Index: head/sys/net/netisr.h =================================================================== --- head/sys/net/netisr.h (revision 271299) +++ head/sys/net/netisr.h (revision 271300) @@ -1,243 +1,245 @@ /*- * Copyright (c) 2007-2009 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * This software was 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. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _NET_NETISR_H_ #define _NET_NETISR_H_ /* * The netisr (network interrupt service routine) provides a deferred * execution evironment in which (generally inbound) network processing can * take place. Protocols register handlers which will be executed directly, * or via deferred dispatch, depending on the circumstances. * * Historically, this was implemented by the BSD software ISR facility; it is * now implemented via a software ithread (SWI). */ /* * Protocol numbers, which are encoded in monitoring applications and kernel * modules. Internally, these are used in bit shift operations so must have * a value 0 < proto < 32; we currently further limit at compile-time to 16 * for array-sizing purposes. */ #define NETISR_IP 1 #define NETISR_IGMP 2 /* IGMPv3 output queue */ #define NETISR_ROUTE 3 /* routing socket */ #define NETISR_ARP 4 /* same as AF_LINK */ #define NETISR_ETHER 5 /* ethernet input */ #define NETISR_IPV6 6 #define NETISR_NATM 7 #define NETISR_EPAIR 8 /* if_epair(4) */ +#define NETISR_IP_DIRECT 9 /* direct-dispatch IPv4 */ +#define NETISR_IPV6_DIRECT 10 /* direct-dispatch IPv6 */ /* * Protocol ordering and affinity policy constants. See the detailed * discussion of policies later in the file. */ #define NETISR_POLICY_SOURCE 1 /* Maintain source ordering. */ #define NETISR_POLICY_FLOW 2 /* Maintain flow ordering. */ #define NETISR_POLICY_CPU 3 /* Protocol determines CPU placement. */ /* * Protocol dispatch policy constants; selects whether and when direct * dispatch is permitted. */ #define NETISR_DISPATCH_DEFAULT 0 /* Use global default. */ #define NETISR_DISPATCH_DEFERRED 1 /* Always defer dispatch. */ #define NETISR_DISPATCH_HYBRID 2 /* Allow hybrid dispatch. */ #define NETISR_DISPATCH_DIRECT 3 /* Always direct dispatch. */ /* * Monitoring data structures, exported by sysctl(2). * * Three sysctls are defined. First, a per-protocol structure exported by * net.isr.proto. */ #define NETISR_NAMEMAXLEN 32 struct sysctl_netisr_proto { u_int snp_version; /* Length of struct. */ char snp_name[NETISR_NAMEMAXLEN]; /* nh_name */ u_int snp_proto; /* nh_proto */ u_int snp_qlimit; /* nh_qlimit */ u_int snp_policy; /* nh_policy */ u_int snp_flags; /* Various flags. */ u_int snp_dispatch; /* Dispatch policy. */ u_int _snp_ispare[6]; }; /* * Flags for sysctl_netisr_proto.snp_flags. */ #define NETISR_SNP_FLAGS_M2FLOW 0x00000001 /* nh_m2flow */ #define NETISR_SNP_FLAGS_M2CPUID 0x00000002 /* nh_m2cpuid */ #define NETISR_SNP_FLAGS_DRAINEDCPU 0x00000004 /* nh_drainedcpu */ /* * Next, a structure per-workstream, with per-protocol data, exported as * net.isr.workstream. */ struct sysctl_netisr_workstream { u_int snws_version; /* Length of struct. */ u_int snws_flags; /* Various flags. */ u_int snws_wsid; /* Workstream ID. */ u_int snws_cpu; /* nws_cpu */ u_int _snws_ispare[12]; }; /* * Flags for sysctl_netisr_workstream.snws_flags */ #define NETISR_SNWS_FLAGS_INTR 0x00000001 /* nws_intr_event */ /* * Finally, a per-workstream-per-protocol structure, exported as * net.isr.work. */ struct sysctl_netisr_work { u_int snw_version; /* Length of struct. */ u_int snw_wsid; /* Workstream ID. */ u_int snw_proto; /* Protocol number. */ u_int snw_len; /* nw_len */ u_int snw_watermark; /* nw_watermark */ u_int _snw_ispare[3]; uint64_t snw_dispatched; /* nw_dispatched */ uint64_t snw_hybrid_dispatched; /* nw_hybrid_dispatched */ uint64_t snw_qdrops; /* nw_qdrops */ uint64_t snw_queued; /* nw_queued */ uint64_t snw_handled; /* nw_handled */ uint64_t _snw_llspare[7]; }; #ifdef _KERNEL /*- * Protocols express ordering constraints and affinity preferences by * implementing one or neither of nh_m2flow and nh_m2cpuid, which are used by * netisr to determine which per-CPU workstream to assign mbufs to. * * The following policies may be used by protocols: * * NETISR_POLICY_SOURCE - netisr should maintain source ordering without * advice from the protocol. netisr will ignore any * flow IDs present on the mbuf for the purposes of * work placement. * * NETISR_POLICY_FLOW - netisr should maintain flow ordering as defined by * the mbuf header flow ID field. If the protocol * implements nh_m2flow, then netisr will query the * protocol in the event that the mbuf doesn't have a * flow ID, falling back on source ordering. * * NETISR_POLICY_CPU - netisr will delegate all work placement decisions to * the protocol, querying nh_m2cpuid for each packet. * * Protocols might make decisions about work placement based on an existing * calculated flow ID on the mbuf, such as one provided in hardware, the * receive interface pointed to by the mbuf (if any), the optional source * identifier passed at some dispatch points, or even parse packet headers to * calculate a flow. Both protocol handlers may return a new mbuf pointer * for the chain, or NULL if the packet proves invalid or m_pullup() fails. * * XXXRW: If we eventually support dynamic reconfiguration, there should be * protocol handlers to notify them of CPU configuration changes so that they * can rebalance work. */ struct mbuf; typedef void netisr_handler_t(struct mbuf *m); typedef struct mbuf *netisr_m2cpuid_t(struct mbuf *m, uintptr_t source, u_int *cpuid); typedef struct mbuf *netisr_m2flow_t(struct mbuf *m, uintptr_t source); typedef void netisr_drainedcpu_t(u_int cpuid); #define NETISR_CPUID_NONE ((u_int)-1) /* No affinity returned. */ /* * Data structure describing a protocol handler. */ struct netisr_handler { const char *nh_name; /* Character string protocol name. */ netisr_handler_t *nh_handler; /* Protocol handler. */ netisr_m2flow_t *nh_m2flow; /* Query flow for untagged packet. */ netisr_m2cpuid_t *nh_m2cpuid; /* Query CPU to process mbuf on. */ netisr_drainedcpu_t *nh_drainedcpu; /* Callback when drained a queue. */ u_int nh_proto; /* Integer protocol ID. */ u_int nh_qlimit; /* Maximum per-CPU queue depth. */ u_int nh_policy; /* Work placement policy. */ u_int nh_dispatch; /* Dispatch policy. */ u_int nh_ispare[4]; /* For future use. */ void *nh_pspare[4]; /* For future use. */ }; /* * Register, unregister, and other netisr handler management functions. */ void netisr_clearqdrops(const struct netisr_handler *nhp); void netisr_getqdrops(const struct netisr_handler *nhp, u_int64_t *qdropsp); void netisr_getqlimit(const struct netisr_handler *nhp, u_int *qlimitp); void netisr_register(const struct netisr_handler *nhp); int netisr_setqlimit(const struct netisr_handler *nhp, u_int qlimit); void netisr_unregister(const struct netisr_handler *nhp); /* * Process a packet destined for a protocol, and attempt direct dispatch. * Supplemental source ordering information can be passed using the _src * variant. */ int netisr_dispatch(u_int proto, struct mbuf *m); int netisr_dispatch_src(u_int proto, uintptr_t source, struct mbuf *m); int netisr_queue(u_int proto, struct mbuf *m); int netisr_queue_src(u_int proto, uintptr_t source, struct mbuf *m); /* * Provide a default implementation of "map an ID to a CPU ID". */ u_int netisr_default_flow2cpu(u_int flowid); /* * Utility routines to return the number of CPUs participting in netisr, and * to return a mapping from a number to a CPU ID that can be used with the * scheduler. */ u_int netisr_get_cpucount(void); u_int netisr_get_cpuid(u_int cpunumber); /* * Interfaces between DEVICE_POLLING and netisr. */ void netisr_sched_poll(void); void netisr_poll(void); void netisr_pollmore(void); #endif /* !_KERNEL */ #endif /* !