diff -ur v2.6.14/linux/include/linux/netfilter_ipv4/ip_nat.h linux/include/linux/netfilter_ipv4/ip_nat.h --- v2.6.14/linux/include/linux/netfilter_ipv4/ip_nat.h 2005-10-29 14:15:09.000000000 +0300 +++ linux/include/linux/netfilter_ipv4/ip_nat.h 2005-10-29 18:11:32.885759304 +0300 @@ -63,6 +63,13 @@ struct ip_conntrack; +/* Call input routing for SNAT-ed traffic */ +extern unsigned int ip_nat_route_input(unsigned int hooknum, + struct sk_buff **pskb, + const struct net_device *in, + const struct net_device *out, + int (*okfn)(struct sk_buff *)); + /* Set up the info structure to map into this range. */ extern unsigned int ip_nat_setup_info(struct ip_conntrack *conntrack, const struct ip_nat_range *range, diff -ur v2.6.14/linux/include/linux/rtnetlink.h linux/include/linux/rtnetlink.h --- v2.6.14/linux/include/linux/rtnetlink.h 2005-10-29 14:15:09.000000000 +0300 +++ linux/include/linux/rtnetlink.h 2005-10-29 18:11:21.299520680 +0300 @@ -292,6 +292,8 @@ #define RTNH_F_DEAD 1 /* Nexthop is dead (used by multipath) */ #define RTNH_F_PERVASIVE 2 /* Do recursive gateway lookup */ #define RTNH_F_ONLINK 4 /* Gateway is forced on link */ +#define RTNH_F_SUSPECT 8 /* We don't know the real state */ +#define RTNH_F_BADSTATE (RTNH_F_DEAD | RTNH_F_SUSPECT) /* Macros to handle hexthops */ diff -ur v2.6.14/linux/include/net/flow.h linux/include/net/flow.h --- v2.6.14/linux/include/net/flow.h 2005-06-18 08:50:52.000000000 +0300 +++ linux/include/net/flow.h 2005-10-29 18:11:32.885759304 +0300 @@ -19,6 +19,8 @@ __u32 daddr; __u32 saddr; __u32 fwmark; + __u32 lsrc; + __u32 gw; __u8 tos; __u8 scope; } ip4_u; @@ -46,6 +48,8 @@ #define fl4_dst nl_u.ip4_u.daddr #define fl4_src nl_u.ip4_u.saddr #define fl4_fwmark nl_u.ip4_u.fwmark +#define fl4_lsrc nl_u.ip4_u.lsrc +#define fl4_gw nl_u.ip4_u.gw #define fl4_tos nl_u.ip4_u.tos #define fl4_scope nl_u.ip4_u.scope diff -ur v2.6.14/linux/include/net/ip_fib.h linux/include/net/ip_fib.h --- v2.6.14/linux/include/net/ip_fib.h 2005-10-29 14:15:09.000000000 +0300 +++ linux/include/net/ip_fib.h 2005-10-29 18:11:21.300520528 +0300 @@ -195,7 +195,8 @@ static inline void fib_select_default(const struct flowi *flp, struct fib_result *res) { - if (FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) + if ((FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) || + FIB_RES_NH(*res).nh_scope == RT_SCOPE_HOST) ip_fib_main_table->tb_select_default(ip_fib_main_table, flp, res); } @@ -207,6 +208,7 @@ extern int fib_lookup(const struct flowi *flp, struct fib_result *res); extern struct fib_table *__fib_new_table(int id); extern void fib_rule_put(struct fib_rule *r); +extern int fib_result_table(struct fib_result *res); static inline struct fib_table *fib_get_table(int id) { @@ -300,4 +302,6 @@ extern void fib_proc_exit(void); #endif +extern rwlock_t fib_nhflags_lock; + #endif /* _NET_FIB_H */ diff -ur v2.6.14/linux/include/net/route.h linux/include/net/route.h --- v2.6.14/linux/include/net/route.h 2005-10-29 14:15:09.000000000 +0300 +++ linux/include/net/route.h 2005-10-29 18:11:32.885759304 +0300 @@ -117,6 +117,7 @@ extern int ip_route_output_key(struct rtable **, struct flowi *flp); extern int ip_route_output_flow(struct rtable **rp, struct flowi *flp, struct sock *sk, int flags); extern int ip_route_input(struct sk_buff*, u32 dst, u32 src, u8 tos, struct net_device *devin); +extern int ip_route_input_lookup(struct sk_buff*, u32 dst, u32 src, u8 tos, struct net_device *devin, u32 lsrc); extern unsigned short ip_rt_frag_needed(struct iphdr *iph, unsigned short new_mtu); extern void ip_rt_send_redirect(struct sk_buff *skb); diff -ur v2.6.14/linux/net/ipv4/fib_frontend.c linux/net/ipv4/fib_frontend.c --- v2.6.14/linux/net/ipv4/fib_frontend.c 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/fib_frontend.c 2005-10-29 18:11:21.300520528 +0300 @@ -54,6 +54,8 @@ struct fib_table *ip_fib_local_table; struct fib_table *ip_fib_main_table; +#define FIB_RES_TABLE(r) (RT_TABLE_MAIN) + #else #define RT_TABLE_MIN 1 @@ -71,6 +73,7 @@ return tb; } +#define FIB_RES_TABLE(r) (fib_result_table(r)) #endif /* CONFIG_IP_MULTIPLE_TABLES */ @@ -168,6 +171,9 @@ .tos = tos } }, .iif = oif }; struct fib_result res; + int table; + unsigned char prefixlen; + unsigned char scope; int no_addr, rpf; int ret; @@ -189,31 +195,35 @@ goto e_inval_res; *spec_dst = FIB_RES_PREFSRC(res); fib_combine_itag(itag, &res); -#ifdef CONFIG_IP_ROUTE_MULTIPATH - if (FIB_RES_DEV(res) == dev || res.fi->fib_nhs > 1) -#else if (FIB_RES_DEV(res) == dev) -#endif { ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; fib_res_put(&res); return ret; } + table = FIB_RES_TABLE(&res); + prefixlen = res.prefixlen; + scope = res.scope; fib_res_put(&res); if (no_addr) goto last_resort; - if (rpf) - goto e_inval; fl.oif = dev->ifindex; ret = 0; if (fib_lookup(&fl, &res) == 0) { - if (res.type == RTN_UNICAST) { + if (res.type == RTN_UNICAST && + ((table == FIB_RES_TABLE(&res) && + res.prefixlen >= prefixlen && res.scope >= scope) || + !rpf)) { *spec_dst = FIB_RES_PREFSRC(res); ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST; + fib_res_put(&res); + return ret; } fib_res_put(&res); } + if (rpf) + goto e_inval; return ret; last_resort: @@ -584,9 +594,7 @@ switch (event) { case NETDEV_UP: fib_add_ifaddr(ifa); -#ifdef CONFIG_IP_ROUTE_MULTIPATH fib_sync_up(ifa->ifa_dev->dev); -#endif rt_cache_flush(-1); break; case NETDEV_DOWN: @@ -622,9 +630,7 @@ for_ifa(in_dev) { fib_add_ifaddr(ifa); } endfor_ifa(in_dev); -#ifdef CONFIG_IP_ROUTE_MULTIPATH fib_sync_up(dev); -#endif rt_cache_flush(-1); break; case NETDEV_DOWN: diff -ur v2.6.14/linux/net/ipv4/fib_hash.c linux/net/ipv4/fib_hash.c --- v2.6.14/linux/net/ipv4/fib_hash.c 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/fib_hash.c 2005-10-29 18:11:21.301520376 +0300 @@ -276,30 +276,38 @@ return err; } -static int fn_hash_last_dflt=-1; - static void fn_hash_select_default(struct fib_table *tb, const struct flowi *flp, struct fib_result *res) { - int order, last_idx; + int order, last_idx, last_dflt, last_nhsel; + struct fib_alias *first_fa = NULL; + struct hlist_head *head; struct hlist_node *node; struct fib_node *f; struct fib_info *fi = NULL; struct fib_info *last_resort; struct fn_hash *t = (struct fn_hash*)tb->tb_data; - struct fn_zone *fz = t->fn_zones[0]; + struct fn_zone *fz = t->fn_zones[res->prefixlen]; + u32 k; if (fz == NULL) return; + k = fz_key(flp->fl4_dst, fz); + last_dflt = -2; + last_nhsel = 0; last_idx = -1; last_resort = NULL; order = -1; read_lock(&fib_hash_lock); - hlist_for_each_entry(f, node, &fz->fz_hash[0], fn_hash) { + head = &fz->fz_hash[fn_hash(k, fz)]; + hlist_for_each_entry(f, node, head, fn_hash) { struct fib_alias *fa; + if (f->fn_key != k) + continue; + list_for_each_entry(fa, &f->fn_alias, fa_list) { struct fib_info *next_fi = fa->fa_info; @@ -307,41 +315,52 @@ fa->fa_type != RTN_UNICAST) continue; + if (fa->fa_tos && + fa->fa_tos != flp->fl4_tos) + continue; if (next_fi->fib_priority > res->fi->fib_priority) break; - if (!next_fi->fib_nh[0].nh_gw || - next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK) - continue; fa->fa_state |= FA_S_ACCESSED; - if (fi == NULL) { - if (next_fi != res->fi) - break; - } else if (!fib_detect_death(fi, order, &last_resort, - &last_idx, &fn_hash_last_dflt)) { + if (!first_fa) { + last_dflt = fa->fa_last_dflt; + first_fa = fa; + } + if (fi && !fib_detect_death(fi, order, &last_resort, + &last_idx, &last_dflt, &last_nhsel, flp)) { if (res->fi) fib_info_put(res->fi); res->fi = fi; atomic_inc(&fi->fib_clntref); - fn_hash_last_dflt = order; + first_fa->fa_last_dflt = order; goto out; } fi = next_fi; order++; } + break; } if (order <= 0 || fi == NULL) { - fn_hash_last_dflt = -1; + if (fi && fi->fib_nhs > 1 && + fib_detect_death(fi, order, &last_resort, &last_idx, + &last_dflt, &last_nhsel, flp) && + last_resort == fi) { + read_lock_bh(&fib_nhflags_lock); + fi->fib_nh[last_nhsel].nh_flags &= ~RTNH_F_SUSPECT; + read_unlock_bh(&fib_nhflags_lock); + } + if (first_fa) first_fa->fa_last_dflt = -1; goto out; } - if (!fib_detect_death(fi, order, &last_resort, &last_idx, &fn_hash_last_dflt)) { + if (!fib_detect_death(fi, order, &last_resort, &last_idx, + &last_dflt, &last_nhsel, flp)) { if (res->fi) fib_info_put(res->fi); res->fi = fi; atomic_inc(&fi->fib_clntref); - fn_hash_last_dflt = order; + first_fa->fa_last_dflt = order; goto out; } @@ -351,8 +370,11 @@ res->fi = last_resort; if (last_resort) atomic_inc(&last_resort->fib_clntref); + read_lock_bh(&fib_nhflags_lock); + last_resort->fib_nh[last_nhsel].nh_flags &= ~RTNH_F_SUSPECT; + read_unlock_bh(&fib_nhflags_lock); + first_fa->fa_last_dflt = last_idx; } - fn_hash_last_dflt = last_idx; out: read_unlock(&fib_hash_lock); } @@ -451,6 +473,7 @@ write_lock_bh(&fib_hash_lock); fi_drop = fa->fa_info; fa->fa_info = fi; + fa->fa_last_dflt = -1; fa->fa_type = type; fa->fa_scope = r->rtm_scope; state = fa->fa_state; @@ -510,6 +533,7 @@ new_fa->fa_type = type; new_fa->fa_scope = r->rtm_scope; new_fa->fa_state = 0; + new_fa->fa_last_dflt = -1; /* * Insert new entry to the list. diff -ur v2.6.14/linux/net/ipv4/fib_lookup.h linux/net/ipv4/fib_lookup.h --- v2.6.14/linux/net/ipv4/fib_lookup.h 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/fib_lookup.h 2005-10-29 18:11:21.302520224 +0300 @@ -9,6 +9,7 @@ struct list_head fa_list; struct rcu_head rcu; struct fib_info *fa_info; + int fa_last_dflt; u8 fa_tos; u8 fa_type; u8 fa_scope; @@ -40,6 +41,7 @@ u8 tos, u32 prio); extern int fib_detect_death(struct fib_info *fi, int order, struct fib_info **last_resort, - int *last_idx, int *dflt); + int *last_idx, int *dflt, int *last_nhsel, + const struct flowi *flp); #endif /* _FIB_LOOKUP_H */ diff -ur v2.6.14/linux/net/ipv4/fib_rules.c linux/net/ipv4/fib_rules.c --- v2.6.14/linux/net/ipv4/fib_rules.c 2005-08-29 07:51:29.000000000 +0300 +++ linux/net/ipv4/fib_rules.c 2005-10-29 18:11:21.302520224 +0300 @@ -280,6 +280,11 @@ } } +int fib_result_table(struct fib_result *res) +{ + return res->r->r_table; +} + int fib_lookup(const struct flowi *flp, struct fib_result *res) { int err; @@ -342,7 +347,8 @@ void fib_select_default(const struct flowi *flp, struct fib_result *res) { if (res->r && res->r->r_action == RTN_UNICAST && - FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) { + ((FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) || + FIB_RES_NH(*res).nh_scope == RT_SCOPE_HOST)) { struct fib_table *tb; if ((tb = fib_get_table(res->r->r_table)) != NULL) tb->tb_select_default(tb, flp, res); diff -ur v2.6.14/linux/net/ipv4/fib_semantics.c linux/net/ipv4/fib_semantics.c --- v2.6.14/linux/net/ipv4/fib_semantics.c 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/fib_semantics.c 2005-10-29 18:11:32.886759152 +0300 @@ -53,6 +53,7 @@ static struct hlist_head *fib_info_laddrhash; static unsigned int fib_hash_size; static unsigned int fib_info_cnt; +rwlock_t fib_nhflags_lock = RW_LOCK_UNLOCKED; #define DEVINDEX_HASHBITS 8 #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS) @@ -188,7 +189,7 @@ #ifdef CONFIG_NET_CLS_ROUTE nh->nh_tclassid != onh->nh_tclassid || #endif - ((nh->nh_flags^onh->nh_flags)&~RTNH_F_DEAD)) + ((nh->nh_flags^onh->nh_flags)&~RTNH_F_BADSTATE)) return -1; onh++; } endfor_nexthops(fi); @@ -225,7 +226,7 @@ nfi->fib_priority == fi->fib_priority && memcmp(nfi->fib_metrics, fi->fib_metrics, sizeof(fi->fib_metrics)) == 0 && - ((nfi->fib_flags^fi->fib_flags)&~RTNH_F_DEAD) == 0 && + ((nfi->fib_flags^fi->fib_flags)&~RTNH_F_BADSTATE) == 0 && (nfi->fib_nhs == 0 || nh_comp(fi, nfi) == 0)) return fi; } @@ -317,26 +318,70 @@ } int fib_detect_death(struct fib_info *fi, int order, - struct fib_info **last_resort, int *last_idx, int *dflt) + struct fib_info **last_resort, int *last_idx, int *dflt, + int *last_nhsel, const struct flowi *flp) { struct neighbour *n; - int state = NUD_NONE; + int nhsel; + int state; + struct fib_nh * nh; + u32 dst; + int flag, dead = 1; + + /* change_nexthops(fi) { */ + for (nhsel = 0, nh = fi->fib_nh; nhsel < fi->fib_nhs; nh++, nhsel++) { + if (flp->oif && flp->oif != nh->nh_oif) + continue; + if (flp->fl4_gw && flp->fl4_gw != nh->nh_gw && nh->nh_gw && + nh->nh_scope == RT_SCOPE_LINK) + continue; + if (nh->nh_flags & RTNH_F_DEAD) + continue; - n = neigh_lookup(&arp_tbl, &fi->fib_nh[0].nh_gw, fi->fib_dev); - if (n) { - state = n->nud_state; - neigh_release(n); - } - if (state==NUD_REACHABLE) - return 0; - if ((state&NUD_VALID) && order != *dflt) - return 0; - if ((state&NUD_VALID) || - (*last_idx<0 && order > *dflt)) { - *last_resort = fi; - *last_idx = order; + flag = 0; + if (nh->nh_dev->flags & IFF_NOARP) { + dead = 0; + goto setfl; + } + + dst = nh->nh_gw; + if (!nh->nh_gw || nh->nh_scope != RT_SCOPE_LINK) + dst = flp->fl4_dst; + + state = NUD_NONE; + n = neigh_lookup(&arp_tbl, &dst, nh->nh_dev); + if (n) { + state = n->nud_state; + neigh_release(n); + } + if (state==NUD_REACHABLE || + ((state&NUD_VALID) && order != *dflt)) { + dead = 0; + goto setfl; + } + if (!(state&NUD_VALID)) + flag = 1; + if (!dead) + goto setfl; + if ((state&NUD_VALID) || + (*last_idx<0 && order >= *dflt)) { + *last_resort = fi; + *last_idx = order; + *last_nhsel = nhsel; + } + + setfl: + + read_lock_bh(&fib_nhflags_lock); + if (flag) + nh->nh_flags |= RTNH_F_SUSPECT; + else + nh->nh_flags &= ~RTNH_F_SUSPECT; + read_unlock_bh(&fib_nhflags_lock); } - return 1; + /* } endfor_nexthops(fi) */ + + return dead; } #ifdef CONFIG_IP_ROUTE_MULTIPATH @@ -507,8 +552,11 @@ return -EINVAL; if ((dev = __dev_get_by_index(nh->nh_oif)) == NULL) return -ENODEV; - if (!(dev->flags&IFF_UP)) - return -ENETDOWN; + if (!(dev->flags&IFF_UP)) { + if (fi->fib_protocol != RTPROT_STATIC) + return -ENETDOWN; + nh->nh_flags |= RTNH_F_DEAD; + } nh->nh_dev = dev; dev_hold(dev); nh->nh_scope = RT_SCOPE_LINK; @@ -523,24 +571,48 @@ /* It is not necessary, but requires a bit of thinking */ if (fl.fl4_scope < RT_SCOPE_LINK) fl.fl4_scope = RT_SCOPE_LINK; - if ((err = fib_lookup(&fl, &res)) != 0) - return err; + err = fib_lookup(&fl, &res); } - err = -EINVAL; - if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) - goto out; - nh->nh_scope = res.scope; - nh->nh_oif = FIB_RES_OIF(res); - if ((nh->nh_dev = FIB_RES_DEV(res)) == NULL) - goto out; - dev_hold(nh->nh_dev); - err = -ENETDOWN; - if (!(nh->nh_dev->flags & IFF_UP)) - goto out; - err = 0; + if (err) { + struct in_device *in_dev; + + if (err != -ENETUNREACH || + fi->fib_protocol != RTPROT_STATIC) + return err; + + in_dev = inetdev_by_index(nh->nh_oif); + if (in_dev == NULL || + in_dev->dev->flags & IFF_UP) { + if (in_dev) + in_dev_put(in_dev); + return err; + } + nh->nh_flags |= RTNH_F_DEAD; + nh->nh_scope = RT_SCOPE_LINK; + nh->nh_dev = in_dev->dev; + dev_hold(nh->nh_dev); + in_dev_put(in_dev); + } else { + err = -EINVAL; + if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) + goto out; + nh->nh_scope = res.scope; + nh->nh_oif = FIB_RES_OIF(res); + if ((nh->nh_dev = FIB_RES_DEV(res)) == NULL) + goto out; + dev_hold(nh->nh_dev); + if (!(nh->nh_dev->flags & IFF_UP)) { + if (fi->fib_protocol != RTPROT_STATIC) { + err = -ENETDOWN; + goto out; + } + nh->nh_flags |= RTNH_F_DEAD; + } + err = 0; out: - fib_res_put(&res); - return err; + fib_res_put(&res); + return err; + } } else { struct in_device *in_dev; @@ -551,8 +623,11 @@ if (in_dev == NULL) return -ENODEV; if (!(in_dev->dev->flags&IFF_UP)) { - in_dev_put(in_dev); - return -ENETDOWN; + if (fi->fib_protocol != RTPROT_STATIC) { + in_dev_put(in_dev); + return -ENETDOWN; + } + nh->nh_flags |= RTNH_F_DEAD; } nh->nh_dev = in_dev->dev; dev_hold(nh->nh_dev); @@ -890,8 +965,12 @@ for_nexthops(fi) { if (nh->nh_flags&RTNH_F_DEAD) continue; - if (!flp->oif || flp->oif == nh->nh_oif) - break; + if (flp->oif && flp->oif != nh->nh_oif) + continue; + if (flp->fl4_gw && flp->fl4_gw != nh->nh_gw && + nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) + continue; + break; } #ifdef CONFIG_IP_ROUTE_MULTIPATH if (nhsel < fi->fib_nhs) { @@ -1197,18 +1276,29 @@ prev_fi = fi; dead = 0; change_nexthops(fi) { - if (nh->nh_flags&RTNH_F_DEAD) - dead++; - else if (nh->nh_dev == dev && - nh->nh_scope != scope) { - nh->nh_flags |= RTNH_F_DEAD; + if (nh->nh_flags&RTNH_F_DEAD) { + if (fi->fib_protocol!