Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1 | /* Connection state tracking for netfilter. This is separated from, |
| 2 | but required by, the NAT layer; it can also be used by an iptables |
| 3 | extension. */ |
| 4 | |
| 5 | /* (C) 1999-2001 Paul `Rusty' Russell |
Harald Welte | dc808fe | 2006-03-20 17:56:32 -0800 | [diff] [blame] | 6 | * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org> |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 7 | * (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org> |
| 8 | * |
| 9 | * This program is free software; you can redistribute it and/or modify |
| 10 | * it under the terms of the GNU General Public License version 2 as |
| 11 | * published by the Free Software Foundation. |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 12 | */ |
| 13 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 14 | #include <linux/types.h> |
| 15 | #include <linux/netfilter.h> |
| 16 | #include <linux/module.h> |
Alexey Dobriyan | d43c36d | 2009-10-07 17:09:06 +0400 | [diff] [blame] | 17 | #include <linux/sched.h> |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 18 | #include <linux/skbuff.h> |
| 19 | #include <linux/proc_fs.h> |
| 20 | #include <linux/vmalloc.h> |
| 21 | #include <linux/stddef.h> |
| 22 | #include <linux/slab.h> |
| 23 | #include <linux/random.h> |
| 24 | #include <linux/jhash.h> |
| 25 | #include <linux/err.h> |
| 26 | #include <linux/percpu.h> |
| 27 | #include <linux/moduleparam.h> |
| 28 | #include <linux/notifier.h> |
| 29 | #include <linux/kernel.h> |
| 30 | #include <linux/netdevice.h> |
| 31 | #include <linux/socket.h> |
Al Viro | d7fe0f2 | 2006-12-03 23:15:30 -0500 | [diff] [blame] | 32 | #include <linux/mm.h> |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 33 | #include <linux/nsproxy.h> |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 34 | #include <linux/rculist_nulls.h> |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 35 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 36 | #include <net/netfilter/nf_conntrack.h> |
| 37 | #include <net/netfilter/nf_conntrack_l3proto.h> |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 38 | #include <net/netfilter/nf_conntrack_l4proto.h> |
Martin Josefsson | 77ab9cf | 2006-11-29 02:34:58 +0100 | [diff] [blame] | 39 | #include <net/netfilter/nf_conntrack_expect.h> |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 40 | #include <net/netfilter/nf_conntrack_helper.h> |
| 41 | #include <net/netfilter/nf_conntrack_core.h> |
Yasuyuki Kozakai | ecfab2c | 2007-07-07 22:23:21 -0700 | [diff] [blame] | 42 | #include <net/netfilter/nf_conntrack_extend.h> |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 43 | #include <net/netfilter/nf_conntrack_acct.h> |
Pablo Neira Ayuso | a0891aa | 2009-06-13 12:26:29 +0200 | [diff] [blame] | 44 | #include <net/netfilter/nf_conntrack_ecache.h> |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 45 | #include <net/netfilter/nf_conntrack_zones.h> |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 46 | #include <net/netfilter/nf_conntrack_timestamp.h> |
Pablo Neira Ayuso | e6a7d3c | 2008-10-14 11:58:31 -0700 | [diff] [blame] | 47 | #include <net/netfilter/nf_nat.h> |
Patrick McHardy | e17b666 | 2008-11-18 12:24:17 +0100 | [diff] [blame] | 48 | #include <net/netfilter/nf_nat_core.h> |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 49 | |
Harald Welte | dc808fe | 2006-03-20 17:56:32 -0800 | [diff] [blame] | 50 | #define NF_CONNTRACK_VERSION "0.5.0" |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 51 | |
Patrick McHardy | e17b666 | 2008-11-18 12:24:17 +0100 | [diff] [blame] | 52 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
| 53 | enum nf_nat_manip_type manip, |
Patrick McHardy | 3993832 | 2009-08-25 16:07:58 +0200 | [diff] [blame] | 54 | const struct nlattr *attr) __read_mostly; |
Pablo Neira Ayuso | e6a7d3c | 2008-10-14 11:58:31 -0700 | [diff] [blame] | 55 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
| 56 | |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 57 | DEFINE_SPINLOCK(nf_conntrack_lock); |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 58 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 59 | |
Martin Josefsson | e2b7606 | 2006-11-29 02:35:04 +0100 | [diff] [blame] | 60 | unsigned int nf_conntrack_htable_size __read_mostly; |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 61 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
| 62 | |
