Index: head/cddl/lib/libdtrace/tcp.d =================================================================== --- head/cddl/lib/libdtrace/tcp.d (revision 334718) +++ head/cddl/lib/libdtrace/tcp.d (revision 334719) @@ -1,392 +1,392 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END * * $FreeBSD$ */ /* * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2013 Mark Johnston */ #pragma D depends_on library ip.d #pragma D depends_on module kernel #pragma D depends_on provider tcp /* * Convert a TCP state value to a string. */ #pragma D binding "1.6.3" TCPS_CLOSED inline int TCPS_CLOSED = 0; #pragma D binding "1.6.3" TCPS_LISTEN inline int TCPS_LISTEN = 1; #pragma D binding "1.6.3" TCPS_SYN_SENT inline int TCPS_SYN_SENT = 2; #pragma D binding "1.6.3" TCPS_SYN_RECEIVED inline int TCPS_SYN_RECEIVED = 3; #pragma D binding "1.6.3" TCPS_ESTABLISHED inline int TCPS_ESTABLISHED = 4; #pragma D binding "1.6.3" TCPS_CLOSE_WAIT inline int TCPS_CLOSE_WAIT = 5; #pragma D binding "1.6.3" TCPS_FIN_WAIT_1 inline int TCPS_FIN_WAIT_1 = 6; #pragma D binding "1.6.3" TCPS_CLOSING inline int TCPS_CLOSING = 7; #pragma D binding "1.6.3" TCPS_LAST_ACK inline int TCPS_LAST_ACK = 8; #pragma D binding "1.6.3" TCPS_FIN_WAIT_2 inline int TCPS_FIN_WAIT_2 = 9; #pragma D binding "1.6.3" TCPS_TIME_WAIT inline int TCPS_TIME_WAIT = 10; /* TCP segment flags. */ #pragma D binding "1.6.3" TH_FIN inline uint8_t TH_FIN = 0x01; #pragma D binding "1.6.3" TH_SYN inline uint8_t TH_SYN = 0x02; #pragma D binding "1.6.3" TH_RST inline uint8_t TH_RST = 0x04; #pragma D binding "1.6.3" TH_PUSH inline uint8_t TH_PUSH = 0x08; #pragma D binding "1.6.3" TH_ACK inline uint8_t TH_ACK = 0x10; #pragma D binding "1.6.3" TH_URG inline uint8_t TH_URG = 0x20; #pragma D binding "1.6.3" TH_ECE inline uint8_t TH_ECE = 0x40; #pragma D binding "1.6.3" TH_CWR inline uint8_t TH_CWR = 0x80; /* TCP connection state strings. */ #pragma D binding "1.6.3" tcp_state_string inline string tcp_state_string[int32_t state] = state == TCPS_CLOSED ? "state-closed" : state == TCPS_LISTEN ? "state-listen" : state == TCPS_SYN_SENT ? "state-syn-sent" : state == TCPS_SYN_RECEIVED ? "state-syn-received" : state == TCPS_ESTABLISHED ? "state-established" : state == TCPS_CLOSE_WAIT ? "state-close-wait" : state == TCPS_FIN_WAIT_1 ? "state-fin-wait-1" : state == TCPS_CLOSING ? "state-closing" : state == TCPS_LAST_ACK ? "state-last-ack" : state == TCPS_FIN_WAIT_2 ? "state-fin-wait-2" : state == TCPS_TIME_WAIT ? "state-time-wait" : ""; /* * tcpsinfo contains stable TCP details from tcp_t. */ typedef struct tcpsinfo { uintptr_t tcps_addr; int tcps_local; /* is delivered locally, boolean */ int tcps_active; /* active open (from here), boolean */ uint16_t tcps_lport; /* local port */ uint16_t tcps_rport; /* remote port */ string tcps_laddr; /* local address, as a string */ string tcps_raddr; /* remote address, as a string */ int32_t tcps_state; /* TCP state */ uint32_t tcps_iss; /* Initial sequence # sent */ uint32_t tcps_irs; /* Initial sequence # received */ uint32_t tcps_suna; /* sequence # sent but unacked */ uint32_t tcps_smax; /* highest sequence number sent */ uint32_t tcps_snxt; /* next sequence # to send */ uint32_t tcps_rack; /* sequence # we have acked */ uint32_t tcps_rnxt; /* next sequence # expected */ u_long tcps_swnd; /* send window size */ int32_t tcps_snd_ws; /* send window scaling */ uint32_t tcps_swl1; /* window update seg seq number */ uint32_t tcps_swl2; /* window update seg ack number */ uint32_t tcps_rup; /* receive urgent pointer */ uint32_t tcps_radv; /* advertised window */ u_long tcps_rwnd; /* receive window size */ int32_t tcps_rcv_ws; /* receive window scaling */ u_long tcps_cwnd; /* congestion window */ u_long tcps_cwnd_ssthresh; /* threshold for congestion avoidance */ uint32_t tcps_srecover; /* for use in NewReno Fast Recovery */ uint32_t tcps_sack_fack; /* SACK sequence # we have acked */ uint32_t tcps_sack_snxt; /* next SACK seq # for retransmission */ uint32_t tcps_rto; /* round-trip timeout, msec */ uint32_t tcps_mss; /* max segment size */ int tcps_retransmit; /* retransmit send event, boolean */ int tcps_srtt; /* smoothed RTT in units of (TCP_RTT_SCALE*hz) */ int tcps_debug; /* socket has SO_DEBUG set */ int tcps_cookie; /* expose the socket's SO_USER_COOKIE */ int32_t tcps_dupacks; /* consecutive dup acks received */ uint32_t tcps_rtttime; /* RTT measurement start time */ uint32_t tcps_rtseq; /* sequence # being timed */ uint32_t tcps_ts_recent; /* timestamp echo data */ } tcpsinfo_t; /* * tcplsinfo provides the old tcp state for state changes. */ typedef struct tcplsinfo { int32_t tcps_state; /* previous TCP state */ } tcplsinfo_t; /* * tcpinfo is the TCP header fields. */ typedef struct tcpinfo { uint16_t tcp_sport; /* source port */ uint16_t tcp_dport; /* destination port */ uint32_t tcp_seq; /* sequence number */ uint32_t tcp_ack; /* acknowledgment number */ uint8_t tcp_offset; /* data offset, in bytes */ uint8_t tcp_flags; /* flags */ uint16_t tcp_window; /* window size */ uint16_t tcp_checksum; /* checksum */ uint16_t tcp_urgent; /* urgent data pointer */ struct tcphdr *tcp_hdr; /* raw TCP header */ } tcpinfo_t; /* * A clone of tcpinfo_t used to handle the fact that the TCP input path * overwrites some fields of the TCP header with their host-order equivalents. * Unfortunately, DTrace doesn't let us simply typedef a new name for struct * tcpinfo and define a separate translator for it. */ typedef struct tcpinfoh { uint16_t tcp_sport; /* source port */ uint16_t tcp_dport; /* destination port */ uint32_t tcp_seq; /* sequence number */ uint32_t tcp_ack; /* acknowledgment number */ uint8_t tcp_offset; /* data offset, in bytes */ uint8_t tcp_flags; /* flags */ uint16_t tcp_window; /* window size */ uint16_t tcp_checksum; /* checksum */ uint16_t tcp_urgent; /* urgent data pointer */ struct tcphdr *tcp_hdr; /* raw TCP header */ } tcpinfoh_t; #pragma D binding "1.6.3" translator translator csinfo_t < struct tcpcb *p > { cs_addr = NULL; cs_cid = (uint64_t)(p == NULL ? 0 : p->t_inpcb); cs_pid = 0; cs_zoneid = 0; }; #pragma D binding "1.6.3" translator translator tcpsinfo_t < struct tcpcb *p > { tcps_addr = (uintptr_t)p; tcps_local = -1; /* XXX */ tcps_active = -1; /* XXX */ tcps_lport = p == NULL ? 0 : ntohs(p->t_inpcb->inp_inc.inc_ie.ie_lport); tcps_rport = p == NULL ? 0 : ntohs(p->t_inpcb->inp_inc.inc_ie.ie_fport); tcps_laddr = p == NULL ? 0 : p->t_inpcb->inp_vflag == INP_IPV4 ? - inet_ntoa(&p->t_inpcb->inp_inc.inc_ie.ie_dependladdr.ie46_local.ia46_addr4.s_addr) : - inet_ntoa6(&p->t_inpcb->inp_inc.inc_ie.ie_dependladdr.ie6_local); + inet_ntoa(&p->t_inpcb->inp_inc.inc_ie.ie_dependladdr.id46_addr.ia46_addr4.s_addr) : + inet_ntoa6(&p->t_inpcb->inp_inc.inc_ie.ie_dependladdr.id6_addr); tcps_raddr = p == NULL ? 0 : p->t_inpcb->inp_vflag == INP_IPV4 ? - inet_ntoa(&p->t_inpcb->inp_inc.inc_ie.ie_dependfaddr.ie46_foreign.ia46_addr4.s_addr) : - inet_ntoa6(&p->t_inpcb->inp_inc.inc_ie.ie_dependfaddr.ie6_foreign); + inet_ntoa(&p->t_inpcb->inp_inc.inc_ie.ie_dependfaddr.id46_addr.ia46_addr4.s_addr) : + inet_ntoa6(&p->t_inpcb->inp_inc.inc_ie.ie_dependfaddr.id6_addr); tcps_state = p == NULL ? -1 : p->t_state; tcps_iss = p == NULL ? 0 : p->iss; tcps_irs = p == NULL ? 0 : p->irs; tcps_suna = p == NULL ? 0 : p->snd_una; tcps_smax = p == NULL ? 0 : p->snd_max; tcps_snxt = p == NULL ? 0 : p->snd_nxt; tcps_rack = p == NULL ? 0 : p->last_ack_sent; tcps_rnxt = p == NULL ? 0 : p->rcv_nxt; tcps_swnd = p == NULL ? -1 : p->snd_wnd; tcps_snd_ws = p == NULL ? -1 : p->snd_scale; tcps_swl1 = p == NULL ? -1 : p->snd_wl1; tcps_swl2 = p == NULL ? -1 : p->snd_wl2; tcps_radv = p == NULL ? -1 : p->rcv_adv; tcps_rwnd = p == NULL ? -1 : p->rcv_wnd; tcps_rup = p == NULL ? -1 : p->rcv_up; tcps_rcv_ws = p == NULL ? -1 : p->rcv_scale; tcps_cwnd = p == NULL ? -1 : p->snd_cwnd; tcps_cwnd_ssthresh = p == NULL ? -1 : p->snd_ssthresh; tcps_srecover = p == NULL ? -1 : p->snd_recover; tcps_sack_fack = p == NULL ? 0 : p->snd_fack; tcps_sack_snxt = p == NULL ? 0 : p->sack_newdata; tcps_rto = p == NULL ? -1 : (p->t_rxtcur * 1000) / `hz; tcps_mss = p == NULL ? -1 : p->t_maxseg; tcps_retransmit = p == NULL ? -1 : p->t_rxtshift > 0 ? 1 : 0; tcps_srtt = p == NULL ? -1 : p->t_srtt; /* smoothed RTT in units of (TCP_RTT_SCALE*hz) */ tcps_debug = p == NULL ? 0 : p->t_inpcb->inp_socket->so_options & 1; tcps_cookie = p == NULL ? -1 : p->t_inpcb->inp_socket->so_user_cookie; tcps_dupacks = p == NULL ? -1 : p->t_dupacks; tcps_rtttime = p == NULL ? -1 : p->t_rtttime; tcps_rtseq = p == NULL ? -1 : p->t_rtseq; tcps_ts_recent = p == NULL ? -1 : p->ts_recent; }; #pragma D binding "1.6.3" translator translator tcpinfo_t < struct tcphdr *p > { tcp_sport = p == NULL ? 0 : ntohs(p->th_sport); tcp_dport = p == NULL ? 0 : ntohs(p->th_dport); tcp_seq = p == NULL ? -1 : ntohl(p->th_seq); tcp_ack = p == NULL ? -1 : ntohl(p->th_ack); tcp_offset = p == NULL ? -1 : (p->th_off >> 2); tcp_flags = p == NULL ? 0 : p->th_flags; tcp_window = p == NULL ? 0 : ntohs(p->th_win); tcp_checksum = p == NULL ? 0 : ntohs(p->th_sum); tcp_urgent = p == NULL ? 0 : ntohs(p->th_urp); tcp_hdr = (struct tcphdr *)p; }; /* * This translator differs from the one for tcpinfo_t in that the sequence * number, acknowledgement number, window size and urgent pointer are already * in host order and thus don't need to be converted. */ #pragma D binding "1.6.3" translator translator tcpinfoh_t < struct tcphdr *p > { tcp_sport = p == NULL ? 0 : ntohs(p->th_sport); tcp_dport = p == NULL ? 0 : ntohs(p->th_dport); tcp_seq = p == NULL ? -1 : p->th_seq; tcp_ack = p == NULL ? -1 : p->th_ack; tcp_offset = p == NULL ? -1 : (p->th_off >> 2); tcp_flags = p == NULL ? 0 : p->th_flags; tcp_window = p == NULL ? 0 : p->th_win; tcp_checksum = p == NULL ? 0 : ntohs(p->th_sum); tcp_urgent = p == NULL ? 0 : p->th_urp; tcp_hdr = (struct tcphdr *)p; }; #pragma D binding "1.6.3" translator translator tcplsinfo_t < int s > { tcps_state = s; }; /* Support for TCP debug */ #pragma D binding "1.12.1" TA_INPUT inline int TA_INPUT = 0; #pragma D binding "1.12.1" TA_OUTPUT inline int TA_OUTPUT = 1; #pragma D binding "1.12.1" TA_USER inline int TA_USER = 2; #pragma D binding "1.12.1" TA_RESPOND inline int TA_RESPOND = 3; #pragma D binding "1.12.1" TA_DROP inline int TA_DROP = 4; /* direction strings. */ #pragma D binding "1.12.1" tcpdebug_dir_string inline string tcpdebug_dir_string[uint8_t direction] = direction == TA_INPUT ? "input" : direction == TA_OUTPUT ? "output" : direction == TA_USER ? "user" : direction == TA_RESPOND ? "respond" : direction == TA_OUTPUT ? "drop" : "unknown" ; #pragma D binding "1.12.1" tcpflag_string inline string tcpflag_string[uint8_t flags] = flags & TH_FIN ? "FIN" : flags & TH_SYN ? "SYN" : flags & TH_RST ? "RST" : flags & TH_PUSH ? "PUSH" : flags & TH_ACK ? "ACK" : flags & TH_URG ? "URG" : flags & TH_ECE ? "ECE" : flags & TH_CWR ? "CWR" : "unknown" ; #pragma D binding "1.12.1" PRU_ATTACH inline int PRU_ATTACH = 0; #pragma D binding "1.12.1" PRU_DETACH inline int PRU_DETACH = 1; #pragma D binding "1.12.1" PRU_BIND inline int PRU_BIND = 2; #pragma D binding "1.12.1" PRU_LISTEN inline int PRU_LISTEN = 3; #pragma D binding "1.12.1" PRU_CONNECT inline int PRU_CONNECT = 4; #pragma D binding "1.12.1" PRU_ACCEPT inline int PRU_ACCEPT = 5 ; #pragma D binding "1.12.1" PRU_DISCONNECT inline int PRU_DISCONNECT = 6; #pragma D binding "1.12.1" PRU_SHUTDOWN inline int PRU_SHUTDOWN = 7; #pragma D binding "1.12.1" PRU_RCVD inline int PRU_RCVD = 8; #pragma D binding "1.12.1" PRU_SEND inline int PRU_SEND = 9; #pragma D binding "1.12.1" PRU_ABORT inline int PRU_ABORT = 10; #pragma D binding "1.12.1" PRU_CONTROL inline int PRU_CONTROL = 11; #pragma D binding "1.12.1" PRU_SENSE inline int PRU_SENSE = 12; #pragma D binding "1.12.1" PRU_RCVOOB inline int PRU_RCVOOB = 13; #pragma D binding "1.12.1" PRU_SENDOOB inline int PRU_SENDOOB = 14; #pragma D binding "1.12.1" PRU_SOCKADDR inline int PRU_SOCKADDR = 15; #pragma D binding "1.12.1" PRU_PEERADDR inline int PRU_PEERADDR = 16; #pragma D binding "1.12.1" PRU_CONNECT2 inline int PRU_CONNECT2 = 17; #pragma D binding "1.12.1" PRU_FASTTIMO inline int PRU_FASTTIMO = 18; #pragma D binding "1.12.1" PRU_SLOWTIMO inline int PRU_SLOWTIMO = 19; #pragma D binding "1.12.1" PRU_PROTORCV inline int PRU_PROTORCV = 20; #pragma D binding "1.12.1" PRU_PROTOSEND inline int PRU_PROTOSEND = 21; #pragma D binding "1.12.1" PRU_SEND_EOF inline int PRU_SEND_EOF = 22; #pragma D binding "1.12.1" PRU_SOSETLABEL inline int PRU_SOSETLABEL = 23; #pragma D binding "1.12.1" PRU_CLOSE inline int PRU_CLOSE = 24; #pragma D binding "1.12.1" PRU_FLUSH inline int PRU_FLUSH = 25; #pragma D binding "1.12.1" prureq_string inline string prureq_string[uint8_t req] = req == PRU_ATTACH ? "ATTACH" : req == PRU_DETACH ? "DETACH" : req == PRU_BIND ? "BIND" : req == PRU_LISTEN ? "LISTEN" : req == PRU_CONNECT ? "CONNECT" : req == PRU_ACCEPT ? "ACCEPT" : req == PRU_DISCONNECT ? "DISCONNECT" : req == PRU_SHUTDOWN ? "SHUTDOWN" : req == PRU_RCVD ? "RCVD" : req == PRU_SEND ? "SEND" : req == PRU_ABORT ? "ABORT" : req == PRU_CONTROL ? "CONTROL" : req == PRU_SENSE ? "SENSE" : req == PRU_RCVOOB ? "RCVOOB" : req == PRU_SENDOOB ? "SENDOOB" : req == PRU_SOCKADDR ? "SOCKADDR" : req == PRU_PEERADDR ? "PEERADDR" : req == PRU_CONNECT2 ? "CONNECT2" : req == PRU_FASTTIMO ? "FASTTIMO" : req == PRU_SLOWTIMO ? "SLOWTIMO" : req == PRU_PROTORCV ? "PROTORCV" : req == PRU_PROTOSEND ? "PROTOSEND" : req == PRU_SEND ? "SEND_EOF" : req == PRU_SOSETLABEL ? "SOSETLABEL" : req == PRU_CLOSE ? "CLOSE" : req == PRU_FLUSH ? "FLUSE" : "unknown" ; Index: head/lib/libc/sys/getsockopt.2 =================================================================== --- head/lib/libc/sys/getsockopt.2 (revision 334718) +++ head/lib/libc/sys/getsockopt.2 (revision 334719) @@ -1,578 +1,588 @@ .\" Copyright (c) 1983, 1991, 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. .\" 3. 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. .\" .\" @(#)getsockopt.2 8.4 (Berkeley) 5/2/95 .\" $FreeBSD$ .\" .Dd May 9, 2018 .Dt GETSOCKOPT 2 .Os .Sh NAME .Nm getsockopt , .Nm setsockopt .Nd get and set options on sockets .Sh LIBRARY .Lb libc .Sh SYNOPSIS .In sys/types.h .In sys/socket.h .Ft int .Fn getsockopt "int s" "int level" "int optname" "void * restrict optval" "socklen_t * restrict optlen" .Ft int .Fn setsockopt "int s" "int level" "int optname" "const void *optval" "socklen_t optlen" .Sh DESCRIPTION The .Fn getsockopt and .Fn setsockopt system calls manipulate the .Em options associated with a socket. Options may exist at multiple protocol levels; they are always present at the uppermost .Dq socket level. .Pp When manipulating socket options the level at which the option resides and the name of the option must be specified. To manipulate options at the socket level, .Fa level is specified as .Dv SOL_SOCKET . To manipulate options at any other level the protocol number of the appropriate protocol controlling the option is supplied. For example, to indicate that an option is to be interpreted by the .Tn TCP protocol, .Fa level should be set to the protocol number of .Tn TCP ; see .Xr getprotoent 3 . .Pp The .Fa optval and .Fa optlen arguments are used to access option values for .Fn setsockopt . For .Fn getsockopt they identify a buffer in which the value for the requested option(s) are to be returned. For .Fn getsockopt , .Fa optlen is a value-result argument, initially containing the size of the buffer pointed to by .Fa optval , and modified on return to indicate the actual size of the value returned. If no option value is to be supplied or returned, .Fa optval may be NULL. .Pp The .Fa optname argument and any specified options are passed uninterpreted to the appropriate protocol module for interpretation. The include file .In sys/socket.h contains definitions for socket level options, described below. Options at other protocol levels vary in format and name; consult the appropriate entries in section 4 of the manual. .Pp Most socket-level options utilize an .Vt int argument for .Fa optval . For .Fn setsockopt , the argument should be non-zero to enable a boolean option, or zero if the option is to be disabled. .Dv SO_LINGER uses a .Vt "struct linger" argument, defined in .In sys/socket.h , which specifies the desired state of the option and the linger interval (see below). .Dv SO_SNDTIMEO and .Dv SO_RCVTIMEO use a .Vt "struct timeval" argument, defined in .In sys/time.h . .Pp The following options are recognized at the socket level. For protocol-specific options, see protocol manual pages, e.g. .Xr ip 4 or .Xr tcp 4 . Except as noted, each may be examined with .Fn getsockopt and set with .Fn setsockopt . .Bl -column SO_ACCEPTFILTER -offset indent .It Dv SO_DEBUG Ta "enables recording of debugging information" .It Dv SO_REUSEADDR Ta "enables local address reuse" .It Dv SO_REUSEPORT Ta "enables duplicate address and port bindings" +.It Dv SO_REUSEPORT_LB Ta "enables duplicate address and port bindings with load balancing" .It Dv SO_KEEPALIVE Ta "enables keep connections alive" .It Dv SO_DONTROUTE Ta "enables routing bypass for outgoing messages" .It Dv SO_LINGER Ta "linger on close if data present" .It Dv SO_BROADCAST Ta "enables permission to transmit broadcast messages" .It Dv SO_OOBINLINE Ta "enables reception of out-of-band data in band" .It Dv SO_SNDBUF Ta "set buffer size for output" .It Dv SO_RCVBUF Ta "set buffer size for input" .It Dv SO_SNDLOWAT Ta "set minimum count for output" .It Dv SO_RCVLOWAT Ta "set minimum count for input" .It Dv SO_SNDTIMEO Ta "set timeout value for output" .It Dv SO_RCVTIMEO Ta "set timeout value for input" .It Dv SO_ACCEPTFILTER Ta "set accept filter on listening socket" .It Dv SO_NOSIGPIPE Ta controls generation of .Dv SIGPIPE for the socket .It Dv SO_TIMESTAMP Ta "enables reception of a timestamp with datagrams" .It Dv SO_BINTIME Ta "enables reception of a timestamp with datagrams" .It Dv SO_ACCEPTCONN Ta "get listening status of the socket (get only)" .It Dv SO_TYPE Ta "get the type of the socket (get only)" .It Dv SO_PROTOCOL Ta "get the protocol number for the socket (get only)" .It Dv SO_PROTOTYPE Ta "SunOS alias for the Linux SO_PROTOCOL (get only)" .It Dv SO_ERROR Ta "get and clear error on the socket (get only)" .It Dv SO_SETFIB Ta "set the associated FIB (routing table) for the socket (set only)" .El .Pp The following options are recognized in .Fx : .Bl -column SO_LISTENINCQLEN -offset indent .It Dv SO_LABEL Ta "get MAC label of the socket (get only)" .It Dv SO_PEERLABEL Ta "get socket's peer's MAC label (get only)" .It Dv SO_LISTENQLIMIT Ta "get backlog limit of the socket (get only)" .It Dv SO_LISTENQLEN Ta "get complete queue length of the socket (get only)" .It Dv SO_LISTENINCQLEN Ta "get incomplete queue length of the socket (get only)" .It Dv SO_USER_COOKIE Ta "set the 'so_user_cookie' value for the socket (uint32_t, set only)" .It Dv SO_TS_CLOCK Ta "set specific format of timestamp returned by SO_TIMESTAMP" .It Dv SO_MAX_PACING_RATE Ta "set the maximum transmit rate in bytes per second for the socket" .El .Pp .Dv SO_DEBUG enables debugging in the underlying protocol modules. .Pp .Dv SO_REUSEADDR indicates that the rules used in validating addresses supplied in a .Xr bind 2 system call should allow reuse of local addresses. .Pp .Dv SO_REUSEPORT allows completely duplicate bindings by multiple processes if they all set .Dv SO_REUSEPORT before binding the port. This option permits multiple instances of a program to each receive UDP/IP multicast or broadcast datagrams destined for the bound port. +.Pp +.Dv SO_REUSEPORT_LB +allows completely duplicate bindings by multiple processes +if they all set +.Dv SO_REUSEPORT_LB +before binding the port. +Incoming TCP and UDP connections are distributed among the sharing +processes based on a hash function of local port number, foreign IP +address and port number. A maximum of 256 processes can share one socket. .Pp .Dv SO_KEEPALIVE enables the periodic transmission of messages on a connected socket. Should the connected party fail to respond to these messages, the connection is considered broken and processes using the socket are notified via a .Dv SIGPIPE signal when attempting to send data. .Pp .Dv SO_DONTROUTE indicates that outgoing messages should bypass the standard routing facilities. Instead, messages are directed to the appropriate network interface according to the network portion of the destination address. .Pp .Dv SO_LINGER controls the action taken when unsent messages are queued on socket and a .Xr close 2 is performed. If the socket promises reliable delivery of data and .Dv SO_LINGER is set, the system will block the process on the .Xr close 2 attempt until it is able to transmit the data or until it decides it is unable to deliver the information (a timeout period, termed the linger interval, is specified in seconds in the .Fn setsockopt system call when .Dv SO_LINGER is requested). If .Dv SO_LINGER is disabled and a .Xr close 2 is issued, the system will process the close in a manner that allows the process to continue as quickly as possible. .Pp The option .Dv SO_BROADCAST requests permission to send broadcast datagrams on the socket. Broadcast was a privileged operation in earlier versions of the system. .Pp With protocols that support out-of-band data, the .Dv SO_OOBINLINE option requests that out-of-band data be placed in the normal data input queue as received; it will then be accessible with .Xr recv 2 or .Xr read 2 calls without the .Dv MSG_OOB flag. Some protocols always behave as if this option is set. .Pp .Dv SO_SNDBUF and .Dv SO_RCVBUF are options to adjust the normal buffer sizes allocated for output and input buffers, respectively. The buffer size may be increased for high-volume connections, or may be decreased to limit the possible backlog of incoming data. The system places an absolute maximum on these values, which is accessible through the .Xr sysctl 3 MIB variable .Dq Li kern.ipc.maxsockbuf . .Pp .Dv SO_SNDLOWAT is an option to set the minimum count for output operations. Most output operations process all of the data supplied by the call, delivering data to the protocol for transmission and blocking as necessary for flow control. Nonblocking output operations will process as much data as permitted subject to flow control without blocking, but will process no data if flow control does not allow the smaller of the low water mark value or the entire request to be processed. A .Xr select 2 operation testing the ability to write to a socket will return true only if the low water mark amount could be processed. The default value for .Dv SO_SNDLOWAT is set to a convenient size for network efficiency, often 1024. .Pp .Dv SO_RCVLOWAT is an option to set the minimum count for input operations. In general, receive calls will block until any (non-zero) amount of data is received, then return with the smaller of the amount available or the amount requested. The default value for .Dv SO_RCVLOWAT is 1. If .Dv SO_RCVLOWAT is set to a larger value, blocking receive calls normally wait until they have received the smaller of the low water mark value or the requested amount. Receive calls may still return less than the low water mark if an error occurs, a signal is caught, or the type of data next in the receive queue is different from that which was returned. .Pp .Dv SO_SNDTIMEO is an option to set a timeout value for output operations. It accepts a .Vt "struct timeval" argument with the number of seconds and microseconds used to limit waits for output operations to complete. If a send operation has blocked for this much time, it returns with a partial count or with the error .Er EWOULDBLOCK if no data were sent. In the current implementation, this timer is restarted each time additional data are delivered to the protocol, implying that the limit applies to output portions ranging in size from the low water mark to the high water mark for output. .Pp .Dv SO_RCVTIMEO is an option to set a timeout value for input operations. It accepts a .Vt "struct timeval" argument with the number of seconds and microseconds used to limit waits for input operations to complete. In the current implementation, this timer is restarted each time additional data are received by the protocol, and thus the limit is in effect an inactivity timer. If a receive operation has been blocked for this much time without receiving additional data, it returns with a short count or with the error .Er EWOULDBLOCK if no data were received. .Pp .Dv SO_SETFIB can be used to over-ride the default FIB (routing table) for the given socket. The value must be from 0 to one less than the number returned from the sysctl .Em net.fibs . .Pp .Dv SO_USER_COOKIE can be used to set the uint32_t so_user_cookie field in the socket. The value is an uint32_t, and can be used in the kernel code that manipulates traffic related to the socket. The default value for the field is 0. As an example, the value can be used as the skipto target or pipe number in .Nm ipfw/dummynet . .Pp .Dv SO_ACCEPTFILTER places an .Xr accept_filter 9 on the socket, which will filter incoming connections on a listening stream socket before being presented for .Xr accept 2 . Once more, .Xr listen 2 must be called on the socket before trying to install the filter on it, or else the .Fn setsockopt system call will fail. .Bd -literal struct accept_filter_arg { char af_name[16]; char af_arg[256-16]; }; .Ed .Pp The .Fa optval argument should point to a .Fa struct accept_filter_arg that will select and configure the .Xr accept_filter 9 . The .Fa af_name argument should be filled with the name of the accept filter that the application wishes to place on the listening socket. The optional argument .Fa af_arg can be passed to the accept filter specified by .Fa af_name to provide additional configuration options at attach time. Passing in an .Fa optval of NULL will remove the filter. .Pp The .Dv SO_NOSIGPIPE option controls generation of the .Dv SIGPIPE signal normally sent when writing to a connected socket where the other end has been closed returns with the error .Er EPIPE . .Pp If the .Dv SO_TIMESTAMP or .Dv SO_BINTIME option is enabled on a .Dv SOCK_DGRAM socket, the .Xr recvmsg 2 call will return a timestamp corresponding to when the datagram was received. The .Va msg_control field in the .Vt msghdr structure points to a buffer that contains a .Vt cmsghdr structure followed by a .Vt "struct timeval" for .Dv SO_TIMESTAMP and .Vt "struct bintime" for .Dv SO_BINTIME . The .Vt cmsghdr fields have the following values for TIMESTAMP by default: .Bd -literal cmsg_len = CMSG_LEN(sizeof(struct timeval)); cmsg_level = SOL_SOCKET; cmsg_type = SCM_TIMESTAMP; .Ed .Pp and for .Dv SO_BINTIME : .Bd -literal cmsg_len = CMSG_LEN(sizeof(struct bintime)); cmsg_level = SOL_SOCKET; cmsg_type = SCM_BINTIME; .Ed .Pp Additional timestamp types are available by following .Dv SO_TIMESTAMP with .Dv SO_TS_CLOCK , which requests a specific timestamp format to be returned instead of .Dv SCM_TIMESTAMP when .Dv SO_TIMESTAMP is enabled. These .Dv SO_TS_CLOCK values are recognized in .Fx : .Bl -column SO_TS_CLOCK -offset indent .It Dv SO_TS_REALTIME_MICRO Ta "realtime (SCM_TIMESTAMP, struct timeval), default" .It Dv SO_TS_BINTIME Ta "realtime (SCM_BINTIME, struct bintime)" .It Dv SO_TS_REALTIME Ta "realtime (SCM_REALTIME, struct timespec)" .It Dv SO_TS_MONOTONIC Ta "monotonic time (SCM_MONOTONIC, struct timespec)" .El .Pp .Dv SO_ACCEPTCONN , .Dv SO_TYPE , .Dv SO_PROTOCOL (and its alias .Dv SO_PROTOTYPE ) and .Dv SO_ERROR are options used only with .Fn getsockopt . .Dv SO_ACCEPTCONN returns whether the socket is currently accepting connections, that is, whether or not the .Xr listen 2 system call was invoked on the socket. .Dv SO_TYPE returns the type of the socket, such as .Dv SOCK_STREAM ; it is useful for servers that inherit sockets on startup. .Dv SO_PROTOCOL returns the protocol number for the socket, for .Dv AF_INET and .Dv AF_INET6 address families. .Dv SO_ERROR returns any pending error on the socket and clears the error status. It may be used to check for asynchronous errors on connected datagram sockets or for other asynchronous errors. .Pp Finally, .Dv SO_LABEL returns the MAC label of the socket. .Dv SO_PEERLABEL returns the MAC label of the socket's peer. Note that your kernel must be compiled with MAC support. See .Xr mac 3 for more information. .Dv SO_LISTENQLIMIT returns the maximal number of queued connections, as set by .Xr listen 2 . .Dv SO_LISTENQLEN returns the number of unaccepted complete connections. .Dv SO_LISTENINCQLEN returns the number of unaccepted incomplete connections. .Pp .Dv SO_MAX_PACING_RATE instruct the socket and underlying network adapter layers to limit the transfer rate to the given unsigned 32-bit value in bytes per second. .Sh RETURN VALUES .Rv -std .Sh ERRORS The call succeeds unless: .Bl -tag -width Er .It Bq Er EBADF The argument .Fa s is not a valid descriptor. .It Bq Er ENOTSOCK The argument .Fa s is a file, not a socket. .It Bq Er ENOPROTOOPT The option is unknown at the level indicated. .It Bq Er EFAULT The address pointed to by .Fa optval is not in a valid part of the process address space. For .Fn getsockopt , this error may also be returned if .Fa optlen is not in a valid part of the process address space. .It Bq Er EINVAL Installing an .Xr accept_filter 9 on a non-listening socket was attempted. .It Bq Er ENOMEM A memory allocation failed that was required to service the request. .El .Sh SEE ALSO .Xr ioctl 2 , .Xr listen 2 , .Xr recvmsg 2 , .Xr socket 2 , .Xr getprotoent 3 , .Xr mac 3 , .Xr sysctl 3 , .Xr ip 4 , .Xr ip6 4 , .Xr sctp 4 , .Xr tcp 4 , .Xr protocols 5 , .Xr sysctl 8 , .Xr accept_filter 9 , .Xr bintime 9 .Sh HISTORY The .Fn getsockopt and .Fn setsockopt system calls appeared in .Bx 4.2 . .Sh BUGS Several of the socket options should be handled at lower levels of the system. Index: head/sys/kern/uipc_debug.c =================================================================== --- head/sys/kern/uipc_debug.c (revision 334718) +++ head/sys/kern/uipc_debug.c (revision 334719) @@ -1,536 +1,540 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2007 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. */ /* * Debugger routines relating to sockets, protocols, etc, for use in DDB. */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include #include #include #include #include #include #ifdef DDB #include static void db_print_sotype(short so_type) { switch (so_type) { case SOCK_STREAM: db_printf("SOCK_STREAM"); break; case SOCK_DGRAM: db_printf("SOCK_DGRAM"); break; case SOCK_RAW: db_printf("SOCK_RAW"); break; case SOCK_RDM: db_printf("SOCK_RDM"); break; case SOCK_SEQPACKET: db_printf("SOCK_SEQPACKET"); break; default: db_printf("unknown"); break; } } static void -db_print_sooptions(short so_options) +db_print_sooptions(int so_options) { int comma; comma = 0; if (so_options & SO_DEBUG) { db_printf("%sSO_DEBUG", comma ? ", " : ""); comma = 1; } if (so_options & SO_ACCEPTCONN) { db_printf("%sSO_ACCEPTCONN", comma ? ", " : ""); comma = 1; } if (so_options & SO_REUSEADDR) { db_printf("%sSO_REUSEADDR", comma ? ", " : ""); comma = 1; } if (so_options & SO_KEEPALIVE) { db_printf("%sSO_KEEPALIVE", comma ? ", " : ""); comma = 1; } if (so_options & SO_DONTROUTE) { db_printf("%sSO_DONTROUTE", comma ? ", " : ""); comma = 1; } if (so_options & SO_BROADCAST) { db_printf("%sSO_BROADCAST", comma ? ", " : ""); comma = 1; } if (so_options & SO_USELOOPBACK) { db_printf("%sSO_USELOOPBACK", comma ? ", " : ""); comma = 1; } if (so_options & SO_LINGER) { db_printf("%sSO_LINGER", comma ? ", " : ""); comma = 1; } if (so_options & SO_OOBINLINE) { db_printf("%sSO_OOBINLINE", comma ? ", " : ""); comma = 1; } if (so_options & SO_REUSEPORT) { db_printf("%sSO_REUSEPORT", comma ? ", " : ""); + comma = 1; + } + if (so_options & SO_REUSEPORT_LB) { + db_printf("%sSO_REUSEPORT_LB", comma ? ", " : ""); comma = 1; } if (so_options & SO_TIMESTAMP) { db_printf("%sSO_TIMESTAMP", comma ? ", " : ""); comma = 1; } if (so_options & SO_NOSIGPIPE) { db_printf("%sSO_NOSIGPIPE", comma ? ", " : ""); comma = 1; } if (so_options & SO_ACCEPTFILTER) { db_printf("%sSO_ACCEPTFILTER", comma ? ", " : ""); comma = 1; } if (so_options & SO_BINTIME) { db_printf("%sSO_BINTIME", comma ? ", " : ""); comma = 1; } if (so_options & SO_NO_OFFLOAD) { db_printf("%sSO_NO_OFFLOAD", comma ? ", " : ""); comma = 1; } if (so_options & SO_NO_DDP) { db_printf("%sSO_NO_DDP", comma ? ", " : ""); comma = 1; } } static void db_print_sostate(short so_state) { int comma; comma = 0; if (so_state & SS_NOFDREF) { db_printf("%sSS_NOFDREF", comma ? ", " : ""); comma = 1; } if (so_state & SS_ISCONNECTED) { db_printf("%sSS_ISCONNECTED", comma ? ", " : ""); comma = 1; } if (so_state & SS_ISCONNECTING) { db_printf("%sSS_ISCONNECTING", comma ? ", " : ""); comma = 1; } if (so_state & SS_ISDISCONNECTING) { db_printf("%sSS_ISDISCONNECTING", comma ? ", " : ""); comma = 1; } if (so_state & SS_NBIO) { db_printf("%sSS_NBIO", comma ? ", " : ""); comma = 1; } if (so_state & SS_ASYNC) { db_printf("%sSS_ASYNC", comma ? ", " : ""); comma = 1; } if (so_state & SS_ISCONFIRMING) { db_printf("%sSS_ISCONFIRMING", comma ? ", " : ""); comma = 1; } if (so_state & SS_PROTOREF) { db_printf("%sSS_PROTOREF", comma ? ", " : ""); comma = 1; } } static void db_print_soqstate(int so_qstate) { int comma; comma = 0; if (so_qstate & SQ_INCOMP) { db_printf("%sSQ_INCOMP", comma ? ", " : ""); comma = 1; } if (so_qstate & SQ_COMP) { db_printf("%sSQ_COMP", comma ? ", " : ""); comma = 1; } } static void db_print_sbstate(short sb_state) { int comma; comma = 0; if (sb_state & SBS_CANTSENDMORE) { db_printf("%sSBS_CANTSENDMORE", comma ? ", " : ""); comma = 1; } if (sb_state & SBS_CANTRCVMORE) { db_printf("%sSBS_CANTRCVMORE", comma ? ", " : ""); comma = 1; } if (sb_state & SBS_RCVATMARK) { db_printf("%sSBS_RCVATMARK", comma ? ", " : ""); comma = 1; } } static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_domain(struct domain *d, const char *domain_name, int indent) { db_print_indent(indent); db_printf("%s at %p\n", domain_name, d); indent += 2; db_print_indent(indent); db_printf("dom_family: %d ", d->dom_family); db_printf("dom_name: %s\n", d->dom_name); db_print_indent(indent); db_printf("dom_init: %p ", d->dom_init); db_printf("dom_externalize: %p ", d->dom_externalize); db_printf("dom_dispose: %p\n", d->dom_dispose); db_print_indent(indent); db_printf("dom_protosw: %p ", d->dom_protosw); db_printf("dom_next: %p\n", d->dom_next); db_print_indent(indent); db_printf("dom_rtattach: %p ", d->dom_rtattach); db_print_indent(indent); db_printf("dom_ifattach: %p ", d->dom_ifattach); db_printf("dom_ifdetach: %p\n", d->dom_ifdetach); } static void db_print_prflags(short pr_flags) { int comma; comma = 0; if (pr_flags & PR_ATOMIC) { db_printf("%sPR_ATOMIC", comma ? ", " : ""); comma = 1; } if (pr_flags & PR_ADDR) { db_printf("%sPR_ADDR", comma ? ", " : ""); comma = 1; } if (pr_flags & PR_CONNREQUIRED) { db_printf("%sPR_CONNREQUIRED", comma ? ", " : ""); comma = 1; } if (pr_flags & PR_WANTRCVD) { db_printf("%sPR_WANTRCVD", comma ? ", " : ""); comma = 1; } if (pr_flags & PR_RIGHTS) { db_printf("%sPR_RIGHTS", comma ? ", " : ""); comma = 1; } if (pr_flags & PR_IMPLOPCL) { db_printf("%sPR_IMPLOPCL", comma ? ", " : ""); comma = 1; } if (pr_flags & PR_LASTHDR) { db_printf("%sPR_LASTHDR", comma ? ", " : ""); comma = 1; } } static void db_print_protosw(struct protosw *pr, const char *prname, int indent) { db_print_indent(indent); db_printf("%s at %p\n", prname, pr); indent += 2; db_print_indent(indent); db_printf("pr_type: %d ", pr->pr_type); db_printf("pr_domain: %p\n", pr->pr_domain); if (pr->pr_domain != NULL) db_print_domain(pr->pr_domain, "pr_domain", indent); db_print_indent(indent); db_printf("pr_protocol: %d\n", pr->pr_protocol); db_print_indent(indent); db_printf("pr_flags: %d (", pr->pr_flags); db_print_prflags(pr->pr_flags); db_printf(")\n"); db_print_indent(indent); db_printf("pr_input: %p ", pr->pr_input); db_printf("pr_output: %p ", pr->pr_output); db_printf("pr_ctlinput: %p\n", pr->pr_ctlinput); db_print_indent(indent); db_printf("pr_ctloutput: %p ", pr->pr_ctloutput); db_printf("pr_init: %p\n", pr->pr_init); db_print_indent(indent); db_printf("pr_fasttimo: %p ", pr->pr_fasttimo); db_printf("pr_slowtimo: %p ", pr->pr_slowtimo); db_printf("pr_drain: %p\n", pr->pr_drain); } static void db_print_sbflags(short sb_flags) { int comma; comma = 0; if (sb_flags & SB_WAIT) { db_printf("%sSB_WAIT", comma ? ", " : ""); comma = 1; } if (sb_flags & SB_SEL) { db_printf("%sSB_SEL", comma ? ", " : ""); comma = 1; } if (sb_flags & SB_ASYNC) { db_printf("%sSB_ASYNC", comma ? ", " : ""); comma = 1; } if (sb_flags & SB_UPCALL) { db_printf("%sSB_UPCALL", comma ? ", " : ""); comma = 1; } if (sb_flags & SB_NOINTR) { db_printf("%sSB_NOINTR", comma ? ", " : ""); comma = 1; } if (sb_flags & SB_AIO) { db_printf("%sSB_AIO", comma ? ", " : ""); comma = 1; } if (sb_flags & SB_KNOTE) { db_printf("%sSB_KNOTE", comma ? ", " : ""); comma = 1; } if (sb_flags & SB_AUTOSIZE) { db_printf("%sSB_AUTOSIZE", comma ? ", " : ""); comma = 1; } } static void db_print_sockbuf(struct sockbuf *sb, const char *sockbufname, int indent) { db_print_indent(indent); db_printf("%s at %p\n", sockbufname, sb); indent += 2; db_print_indent(indent); db_printf("sb_state: 0x%x (", sb->sb_state); db_print_sbstate(sb->sb_state); db_printf(")\n"); db_print_indent(indent); db_printf("sb_mb: %p ", sb->sb_mb); db_printf("sb_mbtail: %p ", sb->sb_mbtail); db_printf("sb_lastrecord: %p\n", sb->sb_lastrecord); db_print_indent(indent); db_printf("sb_sndptr: %p ", sb->sb_sndptr); db_printf("sb_sndptroff: %u\n", sb->sb_sndptroff); db_print_indent(indent); db_printf("sb_acc: %u ", sb->sb_acc); db_printf("sb_ccc: %u ", sb->sb_ccc); db_printf("sb_hiwat: %u ", sb->sb_hiwat); db_printf("sb_mbcnt: %u ", sb->sb_mbcnt); db_printf("sb_mbmax: %u\n", sb->sb_mbmax); db_print_indent(indent); db_printf("sb_mcnt: %u ", sb->sb_mcnt); db_printf("sb_ccnt: %u ", sb->sb_ccnt); db_printf("sb_ctl: %u ", sb->sb_ctl); db_printf("sb_lowat: %d ", sb->sb_lowat); db_printf("sb_timeo: %jd\n", sb->sb_timeo); db_print_indent(indent); db_printf("sb_flags: 0x%x (", sb->sb_flags); db_print_sbflags(sb->sb_flags); db_printf(")\n"); db_print_indent(indent); db_printf("sb_aiojobq first: %p\n", TAILQ_FIRST(&sb->sb_aiojobq)); } static void db_print_socket(struct socket *so, const char *socketname, int indent) { db_print_indent(indent); db_printf("%s at %p\n", socketname, so); indent += 2; db_print_indent(indent); db_printf("so_count: %d ", so->so_count); db_printf("so_type: %d (", so->so_type); db_print_sotype(so->so_type); db_printf(")\n"); db_print_indent(indent); db_printf("so_options: 0x%x (", so->so_options); db_print_sooptions(so->so_options); db_printf(")\n"); db_print_indent(indent); db_printf("so_linger: %d ", so->so_linger); db_printf("so_state: 0x%x (", so->so_state); db_print_sostate(so->so_state); db_printf(")\n"); db_print_indent(indent); db_printf("so_pcb: %p ", so->so_pcb); db_printf("so_proto: %p\n", so->so_proto); if (so->so_proto != NULL) db_print_protosw(so->so_proto, "so_proto", indent); db_print_indent(indent); if (so->so_options & SO_ACCEPTCONN) { db_printf("sol_incomp first: %p ", TAILQ_FIRST(&so->sol_incomp)); db_printf("sol_comp first: %p\n", TAILQ_FIRST(&so->sol_comp)); db_printf("sol_qlen: %d ", so->sol_qlen); db_printf("sol_incqlen: %d ", so->sol_incqlen); db_printf("sol_qlimit: %d ", so->sol_qlimit); } else { db_printf("so_qstate: 0x%x (", so->so_qstate); db_print_soqstate(so->so_qstate); db_printf(") "); db_printf("so_listen: %p ", so->so_listen); /* so_list skipped */ db_printf("so_timeo: %d ", so->so_timeo); db_printf("so_error: %d\n", so->so_error); db_print_indent(indent); db_printf("so_sigio: %p ", so->so_sigio); db_printf("so_oobmark: %lu\n", so->so_oobmark); db_print_sockbuf(&so->so_rcv, "so_rcv", indent); db_print_sockbuf(&so->so_snd, "so_snd", indent); } } DB_SHOW_COMMAND(socket, db_show_socket) { struct socket *so; if (!have_addr) { db_printf("usage: show socket \n"); return; } so = (struct socket *)addr; db_print_socket(so, "socket", 0); } DB_SHOW_COMMAND(sockbuf, db_show_sockbuf) { struct sockbuf *sb; if (!have_addr) { db_printf("usage: show sockbuf \n"); return; } sb = (struct sockbuf *)addr; db_print_sockbuf(sb, "sockbuf", 0); } DB_SHOW_COMMAND(protosw, db_show_protosw) { struct protosw *pr; if (!have_addr) { db_printf("usage: show protosw \n"); return; } pr = (struct protosw *)addr; db_print_protosw(pr, "protosw", 0); } DB_SHOW_COMMAND(domain, db_show_domain) { struct domain *d; if (!have_addr) { db_printf("usage: show protosw \n"); return; } d = (struct domain *)addr; db_print_domain(d, "domain", 0); } #endif Index: head/sys/kern/uipc_socket.c =================================================================== --- head/sys/kern/uipc_socket.c (revision 334718) +++ head/sys/kern/uipc_socket.c (revision 334719) @@ -1,4126 +1,4128 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1988, 1990, 1993 * The Regents of the University of California. * Copyright (c) 2004 The FreeBSD Foundation * Copyright (c) 2004-2008 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. * 3. 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. * * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94 */ /* * Comments on the socket life cycle: * * soalloc() sets of socket layer state for a socket, called only by * socreate() and sonewconn(). Socket layer private. * * sodealloc() tears down socket layer state for a socket, called only by * sofree() and sonewconn(). Socket layer private. * * pru_attach() associates protocol layer state with an allocated socket; * called only once, may fail, aborting socket allocation. This is called * from socreate() and sonewconn(). Socket layer private. * * pru_detach() disassociates protocol layer state from an attached socket, * and will be called exactly once for sockets in which pru_attach() has * been successfully called. If pru_attach() returned an error, * pru_detach() will not be called. Socket layer private. * * pru_abort() and pru_close() notify the protocol layer that the last * consumer of a socket is starting to tear down the socket, and that the * protocol should terminate the connection. Historically, pru_abort() also * detached protocol state from the socket state, but this is no longer the * case. * * socreate() creates a socket and attaches protocol state. This is a public * interface that may be used by socket layer consumers to create new * sockets. * * sonewconn() creates a socket and attaches protocol state. This is a * public interface that may be used by protocols to create new sockets when * a new connection is received and will be available for accept() on a * listen socket. * * soclose() destroys a socket after possibly waiting for it to disconnect. * This is a public interface that socket consumers should use to close and * release a socket when done with it. * * soabort() destroys a socket without waiting for it to disconnect (used * only for incoming connections that are already partially or fully * connected). This is used internally by the socket layer when clearing * listen socket queues (due to overflow or close on the listen socket), but * is also a public interface protocols may use to abort connections in * their incomplete listen queues should they no longer be required. Sockets * placed in completed connection listen queues should not be aborted for * reasons described in the comment above the soclose() implementation. This * is not a general purpose close routine, and except in the specific * circumstances described here, should not be used. * * sofree() will free a socket and its protocol state if all references on * the socket have been released, and is the public interface to attempt to * free a socket when a reference is removed. This is a socket layer private * interface. * * NOTE: In addition to socreate() and soclose(), which provide a single * socket reference to the consumer to be managed as required, there are two * calls to explicitly manage socket references, soref(), and sorele(). * Currently, these are generally required only when transitioning a socket * from a listen queue to a file descriptor, in order to prevent garbage * collection of the socket at an untimely moment. For a number of reasons, * these interfaces are not preferred, and should be avoided. * * NOTE: With regard to VNETs the general rule is that callers do not set * curvnet. Exceptions to this rule include soabort(), sodisconnect(), * sofree() (and with that sorele(), sotryfree()), as well as sonewconn() * and sorflush(), which are usually called from a pre-set VNET context. * sopoll() currently does not need a VNET context to be set. */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_sctp.h" #include #include #include #include #include #include #include #include #include #include #include /* for struct knote */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef COMPAT_FREEBSD32 #include #include #include #endif static int soreceive_rcvoob(struct socket *so, struct uio *uio, int flags); static void so_rdknl_lock(void *); static void so_rdknl_unlock(void *); static void so_rdknl_assert_locked(void *); static void so_rdknl_assert_unlocked(void *); static void so_wrknl_lock(void *); static void so_wrknl_unlock(void *); static void so_wrknl_assert_locked(void *); static void so_wrknl_assert_unlocked(void *); static void filt_sordetach(struct knote *kn); static int filt_soread(struct knote *kn, long hint); static void filt_sowdetach(struct knote *kn); static int filt_sowrite(struct knote *kn, long hint); static int filt_soempty(struct knote *kn, long hint); static int inline hhook_run_socket(struct socket *so, void *hctx, int32_t h_id); fo_kqfilter_t soo_kqfilter; static struct filterops soread_filtops = { .f_isfd = 1, .f_detach = filt_sordetach, .f_event = filt_soread, }; static struct filterops sowrite_filtops = { .f_isfd = 1, .f_detach = filt_sowdetach, .f_event = filt_sowrite, }; static struct filterops soempty_filtops = { .f_isfd = 1, .f_detach = filt_sowdetach, .f_event = filt_soempty, }; so_gen_t so_gencnt; /* generation count for sockets */ MALLOC_DEFINE(M_SONAME, "soname", "socket name"); MALLOC_DEFINE(M_PCB, "pcb", "protocol control block"); #define VNET_SO_ASSERT(so) \ VNET_ASSERT(curvnet != NULL, \ ("%s:%d curvnet is NULL, so=%p", __func__, __LINE__, (so))); VNET_DEFINE(struct hhook_head *, socket_hhh[HHOOK_SOCKET_LAST + 1]); #define V_socket_hhh VNET(socket_hhh) /* * Limit on the number of connections in the listen queue waiting * for accept(2). * NB: The original sysctl somaxconn is still available but hidden * to prevent confusion about the actual purpose of this number. */ static u_int somaxconn = SOMAXCONN; static int sysctl_somaxconn(SYSCTL_HANDLER_ARGS) { int error; int val; val = somaxconn; error = sysctl_handle_int(oidp, &val, 0, req); if (error || !req->newptr ) return (error); /* * The purpose of the UINT_MAX / 3 limit, is so that the formula * 3 * so_qlimit / 2 * below, will not overflow. */ if (val < 1 || val > UINT_MAX / 3) return (EINVAL); somaxconn = val; return (0); } SYSCTL_PROC(_kern_ipc, OID_AUTO, soacceptqueue, CTLTYPE_UINT | CTLFLAG_RW, 0, sizeof(int), sysctl_somaxconn, "I", "Maximum listen socket pending connection accept queue size"); SYSCTL_PROC(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_SKIP, 0, sizeof(int), sysctl_somaxconn, "I", "Maximum listen socket pending connection accept queue size (compat)"); static int numopensockets; SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD, &numopensockets, 0, "Number of open sockets"); /* * accept_mtx locks down per-socket fields relating to accept queues. See * socketvar.h for an annotation of the protected fields of struct socket. */ struct mtx accept_mtx; MTX_SYSINIT(accept_mtx, &accept_mtx, "accept", MTX_DEF); /* * so_global_mtx protects so_gencnt, numopensockets, and the per-socket * so_gencnt field. */ static struct mtx so_global_mtx; MTX_SYSINIT(so_global_mtx, &so_global_mtx, "so_glabel", MTX_DEF); /* * General IPC sysctl name space, used by sockets and a variety of other IPC * types. */ SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC"); /* * Initialize the socket subsystem and set up the socket * memory allocator. */ static uma_zone_t socket_zone; int maxsockets; static void socket_zone_change(void *tag) { maxsockets = uma_zone_set_max(socket_zone, maxsockets); } static void socket_hhook_register(int subtype) { if (hhook_head_register(HHOOK_TYPE_SOCKET, subtype, &V_socket_hhh[subtype], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register hook\n", __func__); } static void socket_hhook_deregister(int subtype) { if (hhook_head_deregister(V_socket_hhh[subtype]) != 0) printf("%s: WARNING: unable to deregister hook\n", __func__); } static void socket_init(void *tag) { socket_zone = uma_zcreate("socket", sizeof(struct socket), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); maxsockets = uma_zone_set_max(socket_zone, maxsockets); uma_zone_set_warning(socket_zone, "kern.ipc.maxsockets limit reached"); EVENTHANDLER_REGISTER(maxsockets_change, socket_zone_change, NULL, EVENTHANDLER_PRI_FIRST); } SYSINIT(socket, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, socket_init, NULL); static void socket_vnet_init(const void *unused __unused) { int i; /* We expect a contiguous range */ for (i = 0; i <= HHOOK_SOCKET_LAST; i++) socket_hhook_register(i); } VNET_SYSINIT(socket_vnet_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, socket_vnet_init, NULL); static void socket_vnet_uninit(const void *unused __unused) { int i; for (i = 0; i <= HHOOK_SOCKET_LAST; i++) socket_hhook_deregister(i); } VNET_SYSUNINIT(socket_vnet_uninit, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, socket_vnet_uninit, NULL); /* * Initialise maxsockets. This SYSINIT must be run after * tunable_mbinit(). */ static void init_maxsockets(void *ignored) { TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets); maxsockets = imax(maxsockets, maxfiles); } SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL); /* * Sysctl to get and set the maximum global sockets limit. Notify protocols * of the change so that they can update their dependent limits as required. */ static int sysctl_maxsockets(SYSCTL_HANDLER_ARGS) { int error, newmaxsockets; newmaxsockets = maxsockets; error = sysctl_handle_int(oidp, &newmaxsockets, 0, req); if (error == 0 && req->newptr) { if (newmaxsockets > maxsockets && newmaxsockets <= maxfiles) { maxsockets = newmaxsockets; EVENTHANDLER_INVOKE(maxsockets_change); } else error = EINVAL; } return (error); } SYSCTL_PROC(_kern_ipc, OID_AUTO, maxsockets, CTLTYPE_INT|CTLFLAG_RW, &maxsockets, 0, sysctl_maxsockets, "IU", "Maximum number of sockets available"); /* * Socket operation routines. These routines are called by the routines in * sys_socket.c or from a system process, and implement the semantics of * socket operations by switching out to the protocol specific routines. */ /* * Get a socket structure from our zone, and initialize it. Note that it * would probably be better to allocate socket and PCB at the same time, but * I'm not convinced that all the protocols can be easily modified to do * this. * * soalloc() returns a socket with a ref count of 0. */ static struct socket * soalloc(struct vnet *vnet) { struct socket *so; so = uma_zalloc(socket_zone, M_NOWAIT | M_ZERO); if (so == NULL) return (NULL); #ifdef MAC if (mac_socket_init(so, M_NOWAIT) != 0) { uma_zfree(socket_zone, so); return (NULL); } #endif if (khelp_init_osd(HELPER_CLASS_SOCKET, &so->osd)) { uma_zfree(socket_zone, so); return (NULL); } /* * The socket locking protocol allows to lock 2 sockets at a time, * however, the first one must be a listening socket. WITNESS lacks * a feature to change class of an existing lock, so we use DUPOK. */ mtx_init(&so->so_lock, "socket", NULL, MTX_DEF | MTX_DUPOK); SOCKBUF_LOCK_INIT(&so->so_snd, "so_snd"); SOCKBUF_LOCK_INIT(&so->so_rcv, "so_rcv"); so->so_rcv.sb_sel = &so->so_rdsel; so->so_snd.sb_sel = &so->so_wrsel; sx_init(&so->so_snd.sb_sx, "so_snd_sx"); sx_init(&so->so_rcv.sb_sx, "so_rcv_sx"); TAILQ_INIT(&so->so_snd.sb_aiojobq); TAILQ_INIT(&so->so_rcv.sb_aiojobq); TASK_INIT(&so->so_snd.sb_aiotask, 0, soaio_snd, so); TASK_INIT(&so->so_rcv.sb_aiotask, 0, soaio_rcv, so); #ifdef VIMAGE VNET_ASSERT(vnet != NULL, ("%s:%d vnet is NULL, so=%p", __func__, __LINE__, so)); so->so_vnet = vnet; #endif /* We shouldn't need the so_global_mtx */ if (hhook_run_socket(so, NULL, HHOOK_SOCKET_CREATE)) { /* Do we need more comprehensive error returns? */ uma_zfree(socket_zone, so); return (NULL); } mtx_lock(&so_global_mtx); so->so_gencnt = ++so_gencnt; ++numopensockets; #ifdef VIMAGE vnet->vnet_sockcnt++; #endif mtx_unlock(&so_global_mtx); return (so); } /* * Free the storage associated with a socket at the socket layer, tear down * locks, labels, etc. All protocol state is assumed already to have been * torn down (and possibly never set up) by the caller. */ static void sodealloc(struct socket *so) { KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count)); KASSERT(so->so_pcb == NULL, ("sodealloc(): so_pcb != NULL")); mtx_lock(&so_global_mtx); so->so_gencnt = ++so_gencnt; --numopensockets; /* Could be below, but faster here. */ #ifdef VIMAGE VNET_ASSERT(so->so_vnet != NULL, ("%s:%d so_vnet is NULL, so=%p", __func__, __LINE__, so)); so->so_vnet->vnet_sockcnt--; #endif mtx_unlock(&so_global_mtx); #ifdef MAC mac_socket_destroy(so); #endif hhook_run_socket(so, NULL, HHOOK_SOCKET_CLOSE); crfree(so->so_cred); khelp_destroy_osd(&so->osd); if (SOLISTENING(so)) { if (so->sol_accept_filter != NULL) accept_filt_setopt(so, NULL); } else { if (so->so_rcv.sb_hiwat) (void)chgsbsize(so->so_cred->cr_uidinfo, &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY); if (so->so_snd.sb_hiwat) (void)chgsbsize(so->so_cred->cr_uidinfo, &so->so_snd.sb_hiwat, 0, RLIM_INFINITY); sx_destroy(&so->so_snd.sb_sx); sx_destroy(&so->so_rcv.sb_sx); SOCKBUF_LOCK_DESTROY(&so->so_snd); SOCKBUF_LOCK_DESTROY(&so->so_rcv); } mtx_destroy(&so->so_lock); uma_zfree(socket_zone, so); } /* * socreate returns a socket with a ref count of 1. The socket should be * closed with soclose(). */ int socreate(int dom, struct socket **aso, int type, int proto, struct ucred *cred, struct thread *td) { struct protosw *prp; struct socket *so; int error; if (proto) prp = pffindproto(dom, proto, type); else prp = pffindtype(dom, type); if (prp == NULL) { /* No support for domain. */ if (pffinddomain(dom) == NULL) return (EAFNOSUPPORT); /* No support for socket type. */ if (proto == 0 && type != 0) return (EPROTOTYPE); return (EPROTONOSUPPORT); } if (prp->pr_usrreqs->pru_attach == NULL || prp->pr_usrreqs->pru_attach == pru_attach_notsupp) return (EPROTONOSUPPORT); if (prison_check_af(cred, prp->pr_domain->dom_family) != 0) return (EPROTONOSUPPORT); if (prp->pr_type != type) return (EPROTOTYPE); so = soalloc(CRED_TO_VNET(cred)); if (so == NULL) return (ENOBUFS); so->so_type = type; so->so_cred = crhold(cred); if ((prp->pr_domain->dom_family == PF_INET) || (prp->pr_domain->dom_family == PF_INET6) || (prp->pr_domain->dom_family == PF_ROUTE)) so->so_fibnum = td->td_proc->p_fibnum; else so->so_fibnum = 0; so->so_proto = prp; #ifdef MAC mac_socket_create(cred, so); #endif knlist_init(&so->so_rdsel.si_note, so, so_rdknl_lock, so_rdknl_unlock, so_rdknl_assert_locked, so_rdknl_assert_unlocked); knlist_init(&so->so_wrsel.si_note, so, so_wrknl_lock, so_wrknl_unlock, so_wrknl_assert_locked, so_wrknl_assert_unlocked); /* * Auto-sizing of socket buffers is managed by the protocols and * the appropriate flags must be set in the pru_attach function. */ CURVNET_SET(so->so_vnet); error = (*prp->pr_usrreqs->pru_attach)(so, proto, td); CURVNET_RESTORE(); if (error) { sodealloc(so); return (error); } soref(so); *aso = so; return (0); } #ifdef REGRESSION static int regression_sonewconn_earlytest = 1; SYSCTL_INT(_regression, OID_AUTO, sonewconn_earlytest, CTLFLAG_RW, ®ression_sonewconn_earlytest, 0, "Perform early sonewconn limit test"); #endif /* * When an attempt at a new connection is noted on a socket which accepts * connections, sonewconn is called. If the connection is possible (subject * to space constraints, etc.) then we allocate a new structure, properly * linked into the data structure of the original socket, and return this. * Connstatus may be 0, or SS_ISCONFIRMING, or SS_ISCONNECTED. * * Note: the ref count on the socket is 0 on return. */ struct socket * sonewconn(struct socket *head, int connstatus) { static struct timeval lastover; static struct timeval overinterval = { 60, 0 }; static int overcount; struct socket *so; u_int over; SOLISTEN_LOCK(head); over = (head->sol_qlen > 3 * head->sol_qlimit / 2); SOLISTEN_UNLOCK(head); #ifdef REGRESSION if (regression_sonewconn_earlytest && over) { #else if (over) { #endif overcount++; if (ratecheck(&lastover, &overinterval)) { log(LOG_DEBUG, "%s: pcb %p: Listen queue overflow: " "%i already in queue awaiting acceptance " "(%d occurrences)\n", __func__, head->so_pcb, head->sol_qlen, overcount); overcount = 0; } return (NULL); } VNET_ASSERT(head->so_vnet != NULL, ("%s: so %p vnet is NULL", __func__, head)); so = soalloc(head->so_vnet); if (so == NULL) { log(LOG_DEBUG, "%s: pcb %p: New socket allocation failure: " "limit reached or out of memory\n", __func__, head->so_pcb); return (NULL); } so->so_listen = head; so->so_type = head->so_type; so->so_linger = head->so_linger; so->so_state = head->so_state | SS_NOFDREF; so->so_fibnum = head->so_fibnum; so->so_proto = head->so_proto; so->so_cred = crhold(head->so_cred); #ifdef MAC mac_socket_newconn(head, so); #endif knlist_init(&so->so_rdsel.si_note, so, so_rdknl_lock, so_rdknl_unlock, so_rdknl_assert_locked, so_rdknl_assert_unlocked); knlist_init(&so->so_wrsel.si_note, so, so_wrknl_lock, so_wrknl_unlock, so_wrknl_assert_locked, so_wrknl_assert_unlocked); VNET_SO_ASSERT(head); if (soreserve(so, head->sol_sbsnd_hiwat, head->sol_sbrcv_hiwat)) { sodealloc(so); log(LOG_DEBUG, "%s: pcb %p: soreserve() failed\n", __func__, head->so_pcb); return (NULL); } if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) { sodealloc(so); log(LOG_DEBUG, "%s: pcb %p: pru_attach() failed\n", __func__, head->so_pcb); return (NULL); } so->so_rcv.sb_lowat = head->sol_sbrcv_lowat; so->so_snd.sb_lowat = head->sol_sbsnd_lowat; so->so_rcv.sb_timeo = head->sol_sbrcv_timeo; so->so_snd.sb_timeo = head->sol_sbsnd_timeo; so->so_rcv.sb_flags |= head->sol_sbrcv_flags & SB_AUTOSIZE; so->so_snd.sb_flags |= head->sol_sbsnd_flags & SB_AUTOSIZE; SOLISTEN_LOCK(head); if (head->sol_accept_filter != NULL) connstatus = 0; so->so_state |= connstatus; so->so_options = head->so_options & ~SO_ACCEPTCONN; soref(head); /* A socket on (in)complete queue refs head. */ if (connstatus) { TAILQ_INSERT_TAIL(&head->sol_comp, so, so_list); so->so_qstate = SQ_COMP; head->sol_qlen++; solisten_wakeup(head); /* unlocks */ } else { /* * Keep removing sockets from the head until there's room for * us to insert on the tail. In pre-locking revisions, this * was a simple if(), but as we could be racing with other * threads and soabort() requires dropping locks, we must * loop waiting for the condition to be true. */ while (head->sol_incqlen > head->sol_qlimit) { struct socket *sp; sp = TAILQ_FIRST(&head->sol_incomp); TAILQ_REMOVE(&head->sol_incomp, sp, so_list); head->sol_incqlen--; SOCK_LOCK(sp); sp->so_qstate = SQ_NONE; sp->so_listen = NULL; SOCK_UNLOCK(sp); sorele(head); /* does SOLISTEN_UNLOCK, head stays */ soabort(sp); SOLISTEN_LOCK(head); } TAILQ_INSERT_TAIL(&head->sol_incomp, so, so_list); so->so_qstate = SQ_INCOMP; head->sol_incqlen++; SOLISTEN_UNLOCK(head); } return (so); } #ifdef SCTP /* * Socket part of sctp_peeloff(). Detach a new socket from an * association. The new socket is returned with a reference. */ struct socket * sopeeloff(struct socket *head) { struct socket *so; VNET_ASSERT(head->so_vnet != NULL, ("%s:%d so_vnet is NULL, head=%p", __func__, __LINE__, head)); so = soalloc(head->so_vnet); if (so == NULL) { log(LOG_DEBUG, "%s: pcb %p: New socket allocation failure: " "limit reached or out of memory\n", __func__, head->so_pcb); return (NULL); } so->so_type = head->so_type; so->so_options = head->so_options; so->so_linger = head->so_linger; so->so_state = (head->so_state & SS_NBIO) | SS_ISCONNECTED; so->so_fibnum = head->so_fibnum; so->so_proto = head->so_proto; so->so_cred = crhold(head->so_cred); #ifdef MAC mac_socket_newconn(head, so); #endif knlist_init(&so->so_rdsel.si_note, so, so_rdknl_lock, so_rdknl_unlock, so_rdknl_assert_locked, so_rdknl_assert_unlocked); knlist_init(&so->so_wrsel.si_note, so, so_wrknl_lock, so_wrknl_unlock, so_wrknl_assert_locked, so_wrknl_assert_unlocked); VNET_SO_ASSERT(head); if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat)) { sodealloc(so); log(LOG_DEBUG, "%s: pcb %p: soreserve() failed\n", __func__, head->so_pcb); return (NULL); } if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) { sodealloc(so); log(LOG_DEBUG, "%s: pcb %p: pru_attach() failed\n", __func__, head->so_pcb); return (NULL); } so->so_rcv.sb_lowat = head->so_rcv.sb_lowat; so->so_snd.sb_lowat = head->so_snd.sb_lowat; so->so_rcv.sb_timeo = head->so_rcv.sb_timeo; so->so_snd.sb_timeo = head->so_snd.sb_timeo; so->so_rcv.sb_flags |= head->so_rcv.sb_flags & SB_AUTOSIZE; so->so_snd.sb_flags |= head->so_snd.sb_flags & SB_AUTOSIZE; soref(so); return (so); } #endif /* SCTP */ int sobind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error; CURVNET_SET(so->so_vnet); error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td); CURVNET_RESTORE(); return (error); } int sobindat(int fd, struct socket *so, struct sockaddr *nam, struct thread *td) { int error; CURVNET_SET(so->so_vnet); error = (*so->so_proto->pr_usrreqs->pru_bindat)(fd, so, nam, td); CURVNET_RESTORE(); return (error); } /* * solisten() transitions a socket from a non-listening state to a listening * state, but can also be used to update the listen queue depth on an * existing listen socket. The protocol will call back into the sockets * layer using solisten_proto_check() and solisten_proto() to check and set * socket-layer listen state. Call backs are used so that the protocol can * acquire both protocol and socket layer locks in whatever order is required * by the protocol. * * Protocol implementors are advised to hold the socket lock across the * socket-layer test and set to avoid races at the socket layer. */ int solisten(struct socket *so, int backlog, struct thread *td) { int error; CURVNET_SET(so->so_vnet); error = (*so->so_proto->pr_usrreqs->pru_listen)(so, backlog, td); CURVNET_RESTORE(); return (error); } int solisten_proto_check(struct socket *so) { SOCK_LOCK_ASSERT(so); if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | SS_ISDISCONNECTING)) return (EINVAL); return (0); } void solisten_proto(struct socket *so, int backlog) { int sbrcv_lowat, sbsnd_lowat; u_int sbrcv_hiwat, sbsnd_hiwat; short sbrcv_flags, sbsnd_flags; sbintime_t sbrcv_timeo, sbsnd_timeo; SOCK_LOCK_ASSERT(so); if (SOLISTENING(so)) goto listening; /* * Change this socket to listening state. */ sbrcv_lowat = so->so_rcv.sb_lowat; sbsnd_lowat = so->so_snd.sb_lowat; sbrcv_hiwat = so->so_rcv.sb_hiwat; sbsnd_hiwat = so->so_snd.sb_hiwat; sbrcv_flags = so->so_rcv.sb_flags; sbsnd_flags = so->so_snd.sb_flags; sbrcv_timeo = so->so_rcv.sb_timeo; sbsnd_timeo = so->so_snd.sb_timeo; sbdestroy(&so->so_snd, so); sbdestroy(&so->so_rcv, so); sx_destroy(&so->so_snd.sb_sx); sx_destroy(&so->so_rcv.sb_sx); SOCKBUF_LOCK_DESTROY(&so->so_snd); SOCKBUF_LOCK_DESTROY(&so->so_rcv); #ifdef INVARIANTS bzero(&so->so_rcv, sizeof(struct socket) - offsetof(struct socket, so_rcv)); #endif so->sol_sbrcv_lowat = sbrcv_lowat; so->sol_sbsnd_lowat = sbsnd_lowat; so->sol_sbrcv_hiwat = sbrcv_hiwat; so->sol_sbsnd_hiwat = sbsnd_hiwat; so->sol_sbrcv_flags = sbrcv_flags; so->sol_sbsnd_flags = sbsnd_flags; so->sol_sbrcv_timeo = sbrcv_timeo; so->sol_sbsnd_timeo = sbsnd_timeo; so->sol_qlen = so->sol_incqlen = 0; TAILQ_INIT(&so->sol_incomp); TAILQ_INIT(&so->sol_comp); so->sol_accept_filter = NULL; so->sol_accept_filter_arg = NULL; so->sol_accept_filter_str = NULL; so->sol_upcall = NULL; so->sol_upcallarg = NULL; so->so_options |= SO_ACCEPTCONN; listening: if (backlog < 0 || backlog > somaxconn) backlog = somaxconn; so->sol_qlimit = backlog; } /* * Wakeup listeners/subsystems once we have a complete connection. * Enters with lock, returns unlocked. */ void solisten_wakeup(struct socket *sol) { if (sol->sol_upcall != NULL) (void )sol->sol_upcall(sol, sol->sol_upcallarg, M_NOWAIT); else { selwakeuppri(&sol->so_rdsel, PSOCK); KNOTE_LOCKED(&sol->so_rdsel.si_note, 0); } SOLISTEN_UNLOCK(sol); wakeup_one(&sol->sol_comp); } /* * Return single connection off a listening socket queue. Main consumer of * the function is kern_accept4(). Some modules, that do their own accept * management also use the function. * * Listening socket must be locked on entry and is returned unlocked on * return. * The flags argument is set of accept4(2) flags and ACCEPT4_INHERIT. */ int solisten_dequeue(struct socket *head, struct socket **ret, int flags) { struct socket *so; int error; SOLISTEN_LOCK_ASSERT(head); while (!