Index: linux-2.4.35.4/include/linux/netfilter_ipv4/ip_nat.h =================================================================== --- linux-2.4.35.4.orig/include/linux/netfilter_ipv4/ip_nat.h +++ linux-2.4.35.4/include/linux/netfilter_ipv4/ip_nat.h @@ -121,5 +121,13 @@ extern int ip_nat_used_tuple(const struc extern u_int16_t ip_nat_cheat_check(u_int32_t oldvalinv, u_int32_t newval, u_int16_t oldcheck); + +/* 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 *)); + #endif /*__KERNEL__*/ #endif Index: linux-2.4.35.4/include/linux/rtnetlink.h =================================================================== --- linux-2.4.35.4.orig/include/linux/rtnetlink.h +++ linux-2.4.35.4/include/linux/rtnetlink.h @@ -234,6 +234,8 @@ struct rtnexthop #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 */ Index: linux-2.4.35.4/include/net/ip_fib.h =================================================================== --- linux-2.4.35.4.orig/include/net/ip_fib.h +++ linux-2.4.35.4/include/net/ip_fib.h @@ -162,7 +162,8 @@ static inline int fib_lookup(const struc static inline void fib_select_default(const struct rt_key *key, 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) main_table->tb_select_default(main_table, key, res); } @@ -174,6 +175,7 @@ extern struct fib_table * fib_tables[RT_ extern int fib_lookup(const struct rt_key *key, 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) { @@ -275,5 +277,6 @@ static inline void fib_res_put(struct fi #endif } +extern rwlock_t fib_nhflags_lock; #endif /* _NET_FIB_H */ Index: linux-2.4.35.4/include/net/route.h =================================================================== --- linux-2.4.35.4.orig/include/net/route.h +++ linux-2.4.35.4/include/net/route.h @@ -49,6 +49,8 @@ struct rt_key { __u32 dst; __u32 src; + __u32 lsrc; + __u32 gw; int iif; int oif; #ifdef CONFIG_IP_ROUTE_FWMARK @@ -128,6 +130,7 @@ extern void ip_rt_advice(struct rtable extern void rt_cache_flush(int how); extern int ip_route_output_key(struct rtable **, const struct rt_key *key); 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_update_pmtu(struct dst_entry *dst, unsigned mtu); extern void ip_rt_send_redirect(struct sk_buff *skb); @@ -148,6 +151,15 @@ static inline int ip_route_output(struct } +static inline int +ip_route_output_lookup(struct rtable **rp, + u32 daddr, u32 saddr, u32 tos, int oif, u32 gw) +{ + struct rt_key key = { dst:daddr, src:saddr, gw:gw, oif:oif, tos:tos }; + + return ip_route_output_key(rp, &key); +} + static inline void ip_rt_put(struct rtable * rt) { if (rt) Index: linux-2.4.35.4/net/ipv4/fib_frontend.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/fib_frontend.c +++ linux-2.4.35.4/net/ipv4/fib_frontend.c @@ -54,6 +54,8 @@ struct fib_table *local_table; struct fib_table *main_table; +#define FIB_RES_TABLE(r) (RT_TABLE_MAIN) + #else #define RT_TABLE_MIN 1 @@ -71,6 +73,7 @@ struct fib_table *__fib_new_table(int id return tb; } +#define FIB_RES_TABLE(r) (fib_result_table(r)) #endif /* CONFIG_IP_MULTIPLE_TABLES */ @@ -209,6 +212,9 @@ int fib_validate_source(u32 src, u32 dst struct in_device *in_dev; struct rt_key key; struct fib_result res; + int table; + unsigned char prefixlen; + unsigned char scope; int no_addr, rpf; int ret; @@ -216,6 +222,7 @@ int fib_validate_source(u32 src, u32 dst key.src = dst; key.tos = tos; key.oif = 0; + key.gw = 0; key.iif = oif; key.scope = RT_SCOPE_UNIVERSE; @@ -237,31 +244,35 @@ int fib_validate_source(u32 src, u32 dst 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; key.oif = dev->ifindex; ret = 0; if (fib_lookup(&key, &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: @@ -579,9 +590,7 @@ static int fib_inetaddr_event(struct not 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: @@ -617,9 +626,7 @@ static int fib_netdev_event(struct notif 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: Index: linux-2.4.35.4/net/ipv4/fib_hash.