Hagen Paul Pfeifer | e478075 | 2009-02-20 10:47:09 +0100 | [diff] [blame] | 63 | unsigned int nf_conntrack_max __read_mostly; |
Patrick McHardy | a999e68 | 2006-11-29 02:35:20 +0100 | [diff] [blame] | 64 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 65 | |
Eric Dumazet | b3c5163 | 2010-06-09 14:43:38 +0200 | [diff] [blame] | 66 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
| 67 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 68 | |
Changli Gao | f682cef | 2011-01-05 04:23:23 +0000 | [diff] [blame] | 69 | unsigned int nf_conntrack_hash_rnd __read_mostly; |
Patrick McHardy | 4d4e61c | 2011-12-23 14:00:13 +0100 | [diff] [blame] | 70 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_rnd); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 71 | |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 72 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, u16 zone) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 73 | { |
Patrick McHardy | 0794935 | 2008-01-31 04:40:52 -0800 | [diff] [blame] | 74 | unsigned int n; |
Sami Farin | 9b88790 | 2007-03-14 16:43:00 -0700 | [diff] [blame] | 75 | |
Patrick McHardy | 0794935 | 2008-01-31 04:40:52 -0800 | [diff] [blame] | 76 | /* The direction must be ignored, so we hash everything up to the |
| 77 | * destination ports (which is a multiple of 4) and treat the last |
| 78 | * three bytes manually. |
| 79 | */ |
| 80 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 81 | return jhash2((u32 *)tuple, n, zone ^ nf_conntrack_hash_rnd ^ |
| 82 | (((__force __u16)tuple->dst.u.all << 16) | |
| 83 | tuple->dst.protonum)); |
| 84 | } |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 85 | |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 86 | static u32 __hash_bucket(u32 hash, unsigned int size) |
| 87 | { |
| 88 | return ((u64)hash * size) >> 32; |
| 89 | } |
| 90 | |
| 91 | static u32 hash_bucket(u32 hash, const struct net *net) |
| 92 | { |
| 93 | return __hash_bucket(hash, net->ct.htable_size); |
| 94 | } |
| 95 | |
| 96 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, |
| 97 | u16 zone, unsigned int size) |
| 98 | { |
| 99 | return __hash_bucket(hash_conntrack_raw(tuple, zone), size); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 100 | } |
| 101 | |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 102 | static inline u_int32_t hash_conntrack(const struct net *net, u16 zone, |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 103 | const struct nf_conntrack_tuple *tuple) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 104 | { |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 105 | return __hash_conntrack(tuple, zone, net->ct.htable_size); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 106 | } |
| 107 | |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 108 | bool |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 109 | nf_ct_get_tuple(const struct sk_buff *skb, |
| 110 | unsigned int nhoff, |
| 111 | unsigned int dataoff, |
| 112 | u_int16_t l3num, |
| 113 | u_int8_t protonum, |
| 114 | struct nf_conntrack_tuple *tuple, |
| 115 | const struct nf_conntrack_l3proto *l3proto, |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 116 | const struct nf_conntrack_l4proto *l4proto) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 117 | { |
Philip Craig | 443a70d | 2008-04-29 03:35:10 -0700 | [diff] [blame] | 118 | memset(tuple, 0, sizeof(*tuple)); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 119 | |
| 120 | tuple->src.l3num = l3num; |
| 121 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 122 | return false; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 123 | |
| 124 | tuple->dst.protonum = protonum; |
| 125 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; |
| 126 | |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 127 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 128 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 129 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 130 | |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 131 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
| 132 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) |
Yasuyuki Kozakai | e2a3123 | 2007-07-14 20:45:14 -0700 | [diff] [blame] | 133 | { |
| 134 | struct nf_conntrack_l3proto *l3proto; |
| 135 | struct nf_conntrack_l4proto *l4proto; |
| 136 | unsigned int protoff; |
| 137 | u_int8_t protonum; |
| 138 | int ret; |
| 139 | |
| 140 | rcu_read_lock(); |
| 141 | |
| 142 | l3proto = __nf_ct_l3proto_find(l3num); |