(head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->sol_comp) && head->so_error == 0) { error = msleep(&head->sol_comp, &head->so_lock, PSOCK | PCATCH, "accept", 0); if (error != 0) { SOLISTEN_UNLOCK(head); return (error); } } if (head->so_error) { error = head->so_error; head->so_error = 0; SOLISTEN_UNLOCK(head); return (error); } if ((head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->sol_comp)) { SOLISTEN_UNLOCK(head); return (EWOULDBLOCK); } so = TAILQ_FIRST(&head->sol_comp); SOCK_LOCK(so); KASSERT(so->so_qstate == SQ_COMP, ("%s: so %p not SQ_COMP", __func__, so)); soref(so); head->sol_qlen--; so->so_qstate = SQ_NONE; so->so_listen = NULL; TAILQ_REMOVE(&head->sol_comp, so, so_list); if (flags & ACCEPT4_INHERIT) so->so_state |= (head->so_state & SS_NBIO); else so->so_state |= (flags & SOCK_NONBLOCK) ? SS_NBIO : 0; SOCK_UNLOCK(so); sorele(head); *ret = so; return (0); } /* * Evaluate the reference count and named references on a socket; if no * references remain, free it. This should be called whenever a reference is * released, such as in sorele(), but also when named reference flags are * cleared in socket or protocol code. * * sofree() will free the socket if: * * - There are no outstanding file descriptor references or related consumers * (so_count == 0). * * - The socket has been closed by user space, if ever open (SS_NOFDREF). * * - The protocol does not have an outstanding strong reference on the socket * (SS_PROTOREF). * * - The socket is not in a completed connection queue, so a process has been * notified that it is present. If it is removed, the user process may * block in accept() despite select() saying the socket was ready. */ void sofree(struct socket *so) { struct protosw *pr = so->so_proto; SOCK_LOCK_ASSERT(so); if ((so->so_state & SS_NOFDREF) == 0 || so->so_count != 0 || (so->so_state & SS_PROTOREF) || (so->so_qstate == SQ_COMP)) { SOCK_UNLOCK(so); return; } if (!SOLISTENING(so) && so->so_qstate == SQ_INCOMP) { struct socket *sol; sol = so->so_listen; KASSERT(sol, ("%s: so %p on incomp of NULL", __func__, so)); /* * To solve race between close of a listening socket and * a socket on its incomplete queue, we need to lock both. * The order is first listening socket, then regular. * Since we don't have SS_NOFDREF neither SS_PROTOREF, this * function and the listening socket are the only pointers * to so. To preserve so and sol, we reference both and then * relock. * After relock the socket may not move to so_comp since it * doesn't have PCB already, but it may be removed from * so_incomp. If that happens, we share responsiblity on * freeing the socket, but soclose() has already removed * it from queue. */ soref(sol); soref(so); SOCK_UNLOCK(so); SOLISTEN_LOCK(sol); SOCK_LOCK(so); if (so->so_qstate == SQ_INCOMP) { KASSERT(so->so_listen == sol, ("%s: so %p migrated out of sol %p", __func__, so, sol)); TAILQ_REMOVE(&sol->sol_incomp, so, so_list); sol->sol_incqlen--; /* This is guarenteed not to be the last. */ refcount_release(&sol->so_count); so->so_qstate = SQ_NONE; so->so_listen = NULL; } else KASSERT(so->so_listen == NULL, ("%s: so %p not on (in)comp with so_listen", __func__, so)); sorele(sol); KASSERT(so->so_count == 1, ("%s: so %p count %u", __func__, so, so->so_count)); so->so_count = 0; } if (SOLISTENING(so)) so->so_error = ECONNABORTED; SOCK_UNLOCK(so); VNET_SO_ASSERT(so); if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose != NULL) (*pr->pr_domain->dom_dispose)(so); if (pr->pr_usrreqs->pru_detach != NULL) (*pr->pr_usrreqs->pru_detach)(so); /* * From this point on, we assume that no other references to this * socket exist anywhere else in the stack. Therefore, no locks need * to be acquired or held. * * We used to do a lot of socket buffer and socket locking here, as * well as invoke sorflush() and perform wakeups. The direct call to * dom_dispose() and sbrelease_internal() are an inlining of what was * necessary from sorflush(). * * Notice that the socket buffer and kqueue state are torn down * before calling pru_detach. This means that protocols shold not * assume they can perform socket wakeups, etc, in their detach code. */ if (!SOLISTENING(so)) { sbdestroy(&so->so_snd, so); sbdestroy(&so->so_rcv, so); } seldrain(&so->so_rdsel); seldrain(&so->so_wrsel); knlist_destroy(&so->so_rdsel.si_note); knlist_destroy(&so->so_wrsel.si_note); sodealloc(so); } /* * Close a socket on last file table reference removal. Initiate disconnect * if connected. Free socket when disconnect complete. * * This function will sorele() the socket. Note that soclose() may be called * prior to the ref count reaching zero. The actual socket structure will * not be freed until the ref count reaches zero. */ int soclose(struct socket *so) { struct accept_queue lqueue; bool listening; int error = 0; KASSERT(!(so->so_state & SS_NOFDREF), ("soclose: SS_NOFDREF on enter")); CURVNET_SET(so->so_vnet); funsetown(&so->so_sigio); if (so->so_state & SS_ISCONNECTED) { if ((so->so_state & SS_ISDISCONNECTING) == 0) { error = sodisconnect(so); if (error) { if (error == ENOTCONN) error = 0; goto drop; } } if (so->so_options & SO_LINGER) { if ((so->so_state & SS_ISDISCONNECTING) && (so->so_state & SS_NBIO)) goto drop; while (so->so_state & SS_ISCONNECTED) { error = tsleep(&so->so_timeo, PSOCK | PCATCH, "soclos", so->so_linger * hz); if (error) break; } } } drop: if (so->so_proto->pr_usrreqs->pru_close != NULL) (*so->so_proto->pr_usrreqs->pru_close)(so); SOCK_LOCK(so); if ((listening = (so->so_options & SO_ACCEPTCONN))) { struct socket *sp; TAILQ_INIT(&lqueue); TAILQ_SWAP(&lqueue, &so->sol_incomp, socket, so_list); TAILQ_CONCAT(&lqueue, &so->sol_comp, so_list); so->sol_qlen = so->sol_incqlen = 0; TAILQ_FOREACH(sp, &lqueue, so_list) { SOCK_LOCK(sp); sp->so_qstate = SQ_NONE; sp->so_listen = NULL; SOCK_UNLOCK(sp); /* Guaranteed not to be the last. */ refcount_release(&so->so_count); } } KASSERT((so->so_state & SS_NOFDREF) == 0, ("soclose: NOFDREF")); so->so_state |= SS_NOFDREF; sorele(so); if (listening) { struct socket *sp; TAILQ_FOREACH(sp, &lqueue, so_list) { SOCK_LOCK(sp); if (sp->so_count == 0) { SOCK_UNLOCK(sp); soabort(sp); } else /* sp is now in sofree() */ SOCK_UNLOCK(sp); } } CURVNET_RESTORE(); return (error); } /* * soabort() is used to abruptly tear down a connection, such as when a * resource limit is reached (listen queue depth exceeded), or if a listen * socket is closed while there are sockets waiting to be accepted. * * This interface is tricky, because it is called on an unreferenced socket, * and must be called only by a thread that has actually removed the socket * from the listen queue it was on, or races with other threads are risked. * * This interface will call into the protocol code, so must not be called * with any socket locks held. Protocols do call it while holding their own * recursible protocol mutexes, but this is something that should be subject * to review in the future. */ void soabort(struct socket *so) { /* * In as much as is possible, assert that no references to this * socket are held. This is not quite the same as asserting that the * current thread is responsible for arranging for no references, but * is as close as we can get for now. */ KASSERT(so->so_count == 0, ("soabort: so_count")); KASSERT((so->so_state & SS_PROTOREF) == 0, ("soabort: SS_PROTOREF")); KASSERT(so->so_state & SS_NOFDREF, ("soabort: !SS_NOFDREF")); KASSERT(so->so_qstate == SQ_NONE, ("soabort: !SQ_NONE")); VNET_SO_ASSERT(so); if (so->so_proto->pr_usrreqs->pru_abort != NULL) (*so->so_proto->pr_usrreqs->pru_abort)(so); SOCK_LOCK(so); sofree(so); } int soaccept(struct socket *so, struct sockaddr **nam) { int error; SOCK_LOCK(so); KASSERT((so->so_state & SS_NOFDREF) != 0, ("soaccept: !NOFDREF")); so->so_state &= ~SS_NOFDREF; SOCK_UNLOCK(so); CURVNET_SET(so->so_vnet); error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam); CURVNET_RESTORE(); return (error); } int soconnect(struct socket *so, struct sockaddr *nam, struct thread *td) { return (soconnectat(AT_FDCWD, so, nam, td)); } int soconnectat(int fd, struct socket *so, struct sockaddr *nam, struct thread *td) { int error; if (so->so_options & SO_ACCEPTCONN) return (EOPNOTSUPP); CURVNET_SET(so->so_vnet); /* * If protocol is connection-based, can only connect once. * Otherwise, if connected, try to disconnect first. This allows * user to disconnect by connecting to, e.g., a null address. */ if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && ((so->so_proto->pr_flags & PR_CONNREQUIRED) || (error = sodisconnect(so)))) { error = EISCONN; } else { /* * Prevent accumulated error from previous connection from * biting us. */ so->so_error = 0; if (fd == AT_FDCWD) { error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td); } else { error = (*so->so_proto->pr_usrreqs->pru_connectat)(fd, so, nam, td); } } CURVNET_RESTORE(); return (error); } int soconnect2(struct socket *so1, struct socket *so2) { int error; CURVNET_SET(so1->so_vnet); error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2); CURVNET_RESTORE(); return (error); } int sodisconnect(struct socket *so) { int error; if ((so->so_state & SS_ISCONNECTED) == 0) return (ENOTCONN); if (so->so_state & SS_ISDISCONNECTING) return (EALREADY); VNET_SO_ASSERT(so); error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so); return (error); } #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? 0 : SBL_WAIT) int sosend_dgram(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td) { long space; ssize_t resid; int clen = 0, error, dontroute; KASSERT(so->so_type == SOCK_DGRAM, ("sosend_dgram: !SOCK_DGRAM")); KASSERT(so->so_proto->pr_flags & PR_ATOMIC, ("sosend_dgram: !PR_ATOMIC")); if (uio != NULL) resid = uio->uio_resid; else resid = top->m_pkthdr.len; /* * In theory resid should be unsigned. However, space must be * signed, as it might be less than 0 if we over-committed, and we * must use a signed comparison of space and resid. On the other * hand, a negative resid causes us to loop sending 0-length * segments to the protocol. */ if (resid < 0) { error = EINVAL; goto out; } dontroute = (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0; if (td != NULL) td->td_ru.ru_msgsnd++; if (control != NULL) clen = control->m_len; SOCKBUF_LOCK(&so->so_snd); if (so->so_snd.sb_state & SBS_CANTSENDMORE) { SOCKBUF_UNLOCK(&so->so_snd); error = EPIPE; goto out; } if (so->so_error) { error = so->so_error; so->so_error = 0; SOCKBUF_UNLOCK(&so->so_snd); goto out; } if ((so->so_state & SS_ISCONNECTED) == 0) { /* * `sendto' and `sendmsg' is allowed on a connection-based * socket if it supports implied connect. Return ENOTCONN if * not connected and no address is supplied. */ if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { if ((so->so_state & SS_ISCONFIRMING) == 0 && !(resid == 0 && clen != 0)) { SOCKBUF_UNLOCK(&so->so_snd); error = ENOTCONN; goto out; } } else if (addr == NULL) { if (so->so_proto->pr_flags & PR_CONNREQUIRED) error = ENOTCONN; else error = EDESTADDRREQ; SOCKBUF_UNLOCK(&so->so_snd); goto out; } } /* * Do we need MSG_OOB support in SOCK_DGRAM? Signs here may be a * problem and need fixing. */ space = sbspace(&so->so_snd); if (flags & MSG_OOB) space += 1024; space -= clen; SOCKBUF_UNLOCK(&so->so_snd); if (resid > space) { error = EMSGSIZE; goto out; } if (uio == NULL) { resid = 0; if (flags & MSG_EOR) top->m_flags |= M_EOR; } else { /* * Copy the data from userland into a mbuf chain. * If no data is to be copied in, a single empty mbuf * is returned. */ top = m_uiotombuf(uio, M_WAITOK, space, max_hdr, (M_PKTHDR | ((flags & MSG_EOR) ? M_EOR : 0))); if (top == NULL) { error = EFAULT; /* only possible error */ goto out; } space -= resid - uio->uio_resid; resid = uio->uio_resid; } KASSERT(resid == 0, ("sosend_dgram: resid != 0")); /* * XXXRW: Frobbing SO_DONTROUTE here is even worse without sblock * than with. */ if (dontroute) { SOCK_LOCK(so); so->so_options |= SO_DONTROUTE; SOCK_UNLOCK(so); } /* * XXX all the SBS_CANTSENDMORE checks previously done could be out * of date. We could have received a reset packet in an interrupt or * maybe we slept while doing page faults in uiomove() etc. We could * probably recheck again inside the locking protection here, but * there are probably other places that this also happens. We must * rethink this. */ VNET_SO_ASSERT(so); error = (*so->so_proto->pr_usrreqs->pru_send)(so, (flags & MSG_OOB) ? PRUS_OOB : /* * If the user set MSG_EOF, the protocol understands this flag and * nothing left to send then use PRU_SEND_EOF instead of PRU_SEND. */ ((flags & MSG_EOF) && (so->so_proto->pr_flags & PR_IMPLOPCL) && (resid <= 0)) ? PRUS_EOF : /* If there is more to send set PRUS_MORETOCOME */ (flags & MSG_MORETOCOME) || (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0, top, addr, control, td); if (dontroute) { SOCK_LOCK(so); so->so_options &= ~SO_DONTROUTE; SOCK_UNLOCK(so); } clen = 0; control = NULL; top = NULL; out: if (top != NULL) m_freem(top); if (control != NULL) m_freem(control); return (error); } /* * Send on a socket. If send must go all at once and message is larger than * send buffering, then hard error. Lock against other senders. If must go * all at once and not enough room now, then inform user that this would * block and do nothing. Otherwise, if nonblocking, send as much as * possible. The data to be sent is described by "uio" if nonzero, otherwise * by the mbuf chain "top" (which must be null if uio is not). Data provided * in mbuf chain must be small enough to send all at once. * * Returns nonzero on error, timeout or signal; callers must check for short * counts if EINTR/ERESTART are returned. Data and control buffers are freed * on return. */ int sosend_generic(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td) { long space; ssize_t resid; int clen = 0, error, dontroute; int atomic = sosendallatonce(so) || top; if (uio != NULL) resid = uio->uio_resid; else resid = top->m_pkthdr.len; /* * In theory resid should be unsigned. However, space must be * signed, as it might be less than 0 if we over-committed, and we * must use a signed comparison of space and resid. On the other * hand, a negative resid causes us to loop sending 0-length * segments to the protocol. * * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM * type sockets since that's an error. */ if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) { error = EINVAL; goto out; } dontroute = (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && (so->so_proto->pr_flags & PR_ATOMIC); if (td != NULL) td->td_ru.ru_msgsnd++; if (control != NULL) clen = control->m_len; error = sblock(&so->so_snd, SBLOCKWAIT(flags)); if (error) goto out; restart: do { SOCKBUF_LOCK(&so->so_snd); if (so->so_snd.sb_state & SBS_CANTSENDMORE) { SOCKBUF_UNLOCK(&so->so_snd); error = EPIPE; goto release; } if (so->so_error) { error = so->so_error; so->so_error = 0; SOCKBUF_UNLOCK(&so->so_snd); goto release; } if ((so->so_state & SS_ISCONNECTED) == 0) { /* * `sendto' and `sendmsg' is allowed on a connection- * based socket if it supports implied connect. * Return ENOTCONN if not connected and no address is * supplied. */ if ((so->so_proto->pr_flags & PR_CONNREQUIRED) && (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) { if ((so->so_state & SS_ISCONFIRMING) == 0 && !(resid == 0 && clen != 0)) { SOCKBUF_UNLOCK(&so->so_snd); error = ENOTCONN; goto release; } } else if (addr == NULL) { SOCKBUF_UNLOCK(&so->so_snd); if (so->so_proto->pr_flags & PR_CONNREQUIRED) error = ENOTCONN; else error = EDESTADDRREQ; goto release; } } space = sbspace(&so->so_snd); if (flags & MSG_OOB) space += 1024; if ((atomic && resid > so->so_snd.sb_hiwat) || clen > so->so_snd.sb_hiwat) { SOCKBUF_UNLOCK(&so->so_snd); error = EMSGSIZE; goto release; } if (space < resid + clen && (atomic || space < so->so_snd.sb_lowat || space < clen)) { if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) { SOCKBUF_UNLOCK(&so->so_snd); error = EWOULDBLOCK; goto release; } error = sbwait(&so->so_snd); SOCKBUF_UNLOCK(&so->so_snd); if (error) goto release; goto restart; } SOCKBUF_UNLOCK(&so->so_snd); space -= clen; do { if (uio == NULL) { resid = 0; if (flags & MSG_EOR) top->m_flags |= M_EOR; } else { /* * Copy the data from userland into a mbuf * chain. If resid is 0, which can happen * only if we have control to send, then * a single empty mbuf is returned. This * is a workaround to prevent protocol send * methods to panic. */ top = m_uiotombuf(uio, M_WAITOK, space, (atomic ? max_hdr : 0), (atomic ? M_PKTHDR : 0) | ((flags & MSG_EOR) ? M_EOR : 0)); if (top == NULL) { error = EFAULT; /* only possible error */ goto release; } space -= resid - uio->uio_resid; resid = uio->uio_resid; } if (dontroute) { SOCK_LOCK(so); so->so_options |= SO_DONTROUTE; SOCK_UNLOCK(so); } /* * XXX all the SBS_CANTSENDMORE checks previously * done could be out of date. We could have received * a reset packet in an interrupt or maybe we slept * while doing page faults in uiomove() etc. We * could probably recheck again inside the locking * protection here, but there are probably other * places that this also happens. We must rethink * this. */ VNET_SO_ASSERT(so); error = (*so->so_proto->pr_usrreqs->pru_send)(so, (flags & MSG_OOB) ? PRUS_OOB : /* * If the user set MSG_EOF, the protocol understands * this flag and nothing left to send then use * PRU_SEND_EOF instead of PRU_SEND. */ ((flags & MSG_EOF) && (so->so_proto->pr_flags & PR_IMPLOPCL) && (resid <= 0)) ? PRUS_EOF : /* If there is more to send set PRUS_MORETOCOME. */ (flags & MSG_MORETOCOME) || (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0, top, addr, control, td); if (dontroute) { SOCK_LOCK(so); so->so_options &= ~SO_DONTROUTE; SOCK_UNLOCK(so); } clen = 0; control = NULL; top = NULL; if (error) goto release; } while (resid && space > 0); } while (resid); release: sbunlock(&so->so_snd); out: if (top != NULL) m_freem(top); if (control != NULL) m_freem(control); return (error); } int sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td) { int error; CURVNET_SET(so->so_vnet); if (!SOLISTENING(so)) error = so->so_proto->pr_usrreqs->pru_sosend(so, addr, uio, top, control, flags, td); else { m_freem(top); m_freem(control); error = ENOTCONN; } CURVNET_RESTORE(); return (error); } /* * The part of soreceive() that implements reading non-inline out-of-band * data from a socket. For more complete comments, see soreceive(), from * which this code originated. * * Note that soreceive_rcvoob(), unlike the remainder of soreceive(), is * unable to return an mbuf chain to the caller. */ static int soreceive_rcvoob(struct socket *so, struct uio *uio, int flags) { struct protosw *pr = so->so_proto; struct mbuf *m; int error; KASSERT(flags & MSG_OOB, ("soreceive_rcvoob: (flags & MSG_OOB) == 0")); VNET_SO_ASSERT(so); m = m_get(M_WAITOK, MT_DATA); error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK); if (error) goto bad; do { error = uiomove(mtod(m, void *), (int) min(uio->uio_resid, m->m_len), uio); m = m_free(m); } while (uio->uio_resid && error == 0 && m); bad: if (m != NULL) m_freem(m); return (error); } /* * Following replacement or removal of the first mbuf on the first mbuf chain * of a socket buffer, push necessary state changes back into the socket * buffer so that other consumers see the values consistently. 'nextrecord' * is the callers locally stored value of the original value of * sb->sb_mb->m_nextpkt which must be restored when the lead mbuf changes. * NOTE: 'nextrecord' may be NULL. */ static __inline void sockbuf_pushsync(struct sockbuf *sb, struct mbuf *nextrecord) { SOCKBUF_LOCK_ASSERT(sb); /* * First, update for the new value of nextrecord. If necessary, make * it the first record. */ if (sb->sb_mb != NULL) sb->sb_mb->m_nextpkt = nextrecord; else sb->sb_mb = nextrecord; /* * Now update any dependent socket buffer fields to reflect the new * state. This is an expanded inline of SB_EMPTY_FIXUP(), with the * addition of a second clause that takes care of the case where * sb_mb has been updated, but remains the last record. */ if (sb->sb_mb == NULL) { sb->sb_mbtail = NULL; sb->sb_lastrecord = NULL; } else if (sb->sb_mb->m_nextpkt == NULL) sb->sb_lastrecord = sb->sb_mb; } /* * Implement receive operations on a socket. We depend on the way that * records are added to the sockbuf by sbappend. In particular, each record * (mbufs linked through m_next) must begin with an address if the protocol * so specifies, followed by an optional mbuf or mbufs containing ancillary * data, and then zero or more mbufs of data. In order to allow parallelism * between network receive and copying to user space, as well as avoid * sleeping with a mutex held, we release the socket buffer mutex during the * user space copy. Although the sockbuf is locked, new data may still be * appended, and thus we must maintain consistency of the sockbuf during that * time. * * The caller may receive the data as a single mbuf chain by supplying an * mbuf **mp0 for use in returning the chain. The uio is then used only for * the count in uio_resid. */ int soreceive_generic(struct socket *so, struct sockaddr **psa, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { struct mbuf *m, **mp; int flags, error, offset; ssize_t len; struct protosw *pr = so->so_proto; struct mbuf *nextrecord; int moff, type = 0; ssize_t orig_resid = uio->uio_resid; mp = mp0; if (psa != NULL) *psa = NULL; if (controlp != NULL) *controlp = NULL; if (flagsp != NULL) flags = *flagsp &~ MSG_EOR; else flags = 0; if (flags & MSG_OOB) return (soreceive_rcvoob(so, uio, flags)); if (mp != NULL) *mp = NULL; if ((pr->pr_flags & PR_WANTRCVD) && (so->so_state & SS_ISCONFIRMING) && uio->uio_resid) { VNET_SO_ASSERT(so); (*pr->pr_usrreqs->pru_rcvd)(so, 0); } error = sblock(&so->so_rcv, SBLOCKWAIT(flags)); if (error) return (error); restart: SOCKBUF_LOCK(&so->so_rcv); m = so->so_rcv.sb_mb; /* * If we have less data than requested, block awaiting more (subject * to any timeout) if: * 1. the current count is less than the low water mark, or * 2. MSG_DONTWAIT is not set */ if (m == NULL || (((flags & MSG_DONTWAIT) == 0 && sbavail(&so->so_rcv) < uio->uio_resid) && sbavail(&so->so_rcv) < so->so_rcv.sb_lowat && m->m_nextpkt == NULL && (pr->pr_flags & PR_ATOMIC) == 0)) { KASSERT(m != NULL || !sbavail(&so->so_rcv), ("receive: m == %p sbavail == %u", m, sbavail(&so->so_rcv))); if (so->so_error) { if (m != NULL) goto dontblock; error = so->so_error; if ((flags & MSG_PEEK) == 0) so->so_error = 0; SOCKBUF_UNLOCK(&so->so_rcv); goto release; } SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { if (m == NULL) { SOCKBUF_UNLOCK(&so->so_rcv); goto release; } else goto dontblock; } for (; m != NULL; m = m->m_next) if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { m = so->so_rcv.sb_mb; goto dontblock; } if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && (so->so_proto->pr_flags & PR_CONNREQUIRED)) { SOCKBUF_UNLOCK(&so->so_rcv); error = ENOTCONN; goto release; } if (uio->uio_resid == 0) { SOCKBUF_UNLOCK(&so->so_rcv); goto release; } if ((so->so_state & SS_NBIO) || (flags & (MSG_DONTWAIT|MSG_NBIO))) { SOCKBUF_UNLOCK(&so->so_rcv); error = EWOULDBLOCK; goto release; } SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); error = sbwait(&so->so_rcv); SOCKBUF_UNLOCK(&so->so_rcv); if (error) goto release; goto restart; } dontblock: /* * From this point onward, we maintain 'nextrecord' as a cache of the * pointer to the next record in the socket buffer. We must keep the * various socket buffer pointers and local stack versions of the * pointers in sync, pushing out modifications before dropping the * socket buffer mutex, and re-reading them when picking it up. * * Otherwise, we will race with the network stack appending new data * or records onto the socket buffer by using inconsistent/stale * versions of the field, possibly resulting in socket buffer * corruption. * * By holding the high-level sblock(), we prevent simultaneous * readers from pulling off the front of the socket buffer. */ SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (uio->uio_td) uio->uio_td->td_ru.ru_msgrcv++; KASSERT(m == so->so_rcv.sb_mb, ("soreceive: m != so->so_rcv.sb_mb")); SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); nextrecord = m->m_nextpkt; if (pr->pr_flags & PR_ADDR) { KASSERT(m->m_type == MT_SONAME, ("m->m_type == %d", m->m_type)); orig_resid = 0; if (psa != NULL) *psa = sodupsockaddr(mtod(m, struct sockaddr *), M_NOWAIT); if (flags & MSG_PEEK) { m = m->m_next; } else { sbfree(&so->so_rcv, m); so->so_rcv.sb_mb = m_free(m); m = so->so_rcv.sb_mb; sockbuf_pushsync(&so->so_rcv, nextrecord); } } /* * Process one or more MT_CONTROL mbufs present before any data mbufs * in the first mbuf chain on the socket buffer. If MSG_PEEK, we * just copy the data; if !MSG_PEEK, we call into the protocol to * perform externalization (or freeing if controlp == NULL). */ if (m != NULL && m->m_type == MT_CONTROL) { struct mbuf *cm = NULL, *cmn; struct mbuf **cme = &cm; do { if (flags & MSG_PEEK) { if (controlp != NULL) { *controlp = m_copym(m, 0, m->m_len, M_NOWAIT); controlp = &(*controlp)->m_next; } m = m->m_next; } else { sbfree(&so->so_rcv, m); so->so_rcv.sb_mb = m->m_next; m->m_next = NULL; *cme = m; cme = &(*cme)->m_next; m = so->so_rcv.sb_mb; } } while (m != NULL && m->m_type == MT_CONTROL); if ((flags & MSG_PEEK) == 0) sockbuf_pushsync(&so->so_rcv, nextrecord); while (cm != NULL) { cmn = cm->m_next; cm->m_next = NULL; if (pr->pr_domain->dom_externalize != NULL) { SOCKBUF_UNLOCK(&so->so_rcv); VNET_SO_ASSERT(so); error = (*pr->pr_domain->dom_externalize) (cm, controlp, flags); SOCKBUF_LOCK(&so->so_rcv); } else if (controlp != NULL) *controlp = cm; else m_freem(cm); if (controlp != NULL) { orig_resid = 0; while (*controlp != NULL) controlp = &(*controlp)->m_next; } cm = cmn; } if (m != NULL) nextrecord = so->so_rcv.sb_mb->m_nextpkt; else nextrecord = so->so_rcv.sb_mb; orig_resid = 0; } if (m != NULL) { if ((flags & MSG_PEEK) == 0) { KASSERT(m->m_nextpkt == nextrecord, ("soreceive: post-control, nextrecord !sync")); if (nextrecord == NULL) { KASSERT(so->so_rcv.sb_mb == m, ("soreceive: post-control, sb_mb!=m")); KASSERT(so->so_rcv.sb_lastrecord == m, ("soreceive: post-control, lastrecord!=m")); } } type = m->m_type; if (type == MT_OOBDATA) flags |= MSG_OOB; } else { if ((flags & MSG_PEEK) == 0) { KASSERT(so->so_rcv.sb_mb == nextrecord, ("soreceive: sb_mb != nextrecord")); if (so->so_rcv.sb_mb == NULL) { KASSERT(so->so_rcv.sb_lastrecord == NULL, ("soreceive: sb_lastercord != NULL")); } } } SOCKBUF_LOCK_ASSERT(&so->so_rcv); SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); /* * Now continue to read any data mbufs off of the head of the socket * buffer until the read request is satisfied. Note that 'type' is * used to store the type of any mbuf reads that have happened so far * such that soreceive() can stop reading if the type changes, which * causes soreceive() to return only one of regular data and inline * out-of-band data in a single socket receive operation. */ moff = 0; offset = 0; while (m != NULL && !(m->m_flags & M_NOTAVAIL) && uio->uio_resid > 0 && error == 0) { /* * If the type of mbuf has changed since the last mbuf * examined ('type'), end the receive operation. */ SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (m->m_type == MT_OOBDATA || m->m_type == MT_CONTROL) { if (type != m->m_type) break; } else if (type == MT_OOBDATA) break; else KASSERT(m->m_type == MT_DATA, ("m->m_type == %d", m->m_type)); so->so_rcv.sb_state &= ~SBS_RCVATMARK; len = uio->uio_resid; if (so->so_oobmark && len > so->so_oobmark - offset) len = so->so_oobmark - offset; if (len > m->m_len - moff) len = m->m_len - moff; /* * If mp is set, just pass back the mbufs. Otherwise copy * them out via the uio, then free. Sockbuf must be * consistent here (points to current mbuf, it points to next * record) when we drop priority; we must note any additions * to the sockbuf when we block interrupts again. */ if (mp == NULL) { SOCKBUF_LOCK_ASSERT(&so->so_rcv); SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); SOCKBUF_UNLOCK(&so->so_rcv); error = uiomove(mtod(m, char *) + moff, (int)len, uio); SOCKBUF_LOCK(&so->so_rcv); if (error) { /* * The MT_SONAME mbuf has already been removed * from the record, so it is necessary to * remove the data mbufs, if any, to preserve * the invariant in the case of PR_ADDR that * requires MT_SONAME mbufs at the head of * each record. */ if (pr->pr_flags & PR_ATOMIC && ((flags & MSG_PEEK) == 0)) (void)sbdroprecord_locked(&so->so_rcv); SOCKBUF_UNLOCK(&so->so_rcv); goto release; } } else uio->uio_resid -= len; SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (len == m->m_len - moff) { if (m->m_flags & M_EOR) flags |= MSG_EOR; if (flags & MSG_PEEK) { m = m->m_next; moff = 0; } else { nextrecord = m->m_nextpkt; sbfree(&so->so_rcv, m); if (mp != NULL) { m->m_nextpkt = NULL; *mp = m; mp = &m->m_next; so->so_rcv.sb_mb = m = m->m_next; *mp = NULL; } else { so->so_rcv.sb_mb = m_free(m); m = so->so_rcv.sb_mb; } sockbuf_pushsync(&so->so_rcv, nextrecord); SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); } } else { if (flags & MSG_PEEK) moff += len; else { if (mp != NULL) { if (flags & MSG_DONTWAIT) { *mp = m_copym(m, 0, len, M_NOWAIT); if (*mp == NULL) { /* * m_copym() couldn't * allocate an mbuf. * Adjust uio_resid back * (it was adjusted * down by len bytes, * which we didn't end * up "copying" over). */ uio->uio_resid += len; break; } } else { SOCKBUF_UNLOCK(&so->so_rcv); *mp = m_copym(m, 0, len, M_WAITOK); SOCKBUF_LOCK(&so->so_rcv); } } sbcut_locked(&so->so_rcv, len); } } SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (so->so_oobmark) { if ((flags & MSG_PEEK) == 0) { so->so_oobmark -= len; if (so->so_oobmark == 0) { so->so_rcv.sb_state |= SBS_RCVATMARK; break; } } else { offset += len; if (offset == so->so_oobmark) break; } } if (flags & MSG_EOR) break; /* * If the MSG_WAITALL flag is set (for non-atomic socket), we * must not quit until "uio->uio_resid == 0" or an error * termination. If a signal/timeout occurs, return with a * short count but without error. Keep sockbuf locked * against other readers. */ while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 && !sosendallatonce(so) && nextrecord == NULL) { SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (so->so_error || so->so_rcv.sb_state & SBS_CANTRCVMORE) break; /* * Notify the protocol that some data has been * drained before blocking. */ if (pr->pr_flags & PR_WANTRCVD) { SOCKBUF_UNLOCK(&so->so_rcv); VNET_SO_ASSERT(so); (*pr->pr_usrreqs->pru_rcvd)(so, flags); SOCKBUF_LOCK(&so->so_rcv); } SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); /* * We could receive some data while was notifying * the protocol. Skip blocking in this case. */ if (so->so_rcv.sb_mb == NULL) { error = sbwait(&so->so_rcv); if (error) { SOCKBUF_UNLOCK(&so->so_rcv); goto release; } } m = so->so_rcv.sb_mb; if (m != NULL) nextrecord = m->m_nextpkt; } } SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (m != NULL && pr->pr_flags & PR_ATOMIC) { flags |= MSG_TRUNC; if ((flags & MSG_PEEK) == 0) (void) sbdroprecord_locked(&so->so_rcv); } if ((flags & MSG_PEEK) == 0) { if (m == NULL) { /* * First part is an inline SB_EMPTY_FIXUP(). Second * part makes sure sb_lastrecord is up-to-date if * there is still data in the socket buffer. */ so->so_rcv.sb_mb = nextrecord; if (so->so_rcv.sb_mb == NULL) { so->so_rcv.sb_mbtail = NULL; so->so_rcv.sb_lastrecord = NULL; } else if (nextrecord->m_nextpkt == NULL) so->so_rcv.sb_lastrecord = nextrecord; } SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); /* * If soreceive() is being done from the socket callback, * then don't need to generate ACK to peer to update window, * since ACK will be generated on return to TCP. */ if (!(flags & MSG_SOCALLBCK) && (pr->pr_flags & PR_WANTRCVD)) { SOCKBUF_UNLOCK(&so->so_rcv); VNET_SO_ASSERT(so); (*pr->pr_usrreqs->pru_rcvd)(so, flags); SOCKBUF_LOCK(&so->so_rcv); } } SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (orig_resid == uio->uio_resid && orig_resid && (flags & MSG_EOR) == 0 && (so->so_rcv.sb_state & SBS_CANTRCVMORE) == 0) { SOCKBUF_UNLOCK(&so->so_rcv); goto restart; } SOCKBUF_UNLOCK(&so->so_rcv); if (flagsp != NULL) *flagsp |= flags; release: sbunlock(&so->so_rcv); return (error); } /* * Optimized version of soreceive() for stream (TCP) sockets. * XXXAO: (MSG_WAITALL | MSG_PEEK) isn't properly handled. */ int soreceive_stream(struct socket *so, struct sockaddr **psa, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { int len = 0, error = 0, flags, oresid; struct sockbuf *sb; struct mbuf *m, *n = NULL; /* We only do stream sockets. */ if (so->so_type != SOCK_STREAM) return (EINVAL); if (psa != NULL) *psa = NULL; if (controlp != NULL) return (EINVAL); if (flagsp != NULL) flags = *flagsp &~ MSG_EOR; else flags = 0; if (flags & MSG_OOB) return (soreceive_rcvoob(so, uio, flags)); if (mp0 != NULL) *mp0 = NULL; sb = &so->so_rcv; /* Prevent other readers from entering the socket. */ error = sblock(sb, SBLOCKWAIT(flags)); if (error) goto out; SOCKBUF_LOCK(sb); /* Easy one, no space to copyout anything. */ if (uio->uio_resid == 0) { error = EINVAL; goto out; } oresid = uio->uio_resid; /* We will never ever get anything unless we are or were connected. */ if (!(so->so_state & (SS_ISCONNECTED|SS_ISDISCONNECTED))) { error = ENOTCONN; goto out; } restart: SOCKBUF_LOCK_ASSERT(&so->so_rcv); /* Abort if socket has reported problems. */ if (so->so_error) { if (sbavail(sb) > 0) goto deliver; if (oresid > uio->uio_resid) goto out; error = so->so_error; if (!(flags & MSG_PEEK)) so->so_error = 0; goto out; } /* Door is closed. Deliver what is left, if any. */ if (sb->sb_state & SBS_CANTRCVMORE) { if (sbavail(sb) > 0) goto deliver; else goto out; } /* Socket buffer is empty and we shall not block. */ if (sbavail(sb) == 0 && ((so->so_state & SS_NBIO) || (flags & (MSG_DONTWAIT|MSG_NBIO)))) { error = EAGAIN; goto out; } /* Socket buffer got some data that we shall deliver now. */ if (sbavail(sb) > 0 && !(flags & MSG_WAITALL) && ((so->so_state & SS_NBIO) || (flags & (MSG_DONTWAIT|MSG_NBIO)) || sbavail(sb) >= sb->sb_lowat || sbavail(sb) >= uio->uio_resid || sbavail(sb) >= sb->sb_hiwat) ) { goto deliver; } /* On MSG_WAITALL we must wait until all data or error arrives. */ if ((flags & MSG_WAITALL) && (sbavail(sb) >= uio->uio_resid || sbavail(sb) >= sb->sb_hiwat)) goto deliver; /* * Wait and block until (more) data comes in. * NB: Drops the sockbuf lock during wait. */ error = sbwait(sb); if (error) goto out; goto restart; deliver: SOCKBUF_LOCK_ASSERT(&so->so_rcv); KASSERT(sbavail(sb) > 0, ("%s: sockbuf empty", __func__)); KASSERT(sb->sb_mb != NULL, ("%s: sb_mb == NULL", __func__)); /* Statistics. */ if (uio->uio_td) uio->uio_td->td_ru.ru_msgrcv++; /* Fill uio until full or current end of socket buffer is reached. */ len = min(uio->uio_resid, sbavail(sb)); if (mp0 != NULL) { /* Dequeue as many mbufs as possible. */ if (!(flags & MSG_PEEK) && len >= sb->sb_mb->m_len) { if (*mp0 == NULL) *mp0 = sb->sb_mb; else m_cat(*mp0, sb->sb_mb); for (m = sb->sb_mb; m != NULL && m->m_len <= len; m = m->m_next) { KASSERT(!(m->m_flags & M_NOTAVAIL), ("%s: m %p not available", __func__, m)); len -= m->m_len; uio->uio_resid -= m->m_len; sbfree(sb, m); n = m; } n->m_next = NULL; sb->sb_mb = m; sb->sb_lastrecord = sb->sb_mb; if (sb->sb_mb == NULL) SB_EMPTY_FIXUP(sb); } /* Copy the remainder. */ if (len > 0) { KASSERT(sb->sb_mb != NULL, ("%s: len > 0 && sb->sb_mb empty", __func__)); m = m_copym(sb->sb_mb, 0, len, M_NOWAIT); if (m == NULL) len = 0; /* Don't flush data from sockbuf. */ else uio->uio_resid -= len; if (*mp0 != NULL) m_cat(*mp0, m); else *mp0 = m; if (*mp0 == NULL) { error = ENOBUFS; goto out; } } } else { /* NB: Must unlock socket buffer as uiomove may sleep. */ SOCKBUF_UNLOCK(sb); error = m_mbuftouio(uio, sb->sb_mb, len); SOCKBUF_LOCK(sb); if (error) goto out; } SBLASTRECORDCHK(sb); SBLASTMBUFCHK(sb); /* * Remove the delivered data from the socket buffer unless we * were only peeking. */ if (!(flags & MSG_PEEK)) { if (len > 0) sbdrop_locked(sb, len); /* Notify protocol that we drained some data. */ if ((so->so_proto->pr_flags & PR_WANTRCVD) && (((flags & MSG_WAITALL) && uio->uio_resid > 0) || !(flags & MSG_SOCALLBCK))) { SOCKBUF_UNLOCK(sb); VNET_SO_ASSERT(so); (*so->so_proto->pr_usrreqs->pru_rcvd)(so, flags); SOCKBUF_LOCK(sb); } } /* * For MSG_WAITALL we may have to loop again and wait for * more data to come in. */ if ((flags & MSG_WAITALL) && uio->uio_resid > 0) goto restart; out: SOCKBUF_LOCK_ASSERT(sb); SBLASTRECORDCHK(sb); SBLASTMBUFCHK(sb); SOCKBUF_UNLOCK(sb); sbunlock(sb); return (error); } /* * Optimized version of soreceive() for simple datagram cases from userspace. * Unlike in the stream case, we're able to drop a datagram if copyout() * fails, and because we handle datagrams atomically, we don't need to use a * sleep lock to prevent I/O interlacing. */ int soreceive_dgram(struct socket *so, struct sockaddr **psa, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { struct mbuf *m, *m2; int flags, error; ssize_t len; struct protosw *pr = so->so_proto; struct mbuf *nextrecord; if (psa != NULL) *psa = NULL; if (controlp != NULL) *controlp = NULL; if (flagsp != NULL) flags = *flagsp &~ MSG_EOR; else flags = 0; /* * For any complicated cases, fall back to the full * soreceive_generic(). */ if (mp0 != NULL || (flags & MSG_PEEK) || (flags & MSG_OOB)) return (soreceive_generic(so, psa, uio, mp0, controlp, flagsp)); /* * Enforce restrictions on use. */ KASSERT((pr->pr_flags & PR_WANTRCVD) == 0, ("soreceive_dgram: wantrcvd")); KASSERT(pr->pr_flags & PR_ATOMIC, ("soreceive_dgram: !atomic")); KASSERT((so->so_rcv.sb_state & SBS_RCVATMARK) == 0, ("soreceive_dgram: SBS_RCVATMARK")); KASSERT((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0, ("soreceive_dgram: P_CONNREQUIRED")); /* * Loop blocking while waiting for a datagram. */ SOCKBUF_LOCK(&so->so_rcv); while ((m = so->so_rcv.sb_mb) == NULL) { KASSERT(sbavail(&so->so_rcv) == 0, ("soreceive_dgram: sb_mb NULL but sbavail %u", sbavail(&so->so_rcv))); if (so->so_error) { error = so->so_error; so->so_error = 0; SOCKBUF_UNLOCK(&so->so_rcv); return (error); } if (so->so_rcv.sb_state & SBS_CANTRCVMORE || uio->uio_resid == 0) { SOCKBUF_UNLOCK(&so->so_rcv); return (0); } if ((so->so_state & SS_NBIO) || (flags & (MSG_DONTWAIT|MSG_NBIO))) { SOCKBUF_UNLOCK(&so->so_rcv); return (EWOULDBLOCK); } SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); error = sbwait(&so->so_rcv); if (error) { SOCKBUF_UNLOCK(&so->so_rcv); return (error); } } SOCKBUF_LOCK_ASSERT(&so->so_rcv); if (uio->uio_td) uio->uio_td->td_ru.ru_msgrcv++; SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); nextrecord = m->m_nextpkt; if (nextrecord == NULL) { KASSERT(so->so_rcv.sb_lastrecord == m, ("soreceive_dgram: lastrecord != m")); } KASSERT(so->so_rcv.sb_mb->m_nextpkt == nextrecord, ("soreceive_dgram: m_nextpkt != nextrecord")); /* * Pull 'm' and its chain off the front of the packet queue. */ so->so_rcv.sb_mb = NULL; sockbuf_pushsync(&so->so_rcv, nextrecord); /* * Walk 'm's chain and free that many bytes from the socket buffer. */ for (m2 = m; m2 != NULL; m2 = m2->m_next) sbfree(&so->so_rcv, m2); /* * Do a few last checks before we let go of the lock. */ SBLASTRECORDCHK(&so->so_rcv); SBLASTMBUFCHK(&so->so_rcv); SOCKBUF_UNLOCK(&so->so_rcv); if (pr->pr_flags & PR_ADDR) { KASSERT(m->m_type == MT_SONAME, ("m->m_type == %d", m->m_type)); if (psa != NULL) *psa = sodupsockaddr(mtod(m, struct sockaddr *), M_NOWAIT); m = m_free(m); } if (m == NULL) { /* XXXRW: Can this happen? */ return (0); } /* * Packet to copyout() is now in 'm' and it is disconnected from the * queue. * * Process one or more MT_CONTROL mbufs present before any data mbufs * in the first mbuf chain on the socket buffer. We call into the * protocol to perform externalization (or freeing if controlp == * NULL). In some cases there can be only MT_CONTROL mbufs without * MT_DATA mbufs. */ if (m->m_type == MT_CONTROL) { struct mbuf *cm = NULL, *cmn; struct mbuf **cme = &cm; do { m2 = m->m_next; m->m_next = NULL; *cme = m; cme = &(*cme)->m_next; m = m2; } while (m != NULL && m->m_type == MT_CONTROL); while (cm != NULL) { cmn = cm->m_next; cm->m_next = NULL; if (pr->pr_domain->dom_externalize != NULL) { error = (*pr->pr_domain->dom_externalize) (cm, controlp, flags); } else if (controlp != NULL) *controlp = cm; else m_freem(cm); if (controlp != NULL) { while (*controlp != NULL) controlp = &(*controlp)->m_next; } cm = cmn; } } KASSERT(m == NULL || m->m_type == MT_DATA, ("soreceive_dgram: !data")); while (m != NULL && uio->uio_resid > 0) { len = uio->uio_resid; if (len > m->m_len) len = m->m_len; error = uiomove(mtod(m, char *), (int)len, uio); if (error) { m_freem(m); return (error); } if (len == m->m_len) m = m_free(m); else { m->m_data += len; m->m_len -= len; } } if (m != NULL) { flags |= MSG_TRUNC; m_freem(m); } if (flagsp != NULL) *flagsp |= flags; return (0); } int soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { int error; CURVNET_SET(so->so_vnet); if (!SOLISTENING(so)) error = (so->so_proto->pr_usrreqs->pru_soreceive(so, psa, uio, mp0, controlp, flagsp)); else error = ENOTCONN; CURVNET_RESTORE(); return (error); } int soshutdown(struct socket *so, int how) { struct protosw *pr = so->so_proto; int error, soerror_enotconn; if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) return (EINVAL); soerror_enotconn = 0; if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | SS_ISDISCONNECTING)) == 0) { /* * POSIX mandates us to return ENOTCONN when shutdown(2) is * invoked on a datagram sockets, however historically we would * actually tear socket down. This is known to be leveraged by * some applications to unblock process waiting in recvXXX(2) * by other process that it shares that socket with. Try to meet * both backward-compatibility and POSIX requirements by forcing * ENOTCONN but still asking protocol to perform pru_shutdown(). */ if (so->so_type != SOCK_DGRAM) return (ENOTCONN); soerror_enotconn = 1; } CURVNET_SET(so->so_vnet); if (pr->pr_usrreqs->pru_flush != NULL) (*pr->pr_usrreqs->pru_flush)(so, how); if (how != SHUT_WR) sorflush(so); if (how != SHUT_RD) { error = (*pr->pr_usrreqs->pru_shutdown)(so); wakeup(&so->so_timeo); CURVNET_RESTORE(); return ((error == 0 && soerror_enotconn) ? ENOTCONN : error); } wakeup(&so->so_timeo); CURVNET_RESTORE(); return (soerror_enotconn ? ENOTCONN : 0); } void sorflush(struct socket *so) { struct sockbuf *sb = &so->so_rcv; struct protosw *pr = so->so_proto; struct socket aso; VNET_SO_ASSERT(so); /* * In order to avoid calling dom_dispose with the socket buffer mutex * held, and in order to generally avoid holding the lock for a long * time, we make a copy of the socket buffer and clear the original * (except locks, state). The new socket buffer copy won't have * initialized locks so we can only call routines that won't use or * assert those locks. * * Dislodge threads currently blocked in receive and wait to acquire * a lock against other simultaneous readers before clearing the * socket buffer. Don't let our acquire be interrupted by a signal * despite any existing socket disposition on interruptable waiting. */ socantrcvmore(so); (void) sblock(sb, SBL_WAIT | SBL_NOINTR); /* * Invalidate/clear most of the sockbuf structure, but leave selinfo * and mutex data unchanged. */ SOCKBUF_LOCK(sb); bzero(&aso, sizeof(aso)); aso.so_pcb = so->so_pcb; bcopy(&sb->sb_startzero, &aso.so_rcv.sb_startzero, sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); bzero(&sb->sb_startzero, sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); SOCKBUF_UNLOCK(sb); sbunlock(sb); /* * Dispose of special rights and flush the copied socket. Don't call * any unsafe routines (that rely on locks being initialized) on aso. */ if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose != NULL) (*pr->pr_domain->dom_dispose)(&aso); sbrelease_internal(&aso.so_rcv, so); } /* * Wrapper for Socket established helper hook. * Parameters: socket, context of the hook point, hook id. */ static int inline hhook_run_socket(struct socket *so, void *hctx, int32_t h_id) { struct socket_hhook_data hhook_data = { .so = so, .hctx = hctx, .m = NULL, .status = 0 }; CURVNET_SET(so->so_vnet); HHOOKS_RUN_IF(V_socket_hhh[h_id], &hhook_data, &so->osd); CURVNET_RESTORE(); /* Ugly but needed, since hhooks return void for now */ return (hhook_data.status); } /* * Perhaps this routine, and sooptcopyout(), below, ought to come in an * additional variant to handle the case where the option value needs to be * some kind of integer, but not a specific size. In addition to their use * here, these functions are also called by the protocol-level pr_ctloutput() * routines. */ int sooptcopyin(struct sockopt *sopt, void *buf, size_t len, size_t minlen) { size_t valsize; /* * If the user gives us more than we wanted, we ignore it, but if we * don't get the minimum length the caller wants, we return EINVAL. * On success, sopt->sopt_valsize is set to however much we actually * retrieved. */ if ((valsize = sopt->sopt_valsize) < minlen) return EINVAL; if (valsize > len) sopt->sopt_valsize = valsize = len; if (sopt->sopt_td != NULL) return (copyin(sopt->sopt_val, buf, valsize)); bcopy(sopt->sopt_val, buf, valsize); return (0); } /* * Kernel version of setsockopt(2). * * XXX: optlen is size_t, not socklen_t */ int so_setsockopt(struct socket *so, int level, int optname, void *optval, size_t optlen) { struct sockopt sopt; sopt.sopt_level = level; sopt.sopt_name = optname; sopt.sopt_dir = SOPT_SET; sopt.sopt_val = optval; sopt.sopt_valsize = optlen; sopt.sopt_td = NULL; return (sosetopt(so, &sopt)); } int sosetopt(struct socket *so, struct sockopt *sopt) { int error, optval; struct linger l; struct timeval tv; sbintime_t val; uint32_t val32; #ifdef MAC struct mac extmac; #endif CURVNET_SET(so->so_vnet); error = 0; if (sopt->sopt_level != SOL_SOCKET) { if (so->so_proto->pr_ctloutput != NULL) { error = (*so->so_proto->pr_ctloutput)(so, sopt); CURVNET_RESTORE(); return (error); } error = ENOPROTOOPT; } else { switch (sopt->sopt_name) { case SO_ACCEPTFILTER: error = accept_filt_setopt(so, sopt); if (error) goto bad; break; case SO_LINGER: error = sooptcopyin(sopt, &l, sizeof l, sizeof l); if (error) goto bad; SOCK_LOCK(so); so->so_linger = l.l_linger; if (l.l_onoff) so->so_options |= SO_LINGER; else so->so_options &= ~SO_LINGER; SOCK_UNLOCK(so); break; case SO_DEBUG: case SO_KEEPALIVE: case SO_DONTROUTE: case SO_USELOOPBACK: case SO_BROADCAST: case SO_REUSEADDR: case SO_REUSEPORT: + case SO_REUSEPORT_LB: case SO_OOBINLINE: case SO_TIMESTAMP: case SO_BINTIME: case SO_NOSIGPIPE: case SO_NO_DDP: case SO_NO_OFFLOAD: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) goto bad; SOCK_LOCK(so); if (optval) so->so_options |= sopt->sopt_name; else so->so_options &= ~sopt->sopt_name; SOCK_UNLOCK(so); break; case SO_SETFIB: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) goto bad; if (optval < 0 || optval >= rt_numfibs) { error = EINVAL; goto bad; } if (((so->so_proto->pr_domain->dom_family == PF_INET) || (so->so_proto->pr_domain->dom_family == PF_INET6) || (so->so_proto->pr_domain->dom_family == PF_ROUTE))) so->so_fibnum = optval; else so->so_fibnum = 0; break; case SO_USER_COOKIE: error = sooptcopyin(sopt, &val32, sizeof val32, sizeof val32); if (error) goto bad; so->so_user_cookie = val32; break; case SO_SNDBUF: case SO_RCVBUF: case SO_SNDLOWAT: case SO_RCVLOWAT: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) goto bad; /* * Values < 1 make no sense for any of these options, * so disallow them. */ if (optval < 1) { error = EINVAL; goto bad; } error = sbsetopt(so, sopt->sopt_name, optval); break; case SO_SNDTIMEO: case SO_RCVTIMEO: #ifdef COMPAT_FREEBSD32 if (SV_CURPROC_FLAG(SV_ILP32)) { struct timeval32 tv32; error = sooptcopyin(sopt, &tv32, sizeof tv32, sizeof tv32); CP(tv32, tv, tv_sec); CP(tv32, tv, tv_usec); } else #endif error = sooptcopyin(sopt, &tv, sizeof tv, sizeof tv); if (error) goto bad; if (tv.tv_sec < 0 || tv.tv_usec < 0 || tv.tv_usec >= 1000000) { error = EDOM; goto bad; } if (tv.tv_sec > INT32_MAX) val = SBT_MAX; else val = tvtosbt(tv); switch (sopt->sopt_name) { case SO_SNDTIMEO: so->so_snd.sb_timeo = val; break; case SO_RCVTIMEO: so->so_rcv.sb_timeo = val; break; } break; case SO_LABEL: #ifdef MAC error = sooptcopyin(sopt, &extmac, sizeof extmac, sizeof extmac); if (error) goto bad; error = mac_setsockopt_label(sopt->sopt_td->td_ucred, so, &extmac); #else error = EOPNOTSUPP; #endif break; case SO_TS_CLOCK: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) goto bad; if (optval < 0 || optval > SO_TS_CLOCK_MAX) { error = EINVAL; goto bad; } so->so_ts_clock = optval; break; case SO_MAX_PACING_RATE: error = sooptcopyin(sopt, &val32, sizeof(val32), sizeof(val32)); if (error) goto bad; so->so_max_pacing_rate = val32; break; default: if (V_socket_hhh[HHOOK_SOCKET_OPT]->hhh_nhooks > 0) error = hhook_run_socket(so, sopt, HHOOK_SOCKET_OPT); else error = ENOPROTOOPT; break; } if (error == 0 && so->so_proto->pr_ctloutput != NULL) (void)(*so->so_proto->pr_ctloutput)(so, sopt); } bad: CURVNET_RESTORE(); return (error); } /* * Helper routine for getsockopt. */ int sooptcopyout(struct sockopt *sopt, const void *buf, size_t len) { int error; size_t valsize; error = 0; /* * Documented get behavior is that we always return a value, possibly * truncated to fit in the user's buffer. Traditional behavior is * that we always tell the user precisely how much we copied, rather * than something useful like the total amount we had available for * her. Note that this interface is not idempotent; the entire * answer must be generated ahead of time. */ valsize = min(len, sopt->sopt_valsize); sopt->sopt_valsize = valsize; if (sopt->sopt_val != NULL) { if (sopt->sopt_td != NULL) error = copyout(buf, sopt->sopt_val, valsize); else bcopy(buf, sopt->sopt_val, valsize); } return (error); } int sogetopt(struct socket *so, struct sockopt *sopt) { int error, optval; struct linger l; struct timeval tv; #ifdef MAC struct mac extmac; #endif CURVNET_SET(so->so_vnet); error = 0; if (sopt->sopt_level != SOL_SOCKET) { if (so->so_proto->pr_ctloutput != NULL) error = (*so->so_proto->pr_ctloutput)(so, sopt); else error = ENOPROTOOPT; CURVNET_RESTORE(); return (error); } else { switch (sopt->sopt_name) { case SO_ACCEPTFILTER: error = accept_filt_getopt(so, sopt); break; case SO_LINGER: SOCK_LOCK(so); l.l_onoff = so->so_options & SO_LINGER; l.l_linger = so->so_linger; SOCK_UNLOCK(so); error = sooptcopyout(sopt, &l, sizeof l); break; case SO_USELOOPBACK: case SO_DONTROUTE: case SO_DEBUG: case SO_KEEPALIVE: case SO_REUSEADDR: case SO_REUSEPORT: + case SO_REUSEPORT_LB: case SO_BROADCAST: case SO_OOBINLINE: case SO_ACCEPTCONN: case SO_TIMESTAMP: case SO_BINTIME: case SO_NOSIGPIPE: optval = so->so_options & sopt->sopt_name; integer: error = sooptcopyout(sopt, &optval, sizeof optval); break; case SO_TYPE: optval = so->so_type; goto integer; case SO_PROTOCOL: optval = so->so_proto->pr_protocol; goto integer; case SO_ERROR: SOCK_LOCK(so); optval = so->so_error; so->so_error = 0; SOCK_UNLOCK(so); goto integer; case SO_SNDBUF: optval = SOLISTENING(so) ? so->sol_sbsnd_hiwat : so->so_snd.sb_hiwat; goto integer; case SO_RCVBUF: optval = SOLISTENING(so) ? so->sol_sbrcv_hiwat : so->so_rcv.sb_hiwat; goto integer; case SO_SNDLOWAT: optval = SOLISTENING(so) ? so->sol_sbsnd_lowat : so->so_snd.sb_lowat; goto integer; case SO_RCVLOWAT: optval = SOLISTENING(so) ? so->sol_sbrcv_lowat : so->so_rcv.sb_lowat; goto integer; case SO_SNDTIMEO: case SO_RCVTIMEO: tv = sbttotv(sopt->sopt_name == SO_SNDTIMEO ? so->so_snd.sb_timeo : so->so_rcv.sb_timeo); #ifdef COMPAT_FREEBSD32 if (SV_CURPROC_FLAG(SV_ILP32)) { struct timeval32 tv32; CP(tv, tv32, tv_sec); CP(tv, tv32, tv_usec); error = sooptcopyout(sopt, &tv32, sizeof tv32); } else #endif error = sooptcopyout(sopt, &tv, sizeof tv); break; case SO_LABEL: #ifdef MAC error = sooptcopyin(sopt, &extmac, sizeof(extmac), sizeof(extmac)); if (error) goto bad; error = mac_getsockopt_label(sopt->sopt_td->td_ucred, so, &extmac); if (error) goto bad; error = sooptcopyout(sopt, &extmac, sizeof extmac); #else error = EOPNOTSUPP; #endif break; case SO_PEERLABEL: #ifdef MAC error = sooptcopyin(sopt, &extmac, sizeof(extmac), sizeof(extmac)); if (error) goto bad; error = mac_getsockopt_peerlabel( sopt->sopt_td->td_ucred, so, &extmac); if (error) goto bad; error = sooptcopyout(sopt, &extmac, sizeof extmac); #else error = EOPNOTSUPP; #endif break; case SO_LISTENQLIMIT: optval = SOLISTENING(so) ? so->sol_qlimit : 0; goto integer; case SO_LISTENQLEN: optval = SOLISTENING(so) ? so->sol_qlen : 0; goto integer; case SO_LISTENINCQLEN: optval = SOLISTENING(so) ? so->sol_incqlen : 0; goto integer; case SO_TS_CLOCK: optval = so->so_ts_clock; goto integer; case SO_MAX_PACING_RATE: optval = so->so_max_pacing_rate; goto integer; default: if (V_socket_hhh[HHOOK_SOCKET_OPT]->hhh_nhooks > 0) error = hhook_run_socket(so, sopt, HHOOK_SOCKET_OPT); else error = ENOPROTOOPT; break; } } #ifdef MAC bad: #endif CURVNET_RESTORE(); return (error); } int soopt_getm(struct sockopt *sopt, struct mbuf **mp) { struct mbuf *m, *m_prev; int sopt_size = sopt->sopt_valsize; MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); if (m == NULL) return ENOBUFS; if (sopt_size > MLEN) { MCLGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT); if ((m->m_flags & M_EXT) == 0) { m_free(m); return ENOBUFS; } m->m_len = min(MCLBYTES, sopt_size); } else { m->m_len = min(MLEN, sopt_size); } sopt_size -= m->m_len; *mp = m; m_prev = m; while (sopt_size) { MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); if (m == NULL) { m_freem(*mp); return ENOBUFS; } if (sopt_size > MLEN) { MCLGET(m, sopt->sopt_td != NULL ? M_WAITOK : M_NOWAIT); if ((m->m_flags & M_EXT) == 0) { m_freem(m); m_freem(*mp); return ENOBUFS; } m->m_len = min(MCLBYTES, sopt_size); } else { m->m_len = min(MLEN, sopt_size); } sopt_size -= m->m_len; m_prev->m_next = m; m_prev = m; } return (0); } int soopt_mcopyin(struct sockopt *sopt, struct mbuf *m) { struct mbuf *m0 = m; if (sopt->sopt_val == NULL) return (0); while (m != NULL && sopt->sopt_valsize >= m->m_len) { if (sopt->sopt_td != NULL) { int error; error = copyin(sopt->sopt_val, mtod(m, char *), m->m_len); if (error != 0) { m_freem(m0); return(error); } } else bcopy(sopt->sopt_val, mtod(m, char *), m->m_len); sopt->sopt_valsize -= m->m_len; sopt->sopt_val = (char *)sopt->sopt_val + m->m_len; m = m->m_next; } if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */ panic("ip6_sooptmcopyin"); return (0); } int soopt_mcopyout(struct sockopt *sopt, struct mbuf *m) { struct mbuf *m0 = m; size_t valsize = 0; if (sopt->sopt_val == NULL) return (0); while (m != NULL && sopt->sopt_valsize >= m->m_len) { if (sopt->sopt_td != NULL) { int error; error = copyout(mtod(m, char *), sopt->sopt_val, m->m_len); if (error != 0) { m_freem(m0); return(error); } } else bcopy(mtod(m, char *), sopt->sopt_val, m->m_len); sopt->sopt_valsize -= m->m_len; sopt->sopt_val = (char *)sopt->sopt_val + m->m_len; valsize += m->m_len; m = m->m_next; } if (m != NULL) { /* enough soopt buffer should be given from user-land */ m_freem(m0); return(EINVAL); } sopt->sopt_valsize = valsize; return (0); } /* * sohasoutofband(): protocol notifies socket layer of the arrival of new * out-of-band data, which will then notify socket consumers. */ void sohasoutofband(struct socket *so) { if (so->so_sigio != NULL) pgsigio(&so->so_sigio, SIGURG, 0); selwakeuppri(&so->so_rdsel, PSOCK); } int sopoll(struct socket *so, int events, struct ucred *active_cred, struct thread *td) { /* * We do not need to set or assert curvnet as long as everyone uses * sopoll_generic(). */ return (so->so_proto->pr_usrreqs->pru_sopoll(so, events, active_cred, td)); } int sopoll_generic(struct socket *so, int events, struct ucred *active_cred, struct thread *td) { int revents; SOCK_LOCK(so); if (SOLISTENING(so)) { if (!