From owner-svn-src-all@FreeBSD.ORG Tue Mar 4 15:14:48 2014 Return-Path: Delivered-To: svn-src-all@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [8.8.178.115]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 8B7B5438; Tue, 4 Mar 2014 15:14:48 +0000 (UTC) Received: from svn.freebsd.org (svn.freebsd.org [IPv6:2001:1900:2254:2068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.freebsd.org (Postfix) with ESMTPS id 75F24144; Tue, 4 Mar 2014 15:14:48 +0000 (UTC) Received: from svn.freebsd.org ([127.0.1.70]) by svn.freebsd.org (8.14.8/8.14.8) with ESMTP id s24FEm1X063018; Tue, 4 Mar 2014 15:14:48 GMT (envelope-from glebius@svn.freebsd.org) Received: (from glebius@localhost) by svn.freebsd.org (8.14.8/8.14.8/Submit) id s24FEmr0063015; Tue, 4 Mar 2014 15:14:48 GMT (envelope-from glebius@svn.freebsd.org) Message-Id: <201403041514.s24FEmr0063015@svn.freebsd.org> From: Gleb Smirnoff Date: Tue, 4 Mar 2014 15:14:48 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-10@freebsd.org Subject: svn commit: r262743 - in stable/10: sys/conf sys/net sys/netinet sys/netinet6 usr.bin/netstat X-SVN-Group: stable-10 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.17 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 04 Mar 2014 15:14:48 -0000 Author: glebius Date: Tue Mar 4 15:14:47 2014 New Revision: 262743 URL: http://svnweb.freebsd.org/changeset/base/262743 Log: Merge r261582, r261601, r261610, r261613, r261627, r261640, r261641, r261823, r261825, r261859, r261875, r261883, r261911, r262027, r262028, r262029, r262030, r262162 from head. Large flowtable revamp. See commit messages for merged revisions for details. Sponsored by: Netflix Added: stable/10/usr.bin/netstat/flowtable.c - copied, changed from r261601, head/usr.bin/netstat/flowtable.c Modified: stable/10/sys/conf/options stable/10/sys/net/flowtable.c stable/10/sys/net/flowtable.h stable/10/sys/net/route.c stable/10/sys/netinet/ip_input.c stable/10/sys/netinet/ip_output.c stable/10/sys/netinet6/in6_proto.c stable/10/sys/netinet6/ip6_input.c stable/10/sys/netinet6/ip6_output.c stable/10/usr.bin/netstat/Makefile stable/10/usr.bin/netstat/main.c stable/10/usr.bin/netstat/netstat.h Directory Properties: stable/10/ (props changed) Modified: stable/10/sys/conf/options ============================================================================== --- stable/10/sys/conf/options Tue Mar 4 15:09:57 2014 (r262742) +++ stable/10/sys/conf/options Tue Mar 4 15:14:47 2014 (r262743) @@ -438,6 +438,7 @@ TCP_SIGNATURE opt_inet.h VLAN_ARRAY opt_vlan.h XBONEHACK FLOWTABLE opt_route.h +FLOWTABLE_HASH_ALL opt_route.h # # SCTP Modified: stable/10/sys/net/flowtable.c ============================================================================== --- stable/10/sys/net/flowtable.c Tue Mar 4 15:09:57 2014 (r262742) +++ stable/10/sys/net/flowtable.c Tue Mar 4 15:14:47 2014 (r262743) @@ -1,31 +1,30 @@ -/************************************************************************** - -Copyright (c) 2008-2010, BitGravity Inc. -All rights reserved. - -Redistribution and use in source and binary forms, with or without -modification, are permitted provided that the following conditions are met: - - 1. Redistributions of source code must retain the above copyright notice, - this list of conditions and the following disclaimer. - - 2. Neither the name of the BitGravity Corporation 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 COPYRIGHT HOLDERS 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 COPYRIGHT OWNER OR CONTRIBUTORS BE -LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR -CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF -SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS -INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN -CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) -ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE -POSSIBILITY OF SUCH DAMAGE. - -***************************************************************************/ +/*- + * Copyright (c) 2014 Gleb Smirnoff + * Copyright (c) 2008-2010, BitGravity Inc. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * 2. Neither the name of the BitGravity Corporation 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 COPYRIGHT HOLDERS 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 COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ #include "opt_route.h" #include "opt_mpath.h" @@ -36,29 +35,32 @@ POSSIBILITY OF SUCH DAMAGE. #include __FBSDID("$FreeBSD$"); -#include +#include #include #include #include #include #include -#include +#include #include #include #include #include +#include #include +#include #include #include #include #include #include #include +#include #include #include #include -#include +#include #include #include @@ -70,156 +72,79 @@ __FBSDID("$FreeBSD$"); #ifdef INET6 #include #endif +#ifdef FLOWTABLE_HASH_ALL #include #include #include +#endif #include -struct ipv4_tuple { - uint16_t ip_sport; /* source port */ - uint16_t ip_dport; /* destination port */ - in_addr_t ip_saddr; /* source address */ - in_addr_t ip_daddr; /* destination address */ -}; - -union ipv4_flow { - struct ipv4_tuple ipf_ipt; - uint32_t ipf_key[3]; -}; +#ifdef FLOWTABLE_HASH_ALL +#define KEY_PORTS (sizeof(uint16_t) * 2) +#define KEY_ADDRS 2 +#else +#define KEY_PORTS 0 +#define KEY_ADDRS 1 +#endif -struct ipv6_tuple { - uint16_t ip_sport; /* source port */ - uint16_t ip_dport; /* destination port */ - struct in6_addr ip_saddr; /* source address */ - struct in6_addr ip_daddr; /* destination address */ -}; +#ifdef INET6 +#define KEY_ADDR_LEN sizeof(struct in6_addr) +#else +#define KEY_ADDR_LEN sizeof(struct in_addr) +#endif -union ipv6_flow { - struct ipv6_tuple ipf_ipt; - uint32_t ipf_key[9]; -}; +#define KEYLEN ((KEY_ADDR_LEN * KEY_ADDRS + KEY_PORTS) / sizeof(uint32_t)) struct flentry { - volatile uint32_t f_fhash; /* hash flowing forward */ - uint16_t f_flags; /* flow flags */ - uint8_t f_pad; - uint8_t f_proto; /* protocol */ - uint32_t f_fibnum; /* fib index */ + uint32_t f_hash; /* hash flowing forward */ + uint32_t f_key[KEYLEN]; /* address(es and ports) */ uint32_t f_uptime; /* uptime at last access */ - struct flentry *f_next; /* pointer to collision entry */ - volatile struct rtentry *f_rt; /* rtentry for flow */ - volatile struct llentry *f_lle; /* llentry for flow */ -}; - -struct flentry_v4 { - struct flentry fl_entry; - union ipv4_flow fl_flow; -}; - -struct flentry_v6 { - struct flentry fl_entry; - union ipv6_flow fl_flow; -}; - -#define fl_fhash fl_entry.fl_fhash -#define fl_flags fl_entry.fl_flags -#define fl_proto fl_entry.fl_proto -#define fl_uptime fl_entry.fl_uptime -#define fl_rt fl_entry.fl_rt -#define fl_lle fl_entry.fl_lle - -#define SECS_PER_HOUR 3600 -#define SECS_PER_DAY (24*SECS_PER_HOUR) - -#define SYN_IDLE 300 -#define UDP_IDLE 300 -#define FIN_WAIT_IDLE 600 -#define TCP_IDLE SECS_PER_DAY - - -typedef void fl_lock_t(struct flowtable *, uint32_t); -typedef void fl_rtalloc_t(struct route *, uint32_t, u_int); - -union flentryp { - struct flentry **global; - struct flentry **pcpu[MAXCPU]; + uint16_t f_fibnum; /* fib index */ +#ifdef FLOWTABLE_HASH_ALL + uint8_t f_proto; /* protocol */ + uint8_t f_flags; /* stale? */ +#define FL_STALE 1 +#endif + SLIST_ENTRY(flentry) f_next; /* pointer to collision entry */ + struct rtentry *f_rt; /* rtentry for flow */ + struct llentry *f_lle; /* llentry for flow */ }; +#undef KEYLEN -struct flowtable_stats { - uint64_t ft_collisions; - uint64_t ft_allocated; - uint64_t ft_misses; - uint64_t ft_max_depth; - uint64_t ft_free_checks; - uint64_t ft_frees; - uint64_t ft_hits; - uint64_t ft_lookups; -} __aligned(CACHE_LINE_SIZE); +SLIST_HEAD(flist, flentry); +/* Make sure we can use pcpu_zone_ptr for struct flist. */ +CTASSERT(sizeof(struct flist) == sizeof(void *)); struct flowtable { - struct flowtable_stats ft_stats[MAXCPU]; + counter_u64_t *ft_stat; int ft_size; - int ft_lock_count; - uint32_t ft_flags; - char *ft_name; - fl_lock_t *ft_lock; - fl_lock_t *ft_unlock; - fl_rtalloc_t *ft_rtalloc; /* - * XXX need to pad out - */ - struct mtx *ft_locks; - union flentryp ft_table; - bitstr_t *ft_masks[MAXCPU]; + * ft_table is a malloc(9)ed array of pointers. Pointers point to + * memory from UMA_ZONE_PCPU zone. + * ft_masks is per-cpu pointer itself. Each instance points + * to a malloc(9)ed bitset, that is private to corresponding CPU. + */ + struct flist **ft_table; + bitstr_t **ft_masks; bitstr_t *ft_tmpmask; - struct flowtable *ft_next; +}; - uint32_t ft_count __aligned(CACHE_LINE_SIZE); - uint32_t ft_udp_idle __aligned(CACHE_LINE_SIZE); - uint32_t ft_fin_wait_idle; - uint32_t ft_syn_idle; - uint32_t ft_tcp_idle; - boolean_t ft_full; -} __aligned(CACHE_LINE_SIZE); +#define FLOWSTAT_ADD(ft, name, v) \ + counter_u64_add((ft)->ft_stat[offsetof(struct flowtable_stat, name) / sizeof(uint64_t)], (v)) +#define FLOWSTAT_INC(ft, name) FLOWSTAT_ADD(ft, name, 1) static struct proc *flowcleanerproc; -static VNET_DEFINE(struct flowtable *, flow_list_head); -static VNET_DEFINE(uint32_t, flow_hashjitter); -static VNET_DEFINE(uma_zone_t, flow_ipv4_zone); -static VNET_DEFINE(uma_zone_t, flow_ipv6_zone); - -#define V_flow_list_head VNET(flow_list_head) -#define V_flow_hashjitter VNET(flow_hashjitter) -#define V_flow_ipv4_zone VNET(flow_ipv4_zone) -#define V_flow_ipv6_zone VNET(flow_ipv6_zone) - +static uint32_t flow_hashjitter; static struct cv flowclean_f_cv; static struct cv flowclean_c_cv; static struct mtx flowclean_lock; static uint32_t flowclean_cycles; -static uint32_t flowclean_freq; - -#ifdef FLOWTABLE_DEBUG -#define FLDPRINTF(ft, flags, fmt, ...) \ -do { \ - if ((ft)->ft_flags & (flags)) \ - printf((fmt), __VA_ARGS__); \ -} while (0); \ - -#else -#define FLDPRINTF(ft, flags, fmt, ...) - -#endif - /* * TODO: - * - Make flowtable stats per-cpu, aggregated at sysctl call time, - * to avoid extra cache evictions caused by incrementing a shared - * counter - * - add sysctls to resize && flush flow tables + * - add sysctls to resize && flush flow tables * - Add per flowtable sysctls for statistics and configuring timeouts * - add saturation counter to rtentry to support per-packet load-balancing * add flag to indicate round-robin flow, add list lookup from head @@ -230,396 +155,117 @@ do { \ * - support explicit connection state (currently only ad-hoc for DSR) * - idetach() cleanup for options VIMAGE builds. */ -VNET_DEFINE(int, flowtable_enable) = 1; -static VNET_DEFINE(int, flowtable_debug); -static VNET_DEFINE(int, flowtable_syn_expire) = SYN_IDLE; -static VNET_DEFINE(int, flowtable_udp_expire) = UDP_IDLE; -static VNET_DEFINE(int, flowtable_fin_wait_expire) = FIN_WAIT_IDLE; -static