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/fib_hash.c +++ linux-2.4.35.4/net/ipv4/fib_hash.c @@ -71,6 +71,7 @@ struct fib_node struct fib_info *fn_info; #define FIB_INFO(f) ((f)->fn_info) fn_key_t fn_key; + int fn_last_dflt; u8 fn_tos; u8 fn_type; u8 fn_scope; @@ -336,72 +337,123 @@ out: return err; } -static int fn_hash_last_dflt=-1; - -static int fib_detect_death(struct fib_info *fi, int order, - struct fib_info **last_resort, int *last_idx) +static int fib_detect_death(struct fib_info *fi, int order, int last_dflt, + struct fib_info **last_resort, int *last_idx, + int *last_nhsel, const struct rt_key *key) { 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 (key->oif && key->oif != nh->nh_oif) + continue; + if (key->gw && key->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); + 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 = key->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 != last_dflt)) { + dead = 0; + goto setfl; + } + if (!(state&NUD_VALID)) + flag = 1; + if (!dead) + goto setfl; + if ((state&NUD_VALID) || + (*last_idx<0 && order >= last_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); } - if (state==NUD_REACHABLE) - return 0; - if ((state&NUD_VALID) && order != fn_hash_last_dflt) - return 0; - if ((state&NUD_VALID) || - (*last_idx<0 && order > fn_hash_last_dflt)) { - *last_resort = fi; - *last_idx = order; - } - return 1; + /* } endfor_nexthops(fi) */ + + return dead; } static void fn_hash_select_default(struct fib_table *tb, const struct rt_key *key, struct fib_result *res) { - int order, last_idx; - struct fib_node *f; + int order, last_idx, last_dflt, last_nhsel; + struct fib_node *f, *first_node; 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]; + fn_key_t k; if (fz == NULL) return; + k = fz_key(key->dst, fz); + last_dflt = -2; + first_node = NULL; last_idx = -1; last_resort = NULL; + last_nhsel = 0; order = -1; read_lock(&fib_hash_lock); - for (f = fz->fz_hash[0]; f; f = f->fn_next) { + for (f = fz_chain(k, fz); f; f = f->fn_next) { struct fib_info *next_fi = FIB_INFO(f); - if ((f->fn_state&FN_S_ZOMBIE) || + if (!fn_key_eq(k, f->fn_key) || + (f->fn_state&FN_S_ZOMBIE) || f->fn_scope != res->scope || +#ifdef CONFIG_IP_ROUTE_TOS + (f->fn_tos && f->fn_tos != key->tos) || +#endif f->fn_type != RTN_UNICAST) 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; f->fn_state |= FN_S_ACCESSED; - if (fi == NULL) { - if (next_fi != res->fi) - break; - } else if (!fib_detect_death(fi, order, &last_resort, &last_idx)) { + if (!first_node) { + last_dflt = f->fn_last_dflt; + first_node = f; + } + if (fi && !fib_detect_death(fi, order, last_dflt, + &last_resort, &last_idx, &last_nhsel, key)) { if (res->fi) fib_info_put(res->fi); res->fi = fi; atomic_inc(&fi->fib_clntref); - fn_hash_last_dflt = order; + first_node->fn_last_dflt = order; goto out; } fi = next_fi; @@ -409,16 +461,25 @@ fn_hash_select_default(struct fib_table } if (order<=0 || fi==NULL) { - fn_hash_last_dflt = -1; + if (fi && fi->fib_nhs > 1 && + fib_detect_death(fi, order, last_dflt, + &last_resort, &last_idx, &last_nhsel, key) && + 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_node) first_node->fn_last_dflt = -1; goto