(events & (POLLIN | POLLRDNORM))) revents = 0; else if (!TAILQ_EMPTY(&so->sol_comp)) revents = events & (POLLIN | POLLRDNORM); else { selrecord(td, &so->so_rdsel); revents = 0; } } else { revents = 0; SOCKBUF_LOCK(&so->so_snd); SOCKBUF_LOCK(&so->so_rcv); if (events & (POLLIN | POLLRDNORM)) if (soreadabledata(so)) revents |= events & (POLLIN | POLLRDNORM); if (events & (POLLOUT | POLLWRNORM)) if (sowriteable(so)) revents |= events & (POLLOUT | POLLWRNORM); if (events & (POLLPRI | POLLRDBAND)) if (so->so_oobmark || (so->so_rcv.sb_state & SBS_RCVATMARK)) revents |= events & (POLLPRI | POLLRDBAND); if ((events & POLLINIGNEOF) == 0) { if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { revents |= events & (POLLIN | POLLRDNORM); if (so->so_snd.sb_state & SBS_CANTSENDMORE) revents |= POLLHUP; } } if (revents == 0) { if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { selrecord(td, &so->so_rdsel); so->so_rcv.sb_flags |= SB_SEL; } if (events & (POLLOUT | POLLWRNORM)) { selrecord(td, &so->so_wrsel); so->so_snd.sb_flags |= SB_SEL; } } SOCKBUF_UNLOCK(&so->so_rcv); SOCKBUF_UNLOCK(&so->so_snd); } SOCK_UNLOCK(so); return (revents); } int soo_kqfilter(struct file *fp, struct knote *kn) { struct socket *so = kn->kn_fp->f_data; struct sockbuf *sb; struct knlist *knl; switch (kn->kn_filter) { case EVFILT_READ: kn->kn_fop = &soread_filtops; knl = &so->so_rdsel.si_note; sb = &so->so_rcv; break; case EVFILT_WRITE: kn->kn_fop = &sowrite_filtops; knl = &so->so_wrsel.si_note; sb = &so->so_snd; break; case EVFILT_EMPTY: kn->kn_fop = &soempty_filtops; knl = &so->so_wrsel.si_note; sb = &so->so_snd; break; default: return (EINVAL); } SOCK_LOCK(so); if (SOLISTENING(so)) { knlist_add(knl, kn, 1); } else { SOCKBUF_LOCK(sb); knlist_add(knl, kn, 1); sb->sb_flags |= SB_KNOTE; SOCKBUF_UNLOCK(sb); } SOCK_UNLOCK(so); return (0); } /* * Some routines that return EOPNOTSUPP for entry points that are not * supported by a protocol. Fill in as needed. */ int pru_accept_notsupp(struct socket *so, struct sockaddr **nam) { return EOPNOTSUPP; } int pru_aio_queue_notsupp(struct socket *so, struct kaiocb *job) { return EOPNOTSUPP; } int pru_attach_notsupp(struct socket *so, int proto, struct thread *td) { return EOPNOTSUPP; } int pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td) { return EOPNOTSUPP; } int pru_bindat_notsupp(int fd, struct socket *so, struct sockaddr *nam, struct thread *td) { return EOPNOTSUPP; } int pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td) { return EOPNOTSUPP; } int pru_connectat_notsupp(int fd, struct socket *so, struct sockaddr *nam, struct thread *td) { return EOPNOTSUPP; } int pru_connect2_notsupp(struct socket *so1, struct socket *so2) { return EOPNOTSUPP; } int pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td) { return EOPNOTSUPP; } int pru_disconnect_notsupp(struct socket *so) { return EOPNOTSUPP; } int pru_listen_notsupp(struct socket *so, int backlog, struct thread *td) { return EOPNOTSUPP; } int pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam) { return EOPNOTSUPP; } int pru_rcvd_notsupp(struct socket *so, int flags) { return EOPNOTSUPP; } int pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags) { return EOPNOTSUPP; } int pru_send_notsupp(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { return EOPNOTSUPP; } int pru_ready_notsupp(struct socket *so, struct mbuf *m, int count) { return (EOPNOTSUPP); } /* * This isn't really a ``null'' operation, but it's the default one and * doesn't do anything destructive. */ int pru_sense_null(struct socket *so, struct stat *sb) { sb->st_blksize = so->so_snd.sb_hiwat; return 0; } int pru_shutdown_notsupp(struct socket *so) { return EOPNOTSUPP; } int pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam) { return EOPNOTSUPP; } int pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td) { return EOPNOTSUPP; } int pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { return EOPNOTSUPP; } int pru_sopoll_notsupp(struct socket *so, int events, struct ucred *cred, struct thread *td) { return EOPNOTSUPP; } static void filt_sordetach(struct knote *kn) { struct socket *so = kn->kn_fp->f_data; so_rdknl_lock(so); knlist_remove(&so->so_rdsel.si_note, kn, 1); if (!SOLISTENING(so) && knlist_empty(&so->so_rdsel.si_note)) so->so_rcv.sb_flags &= ~SB_KNOTE; so_rdknl_unlock(so); } /*ARGSUSED*/ static int filt_soread(struct knote *kn, long hint) { struct socket *so; so = kn->kn_fp->f_data; if (SOLISTENING(so)) { SOCK_LOCK_ASSERT(so); kn->kn_data = so->sol_qlen; return (!TAILQ_EMPTY(&so->sol_comp)); } SOCKBUF_LOCK_ASSERT(&so->so_rcv); kn->kn_data = sbavail(&so->so_rcv) - so->so_rcv.sb_ctl; if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { kn->kn_flags |= EV_EOF; kn->kn_fflags = so->so_error; return (1); } else if (so->so_error) /* temporary udp error */ return (1); if (kn->kn_sfflags & NOTE_LOWAT) { if (kn->kn_data >= kn->kn_sdata) return (1); } else if (sbavail(&so->so_rcv) >= so->so_rcv.sb_lowat) return (1); /* This hook returning non-zero indicates an event, not error */ return (hhook_run_socket(so, NULL, HHOOK_FILT_SOREAD)); } static void filt_sowdetach(struct knote *kn) { struct socket *so = kn->kn_fp->f_data; so_wrknl_lock(so); knlist_remove(&so->so_wrsel.si_note, kn, 1); if (!SOLISTENING(so) && knlist_empty(&so->so_wrsel.si_note)) so->so_snd.sb_flags &= ~SB_KNOTE; so_wrknl_unlock(so); } /*ARGSUSED*/ static int filt_sowrite(struct knote *kn, long hint) { struct socket *so; so = kn->kn_fp->f_data; if (SOLISTENING(so)) return (0); SOCKBUF_LOCK_ASSERT(&so->so_snd); kn->kn_data = sbspace(&so->so_snd); hhook_run_socket(so, kn, HHOOK_FILT_SOWRITE); if (so->so_snd.sb_state & SBS_CANTSENDMORE) { kn->kn_flags |= EV_EOF; kn->kn_fflags = so->so_error; return (1); } else if (so->so_error) /* temporary udp error */ return (1); else if (((so->so_state & SS_ISCONNECTED) == 0) && (so->so_proto->pr_flags & PR_CONNREQUIRED)) return (0); else if (kn->kn_sfflags & NOTE_LOWAT) return (kn->kn_data >= kn->kn_sdata); else return (kn->kn_data >= so->so_snd.sb_lowat); } static int filt_soempty(struct knote *kn, long hint) { struct socket *so; so = kn->kn_fp->f_data; if (SOLISTENING(so)) return (1); SOCKBUF_LOCK_ASSERT(&so->so_snd); kn->kn_data = sbused(&so->so_snd); if (kn->kn_data == 0) return (1); else return (0); } int socheckuid(struct socket *so, uid_t uid) { if (so == NULL) return (EPERM); if (so->so_cred->cr_uid != uid) return (EPERM); return (0); } /* * These functions are used by protocols to notify the socket layer (and its * consumers) of state changes in the sockets driven by protocol-side events. */ /* * Procedures to manipulate state flags of socket and do appropriate wakeups. * * Normal sequence from the active (originating) side is that * soisconnecting() is called during processing of connect() call, resulting * in an eventual call to soisconnected() if/when the connection is * established. When the connection is torn down soisdisconnecting() is * called during processing of disconnect() call, and soisdisconnected() is * called when the connection to the peer is totally severed. The semantics * of these routines are such that connectionless protocols can call * soisconnected() and soisdisconnected() only, bypassing the in-progress * calls when setting up a ``connection'' takes no time. * * From the passive side, a socket is created with two queues of sockets: * so_incomp for connections in progress and so_comp for connections already * made and awaiting user acceptance. As a protocol is preparing incoming * connections, it creates a socket structure queued on so_incomp by calling * sonewconn(). When the connection is established, soisconnected() is * called, and transfers the socket structure to so_comp, making it available * to accept(). * * If a socket is closed with sockets on either so_incomp or so_comp, these * sockets are dropped. * * If higher-level protocols are implemented in the kernel, the wakeups done * here will sometimes cause software-interrupt process scheduling. */ void soisconnecting(struct socket *so) { SOCK_LOCK(so); so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); so->so_state |= SS_ISCONNECTING; SOCK_UNLOCK(so); } void soisconnected(struct socket *so) { SOCK_LOCK(so); so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); so->so_state |= SS_ISCONNECTED; if (so->so_qstate == SQ_INCOMP) { struct socket *head = so->so_listen; int ret; KASSERT(head, ("%s: so %p on incomp of NULL", __func__, so)); /* * Promoting a socket from incomplete queue to complete, we * need to go through reverse order of locking. We first do * trylock, and if that doesn't succeed, we go the hard way * leaving a reference and rechecking consistency after proper * locking. */ if (__predict_false(SOLISTEN_TRYLOCK(head) == 0)) { soref(head); SOCK_UNLOCK(so); SOLISTEN_LOCK(head); SOCK_LOCK(so); if (__predict_false(head != so->so_listen)) { /* * The socket went off the listen queue, * should be lost race to close(2) of sol. * The socket is about to soabort(). */ SOCK_UNLOCK(so); sorele(head); return; } /* Not the last one, as so holds a ref. */ refcount_release(&head->so_count); } again: if ((so->so_options & SO_ACCEPTFILTER) == 0) { TAILQ_REMOVE(&head->sol_incomp, so, so_list); head->sol_incqlen--; TAILQ_INSERT_TAIL(&head->sol_comp, so, so_list); head->sol_qlen++; so->so_qstate = SQ_COMP; SOCK_UNLOCK(so); solisten_wakeup(head); /* unlocks */ } else { SOCKBUF_LOCK(&so->so_rcv); soupcall_set(so, SO_RCV, head->sol_accept_filter->accf_callback, head->sol_accept_filter_arg); so->so_options &= ~SO_ACCEPTFILTER; ret = head->sol_accept_filter->accf_callback(so, head->sol_accept_filter_arg, M_NOWAIT); if (ret == SU_ISCONNECTED) { soupcall_clear(so, SO_RCV); SOCKBUF_UNLOCK(&so->so_rcv); goto again; } SOCKBUF_UNLOCK(&so->so_rcv); SOCK_UNLOCK(so); SOLISTEN_UNLOCK(head); } return; } SOCK_UNLOCK(so); wakeup(&so->so_timeo); sorwakeup(so); sowwakeup(so); } void soisdisconnecting(struct socket *so) { SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTING; so->so_state |= SS_ISDISCONNECTING; if (!SOLISTENING(so)) { SOCKBUF_LOCK(&so->so_rcv); socantrcvmore_locked(so); SOCKBUF_LOCK(&so->so_snd); socantsendmore_locked(so); } SOCK_UNLOCK(so); wakeup(&so->so_timeo); } void soisdisconnected(struct socket *so) { SOCK_LOCK(so); so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); so->so_state |= SS_ISDISCONNECTED; if (!SOLISTENING(so)) { SOCK_UNLOCK(so); SOCKBUF_LOCK(&so->so_rcv); socantrcvmore_locked(so); SOCKBUF_LOCK(&so->so_snd); sbdrop_locked(&so->so_snd, sbused(&so->so_snd)); socantsendmore_locked(so); } else SOCK_UNLOCK(so); wakeup(&so->so_timeo); } /* * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. */ struct sockaddr * sodupsockaddr(const struct sockaddr *sa, int mflags) { struct sockaddr *sa2; sa2 = malloc(sa->sa_len, M_SONAME, mflags); if (sa2) bcopy(sa, sa2, sa->sa_len); return sa2; } /* * Register per-socket buffer upcalls. */ void soupcall_set(struct socket *so, int which, so_upcall_t func, void *arg) { struct sockbuf *sb; KASSERT(!SOLISTENING(so), ("%s: so %p listening", __func__, so)); switch (which) { case SO_RCV: sb = &so->so_rcv; break; case SO_SND: sb = &so->so_snd; break; default: panic("soupcall_set: bad which"); } SOCKBUF_LOCK_ASSERT(sb); sb->sb_upcall = func; sb->sb_upcallarg = arg; sb->sb_flags |= SB_UPCALL; } void soupcall_clear(struct socket *so, int which) { struct sockbuf *sb; KASSERT(!SOLISTENING(so), ("%s: so %p listening", __func__, so)); switch (which) { case SO_RCV: sb = &so->so_rcv; break; case SO_SND: sb = &so->so_snd; break; default: panic("soupcall_clear: bad which"); } SOCKBUF_LOCK_ASSERT(sb); KASSERT(sb->sb_upcall != NULL, ("%s: so %p no upcall to clear", __func__, so)); sb->sb_upcall = NULL; sb->sb_upcallarg = NULL; sb->sb_flags &= ~SB_UPCALL; } void solisten_upcall_set(struct socket *so, so_upcall_t func, void *arg) { SOLISTEN_LOCK_ASSERT(so); so->sol_upcall = func; so->sol_upcallarg = arg; } static void so_rdknl_lock(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_LOCK(so); else SOCKBUF_LOCK(&so->so_rcv); } static void so_rdknl_unlock(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_UNLOCK(so); else SOCKBUF_UNLOCK(&so->so_rcv); } static void so_rdknl_assert_locked(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_LOCK_ASSERT(so); else SOCKBUF_LOCK_ASSERT(&so->so_rcv); } static void so_rdknl_assert_unlocked(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_UNLOCK_ASSERT(so); else SOCKBUF_UNLOCK_ASSERT(&so->so_rcv); } static void so_wrknl_lock(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_LOCK(so); else SOCKBUF_LOCK(&so->so_snd); } static void so_wrknl_unlock(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_UNLOCK(so); else SOCKBUF_UNLOCK(&so->so_snd); } static void so_wrknl_assert_locked(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_LOCK_ASSERT(so); else SOCKBUF_LOCK_ASSERT(&so->so_snd); } static void so_wrknl_assert_unlocked(void *arg) { struct socket *so = arg; if (SOLISTENING(so)) SOCK_UNLOCK_ASSERT(so); else SOCKBUF_UNLOCK_ASSERT(&so->so_snd); } /* * Create an external-format (``xsocket'') structure using the information in * the kernel-format socket structure pointed to by so. This is done to * reduce the spew of irrelevant information over this interface, to isolate * user code from changes in the kernel structure, and potentially to provide * information-hiding if we decide that some of this information should be * hidden from users. */ void sotoxsocket(struct socket *so, struct xsocket *xso) { xso->xso_len = sizeof *xso; xso->xso_so = so; xso->so_type = so->so_type; xso->so_options = so->so_options; xso->so_linger = so->so_linger; xso->so_state = so->so_state; xso->so_pcb = so->so_pcb; xso->xso_protocol = so->so_proto->pr_protocol; xso->xso_family = so->so_proto->pr_domain->dom_family; xso->so_timeo = so->so_timeo; xso->so_error = so->so_error; xso->so_uid = so->so_cred->cr_uid; xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0; if (SOLISTENING(so)) { xso->so_qlen = so->sol_qlen; xso->so_incqlen = so->sol_incqlen; xso->so_qlimit = so->sol_qlimit; xso->so_oobmark = 0; bzero(&xso->so_snd, sizeof(xso->so_snd)); bzero(&xso->so_rcv, sizeof(xso->so_rcv)); } else { xso->so_state |= so->so_qstate; xso->so_qlen = xso->so_incqlen = xso->so_qlimit = 0; xso->so_oobmark = so->so_oobmark; sbtoxsockbuf(&so->so_snd, &xso->so_snd); sbtoxsockbuf(&so->so_rcv, &xso->so_rcv); } } struct sockbuf * so_sockbuf_rcv(struct socket *so) { return (&so->so_rcv); } struct sockbuf * so_sockbuf_snd(struct socket *so) { return (&so->so_snd); } int so_state_get(const struct socket *so) { return (so->so_state); } void so_state_set(struct socket *so, int val) { so->so_state = val; } int so_options_get(const struct socket *so) { return (so->so_options); } void so_options_set(struct socket *so, int val) { so->so_options = val; } int so_error_get(const struct socket *so) { return (so->so_error); } void so_error_set(struct socket *so, int val) { so->so_error = val; } int so_linger_get(const struct socket *so) { return (so->so_linger); } void so_linger_set(struct socket *so, int val) { so->so_linger = val; } struct protosw * so_protosw_get(const struct socket *so) { return (so->so_proto); } void so_protosw_set(struct socket *so, struct protosw *val) { so->so_proto = val; } void so_sorwakeup(struct socket *so) { sorwakeup(so); } void so_sowwakeup(struct socket *so) { sowwakeup(so); } void so_sorwakeup_locked(struct socket *so) { sorwakeup_locked(so); } void so_sowwakeup_locked(struct socket *so) { sowwakeup_locked(so); } void so_lock(struct socket *so) { SOCK_LOCK(so); } void so_unlock(struct socket *so) { SOCK_UNLOCK(so); } Index: head/sys/netinet/in_pcb.c =================================================================== --- head/sys/netinet/in_pcb.c (revision 334718) +++ head/sys/netinet/in_pcb.c (revision 334719) @@ -1,3129 +1,3449 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1991, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2007-2009 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, 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. * 3. 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. * * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_ipsec.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ratelimit.h" #include "opt_pcbgroup.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #include #include #include #include #if defined(INET) || defined(INET6) #include #include #include #include #ifdef TCPHPTS #include #endif #include #include #endif #ifdef INET #include #endif #ifdef INET6 #include #include #include #include #endif /* INET6 */ #include #include +#define INPCBLBGROUP_SIZMIN 8 +#define INPCBLBGROUP_SIZMAX 256 + static struct callout ipport_tick_callout; /* * These configure the range of local port addresses assigned to * "unspecified" outgoing connections/packets/whatever. */ VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */ VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */ VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */ VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */ VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */ /* * Reserved ports accessible only to root. There are significant * security considerations that must be accounted for when changing these, * but the security benefits can be great. Please be careful. */ VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_reservedlow); /* Variables dealing with random ephemeral port allocation. */ VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */ VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */ VNET_DEFINE(int, ipport_tcpallocs); static VNET_DEFINE(int, ipport_tcplastcount); #define V_ipport_tcplastcount VNET(ipport_tcplastcount) static void in_pcbremlists(struct inpcb *inp); #ifdef INET static struct inpcb *in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp); #define RANGECHK(var, min, max) \ if ((var) < (min)) { (var) = (min); } \ else if ((var) > (max)) { (var) = (max); } static int sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) { int error; error = sysctl_handle_int(oidp, arg1, arg2, req); if (error == 0) { RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); } return (error); } #undef RANGECHK static SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports"); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipport_randomized), 0, "Enable random port allocation"); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port " "allocations before switching to a sequental one"); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipport_randomtime), 0, "Minimum time to keep sequental port " "allocation before switching to a random one"); #endif /* INET */ /* * in_pcb.c: manage the Protocol Control Blocks. * * NOTE: It is assumed that most of these functions will be called with * the pcbinfo lock held, and often, the inpcb lock held, as these utility * functions often modify hash chains or addresses in pcbs. */ +static struct inpcblbgroup * +in_pcblbgroup_alloc(struct inpcblbgrouphead *hdr, u_char vflag, + uint16_t port, const union in_dependaddr *addr, int size) +{ + struct inpcblbgroup *grp; + size_t bytes; + + bytes = __offsetof(struct inpcblbgroup, il_inp[size]); + grp = malloc(bytes, M_PCB, M_ZERO | M_NOWAIT); + if (!grp) + return (NULL); + grp->il_vflag = vflag; + grp->il_lport = port; + grp->il_dependladdr = *addr; + grp->il_inpsiz = size; + LIST_INSERT_HEAD(hdr, grp, il_list); + return (grp); +} + +static void +in_pcblbgroup_free(struct inpcblbgroup *grp) +{ + + LIST_REMOVE(grp, il_list); + free(grp, M_TEMP); +} + +static struct inpcblbgroup * +in_pcblbgroup_resize(struct inpcblbgrouphead *hdr, + struct inpcblbgroup *old_grp, int size) +{ + struct inpcblbgroup *grp; + int i; + + grp = in_pcblbgroup_alloc(hdr, old_grp->il_vflag, + old_grp->il_lport, &old_grp->il_dependladdr, size); + if (!grp) + return (NULL); + + KASSERT(old_grp->il_inpcnt < grp->il_inpsiz, + ("invalid new local group size %d and old local group count %d", + grp->il_inpsiz, old_grp->il_inpcnt)); + + for (i = 0; i < old_grp->il_inpcnt; ++i) + grp->il_inp[i] = old_grp->il_inp[i]; + grp->il_inpcnt = old_grp->il_inpcnt; + in_pcblbgroup_free(old_grp); + return (grp); +} + /* + * PCB at index 'i' is removed from the group. Pull up the ones below il_inp[i] + * and shrink group if possible. + */ +static void +in_pcblbgroup_reorder(struct inpcblbgrouphead *hdr, struct inpcblbgroup **grpp, + int i) +{ + struct inpcblbgroup *grp = *grpp; + + for (; i + 1 < grp->il_inpcnt; ++i) + grp->il_inp[i] = grp->il_inp[i + 1]; + grp->il_inpcnt--; + + if (grp->il_inpsiz > INPCBLBGROUP_SIZMIN && + grp->il_inpcnt <= (grp->il_inpsiz / 4)) { + /* Shrink this group. */ + struct inpcblbgroup *new_grp = + in_pcblbgroup_resize(hdr, grp, grp->il_inpsiz / 2); + if (new_grp) + *grpp = new_grp; + } + return; +} + +/* + * Add PCB to load balance group for SO_REUSEPORT_LB option. + */ +static int +in_pcbinslbgrouphash(struct inpcb *inp) +{ + struct inpcbinfo *pcbinfo; + struct inpcblbgrouphead *hdr; + struct inpcblbgroup *grp; + uint16_t hashmask, lport; + uint32_t group_index; + struct ucred *cred; + static int limit_logged = 0; + + pcbinfo = inp->inp_pcbinfo; + + INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(pcbinfo); + + if (pcbinfo->ipi_lbgrouphashbase == NULL) + return (0); + + hashmask = pcbinfo->ipi_lbgrouphashmask; + lport = inp->inp_lport; + group_index = INP_PCBLBGROUP_PORTHASH(lport, hashmask); + hdr = &pcbinfo->ipi_lbgrouphashbase[group_index]; + + /* + * Don't allow jailed socket to join local group. + */ + if (inp->inp_socket != NULL) + cred = inp->inp_socket->so_cred; + else + cred = NULL; + if (cred != NULL && jailed(cred)) + return (0); + +#ifdef INET6 + /* + * Don't allow IPv4 mapped INET6 wild socket. + */ + if ((inp->inp_vflag & INP_IPV4) && + inp->inp_laddr.s_addr == INADDR_ANY && + INP_CHECK_SOCKAF(inp->inp_socket, AF_INET6)) { + return (0); + } +#endif + + hdr = &pcbinfo->ipi_lbgrouphashbase[ + INP_PCBLBGROUP_PORTHASH(inp->inp_lport, + pcbinfo->ipi_lbgrouphashmask)]; + LIST_FOREACH(grp, hdr, il_list) { + if (grp->il_vflag == inp->inp_vflag && + grp->il_lport == inp->inp_lport && + memcmp(&grp->il_dependladdr, + &inp->inp_inc.inc_ie.ie_dependladdr, + sizeof(grp->il_dependladdr)) == 0) { + break; + } + } + if (grp == NULL) { + /* Create new load balance group. */ + grp = in_pcblbgroup_alloc(hdr, inp->inp_vflag, + inp->inp_lport, &inp->inp_inc.inc_ie.ie_dependladdr, + INPCBLBGROUP_SIZMIN); + if (!grp) + return (ENOBUFS); + } else if (grp->il_inpcnt == grp->il_inpsiz) { + if (grp->il_inpsiz >= INPCBLBGROUP_SIZMAX) { + if (!limit_logged) { + limit_logged = 1; + printf("lb group port %d, limit reached\n", + ntohs(grp->il_lport)); + } + return (0); + } + + /* Expand this local group. */ + grp = in_pcblbgroup_resize(hdr, grp, grp->il_inpsiz * 2); + if (!grp) + return (ENOBUFS); + } + + KASSERT(grp->il_inpcnt < grp->il_inpsiz, + ("invalid local group size %d and count %d", + grp->il_inpsiz, grp->il_inpcnt)); + + grp->il_inp[grp->il_inpcnt] = inp; + grp->il_inpcnt++; + return (0); +} + +/* + * Remove PCB from load balance group. + */ +static void +in_pcbremlbgrouphash(struct inpcb *inp) +{ + struct inpcbinfo *pcbinfo; + struct inpcblbgrouphead *hdr; + struct inpcblbgroup *grp; + int i; + + pcbinfo = inp->inp_pcbinfo; + + INP_WLOCK_ASSERT(inp); + INP_HASH_WLOCK_ASSERT(pcbinfo); + + if (pcbinfo->ipi_lbgrouphashbase == NULL) + return; + + hdr = &pcbinfo->ipi_lbgrouphashbase[ + INP_PCBLBGROUP_PORTHASH(inp->inp_lport, + pcbinfo->ipi_lbgrouphashmask)]; + + LIST_FOREACH(grp, hdr, il_list) { + for (i = 0; i < grp->il_inpcnt; ++i) { + if (grp->il_inp[i] != inp) + continue; + + if (grp->il_inpcnt == 1) { + /* We are the last, free this local group. */ + in_pcblbgroup_free(grp); + } else { + /* Pull up inpcbs, shrink group if possible. */ + in_pcblbgroup_reorder(hdr, &grp, i); + } + return; + } + } +} + +/* * Different protocols initialize their inpcbs differently - giving * different name to the lock. But they all are disposed the same. */ static void inpcb_fini(void *mem, int size) { struct inpcb *inp = mem; INP_LOCK_DESTROY(inp); } /* * Initialize an inpcbinfo -- we should be able to reduce the number of * arguments in time. */ void in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name, struct inpcbhead *listhead, int hash_nelements, int porthash_nelements, char *inpcbzone_name, uma_init inpcbzone_init, u_int hashfields) { INP_INFO_LOCK_INIT(pcbinfo, name); INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash"); /* XXXRW: argument? */ INP_LIST_LOCK_INIT(pcbinfo, "pcbinfolist"); #ifdef VIMAGE pcbinfo->ipi_vnet = curvnet; #endif pcbinfo->ipi_listhead = listhead; LIST_INIT(pcbinfo->ipi_listhead); pcbinfo->ipi_count = 0; pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB, &pcbinfo->ipi_hashmask); pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB, &pcbinfo->ipi_porthashmask); + pcbinfo->ipi_lbgrouphashbase = hashinit(hash_nelements, M_PCB, + &pcbinfo->ipi_lbgrouphashmask); #ifdef PCBGROUP in_pcbgroup_init(pcbinfo, hashfields, hash_nelements); #endif pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb), NULL, NULL, inpcbzone_init, inpcb_fini, UMA_ALIGN_PTR, 0); uma_zone_set_max(pcbinfo->ipi_zone, maxsockets); uma_zone_set_warning(pcbinfo->ipi_zone, "kern.ipc.maxsockets limit reached"); } /* * Destroy an inpcbinfo. */ void in_pcbinfo_destroy(struct inpcbinfo *pcbinfo) { KASSERT(pcbinfo->ipi_count == 0, ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count)); hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask); hashdestroy(pcbinfo->ipi_porthashbase, M_PCB, pcbinfo->ipi_porthashmask); + hashdestroy(pcbinfo->ipi_lbgrouphashbase, M_PCB, + pcbinfo->ipi_lbgrouphashmask); #ifdef PCBGROUP in_pcbgroup_destroy(pcbinfo); #endif uma_zdestroy(pcbinfo->ipi_zone); INP_LIST_LOCK_DESTROY(pcbinfo); INP_HASH_LOCK_DESTROY(pcbinfo); INP_INFO_LOCK_DESTROY(pcbinfo); } /* * Allocate a PCB and associate it with the socket. * On success return with the PCB locked. */ int in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo) { struct inpcb *inp; int error; #ifdef INVARIANTS if (pcbinfo == &V_tcbinfo) { INP_INFO_RLOCK_ASSERT(pcbinfo); } else { INP_INFO_WLOCK_ASSERT(pcbinfo); } #endif error = 0; inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT); if (inp == NULL) return (ENOBUFS); bzero(&inp->inp_start_zero, inp_zero_size); inp->inp_pcbinfo = pcbinfo; inp->inp_socket = so; inp->inp_cred = crhold(so->so_cred); inp->inp_inc.inc_fibnum = so->so_fibnum; #ifdef MAC error = mac_inpcb_init(inp, M_NOWAIT); if (error != 0) goto out; mac_inpcb_create(so, inp); #endif #if defined(IPSEC) || defined(IPSEC_SUPPORT) error = ipsec_init_pcbpolicy(inp); if (error != 0) { #ifdef MAC mac_inpcb_destroy(inp); #endif goto out; } #endif /*IPSEC*/ #ifdef INET6 if (INP_SOCKAF(so) == AF_INET6) { inp->inp_vflag |= INP_IPV6PROTO; if (V_ip6_v6only) inp->inp_flags |= IN6P_IPV6_V6ONLY; } #endif INP_WLOCK(inp); INP_LIST_WLOCK(pcbinfo); LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list); pcbinfo->ipi_count++; so->so_pcb = (caddr_t)inp; #ifdef INET6 if (V_ip6_auto_flowlabel) inp->inp_flags |= IN6P_AUTOFLOWLABEL; #endif inp->inp_gencnt = ++pcbinfo->ipi_gencnt; refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */ /* * Routes in inpcb's can cache L2 as well; they are guaranteed * to be cleaned up. */ inp->inp_route.ro_flags = RT_LLE_CACHE; INP_LIST_WUNLOCK(pcbinfo); #if defined(IPSEC) || defined(IPSEC_SUPPORT) || defined(MAC) out: if (error != 0) { crfree(inp->inp_cred); uma_zfree(pcbinfo->ipi_zone, inp); } #endif return (error); } #ifdef INET int in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { int anonport, error; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) return (EINVAL); anonport = nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0; error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, &inp->inp_lport, cred); if (error) return (error); if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; return (EAGAIN); } if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } #endif /* * Select a local port (number) to use. */ #if defined(INET) || defined(INET6) int in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp, struct ucred *cred, int lookupflags) { struct inpcbinfo *pcbinfo; struct inpcb *tmpinp; unsigned short *lastport; int count, dorandom, error; u_short aux, first, last, lport; #ifdef INET struct in_addr laddr; #endif pcbinfo = inp->inp_pcbinfo; /* * Because no actual state changes occur here, a global write lock on * the pcbinfo isn't required. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(pcbinfo); if (inp->inp_flags & INP_HIGHPORT) { first = V_ipport_hifirstauto; /* sysctl */ last = V_ipport_hilastauto; lastport = &pcbinfo->ipi_lasthi; } else if (inp->inp_flags & INP_LOWPORT) { error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); if (error) return (error); first = V_ipport_lowfirstauto; /* 1023 */ last = V_ipport_lowlastauto; /* 600 */ lastport = &pcbinfo->ipi_lastlow; } else { first = V_ipport_firstauto; /* sysctl */ last = V_ipport_lastauto; lastport = &pcbinfo->ipi_lastport; } /* * For UDP(-Lite), use random port allocation as long as the user * allows it. For TCP (and as of yet unknown) connections, * use random port allocation only if the user allows it AND * ipport_tick() allows it. */ if (V_ipport_randomized && (!V_ipport_stoprandom || pcbinfo == &V_udbinfo || pcbinfo == &V_ulitecbinfo)) dorandom = 1; else dorandom = 0; /* * It makes no sense to do random port allocation if * we have the only port available. */ if (first == last) dorandom = 0; /* Make sure to not include UDP(-Lite) packets in the count. */ if (pcbinfo != &V_udbinfo || pcbinfo != &V_ulitecbinfo) V_ipport_tcpallocs++; /* * Instead of having two loops further down counting up or down * make sure that first is always <= last and go with only one * code path implementing all logic. */ if (first > last) { aux = first; first = last; last = aux; } #ifdef INET /* Make the compiler happy. */ laddr.s_addr = 0; if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) { KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p", __func__, inp)); laddr = *laddrp; } #endif tmpinp = NULL; /* Make compiler happy. */ lport = *lportp; if (dorandom) *lastport = first + (arc4random() % (last - first)); count = last - first; do { if (count-- < 0) /* completely used? */ return (EADDRNOTAVAIL); ++*lastport; if (*lastport < first || *lastport > last) *lastport = first; lport = htons(*lastport); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) tmpinp = in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, lport, lookupflags, cred); #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET tmpinp = in_pcblookup_local(pcbinfo, laddr, lport, lookupflags, cred); #endif } while (tmpinp != NULL); #ifdef INET if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) laddrp->s_addr = laddr.s_addr; #endif *lportp = lport; return (0); } /* * Return cached socket options. */ -short +int inp_so_options(const struct inpcb *inp) { - short so_options; + int so_options; - so_options = 0; + so_options = 0; - if ((inp->inp_flags2 & INP_REUSEPORT) != 0) - so_options |= SO_REUSEPORT; - if ((inp->inp_flags2 & INP_REUSEADDR) != 0) - so_options |= SO_REUSEADDR; - return (so_options); + if ((inp->inp_flags2 & INP_REUSEPORT_LB) != 0) + so_options |= SO_REUSEPORT_LB; + if ((inp->inp_flags2 & INP_REUSEPORT) != 0) + so_options |= SO_REUSEPORT; + if ((inp->inp_flags2 & INP_REUSEADDR) != 0) + so_options |= SO_REUSEADDR; + return (so_options); } #endif /* INET || INET6 */ /* * Check if a new BINDMULTI socket is allowed to be created. * * ni points to the new inp. * oi points to the exisitng inp. * * This checks whether the existing inp also has BINDMULTI and * whether the credentials match. */ int in_pcbbind_check_bindmulti(const struct inpcb *ni, const struct inpcb *oi) { /* Check permissions match */ if ((ni->inp_flags2 & INP_BINDMULTI) && (ni->inp_cred->cr_uid != oi->inp_cred->cr_uid)) return (0); /* Check the existing inp has BINDMULTI set */ if ((ni->inp_flags2 & INP_BINDMULTI) && ((oi->inp_flags2 & INP_BINDMULTI) == 0)) return (0); /* * We're okay - either INP_BINDMULTI isn't set on ni, or * it is and it matches the checks. */ return (1); } #ifdef INET /* * Set up a bind operation on a PCB, performing port allocation * as required, but do not actually modify the PCB. Callers can * either complete the bind by setting inp_laddr/inp_lport and * calling in_pcbinshash(), or they can just use the resulting * port and address to authorise the sending of a once-off packet. * * On error, the values of *laddrp and *lportp are not changed. */ int in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, u_short *lportp, struct ucred *cred) { struct socket *so = inp->inp_socket; struct sockaddr_in *sin; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct in_addr laddr; u_short lport = 0; int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT); int error; /* + * XXX: Maybe we could let SO_REUSEPORT_LB set SO_REUSEPORT bit here + * so that we don't have to add to the (already messy) code below. + */ + int reuseport_lb = (so->so_options & SO_REUSEPORT_LB); + + /* * No state changes, so read locks are sufficient here. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(pcbinfo); if (CK_STAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */ return (EADDRNOTAVAIL); laddr.s_addr = *laddrp; if (nam != NULL && laddr.s_addr != INADDR_ANY) return (EINVAL); - if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) + if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0) lookupflags = INPLOOKUP_WILDCARD; if (nam == NULL) { if ((error = prison_local_ip4(cred, &laddr)) != 0) return (error); } else { sin = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sin)) return (EINVAL); #ifdef notdef /* * We should check the family, but old programs * incorrectly fail to initialize it. */ if (sin->sin_family != AF_INET) return (EAFNOSUPPORT); #endif error = prison_local_ip4(cred, &sin->sin_addr); if (error) return (error); if (sin->sin_port != *lportp) { /* Don't allow the port to change. */ if (*lportp != 0) return (EINVAL); lport = sin->sin_port; } /* NB: lport is left as 0 if the port isn't being changed. */ if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { /* * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; * allow complete duplication of binding if * SO_REUSEPORT is set, or if SO_REUSEADDR is set * and a multicast address is bound on both * new and duplicated sockets. */ if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0) reuseport = SO_REUSEADDR|SO_REUSEPORT; + /* + * XXX: How to deal with SO_REUSEPORT_LB here? + * Treat same as SO_REUSEPORT for now. + */ + if ((so->so_options & + (SO_REUSEADDR|SO_REUSEPORT_LB)) != 0) + reuseport_lb = SO_REUSEADDR|SO_REUSEPORT_LB; } else if (sin->sin_addr.s_addr != INADDR_ANY) { sin->sin_port = 0; /* yech... */ bzero(&sin->sin_zero, sizeof(sin->sin_zero)); /* - * Is the address a local IP address? + * Is the address a local IP address? * If INP_BINDANY is set, then the socket may be bound * to any endpoint address, local or not. */ if ((inp->inp_flags & INP_BINDANY) == 0 && - ifa_ifwithaddr_check((struct sockaddr *)sin) == 0) + ifa_ifwithaddr_check((struct sockaddr *)sin) == 0) return (EADDRNOTAVAIL); } laddr = sin->sin_addr; if (lport) { struct inpcb *t; struct tcptw *tw; /* GROSS */ if (ntohs(lport) <= V_ipport_reservedhigh && ntohs(lport) >= V_ipport_reservedlow && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) return (EACCES); if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) && priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT, 0) != 0) { t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport, INPLOOKUP_WILDCARD, cred); /* * XXX * This entire block sorely needs a rewrite. */ if (t && ((inp->inp_flags2 & INP_BINDMULTI) == 0) && ((t->inp_flags & INP_TIMEWAIT) == 0) && (so->so_type != SOCK_STREAM || ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || ntohl(t->inp_laddr.s_addr) != INADDR_ANY || - (t->inp_flags2 & INP_REUSEPORT) == 0) && + (t->inp_flags2 & INP_REUSEPORT) || + (t->inp_flags2 & INP_REUSEPORT_LB) == 0) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); /* * If the socket is a BINDMULTI socket, then * the credentials need to match and the * original socket also has to have been bound * with BINDMULTI. */ if (t && (! in_pcbbind_check_bindmulti(inp, t))) return (EADDRINUSE); } t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport, lookupflags, cred); if (t && (t->inp_flags & INP_TIMEWAIT)) { /* * XXXRW: If an incpb has had its timewait * state recycled, we treat the address as * being in use (for now). This is better * than a panic, but not desirable. */ tw = intotw(t); if (tw == NULL || - (reuseport & tw->tw_so_options) == 0) + ((reuseport & tw->tw_so_options) == 0 && + (reuseport_lb & + tw->tw_so_options) == 0)) { return (EADDRINUSE); + } } else if (t && - ((inp->inp_flags2 & INP_BINDMULTI) == 0) && - (reuseport & inp_so_options(t)) == 0) { + ((inp->inp_flags2 & INP_BINDMULTI) == 0) && + (reuseport & inp_so_options(t)) == 0 && + (reuseport_lb & inp_so_options(t)) == 0) { #ifdef INET6 if (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || ntohl(t->inp_laddr.s_addr) != INADDR_ANY || (inp->inp_vflag & INP_IPV6PROTO) == 0 || (t->inp_vflag & INP_IPV6PROTO) == 0) #endif - return (EADDRINUSE); + return (EADDRINUSE); if (t && (! in_pcbbind_check_bindmulti(inp, t))) return (EADDRINUSE); } } } if (*lportp != 0) lport = *lportp; if (lport == 0) { error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags); if (error != 0) return (error); } *laddrp = laddr.s_addr; *lportp = lport; return (0); } /* * Connect from a socket to a specified address. * Both address and port must be specified in argument sin. * If don't have a local address for this socket yet, * then pick one. */ int in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred, struct mbuf *m) { u_short lport, fport; in_addr_t laddr, faddr; int anonport, error; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); lport = inp->inp_lport; laddr = inp->inp_laddr.s_addr; anonport = (lport == 0); error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, NULL, cred); if (error) return (error); /* Do the initial binding of the local address if required. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; return (EAGAIN); } } /* Commit the remaining changes. */ inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; inp->inp_faddr.s_addr = faddr; inp->inp_fport = fport; in_pcbrehash_mbuf(inp, m); if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } int in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { return (in_pcbconnect_mbuf(inp, nam, cred, NULL)); } /* * Do proper source address selection on an unbound socket in case * of connect. Take jails into account as well. */ int in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, struct ucred *cred) { struct ifaddr *ifa; struct sockaddr *sa; struct sockaddr_in *sin; struct route sro; int error; KASSERT(laddr != NULL, ("%s: laddr NULL", __func__)); /* * Bypass source address selection and use the primary jail IP * if requested. */ if (cred != NULL && !prison_saddrsel_ip4(cred, laddr)) return (0); error = 0; bzero(&sro, sizeof(sro)); sin = (struct sockaddr_in *)&sro.ro_dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(struct sockaddr_in); sin->sin_addr.s_addr = faddr->s_addr; /* * If route is known our src addr is taken from the i/f, * else punt. * * Find out route to destination. */ if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum); /* * If we found a route, use the address corresponding to * the outgoing interface. * * Otherwise assume faddr is reachable on a directly connected * network and try to find a corresponding interface to take * the source address from. */ NET_EPOCH_ENTER(); if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) { struct in_ifaddr *ia; struct ifnet *ifp; ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin, inp->inp_socket->so_fibnum)); if (ia == NULL) { ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0, inp->inp_socket->so_fibnum)); } if (ia == NULL) { error = ENETUNREACH; goto done; } if (cred == NULL || !prison_flag(cred, PR_IP4)) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } ifp = ia->ia_ifp; ia = NULL; IF_ADDR_RLOCK(ifp); CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_RUNLOCK(ifp); goto done; } IF_ADDR_RUNLOCK(ifp); /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * If the outgoing interface on the route found is not * a loopback interface, use the address from that interface. * In case of jails do those three steps: * 1. check if the interface address belongs to the jail. If so use it. * 2. check if we have any address on the outgoing interface * belonging to this jail. If so use it. * 3. as a last resort return the 'default' jail address. */ if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { struct in_ifaddr *ia; struct ifnet *ifp; /* If not jailed, use the default returned. */ if (cred == NULL || !prison_flag(cred, PR_IP4)) { ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* Jailed. */ /* 1. Check if the iface address belongs to the jail. */ sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* * 2. Check if we have any address on the outgoing interface * belonging to this jail. */ ia = NULL; ifp = sro.ro_rt->rt_ifp; IF_ADDR_RLOCK(ifp); CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_RUNLOCK(ifp); goto done; } IF_ADDR_RUNLOCK(ifp); /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * The outgoing interface is marked with 'loopback net', so a route * to ourselves is here. * Try to find the interface of the destination address and then * take the address from there. That interface is not necessarily * a loopback interface. * In case of jails, check that it is an address of the jail * and if we cannot find, fall back to the 'default' jail address. */ if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { struct sockaddr_in sain; struct in_ifaddr *ia; bzero(&sain, sizeof(struct sockaddr_in)); sain.sin_family = AF_INET; sain.sin_len = sizeof(struct sockaddr_in); sain.sin_addr.s_addr = faddr->s_addr; ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain), inp->inp_socket->so_fibnum)); if (ia == NULL) ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0, inp->inp_socket->so_fibnum)); if (ia == NULL) ia = ifatoia(ifa_ifwithaddr(sintosa(&sain))); if (cred == NULL || !prison_flag(cred, PR_IP4)) { if (ia == NULL) { error = ENETUNREACH; goto done; } laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* Jailed. */ if (ia != NULL) { struct ifnet *ifp; ifp = ia->ia_ifp; ia = NULL; IF_ADDR_RLOCK(ifp); CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; IF_ADDR_RUNLOCK(ifp); goto done; } IF_ADDR_RUNLOCK(ifp); } /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } done: NET_EPOCH_EXIT(); if (sro.ro_rt != NULL) RTFREE(sro.ro_rt); return (error); } /* * Set up for a connect from a socket to the specified address. * On entry, *laddrp and *lportp should contain the current local * address and port for the PCB; these are updated to the values * that should be placed in inp_laddr and inp_lport to complete * the connect. * * On success, *faddrp and *fportp will be set to the remote address * and port. These are not updated in the error case. * * If the operation fails because the connection already exists, * *oinpp will be set to the PCB of that connection so that the * caller can decide to override it. In all other cases, *oinpp * is set to NULL. */ int in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp, struct inpcb **oinpp, struct ucred *cred) { struct rm_priotracker in_ifa_tracker; struct sockaddr_in *sin = (struct sockaddr_in *)nam; struct in_ifaddr *ia; struct inpcb *oinp; struct in_addr laddr, faddr; u_short lport, fport; int error; /* * Because a global state change doesn't actually occur here, a read * lock is sufficient. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo); if (oinpp != NULL) *oinpp = NULL; if (nam->sa_len != sizeof (*sin)) return (EINVAL); if (sin->sin_family != AF_INET) return (EAFNOSUPPORT); if (sin->sin_port == 0) return (EADDRNOTAVAIL); laddr.s_addr = *laddrp; lport = *lportp; faddr = sin->sin_addr; fport = sin->sin_port; if (!CK_STAILQ_EMPTY(&V_in_ifaddrhead)) { /* * If the destination address is INADDR_ANY, * use the primary local address. * If the supplied address is INADDR_BROADCAST, * and the primary interface supports broadcast, * choose the broadcast address for that interface. */ if (faddr.s_addr == INADDR_ANY) { IN_IFADDR_RLOCK(&in_ifa_tracker); faddr = IA_SIN(CK_STAILQ_FIRST(&V_in_ifaddrhead))->sin_addr; IN_IFADDR_RUNLOCK(&in_ifa_tracker); if (cred != NULL && (error = prison_get_ip4(cred, &faddr)) != 0) return (error); } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) { IN_IFADDR_RLOCK(&in_ifa_tracker); if (CK_STAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags & IFF_BROADCAST) faddr = satosin(&CK_STAILQ_FIRST( &V_in_ifaddrhead)->ia_broadaddr)->sin_addr; IN_IFADDR_RUNLOCK(&in_ifa_tracker); } } if (laddr.s_addr == INADDR_ANY) { error = in_pcbladdr(inp, &faddr, &laddr, cred); /* * If the destination address is multicast and an outgoing * interface has been set as a multicast option, prefer the * address of that interface as our source address. */ if (IN_MULTICAST(ntohl(faddr.s_addr)) && inp->inp_moptions != NULL) { struct ip_moptions *imo; struct ifnet *ifp; imo = inp->inp_moptions; if (imo->imo_multicast_ifp != NULL) { ifp = imo->imo_multicast_ifp; IN_IFADDR_RLOCK(&in_ifa_tracker); CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if ((ia->ia_ifp == ifp) && (cred == NULL || prison_check_ip4(cred, &ia->ia_addr.sin_addr) == 0)) break; } if (ia == NULL) error = EADDRNOTAVAIL; else { laddr = ia->ia_addr.sin_addr; error = 0; } IN_IFADDR_RUNLOCK(&in_ifa_tracker); } } if (error) return (error); } oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport, laddr, lport, 0, NULL); if (oinp != NULL) { if (oinpp != NULL) *oinpp = oinp; return (EADDRINUSE); } if (lport == 0) { error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, cred); if (error) return (error); } *laddrp = laddr.s_addr; *lportp = lport; *faddrp = faddr.s_addr; *fportp = fport; return (0); } void in_pcbdisconnect(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); inp->inp_faddr.s_addr = INADDR_ANY; inp->inp_fport = 0; in_pcbrehash(inp); } #endif /* INET */ /* * in_pcbdetach() is responsibe for disassociating a socket from an inpcb. * For most protocols, this will be invoked immediately prior to calling * in_pcbfree(). However, with TCP the inpcb may significantly outlive the * socket, in which case in_pcbfree() is deferred. */ void in_pcbdetach(struct inpcb *inp) { KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__)); #ifdef RATELIMIT if (inp->inp_snd_tag != NULL) in_pcbdetach_txrtlmt(inp); #endif inp->inp_socket->so_pcb = NULL; inp->inp_socket = NULL; } /* * in_pcbref() bumps the reference count on an inpcb in order to maintain * stability of an inpcb pointer despite the inpcb lock being released. This * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded, * but where the inpcb lock may already held, or when acquiring a reference * via a pcbgroup. * * in_pcbref() should be used only to provide brief memory stability, and * must always be followed by a call to INP_WLOCK() and in_pcbrele() to * garbage collect the inpcb if it has been in_pcbfree()'d from another * context. Until in_pcbrele() has returned that the inpcb is still valid, * lock and rele are the *only* safe operations that may be performed on the * inpcb. * * While the inpcb will not be freed, releasing the inpcb lock means that the * connection's state may change, so the caller should be careful to * revalidate any cached state on reacquiring the lock. Drop the reference * using in_pcbrele(). */ void in_pcbref(struct inpcb *inp) { KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); refcount_acquire(&inp->inp_refcount); } /* * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we * return a flag indicating whether or not the inpcb remains valid. If it is * valid, we return with the inpcb lock held. * * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a * reference on an inpcb. Historically more work was done here (actually, in * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely * about memory stability (and continued use of the write lock). */ int in_pcbrele_rlocked(struct inpcb *inp) { struct inpcbinfo *pcbinfo; KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); INP_RLOCK_ASSERT(inp); if (refcount_release(&inp->inp_refcount) == 0) { /* * If the inpcb has been freed, let the caller know, even if * this isn't the last reference. */ if (inp->inp_flags2 & INP_FREED) { INP_RUNLOCK(inp); return (1); } return (0); } KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); #ifdef TCPHPTS if (inp->inp_in_hpts || inp->inp_in_input) { struct tcp_hpts_entry *hpts; /* * We should not be on the hpts at * this point in any form. we must * get the lock to be sure. */ hpts = tcp_hpts_lock(inp); if (inp->inp_in_hpts) panic("Hpts:%p inp:%p at free still on hpts", hpts, inp); mtx_unlock(&hpts->p_mtx); hpts = tcp_input_lock(inp); if (inp->inp_in_input) panic("Hpts:%p inp:%p at free still on input hpts", hpts, inp); mtx_unlock(&hpts->p_mtx); } #endif INP_RUNLOCK(inp); pcbinfo = inp->inp_pcbinfo; uma_zfree(pcbinfo->ipi_zone, inp); return (1); } int in_pcbrele_wlocked(struct inpcb *inp) { struct inpcbinfo *pcbinfo; KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); INP_WLOCK_ASSERT(inp); if (refcount_release(&inp->inp_refcount) == 0) { /* * If the inpcb has been freed, let the caller know, even if * this isn't the last reference. */ if (inp->inp_flags2 & INP_FREED) { INP_WUNLOCK(inp); return (1); } return (0); } KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); #ifdef TCPHPTS if (inp->inp_in_hpts || inp->inp_in_input) { struct tcp_hpts_entry *hpts; /* * We should not be on the hpts at * this point in any form. we must * get the lock to be sure. */ hpts = tcp_hpts_lock(inp); if (inp->inp_in_hpts) panic("Hpts:%p inp:%p at free still on hpts", hpts, inp); mtx_unlock(&hpts->p_mtx); hpts = tcp_input_lock(inp); if (inp->inp_in_input) panic("Hpts:%p inp:%p at free still on input hpts", hpts, inp); mtx_unlock(&hpts->p_mtx); } #endif INP_WUNLOCK(inp); pcbinfo = inp->inp_pcbinfo; uma_zfree(pcbinfo->ipi_zone, inp); return (1); } /* * Temporary wrapper. */ int in_pcbrele(struct inpcb *inp) { return (in_pcbrele_wlocked(inp)); } void in_pcblist_rele_rlocked(epoch_context_t ctx) { struct in_pcblist *il; struct inpcb *inp; struct inpcbinfo *pcbinfo; int i, n; il = __containerof(ctx, struct in_pcblist, il_epoch_ctx); pcbinfo = il->il_pcbinfo; n = il->il_count; INP_INFO_WLOCK(pcbinfo); for (i = 0; i < n; i++) { inp = il->il_inp_list[i]; INP_RLOCK(inp); if (!in_pcbrele_rlocked(inp)) INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); free(il, M_TEMP); } /* * Unconditionally schedule an inpcb to be freed by decrementing its * reference count, which should occur only after the inpcb has been detached * from its socket. If another thread holds a temporary reference (acquired * using in_pcbref()) then the free is deferred until that reference is * released using in_pcbrele(), but the inpcb is still unlocked. Almost all * work, including removal from global lists, is done in this context, where * the pcbinfo lock is held. */ void in_pcbfree(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; #ifdef INET6 struct ip6_moptions *im6o = NULL; #endif #ifdef INET struct ip_moptions *imo = NULL; #endif KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); KASSERT((inp->inp_flags2 & INP_FREED) == 0, ("%s: called twice for pcb %p", __func__, inp)); if (inp->inp_flags2 & INP_FREED) { INP_WUNLOCK(inp); return; } #ifdef INVARIANTS if (pcbinfo == &V_tcbinfo) { INP_INFO_LOCK_ASSERT(pcbinfo); } else { INP_INFO_WLOCK_ASSERT(pcbinfo); } #endif INP_WLOCK_ASSERT(inp); #ifdef INET imo = inp->inp_moptions; inp->inp_moptions = NULL; #endif /* XXXRW: Do as much as possible here. */ #if defined(IPSEC) || defined(IPSEC_SUPPORT) if (inp->inp_sp != NULL) ipsec_delete_pcbpolicy(inp); #endif INP_LIST_WLOCK(pcbinfo); inp->inp_gencnt = ++pcbinfo->ipi_gencnt; in_pcbremlists(inp); INP_LIST_WUNLOCK(pcbinfo); #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) { ip6_freepcbopts(inp->in6p_outputopts); im6o = inp->in6p_moptions; inp->in6p_moptions = NULL; } #endif if (inp->inp_options) (void)m_free(inp->inp_options); RO_INVALIDATE_CACHE(&inp->inp_route); inp->inp_vflag = 0; inp->inp_flags2 |= INP_FREED; crfree(inp->inp_cred); #ifdef MAC mac_inpcb_destroy(inp); #endif #ifdef INET6 ip6_freemoptions(im6o); #endif #ifdef INET inp_freemoptions(imo); #endif if (!in_pcbrele_wlocked(inp)) INP_WUNLOCK(inp); } /* * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and * port reservation, and preventing it from being returned by inpcb lookups. * * It is used by TCP to mark an inpcb as unused and avoid future packet * delivery or event notification when a socket remains open but TCP has * closed. This might occur as a result of a shutdown()-initiated TCP close * or a RST on the wire, and allows the port binding to be reused while still * maintaining the invariant that so_pcb always points to a valid inpcb until * in_pcbdetach(). * * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by * in_pcbnotifyall() and in_pcbpurgeif0()? */ void in_pcbdrop(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); /* * XXXRW: Possibly we should protect the setting of INP_DROPPED with * the hash lock...? */ inp->inp_flags |= INP_DROPPED; if (inp->inp_flags & INP_INHASHLIST) { struct inpcbport *phd = inp->inp_phd; INP_HASH_WLOCK(inp->inp_pcbinfo); + in_pcbremlbgrouphash(inp); LIST_REMOVE(inp, inp_hash); LIST_REMOVE(inp, inp_portlist); if (LIST_FIRST(&phd->phd_pcblist) == NULL) { LIST_REMOVE(phd, phd_hash); free(phd, M_PCB); } INP_HASH_WUNLOCK(inp->inp_pcbinfo); inp->inp_flags &= ~INP_INHASHLIST; #ifdef PCBGROUP in_pcbgroup_remove(inp); #endif } } #ifdef INET /* * Common routines to return the socket addresses associated with inpcbs. */ struct sockaddr * in_sockaddr(in_port_t port, struct in_addr *addr_p) { struct sockaddr_in *sin; sin = malloc(sizeof *sin, M_SONAME, M_WAITOK | M_ZERO); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = *addr_p; sin->sin_port = port; return (struct sockaddr *)sin; } int in_getsockaddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_lport; addr = inp->inp_laddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } int in_getpeeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_fport; addr = inp->inp_faddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno, struct inpcb *(*notify)(struct inpcb *, int)) { struct inpcb *inp, *inp_temp; INP_INFO_WLOCK(pcbinfo); LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { INP_WLOCK(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) { INP_WUNLOCK(inp); continue; } #endif if (inp->inp_faddr.s_addr != faddr.s_addr || inp->inp_socket == NULL) { INP_WUNLOCK(inp); continue; } if ((*notify)(inp, errno)) INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); } void in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) { struct inpcb *inp; struct ip_moptions *imo; int i, gap; INP_INFO_WLOCK(pcbinfo); LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) { INP_WLOCK(inp); imo = inp->inp_moptions; if ((inp->inp_vflag & INP_IPV4) && imo != NULL) { /* * Unselect the outgoing interface if it is being * detached. */ if (imo->imo_multicast_ifp == ifp) imo->imo_multicast_ifp = NULL; /* * Drop multicast group membership if we joined * through the interface being detached. * * XXX This can all be deferred to an epoch_call */ for (i = 0, gap = 0; i < imo->imo_num_memberships; i++) { if (imo->imo_membership[i]->inm_ifp == ifp) { IN_MULTI_LOCK_ASSERT(); in_leavegroup_locked(imo->imo_membership[i], NULL); gap++; } else if (gap != 0) imo->imo_membership[i - gap] = imo->imo_membership[i]; } imo->imo_num_memberships -= gap; } INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); } /* * Lookup a PCB based on the local address and port. Caller must hold the * hash lock. No inpcb locks or references are acquired. */ #define INP_LOOKUP_MAPPED_PCB_COST 3 struct inpcb * in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr, u_short lport, int lookupflags, struct ucred *cred) { struct inpcb *inp; #ifdef INET6 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST; #else int matchwild = 3; #endif int wildcard; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); INP_HASH_LOCK_ASSERT(pcbinfo); if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { struct inpcbhead *head; /* * Look for an unconnected (wildcard foreign addr) PCB that * matches the local address and port we're looking for. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == INADDR_ANY && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_lport == lport) { /* * Found? */ if (cred == NULL || prison_equal_ip4(cred->cr_prison, inp->inp_cred->cr_prison)) return (inp); } } /* * Not found. */ return (NULL); } else { struct inpcbporthead *porthash; struct inpcbport *phd; struct inpcb *match = NULL; /* * Best fit PCB lookup. * * First see if this local port is in use by looking on the * port hash list. */ porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, pcbinfo->ipi_porthashmask)]; LIST_FOREACH(phd, porthash, phd_hash) { if (phd->phd_port == lport) break; } if (phd != NULL) { /* * Port is in use by one or more PCBs. Look for best * fit. */ LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { wildcard = 0; if (cred != NULL && !prison_equal_ip4(inp->inp_cred->cr_prison, cred->cr_prison)) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; /* * We never select the PCB that has * INP_IPV6 flag and is bound to :: if * we have another PCB which is bound * to 0.0.0.0. If a PCB has the * INP_IPV6 flag, then we set its cost * higher than IPv4 only PCBs. * * Note that the case only happens * when a socket is bound to ::, under * the condition that the use of the * mapped address is allowed. */ if ((inp->inp_vflag & INP_IPV6) != 0) wildcard += INP_LOOKUP_MAPPED_PCB_COST; #endif if (inp->inp_faddr.s_addr != INADDR_ANY) wildcard++; if (inp->inp_laddr.s_addr != INADDR_ANY) { if (laddr.s_addr == INADDR_ANY) wildcard++; else if (inp->inp_laddr.s_addr != laddr.s_addr) continue; } else { if (laddr.s_addr != INADDR_ANY) wildcard++; } if (wildcard < matchwild) { match = inp; matchwild = wildcard; if (matchwild == 0) break; } } } return (match); } } #undef INP_LOOKUP_MAPPED_PCB_COST +static struct inpcb * +in_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo, + const struct in_addr *laddr, uint16_t lport, const struct in_addr *faddr, + uint16_t fport, int lookupflags) +{ + struct inpcb *local_wild = NULL; + const struct inpcblbgrouphead *hdr; + struct inpcblbgroup *grp; + struct inpcblbgroup *grp_local_wild; + + INP_HASH_LOCK_ASSERT(pcbinfo); + + hdr = &pcbinfo->ipi_lbgrouphashbase[ + INP_PCBLBGROUP_PORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)]; + + /* + * Order of socket selection: + * 1. non-wild. + * 2. wild (if lookupflags contains INPLOOKUP_WILDCARD). + * + * NOTE: + * - Load balanced group does not contain jailed sockets + * - Load balanced group does not contain IPv4 mapped INET6 wild sockets + */ + LIST_FOREACH(grp, hdr, il_list) { +#ifdef INET6 + if (!(grp->il_vflag & INP_IPV4)) + continue; +#endif + + if (grp->il_lport == lport) { + + uint32_t idx = 0; + int pkt_hash = INP_PCBLBGROUP_PKTHASH(faddr->s_addr, + lport, fport); + + idx = pkt_hash % grp->il_inpcnt; + + if (grp->il_laddr.s_addr == laddr->s_addr) { + return (grp->il_inp[idx]); + } else { + if (grp->il_laddr.s_addr == INADDR_ANY && + (lookupflags & INPLOOKUP_WILDCARD)) { + local_wild = grp->il_inp[idx]; + grp_local_wild = grp; + } + } + } + } + if (local_wild != NULL) { + return (local_wild); + } + return (NULL); +} + #ifdef PCBGROUP /* * Lookup PCB in hash list, using pcbgroup tables. */ static struct inpcb * in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; bool locked; /* * First look for an exact match. */ tmpinp = NULL; INP_GROUP_LOCK(pcbgroup); head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbgroup->ipg_hashmask)]; LIST_FOREACH(inp, head, inp_pcbgrouphash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == faddr.s_addr && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP4)) goto found; if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) { inp = tmpinp; goto found; } #ifdef RSS /* * For incoming connections, we may wish to do a wildcard * match for an RSS-local socket. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; #ifdef INET6 struct inpcb *local_wild_mapped = NULL; #endif struct inpcb *jail_wild = NULL; struct inpcbhead *head; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbgroup->ipg_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)]; LIST_FOREACH(inp, head, inp_pcbgrouphash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) continue; injail = prison_flag(inp->inp_cred, PR_IP4); if (injail) { if (prison_check_ip4(inp->inp_cred, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (inp->inp_laddr.s_addr == laddr.s_addr) { if (injail) goto found; else local_exact = inp; } else if (inp->inp_laddr.s_addr == INADDR_ANY) { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ inp = jail_wild; if (inp == NULL) inp = local_exact; if (inp == NULL) inp = local_wild; #ifdef INET6 if (inp == NULL) inp = local_wild_mapped; #endif if (inp != NULL) goto found; } #endif /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; #ifdef INET6 struct inpcb *local_wild_mapped = NULL; #endif struct inpcb *jail_wild = NULL; struct inpcbhead *head; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_wildmask)]; LIST_FOREACH(inp, head, inp_pcbgroup_wild) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) continue; injail = prison_flag(inp->inp_cred, PR_IP4); if (injail) { if (prison_check_ip4(inp->inp_cred, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (inp->inp_laddr.s_addr == laddr.s_addr) { if (injail) goto found; else local_exact = inp; } else if (inp->inp_laddr.s_addr == INADDR_ANY) { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ inp = jail_wild; if (inp == NULL) inp = local_exact; if (inp == NULL) inp = local_wild; #ifdef INET6 if (inp == NULL) inp = local_wild_mapped; #endif if (inp != NULL) goto found; } /* if (lookupflags & INPLOOKUP_WILDCARD) */ INP_GROUP_UNLOCK(pcbgroup); return (NULL); found: if (lookupflags & INPLOOKUP_WLOCKPCB) locked = INP_TRY_WLOCK(inp); else if (lookupflags & INPLOOKUP_RLOCKPCB) locked = INP_TRY_RLOCK(inp); else panic("%s: locking bug", __func__); if (!locked) in_pcbref(inp); INP_GROUP_UNLOCK(pcbgroup); if (!locked) { if (lookupflags & INPLOOKUP_WLOCKPCB) { INP_WLOCK(inp); if (in_pcbrele_wlocked(inp)) return (NULL); } else { INP_RLOCK(inp); if (in_pcbrele_rlocked(inp)) return (NULL); } } #ifdef INVARIANTS if (lookupflags & INPLOOKUP_WLOCKPCB) INP_WLOCK_ASSERT(inp); else INP_RLOCK_ASSERT(inp); #endif return (inp); } #endif /* PCBGROUP */ /* * Lookup PCB in hash list, using pcbinfo tables. This variation assumes * that the caller has locked the hash list, and will not perform any further * locking or reference operations on either the hash list or the connection. */ static struct inpcb * in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); INP_HASH_LOCK_ASSERT(pcbinfo); /* * First look for an exact match. */ tmpinp = NULL; head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == faddr.s_addr && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP4)) return (inp); if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) return (tmpinp); /* + * Then look in lb group (for wildcard match). + */ + if (pcbinfo->ipi_lbgrouphashbase != NULL && + (lookupflags & INPLOOKUP_WILDCARD)) { + inp = in_pcblookup_lbgroup(pcbinfo, &laddr, lport, &faddr, + fport, lookupflags); + if (inp != NULL) { + return (inp); + } + } + + /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; #ifdef INET6 struct inpcb *local_wild_mapped = NULL; #endif struct inpcb *jail_wild = NULL; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) continue; injail = prison_flag(inp->inp_cred, PR_IP4); if (injail) { if (prison_check_ip4(inp->inp_cred, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (inp->inp_laddr.s_addr == laddr.s_addr) { if (injail) return (inp); else local_exact = inp; } else if (inp->inp_laddr.s_addr == INADDR_ANY) { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ if (jail_wild != NULL) return (jail_wild); if (local_exact != NULL) return (local_exact); if (local_wild != NULL) return (local_wild); #ifdef INET6 if (local_wild_mapped != NULL) return (local_wild_mapped); #endif } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ return (NULL); } /* * Lookup PCB in hash list, using pcbinfo tables. This variation locks the * hash list lock, and will return the inpcb locked (i.e., requires * INPLOOKUP_LOCKPCB). */ static struct inpcb * in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp) { struct inpcb *inp; bool locked; INP_HASH_RLOCK(pcbinfo); inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); if (inp != NULL) { if (lookupflags & INPLOOKUP_WLOCKPCB) locked = INP_TRY_WLOCK(inp); else if (lookupflags & INPLOOKUP_RLOCKPCB) locked = INP_TRY_RLOCK(inp); else panic("%s: locking bug", __func__); if (!locked) in_pcbref(inp); INP_HASH_RUNLOCK(pcbinfo); if (!locked) { if (lookupflags & INPLOOKUP_WLOCKPCB) { INP_WLOCK(inp); if (in_pcbrele_wlocked(inp)) return (NULL); } else { INP_RLOCK(inp); if (in_pcbrele_rlocked(inp)) return (NULL); } } #ifdef INVARIANTS if (lookupflags & INPLOOKUP_WLOCKPCB) INP_WLOCK_ASSERT(inp); else INP_RLOCK_ASSERT(inp); #endif } else INP_HASH_RUNLOCK(pcbinfo); return (inp); } /* * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf * from which a pre-calculated hash value may be extracted. * * Possibly more of this logic should be in in_pcbgroup.c. */ struct inpcb * in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp) { #if defined(PCBGROUP) && !defined(RSS) struct inpcbgroup *pcbgroup; #endif KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); /* * When not using RSS, use connection groups in preference to the * reservation table when looking up 4-tuples. When using RSS, just * use the reservation table, due to the cost of the Toeplitz hash * in software. * * XXXRW: This policy belongs in the pcbgroup code, as in principle * we could be doing RSS with a non-Toeplitz hash that is affordable * in software. */ #if defined(PCBGROUP) && !defined(RSS) if (in_pcbgroup_enabled(pcbinfo)) { pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, fport); return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); } #endif return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, ifp)); } struct inpcb * in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp, struct mbuf *m) { #ifdef PCBGROUP struct inpcbgroup *pcbgroup; #endif KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); #ifdef PCBGROUP /* * If we can use a hardware-generated hash to look up the connection * group, use that connection group to find the inpcb. Otherwise * fall back on a software hash -- or the reservation table if we're * using RSS. * * XXXRW: As above, that policy belongs in the pcbgroup code. */ if (in_pcbgroup_enabled(pcbinfo) && !