| 143 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); |
| 144 | if (ret != NF_ACCEPT) { |
| 145 | rcu_read_unlock(); |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 146 | return false; |
Yasuyuki Kozakai | e2a3123 | 2007-07-14 20:45:14 -0700 | [diff] [blame] | 147 | } |
| 148 | |
| 149 | l4proto = __nf_ct_l4proto_find(l3num, protonum); |
| 150 | |
| 151 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, |
| 152 | l3proto, l4proto); |
| 153 | |
| 154 | rcu_read_unlock(); |
| 155 | return ret; |
| 156 | } |
| 157 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); |
| 158 | |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 159 | bool |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 160 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
| 161 | const struct nf_conntrack_tuple *orig, |
| 162 | const struct nf_conntrack_l3proto *l3proto, |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 163 | const struct nf_conntrack_l4proto *l4proto) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 164 | { |
Philip Craig | 443a70d | 2008-04-29 03:35:10 -0700 | [diff] [blame] | 165 | memset(inverse, 0, sizeof(*inverse)); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 166 | |
| 167 | inverse->src.l3num = orig->src.l3num; |
| 168 | if (l3proto->invert_tuple(inverse, orig) == 0) |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 169 | return false; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 170 | |
| 171 | inverse->dst.dir = !orig->dst.dir; |
| 172 | |
| 173 | inverse->dst.protonum = orig->dst.protonum; |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 174 | return l4proto->invert_tuple(inverse, orig); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 175 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 176 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 177 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 178 | static void |
| 179 | clean_from_lists(struct nf_conn *ct) |
| 180 | { |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 181 | pr_debug("clean_from_lists(%p)\n", ct); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 182 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
| 183 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 184 | |
| 185 | /* Destroy all pending expectations */ |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 186 | nf_ct_remove_expectations(ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 187 | } |
| 188 | |
| 189 | static void |
| 190 | destroy_conntrack(struct nf_conntrack *nfct) |
| 191 | { |
| 192 | struct nf_conn *ct = (struct nf_conn *)nfct; |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 193 | struct net *net = nf_ct_net(ct); |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 194 | struct nf_conntrack_l4proto *l4proto; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 195 | |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 196 | pr_debug("destroy_conntrack(%p)\n", ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 197 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
| 198 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); |
| 199 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 200 | /* To make sure we don't get any weird locking issues here: |
| 201 | * destroy_conntrack() MUST NOT be called with a write lock |
| 202 | * to nf_conntrack_lock!!! -HW */ |
Patrick McHardy | 923f490 | 2007-02-12 11:12:57 -0800 | [diff] [blame] | 203 | rcu_read_lock(); |
Patrick McHardy | 5e8fbe2 | 2008-04-14 11:15:52 +0200 | [diff] [blame] | 204 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 205 | if (l4proto && l4proto->destroy) |
| 206 | l4proto->destroy(ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 207 | |
Patrick McHardy | 982d9a9 | 2007-02-12 11:14:11 -0800 | [diff] [blame] | 208 | rcu_read_unlock(); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 209 | |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 210 | spin_lock_bh(&nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 211 | /* Expectations will have been removed in clean_from_lists, |
| 212 | * except TFTP can create an expectation on the first packet, |
| 213 | * before connection is in the list, so we need to clean here, |
| 214 | * too. */ |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 215 | nf_ct_remove_expectations(ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 216 | |
| 217 | /* We overload first tuple to link into unconfirmed list. */ |