VNET_DEFINE(int, flowtable_tcp_expire) = TCP_IDLE; -static VNET_DEFINE(int, flowtable_nmbflows); -static VNET_DEFINE(int, flowtable_ready) = 0; +#ifdef INET +static VNET_DEFINE(struct flowtable, ip4_ft); +#define V_ip4_ft VNET(ip4_ft) +#endif +#ifdef INET6 +static VNET_DEFINE(struct flowtable, ip6_ft); +#define V_ip6_ft VNET(ip6_ft) +#endif +static uma_zone_t flow_zone; + +static VNET_DEFINE(int, flowtable_enable) = 1; #define V_flowtable_enable VNET(flowtable_enable) -#define V_flowtable_debug VNET(flowtable_debug) -#define V_flowtable_syn_expire VNET(flowtable_syn_expire) -#define V_flowtable_udp_expire VNET(flowtable_udp_expire) -#define V_flowtable_fin_wait_expire VNET(flowtable_fin_wait_expire) -#define V_flowtable_tcp_expire VNET(flowtable_tcp_expire) -#define V_flowtable_nmbflows VNET(flowtable_nmbflows) -#define V_flowtable_ready VNET(flowtable_ready) -static SYSCTL_NODE(_net_inet, OID_AUTO, flowtable, CTLFLAG_RD, NULL, +static SYSCTL_NODE(_net, OID_AUTO, flowtable, CTLFLAG_RD, NULL, "flowtable"); -SYSCTL_VNET_INT(_net_inet_flowtable, OID_AUTO, debug, CTLFLAG_RW, - &VNET_NAME(flowtable_debug), 0, "print debug info."); -SYSCTL_VNET_INT(_net_inet_flowtable, OID_AUTO, enable, CTLFLAG_RW, +SYSCTL_VNET_INT(_net_flowtable, OID_AUTO, enable, CTLFLAG_RW, &VNET_NAME(flowtable_enable), 0, "enable flowtable caching."); +SYSCTL_UMA_MAX(_net_flowtable, OID_AUTO, maxflows, CTLFLAG_RW, + &flow_zone, "Maximum number of flows allowed"); -/* - * XXX This does not end up updating timeouts at runtime - * and only reflects the value for the last table added :-/ - */ -SYSCTL_VNET_INT(_net_inet_flowtable, OID_AUTO, syn_expire, CTLFLAG_RW, - &VNET_NAME(flowtable_syn_expire), 0, - "seconds after which to remove syn allocated flow."); -SYSCTL_VNET_INT(_net_inet_flowtable, OID_AUTO, udp_expire, CTLFLAG_RW, - &VNET_NAME(flowtable_udp_expire), 0, - "seconds after which to remove flow allocated to UDP."); -SYSCTL_VNET_INT(_net_inet_flowtable, OID_AUTO, fin_wait_expire, CTLFLAG_RW, - &VNET_NAME(flowtable_fin_wait_expire), 0, - "seconds after which to remove a flow in FIN_WAIT."); -SYSCTL_VNET_INT(_net_inet_flowtable, OID_AUTO, tcp_expire, CTLFLAG_RW, - &VNET_NAME(flowtable_tcp_expire), 0, - "seconds after which to remove flow allocated to a TCP connection."); - - -/* - * Maximum number of flows that can be allocated of a given type. - * - * The table is allocated at boot time (for the pure caching case - * there is no reason why this could not be changed at runtime) - * and thus (currently) needs to be set with a tunable. - */ -static int -sysctl_nmbflows(SYSCTL_HANDLER_ARGS) -{ - int error, newnmbflows; - - newnmbflows = V_flowtable_nmbflows; - error = sysctl_handle_int(oidp, &newnmbflows, 0, req); - if (error == 0 && req->newptr) { - if (newnmbflows > V_flowtable_nmbflows) { - V_flowtable_nmbflows = newnmbflows; - uma_zone_set_max(V_flow_ipv4_zone, - V_flowtable_nmbflows); - uma_zone_set_max(V_flow_ipv6_zone, - V_flowtable_nmbflows); - } else - error = EINVAL; - } - return (error); -} -SYSCTL_VNET_PROC(_net_inet_flowtable, OID_AUTO, nmbflows, - CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_nmbflows, "IU", - "Maximum number of flows allowed"); - - - -#define FS_PRINT(sb, field) sbuf_printf((sb), "\t%s: %jd\n", #field, fs->ft_##field) - -static void -fs_print(struct sbuf *sb, struct flowtable_stats *fs) -{ - - FS_PRINT(sb, collisions); - FS_PRINT(sb, allocated); - FS_PRINT(sb, misses); - FS_PRINT(sb, max_depth); - FS_PRINT(sb, free_checks); - FS_PRINT(sb, frees); - FS_PRINT(sb, hits); - FS_PRINT(sb, lookups); -} - -static