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) { pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m), m->m_pkthdr.flowid); if (pcbgroup != NULL) return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); #ifndef RSS pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, fport); return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); #endif } #endif return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, ifp)); } #endif /* INET */ /* * Insert PCB onto various hash lists. */ static int in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update) { struct inpcbhead *pcbhash; struct inpcbporthead *pcbporthash; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbport *phd; u_int32_t hashkey_faddr; + int so_options; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); KASSERT((inp->inp_flags & INP_INHASHLIST) == 0, ("in_pcbinshash: INP_INHASHLIST")); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr); else #endif hashkey_faddr = inp->inp_faddr.s_addr; pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; pcbporthash = &pcbinfo->ipi_porthashbase[ INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)]; /* + * Add entry to load balance group. + * Only do this if SO_REUSEPORT_LB is set. + */ + so_options = inp_so_options(inp); + if (so_options & SO_REUSEPORT_LB) { + int ret = in_pcbinslbgrouphash(inp); + if (ret) { + /* pcb lb group malloc fail (ret=ENOBUFS). */ + return (ret); + } + } + + /* * Go through port list and look for a head for this lport. */ LIST_FOREACH(phd, pcbporthash, phd_hash) { if (phd->phd_port == inp->inp_lport) break; } /* * If none exists, malloc one and tack it on. */ if (phd == NULL) { phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT); if (phd == NULL) { return (ENOBUFS); /* XXX */ } phd->phd_port = inp->inp_lport; LIST_INIT(&phd->phd_pcblist); LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); } inp->inp_phd = phd; LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); LIST_INSERT_HEAD(pcbhash, inp, inp_hash); inp->inp_flags |= INP_INHASHLIST; #ifdef PCBGROUP if (do_pcbgroup_update) in_pcbgroup_update(inp); #endif return (0); } /* * For now, there are two public interfaces to insert an inpcb into the hash * lists -- one that does update pcbgroups, and one that doesn't. The latter * is used only in the TCP syncache, where in_pcbinshash is called before the * full 4-tuple is set for the inpcb, and we don't want to install in the * pcbgroup until later. * * XXXRW: This seems like a misfeature. in_pcbinshash should always update * connection groups, and partially initialised inpcbs should not be exposed * to either reservation hash tables or pcbgroups. */ int in_pcbinshash(struct inpcb *inp) { return (in_pcbinshash_internal(inp, 1)); } int in_pcbinshash_nopcbgroup(struct inpcb *inp) { return (in_pcbinshash_internal(inp, 0)); } /* * Move PCB to the proper hash bucket when { faddr, fport } have been * changed. NOTE: This does not handle the case of the lport changing (the * hashed port list would have to be updated as well), so the lport must * not change after in_pcbinshash() has been called. */ void in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbhead *head; u_int32_t hashkey_faddr; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); KASSERT(inp->inp_flags & INP_INHASHLIST, ("in_pcbrehash: !INP_INHASHLIST")); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr); else #endif hashkey_faddr = inp->inp_faddr.s_addr; head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; LIST_REMOVE(inp, inp_hash); LIST_INSERT_HEAD(head, inp, inp_hash); #ifdef PCBGROUP if (m != NULL) in_pcbgroup_update_mbuf(inp, m); else in_pcbgroup_update(inp); #endif } void in_pcbrehash(struct inpcb *inp) { in_pcbrehash_mbuf(inp, NULL); } /* * Remove PCB from various lists. */ static void in_pcbremlists(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; #ifdef INVARIANTS if (pcbinfo == &V_tcbinfo) { INP_INFO_RLOCK_ASSERT(pcbinfo); } else { INP_INFO_WLOCK_ASSERT(pcbinfo); } #endif INP_WLOCK_ASSERT(inp); INP_LIST_WLOCK_ASSERT(pcbinfo); inp->inp_gencnt = ++pcbinfo->ipi_gencnt; if (inp->inp_flags & INP_INHASHLIST) { struct inpcbport *phd = inp->inp_phd; INP_HASH_WLOCK(pcbinfo); + + /* XXX: Only do if SO_REUSEPORT_LB set? */ + in_pcbremlbgrouphash(inp); + LIST_REMOVE(inp, inp_hash); LIST_REMOVE(inp, inp_portlist); if (LIST_FIRST(&phd->phd_pcblist) == NULL) { LIST_REMOVE(phd, phd_hash); free(phd, M_PCB); } INP_HASH_WUNLOCK(pcbinfo); inp->inp_flags &= ~INP_INHASHLIST; } LIST_REMOVE(inp, inp_list); pcbinfo->ipi_count--; #ifdef PCBGROUP in_pcbgroup_remove(inp); #endif } /* * Check for alternatives when higher level complains * about service problems. For now, invalidate cached * routing information. If the route was created dynamically * (by a redirect), time to try a default gateway again. */ void in_losing(struct inpcb *inp) { RO_INVALIDATE_CACHE(&inp->inp_route); return; } /* * A set label operation has occurred at the socket layer, propagate the * label change into the in_pcb for the socket. */ void in_pcbsosetlabel(struct socket *so) { #ifdef MAC struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL")); INP_WLOCK(inp); SOCK_LOCK(so); mac_inpcb_sosetlabel(so, inp); SOCK_UNLOCK(so); INP_WUNLOCK(inp); #endif } /* * ipport_tick runs once per second, determining if random port allocation * should be continued. If more than ipport_randomcps ports have been * allocated in the last second, then we return to sequential port * allocation. We return to random allocation only once we drop below * ipport_randomcps for at least ipport_randomtime seconds. */ static void ipport_tick(void *xtp) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */ if (V_ipport_tcpallocs <= V_ipport_tcplastcount + V_ipport_randomcps) { if (V_ipport_stoprandom > 0) V_ipport_stoprandom--; } else V_ipport_stoprandom = V_ipport_randomtime; V_ipport_tcplastcount = V_ipport_tcpallocs; CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL); } static void ip_fini(void *xtp) { callout_stop(&ipport_tick_callout); } /* * The ipport_callout should start running at about the time we attach the * inet or inet6 domains. */ static void ipport_tick_init(const void *unused __unused) { /* Start ipport_tick. */ callout_init(&ipport_tick_callout, 1); callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, SHUTDOWN_PRI_DEFAULT); } SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipport_tick_init, NULL); void inp_wlock(struct inpcb *inp) { INP_WLOCK(inp); } void inp_wunlock(struct inpcb *inp) { INP_WUNLOCK(inp); } void inp_rlock(struct inpcb *inp) { INP_RLOCK(inp); } void inp_runlock(struct inpcb *inp) { INP_RUNLOCK(inp); } #ifdef INVARIANT_SUPPORT void inp_lock_assert(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); } void inp_unlock_assert(struct inpcb *inp) { INP_UNLOCK_ASSERT(inp); } #endif void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg) { struct inpcb *inp; INP_INFO_WLOCK(&V_tcbinfo); LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) { INP_WLOCK(inp); func(inp, arg); INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(&V_tcbinfo); } struct socket * inp_inpcbtosocket(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return (inp->inp_socket); } struct tcpcb * inp_inpcbtotcpcb(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return ((struct tcpcb *)inp->inp_ppcb); } int inp_ip_tos_get(const struct inpcb *inp) { return (inp->inp_ip_tos); } void inp_ip_tos_set(struct inpcb *inp, int val) { inp->inp_ip_tos = val; } void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, uint32_t *faddr, uint16_t *fp) { INP_LOCK_ASSERT(inp); *laddr = inp->inp_laddr.s_addr; *faddr = inp->inp_faddr.s_addr; *lp = inp->inp_lport; *fp = inp->inp_fport; } struct inpcb * so_sotoinpcb(struct socket *so) { return (sotoinpcb(so)); } struct tcpcb * so_sototcpcb(struct socket *so) { return (sototcpcb(so)); } /* * Create an external-format (``xinpcb'') structure using the information in * the kernel-format in_pcb structure pointed to by inp. This is done to * reduce the spew of irrelevant information over this interface, to isolate * user code from changes in the kernel structure, and potentially to provide * information-hiding if we decide that some of this information should be * hidden from users. */ void in_pcbtoxinpcb(const struct inpcb *inp, struct xinpcb *xi) { xi->xi_len = sizeof(struct xinpcb); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi->xi_socket); else bzero(&xi->xi_socket, sizeof(struct xsocket)); bcopy(&inp->inp_inc, &xi->inp_inc, sizeof(struct in_conninfo)); xi->inp_gencnt = inp->inp_gencnt; xi->inp_ppcb = inp->inp_ppcb; xi->inp_flow = inp->inp_flow; xi->inp_flowid = inp->inp_flowid; xi->inp_flowtype = inp->inp_flowtype; xi->inp_flags = inp->inp_flags; xi->inp_flags2 = inp->inp_flags2; xi->inp_rss_listen_bucket = inp->inp_rss_listen_bucket; xi->in6p_cksum = inp->in6p_cksum; xi->in6p_hops = inp->in6p_hops; xi->inp_ip_tos = inp->inp_ip_tos; xi->inp_vflag = inp->inp_vflag; xi->inp_ip_ttl = inp->inp_ip_ttl; xi->inp_ip_p = inp->inp_ip_p; xi->inp_ip_minttl = inp->inp_ip_minttl; } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent) { char faddr_str[48], laddr_str[48]; db_print_indent(indent); db_printf("%s at %p\n", name, inc); indent += 2; #ifdef INET6 if (inc->inc_flags & INC_ISIPV6) { /* IPv6. */ ip6_sprintf(laddr_str, &inc->inc6_laddr); ip6_sprintf(faddr_str, &inc->inc6_faddr); } else #endif { /* IPv4. */ inet_ntoa_r(inc->inc_laddr, laddr_str); inet_ntoa_r(inc->inc_faddr, faddr_str); } db_print_indent(indent); db_printf("inc_laddr %s inc_lport %u\n", laddr_str, ntohs(inc->inc_lport)); db_print_indent(indent); db_printf("inc_faddr %s inc_fport %u\n", faddr_str, ntohs(inc->inc_fport)); } static void db_print_inpflags(int inp_flags) { int comma; comma = 0; if (inp_flags & INP_RECVOPTS) { db_printf("%sINP_RECVOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVRETOPTS) { db_printf("%sINP_RECVRETOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVDSTADDR) { db_printf("%sINP_RECVDSTADDR", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ORIGDSTADDR) { db_printf("%sINP_ORIGDSTADDR", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HDRINCL) { db_printf("%sINP_HDRINCL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HIGHPORT) { db_printf("%sINP_HIGHPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_LOWPORT) { db_printf("%sINP_LOWPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ANONPORT) { db_printf("%sINP_ANONPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVIF) { db_printf("%sINP_RECVIF", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_MTUDISC) { db_printf("%sINP_MTUDISC", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVTTL) { db_printf("%sINP_RECVTTL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DONTFRAG) { db_printf("%sINP_DONTFRAG", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVTOS) { db_printf("%sINP_RECVTOS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_IPV6_V6ONLY) { db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_PKTINFO) { db_printf("%sIN6P_PKTINFO", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPLIMIT) { db_printf("%sIN6P_HOPLIMIT", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPOPTS) { db_printf("%sIN6P_HOPOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_DSTOPTS) { db_printf("%sIN6P_DSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDR) { db_printf("%sIN6P_RTHDR", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDRDSTOPTS) { db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_TCLASS) { db_printf("%sIN6P_TCLASS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_AUTOFLOWLABEL) { db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_TIMEWAIT) { db_printf("%sINP_TIMEWAIT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ONESBCAST) { db_printf("%sINP_ONESBCAST", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DROPPED) { db_printf("%sINP_DROPPED", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_SOCKREF) { db_printf("%sINP_SOCKREF", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RFC2292) { db_printf("%sIN6P_RFC2292", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_MTU) { db_printf("IN6P_MTU%s", comma ? ", " : ""); comma = 1; } } static void db_print_inpvflag(u_char inp_vflag) { int comma; comma = 0; if (inp_vflag & INP_IPV4) { db_printf("%sINP_IPV4", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6) { db_printf("%sINP_IPV6", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6PROTO) { db_printf("%sINP_IPV6PROTO", comma ? ", " : ""); comma = 1; } } static void db_print_inpcb(struct inpcb *inp, const char *name, int indent) { db_print_indent(indent); db_printf("%s at %p\n", name, inp); indent += 2; db_print_indent(indent); db_printf("inp_flow: 0x%x\n", inp->inp_flow); db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent); db_print_indent(indent); db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n", inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket); db_print_indent(indent); db_printf("inp_label: %p inp_flags: 0x%x (", inp->inp_label, inp->inp_flags); db_print_inpflags(inp->inp_flags); db_printf(")\n"); db_print_indent(indent); db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp, inp->inp_vflag); db_print_inpvflag(inp->inp_vflag); db_printf(")\n"); db_print_indent(indent); db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n", inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl); db_print_indent(indent); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { db_printf("in6p_options: %p in6p_outputopts: %p " "in6p_moptions: %p\n", inp->in6p_options, inp->in6p_outputopts, inp->in6p_moptions); db_printf("in6p_icmp6filt: %p in6p_cksum %d " "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum, inp->in6p_hops); } else #endif { db_printf("inp_ip_tos: %d inp_ip_options: %p " "inp_ip_moptions: %p\n", inp->inp_ip_tos, inp->inp_options, inp->inp_moptions); } db_print_indent(indent); db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd, (uintmax_t)inp->inp_gencnt); } DB_SHOW_COMMAND(inpcb, db_show_inpcb) { struct inpcb *inp; if (!have_addr) { db_printf("usage: show inpcb \n"); return; } inp = (struct inpcb *)addr; db_print_inpcb(inp, "inpcb", 0); } #endif /* DDB */ #ifdef RATELIMIT /* * Modify TX rate limit based on the existing "inp->inp_snd_tag", * if any. */ int in_pcbmodify_txrtlmt(struct inpcb *inp, uint32_t max_pacing_rate) { union if_snd_tag_modify_params params = { .rate_limit.max_rate = max_pacing_rate, }; struct m_snd_tag *mst; struct ifnet *ifp; int error; mst = inp->inp_snd_tag; if (mst == NULL) return (EINVAL); ifp = mst->ifp; if (ifp == NULL) return (EINVAL); if (ifp->if_snd_tag_modify == NULL) { error = EOPNOTSUPP; } else { error = ifp->if_snd_tag_modify(mst, ¶ms); } return (error); } /* * Query existing TX rate limit based on the existing * "inp->inp_snd_tag", if any. */ int in_pcbquery_txrtlmt(struct inpcb *inp, uint32_t *p_max_pacing_rate) { union if_snd_tag_query_params params = { }; struct m_snd_tag *mst; struct ifnet *ifp; int error; mst = inp->inp_snd_tag; if (mst == NULL) return (EINVAL); ifp = mst->ifp; if (ifp == NULL) return (EINVAL); if (ifp->if_snd_tag_query == NULL) { error = EOPNOTSUPP; } else { error = ifp->if_snd_tag_query(mst, ¶ms); if (error == 0 && p_max_pacing_rate != NULL) *p_max_pacing_rate = params.rate_limit.max_rate; } return (error); } /* * Query existing TX queue level based on the existing * "inp->inp_snd_tag", if any. */ int in_pcbquery_txrlevel(struct inpcb *inp, uint32_t *p_txqueue_level) { union if_snd_tag_query_params params = { }; struct m_snd_tag *mst; struct ifnet *ifp; int error; mst = inp->inp_snd_tag; if (mst == NULL) return (EINVAL); ifp = mst->ifp; if (ifp == NULL) return (EINVAL); if (ifp->if_snd_tag_query == NULL) return (EOPNOTSUPP); error = ifp->if_snd_tag_query(mst, ¶ms); if (error == 0 && p_txqueue_level != NULL) *p_txqueue_level = params.rate_limit.queue_level; return (error); } /* * Allocate a new TX rate limit send tag from the network interface * given by the "ifp" argument and save it in "inp->inp_snd_tag": */ int in_pcbattach_txrtlmt(struct inpcb *inp, struct ifnet *ifp, uint32_t flowtype, uint32_t flowid, uint32_t max_pacing_rate) { union if_snd_tag_alloc_params params = { .rate_limit.hdr.type = (max_pacing_rate == -1U) ? IF_SND_TAG_TYPE_UNLIMITED : IF_SND_TAG_TYPE_RATE_LIMIT, .rate_limit.hdr.flowid = flowid, .rate_limit.hdr.flowtype = flowtype, .rate_limit.max_rate = max_pacing_rate, }; int error; INP_WLOCK_ASSERT(inp); if (inp->inp_snd_tag != NULL) return (EINVAL); if (ifp->if_snd_tag_alloc == NULL) { error = EOPNOTSUPP; } else { error = ifp->if_snd_tag_alloc(ifp, ¶ms, &inp->inp_snd_tag); /* * At success increment the refcount on * the send tag's network interface: */ if (error == 0) if_ref(inp->inp_snd_tag->ifp); } return (error); } /* * Free an existing TX rate limit tag based on the "inp->inp_snd_tag", * if any: */ void in_pcbdetach_txrtlmt(struct inpcb *inp) { struct m_snd_tag *mst; struct ifnet *ifp; INP_WLOCK_ASSERT(inp); mst = inp->inp_snd_tag; inp->inp_snd_tag = NULL; if (mst == NULL) return; ifp = mst->ifp; if (ifp == NULL) return; /* * If the device was detached while we still had reference(s) * on the ifp, we assume if_snd_tag_free() was replaced with * stubs. */ ifp->if_snd_tag_free(mst); /* release reference count on network interface */ if_rele(ifp); } /* * This function should be called when the INP_RATE_LIMIT_CHANGED flag * is set in the fast path and will attach/detach/modify the TX rate * limit send tag based on the socket's so_max_pacing_rate value. */ void in_pcboutput_txrtlmt(struct inpcb *inp, struct ifnet *ifp, struct mbuf *mb) { struct socket *socket; uint32_t max_pacing_rate; bool did_upgrade; int error; if (inp == NULL) return; socket = inp->inp_socket; if (socket == NULL) return; if (!INP_WLOCKED(inp)) { /* * NOTE: If the write locking fails, we need to bail * out and use the non-ratelimited ring for the * transmit until there is a new chance to get the * write lock. */ if (!INP_TRY_UPGRADE(inp)) return; did_upgrade = 1; } else { did_upgrade = 0; } /* * NOTE: The so_max_pacing_rate value is read unlocked, * because atomic updates are not required since the variable * is checked at every mbuf we send. It is assumed that the * variable read itself will be atomic. */ max_pacing_rate = socket->so_max_pacing_rate; /* * NOTE: When attaching to a network interface a reference is * made to ensure the network interface doesn't go away until * all ratelimit connections are gone. The network interface * pointers compared below represent valid network interfaces, * except when comparing towards NULL. */ if (max_pacing_rate == 0 && inp->inp_snd_tag == NULL) { error = 0; } else if (!(ifp->if_capenable & IFCAP_TXRTLMT)) { if (inp->inp_snd_tag != NULL) in_pcbdetach_txrtlmt(inp); error = 0; } else if (inp->inp_snd_tag == NULL) { /* * In order to utilize packet pacing with RSS, we need * to wait until there is a valid RSS hash before we * can proceed: */ if (M_HASHTYPE_GET(mb) == M_HASHTYPE_NONE) { error = EAGAIN; } else { error = in_pcbattach_txrtlmt(inp, ifp, M_HASHTYPE_GET(mb), mb->m_pkthdr.flowid, max_pacing_rate); } } else { error = in_pcbmodify_txrtlmt(inp, max_pacing_rate); } if (error == 0 || error == EOPNOTSUPP) inp->inp_flags2 &= ~INP_RATE_LIMIT_CHANGED; if (did_upgrade) INP_DOWNGRADE(inp); } /* * Track route changes for TX rate limiting. */ void in_pcboutput_eagain(struct inpcb *inp) { struct socket *socket; bool did_upgrade; if (inp == NULL) return; socket = inp->inp_socket; if (socket == NULL) return; if (inp->inp_snd_tag == NULL) return; if (!INP_WLOCKED(inp)) { /* * NOTE: If the write locking fails, we need to bail * out and use the non-ratelimited ring for the * transmit until there is a new chance to get the * write lock. */ if (!INP_TRY_UPGRADE(inp)) return; did_upgrade = 1; } else { did_upgrade = 0; } /* detach rate limiting */ in_pcbdetach_txrtlmt(inp); /* make sure new mbuf send tag allocation is made */ inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED; if (did_upgrade) INP_DOWNGRADE(inp); } #endif /* RATELIMIT */ Index: head/sys/netinet/in_pcb.h =================================================================== --- head/sys/netinet/in_pcb.h (revision 334718) +++ head/sys/netinet/in_pcb.h (revision 334719) @@ -1,859 +1,889 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1990, 1993 * The Regents of the University of California. * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, 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. * 3. 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. * * @(#)in_pcb.h 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ #ifndef _NETINET_IN_PCB_H_ #define _NETINET_IN_PCB_H_ #include #include #include #include #include #include #ifdef _KERNEL #include #include #include #include #endif #define in6pcb inpcb /* for KAME src sync over BSD*'s */ #define in6p_sp inp_sp /* for KAME src sync over BSD*'s */ /* * struct inpcb is the common protocol control block structure used in most * IP transport protocols. * * Pointers to local and foreign host table entries, local and foreign socket * numbers, and pointers up (to a socket structure) and down (to a * protocol-specific control block) are stored here. */ LIST_HEAD(inpcbhead, inpcb); LIST_HEAD(inpcbporthead, inpcbport); typedef uint64_t inp_gen_t; /* * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet. * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing * the following structure. */ struct in_addr_4in6 { u_int32_t ia46_pad32[3]; struct in_addr ia46_addr4; }; +union in_dependaddr { + struct in_addr_4in6 id46_addr; + struct in6_addr id6_addr; +}; + /* * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has * some extra padding to accomplish this. * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport, * lport, faddr to generate hash, so these fields shouldn't be moved. */ struct in_endpoints { u_int16_t ie_fport; /* foreign port */ u_int16_t ie_lport; /* local port */ /* protocol dependent part, local and foreign addr */ - union { - /* foreign host table entry */ - struct in_addr_4in6 ie46_foreign; - struct in6_addr ie6_foreign; - } ie_dependfaddr; - union { - /* local host table entry */ - struct in_addr_4in6 ie46_local; - struct in6_addr ie6_local; - } ie_dependladdr; + union in_dependaddr ie_dependfaddr; /* foreign host table entry */ + union in_dependaddr ie_dependladdr; /* local host table entry */ +#define ie_faddr ie_dependfaddr.id46_addr.ia46_addr4 +#define ie_laddr ie_dependladdr.id46_addr.ia46_addr4 +#define ie6_faddr ie_dependfaddr.id6_addr +#define ie6_laddr ie_dependladdr.id6_addr u_int32_t ie6_zoneid; /* scope zone id */ }; -#define ie_faddr ie_dependfaddr.ie46_foreign.ia46_addr4 -#define ie_laddr ie_dependladdr.ie46_local.ia46_addr4 -#define ie6_faddr ie_dependfaddr.ie6_foreign -#define ie6_laddr ie_dependladdr.ie6_local /* * XXX The defines for inc_* are hacks and should be changed to direct * references. */ struct in_conninfo { u_int8_t inc_flags; u_int8_t inc_len; u_int16_t inc_fibnum; /* XXX was pad, 16 bits is plenty */ /* protocol dependent part */ struct in_endpoints inc_ie; }; /* * Flags for inc_flags. */ #define INC_ISIPV6 0x01 #define inc_isipv6 inc_flags /* temp compatibility */ #define inc_fport inc_ie.ie_fport #define inc_lport inc_ie.ie_lport #define inc_faddr inc_ie.ie_faddr #define inc_laddr inc_ie.ie_laddr #define inc6_faddr inc_ie.ie6_faddr #define inc6_laddr inc_ie.ie6_laddr #define inc6_zoneid inc_ie.ie6_zoneid #if defined(_KERNEL) || defined(_WANT_INPCB) /* * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and * IPv6 sockets. In the case of TCP and UDP, further per-connection state is * hung off of inp_ppcb most of the time. Almost all fields of struct inpcb * are static after creation or protected by a per-inpcb rwlock, inp_lock. A * few fields are protected by multiple locks as indicated in the locking notes * below. For these fields, all of the listed locks must be write-locked for * any modifications. However, these fields can be safely read while any one of * the listed locks are read-locked. This model can permit greater concurrency * for read operations. For example, connections can be looked up while only * holding a read lock on the global pcblist lock. This is important for * performance when attempting to find the connection for a packet given its IP * and port tuple. * * One noteworthy exception is that the global pcbinfo lock follows a different * set of rules in relation to the inp_list field. Rather than being * write-locked for modifications and read-locked for list iterations, it must * be read-locked during modifications and write-locked during list iterations. * This ensures that the relatively rare global list iterations safely walk a * stable snapshot of connections while allowing more common list modifications * to safely grab the pcblist lock just while adding or removing a connection * from the global list. * * Key: * (b) - Protected by the hpts lock. * (c) - Constant after initialization * (g) - Protected by the pcbgroup lock * (i) - Protected by the inpcb lock * (p) - Protected by the pcbinfo lock for the inpcb * (l) - Protected by the pcblist lock for the inpcb * (h) - Protected by the pcbhash lock for the inpcb * (s) - Protected by another subsystem's locks * (x) - Undefined locking * * Notes on the tcp_hpts: * * First Hpts lock order is * 1) INP_WLOCK() * 2) HPTS_LOCK() i.e. hpts->pmtx * * To insert a TCB on the hpts you *must* be holding the INP_WLOCK(). * You may check the inp->inp_in_hpts flag without the hpts lock. * The hpts is the only one that will clear this flag holding * only the hpts lock. This means that in your tcp_output() * routine when you test for the inp_in_hpts flag to be 1 * it may be transitioning to 0 (by the hpts). * That's ok since that will just mean an extra call to tcp_output * that most likely will find the call you executed * (when the mis-match occured) will have put the TCB back * on the hpts and it will return. If your * call did not add the inp back to the hpts then you will either * over-send or the cwnd will block you from sending more. * * Note you should also be holding the INP_WLOCK() when you * call the remove from the hpts as well. Though usually * you are either doing this from a timer, where you need and have * the INP_WLOCK() or from destroying your TCB where again * you should already have the INP_WLOCK(). * * The inp_hpts_cpu, inp_hpts_cpu_set, inp_input_cpu and * inp_input_cpu_set fields are controlled completely by * the hpts. Do not ever set these. The inp_hpts_cpu_set * and inp_input_cpu_set fields indicate if the hpts has * setup the respective cpu field. It is advised if this * field is 0, to enqueue the packet with the appropriate * hpts_immediate() call. If the _set field is 1, then * you may compare the inp_*_cpu field to the curcpu and * may want to again insert onto the hpts if these fields * are not equal (i.e. you are not on the expected CPU). * * A note on inp_hpts_calls and inp_input_calls, these * flags are set when the hpts calls either the output * or do_segment routines respectively. If the routine * being called wants to use this, then it needs to * clear the flag before returning. The hpts will not * clear the flag. The flags can be used to tell if * the hpts is the function calling the respective * routine. * * A few other notes: * * When a read lock is held, stability of the field is guaranteed; to write * to a field, a write lock must generally be held. * * netinet/netinet6-layer code should not assume that the inp_socket pointer * is safe to dereference without inp_lock being held, even for protocols * other than TCP (where the inpcb persists during TIMEWAIT even after the * socket has been freed), or there may be close(2)-related races. * * The inp_vflag field is overloaded, and would otherwise ideally be (c). * * TODO: Currently only the TCP stack is leveraging the global pcbinfo lock * read-lock usage during modification, this model can be applied to other * protocols (especially SCTP). */ struct icmp6_filter; struct inpcbpolicy; struct m_snd_tag; struct inpcb { /* Cache line #1 (amd64) */ LIST_ENTRY(inpcb) inp_hash; /* (h/i) hash list */ LIST_ENTRY(inpcb) inp_pcbgrouphash; /* (g/i) hash list */ struct rwlock inp_lock; /* Cache line #2 (amd64) */ #define inp_start_zero inp_hpts #define inp_zero_size (sizeof(struct inpcb) - \ offsetof(struct inpcb, inp_start_zero)) TAILQ_ENTRY(inpcb) inp_hpts; /* pacing out queue next lock(b) */ uint32_t inp_hpts_request; /* Current hpts request, zero if * fits in the pacing window (i&b). */ /* * Note the next fields are protected by a * different lock (hpts-lock). This means that * they must correspond in size to the smallest * protectable bit field (uint8_t on x86, and * other platfomrs potentially uint32_t?). Also * since CPU switches can occur at different times the two * fields can *not* be collapsed into a signal bit field. */ #if defined(__amd64__) || defined(__i386__) volatile uint8_t inp_in_hpts; /* on output hpts (lock b) */ volatile uint8_t inp_in_input; /* on input hpts (lock b) */ #else volatile uint32_t inp_in_hpts; /* on output hpts (lock b) */ volatile uint32_t inp_in_input; /* on input hpts (lock b) */ #endif volatile uint16_t inp_hpts_cpu; /* Lock (i) */ u_int inp_refcount; /* (i) refcount */ int inp_flags; /* (i) generic IP/datagram flags */ int inp_flags2; /* (i) generic IP/datagram flags #2*/ volatile uint16_t inp_input_cpu; /* Lock (i) */ volatile uint8_t inp_hpts_cpu_set :1, /* on output hpts (i) */ inp_input_cpu_set : 1, /* on input hpts (i) */ inp_hpts_calls :1, /* (i) from output hpts */ inp_input_calls :1, /* (i) from input hpts */ inp_spare_bits2 : 4; uint8_t inp_spare_byte; /* Compiler hole */ void *inp_ppcb; /* (i) pointer to per-protocol pcb */ struct socket *inp_socket; /* (i) back pointer to socket */ uint32_t inp_hptsslot; /* Hpts wheel slot this tcb is Lock(i&b) */ uint32_t inp_hpts_drop_reas; /* reason we are dropping the PCB (lock i&b) */ TAILQ_ENTRY(inpcb) inp_input; /* pacing in queue next lock(b) */ struct inpcbinfo *inp_pcbinfo; /* (c) PCB list info */ struct inpcbgroup *inp_pcbgroup; /* (g/i) PCB group list */ LIST_ENTRY(inpcb) inp_pcbgroup_wild; /* (g/i/h) group wildcard entry */ struct ucred *inp_cred; /* (c) cache of socket cred */ u_int32_t inp_flow; /* (i) IPv6 flow information */ u_char inp_vflag; /* (i) IP version flag (v4/v6) */ u_char inp_ip_ttl; /* (i) time to live proto */ u_char inp_ip_p; /* (c) protocol proto */ u_char inp_ip_minttl; /* (i) minimum TTL or drop */ uint32_t inp_flowid; /* (x) flow id / queue id */ struct m_snd_tag *inp_snd_tag; /* (i) send tag for outgoing mbufs */ uint32_t inp_flowtype; /* (x) M_HASHTYPE value */ uint32_t inp_rss_listen_bucket; /* (x) overridden RSS listen bucket */ /* Local and foreign ports, local and foreign addr. */ struct in_conninfo inp_inc; /* (i) list for PCB's local port */ /* MAC and IPSEC policy information. */ struct label *inp_label; /* (i) MAC label */ struct inpcbpolicy *inp_sp; /* (s) for IPSEC */ /* Protocol-dependent part; options. */ struct { u_char inp_ip_tos; /* (i) type of service proto */ struct mbuf *inp_options; /* (i) IP options */ struct ip_moptions *inp_moptions; /* (i) mcast options */ }; struct { /* (i) IP options */ struct mbuf *in6p_options; /* (i) IP6 options for outgoing packets */ struct ip6_pktopts *in6p_outputopts; /* (i) IP multicast options */ struct ip6_moptions *in6p_moptions; /* (i) ICMPv6 code type filter */ struct icmp6_filter *in6p_icmp6filt; /* (i) IPV6_CHECKSUM setsockopt */ int in6p_cksum; short in6p_hops; }; LIST_ENTRY(inpcb) inp_portlist; /* (i/h) */ struct inpcbport *inp_phd; /* (i/h) head of this list */ inp_gen_t inp_gencnt; /* (c) generation count */ struct llentry *inp_lle; /* cached L2 information */ rt_gen_t inp_rt_cookie; /* generation for route entry */ union { /* cached L3 information */ struct route inp_route; struct route_in6 inp_route6; }; LIST_ENTRY(inpcb) inp_list; /* (p/l) list for all PCBs for proto */ /* (p[w]) for list iteration */ /* (p[r]/l) for addition/removal */ }; #endif /* _KERNEL */ #define inp_fport inp_inc.inc_fport #define inp_lport inp_inc.inc_lport #define inp_faddr inp_inc.inc_faddr #define inp_laddr inp_inc.inc_laddr #define in6p_faddr inp_inc.inc6_faddr #define in6p_laddr inp_inc.inc6_laddr #define in6p_zoneid inp_inc.inc6_zoneid #define in6p_flowinfo inp_flow #define inp_vnet inp_pcbinfo->ipi_vnet /* * The range of the generation count, as used in this implementation, is 9e19. * We would have to create 300 billion connections per second for this number * to roll over in a year. This seems sufficiently unlikely that we simply * don't concern ourselves with that possibility. */ /* * Interface exported to userland by various protocols which use inpcbs. Hack * alert -- only define if struct xsocket is in scope. * Fields prefixed with "xi_" are unique to this structure, and the rest * match fields in the struct inpcb, to ease coding and porting. * * Legend: * (s) - used by userland utilities in src * (p) - used by utilities in ports * (3) - is known to be used by third party software not in ports * (n) - no known usage */ #ifdef _SYS_SOCKETVAR_H_ struct xinpcb { size_t xi_len; /* length of this structure */ struct xsocket xi_socket; /* (s,p) */ struct in_conninfo inp_inc; /* (s,p) */ uint64_t inp_gencnt; /* (s,p) */ union { void *inp_ppcb; /* (s) netstat(1) */ int64_t ph_ppcb; }; int64_t inp_spare64[4]; uint32_t inp_flow; /* (s) */ uint32_t inp_flowid; /* (s) */ uint32_t inp_flowtype; /* (s) */ int32_t inp_flags; /* (s,p) */ int32_t inp_flags2; /* (s) */ int32_t inp_rss_listen_bucket; /* (n) */ int32_t in6p_cksum; /* (n) */ int32_t inp_spare32[4]; uint16_t in6p_hops; /* (n) */ uint8_t inp_ip_tos; /* (n) */ int8_t pad8; uint8_t inp_vflag; /* (s,p) */ uint8_t inp_ip_ttl; /* (n) */ uint8_t inp_ip_p; /* (n) */ uint8_t inp_ip_minttl; /* (n) */ int8_t inp_spare8[4]; } __aligned(8); struct xinpgen { size_t xig_len; /* length of this structure */ u_int xig_count; /* number of PCBs at this time */ inp_gen_t xig_gen; /* generation count at this time */ so_gen_t xig_sogen; /* socket generation count this time */ } __aligned(8); #ifdef _KERNEL void in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *); #endif #endif /* _SYS_SOCKETVAR_H_ */ struct inpcbport { LIST_ENTRY(inpcbport) phd_hash; struct inpcbhead phd_pcblist; u_short phd_port; }; struct in_pcblist { int il_count; struct epoch_context il_epoch_ctx; struct inpcbinfo *il_pcbinfo; struct inpcb *il_inp_list[0]; }; /*- * Global data structure for each high-level protocol (UDP, TCP, ...) in both * IPv4 and IPv6. Holds inpcb lists and information for managing them. * * Each pcbinfo is protected by three locks: ipi_lock, ipi_hash_lock and * ipi_list_lock: * - ipi_lock covering the global pcb list stability during loop iteration, * - ipi_hash_lock covering the hashed lookup tables, * - ipi_list_lock covering mutable global fields (such as the global * pcb list) * * The lock order is: * * ipi_lock (before) * inpcb locks (before) * ipi_list locks (before) * {ipi_hash_lock, pcbgroup locks} * * Locking key: * * (c) Constant or nearly constant after initialisation * (g) Locked by ipi_lock * (l) Locked by ipi_list_lock * (h) Read using either ipi_hash_lock or inpcb lock; write requires both * (p) Protected by one or more pcbgroup locks * (x) Synchronisation properties poorly defined */ struct inpcbinfo { /* * Global lock protecting full inpcb list traversal */ struct rwlock ipi_lock; /* * Global list of inpcbs on the protocol. */ struct inpcbhead *ipi_listhead; /* (g/l) */ u_int ipi_count; /* (l) */ /* * Generation count -- incremented each time a connection is allocated * or freed. */ u_quad_t ipi_gencnt; /* (l) */ /* * Fields associated with port lookup and allocation. */ u_short ipi_lastport; /* (x) */ u_short ipi_lastlow; /* (x) */ u_short ipi_lasthi; /* (x) */ /* * UMA zone from which inpcbs are allocated for this protocol. */ struct uma_zone *ipi_zone; /* (c) */ /* * Connection groups associated with this protocol. These fields are * constant, but pcbgroup structures themselves are protected by * per-pcbgroup locks. */ struct inpcbgroup *ipi_pcbgroups; /* (c) */ u_int ipi_npcbgroups; /* (c) */ u_int ipi_hashfields; /* (c) */ /* * Global lock protecting non-pcbgroup hash lookup tables. */ struct rwlock ipi_hash_lock; /* * Global hash of inpcbs, hashed by local and foreign addresses and * port numbers. */ struct inpcbhead *ipi_hashbase; /* (h) */ u_long ipi_hashmask; /* (h) */ /* * Global hash of inpcbs, hashed by only local port number. */ struct inpcbporthead *ipi_porthashbase; /* (h) */ u_long ipi_porthashmask; /* (h) */ /* * List of wildcard inpcbs for use with pcbgroups. In the past, was * per-pcbgroup but is now global. All pcbgroup locks must be held * to modify the list, so any is sufficient to read it. */ struct inpcbhead *ipi_wildbase; /* (p) */ u_long ipi_wildmask; /* (p) */ /* + * Load balance groups used for the SO_REUSEPORT_LB option, + * hashed by local port. + */ + struct inpcblbgrouphead *ipi_lbgrouphashbase; /* (h) */ + u_long ipi_lbgrouphashmask; /* (h) */ + + /* * Pointer to network stack instance */ struct vnet *ipi_vnet; /* (c) */ /* * general use 2 */ void *ipi_pspare[2]; /* * Global lock protecting global inpcb list, inpcb count, etc. */ struct rwlock ipi_list_lock; }; #ifdef _KERNEL /* * Connection groups hold sets of connections that have similar CPU/thread * affinity. Each connection belongs to exactly one connection group. */ struct inpcbgroup { /* * Per-connection group hash of inpcbs, hashed by local and foreign * addresses and port numbers. */ struct inpcbhead *ipg_hashbase; /* (c) */ u_long ipg_hashmask; /* (c) */ /* * Notional affinity of this pcbgroup. */ u_int ipg_cpu; /* (p) */ /* * Per-connection group lock, not to be confused with ipi_lock. * Protects the hash table hung off the group, but also the global * wildcard list in inpcbinfo. */ struct mtx ipg_lock; } __aligned(CACHE_LINE_SIZE); +/* + * Load balance groups used for the SO_REUSEPORT_LB socket option. Each group + * (or unique address:port combination) can be re-used at most + * INPCBLBGROUP_SIZMAX (256) times. The inpcbs are stored in il_inp which + * is dynamically resized as processes bind/unbind to that specific group. + */ +struct inpcblbgroup { + LIST_ENTRY(inpcblbgroup) il_list; + uint16_t il_lport; /* (c) */ + u_char il_vflag; /* (c) */ + u_char il_pad; + uint32_t il_pad2; + union in_dependaddr il_dependladdr; /* (c) */ +#define il_laddr il_dependladdr.id46_addr.ia46_addr4 +#define il6_laddr il_dependladdr.id6_addr + uint32_t il_inpsiz; /* max count in il_inp[] (h) */ + uint32_t il_inpcnt; /* cur count in il_inp[] (h) */ + struct inpcb *il_inp[]; /* (h) */ +}; +LIST_HEAD(inpcblbgrouphead, inpcblbgroup); + #define INP_LOCK_INIT(inp, d, t) \ rw_init_flags(&(inp)->inp_lock, (t), RW_RECURSE | RW_DUPOK) #define INP_LOCK_DESTROY(inp) rw_destroy(&(inp)->inp_lock) #define INP_RLOCK(inp) rw_rlock(&(inp)->inp_lock) #define INP_WLOCK(inp) rw_wlock(&(inp)->inp_lock) #define INP_TRY_RLOCK(inp) rw_try_rlock(&(inp)->inp_lock) #define INP_TRY_WLOCK(inp) rw_try_wlock(&(inp)->inp_lock) #define INP_RUNLOCK(inp) rw_runlock(&(inp)->inp_lock) #define INP_WUNLOCK(inp) rw_wunlock(&(inp)->inp_lock) #define INP_TRY_UPGRADE(inp) rw_try_upgrade(&(inp)->inp_lock) #define INP_DOWNGRADE(inp) rw_downgrade(&(inp)->inp_lock) #define INP_WLOCKED(inp) rw_wowned(&(inp)->inp_lock) #define INP_LOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_LOCKED) #define INP_RLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_RLOCKED) #define INP_WLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_WLOCKED) #define INP_UNLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_UNLOCKED) /* * These locking functions are for inpcb consumers outside of sys/netinet, * more specifically, they were added for the benefit of TOE drivers. The * macros are reserved for use by the stack. */ void inp_wlock(struct inpcb *); void inp_wunlock(struct inpcb *); void inp_rlock(struct inpcb *); void inp_runlock(struct inpcb *); #ifdef INVARIANT_SUPPORT void inp_lock_assert(struct inpcb *); void inp_unlock_assert(struct inpcb *); #else #define inp_lock_assert(inp) do {} while (0) #define inp_unlock_assert(inp) do {} while (0) #endif void inp_apply_all(void (*func)(struct inpcb *, void *), void *arg); int inp_ip_tos_get(const struct inpcb *inp); void inp_ip_tos_set(struct inpcb *inp, int val); struct socket * inp_inpcbtosocket(struct inpcb *inp); struct tcpcb * inp_inpcbtotcpcb(struct inpcb *inp); void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, uint32_t *faddr, uint16_t *fp); -short inp_so_options(const struct inpcb *inp); +int inp_so_options(const struct inpcb *inp); #endif /* _KERNEL */ #define INP_INFO_LOCK_INIT(ipi, d) \ rw_init_flags(&(ipi)->ipi_lock, (d), RW_RECURSE) #define INP_INFO_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_lock) #define INP_INFO_RLOCK(ipi) rw_rlock(&(ipi)->ipi_lock) #define INP_INFO_WLOCK(ipi) rw_wlock(&(ipi)->ipi_lock) #define INP_INFO_TRY_RLOCK(ipi) rw_try_rlock(&(ipi)->ipi_lock) #define INP_INFO_TRY_WLOCK(ipi) rw_try_wlock(&(ipi)->ipi_lock) #define INP_INFO_TRY_UPGRADE(ipi) rw_try_upgrade(&(ipi)->ipi_lock) #define INP_INFO_WLOCKED(ipi) rw_wowned(&(ipi)->ipi_lock) #define INP_INFO_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_lock) #define INP_INFO_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_lock) #define INP_INFO_LOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_LOCKED) #define INP_INFO_RLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_RLOCKED) #define INP_INFO_WLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_WLOCKED) #define INP_INFO_UNLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_lock, RA_UNLOCKED) #define INP_LIST_LOCK_INIT(ipi, d) \ rw_init_flags(&(ipi)->ipi_list_lock, (d), 0) #define INP_LIST_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_list_lock) #define INP_LIST_RLOCK(ipi) rw_rlock(&(ipi)->ipi_list_lock) #define INP_LIST_WLOCK(ipi) rw_wlock(&(ipi)->ipi_list_lock) #define INP_LIST_TRY_RLOCK(ipi) rw_try_rlock(&(ipi)->ipi_list_lock) #define INP_LIST_TRY_WLOCK(ipi) rw_try_wlock(&(ipi)->ipi_list_lock) #define INP_LIST_TRY_UPGRADE(ipi) rw_try_upgrade(&(ipi)->ipi_list_lock) #define INP_LIST_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_list_lock) #define INP_LIST_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_list_lock) #define INP_LIST_LOCK_ASSERT(ipi) \ rw_assert(&(ipi)->ipi_list_lock, RA_LOCKED) #define INP_LIST_RLOCK_ASSERT(ipi) \ rw_assert(&(ipi)->ipi_list_lock, RA_RLOCKED) #define INP_LIST_WLOCK_ASSERT(ipi) \ rw_assert(&(ipi)->ipi_list_lock, RA_WLOCKED) #define INP_LIST_UNLOCK_ASSERT(ipi) \ rw_assert(&(ipi)->ipi_list_lock, RA_UNLOCKED) #define INP_HASH_LOCK_INIT(ipi, d) \ rw_init_flags(&(ipi)->ipi_hash_lock, (d), 0) #define INP_HASH_LOCK_DESTROY(ipi) rw_destroy(&(ipi)->ipi_hash_lock) #define INP_HASH_RLOCK(ipi) rw_rlock(&(ipi)->ipi_hash_lock) #define INP_HASH_WLOCK(ipi) rw_wlock(&(ipi)->ipi_hash_lock) #define INP_HASH_RUNLOCK(ipi) rw_runlock(&(ipi)->ipi_hash_lock) #define INP_HASH_WUNLOCK(ipi) rw_wunlock(&(ipi)->ipi_hash_lock) #define INP_HASH_LOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_hash_lock, \ RA_LOCKED) #define INP_HASH_WLOCK_ASSERT(ipi) rw_assert(&(ipi)->ipi_hash_lock, \ RA_WLOCKED) #define INP_GROUP_LOCK_INIT(ipg, d) mtx_init(&(ipg)->ipg_lock, (d), NULL, \ MTX_DEF | MTX_DUPOK) #define INP_GROUP_LOCK_DESTROY(ipg) mtx_destroy(&(ipg)->ipg_lock) #define INP_GROUP_LOCK(ipg) mtx_lock(&(ipg)->ipg_lock) #define INP_GROUP_LOCK_ASSERT(ipg) mtx_assert(&(ipg)->ipg_lock, MA_OWNED) #define INP_GROUP_UNLOCK(ipg) mtx_unlock(&(ipg)->ipg_lock) #define INP_PCBHASH(faddr, lport, fport, mask) \ (((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) & (mask)) #define INP_PCBPORTHASH(lport, mask) \ (ntohs((lport)) & (mask)) +#define INP_PCBLBGROUP_PORTHASH(lport, mask) \ + (ntohs((lport)) & (mask)) +#define INP_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ + ((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) #define INP6_PCBHASHKEY(faddr) ((faddr)->s6_addr32[3]) /* * Flags for inp_vflags -- historically version flags only */ #define INP_IPV4 0x1 #define INP_IPV6 0x2 #define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ /* * Flags for inp_flags. */ #define INP_RECVOPTS 0x00000001 /* receive incoming IP options */ #define INP_RECVRETOPTS 0x00000002 /* receive IP options for reply */ #define INP_RECVDSTADDR 0x00000004 /* receive IP dst address */ #define INP_HDRINCL 0x00000008 /* user supplies entire IP header */ #define INP_HIGHPORT 0x00000010 /* user wants "high" port binding */ #define INP_LOWPORT 0x00000020 /* user wants "low" port binding */ #define INP_ANONPORT 0x00000040 /* port chosen for user */ #define INP_RECVIF 0x00000080 /* receive incoming interface */ #define INP_MTUDISC 0x00000100 /* user can do MTU discovery */ /* 0x000200 unused: was INP_FAITH */ #define INP_RECVTTL 0x00000400 /* receive incoming IP TTL */ #define INP_DONTFRAG 0x00000800 /* don't fragment packet */ #define INP_BINDANY 0x00001000 /* allow bind to any address */ #define INP_INHASHLIST 0x00002000 /* in_pcbinshash() has been called */ #define INP_RECVTOS 0x00004000 /* receive incoming IP TOS */ #define IN6P_IPV6_V6ONLY 0x00008000 /* restrict AF_INET6 socket for v6 */ #define IN6P_PKTINFO 0x00010000 /* receive IP6 dst and I/F */ #define IN6P_HOPLIMIT 0x00020000 /* receive hoplimit */ #define IN6P_HOPOPTS 0x00040000 /* receive hop-by-hop options */ #define IN6P_DSTOPTS 0x00080000 /* receive dst options after rthdr */ #define IN6P_RTHDR 0x00100000 /* receive routing header */ #define IN6P_RTHDRDSTOPTS 0x00200000 /* receive dstoptions before rthdr */ #define IN6P_TCLASS 0x00400000 /* receive traffic class value */ #define IN6P_AUTOFLOWLABEL 0x00800000 /* attach flowlabel automatically */ #define INP_TIMEWAIT 0x01000000 /* in TIMEWAIT, ppcb is tcptw */ #define INP_ONESBCAST 0x02000000 /* send all-ones broadcast */ #define INP_DROPPED 0x04000000 /* protocol drop flag */ #define INP_SOCKREF 0x08000000 /* strong socket reference */ #define INP_RESERVED_0 0x10000000 /* reserved field */ #define INP_RESERVED_1 0x20000000 /* reserved field */ #define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ #define IN6P_MTU 0x80000000 /* receive path MTU */ #define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\ IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ IN6P_MTU) /* * Flags for inp_flags2. */ #define INP_LLE_VALID 0x00000001 /* cached lle is valid */ #define INP_RT_VALID 0x00000002 /* cached rtentry is valid */ #define INP_PCBGROUPWILD 0x00000004 /* in pcbgroup wildcard list */ #define INP_REUSEPORT 0x00000008 /* SO_REUSEPORT option is set */ #define INP_FREED 0x00000010 /* inp itself is not valid */ #define INP_REUSEADDR 0x00000020 /* SO_REUSEADDR option is set */ #define INP_BINDMULTI 0x00000040 /* IP_BINDMULTI option is set */ #define INP_RSS_BUCKET_SET 0x00000080 /* IP_RSS_LISTEN_BUCKET is set */ #define INP_RECVFLOWID 0x00000100 /* populate recv datagram with flow info */ #define INP_RECVRSSBUCKETID 0x00000200 /* populate recv datagram with bucket id */ #define INP_RATE_LIMIT_CHANGED 0x00000400 /* rate limit needs attention */ #define INP_ORIGDSTADDR 0x00000800 /* receive IP dst address/port */ #define INP_CANNOT_DO_ECN 0x00001000 /* The stack does not do ECN */ +#define INP_REUSEPORT_LB 0x00002000 /* SO_REUSEPORT_LB option is set */ /* * Flags passed to in_pcblookup*() functions. */ #define INPLOOKUP_WILDCARD 0x00000001 /* Allow wildcard sockets. */ #define INPLOOKUP_RLOCKPCB 0x00000002 /* Return inpcb read-locked. */ #define INPLOOKUP_WLOCKPCB 0x00000004 /* Return inpcb write-locked. */ #define INPLOOKUP_MASK (INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \ INPLOOKUP_WLOCKPCB) #define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) #define sotoin6pcb(so) sotoinpcb(so) /* for KAME src sync over BSD*'s */ #define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family #define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) /* * Constants for pcbinfo.ipi_hashfields. */ #define IPI_HASHFIELDS_NONE 0 #define IPI_HASHFIELDS_2TUPLE 1 #define IPI_HASHFIELDS_4TUPLE 2 #ifdef _KERNEL VNET_DECLARE(int, ipport_reservedhigh); VNET_DECLARE(int, ipport_reservedlow); VNET_DECLARE(int, ipport_lowfirstauto); VNET_DECLARE(int, ipport_lowlastauto); VNET_DECLARE(int, ipport_firstauto); VNET_DECLARE(int, ipport_lastauto); VNET_DECLARE(int, ipport_hifirstauto); VNET_DECLARE(int, ipport_hilastauto); VNET_DECLARE(int, ipport_randomized); VNET_DECLARE(int, ipport_randomcps); VNET_DECLARE(int, ipport_randomtime); VNET_DECLARE(int, ipport_stoprandom); VNET_DECLARE(int, ipport_tcpallocs); #define V_ipport_reservedhigh VNET(ipport_reservedhigh) #define V_ipport_reservedlow VNET(ipport_reservedlow) #define V_ipport_lowfirstauto VNET(ipport_lowfirstauto) #define V_ipport_lowlastauto VNET(ipport_lowlastauto) #define V_ipport_firstauto VNET(ipport_firstauto) #define V_ipport_lastauto VNET(ipport_lastauto) #define V_ipport_hifirstauto VNET(ipport_hifirstauto) #define V_ipport_hilastauto VNET(ipport_hilastauto) #define V_ipport_randomized VNET(ipport_randomized) #define V_ipport_randomcps VNET(ipport_randomcps) #define V_ipport_randomtime VNET(ipport_randomtime) #define V_ipport_stoprandom VNET(ipport_stoprandom) #define V_ipport_tcpallocs VNET(ipport_tcpallocs) void in_pcbinfo_destroy(struct inpcbinfo *); void in_pcbinfo_init(struct inpcbinfo *, const char *, struct inpcbhead *, int, int, char *, uma_init, u_int); int in_pcbbind_check_bindmulti(const struct inpcb *ni, const struct inpcb *oi); struct inpcbgroup * in_pcbgroup_byhash(struct inpcbinfo *, u_int, uint32_t); struct inpcbgroup * in_pcbgroup_byinpcb(struct inpcb *); struct inpcbgroup * in_pcbgroup_bytuple(struct inpcbinfo *, struct in_addr, u_short, struct in_addr, u_short); void in_pcbgroup_destroy(struct inpcbinfo *); int in_pcbgroup_enabled(struct inpcbinfo *); void in_pcbgroup_init(struct inpcbinfo *, u_int, int); void in_pcbgroup_remove(struct inpcb *); void in_pcbgroup_update(struct inpcb *); void in_pcbgroup_update_mbuf(struct inpcb *, struct mbuf *); void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); int in_pcballoc(struct socket *, struct inpcbinfo *); int in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *); int in_pcb_lport(struct inpcb *, struct in_addr *, u_short *, struct ucred *, int); int in_pcbbind_setup(struct inpcb *, struct sockaddr *, in_addr_t *, u_short *, struct ucred *); int in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *); int in_pcbconnect_mbuf(struct inpcb *, struct sockaddr *, struct ucred *, struct mbuf *); int in_pcbconnect_setup(struct inpcb *, struct sockaddr *, in_addr_t *, u_short *, in_addr_t *, u_short *, struct inpcb **, struct ucred *); void in_pcbdetach(struct inpcb *); void in_pcbdisconnect(struct inpcb *); void in_pcbdrop(struct inpcb *); void in_pcbfree(struct inpcb *); int in_pcbinshash(struct inpcb *); int in_pcbinshash_nopcbgroup(struct inpcb *); int in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *, struct ucred *); struct inpcb * in_pcblookup_local(struct inpcbinfo *, struct in_addr, u_short, int, struct ucred *); struct inpcb * in_pcblookup(struct inpcbinfo *, struct in_addr, u_int, struct in_addr, u_int, int, struct ifnet *); struct inpcb * in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int, struct in_addr, u_int, int, struct ifnet *, struct mbuf *); void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, int, struct inpcb *(*)(struct inpcb *, int)); void in_pcbref(struct inpcb *); void in_pcbrehash(struct inpcb *); void in_pcbrehash_mbuf(struct inpcb *, struct mbuf *); int in_pcbrele(struct inpcb *); int in_pcbrele_rlocked(struct inpcb *); int in_pcbrele_wlocked(struct inpcb *); void in_pcblist_rele_rlocked(epoch_context_t ctx); void in_losing(struct inpcb *); void in_pcbsetsolabel(struct socket *so); int in_getpeeraddr(struct socket *so, struct sockaddr **nam); int in_getsockaddr(struct socket *so, struct sockaddr **nam); struct sockaddr * in_sockaddr(in_port_t port, struct in_addr *addr); void in_pcbsosetlabel(struct socket *so); #ifdef RATELIMIT int in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t, uint32_t); void in_pcbdetach_txrtlmt(struct inpcb *); int in_pcbmodify_txrtlmt(struct inpcb *, uint32_t); int in_pcbquery_txrtlmt(struct inpcb *, uint32_t *); int in_pcbquery_txrlevel(struct inpcb *, uint32_t *); void in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *); void in_pcboutput_eagain(struct inpcb *); #endif #endif /* _KERNEL */ #endif /* !_NETINET_IN_PCB_H_ */ Index: head/sys/netinet/ip_output.c =================================================================== --- head/sys/netinet/ip_output.c (revision 334718) +++ head/sys/netinet/ip_output.c (revision 334719) @@ -1,1441 +1,1450 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1988, 1990, 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. * 3. 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. * * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_ratelimit.h" #include "opt_ipsec.h" #include "opt_mbuf_stress_test.h" #include "opt_mpath.h" #include "opt_route.h" #include "opt_sctp.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SCTP #include #include #endif #include #include #include #ifdef MBUF_STRESS_TEST static int mbuf_frag_size = 0; SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); #endif static void ip_mloopback(struct ifnet *, const struct mbuf *, int); extern int in_mcast_loop; extern struct protosw inetsw[]; static inline int ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error) { struct m_tag *fwd_tag = NULL; struct mbuf *m; struct in_addr odst; struct ip *ip; m = *mp; ip = mtod(m, struct ip *); /* Run through list of hooks for output packets. */ odst.s_addr = ip->ip_dst.s_addr; *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp); m = *mp; if ((*error) != 0 || m == NULL) return 1; /* Finished */ ip = mtod(m, struct ip *); /* See if destination IP address was changed by packet filter. */ if (odst.s_addr != ip->ip_dst.s_addr) { m->m_flags |= M_SKIP_FIREWALL; /* If destination is now ourself drop to ip_input(). */ if (in_localip(ip->ip_dst)) { m->m_flags |= M_FASTFWD_OURS; if (m->m_pkthdr.rcvif == NULL) m->m_pkthdr.rcvif = V_loif; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { m->m_pkthdr.csum_flags |= CSUM_DATA_VALID | CSUM_PSEUDO_HDR; m->m_pkthdr.csum_data = 0xffff; } m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID; #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP) m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; #endif *error = netisr_queue(NETISR_IP, m); return 1; /* Finished */ } bzero(dst, sizeof(*dst)); dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = ip->ip_dst; return -1; /* Reloop */ } /* See if fib was changed by packet filter. */ if ((*fibnum) != M_GETFIB(m)) { m->m_flags |= M_SKIP_FIREWALL; *fibnum = M_GETFIB(m); return -1; /* Reloop for FIB change */ } /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ if (m->m_flags & M_FASTFWD_OURS) { if (m->m_pkthdr.rcvif == NULL) m->m_pkthdr.rcvif = V_loif; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { m->m_pkthdr.csum_flags |= CSUM_DATA_VALID | CSUM_PSEUDO_HDR; m->m_pkthdr.csum_data = 0xffff; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP) m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; #endif m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID; *error = netisr_queue(NETISR_IP, m); return 1; /* Finished */ } /* Or forward to some other address? */ if ((m->m_flags & M_IP_NEXTHOP) && ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) { bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); m->m_flags |= M_SKIP_FIREWALL; m->m_flags &= ~M_IP_NEXTHOP; m_tag_delete(m, fwd_tag); return -1; /* Reloop for CHANGE of dst */ } return 0; } /* * IP output. The packet in mbuf chain m contains a skeletal IP * header (with len, off, ttl, proto, tos, src, dst). * The mbuf chain containing the packet will be freed. * The mbuf opt, if present, will not be freed. * If route ro is present and has ro_rt initialized, route lookup would be * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, * then result of route lookup is stored in ro->ro_rt. * * In the IP forwarding case, the packet will arrive with options already * inserted, so must have a NULL opt pointer. */ int ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, struct ip_moptions *imo, struct inpcb *inp) { struct rm_priotracker in_ifa_tracker; struct ip *ip; struct ifnet *ifp = NULL; /* keep compiler happy */ struct mbuf *m0; int hlen = sizeof (struct ip); int mtu; int error = 0; struct sockaddr_in *dst; const struct sockaddr_in *gw; struct in_ifaddr *ia; int isbroadcast; uint16_t ip_len, ip_off; struct route iproute; struct rtentry *rte; /* cache for ro->ro_rt */ uint32_t fibnum; #if defined(IPSEC) || defined(IPSEC_SUPPORT) int no_route_but_check_spd = 0; #endif M_ASSERTPKTHDR(m); if (inp != NULL) { INP_LOCK_ASSERT(inp); M_SETFIB(m, inp->inp_inc.inc_fibnum); if ((flags & IP_NODEFAULTFLOWID) == 0) { m->m_pkthdr.flowid = inp->inp_flowid; M_HASHTYPE_SET(m, inp->inp_flowtype); } } if (ro == NULL) { ro = &iproute; bzero(ro, sizeof (*ro)); } if (opt) { int len = 0; m = ip_insertoptions(m, opt, &len); if (len != 0) hlen = len; /* ip->ip_hl is updated above */ } ip = mtod(m, struct ip *); ip_len = ntohs(ip->ip_len); ip_off = ntohs(ip->ip_off); if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { ip->ip_v = IPVERSION; ip->ip_hl = hlen >> 2; ip_fillid(ip); IPSTAT_INC(ips_localout); } else { /* Header already set, fetch hlen from there */ hlen = ip->ip_hl << 2; } /* * dst/gw handling: * * dst can be rewritten but always points to &ro->ro_dst. * gw is readonly but can point either to dst OR rt_gateway, * therefore we need restore gw if we're redoing lookup. */ gw = dst = (struct sockaddr_in *)&ro->ro_dst; fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m); rte = ro->ro_rt; if (rte == NULL) { bzero(dst, sizeof(*dst)); dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = ip->ip_dst; } NET_EPOCH_ENTER(); again: /* * Validate route against routing table additions; * a better/more specific route might have been added. */ if (inp) RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum); /* * If there is a cached route, * check that it is to the same destination * and is still up. If not, free it and try again. * The address family should also be checked in case of sharing the * cache with IPv6. * Also check whether routing cache needs invalidation. */ rte = ro->ro_rt; if (rte && ((rte->rt_flags & RTF_UP) == 0 || rte->rt_ifp == NULL || !RT_LINK_IS_UP(rte->rt_ifp) || dst->sin_family != AF_INET || dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { RO_INVALIDATE_CACHE(ro); rte = NULL; } ia = NULL; /* * If routing to interface only, short circuit routing lookup. * The use of an all-ones broadcast address implies this; an * interface is specified by the broadcast address of an interface, * or the destination address of a ptp interface. */ if (flags & IP_SENDONES) { if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst), M_GETFIB(m)))) == NULL && (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst), M_GETFIB(m)))) == NULL) { IPSTAT_INC(ips_noroute); error = ENETUNREACH; goto bad; } ip->ip_dst.s_addr = INADDR_BROADCAST; dst->sin_addr = ip->ip_dst; ifp = ia->ia_ifp; ip->ip_ttl = 1; isbroadcast = 1; } else if (flags & IP_ROUTETOIF) { if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst), M_GETFIB(m)))) == NULL && (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0, M_GETFIB(m)))) == NULL) { IPSTAT_INC(ips_noroute); error = ENETUNREACH; goto bad; } ifp = ia->ia_ifp; ip->ip_ttl = 1; isbroadcast = ifp->if_flags & IFF_BROADCAST ? in_ifaddr_broadcast(dst->sin_addr, ia) : 0; } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && imo != NULL && imo->imo_multicast_ifp != NULL) { /* * Bypass the normal routing lookup for multicast * packets if the interface is specified. */ ifp = imo->imo_multicast_ifp; IFP_TO_IA(ifp, ia, &in_ifa_tracker); isbroadcast = 0; /* fool gcc */ } else { /* * We want to do any cloning requested by the link layer, * as this is probably required in all cases for correct * operation (as it is for ARP). */ if (rte == NULL) { #ifdef RADIX_MPATH rtalloc_mpath_fib(ro, ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), fibnum); #else in_rtalloc_ign(ro, 0, fibnum); #endif rte = ro->ro_rt; } if (rte == NULL || (rte->rt_flags & RTF_UP) == 0 || rte->rt_ifp == NULL || !RT_LINK_IS_UP(rte->rt_ifp)) { #if defined(IPSEC) || defined(IPSEC_SUPPORT) /* * There is no route for this packet, but it is * possible that a matching SPD entry exists. */ no_route_but_check_spd = 1; mtu = 0; /* Silence GCC warning. */ goto sendit; #endif IPSTAT_INC(ips_noroute); error = EHOSTUNREACH; goto bad; } ia = ifatoia(rte->rt_ifa); ifp = rte->rt_ifp; counter_u64_add(rte->rt_pksent, 1); rt_update_ro_flags(ro); if (rte->rt_flags & RTF_GATEWAY) gw = (struct sockaddr_in *)rte->rt_gateway; if (rte->rt_flags & RTF_HOST) isbroadcast = (rte->rt_flags & RTF_BROADCAST); else if (ifp->if_flags & IFF_BROADCAST) isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia); else isbroadcast = 0; } /* * Calculate MTU. If we have a route that is up, use that, * otherwise use the interface's MTU. */ if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) mtu = rte->rt_mtu; else mtu = ifp->if_mtu; /* Catch a possible divide by zero later. */ KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp)); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { m->m_flags |= M_MCAST; /* * IP destination address is multicast. Make sure "gw" * still points to the address in "ro". (It may have been * changed to point to a gateway address, above.) */ gw = dst; /* * See if the caller provided any multicast options */ if (imo != NULL) { ip->ip_ttl = imo->imo_multicast_ttl; if (imo->imo_multicast_vif != -1) ip->ip_src.s_addr = ip_mcast_src ? ip_mcast_src(imo->imo_multicast_vif) : INADDR_ANY; } else ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; /* * Confirm that the outgoing interface supports multicast. */ if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { if ((ifp->if_flags & IFF_MULTICAST) == 0) { IPSTAT_INC(ips_noroute); error = ENETUNREACH; goto bad; } } /* * If source address not specified yet, use address * of outgoing interface. */ if (ip->ip_src.s_addr == INADDR_ANY) { /* Interface may have no addresses. */ if (ia != NULL) ip->ip_src = IA_SIN(ia)->sin_addr; } if ((imo == NULL && in_mcast_loop) || (imo && imo->imo_multicast_loop)) { /* * Loop back multicast datagram if not expressly * forbidden to do so, even if we are not a member * of the group; ip_input() will filter it later, * thus deferring a hash lookup and mutex acquisition * at the expense of a cheap copy using m_copym(). */ ip_mloopback(ifp, m, hlen); } else { /* * If we are acting as a multicast router, perform * multicast forwarding as if the packet had just * arrived on the interface to which we are about * to send. The multicast forwarding function * recursively calls this function, using the * IP_FORWARDING flag to prevent infinite recursion. * * Multicasts that are looped back by ip_mloopback(), * above, will be forwarded by the ip_input() routine, * if necessary. */ if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { /* * If rsvp daemon is not running, do not * set ip_moptions. This ensures that the packet * is multicast and not just sent down one link * as prescribed by rsvpd. */ if (!V_rsvp_on) imo = NULL; if (ip_mforward && ip_mforward(ip, ifp, m, imo) != 0) { m_freem(m); goto done; } } } /* * Multicasts with a time-to-live of zero may be looped- * back, above, but must not be transmitted on a network. * Also, multicasts addressed to the loopback interface * are not sent -- the above call to ip_mloopback() will * loop back a copy. ip_input() will drop the copy if * this host does not belong to the destination group on * the loopback interface. */ if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { m_freem(m); goto done; } goto sendit; } /* * If the source address is not specified yet, use the address * of the outoing interface. */ if (ip->ip_src.s_addr == INADDR_ANY) { /* Interface may have no addresses. */ if (ia != NULL) { ip->ip_src = IA_SIN(ia)->sin_addr; } } /* * Look for broadcast address and * verify user is allowed to send * such a packet. */ if (isbroadcast) { if ((ifp->if_flags & IFF_BROADCAST) == 0) { error = EADDRNOTAVAIL; goto bad; } if ((flags & IP_ALLOWBROADCAST) == 0) { error = EACCES; goto bad; } /* don't allow broadcast messages to be fragmented */ if (ip_len > mtu) { error = EMSGSIZE; goto bad; } m->m_flags |= M_BCAST; } else { m->m_flags &= ~M_BCAST; } sendit: #if defined(IPSEC) || defined(IPSEC_SUPPORT) if (IPSEC_ENABLED(ipv4)) { if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) { if (error == EINPROGRESS) error = 0; goto done; } } /* * Check if there was a route for this packet; return error if not. */ if (no_route_but_check_spd) { IPSTAT_INC(ips_noroute); error = EHOSTUNREACH; goto bad; } /* Update variables that are affected by ipsec4_output(). */ ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; #endif /* IPSEC */ /* Jump over all PFIL processing if hooks are not active. */ if (PFIL_HOOKED(&V_inet_pfil_hook)) { switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) { case 1: /* Finished */ goto done; case 0: /* Continue normally */ ip = mtod(m, struct ip *); break; case -1: /* Need to try again */ /* Reset everything for a new round */ RO_RTFREE(ro); ro->ro_prepend = NULL; rte = NULL; gw = dst; ip = mtod(m, struct ip *); goto again; } } /* 127/8 must not appear on wire - RFC1122. */ if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { if ((ifp->if_flags & IFF_LOOPBACK) == 0) { IPSTAT_INC(ips_badaddr); error = EADDRNOTAVAIL; goto bad; } } m->m_pkthdr.csum_flags |= CSUM_IP; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) { in_delayed_cksum(m); m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) { sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); m->m_pkthdr.csum_flags &= ~CSUM_SCTP; } #endif /* * If small enough for interface, or the interface will take * care of the fragmentation for us, we can just send directly. */ if (ip_len <= mtu || (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) { ip->ip_sum = 0; if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) { ip->ip_sum = in_cksum(m, hlen); m->m_pkthdr.csum_flags &= ~CSUM_IP; } /* * Record statistics for this interface address. * With CSUM_TSO the byte/packet count will be slightly * incorrect because we count the IP+TCP headers only * once instead of for every generated packet. */ if (!