_NET_NETISR_H_ */ Index: head/sys/netinet/in_var.h =================================================================== --- head/sys/netinet/in_var.h (revision 271299) +++ head/sys/netinet/in_var.h (revision 271300) @@ -1,440 +1,441 @@ /*- * Copyright (c) 1985, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_var.h 8.2 (Berkeley) 1/9/95 * $FreeBSD$ */ #ifndef _NETINET_IN_VAR_H_ #define _NETINET_IN_VAR_H_ #include #include #include struct igmp_ifinfo; struct in_multi; struct lltable; /* * IPv4 per-interface state. */ struct in_ifinfo { struct lltable *ii_llt; /* ARP state */ struct igmp_ifinfo *ii_igmp; /* IGMP state */ struct in_multi *ii_allhosts; /* 224.0.0.1 membership */ }; #if defined(_KERNEL) || defined(_WANT_IFADDR) /* * Interface address, Internet version. One of these structures * is allocated for each Internet address on an interface. * The ifaddr structure contains the protocol-independent part * of the structure and is assumed to be first. */ struct in_ifaddr { struct ifaddr ia_ifa; /* protocol-independent info */ #define ia_ifp ia_ifa.ifa_ifp #define ia_flags ia_ifa.ifa_flags /* ia_subnet{,mask} in host order */ u_long ia_subnet; /* subnet address */ u_long ia_subnetmask; /* mask of subnet */ LIST_ENTRY(in_ifaddr) ia_hash; /* entry in bucket of inet addresses */ TAILQ_ENTRY(in_ifaddr) ia_link; /* list of internet addresses */ struct sockaddr_in ia_addr; /* reserve space for interface name */ struct sockaddr_in ia_dstaddr; /* reserve space for broadcast addr */ #define ia_broadaddr ia_dstaddr struct sockaddr_in ia_sockmask; /* reserve space for general netmask */ }; #endif struct in_aliasreq { char ifra_name[IFNAMSIZ]; /* if name, e.g. "en0" */ struct sockaddr_in ifra_addr; struct sockaddr_in ifra_broadaddr; #define ifra_dstaddr ifra_broadaddr struct sockaddr_in ifra_mask; int ifra_vhid; }; /* * Given a pointer to an in_ifaddr (ifaddr), * return a pointer to the addr as a sockaddr_in. */ #define IA_SIN(ia) (&(((struct in_ifaddr *)(ia))->ia_addr)) #define IA_DSTSIN(ia) (&(((struct in_ifaddr *)(ia))->ia_dstaddr)) #define IA_MASKSIN(ia) (&(((struct in_ifaddr *)(ia))->ia_sockmask)) #define IN_LNAOF(in, ifa) \ ((ntohl((in).s_addr) & ~((struct in_ifaddr *)(ifa)->ia_subnetmask)) #ifdef _KERNEL extern u_char inetctlerrmap[]; #define LLTABLE(ifp) \ ((struct in_ifinfo *)(ifp)->if_afdata[AF_INET])->ii_llt /* * Hash table for IP addresses. */ TAILQ_HEAD(in_ifaddrhead, in_ifaddr); LIST_HEAD(in_ifaddrhashhead, in_ifaddr); VNET_DECLARE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); VNET_DECLARE(struct in_ifaddrhead, in_ifaddrhead); VNET_DECLARE(u_long, in_ifaddrhmask); /* mask for hash table */ #define V_in_ifaddrhashtbl VNET(in_ifaddrhashtbl) #define V_in_ifaddrhead VNET(in_ifaddrhead) #define V_in_ifaddrhmask VNET(in_ifaddrhmask) #define INADDR_NHASH_LOG2 9 #define INADDR_NHASH (1 << INADDR_NHASH_LOG2) #define INADDR_HASHVAL(x) fnv_32_buf((&(x)), sizeof(x), FNV1_32_INIT) #define INADDR_HASH(x) \ (&V_in_ifaddrhashtbl[INADDR_HASHVAL(x) & V_in_ifaddrhmask]) extern struct rwlock in_ifaddr_lock; #define IN_IFADDR_LOCK_ASSERT() rw_assert(&in_ifaddr_lock, RA_LOCKED) #define IN_IFADDR_RLOCK() rw_rlock(&in_ifaddr_lock) #define IN_IFADDR_RLOCK_ASSERT() rw_assert(&in_ifaddr_lock, RA_RLOCKED) #define IN_IFADDR_RUNLOCK() rw_runlock(&in_ifaddr_lock) #define IN_IFADDR_WLOCK() rw_wlock(&in_ifaddr_lock) #define IN_IFADDR_WLOCK_ASSERT() rw_assert(&in_ifaddr_lock, RA_WLOCKED) #define IN_IFADDR_WUNLOCK() rw_wunlock(&in_ifaddr_lock) /* * Macro for finding the internet address structure (in_ifaddr) * corresponding to one of our IP addresses (in_addr). */ #define INADDR_TO_IFADDR(addr, ia) \ /* struct in_addr addr; */ \ /* struct in_ifaddr *ia; */ \ do { \ \ LIST_FOREACH(ia, INADDR_HASH((addr).s_addr), ia_hash) \ if (IA_SIN(ia)->sin_addr.s_addr == (addr).s_addr) \ break; \ } while (0) /* * Macro for finding the interface (ifnet structure) corresponding to one * of our IP addresses. */ #define INADDR_TO_IFP(addr, ifp) \ /* struct in_addr addr; */ \ /* struct ifnet *ifp; */ \ { \ struct in_ifaddr *ia; \ \ INADDR_TO_IFADDR(addr, ia); \ (ifp) = (ia == NULL) ? NULL : ia->ia_ifp; \ } /* * Macro for finding the internet address structure (in_ifaddr) corresponding * to a given interface (ifnet structure). */ #define IFP_TO_IA(ifp, ia) \ /* struct ifnet *ifp; */ \ /* struct in_ifaddr *ia; */ \ do { \ IN_IFADDR_RLOCK(); \ for ((ia) = TAILQ_FIRST(&V_in_ifaddrhead); \ (ia) != NULL && (ia)->ia_ifp != (ifp); \ (ia) = TAILQ_NEXT((ia), ia_link)) \ continue; \ if ((ia) != NULL) \ ifa_ref(&(ia)->ia_ifa); \ IN_IFADDR_RUNLOCK(); \ } while (0) #endif /* * IP datagram reassembly. */ #define IPREASS_NHASH_LOG2 6 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) #define IPREASS_HMASK (IPREASS_NHASH - 1) #define IPREASS_HASH(x,y) \ (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) /* * Legacy IPv4 IGMP per-link structure. */ struct router_info { struct ifnet *rti_ifp; int rti_type; /* type of router which is querier on this interface */ int rti_time; /* # of slow timeouts since last old query */ SLIST_ENTRY(router_info) rti_list; }; /* * Per-interface IGMP router version information. */ struct igmp_ifinfo { LIST_ENTRY(igmp_ifinfo) igi_link; struct ifnet *igi_ifp; /* interface this instance belongs to */ uint32_t igi_version; /* IGMPv3 Host Compatibility Mode */ uint32_t igi_v1_timer; /* IGMPv1 Querier Present timer (s) */ uint32_t igi_v2_timer; /* IGMPv2 Querier Present timer (s) */ uint32_t igi_v3_timer; /* IGMPv3 General Query (interface) timer (s)*/ uint32_t igi_flags; /* IGMP per-interface flags */ uint32_t igi_rv; /* IGMPv3 Robustness Variable */ uint32_t igi_qi; /* IGMPv3 Query Interval (s) */ uint32_t igi_qri; /* IGMPv3 Query Response Interval (s) */ uint32_t igi_uri; /* IGMPv3 Unsolicited Report Interval (s) */ SLIST_HEAD(,in_multi) igi_relinmhead; /* released groups */ struct ifqueue igi_gq; /* queue of general query responses */ }; #define IGIF_SILENT 0x00000001 /* Do not use IGMP on this ifp */ #define IGIF_LOOPBACK 0x00000002 /* Send IGMP reports to loopback */ /* * IPv4 multicast IGMP-layer source entry. */ struct ip_msource { RB_ENTRY(ip_msource) ims_link; /* RB tree links */ in_addr_t ims_haddr; /* host byte order */ struct ims_st { uint16_t ex; /* # of exclusive members */ uint16_t in; /* # of inclusive members */ } ims_st[2]; /* state at t0, t1 */ uint8_t ims_stp; /* pending query */ }; /* * IPv4 multicast PCB-layer source entry. */ struct in_msource { RB_ENTRY(ip_msource) ims_link; /* RB tree links */ in_addr_t ims_haddr; /* host byte order */ uint8_t imsl_st[2]; /* state before/at commit */ }; RB_HEAD(ip_msource_tree, ip_msource); /* define struct ip_msource_tree */ static __inline int ip_msource_cmp(const struct ip_msource *a, const struct ip_msource *b) { if (a->ims_haddr < b->ims_haddr) return (-1); if (a->ims_haddr == b->ims_haddr) return (0); return (1); } RB_PROTOTYPE(ip_msource_tree, ip_msource, ims_link, ip_msource_cmp); /* * IPv4 multicast PCB-layer group filter descriptor. */ struct in_mfilter { struct ip_msource_tree imf_sources; /* source list for (S,G) */ u_long imf_nsrc; /* # of source entries */ uint8_t imf_st[2]; /* state before/at commit */ }; /* * IPv4 group descriptor. * * For every entry on an ifnet's if_multiaddrs list which represents * an IP multicast group, there is one of these structures. * * If any source filters are present, then a node will exist in the RB-tree * to permit fast lookup by source whenever an operation takes place. * This permits pre-order traversal when we issue reports. * Source filter trees are kept separately from the socket layer to * greatly simplify locking. * * When IGMPv3 is active, inm_timer is the response to group query timer. * The state-change timer inm_sctimer is separate; whenever state changes * for the group the state change record is generated and transmitted, * and kept if retransmissions are necessary. * * FUTURE: inm_link is now only used when groups are being purged * on a detaching ifnet. It could be demoted to a SLIST_ENTRY, but * because it