed. */ si = &bi->slotinfo[pi->pi_slot]; bestpin = 0; bestcount = si->si_intpins[0].ii_count; for (pin = 1; pin < 4; pin++) { if (si->si_intpins[pin].ii_count < bestcount) { bestpin = pin; bestcount = si->si_intpins[pin].ii_count; } } si->si_intpins[bestpin].ii_count++; pi->pi_lintr.pin = bestpin + 1; pci_set_cfgdata8(pi, PCIR_INTPIN, bestpin + 1); } static void pci_lintr_route(struct pci_devinst *pi) { struct businfo *bi; struct intxinfo *ii; if (pi->pi_lintr.pin == 0) return; bi = pci_businfo[pi->pi_bus]; assert(bi != NULL); ii = &bi->slotinfo[pi->pi_slot].si_intpins[pi->pi_lintr.pin - 1]; /* * Attempt to allocate an I/O APIC pin for this intpin if one * is not yet assigned. */ if (ii->ii_ioapic_irq == 0) ii->ii_ioapic_irq = ioapic_pci_alloc_irq(); assert(ii->ii_ioapic_irq > 0); /* * Attempt to allocate a PIRQ pin for this intpin if one is * not yet assigned. */ if (ii->ii_pirq_pin == 0) ii->ii_pirq_pin = pirq_alloc_pin(pi->pi_vmctx); assert(ii->ii_pirq_pin > 0); pi->pi_lintr.ioapic_irq = ii->ii_ioapic_irq; pi->pi_lintr.pirq_pin = ii->ii_pirq_pin; pci_set_cfgdata8(pi, PCIR_INTLINE, pirq_irq(ii->ii_pirq_pin)); } void pci_lintr_assert(struct pci_devinst *pi) { assert(pi->pi_lintr.pin > 0); pthread_mutex_lock(&pi->pi_lintr.lock); if (pi->pi_lintr.state == IDLE) { if (pci_lintr_permitted(pi)) { pi->pi_lintr.state = ASSERTED; pci_irq_assert(pi); } else pi->pi_lintr.state = PENDING; } pthread_mutex_unlock(&pi->pi_lintr.lock); } void pci_lintr_deassert(struct pci_devinst *pi) { assert(pi->pi_lintr.pin > 0); pthread_mutex_lock(&pi->pi_lintr.lock); if (pi->pi_lintr.state == ASSERTED) { pi->pi_lintr.state = IDLE; pci_irq_deassert(pi); } else if (pi->pi_lintr.state == PENDING) pi->pi_lintr.state = IDLE; pthread_mutex_unlock(&pi->pi_lintr.lock); } static void pci_lintr_update(struct pci_devinst *pi) { pthread_mutex_lock(&pi->pi_lintr.lock); if (pi->pi_lintr.state == ASSERTED && !pci_lintr_permitted(pi)) { pci_irq_deassert(pi); pi->pi_lintr.state = PENDING; } else if (pi->pi_lintr.state == PENDING && pci_lintr_permitted(pi)) { pi->pi_lintr.state = ASSERTED; pci_irq_assert(pi); } pthread_mutex_unlock(&pi->pi_lintr.lock); } int pci_count_lintr(int bus) { int count, slot, pin; struct slotinfo *slotinfo; count = 0; if (pci_businfo[bus] != NULL) { for (slot = 0; slot < MAXSLOTS; slot++) { slotinfo = &pci_businfo[bus]->slotinfo[slot]; for (pin = 0; pin < 4; pin++) { if (slotinfo->si_intpins[pin].ii_count != 0) count++; } } } return (count); } void pci_walk_lintr(int bus, pci_lintr_cb cb, void *arg) { struct businfo *bi; struct slotinfo *si; struct intxinfo *ii; int slot, pin; if ((bi = pci_businfo[bus]) == NULL) return; for (slot = 0; slot < MAXSLOTS; slot++) { si = &bi->slotinfo[slot]; for (pin = 0; pin < 4; pin++) { ii = &si->si_intpins[pin]; if (ii->ii_count != 0) cb(bus, slot, pin + 1, ii->ii_pirq_pin, ii->ii_ioapic_irq, arg); } } } /* * Return 1 if the emulated device in 'slot' is a multi-function device. * Return 0 otherwise. */ static int pci_emul_is_mfdev(int bus, int slot) { struct businfo *bi; struct slotinfo *si; int f, numfuncs; numfuncs = 0; if ((bi = pci_businfo[bus]) != NULL) { si = &bi->slotinfo[slot]; for (f = 0; f < MAXFUNCS; f++) { if (si->si_funcs[f].fi_devi != NULL) { numfuncs++; } } } return (numfuncs > 1); } /* * Ensure that the PCIM_MFDEV bit is properly set (or unset) depending on * whether or not is a multi-function being emulated in the pci 'slot'. */ static void pci_emul_hdrtype_fixup(int bus, int slot, int off, int bytes, uint32_t *rv) { int mfdev; if (off <= PCIR_HDRTYPE && off + bytes > PCIR_HDRTYPE) { mfdev = pci_emul_is_mfdev(bus, slot); switch (bytes) { case 1: case 2: *rv &= ~PCIM_MFDEV; if (mfdev) { *rv |= PCIM_MFDEV; } break; case 4: *rv &= ~(PCIM_MFDEV << 16); if (mfdev) { *rv |= (PCIM_MFDEV << 16); } break; } } } -static uint32_t -bits_changed(uint32_t old, uint32_t new, uint32_t mask) -{ - - return ((old ^ new) & mask); -} - static void -pci_emul_cmdwrite(struct pci_devinst *pi, uint32_t new, int bytes) +pci_emul_cmdsts_write(struct pci_devinst *pi, int coff, uint32_t new, int bytes) { - int i; - uint16_t old; + int i, rshift; + uint32_t cmd, cmd2, changed, old, readonly; + cmd = pci_get_cfgdata16(pi, PCIR_COMMAND); /* stash old value */ + /* - * The command register is at an offset of 4 bytes and thus the - * guest could write 1, 2 or 4 bytes starting at this offset. + * From PCI Local Bus Specification 3.0 sections 6.2.2 and 6.2.3. + * + * XXX Bits 8, 11, 12, 13, 14 and 15 in the status register are + * 'write 1 to clear'. However these bits are not set to '1' by + * any device emulation so it is simpler to treat them as readonly. */ + rshift = (coff & 0x3) * 8; + readonly = 0xFFFFF880 >> rshift; - old = pci_get_cfgdata16(pi, PCIR_COMMAND); /* stash old value */ - CFGWRITE(pi, PCIR_COMMAND, new, bytes); /* update config */ - new = pci_get_cfgdata16(pi, PCIR_COMMAND); /* get updated value */ + old = CFGREAD(pi, coff, bytes); + new &= ~readonly; + new |= (old & readonly); + CFGWRITE(pi, coff, new, bytes); /* update config */ + cmd2 = pci_get_cfgdata16(pi, PCIR_COMMAND); /* get updated value */ + changed = cmd ^ cmd2; + /* * If the MMIO or I/O address space decoding has changed then * register/unregister all BARs that decode that address space. */ for (i = 0; i <= PCI_BARMAX; i++) { switch (pi->pi_bar[i].type) { case PCIBAR_NONE: case PCIBAR_MEMHI64: break; case PCIBAR_IO: /* I/O address space decoding changed? */ - if (bits_changed(old, new, PCIM_CMD_PORTEN)) { + if (changed & PCIM_CMD_PORTEN) { if (porten(pi)) register_bar(pi, i); else unregister_bar(pi, i); } break; case PCIBAR_MEM32: case PCIBAR_MEM64: /* MMIO address space decoding changed? */ - if (bits_changed(old, new, PCIM_CMD_MEMEN)) { + if (changed & PCIM_CMD_MEMEN) { if (memen(pi)) register_bar(pi, i); else unregister_bar(pi, i); } break; default: assert(0); } } /* * If INTx has been unmasked and is pending, assert the * interrupt. */ pci_lintr_update(pi); } static void pci_cfgrw(struct vmctx *ctx, int vcpu, int in, int bus, int slot, int func, int coff, int bytes, uint32_t *eax) { struct businfo *bi; struct slotinfo *si; struct pci_devinst *pi; struct pci_devemu *pe; int idx, needcfg; uint64_t addr, bar, mask; if ((bi = pci_businfo[bus]) != NULL) { si = &bi->slotinfo[slot]; pi = si->si_funcs[func].fi_devi; } else pi = NULL; /* * Just return if there is no device at this slot:func or if the * the guest is doing an un-aligned access. */ if (pi == NULL || (bytes != 1 && bytes != 2 && bytes != 4) || (coff & (bytes - 1)) != 0) { if (in) *eax = 0xffffffff; return; } /* * Ignore all writes beyond the standard config space and return all * ones on reads. */ if (coff >= PCI_REGMAX + 1) { if (in) { *eax = 0xffffffff; /* * Extended capabilities begin at offset 256 in config * space. Absence of extended capabilities is signaled * with all 0s in the extended capability header at * offset 256. */ if (coff <= PCI_REGMAX + 4) *eax = 0x00000000; } return; } pe = pi->pi_d; /* * Config read */ if (in) { /* Let the device emulation override the default handler */ if (pe->pe_cfgread != NULL) { needcfg = pe->pe_cfgread(ctx, vcpu, pi, coff, bytes, eax); } else { needcfg = 1; } - if (needcfg) { - if (bytes == 1) - *eax = pci_get_cfgdata8(pi, coff); - else if (bytes == 2) - *eax = pci_get_cfgdata16(pi, coff); - else - *eax = pci_get_cfgdata32(pi, coff); - } + if (needcfg) + *eax = CFGREAD(pi, coff, bytes); pci_emul_hdrtype_fixup(bus, slot, coff, bytes, eax); } else { /* Let the device emulation override the default handler */ if (pe->pe_cfgwrite != NULL && (*pe->pe_cfgwrite)(ctx, vcpu, pi, coff, bytes, *eax) == 0) return; /* * Special handling for write to BAR registers */ if (coff >= PCIR_BAR(0) && coff < PCIR_BAR(PCI_BARMAX + 1)) { /* * Ignore writes to BAR registers that are not * 4-byte aligned. */ if (bytes != 4 || (coff & 0x3) != 0) return; idx = (coff - PCIR_BAR(0)) / 4; mask = ~(pi->pi_bar[idx].size - 1); switch (pi->pi_bar[idx].type) { case PCIBAR_NONE: pi->pi_bar[idx].addr = bar = 0; break; case PCIBAR_IO: addr = *eax & mask; addr &= 0xffff; bar = addr | PCIM_BAR_IO_SPACE; /* * Register the new BAR value for interception */ if (addr != pi->pi_bar[idx].addr) { update_bar_address(pi, addr, idx, PCIBAR_IO); } break; case PCIBAR_MEM32: addr = bar = *eax & mask; bar |= PCIM_BAR_MEM_SPACE | PCIM_BAR_MEM_32; if (addr != pi->pi_bar[idx].addr) { update_bar_address(pi, addr, idx, PCIBAR_MEM32); } break; case PCIBAR_MEM64: addr = bar = *eax & mask; bar |= PCIM_BAR_MEM_SPACE | PCIM_BAR_MEM_64 | PCIM_BAR_MEM_PREFETCH; if (addr != (uint32_t)pi->pi_bar[idx].addr) { update_bar_address(pi, addr, idx, PCIBAR_MEM64); } break; case PCIBAR_MEMHI64: mask = ~(pi->pi_bar[idx - 1].size - 1); addr = ((uint64_t)*eax << 32) & mask; bar = addr >> 32; if (bar != pi->pi_bar[idx - 1].addr >> 32) { update_bar_address(pi, addr, idx - 1, PCIBAR_MEMHI64); } break; default: assert(0); } pci_set_cfgdata32(pi, coff, bar); } else if (pci_emul_iscap(pi, coff)) { pci_emul_capwrite(pi, coff, bytes, *eax); - } else if (coff == PCIR_COMMAND) { - pci_emul_cmdwrite(pi, *eax, bytes); + } else if (coff >= PCIR_COMMAND && coff < PCIR_REVID) { + pci_emul_cmdsts_write(pi, coff, *eax, bytes); } else { CFGWRITE(pi, coff, *eax, bytes); } } } static int cfgenable, cfgbus, cfgslot, cfgfunc, cfgoff; static int pci_emul_cfgaddr(struct vmctx *ctx, int vcpu, int in, int port, int bytes, uint32_t *eax, void *arg) { uint32_t x; if (bytes != 4) { if (in) *eax = (bytes == 2) ? 0xffff : 0xff; return (0); } if (in) { x = (cfgbus << 16) | (cfgslot << 11) | (cfgfunc << 8) | cfgoff; if (cfgenable) x |= CONF1_ENABLE; *eax = x; } else { x = *eax; cfgenable = (x & CONF1_ENABLE) == CONF1_ENABLE; cfgoff = x & PCI_REGMAX; cfgfunc = (x >> 8) & PCI_FUNCMAX; cfgslot = (x >> 11) & PCI_SLOTMAX; cfgbus = (x >> 16) & PCI_BUSMAX; } return (0); } INOUT_PORT(pci_cfgaddr, CONF1_ADDR_PORT, IOPORT_F_INOUT, pci_emul_cfgaddr); static int pci_emul_cfgdata(struct vmctx *ctx, int vcpu, int in, int port, int bytes, uint32_t *eax, void *arg) { int coff; assert(bytes == 1 || bytes == 2 || bytes == 4); coff = cfgoff + (port - CONF1_DATA_PORT); if (cfgenable) { pci_cfgrw(ctx, vcpu, in, cfgbus, cfgslot, cfgfunc, coff, bytes, eax); } else { /* Ignore accesses to cfgdata if not enabled by cfgaddr */ if (in) *eax = 0xffffffff; } return (0); } INOUT_PORT(pci_cfgdata, CONF1_DATA_PORT+0, IOPORT_F_INOUT, pci_emul_cfgdata); INOUT_PORT(pci_cfgdata, CONF1_DATA_PORT+1, IOPORT_F_INOUT, pci_emul_cfgdata); INOUT_PORT(pci_cfgdata, CONF1_DATA_PORT+2, IOPORT_F_INOUT, pci_emul_cfgdata); INOUT_PORT(pci_cfgdata, CONF1_DATA_PORT+3, IOPORT_F_INOUT, pci_emul_cfgdata); #define PCI_EMUL_TEST #ifdef PCI_EMUL_TEST /* * Define a dummy test device */ #define DIOSZ 8 #define DMEMSZ 4096 struct pci_emul_dsoftc { uint8_t ioregs[DIOSZ]; uint8_t memregs[DMEMSZ]; }; #define PCI_EMUL_MSI_MSGS 4 #define PCI_EMUL_MSIX_MSGS 16 static int pci_emul_dinit(struct vmctx *ctx, struct pci_devinst *pi, char *opts) { int error; struct pci_emul_dsoftc *sc; sc = calloc(1, sizeof(struct pci_emul_dsoftc)); pi->pi_arg = sc; pci_set_cfgdata16(pi, PCIR_DEVICE, 0x0001); pci_set_cfgdata16(pi, PCIR_VENDOR, 0x10DD); pci_set_cfgdata8(pi, PCIR_CLASS, 0x02); error = pci_emul_add_msicap(pi, PCI_EMUL_MSI_MSGS); assert(error == 0); error = pci_emul_alloc_bar(pi, 0, PCIBAR_IO, DIOSZ); assert(error == 0); error = pci_emul_alloc_bar(pi, 1, PCIBAR_MEM32, DMEMSZ); assert(error == 0); return (0); } static void pci_emul_diow(struct vmctx *ctx, int vcpu, struct pci_devinst *pi, int baridx, uint64_t offset, int size, uint64_t value) { int i; struct pci_emul_dsoftc *sc = pi->pi_arg; if (baridx == 0) { if (offset + size > DIOSZ) { printf("diow: iow too large, offset %ld size %d\n", offset, size); return; } if (size == 1) { sc->ioregs[offset] = value & 0xff; } else if (size == 2) { *(uint16_t *)&sc->ioregs[offset] = value & 0xffff; } else if (size == 4) { *(uint32_t *)&sc->ioregs[offset] = value; } else { printf("diow: iow unknown size %d\n", size); } /* * Special magic value to generate an interrupt */ if (offset == 4 && size == 4 && pci_msi_enabled(pi)) pci_generate_msi(pi, value % pci_msi_maxmsgnum(pi)); if (value == 0xabcdef) { for (i = 0; i < pci_msi_maxmsgnum(pi); i++) pci_generate_msi(pi, i); } } if (baridx == 1) { if (offset + size > DMEMSZ) { printf("diow: memw too large, offset %ld size %d\n", offset, size); return; } if (size == 1) { sc->memregs[offset] = value; } else if (size == 2) { *(uint16_t *)&sc->memregs[offset] = value; } else if (size == 4) { *(uint32_t *)&sc->memregs[offset] = value; } else if (size == 8) { *(uint64_t *)&sc->memregs[offset] = value; } else { printf("diow: memw unknown size %d\n", size); } /* * magic interrupt ?? */ } if (baridx > 1) { printf("diow: unknown bar idx %d\n", baridx); } } static uint64_t pci_emul_dior(struct vmctx *ctx, int vcpu, struct pci_devinst *pi, int baridx, uint64_t offset, int size) { struct pci_emul_dsoftc *sc = pi->pi_arg; uint32_t value; if (baridx == 0) { if (offset + size > DIOSZ) { printf("dior: ior too large, offset %ld size %d\n", offset, size); return (0); } if (size == 1) { value = sc->ioregs[offset]; } else if (size == 2) { value = *(uint16_t *) &sc->ioregs[offset]; } else if (size == 4) { value = *(uint32_t *) &sc->ioregs[offset]; } else { printf("dior: ior unknown size %d\n", size); } } if (baridx == 1) { if (offset + size > DMEMSZ) { printf("dior: memr too large, offset %ld size %d\n", offset, size); return (0); } if (size == 1) { value = sc->memregs[offset]; } else if (size == 2) { value = *(uint16_t *) &sc->memregs[offset]; } else if (size == 4) { value = *(uint32_t *) &sc->memregs[offset]; } else if (size == 8) { value = *(uint64_t *) &sc->memregs[offset]; } else { printf("dior: ior unknown size %d\n", size); } } if (baridx > 1) { printf("dior: unknown bar idx %d\n", baridx); return (0); } return (value); } struct pci_devemu pci_dummy = { .pe_emu = "dummy", .pe_init = pci_emul_dinit, .pe_barwrite = pci_emul_diow, .pe_barread = pci_emul_dior }; PCI_EMUL_SET(pci_dummy); #endif /* PCI_EMUL_TEST */ Index: user/ngie/more-tests/usr.sbin/bhyve =================================================================== --- user/ngie/more-tests/usr.sbin/bhyve (revision 281974) +++ user/ngie/more-tests/usr.sbin/bhyve (revision 281975) Property changes on: user/ngie/more-tests/usr.sbin/bhyve ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/usr.sbin/bhyve:r281925-281974 Index: user/ngie/more-tests/usr.sbin/iovctl/iovctl.conf.5 =================================================================== --- user/ngie/more-tests/usr.sbin/iovctl/iovctl.conf.5 (revision 281974) +++ user/ngie/more-tests/usr.sbin/iovctl/iovctl.conf.5 (revision 281975) @@ -1,167 +1,167 @@ .