| 218 | if (!nf_ct_is_confirmed(ct)) { |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 219 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); |
| 220 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 221 | } |
| 222 | |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 223 | NF_CT_STAT_INC(net, delete); |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 224 | spin_unlock_bh(&nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 225 | |
| 226 | if (ct->master) |
| 227 | nf_ct_put(ct->master); |
| 228 | |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 229 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 230 | nf_conntrack_free(ct); |
| 231 | } |
| 232 | |
Pablo Neira Ayuso | dd7669a | 2009-06-13 12:30:52 +0200 | [diff] [blame] | 233 | void nf_ct_delete_from_lists(struct nf_conn *ct) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 234 | { |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 235 | struct net *net = nf_ct_net(ct); |
Patrick McHardy | 5397e97 | 2007-05-19 14:23:52 -0700 | [diff] [blame] | 236 | |
Pablo Neira Ayuso | 9858a3a | 2009-06-13 12:28:22 +0200 | [diff] [blame] | 237 | nf_ct_helper_destroy(ct); |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 238 | spin_lock_bh(&nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 239 | /* Inside lock so preempt is disabled on module removal path. |
| 240 | * Otherwise we can get spurious warnings. */ |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 241 | NF_CT_STAT_INC(net, delete_list); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 242 | clean_from_lists(ct); |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 243 | spin_unlock_bh(&nf_conntrack_lock); |
Pablo Neira Ayuso | dd7669a | 2009-06-13 12:30:52 +0200 | [diff] [blame] | 244 | } |
| 245 | EXPORT_SYMBOL_GPL(nf_ct_delete_from_lists); |
| 246 | |
| 247 | static void death_by_event(unsigned long ul_conntrack) |
| 248 | { |
| 249 | struct nf_conn *ct = (void *)ul_conntrack; |
| 250 | struct net *net = nf_ct_net(ct); |
| 251 | |
| 252 | if (nf_conntrack_event(IPCT_DESTROY, ct) < 0) { |
| 253 | /* bad luck, let's retry again */ |
| 254 | ct->timeout.expires = jiffies + |
| 255 | (random32() % net->ct.sysctl_events_retry_timeout); |
| 256 | add_timer(&ct->timeout); |
| 257 | return; |
| 258 | } |
| 259 | /* we've got the event delivered, now it's dying */ |
| 260 | set_bit(IPS_DYING_BIT, &ct->status); |
| 261 | spin_lock(&nf_conntrack_lock); |
| 262 | hlist_nulls_del(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
| 263 | spin_unlock(&nf_conntrack_lock); |
| 264 | nf_ct_put(ct); |
| 265 | } |
| 266 | |
| 267 | void nf_ct_insert_dying_list(struct nf_conn *ct) |
| 268 | { |
| 269 | struct net *net = nf_ct_net(ct); |
| 270 | |
| 271 | /* add this conntrack to the dying list */ |
| 272 | spin_lock_bh(&nf_conntrack_lock); |
| 273 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
| 274 | &net->ct.dying); |
| 275 | spin_unlock_bh(&nf_conntrack_lock); |
| 276 | /* set a new timer to retry event delivery */ |
| 277 | setup_timer(&ct->timeout, death_by_event, (unsigned long)ct); |
| 278 | ct->timeout.expires = jiffies + |
| 279 | (random32() % net->ct.sysctl_events_retry_timeout); |
| 280 | add_timer(&ct->timeout); |
| 281 | } |
| 282 | EXPORT_SYMBOL_GPL(nf_ct_insert_dying_list); |
| 283 | |
| 284 | static void death_by_timeout(unsigned long ul_conntrack) |
| 285 | { |
| 286 | struct nf_conn *ct = (void *)ul_conntrack; |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 287 | struct nf_conn_tstamp *tstamp; |
| 288 | |
| 289 | tstamp = nf_conn_tstamp_find(ct); |
| 290 | if (tstamp && tstamp->stop == 0) |
| 291 | tstamp->stop = ktime_to_ns(ktime_get_real()); |
Pablo Neira Ayuso | dd7669a | 2009-06-13 12:30:52 +0200 | [diff] [blame] | 292 | |
| 293 | if (!test_bit(IPS_DYING_BIT, &ct->status) && |
| 294 | unlikely(nf_conntrack_event(IPCT_DESTROY, ct) < 0)) { |
| 295 | /* destroy event was not delivered */ |
| 296 | nf_ct_delete_from_lists(ct); |
| 297 | nf_ct_insert_dying_list(ct); |
| 298 | return; |
| 299 | } |
| 300 | set_bit(IPS_DYING_BIT, &ct->status); |
| 301 | nf_ct_delete_from_lists(ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 302 | nf_ct_put(ct); |
| 303 | } |
| 304 | |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 305 | /* |
| 306 | * Warning : |
| 307 | * - Caller must take a reference on returned object |
| 308 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) |
| 309 | * OR |
| 310 | * - Caller must lock nf_conntrack_lock before calling this function |