(flags & IP_FORWARDING) && ia) { if (m->m_pkthdr.csum_flags & CSUM_TSO) counter_u64_add(ia->ia_ifa.ifa_opackets, m->m_pkthdr.len / m->m_pkthdr.tso_segsz); else counter_u64_add(ia->ia_ifa.ifa_opackets, 1); counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len); } #ifdef MBUF_STRESS_TEST if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) m = m_fragment(m, M_NOWAIT, mbuf_frag_size); #endif /* * Reset layer specific mbuf flags * to avoid confusing lower layers. */ m_clrprotoflags(m); IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); #ifdef RATELIMIT if (inp != NULL) { if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) in_pcboutput_txrtlmt(inp, ifp, m); /* stamp send tag on mbuf */ m->m_pkthdr.snd_tag = inp->inp_snd_tag; } else { m->m_pkthdr.snd_tag = NULL; } #endif error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro); #ifdef RATELIMIT /* check for route change */ if (error == EAGAIN) in_pcboutput_eagain(inp); #endif goto done; } /* Balk when DF bit is set or the interface didn't support TSO. */ if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { error = EMSGSIZE; IPSTAT_INC(ips_cantfrag); goto bad; } /* * Too large for interface; fragment if possible. If successful, * on return, m will point to a list of packets to be sent. */ error = ip_fragment(ip, &m, mtu, ifp->if_hwassist); if (error) goto bad; for (; m; m = m0) { m0 = m->m_nextpkt; m->m_nextpkt = 0; if (error == 0) { /* Record statistics for this interface address. */ if (ia != NULL) { counter_u64_add(ia->ia_ifa.ifa_opackets, 1); counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len); } /* * Reset layer specific mbuf flags * to avoid confusing upper layers. */ m_clrprotoflags(m); IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp, mtod(m, struct ip *), NULL); #ifdef RATELIMIT if (inp != NULL) { if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) in_pcboutput_txrtlmt(inp, ifp, m); /* stamp send tag on mbuf */ m->m_pkthdr.snd_tag = inp->inp_snd_tag; } else { m->m_pkthdr.snd_tag = NULL; } #endif error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro); #ifdef RATELIMIT /* check for route change */ if (error == EAGAIN) in_pcboutput_eagain(inp); #endif } else m_freem(m); } if (error == 0) IPSTAT_INC(ips_fragmented); done: if (ro == &iproute) RO_RTFREE(ro); else if (rte == NULL) /* * If the caller supplied a route but somehow the reference * to it has been released need to prevent the caller * calling RTFREE on it again. */ ro->ro_rt = NULL; NET_EPOCH_EXIT(); return (error); bad: m_freem(m); goto done; } /* * Create a chain of fragments which fit the given mtu. m_frag points to the * mbuf to be fragmented; on return it points to the chain with the fragments. * Return 0 if no error. If error, m_frag may contain a partially built * chain of fragments that should be freed by the caller. * * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) */ int ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, u_long if_hwassist_flags) { int error = 0; int hlen = ip->ip_hl << 2; int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ int off; struct mbuf *m0 = *m_frag; /* the original packet */ int firstlen; struct mbuf **mnext; int nfrags; uint16_t ip_len, ip_off; ip_len = ntohs(ip->ip_len); ip_off = ntohs(ip->ip_off); if (ip_off & IP_DF) { /* Fragmentation not allowed */ IPSTAT_INC(ips_cantfrag); return EMSGSIZE; } /* * Must be able to put at least 8 bytes per fragment. */ if (len < 8) return EMSGSIZE; /* * If the interface will not calculate checksums on * fragmented packets, then do it here. */ if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { in_delayed_cksum(m0); m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } #ifdef SCTP if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { sctp_delayed_cksum(m0, hlen); m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; } #endif if (len > PAGE_SIZE) { /* * Fragment large datagrams such that each segment * contains a multiple of PAGE_SIZE amount of data, * plus headers. This enables a receiver to perform * page-flipping zero-copy optimizations. * * XXX When does this help given that sender and receiver * could have different page sizes, and also mtu could * be less than the receiver's page size ? */ int newlen; off = MIN(mtu, m0->m_pkthdr.len); /* * firstlen (off - hlen) must be aligned on an * 8-byte boundary */ if (off < hlen) goto smart_frag_failure; off = ((off - hlen) & ~7) + hlen; newlen = (~PAGE_MASK) & mtu; if ((newlen + sizeof (struct ip)) > mtu) { /* we failed, go back the default */ smart_frag_failure: newlen = len; off = hlen + len; } len = newlen; } else { off = hlen + len; } firstlen = off - hlen; mnext = &m0->m_nextpkt; /* pointer to next packet */ /* * Loop through length of segment after first fragment, * make new header and copy data of each part and link onto chain. * Here, m0 is the original packet, m is the fragment being created. * The fragments are linked off the m_nextpkt of the original * packet, which after processing serves as the first fragment. */ for (nfrags = 1; off < ip_len; off += len, nfrags++) { struct ip *mhip; /* ip header on the fragment */ struct mbuf *m; int mhlen = sizeof (struct ip); m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) { error = ENOBUFS; IPSTAT_INC(ips_odropped); goto done; } /* * Make sure the complete packet header gets copied * from the originating mbuf to the newly created * mbuf. This also ensures that existing firewall * classification(s), VLAN tags and so on get copied * to the resulting fragmented packet(s): */ if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) { m_free(m); error = ENOBUFS; IPSTAT_INC(ips_odropped); goto done; } /* * In the first mbuf, leave room for the link header, then * copy the original IP header including options. The payload * goes into an additional mbuf chain returned by m_copym(). */ m->m_data += max_linkhdr; mhip = mtod(m, struct ip *); *mhip = *ip; if (hlen > sizeof (struct ip)) { mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); mhip->ip_v = IPVERSION; mhip->ip_hl = mhlen >> 2; } m->m_len = mhlen; /* XXX do we need to add ip_off below ? */ mhip->ip_off = ((off - hlen) >> 3) + ip_off; if (off + len >= ip_len) len = ip_len - off; else mhip->ip_off |= IP_MF; mhip->ip_len = htons((u_short)(len + mhlen)); m->m_next = m_copym(m0, off, len, M_NOWAIT); if (m->m_next == NULL) { /* copy failed */ m_free(m); error = ENOBUFS; /* ??? */ IPSTAT_INC(ips_odropped); goto done; } m->m_pkthdr.len = mhlen + len; #ifdef MAC mac_netinet_fragment(m0, m); #endif mhip->ip_off = htons(mhip->ip_off); mhip->ip_sum = 0; if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { mhip->ip_sum = in_cksum(m, mhlen); m->m_pkthdr.csum_flags &= ~CSUM_IP; } *mnext = m; mnext = &m->m_nextpkt; } IPSTAT_ADD(ips_ofragments, nfrags); /* * Update first fragment by trimming what's been copied out * and updating header. */ m_adj(m0, hlen + firstlen - ip_len); m0->m_pkthdr.len = hlen + firstlen; ip->ip_len = htons((u_short)m0->m_pkthdr.len); ip->ip_off = htons(ip_off | IP_MF); ip->ip_sum = 0; if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { ip->ip_sum = in_cksum(m0, hlen); m0->m_pkthdr.csum_flags &= ~CSUM_IP; } done: *m_frag = m0; return error; } void in_delayed_cksum(struct mbuf *m) { struct ip *ip; struct udphdr *uh; uint16_t cklen, csum, offset; ip = mtod(m, struct ip *); offset = ip->ip_hl << 2 ; if (m->m_pkthdr.csum_flags & CSUM_UDP) { /* if udp header is not in the first mbuf copy udplen */ if (offset + sizeof(struct udphdr) > m->m_len) m_copydata(m, offset + offsetof(struct udphdr, uh_ulen), sizeof(cklen), (caddr_t)&cklen); else { uh = (struct udphdr *)mtodo(m, offset); cklen = ntohs(uh->uh_ulen); } csum = in_cksum_skip(m, cklen + offset, offset); if (csum == 0) csum = 0xffff; } else { cklen = ntohs(ip->ip_len); csum = in_cksum_skip(m, cklen, offset); } offset += m->m_pkthdr.csum_data; /* checksum offset */ if (offset + sizeof(csum) > m->m_len) m_copyback(m, offset, sizeof(csum), (caddr_t)&csum); else *(u_short *)mtodo(m, offset) = csum; } /* * IP socket option processing. */ int ip_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp = sotoinpcb(so); int error, optval; #ifdef RSS uint32_t rss_bucket; int retval; #endif error = optval = 0; if (sopt->sopt_level != IPPROTO_IP) { error = EINVAL; if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_dir == SOPT_SET) { switch (sopt->sopt_name) { case SO_REUSEADDR: INP_WLOCK(inp); if ((so->so_options & SO_REUSEADDR) != 0) inp->inp_flags2 |= INP_REUSEADDR; else inp->inp_flags2 &= ~INP_REUSEADDR; INP_WUNLOCK(inp); error = 0; break; case SO_REUSEPORT: INP_WLOCK(inp); if ((so->so_options & SO_REUSEPORT) != 0) inp->inp_flags2 |= INP_REUSEPORT; else inp->inp_flags2 &= ~INP_REUSEPORT; INP_WUNLOCK(inp); error = 0; break; + case SO_REUSEPORT_LB: + INP_WLOCK(inp); + if ((so->so_options & SO_REUSEPORT_LB) != 0) + inp->inp_flags2 |= INP_REUSEPORT_LB; + else + inp->inp_flags2 &= ~INP_REUSEPORT_LB; + INP_WUNLOCK(inp); + error = 0; + break; case SO_SETFIB: INP_WLOCK(inp); inp->inp_inc.inc_fibnum = so->so_fibnum; INP_WUNLOCK(inp); error = 0; break; case SO_MAX_PACING_RATE: #ifdef RATELIMIT INP_WLOCK(inp); inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED; INP_WUNLOCK(inp); error = 0; #else error = EOPNOTSUPP; #endif break; default: break; } } return (error); } switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { case IP_OPTIONS: #ifdef notyet case IP_RETOPTS: #endif { struct mbuf *m; if (sopt->sopt_valsize > MLEN) { error = EMSGSIZE; break; } m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); if (m == NULL) { error = ENOBUFS; break; } m->m_len = sopt->sopt_valsize; error = sooptcopyin(sopt, mtod(m, char *), m->m_len, m->m_len); if (error) { m_free(m); break; } INP_WLOCK(inp); error = ip_pcbopts(inp, sopt->sopt_name, m); INP_WUNLOCK(inp); return (error); } case IP_BINDANY: if (sopt->sopt_td != NULL) { error = priv_check(sopt->sopt_td, PRIV_NETINET_BINDANY); if (error) break; } /* FALLTHROUGH */ case IP_BINDMULTI: #ifdef RSS case IP_RSS_LISTEN_BUCKET: #endif case IP_TOS: case IP_TTL: case IP_MINTTL: case IP_RECVOPTS: case IP_RECVRETOPTS: case IP_ORIGDSTADDR: case IP_RECVDSTADDR: case IP_RECVTTL: case IP_RECVIF: case IP_ONESBCAST: case IP_DONTFRAG: case IP_RECVTOS: case IP_RECVFLOWID: #ifdef RSS case IP_RECVRSSBUCKETID: #endif error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; switch (sopt->sopt_name) { case IP_TOS: inp->inp_ip_tos = optval; break; case IP_TTL: inp->inp_ip_ttl = optval; break; case IP_MINTTL: if (optval >= 0 && optval <= MAXTTL) inp->inp_ip_minttl = optval; else error = EINVAL; break; #define OPTSET(bit) do { \ INP_WLOCK(inp); \ if (optval) \ inp->inp_flags |= bit; \ else \ inp->inp_flags &= ~bit; \ INP_WUNLOCK(inp); \ } while (0) #define OPTSET2(bit, val) do { \ INP_WLOCK(inp); \ if (val) \ inp->inp_flags2 |= bit; \ else \ inp->inp_flags2 &= ~bit; \ INP_WUNLOCK(inp); \ } while (0) case IP_RECVOPTS: OPTSET(INP_RECVOPTS); break; case IP_RECVRETOPTS: OPTSET(INP_RECVRETOPTS); break; case IP_RECVDSTADDR: OPTSET(INP_RECVDSTADDR); break; case IP_ORIGDSTADDR: OPTSET2(INP_ORIGDSTADDR, optval); break; case IP_RECVTTL: OPTSET(INP_RECVTTL); break; case IP_RECVIF: OPTSET(INP_RECVIF); break; case IP_ONESBCAST: OPTSET(INP_ONESBCAST); break; case IP_DONTFRAG: OPTSET(INP_DONTFRAG); break; case IP_BINDANY: OPTSET(INP_BINDANY); break; case IP_RECVTOS: OPTSET(INP_RECVTOS); break; case IP_BINDMULTI: OPTSET2(INP_BINDMULTI, optval); break; case IP_RECVFLOWID: OPTSET2(INP_RECVFLOWID, optval); break; #ifdef RSS case IP_RSS_LISTEN_BUCKET: if ((optval >= 0) && (optval < rss_getnumbuckets())) { inp->inp_rss_listen_bucket = optval; OPTSET2(INP_RSS_BUCKET_SET, 1); } else { error = EINVAL; } break; case IP_RECVRSSBUCKETID: OPTSET2(INP_RECVRSSBUCKETID, optval); break; #endif } break; #undef OPTSET #undef OPTSET2 /* * Multicast socket options are processed by the in_mcast * module. */ case IP_MULTICAST_IF: case IP_MULTICAST_VIF: case IP_MULTICAST_TTL: case IP_MULTICAST_LOOP: case IP_ADD_MEMBERSHIP: case IP_DROP_MEMBERSHIP: case IP_ADD_SOURCE_MEMBERSHIP: case IP_DROP_SOURCE_MEMBERSHIP: case IP_BLOCK_SOURCE: case IP_UNBLOCK_SOURCE: case IP_MSFILTER: case MCAST_JOIN_GROUP: case MCAST_LEAVE_GROUP: case MCAST_JOIN_SOURCE_GROUP: case MCAST_LEAVE_SOURCE_GROUP: case MCAST_BLOCK_SOURCE: case MCAST_UNBLOCK_SOURCE: error = inp_setmoptions(inp, sopt); break; case IP_PORTRANGE: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; INP_WLOCK(inp); switch (optval) { case IP_PORTRANGE_DEFAULT: inp->inp_flags &= ~(INP_LOWPORT); inp->inp_flags &= ~(INP_HIGHPORT); break; case IP_PORTRANGE_HIGH: inp->inp_flags &= ~(INP_LOWPORT); inp->inp_flags |= INP_HIGHPORT; break; case IP_PORTRANGE_LOW: inp->inp_flags &= ~(INP_HIGHPORT); inp->inp_flags |= INP_LOWPORT; break; default: error = EINVAL; break; } INP_WUNLOCK(inp); break; #if defined(IPSEC) || defined(IPSEC_SUPPORT) case IP_IPSEC_POLICY: if (IPSEC_ENABLED(ipv4)) { error = IPSEC_PCBCTL(ipv4, inp, sopt); break; } /* FALLTHROUGH */ #endif /* IPSEC */ default: error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (sopt->sopt_name) { case IP_OPTIONS: case IP_RETOPTS: if (inp->inp_options) error = sooptcopyout(sopt, mtod(inp->inp_options, char *), inp->inp_options->m_len); else sopt->sopt_valsize = 0; break; case IP_TOS: case IP_TTL: case IP_MINTTL: case IP_RECVOPTS: case IP_RECVRETOPTS: case IP_ORIGDSTADDR: case IP_RECVDSTADDR: case IP_RECVTTL: case IP_RECVIF: case IP_PORTRANGE: case IP_ONESBCAST: case IP_DONTFRAG: case IP_BINDANY: case IP_RECVTOS: case IP_BINDMULTI: case IP_FLOWID: case IP_FLOWTYPE: case IP_RECVFLOWID: #ifdef RSS case IP_RSSBUCKETID: case IP_RECVRSSBUCKETID: #endif switch (sopt->sopt_name) { case IP_TOS: optval = inp->inp_ip_tos; break; case IP_TTL: optval = inp->inp_ip_ttl; break; case IP_MINTTL: optval = inp->inp_ip_minttl; break; #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) #define OPTBIT2(bit) (inp->inp_flags2 & bit ? 1 : 0) case IP_RECVOPTS: optval = OPTBIT(INP_RECVOPTS); break; case IP_RECVRETOPTS: optval = OPTBIT(INP_RECVRETOPTS); break; case IP_RECVDSTADDR: optval = OPTBIT(INP_RECVDSTADDR); break; case IP_ORIGDSTADDR: optval = OPTBIT2(INP_ORIGDSTADDR); break; case IP_RECVTTL: optval = OPTBIT(INP_RECVTTL); break; case IP_RECVIF: optval = OPTBIT(INP_RECVIF); break; case IP_PORTRANGE: if (inp->inp_flags & INP_HIGHPORT) optval = IP_PORTRANGE_HIGH; else if (inp->inp_flags & INP_LOWPORT) optval = IP_PORTRANGE_LOW; else optval = 0; break; case IP_ONESBCAST: optval = OPTBIT(INP_ONESBCAST); break; case IP_DONTFRAG: optval = OPTBIT(INP_DONTFRAG); break; case IP_BINDANY: optval = OPTBIT(INP_BINDANY); break; case IP_RECVTOS: optval = OPTBIT(INP_RECVTOS); break; case IP_FLOWID: optval = inp->inp_flowid; break; case IP_FLOWTYPE: optval = inp->inp_flowtype; break; case IP_RECVFLOWID: optval = OPTBIT2(INP_RECVFLOWID); break; #ifdef RSS case IP_RSSBUCKETID: retval = rss_hash2bucket(inp->inp_flowid, inp->inp_flowtype, &rss_bucket); if (retval == 0) optval = rss_bucket; else error = EINVAL; break; case IP_RECVRSSBUCKETID: optval = OPTBIT2(INP_RECVRSSBUCKETID); break; #endif case IP_BINDMULTI: optval = OPTBIT2(INP_BINDMULTI); break; } error = sooptcopyout(sopt, &optval, sizeof optval); break; /* * Multicast socket options are processed by the in_mcast * module. */ case IP_MULTICAST_IF: case IP_MULTICAST_VIF: case IP_MULTICAST_TTL: case IP_MULTICAST_LOOP: case IP_MSFILTER: error = inp_getmoptions(inp, sopt); break; #if defined(IPSEC) || defined(IPSEC_SUPPORT) case IP_IPSEC_POLICY: if (IPSEC_ENABLED(ipv4)) { error = IPSEC_PCBCTL(ipv4, inp, sopt); break; } /* FALLTHROUGH */ #endif /* IPSEC */ default: error = ENOPROTOOPT; break; } break; } return (error); } /* * Routine called from ip_output() to loop back a copy of an IP multicast * packet to the input queue of a specified interface. Note that this * calls the output routine of the loopback "driver", but with an interface * pointer that might NOT be a loopback interface -- evil, but easier than * replicating that code here. */ static void ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen) { struct ip *ip; struct mbuf *copym; /* * Make a deep copy of the packet because we're going to * modify the pack in order to generate checksums. */ copym = m_dup(m, M_NOWAIT); if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen)) copym = m_pullup(copym, hlen); if (copym != NULL) { /* If needed, compute the checksum and mark it as valid. */ if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { in_delayed_cksum(copym); copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID | CSUM_PSEUDO_HDR; copym->m_pkthdr.csum_data = 0xffff; } /* * We don't bother to fragment if the IP length is greater * than the interface's MTU. Can this possibly matter? */ ip = mtod(copym, struct ip *); ip->ip_sum = 0; ip->ip_sum = in_cksum(copym, hlen); if_simloop(ifp, copym, AF_INET, 0); } } Index: head/sys/netinet/udp_usrreq.c =================================================================== --- head/sys/netinet/udp_usrreq.c (revision 334718) +++ head/sys/netinet/udp_usrreq.c (revision 334719) @@ -1,1792 +1,1792 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2008 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * Copyright (c) 2014 Kevin Lo * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, 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. * 3. 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. * * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_rss.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 #ifdef INET6 #include #endif #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include /* * UDP and UDP-Lite protocols implementation. * Per RFC 768, August, 1980. * Per RFC 3828, July, 2004. */ /* * BSD 4.2 defaulted the udp checksum to be off. Turning off udp checksums * removes the only data integrity mechanism for packets and malformed * packets that would otherwise be discarded due to bad checksums, and may * cause problems (especially for NFS data blocks). */ VNET_DEFINE(int, udp_cksum) = 1; SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(udp_cksum), 0, "compute udp checksum"); int udp_log_in_vain = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, &udp_log_in_vain, 0, "Log all incoming UDP packets"); VNET_DEFINE(int, udp_blackhole) = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(udp_blackhole), 0, "Do not send port unreachables for refused connects"); u_long udp_sendspace = 9216; /* really max datagram size */ SYSCTL_ULONG(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW, &udp_sendspace, 0, "Maximum outgoing UDP datagram size"); u_long udp_recvspace = 40 * (1024 + #ifdef INET6 sizeof(struct sockaddr_in6) #else sizeof(struct sockaddr_in) #endif ); /* 40 1K datagrams */ SYSCTL_ULONG(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW, &udp_recvspace, 0, "Maximum space for incoming UDP datagrams"); VNET_DEFINE(struct inpcbhead, udb); /* from udp_var.h */ VNET_DEFINE(struct inpcbinfo, udbinfo); VNET_DEFINE(struct inpcbhead, ulitecb); VNET_DEFINE(struct inpcbinfo, ulitecbinfo); static VNET_DEFINE(uma_zone_t, udpcb_zone); #define V_udpcb_zone VNET(udpcb_zone) #ifndef UDBHASHSIZE #define UDBHASHSIZE 128 #endif VNET_PCPUSTAT_DEFINE(struct udpstat, udpstat); /* from udp_var.h */ VNET_PCPUSTAT_SYSINIT(udpstat); SYSCTL_VNET_PCPUSTAT(_net_inet_udp, UDPCTL_STATS, stats, struct udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(udpstat); #endif /* VIMAGE */ #ifdef INET static void udp_detach(struct socket *so); static int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *, struct mbuf *, struct thread *); #endif static void udp_zone_change(void *tag) { uma_zone_set_max(V_udbinfo.ipi_zone, maxsockets); uma_zone_set_max(V_udpcb_zone, maxsockets); } static int udp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp; inp = mem; INP_LOCK_INIT(inp, "inp", "udpinp"); return (0); } static int udplite_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp; inp = mem; INP_LOCK_INIT(inp, "inp", "udpliteinp"); return (0); } void udp_init(void) { /* * For now default to 2-tuple UDP hashing - until the fragment * reassembly code can also update the flowid. * * Once we can calculate the flowid that way and re-establish * a 4-tuple, flip this to 4-tuple. */ in_pcbinfo_init(&V_udbinfo, "udp", &V_udb, UDBHASHSIZE, UDBHASHSIZE, "udp_inpcb", udp_inpcb_init, IPI_HASHFIELDS_2TUPLE); V_udpcb_zone = uma_zcreate("udpcb", sizeof(struct udpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); uma_zone_set_max(V_udpcb_zone, maxsockets); uma_zone_set_warning(V_udpcb_zone, "kern.ipc.maxsockets limit reached"); EVENTHANDLER_REGISTER(maxsockets_change, udp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } void udplite_init(void) { in_pcbinfo_init(&V_ulitecbinfo, "udplite", &V_ulitecb, UDBHASHSIZE, UDBHASHSIZE, "udplite_inpcb", udplite_inpcb_init, IPI_HASHFIELDS_2TUPLE); } /* * Kernel module interface for updating udpstat. The argument is an index * into udpstat treated as an array of u_long. While this encodes the * general layout of udpstat into the caller, it doesn't encode its location, * so that future changes to add, for example, per-CPU stats support won't * cause binary compatibility problems for kernel modules. */ void kmod_udpstat_inc(int statnum) { counter_u64_add(VNET(udpstat)[statnum], 1); } int udp_newudpcb(struct inpcb *inp) { struct udpcb *up; up = uma_zalloc(V_udpcb_zone, M_NOWAIT | M_ZERO); if (up == NULL) return (ENOBUFS); inp->inp_ppcb = up; return (0); } void udp_discardcb(struct udpcb *up) { uma_zfree(V_udpcb_zone, up); } #ifdef VIMAGE static void udp_destroy(void *unused __unused) { in_pcbinfo_destroy(&V_udbinfo); uma_zdestroy(V_udpcb_zone); } VNET_SYSUNINIT(udp, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, udp_destroy, NULL); static void udplite_destroy(void *unused __unused) { in_pcbinfo_destroy(&V_ulitecbinfo); } VNET_SYSUNINIT(udplite, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, udplite_destroy, NULL); #endif #ifdef INET /* * Subroutine of udp_input(), which appends the provided mbuf chain to the * passed pcb/socket. The caller must provide a sockaddr_in via udp_in that * contains the source address. If the socket ends up being an IPv6 socket, * udp_append() will convert to a sockaddr_in6 before passing the address * into the socket code. * * In the normal case udp_append() will return 0, indicating that you * must unlock the inp. However if a tunneling protocol is in place we increment * the inpcb refcnt and unlock the inp, on return from the tunneling protocol we * then decrement the reference count. If the inp_rele returns 1, indicating the * inp is gone, we return that to the caller to tell them *not* to unlock * the inp. In the case of multi-cast this will cause the distribution * to stop (though most tunneling protocols known currently do *not* use * multicast). */ static int udp_append(struct inpcb *inp, struct ip *ip, struct mbuf *n, int off, struct sockaddr_in *udp_in) { struct sockaddr *append_sa; struct socket *so; struct mbuf *tmpopts, *opts = NULL; #ifdef INET6 struct sockaddr_in6 udp_in6; #endif struct udpcb *up; INP_LOCK_ASSERT(inp); /* * Engage the tunneling protocol. */ up = intoudpcb(inp); if (up->u_tun_func != NULL) { in_pcbref(inp); INP_RUNLOCK(inp); (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&udp_in[0], up->u_tun_ctx); INP_RLOCK(inp); return (in_pcbrele_rlocked(inp)); } off += sizeof(struct udphdr); #if defined(IPSEC) || defined(IPSEC_SUPPORT) /* Check AH/ESP integrity. */ if (IPSEC_ENABLED(ipv4) && IPSEC_CHECK_POLICY(ipv4, n, inp) != 0) { m_freem(n); return (0); } if (up->u_flags & UF_ESPINUDP) {/* IPSec UDP encaps. */ if (IPSEC_ENABLED(ipv4) && UDPENCAP_INPUT(n, off, AF_INET) != 0) return (0); /* Consumed. */ } #endif /* IPSEC */ #ifdef MAC if (mac_inpcb_check_deliver(inp, n) != 0) { m_freem(n); return (0); } #endif /* MAC */ if (inp->inp_flags & INP_CONTROLOPTS || inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) { #ifdef INET6 if (inp->inp_vflag & INP_IPV6) (void)ip6_savecontrol_v4(inp, n, &opts, NULL); else #endif /* INET6 */ ip_savecontrol(inp, &opts, ip, n); } if ((inp->inp_vflag & INP_IPV4) && (inp->inp_flags2 & INP_ORIGDSTADDR)) { tmpopts = sbcreatecontrol((caddr_t)&udp_in[1], sizeof(struct sockaddr_in), IP_ORIGDSTADDR, IPPROTO_IP); if (tmpopts) { if (opts) { tmpopts->m_next = opts; opts = tmpopts; } else opts = tmpopts; } } #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { bzero(&udp_in6, sizeof(udp_in6)); udp_in6.sin6_len = sizeof(udp_in6); udp_in6.sin6_family = AF_INET6; in6_sin_2_v4mapsin6(&udp_in[0], &udp_in6); append_sa = (struct sockaddr *)&udp_in6; } else #endif /* INET6 */ append_sa = (struct sockaddr *)&udp_in[0]; m_adj(n, off); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_rcv); if (sbappendaddr_locked(&so->so_rcv, append_sa, n, opts) == 0) { SOCKBUF_UNLOCK(&so->so_rcv); m_freem(n); if (opts) m_freem(opts); UDPSTAT_INC(udps_fullsock); } else sorwakeup_locked(so); return (0); } int udp_input(struct mbuf **mp, int *offp, int proto) { struct ip *ip; struct udphdr *uh; struct ifnet *ifp; struct inpcb *inp; uint16_t len, ip_len; struct inpcbinfo *pcbinfo; struct ip save_ip; struct sockaddr_in udp_in[2]; struct mbuf *m; struct m_tag *fwd_tag; int cscov_partial, iphlen; m = *mp; iphlen = *offp; ifp = m->m_pkthdr.rcvif; *mp = NULL; UDPSTAT_INC(udps_ipackets); /* * Strip IP options, if any; should skip this, make available to * user, and use on returned packets, but we don't yet have a way to * check the checksum with options still present. */ if (iphlen > sizeof (struct ip)) { ip_stripoptions(m); iphlen = sizeof(struct ip); } /* * Get IP and UDP header together in first mbuf. */ ip = mtod(m, struct ip *); if (m->m_len < iphlen + sizeof(struct udphdr)) { if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == NULL) { UDPSTAT_INC(udps_hdrops); return (IPPROTO_DONE); } ip = mtod(m, struct ip *); } uh = (struct udphdr *)((caddr_t)ip + iphlen); cscov_partial = (proto == IPPROTO_UDPLITE) ? 1 : 0; /* * Destination port of 0 is illegal, based on RFC768. */ if (uh->uh_dport == 0) goto badunlocked; /* * Construct sockaddr format source address. Stuff source address * and datagram in user buffer. */ bzero(&udp_in[0], sizeof(struct sockaddr_in) * 2); udp_in[0].sin_len = sizeof(struct sockaddr_in); udp_in[0].sin_family = AF_INET; udp_in[0].sin_port = uh->uh_sport; udp_in[0].sin_addr = ip->ip_src; udp_in[1].sin_len = sizeof(struct sockaddr_in); udp_in[1].sin_family = AF_INET; udp_in[1].sin_port = uh->uh_dport; udp_in[1].sin_addr = ip->ip_dst; /* * Make mbuf data length reflect UDP length. If not enough data to * reflect UDP length, drop. */ len = ntohs((u_short)uh->uh_ulen); ip_len = ntohs(ip->ip_len) - iphlen; if (proto == IPPROTO_UDPLITE && (len == 0 || len == ip_len)) { /* Zero means checksum over the complete packet. */ if (len == 0) len = ip_len; cscov_partial = 0; } if (ip_len != len) { if (len > ip_len || len < sizeof(struct udphdr)) { UDPSTAT_INC(udps_badlen); goto badunlocked; } if (proto == IPPROTO_UDP) m_adj(m, len - ip_len); } /* * Save a copy of the IP header in case we want restore it for * sending an ICMP error message in response. */ if (!V_udp_blackhole) save_ip = *ip; else memset(&save_ip, 0, sizeof(save_ip)); /* * Checksum extended UDP header and data. */ if (uh->uh_sum) { u_short uh_sum; if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID) && !cscov_partial) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) uh_sum = m->m_pkthdr.csum_data; else uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + proto)); uh_sum ^= 0xffff; } else { char b[9]; bcopy(((struct ipovly *)ip)->ih_x1, b, 9); bzero(((struct ipovly *)ip)->ih_x1, 9); ((struct ipovly *)ip)->ih_len = (proto == IPPROTO_UDP) ? uh->uh_ulen : htons(ip_len); uh_sum = in_cksum(m, len + sizeof (struct ip)); bcopy(b, ((struct ipovly *)ip)->ih_x1, 9); } if (uh_sum) { UDPSTAT_INC(udps_badsum); m_freem(m); return (IPPROTO_DONE); } } else { if (proto == IPPROTO_UDP) { UDPSTAT_INC(udps_nosum); } else { /* UDPLite requires a checksum */ /* XXX: What is the right UDPLite MIB counter here? */ m_freem(m); return (IPPROTO_DONE); } } pcbinfo = udp_get_inpcbinfo(proto); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || in_broadcast(ip->ip_dst, ifp)) { struct inpcb *last; struct inpcbhead *pcblist; struct ip_moptions *imo; INP_INFO_RLOCK(pcbinfo); pcblist = udp_get_pcblist(proto); last = NULL; LIST_FOREACH(inp, pcblist, inp_list) { if (inp->inp_lport != uh->uh_dport) continue; #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_laddr.s_addr != INADDR_ANY && inp->inp_laddr.s_addr != ip->ip_dst.s_addr) continue; if (inp->inp_faddr.s_addr != INADDR_ANY && inp->inp_faddr.s_addr != ip->ip_src.s_addr) continue; if (inp->inp_fport != 0 && inp->inp_fport != uh->uh_sport) continue; INP_RLOCK(inp); /* * XXXRW: Because we weren't holding either the inpcb * or the hash lock when we checked for a match * before, we should probably recheck now that the * inpcb lock is held. */ /* * Handle socket delivery policy for any-source * and source-specific multicast. [RFC3678] */ imo = inp->inp_moptions; if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { struct sockaddr_in group; int blocked; if (imo == NULL) { INP_RUNLOCK(inp); continue; } bzero(&group, sizeof(struct sockaddr_in)); group.sin_len = sizeof(struct sockaddr_in); group.sin_family = AF_INET; group.sin_addr = ip->ip_dst; blocked = imo_multi_filter(imo, ifp, (struct sockaddr *)&group, (struct sockaddr *)&udp_in[0]); if (blocked != MCAST_PASS) { if (blocked == MCAST_NOTGMEMBER) IPSTAT_INC(ips_notmember); if (blocked == MCAST_NOTSMEMBER || blocked == MCAST_MUTED) UDPSTAT_INC(udps_filtermcast); INP_RUNLOCK(inp); continue; } } if (last != NULL) { struct mbuf *n; if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) { UDP_PROBE(receive, NULL, last, ip, last, uh); if (udp_append(last, ip, n, iphlen, udp_in)) { goto inp_lost; } } INP_RUNLOCK(last); } last = inp; /* * Don't look for additional matches if this one does * not have either the SO_REUSEPORT or SO_REUSEADDR * socket options set. This heuristic avoids * searching through all pcbs in the common case of a * non-shared port. It assumes that an application * will never clear these options after setting them. */ if ((last->inp_socket->so_options & - (SO_REUSEPORT|SO_REUSEADDR)) == 0) + (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) break; } if (last == NULL) { /* * No matching pcb found; discard datagram. (No need * to send an ICMP Port Unreachable for a broadcast * or multicast datgram.) */ UDPSTAT_INC(udps_noportbcast); if (inp) INP_RUNLOCK(inp); INP_INFO_RUNLOCK(pcbinfo); goto badunlocked; } UDP_PROBE(receive, NULL, last, ip, last, uh); if (udp_append(last, ip, m, iphlen, udp_in) == 0) INP_RUNLOCK(last); inp_lost: INP_INFO_RUNLOCK(pcbinfo); return (IPPROTO_DONE); } /* * Locate pcb for datagram. */ /* * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ if ((m->m_flags & M_IP_NEXTHOP) && (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { struct sockaddr_in *next_hop; next_hop = (struct sockaddr_in *)(fwd_tag + 1); /* * Transparently forwarded. Pretend to be the destination. * Already got one like this? */ inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport, ip->ip_dst, uh->uh_dport, INPLOOKUP_RLOCKPCB, ifp, m); if (!inp) { /* * It's new. Try to find the ambushing socket. * Because we've rewritten the destination address, * any hardware-generated hash is ignored. */ inp = in_pcblookup(pcbinfo, ip->ip_src, uh->uh_sport, next_hop->sin_addr, next_hop->sin_port ? htons(next_hop->sin_port) : uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, ifp); } /* Remove the tag from the packet. We don't need it anymore. */ m_tag_delete(m, fwd_tag); m->m_flags &= ~M_IP_NEXTHOP; } else inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport, ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, ifp, m); if (inp == NULL) { if (udp_log_in_vain) { char src[INET_ADDRSTRLEN]; char dst[INET_ADDRSTRLEN]; log(LOG_INFO, "Connection attempt to UDP %s:%d from %s:%d\n", inet_ntoa_r(ip->ip_dst, dst), ntohs(uh->uh_dport), inet_ntoa_r(ip->ip_src, src), ntohs(uh->uh_sport)); } UDPSTAT_INC(udps_noport); if (m->m_flags & (M_BCAST | M_MCAST)) { UDPSTAT_INC(udps_noportbcast); goto badunlocked; } if (V_udp_blackhole) goto badunlocked; if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0) goto badunlocked; *ip = save_ip; icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); return (IPPROTO_DONE); } /* * Check the minimum TTL for socket. */ INP_RLOCK_ASSERT(inp); if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl) { INP_RUNLOCK(inp); m_freem(m); return (IPPROTO_DONE); } if (cscov_partial) { struct udpcb *up; up = intoudpcb(inp); if (up->u_rxcslen == 0 || up->u_rxcslen > len) { INP_RUNLOCK(inp); m_freem(m); return (IPPROTO_DONE); } } UDP_PROBE(receive, NULL, inp, ip, inp, uh); if (udp_append(inp, ip, m, iphlen, udp_in) == 0) INP_RUNLOCK(inp); return (IPPROTO_DONE); badunlocked: m_freem(m); return (IPPROTO_DONE); } #endif /* INET */ /* * Notify a udp user of an asynchronous error; just wake up so that they can * collect error status. */ struct inpcb * udp_notify(struct inpcb *inp, int errno) { /* * While udp_ctlinput() always calls udp_notify() with a read lock * when invoking it directly, in_pcbnotifyall() currently uses write * locks due to sharing code with TCP. For now, accept either a read * or a write lock, but a read lock is sufficient. */ INP_LOCK_ASSERT(inp); if ((errno == EHOSTUNREACH || errno == ENETUNREACH || errno == EHOSTDOWN) && inp->inp_route.ro_rt) { RTFREE(inp->inp_route.ro_rt); inp->inp_route.ro_rt = (struct rtentry *)NULL; } inp->inp_socket->so_error = errno; sorwakeup(inp->inp_socket); sowwakeup(inp->inp_socket); return (inp); } #ifdef INET static void udp_common_ctlinput(int cmd, struct sockaddr *sa, void *vip, struct inpcbinfo *pcbinfo) { struct ip *ip = vip; struct udphdr *uh; struct in_addr faddr; struct inpcb *inp; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (PRC_IS_REDIRECT(cmd)) { /* signal EHOSTDOWN, as it flushes the cached route */ in_pcbnotifyall(&V_udbinfo, faddr, EHOSTDOWN, udp_notify); 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. */ if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); inp = in_pcblookup(pcbinfo, faddr, uh->uh_dport, ip->ip_src, uh->uh_sport, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { INP_RLOCK_ASSERT(inp); if (inp->inp_socket != NULL) { udp_notify(inp, inetctlerrmap[cmd]); } INP_RUNLOCK(inp); } else { inp = in_pcblookup(pcbinfo, faddr, uh->uh_dport, ip->ip_src, uh->uh_sport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { struct udpcb *up; up = intoudpcb(inp); if (up->u_icmp_func != NULL) { INP_RUNLOCK(inp); (*up->u_icmp_func)(cmd, sa, vip, up->u_tun_ctx); } else { INP_RUNLOCK(inp); } } } } else in_pcbnotifyall(pcbinfo, faddr, inetctlerrmap[cmd], udp_notify); } void udp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { return (udp_common_ctlinput(cmd, sa, vip, &V_udbinfo)); } void udplite_ctlinput(int cmd, struct sockaddr *sa, void *vip) { return (udp_common_ctlinput(cmd, sa, vip, &V_ulitecbinfo)); } #endif /* INET */ static int udp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; struct in_pcblist *il; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the PCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = V_udbinfo.ipi_count; n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); return (0); } if (req->newptr != 0) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_udbinfo); gencnt = V_udbinfo.ipi_gencnt; n = V_udbinfo.ipi_count; INP_INFO_RUNLOCK(&V_udbinfo); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + n * sizeof(struct xinpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); il = malloc(sizeof(struct in_pcblist) + n * sizeof(struct inpcb *), M_TEMP, M_WAITOK|M_ZERO_INVARIANTS); inp_list = il->il_inp_list; INP_INFO_RLOCK(&V_udbinfo); for (inp = LIST_FIRST(V_udbinfo.ipi_listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseeinpcb(req->td->td_ucred, inp) == 0) { in_pcbref(inp); inp_list[i++] = inp; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_udbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; in_pcbtoxinpcb(inp, &xi); INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_RUNLOCK(inp); } il->il_count = n; il->il_pcbinfo = &V_udbinfo; epoch_call(net_epoch_preempt, &il->il_epoch_ctx, in_pcblist_rele_rlocked); 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_udbinfo); xig.xig_gen = V_udbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_udbinfo.ipi_count; INP_INFO_RUNLOCK(&V_udbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } return (error); } SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, udp_pcblist, "S,xinpcb", "List of active UDP sockets"); #ifdef INET static int udp_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_udbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { INP_RLOCK_ASSERT(inp); 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_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, udp_getcred, "S,xucred", "Get the xucred of a UDP connection"); #endif /* INET */ int udp_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp; struct udpcb *up; int isudplite, error, optval; error = 0; isudplite = (so->so_proto->pr_protocol == IPPROTO_UDPLITE) ? 1 : 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); if (sopt->sopt_level != so->so_proto->pr_protocol) { #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) { INP_WUNLOCK(inp); error = ip6_ctloutput(so, sopt); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { INP_WUNLOCK(inp); error = ip_ctloutput(so, sopt); } #endif return (error); } switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { #if defined(IPSEC) || defined(IPSEC_SUPPORT) #ifdef INET case UDP_ENCAP: if (!IPSEC_ENABLED(ipv4)) { INP_WUNLOCK(inp); return (ENOPROTOOPT); } error = UDPENCAP_PCBCTL(inp, sopt); break; #endif /* INET */ #endif /* IPSEC */ case UDPLITE_SEND_CSCOV: case UDPLITE_RECV_CSCOV: if (!isudplite) { INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); if (error != 0) break; inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); if ((optval != 0 && optval < 8) || (optval > 65535)) { INP_WUNLOCK(inp); error = EINVAL; break; } if (sopt->sopt_name == UDPLITE_SEND_CSCOV) up->u_txcslen = optval; else up->u_rxcslen = optval; INP_WUNLOCK(inp); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (sopt->sopt_name) { #if defined(IPSEC) || defined(IPSEC_SUPPORT) #ifdef INET case UDP_ENCAP: if (!IPSEC_ENABLED(ipv4)) { INP_WUNLOCK(inp); return (ENOPROTOOPT); } error = UDPENCAP_PCBCTL(inp, sopt); break; #endif /* INET */ #endif /* IPSEC */ case UDPLITE_SEND_CSCOV: case UDPLITE_RECV_CSCOV: if (!isudplite) { INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); if (sopt->sopt_name == UDPLITE_SEND_CSCOV) optval = up->u_txcslen; else optval = up->u_rxcslen; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof(optval)); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; } return (error); } #ifdef INET #define UH_WLOCKED 2 #define UH_RLOCKED 1 #define UH_UNLOCKED 0 static int udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct udpiphdr *ui; int len = m->m_pkthdr.len; struct in_addr faddr, laddr; struct cmsghdr *cm; struct inpcbinfo *pcbinfo; struct sockaddr_in *sin, src; int cscov_partial = 0; int error = 0; int ipflags; u_short fport, lport; int unlock_udbinfo, unlock_inp; u_char tos; uint8_t pr; uint16_t cscov = 0; uint32_t flowid = 0; uint8_t flowtype = M_HASHTYPE_NONE; /* * udp_output() may need to temporarily bind or connect the current * inpcb. As such, we don't know up front whether we will need the * pcbinfo lock or not. Do any work to decide what is needed up * front before acquiring any locks. */ if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { if (control) m_freem(control); m_freem(m); return (EMSGSIZE); } src.sin_family = 0; sin = (struct sockaddr_in *)addr; if (sin == NULL || (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) { INP_WLOCK(inp); unlock_inp = UH_WLOCKED; } else { INP_RLOCK(inp); unlock_inp = UH_RLOCKED; } tos = inp->inp_ip_tos; if (control != NULL) { /* * XXX: Currently, we assume all the optional information is * stored in a single mbuf. */ if (control->m_next) { if (unlock_inp == UH_WLOCKED) INP_WUNLOCK(inp); else INP_RUNLOCK(inp); m_freem(control); m_freem(m); return (EINVAL); } for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { cm = mtod(control, struct cmsghdr *); if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) { error = EINVAL; break; } if (cm->cmsg_level != IPPROTO_IP) continue; switch (cm->cmsg_type) { case IP_SENDSRCADDR: if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr))) { error = EINVAL; break; } bzero(&src, sizeof(src)); src.sin_family = AF_INET; src.sin_len = sizeof(src); src.sin_port = inp->inp_lport; src.sin_addr = *(struct in_addr *)CMSG_DATA(cm); break; case IP_TOS: if (cm->cmsg_len != CMSG_LEN(sizeof(u_char))) { error = EINVAL; break; } tos = *(u_char *)CMSG_DATA(cm); break; case IP_FLOWID: if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) { error = EINVAL; break; } flowid = *(uint32_t *) CMSG_DATA(cm); break; case IP_FLOWTYPE: if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) { error = EINVAL; break; } flowtype = *(uint32_t *) CMSG_DATA(cm); break; #ifdef RSS case IP_RSSBUCKETID: if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) { error = EINVAL; break; } /* This is just a placeholder for now */ break; #endif /* RSS */ default: error = ENOPROTOOPT; break; } if (error) break; } m_freem(control); } if (error) { if (unlock_inp == UH_WLOCKED) INP_WUNLOCK(inp); else INP_RUNLOCK(inp); m_freem(m); return (error); } /* * Depending on whether or not the application has bound or connected * the socket, we may have to do varying levels of work. The optimal * case is for a connected UDP socket, as a global lock isn't * required at all. * * In order to decide which we need, we require stability of the * inpcb binding, which we ensure by acquiring a read lock on the * inpcb. This doesn't strictly follow the lock order, so we play * the trylock and retry game; note that we may end up with more * conservative locks than required the second time around, so later * assertions have to accept that. Further analysis of the number of * misses under contention is required. * * XXXRW: Check that hash locking update here is correct. */ pr = inp->inp_socket->so_proto->pr_protocol; pcbinfo = udp_get_inpcbinfo(pr); sin = (struct sockaddr_in *)addr; if (sin != NULL && (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) { INP_HASH_WLOCK(pcbinfo); unlock_udbinfo = UH_WLOCKED; } else if ((sin != NULL && ( (sin->sin_addr.s_addr == INADDR_ANY) || (sin->sin_addr.s_addr == INADDR_BROADCAST) || (inp->inp_laddr.s_addr == INADDR_ANY) || (inp->inp_lport == 0))) || (src.sin_family == AF_INET)) { INP_HASH_RLOCK(pcbinfo); unlock_udbinfo = UH_RLOCKED; } else unlock_udbinfo = UH_UNLOCKED; /* * If the IP_SENDSRCADDR control message was specified, override the * source address for this datagram. Its use is invalidated if the * address thus specified is incomplete or clobbers other inpcbs. */ laddr = inp->inp_laddr; lport = inp->inp_lport; if (src.sin_family == AF_INET) { INP_HASH_LOCK_ASSERT(pcbinfo); if ((lport == 0) || (laddr.s_addr == INADDR_ANY && src.sin_addr.s_addr == INADDR_ANY)) { error = EINVAL; goto release; } error = in_pcbbind_setup(inp, (struct sockaddr *)&src, &laddr.s_addr, &lport, td->td_ucred); if (error) goto release; } /* * If a UDP socket has been connected, then a local address/port will * have been selected and bound. * * If a UDP socket has not been connected to, then an explicit * destination address must be used, in which case a local * address/port may not have been selected and bound. */ if (sin != NULL) { INP_LOCK_ASSERT(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { error = EISCONN; goto release; } /* * Jail may rewrite the destination address, so let it do * that before we use it. */ error = prison_remote_ip4(td->td_ucred, &sin->sin_addr); if (error) goto release; /* * If a local address or port hasn't yet been selected, or if * the destination address needs to be rewritten due to using * a special INADDR_ constant, invoke in_pcbconnect_setup() * to do the heavy lifting. Once a port is selected, we * commit the binding back to the socket; we also commit the * binding of the address if in jail. * * If we already have a valid binding and we're not * requesting a destination address rewrite, use a fast path. */ if (inp->inp_laddr.s_addr == INADDR_ANY || inp->inp_lport == 0 || sin->sin_addr.s_addr == INADDR_ANY || sin->sin_addr.s_addr == INADDR_BROADCAST) { INP_HASH_LOCK_ASSERT(pcbinfo); error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport, &faddr.s_addr, &fport, NULL, td->td_ucred); if (error) goto release; /* * XXXRW: Why not commit the port if the address is * !INADDR_ANY? */ /* Commit the local port if newly assigned. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); /* * Remember addr if jailed, to prevent * rebinding. */ if (prison_flag(td->td_ucred, PR_IP4)) inp->inp_laddr = laddr; inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->inp_lport = 0; error = EAGAIN; goto release; } inp->inp_flags |= INP_ANONPORT; } } else { faddr = sin->sin_addr; fport = sin->sin_port; } } else { INP_LOCK_ASSERT(inp); faddr = inp->inp_faddr; fport = inp->inp_fport; if (faddr.s_addr == INADDR_ANY) { error = ENOTCONN; goto release; } } /* * Calculate data length and get a mbuf for UDP, IP, and possible * link-layer headers. Immediate slide the data pointer back forward * since we won't use that space at this layer. */ M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_NOWAIT); if (m == NULL) { error = ENOBUFS; goto release; } m->m_data += max_linkhdr; m->m_len -= max_linkhdr; m->m_pkthdr.len -= max_linkhdr; /* * Fill in mbuf with extended UDP header and addresses and length put * into network format. */ ui = mtod(m, struct udpiphdr *); bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */ ui->ui_pr = pr; ui->ui_src = laddr; ui->ui_dst = faddr; ui->ui_sport = lport; ui->ui_dport = fport; ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); if (pr == IPPROTO_UDPLITE) { struct udpcb *up; uint16_t plen; up = intoudpcb(inp); cscov = up->u_txcslen; plen = (u_short)len + sizeof(struct udphdr); if (cscov >= plen) cscov = 0; ui->ui_len = htons(plen); ui->ui_ulen = htons(cscov); /* * For UDP-Lite, checksum coverage length of zero means * the entire UDPLite packet is covered by the checksum. */ cscov_partial = (cscov == 0) ? 0 : 1; } else ui->ui_v = IPVERSION << 4; /* * Set the Don't Fragment bit in the IP header. */ if (inp->inp_flags & INP_DONTFRAG) { struct ip *ip; ip = (struct ip *)&ui->ui_i; ip->ip_off |= htons(IP_DF); } ipflags = 0; if (inp->inp_socket->so_options & SO_DONTROUTE) ipflags |= IP_ROUTETOIF; if (inp->inp_socket->so_options & SO_BROADCAST) ipflags |= IP_ALLOWBROADCAST; if (inp->inp_flags & INP_ONESBCAST) ipflags |= IP_SENDONES; #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif /* * Set up checksum and output datagram. */ ui->ui_sum = 0; if (pr == IPPROTO_UDPLITE) { if (inp->inp_flags & INP_ONESBCAST) faddr.s_addr = INADDR_BROADCAST; if (cscov_partial) { if ((ui->ui_sum = in_cksum(m, sizeof(struct ip) + cscov)) == 0) ui->ui_sum = 0xffff; } else { if ((ui->ui_sum = in_cksum(m, sizeof(struct udpiphdr) + len)) == 0) ui->ui_sum = 0xffff; } } else if (V_udp_cksum) { if (inp->inp_flags & INP_ONESBCAST) faddr.s_addr = INADDR_BROADCAST; ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr, htons((u_short)len + sizeof(struct udphdr) + pr)); m->m_pkthdr.csum_flags = CSUM_UDP; m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); } ((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len); ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ ((struct ip *)ui)->ip_tos = tos; /* XXX */ UDPSTAT_INC(udps_opackets); /* * Setup flowid / RSS information for outbound socket. * * Once the UDP code decides to set a flowid some other way, * this allows the flowid to be overridden by userland. */ if (flowtype != M_HASHTYPE_NONE) { m->m_pkthdr.flowid = flowid; M_HASHTYPE_SET(m, flowtype); #ifdef RSS } else { uint32_t hash_val, hash_type; /* * Calculate an appropriate RSS hash for UDP and * UDP Lite. * * The called function will take care of figuring out * whether a 2-tuple or 4-tuple hash is required based * on the currently configured scheme. * * Later later on connected socket values should be * cached in the inpcb and reused, rather than constantly * re-calculating it. * * UDP Lite is a different protocol number and will * likely end up being hashed as a 2-tuple until * RSS / NICs grow UDP Lite protocol awareness. */ if (rss_proto_software_hash_v4(faddr, laddr, fport, lport, pr, &hash_val, &hash_type) == 0) { m->m_pkthdr.flowid = hash_val; M_HASHTYPE_SET(m, hash_type); } #endif } #ifdef RSS /* * Don't override with the inp cached flowid value. * * Depending upon the kind of send being done, the inp * flowid/flowtype values may actually not be appropriate * for this particular socket send. * * We should either leave the flowid at zero (which is what is * currently done) or set it to some software generated * hash value based on the packet contents. */ ipflags |= IP_NODEFAULTFLOWID; #endif /* RSS */ if (unlock_udbinfo == UH_WLOCKED) INP_HASH_WUNLOCK(pcbinfo); else if (unlock_udbinfo == UH_RLOCKED) INP_HASH_RUNLOCK(pcbinfo); UDP_PROBE(send, NULL, inp, &ui->ui_i, inp, &ui->ui_u); error = ip_output(m, inp->inp_options, (unlock_inp == UH_WLOCKED ? &inp->inp_route : NULL), ipflags, inp->inp_moptions, inp); if (unlock_inp == UH_WLOCKED) INP_WUNLOCK(inp); else INP_RUNLOCK(inp); return (error); release: if (unlock_udbinfo == UH_WLOCKED) { KASSERT(unlock_inp == UH_WLOCKED, ("%s: excl udbinfo lock, shared inp lock", __func__)); INP_HASH_WUNLOCK(pcbinfo); INP_WUNLOCK(inp); } else if (unlock_udbinfo == UH_RLOCKED) { KASSERT(unlock_inp == UH_RLOCKED, ("%s: shared udbinfo lock, excl inp lock", __func__)); INP_HASH_RUNLOCK(pcbinfo); INP_RUNLOCK(inp); } else if (unlock_inp == UH_WLOCKED) INP_WUNLOCK(inp); else INP_RUNLOCK(inp); m_freem(m); return (error); } static void udp_abort(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_abort: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_HASH_WLOCK(pcbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; INP_HASH_WUNLOCK(pcbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); } static int udp_attach(struct socket *so, int proto, struct thread *td) { static uint32_t udp_flowid; struct inpcb *inp; struct inpcbinfo *pcbinfo; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp == NULL, ("udp_attach: inp != NULL")); error = soreserve(so, udp_sendspace, udp_recvspace); if (error) return (error); INP_INFO_WLOCK(pcbinfo); error = in_pcballoc(so, pcbinfo); if (error) { INP_INFO_WUNLOCK(pcbinfo); return (error); } inp = sotoinpcb(so); inp->inp_vflag |= INP_IPV4; inp->inp_ip_ttl = V_ip_defttl; inp->inp_flowid = atomic_fetchadd_int(&udp_flowid, 1); inp->inp_flowtype = M_HASHTYPE_OPAQUE; error = udp_newudpcb(inp); if (error) { in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(pcbinfo); return (error); } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(pcbinfo); return (0); } #endif /* INET */ int udp_set_kernel_tunneling(struct socket *so, udp_tun_func_t f, udp_tun_icmp_t i, void *ctx) { struct inpcb *inp; struct udpcb *up; KASSERT(so->so_type == SOCK_DGRAM, ("udp_set_kernel_tunneling: !dgram")); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_set_kernel_tunneling: inp == NULL")); INP_WLOCK(inp); up = intoudpcb(inp); if ((up->u_tun_func != NULL) || (up->u_icmp_func != NULL)) { INP_WUNLOCK(inp); return (EBUSY); } up->u_tun_func = f; up->u_icmp_func = i; up->u_tun_ctx = ctx; INP_WUNLOCK(inp); return (0); } #ifdef INET static int udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_bind: inp == NULL")); INP_WLOCK(inp); INP_HASH_WLOCK(pcbinfo); error = in_pcbbind(inp, nam, td->td_ucred); INP_HASH_WUNLOCK(pcbinfo); INP_WUNLOCK(inp); return (error); } static void udp_close(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_close: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_HASH_WLOCK(pcbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; INP_HASH_WUNLOCK(pcbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); } static int udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; struct sockaddr_in *sin; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_connect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_WUNLOCK(inp); return (EISCONN); } sin = (struct sockaddr_in *)nam; error = prison_remote_ip4(td->td_ucred, &sin->sin_addr); if (error != 0) { INP_WUNLOCK(inp); return (error); } INP_HASH_WLOCK(pcbinfo); error = in_pcbconnect(inp, nam, td->td_ucred); INP_HASH_WUNLOCK(pcbinfo); if (error == 0) soisconnected(so); INP_WUNLOCK(inp); return (error); } static void udp_detach(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; struct udpcb *up; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_detach: inp == NULL")); KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, ("udp_detach: not disconnected")); INP_INFO_WLOCK(pcbinfo); INP_WLOCK(inp); up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); /* XXX defer to epoch_call */ inp->inp_ppcb = NULL; in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(pcbinfo); udp_discardcb(up); } static int udp_disconnect(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_disconnect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr == INADDR_ANY) { INP_WUNLOCK(inp); return (ENOTCONN); } INP_HASH_WLOCK(pcbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; INP_HASH_WUNLOCK(pcbinfo); SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; /* XXX */ SOCK_UNLOCK(so); INP_WUNLOCK(inp); return (0); } static int udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_send: inp == NULL")); return (udp_output(inp, m, addr, control, td)); } #endif /* INET */ int udp_shutdown(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_shutdown: inp == NULL")); INP_WLOCK(inp); socantsendmore(so); INP_WUNLOCK(inp); return (0); } #ifdef INET struct pr_usrreqs udp_usrreqs = { .pru_abort = udp_abort, .pru_attach = udp_attach, .pru_bind = udp_bind, .pru_connect = udp_connect, .pru_control = in_control, .pru_detach = udp_detach, .pru_disconnect = udp_disconnect, .pru_peeraddr = in_getpeeraddr, .pru_send = udp_send, .pru_soreceive = soreceive_dgram, .pru_sosend = sosend_dgram, .pru_shutdown = udp_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = udp_close, }; #endif /* INET */ Index: head/sys/netinet6/in6_pcb.c =================================================================== --- head/sys/netinet6/in6_pcb.c (revision 334718) +++ head/sys/netinet6/in6_pcb.c (revision 334719) @@ -1,1315 +1,1396 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, 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. * 3. Neither the name of the project 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 PROJECT 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 PROJECT 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. * * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $ */ /*- * Copyright (c) 1982, 1986, 1991, 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. * 3. 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. * * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_pcbgroup.h" #include "opt_rss.