\" .\" Copyright (c) 2014 Sandvine Inc. .\" All rights reserved. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" .Dd May 21, 2014 .Dt IOVCTL.CONF 5 .Os .Sh NAME .Nm iovctl.conf .Nd IOVCTL configuration file .Sh DESCRIPTION The .Nm file is the configuration file for the .Xr iovctl 8 program. This file specifies configuration parameters for a single Physical Function .Pq PF device. To configure SR-IOV on multiple PF devices, use one configuration file for each PF. The locations of all .Xr iovctl 9 configuration files are specified in .Xr rc.conf 5 . .Pp The .Nm file uses UCL format. UCL syntax is documented at the official UCL website: http://github.com/vstakhov/libucl. .Pp There are three types of sections in the .Nm file. A section is a key at the top level of the file with a list as its value. The list may contain the keys specified in the .Sx OPTIONS section of this manual page. Individual PF driver implementations may specify additional device-specific configuration keys that they will accept. The order in which sections appear in .Nm is ignored. No two sections may have the same key. For example, two sections for VF-1 must not be defined. .Pp The first section type is the PF section. This section always has the key "PF"; therefore, only one such section may be defined. This section defines configuration parameters that apply to the PF as a whole. .Pp The second section type is the VF section. This section has the key "VF-" followed by a VF index. VF indices start at 0 and always increment by 1. Valid VF indices are in the range of 0 to .Po num_vfs - 1 Pc . The VF index must be given as a decimal integer with no leading zeros. This section defines configuration parameters that apply to a single VF. .Pp The third section type is the default section. This section always has the key "DEFAULT"; therefore, only one such section may be specified. This section defines default configuration parameters that apply to all VFs. All configuration keys that are valid to be applied to a VF are valid in this section. An individual VF section may override a default specified in this section by providing a different value for the configuration parameter. Note that the default section applies to ALL VFs. The default section must appear before any VF sections. The default section may appear before or after the PF section. .Pp The following option types are supported: .Bl -tag -width indent .It boolean Accepts a boolean value of true or false. .It mac-addr Accepts a unicast MAC address specified as a string of the form xx:xx:xx:xx:xx:xx, where xx is one or two hexadecimal digits. .It string Accepts any string value. .It uint8_t Accepts any integer in the range 0-255, inclusive. .It uint16_t Accepts any integer in the range 0-65535, inclusive. .It uint32_t Accepts any integer in the range 0-2**32, inclusive. .It uint64_t Accepts any integer in the range 0-2**64, inclusive. .El .Sh OPTIONS The following parameters are accepted by all PF drivers: .Bl -tag -width indent .It device Pq string This parameter specifies the name of the PF device. This parameter is required to be specified. .It num_vfs Pq uint16_t This parameter specifies the number of VF children to create. This parameter may not be zero. The maximum value of this parameter is device-specific. .El .Pp The following parameters are accepted by all VFs: .Bl -tag -width indent .It passthrough Pq boolean This parameter controls whether the VF is reserved for the use of the .Xr bhyve 8 hypervisor as a PCI passthrough device. If this parameter is set to true, then the VF will be reserved as a PCI passthrough device and it will not be accessible from the host OS. The default value of this parameter is false. .El .Pp See the PF driver manual page for configuration parameters specific to particular hardware. .Sh EXAMPLES This sample file will create 3 VFs as children of the ix0 device. VF-1 and VF-2 are set as .Xr bhyve 8 passthrough devices through the use of the default section. VF-0 is not configured as a passthrough device as it explicitly overrides the default. VF-0 also sets a device-specific parameter named mac-addr. -.Bd -literal .offset ident +.Bd -literal -offset ident PF { device : "ix0"; num_vfs : 3; } DEFAULT { passthrough : true; } VF-0 { mac-addr : "02:56:48:7e:d9:f7"; passthrough : false; } .Ed .Sh SEE ALSO -.Xr iovctl 8 , -.Xr rc.conf 5 +.Xr rc.conf 5 , +.Xr iovctl 8 .Sh AUTHORS This manual page was written by .An Ryan Stone Aq Mt rstone@FreeBSD.org . Index: user/ngie/more-tests/usr.sbin/jail/jail.8 =================================================================== --- user/ngie/more-tests/usr.sbin/jail/jail.8 (revision 281974) +++ user/ngie/more-tests/usr.sbin/jail/jail.8 (revision 281975) @@ -1,1288 +1,1287 @@ .\" Copyright (c) 2000, 2003 Robert N. M. Watson .\" Copyright (c) 2008-2012 James Gritton .\" All rights reserved. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" .Dd February 25, 2015 .Dt JAIL 8 .Os .Sh NAME .Nm jail .Nd "manage system jails" .Sh SYNOPSIS .Nm .Op Fl dhilqv .Op Fl J Ar jid_file .Op Fl u Ar username .Op Fl U Ar username .Op Fl cmr .Ar param Ns = Ns Ar value ... .Op Cm command Ns = Ns Ar command ... .Nm .Op Fl dqv .Op Fl f Ar conf_file .Op Fl p Ar limit .Op Fl cmr .Op Ar jail .Nm .Op Fl qv .Op Fl f Ar conf_file .Op Fl rR .Op Cm * | Ar jail ... .Nm .Op Fl dhilqv .Op Fl J Ar jid_file .Op Fl u Ar username .Op Fl U Ar username .Op Fl n Ar jailname .Op Fl s Ar securelevel .Op Ar path hostname [ Ar ip Ns [ Ns Ar ,... Ns ]] Ar command ... .Sh DESCRIPTION The .Nm utility creates new jails, or modifies or removes existing jails. A jail .Pq or Dq prison is specified via parameters on the command line, or in the .Xr jail.conf 5 file. .Pp At least one of the options .Fl c , .Fl m or .Fl r must be specified. These options are used alone or in combination to describe the operation to perform: .Bl -tag -width indent .It Fl c Create a new jail. The jail .Va jid and .Va name parameters (if specified on the command line) must not refer to an existing jail. .It Fl m Modify an existing jail. One of the .Va jid or .Va name parameters must exist and refer to an existing jail. Some parameters may not be changed on a running jail. .It Fl r Remove the .Ar jail specified by jid or name. All jailed processes are killed, and all jails that are children of this jail are also removed. .It Fl rc Restart an existing jail. The jail is first removed and then re-created, as if .Dq Nm Fl r and .Dq Nm Fl c were run in succession. .It Fl cm Create a jail if it does not exist, or modify the jail if it does exist. .It Fl mr Modify an existing jail. The jail may be restarted if necessary to modify parameters than could not otherwise be changed. .It Fl cmr Create a jail if it doesn't exist, or modify (and possibly restart) the jail if it does exist. .El .Pp Other available options are: .Bl -tag -width indent .It Fl d Allow making changes to a dying jail, equivalent to the .Va allow.dying parameter. .It Fl f Ar conf_file Use configuration file .Ar conf_file instead of the default .Pa /etc/jail.conf . .It Fl h Resolve the .Va host.hostname parameter (or .Va hostname ) and add all IP addresses returned by the resolver to the list of addresses for this jail. This is equivalent to the .Va ip_hostname parameter. .It Fl i Output (only) the jail identifier of the newly created jail(s). This implies the .Fl q option. .It Fl J Ar jid_file Write a .Ar jid_file file, containing the parameters used to start the jail. .It Fl l Run commands in a clean environment. This is deprecated and is equivalent to the exec.clean parameter. .It Fl n Ar jailname Set the jail's name. This is deprecated and is equivalent to the .Va name parameter. .It Fl p Ar limit Limit the number of commands from .Va exec.