=RTPROT_STATIC || + nh->nh_dev == NULL || + __in_dev_get_rtnl(nh->nh_dev) == NULL || + nh->nh_dev->flags&IFF_UP) + dead++; + } else if (nh->nh_dev == dev && + nh->nh_scope != scope) { + write_lock_bh(&fib_nhflags_lock); #ifdef CONFIG_IP_ROUTE_MULTIPATH - spin_lock_bh(&fib_multipath_lock); + spin_lock(&fib_multipath_lock); + nh->nh_flags |= RTNH_F_DEAD; fi->fib_power -= nh->nh_power; nh->nh_power = 0; - spin_unlock_bh(&fib_multipath_lock); + spin_unlock(&fib_multipath_lock); +#else + nh->nh_flags |= RTNH_F_DEAD; #endif - dead++; + write_unlock_bh(&fib_nhflags_lock); + if (fi->fib_protocol!=RTPROT_STATIC || + force || + __in_dev_get_rtnl(dev) == NULL) + dead++; } #ifdef CONFIG_IP_ROUTE_MULTIPATH if (force > 1 && nh->nh_dev == dev) { @@ -1227,11 +1317,8 @@ return ret; } -#ifdef CONFIG_IP_ROUTE_MULTIPATH - /* - Dead device goes up. We wake up dead nexthops. - It takes sense only on multipath routes. + Dead device goes up or new address is added. We wake up dead nexthops. */ int fib_sync_up(struct net_device *dev) @@ -1241,8 +1328,10 @@ struct hlist_head *head; struct hlist_node *node; struct fib_nh *nh; - int ret; + struct fib_result res; + int ret, rep; +repeat: if (!(dev->flags&IFF_UP)) return 0; @@ -1250,6 +1339,7 @@ hash = fib_devindex_hashfn(dev->ifindex); head = &fib_info_devhash[hash]; ret = 0; + rep = 0; hlist_for_each_entry(nh, node, head, nh_hash) { struct fib_info *fi = nh->nh_parent; @@ -1262,19 +1352,37 @@ prev_fi = fi; alive = 0; change_nexthops(fi) { - if (!(nh->nh_flags&RTNH_F_DEAD)) { - alive++; + if (!(nh->nh_flags&RTNH_F_DEAD)) continue; - } if (nh->nh_dev == NULL || !(nh->nh_dev->flags&IFF_UP)) continue; if (nh->nh_dev != dev || !__in_dev_get_rtnl(dev)) continue; + if (nh->nh_gw && fi->fib_protocol == RTPROT_STATIC) { + struct flowi fl = { + .nl_u = { .ip4_u = + { .daddr = nh->nh_gw, + .scope = nh->nh_scope } }, + .oif = nh->nh_oif, + }; + if (fib_lookup(&fl, &res) != 0) + continue; + if (res.type != RTN_UNICAST && + res.type != RTN_LOCAL) { + fib_res_put(&res); + continue; + } + nh->nh_scope = res.scope; + fib_res_put(&res); + rep = 1; + } alive++; +#ifdef CONFIG_IP_ROUTE_MULTIPATH spin_lock_bh(&fib_multipath_lock); nh->nh_power = 0; nh->nh_flags &= ~RTNH_F_DEAD; spin_unlock_bh(&fib_multipath_lock); +#endif } endfor_nexthops(fi) if (alive > 0) { @@ -1282,10 +1390,14 @@ ret++; } } + if (rep) + goto repeat; return ret; } +#ifdef CONFIG_IP_ROUTE_MULTIPATH + /* The algorithm is suboptimal, but it provides really fair weighted route distribution. @@ -1294,24 +1406,45 @@ void fib_select_multipath(const struct flowi *flp, struct fib_result *res) { struct fib_info *fi = res->fi; - int w; + int w, alive; spin_lock_bh(&fib_multipath_lock); + if (flp->oif) { + int sel = -1; + w = -1; + change_nexthops(fi) { + if (flp->oif != nh->nh_oif) + continue; + if (flp->fl4_gw && flp->fl4_gw != nh->nh_gw && + nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) + continue; + if (!(nh->nh_flags&RTNH_F_BADSTATE)) { + if (nh->nh_power > w) { + w = nh->nh_power; + sel = nhsel; + } + } + } endfor_nexthops(fi); + if (sel >= 0) { + spin_unlock_bh(&fib_multipath_lock); + res->nh_sel = sel; + return; + } + goto last_resort; + } + +repeat: if (fi->fib_power <= 0) { int power = 0; change_nexthops(fi) { - if (!(nh->nh_flags&RTNH_F_DEAD)) { + if (!(nh->nh_flags&RTNH_F_BADSTATE)) { power += nh->nh_weight; nh->nh_power = nh->nh_weight; } } endfor_nexthops(fi); fi->fib_power = power; - if (power <= 0) { - spin_unlock_bh(&fib_multipath_lock); - /* Race condition: route has just become dead. */ - res->nh_sel = 0; - return; - } + if (power <= 0) + goto last_resort; } @@ -1321,20 +1454,40 @@ w = jiffies % fi->fib_power; + alive = 0; change_nexthops(fi) { - if (!(nh->nh_flags&RTNH_F_DEAD) && nh->nh_power) { + if (!(nh->nh_flags&RTNH_F_BADSTATE) && nh->nh_power) { if ((w -= nh->nh_power) <= 0) { nh->nh_power--; fi->fib_power--; - res->nh_sel = nhsel; spin_unlock_bh(&fib_multipath_lock); + res->nh_sel = nhsel; return; } + alive = 1; + } + } endfor_nexthops(fi); + if (alive) { + fi->fib_power = 0; + goto repeat; + } + +last_resort: + + for_nexthops(fi) { + if (!