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 struct inpcb *in6_pcblookup_hash_locked(struct inpcbinfo *, struct in6_addr *, u_int, struct in6_addr *, u_int, int, struct ifnet *); int in6_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { struct socket *so = inp->inp_socket; struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; u_short lport = 0; int error, lookupflags = 0; int reuseport = (so->so_options & SO_REUSEPORT); + /* + * XXX: Maybe we could let SO_REUSEPORT_LB set SO_REUSEPORT bit here + * so that we don't have to add to the (already messy) code below. + */ + int reuseport_lb = (so->so_options & SO_REUSEPORT_LB); + INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); if (CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */ return (EADDRNOTAVAIL); if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) return (EINVAL); - if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) + if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0) lookupflags = INPLOOKUP_WILDCARD; if (nam == NULL) { if ((error = prison_local_ip6(cred, &inp->in6p_laddr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) return (error); } else { sin6 = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof(*sin6)) return (EINVAL); /* * family check. */ if (nam->sa_family != AF_INET6) return (EAFNOSUPPORT); if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) return(error); if ((error = prison_local_ip6(cred, &sin6->sin6_addr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) return (error); lport = sin6->sin6_port; if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { /* * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; * allow compepte duplication of binding if * SO_REUSEPORT is set, or if SO_REUSEADDR is set * and a multicast address is bound on both * new and duplicated sockets. */ if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0) reuseport = SO_REUSEADDR|SO_REUSEPORT; + /* + * XXX: How to deal with SO_REUSEPORT_LB here? + * Treat same as SO_REUSEPORT for now. + */ + if ((so->so_options & + (SO_REUSEADDR|SO_REUSEPORT_LB)) != 0) + reuseport_lb = SO_REUSEADDR|SO_REUSEPORT_LB; } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { struct ifaddr *ifa; sin6->sin6_port = 0; /* yech... */ NET_EPOCH_ENTER(); if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) == NULL && (inp->inp_flags & INP_BINDANY) == 0) { NET_EPOCH_EXIT(); return (EADDRNOTAVAIL); } /* * XXX: bind to an anycast address might accidentally * cause sending a packet with anycast source address. * We should allow to bind to a deprecated address, since * the application dares to use it. */ if (ifa != NULL && ((struct in6_ifaddr *)ifa)->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) { NET_EPOCH_EXIT(); return (EADDRNOTAVAIL); } NET_EPOCH_EXIT(); } if (lport) { struct inpcb *t; struct tcptw *tw; /* GROSS */ if (ntohs(lport) <= V_ipport_reservedhigh && ntohs(lport) >= V_ipport_reservedlow && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) return (EACCES); if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) && priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT, 0) != 0) { t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr, lport, INPLOOKUP_WILDCARD, cred); if (t && ((inp->inp_flags2 & INP_BINDMULTI) == 0) && ((t->inp_flags & INP_TIMEWAIT) == 0) && (so->so_type != SOCK_STREAM || IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) && (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) || - (t->inp_flags2 & INP_REUSEPORT) == 0) && + (t->inp_flags2 & INP_REUSEPORT) || + (t->inp_flags2 & INP_REUSEPORT_LB) == 0) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); /* * If the socket is a BINDMULTI socket, then * the credentials need to match and the * original socket also has to have been bound * with BINDMULTI. */ if (t && (! in_pcbbind_check_bindmulti(inp, t))) return (EADDRINUSE); #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6); t = in_pcblookup_local(pcbinfo, sin.sin_addr, lport, INPLOOKUP_WILDCARD, cred); if (t && ((inp->inp_flags2 & INP_BINDMULTI) == 0) && ((t->inp_flags & INP_TIMEWAIT) == 0) && (so->so_type != SOCK_STREAM || ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); if (t && (! in_pcbbind_check_bindmulti(inp, t))) return (EADDRINUSE); } #endif } t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr, lport, lookupflags, cred); if (t && (t->inp_flags & INP_TIMEWAIT)) { /* * XXXRW: If an incpb has had its timewait * state recycled, we treat the address as * being in use (for now). This is better * than a panic, but not desirable. */ tw = intotw(t); if (tw == NULL || - (reuseport & tw->tw_so_options) == 0) + ((reuseport & tw->tw_so_options) == 0 && + (reuseport_lb & tw->tw_so_options) == 0)) return (EADDRINUSE); - } else if (t && (reuseport & inp_so_options(t)) == 0) { + } else if (t && (reuseport & inp_so_options(t)) == 0 && + (reuseport_lb & inp_so_options(t)) == 0) { return (EADDRINUSE); } #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 && IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6); t = in_pcblookup_local(pcbinfo, sin.sin_addr, - lport, lookupflags, cred); + lport, lookupflags, cred); if (t && t->inp_flags & INP_TIMEWAIT) { tw = intotw(t); if (tw == NULL) return (EADDRINUSE); if ((reuseport & tw->tw_so_options) == 0 + && (reuseport_lb & tw->tw_so_options) == 0 && (ntohl(t->inp_laddr.s_addr) != - INADDR_ANY || ((inp->inp_vflag & - INP_IPV6PROTO) == - (t->inp_vflag & INP_IPV6PROTO)))) + INADDR_ANY || ((inp->inp_vflag & + INP_IPV6PROTO) == + (t->inp_vflag & INP_IPV6PROTO)))) return (EADDRINUSE); } else if (t && (reuseport & inp_so_options(t)) == 0 && + (reuseport_lb & inp_so_options(t)) == 0 && (ntohl(t->inp_laddr.s_addr) != INADDR_ANY || - (t->inp_vflag & INP_IPV6PROTO) != 0)) + (t->inp_vflag & INP_IPV6PROTO) != 0)) { return (EADDRINUSE); + } } #endif } inp->in6p_laddr = sin6->sin6_addr; } if (lport == 0) { if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) { /* Undo an address bind that may have occurred. */ inp->in6p_laddr = in6addr_any; return (error); } } else { inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->in6p_laddr = in6addr_any; inp->inp_lport = 0; return (EAGAIN); } } return (0); } /* * Transform old in6_pcbconnect() into an inner subroutine for new * in6_pcbconnect(): Do some validity-checking on the remote * address (in mbuf 'nam') and then determine local host address * (i.e., which interface) to use to access that remote host. * * This preserves definition of in6_pcbconnect(), while supporting a * slightly different version for T/TCP. (This is more than * a bit of a kludge, but cleaning up the internal interfaces would * have forced minor changes in every protocol). */ static int in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam, struct in6_addr *plocal_addr6) { struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; int error = 0; int scope_ambiguous = 0; struct in6_addr in6a; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */ if (nam->sa_len != sizeof (*sin6)) return (EINVAL); if (sin6->sin6_family != AF_INET6) return (EAFNOSUPPORT); if (sin6->sin6_port == 0) return (EADDRNOTAVAIL); if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) scope_ambiguous = 1; if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) return(error); if (!CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) { /* * If the destination address is UNSPECIFIED addr, * use the loopback addr, e.g ::1. */ if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) sin6->sin6_addr = in6addr_loopback; } if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0) return (error); error = in6_selectsrc_socket(sin6, inp->in6p_outputopts, inp, inp->inp_cred, scope_ambiguous, &in6a, NULL); if (error) return (error); /* * Do not update this earlier, in case we return with an error. * * XXX: this in6_selectsrc_socket result might replace the bound local * address with the address specified by setsockopt(IPV6_PKTINFO). * Is it the intended behavior? */ *plocal_addr6 = in6a; /* * Don't do pcblookup call here; return interface in * plocal_addr6 * and exit to caller, that will do the lookup. */ return (0); } /* * Outer subroutine: * Connect from a socket to a specified address. * Both address and port must be specified in argument sin. * If don't have a local address for this socket yet, * then pick one. */ int in6_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred, struct mbuf *m) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; struct in6_addr addr6; int error; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); /* * Call inner routine, to assign local interface address. * in6_pcbladdr() may automatically fill in sin6_scope_id. */ if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0) return (error); if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr, sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? &addr6 : &inp->in6p_laddr, inp->inp_lport, 0, NULL) != NULL) { return (EADDRINUSE); } if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (inp->inp_lport == 0) { error = in6_pcbbind(inp, (struct sockaddr *)0, cred); if (error) return (error); } inp->in6p_laddr = addr6; } inp->in6p_faddr = sin6->sin6_addr; inp->inp_fport = sin6->sin6_port; /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; if (inp->inp_flags & IN6P_AUTOFLOWLABEL) inp->inp_flow |= (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); in_pcbrehash_mbuf(inp, m); return (0); } int in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) { return (in6_pcbconnect_mbuf(inp, nam, cred, NULL)); } void in6_pcbdisconnect(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr)); inp->inp_fport = 0; /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; in_pcbrehash(inp); } struct sockaddr * in6_sockaddr(in_port_t port, struct in6_addr *addr_p) { struct sockaddr_in6 *sin6; sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK); bzero(sin6, sizeof *sin6); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = port; sin6->sin6_addr = *addr_p; (void)sa6_recoverscope(sin6); /* XXX: should catch errors */ return (struct sockaddr *)sin6; } struct sockaddr * in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p) { struct sockaddr_in sin; struct sockaddr_in6 *sin6_p; bzero(&sin, sizeof sin); sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_port = port; sin.sin_addr = *addr_p; sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK); in6_sin_2_v4mapsin6(&sin, sin6_p); return (struct sockaddr *)sin6_p; } int in6_getsockaddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in6_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_lport; addr = inp->in6p_laddr; INP_RUNLOCK(inp); *nam = in6_sockaddr(port, &addr); return 0; } int in6_getpeeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in6_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_fport; addr = inp->in6p_faddr; INP_RUNLOCK(inp); *nam = in6_sockaddr(port, &addr); return 0; } int in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL")); #ifdef INET if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { error = in_getsockaddr(so, nam); if (error == 0) in6_sin_2_v4mapsin6_in_sock(nam); } else #endif { /* scope issues will be handled in in6_getsockaddr(). */ error = in6_getsockaddr(so, nam); } return error; } int in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL")); #ifdef INET if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) { error = in_getpeeraddr(so, nam); if (error == 0) in6_sin_2_v4mapsin6_in_sock(nam); } else #endif /* scope issues will be handled in in6_getpeeraddr(). */ error = in6_getpeeraddr(so, nam); return error; } /* * Pass some notification to all connections of a protocol * associated with address dst. The local address and/or port numbers * may be specified to limit the search. The "usual action" will be * taken, depending on the ctlinput cmd. The caller must filter any * cmds that are uninteresting (e.g., no error in the map). * Call the protocol specific routine (if any) to report * any errors for each matching socket. */ void in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd, void *cmdarg, struct inpcb *(*notify)(struct inpcb *, int)) { struct inpcb *inp, *inp_temp; struct sockaddr_in6 sa6_src, *sa6_dst; u_short fport = fport_arg, lport = lport_arg; u_int32_t flowinfo; int errno; if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6) return; sa6_dst = (struct sockaddr_in6 *)dst; if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr)) return; /* * note that src can be NULL when we get notify by local fragmentation. */ sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src; flowinfo = sa6_src.sin6_flowinfo; /* * Redirects go to all references to the destination, * and use in6_rtchange to invalidate the route cache. * Dead host indications: also use in6_rtchange to invalidate * the cache, and deliver the error to all the sockets. * Otherwise, if we have knowledge of the local port and address, * deliver only to that socket. */ if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) { fport = 0; lport = 0; bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr)); if (cmd != PRC_HOSTDEAD) notify = in6_rtchange; } errno = inet6ctlerrmap[cmd]; INP_INFO_WLOCK(pcbinfo); LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { INP_WLOCK(inp); if ((inp->inp_vflag & INP_IPV6) == 0) { INP_WUNLOCK(inp); continue; } /* * If the error designates a new path MTU for a destination * and the application (associated with this socket) wanted to * know the value, notify. * XXX: should we avoid to notify the value to TCP sockets? */ if (cmd == PRC_MSGSIZE && cmdarg != NULL) ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, *(u_int32_t *)cmdarg); /* * Detect if we should notify the error. If no source and * destination ports are specifed, but non-zero flowinfo and * local address match, notify the error. This is the case * when the error is delivered with an encrypted buffer * by ESP. Otherwise, just compare addresses and ports * as usual. */ if (lport == 0 && fport == 0 && flowinfo && inp->inp_socket != NULL && flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr)) goto do_notify; else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr) || inp->inp_socket == 0 || (lport && inp->inp_lport != lport) || (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) && !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr)) || (fport && inp->inp_fport != fport)) { INP_WUNLOCK(inp); continue; } do_notify: if (notify) { if ((*notify)(inp, errno)) INP_WUNLOCK(inp); } else INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); } /* * Lookup a PCB based on the local address and port. Caller must hold the * hash lock. No inpcb locks or references are acquired. */ struct inpcb * in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr, u_short lport, int lookupflags, struct ucred *cred) { struct inpcb *inp; int matchwild = 3, wildcard; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); INP_HASH_WLOCK_ASSERT(pcbinfo); if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { struct inpcbhead *head; /* * Look for an unconnected (wildcard foreign addr) PCB that * matches the local address and port we're looking for. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH( INP6_PCBHASHKEY(&in6addr_any), lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && inp->inp_lport == lport) { /* Found. */ if (cred == NULL || prison_equal_ip6(cred->cr_prison, inp->inp_cred->cr_prison)) return (inp); } } /* * Not found. */ return (NULL); } else { struct inpcbporthead *porthash; struct inpcbport *phd; struct inpcb *match = NULL; /* * Best fit PCB lookup. * * First see if this local port is in use by looking on the * port hash list. */ porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, pcbinfo->ipi_porthashmask)]; LIST_FOREACH(phd, porthash, phd_hash) { if (phd->phd_port == lport) break; } if (phd != NULL) { /* * Port is in use by one or more PCBs. Look for best * fit. */ LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { wildcard = 0; if (cred != NULL && !prison_equal_ip6(cred->cr_prison, inp->inp_cred->cr_prison)) continue; /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) wildcard++; if (!IN6_IS_ADDR_UNSPECIFIED( &inp->in6p_laddr)) { if (IN6_IS_ADDR_UNSPECIFIED(laddr)) wildcard++; else if (!IN6_ARE_ADDR_EQUAL( &inp->in6p_laddr, laddr)) continue; } else { if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) wildcard++; } if (wildcard < matchwild) { match = inp; matchwild = wildcard; if (matchwild == 0) break; } } } return (match); } } void in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) { struct inpcb *in6p; struct ip6_moptions *im6o; int i, gap; INP_INFO_WLOCK(pcbinfo); LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) { INP_WLOCK(in6p); im6o = in6p->in6p_moptions; if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) { /* * Unselect the outgoing ifp for multicast if it * is being detached. */ if (im6o->im6o_multicast_ifp == ifp) im6o->im6o_multicast_ifp = NULL; /* * Drop multicast group membership if we joined * through the interface being detached. */ gap = 0; for (i = 0; i < im6o->im6o_num_memberships; i++) { if (im6o->im6o_membership[i]->in6m_ifp == ifp) { in6_leavegroup(im6o->im6o_membership[i], NULL); gap++; } else if (gap != 0) { im6o->im6o_membership[i - gap] = im6o->im6o_membership[i]; } } im6o->im6o_num_memberships -= gap; } INP_WUNLOCK(in6p); } INP_INFO_WUNLOCK(pcbinfo); } /* * Check for alternatives when higher level complains * about service problems. For now, invalidate cached * routing information. If the route was created dynamically * (by a redirect), time to try a default gateway again. */ void in6_losing(struct inpcb *in6p) { if (in6p->inp_route6.ro_rt) { RTFREE(in6p->inp_route6.ro_rt); in6p->inp_route6.ro_rt = (struct rtentry *)NULL; } if (in6p->inp_route.ro_lle) LLE_FREE(in6p->inp_route.ro_lle); /* zeros ro_lle */ return; } /* * After a routing change, flush old routing * and allocate a (hopefully) better one. */ struct inpcb * in6_rtchange(struct inpcb *inp, int errno) { if (inp->inp_route6.ro_rt) { RTFREE(inp->inp_route6.ro_rt); inp->inp_route6.ro_rt = (struct rtentry *)NULL; } if (inp->inp_route.ro_lle) LLE_FREE(inp->inp_route.ro_lle); /* zeros ro_lle */ return inp; } +static struct inpcb * +in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo, + const struct in6_addr *laddr, uint16_t lport, const struct in6_addr *faddr, + uint16_t fport, int lookupflags) +{ + struct inpcb *local_wild = NULL; + const struct inpcblbgrouphead *hdr; + struct inpcblbgroup *grp; + struct inpcblbgroup *grp_local_wild; + uint32_t idx; + + INP_HASH_LOCK_ASSERT(pcbinfo); + + hdr = &pcbinfo->ipi_lbgrouphashbase[INP_PCBLBGROUP_PORTHASH( + lport, pcbinfo->ipi_lbgrouphashmask)]; + + /* + * Order of socket selection: + * 1. non-wild. + * 2. wild (if lookupflags contains INPLOOKUP_WILDCARD). + * + * NOTE: + * - Load balanced group does not contain jailed sockets. + * - Load balanced does not contain IPv4 mapped INET6 wild sockets. + */ + LIST_FOREACH(grp, hdr, il_list) { + if (grp->il_lport == lport) { + idx = 0; + int pkt_hash = INP_PCBLBGROUP_PKTHASH( + INP6_PCBHASHKEY(faddr), lport, fport); + + idx = pkt_hash % grp->il_inpcnt; + + if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr)) { + return (grp->il_inp[idx]); + } else { + if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr) && + (lookupflags & INPLOOKUP_WILDCARD)) { + local_wild = grp->il_inp[idx]; + grp_local_wild = grp; + } + } + } + } + if (local_wild != NULL) { + return (local_wild); + } + return (NULL); +} + #ifdef PCBGROUP /* * Lookup PCB in hash list, using pcbgroup tables. */ static struct inpcb * in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup, struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; bool locked; /* * First look for an exact match. */ tmpinp = NULL; INP_GROUP_LOCK(pcbgroup); head = &pcbgroup->ipg_hashbase[INP_PCBHASH( INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)]; LIST_FOREACH(inp, head, inp_pcbgrouphash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP6)) goto found; if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) { inp = tmpinp; goto found; } /* * Then look for a wildcard match in the pcbgroup. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; struct inpcb *jail_wild = NULL; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbgroup->ipg_hashbase[ INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)]; LIST_FOREACH(inp, head, inp_pcbgrouphash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || inp->inp_lport != lport) { continue; } injail = prison_flag(inp->inp_cred, PR_IP6); if (injail) { if (prison_check_ip6(inp->inp_cred, laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { if (injail) goto found; else local_exact = inp; } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ inp = jail_wild; if (inp == NULL) inp = jail_wild; if (inp == NULL) inp = local_exact; if (inp == NULL) inp = local_wild; if (inp != NULL) goto found; } /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; struct inpcb *jail_wild = NULL; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_wildbase[INP_PCBHASH( INP6_PCBHASHKEY(&in6addr_any), lport, 0, pcbinfo->ipi_wildmask)]; LIST_FOREACH(inp, head, inp_pcbgroup_wild) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || inp->inp_lport != lport) { continue; } injail = prison_flag(inp->inp_cred, PR_IP6); if (injail) { if (prison_check_ip6(inp->inp_cred, laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { if (injail) goto found; else local_exact = inp; } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ inp = jail_wild; if (inp == NULL) inp = jail_wild; if (inp == NULL) inp = local_exact; if (inp == NULL) inp = local_wild; if (inp != NULL) goto found; } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ INP_GROUP_UNLOCK(pcbgroup); return (NULL); found: if (lookupflags & INPLOOKUP_WLOCKPCB) locked = INP_TRY_WLOCK(inp); else if (lookupflags & INPLOOKUP_RLOCKPCB) locked = INP_TRY_RLOCK(inp); else panic("%s: locking buf", __func__); if (!locked) in_pcbref(inp); INP_GROUP_UNLOCK(pcbgroup); if (!locked) { if (lookupflags & INPLOOKUP_WLOCKPCB) { INP_WLOCK(inp); if (in_pcbrele_wlocked(inp)) return (NULL); } else { INP_RLOCK(inp); if (in_pcbrele_rlocked(inp)) return (NULL); } } #ifdef INVARIANTS if (lookupflags & INPLOOKUP_WLOCKPCB) INP_WLOCK_ASSERT(inp); else INP_RLOCK_ASSERT(inp); #endif return (inp); } #endif /* PCBGROUP */ /* * Lookup PCB in hash list. */ static struct inpcb * in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr, u_int lport_arg, int lookupflags, struct ifnet *ifp) { struct inpcbhead *head; struct inpcb *inp, *tmpinp; u_short fport = fport_arg, lport = lport_arg; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); INP_HASH_LOCK_ASSERT(pcbinfo); /* * First look for an exact match. */ tmpinp = NULL; head = &pcbinfo->ipi_hashbase[INP_PCBHASH( INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) && inp->inp_fport == fport && inp->inp_lport == lport) { /* * XXX We should be able to directly return * the inp here, without any checks. * Well unless both bound with SO_REUSEPORT? */ if (prison_flag(inp->inp_cred, PR_IP6)) return (inp); if (tmpinp == NULL) tmpinp = inp; } } if (tmpinp != NULL) return (tmpinp); + + /* + * Then look in lb group (for wildcard match). + */ + if (pcbinfo->ipi_lbgrouphashbase != NULL && + (lookupflags & INPLOOKUP_WILDCARD)) { + inp = in6_pcblookup_lbgroup(pcbinfo, laddr, lport, faddr, + fport, lookupflags); + if (inp != NULL) { + return (inp); + } + } /* * Then look for a wildcard match, if requested. */ if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { struct inpcb *local_wild = NULL, *local_exact = NULL; struct inpcb *jail_wild = NULL; int injail; /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_hashbase[INP_PCBHASH( INP6_PCBHASHKEY(&in6addr_any), lport, 0, pcbinfo->ipi_hashmask)]; LIST_FOREACH(inp, head, inp_hash) { /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) || inp->inp_lport != lport) { continue; } injail = prison_flag(inp->inp_cred, PR_IP6); if (injail) { if (prison_check_ip6(inp->inp_cred, laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) { if (injail) return (inp); else local_exact = inp; } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (injail) jail_wild = inp; else local_wild = inp; } } /* LIST_FOREACH */ if (jail_wild != NULL) return (jail_wild); if (local_exact != NULL) return (local_exact); if (local_wild != NULL) return (local_wild); } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ /* * Not found. */ return (NULL); } /* * Lookup PCB in hash list, using pcbinfo tables. This variation locks the * hash list lock, and will return the inpcb locked (i.e., requires * INPLOOKUP_LOCKPCB). */ static struct inpcb * in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp) { struct inpcb *inp; bool locked; INP_HASH_RLOCK(pcbinfo); inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); if (inp != NULL) { if (lookupflags & INPLOOKUP_WLOCKPCB) locked = INP_TRY_WLOCK(inp); else if (lookupflags & INPLOOKUP_RLOCKPCB) locked = INP_TRY_RLOCK(inp); else panic("%s: locking bug", __func__); if (!locked) in_pcbref(inp); INP_HASH_RUNLOCK(pcbinfo); if (!locked) { if (lookupflags & INPLOOKUP_WLOCKPCB) { INP_WLOCK(inp); if (in_pcbrele_wlocked(inp)) return (NULL); } else { INP_RLOCK(inp); if (in_pcbrele_rlocked(inp)) return (NULL); } } #ifdef INVARIANTS if (lookupflags & INPLOOKUP_WLOCKPCB) INP_WLOCK_ASSERT(inp); else INP_RLOCK_ASSERT(inp); #endif } else INP_HASH_RUNLOCK(pcbinfo); return (inp); } /* * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf * from which a pre-calculated hash value may be extracted. * * Possibly more of this logic should be in in6_pcbgroup.c. */ struct inpcb * in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp) { #if defined(PCBGROUP) && !defined(RSS) struct inpcbgroup *pcbgroup; #endif KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); /* * When not using RSS, use connection groups in preference to the * reservation table when looking up 4-tuples. When using RSS, just * use the reservation table, due to the cost of the Toeplitz hash * in software. * * XXXRW: This policy belongs in the pcbgroup code, as in principle * we could be doing RSS with a non-Toeplitz hash that is affordable * in software. */ #if defined(PCBGROUP) && !defined(RSS) if (in_pcbgroup_enabled(pcbinfo)) { pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, fport); return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); } #endif return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, ifp)); } struct inpcb * in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp, struct mbuf *m) { #ifdef PCBGROUP struct inpcbgroup *pcbgroup; #endif KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); #ifdef PCBGROUP /* * If we can use a hardware-generated hash to look up the connection * group, use that connection group to find the inpcb. Otherwise * fall back on a software hash -- or the reservation table if we're * using RSS. * * XXXRW: As above, that policy belongs in the pcbgroup code. */ if (in_pcbgroup_enabled(pcbinfo) && M_HASHTYPE_TEST(m, M_HASHTYPE_NONE) == 0) { pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m), m->m_pkthdr.flowid); if (pcbgroup != NULL) return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); #ifndef RSS pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, fport); return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, laddr, lport, lookupflags, ifp)); #endif } #endif return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, ifp)); } void init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst) { struct ip6_hdr *ip; ip = mtod(m, struct ip6_hdr *); bzero(sin6, sizeof(*sin6)); sin6->sin6_len = sizeof(*sin6); sin6->sin6_family = AF_INET6; sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src; (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */ return; } Index: head/sys/netinet6/in6_src.c =================================================================== --- head/sys/netinet6/in6_src.c (revision 334718) +++ head/sys/netinet6/in6_src.c (revision 334719) @@ -1,1243 +1,1243 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * 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. Neither the name of the project 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 PROJECT 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 PROJECT 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. * * $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $ */ /*- * Copyright (c) 1982, 1986, 1991, 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. * 3. 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. * * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_mpath.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static struct mtx addrsel_lock; #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF) #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock) #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock) #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED) static struct sx addrsel_sxlock; #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock") #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock) #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock) #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock) #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock) #define ADDR_LABEL_NOTAPP (-1) static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy); #define V_defaultaddrpolicy VNET(defaultaddrpolicy) VNET_DEFINE(int, ip6_prefer_tempaddr) = 0; static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *, struct ip6_moptions *, struct route_in6 *, struct ifnet **, struct rtentry **, int, u_int); static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *, struct ip6_moptions *, struct ifnet **, struct ifnet *, u_int); static int in6_selectsrc(uint32_t, struct sockaddr_in6 *, struct ip6_pktopts *, struct inpcb *, struct ucred *, struct ifnet **, struct in6_addr *); static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *); static void init_policy_queue(void); static int add_addrsel_policyent(struct in6_addrpolicy *); static int delete_addrsel_policyent(struct in6_addrpolicy *); static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *), void *); static int dump_addrsel_policyent(struct in6_addrpolicy *, void *); static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *); /* * Return an IPv6 address, which is the most appropriate for a given * destination and user specified options. * If necessary, this function lookups the routing table and returns * an entry to the caller for later use. */ #define REPLACE(r) do {\ IP6STAT_INC(ip6s_sources_rule[(r)]); \ /* { \ char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ } */ \ goto replace; \ } while(0) #define NEXT(r) do {\ /* { \ char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \ printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \ } */ \ goto next; /* XXX: we can't use 'continue' here */ \ } while(0) #define BREAK(r) do { \ IP6STAT_INC(ip6s_sources_rule[(r)]); \ goto out; /* XXX: we can't use 'break' here */ \ } while(0) static int in6_selectsrc(uint32_t fibnum, struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct inpcb *inp, struct ucred *cred, struct ifnet **ifpp, struct in6_addr *srcp) { struct rm_priotracker in6_ifa_tracker; struct in6_addr dst, tmp; struct ifnet *ifp = NULL, *oifp = NULL; struct in6_ifaddr *ia = NULL, *ia_best = NULL; struct in6_pktinfo *pi = NULL; int dst_scope = -1, best_scope = -1, best_matchlen = -1; struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL; u_int32_t odstzone; int prefer_tempaddr; int error; struct ip6_moptions *mopts; KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__)); dst = dstsock->sin6_addr; /* make a copy for local operation */ if (ifpp) { /* * Save a possibly passed in ifp for in6_selectsrc. Only * neighbor discovery code should use this feature, where * we may know the interface but not the FIB number holding * the connected subnet in case someone deleted it from the * default FIB and we need to check the interface. */ if (*ifpp != NULL) oifp = *ifpp; *ifpp = NULL; } if (inp != NULL) { INP_LOCK_ASSERT(inp); mopts = inp->in6p_moptions; } else { mopts = NULL; } /* * If the source address is explicitly specified by the caller, * check if the requested source address is indeed a unicast address * assigned to the node, and can be used as the packet's source * address. If everything is okay, use the address as source. */ if (opts && (pi = opts->ip6po_pktinfo) && !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) { /* get the outgoing interface */ if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp, fibnum)) != 0) return (error); /* * determine the appropriate zone id of the source based on * the zone of the destination and the outgoing interface. * If the specified address is ambiguous wrt the scope zone, * the interface must be specified; otherwise, ifa_ifwithaddr() * will fail matching the address. */ tmp = pi->ipi6_addr; if (ifp) { error = in6_setscope(&tmp, ifp, &odstzone); if (error) return (error); } if (cred != NULL && (error = prison_local_ip6(cred, &tmp, (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) return (error); /* * If IPV6_BINDANY socket option is set, we allow to specify * non local addresses as source address in IPV6_PKTINFO * ancillary data. */ if ((inp->inp_flags & INP_BINDANY) == 0) { ia = in6ifa_ifwithaddr(&tmp, 0 /* XXX */); if (ia == NULL || (ia->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) { if (ia != NULL) ifa_free(&ia->ia_ifa); return (EADDRNOTAVAIL); } bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp)); ifa_free(&ia->ia_ifa); } else bcopy(&tmp, srcp, sizeof(*srcp)); pi->ipi6_addr = tmp; /* XXX: this overrides pi */ if (ifpp) *ifpp = ifp; return (0); } /* * Otherwise, if the socket has already bound the source, just use it. */ if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (cred != NULL && (error = prison_local_ip6(cred, &inp->in6p_laddr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0) return (error); bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp)); return (0); } /* * Bypass source address selection and use the primary jail IP * if requested. */ if (cred != NULL && !prison_saddrsel_ip6(cred, srcp)) return (0); /* * If the address is not specified, choose the best one based on * the outgoing interface and the destination address. */ /* get the outgoing interface */ if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp, (inp != NULL) ? inp->inp_inc.inc_fibnum : fibnum)) != 0) return (error); #ifdef DIAGNOSTIC if (ifp == NULL) /* this should not happen */ panic("in6_selectsrc: NULL ifp"); #endif error = in6_setscope(&dst, ifp, &odstzone); if (error) return (error); IN6_IFADDR_RLOCK(&in6_ifa_tracker); CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { int new_scope = -1, new_matchlen = -1; struct in6_addrpolicy *new_policy = NULL; u_int32_t srczone, osrczone, dstzone; struct in6_addr src; struct ifnet *ifp1 = ia->ia_ifp; /* * We'll never take an address that breaks the scope zone * of the destination. We also skip an address if its zone * does not contain the outgoing interface. * XXX: we should probably use sin6_scope_id here. */ if (in6_setscope(&dst, ifp1, &dstzone) || odstzone != dstzone) { continue; } src = ia->ia_addr.sin6_addr; if (in6_setscope(&src, ifp, &osrczone) || in6_setscope(&src, ifp1, &srczone) || osrczone != srczone) { continue; } /* avoid unusable addresses */ if ((ia->ia6_flags & (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) { continue; } if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia)) continue; /* If jailed only take addresses of the jail into account. */ if (cred != NULL && prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0) continue; /* Rule 1: Prefer same address */ if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) { ia_best = ia; BREAK(1); /* there should be no better candidate */ } if (ia_best == NULL) REPLACE(0); /* Rule 2: Prefer appropriate scope */ if (dst_scope < 0) dst_scope = in6_addrscope(&dst); new_scope = in6_addrscope(&ia->ia_addr.sin6_addr); if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) { if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0) REPLACE(2); NEXT(2); } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) { if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0) NEXT(2); REPLACE(2); } /* * Rule 3: Avoid deprecated addresses. Note that the case of * !ip6_use_deprecated is already rejected above. */ if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia)) NEXT(3); if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia)) REPLACE(3); /* Rule 4: Prefer home addresses */ /* * XXX: This is a TODO. We should probably merge the MIP6 * case above. */ /* Rule 5: Prefer outgoing interface */ if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) { if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp) NEXT(5); if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp) REPLACE(5); } /* * Rule 6: Prefer matching label * Note that best_policy should be non-NULL here. */ if (dst_policy == NULL) dst_policy = lookup_addrsel_policy(dstsock); if (dst_policy->label != ADDR_LABEL_NOTAPP) { new_policy = lookup_addrsel_policy(&ia->ia_addr); if (dst_policy->label == best_policy->label && dst_policy->label != new_policy->label) NEXT(6); if (dst_policy->label != best_policy->label && dst_policy->label == new_policy->label) REPLACE(6); } /* * Rule 7: Prefer public addresses. * We allow users to reverse the logic by configuring * a sysctl variable, so that privacy conscious users can * always prefer temporary addresses. */ if (opts == NULL || opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) { prefer_tempaddr = V_ip6_prefer_tempaddr; } else if (opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_NOTPREFER) { prefer_tempaddr = 0; } else prefer_tempaddr = 1; if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) && (ia->ia6_flags & IN6_IFF_TEMPORARY)) { if (prefer_tempaddr) REPLACE(7); else NEXT(7); } if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) && !(ia->ia6_flags & IN6_IFF_TEMPORARY)) { if (prefer_tempaddr) NEXT(7); else REPLACE(7); } /* * Rule 8: prefer addresses on alive interfaces. * This is a KAME specific rule. */ if ((ia_best->ia_ifp->if_flags & IFF_UP) && !(ia->ia_ifp->if_flags & IFF_UP)) NEXT(8); if (!(ia_best->ia_ifp->if_flags & IFF_UP) && (ia->ia_ifp->if_flags & IFF_UP)) REPLACE(8); /* * Rule 9: prefer address with better virtual status. */ if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa)) REPLACE(9); if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa)) NEXT(9); /* * Rule 10: prefer address with `prefer_source' flag. */ if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 && (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0) REPLACE(10); if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 && (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0) NEXT(10); /* * Rule 14: Use longest matching prefix. * Note: in the address selection draft, this rule is * documented as "Rule 8". However, since it is also * documented that this rule can be overridden, we assign * a large number so that it is easy to assign smaller numbers * to more preferred rules. */ new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst); if (best_matchlen < new_matchlen) REPLACE(14); if (new_matchlen < best_matchlen) NEXT(14); /* Rule 15 is reserved. */ /* * Last resort: just keep the current candidate. * Or, do we need more rules? */ continue; replace: ia_best = ia; best_scope = (new_scope >= 0 ? new_scope : in6_addrscope(&ia_best->ia_addr.sin6_addr)); best_policy = (new_policy ? new_policy : lookup_addrsel_policy(&ia_best->ia_addr)); best_matchlen = (new_matchlen >= 0 ? new_matchlen : in6_matchlen(&ia_best->ia_addr.sin6_addr, &dst)); next: continue; out: break; } if ((ia = ia_best) == NULL) { IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); IP6STAT_INC(ip6s_sources_none); return (EADDRNOTAVAIL); } /* * At this point at least one of the addresses belonged to the jail * but it could still be, that we want to further restrict it, e.g. * theoratically IN6_IS_ADDR_LOOPBACK. * It must not be IN6_IS_ADDR_UNSPECIFIED anymore. * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should * let all others previously selected pass. * Use tmp to not change ::1 on lo0 to the primary jail address. */ tmp = ia->ia_addr.sin6_addr; if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL && (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) { IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); IP6STAT_INC(ip6s_sources_none); return (EADDRNOTAVAIL); } if (ifpp) *ifpp = ifp; bcopy(&tmp, srcp, sizeof(*srcp)); if (ia->ia_ifp == ifp) IP6STAT_INC(ip6s_sources_sameif[best_scope]); else IP6STAT_INC(ip6s_sources_otherif[best_scope]); if (dst_scope == best_scope) IP6STAT_INC(ip6s_sources_samescope[best_scope]); else IP6STAT_INC(ip6s_sources_otherscope[best_scope]); if (IFA6_IS_DEPRECATED(ia)) IP6STAT_INC(ip6s_sources_deprecated[best_scope]); IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (0); } /* * Select source address based on @inp, @dstsock and @opts. * Stores selected address to @srcp. If @scope_ambiguous is set, * embed scope from selected outgoing interface. If @hlim pointer * is provided, stores calculated hop limit there. * Returns 0 on success. */ int in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct inpcb *inp, struct ucred *cred, int scope_ambiguous, struct in6_addr *srcp, int *hlim) { struct ifnet *retifp; uint32_t fibnum; int error; fibnum = inp->inp_inc.inc_fibnum; retifp = NULL; error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp); if (error != 0) return (error); if (hlim != NULL) *hlim = in6_selecthlim(inp, retifp); if (retifp == NULL || scope_ambiguous == 0) return (0); /* * Application should provide a proper zone ID or the use of * default zone IDs should be enabled. Unfortunately, some * applications do not behave as it should, so we need a * workaround. Even if an appropriate ID is not determined * (when it's required), if we can determine the outgoing * interface. determine the zone ID based on the interface. */ error = in6_setscope(&dstsock->sin6_addr, retifp, NULL); return (error); } /* * Select source address based on @fibnum, @dst and @scopeid. * Stores selected address to @srcp. * Returns 0 on success. * * Used by non-socket based consumers (ND code mostly) */ int in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst, uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp, int *hlim) { struct ifnet *retifp; struct sockaddr_in6 dst_sa; int error; retifp = ifp; bzero(&dst_sa, sizeof(dst_sa)); dst_sa.sin6_family = AF_INET6; dst_sa.sin6_len = sizeof(dst_sa); dst_sa.sin6_addr = *dst; dst_sa.sin6_scope_id = scopeid; sa6_embedscope(&dst_sa, 0); error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL, NULL, &retifp, srcp); if (hlim != NULL) *hlim = in6_selecthlim(NULL, retifp); return (error); } /* * clone - meaningful only for bsdi and freebsd */ static int selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum) { int error = 0; struct ifnet *ifp = NULL; struct rtentry *rt = NULL; struct sockaddr_in6 *sin6_next; struct in6_pktinfo *pi = NULL; struct in6_addr *dst = &dstsock->sin6_addr; uint32_t zoneid; #if 0 char ip6buf[INET6_ADDRSTRLEN]; if (dstsock->sin6_addr.s6_addr32[0] == 0 && dstsock->sin6_addr.s6_addr32[1] == 0 && !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) { printf("in6_selectroute: strange destination %s\n", ip6_sprintf(ip6buf, &dstsock->sin6_addr)); } else { printf("in6_selectroute: destination = %s%%%d\n", ip6_sprintf(ip6buf, &dstsock->sin6_addr), dstsock->sin6_scope_id); /* for debug */ } #endif /* If the caller specify the outgoing interface explicitly, use it. */ if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) { /* XXX boundary check is assumed to be already done. */ ifp = ifnet_byindex(pi->ipi6_ifindex); if (ifp != NULL && (norouteok || retrt == NULL || IN6_IS_ADDR_MULTICAST(dst))) { /* * we do not have to check or get the route for * multicast. */ goto done; } else goto getroute; } /* * If the destination address is a multicast address and the outgoing * interface for the address is specified by the caller, use it. */ if (IN6_IS_ADDR_MULTICAST(dst) && mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) { goto done; /* we do not need a route for multicast. */ } /* * If destination address is LLA or link- or node-local multicast, * use it's embedded scope zone id to determine outgoing interface. */ if (IN6_IS_ADDR_MC_LINKLOCAL(dst) || IN6_IS_ADDR_MC_NODELOCAL(dst)) { zoneid = ntohs(in6_getscope(dst)); if (zoneid > 0) { ifp = in6_getlinkifnet(zoneid); goto done; } } getroute: /* * If the next hop address for the packet is specified by the caller, * use it as the gateway. */ if (opts && opts->ip6po_nexthop) { struct route_in6 *ron; sin6_next = satosin6(opts->ip6po_nexthop); if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) { /* * Next hop is LLA, thus it should be neighbor. * Determine outgoing interface by zone index. */ zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr)); if (zoneid > 0) { ifp = in6_getlinkifnet(zoneid); goto done; } } ron = &opts->ip6po_nextroute; /* Use a cached route if it exists and is valid. */ if (ron->ro_rt != NULL && ( (ron->ro_rt->rt_flags & RTF_UP) == 0 || ron->ro_dst.sin6_family != AF_INET6 || !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr, &sin6_next->sin6_addr))) RO_RTFREE(ron); if (ron->ro_rt == NULL) { ron->ro_dst = *sin6_next; in6_rtalloc(ron, fibnum); /* multi path case? */ } /* * The node identified by that address must be a * neighbor of the sending host. */ if (ron->ro_rt == NULL || (ron->ro_rt->rt_flags & RTF_GATEWAY) != 0) error = EHOSTUNREACH; goto done; } /* * Use a cached route if it exists and is valid, else try to allocate * a new one. Note that we should check the address family of the * cached destination, in case of sharing the cache with IPv4. */ if (ro) { if (ro->ro_rt && (!(ro->ro_rt->rt_flags & RTF_UP) || ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 || !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst))) { RTFREE(ro->ro_rt); ro->ro_rt = (struct rtentry *)NULL; } if (ro->ro_rt == (struct rtentry *)NULL) { struct sockaddr_in6 *sa6; /* No route yet, so try to acquire one */ bzero(&ro->ro_dst, sizeof(struct sockaddr_in6)); sa6 = (struct sockaddr_in6 *)&ro->ro_dst; *sa6 = *dstsock; sa6->sin6_scope_id = 0; #ifdef RADIX_MPATH rtalloc_mpath_fib((struct route *)ro, ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum); #else ro->ro_rt = in6_rtalloc1((struct sockaddr *) &ro->ro_dst, 0, 0UL, fibnum); if (ro->ro_rt) RT_UNLOCK(ro->ro_rt); #endif } /* * do not care about the result if we have the nexthop * explicitly specified. */ if (opts && opts->ip6po_nexthop) goto done; if (ro->ro_rt) { ifp = ro->ro_rt->rt_ifp; if (ifp == NULL) { /* can this really happen? */ RTFREE(ro->ro_rt); ro->ro_rt = NULL; } } if (ro->ro_rt == NULL) error = EHOSTUNREACH; rt = ro->ro_rt; /* * Check if the outgoing interface conflicts with * the interface specified by ipi6_ifindex (if specified). * Note that loopback interface is always okay. * (this may happen when we are sending a packet to one of * our own addresses.) */ if (ifp && opts && opts->ip6po_pktinfo && opts->ip6po_pktinfo->ipi6_ifindex) { if (!(ifp->if_flags & IFF_LOOPBACK) && ifp->if_index != opts->ip6po_pktinfo->ipi6_ifindex) { error = EHOSTUNREACH; goto done; } } } done: if (ifp == NULL && rt == NULL) { /* * This can happen if the caller did not pass a cached route * nor any other hints. We treat this case an error. */ error = EHOSTUNREACH; } if (error == EHOSTUNREACH) IP6STAT_INC(ip6s_noroute); if (retifp != NULL) { *retifp = ifp; /* * Adjust the "outgoing" interface. If we're going to loop * the packet back to ourselves, the ifp would be the loopback * interface. However, we'd rather know the interface associated * to the destination address (which should probably be one of * our own addresses.) */ if (rt) { if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) && (rt->rt_gateway->sa_family == AF_LINK)) *retifp = ifnet_byindex(((struct sockaddr_dl *) rt->rt_gateway)->sdl_index); } } if (retrt != NULL) *retrt = rt; /* rt may be NULL */ return (error); } static int in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct ifnet **retifp, struct ifnet *oifp, u_int fibnum) { int error; struct route_in6 sro; struct rtentry *rt = NULL; int rt_flags; KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__)); bzero(&sro, sizeof(sro)); rt_flags = 0; error = selectroute(dstsock, opts, mopts, &sro, retifp, &rt, 1, fibnum); if (rt) rt_flags = rt->rt_flags; if (rt && rt == sro.ro_rt) RTFREE(rt); if (error != 0) { /* Help ND. See oifp comment in in6_selectsrc(). */ if (oifp != NULL && fibnum == RT_DEFAULT_FIB) { *retifp = oifp; error = 0; } return (error); } /* * do not use a rejected or black hole route. * XXX: this check should be done in the L2 output routine. * However, if we skipped this check here, we'd see the following * scenario: * - install a rejected route for a scoped address prefix * (like fe80::/10) * - send a packet to a destination that matches the scoped prefix, * with ambiguity about the scope zone. * - pick the outgoing interface from the route, and disambiguate the * scope zone with the interface. * - ip6_output() would try to get another route with the "new" * destination, which may be valid. * - we'd see no error on output. * Although this may not be very harmful, it should still be confusing. * We thus reject the case here. */ if (rt_flags & (RTF_REJECT | RTF_BLACKHOLE)) { error = (rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); return (error); } return (0); } /* * Public wrapper function to selectroute(). * * XXX-BZ in6_selectroute() should and will grow the FIB argument. The * in6_selectroute_fib() function is only there for backward compat on stable. */ int in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, struct rtentry **retrt) { return (selectroute(dstsock, opts, mopts, ro, retifp, retrt, 0, RT_DEFAULT_FIB)); } #ifndef BURN_BRIDGES int in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp, struct rtentry **retrt, u_int fibnum) { return (selectroute(dstsock, opts, mopts, ro, retifp, retrt, 0, fibnum)); } #endif /* * Default hop limit selection. The precedence is as follows: * 1. Hoplimit value specified via ioctl. * 2. (If the outgoing interface is detected) the current * hop limit of the interface specified by router advertisement. * 3. The system default hoplimit. */ int in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp) { if (in6p && in6p->in6p_hops >= 0) return (in6p->in6p_hops); else if (ifp) return (ND_IFINFO(ifp)->chlim); else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { struct nhop6_basic nh6; struct in6_addr dst; uint32_t fibnum, scopeid; int hlim; fibnum = in6p->inp_inc.inc_fibnum; in6_splitscope(&in6p->in6p_faddr, &dst, &scopeid); if (fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6)==0){ hlim = ND_IFINFO(nh6.nh_ifp)->chlim; return (hlim); } } return (V_ip6_defhlim); } /* * XXX: this is borrowed from in6_pcbbind(). If possible, we should * share this function by all *bsd*... */ int in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred) { struct socket *so = inp->inp_socket; u_int16_t lport = 0; int error, lookupflags = 0; #ifdef INVARIANTS struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; #endif INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); error = prison_local_ip6(cred, laddr, ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)); if (error) return(error); /* XXX: this is redundant when called from in6_pcbbind */ - if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) + if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0) lookupflags = INPLOOKUP_WILDCARD; inp->inp_flags |= INP_ANONPORT; error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags); if (error != 0) return (error); inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->in6p_laddr = in6addr_any; inp->inp_lport = 0; return (EAGAIN); } return (0); } void addrsel_policy_init(void) { init_policy_queue(); /* initialize the "last resort" policy */ bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy)); V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP; if (!IS_DEFAULT_VNET(curvnet)) return; ADDRSEL_LOCK_INIT(); ADDRSEL_SXLOCK_INIT(); } static struct in6_addrpolicy * lookup_addrsel_policy(struct sockaddr_in6 *key) { struct in6_addrpolicy *match = NULL; ADDRSEL_LOCK(); match = match_addrsel_policy(key); if (match == NULL) match = &V_defaultaddrpolicy; else match->use++; ADDRSEL_UNLOCK(); return (match); } /* * Subroutines to manage the address selection policy table via sysctl. */ struct walkarg { struct sysctl_req *w_req; }; static int in6_src_sysctl(SYSCTL_HANDLER_ARGS); SYSCTL_DECL(_net_inet6_ip6); static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy, CTLFLAG_RD, in6_src_sysctl, ""); static int in6_src_sysctl(SYSCTL_HANDLER_ARGS) { struct walkarg w; if (req->newptr) return EPERM; bzero(&w, sizeof(w)); w.w_req = req; return (walk_addrsel_policy(dump_addrsel_policyent, &w)); } int in6_src_ioctl(u_long cmd, caddr_t data) { struct in6_addrpolicy ent0; if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY) return (EOPNOTSUPP); /* check for safety */ ent0 = *(struct in6_addrpolicy *)data; if (ent0.label == ADDR_LABEL_NOTAPP) return (EINVAL); /* check if the prefix mask is consecutive. */ if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0) return (EINVAL); /* clear trailing garbages (if any) of the prefix address. */ IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr); ent0.use = 0; switch (cmd) { case SIOCAADDRCTL_POLICY: return (add_addrsel_policyent(&ent0)); case SIOCDADDRCTL_POLICY: return (delete_addrsel_policyent(&ent0)); } return (0); /* XXX: compromise compilers */ } /* * The followings are implementation of the policy table using a * simple tail queue. * XXX such details should be hidden. * XXX implementation using binary tree should be more efficient. */ struct addrsel_policyent { TAILQ_ENTRY(addrsel_policyent) ape_entry; struct in6_addrpolicy ape_policy; }; TAILQ_HEAD(addrsel_policyhead, addrsel_policyent); static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab); #define V_addrsel_policytab VNET(addrsel_policytab) static void init_policy_queue(void) { TAILQ_INIT(&V_addrsel_policytab); } static int add_addrsel_policyent(struct in6_addrpolicy *newpolicy) { struct addrsel_policyent *new, *pol; new = malloc(sizeof(*new), M_IFADDR, M_WAITOK); ADDRSEL_XLOCK(); ADDRSEL_LOCK(); /* duplication check */ TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr, &pol->ape_policy.addr.sin6_addr) && IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr, &pol->ape_policy.addrmask.sin6_addr)) { ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); free(new, M_IFADDR); return (EEXIST); /* or override it? */ } } bzero(new, sizeof(*new)); /* XXX: should validate entry */ new->ape_policy = *newpolicy; TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry); ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); return (0); } static int delete_addrsel_policyent(struct in6_addrpolicy *key) { struct addrsel_policyent *pol; ADDRSEL_XLOCK(); ADDRSEL_LOCK(); /* search for the entry in the table */ TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr, &pol->ape_policy.addr.sin6_addr) && IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr, &pol->ape_policy.addrmask.sin6_addr)) { break; } } if (pol == NULL) { ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); return (ESRCH); } TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry); ADDRSEL_UNLOCK(); ADDRSEL_XUNLOCK(); free(pol, M_IFADDR); return (0); } static int walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w) { struct addrsel_policyent *pol; int error = 0; ADDRSEL_SLOCK(); TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) { if ((error = (*callback)(&pol->ape_policy, w)) != 0) { ADDRSEL_SUNLOCK(); return (error); } } ADDRSEL_SUNLOCK(); return (error); } static int dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg) { int error = 0; struct walkarg *w = arg; error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol)); return (error); } static struct in6_addrpolicy * match_addrsel_policy(struct sockaddr_in6 *key) { struct addrsel_policyent *pent; struct in6_addrpolicy *bestpol = NULL, *pol; int matchlen, bestmatchlen = -1; u_char *mp, *ep, *k, *p, m; TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) { matchlen = 0; pol = &pent->ape_policy; mp = (u_char *)&pol->addrmask.sin6_addr; ep = mp + 16; /* XXX: scope field? */ k = (u_char *)&key->sin6_addr; p = (u_char *)&pol->addr.sin6_addr; for (; mp < ep && *mp; mp++, k++, p++) { m = *mp; if ((*k & m) != *p) goto next; /* not match */ if (m == 0xff) /* short cut for a typical case */ matchlen += 8; else { while (m >= 0x80) { matchlen++; m <<= 1; } } } /* matched. check if this is better than the current best. */ if (bestpol == NULL || matchlen > bestmatchlen) { bestpol = pol; bestmatchlen = matchlen; } next: continue; } return (bestpol); } Index: head/sys/netinet6/ip6_output.c =================================================================== --- head/sys/netinet6/ip6_output.c (revision 334718) +++ head/sys/netinet6/ip6_output.c (revision 334719) @@ -1,3098 +1,3107 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * 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. Neither the name of the project 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 PROJECT 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 PROJECT 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. * * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ */ /*- * Copyright (c) 1982, 1986, 1988, 1990, 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. * 3. 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. * * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ratelimit.h" #include "opt_ipsec.h" #include "opt_sctp.h" #include "opt_route.h" #include "opt_rss.