void -flowtable_show_stats(struct sbuf *sb, struct flowtable *ft) -{ - int i; - struct flowtable_stats fs, *pfs; - - if (ft->ft_flags & FL_PCPU) { - bzero(&fs, sizeof(fs)); - pfs = &fs; - CPU_FOREACH(i) { - pfs->ft_collisions += ft->ft_stats[i].ft_collisions; - pfs->ft_allocated += ft->ft_stats[i].ft_allocated; - pfs->ft_misses += ft->ft_stats[i].ft_misses; - pfs->ft_free_checks += ft->ft_stats[i].ft_free_checks; - pfs->ft_frees += ft->ft_stats[i].ft_frees; - pfs->ft_hits += ft->ft_stats[i].ft_hits; - pfs->ft_lookups += ft->ft_stats[i].ft_lookups; - if (ft->ft_stats[i].ft_max_depth > pfs->ft_max_depth) - pfs->ft_max_depth = ft->ft_stats[i].ft_max_depth; - } - } else { - pfs = &ft->ft_stats[0]; - } - fs_print(sb, pfs); -} - -static int -sysctl_flowtable_stats(SYSCTL_HANDLER_ARGS) -{ - struct flowtable *ft; - struct sbuf *sb; - int error; - - sb = sbuf_new(NULL, NULL, 64*1024, SBUF_FIXEDLEN); +static MALLOC_DEFINE(M_FTABLE, "flowtable", "flowtable hashes and bitstrings"); - ft = V_flow_list_head; - while (ft != NULL) { - sbuf_printf(sb, "\ntable name: %s\n", ft->ft_name); - flowtable_show_stats(sb, ft); - ft = ft->ft_next; - } - sbuf_finish(sb); - error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); - sbuf_delete(sb); - - return (error); -} -SYSCTL_VNET_PROC(_net_inet_flowtable, OID_AUTO, stats, CTLTYPE_STRING|CTLFLAG_RD, - NULL, 0, sysctl_flowtable_stats, "A", "flowtable statistics"); - - -#ifndef RADIX_MPATH -static void -rtalloc_ign_wrapper(struct route *ro, uint32_t hash, u_int fibnum) -{ - - rtalloc_ign_fib(ro, 0, fibnum); -} -#endif - -static void -flowtable_global_lock(struct flowtable *table, uint32_t hash) -{ - int lock_index = (hash)&(table->ft_lock_count - 1); - - mtx_lock(&table->ft_locks[lock_index]); -} - -static void -flowtable_global_unlock(struct flowtable *table, uint32_t hash) -{ - int lock_index = (hash)&(table->ft_lock_count - 1); - - mtx_unlock(&table->ft_locks[lock_index]); -} - -static void -flowtable_pcpu_lock(struct flowtable *table, uint32_t hash) -{ - - critical_enter(); -} - -static void -flowtable_pcpu_unlock(struct flowtable *table, uint32_t hash) -{ - - critical_exit(); -} - -#define FL_ENTRY_INDEX(table, hash)((hash) % (table)->ft_size) -#define FL_ENTRY(table, hash) *flowtable_entry((table), (hash)) -#define FL_ENTRY_LOCK(table, hash) (table)->ft_lock((table), (hash)) -#define FL_ENTRY_UNLOCK(table, hash) (table)->ft_unlock((table), (hash)) - -#define FL_STALE (1<<8) -#define FL_OVERWRITE (1<<10) - -void -flow_invalidate(struct flentry *fle) -{ - - fle->f_flags |= FL_STALE; -} - -static __inline int -proto_to_flags(uint8_t proto) -{ - int flag; - - switch (proto) { - case IPPROTO_TCP: - flag = FL_TCP; - break; - case IPPROTO_SCTP: - flag = FL_SCTP; - break; - case IPPROTO_UDP: - flag = FL_UDP; - break; - default: - flag = 0; - break; - } - - return (flag); -} - -static __inline int -flags_to_proto(int flags) -{ - int proto, protoflags; - - protoflags = flags & (FL_TCP|FL_SCTP|FL_UDP); - switch (protoflags) { - case FL_TCP: - proto = IPPROTO_TCP; - break; - case FL_SCTP: - proto = IPPROTO_SCTP; - break; - case FL_UDP: - proto = IPPROTO_UDP; - break; - default: - proto = 0; - break; - } - return (proto); -} +static struct flentry * +flowtable_lookup_common(struct flowtable *, uint32_t *, int, uint32_t); #ifdef INET -#ifdef FLOWTABLE_DEBUG -static void -ipv4_flow_print_tuple(int flags, int proto, struct sockaddr_in *ssin, - struct sockaddr_in *dsin) -{ - char saddr[4*sizeof "123"], daddr[4*sizeof "123"]; - - if (flags & FL_HASH_ALL) { - inet_ntoa_r(ssin->sin_addr, saddr); - inet_ntoa_r(dsin->sin_addr, daddr); - printf("proto=%d %s:%d->%s:%d\n", - proto, saddr, ntohs(ssin->sin_port), daddr, - ntohs(dsin->sin_port)); - } else { - inet_ntoa_r(*(struct in_addr *) &dsin->sin_addr, daddr); - printf("proto=%d %s\n", proto, daddr); - } - -} -#endif - -static int -ipv4_mbuf_demarshal(struct flowtable *ft, struct mbuf *m, - struct sockaddr_in *ssin, struct sockaddr_in *dsin, uint16_t *flags) +static struct flentry * +flowtable_lookup_ipv4(struct mbuf *m, struct route *ro) { + struct flentry *fle; + struct sockaddr_in *sin; struct ip *ip; - uint8_t proto; + uint32_t fibnum; +#ifdef FLOWTABLE_HASH_ALL + uint32_t key[3]; int iphlen; - struct tcphdr *th; - struct udphdr *uh; - struct sctphdr *sh; uint16_t sport, dport; + uint8_t proto; +#endif - proto = sport = dport = 0; ip = mtod(m, struct ip *); - dsin->sin_family = AF_INET; - dsin->sin_len = sizeof(*dsin); - dsin->sin_addr = ip->ip_dst; - ssin->sin_family = AF_INET; - ssin->sin_len = sizeof(*ssin); - ssin->sin_addr = ip->ip_src; - proto = ip->ip_p; - if ((*flags & FL_HASH_ALL) == 0) { - FLDPRINTF(ft, FL_DEBUG_ALL, "skip port check flags=0x%x ", - *flags); - goto skipports; - } + if (ip->ip_src.s_addr == ip->ip_dst.s_addr || + (ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || + (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) + return (NULL); - iphlen = ip->ip_hl << 2; /* XXX options? */ + fibnum = M_GETFIB(m); + +#ifdef FLOWTABLE_HASH_ALL + iphlen = ip->ip_hl << 2; + proto = ip->ip_p; switch (proto) { - case IPPROTO_TCP: - th = (struct tcphdr *)((caddr_t)ip + iphlen); + case IPPROTO_TCP: { + struct tcphdr *th; + + th = (struct tcphdr *)((char *)ip + iphlen); sport = th->th_sport; dport = th->th_dport; - if ((*flags & FL_HASH_ALL) && - (th->th_flags & (TH_RST|TH_FIN))) - *flags |= FL_STALE; - break; - case IPPROTO_UDP: - uh = (struct udphdr *)((caddr_t)ip + iphlen); + if (th->th_flags & (TH_RST|TH_FIN)) + fibnum |= (FL_STALE << 24); + break; + } + case IPPROTO_UDP: { + struct udphdr *uh; + + uh = (struct udphdr *)((char *)ip + iphlen); sport = uh->uh_sport; dport = uh->uh_dport; - break; - case IPPROTO_SCTP: - sh = (struct sctphdr *)((caddr_t)ip + iphlen); + break; + } + case IPPROTO_SCTP: { + struct sctphdr *sh; + + sh = (struct sctphdr *)((char *)ip + iphlen); sport = sh->src_port; dport = sh->dest_port; - break; + /* XXXGL: handle stale? */ + break; + } default: - FLDPRINTF(ft, FL_DEBUG_ALL, "proto=0x%x not supported\n", proto); - return (ENOTSUP); - /* no port - hence not a protocol we care about */ + sport = dport = 0; break; - } -skipports: - *flags |= proto_to_flags(proto); - ssin->sin_port = sport; - dsin->sin_port = dport; - return (0); -} - -static uint32_t -ipv4_flow_lookup_hash_internal( - struct sockaddr_in *ssin, struct sockaddr_in *dsin, - uint32_t *key, uint16_t flags) -{ - uint16_t sport, dport; - uint8_t proto; - int offset = 0; + key[0] = ip->ip_dst.s_addr; + key[1] = ip->ip_src.s_addr; + key[2] = (dport << 16) | sport; + fibnum |= proto << 16; - if ((V_flowtable_enable == 0) || (V_flowtable_ready == 0)) - return (0); - proto = flags_to_proto(flags); - sport = dport = key[2] = key[1] = key[0] = 0; - if ((ssin != NULL) && (flags & FL_HASH_ALL)) { - key[1] = ssin->sin_addr.s_addr; - sport = ssin->sin_port; - } - if (dsin != NULL) { - key[2] = dsin->sin_addr.s_addr; - dport = dsin->sin_port; - } - if (flags & FL_HASH_ALL) { - ((uint16_t *)key)[0] = sport; - ((uint16_t *)key)[1] = dport; - } else - offset = V_flow_hashjitter + proto; + fle = flowtable_lookup_common(&V_ip4_ft, key, 3 * sizeof(uint32_t), + fibnum); - return (jenkins_hash32(key, 3, offset)); -} +#else /* !FLOWTABLE_HASH_ALL */ -static struct flentry * -flowtable_lookup_mbuf4(struct flowtable *ft, struct mbuf *m) -{ - struct sockaddr_storage ssa, dsa; - uint16_t flags; - struct sockaddr_in *dsin, *ssin; - - dsin = (struct sockaddr_in *)&dsa; - ssin = (struct sockaddr_in *)&ssa; - bzero(dsin, sizeof(*dsin)); - bzero(ssin, sizeof(*ssin)); - flags = ft->ft_flags; - if (ipv4_mbuf_demarshal(ft, m, ssin, dsin, &flags) != 0) - return (NULL); + fle = flowtable_lookup_common(&V_ip4_ft, (uint32_t *)&ip->ip_dst, + sizeof(struct in_addr), fibnum); - return (flowtable_lookup(ft, &ssa, &dsa, M_GETFIB(m), flags)); -} +#endif /* FLOWTABLE_HASH_ALL */ -void -flow_to_route(struct flentry *fle, struct route *ro) -{ - uint32_t *hashkey = NULL; - struct sockaddr_in *sin; + if (fle == NULL) + return (NULL); sin = (struct sockaddr_in *)&ro->ro_dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); - hashkey = ((struct flentry_v4 *)fle)->fl_flow.ipf_key; - sin->sin_addr.s_addr = hashkey[2]; - ro->ro_rt = __DEVOLATILE(struct rtentry *, fle->f_rt); - ro->ro_lle = __DEVOLATILE(struct llentry *, fle->f_lle); - ro->ro_flags |= RT_NORTREF; + sin->sin_addr = ip->ip_dst; + + return (fle); } #endif /* INET */ @@ -633,9 +279,8 @@ flow_to_route(struct flentry *fle, struc #define PULLUP_TO(_len, p, T) \ do { \ int x = (_len) + sizeof(T); \ - if ((m)->m_len < x) { \ - goto receive_failed; \ - } \ + if ((m)->m_len < x) \ + return (NULL); \ p = (mtod(m, char *) + (_len)); \ } while (0) @@ -643,26 +288,35 @@ do { \ #define SCTP(p) ((struct sctphdr *)(p)) #define UDP(p) ((struct udphdr *)(p)) -static int -ipv6_mbuf_demarshal(struct flowtable *ft, struct mbuf *m, - struct sockaddr_in6 *ssin6, struct sockaddr_in6 *dsin6, uint16_t *flags) +static struct flentry * +flowtable_lookup_ipv6(struct mbuf *m, struct route *ro) { + struct flentry *fle; + struct sockaddr_in6 *sin6; struct ip6_hdr *ip6; - uint8_t proto; + uint32_t fibnum; +#ifdef FLOWTABLE_HASH_ALL + uint32_t key[9]; + void *ulp; int hlen; - uint16_t src_port, dst_port; + uint16_t sport, dport; u_short offset; - void *ulp; + uint8_t proto; +#else + uint32_t key[4]; +#endif - offset = hlen = src_port = dst_port = 0; - ulp = NULL; ip6 = mtod(m, struct ip6_hdr *); - hlen = sizeof(struct ip6_hdr); - proto = ip6->ip6_nxt; + if (in6_localaddr(&ip6->ip6_dst)) + return (NULL); - if ((*flags & FL_HASH_ALL) == 0) - goto skipports; + fibnum = M_GETFIB(m); +#ifdef FLOWTABLE_HASH_ALL + hlen = sizeof(struct ip6_hdr); + proto = ip6->ip6_nxt; + offset = sport = dport = 0; + ulp = NULL; while (ulp == NULL) { switch (proto) { case IPPROTO_ICMPV6: @@ -675,21 +329,21 @@ ipv6_mbuf_demarshal(struct flowtable *ft break; case IPPROTO_TCP: PULLUP_TO(hlen, ulp, struct tcphdr); - dst_port = TCP(ulp)->th_dport; - src_port = TCP(ulp)->th_sport; - if ((*flags & FL_HASH_ALL) && - (TCP(ulp)->th_flags & (TH_RST|TH_FIN))) - *flags |= FL_STALE; + dport = TCP(ulp)->th_dport; + sport = TCP(ulp)->th_sport; + if (TCP(ulp)->th_flags & (TH_RST|TH_FIN)) + fibnum |= (FL_STALE << 24); break; case IPPROTO_SCTP: PULLUP_TO(hlen, ulp, struct sctphdr); - src_port = SCTP(ulp)->src_port; - dst_port = SCTP(ulp)->dest_port; + dport = SCTP(ulp)->src_port; + sport = SCTP(ulp)->dest_port; + /* XXXGL: handle stale? */ break; case IPPROTO_UDP: PULLUP_TO(hlen, ulp, struct udphdr); - dst_port = UDP(ulp)->uh_dport; - src_port = UDP(ulp)->uh_sport; + dport = UDP(ulp)->uh_dport; + sport = UDP(ulp)->uh_sport; break; case