* that can run simultaneously. .It Fl q Suppress the message printed whenever a jail is created, modified or removed. Only error messages will be printed. .It Fl R A variation of the .Fl r option that removes an existing jail without using the configuration file. No removal-related parameters for this jail will be used \(em the jail will simply be removed. .It Fl s Ar securelevel Set the .Va kern.securelevel MIB entry to the specified value inside the newly created jail. This is deprecated and is equivalent to the .Va securelevel parameter. .It Fl u Ar username The user name from host environment as whom jailed commands should run. This is deprecated and is equivalent to the .Va exec.jail_user and .Va exec.system_jail_user parameters. .It Fl U Ar username The user name from the jailed environment as whom jailed commands should run. This is deprecated and is equivalent to the .Va exec.jail_user parameter. .It Fl v Print a message on every operation, such as running commands and mounting filesystems. .El .Pp If no arguments are given after the options, the operation (except remove) will be performed on all jails specified in the .Xr jail.conf 5 file. A single argument of a jail name will operate only on the specified jail. The .Fl r and .Fl R options can also remove running jails that aren't in the .Xr jail.conf 5 file, specified by name or jid. .Pp An argument of .Dq * is a wildcard that will operate on all jails, regardless of whether they appear in .Xr jail.conf 5 ; this is the surest way for .Fl r to remove all jails. If hierarchical jails exist, a partial-matching wildcard definition may be specified. For example, an argument of .Dq foo.* would apply to jails with names like .Dq foo.bar and .Dq foo.bar.baz . .Pp A jail may be specified with parameters directly on the command line. In this case, the .Xr jail.conf 5 file will not be used. For backward compatibility, the command line may also have four fixed parameters, without names: .Ar path , .Ar hostname , .Ar ip , and .Ar command . This mode will always create a new jail, and the .Fl c and .Fl m options do not apply (and must not be present). .Ss Jail Parameters Parameters in the .Xr jail.conf 5 file, or on the command line, are generally of the form .Dq name=value . Some parameters are boolean, and do not have a value but are set by the name alone with or without a .Dq no prefix, e.g. .Va persist or .Va nopersist . They can also be given the values .Dq true and .Dq false . Other parameters may have more than one value, specified as a comma-separated list or with .Dq += in the configuration file (see .Xr jail.conf 5 for details). .Pp The .Nm utility recognizes two classes of parameters. There are the true jail parameters that are passed to the kernel when the jail is created, which can be seen with .Xr jls 8 , and can (usually) be changed with .Dq Nm Fl m . Then there are pseudo-parameters that are only used by .Nm itself. .Pp Jails have a set of core parameters, and kernel modules can add their own jail parameters. The current set of available parameters can be retrieved via .Dq Nm sysctl Fl d Va security.jail.param . Any parameters not set will be given default values, often based on the current environment. The core parameters are: .Bl -tag -width indent .It Va jid The jail identifier. This will be assigned automatically to a new jail (or can be explicitly set), and can be used to identify the jail for later modification, or for such commands as .Xr jls 8 or .Xr jexec 8 . .It Va name The jail name. This is an arbitrary string that identifies a jail (except it may not contain a .Sq \&. ) . Like the .Va jid , it can be passed to later .Nm commands, or to .Xr jls 8 or .Xr jexec 8 . If no .Va name is supplied, a default is assumed that is the same as the .Va jid . The .Va name parameter is implied by the .Xr jail.conf 5 file format, and need not be explicitly set when using the configuration file. .It Va path The directory which is to be the root of the jail. Any commands run inside the jail, either by .Nm or from .Xr jexec 8 , are run from this directory. .It Va ip4.addr A list of IPv4 addresses assigned to the jail. If this is set, the jail is restricted to using only these addresses. Any attempts to use other addresses fail, and attempts to use wildcard addresses silently use the jailed address instead. For IPv4 the first address given will be used as the source address when source address selection on unbound sockets cannot find a better match. It is only possible to start multiple jails with the same IP address if none of the jails has more than this single overlapping IP address assigned to itself. .It Va ip4.saddrsel A boolean option to change the formerly mentioned behaviour and disable IPv4 source address selection for the jail in favour of the primary IPv4 address of the jail. Source address selection is enabled by default for all jails and the .Va ip4.nosaddrsel setting of a parent jail is not inherited for any child jails. .It Va ip4 Control the availability of IPv4 addresses. Possible values are .Dq inherit to allow unrestricted access to all system addresses, .Dq new to restrict addresses via .Va ip4.addr , and .Dq disable to stop the jail from using IPv4 entirely. Setting the .Va ip4.addr parameter implies a value of .Dq new . .It Va ip6.addr , Va ip6.saddrsel , Va ip6 A set of IPv6 options for the jail, the counterparts to .Va ip4.addr , .Va ip4.saddrsel and .Va ip4 above. .It Va vnet Create the jail with its own virtual network stack, with its own network interfaces, addresses, routing table, etc. The kernel must have been compiled with the .Sy VIMAGE option for this to be available. Possible values are .Dq inherit to use the system network stack, possibly with restricted IP addresses, and .Dq new to create a new network stack. .It Va host.hostname The hostname of the jail. Other similar parameters are .Va host.domainname , .Va host.hostuuid and .Va host.hostid . .It Va host Set the origin of hostname and related information. Possible values are .Dq inherit to use the system information