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 #ifdef SCTP #include #include #endif #include #include extern int in6_mcast_loop; struct ip6_exthdrs { struct mbuf *ip6e_ip6; struct mbuf *ip6e_hbh; struct mbuf *ip6e_dest1; struct mbuf *ip6e_rthdr; struct mbuf *ip6e_dest2; }; static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options"); static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, struct ucred *, int); static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *, struct sockopt *); static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *); static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, struct ucred *, int, int, int); static int ip6_copyexthdr(struct mbuf **, caddr_t, int); static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, struct ip6_frag **); static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); static int ip6_getpmtu(struct route_in6 *, int, struct ifnet *, const struct in6_addr *, u_long *, int *, u_int, u_int); static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long, u_long *, int *, u_int); static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *); static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); /* * Make an extension header from option data. hp is the source, and * mp is the destination. */ #define MAKE_EXTHDR(hp, mp) \ do { \ if (hp) { \ struct ip6_ext *eh = (struct ip6_ext *)(hp); \ error = ip6_copyexthdr((mp), (caddr_t)(hp), \ ((eh)->ip6e_len + 1) << 3); \ if (error) \ goto freehdrs; \ } \ } while (/*CONSTCOND*/ 0) /* * Form a chain of extension headers. * m is the extension header mbuf * mp is the previous mbuf in the chain * p is the next header * i is the type of option. */ #define MAKE_CHAIN(m, mp, p, i)\ do {\ if (m) {\ if (!hdrsplit) \ panic("assumption failed: hdr not split"); \ *mtod((m), u_char *) = *(p);\ *(p) = (i);\ p = mtod((m), u_char *);\ (m)->m_next = (mp)->m_next;\ (mp)->m_next = (m);\ (mp) = (m);\ }\ } while (/*CONSTCOND*/ 0) void in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset) { u_short csum; csum = in_cksum_skip(m, offset + plen, offset); if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0) csum = 0xffff; offset += m->m_pkthdr.csum_data; /* checksum offset */ if (offset + sizeof(csum) > m->m_len) m_copyback(m, offset, sizeof(csum), (caddr_t)&csum); else *(u_short *)mtodo(m, offset) = csum; } int ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto, int fraglen , uint32_t id) { struct mbuf *m, **mnext, *m_frgpart; struct ip6_hdr *ip6, *mhip6; struct ip6_frag *ip6f; int off; int error; int tlen = m0->m_pkthdr.len; KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8")); m = m0; ip6 = mtod(m, struct ip6_hdr *); mnext = &m->m_nextpkt; for (off = hlen; off < tlen; off += fraglen) { m = m_gethdr(M_NOWAIT, MT_DATA); if (!m) { IP6STAT_INC(ip6s_odropped); return (ENOBUFS); } m->m_flags = m0->m_flags & M_COPYFLAGS; *mnext = m; mnext = &m->m_nextpkt; m->m_data += max_linkhdr; mhip6 = mtod(m, struct ip6_hdr *); *mhip6 = *ip6; m->m_len = sizeof(*mhip6); error = ip6_insertfraghdr(m0, m, hlen, &ip6f); if (error) { IP6STAT_INC(ip6s_odropped); return (error); } ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); if (off + fraglen >= tlen) fraglen = tlen - off; else ip6f->ip6f_offlg |= IP6F_MORE_FRAG; mhip6->ip6_plen = htons((u_short)(fraglen + hlen + sizeof(*ip6f) - sizeof(struct ip6_hdr))); if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) { IP6STAT_INC(ip6s_odropped); return (ENOBUFS); } m_cat(m, m_frgpart); m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f); m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum; m->m_pkthdr.rcvif = NULL; ip6f->ip6f_reserved = 0; ip6f->ip6f_ident = id; ip6f->ip6f_nxt = nextproto; IP6STAT_INC(ip6s_ofragments); in6_ifstat_inc(ifp, ifs6_out_fragcreat); } return (0); } /* * IP6 output. The packet in mbuf chain m contains a skeletal IP6 * header (with pri, len, nxt, hlim, src, dst). * This function may modify ver and hlim only. * The mbuf chain containing the packet will be freed. * The mbuf opt, if present, will not be freed. * If route_in6 ro is present and has ro_rt initialized, route lookup would be * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, * then result of route lookup is stored in ro->ro_rt. * * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, * which is rt_mtu. * * ifpp - XXX: just for statistics */ /* * XXX TODO: no flowid is assigned for outbound flows? */ int ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro, int flags, struct ip6_moptions *im6o, struct ifnet **ifpp, struct inpcb *inp) { struct ip6_hdr *ip6; struct ifnet *ifp, *origifp; struct mbuf *m = m0; struct mbuf *mprev = NULL; int hlen, tlen, len; struct route_in6 ip6route; struct rtentry *rt = NULL; struct sockaddr_in6 *dst, src_sa, dst_sa; struct in6_addr odst; int error = 0; struct in6_ifaddr *ia = NULL; u_long mtu; int alwaysfrag, dontfrag; u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; struct ip6_exthdrs exthdrs; struct in6_addr src0, dst0; u_int32_t zone; struct route_in6 *ro_pmtu = NULL; int hdrsplit = 0; int sw_csum, tso; int needfiblookup; uint32_t fibnum; struct m_tag *fwd_tag = NULL; uint32_t id; if (inp != NULL) { INP_LOCK_ASSERT(inp); M_SETFIB(m, inp->inp_inc.inc_fibnum); if ((flags & IP_NODEFAULTFLOWID) == 0) { /* unconditionally set flowid */ m->m_pkthdr.flowid = inp->inp_flowid; M_HASHTYPE_SET(m, inp->inp_flowtype); } } #if defined(IPSEC) || defined(IPSEC_SUPPORT) /* * IPSec checking which handles several cases. * FAST IPSEC: We re-injected the packet. * XXX: need scope argument. */ if (IPSEC_ENABLED(ipv6)) { if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) { if (error == EINPROGRESS) error = 0; goto done; } } #endif /* IPSEC */ bzero(&exthdrs, sizeof(exthdrs)); if (opt) { /* Hop-by-Hop options header */ MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); /* Destination options header(1st part) */ if (opt->ip6po_rthdr) { /* * Destination options header(1st part) * This only makes sense with a routing header. * See Section 9.2 of RFC 3542. * Disabling this part just for MIP6 convenience is * a bad idea. We need to think carefully about a * way to make the advanced API coexist with MIP6 * options, which might automatically be inserted in * the kernel. */ MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); } /* Routing header */ MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); /* Destination options header(2nd part) */ MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); } /* * Calculate the total length of the extension header chain. * Keep the length of the unfragmentable part for fragmentation. */ optlen = 0; if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; unfragpartlen = optlen + sizeof(struct ip6_hdr); /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */ if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; /* * If there is at least one extension header, * separate IP6 header from the payload. */ if (optlen && !hdrsplit) { if ((error = ip6_splithdr(m, &exthdrs)) != 0) { m = NULL; goto freehdrs; } m = exthdrs.ip6e_ip6; hdrsplit++; } ip6 = mtod(m, struct ip6_hdr *); /* adjust mbuf packet header length */ m->m_pkthdr.len += optlen; plen = m->m_pkthdr.len - sizeof(*ip6); /* If this is a jumbo payload, insert a jumbo payload option. */ if (plen > IPV6_MAXPACKET) { if (!hdrsplit) { if ((error = ip6_splithdr(m, &exthdrs)) != 0) { m = NULL; goto freehdrs; } m = exthdrs.ip6e_ip6; hdrsplit++; } /* adjust pointer */ ip6 = mtod(m, struct ip6_hdr *); if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) goto freehdrs; ip6->ip6_plen = 0; } else ip6->ip6_plen = htons(plen); /* * Concatenate headers and fill in next header fields. * Here we have, on "m" * IPv6 payload * and we insert headers accordingly. Finally, we should be getting: * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] * * during the header composing process, "m" points to IPv6 header. * "mprev" points to an extension header prior to esp. */ u_char *nexthdrp = &ip6->ip6_nxt; mprev = m; /* * we treat dest2 specially. this makes IPsec processing * much easier. the goal here is to make mprev point the * mbuf prior to dest2. * * result: IPv6 dest2 payload * m and mprev will point to IPv6 header. */ if (exthdrs.ip6e_dest2) { if (!hdrsplit) panic("assumption failed: hdr not split"); exthdrs.ip6e_dest2->m_next = m->m_next; m->m_next = exthdrs.ip6e_dest2; *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; ip6->ip6_nxt = IPPROTO_DSTOPTS; } /* * result: IPv6 hbh dest1 rthdr dest2 payload * m will point to IPv6 header. mprev will point to the * extension header prior to dest2 (rthdr in the above case). */ MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS); MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING); /* * If there is a routing header, discard the packet. */ if (exthdrs.ip6e_rthdr) { error = EINVAL; goto bad; } /* Source address validation */ if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && (flags & IPV6_UNSPECSRC) == 0) { error = EOPNOTSUPP; IP6STAT_INC(ip6s_badscope); goto bad; } if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { error = EOPNOTSUPP; IP6STAT_INC(ip6s_badscope); goto bad; } IP6STAT_INC(ip6s_localout); /* * Route packet. */ if (ro == NULL) { ro = &ip6route; bzero((caddr_t)ro, sizeof(*ro)); } ro_pmtu = ro; if (opt && opt->ip6po_rthdr) ro = &opt->ip6po_route; dst = (struct sockaddr_in6 *)&ro->ro_dst; fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m); again: /* * if specified, try to fill in the traffic class field. * do not override if a non-zero value is already set. * we check the diffserv field and the ecn field separately. */ if (opt && opt->ip6po_tclass >= 0) { int mask = 0; if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) mask |= 0xfc; if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) mask |= 0x03; if (mask != 0) ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); } /* fill in or override the hop limit field, if necessary. */ if (opt && opt->ip6po_hlim != -1) ip6->ip6_hlim = opt->ip6po_hlim & 0xff; else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { if (im6o != NULL) ip6->ip6_hlim = im6o->im6o_multicast_hlim; else ip6->ip6_hlim = V_ip6_defmcasthlim; } /* * Validate route against routing table additions; * a better/more specific route might have been added. * Make sure address family is set in route. */ if (inp) { ro->ro_dst.sin6_family = AF_INET6; RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum); } if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) && ro->ro_dst.sin6_family == AF_INET6 && IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) { rt = ro->ro_rt; ifp = ro->ro_rt->rt_ifp; } else { if (ro->ro_lle) LLE_FREE(ro->ro_lle); /* zeros ro_lle */ ro->ro_lle = NULL; if (fwd_tag == NULL) { bzero(&dst_sa, sizeof(dst_sa)); dst_sa.sin6_family = AF_INET6; dst_sa.sin6_len = sizeof(dst_sa); dst_sa.sin6_addr = ip6->ip6_dst; } error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp, &rt, fibnum); if (error != 0) { if (ifp != NULL) in6_ifstat_inc(ifp, ifs6_out_discard); goto bad; } } if (rt == NULL) { /* * If in6_selectroute() does not return a route entry, * dst may not have been updated. */ *dst = dst_sa; /* XXX */ } /* * then rt (for unicast) and ifp must be non-NULL valid values. */ if ((flags & IPV6_FORWARDING) == 0) { /* XXX: the FORWARDING flag can be set for mrouting. */ in6_ifstat_inc(ifp, ifs6_out_request); } if (rt != NULL) { ia = (struct in6_ifaddr *)(rt->rt_ifa); counter_u64_add(rt->rt_pksent, 1); } /* * The outgoing interface must be in the zone of source and * destination addresses. */ origifp = ifp; src0 = ip6->ip6_src; if (in6_setscope(&src0, origifp, &zone)) goto badscope; bzero(&src_sa, sizeof(src_sa)); src_sa.sin6_family = AF_INET6; src_sa.sin6_len = sizeof(src_sa); src_sa.sin6_addr = ip6->ip6_src; if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) goto badscope; dst0 = ip6->ip6_dst; if (in6_setscope(&dst0, origifp, &zone)) goto badscope; /* re-initialize to be sure */ bzero(&dst_sa, sizeof(dst_sa)); dst_sa.sin6_family = AF_INET6; dst_sa.sin6_len = sizeof(dst_sa); dst_sa.sin6_addr = ip6->ip6_dst; if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) { goto badscope; } /* We should use ia_ifp to support the case of * sending packets to an address of our own. */ if (ia != NULL && ia->ia_ifp) ifp = ia->ia_ifp; /* scope check is done. */ goto routefound; badscope: IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(origifp, ifs6_out_discard); if (error == 0) error = EHOSTUNREACH; /* XXX */ goto bad; routefound: if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { if (opt && opt->ip6po_nextroute.ro_rt) { /* * The nexthop is explicitly specified by the * application. We assume the next hop is an IPv6 * address. */ dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; } else if ((rt->rt_flags & RTF_GATEWAY)) dst = (struct sockaddr_in6 *)rt->rt_gateway; } if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ } else { m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; in6_ifstat_inc(ifp, ifs6_out_mcast); /* * Confirm that the outgoing interface supports multicast. */ if (!(ifp->if_flags & IFF_MULTICAST)) { IP6STAT_INC(ip6s_noroute); in6_ifstat_inc(ifp, ifs6_out_discard); error = ENETUNREACH; goto bad; } if ((im6o == NULL && in6_mcast_loop) || (im6o && im6o->im6o_multicast_loop)) { /* * Loop back multicast datagram if not expressly * forbidden to do so, even if we have not joined * the address; protocols will filter it later, * thus deferring a hash lookup and lock acquisition * at the expense of an m_copym(). */ ip6_mloopback(ifp, m); } else { /* * If we are acting as a multicast router, perform * multicast forwarding as if the packet had just * arrived on the interface to which we are about * to send. The multicast forwarding function * recursively calls this function, using the * IPV6_FORWARDING flag to prevent infinite recursion. * * Multicasts that are looped back by ip6_mloopback(), * above, will be forwarded by the ip6_input() routine, * if necessary. */ if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { /* * XXX: ip6_mforward expects that rcvif is NULL * when it is called from the originating path. * However, it may not always be the case. */ m->m_pkthdr.rcvif = NULL; if (ip6_mforward(ip6, ifp, m) != 0) { m_freem(m); goto done; } } } /* * Multicasts with a hoplimit of zero may be looped back, * above, but must not be transmitted on a network. * Also, multicasts addressed to the loopback interface * are not sent -- the above call to ip6_mloopback() will * loop back a copy if this host actually belongs to the * destination group on the loopback interface. */ if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { m_freem(m); goto done; } } /* * Fill the outgoing inteface to tell the upper layer * to increment per-interface statistics. */ if (ifpp) *ifpp = ifp; /* Determine path MTU. */ if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0) goto bad; /* * The caller of this function may specify to use the minimum MTU * in some cases. * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU * setting. The logic is a bit complicated; by default, unicast * packets will follow path MTU while multicast packets will be sent at * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets * including unicast ones will be sent at the minimum MTU. Multicast * packets will always be sent at the minimum MTU unless * IP6PO_MINMTU_DISABLE is explicitly specified. * See RFC 3542 for more details. */ if (mtu > IPV6_MMTU) { if ((flags & IPV6_MINMTU)) mtu = IPV6_MMTU; else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) mtu = IPV6_MMTU; else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && (opt == NULL || opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { mtu = IPV6_MMTU; } } /* * clear embedded scope identifiers if necessary. * in6_clearscope will touch the addresses only when necessary. */ in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); /* * If the outgoing packet contains a hop-by-hop options header, * it must be examined and processed even by the source node. * (RFC 2460, section 4.) */ if (exthdrs.ip6e_hbh) { struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); u_int32_t dummy; /* XXX unused */ u_int32_t plen = 0; /* XXX: ip6_process will check the value */ #ifdef DIAGNOSTIC if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) panic("ip6e_hbh is not contiguous"); #endif /* * XXX: if we have to send an ICMPv6 error to the sender, * we need the M_LOOP flag since icmp6_error() expects * the IPv6 and the hop-by-hop options header are * contiguous unless the flag is set. */ m->m_flags |= M_LOOP; m->m_pkthdr.rcvif = ifp; if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), &dummy, &plen) < 0) { /* m was already freed at this point */ error = EINVAL;/* better error? */ goto done; } m->m_flags &= ~M_LOOP; /* XXX */ m->m_pkthdr.rcvif = NULL; } /* Jump over all PFIL processing if hooks are not active. */ if (!PFIL_HOOKED(&V_inet6_pfil_hook)) goto passout; odst = ip6->ip6_dst; /* Run through list of hooks for output packets. */ error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, 0, inp); if (error != 0 || m == NULL) goto done; /* adjust pointer */ ip6 = mtod(m, struct ip6_hdr *); needfiblookup = 0; /* See if destination IP address was changed by packet filter. */ if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { m->m_flags |= M_SKIP_FIREWALL; /* If destination is now ourself drop to ip6_input(). */ if (in6_localip(&ip6->ip6_dst)) { m->m_flags |= M_FASTFWD_OURS; if (m->m_pkthdr.rcvif == NULL) m->m_pkthdr.rcvif = V_loif; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { m->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; m->m_pkthdr.csum_data = 0xffff; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; #endif error = netisr_queue(NETISR_IPV6, m); goto done; } else { RO_RTFREE(ro); needfiblookup = 1; /* Redo the routing table lookup. */ if (ro->ro_lle) LLE_FREE(ro->ro_lle); /* zeros ro_lle */ ro->ro_lle = NULL; } } /* See if fib was changed by packet filter. */ if (fibnum != M_GETFIB(m)) { m->m_flags |= M_SKIP_FIREWALL; fibnum = M_GETFIB(m); RO_RTFREE(ro); needfiblookup = 1; if (ro->ro_lle) LLE_FREE(ro->ro_lle); /* zeros ro_lle */ ro->ro_lle = NULL; } if (needfiblookup) goto again; /* See if local, if yes, send it to netisr. */ if (m->m_flags & M_FASTFWD_OURS) { if (m->m_pkthdr.rcvif == NULL) m->m_pkthdr.rcvif = V_loif; if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { m->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; m->m_pkthdr.csum_data = 0xffff; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; #endif error = netisr_queue(NETISR_IPV6, m); goto done; } /* Or forward to some other address? */ if ((m->m_flags & M_IP6_NEXTHOP) && (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { dst = (struct sockaddr_in6 *)&ro->ro_dst; bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6)); m->m_flags |= M_SKIP_FIREWALL; m->m_flags &= ~M_IP6_NEXTHOP; m_tag_delete(m, fwd_tag); goto again; } passout: /* * Send the packet to the outgoing interface. * If necessary, do IPv6 fragmentation before sending. * * the logic here is rather complex: * 1: normal case (dontfrag == 0, alwaysfrag == 0) * 1-a: send as is if tlen <= path mtu * 1-b: fragment if tlen > path mtu * * 2: if user asks us not to fragment (dontfrag == 1) * 2-a: send as is if tlen <= interface mtu * 2-b: error if tlen > interface mtu * * 3: if we always need to attach fragment header (alwaysfrag == 1) * always fragment * * 4: if dontfrag == 1 && alwaysfrag == 1 * error, as we cannot handle this conflicting request */ sw_csum = m->m_pkthdr.csum_flags; if (!hdrsplit) { tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0; sw_csum &= ~ifp->if_hwassist; } else tso = 0; /* * If we added extension headers, we will not do TSO and calculate the * checksums ourselves for now. * XXX-BZ Need a framework to know when the NIC can handle it, even * with ext. hdrs. */ if (sw_csum & CSUM_DELAY_DATA_IPV6) { sw_csum &= ~CSUM_DELAY_DATA_IPV6; in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr)); } #ifdef SCTP if (sw_csum & CSUM_SCTP_IPV6) { sw_csum &= ~CSUM_SCTP_IPV6; sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); } #endif m->m_pkthdr.csum_flags &= ifp->if_hwassist; tlen = m->m_pkthdr.len; if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso) dontfrag = 1; else dontfrag = 0; if (dontfrag && alwaysfrag) { /* case 4 */ /* conflicting request - can't transmit */ error = EMSGSIZE; goto bad; } if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */ /* * Even if the DONTFRAG option is specified, we cannot send the * packet when the data length is larger than the MTU of the * outgoing interface. * Notify the error by sending IPV6_PATHMTU ancillary data if * application wanted to know the MTU value. Also return an * error code (this is not described in the API spec). */ if (inp != NULL) ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu); error = EMSGSIZE; goto bad; } /* * transmit packet without fragmentation */ if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ struct in6_ifaddr *ia6; ip6 = mtod(m, struct ip6_hdr *); ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); if (ia6) { /* Record statistics for this interface address. */ counter_u64_add(ia6->ia_ifa.ifa_opackets, 1); counter_u64_add(ia6->ia_ifa.ifa_obytes, m->m_pkthdr.len); ifa_free(&ia6->ia_ifa); } #ifdef RATELIMIT if (inp != NULL) { if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) in_pcboutput_txrtlmt(inp, ifp, m); /* stamp send tag on mbuf */ m->m_pkthdr.snd_tag = inp->inp_snd_tag; } else { m->m_pkthdr.snd_tag = NULL; } #endif error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro); #ifdef RATELIMIT /* check for route change */ if (error == EAGAIN) in_pcboutput_eagain(inp); #endif goto done; } /* * try to fragment the packet. case 1-b and 3 */ if (mtu < IPV6_MMTU) { /* path MTU cannot be less than IPV6_MMTU */ error = EMSGSIZE; in6_ifstat_inc(ifp, ifs6_out_fragfail); goto bad; } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */ error = EMSGSIZE; in6_ifstat_inc(ifp, ifs6_out_fragfail); goto bad; } else { u_char nextproto; /* * Too large for the destination or interface; * fragment if possible. * Must be able to put at least 8 bytes per fragment. */ hlen = unfragpartlen; if (mtu > IPV6_MAXPACKET) mtu = IPV6_MAXPACKET; len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; if (len < 8) { error = EMSGSIZE; in6_ifstat_inc(ifp, ifs6_out_fragfail); goto bad; } /* * If the interface will not calculate checksums on * fragmented packets, then do it here. * XXX-BZ handle the hw offloading case. Need flags. */ if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { in6_delayed_cksum(m, plen, hlen); m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { sctp_delayed_cksum(m, hlen); m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; } #endif /* * Change the next header field of the last header in the * unfragmentable part. */ if (exthdrs.ip6e_rthdr) { nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; } else if (exthdrs.ip6e_dest1) { nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; } else if (exthdrs.ip6e_hbh) { nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; } else { nextproto = ip6->ip6_nxt; ip6->ip6_nxt = IPPROTO_FRAGMENT; } /* * Loop through length of segment after first fragment, * make new header and copy data of each part and link onto * chain. */ m0 = m; id = htonl(ip6_randomid()); if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id))) goto sendorfree; in6_ifstat_inc(ifp, ifs6_out_fragok); } /* * Remove leading garbages. */ sendorfree: m = m0->m_nextpkt; m0->m_nextpkt = 0; m_freem(m0); for (; m; m = m0) { m0 = m->m_nextpkt; m->m_nextpkt = 0; if (error == 0) { /* Record statistics for this interface address. */ if (ia) { counter_u64_add(ia->ia_ifa.ifa_opackets, 1); counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len); } #ifdef RATELIMIT if (inp != NULL) { if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) in_pcboutput_txrtlmt(inp, ifp, m); /* stamp send tag on mbuf */ m->m_pkthdr.snd_tag = inp->inp_snd_tag; } else { m->m_pkthdr.snd_tag = NULL; } #endif error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro); #ifdef RATELIMIT /* check for route change */ if (error == EAGAIN) in_pcboutput_eagain(inp); #endif } else m_freem(m); } if (error == 0) IP6STAT_INC(ip6s_fragmented); done: if (ro == &ip6route) RO_RTFREE(ro); return (error); freehdrs: m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ m_freem(exthdrs.ip6e_dest1); m_freem(exthdrs.ip6e_rthdr); m_freem(exthdrs.ip6e_dest2); /* FALLTHROUGH */ bad: if (m) m_freem(m); goto done; } static int ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) { struct mbuf *m; if (hlen > MCLBYTES) return (ENOBUFS); /* XXX */ if (hlen > MLEN) m = m_getcl(M_NOWAIT, MT_DATA, 0); else m = m_get(M_NOWAIT, MT_DATA); if (m == NULL) return (ENOBUFS); m->m_len = hlen; if (hdr) bcopy(hdr, mtod(m, caddr_t), hlen); *mp = m; return (0); } /* * Insert jumbo payload option. */ static int ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) { struct mbuf *mopt; u_char *optbuf; u_int32_t v; #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ /* * If there is no hop-by-hop options header, allocate new one. * If there is one but it doesn't have enough space to store the * jumbo payload option, allocate a cluster to store the whole options. * Otherwise, use it to store the options. */ if (exthdrs->ip6e_hbh == NULL) { mopt = m_get(M_NOWAIT, MT_DATA); if (mopt == NULL) return (ENOBUFS); mopt->m_len = JUMBOOPTLEN; optbuf = mtod(mopt, u_char *); optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ exthdrs->ip6e_hbh = mopt; } else { struct ip6_hbh *hbh; mopt = exthdrs->ip6e_hbh; if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { /* * XXX assumption: * - exthdrs->ip6e_hbh is not referenced from places * other than exthdrs. * - exthdrs->ip6e_hbh is not an mbuf chain. */ int oldoptlen = mopt->m_len; struct mbuf *n; /* * XXX: give up if the whole (new) hbh header does * not fit even in an mbuf cluster. */ if (oldoptlen + JUMBOOPTLEN > MCLBYTES) return (ENOBUFS); /* * As a consequence, we must always prepare a cluster * at this point. */ n = m_getcl(M_NOWAIT, MT_DATA, 0); if (n == NULL) return (ENOBUFS); n->m_len = oldoptlen + JUMBOOPTLEN; bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen); optbuf = mtod(n, caddr_t) + oldoptlen; m_freem(mopt); mopt = exthdrs->ip6e_hbh = n; } else { optbuf = mtod(mopt, u_char *) + mopt->m_len; mopt->m_len += JUMBOOPTLEN; } optbuf[0] = IP6OPT_PADN; optbuf[1] = 1; /* * Adjust the header length according to the pad and * the jumbo payload option. */ hbh = mtod(mopt, struct ip6_hbh *); hbh->ip6h_len += (JUMBOOPTLEN >> 3); } /* fill in the option. */ optbuf[2] = IP6OPT_JUMBO; optbuf[3] = 4; v = (u_int32_t)htonl(plen + JUMBOOPTLEN); bcopy(&v, &optbuf[4], sizeof(u_int32_t)); /* finally, adjust the packet header length */ exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; return (0); #undef JUMBOOPTLEN } /* * Insert fragment header and copy unfragmentable header portions. */ static int ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, struct ip6_frag **frghdrp) { struct mbuf *n, *mlast; if (hlen > sizeof(struct ip6_hdr)) { n = m_copym(m0, sizeof(struct ip6_hdr), hlen - sizeof(struct ip6_hdr), M_NOWAIT); if (n == NULL) return (ENOBUFS); m->m_next = n; } else n = m; /* Search for the last mbuf of unfragmentable part. */ for (mlast = n; mlast->m_next; mlast = mlast->m_next) ; if (M_WRITABLE(mlast) && M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { /* use the trailing space of the last mbuf for the fragment hdr */ *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len); mlast->m_len += sizeof(struct ip6_frag); m->m_pkthdr.len += sizeof(struct ip6_frag); } else { /* allocate a new mbuf for the fragment header */ struct mbuf *mfrg; mfrg = m_get(M_NOWAIT, MT_DATA); if (mfrg == NULL) return (ENOBUFS); mfrg->m_len = sizeof(struct ip6_frag); *frghdrp = mtod(mfrg, struct ip6_frag *); mlast->m_next = mfrg; } return (0); } /* * Calculates IPv6 path mtu for destination @dst. * Resulting MTU is stored in @mtup. * * Returns 0 on success. */ static int ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup) { struct nhop6_extended nh6; struct in6_addr kdst; uint32_t scopeid; struct ifnet *ifp; u_long mtu; int error; in6_splitscope(dst, &kdst, &scopeid); if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0) return (EHOSTUNREACH); ifp = nh6.nh_ifp; mtu = nh6.nh_mtu; error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0); fib6_free_nh_ext(fibnum, &nh6); return (error); } /* * Calculates IPv6 path MTU for @dst based on transmit @ifp, * and cached data in @ro_pmtu. * MTU from (successful) route lookup is saved (along with dst) * inside @ro_pmtu to avoid subsequent route lookups after packet * filter processing. * * Stores mtu and always-frag value into @mtup and @alwaysfragp. * Returns 0 on success. */ static int ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup, struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup, int *alwaysfragp, u_int fibnum, u_int proto) { struct nhop6_basic nh6; struct in6_addr kdst; uint32_t scopeid; struct sockaddr_in6 *sa6_dst; u_long mtu; mtu = 0; if (do_lookup) { /* * Here ro_pmtu has final destination address, while * ro might represent immediate destination. * Use ro_pmtu destination since mtu might differ. */ sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst; if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) ro_pmtu->ro_mtu = 0; if (ro_pmtu->ro_mtu == 0) { bzero(sa6_dst, sizeof(*sa6_dst)); sa6_dst->sin6_family = AF_INET6; sa6_dst->sin6_len = sizeof(struct sockaddr_in6); sa6_dst->sin6_addr = *dst; in6_splitscope(dst, &kdst, &scopeid); if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0, &nh6) == 0) ro_pmtu->ro_mtu = nh6.nh_mtu; } mtu = ro_pmtu->ro_mtu; } if (ro_pmtu->ro_rt) mtu = ro_pmtu->ro_rt->rt_mtu; return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto)); } /* * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and * hostcache data for @dst. * Stores mtu and always-frag value into @mtup and @alwaysfragp. * * Returns 0 on success. */ static int ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu, u_long *mtup, int *alwaysfragp, u_int proto) { u_long mtu = 0; int alwaysfrag = 0; int error = 0; if (rt_mtu > 0) { u_int32_t ifmtu; struct in_conninfo inc; bzero(&inc, sizeof(inc)); inc.inc_flags |= INC_ISIPV6; inc.inc6_faddr = *dst; ifmtu = IN6_LINKMTU(ifp); /* TCP is known to react to pmtu changes so skip hc */ if (proto != IPPROTO_TCP) mtu = tcp_hc_getmtu(&inc); if (mtu) mtu = min(mtu, rt_mtu); else mtu = rt_mtu; if (mtu == 0) mtu = ifmtu; else if (mtu < IPV6_MMTU) { /* * RFC2460 section 5, last paragraph: * if we record ICMPv6 too big message with * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU * or smaller, with framgent header attached. * (fragment header is needed regardless from the * packet size, for translators to identify packets) */ alwaysfrag = 1; mtu = IPV6_MMTU; } } else if (ifp) { mtu = IN6_LINKMTU(ifp); } else error = EHOSTUNREACH; /* XXX */ *mtup = mtu; if (alwaysfragp) *alwaysfragp = alwaysfrag; return (error); } /* * IP6 socket option processing. */ int ip6_ctloutput(struct socket *so, struct sockopt *sopt) { int optdatalen, uproto; void *optdata; struct inpcb *in6p = sotoinpcb(so); int error, optval; int level, op, optname; int optlen; struct thread *td; #ifdef RSS uint32_t rss_bucket; int retval; #endif /* * Don't use more than a quarter of mbuf clusters. N.B.: * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow * on LP64 architectures, so cast to u_long to avoid undefined * behavior. ILP32 architectures cannot have nmbclusters * large enough to overflow for other reasons. */ #define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4) level = sopt->sopt_level; op = sopt->sopt_dir; optname = sopt->sopt_name; optlen = sopt->sopt_valsize; td = sopt->sopt_td; error = 0; optval = 0; uproto = (int)so->so_proto->pr_protocol; if (level != IPPROTO_IPV6) { error = EINVAL; if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_dir == SOPT_SET) { switch (sopt->sopt_name) { case SO_REUSEADDR: INP_WLOCK(in6p); if ((so->so_options & SO_REUSEADDR) != 0) in6p->inp_flags2 |= INP_REUSEADDR; else in6p->inp_flags2 &= ~INP_REUSEADDR; INP_WUNLOCK(in6p); error = 0; break; case SO_REUSEPORT: INP_WLOCK(in6p); if ((so->so_options & SO_REUSEPORT) != 0) in6p->inp_flags2 |= INP_REUSEPORT; else in6p->inp_flags2 &= ~INP_REUSEPORT; INP_WUNLOCK(in6p); error = 0; break; + case SO_REUSEPORT_LB: + INP_WLOCK(in6p); + if ((so->so_options & SO_REUSEPORT_LB) != 0) + in6p->inp_flags2 |= INP_REUSEPORT_LB; + else + in6p->inp_flags2 &= ~INP_REUSEPORT_LB; + INP_WUNLOCK(in6p); + error = 0; + break; case SO_SETFIB: INP_WLOCK(in6p); in6p->inp_inc.inc_fibnum = so->so_fibnum; INP_WUNLOCK(in6p); error = 0; break; case SO_MAX_PACING_RATE: #ifdef RATELIMIT INP_WLOCK(in6p); in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED; INP_WUNLOCK(in6p); error = 0; #else error = EOPNOTSUPP; #endif break; default: break; } } } else { /* level == IPPROTO_IPV6 */ switch (op) { case SOPT_SET: switch (optname) { case IPV6_2292PKTOPTIONS: #ifdef IPV6_PKTOPTIONS case IPV6_PKTOPTIONS: #endif { struct mbuf *m; if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) { printf("ip6_ctloutput: mbuf limit hit\n"); error = ENOBUFS; break; } error = soopt_getm(sopt, &m); /* XXX */ if (error != 0) break; error = soopt_mcopyin(sopt, m); /* XXX */ if (error != 0) break; error = ip6_pcbopts(&in6p->in6p_outputopts, m, so, sopt); m_freem(m); /* XXX */ break; } /* * Use of some Hop-by-Hop options or some * Destination options, might require special * privilege. That is, normal applications * (without special privilege) might be forbidden * from setting certain options in outgoing packets, * and might never see certain options in received * packets. [RFC 2292 Section 6] * KAME specific note: * KAME prevents non-privileged users from sending or * receiving ANY hbh/dst options in order to avoid * overhead of parsing options in the kernel. */ case IPV6_RECVHOPOPTS: case IPV6_RECVDSTOPTS: case IPV6_RECVRTHDRDSTOPTS: if (td != NULL) { error = priv_check(td, PRIV_NETINET_SETHDROPTS); if (error) break; } /* FALLTHROUGH */ case IPV6_UNICAST_HOPS: case IPV6_HOPLIMIT: case IPV6_RECVPKTINFO: case IPV6_RECVHOPLIMIT: case IPV6_RECVRTHDR: case IPV6_RECVPATHMTU: case IPV6_RECVTCLASS: case IPV6_RECVFLOWID: #ifdef RSS case IPV6_RECVRSSBUCKETID: #endif case IPV6_V6ONLY: case IPV6_AUTOFLOWLABEL: case IPV6_ORIGDSTADDR: case IPV6_BINDANY: case IPV6_BINDMULTI: #ifdef RSS case IPV6_RSS_LISTEN_BUCKET: #endif if (optname == IPV6_BINDANY && td != NULL) { error = priv_check(td, PRIV_NETINET_BINDANY); if (error) break; } if (optlen != sizeof(int)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; switch (optname) { case IPV6_UNICAST_HOPS: if (optval < -1 || optval >= 256) error = EINVAL; else { /* -1 = kernel default */ in6p->in6p_hops = optval; if ((in6p->inp_vflag & INP_IPV4) != 0) in6p->inp_ip_ttl = optval; } break; #define OPTSET(bit) \ do { \ INP_WLOCK(in6p); \ if (optval) \ in6p->inp_flags |= (bit); \ else \ in6p->inp_flags &= ~(bit); \ INP_WUNLOCK(in6p); \ } while (/*CONSTCOND*/ 0) #define OPTSET2292(bit) \ do { \ INP_WLOCK(in6p); \ in6p->inp_flags |= IN6P_RFC2292; \ if (optval) \ in6p->inp_flags |= (bit); \ else \ in6p->inp_flags &= ~(bit); \ INP_WUNLOCK(in6p); \ } while (/*CONSTCOND*/ 0) #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) #define OPTSET2_N(bit, val) do { \ if (val) \ in6p->inp_flags2 |= bit; \ else \ in6p->inp_flags2 &= ~bit; \ } while (0) #define OPTSET2(bit, val) do { \ INP_WLOCK(in6p); \ OPTSET2_N(bit, val); \ INP_WUNLOCK(in6p); \ } while (0) #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0) #define OPTSET2292_EXCLUSIVE(bit) \ do { \ INP_WLOCK(in6p); \ if (OPTBIT(IN6P_RFC2292)) { \ error = EINVAL; \ } else { \ if (optval) \ in6p->inp_flags |= (bit); \ else \ in6p->inp_flags &= ~(bit); \ } \ INP_WUNLOCK(in6p); \ } while (/*CONSTCOND*/ 0) case IPV6_RECVPKTINFO: OPTSET2292_EXCLUSIVE(IN6P_PKTINFO); break; case IPV6_HOPLIMIT: { struct ip6_pktopts **optp; /* cannot mix with RFC2292 */ if (OPTBIT(IN6P_RFC2292)) { error = EINVAL; break; } optp = &in6p->in6p_outputopts; error = ip6_pcbopt(IPV6_HOPLIMIT, (u_char *)&optval, sizeof(optval), optp, (td != NULL) ? td->td_ucred : NULL, uproto); break; } case IPV6_RECVHOPLIMIT: OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT); break; case IPV6_RECVHOPOPTS: OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS); break; case IPV6_RECVDSTOPTS: OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS); break; case IPV6_RECVRTHDRDSTOPTS: OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS); break; case IPV6_RECVRTHDR: OPTSET2292_EXCLUSIVE(IN6P_RTHDR); break; case IPV6_RECVPATHMTU: /* * We ignore this option for TCP * sockets. * (RFC3542 leaves this case * unspecified.) */ if (uproto != IPPROTO_TCP) OPTSET(IN6P_MTU); break; case IPV6_RECVFLOWID: OPTSET2(INP_RECVFLOWID, optval); break; #ifdef RSS case IPV6_RECVRSSBUCKETID: OPTSET2(INP_RECVRSSBUCKETID, optval); break; #endif case IPV6_V6ONLY: /* * make setsockopt(IPV6_V6ONLY) * available only prior to bind(2). * see ipng mailing list, Jun 22 2001. */ if (in6p->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { error = EINVAL; break; } OPTSET(IN6P_IPV6_V6ONLY); if (optval) in6p->inp_vflag &= ~INP_IPV4; else in6p->inp_vflag |= INP_IPV4; break; case IPV6_RECVTCLASS: /* cannot mix with RFC2292 XXX */ OPTSET2292_EXCLUSIVE(IN6P_TCLASS); break; case IPV6_AUTOFLOWLABEL: OPTSET(IN6P_AUTOFLOWLABEL); break; case IPV6_ORIGDSTADDR: OPTSET2(INP_ORIGDSTADDR, optval); break; case IPV6_BINDANY: OPTSET(INP_BINDANY); break; case IPV6_BINDMULTI: OPTSET2(INP_BINDMULTI, optval); break; #ifdef RSS case IPV6_RSS_LISTEN_BUCKET: if ((optval >= 0) && (optval < rss_getnumbuckets())) { INP_WLOCK(in6p); in6p->inp_rss_listen_bucket = optval; OPTSET2_N(INP_RSS_BUCKET_SET, 1); INP_WUNLOCK(in6p); } else { error = EINVAL; } break; #endif } break; case IPV6_TCLASS: case IPV6_DONTFRAG: case IPV6_USE_MIN_MTU: case IPV6_PREFER_TEMPADDR: if (optlen != sizeof(optval)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; { struct ip6_pktopts **optp; optp = &in6p->in6p_outputopts; error = ip6_pcbopt(optname, (u_char *)&optval, sizeof(optval), optp, (td != NULL) ? td->td_ucred : NULL, uproto); break; } case IPV6_2292PKTINFO: case IPV6_2292HOPLIMIT: case IPV6_2292HOPOPTS: case IPV6_2292DSTOPTS: case IPV6_2292RTHDR: /* RFC 2292 */ if (optlen != sizeof(int)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; switch (optname) { case IPV6_2292PKTINFO: OPTSET2292(IN6P_PKTINFO); break; case IPV6_2292HOPLIMIT: OPTSET2292(IN6P_HOPLIMIT); break; case IPV6_2292HOPOPTS: /* * Check super-user privilege. * See comments for IPV6_RECVHOPOPTS. */ if (td != NULL) { error = priv_check(td, PRIV_NETINET_SETHDROPTS); if (error) return (error); } OPTSET2292(IN6P_HOPOPTS); break; case IPV6_2292DSTOPTS: if (td != NULL) { error = priv_check(td, PRIV_NETINET_SETHDROPTS); if (error) return (error); } OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ break; case IPV6_2292RTHDR: OPTSET2292(IN6P_RTHDR); break; } break; case IPV6_PKTINFO: case IPV6_HOPOPTS: case IPV6_RTHDR: case IPV6_DSTOPTS: case IPV6_RTHDRDSTOPTS: case IPV6_NEXTHOP: { /* new advanced API (RFC3542) */ u_char *optbuf; u_char optbuf_storage[MCLBYTES]; int optlen; struct ip6_pktopts **optp; /* cannot mix with RFC2292 */ if (OPTBIT(IN6P_RFC2292)) { error = EINVAL; break; } /* * We only ensure valsize is not too large * here. Further validation will be done * later. */ error = sooptcopyin(sopt, optbuf_storage, sizeof(optbuf_storage), 0); if (error) break; optlen = sopt->sopt_valsize; optbuf = optbuf_storage; optp = &in6p->in6p_outputopts; error = ip6_pcbopt(optname, optbuf, optlen, optp, (td != NULL) ? td->td_ucred : NULL, uproto); break; } #undef OPTSET case IPV6_MULTICAST_IF: case IPV6_MULTICAST_HOPS: case IPV6_MULTICAST_LOOP: case IPV6_JOIN_GROUP: case IPV6_LEAVE_GROUP: case IPV6_MSFILTER: case MCAST_BLOCK_SOURCE: case MCAST_UNBLOCK_SOURCE: case MCAST_JOIN_GROUP: case MCAST_LEAVE_GROUP: case MCAST_JOIN_SOURCE_GROUP: case MCAST_LEAVE_SOURCE_GROUP: error = ip6_setmoptions(in6p, sopt); break; case IPV6_PORTRANGE: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; INP_WLOCK(in6p); switch (optval) { case IPV6_PORTRANGE_DEFAULT: in6p->inp_flags &= ~(INP_LOWPORT); in6p->inp_flags &= ~(INP_HIGHPORT); break; case IPV6_PORTRANGE_HIGH: in6p->inp_flags &= ~(INP_LOWPORT); in6p->inp_flags |= INP_HIGHPORT; break; case IPV6_PORTRANGE_LOW: in6p->inp_flags &= ~(INP_HIGHPORT); in6p->inp_flags |= INP_LOWPORT; break; default: error = EINVAL; break; } INP_WUNLOCK(in6p); break; #if defined(IPSEC) || defined(IPSEC_SUPPORT) case IPV6_IPSEC_POLICY: if (IPSEC_ENABLED(ipv6)) { error = IPSEC_PCBCTL(ipv6, in6p, sopt); break; } /* FALLTHROUGH */ #endif /* IPSEC */ default: error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (optname) { case IPV6_2292PKTOPTIONS: #ifdef IPV6_PKTOPTIONS case IPV6_PKTOPTIONS: #endif /* * RFC3542 (effectively) deprecated the * semantics of the 2292-style pktoptions. * Since it was not reliable in nature (i.e., * applications had to expect the lack of some * information after all), it would make sense * to simplify this part by always returning * empty data. */ sopt->sopt_valsize = 0; break; case IPV6_RECVHOPOPTS: case IPV6_RECVDSTOPTS: case IPV6_RECVRTHDRDSTOPTS: case IPV6_UNICAST_HOPS: case IPV6_RECVPKTINFO: case IPV6_RECVHOPLIMIT: case IPV6_RECVRTHDR: case IPV6_RECVPATHMTU: case IPV6_V6ONLY: case IPV6_PORTRANGE: case IPV6_RECVTCLASS: case IPV6_AUTOFLOWLABEL: case IPV6_BINDANY: case IPV6_FLOWID: case IPV6_FLOWTYPE: case IPV6_RECVFLOWID: #ifdef RSS case IPV6_RSSBUCKETID: case IPV6_RECVRSSBUCKETID: #endif case IPV6_BINDMULTI: switch (optname) { case IPV6_RECVHOPOPTS: optval = OPTBIT(IN6P_HOPOPTS); break; case IPV6_RECVDSTOPTS: optval = OPTBIT(IN6P_DSTOPTS); break; case IPV6_RECVRTHDRDSTOPTS: optval = OPTBIT(IN6P_RTHDRDSTOPTS); break; case IPV6_UNICAST_HOPS: optval = in6p->in6p_hops; break; case IPV6_RECVPKTINFO: optval = OPTBIT(IN6P_PKTINFO); break; case IPV6_RECVHOPLIMIT: optval = OPTBIT(IN6P_HOPLIMIT); break; case IPV6_RECVRTHDR: optval = OPTBIT(IN6P_RTHDR); break; case IPV6_RECVPATHMTU: optval = OPTBIT(IN6P_MTU); break; case IPV6_V6ONLY: optval = OPTBIT(IN6P_IPV6_V6ONLY); break; case IPV6_PORTRANGE: { int flags; flags = in6p->inp_flags; if (flags & INP_HIGHPORT) optval = IPV6_PORTRANGE_HIGH; else if (flags & INP_LOWPORT) optval = IPV6_PORTRANGE_LOW; else optval = 0; break; } case IPV6_RECVTCLASS: optval = OPTBIT(IN6P_TCLASS); break; case IPV6_AUTOFLOWLABEL: optval = OPTBIT(IN6P_AUTOFLOWLABEL); break; case IPV6_ORIGDSTADDR: optval = OPTBIT2(INP_ORIGDSTADDR); break; case IPV6_BINDANY: optval = OPTBIT(INP_BINDANY); break; case IPV6_FLOWID: optval = in6p->inp_flowid; break; case IPV6_FLOWTYPE: optval = in6p->inp_flowtype; break; case IPV6_RECVFLOWID: optval = OPTBIT2(INP_RECVFLOWID); break; #ifdef RSS case IPV6_RSSBUCKETID: retval = rss_hash2bucket(in6p->inp_flowid, in6p->inp_flowtype, &rss_bucket); if (retval == 0) optval = rss_bucket; else error = EINVAL; break; case IPV6_RECVRSSBUCKETID: optval = OPTBIT2(INP_RECVRSSBUCKETID); break; #endif case IPV6_BINDMULTI: optval = OPTBIT2(INP_BINDMULTI); break; } if (error) break; error = sooptcopyout(sopt, &optval, sizeof optval); break; case IPV6_PATHMTU: { u_long pmtu = 0; struct ip6_mtuinfo mtuinfo; struct in6_addr addr; if (!(so->so_state & SS_ISCONNECTED)) return (ENOTCONN); /* * XXX: we dot not consider the case of source * routing, or optional information to specify * the outgoing interface. * Copy faddr out of in6p to avoid holding lock * on inp during route lookup. */ INP_RLOCK(in6p); bcopy(&in6p->in6p_faddr, &addr, sizeof(addr)); INP_RUNLOCK(in6p); error = ip6_getpmtu_ctl(so->so_fibnum, &addr, &pmtu); if (error) break; if (pmtu > IPV6_MAXPACKET) pmtu = IPV6_MAXPACKET; bzero(&mtuinfo, sizeof(mtuinfo)); mtuinfo.ip6m_mtu = (u_int32_t)pmtu; optdata = (void *)&mtuinfo; optdatalen = sizeof(mtuinfo); error = sooptcopyout(sopt, optdata, optdatalen); break; } case IPV6_2292PKTINFO: case IPV6_2292HOPLIMIT: case IPV6_2292HOPOPTS: case IPV6_2292RTHDR: case IPV6_2292DSTOPTS: switch (optname) { case IPV6_2292PKTINFO: optval = OPTBIT(IN6P_PKTINFO); break; case IPV6_2292HOPLIMIT: optval = OPTBIT(IN6P_HOPLIMIT); break; case IPV6_2292HOPOPTS: optval = OPTBIT(IN6P_HOPOPTS); break; case IPV6_2292RTHDR: optval = OPTBIT(IN6P_RTHDR); break; case IPV6_2292DSTOPTS: optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); break; } error = sooptcopyout(sopt, &optval, sizeof optval); break; case IPV6_PKTINFO: case IPV6_HOPOPTS: case IPV6_RTHDR: case IPV6_DSTOPTS: case IPV6_RTHDRDSTOPTS: case IPV6_NEXTHOP: case IPV6_TCLASS: case IPV6_DONTFRAG: case IPV6_USE_MIN_MTU: case IPV6_PREFER_TEMPADDR: error = ip6_getpcbopt(in6p, optname, sopt); break; case IPV6_MULTICAST_IF: case IPV6_MULTICAST_HOPS: case IPV6_MULTICAST_LOOP: case IPV6_MSFILTER: error = ip6_getmoptions(in6p, sopt); break; #if defined(IPSEC) || defined(IPSEC_SUPPORT) case IPV6_IPSEC_POLICY: if (IPSEC_ENABLED(ipv6)) { error = IPSEC_PCBCTL(ipv6, in6p, sopt); break; } /* FALLTHROUGH */ #endif /* IPSEC */ default: error = ENOPROTOOPT; break; } break; } } return (error); } int ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) { int error = 0, optval, optlen; const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); struct inpcb *in6p = sotoinpcb(so); int level, op, optname; level = sopt->sopt_level; op = sopt->sopt_dir; optname = sopt->sopt_name; optlen = sopt->sopt_valsize; if (level != IPPROTO_IPV6) { return (EINVAL); } switch (optname) { case IPV6_CHECKSUM: /* * For ICMPv6 sockets, no modification allowed for checksum * offset, permit "no change" values to help existing apps. * * RFC3542 says: "An attempt to set IPV6_CHECKSUM * for an ICMPv6 socket will fail." * The current behavior does not meet RFC3542. */ switch (op) { case SOPT_SET: if (optlen != sizeof(int)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); if (error) break; if ((optval % 2) != 0) { /* the API assumes even offset values */ error = EINVAL; } else if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) { if (optval != icmp6off) error = EINVAL; } else in6p->in6p_cksum = optval; break; case SOPT_GET: if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) optval = icmp6off; else optval = in6p->in6p_cksum; error = sooptcopyout(sopt, &optval, sizeof(optval)); break; default: error = EINVAL; break; } break; default: error = ENOPROTOOPT; break; } return (error); } /* * Set up IP6 options in pcb for insertion in output packets or * specifying behavior of outgoing packets. */ static int ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so, struct sockopt *sopt) { struct ip6_pktopts *opt = *pktopt; int error = 0; struct thread *td = sopt->sopt_td; /* turn off any old options. */ if (opt) { #ifdef DIAGNOSTIC if (opt->ip6po_pktinfo || opt->ip6po_nexthop || opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || opt->ip6po_rhinfo.ip6po_rhi_rthdr) printf("ip6_pcbopts: all specified options are cleared.\n"); #endif ip6_clearpktopts(opt, -1); } else opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); *pktopt = NULL; if (!m || m->m_len == 0) { /* * Only turning off any previous options, regardless of * whether the opt is just created or given. */ free(opt, M_IP6OPT); return (0); } /* set options specified by user. */ if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ? td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) { ip6_clearpktopts(opt, -1); /* XXX: discard all options */ free(opt, M_IP6OPT); return (error); } *pktopt = opt; return (0); } /* * initialize ip6_pktopts. beware that there are non-zero default values in * the struct. */ void ip6_initpktopts(struct ip6_pktopts *opt) { bzero(opt, sizeof(*opt)); opt->ip6po_hlim = -1; /* -1 means default hop limit */ opt->ip6po_tclass = -1; /* -1 means default traffic class */ opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; } static int ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, struct ucred *cred, int uproto) { struct ip6_pktopts *opt; if (*pktopt == NULL) { *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, M_WAITOK); ip6_initpktopts(*pktopt); } opt = *pktopt; return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); } #define GET_PKTOPT_VAR(field, lenexpr) do { \ if (pktopt && pktopt->field) { \ INP_RUNLOCK(in6p); \ optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK); \ malloc_optdata = true; \ INP_RLOCK(in6p); \ if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \ INP_RUNLOCK(in6p); \ free(optdata, M_TEMP); \ return (ECONNRESET); \ } \ pktopt = in6p->in6p_outputopts; \ if (pktopt && pktopt->field) { \ optdatalen = min(lenexpr, sopt->sopt_valsize); \ bcopy(&pktopt->field, optdata, optdatalen); \ } else { \ free(optdata, M_TEMP); \ optdata = NULL; \ malloc_optdata = false; \ } \ } \ } while(0) #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field, \ (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3) #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field, \ pktopt->field->sa_len) static int ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt) { void *optdata = NULL; bool malloc_optdata = false; int optdatalen = 0; int error = 0; struct in6_pktinfo null_pktinfo; int deftclass = 0, on; int defminmtu = IP6PO_MINMTU_MCASTONLY; int defpreftemp = IP6PO_TEMPADDR_SYSTEM; struct ip6_pktopts *pktopt; INP_RLOCK(in6p); pktopt = in6p->in6p_outputopts; switch (optname) { case IPV6_PKTINFO: optdata = (void *)&null_pktinfo; if (pktopt && pktopt->ip6po_pktinfo) { bcopy(pktopt->ip6po_pktinfo, &null_pktinfo, sizeof(null_pktinfo)); in6_clearscope(&null_pktinfo.ipi6_addr); } else { /* XXX: we don't have to do this every time... */ bzero(&null_pktinfo, sizeof(null_pktinfo)); } optdatalen = sizeof(struct in6_pktinfo); break; case IPV6_TCLASS: if (pktopt && pktopt->ip6po_tclass >= 0) deftclass = pktopt->ip6po_tclass; optdata = (void *)&deftclass; optdatalen = sizeof(int); break; case IPV6_HOPOPTS: GET_PKTOPT_EXT_HDR(ip6po_hbh); break; case IPV6_RTHDR: GET_PKTOPT_EXT_HDR(ip6po_rthdr); break; case IPV6_RTHDRDSTOPTS: GET_PKTOPT_EXT_HDR(ip6po_dest1); break; case IPV6_DSTOPTS: GET_PKTOPT_EXT_HDR(ip6po_dest2); break; case IPV6_NEXTHOP: GET_PKTOPT_SOCKADDR(ip6po_nexthop); break; case IPV6_USE_MIN_MTU: if (pktopt) defminmtu = pktopt->ip6po_minmtu; optdata = (void *)&defminmtu; optdatalen = sizeof(int); break; case IPV6_DONTFRAG: if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) on = 1; else on = 0; optdata = (void *)&on; optdatalen = sizeof(on); break; case IPV6_PREFER_TEMPADDR: if (pktopt) defpreftemp = pktopt->ip6po_prefer_tempaddr; optdata = (void *)&defpreftemp; optdatalen = sizeof(int); break; default: /* should not happen */ #ifdef DIAGNOSTIC panic("ip6_getpcbopt: unexpected option\n"); #endif INP_RUNLOCK(in6p); return (ENOPROTOOPT); } INP_RUNLOCK(in6p); error = sooptcopyout(sopt, optdata, optdatalen); if (malloc_optdata) free(optdata, M_TEMP); return (error); } void ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) { if (pktopt == NULL) return; if (optname == -1 || optname == IPV6_PKTINFO) { if (pktopt->ip6po_pktinfo) free(pktopt->ip6po_pktinfo, M_IP6OPT); pktopt->ip6po_pktinfo = NULL; } if (optname == -1 || optname == IPV6_HOPLIMIT) pktopt->ip6po_hlim = -1; if (optname == -1 || optname == IPV6_TCLASS) pktopt->ip6po_tclass = -1; if (optname == -1 || optname == IPV6_NEXTHOP) { if (pktopt->ip6po_nextroute.ro_rt) { RTFREE(pktopt->ip6po_nextroute.ro_rt); pktopt->ip6po_nextroute.ro_rt = NULL; } if (pktopt->ip6po_nexthop) free(pktopt->ip6po_nexthop, M_IP6OPT); pktopt->ip6po_nexthop = NULL; } if (optname == -1 || optname == IPV6_HOPOPTS) { if (pktopt->ip6po_hbh) free(pktopt->ip6po_hbh, M_IP6OPT); pktopt->ip6po_hbh = NULL; } if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { if (pktopt->ip6po_dest1) free(pktopt->ip6po_dest1, M_IP6OPT); pktopt->ip6po_dest1 = NULL; } if (optname == -1 || optname == IPV6_RTHDR) { if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; if (pktopt->ip6po_route.ro_rt) { RTFREE(pktopt->ip6po_route.ro_rt); pktopt->ip6po_route.ro_rt = NULL; } } if (optname == -1 || optname == IPV6_DSTOPTS) { if (pktopt->ip6po_dest2) free(pktopt->ip6po_dest2, M_IP6OPT); pktopt->ip6po_dest2 = NULL; } } #define PKTOPT_EXTHDRCPY(type) \ do {\ if (src->type) {\ int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ dst->type = malloc(hlen, M_IP6OPT, canwait);\ if (dst->type == NULL)\ goto bad;\ bcopy(src->type, dst->type, hlen);\ }\ } while (/*CONSTCOND*/ 0) static int copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) { if (dst == NULL || src == NULL) { printf("ip6_clearpktopts: invalid argument\n"); return (EINVAL); } dst->ip6po_hlim = src->ip6po_hlim; dst->ip6po_tclass = src->ip6po_tclass; dst->ip6po_flags = src->ip6po_flags; dst->ip6po_minmtu = src->ip6po_minmtu; dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; if (src->ip6po_pktinfo) { dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), M_IP6OPT, canwait); if (dst->ip6po_pktinfo == NULL) goto bad; *dst->ip6po_pktinfo = *src->ip6po_pktinfo; } if (src->ip6po_nexthop) { dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, M_IP6OPT, canwait); if (dst->ip6po_nexthop == NULL) goto bad; bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, src->ip6po_nexthop->sa_len); } PKTOPT_EXTHDRCPY(ip6po_hbh); PKTOPT_EXTHDRCPY(ip6po_dest1); PKTOPT_EXTHDRCPY(ip6po_dest2); PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ return (0); bad: ip6_clearpktopts(dst, -1); return (ENOBUFS); } #undef PKTOPT_EXTHDRCPY struct ip6_pktopts * ip6_copypktopts(struct ip6_pktopts *src, int canwait) { int error; struct ip6_pktopts *dst; dst = malloc(sizeof(*dst), M_IP6OPT, canwait); if (dst == NULL) return (NULL); ip6_initpktopts(dst); if ((error = copypktopts(dst, src, canwait)) != 0) { free(dst, M_IP6OPT); return (NULL); } return (dst); } void ip6_freepcbopts(struct ip6_pktopts *pktopt) { if (pktopt == NULL) return; ip6_clearpktopts(pktopt, -1); free(pktopt, M_IP6OPT); } /* * Set IPv6 outgoing packet options based on advanced API. */ int ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto) { struct cmsghdr *cm = NULL; if (control == NULL || opt == NULL) return (EINVAL); ip6_initpktopts(opt); if (stickyopt) { int error; /* * If stickyopt is provided, make a local copy of the options * for this particular packet, then override them by ancillary * objects. * XXX: copypktopts() does not copy the cached route to a next * hop (if any). This is not very good in terms of efficiency, * but we can allow this since this option should be rarely * used. */ if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) return (error); } /* * XXX: Currently, we assume all the optional information is stored * in a single mbuf. */ if (control->m_next) return (EINVAL); for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { int error; if (control->m_len < CMSG_LEN(0)) return (EINVAL); cm = mtod(control, struct cmsghdr *); if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) return (EINVAL); if (cm->cmsg_level != IPPROTO_IPV6) continue; error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); if (error) return (error); } return (0); } /* * Set a particular packet option, as a sticky option or an ancillary data * item. "len" can be 0 only when it's a sticky option. * We have 4 cases of combination of "sticky" and "cmsg": * "sticky=0, cmsg=0": impossible * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data * "sticky=1, cmsg=0": RFC3542 socket option * "sticky=1, cmsg=1": RFC2292 socket option */ static int ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, struct ucred *cred, int sticky, int cmsg, int uproto) { int minmtupolicy, preftemp; int error; if (!sticky && !cmsg) { #ifdef DIAGNOSTIC printf("ip6_setpktopt: impossible case\n"); #endif return (EINVAL); } /* * IPV6_2292xxx is for backward compatibility to RFC2292, and should * not be specified in the context of RFC3542. Conversely, * RFC3542 types should not be specified in the context of RFC2292. */ if (!cmsg) { switch (optname) { case IPV6_2292PKTINFO: case IPV6_2292HOPLIMIT: case IPV6_2292NEXTHOP: case IPV6_2292HOPOPTS: case IPV6_2292DSTOPTS: case IPV6_2292RTHDR: case IPV6_2292PKTOPTIONS: return (ENOPROTOOPT); } } if (sticky && cmsg) { switch (optname) { case IPV6_PKTINFO: case IPV6_HOPLIMIT: case IPV6_NEXTHOP: case IPV6_HOPOPTS: case IPV6_DSTOPTS: case IPV6_RTHDRDSTOPTS: case IPV6_RTHDR: case IPV6_USE_MIN_MTU: case IPV6_DONTFRAG: case IPV6_TCLASS: case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ return (ENOPROTOOPT); } } switch (optname) { case IPV6_2292PKTINFO: case IPV6_PKTINFO: { struct ifnet *ifp = NULL; struct in6_pktinfo *pktinfo; if (len != sizeof(struct in6_pktinfo)) return (EINVAL); pktinfo = (struct in6_pktinfo *)buf; /* * An application can clear any sticky IPV6_PKTINFO option by * doing a "regular" setsockopt with ipi6_addr being * in6addr_any and ipi6_ifindex being zero. * [RFC 3542, Section 6] */ if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && pktinfo->ipi6_ifindex == 0 && IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { ip6_clearpktopts(opt, optname); break; } if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { return (EINVAL); } if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr)) return (EINVAL); /* validate the interface index if specified. */ if (pktinfo->ipi6_ifindex > V_if_index) return (ENXIO); if (pktinfo->ipi6_ifindex) { ifp = ifnet_byindex(pktinfo->ipi6_ifindex); if (ifp == NULL) return (ENXIO); } if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)) return (ENETDOWN); if (ifp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { struct in6_ifaddr *ia; in6_setscope(&pktinfo->ipi6_addr, ifp, NULL); ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr); if (ia == NULL) return (EADDRNOTAVAIL); ifa_free(&ia->ia_ifa); } /* * We store the address anyway, and let in6_selectsrc() * validate the specified address. This is because ipi6_addr * may not have enough information about its scope zone, and * we may need additional information (such as outgoing * interface or the scope zone of a destination address) to * disambiguate the scope. * XXX: the delay of the validation may confuse the * application when it is used as a sticky option. */ if (opt->ip6po_pktinfo == NULL) { opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), M_IP6OPT, M_NOWAIT); if (opt->ip6po_pktinfo == NULL) return (ENOBUFS); } bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); break; } case IPV6_2292HOPLIMIT: case IPV6_HOPLIMIT: { int *hlimp; /* * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT * to simplify the ordering among hoplimit options. */ if (optname == IPV6_HOPLIMIT && sticky) return (ENOPROTOOPT); if (len != sizeof(int)) return (EINVAL); hlimp = (int *)buf; if (*hlimp < -1 || *hlimp > 255) return (EINVAL); opt->ip6po_hlim = *hlimp; break; } case IPV6_TCLASS: { int tclass; if (len != sizeof(int)) return (EINVAL); tclass = *(int *)buf; if (tclass < -1 || tclass > 255) return (EINVAL); opt->ip6po_tclass = tclass; break; } case IPV6_2292NEXTHOP: case IPV6_NEXTHOP: if (cred != NULL) { error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS, 0); if (error) return (error); } if (len == 0) { /* just remove the option */ ip6_clearpktopts(opt, IPV6_NEXTHOP); break; } /* check if cmsg_len is large enough for sa_len */ if (len < sizeof(struct sockaddr) || len < *buf) return (EINVAL); switch (((struct sockaddr *)buf)->sa_family) { case AF_INET6: { struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; int error; if (sa6->sin6_len != sizeof(struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { return (EINVAL); } if ((error = sa6_embedscope(sa6, V_ip6_use_defzone)) != 0) { return (error); } break; } case AF_LINK: /* should eventually be supported */ default: return (EAFNOSUPPORT); } /* turn off the previous option, then set the new option. */ ip6_clearpktopts(opt, IPV6_NEXTHOP); opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); if (opt->ip6po_nexthop == NULL) return (ENOBUFS); bcopy(buf, opt->ip6po_nexthop, *buf); break; case IPV6_2292HOPOPTS: case IPV6_HOPOPTS: { struct ip6_hbh *hbh; int hbhlen; /* * XXX: We don't allow a non-privileged user to set ANY HbH * options, since per-option restriction has too much * overhead. */ if (cred != NULL) { error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS, 0); if (error) return (error); } if (len == 0) { ip6_clearpktopts(opt, IPV6_HOPOPTS); break; /* just remove the option */ } /* message length validation */ if (len < sizeof(struct ip6_hbh)) return (EINVAL); hbh = (struct ip6_hbh *)buf; hbhlen = (hbh->ip6h_len + 1) << 3; if (len != hbhlen) return (EINVAL); /* turn off the previous option, then set the new option. */ ip6_clearpktopts(opt, IPV6_HOPOPTS); opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); if (opt->ip6po_hbh == NULL) return (ENOBUFS); bcopy(hbh, opt->ip6po_hbh, hbhlen); break; } case IPV6_2292DSTOPTS: case IPV6_DSTOPTS: case IPV6_RTHDRDSTOPTS: { struct ip6_dest *dest, **newdest = NULL; int destlen; if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */ error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS, 0); if (error) return (error); } if (len == 0) { ip6_clearpktopts(opt, optname); break; /* just remove the option */ } /* message length validation */ if (len < sizeof(struct ip6_dest)) return (EINVAL); dest = (struct ip6_dest *)buf; destlen = (dest->ip6d_len + 1) << 3; if (len != destlen) return (EINVAL); /* * Determine the position that the destination options header * should be inserted; before or after the routing header. */ switch (optname) { case IPV6_2292DSTOPTS: /* * The old advacned API is ambiguous on this point. * Our approach is to determine the position based * according to the existence of a routing header. * Note, however, that this depends on the order of the * extension headers in the ancillary data; the 1st * part of the destination options header must appear * before the routing header in the ancillary data, * too. * RFC3542 solved the ambiguity by introducing * separate ancillary data or option types. */ if (opt->ip6po_rthdr == NULL) newdest = &opt->ip6po_dest1; else newdest = &opt->ip6po_dest2; break; case IPV6_RTHDRDSTOPTS: newdest = &opt->ip6po_dest1; break; case IPV6_DSTOPTS: newdest = &opt->ip6po_dest2; break; } /* turn off the previous option, then set the new option. */ ip6_clearpktopts(opt, optname); *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); if (*newdest == NULL) return (ENOBUFS); bcopy(dest, *newdest, destlen); break; } case IPV6_2292RTHDR: case IPV6_RTHDR: { struct ip6_rthdr *rth; int rthlen; if (len == 0) { ip6_clearpktopts(opt, IPV6_RTHDR); break; /* just remove the option */ } /* message length validation */ if (len < sizeof(struct ip6_rthdr)) return (EINVAL); rth = (struct ip6_rthdr *)buf; rthlen = (rth->ip6r_len + 1) << 3; if (len != rthlen) return (EINVAL); switch (rth->ip6r_type) { case IPV6_RTHDR_TYPE_0: if (rth->ip6r_len == 0) /* must contain one addr */ return (EINVAL); if (rth->ip6r_len % 2) /* length must be even */ return (EINVAL); if (rth->ip6r_len / 2 != rth->ip6r_segleft) return (EINVAL); break; default: return (EINVAL); /* not supported */ } /* turn off the previous option */ ip6_clearpktopts(opt, IPV6_RTHDR); opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); if (opt->ip6po_rthdr == NULL) return (ENOBUFS); bcopy(rth, opt->ip6po_rthdr, rthlen); break; } case IPV6_USE_MIN_MTU: if (len != sizeof(int)) return (EINVAL); minmtupolicy = *(int *)buf; if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && minmtupolicy != IP6PO_MINMTU_DISABLE && minmtupolicy != IP6PO_MINMTU_ALL) { return (EINVAL); } opt->ip6po_minmtu = minmtupolicy; break; case IPV6_DONTFRAG: if (len != sizeof(int)) return (EINVAL); if (uproto == IPPROTO_TCP || *(int *)buf == 0) { /* * we ignore this option for TCP sockets. * (RFC3542 leaves this case unspecified.) */ opt->ip6po_flags &= ~IP6PO_DONTFRAG; } else opt->ip6po_flags |= IP6PO_DONTFRAG; break; case IPV6_PREFER_TEMPADDR: if (len != sizeof(int)) return (EINVAL); preftemp = *(int *)buf; if (preftemp != IP6PO_TEMPADDR_SYSTEM && preftemp != IP6PO_TEMPADDR_NOTPREFER && preftemp != IP6PO_TEMPADDR_PREFER) { return (EINVAL); } opt->ip6po_prefer_tempaddr = preftemp; break; default: return (ENOPROTOOPT); } /* end of switch */ return (0); } /* * Routine called from ip6_output() to loop back a copy of an IP6 multicast * packet to the input queue of a specified interface. Note that this * calls the output routine of the loopback "driver", but with an interface * pointer that might NOT be &loif -- easier than replicating that code here. */ void ip6_mloopback(struct ifnet *ifp, struct mbuf *m) { struct mbuf *copym; struct ip6_hdr *ip6; copym = m_copym(m, 0, M_COPYALL, M_NOWAIT); if (copym == NULL) return; /* * Make sure to deep-copy IPv6 header portion in case the data * is in an mbuf cluster, so that we can safely override the IPv6 * header portion later. */ if (!M_WRITABLE(copym) || copym->m_len < sizeof(struct ip6_hdr)) { copym = m_pullup(copym, sizeof(struct ip6_hdr)); if (copym == NULL) return; } ip6 = mtod(copym, struct ip6_hdr *); /* * clear embedded scope identifiers if necessary. * in6_clearscope will touch the addresses only when necessary. */ in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; copym->m_pkthdr.csum_data = 0xffff; } if_simloop(ifp, copym, AF_INET6, 0); } /* * Chop IPv6 header off from the payload. */ static int ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) { struct mbuf *mh; struct ip6_hdr *ip6; ip6 = mtod(m, struct ip6_hdr *); if (m->m_len > sizeof(*ip6)) { mh = m_gethdr(M_NOWAIT, MT_DATA); if (mh == NULL) { m_freem(m); return ENOBUFS; } m_move_pkthdr(mh, m); M_ALIGN(mh, sizeof(*ip6)); m->m_len -= sizeof(*ip6); m->m_data += sizeof(*ip6); mh->m_next = m; m = mh; m->m_len = sizeof(*ip6); bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); } exthdrs->ip6e_ip6 = m; return 0; } /* * Compute IPv6 extension header length. */ int ip6_optlen(struct inpcb *in6p) { int len; if (!in6p->in6p_outputopts) return 0; len = 0; #define elen(x) \ (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) len += elen(in6p->in6p_outputopts->ip6po_hbh); if (in6p->in6p_outputopts->ip6po_rthdr) /* dest1 is valid with rthdr only */ len += elen(in6p->in6p_outputopts->ip6po_dest1); len += elen(in6p->in6p_outputopts->ip6po_rthdr); len += elen(in6p->in6p_outputopts->ip6po_dest2); return len; #undef elen } Index: head/sys/netinet6/udp6_usrreq.c =================================================================== --- head/sys/netinet6/udp6_usrreq.c (revision 334718) +++ head/sys/netinet6/udp6_usrreq.c (revision 334719) @@ -1,1320 +1,1320 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * Copyright (c) 2010-2011 Juniper Networks, Inc. * Copyright (c) 2014 Kevin Lo * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, 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. * 3. Neither the name of the project 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 PROJECT 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 PROJECT 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. * * $KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $ * $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $ */ /*- * 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. * 3. 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. * * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_rss.