out; } - if (!fib_detect_death(fi, order, &last_resort, &last_idx)) { + if (!fib_detect_death(fi, order, last_dflt, &last_resort, &last_idx, + &last_nhsel, key)) { if (res->fi) fib_info_put(res->fi); res->fi = fi; atomic_inc(&fi->fib_clntref); - fn_hash_last_dflt = order; + first_node->fn_last_dflt = order; goto out; } @@ -428,8 +489,11 @@ fn_hash_select_default(struct fib_table 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_node->fn_last_dflt = last_idx; } - fn_hash_last_dflt = last_idx; out: read_unlock(&fib_hash_lock); } @@ -589,6 +653,7 @@ replace: memset(new_f, 0, sizeof(struct fib_node)); + new_f->fn_last_dflt = -1; new_f->fn_key = key; #ifdef CONFIG_IP_ROUTE_TOS new_f->fn_tos = tos; Index: linux-2.4.35.4/net/ipv4/fib_rules.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/fib_rules.c +++ linux-2.4.35.4/net/ipv4/fib_rules.c @@ -307,6 +307,11 @@ static void fib_rules_attach(struct net_ } } +int fib_result_table(struct fib_result *res) +{ + return res->r->r_table; +} + int fib_lookup(const struct rt_key *key, struct fib_result *res) { int err; @@ -371,8 +376,10 @@ FRprintk("FAILURE\n"); void fib_select_default(const struct rt_key *key, 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) { + if (res->r && + (res->r->r_action == RTN_UNICAST || res->r->r_action == RTN_NAT) && + ((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, key, res); Index: linux-2.4.35.4/net/ipv4/fib_semantics.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/fib_semantics.c +++ linux-2.4.35.4/net/ipv4/fib_semantics.c @@ -48,6 +48,7 @@ static struct fib_info *fib_info_list; static rwlock_t fib_info_lock = RW_LOCK_UNLOCKED; int fib_info_cnt; +rwlock_t fib_nhflags_lock = RW_LOCK_UNLOCKED; #define for_fib_info() { struct fib_info *fi; \ for (fi = fib_info_list; fi; fi = fi->fib_next) @@ -150,7 +151,7 @@ static __inline__ int nh_comp(const stru #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); @@ -166,7 +167,7 @@ static __inline__ struct fib_info * fib_ nfi->fib_prefsrc == fi->fib_prefsrc && 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; } endfor_fib_info(); @@ -365,8 +366,11 @@ static int fib_check_nh(const struct rtm 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; @@ -380,23 +384,48 @@ static int fib_check_nh(const struct rtm /* It is not necessary, but requires a bit of thinking */ if (key.scope < RT_SCOPE_LINK) key.scope = RT_SCOPE_LINK; - if ((err = fib_lookup(&key, &res)) != 0) - return err; - 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; + + err = fib_lookup(&key, &res); + 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; @@ -407,8 +436,11 @@ out: 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); @@ -606,8 +638,12 @@ fib_semantic_match(int type, struct fib_ for_nexthops(fi) { if (nh->nh_flags&RTNH_F_DEAD) continue; - if (!key->oif || key->oif == nh->nh_oif) - break; + if (key->oif && key->oif != nh->nh_oif) + continue; + if (key->gw && key->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) { @@ -873,22 +909,35 @@ int fib_sync_down(u32 local, struct net_ if (local && fi->fib_prefsrc == local) { fi->fib_flags |= RTNH_F_DEAD; ret++; - } else if (dev && fi->fib_nhs) { + } else if (fi->fib_nhs) { int 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(nh->nh_dev) || + nh->nh_dev->flags&IFF_UP) + dead++; + } else if ((nh->nh_dev == dev && dev && + nh->nh_scope != scope) || + (local == nh->nh_gw && local && + nh->nh_oif)) { + 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 || + (dev && __in_dev_get(dev) == NULL)) + dead++; } #ifdef CONFIG_IP_ROUTE_MULTIPATH if (force > 1 && nh->nh_dev == dev) { @@ -906,37 +955,55 @@ int fib_sync_down(u32 local, struct net_ 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) { - int ret = 0; + struct rt_key key; + struct fib_result res; + int ret, rep; +repeat: if (!