and .Dq new for the jail to use the information from the above fields. Setting any of the above fields implies a value of .Dq new . .It Va securelevel The value of the jail's .Va kern.securelevel sysctl. A jail never has a lower securelevel than its parent system, but by setting this parameter it may have a higher one. If the system securelevel is changed, any jail securelevels will be at least as secure. .It Va devfs_ruleset The number of the devfs ruleset that is enforced for mounting devfs in this jail. A value of zero (default) means no ruleset is enforced. Descendant jails inherit the parent jail's devfs ruleset enforcement. Mounting devfs inside a jail is possible only if the .Va allow.mount and .Va allow.mount.devfs permissions are effective and .Va enforce_statfs is set to a value lower than 2. Devfs rules and rulesets cannot be viewed or modified from inside a jail. .Pp NOTE: It is important that only appropriate device nodes in devfs be exposed to a jail; access to disk devices in the jail may permit processes in the jail to bypass the jail sandboxing by modifying files outside of the jail. See .Xr devfs 8 for information on how to use devfs rules to limit access to entries in the per-jail devfs. A simple devfs ruleset for jails is available as ruleset #4 in .Pa /etc/defaults/devfs.rules . .It Va children.max The number of child jails allowed to be created by this jail (or by other jails under this jail). This limit is zero by default, indicating the jail is not allowed to create child jails. See the .Sx "Hierarchical Jails" section for more information. .It Va children.cur The number of descendants of this jail, including its own child jails and any jails created under them. .It Va enforce_statfs This determines what information processes in a jail are able to get about mount points. It affects the behaviour of the following syscalls: .Xr statfs 2 , .Xr fstatfs 2 , .Xr getfsstat 2 , and .Xr fhstatfs 2 (as well as similar compatibility syscalls). When set to 0, all mount points are available without any restrictions. When set to 1, only mount points below the jail's chroot directory are visible. In addition to that, the path to the jail's chroot directory is removed from the front of their pathnames. When set to 2 (default), above syscalls can operate only on a mount-point where the jail's chroot directory is located. .It Va persist Setting this boolean parameter allows a jail to exist without any processes. Normally, a command is run as part of jail creation, and then the jail is destroyed as its last process exits. A new jail must have either the .Va persist parameter or .Va exec.start or .Va command pseudo-parameter set. .It Va cpuset.id The ID of the cpuset associated with this jail (read-only). .It Va dying This is true if the jail is in the process of shutting down (read-only). .It Va parent The .Va jid of the parent of this jail, or zero if this is a top-level jail (read-only). .It Va osrelease The string for the jail's .Va kern.osrelease sysctl and uname -r. .It Va osreldate The number for the jail's .Va kern.osreldate and uname -K. .It Va allow.* Some restrictions of the jail environment may be set on a per-jail basis. With the exception of .Va allow.set_hostname , these boolean parameters are off by default. .Bl -tag -width indent .It Va allow.set_hostname The jail's hostname may be changed via .Xr hostname 1 or .Xr sethostname 3 . .It Va allow.sysvipc A process within the jail has access to System V IPC primitives. In the current jail implementation, System V primitives share a single namespace across the host and jail environments, meaning that processes within a jail would be able to communicate with (and potentially interfere with) processes outside of the jail, and in other jails. .It Va allow.raw_sockets The jail root is allowed to create raw sockets. Setting this parameter allows utilities like .Xr ping 8 and .Xr traceroute 8 to operate inside the jail. If this is set, the source IP addresses are enforced to comply with the IP address bound to the jail, regardless of whether or not the .Dv IP_HDRINCL flag has been set on the socket. Since raw sockets can be used to configure and interact with various network subsystems, extra caution should be used where privileged access to jails is given out to untrusted parties. .It Va allow.chflags Normally, privileged users inside a jail are treated as unprivileged by .Xr chflags 2 . When this parameter is set, such users are treated as privileged, and may manipulate system file flags subject to the usual constraints on .Va kern.securelevel . .It Va allow.mount privileged users inside the jail will be able to mount and unmount file system types marked as jail-friendly. The .Xr lsvfs 1 command can be used to find file system types available for mount from within a jail. This permission is effective only if .Va enforce_statfs is set to a value lower than 2. .It Va allow.mount.devfs privileged users inside the jail will be able to mount and unmount the devfs file system. This permission is effective only together with .Va allow.mount and only when .Va enforce_statfs is set to a value lower than 2. The devfs ruleset should be restricted from the default by using the .Va devfs_ruleset option. .It Va allow.mount.fdescfs privileged users inside the jail will be able to mount and unmount the fdescfs file system. This permission is effective only together with .Va allow.mount and only when .Va enforce_statfs is set to a value lower than 2. .It Va allow.mount.nullfs privileged users inside the jail will be able to mount and unmount the nullfs file system. This permission is effective only together with .Va allow.mount and only when .Va enforce_statfs is set to a value lower than 2. .It Va allow.mount.procfs privileged users inside the jail will be able to mount and unmount the procfs file system. This permission is effective only together with .Va allow.mount and only when .Va enforce_statfs is set to a value lower than 2. .It Va allow.mount.tmpfs privileged users inside the jail will be able to mount and unmount the tmpfs file system. This permission is effective only together with .Va allow.mount and only when .Va enforce_statfs is set to a value lower than 2. .It Va allow.mount.zfs privileged users inside the jail will be able to mount and unmount the ZFS file system. This permission is effective only together with .Va allow.mount and only when .Va enforce_statfs is set to a value lower than 2. See .Xr zfs 8 for information on how to configure the ZFS filesystem to operate from within a jail. .It Va allow.quotas The jail root may administer quotas on the jail's filesystem(s). This includes filesystems that the jail may share with other jails or with non-jailed parts of the system. .It Va allow.socket_af Sockets within a jail are normally restricted to IPv4, IPv6, local (UNIX), and route. This allows access to other protocol stacks that have not had jail functionality added to them. .El .El .Pp There are pseudo-parameters that are not passed to the kernel, but are used by .Nm to set up the jail environment, often by running specified commands when jails are created or removed. The .Va exec.