is at the very start of the struct, we can't do this * w/o breaking the ABI for ifmcstat. */ struct in_multi { LIST_ENTRY(in_multi) inm_link; /* to-be-released by in_ifdetach */ struct in_addr inm_addr; /* IP multicast address, convenience */ struct ifnet *inm_ifp; /* back pointer to ifnet */ struct ifmultiaddr *inm_ifma; /* back pointer to ifmultiaddr */ u_int inm_timer; /* IGMPv1/v2 group / v3 query timer */ u_int inm_state; /* state of the membership */ void *inm_rti; /* unused, legacy field */ u_int inm_refcount; /* reference count */ /* New fields for IGMPv3 follow. */ struct igmp_ifinfo *inm_igi; /* IGMP info */ SLIST_ENTRY(in_multi) inm_nrele; /* to-be-released by IGMP */ struct ip_msource_tree inm_srcs; /* tree of sources */ u_long inm_nsrc; /* # of tree entries */ struct ifqueue inm_scq; /* queue of pending * state-change packets */ struct timeval inm_lastgsrtv; /* Time of last G-S-R query */ uint16_t inm_sctimer; /* state-change timer */ uint16_t inm_scrv; /* state-change rexmit count */ /* * SSM state counters which track state at T0 (the time the last * state-change report's RV timer went to zero) and T1 * (time of pending report, i.e. now). * Used for computing IGMPv3 state-change reports. Several refcounts * are maintained here to optimize for common use-cases. */ struct inm_st { uint16_t iss_fmode; /* IGMP filter mode */ uint16_t iss_asm; /* # of ASM listeners */ uint16_t iss_ex; /* # of exclusive members */ uint16_t iss_in; /* # of inclusive members */ uint16_t iss_rec; /* # of recorded sources */ } inm_st[2]; /* state at t0, t1 */ }; /* * Helper function to derive the filter mode on a source entry * from its internal counters. Predicates are: * A source is only excluded if all listeners exclude it. * A source is only included if no listeners exclude it, * and at least one listener includes it. * May be used by ifmcstat(8). */ static __inline uint8_t ims_get_mode(const struct in_multi *inm, const struct ip_msource *ims, uint8_t t) { t = !!t; if (inm->inm_st[t].iss_ex > 0 && inm->inm_st[t].iss_ex == ims->ims_st[t].ex) return (MCAST_EXCLUDE); else if (ims->ims_st[t].in > 0 && ims->ims_st[t].ex == 0) return (MCAST_INCLUDE); return (MCAST_UNDEFINED); } #ifdef _KERNEL #ifdef SYSCTL_DECL SYSCTL_DECL(_net_inet); SYSCTL_DECL(_net_inet_ip); SYSCTL_DECL(_net_inet_raw); #endif /* * Lock macros for IPv4 layer multicast address lists. IPv4 lock goes * before link layer multicast locks in the lock order. In most cases, * consumers of IN_*_MULTI() macros should acquire the locks before * calling them; users of the in_{add,del}multi() functions should not. */ extern struct mtx in_multi_mtx; #define IN_MULTI_LOCK() mtx_lock(&in_multi_mtx) #define IN_MULTI_UNLOCK() mtx_unlock(&in_multi_mtx) #define IN_MULTI_LOCK_ASSERT() mtx_assert(&in_multi_mtx, MA_OWNED) #define IN_MULTI_UNLOCK_ASSERT() mtx_assert(&in_multi_mtx, MA_NOTOWNED) /* Acquire an in_multi record. */ static __inline void inm_acquire_locked(struct in_multi *inm) { IN_MULTI_LOCK_ASSERT(); ++inm->inm_refcount; } /* * Return values for imo_multi_filter(). */ #define MCAST_PASS 0 /* Pass */ #define MCAST_NOTGMEMBER 1 /* This host not a member of group */ #define MCAST_NOTSMEMBER 2 /* This host excluded source */ #define MCAST_MUTED 3 /* [deprecated] */ struct rtentry; struct route; struct ip_moptions; struct radix_node_head; struct in_multi *inm_lookup_locked(struct ifnet *, const struct in_addr); struct in_multi *inm_lookup(struct ifnet *, const struct in_addr); int imo_multi_filter(const struct ip_moptions *, const struct ifnet *, const struct sockaddr *, const struct sockaddr *); void inm_commit(struct in_multi *); void inm_clear_recorded(struct in_multi *); void inm_print(const struct in_multi *); int inm_record_source(struct in_multi *inm, const in_addr_t); void inm_release(struct in_multi *); void inm_release_locked(struct in_multi *); struct in_multi * in_addmulti(struct in_addr *, struct ifnet *); void in_delmulti(struct in_multi *); int in_joingroup(struct ifnet *, const struct in_addr *, /*const*/ struct in_mfilter *, struct in_multi **); int in_joingroup_locked(struct ifnet *, const struct in_addr *, /*const*/ struct in_mfilter *, struct in_multi **); int in_leavegroup(struct in_multi *, /*const*/ struct in_mfilter *); int in_leavegroup_locked(struct in_multi *, /*const*/ struct in_mfilter *); int in_control(struct socket *, u_long, caddr_t, struct ifnet *, struct thread *); void in_rtqdrain(void); int in_addprefix(struct in_ifaddr *, int); int in_scrubprefix(struct in_ifaddr *, u_int); void ip_input(struct mbuf *); +void ip_direct_input(struct mbuf *); void in_ifadown(struct ifaddr *ifa, int); struct mbuf *ip_fastforward(struct mbuf *); void *in_domifattach(struct ifnet *); void in_domifdetach(struct ifnet *, void *); /* XXX */ void in_rtalloc_ign(struct route *ro, u_long ignflags, u_int fibnum); void in_rtalloc(struct route *ro, u_int fibnum); struct rtentry *in_rtalloc1(struct sockaddr *, int, u_long, u_int); void in_rtredirect(struct sockaddr *, struct sockaddr *, struct sockaddr *, int, struct sockaddr *, u_int); int in_rtrequest(int, struct sockaddr *, struct sockaddr *, struct sockaddr *, int, struct rtentry **, u_int); void in_setmatchfunc(struct radix_node_head *, int); #if 0 int in_rt_getifa(struct rt_addrinfo *, u_int fibnum); int in_rtioctl(u_long, caddr_t, u_int); int in_rtrequest1(int, struct rt_addrinfo *, struct rtentry **, u_int); #endif #endif /* _KERNEL */ /* INET6 stuff */ #include #endif /* _NETINET_IN_VAR_H_ */ Index: head/sys/netinet/ip_input.c =================================================================== --- head/sys/netinet/ip_input.c (revision 271299) +++ head/sys/netinet/ip_input.c (revision 271300) @@ -1,1787 +1,1918 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_bootp.h" #include "opt_ipfw.h" #include "opt_ipstealth.h" #include "opt_ipsec.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 IPSEC #include #endif /* IPSEC */ +#include #include #include #ifdef CTASSERT CTASSERT(sizeof(struct ip) == 20); #endif struct rwlock in_ifaddr_lock; RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); VNET_DEFINE(int, rsvp_on); VNET_DEFINE(int, ipforwarding); SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, &VNET_NAME(ipforwarding), 0, "Enable IP forwarding between interfaces"); static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ #define V_ipsendredirects VNET(ipsendredirects) SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, &VNET_NAME(ipsendredirects), 0, "Enable sending IP redirects"); static VNET_DEFINE(int, ip_keepfaith); #define V_ip_keepfaith VNET(ip_keepfaith) SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, &VNET_NAME(ip_keepfaith), 0, "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); static VNET_DEFINE(int, ip_sendsourcequench); #define V_ip_sendsourcequench VNET(ip_sendsourcequench) SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, &VNET_NAME(ip_sendsourcequench), 0, "Enable the transmission of source quench packets"); VNET_DEFINE(int, ip_do_randomid); SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, &VNET_NAME(ip_do_randomid), 0, "Assign random ip_id values"); /* * XXX - Setting ip_checkinterface mostly implements the receive side of * the Strong ES model described in RFC 1122, but since the routing table * and transmit implementation do not implement the Strong ES model, * setting this to 1 results in an odd hybrid. * * XXX - ip_checkinterface currently must be disabled if you use ipnat * to translate the destination address to another local interface. * * XXX - ip_checkinterface must be disabled if you add IP aliases * to the loopback interface instead of the interface where the * packets for those addresses are received. */ static VNET_DEFINE(int, ip_checkinterface); #define V_ip_checkinterface VNET(ip_checkinterface) SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, &VNET_NAME(ip_checkinterface), 0, "Verify packet arrives