(nh->nh_flags&RTNH_F_DEAD)) { + if (flp->oif && flp->oif != nh->nh_oif) + continue; + if (flp->fl4_gw && flp->fl4_gw != nh->nh_gw && + nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) + continue; + spin_unlock_bh(&fib_multipath_lock); + res->nh_sel = nhsel; + return; } } endfor_nexthops(fi); /* Race condition: route has just become dead. */ - res->nh_sel = 0; spin_unlock_bh(&fib_multipath_lock); } #endif diff -ur v2.6.14/linux/net/ipv4/netfilter/ip_nat_core.c linux/net/ipv4/netfilter/ip_nat_core.c --- v2.6.14/linux/net/ipv4/netfilter/ip_nat_core.c 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/netfilter/ip_nat_core.c 2005-10-29 18:11:32.887759000 +0300 @@ -591,6 +591,53 @@ EXPORT_SYMBOL_GPL(ip_nat_port_range_to_nfattr); #endif +unsigned int +ip_nat_route_input(unsigned int hooknum, + struct sk_buff **pskb, + const struct net_device *in, + const struct net_device *out, + int (*okfn)(struct sk_buff *)) +{ + struct sk_buff *skb = *pskb; + struct iphdr *iph; + struct ip_conntrack *conn; + enum ip_conntrack_info ctinfo; + enum ip_conntrack_dir dir; + unsigned long statusbit; + u32 saddr; + + if (!(conn = ip_conntrack_get(skb, &ctinfo))) + return NF_ACCEPT; + + if (!(conn->status & IPS_NAT_DONE_MASK)) + return NF_ACCEPT; + dir = CTINFO2DIR(ctinfo); + statusbit = IPS_SRC_NAT; + if (dir == IP_CT_DIR_REPLY) + statusbit ^= IPS_NAT_MASK; + if (!(conn->status & statusbit)) + return NF_ACCEPT; + + if (skb->dst) + return NF_ACCEPT; + + if (skb->len < sizeof(struct iphdr)) + return NF_ACCEPT; + + /* use daddr in other direction as masquerade address (lsrc) */ + iph = skb->nh.iph; + saddr = conn->tuplehash[!dir].tuple.dst.ip; + if (saddr == iph->saddr) + return NF_ACCEPT; + + if (ip_route_input_lookup(skb, iph->daddr, iph->saddr, iph->tos, + skb->dev, saddr)) + return NF_DROP; + + return NF_ACCEPT; +} +EXPORT_SYMBOL_GPL(ip_nat_route_input); + static int __init ip_nat_init(void) { size_t i; diff -ur v2.6.14/linux/net/ipv4/netfilter/ip_nat_standalone.c linux/net/ipv4/netfilter/ip_nat_standalone.c --- v2.6.14/linux/net/ipv4/netfilter/ip_nat_standalone.c 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/netfilter/ip_nat_standalone.c 2005-10-29 18:11:32.887759000 +0300 @@ -266,6 +266,14 @@ .priority = NF_IP_PRI_NAT_DST, }; +/* Before routing, route before mangling */ +static struct nf_hook_ops ip_nat_inr_ops = { + .hook = ip_nat_route_input, + .pf = PF_INET, + .hooknum = NF_IP_PRE_ROUTING, + .priority = NF_IP_PRI_LAST-1, +}; + /* After packet filtering, change source */ static struct nf_hook_ops ip_nat_out_ops = { .hook = ip_nat_out, @@ -330,10 +338,15 @@ printk("ip_nat_init: can't register in hook.\n"); goto cleanup_rule_init; } + ret = nf_register_hook(&ip_nat_inr_ops); + if (ret < 0) { + printk("ip_nat_init: can't register inr hook.\n"); + goto cleanup_inops; + } ret = nf_register_hook(&ip_nat_out_ops); if (ret < 0) { printk("ip_nat_init: can't register out hook.\n"); - goto cleanup_inops; + goto cleanup_inrops; } ret = nf_register_hook(&ip_nat_adjust_in_ops); if (ret < 0) { @@ -367,6 +380,8 @@ nf_unregister_hook(&ip_nat_adjust_in_ops); cleanup_outops: nf_unregister_hook(&ip_nat_out_ops); + cleanup_inrops: + nf_unregister_hook(&ip_nat_inr_ops); cleanup_inops: nf_unregister_hook(&ip_nat_in_ops); cleanup_rule_init: diff -ur v2.6.14/linux/net/ipv4/netfilter/ipt_MASQUERADE.c linux/net/ipv4/netfilter/ipt_MASQUERADE.c --- v2.6.14/linux/net/ipv4/netfilter/ipt_MASQUERADE.c 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/netfilter/ipt_MASQUERADE.c 2005-10-29 18:11:32.887759000 +0300 @@ -97,13 +97,31 @@ return NF_ACCEPT; mr = targinfo; - rt = (struct rtable *)(*pskb)->dst; - newsrc = inet_select_addr(out, rt->rt_gateway, RT_SCOPE_UNIVERSE); - if (!newsrc) { - printk("MASQUERADE: %s ate my IP address\n", out->name); - return NF_DROP; + + { + struct flowi fl = { .nl_u = { .ip4_u = + { .daddr = (*pskb)->nh.iph->daddr, + .tos = (RT_TOS((*pskb)->nh.iph->tos) | + RTO_CONN), + .gw = ((struct rtable *) (*pskb)->dst)->rt_gateway, +#ifdef CONFIG_IP_ROUTE_FWMARK + .fwmark = (*pskb)->nfmark +#endif + } }, + .oif = out->ifindex }; + if (ip_route_output_key(&rt, &fl) != 0) { + /* Funky routing can do this. */ + if (net_ratelimit()) + printk("MASQUERADE:" + " No route: Rusty's brain broke!