(dev->flags&IFF_UP)) return 0; + ret = 0; + rep = 0; for_fib_info() { int 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(dev) == NULL) continue; + if (nh->nh_gw && fi->fib_protocol == RTPROT_STATIC) { + memset(&key, 0, sizeof(key)); + key.dst = nh->nh_gw; + key.oif = nh->nh_oif; + key.scope = nh->nh_scope; + if (fib_lookup(&key, &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) { @@ -944,9 +1011,13 @@ int fib_sync_up(struct net_device *dev) ret++; } } endfor_fib_info(); + if (rep) + goto repeat; return ret; } +#ifdef CONFIG_IP_ROUTE_MULTIPATH + /* The algorithm is suboptimal, but it provides really fair weighted route distribution. @@ -955,24 +1026,45 @@ int fib_sync_up(struct net_device *dev) void fib_select_multipath(const struct rt_key *key, struct fib_result *res) { struct fib_info *fi = res->fi; - int w; + int w, alive; spin_lock_bh(&fib_multipath_lock); + if (key->oif) { + int sel = -1; + w = -1; + change_nexthops(fi) { + if (key->oif != nh->nh_oif) + continue; + if (key->gw && key->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; } @@ -982,20 +1074,40 @@ void fib_select_multipath(const struct r 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 (key->oif && key->oif != nh->nh_oif) + continue; + if (key->gw && key->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 Index: linux-2.4.35.4/net/ipv4/ip_nat_dumb.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/ip_nat_dumb.c +++ linux-2.4.35.4/net/ipv4/ip_nat_dumb.c @@ -124,6 +124,7 @@ ip_do_nat(struct sk_buff *skb) key.dst = ciph->saddr; key.iif = skb->dev->ifindex; key.oif = 0; + key.gw = 0; #ifdef CONFIG_IP_ROUTE_TOS key.tos = RT_TOS(ciph->tos); #endif Index: linux-2.4.35.4/net/ipv4/netfilter/ip_fw_compat_masq.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/netfilter/ip_fw_compat_masq.c +++ linux-2.4.35.4/net/ipv4/netfilter/ip_fw_compat_masq.c @@ -41,6 +41,10 @@ do_masquerade(struct sk_buff **pskb, con enum ip_conntrack_info ctinfo; struct ip_conntrack *ct; unsigned int ret; + struct rtable *rt, *skb_rt; + struct net_device *skb_dev; + __u32 saddr; + int new; /* Sorry, only ICMP, TCP and UDP. */ if (iph->protocol != IPPROTO_ICMP @@ -64,22 +68,28 @@ do_masquerade(struct sk_buff **pskb, con } info = &ct->nat.info; + iph = (*pskb)->nh.iph; + saddr = iph->saddr; + new = 0; WRITE_LOCK(&ip_nat_lock); /* Setup the masquerade, if not already */ if (!info->initialized) { u_int32_t newsrc; - struct rtable *rt; struct ip_nat_multi_range range; + skb_rt = (struct rtable *) (*pskb)->dst; + skb_dev = skb_rt->u.dst.dev; /* Pass 0 instead of saddr, since it's going to be changed anyway. */ - if (ip_route_output(&rt, iph->daddr, 0, 0, 0) != 0) { + if (ip_route_output_lookup(&rt, iph->daddr, 0, RT_TOS(iph->tos), + skb_dev? skb_dev->ifindex : 0, + skb_dev? skb_rt->rt_gateway : 0) != 0) { + WRITE_UNLOCK(&ip_nat_lock); DEBUGP("ipnat_rule_masquerade: Can't reroute.\n"); return NF_DROP; } - newsrc = inet_select_addr(rt->u.dst.dev, rt->rt_gateway, - RT_SCOPE_UNIVERSE); + newsrc = rt->rt_src; ip_rt_put(rt); range = ((struct ip_nat_multi_range) { 1, @@ -92,11 +102,31 @@ do_masquerade(struct sk_buff **pskb, con WRITE_UNLOCK(&ip_nat_lock); return ret; } + new = 1; } else DEBUGP("Masquerading already done on this conn.