* command parameters are .Xr sh 1 command lines that are run in either the system or jail environment. They may be given multiple values, which run would the specified commands in sequence. All commands must succeed (return a zero exit status), or the jail will not be created or removed, as appropriate. .Pp The pseudo-parameters are: .Bl -tag -width indent .It Va exec.prestart Command(s) to run in the system environment before a jail is created. .It Va exec.start Command(s) to run in the jail environment when a jail is created. A typical command to run is .Dq sh /etc/rc . .It Va command A synonym for .Va exec.start for use when specifying a jail directly on the command line. Unlike other parameters whose value is a single string, .Va command uses the remainder of the .Nm command line as its own arguments. .It Va exec.poststart Command(s) to run in the system environment after a jail is created, and after any .Va exec.start commands have completed. .It Va exec.prestop Command(s) to run in the system environment before a jail is removed. .It Va exec.stop Command(s) to run in the jail environment before a jail is removed, and after any .Va exec.prestop commands have completed. A typical command to run is .Dq sh /etc/rc.shutdown . .It Va exec.poststop Command(s) to run in the system environment after a jail is removed. .It Va exec.clean Run commands in a clean environment. The environment is discarded except for .Ev HOME , SHELL , TERM and .Ev USER . .Ev HOME and .Ev SHELL are set to the target login's default values. .Ev USER is set to the target login. .Ev TERM is imported from the current environment. The environment variables from the login class capability database for the target login are also set. .It Va exec.jail_user The user to run commands as, when running in the jail environment. The default is to run the commands as the current user. .It Va exec.system_jail_user This boolean option looks for the .Va exec.jail_user in the system .Xr passwd 5 file, instead of in the jail's file. .It Va exec.system_user The user to run commands as, when running in the system environment. The default is to run the commands as the current user. .It Va exec.timeout The maximum amount of time to wait for a command to complete, in seconds. If a command is still running after this timeout has passed, the jail will not be created or removed, as appropriate. .It Va exec.consolelog A file to direct command output (stdout and stderr) to. .It Va exec.fib The FIB (routing table) to set when running commands inside the jail. .It Va stop.timeout The maximum amount of time to wait for a jail's processes to exit after sending them a .Dv SIGTERM signal (which happens after the .Va exec.stop commands have completed). After this many seconds have passed, the jail will be removed, which will kill any remaining processes. If this is set to zero, no .Dv SIGTERM is sent and the jail is immediately removed. The default is 10 seconds. .It Va interface A network interface to add the jail's IP addresses .Va ( ip4.addr and .Va ip6.addr ) to. An alias for each address will be added to the interface before the jail is created, and will be removed from the interface after the jail is removed. .It Va ip4.addr In addition to the IP addresses that are passed to the kernel, an interface, netmask and additional parameters (as supported by .Xr ifconfig 8 Ns ) may also be specified, in the form .Dq Ar interface Ns | Ns Ar ip-address Ns / Ns Ar netmask param ... . If an interface is given before the IP address, an alias for the address will be added to that interface, as it is with the .Va interface parameter. If a netmask in either dotted-quad or CIDR form is given after an IP address, it will be used when adding the IP alias. If additional parameters are specified then they will also be used when adding the IP alias. .It Va ip6.addr In addition to the IP addresses that are passed to the kernel, an interface, prefix and additional parameters (as supported by .Xr ifconfig 8 Ns ) may also be specified, in the form .Dq Ar interface Ns | Ns Ar ip-address Ns / Ns Ar prefix param ... . .It Va vnet.interface A network interface to give to a vnet-enabled jail after is it created. The interface will automatically be released when the jail is removed. .It Va ip_hostname Resolve the .Va host.hostname parameter and add all IP addresses returned by the resolver to the list of addresses .Po Va ip4.addr or .Va ip6.addr Pc for this jail. This may affect default address selection for outgoing IPv4 connections from jails. The address first returned by the resolver for each address family will be used as the primary address. .It Va mount A filesystem to mount before creating the jail (and to unmount after removing it), given as a single .Xr fstab 5 line. .It Va mount.fstab An .Xr fstab 5 format file containing filesystems to mount before creating a jail. .It Va mount.devfs Mount a .Xr devfs 5 filesystem on the chrooted .Pa /dev directory, and apply the ruleset in the .Va devfs_ruleset parameter (or a default of ruleset 4: devfsrules_jail) to restrict the devices visible inside the jail. .It Va mount.fdescfs Mount a .Xr fdescfs 5 filesystem on the chrooted .Pa /dev/fd directory. .It Va mount.procfs Mount a .Xr procfs 5 filesystem on the chrooted .Pa /proc directory. .It Va allow.dying Allow making changes to a .Va dying jail. .It Va depend Specify a jail (or jails) that this jail depends on. Any such jails must be fully created, up to the last .Va exec.poststart command, before any action will taken to create this jail. When jails are removed the opposite is true: this jail must be fully removed, up to the last .Va exec.poststop command, before the jail(s) it depends on are stopped. .El .Sh EXAMPLES Jails are typically set up using one of two philosophies: either to constrain a specific application (possibly running with privilege), or to create a .Dq "virtual system image" running a variety of daemons and services. In both cases, a fairly complete file system install of .Fx is required, so as to provide the necessary command line tools, daemons, libraries, application configuration files, etc. However, for a virtual server configuration, a fair amount of additional work is required so as to replace the .Dq boot process. This manual page documents the configuration steps necessary to support either of these steps, although the configuration steps may need to be refined based on local requirements. .Ss "Setting up a Jail Directory Tree" To set up a jail directory tree containing an entire .Fx distribution, the following .Xr sh 1 command script can be used: .Bd -literal D=/here/is/the/jail cd /usr/src mkdir -p $D make world DESTDIR=$D make distribution DESTDIR=$D .Ed .Pp In many cases this example would put far more in the jail than needed. In the other extreme case a jail might contain only one file: the executable to be run in the jail. .Pp We recommend experimentation, and caution that it is a lot easier to start with a .Dq fat jail and remove things until it stops working, than it is to start with a .Dq thin jail and