on correct interface"); VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ static struct netisr_handler ip_nh = { .nh_name = "ip", .nh_handler = ip_input, .nh_proto = NETISR_IP, +#ifdef RSS + .nh_m2cpuid = rss_soft_m2cpuid, + .nh_policy = NETISR_POLICY_CPU, + .nh_dispatch = NETISR_DISPATCH_HYBRID, +#else .nh_policy = NETISR_POLICY_FLOW, +#endif }; +#ifdef RSS +/* + * Directly dispatched frames are currently assumed + * to have a flowid already calculated. + * + * It should likely have something that assert it + * actually has valid flow details. + */ +static struct netisr_handler ip_direct_nh = { + .nh_name = "ip_direct", + .nh_handler = ip_direct_input, + .nh_proto = NETISR_IP_DIRECT, + .nh_m2cpuid = rss_m2cpuid, + .nh_policy = NETISR_POLICY_CPU, + .nh_dispatch = NETISR_DISPATCH_HYBRID, +}; +#endif + extern struct domain inetdomain; extern struct protosw inetsw[]; u_char ip_protox[IPPROTO_MAX]; VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ static VNET_DEFINE(uma_zone_t, ipq_zone); static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]); static struct mtx ipqlock; #define V_ipq_zone VNET(ipq_zone) #define V_ipq VNET(ipq) #define IPQ_LOCK() mtx_lock(&ipqlock) #define IPQ_UNLOCK() mtx_unlock(&ipqlock) #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) static void maxnipq_update(void); static void ipq_zone_change(void *); static void ip_drain_locked(void); static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */ static VNET_DEFINE(int, nipq); /* Total # of reass queues */ #define V_maxnipq VNET(maxnipq) #define V_nipq VNET(nipq) SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &VNET_NAME(nipq), 0, "Current number of IPv4 fragment reassembly queue entries"); static VNET_DEFINE(int, maxfragsperpacket); #define V_maxfragsperpacket VNET(maxfragsperpacket) SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, &VNET_NAME(maxfragsperpacket), 0, "Maximum number of IPv4 fragments allowed per packet"); #ifdef IPCTL_DEFMTU SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, &ip_mtu, 0, "Default MTU"); #endif #ifdef IPSTEALTH VNET_DEFINE(int, ipstealth); SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &VNET_NAME(ipstealth), 0, "IP stealth mode, no TTL decrementation on forwarding"); #endif static void ip_freef(struct ipqhead *, struct ipq *); /* * IP statistics are stored in the "array" of counter(9)s. */ VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); VNET_PCPUSTAT_SYSINIT(ipstat); SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipstat); #endif /* VIMAGE */ /* * Kernel module interface for updating ipstat. The argument is an index * into ipstat treated as an array. */ void kmod_ipstat_inc(int statnum) { counter_u64_add(VNET(ipstat)[statnum], 1); } void kmod_ipstat_dec(int statnum) { counter_u64_add(VNET(ipstat)[statnum], -1); } static int sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) { int error, qlimit; netisr_getqlimit(&ip_nh, &qlimit); error = sysctl_handle_int(oidp, &qlimit, 0, req); if (error || !req->newptr) return (error); if (qlimit < 1) return (EINVAL); return (netisr_setqlimit(&ip_nh, qlimit)); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", "Maximum size of the IP input queue"); static int sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) { u_int64_t qdrops_long; int error, qdrops; netisr_getqdrops(&ip_nh, &qdrops_long); qdrops = qdrops_long; error = sysctl_handle_int(oidp, &qdrops, 0, req); if (error || !req->newptr) return (error); if (qdrops != 0) return (EINVAL); netisr_clearqdrops(&ip_nh); return (0); } SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", "Number of packets dropped from the IP input queue"); +#ifdef RSS +static int +sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) +{ + int error, qlimit; + + netisr_getqlimit(&ip_direct_nh, &qlimit); + error = sysctl_handle_int(oidp, &qlimit, 0, req); + if (error || !req->newptr) + return (error); + if (qlimit < 1) + return (EINVAL); + return (netisr_setqlimit(&ip_direct_nh, qlimit)); +} +SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen, + CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I", + "Maximum size of the IP direct input queue"); + +static int +sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) +{ + u_int64_t qdrops_long; + int error, qdrops; + + netisr_getqdrops(&ip_direct_nh, &qdrops_long); + qdrops = qdrops_long; + error = sysctl_handle_int(oidp, &qdrops, 0, req); + if (error || !req->newptr) + return (error); + if (qdrops != 0) + return (EINVAL); + netisr_clearqdrops(&ip_direct_nh); + return (0); +} + +SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops, + CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I", + "Number of packets dropped from the IP direct input queue"); +#endif /* RSS */ + /* * IP initialization: fill in IP protocol switch table. * All protocols not implemented in kernel go to raw IP protocol handler. */ void ip_init(void) { struct protosw *pr; int i; V_ip_id = time_second & 0xffff; TAILQ_INIT(&V_in_ifaddrhead); V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); /* Initialize IP reassembly queue. */ for (i = 0; i < IPREASS_NHASH; i++) TAILQ_INIT(&V_ipq[i]); V_maxnipq = nmbclusters / 32; V_maxfragsperpacket = 16; V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); maxnipq_update(); /* Initialize packet filter hooks. */ V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; V_inet_pfil_hook.ph_af = AF_INET; if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) printf("%s: WARNING: unable to register pfil hook, " "error %d\n", __func__, i); /* Skip initialization of globals for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) panic("ip_init: PF_INET not found"); /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ for (i = 0; i < IPPROTO_MAX; i++) ip_protox[i] = pr - inetsw; /* * Cycle through IP protocols and put them into the appropriate place * in ip_protox[]. */ for (pr = inetdomain.dom_protosw; pr < inetdomain.dom_protoswNPROTOSW; pr++) if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { /* Be careful to only index valid IP protocols. */ if (pr->pr_protocol < IPPROTO_MAX) ip_protox[pr->pr_protocol] = pr - inetsw; } EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change, NULL, EVENTHANDLER_PRI_ANY); /* Initialize various other remaining things. */ IPQ_LOCK_INIT(); netisr_register(&ip_nh); +#ifdef RSS + netisr_register(&ip_direct_nh); +#endif } #ifdef VIMAGE void ip_destroy(void) { int i; if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0) printf("%s: WARNING: unable to unregister pfil hook, " "error %d\n", __func__, i); /* Cleanup in_ifaddr hash table; should be empty. */ hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); IPQ_LOCK(); ip_drain_locked(); IPQ_UNLOCK(); uma_zdestroy(V_ipq_zone); } #endif +#ifdef RSS /* + * IP direct input routine. + * + * This is called when reinjecting completed fragments where + * all of the previous checking and book-keeping has been done. + */ +void +ip_direct_input(struct mbuf *m) +{ + struct ip *ip; + int hlen; + + ip = mtod(m, struct ip *); + hlen = ip->ip_hl << 2; + + IPSTAT_INC(ips_delivered); + (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); + return; +} +#endif + +/* * Ip input routine. Checksum and byte swap header. If fragmented * try to reassemble. Process options. Pass to next level. */ void ip_input(struct mbuf *m) { struct ip *ip = NULL; struct in_ifaddr *ia = NULL; struct ifaddr *ifa; struct ifnet *ifp; int checkif, hlen = 0; uint16_t sum, ip_len; int dchg = 0; /* dest changed after fw */ struct in_addr odst; /* original dst address */ M_ASSERTPKTHDR(m); if (m->m_flags & M_FASTFWD_OURS) { m->m_flags &= ~M_FASTFWD_OURS; /* Set up some basics that will be used later. */ ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; ip_len = ntohs(ip->ip_len); goto ours; } IPSTAT_INC(ips_total); if (m->m_pkthdr.len < sizeof(struct ip)) goto tooshort; if (m->m_len < sizeof (struct ip) && (m = m_pullup(m, sizeof (struct ip))) == NULL) { IPSTAT_INC(ips_toosmall); return; } ip = mtod(m, struct ip *); if (ip->ip_v != IPVERSION) { IPSTAT_INC(ips_badvers); goto bad; } hlen = ip->ip_hl << 2; if (hlen < sizeof(struct ip)) { /* minimum header length */ IPSTAT_INC(ips_badhlen); goto bad; } if (hlen > m->m_len) { if ((m = m_pullup(m, hlen)) == NULL) { IPSTAT_INC(ips_badhlen); return; } ip = mtod(m, struct ip *); } IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); /* 127/8 must not appear on wire - RFC1122 */ ifp = m->m_pkthdr.rcvif; 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); goto bad; } } if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); } else { if (hlen == sizeof(struct ip)) { sum = in_cksum_hdr(ip); } else { sum = in_cksum(m, hlen); } } if (sum) { IPSTAT_INC(ips_badsum); goto bad; } #ifdef ALTQ if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) /* packet is dropped by traffic conditioner */ return; #endif ip_len = ntohs(ip->ip_len); if (ip_len < hlen) { IPSTAT_INC(ips_badlen); goto bad; } /* * Check that the amount of data in the buffers * is as at least much as the IP header would have us expect. * Trim mbufs if longer than we expect. * Drop packet if shorter than we expect. */ if (m->m_pkthdr.len < ip_len) { tooshort: IPSTAT_INC(ips_tooshort); goto bad; } if (m->m_pkthdr.len > ip_len) { if (m->m_len == m->m_pkthdr.len) { m->m_len = ip_len; m->m_pkthdr.len = ip_len; } else m_adj(m, ip_len - m->m_pkthdr.len); } + #ifdef IPSEC /* * Bypass packet filtering for packets previously handled by IPsec. */ if (ip_ipsec_filtertunnel(m)) goto passin; #endif /* IPSEC */ /* * Run through list of hooks for input packets. * * NB: Beware of the destination address changing (e.g. * by NAT rewriting). When this happens, tell * ip_forward to do the right thing. */ /* Jump over all PFIL processing if hooks are not active. */ if (!PFIL_HOOKED(&V_inet_pfil_hook)) goto passin; odst = ip->ip_dst; if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) return; if (m == NULL) /* consumed by filter */ return; ip = mtod(m, struct ip *); dchg = (odst.s_addr != ip->ip_dst.s_addr); ifp = m->m_pkthdr.rcvif; if (m->m_flags & M_FASTFWD_OURS) { m->m_flags &= ~M_FASTFWD_OURS; goto ours; } if (m->m_flags & M_IP_NEXTHOP) { dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL); if (dchg != 0) { /* * Directly ship the packet on. This allows * forwarding packets originally destined to us * to some other directly connected host. */ ip_forward(m, 1); return; } } passin: /* * Process options and, if not destined for us, * ship it on. ip_dooptions returns 1 when an * error was detected (causing an icmp message * to be sent and the original packet to be freed). */ if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) return; /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no * matter if it is destined to another node, or whether it is * a multicast one, RSVP wants it! and prevents it from being forwarded * anywhere else. Also checks if the rsvp daemon is running before * grabbing the packet. */ if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) goto ours; /* * Check our list of addresses, to see if the packet is for us. * If we don't have any addresses, assume any unicast packet * we receive might be for us (and let the upper layers deal * with it). */ if (TAILQ_EMPTY(&V_in_ifaddrhead) && (m->m_flags & (M_MCAST|M_BCAST)) == 0) goto ours; /* * Enable a consistency check between the destination address * and the arrival interface for a unicast packet (the RFC 1122 * strong ES model) if IP forwarding is disabled and the packet * is not locally generated and the packet is not subject to * 'ipfw fwd'. * * XXX - Checking also should be disabled if the destination * address is ipnat'ed to a different interface. * * XXX - Checking is incompatible with IP aliases added * to the loopback interface instead of the interface where * the packets are received. * * XXX - This is the case for carp vhost IPs as well so we * insert a workaround. If the packet got here, we already * checked with carp_iamatch() and carp_forus(). */ checkif = V_ip_checkinterface && (V_ipforwarding == 0) && ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && ifp->if_carp == NULL && (dchg == 0); /* * Check for exact addresses in the hash bucket. */ /* IN_IFADDR_RLOCK(); */ LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { /* * If the address matches, verify that the packet * arrived via the correct interface if checking is * enabled. */ if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && (!checkif || ia->ia_ifp == ifp)) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); /* IN_IFADDR_RUNLOCK(); */ goto ours; } } /* IN_IFADDR_RUNLOCK(); */ /* * Check for broadcast addresses. * * Only accept broadcast packets that arrive via the matching * interface. Reception of forwarded directed broadcasts would * be handled via ip_forward() and ether_output() with the loopback * into the stack for SIMPLEX interfaces handled by ether_output(). */ if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == ip->ip_dst.s_addr) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); IF_ADDR_RUNLOCK(ifp); goto ours; } #ifdef BOOTP_COMPAT if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); IF_ADDR_RUNLOCK(ifp); goto ours; } #endif } IF_ADDR_RUNLOCK(ifp); ia = NULL; } /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { if (V_ip_mrouter) { /* * If we are acting as a multicast router, all * incoming multicast packets are passed to the * kernel-level multicast forwarding function. * The packet is returned (relatively) intact; if * ip_mforward() returns a non-zero value, the packet * must be discarded, else it may be accepted below. */ if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } /* * The process-level routing daemon needs to receive * all multicast IGMP packets, whether or not this * host belongs to their destination groups. */ if (ip->ip_p == IPPROTO_IGMP) goto ours; IPSTAT_INC(ips_forward); } /* * Assume the packet is for us, to avoid prematurely taking * a lock on the in_multi hash. Protocols must perform * their own filtering and update statistics accordingly. */ goto ours; } if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) goto ours; if (ip->ip_dst.s_addr == INADDR_ANY) goto ours; /* * FAITH(Firewall Aided Internet Translator) */ if (ifp && ifp->if_type == IFT_FAITH) { if (V_ip_keepfaith) { if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) goto ours; } m_freem(m); return; } /* * Not for us; forward if possible and desirable. */ if (V_ipforwarding == 0) { IPSTAT_INC(ips_cantforward); m_freem(m); } else { #ifdef IPSEC if (ip_ipsec_fwd(m)) goto bad; #endif /* IPSEC */ ip_forward(m, dchg); } return; ours: #ifdef IPSTEALTH /* * IPSTEALTH: Process non-routing options only * if the packet is destined for us. */ if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) return; #endif /* IPSTEALTH */ /* * Attempt reassembly; if it succeeds, proceed. * ip_reass() will return a different mbuf. */ if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { /* XXXGL: shouldn't we save & set m_flags? */ m = ip_reass(m); if (m == NULL) return; ip = mtod(m, struct ip *); /* Get the header length of the reassembled packet */ hlen = ip->ip_hl << 2; } #ifdef IPSEC /* * enforce IPsec policy checking if we are seeing last header. * note that we do not visit this with protocols with pcb layer * code - like udp/tcp/raw ip. */ if (ip_ipsec_input(m)) goto bad; #endif /* IPSEC */ /* * Switch out to protocol's input routine. */ IPSTAT_INC(ips_delivered); (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); return; bad: m_freem(m); } /* * After maxnipq has been updated, propagate the change to UMA. The UMA zone * max has slightly different semantics than the sysctl, for historical * reasons. */ static void maxnipq_update(void) { /* * -1 for unlimited allocation. */ if (V_maxnipq < 0) uma_zone_set_max(V_ipq_zone, 