\n"); WRITE_UNLOCK(&ip_nat_lock); - return do_bindings(ct, ctinfo, info, NF_IP_POST_ROUTING, pskb); + ret = do_bindings(ct, ctinfo, info, NF_IP_POST_ROUTING, pskb); + if (ret != NF_ACCEPT || saddr == (*pskb)->nh.iph->saddr || new) + return ret; + + iph = (*pskb)->nh.iph; + if (ip_route_output(&rt, iph->daddr, iph->saddr, RT_TOS(iph->tos), 0) != 0) + return NF_DROP; + + skb_rt = (struct rtable *) (*pskb)->dst; + skb_dev = skb_rt->u.dst.dev; + if (skb_dev != rt->u.dst.dev || rt->rt_gateway != skb_rt->rt_gateway) { + if (skb_dev != rt->u.dst.dev) { + /* TODO: check the new mtu and reply FRAG_NEEDED */ + } + dst_release((*pskb)->dst); + (*pskb)->dst = &rt->u.dst; + } else { + ip_rt_put(rt); + } + return NF_ACCEPT; } void Index: linux-2.4.35.4/net/ipv4/netfilter/ip_nat_core.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/netfilter/ip_nat_core.c +++ linux-2.4.35.4/net/ipv4/netfilter/ip_nat_core.c @@ -994,6 +994,60 @@ icmp_reply_translation(struct sk_buff *s return NF_ACCEPT; } +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 *ct; + enum ip_conntrack_info ctinfo; + struct ip_nat_info *info; + enum ip_conntrack_dir dir; + __u32 saddr; + int i; + + if (!(ct = ip_conntrack_get(skb, &ctinfo))) + return NF_ACCEPT; + + info = &ct->nat.info; + if (!info->initialized) + return NF_ACCEPT; + + if (skb->dst) + return NF_ACCEPT; + + if (skb->len < sizeof(struct iphdr)) + return NF_ACCEPT; + + iph = skb->nh.iph; + saddr = iph->saddr; + hooknum = NF_IP_POST_ROUTING; + dir = CTINFO2DIR(ctinfo); + + READ_LOCK(&ip_nat_lock); + for (i = 0; i < info->num_manips; i++) { + if (info->manips[i].direction == dir + && info->manips[i].hooknum == hooknum + && info->manips[i].maniptype == IP_NAT_MANIP_SRC) { + saddr = info->manips[i].manip.ip; + } + } + READ_UNLOCK(&ip_nat_lock); + + 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; +} + int __init ip_nat_init(void) { size_t i; Index: linux-2.4.35.4/net/ipv4/netfilter/ip_nat_standalone.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/netfilter/ip_nat_standalone.c +++ linux-2.4.35.4/net/ipv4/netfilter/ip_nat_standalone.c @@ -245,6 +245,9 @@ ip_nat_local_fn(unsigned int hooknum, /* Before packet filtering, change destination */ static struct nf_hook_ops ip_nat_in_ops = { { NULL, NULL }, ip_nat_in, PF_INET, NF_IP_PRE_ROUTING, NF_IP_PRI_NAT_DST }; +/* Before routing, route before mangling */ +static struct nf_hook_ops ip_nat_inr_ops += { { NULL, NULL }, ip_nat_route_input, PF_INET, NF_IP_PRE_ROUTING, NF_IP_PRI_LAST-1 }; /* After packet filtering, change source */ static struct nf_hook_ops ip_nat_out_ops = { { NULL, NULL }, ip_nat_out, PF_INET, NF_IP_POST_ROUTING, NF_IP_PRI_NAT_SRC}; @@ -313,10 +316,15 @@ static int init_or_cleanup(int init) printk("ip_nat_init: can't register in hook.\n"); goto cleanup_nat; } + 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_local_out_ops); if (ret < 0) { @@ -336,6 +344,8 @@ static int init_or_cleanup(int init) nf_unregister_hook(&ip_nat_local_out_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_nat: Index: linux-2.4.35.4/net/ipv4/netfilter/ipt_MASQUERADE.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/netfilter/ipt_MASQUERADE.c +++ linux-2.4.35.4/net/ipv4/netfilter/ipt_MASQUERADE.c @@ -87,7 +87,8 @@ masquerade_target(struct sk_buff **pskb, key.dst = (*pskb)->nh.iph->daddr; key.src = 0; /* Unknown: that's what we're trying to establish */ key.tos = RT_TOS((*pskb)->nh.iph->tos)|RTO_CONN; - key.oif = 0; + key.oif = out->ifindex; + key.gw = ((struct rtable *) (*pskb)->dst)->rt_gateway; #ifdef CONFIG_IP_ROUTE_FWMARK key.fwmark = (*pskb)->nfmark; #endif @@ -98,13 +99,6 @@ masquerade_target(struct sk_buff **pskb, " No route: Rusty's brain broke!