| 311 | */ |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 312 | static struct nf_conntrack_tuple_hash * |
| 313 | ____nf_conntrack_find(struct net *net, u16 zone, |
| 314 | const struct nf_conntrack_tuple *tuple, u32 hash) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 315 | { |
| 316 | struct nf_conntrack_tuple_hash *h; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 317 | struct hlist_nulls_node *n; |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 318 | unsigned int bucket = hash_bucket(hash, net); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 319 | |
Patrick McHardy | 4e29e9e | 2008-02-27 12:07:47 -0800 | [diff] [blame] | 320 | /* Disable BHs the entire time since we normally need to disable them |
| 321 | * at least once for the stats anyway. |
| 322 | */ |
| 323 | local_bh_disable(); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 324 | begin: |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 325 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[bucket], hnnode) { |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 326 | if (nf_ct_tuple_equal(tuple, &h->tuple) && |
| 327 | nf_ct_zone(nf_ct_tuplehash_to_ctrack(h)) == zone) { |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 328 | NF_CT_STAT_INC(net, found); |
Patrick McHardy | 4e29e9e | 2008-02-27 12:07:47 -0800 | [diff] [blame] | 329 | local_bh_enable(); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 330 | return h; |
| 331 | } |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 332 | NF_CT_STAT_INC(net, searched); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 333 | } |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 334 | /* |
| 335 | * if the nulls value we got at the end of this lookup is |
| 336 | * not the expected one, we must restart lookup. |
| 337 | * We probably met an item that was moved to another chain. |
| 338 | */ |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 339 | if (get_nulls_value(n) != bucket) { |
Jesper Dangaard Brouer | af740b2 | 2010-04-23 12:34:56 +0200 | [diff] [blame] | 340 | NF_CT_STAT_INC(net, search_restart); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 341 | goto begin; |
Jesper Dangaard Brouer | af740b2 | 2010-04-23 12:34:56 +0200 | [diff] [blame] | 342 | } |
Patrick McHardy | 4e29e9e | 2008-02-27 12:07:47 -0800 | [diff] [blame] | 343 | local_bh_enable(); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 344 | |
| 345 | return NULL; |
| 346 | } |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 347 | |
| 348 | struct nf_conntrack_tuple_hash * |
| 349 | __nf_conntrack_find(struct net *net, u16 zone, |
| 350 | const struct nf_conntrack_tuple *tuple) |
| 351 | { |
| 352 | return ____nf_conntrack_find(net, zone, tuple, |
| 353 | hash_conntrack_raw(tuple, zone)); |
| 354 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 355 | EXPORT_SYMBOL_GPL(__nf_conntrack_find); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 356 | |
| 357 | /* Find a connection corresponding to a tuple. */ |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 358 | static struct nf_conntrack_tuple_hash * |
| 359 | __nf_conntrack_find_get(struct net *net, u16 zone, |
| 360 | const struct nf_conntrack_tuple *tuple, u32 hash) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 361 | { |
| 362 | struct nf_conntrack_tuple_hash *h; |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 363 | struct nf_conn *ct; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 364 | |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 365 | rcu_read_lock(); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 366 | begin: |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 367 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 368 | if (h) { |
| 369 | ct = nf_ct_tuplehash_to_ctrack(h); |
Patrick McHardy | 8d8890b | 2009-06-22 14:14:41 +0200 | [diff] [blame] | 370 | if (unlikely(nf_ct_is_dying(ct) || |
| 371 | !atomic_inc_not_zero(&ct->ct_general.use))) |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 372 | h = NULL; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 373 | else { |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 374 | if (unlikely(!nf_ct_tuple_equal(tuple, &h->tuple) || |
| 375 | nf_ct_zone(ct) != zone)) { |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 376 | nf_ct_put(ct); |
| 377 | goto begin; |
| 378 | } |
| 379 | } |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 380 | } |
| 381 | rcu_read_unlock(); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 382 | |