IPPROTO_HOPOPTS: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_hbh); @@ -698,7 +352,7 @@ ipv6_mbuf_demarshal(struct flowtable *ft ulp = NULL; break; case IPPROTO_ROUTING: /* RFC 2460 */ - PULLUP_TO(hlen, ulp, struct ip6_rthdr); + PULLUP_TO(hlen, ulp, struct ip6_rthdr); hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; ulp = NULL; @@ -729,689 +383,395 @@ ipv6_mbuf_demarshal(struct flowtable *ft } } - if (src_port == 0) { - receive_failed: - return (ENOTSUP); - } - -skipports: - dsin6->sin6_family = AF_INET6; - dsin6->sin6_len = sizeof(*dsin6); - dsin6->sin6_port = dst_port; - memcpy(&dsin6->sin6_addr, &ip6->ip6_dst, sizeof(struct in6_addr)); - - ssin6->sin6_family = AF_INET6; - ssin6->sin6_len = sizeof(*ssin6); - ssin6->sin6_port = src_port; - memcpy(&ssin6->sin6_addr, &ip6->ip6_src, sizeof(struct in6_addr)); - *flags |= proto_to_flags(proto); - - return (0); -} - -#define zero_key(key) \ -do { \ - key[0] = 0; \ - key[1] = 0; \ - key[2] = 0; \ - key[3] = 0; \ - key[4] = 0; \ - key[5] = 0; \ - key[6] = 0; \ - key[7] = 0; \ - key[8] = 0; \ -} while (0) - -static uint32_t -ipv6_flow_lookup_hash_internal( - struct sockaddr_in6 *ssin6, struct sockaddr_in6 *dsin6, - uint32_t *key, uint16_t flags) -{ - uint16_t sport, dport; - uint8_t proto; - int offset = 0; - - if ((V_flowtable_enable == 0) || (V_flowtable_ready == 0)) - return (0); - - proto = flags_to_proto(flags); - zero_key(key); - sport = dport = 0; - if (dsin6 != NULL) { - memcpy(&key[1], &dsin6->sin6_addr, sizeof(struct in6_addr)); - dport = dsin6->sin6_port; - } - if ((ssin6 != NULL) && (flags & FL_HASH_ALL)) { - memcpy(&key[5], &ssin6->sin6_addr, sizeof(struct in6_addr)); - sport = ssin6->sin6_port; - } - if (flags & FL_HASH_ALL) { - ((uint16_t *)key)[0] = sport; - ((uint16_t *)key)[1] = dport; - } else - offset = V_flow_hashjitter + proto; - - return (jenkins_hash32(key, 9, offset)); -} + bcopy(&ip6->ip6_dst, &key[0], sizeof(struct in6_addr)); + bcopy(&ip6->ip6_src, &key[4], sizeof(struct in6_addr)); + key[8] = (dport << 16) | sport; + fibnum |= proto << 16; + + fle = flowtable_lookup_common(&V_ip6_ft, key, 9 * sizeof(uint32_t), + fibnum); +#else /* !FLOWTABLE_HASH_ALL */ + bcopy(&ip6->ip6_dst, &key[0], sizeof(struct in6_addr)); + fle = flowtable_lookup_common(&V_ip6_ft, key, sizeof(struct in6_addr), + fibnum); +#endif /* FLOWTABLE_HASH_ALL */ -static struct flentry * -flowtable_lookup_mbuf6(struct flowtable *ft, struct mbuf *m) -{ - struct sockaddr_storage ssa, dsa; - struct sockaddr_in6 *dsin6, *ssin6; - uint16_t flags; - - dsin6 = (struct sockaddr_in6 *)&dsa; - ssin6 = (struct sockaddr_in6 *)&ssa; - bzero(dsin6, sizeof(*dsin6)); - bzero(ssin6, sizeof(*ssin6)); - flags = ft->ft_flags; - - if (ipv6_mbuf_demarshal(ft, m, ssin6, dsin6, &flags) != 0) + if (fle == NULL) return (NULL); - return (flowtable_lookup(ft, &ssa, &dsa, M_GETFIB(m), flags)); -} - -void -flow_to_route_in6(struct flentry *fle, struct route_in6 *ro) -{ - uint32_t *hashkey = NULL; - struct sockaddr_in6 *sin6; - sin6 = (struct sockaddr_in6 *)&ro->ro_dst; - sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); - hashkey = ((struct flentry_v6 *)fle)->fl_flow.ipf_key; - memcpy(&sin6->sin6_addr, &hashkey[5], sizeof (struct in6_addr)); - ro->ro_rt = __DEVOLATILE(struct rtentry *, fle->f_rt); - ro->ro_lle = __DEVOLATILE(struct llentry *, fle->f_lle); - ro->ro_flags |= RT_NORTREF; + bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(struct in6_addr)); + + return (fle); } #endif /* INET6 */ *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***