\n"); return NF_DROP; } - if (rt->u.dst.dev != out) { - if (net_ratelimit()) - printk("MASQUERADE:" - " Route sent us somewhere else.\n"); - ip_rt_put(rt); - return NF_DROP; - } newsrc = rt->rt_src; DEBUGP("newsrc = %u.%u.%u.%u\n", NIPQUAD(newsrc)); Index: linux-2.4.35.4/net/ipv4/route.c =================================================================== --- linux-2.4.35.4.orig/net/ipv4/route.c +++ linux-2.4.35.4/net/ipv4/route.c @@ -919,6 +919,7 @@ void ip_rt_redirect(u32 old_gw, u32 dadd /* Gateway is different ... */ rt->rt_gateway = new_gw; + if (rt->key.gw) rt->key.gw = new_gw; /* Redirect received -> path was valid */ dst_confirm(&rth->u.dst); @@ -1343,6 +1344,7 @@ static int ip_route_input_mc(struct sk_b rth->key.fwmark = skb->nfmark; #endif rth->key.src = saddr; + rth->key.lsrc = 0; rth->rt_src = saddr; #ifdef CONFIG_IP_ROUTE_NAT rth->rt_dst_map = daddr; @@ -1356,6 +1358,7 @@ static int ip_route_input_mc(struct sk_b rth->u.dst.dev = &loopback_dev; dev_hold(rth->u.dst.dev); rth->key.oif = 0; + rth->key.gw = 0; rth->rt_gateway = daddr; rth->rt_spec_dst= spec_dst; rth->rt_type = RTN_MULTICAST; @@ -1395,7 +1398,7 @@ e_inval: */ 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 rt_key key; struct fib_result res; @@ -1415,16 +1418,17 @@ int ip_route_input_slow(struct sk_buff * goto out; key.dst = daddr; - key.src = saddr; + key.src = lsrc? : saddr; key.tos = tos; #ifdef CONFIG_IP_ROUTE_FWMARK key.fwmark = skb->nfmark; #endif - key.iif = dev->ifindex; + key.iif = lsrc? loopback_dev.ifindex : dev->ifindex; key.oif = 0; + key.gw = 0; key.scope = RT_SCOPE_UNIVERSE; - hash = rt_hash_code(daddr, saddr ^ (key.iif << 5), tos); + hash = rt_hash_code(daddr, saddr ^ (dev->ifindex << 5), tos); /* Check for the most weird martians, which can be not detected by fib_lookup. @@ -1445,6 +1449,12 @@ int ip_route_input_slow(struct sk_buff * 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. */ @@ -1454,6 +1464,10 @@ int ip_route_input_slow(struct sk_buff * goto no_route; } free_res = 1; + if (lsrc && res.type != RTN_UNICAST && res.type != RTN_NAT) + goto e_inval; + key.iif = dev->ifindex; + key.src = saddr; rt_cache_stat[smp_processor_id()].in_slow_tot++; @@ -1464,7 +1478,7 @@ int ip_route_input_slow(struct sk_buff * if (1) { u32 src_map = saddr; - if (res.r) + if (res.r && !lsrc) src_map = fib_rules_policy(saddr, &res, &flags); if (res.type == RTN_NAT) { @@ -1503,8 +1517,9 @@ int ip_route_input_slow(struct sk_buff * if (res.type != RTN_UNICAST) goto martian_destination; + fib_select_default(&key, &res); #ifdef CONFIG_IP_ROUTE_MULTIPATH - if (res.fi->fib_nhs > 1 && key.oif == 0) + if (res.fi->fib_nhs > 1) fib_select_multipath(&key, &res); #endif out_dev = in_dev_get(FIB_RES_DEV(res)); @@ -1524,6 +1539,7 @@ int ip_route_input_slow(struct sk_buff * 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; @@ -1550,6 +1566,7 @@ int ip_route_input_slow(struct sk_buff * #endif rth->key.src = saddr; rth->rt_src = saddr; + rth->key.lsrc = lsrc; rth->rt_gateway = daddr; #ifdef CONFIG_IP_ROUTE_NAT rth->rt_src_map = key.src; @@ -1562,6 +1579,7 @@ int ip_route_input_slow(struct sk_buff * rth->u.dst.dev = out_dev->dev; dev_hold(rth->u.dst.dev); rth->key.oif = 0; + rth->key.gw = 0; rth->rt_spec_dst= spec_dst; rth->u.dst.input = ip_forward; @@ -1572,7 +1590,8 @@ int ip_route_input_slow(struct sk_buff * rth->rt_flags = flags; #ifdef CONFIG_NET_FASTROUTE - if (netdev_fastroute && !