| 383 | return h; |
| 384 | } |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 385 | |
| 386 | struct nf_conntrack_tuple_hash * |
| 387 | nf_conntrack_find_get(struct net *net, u16 zone, |
| 388 | const struct nf_conntrack_tuple *tuple) |
| 389 | { |
| 390 | return __nf_conntrack_find_get(net, zone, tuple, |
| 391 | hash_conntrack_raw(tuple, zone)); |
| 392 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 393 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 394 | |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 395 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
| 396 | unsigned int hash, |
YOSHIFUJI Hideaki | 601e68e | 2007-02-12 11:15:49 -0800 | [diff] [blame] | 397 | unsigned int repl_hash) |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 398 | { |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 399 | struct net *net = nf_ct_net(ct); |
| 400 | |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 401 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 402 | &net->ct.hash[hash]); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 403 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 404 | &net->ct.hash[repl_hash]); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 405 | } |
| 406 | |
| 407 | void nf_conntrack_hash_insert(struct nf_conn *ct) |
| 408 | { |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 409 | struct net *net = nf_ct_net(ct); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 410 | unsigned int hash, repl_hash; |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 411 | u16 zone; |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 412 | |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 413 | zone = nf_ct_zone(ct); |
| 414 | hash = hash_conntrack(net, zone, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
| 415 | repl_hash = hash_conntrack(net, zone, &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 416 | |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 417 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 418 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 419 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_insert); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 420 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 421 | /* Confirm a connection given skb; places it in hash table */ |
| 422 | int |
Herbert Xu | 3db05fe | 2007-10-15 00:53:15 -0700 | [diff] [blame] | 423 | __nf_conntrack_confirm(struct sk_buff *skb) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 424 | { |
| 425 | unsigned int hash, repl_hash; |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 426 | struct nf_conntrack_tuple_hash *h; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 427 | struct nf_conn *ct; |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 428 | struct nf_conn_help *help; |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 429 | struct nf_conn_tstamp *tstamp; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 430 | struct hlist_nulls_node *n; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 431 | enum ip_conntrack_info ctinfo; |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 432 | struct net *net; |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 433 | u16 zone; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 434 | |
Herbert Xu | 3db05fe | 2007-10-15 00:53:15 -0700 | [diff] [blame] | 435 | ct = nf_ct_get(skb, &ctinfo); |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 436 | net = nf_ct_net(ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 437 | |
| 438 | /* ipt_REJECT uses nf_conntrack_attach to attach related |
| 439 | ICMP/TCP RST packets in other direction. Actual packet |
| 440 | which created connection will be IP_CT_NEW or for an |
| 441 | expected connection, IP_CT_RELATED. */ |
| 442 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) |
| 443 | return NF_ACCEPT; |
| 444 | |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 445 | zone = nf_ct_zone(ct); |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 446 | /* reuse the hash saved before */ |
| 447 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; |
| 448 | hash = hash_bucket(hash, net); |
| 449 | repl_hash = hash_conntrack(net, zone, |
| 450 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 451 | |
| 452 | /* We're not in hash table, and we refuse to set up related |
| 453 | connections for unconfirmed conns. But packet copies and |