add things until it works. .Ss "Setting Up a Jail" Do what was described in .Sx "Setting Up a Jail Directory Tree" to build the jail directory tree. For the sake of this example, we will assume you built it in .Pa /data/jail/testjail , for a jail named .Dq testjail . Substitute below as needed with your own directory, IP address, and hostname. .Ss "Setting up the Host Environment" First, set up the real system's environment to be .Dq jail-friendly . For consistency, we will refer to the parent box as the .Dq "host environment" , and to the jailed virtual machine as the .Dq "jail environment" . Since jails are implemented using IP aliases, one of the first things to do is to disable IP services on the host system that listen on all local IP addresses for a service. If a network service is present in the host environment that binds all available IP addresses rather than specific IP addresses, it may service requests sent to jail IP addresses if the jail did not bind the port. This means changing .Xr inetd 8 to only listen on the appropriate IP address, and so forth. Add the following to .Pa /etc/rc.conf in the host environment: .Bd -literal -offset indent sendmail_enable="NO" inetd_flags="-wW -a 192.0.2.23" rpcbind_enable="NO" .Ed .Pp .Li 192.0.2.23 is the native IP address for the host system, in this example. Daemons that run out of .Xr inetd 8 can be easily configured to use only the specified host IP address. Other daemons will need to be manually configured \(em for some this is possible through .Xr rc.conf 5 flags entries; for others it is necessary to modify per-application configuration files, or to recompile the application. The following frequently deployed services must have their individual configuration files modified to limit the application to listening to a specific IP address: .Pp To configure .Xr sshd 8 , it is necessary to modify .Pa /etc/ssh/sshd_config . .Pp To configure .Xr sendmail 8 , it is necessary to modify .Pa /etc/mail/sendmail.cf . .Pp For .Xr named 8 , it is necessary to modify .Pa /etc/namedb/named.conf . .Pp In addition, a number of services must be recompiled in order to run them in the host environment. This includes most applications providing services using .Xr rpc 3 , such as .Xr rpcbind 8 , .Xr nfsd 8 , and .Xr mountd 8 . In general, applications for which it is not possible to specify which IP address to bind should not be run in the host environment unless they should also service requests sent to jail IP addresses. Attempting to serve NFS from the host environment may also cause confusion, and cannot be easily reconfigured to use only specific IPs, as some NFS services are hosted directly from the kernel. Any third-party network software running in the host environment should also be checked and configured so that it does not bind all IP addresses, which would result in those services also appearing to be offered by the jail environments. .Pp Once these daemons have been disabled or fixed in the host environment, it is best to reboot so that all daemons are in a known state, to reduce the potential for confusion later (such as finding that when you send mail to a jail, and its sendmail is down, the mail is delivered to the host, etc.). .Ss "Configuring the Jail" Start any jail for the first time without configuring the network interface so that you can clean it up a little and set up accounts. As with any machine (virtual or not), you will need to set a root password, time zone, etc. Some of these steps apply only if you intend to run a full virtual server inside the jail; others apply both for constraining a particular application or for running a virtual server. .Pp Start a shell in the jail: .Bd -literal -offset indent jail -c path=/data/jail/testjail mount.devfs \\ host.hostname=testhostname ip4.addr=192.0.2.100 \\ command=/bin/sh .Ed .Pp Assuming no errors, you will end up with a shell prompt within the jail. You can now run .Pa /usr/sbin/sysinstall and do the post-install configuration to set various configuration options, or perform these actions manually by editing .Pa /etc/rc.conf , etc. .Pp .Bl -bullet -offset indent -compact .It Configure .Pa /etc/resolv.conf so that name resolution within the jail will work correctly. .It Run .Xr newaliases 1 to quell .Xr sendmail 8 warnings. .It Set a root password, probably different from the real host system. .It Set the timezone. .It Add accounts for users in the jail environment. .It Install any packages the environment requires. .El .Pp You may also want to perform any package-specific configuration (web servers, SSH servers, etc), patch up .Pa /etc/syslog.conf so it logs as you would like, etc. If you are not using a virtual server, you may wish to modify .Xr syslogd 8 in the host environment to listen on the syslog socket in the jail environment; in this example, the syslog socket would be stored in .Pa /data/jail/testjail/var/run/log . .Pp Exit from the shell, and the jail will be shut down. .Ss "Starting the Jail" You are now ready to restart the jail and bring up the environment with all of its daemons and other programs. Create an entry for the jail in .Pa /etc/jail.conf : .Bd -literal -offset indent testjail { path = /tmp/jail/testjail; mount.devfs; host.hostname = testhostname; ip4.addr = 192.0.2.100; interface = ed0; exec.start = "/bin/sh /etc/rc"; exec.stop = "/bin/sh /etc/rc.shutdown"; } .Ed .Pp To start a virtual server environment, .Pa /etc/rc is run to launch various daemons and services, and .Pa /etc/rc.shutdown is run to shut them down when the jail is removed. If you are running a single application in the jail, substitute the command used to start the application for .Dq /bin/sh /etc/rc ; there may be some script available to cleanly shut down the application, or it may be sufficient to go without a stop command, and have .Nm send .Dv SIGTERM to the application. .Pp Start the jail by running: .Bd -literal -offset indent jail -c testjail .Ed .Pp A few warnings may be produced; however, it should all work properly. You should be able to see .Xr inetd 8 , .Xr syslogd 8 , and other processes running within the jail using .Xr ps 1 , with the .Ql J flag appearing beside jailed processes. To see an active list of jails, use .Xr jls 8 . If .Xr sshd 8 is enabled in the jail environment, you should be able to .Xr ssh 1 to the hostname or IP address of the jailed environment, and log in using the accounts you created previously. .Pp It is possible to have jails started at boot time. Please refer to the .Dq jail_* variables in .Xr rc.conf 5 for more information. .Ss "Managing the Jail" Normal machine shutdown commands, such as .Xr halt 8 , .Xr reboot 8 , and .Xr shutdown 8 , cannot be used successfully within the jail. To kill all processes from within a jail, you may use one of the following commands, depending on what you want to accomplish: .Bd -literal -offset indent kill -TERM -1 kill -KILL -1 .Ed .Pp This will send the .Dv SIGTERM or .Dv SIGKILL signals to all processes in the jail \(em be careful not to run this from the host environment! Once all of the jail's processes have died, unless the