0); /* * Positive number for specific bound. */ if (V_maxnipq > 0) uma_zone_set_max(V_ipq_zone, V_maxnipq); /* * Zero specifies no further fragment queue allocation -- set the * bound very low, but rely on implementation elsewhere to actually * prevent allocation and reclaim current queues. */ if (V_maxnipq == 0) uma_zone_set_max(V_ipq_zone, 1); } static void ipq_zone_change(void *tag) { if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) { V_maxnipq = nmbclusters / 32; maxnipq_update(); } } static int sysctl_maxnipq(SYSCTL_HANDLER_ARGS) { int error, i; i = V_maxnipq; error = sysctl_handle_int(oidp, &i, 0, req); if (error || !req->newptr) return (error); /* * XXXRW: Might be a good idea to sanity check the argument and place * an extreme upper bound. */ if (i < -1) return (EINVAL); V_maxnipq = i; maxnipq_update(); return (0); } SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW, NULL, 0, sysctl_maxnipq, "I", "Maximum number of IPv4 fragment reassembly queue entries"); #define M_IP_FRAG M_PROTO9 /* * Take incoming datagram fragment and try to reassemble it into * whole datagram. If the argument is the first fragment or one * in between the function will return NULL and store the mbuf * in the fragment chain. If the argument is the last fragment * the packet will be reassembled and the pointer to the new * mbuf returned for further processing. Only m_tags attached * to the first packet/fragment are preserved. * The IP header is *NOT* adjusted out of iplen. */ struct mbuf * ip_reass(struct mbuf *m) { struct ip *ip; struct mbuf *p, *q, *nq, *t; struct ipq *fp = NULL; struct ipqhead *head; int i, hlen, next; u_int8_t ecn, ecn0; u_short hash; +#ifdef RSS + uint32_t rss_hash, rss_type; +#endif /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ if (V_maxnipq == 0 || V_maxfragsperpacket == 0) { IPSTAT_INC(ips_fragments); IPSTAT_INC(ips_fragdropped); m_freem(m); return (NULL); } ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); head = &V_ipq[hash]; IPQ_LOCK(); /* * Look for queue of fragments * of this datagram. */ TAILQ_FOREACH(fp, head, ipq_list) if (ip->ip_id == fp->ipq_id && ip->ip_src.s_addr == fp->ipq_src.s_addr && ip->ip_dst.s_addr == fp->ipq_dst.s_addr && #ifdef MAC mac_ipq_match(m, fp) && #endif ip->ip_p == fp->ipq_p) goto found; fp = NULL; /* * Attempt to trim the number of allocated fragment queues if it * exceeds the administrative limit. */ if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) { /* * drop something from the tail of the current queue * before proceeding further */ struct ipq *q = TAILQ_LAST(head, ipqhead); if (q == NULL) { /* gak */ for (i = 0; i < IPREASS_NHASH; i++) { struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead); if (r) { IPSTAT_ADD(ips_fragtimeout, r->ipq_nfrags); ip_freef(&V_ipq[i], r); break; } } } else { IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags); ip_freef(head, q); } } found: /* * Adjust ip_len to not reflect header, * convert offset of this to bytes. */ ip->ip_len = htons(ntohs(ip->ip_len) - hlen); if (ip->ip_off & htons(IP_MF)) { /* * Make sure that fragments have a data length * that's a non-zero multiple of 8 bytes. */ if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) { IPSTAT_INC(ips_toosmall); /* XXX */ goto dropfrag; } m->m_flags |= M_IP_FRAG; } else m->m_flags &= ~M_IP_FRAG; ip->ip_off = htons(ntohs(ip->ip_off) << 3); /* * Attempt reassembly; if it succeeds, proceed. * ip_reass() will return a different mbuf. */ IPSTAT_INC(ips_fragments); m->m_pkthdr.PH_loc.ptr = ip; /* Previous ip_reass() started here. */ /* * Presence of header sizes in mbufs * would confuse code below. */ m->m_data += hlen; m->m_len -= hlen; /* * If first fragment to arrive, create a reassembly queue. */ if (fp == NULL) { fp = uma_zalloc(V_ipq_zone, M_NOWAIT); if (fp == NULL) goto dropfrag; #ifdef MAC if (mac_ipq_init(fp, M_NOWAIT) != 0) { uma_zfree(V_ipq_zone, fp); fp = NULL; goto dropfrag; } mac_ipq_create(m, fp); #endif TAILQ_INSERT_HEAD(head, fp, ipq_list); V_nipq++; fp->ipq_nfrags = 1; fp->ipq_ttl = IPFRAGTTL; fp->ipq_p = ip->ip_p; fp->ipq_id = ip->ip_id; fp->ipq_src = ip->ip_src; fp->ipq_dst = ip->ip_dst; fp->ipq_frags = m; m->m_nextpkt = NULL; goto done; } else { fp->ipq_nfrags++; #ifdef MAC mac_ipq_update(m, fp); #endif } #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr)) /* * Handle ECN by comparing this segment with the first one; * if CE is set, do not lose CE. * drop if CE and not-ECT are mixed for the same packet. */ ecn = ip->ip_tos & IPTOS_ECN_MASK; ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; if (ecn == IPTOS_ECN_CE) { if (ecn0 == IPTOS_ECN_NOTECT) goto dropfrag; if (ecn0 != IPTOS_ECN_CE) GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; } if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) goto dropfrag; /* * Find a segment which begins after this one does. */ for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off)) break; /* * If there is a preceding segment, it may provide some of * our data already. If so, drop the data from the incoming * segment. If it provides all of our data, drop us, otherwise * stick new segment in the proper place. * * If some of the data is dropped from the preceding * segment, then it's checksum is invalidated. */ if (p) { i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) - ntohs(ip->ip_off); if (i > 0) { if (i >= ntohs(ip->ip_len)) goto dropfrag; m_adj(m, i); m->m_pkthdr.csum_flags = 0; ip->ip_off = htons(ntohs(ip->ip_off) + i); ip->ip_len = htons(ntohs(ip->ip_len) - i); } m->m_nextpkt = p->m_nextpkt; p->m_nextpkt = m; } else { m->m_nextpkt = fp->ipq_frags; fp->ipq_frags = m; } /* * While we overlap succeeding segments trim them or, * if they are completely covered, dequeue them. */ for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) > ntohs(GETIP(q)->ip_off); q = nq) { i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) - ntohs(GETIP(q)->ip_off); if (i < ntohs(GETIP(q)->ip_len)) { GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i); GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i); m_adj(q, i); q->m_pkthdr.csum_flags = 0; break; } nq = q->m_nextpkt; m->m_nextpkt = nq; IPSTAT_INC(ips_fragdropped); fp->ipq_nfrags--; m_freem(q); } /* * Check for complete reassembly and perform frag per packet * limiting. * * Frag limiting is performed here so that the nth frag has * a chance to complete the packet before we drop the packet. * As a result, n+1 frags are actually allowed per packet, but * only n will ever be stored. (n = maxfragsperpacket.) * */ next = 0; for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { if (ntohs(GETIP(q)->ip_off) != next) { if (fp->ipq_nfrags > V_maxfragsperpacket) { IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); ip_freef(head, fp); } goto done; } next += ntohs(GETIP(q)->ip_len); } /* Make sure the last packet didn't have the IP_MF flag */ if (p->m_flags & M_IP_FRAG) { if (fp->ipq_nfrags > V_maxfragsperpacket) { IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); ip_freef(head, fp); } goto done; } /* * Reassembly is complete. Make sure the packet is a sane size. */ q = fp->ipq_frags; ip = GETIP(q); if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { IPSTAT_INC(ips_toolong); IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags); ip_freef(head, fp); goto done; } /* * Concatenate fragments. */ m = q; t = m->m_next; m->m_next = NULL; m_cat(m, t); nq = q->m_nextpkt; q->m_nextpkt = NULL; for (q = nq; q != NULL; q = nq) { nq = q->m_nextpkt; q->m_nextpkt = NULL; m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; m_cat(m, q); } /* * In order to do checksumming faster we do 'end-around carry' here * (and not in for{} loop), though it implies we are not going to * reassemble more than 64k fragments. */ while (m->m_pkthdr.csum_data & 0xffff0000) m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16); #ifdef MAC mac_ipq_reassemble(fp, m); mac_ipq_destroy(fp); #endif /* * Create