| 454 | REJECT will give spurious warnings here. */ |
| 455 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ |
| 456 | |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 457 | /* No external references means no one else could have |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 458 | confirmed us. */ |
| 459 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 460 | pr_debug("Confirming conntrack %p\n", ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 461 | |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 462 | spin_lock_bh(&nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 463 | |
Joerg Marx | fc35077 | 2010-05-20 15:55:30 +0200 | [diff] [blame] | 464 | /* We have to check the DYING flag inside the lock to prevent |
| 465 | a race against nf_ct_get_next_corpse() possibly called from |
| 466 | user context, else we insert an already 'dead' hash, blocking |
| 467 | further use of that particular connection -JM */ |
| 468 | |
| 469 | if (unlikely(nf_ct_is_dying(ct))) { |
| 470 | spin_unlock_bh(&nf_conntrack_lock); |
| 471 | return NF_ACCEPT; |
| 472 | } |
| 473 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 474 | /* See if there's one in the list already, including reverse: |
| 475 | NAT could have grabbed it without realizing, since we're |
| 476 | not in the hash. If there is, we lost race. */ |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 477 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 478 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 479 | &h->tuple) && |
| 480 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 481 | goto out; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 482 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 483 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 484 | &h->tuple) && |
| 485 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 486 | goto out; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 487 | |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 488 | /* Remove from unconfirmed list */ |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 489 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 490 | |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 491 | /* Timer relative to confirmation time, not original |
| 492 | setting time, otherwise we'd get timer wrap in |
| 493 | weird delay cases. */ |
| 494 | ct->timeout.expires += jiffies; |
| 495 | add_timer(&ct->timeout); |
| 496 | atomic_inc(&ct->ct_general.use); |
Changli Gao | 45eec34 | 2011-01-18 15:08:13 +0100 | [diff] [blame] | 497 | ct->status |= IPS_CONFIRMED; |
Patrick McHardy | 5c8ec91 | 2009-06-22 14:14:16 +0200 | [diff] [blame] | 498 | |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 499 | /* set conntrack timestamp, if enabled. */ |
| 500 | tstamp = nf_conn_tstamp_find(ct); |
| 501 | if (tstamp) { |
| 502 | if (skb->tstamp.tv64 == 0) |
| 503 | __net_timestamp((struct sk_buff *)skb); |
| 504 | |
| 505 | tstamp->start = ktime_to_ns(skb->tstamp); |
| 506 | } |
Patrick McHardy | 5c8ec91 | 2009-06-22 14:14:16 +0200 | [diff] [blame] | 507 | /* Since the lookup is lockless, hash insertion must be done after |
| 508 | * starting the timer and setting the CONFIRMED bit. The RCU barriers |
| 509 | * guarantee that no other CPU can find the conntrack before the above |
| 510 | * stores are visible. |
| 511 | */ |
| 512 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 513 | NF_CT_STAT_INC(net, insert); |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 514 | spin_unlock_bh(&nf_conntrack_lock); |
Patrick McHardy | 5c8ec91 | 2009-06-22 14:14:16 +0200 | [diff] [blame] | 515 | |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 516 | help = nfct_help(ct); |
| 517 | if (help && help->helper) |
Alexey Dobriyan | a71996f | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 518 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
Pablo Neira Ayuso | 17e6e4e | 2009-06-02 20:08:46 +0200 | [diff] [blame] | 519 | |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 520 | nf_conntrack_event_cache(master_ct(ct) ? |
Alexey Dobriyan | a71996f | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 521 | IPCT_RELATED : IPCT_NEW, ct); |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 522 | return NF_ACCEPT; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 523 | |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 524 | out: |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 525 | NF_CT_STAT_INC(net, insert_failed); |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 526 | spin_unlock_bh(&nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 527 | return NF_DROP; |