(flags&(RTCF_NAT|RTCF_MASQ|RTCF_DOREDIRECT))) { + if (netdev_fastroute && !(flags&(RTCF_NAT|RTCF_MASQ|RTCF_DOREDIRECT)) && + !lsrc) { struct net_device *odev = rth->u.dst.dev; if (odev != dev && dev->accept_fastpath && @@ -1595,6 +1614,8 @@ out: return err; 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); @@ -1627,6 +1648,7 @@ local_input: #endif rth->key.src = saddr; rth->rt_src = saddr; + rth->key.lsrc = 0; #ifdef CONFIG_IP_ROUTE_NAT rth->rt_dst_map = key.dst; rth->rt_src_map = key.src; @@ -1639,6 +1661,7 @@ local_input: rth->u.dst.dev = &loopback_dev; dev_hold(rth->u.dst.dev); rth->key.oif = 0; + rth->key.gw = 0; rth->rt_gateway = daddr; rth->rt_spec_dst= spec_dst; rth->u.dst.input= ip_local_deliver; @@ -1704,8 +1727,9 @@ martian_source: 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; @@ -1719,6 +1743,7 @@ int ip_route_input(struct sk_buff *skb, if (rth->key.dst == daddr && rth->key.src == saddr && rth->key.iif == iif && + rth->key.lsrc == lsrc && rth->key.oif == 0 && #ifdef CONFIG_IP_ROUTE_FWMARK rth->key.fwmark == skb->nfmark && @@ -1766,9 +1791,21 @@ int ip_route_input(struct sk_buff *skb, read_unlock(&inetdev_lock); 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); +} + /* * Major route resolver routine. */ @@ -1791,6 +1828,7 @@ int ip_route_output_slow(struct rtable * key.tos = tos & IPTOS_RT_MASK; key.iif = loopback_dev.ifindex; key.oif = oldkey->oif; + key.gw = oldkey->gw; #ifdef CONFIG_IP_ROUTE_FWMARK key.fwmark = oldkey->fwmark; #endif @@ -1880,6 +1918,7 @@ int ip_route_output_slow(struct rtable * dev_out = &loopback_dev; dev_hold(dev_out); key.oif = loopback_dev.ifindex; + key.gw = 0; res.type = RTN_LOCAL; flags |= RTCF_LOCAL; goto make_route; @@ -1887,7 +1926,7 @@ int ip_route_output_slow(struct rtable * if (fib_lookup(&key, &res)) { res.fi = NULL; - if (oldkey->oif) { + if (oldkey->oif && dev_out->flags&IFF_UP) { /* Apparently, routing tables are wrong. Assume, that the destination is on link. @@ -1930,6 +1969,7 @@ int ip_route_output_slow(struct rtable * dev_out = &loopback_dev; dev_hold(dev_out); key.oif = dev_out->ifindex; + key.gw = 0; if (res.fi) fib_info_put(res.fi); res.fi = NULL; @@ -1937,13 +1977,12 @@ int ip_route_output_slow(struct rtable * goto make_route; } + if (res.type == RTN_UNICAST) + fib_select_default(&key, &res); #ifdef CONFIG_IP_ROUTE_MULTIPATH - if (res.fi->fib_nhs > 1 && key.oif == 0) + if (res.fi->fib_nhs > 1) fib_select_multipath(&key, &res); - else #endif - if (!res.prefixlen && res.type == RTN_UNICAST && !key.oif) - fib_select_default(&key, &res); if (!key.src) key.src = FIB_RES_PREFSRC(res); @@ -2001,7 +2040,9 @@ make_route: rth->key.tos = tos; rth->key.src = oldkey->src; rth->key.iif = 0; + rth->key.lsrc = 0; rth->key.oif = oldkey->oif; + rth->key.gw = oldkey->gw; #ifdef CONFIG_IP_ROUTE_FWMARK rth->key.fwmark = oldkey->fwmark; #endif @@ -2080,6 +2121,7 @@ int ip_route_output_key(struct rtable ** rth->key.src == key->src && rth->key.iif == 0 && rth->key.oif == key->oif && + rth->key.gw == key->gw && #ifdef CONFIG_IP_ROUTE_FWMARK rth->key.fwmark == key->fwmark && #endif Index: linux-2.4.35.4/net/netsyms.c =================================================================== --- linux-2.4.35.4.orig/net/netsyms.c +++ linux-2.4.35.4/net/netsyms.c @@ -260,6 +260,7 @@ EXPORT_SYMBOL(inet_register_protosw); EXPORT_SYMBOL(inet_unregister_protosw); EXPORT_SYMBOL(ip_route_output_key); EXPORT_SYMBOL(ip_route_input); +EXPORT_SYMBOL(ip_route_input_lookup); EXPORT_SYMBOL(icmp_send); EXPORT_SYMBOL(icmp_statistics); EXPORT_SYMBOL(icmp_err_convert);