\n"); + return NF_DROP; + } } + newsrc = rt->rt_src; + DEBUGP("newsrc = %u.%u.%u.%u\n", NIPQUAD(newsrc)); + ip_rt_put(rt); + write_lock_bh(&masq_lock); ct->nat.masq_index = out->ifindex; write_unlock_bh(&masq_lock); diff -ur v2.6.14/linux/net/ipv4/route.c linux/net/ipv4/route.c --- v2.6.14/linux/net/ipv4/route.c 2005-10-29 14:15:09.000000000 +0300 +++ linux/net/ipv4/route.c 2005-10-29 18:11:32.889758696 +0300 @@ -1197,6 +1197,7 @@ /* Gateway is different ... */ rt->rt_gateway = new_gw; + if (rt->fl.fl4_gw) rt->fl.fl4_gw = new_gw; /* Redirect received -> path was valid */ dst_confirm(&rth->u.dst); @@ -1632,6 +1633,7 @@ rth->fl.fl4_fwmark= skb->nfmark; #endif rth->fl.fl4_src = saddr; + rth->fl.fl4_lsrc = 0; rth->rt_src = saddr; #ifdef CONFIG_NET_CLS_ROUTE rth->u.dst.tclassid = itag; @@ -1642,6 +1644,7 @@ dev_hold(rth->u.dst.dev); rth->idev = in_dev_get(rth->u.dst.dev); rth->fl.oif = 0; + rth->fl.fl4_gw = 0; rth->rt_gateway = daddr; rth->rt_spec_dst= spec_dst; rth->rt_type = RTN_MULTICAST; @@ -1706,7 +1709,7 @@ struct fib_result* res, struct in_device *in_dev, u32 daddr, u32 saddr, u32 tos, - struct rtable **result) + u32 lsrc, struct rtable **result) { struct rtable *rth; @@ -1739,6 +1742,7 @@ flags |= RTCF_DIRECTSRC; if (out_dev == in_dev && err && !(flags & (RTCF_NAT | RTCF_MASQ)) && + !lsrc && (IN_DEV_SHARED_MEDIA(out_dev) || inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) flags |= RTCF_DOREDIRECT; @@ -1778,6 +1782,7 @@ #endif rth->fl.fl4_src = saddr; rth->rt_src = saddr; + rth->fl.fl4_lsrc = lsrc; rth->rt_gateway = daddr; rth->rt_iif = rth->fl.iif = in_dev->dev->ifindex; @@ -1785,6 +1790,7 @@ dev_hold(rth->u.dst.dev); rth->idev = in_dev_get(rth->u.dst.dev); rth->fl.oif = 0; + rth->fl.fl4_gw = 0; rth->rt_spec_dst= spec_dst; rth->u.dst.input = ip_forward; @@ -1806,19 +1812,20 @@ struct fib_result* res, const struct flowi *fl, struct in_device *in_dev, - u32 daddr, u32 saddr, u32 tos) + u32 daddr, u32 saddr, u32 tos, u32 lsrc) { struct rtable* rth = NULL; int err; unsigned hash; + fib_select_default(fl, res); #ifdef CONFIG_IP_ROUTE_MULTIPATH - if (res->fi && res->fi->fib_nhs > 1 && fl->oif == 0) + if (res->fi && res->fi->fib_nhs > 1) fib_select_multipath(fl, res); #endif /* create a routing cache entry */ - err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth); + err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, lsrc, &rth); if (err) return err; @@ -1831,7 +1838,7 @@ struct fib_result* res, const struct flowi *fl, struct in_device *in_dev, - u32 daddr, u32 saddr, u32 tos) + u32 daddr, u32 saddr, u32 tos, u32 lsrc) { #ifdef CONFIG_IP_ROUTE_MULTIPATH_CACHED struct rtable* rth = NULL, *rtres; @@ -1847,7 +1854,7 @@ /* distinguish between multipath and singlepath */ if (hopcount < 2) return ip_mkroute_input_def(skb, res, fl, in_dev, daddr, - saddr, tos); + saddr, tos, 0); /* add all alternatives to the routing cache */ for (hop = 0; hop < hopcount; hop++) { @@ -1859,7 +1866,7 @@ /* create a routing cache entry */ err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, - &rth); + 0, &rth); if (err) return err; @@ -1879,7 +1886,7 @@ skb->dst = &rtres->u.dst; return err; #else /* CONFIG_IP_ROUTE_MULTIPATH_CACHED */ - return ip_mkroute_input_def(skb, res, fl, in_dev, daddr, saddr, tos); + return ip_mkroute_input_def(skb, res, fl, in_dev, daddr, saddr, tos, lsrc); #endif /* CONFIG_IP_ROUTE_MULTIPATH_CACHED */ } @@ -1895,20 +1902,20 @@ */ static int ip_route_input_slow(struct sk_buff *skb, u32 daddr, u32 saddr, - u8 tos, struct net_device *dev) + u8 tos, struct net_device *dev, u32 lsrc) { struct fib_result res; struct in_device *in_dev = in_dev_get(dev); struct flowi fl = { .nl_u = { .ip4_u = { .daddr = daddr, - .saddr = saddr, + .saddr = lsrc? : saddr, .tos = tos, .scope = RT_SCOPE_UNIVERSE, #ifdef CONFIG_IP_ROUTE_FWMARK .fwmark = skb->nfmark #endif } }, - .iif = dev->ifindex }; + .iif = lsrc? loopback_dev.ifindex : dev->ifindex }; unsigned flags = 0; u32 itag = 0; struct rtable * rth; @@ -1941,6 +1948,12 @@ if (BADCLASS(daddr) || ZERONET(daddr) || LOOPBACK(daddr)) goto martian_destination; + if (lsrc) { + if (MULTICAST(lsrc) || BADCLASS(lsrc) || + ZERONET(lsrc) || LOOPBACK(lsrc)) + goto e_inval; + } + /* * Now we are ready to route packet. */ @@ -1950,6 +1963,10 @@ goto no_route; } free_res = 1; + if (lsrc && res.type != RTN_UNICAST && res.type != RTN_NAT) + goto e_inval; + fl.iif = dev->ifindex; + fl.fl4_src = saddr; RT_CACHE_STAT_INC(in_slow_tot); @@ -1974,7 +1991,7 @@ if (res.type != RTN_UNICAST) goto martian_destination; - err = ip_mkroute_input(skb, &res, &fl, in_dev, daddr, saddr, tos); + err = ip_mkroute_input(skb, &res, &fl, in_dev, daddr, saddr, tos, lsrc); if (err == -ENOBUFS) goto e_nobufs; if (err == -EINVAL) @@ -1989,6 +2006,8 @@ brd_input: if (skb->protocol != htons(ETH_P_IP)) goto e_inval; + if (lsrc) + goto e_inval; if (ZERONET(saddr)) spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); @@ -2031,6 +2050,7 @@ rth->u.dst.dev = &loopback_dev; dev_hold(rth->u.dst.dev); rth->idev = in_dev_get(rth->u.dst.dev); + rth->fl.fl4_gw = 0; rth->rt_gateway = daddr; rth->rt_spec_dst= spec_dst; rth->u.dst.input= ip_local_deliver; @@ -2080,8 +2100,9 @@ goto e_inval; } -int ip_route_input(struct sk_buff *skb, u32 daddr, u32 saddr, - u8 tos, struct net_device *dev) +static inline int +ip_route_input_cached(struct sk_buff *skb, u32 daddr, u32 saddr, + u8 tos, struct net_device *dev, u32 lsrc) { struct rtable * rth; unsigned hash; @@ -2096,6 +2117,7 @@ if (rth->fl.fl4_dst == daddr && rth->fl.fl4_src == saddr && rth->fl.iif == iif && + rth->fl.fl4_lsrc == lsrc && rth->fl.oif == 0 && #ifdef CONFIG_IP_ROUTE_FWMARK rth->fl.fl4_fwmark == skb->nfmark && @@ -2144,7 +2166,19 @@ rcu_read_unlock(); return -EINVAL; } - return ip_route_input_slow(skb, daddr, saddr, tos, dev); + return ip_route_input_slow(skb, daddr, saddr, tos, dev, lsrc); +} + +int ip_route_input(struct sk_buff *skb, u32 daddr, u32 saddr, + u8 tos, struct net_device *dev) +{ + return ip_route_input_cached(skb, daddr, saddr, tos, dev, 0); +} + +int ip_route_input_lookup(struct sk_buff *skb, u32 daddr, u32 saddr, + u8 tos, struct net_device *dev, u32 lsrc) +{ + return ip_route_input_cached(skb, daddr, saddr, tos, dev, lsrc); } static inline int __mkroute_output(struct rtable **result, @@ -2223,6 +2257,7 @@ rth->fl.fl4_tos = tos; rth->fl.fl4_src = oldflp->fl4_src; rth->fl.oif = oldflp->oif; + rth->fl.fl4_gw = oldflp->fl4_gw; #ifdef CONFIG_IP_ROUTE_FWMARK rth->fl.fl4_fwmark= oldflp->fl4_fwmark; #endif @@ -2370,6 +2405,7 @@ struct flowi fl = { .nl_u = { .ip4_u = { .daddr = oldflp->fl4_dst, .saddr = oldflp->fl4_src, + .gw = oldflp->fl4_gw, .tos = tos & IPTOS_RT_MASK, .scope = ((tos & RTO_ONLINK) ? RT_SCOPE_LINK : @@ -2475,6 +2511,7 @@ dev_out = &loopback_dev; dev_hold(dev_out); fl.oif = loopback_dev.ifindex; + fl.fl4_gw = 0; res.type = RTN_LOCAL; flags |= RTCF_LOCAL; goto make_route; @@ -2482,7 +2519,7 @@ if (fib_lookup(&fl, &res)) { res.fi = NULL; - if (oldflp->oif) { + if (oldflp->oif && dev_out->flags & IFF_UP) { /* Apparently, routing tables are wrong. Assume, that the destination is on link. @@ -2522,6 +2559,7 @@ dev_out = &loopback_dev; dev_hold(dev_out); fl.oif = dev_out->ifindex; + fl.fl4_gw = 0; if (res.fi) fib_info_put(res.fi); res.fi = NULL; @@ -2529,13 +2567,12 @@ goto make_route; } + if (res.type == RTN_UNICAST) + fib_select_default(&fl, &res); #ifdef CONFIG_IP_ROUTE_MULTIPATH - if (res.fi->fib_nhs > 1 && fl.oif == 0) + if (res.fi->fib_nhs > 1) fib_select_multipath(&fl, &res); - else #endif - if (!res.prefixlen && res.type == RTN_UNICAST && !fl.oif) - fib_select_default(&fl, &res); if (!fl.fl4_src) fl.fl4_src = FIB_RES_PREFSRC(res); @@ -2572,6 +2609,7 @@ rth->fl.fl4_src == flp->fl4_src && rth->fl.iif == 0 && rth->fl.oif == flp->oif && + rth->fl.fl4_gw == flp->fl4_gw && #ifdef CONFIG_IP_ROUTE_FWMARK rth->fl.fl4_fwmark == flp->fl4_fwmark && #endif @@ -3211,3 +3249,4 @@ EXPORT_SYMBOL(__ip_select_ident); EXPORT_SYMBOL(ip_route_input); EXPORT_SYMBOL(ip_route_output_key); +EXPORT_SYMBOL(ip_route_input_lookup);