header for new ip packet by modifying header of first * packet; dequeue and discard fragment reassembly header. * Make header visible. */ ip->ip_len = htons((ip->ip_hl << 2) + next); ip->ip_src = fp->ipq_src; ip->ip_dst = fp->ipq_dst; TAILQ_REMOVE(head, fp, ipq_list); V_nipq--; uma_zfree(V_ipq_zone, fp); m->m_len += (ip->ip_hl << 2); m->m_data -= (ip->ip_hl << 2); /* some debugging cruft by sklower, below, will go away soon */ if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ m_fixhdr(m); IPSTAT_INC(ips_reassembled); IPQ_UNLOCK(); + +#ifdef RSS + /* + * Query the RSS layer for the flowid / flowtype for the + * mbuf payload. + * + * For now, just assume we have to calculate a new one. + * Later on we should check to see if the assigned flowid matches + * what RSS wants for the given IP protocol and if so, just keep it. + * + * We then queue into the relevant netisr so it can be dispatched + * to the correct CPU. + * + * Note - this may return 1, which means the flowid in the mbuf + * is correct for the configured RSS hash types and can be used. + */ + if (rss_mbuf_software_hash_v4(m, 0, &rss_hash, &rss_type) == 0) { + m->m_pkthdr.flowid = rss_hash; + M_HASHTYPE_SET(m, rss_type); + m->m_flags |= M_FLOWID; + } +#endif + +#ifdef RSS + /* + * Queue/dispatch for reprocessing. + * + * Note: this is much slower than just handling the frame in the + * current receive context. It's likely worth investigating + * why this is. + */ + netisr_dispatch(NETISR_IP_DIRECT, m); + return (NULL); +#endif + + /* Handle in-line */ return (m); dropfrag: IPSTAT_INC(ips_fragdropped); if (fp != NULL) fp->ipq_nfrags--; m_freem(m); done: IPQ_UNLOCK(); return (NULL); #undef GETIP } /* * Free a fragment reassembly header and all * associated datagrams. */ static void ip_freef(struct ipqhead *fhp, struct ipq *fp) { struct mbuf *q; IPQ_LOCK_ASSERT(); while (fp->ipq_frags) { q = fp->ipq_frags; fp->ipq_frags = q->m_nextpkt; m_freem(q); } TAILQ_REMOVE(fhp, fp, ipq_list); uma_zfree(V_ipq_zone, fp); V_nipq--; } /* * IP timer processing; * if a timer expires on a reassembly * queue, discard it. */ void ip_slowtimo(void) { VNET_ITERATOR_DECL(vnet_iter); struct ipq *fp; int i; VNET_LIST_RLOCK_NOSLEEP(); IPQ_LOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); for (i = 0; i < IPREASS_NHASH; i++) { for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) { struct ipq *fpp; fpp = fp; fp = TAILQ_NEXT(fp, ipq_list); if(--fpp->ipq_ttl == 0) { IPSTAT_ADD(ips_fragtimeout, fpp->ipq_nfrags); ip_freef(&V_ipq[i], fpp); } } } /* * If we are over the maximum number of fragments * (due to the limit being lowered), drain off * enough to get down to the new limit. */ if (V_maxnipq >= 0 && V_nipq > V_maxnipq) { for (i = 0; i < IPREASS_NHASH; i++) { while (V_nipq > V_maxnipq && !TAILQ_EMPTY(&V_ipq[i])) { IPSTAT_ADD(ips_fragdropped, TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); } } } CURVNET_RESTORE(); } IPQ_UNLOCK(); VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Drain off all datagram fragments. */ static void ip_drain_locked(void) { int i; IPQ_LOCK_ASSERT(); for (i = 0; i < IPREASS_NHASH; i++) { while(!TAILQ_EMPTY(&V_ipq[i])) { IPSTAT_ADD(ips_fragdropped, TAILQ_FIRST(&V_ipq[i])->ipq_nfrags); ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i])); } } } void ip_drain(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); IPQ_LOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); ip_drain_locked(); CURVNET_RESTORE(); } IPQ_UNLOCK(); VNET_LIST_RUNLOCK_NOSLEEP(); in_rtqdrain(); } /* * The protocol to be inserted into ip_protox[] must be already registered * in inetsw[], either statically or through pf_proto_register(). */ int ipproto_register(short ipproto) { struct protosw *pr; /* Sanity checks. */ if (ipproto <= 0 || ipproto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* * The protocol slot must not be occupied by another protocol * already. An index pointing to IPPROTO_RAW is unused. */ pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ return (EEXIST); /* Find the protocol position in inetsw[] and set the index. */ for (pr = inetdomain.dom_protosw; pr < inetdomain.dom_protoswNPROTOSW; pr++) { if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol && pr->pr_protocol == ipproto) { ip_protox[pr->pr_protocol] = pr - inetsw; return (0); } } return (EPROTONOSUPPORT); } int ipproto_unregister(short ipproto) { struct protosw *pr; /* Sanity checks. */ if (ipproto <= 0 || ipproto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* Check if the protocol was indeed registered. */ pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ return (ENOENT); /* Reset the protocol slot to IPPROTO_RAW. */ ip_protox[ipproto] = pr - inetsw; return (0); } /* * Given address of next destination (final or next hop), return (referenced) * internet address info of interface to be used to get there. */ struct in_ifaddr * ip_rtaddr(struct in_addr dst, u_int fibnum) { struct route sro; struct sockaddr_in *sin; struct in_ifaddr *ia; bzero(&sro, sizeof(sro)); sin = (struct sockaddr_in *)&sro.ro_dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = dst; in_rtalloc_ign(&sro, 0, fibnum); if (sro.ro_rt == NULL) return (NULL); ia = ifatoia(sro.ro_rt->rt_ifa); ifa_ref(&ia->ia_ifa); RTFREE(sro.ro_rt); return (ia); } u_char inetctlerrmap[PRC_NCMDS] = { 0, 0, 0, 0, 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, EMSGSIZE, EHOSTUNREACH, 0, 0, 0, 0, EHOSTUNREACH, 0, ENOPROTOOPT, ECONNREFUSED }; /* * Forward a packet. If some error occurs return the sender * an icmp packet. Note we can't always generate a meaningful * icmp message because icmp doesn't have a large enough repertoire * of codes and types. * * If not forwarding, just drop the packet. This could be confusing * if ipforwarding was zero but some routing protocol was advancing * us as a gateway to somewhere. However, we must let the routing * protocol deal with that. * * The srcrt parameter indicates whether the packet is being forwarded * via a source route. */ void ip_forward(struct mbuf *m, int srcrt) { struct ip *ip = mtod(m, struct ip *); struct in_ifaddr *ia; struct mbuf *mcopy; struct in_addr dest; struct route ro; int error, type = 0, code = 0, mtu = 0; if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { IPSTAT_INC(ips_cantforward); m_freem(m); return; } #ifdef IPSTEALTH if (!V_ipstealth) { #endif if (ip->ip_ttl <= IPTTLDEC) { icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); return; } #ifdef IPSTEALTH } #endif ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); #ifndef IPSEC /* * 'ia' may be NULL if there is no route for this destination. * In case of IPsec, Don't discard it just yet, but pass it to * ip_output in case of outgoing IPsec policy. */ if (!srcrt && ia == NULL) { icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); return; } #endif /* * Save the IP header and at most 8 bytes of the payload, * in case we need to generate an ICMP message to the src. * * XXX this can be optimized a lot by saving the data in a local * buffer on the stack (72 bytes at most), and only allocating the * mbuf if really necessary. The vast majority of the packets * are forwarded without having to send an ICMP back (either * because unnecessary, or because rate limited), so we are * really we are wasting a lot of work here. * * We don't use m_copy() because it might return a reference * to a shared cluster. Both this function and ip_output() * assume exclusive access to the IP header in `m', so any * data in a cluster may change before we reach icmp_error(). */ mcopy = m_gethdr(M_NOWAIT, m->m_type); if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { /* * It's probably ok if the pkthdr dup fails (because * the deep copy of the tag chain failed), but for now * be conservative and just discard the copy since * code below may some day want the tags. */ m_free(mcopy); mcopy = NULL; } if (mcopy != NULL) { mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); mcopy->m_pkthdr.len = mcopy->m_len; m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); } #ifdef IPSTEALTH if (!V_ipstealth) { #endif ip->ip_ttl -= IPTTLDEC; #ifdef IPSTEALTH } #endif /* * If forwarding packet using same interface that it came in on, * perhaps should send a redirect to sender to shortcut a hop. * Only send redirect if source is sending directly to us, * and if packet was not source routed (or has any options). * Also, don't send redirect if forwarding using a default route * or a route modified by a redirect. */ dest.s_addr = 0; if (!srcrt && V_ipsendredirects && ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { struct sockaddr_in *sin; struct rtentry *rt; bzero(&ro, sizeof(ro)); sin = (struct sockaddr_in *)&ro.ro_dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = ip->ip_dst; in_rtalloc_ign(&ro, 0, M_GETFIB(m)); rt = ro.ro_rt; if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && satosin(rt_key(rt))->sin_addr.s_addr != 0) { #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) u_long src = ntohl(ip->ip_src.s_addr); if (RTA(rt) && (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { if (rt->rt_flags & RTF_GATEWAY) dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; else dest.s_addr = ip->ip_dst.s_addr; /* Router requirements says to only send host redirects */ type = ICMP_REDIRECT; code = ICMP_REDIRECT_HOST; } } if (rt) RTFREE(rt); } /* * Try to cache the route MTU from ip_output so we can consider it for * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191. */ bzero(&ro, sizeof(ro)); error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); if (error == EMSGSIZE && ro.ro_rt) mtu = ro.ro_rt->rt_mtu; RO_RTFREE(&ro); if (error) IPSTAT_INC(ips_cantforward); else { IPSTAT_INC(ips_forward); if (type) IPSTAT_INC(ips_redirectsent); else { if (mcopy) m_freem(mcopy); if (ia != NULL) ifa_free(&ia->ia_ifa); return; } } if (mcopy == NULL) { if (ia != NULL) ifa_free(&ia->ia_ifa); return; } switch (error) { case 0: /* forwarded, but need redirect */ /* type, code set above */ break; case ENETUNREACH: case EHOSTUNREACH: case ENETDOWN: case EHOSTDOWN: default: type = ICMP_UNREACH; code = ICMP_UNREACH_HOST; break; case EMSGSIZE: type = ICMP_UNREACH; code = ICMP_UNREACH_NEEDFRAG; #ifdef IPSEC /* * If IPsec is configured for this path, * override any possibly mtu value set by ip_output. */ mtu = ip_ipsec_mtu(mcopy, mtu); #endif /* IPSEC */ /* * If the MTU was set before make sure we are below the * interface MTU. * If the MTU wasn't set before use the interface mtu or * fall back to the next smaller mtu step compared to the * current packet size. */ if (mtu != 0) { if (ia != NULL) mtu = min(mtu, ia->ia_ifp->if_mtu); } else { if (ia != NULL) mtu = ia->ia_ifp->if_mtu; else mtu = ip_next_mtu(ntohs(ip->ip_len), 0); } IPSTAT_INC(ips_cantfrag); break; case ENOBUFS: /* * A router should not generate ICMP_SOURCEQUENCH as * required in RFC1812 Requirements for IP Version 4 Routers. * Source quench could be a big problem under DoS attacks, * or if the underlying interface is rate-limited. * Those who need source quench packets may re-enable them * via the net.inet.ip.sendsourcequench sysctl. */ if (V_ip_sendsourcequench == 0) { m_freem(mcopy); if (ia != NULL) ifa_free(&ia->ia_ifa); return; } else { type = ICMP_SOURCEQUENCH; code = 0; } break; case EACCES: /* ipfw denied packet */ m_freem(mcopy); if (ia != NULL) ifa_free(&ia->ia_ifa); return; } if (ia != NULL) ifa_free(&ia->ia_ifa); icmp_error(mcopy, type, code, dest.s_addr, mtu); } void ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, struct mbuf *m) { if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { struct bintime bt; bintime(&bt); if (inp->inp_socket->so_options & SO_BINTIME) { *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), SCM_BINTIME, SOL_SOCKET); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_socket->so_options & SO_TIMESTAMP) { struct timeval tv; bintime2timeval(&bt, &tv); *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP, SOL_SOCKET); if (*mp) mp = &(*mp)->m_next; } } if (inp->inp_flags & INP_RECVDSTADDR) { *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_flags & INP_RECVTTL) { *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, sizeof(u_char), IP_RECVTTL, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } #ifdef notyet /* XXX * Moving these out of udp_input() made them even more broken * than they already were. */ /* options were tossed already */ if (inp->inp_flags & INP_RECVOPTS) { *mp = sbcreatecontrol((caddr_t)opts_deleted_above, sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } /* ip_srcroute doesn't do what we want here, need to fix */ if (inp->inp_flags & INP_RECVRETOPTS) { *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } #endif if (inp->inp_flags & INP_RECVIF) { struct ifnet *ifp; struct sdlbuf { struct sockaddr_dl sdl; u_char pad[32]; } sdlbuf; struct sockaddr_dl *sdp; struct sockaddr_dl *sdl2 = &sdlbuf.sdl; if ((ifp = m->m_pkthdr.rcvif) && ifp->if_index && ifp->if_index <= V_if_index) { sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; /* * Change our mind and don't try copy. */ if (sdp->sdl_family != AF_LINK || sdp->sdl_len > sizeof(sdlbuf)) { goto makedummy; } bcopy(sdp, sdl2, sdp->sdl_len); } else { makedummy: sdl2->sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); sdl2->sdl_family = AF_LINK; sdl2->sdl_index = 0; sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; } *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len, IP_RECVIF, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_flags & INP_RECVTOS) { *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, sizeof(u_char), IP_RECVTOS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_flags2 & INP_RECVFLOWID) { uint32_t flowid, flow_type; flowid = m->m_pkthdr.flowid; flow_type = M_HASHTYPE_GET(m); /* * XXX should handle the failure of one or the * other - don't populate both? */ *mp = sbcreatecontrol((caddr_t) &flowid, sizeof(uint32_t), IP_FLOWID, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; *mp = sbcreatecontrol((caddr_t) &flow_type, sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } #ifdef RSS if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { uint32_t flowid, flow_type; uint32_t rss_bucketid; flowid = m->m_pkthdr.flowid; flow_type = M_HASHTYPE_GET(m); if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { *mp = sbcreatecontrol((caddr_t) &rss_bucketid, sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } } #endif } /* * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on * locking. This code remains in ip_input.c as ip_mroute.c is optionally * compiled. */ static VNET_DEFINE(int, ip_rsvp_on); VNET_DEFINE(struct socket *, ip_rsvpd); #define V_ip_rsvp_on VNET(ip_rsvp_on) int ip_rsvp_init(struct socket *so) { if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) return EOPNOTSUPP; if (V_ip_rsvpd != NULL) return EADDRINUSE; V_ip_rsvpd = so; /* * This may seem silly, but we need to be sure we don't over-increment * the RSVP counter, in case something slips up. */ if (!V_ip_rsvp_on) { V_ip_rsvp_on = 1; V_rsvp_on++; } return 0; } int ip_rsvp_done(void) { V_ip_rsvpd = NULL; /* * This may seem silly, but we need to be sure we don't over-decrement * the RSVP counter, in case something slips up. */ if (V_ip_rsvp_on) { V_ip_rsvp_on = 0; V_rsvp_on--; } return 0; } int rsvp_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m; m = *mp; *mp = NULL; if (rsvp_input_p) { /* call the real one if loaded */ *mp = m; rsvp_input_p(mp, offp, proto); return (IPPROTO_DONE); } /* Can still get packets with rsvp_on = 0 if there is a local member * of the group to which the RSVP packet is addressed. But in this * case we want to throw the packet away. */ if (!V_rsvp_on) { m_freem(m); return (IPPROTO_DONE); } if (V_ip_rsvpd != NULL) { *mp = m; rip_input(mp, offp, proto); return (IPPROTO_DONE); } /* Drop the packet */ m_freem(m); return (IPPROTO_DONE); }