jail was created with the .Va persist parameter, the jail will be removed. Depending on the intended use of the jail, you may also want to run .Pa /etc/rc.shutdown from within the jail. .Pp To shut down the jail from the outside, simply remove it with .Nm .Ar -r , which will run any commands specified by .Va exec.stop , and then send .Dv SIGTERM and eventually .Dv SIGKILL to any remaining jailed processes. .Pp The .Pa /proc/ Ns Ar pid Ns Pa /status file contains, as its last field, the name of the jail in which the process runs, or .Dq Li - to indicate that the process is not running within a jail. The .Xr ps 1 command also shows a .Ql J flag for processes in a jail. .Pp You can also list/kill processes based on their jail ID. To show processes and their jail ID, use the following command: .Pp .Dl "ps ax -o pid,jid,args" .Pp To show and then kill processes in jail number 3 use the following commands: .Bd -literal -offset indent pgrep -lfj 3 pkill -j 3 .Ed or: .Pp .Dl "killall -j 3" .Ss "Jails and File Systems" It is not possible to .Xr mount 8 or .Xr umount 8 any file system inside a jail unless the file system is marked jail-friendly, the jail's .Va allow.mount parameter is set, and the jail's .Va enforce_statfs parameter is lower than 2. .Pp Multiple jails sharing the same file system can influence each other. For example, a user in one jail can fill the file system, leaving no space for processes in the other jail. Trying to use .Xr quota 1 to prevent this will not work either, as the file system quotas are not aware of jails but only look at the user and group IDs. This means the same user ID in two jails share a single file system quota. One would need to use one file system per jail to make this work. .Ss "Sysctl MIB Entries" The read-only entry .Va security.jail.jailed can be used to determine if a process is running inside a jail (value is one) or not (value is zero). .Pp The variable .Va security.jail.max_af_ips determines how may address per address family a jail may have. The default is 255. .Pp Some MIB variables have per-jail settings. Changes to these variables by a jailed process do not affect the host environment, only the jail environment. These variables are .Va kern.securelevel , .Va kern.hostname , .Va kern.domainname , .Va kern.hostid , and .Va kern.hostuuid . .Ss "Hierarchical Jails" By setting a jail's .Va children.max parameter, processes within a jail may be able to create jails of their own. These child jails are kept in a hierarchy, with jails only able to see and/or modify the jails they created (or those jails' children). Each jail has a read-only .Va parent parameter, containing the .Va jid of the jail that created it; a .Va jid of 0 indicates the jail is a child of the current jail (or is a top-level jail if the current process isn't jailed). .Pp Jailed processes are not allowed to confer greater permissions than they themselves are given, e.g., if a jail is created with .Va allow.nomount , it is not able to create a jail with .Va allow.mount set. Similarly, such restrictions as .Va ip4.addr and .Va securelevel may not be bypassed in child jails. .Pp A child jail may in turn create its own child jails if its own .Va children.max parameter is set (remember it is zero by default). These jails are visible to and can be modified by their parent and all ancestors. .Pp Jail names reflect this hierarchy, with a full name being an MIB-type string separated by dots. For example, if a base system process creates a jail .Dq foo , and a process under that jail creates another jail .Dq bar , then the second jail will be seen as .Dq foo.bar in the base system (though it is only seen as .Dq bar to any processes inside jail .Dq foo ) . Jids on the other hand exist in a single space, and each jail must have a unique jid. .Pp Like the names, a child jail's .Va path appears relative to its creator's own .Va path . This is by virtue of the child jail being created in the chrooted environment of the first jail. .Sh SEE ALSO .Xr killall 1 , .Xr lsvfs 1 , .Xr newaliases 1 , .Xr pgrep 1 , .Xr pkill 1 , .Xr ps 1 , .Xr quota 1 , .Xr jail_set 2 , .Xr devfs 5 , .Xr fdescfs 5 , .Xr jail.conf 5 , .Xr procfs 5 , .Xr rc.conf 5 , .Xr sysctl.conf 5 , .Xr chroot 8 , .Xr devfs 8 , .Xr halt 8 , .Xr ifconfig 8 , .Xr inetd 8 , .Xr jexec 8 , .Xr jls 8 , .Xr mount 8 , .Xr named 8 , -.Xr procfs 5 , .Xr reboot 8 , .Xr rpcbind 8 , .Xr sendmail 8 , .Xr shutdown 8 , .Xr sysctl 8 , .Xr syslogd 8 , .Xr umount 8 .Sh HISTORY The .Nm utility appeared in .Fx 4.0 . Hierarchical/extensible jails were introduced in .Fx 8.0 . The configuration file was introduced in .Fx 9.1 . .Sh AUTHORS .An -nosplit The jail feature was written by .An Poul-Henning Kamp for R&D Associates .Pa http://www.rndassociates.com/ who contributed it to .Fx . .Pp .An Robert Watson wrote the extended documentation, found a few bugs, added a few new features, and cleaned up the userland jail environment. .Pp .An Bjoern A. Zeeb added multi-IP jail support for IPv4 and IPv6 based on a patch originally done by .An Pawel Jakub Dawidek for IPv4. .Pp .An James Gritton added the extensible jail parameters, hierarchical jails, and the configuration file. .Sh BUGS It might be a good idea to add an address alias flag such that daemons listening on all IPs .Pq Dv INADDR_ANY will not bind on that address, which would facilitate building a safe host environment such that host daemons do not impose on services offered from within jails. Currently, the simplest answer is to minimize services offered on the host, possibly limiting it to services offered from .Xr inetd 8 which is easily configurable. .Sh NOTES Great care should be taken when managing directories visible within the jail. For example, if a jailed process has its current working directory set to a directory that is moved out of the jail's chroot, then the process may gain access to the file space outside of the jail. It is recommended that directories always be copied, rather than moved, out of a jail. .Pp In addition, there are several ways in which an unprivileged user outside the jail can cooperate with a privileged user inside the jail and thereby obtain elevated privileges in the host environment. Most of these attacks can be mitigated by ensuring that the jail root is not accessible to unprivileged users in the host environment. Regardless, as a general rule, untrusted users with privileged access to a jail should not be given access to the host environment. Index: user/ngie/more-tests/usr.sbin/jail =================================================================== --- user/ngie/more-tests/usr.sbin/jail (revision 281974) +++ user/ngie/more-tests/usr.sbin/jail (revision 281975) Property changes on: user/ngie/more-tests/usr.sbin/jail ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/usr.sbin/jail:r281414-281974 Index: user/ngie/more-tests =================================================================== --- user/ngie/more-tests (revision 281974) +++ user/ngie/more-tests (revision 281975) Property changes on: user/ngie/more-tests ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r281925-281974