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 /* * UDP protocol implementation. * Per RFC 768, August, 1980. */ extern struct protosw inetsw[]; static void udp6_detach(struct socket *so); static int udp6_append(struct inpcb *inp, struct mbuf *n, int off, struct sockaddr_in6 *fromsa) { struct socket *so; struct mbuf *opts = NULL, *tmp_opts; struct udpcb *up; INP_LOCK_ASSERT(inp); /* * Engage the tunneling protocol. */ up = intoudpcb(inp); if (up->u_tun_func != NULL) { in_pcbref(inp); INP_RUNLOCK(inp); (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0], up->u_tun_ctx); INP_RLOCK(inp); return (in_pcbrele_rlocked(inp)); } #if defined(IPSEC) || defined(IPSEC_SUPPORT) /* Check AH/ESP integrity. */ if (IPSEC_ENABLED(ipv6)) { if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) { m_freem(n); return (0); } } #endif /* IPSEC */ #ifdef MAC if (mac_inpcb_check_deliver(inp, n) != 0) { m_freem(n); return (0); } #endif opts = NULL; if (inp->inp_flags & INP_CONTROLOPTS || inp->inp_socket->so_options & SO_TIMESTAMP) ip6_savecontrol(inp, n, &opts); if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) { tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1], sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6); if (tmp_opts) { if (opts) { tmp_opts->m_next = opts; opts = tmp_opts; } else opts = tmp_opts; } } m_adj(n, off + sizeof(struct udphdr)); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_rcv); if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n, opts) == 0) { SOCKBUF_UNLOCK(&so->so_rcv); m_freem(n); if (opts) m_freem(opts); UDPSTAT_INC(udps_fullsock); } else sorwakeup_locked(so); return (0); } int udp6_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; struct ifnet *ifp; struct ip6_hdr *ip6; struct udphdr *uh; struct inpcb *inp; struct inpcbinfo *pcbinfo; struct udpcb *up; int off = *offp; int cscov_partial; int plen, ulen; struct sockaddr_in6 fromsa[2]; struct m_tag *fwd_tag; uint16_t uh_sum; uint8_t nxt; ifp = m->m_pkthdr.rcvif; #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); ip6 = mtod(m, struct ip6_hdr *); uh = (struct udphdr *)((caddr_t)ip6 + off); #else IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh)); if (!uh) return (IPPROTO_DONE); ip6 = mtod(m, struct ip6_hdr *); #endif UDPSTAT_INC(udps_ipackets); /* * Destination port of 0 is illegal, based on RFC768. */ if (uh->uh_dport == 0) goto badunlocked; plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6); ulen = ntohs((u_short)uh->uh_ulen); nxt = proto; cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0; if (nxt == IPPROTO_UDPLITE) { /* Zero means checksum over the complete packet. */ if (ulen == 0) ulen = plen; if (ulen == plen) cscov_partial = 0; if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) { /* XXX: What is the right UDPLite MIB counter? */ goto badunlocked; } if (uh->uh_sum == 0) { /* XXX: What is the right UDPLite MIB counter? */ goto badunlocked; } } else { if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) { UDPSTAT_INC(udps_badlen); goto badunlocked; } if (uh->uh_sum == 0) { UDPSTAT_INC(udps_nosum); goto badunlocked; } } if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) && !cscov_partial) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) uh_sum = m->m_pkthdr.csum_data; else uh_sum = in6_cksum_pseudo(ip6, ulen, nxt, m->m_pkthdr.csum_data); uh_sum ^= 0xffff; } else uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen); if (uh_sum != 0) { UDPSTAT_INC(udps_badsum); goto badunlocked; } /* * Construct sockaddr format source address. */ init_sin6(&fromsa[0], m, 0); fromsa[0].sin6_port = uh->uh_sport; init_sin6(&fromsa[1], m, 1); fromsa[1].sin6_port = uh->uh_dport; pcbinfo = udp_get_inpcbinfo(nxt); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { struct inpcb *last; struct inpcbhead *pcblist; struct ip6_moptions *imo; INP_INFO_RLOCK(pcbinfo); /* * In the event that laddr should be set to the link-local * address (this happens in RIPng), the multicast address * specified in the received packet will not match laddr. To * handle this situation, matching is relaxed if the * receiving interface is the same as one specified in the * socket and if the destination multicast address matches * one of the multicast groups specified in the socket. */ /* * KAME note: traditionally we dropped udpiphdr from mbuf * here. We need udphdr for IPsec processing so we do that * later. */ pcblist = udp_get_pcblist(nxt); last = NULL; LIST_FOREACH(inp, pcblist, inp_list) { if ((inp->inp_vflag & INP_IPV6) == 0) continue; if (inp->inp_lport != uh->uh_dport) continue; if (inp->inp_fport != 0 && inp->inp_fport != uh->uh_sport) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) { if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst)) continue; } if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src) || inp->inp_fport != uh->uh_sport) continue; } /* * XXXRW: Because we weren't holding either the inpcb * or the hash lock when we checked for a match * before, we should probably recheck now that the * inpcb lock is (supposed to be) held. */ /* * Handle socket delivery policy for any-source * and source-specific multicast. [RFC3678] */ imo = inp->in6p_moptions; if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { struct sockaddr_in6 mcaddr; int blocked; INP_RLOCK(inp); bzero(&mcaddr, sizeof(struct sockaddr_in6)); mcaddr.sin6_len = sizeof(struct sockaddr_in6); mcaddr.sin6_family = AF_INET6; mcaddr.sin6_addr = ip6->ip6_dst; blocked = im6o_mc_filter(imo, ifp, (struct sockaddr *)&mcaddr, (struct sockaddr *)&fromsa[0]); if (blocked != MCAST_PASS) { if (blocked == MCAST_NOTGMEMBER) IP6STAT_INC(ip6s_notmember); if (blocked == MCAST_NOTSMEMBER || blocked == MCAST_MUTED) UDPSTAT_INC(udps_filtermcast); INP_RUNLOCK(inp); /* XXX */ continue; } INP_RUNLOCK(inp); } if (last != NULL) { struct mbuf *n; if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) { INP_RLOCK(last); UDP_PROBE(receive, NULL, last, ip6, last, uh); if (udp6_append(last, n, off, fromsa)) goto inp_lost; INP_RUNLOCK(last); } } last = inp; /* * Don't look for additional matches if this one does * not have either the SO_REUSEPORT or SO_REUSEADDR * socket options set. This heuristic avoids * searching through all pcbs in the common case of a * non-shared port. It assumes that an application * will never clear these options after setting them. */ if ((last->inp_socket->so_options & - (SO_REUSEPORT|SO_REUSEADDR)) == 0) + (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) break; } if (last == NULL) { /* * No matching pcb found; discard datagram. (No need * to send an ICMP Port Unreachable for a broadcast * or multicast datgram.) */ UDPSTAT_INC(udps_noport); UDPSTAT_INC(udps_noportmcast); goto badheadlocked; } INP_RLOCK(last); INP_INFO_RUNLOCK(pcbinfo); UDP_PROBE(receive, NULL, last, ip6, last, uh); if (udp6_append(last, m, off, fromsa) == 0) INP_RUNLOCK(last); inp_lost: return (IPPROTO_DONE); } /* * Locate pcb for datagram. */ /* * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ if ((m->m_flags & M_IP6_NEXTHOP) && (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { struct sockaddr_in6 *next_hop6; next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1); /* * Transparently forwarded. Pretend to be the destination. * Already got one like this? */ inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src, uh->uh_sport, &ip6->ip6_dst, uh->uh_dport, INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m); if (!inp) { /* * It's new. Try to find the ambushing socket. * Because we've rewritten the destination address, * any hardware-generated hash is ignored. */ inp = in6_pcblookup(pcbinfo, &ip6->ip6_src, uh->uh_sport, &next_hop6->sin6_addr, next_hop6->sin6_port ? htons(next_hop6->sin6_port) : uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif); } /* Remove the tag from the packet. We don't need it anymore. */ m_tag_delete(m, fwd_tag); m->m_flags &= ~M_IP6_NEXTHOP; } else inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src, uh->uh_sport, &ip6->ip6_dst, uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m); if (inp == NULL) { if (udp_log_in_vain) { char ip6bufs[INET6_ADDRSTRLEN]; char ip6bufd[INET6_ADDRSTRLEN]; log(LOG_INFO, "Connection attempt to UDP [%s]:%d from [%s]:%d\n", ip6_sprintf(ip6bufd, &ip6->ip6_dst), ntohs(uh->uh_dport), ip6_sprintf(ip6bufs, &ip6->ip6_src), ntohs(uh->uh_sport)); } UDPSTAT_INC(udps_noport); if (m->m_flags & M_MCAST) { printf("UDP6: M_MCAST is set in a unicast packet.\n"); UDPSTAT_INC(udps_noportmcast); goto badunlocked; } if (V_udp_blackhole) goto badunlocked; icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); return (IPPROTO_DONE); } INP_RLOCK_ASSERT(inp); up = intoudpcb(inp); if (cscov_partial) { if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) { INP_RUNLOCK(inp); m_freem(m); return (IPPROTO_DONE); } } UDP_PROBE(receive, NULL, inp, ip6, inp, uh); if (udp6_append(inp, m, off, fromsa) == 0) INP_RUNLOCK(inp); return (IPPROTO_DONE); badheadlocked: INP_INFO_RUNLOCK(pcbinfo); badunlocked: if (m) m_freem(m); return (IPPROTO_DONE); } static void udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d, struct inpcbinfo *pcbinfo) { struct udphdr uh; struct ip6_hdr *ip6; struct mbuf *m; int off = 0; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; void *cmdarg; struct inpcb *(*notify)(struct inpcb *, int) = udp_notify; struct udp_portonly { u_int16_t uh_sport; u_int16_t uh_dport; } *uhp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if ((unsigned)cmd >= PRC_NCMDS) return; if (PRC_IS_REDIRECT(cmd)) notify = in6_rtchange, d = NULL; else if (cmd == PRC_HOSTDEAD) d = NULL; else if (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; cmdarg = ip6cp->ip6c_cmdarg; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; cmdarg = NULL; sa6_src = &sa6_any; } if (ip6) { /* * 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(*uhp)) return; bzero(&uh, sizeof(uh)); m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh); if (!PRC_IS_REDIRECT(cmd)) { /* Check to see if its tunneled */ struct inpcb *inp; inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst, uh.uh_dport, &ip6->ip6_src, uh.uh_sport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m); if (inp != NULL) { struct udpcb *up; up = intoudpcb(inp); if (up->u_icmp_func) { /* Yes it is. */ INP_RUNLOCK(inp); (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src, d, up->u_tun_ctx); return; } else { /* Can't find it. */ INP_RUNLOCK(inp); } } } (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport, (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd, cmdarg, notify); } else (void)in6_pcbnotify(pcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify); } void udp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo)); } void udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d) { return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo)); } static int udp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); if (req->newlen != sizeof(addrs)) return (EINVAL); if (req->oldlen != sizeof(struct xucred)) return (EINVAL); 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); } inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { INP_RLOCK_ASSERT(inp); if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); 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_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0, 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection"); static int udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6, struct mbuf *control, struct thread *td) { u_int32_t ulen = m->m_pkthdr.len; u_int32_t plen = sizeof(struct udphdr) + ulen; struct ip6_hdr *ip6; struct udphdr *udp6; struct in6_addr *laddr, *faddr, in6a; struct sockaddr_in6 *sin6 = NULL; int cscov_partial = 0; int scope_ambiguous = 0; u_short fport; int error = 0; uint8_t nxt; uint16_t cscov = 0; struct ip6_pktopts *optp, opt; int af = AF_INET6, hlen = sizeof(struct ip6_hdr); int flags; struct sockaddr_in6 tmp; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); if (addr6) { /* addr6 has been validated in udp6_send(). */ sin6 = (struct sockaddr_in6 *)addr6; /* protect *sin6 from overwrites */ tmp = *sin6; sin6 = &tmp; /* * Application should provide a proper zone ID or the use of * default zone IDs should be enabled. Unfortunately, some * applications do not behave as it should, so we need a * workaround. Even if an appropriate ID is not determined, * we'll see if we can determine the outgoing interface. If we * can, determine the zone ID based on the interface below. */ if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone) scope_ambiguous = 1; if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) return (error); } nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE; if (control) { if ((error = ip6_setpktopts(control, &opt, inp->in6p_outputopts, td->td_ucred, nxt)) != 0) goto release; optp = &opt; } else optp = inp->in6p_outputopts; if (sin6) { faddr = &sin6->sin6_addr; /* * Since we saw no essential reason for calling in_pcbconnect, * we get rid of such kind of logic, and call in6_selectsrc * and in6_pcbsetport in order to fill in the local address * and the local port. */ if (sin6->sin6_port == 0) { error = EADDRNOTAVAIL; goto release; } if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { /* how about ::ffff:0.0.0.0 case? */ error = EISCONN; goto release; } fport = sin6->sin6_port; /* allow 0 port */ if (IN6_IS_ADDR_V4MAPPED(faddr)) { if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { /* * I believe we should explicitly discard the * packet when mapped addresses are disabled, * rather than send the packet as an IPv6 one. * If we chose the latter approach, the packet * might be sent out on the wire based on the * default route, the situation which we'd * probably want to avoid. * (20010421 jinmei@kame.net) */ error = EINVAL; goto release; } if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) { /* * when remote addr is an IPv4-mapped address, * local addr should not be an IPv6 address, * since you cannot determine how to map IPv6 * source address to IPv4. */ error = EINVAL; goto release; } af = AF_INET; } if (!IN6_IS_ADDR_V4MAPPED(faddr)) { error = in6_selectsrc_socket(sin6, optp, inp, td->td_ucred, scope_ambiguous, &in6a, NULL); if (error) goto release; laddr = &in6a; } else laddr = &inp->in6p_laddr; /* XXX */ if (laddr == NULL) { if (error == 0) error = EADDRNOTAVAIL; goto release; } if (inp->inp_lport == 0 && (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) { /* Undo an address bind that may have occurred. */ inp->in6p_laddr = in6addr_any; goto release; } } else { if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { error = ENOTCONN; goto release; } if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) { if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) { /* * XXX: this case would happen when the * application sets the V6ONLY flag after * connecting the foreign address. * Such applications should be fixed, * so we bark here. */ log(LOG_INFO, "udp6_output: IPV6_V6ONLY " "option was set for a connected socket\n"); error = EINVAL; goto release; } else af = AF_INET; } laddr = &inp->in6p_laddr; faddr = &inp->in6p_faddr; fport = inp->inp_fport; } if (af == AF_INET) hlen = sizeof(struct ip); /* * Calculate data length and get a mbuf * for UDP and IP6 headers. */ M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT); if (m == NULL) { error = ENOBUFS; goto release; } /* * Stuff checksum and output datagram. */ udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen); udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */ udp6->uh_dport = fport; if (nxt == IPPROTO_UDPLITE) { struct udpcb *up; up = intoudpcb(inp); cscov = up->u_txcslen; if (cscov >= plen) cscov = 0; udp6->uh_ulen = htons(cscov); /* * For UDP-Lite, checksum coverage length of zero means * the entire UDPLite packet is covered by the checksum. */ cscov_partial = (cscov == 0) ? 0 : 1; } else if (plen <= 0xffff) udp6->uh_ulen = htons((u_short)plen); else udp6->uh_ulen = 0; udp6->uh_sum = 0; switch (af) { case AF_INET6: ip6 = mtod(m, struct ip6_hdr *); ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK; ip6->ip6_vfc &= ~IPV6_VERSION_MASK; ip6->ip6_vfc |= IPV6_VERSION; ip6->ip6_plen = htons((u_short)plen); ip6->ip6_nxt = nxt; ip6->ip6_hlim = in6_selecthlim(inp, NULL); ip6->ip6_src = *laddr; ip6->ip6_dst = *faddr; if (cscov_partial) { if ((udp6->uh_sum = in6_cksum_partial(m, nxt, sizeof(struct ip6_hdr), plen, cscov)) == 0) udp6->uh_sum = 0xffff; } else { udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0); m->m_pkthdr.csum_flags = CSUM_UDP_IPV6; m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); } #ifdef RSS { uint32_t hash_val, hash_type; uint8_t pr; pr = inp->inp_socket->so_proto->pr_protocol; /* * Calculate an appropriate RSS hash for UDP and * UDP Lite. * * The called function will take care of figuring out * whether a 2-tuple or 4-tuple hash is required based * on the currently configured scheme. * * Later later on connected socket values should be * cached in the inpcb and reused, rather than constantly * re-calculating it. * * UDP Lite is a different protocol number and will * likely end up being hashed as a 2-tuple until * RSS / NICs grow UDP Lite protocol awareness. */ if (rss_proto_software_hash_v6(faddr, laddr, fport, inp->inp_lport, pr, &hash_val, &hash_type) == 0) { m->m_pkthdr.flowid = hash_val; M_HASHTYPE_SET(m, hash_type); } } #endif flags = 0; #ifdef RSS /* * Don't override with the inp cached flowid. * * Until the whole UDP path is vetted, it may actually * be incorrect. */ flags |= IP_NODEFAULTFLOWID; #endif UDP_PROBE(send, NULL, inp, ip6, inp, udp6); UDPSTAT_INC(udps_opackets); error = ip6_output(m, optp, &inp->inp_route6, flags, inp->in6p_moptions, NULL, inp); break; case AF_INET: error = EAFNOSUPPORT; goto release; } goto releaseopt; release: m_freem(m); releaseopt: if (control) { ip6_clearpktopts(&opt, -1); m_freem(control); } return (error); } static void udp6_abort(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_abort: inp == NULL")); INP_WLOCK(inp); #ifdef INET if (inp->inp_vflag & INP_IPV4) { struct pr_usrreqs *pru; uint8_t nxt; nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE; INP_WUNLOCK(inp); pru = inetsw[ip_protox[nxt]].pr_usrreqs; (*pru->pru_abort)(so); return; } #endif if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { INP_HASH_WLOCK(pcbinfo); in6_pcbdisconnect(inp); inp->in6p_laddr = in6addr_any; INP_HASH_WUNLOCK(pcbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); } static int udp6_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp == NULL, ("udp6_attach: inp != NULL")); if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { error = soreserve(so, udp_sendspace, udp_recvspace); if (error) return (error); } INP_INFO_WLOCK(pcbinfo); error = in_pcballoc(so, pcbinfo); if (error) { INP_INFO_WUNLOCK(pcbinfo); return (error); } inp = (struct inpcb *)so->so_pcb; inp->inp_vflag |= INP_IPV6; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) inp->inp_vflag |= INP_IPV4; inp->in6p_hops = -1; /* use kernel default */ inp->in6p_cksum = -1; /* just to be sure */ /* * XXX: ugly!! * 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; error = udp_newudpcb(inp); if (error) { in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(pcbinfo); return (error); } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(pcbinfo); return (0); } static int udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_bind: inp == NULL")); INP_WLOCK(inp); INP_HASH_WLOCK(pcbinfo); inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { struct sockaddr_in6 *sin6_p; sin6_p = (struct sockaddr_in6 *)nam; if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr)) inp->inp_vflag |= INP_IPV4; #ifdef INET else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6_p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = in_pcbbind(inp, (struct sockaddr *)&sin, td->td_ucred); goto out; } #endif } error = in6_pcbbind(inp, nam, td->td_ucred); #ifdef INET out: #endif INP_HASH_WUNLOCK(pcbinfo); INP_WUNLOCK(inp); return (error); } static void udp6_close(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_close: inp == NULL")); INP_WLOCK(inp); #ifdef INET if (inp->inp_vflag & INP_IPV4) { struct pr_usrreqs *pru; uint8_t nxt; nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE; INP_WUNLOCK(inp); pru = inetsw[ip_protox[nxt]].pr_usrreqs; (*pru->pru_disconnect)(so); return; } #endif if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { INP_HASH_WLOCK(pcbinfo); in6_pcbdisconnect(inp); inp->in6p_laddr = in6addr_any; INP_HASH_WUNLOCK(pcbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); } static int udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; struct sockaddr_in6 *sin6; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); sin6 = (struct sockaddr_in6 *)nam; KASSERT(inp != NULL, ("udp6_connect: inp == NULL")); /* * XXXRW: Need to clarify locking of v4/v6 flags. */ INP_WLOCK(inp); #ifdef INET if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { struct sockaddr_in sin; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { error = EINVAL; goto out; } if ((inp->inp_vflag & INP_IPV4) == 0) { error = EAFNOSUPPORT; goto out; } if (inp->inp_faddr.s_addr != INADDR_ANY) { error = EISCONN; goto out; } in6_sin6_2_sin(&sin, sin6); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = prison_remote_ip4(td->td_ucred, &sin.sin_addr); if (error != 0) goto out; INP_HASH_WLOCK(pcbinfo); error = in_pcbconnect(inp, (struct sockaddr *)&sin, td->td_ucred); INP_HASH_WUNLOCK(pcbinfo); if (error == 0) soisconnected(so); goto out; } else { if ((inp->inp_vflag & INP_IPV6) == 0) { error = EAFNOSUPPORT; goto out; } } #endif if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { error = EISCONN; goto out; } inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr); if (error != 0) goto out; INP_HASH_WLOCK(pcbinfo); error = in6_pcbconnect(inp, nam, td->td_ucred); INP_HASH_WUNLOCK(pcbinfo); if (error == 0) soisconnected(so); out: INP_WUNLOCK(inp); return (error); } static void udp6_detach(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; struct udpcb *up; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_detach: inp == NULL")); INP_INFO_WLOCK(pcbinfo); INP_WLOCK(inp); up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(pcbinfo); udp_discardcb(up); } static int udp6_disconnect(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL")); INP_WLOCK(inp); #ifdef INET if (inp->inp_vflag & INP_IPV4) { struct pr_usrreqs *pru; uint8_t nxt; nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE; INP_WUNLOCK(inp); pru = inetsw[ip_protox[nxt]].pr_usrreqs; (void)(*pru->pru_disconnect)(so); return (0); } #endif if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { INP_WUNLOCK(inp); return (ENOTCONN); } INP_HASH_WLOCK(pcbinfo); in6_pcbdisconnect(inp); inp->in6p_laddr = in6addr_any; INP_HASH_WUNLOCK(pcbinfo); SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; /* XXX */ SOCK_UNLOCK(so); INP_WUNLOCK(inp); return (0); } static int udp6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; int error = 0; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp6_send: inp == NULL")); INP_WLOCK(inp); if (addr) { if (addr->sa_len != sizeof(struct sockaddr_in6)) { error = EINVAL; goto bad; } if (addr->sa_family != AF_INET6) { error = EAFNOSUPPORT; goto bad; } } #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { int hasv4addr; struct sockaddr_in6 *sin6 = NULL; if (addr == NULL) hasv4addr = (inp->inp_vflag & INP_IPV4); else { sin6 = (struct sockaddr_in6 *)addr; hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr) ? 1 : 0; } if (hasv4addr) { struct pr_usrreqs *pru; uint8_t nxt; nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE; /* * XXXRW: We release UDP-layer locks before calling * udp_send() in order to avoid recursion. However, * this does mean there is a short window where inp's * fields are unstable. Could this lead to a * potential race in which the factors causing us to * select the UDPv4 output routine are invalidated? */ INP_WUNLOCK(inp); if (sin6) in6_sin6_2_sin_in_sock(addr); pru = inetsw[ip_protox[nxt]].pr_usrreqs; /* addr will just be freed in sendit(). */ return ((*pru->pru_send)(so, flags, m, addr, control, td)); } } #endif #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif INP_HASH_WLOCK(pcbinfo); error = udp6_output(inp, m, addr, control, td); INP_HASH_WUNLOCK(pcbinfo); INP_WUNLOCK(inp); return (error); bad: INP_WUNLOCK(inp); m_freem(m); return (error); } struct pr_usrreqs udp6_usrreqs = { .pru_abort = udp6_abort, .pru_attach = udp6_attach, .pru_bind = udp6_bind, .pru_connect = udp6_connect, .pru_control = in6_control, .pru_detach = udp6_detach, .pru_disconnect = udp6_disconnect, .pru_peeraddr = in6_mapped_peeraddr, .pru_send = udp6_send, .pru_shutdown = udp_shutdown, .pru_sockaddr = in6_mapped_sockaddr, .pru_soreceive = soreceive_dgram, .pru_sosend = sosend_dgram, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = udp6_close }; Index: head/sys/sys/socket.h =================================================================== --- head/sys/sys/socket.h (revision 334718) +++ head/sys/sys/socket.h (revision 334719) @@ -1,729 +1,730 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1985, 1986, 1988, 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. * 3. 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. * * @(#)socket.h 8.4 (Berkeley) 2/21/94 * $FreeBSD$ */ #ifndef _SYS_SOCKET_H_ #define _SYS_SOCKET_H_ #include #include #include #include /* * Definitions related to sockets: types, address families, options. */ /* * Data types. */ #if __BSD_VISIBLE #ifndef _GID_T_DECLARED typedef __gid_t gid_t; #define _GID_T_DECLARED #endif #ifndef _OFF_T_DECLARED typedef __off_t off_t; #define _OFF_T_DECLARED #endif #ifndef _PID_T_DECLARED typedef __pid_t pid_t; #define _PID_T_DECLARED #endif #endif #ifndef _SA_FAMILY_T_DECLARED typedef __sa_family_t sa_family_t; #define _SA_FAMILY_T_DECLARED #endif #ifndef _SOCKLEN_T_DECLARED typedef __socklen_t socklen_t; #define _SOCKLEN_T_DECLARED #endif #ifndef _SSIZE_T_DECLARED typedef __ssize_t ssize_t; #define _SSIZE_T_DECLARED #endif #if __BSD_VISIBLE #ifndef _UID_T_DECLARED typedef __uid_t uid_t; #define _UID_T_DECLARED #endif #endif #ifndef _UINT32_T_DECLARED typedef __uint32_t uint32_t; #define _UINT32_T_DECLARED #endif #ifndef _UINTPTR_T_DECLARED typedef __uintptr_t uintptr_t; #define _UINTPTR_T_DECLARED #endif /* * Types */ #define SOCK_STREAM 1 /* stream socket */ #define SOCK_DGRAM 2 /* datagram socket */ #define SOCK_RAW 3 /* raw-protocol interface */ #if __BSD_VISIBLE #define SOCK_RDM 4 /* reliably-delivered message */ #endif #define SOCK_SEQPACKET 5 /* sequenced packet stream */ #if __BSD_VISIBLE /* * Creation flags, OR'ed into socket() and socketpair() type argument. */ #define SOCK_CLOEXEC 0x10000000 #define SOCK_NONBLOCK 0x20000000 #ifdef _KERNEL /* * Flags for accept1(), kern_accept4() and solisten_dequeue, in addition * to SOCK_CLOEXEC and SOCK_NONBLOCK. */ #define ACCEPT4_INHERIT 0x1 #define ACCEPT4_COMPAT 0x2 #endif /* _KERNEL */ #endif /* __BSD_VISIBLE */ /* * Option flags per-socket. */ -#define SO_DEBUG 0x0001 /* turn on debugging info recording */ -#define SO_ACCEPTCONN 0x0002 /* socket has had listen() */ -#define SO_REUSEADDR 0x0004 /* allow local address reuse */ -#define SO_KEEPALIVE 0x0008 /* keep connections alive */ -#define SO_DONTROUTE 0x0010 /* just use interface addresses */ -#define SO_BROADCAST 0x0020 /* permit sending of broadcast msgs */ +#define SO_DEBUG 0x00000001 /* turn on debugging info recording */ +#define SO_ACCEPTCONN 0x00000002 /* socket has had listen() */ +#define SO_REUSEADDR 0x00000004 /* allow local address reuse */ +#define SO_KEEPALIVE 0x00000008 /* keep connections alive */ +#define SO_DONTROUTE 0x00000010 /* just use interface addresses */ +#define SO_BROADCAST 0x00000020 /* permit sending of broadcast msgs */ #if __BSD_VISIBLE -#define SO_USELOOPBACK 0x0040 /* bypass hardware when possible */ +#define SO_USELOOPBACK 0x00000040 /* bypass hardware when possible */ #endif -#define SO_LINGER 0x0080 /* linger on close if data present */ -#define SO_OOBINLINE 0x0100 /* leave received OOB data in line */ +#define SO_LINGER 0x00000080 /* linger on close if data present */ +#define SO_OOBINLINE 0x00000100 /* leave received OOB data in line */ #if __BSD_VISIBLE -#define SO_REUSEPORT 0x0200 /* allow local address & port reuse */ -#define SO_TIMESTAMP 0x0400 /* timestamp received dgram traffic */ -#define SO_NOSIGPIPE 0x0800 /* no SIGPIPE from EPIPE */ -#define SO_ACCEPTFILTER 0x1000 /* there is an accept filter */ -#define SO_BINTIME 0x2000 /* timestamp received dgram traffic */ +#define SO_REUSEPORT 0x00000200 /* allow local address & port reuse */ +#define SO_TIMESTAMP 0x00000400 /* timestamp received dgram traffic */ +#define SO_NOSIGPIPE 0x00000800 /* no SIGPIPE from EPIPE */ +#define SO_ACCEPTFILTER 0x00001000 /* there is an accept filter */ +#define SO_BINTIME 0x00002000 /* timestamp received dgram traffic */ #endif -#define SO_NO_OFFLOAD 0x4000 /* socket cannot be offloaded */ -#define SO_NO_DDP 0x8000 /* disable direct data placement */ +#define SO_NO_OFFLOAD 0x00004000 /* socket cannot be offloaded */ +#define SO_NO_DDP 0x00008000 /* disable direct data placement */ +#define SO_REUSEPORT_LB 0x00010000 /* reuse with load balancing */ /* * Additional options, not kept in so_options. */ #define SO_SNDBUF 0x1001 /* send buffer size */ #define SO_RCVBUF 0x1002 /* receive buffer size */ #define SO_SNDLOWAT 0x1003 /* send low-water mark */ #define SO_RCVLOWAT 0x1004 /* receive low-water mark */ #define SO_SNDTIMEO 0x1005 /* send timeout */ #define SO_RCVTIMEO 0x1006 /* receive timeout */ #define SO_ERROR 0x1007 /* get error status and clear */ #define SO_TYPE 0x1008 /* get socket type */ #if __BSD_VISIBLE #define SO_LABEL 0x1009 /* socket's MAC label */ #define SO_PEERLABEL 0x1010 /* socket's peer's MAC label */ #define SO_LISTENQLIMIT 0x1011 /* socket's backlog limit */ #define SO_LISTENQLEN 0x1012 /* socket's complete queue length */ #define SO_LISTENINCQLEN 0x1013 /* socket's incomplete queue length */ #define SO_SETFIB 0x1014 /* use this FIB to route */ #define SO_USER_COOKIE 0x1015 /* user cookie (dummynet etc.) */ #define SO_PROTOCOL 0x1016 /* get socket protocol (Linux name) */ #define SO_PROTOTYPE SO_PROTOCOL /* alias for SO_PROTOCOL (SunOS name) */ #define SO_TS_CLOCK 0x1017 /* clock type used for SO_TIMESTAMP */ #define SO_MAX_PACING_RATE 0x1018 /* socket's max TX pacing rate (Linux name) */ #endif #if __BSD_VISIBLE #define SO_TS_REALTIME_MICRO 0 /* microsecond resolution, realtime */ #define SO_TS_BINTIME 1 /* sub-nanosecond resolution, realtime */ #define SO_TS_REALTIME 2 /* nanosecond resolution, realtime */ #define SO_TS_MONOTONIC 3 /* nanosecond resolution, monotonic */ #define SO_TS_DEFAULT SO_TS_REALTIME_MICRO #define SO_TS_CLOCK_MAX SO_TS_MONOTONIC #endif /* * Space reserved for new socket options added by third-party vendors. * This range applies to all socket option levels. New socket options * in FreeBSD should always use an option value less than SO_VENDOR. */ #if __BSD_VISIBLE #define SO_VENDOR 0x80000000 #endif /* * Structure used for manipulating linger option. */ struct linger { int l_onoff; /* option on/off */ int l_linger; /* linger time */ }; #if __BSD_VISIBLE struct accept_filter_arg { char af_name[16]; char af_arg[256-16]; }; #endif /* * Level number for (get/set)sockopt() to apply to socket itself. */ #define SOL_SOCKET 0xffff /* options for socket level */ /* * Address families. */ #define AF_UNSPEC 0 /* unspecified */ #if __BSD_VISIBLE #define AF_LOCAL AF_UNIX /* local to host (pipes, portals) */ #endif #define AF_UNIX 1 /* standardized name for AF_LOCAL */ #define AF_INET 2 /* internetwork: UDP, TCP, etc. */ #if __BSD_VISIBLE #define AF_IMPLINK 3 /* arpanet imp addresses */ #define AF_PUP 4 /* pup protocols: e.g. BSP */ #define AF_CHAOS 5 /* mit CHAOS protocols */ #define AF_NETBIOS 6 /* SMB protocols */ #define AF_ISO 7 /* ISO protocols */ #define AF_OSI AF_ISO #define AF_ECMA 8 /* European computer manufacturers */ #define AF_DATAKIT 9 /* datakit protocols */ #define AF_CCITT 10 /* CCITT protocols, X.25 etc */ #define AF_SNA 11 /* IBM SNA */ #define AF_DECnet 12 /* DECnet */ #define AF_DLI 13 /* DEC Direct data link interface */ #define AF_LAT 14 /* LAT */ #define AF_HYLINK 15 /* NSC Hyperchannel */ #define AF_APPLETALK 16 /* Apple Talk */ #define AF_ROUTE 17 /* Internal Routing Protocol */ #define AF_LINK 18 /* Link layer interface */ #define pseudo_AF_XTP 19 /* eXpress Transfer Protocol (no AF) */ #define AF_COIP 20 /* connection-oriented IP, aka ST II */ #define AF_CNT 21 /* Computer Network Technology */ #define pseudo_AF_RTIP 22 /* Help Identify RTIP packets */ #define AF_IPX 23 /* Novell Internet Protocol */ #define AF_SIP 24 /* Simple Internet Protocol */ #define pseudo_AF_PIP 25 /* Help Identify PIP packets */ #define AF_ISDN 26 /* Integrated Services Digital Network*/ #define AF_E164 AF_ISDN /* CCITT E.164 recommendation */ #define pseudo_AF_KEY 27 /* Internal key-management function */ #endif #define AF_INET6 28 /* IPv6 */ #if __BSD_VISIBLE #define AF_NATM 29 /* native ATM access */ #define AF_ATM 30 /* ATM */ #define pseudo_AF_HDRCMPLT 31 /* Used by BPF to not rewrite headers * in interface output routine */ #define AF_NETGRAPH 32 /* Netgraph sockets */ #define AF_SLOW 33 /* 802.3ad slow protocol */ #define AF_SCLUSTER 34 /* Sitara cluster protocol */ #define AF_ARP 35 #define AF_BLUETOOTH 36 /* Bluetooth sockets */ #define AF_IEEE80211 37 /* IEEE 802.11 protocol */ #define AF_INET_SDP 40 /* OFED Socket Direct Protocol ipv4 */ #define AF_INET6_SDP 42 /* OFED Socket Direct Protocol ipv6 */ #define AF_MAX 42 /* * When allocating a new AF_ constant, please only allocate * even numbered constants for FreeBSD until 134 as odd numbered AF_ * constants 39-133 are now reserved for vendors. */ #define AF_VENDOR00 39 #define AF_VENDOR01 41 #define AF_VENDOR02 43 #define AF_VENDOR03 45 #define AF_VENDOR04 47 #define AF_VENDOR05 49 #define AF_VENDOR06 51 #define AF_VENDOR07 53 #define AF_VENDOR08 55 #define AF_VENDOR09 57 #define AF_VENDOR10 59 #define AF_VENDOR11 61 #define AF_VENDOR12 63 #define AF_VENDOR13 65 #define AF_VENDOR14 67 #define AF_VENDOR15 69 #define AF_VENDOR16 71 #define AF_VENDOR17 73 #define AF_VENDOR18 75 #define AF_VENDOR19 77 #define AF_VENDOR20 79 #define AF_VENDOR21 81 #define AF_VENDOR22 83 #define AF_VENDOR23 85 #define AF_VENDOR24 87 #define AF_VENDOR25 89 #define AF_VENDOR26 91 #define AF_VENDOR27 93 #define AF_VENDOR28 95 #define AF_VENDOR29 97 #define AF_VENDOR30 99 #define AF_VENDOR31 101 #define AF_VENDOR32 103 #define AF_VENDOR33 105 #define AF_VENDOR34 107 #define AF_VENDOR35 109 #define AF_VENDOR36 111 #define AF_VENDOR37 113 #define AF_VENDOR38 115 #define AF_VENDOR39 117 #define AF_VENDOR40 119 #define AF_VENDOR41 121 #define AF_VENDOR42 123 #define AF_VENDOR43 125 #define AF_VENDOR44 127 #define AF_VENDOR45 129 #define AF_VENDOR46 131 #define AF_VENDOR47 133 #endif /* * Structure used by kernel to store most * addresses. */ struct sockaddr { unsigned char sa_len; /* total length */ sa_family_t sa_family; /* address family */ char sa_data[14]; /* actually longer; address value */ }; #if __BSD_VISIBLE #define SOCK_MAXADDRLEN 255 /* longest possible addresses */ /* * Structure used by kernel to pass protocol * information in raw sockets. */ struct sockproto { unsigned short sp_family; /* address family */ unsigned short sp_protocol; /* protocol */ }; #endif #include #if __BSD_VISIBLE /* * Protocol families, same as address families for now. */ #define PF_UNSPEC AF_UNSPEC #define PF_LOCAL AF_LOCAL #define PF_UNIX PF_LOCAL /* backward compatibility */ #define PF_INET AF_INET #define PF_IMPLINK AF_IMPLINK #define PF_PUP AF_PUP #define PF_CHAOS AF_CHAOS #define PF_NETBIOS AF_NETBIOS #define PF_ISO AF_ISO #define PF_OSI AF_ISO #define PF_ECMA AF_ECMA #define PF_DATAKIT AF_DATAKIT #define PF_CCITT AF_CCITT #define PF_SNA AF_SNA #define PF_DECnet AF_DECnet #define PF_DLI AF_DLI #define PF_LAT AF_LAT #define PF_HYLINK AF_HYLINK #define PF_APPLETALK AF_APPLETALK #define PF_ROUTE AF_ROUTE #define PF_LINK AF_LINK #define PF_XTP pseudo_AF_XTP /* really just proto family, no AF */ #define PF_COIP AF_COIP #define PF_CNT AF_CNT #define PF_SIP AF_SIP #define PF_IPX AF_IPX #define PF_RTIP pseudo_AF_RTIP /* same format as AF_INET */ #define PF_PIP pseudo_AF_PIP #define PF_ISDN AF_ISDN #define PF_KEY pseudo_AF_KEY #define PF_INET6 AF_INET6 #define PF_NATM AF_NATM #define PF_ATM AF_ATM #define PF_NETGRAPH AF_NETGRAPH #define PF_SLOW AF_SLOW #define PF_SCLUSTER AF_SCLUSTER #define PF_ARP AF_ARP #define PF_BLUETOOTH AF_BLUETOOTH #define PF_IEEE80211 AF_IEEE80211 #define PF_INET_SDP AF_INET_SDP #define PF_INET6_SDP AF_INET6_SDP #define PF_MAX AF_MAX /* * Definitions for network related sysctl, CTL_NET. * * Second level is protocol family. * Third level is protocol number. * * Further levels are defined by the individual families. */ /* * PF_ROUTE - Routing table * * Three additional levels are defined: * Fourth: address family, 0 is wildcard * Fifth: type of info, defined below * Sixth: flag(s) to mask with for NET_RT_FLAGS */ #define NET_RT_DUMP 1 /* dump; may limit to a.f. */ #define NET_RT_FLAGS 2 /* by flags, e.g. RESOLVING */ #define NET_RT_IFLIST 3 /* survey interface list */ #define NET_RT_IFMALIST 4 /* return multicast address list */ #define NET_RT_IFLISTL 5 /* Survey interface list, using 'l'en * versions of msghdr structs. */ #endif /* __BSD_VISIBLE */ /* * Maximum queue length specifiable by listen. */ #define SOMAXCONN 128 /* * Message header for recvmsg and sendmsg calls. * Used value-result for recvmsg, value only for sendmsg. */ struct msghdr { void *msg_name; /* optional address */ socklen_t msg_namelen; /* size of address */ struct iovec *msg_iov; /* scatter/gather array */ int msg_iovlen; /* # elements in msg_iov */ void *msg_control; /* ancillary data, see below */ socklen_t msg_controllen; /* ancillary data buffer len */ int msg_flags; /* flags on received message */ }; #define MSG_OOB 0x00000001 /* process out-of-band data */ #define MSG_PEEK 0x00000002 /* peek at incoming message */ #define MSG_DONTROUTE 0x00000004 /* send without using routing tables */ #define MSG_EOR 0x00000008 /* data completes record */ #define MSG_TRUNC 0x00000010 /* data discarded before delivery */ #define MSG_CTRUNC 0x00000020 /* control data lost before delivery */ #define MSG_WAITALL 0x00000040 /* wait for full request or error */ #if __BSD_VISIBLE #define MSG_DONTWAIT 0x00000080 /* this message should be nonblocking */ #define MSG_EOF 0x00000100 /* data completes connection */ /* 0x00000200 unused */ /* 0x00000400 unused */ /* 0x00000800 unused */ /* 0x00001000 unused */ #define MSG_NOTIFICATION 0x00002000 /* SCTP notification */ #define MSG_NBIO 0x00004000 /* FIONBIO mode, used by fifofs */ #define MSG_COMPAT 0x00008000 /* used in sendit() */ #endif #ifdef _KERNEL #define MSG_SOCALLBCK 0x00010000 /* for use by socket callbacks - soreceive (TCP) */ #endif #if __POSIX_VISIBLE >= 200809 #define MSG_NOSIGNAL 0x00020000 /* do not generate SIGPIPE on EOF */ #endif #if __BSD_VISIBLE #define MSG_CMSG_CLOEXEC 0x00040000 /* make received fds close-on-exec */ #define MSG_WAITFORONE 0x00080000 /* for recvmmsg() */ #endif #ifdef _KERNEL #define MSG_MORETOCOME 0x00100000 /* additional data pending */ #endif /* * Header for ancillary data objects in msg_control buffer. * Used for additional information with/about a datagram * not expressible by flags. The format is a sequence * of message elements headed by cmsghdr structures. */ struct cmsghdr { socklen_t cmsg_len; /* data byte count, including hdr */ int cmsg_level; /* originating protocol */ int cmsg_type; /* protocol-specific type */ /* followed by u_char cmsg_data[]; */ }; #if __BSD_VISIBLE /* * While we may have more groups than this, the cmsgcred struct must * be able to fit in an mbuf and we have historically supported a * maximum of 16 groups. */ #define CMGROUP_MAX 16 /* * Credentials structure, used to verify the identity of a peer * process that has sent us a message. This is allocated by the * peer process but filled in by the kernel. This prevents the * peer from lying about its identity. (Note that cmcred_groups[0] * is the effective GID.) */ struct cmsgcred { pid_t cmcred_pid; /* PID of sending process */ uid_t cmcred_uid; /* real UID of sending process */ uid_t cmcred_euid; /* effective UID of sending process */ gid_t cmcred_gid; /* real GID of sending process */ short cmcred_ngroups; /* number or groups */ gid_t cmcred_groups[CMGROUP_MAX]; /* groups */ }; /* * Socket credentials. */ struct sockcred { uid_t sc_uid; /* real user id */ uid_t sc_euid; /* effective user id */ gid_t sc_gid; /* real group id */ gid_t sc_egid; /* effective group id */ int sc_ngroups; /* number of supplemental groups */ gid_t sc_groups[1]; /* variable length */ }; /* * Compute size of a sockcred structure with groups. */ #define SOCKCREDSIZE(ngrps) \ (sizeof(struct sockcred) + (sizeof(gid_t) * ((ngrps) - 1))) #endif /* __BSD_VISIBLE */ /* given pointer to struct cmsghdr, return pointer to data */ #define CMSG_DATA(cmsg) ((unsigned char *)(cmsg) + \ _ALIGN(sizeof(struct cmsghdr))) /* given pointer to struct cmsghdr, return pointer to next cmsghdr */ #define CMSG_NXTHDR(mhdr, cmsg) \ ((char *)(cmsg) == (char *)0 ? CMSG_FIRSTHDR(mhdr) : \ ((char *)(cmsg) + _ALIGN(((struct cmsghdr *)(cmsg))->cmsg_len) + \ _ALIGN(sizeof(struct cmsghdr)) > \ (char *)(mhdr)->msg_control + (mhdr)->msg_controllen) ? \ (struct cmsghdr *)0 : \ (struct cmsghdr *)(void *)((char *)(cmsg) + \ _ALIGN(((struct cmsghdr *)(cmsg))->cmsg_len))) /* * RFC 2292 requires to check msg_controllen, in case that the kernel returns * an empty list for some reasons. */ #define CMSG_FIRSTHDR(mhdr) \ ((mhdr)->msg_controllen >= sizeof(struct cmsghdr) ? \ (struct cmsghdr *)(mhdr)->msg_control : \ (struct cmsghdr *)0) #if __BSD_VISIBLE /* RFC 2292 additions */ #define CMSG_SPACE(l) (_ALIGN(sizeof(struct cmsghdr)) + _ALIGN(l)) #define CMSG_LEN(l) (_ALIGN(sizeof(struct cmsghdr)) + (l)) #endif #ifdef _KERNEL #define CMSG_ALIGN(n) _ALIGN(n) #endif /* "Socket"-level control message types: */ #define SCM_RIGHTS 0x01 /* access rights (array of int) */ #if __BSD_VISIBLE #define SCM_TIMESTAMP 0x02 /* timestamp (struct timeval) */ #define SCM_CREDS 0x03 /* process creds (struct cmsgcred) */ #define SCM_BINTIME 0x04 /* timestamp (struct bintime) */ #define SCM_REALTIME 0x05 /* timestamp (struct timespec) */ #define SCM_MONOTONIC 0x06 /* timestamp (struct timespec) */ #define SCM_TIME_INFO 0x07 /* timestamp info */ struct sock_timestamp_info { __uint32_t st_info_flags; __uint32_t st_info_pad0; __uint64_t st_info_rsv[7]; }; #define ST_INFO_HW 0x0001 /* SCM_TIMESTAMP was hw */ #define ST_INFO_HW_HPREC 0x0002 /* SCM_TIMESTAMP was hw-assisted on entrance */ #endif #if __BSD_VISIBLE /* * 4.3 compat sockaddr, move to compat file later */ struct osockaddr { unsigned short sa_family; /* address family */ char sa_data[14]; /* up to 14 bytes of direct address */ }; /* * 4.3-compat message header (move to compat file later). */ struct omsghdr { char *msg_name; /* optional address */ int msg_namelen; /* size of address */ struct iovec *msg_iov; /* scatter/gather array */ int msg_iovlen; /* # elements in msg_iov */ char *msg_accrights; /* access rights sent/received */ int msg_accrightslen; }; #endif /* * howto arguments for shutdown(2), specified by Posix.1g. */ #define SHUT_RD 0 /* shut down the reading side */ #define SHUT_WR 1 /* shut down the writing side */ #define SHUT_RDWR 2 /* shut down both sides */ #if __BSD_VISIBLE /* for SCTP */ /* we cheat and use the SHUT_XX defines for these */ #define PRU_FLUSH_RD SHUT_RD #define PRU_FLUSH_WR SHUT_WR #define PRU_FLUSH_RDWR SHUT_RDWR #endif #if __BSD_VISIBLE /* * sendfile(2) header/trailer struct */ struct sf_hdtr { struct iovec *headers; /* pointer to an array of header struct iovec's */ int hdr_cnt; /* number of header iovec's */ struct iovec *trailers; /* pointer to an array of trailer struct iovec's */ int trl_cnt; /* number of trailer iovec's */ }; /* * Sendfile-specific flag(s) */ #define SF_NODISKIO 0x00000001 #define SF_MNOWAIT 0x00000002 /* obsolete */ #define SF_SYNC 0x00000004 #define SF_USER_READAHEAD 0x00000008 #define SF_NOCACHE 0x00000010 #define SF_FLAGS(rh, flags) (((rh) << 16) | (flags)) #ifdef _KERNEL #define SF_READAHEAD(flags) ((flags) >> 16) #endif /* _KERNEL */ /* * Sendmmsg/recvmmsg specific structure(s) */ struct mmsghdr { struct msghdr msg_hdr; /* message header */ ssize_t msg_len; /* message length */ }; #endif /* __BSD_VISIBLE */ #ifndef _KERNEL #include __BEGIN_DECLS int accept(int, struct sockaddr * __restrict, socklen_t * __restrict); int bind(int, const struct sockaddr *, socklen_t); int connect(int, const struct sockaddr *, socklen_t); #if __BSD_VISIBLE int accept4(int, struct sockaddr * __restrict, socklen_t * __restrict, int); int bindat(int, int, const struct sockaddr *, socklen_t); int connectat(int, int, const struct sockaddr *, socklen_t); #endif int getpeername(int, struct sockaddr * __restrict, socklen_t * __restrict); int getsockname(int, struct sockaddr * __restrict, socklen_t * __restrict); int getsockopt(int, int, int, void * __restrict, socklen_t * __restrict); int listen(int, int); ssize_t recv(int, void *, size_t, int); ssize_t recvfrom(int, void *, size_t, int, struct sockaddr * __restrict, socklen_t * __restrict); ssize_t recvmsg(int, struct msghdr *, int); #if __BSD_VISIBLE struct timespec; ssize_t recvmmsg(int, struct mmsghdr * __restrict, size_t, int, const struct timespec * __restrict); #endif ssize_t send(int, const void *, size_t, int); ssize_t sendto(int, const void *, size_t, int, const struct sockaddr *, socklen_t); ssize_t sendmsg(int, const struct msghdr *, int); #if __BSD_VISIBLE int sendfile(int, int, off_t, size_t, struct sf_hdtr *, off_t *, int); ssize_t sendmmsg(int, struct mmsghdr * __restrict, size_t, int); int setfib(int); #endif int setsockopt(int, int, int, const void *, socklen_t); int shutdown(int, int); int sockatmark(int); int socket(int, int, int); int socketpair(int, int, int, int *); __END_DECLS #endif /* !_KERNEL */ #ifdef _KERNEL struct socket; struct tcpcb *so_sototcpcb(struct socket *so); struct inpcb *so_sotoinpcb(struct socket *so); struct sockbuf *so_sockbuf_snd(struct socket *); struct sockbuf *so_sockbuf_rcv(struct socket *); int so_state_get(const struct socket *); void so_state_set(struct socket *, int); int so_options_get(const struct socket *); void so_options_set(struct socket *, int); int so_error_get(const struct socket *); void so_error_set(struct socket *, int); int so_linger_get(const struct socket *); void so_linger_set(struct socket *, int); struct protosw *so_protosw_get(const struct socket *); void so_protosw_set(struct socket *, struct protosw *); void so_sorwakeup_locked(struct socket *so); void so_sowwakeup_locked(struct socket *so); void so_sorwakeup(struct socket *so); void so_sowwakeup(struct socket *so); void so_lock(struct socket *so); void so_unlock(struct socket *so); #endif /* _KERNEL */ #endif /* !_SYS_SOCKET_H_ */ Index: head/sys/sys/socketvar.h =================================================================== --- head/sys/sys/socketvar.h (revision 334718) +++ head/sys/sys/socketvar.h (revision 334719) @@ -1,524 +1,524 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1990, 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. * 3. 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. * * @(#)socketvar.h 8.3 (Berkeley) 2/19/95 * * $FreeBSD$ */ #ifndef _SYS_SOCKETVAR_H_ #define _SYS_SOCKETVAR_H_ /* * Socket generation count type. Also used in xinpcb, xtcpcb, xunpcb. */ typedef uint64_t so_gen_t; #if defined(_KERNEL) || defined(_WANT_SOCKET) #include /* for TAILQ macros */ #include /* for struct selinfo */ #include #include #include #include #include #ifdef _KERNEL #include #include #endif struct vnet; /* * Kernel structure per socket. * Contains send and receive buffer queues, * handle on protocol and pointer to protocol * private data and error information. */ typedef int so_upcall_t(struct socket *, void *, int); struct socket; /*- * Locking key to struct socket: * (a) constant after allocation, no locking required. * (b) locked by SOCK_LOCK(so). * (cr) locked by SOCKBUF_LOCK(&so->so_rcv). * (cs) locked by SOCKBUF_LOCK(&so->so_snd). * (e) locked by SOLISTEN_LOCK() of corresponding listening socket. * (f) not locked since integer reads/writes are atomic. * (g) used only as a sleep/wakeup address, no value. * (h) locked by global mutex so_global_mtx. */ TAILQ_HEAD(accept_queue, socket); struct socket { struct mtx so_lock; volatile u_int so_count; /* (b / refcount) */ struct selinfo so_rdsel; /* (b/cr) for so_rcv/so_comp */ struct selinfo so_wrsel; /* (b/cs) for so_snd */ short so_type; /* (a) generic type, see socket.h */ - short so_options; /* (b) from socket call, see socket.h */ + int so_options; /* (b) from socket call, see socket.h */ short so_linger; /* time to linger close(2) */ short so_state; /* (b) internal state flags SS_* */ void *so_pcb; /* protocol control block */ struct vnet *so_vnet; /* (a) network stack instance */ struct protosw *so_proto; /* (a) protocol handle */ short so_timeo; /* (g) connection timeout */ u_short so_error; /* (f) error affecting connection */ struct sigio *so_sigio; /* [sg] information for async I/O or out of band data (SIGURG) */ struct ucred *so_cred; /* (a) user credentials */ struct label *so_label; /* (b) MAC label for socket */ /* NB: generation count must not be first. */ so_gen_t so_gencnt; /* (h) generation count */ void *so_emuldata; /* (b) private data for emulators */ struct osd osd; /* Object Specific extensions */ /* * so_fibnum, so_user_cookie and friends can be used to attach * some user-specified metadata to a socket, which then can be * used by the kernel for various actions. * so_user_cookie is used by ipfw/dummynet. */ int so_fibnum; /* routing domain for this socket */ uint32_t so_user_cookie; int so_ts_clock; /* type of the clock used for timestamps */ uint32_t so_max_pacing_rate; /* (f) TX rate limit in bytes/s */ union { /* Regular (data flow) socket. */ struct { /* (cr, cs) Receive and send buffers. */ struct sockbuf so_rcv, so_snd; /* (e) Our place on accept queue. */ TAILQ_ENTRY(socket) so_list; struct socket *so_listen; /* (b) */ enum { SQ_NONE = 0, SQ_INCOMP = 0x0800, /* on sol_incomp */ SQ_COMP = 0x1000, /* on sol_comp */ } so_qstate; /* (b) */ /* (b) cached MAC label for peer */ struct label *so_peerlabel; u_long so_oobmark; /* chars to oob mark */ }; /* * Listening socket, where accepts occur, is so_listen in all * subsidiary sockets. If so_listen is NULL, socket is not * related to an accept. For a listening socket itself * sol_incomp queues partially completed connections, while * sol_comp is a queue of connections ready to be accepted. * If a connection is aborted and it has so_listen set, then * it has to be pulled out of either sol_incomp or sol_comp. * We allow connections to queue up based on current queue * lengths and limit on number of queued connections for this * socket. */ struct { /* (e) queue of partial unaccepted connections */ struct accept_queue sol_incomp; /* (e) queue of complete unaccepted connections */ struct accept_queue sol_comp; u_int sol_qlen; /* (e) sol_comp length */ u_int sol_incqlen; /* (e) sol_incomp length */ u_int sol_qlimit; /* (e) queue limit */ /* accept_filter(9) optional data */ struct accept_filter *sol_accept_filter; void *sol_accept_filter_arg; /* saved filter args */ char *sol_accept_filter_str; /* saved user args */ /* Optional upcall, for kernel socket. */ so_upcall_t *sol_upcall; /* (e) */ void *sol_upcallarg; /* (e) */ /* Socket buffer parameters, to be copied to * dataflow sockets, accepted from this one. */ int sol_sbrcv_lowat; int sol_sbsnd_lowat; u_int sol_sbrcv_hiwat; u_int sol_sbsnd_hiwat; short sol_sbrcv_flags; short sol_sbsnd_flags; sbintime_t sol_sbrcv_timeo; sbintime_t sol_sbsnd_timeo; }; }; }; #endif /* defined(_KERNEL) || defined(_WANT_SOCKET) */ /* * Socket state bits. * * Historically, this bits were all kept in the so_state field. For * locking reasons, they are now in multiple fields, as they are * locked differently. so_state maintains basic socket state protected * by the socket lock. so_qstate holds information about the socket * accept queues. Each socket buffer also has a state field holding * information relevant to that socket buffer (can't send, rcv). Many * fields will be read without locks to improve performance and avoid * lock order issues. However, this approach must be used with caution. */ #define SS_NOFDREF 0x0001 /* no file table ref any more */ #define SS_ISCONNECTED 0x0002 /* socket connected to a peer */ #define SS_ISCONNECTING 0x0004 /* in process of connecting to peer */ #define SS_ISDISCONNECTING 0x0008 /* in process of disconnecting */ #define SS_NBIO 0x0100 /* non-blocking ops */ #define SS_ASYNC 0x0200 /* async i/o notify */ #define SS_ISCONFIRMING 0x0400 /* deciding to accept connection req */ #define SS_ISDISCONNECTED 0x2000 /* socket disconnected from peer */ /* * Protocols can mark a socket as SS_PROTOREF to indicate that, following * pru_detach, they still want the socket to persist, and will free it * themselves when they are done. Protocols should only ever call sofree() * following setting this flag in pru_detach(), and never otherwise, as * sofree() bypasses socket reference counting. */ #define SS_PROTOREF 0x4000 /* strong protocol reference */ #ifdef _KERNEL #define SOCK_MTX(so) &(so)->so_lock #define SOCK_LOCK(so) mtx_lock(&(so)->so_lock) #define SOCK_OWNED(so) mtx_owned(&(so)->so_lock) #define SOCK_UNLOCK(so) mtx_unlock(&(so)->so_lock) #define SOCK_LOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_OWNED) #define SOCK_UNLOCK_ASSERT(so) mtx_assert(&(so)->so_lock, MA_NOTOWNED) #define SOLISTENING(sol) (((sol)->so_options & SO_ACCEPTCONN) != 0) #define SOLISTEN_LOCK(sol) do { \ mtx_lock(&(sol)->so_lock); \ KASSERT(SOLISTENING(sol), \ ("%s: %p not listening", __func__, (sol))); \ } while (0) #define SOLISTEN_TRYLOCK(sol) mtx_trylock(&(sol)->so_lock) #define SOLISTEN_UNLOCK(sol) do { \ KASSERT(SOLISTENING(sol), \ ("%s: %p not listening", __func__, (sol))); \ mtx_unlock(&(sol)->so_lock); \ } while (0) #define SOLISTEN_LOCK_ASSERT(sol) do { \ mtx_assert(&(sol)->so_lock, MA_OWNED); \ KASSERT(SOLISTENING(sol), \ ("%s: %p not listening", __func__, (sol))); \ } while (0) /* * Macros for sockets and socket buffering. */ /* * Flags to sblock(). */ #define SBL_WAIT 0x00000001 /* Wait if not immediately available. */ #define SBL_NOINTR 0x00000002 /* Force non-interruptible sleep. */ #define SBL_VALID (SBL_WAIT | SBL_NOINTR) /* * Do we need to notify the other side when I/O is possible? */ #define sb_notify(sb) (((sb)->sb_flags & (SB_WAIT | SB_SEL | SB_ASYNC | \ SB_UPCALL | SB_AIO | SB_KNOTE)) != 0) /* do we have to send all at once on a socket? */ #define sosendallatonce(so) \ ((so)->so_proto->pr_flags & PR_ATOMIC) /* can we read something from so? */ #define soreadabledata(so) \ (sbavail(&(so)->so_rcv) >= (so)->so_rcv.sb_lowat || (so)->so_error) #define soreadable(so) \ (soreadabledata(so) || ((so)->so_rcv.sb_state & SBS_CANTRCVMORE)) /* can we write something to so? */ #define sowriteable(so) \ ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && \ (((so)->so_state&SS_ISCONNECTED) || \ ((so)->so_proto->pr_flags&PR_CONNREQUIRED)==0)) || \ ((so)->so_snd.sb_state & SBS_CANTSENDMORE) || \ (so)->so_error) /* * soref()/sorele() ref-count the socket structure. * soref() may be called without owning socket lock, but in that case a * caller must own something that holds socket, and so_count must be not 0. * Note that you must still explicitly close the socket, but the last ref * count will free the structure. */ #define soref(so) refcount_acquire(&(so)->so_count) #define sorele(so) do { \ SOCK_LOCK_ASSERT(so); \ if (refcount_release(&(so)->so_count)) \ sofree(so); \ else \ SOCK_UNLOCK(so); \ } while (0) /* * In sorwakeup() and sowwakeup(), acquire the socket buffer lock to * avoid a non-atomic test-and-wakeup. However, sowakeup is * responsible for releasing the lock if it is called. We unlock only * if we don't call into sowakeup. If any code is introduced that * directly invokes the underlying sowakeup() primitives, it must * maintain the same semantics. */ #define sorwakeup_locked(so) do { \ SOCKBUF_LOCK_ASSERT(&(so)->so_rcv); \ if (sb_notify(&(so)->so_rcv)) \ sowakeup((so), &(so)->so_rcv); \ else \ SOCKBUF_UNLOCK(&(so)->so_rcv); \ } while (0) #define sorwakeup(so) do { \ SOCKBUF_LOCK(&(so)->so_rcv); \ sorwakeup_locked(so); \ } while (0) #define sowwakeup_locked(so) do { \ SOCKBUF_LOCK_ASSERT(&(so)->so_snd); \ if (sb_notify(&(so)->so_snd)) \ sowakeup((so), &(so)->so_snd); \ else \ SOCKBUF_UNLOCK(&(so)->so_snd); \ } while (0) #define sowwakeup(so) do { \ SOCKBUF_LOCK(&(so)->so_snd); \ sowwakeup_locked(so); \ } while (0) struct accept_filter { char accf_name[16]; int (*accf_callback) (struct socket *so, void *arg, int waitflag); void * (*accf_create) (struct socket *so, char *arg); void (*accf_destroy) (struct socket *so); SLIST_ENTRY(accept_filter) accf_next; }; #ifdef MALLOC_DECLARE MALLOC_DECLARE(M_ACCF); MALLOC_DECLARE(M_PCB); MALLOC_DECLARE(M_SONAME); #endif /* * Socket specific helper hook point identifiers * Do not leave holes in the sequence, hook registration is a loop. */ #define HHOOK_SOCKET_OPT 0 #define HHOOK_SOCKET_CREATE 1 #define HHOOK_SOCKET_RCV 2 #define HHOOK_SOCKET_SND 3 #define HHOOK_FILT_SOREAD 4 #define HHOOK_FILT_SOWRITE 5 #define HHOOK_SOCKET_CLOSE 6 #define HHOOK_SOCKET_LAST HHOOK_SOCKET_CLOSE struct socket_hhook_data { struct socket *so; struct mbuf *m; void *hctx; /* hook point specific data*/ int status; }; extern int maxsockets; extern u_long sb_max; extern so_gen_t so_gencnt; struct file; struct filecaps; struct filedesc; struct mbuf; struct sockaddr; struct ucred; struct uio; /* 'which' values for socket upcalls. */ #define SO_RCV 1 #define SO_SND 2 /* Return values for socket upcalls. */ #define SU_OK 0 #define SU_ISCONNECTED 1 /* * From uipc_socket and friends */ int getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len); int getsock_cap(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp, u_int *fflagp, struct filecaps *havecaps); void soabort(struct socket *so); int soaccept(struct socket *so, struct sockaddr **nam); void soaio_enqueue(struct task *task); void soaio_rcv(void *context, int pending); void soaio_snd(void *context, int pending); int socheckuid(struct socket *so, uid_t uid); int sobind(struct socket *so, struct sockaddr *nam, struct thread *td); int sobindat(int fd, struct socket *so, struct sockaddr *nam, struct thread *td); int soclose(struct socket *so); int soconnect(struct socket *so, struct sockaddr *nam, struct thread *td); int soconnectat(int fd, struct socket *so, struct sockaddr *nam, struct thread *td); int soconnect2(struct socket *so1, struct socket *so2); int socreate(int dom, struct socket **aso, int type, int proto, struct ucred *cred, struct thread *td); int sodisconnect(struct socket *so); struct sockaddr *sodupsockaddr(const struct sockaddr *sa, int mflags); void sofree(struct socket *so); void sohasoutofband(struct socket *so); int solisten(struct socket *so, int backlog, struct thread *td); void solisten_proto(struct socket *so, int backlog); int solisten_proto_check(struct socket *so); int solisten_dequeue(struct socket *, struct socket **, int); struct socket * sonewconn(struct socket *head, int connstatus); struct socket * sopeeloff(struct socket *); int sopoll(struct socket *so, int events, struct ucred *active_cred, struct thread *td); int sopoll_generic(struct socket *so, int events, struct ucred *active_cred, struct thread *td); int soreceive(struct socket *so, struct sockaddr **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp); int soreceive_stream(struct socket *so, struct sockaddr **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp); int soreceive_dgram(struct socket *so, struct sockaddr **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp); int soreceive_generic(struct socket *so, struct sockaddr **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp); int soreserve(struct socket *so, u_long sndcc, u_long rcvcc); void sorflush(struct socket *so); int sosend(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td); int sosend_dgram(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td); int sosend_generic(struct socket *so, struct sockaddr *addr, struct uio *uio, struct mbuf *top, struct mbuf *control, int flags, struct thread *td); int soshutdown(struct socket *so, int how); void soupcall_clear(struct socket *, int); void soupcall_set(struct socket *, int, so_upcall_t, void *); void solisten_upcall_set(struct socket *, so_upcall_t, void *); void sowakeup(struct socket *so, struct sockbuf *sb); void sowakeup_aio(struct socket *so, struct sockbuf *sb); void solisten_wakeup(struct socket *); int selsocket(struct socket *so, int events, struct timeval *tv, struct thread *td); void soisconnected(struct socket *so); void soisconnecting(struct socket *so); void soisdisconnected(struct socket *so); void soisdisconnecting(struct socket *so); void socantrcvmore(struct socket *so); void socantrcvmore_locked(struct socket *so); void socantsendmore(struct socket *so); void socantsendmore_locked(struct socket *so); /* * Accept filter functions (duh). */ int accept_filt_add(struct accept_filter *filt); int accept_filt_del(char *name); struct accept_filter *accept_filt_get(char *name); #ifdef ACCEPT_FILTER_MOD #ifdef SYSCTL_DECL SYSCTL_DECL(_net_inet_accf); #endif int accept_filt_generic_mod_event(module_t mod, int event, void *data); #endif #endif /* _KERNEL */ /* * Structure to export socket from kernel to utilities, via sysctl(3). */ struct xsocket { size_t xso_len; /* length of this structure */ union { void *xso_so; /* kernel address of struct socket */ int64_t ph_so; }; union { void *so_pcb; /* kernel address of struct inpcb */ int64_t ph_pcb; }; uint64_t so_oobmark; int64_t so_spare64[8]; int32_t xso_protocol; int32_t xso_family; uint32_t so_qlen; uint32_t so_incqlen; uint32_t so_qlimit; pid_t so_pgid; uid_t so_uid; int32_t so_spare32[8]; int16_t so_type; int16_t so_options; int16_t so_linger; int16_t so_state; int16_t so_timeo; uint16_t so_error; struct xsockbuf { uint32_t sb_cc; uint32_t sb_hiwat; uint32_t sb_mbcnt; uint32_t sb_mcnt; uint32_t sb_ccnt; uint32_t sb_mbmax; int32_t sb_lowat; int32_t sb_timeo; int16_t sb_flags; } so_rcv, so_snd; }; #ifdef _KERNEL void sotoxsocket(struct socket *so, struct xsocket *xso); void sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb); #endif /* * Socket buffer state bits. Exported via libprocstat(3). */ #define SBS_CANTSENDMORE 0x0010 /* can't send more data to peer */ #define SBS_CANTRCVMORE 0x0020 /* can't receive more data from peer */ #define SBS_RCVATMARK 0x0040 /* at mark on input */ #endif /* !_SYS_SOCKETVAR_H_ */