| 528 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 529 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 530 | |
| 531 | /* Returns true if a connection correspondings to the tuple (required |
| 532 | for NAT). */ |
| 533 | int |
| 534 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, |
| 535 | const struct nf_conn *ignored_conntrack) |
| 536 | { |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 537 | struct net *net = nf_ct_net(ignored_conntrack); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 538 | struct nf_conntrack_tuple_hash *h; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 539 | struct hlist_nulls_node *n; |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 540 | struct nf_conn *ct; |
| 541 | u16 zone = nf_ct_zone(ignored_conntrack); |
| 542 | unsigned int hash = hash_conntrack(net, zone, tuple); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 543 | |
Patrick McHardy | 4e29e9e | 2008-02-27 12:07:47 -0800 | [diff] [blame] | 544 | /* Disable BHs the entire time since we need to disable them at |
| 545 | * least once for the stats anyway. |
| 546 | */ |
| 547 | rcu_read_lock_bh(); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 548 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 549 | ct = nf_ct_tuplehash_to_ctrack(h); |
| 550 | if (ct != ignored_conntrack && |
| 551 | nf_ct_tuple_equal(tuple, &h->tuple) && |
| 552 | nf_ct_zone(ct) == zone) { |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 553 | NF_CT_STAT_INC(net, found); |
Patrick McHardy | 4e29e9e | 2008-02-27 12:07:47 -0800 | [diff] [blame] | 554 | rcu_read_unlock_bh(); |
Patrick McHardy | ba419af | 2008-01-31 04:39:23 -0800 | [diff] [blame] | 555 | return 1; |
| 556 | } |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 557 | NF_CT_STAT_INC(net, searched); |
Patrick McHardy | ba419af | 2008-01-31 04:39:23 -0800 | [diff] [blame] | 558 | } |
Patrick McHardy | 4e29e9e | 2008-02-27 12:07:47 -0800 | [diff] [blame] | 559 | rcu_read_unlock_bh(); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 560 | |
Patrick McHardy | ba419af | 2008-01-31 04:39:23 -0800 | [diff] [blame] | 561 | return 0; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 562 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 563 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 564 | |
Patrick McHardy | 7ae7730 | 2007-07-07 22:37:38 -0700 | [diff] [blame] | 565 | #define NF_CT_EVICTION_RANGE 8 |
| 566 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 567 | /* There's a small race here where we may free a just-assured |
| 568 | connection. Too bad: we're in trouble anyway. */ |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 569 | static noinline int early_drop(struct net *net, unsigned int hash) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 570 | { |
Patrick McHardy | f205c5e | 2007-07-07 22:28:14 -0700 | [diff] [blame] | 571 | /* Use oldest entry, which is roughly LRU */ |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 572 | struct nf_conntrack_tuple_hash *h; |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 573 | struct nf_conn *ct = NULL, *tmp; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 574 | struct hlist_nulls_node *n; |
Patrick McHardy | 7ae7730 | 2007-07-07 22:37:38 -0700 | [diff] [blame] | 575 | unsigned int i, cnt = 0; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 576 | int dropped = 0; |
| 577 | |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 578 | rcu_read_lock(); |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 579 | for (i = 0; i < net->ct.htable_size; i++) { |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 580 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
| 581 | hnnode) { |
Patrick McHardy | 7ae7730 | 2007-07-07 22:37:38 -0700 | [diff] [blame] | 582 | tmp = nf_ct_tuplehash_to_ctrack(h); |
| 583 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) |
| 584 | ct = tmp; |
| 585 | cnt++; |
| 586 | } |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 587 | |
Changli Gao | 5ae27aa | 2009-11-05 14:51:31 +0100 | [diff] [blame] | 588 | if (ct != NULL) { |
| 589 | if (likely(!nf_ct_is_dying(ct) && |
| 590 | atomic_inc_not_zero(&ct->ct_general.use))) |
| 591 | break; |
| 592 | else |
| 593 | ct = NULL; |
| 594 | } |
| 595 | |
| 596 | if (cnt >= NF_CT_EVICTION_RANGE) |
Patrick McHardy | 7ae7730 | 2007-07-07 22:37:38 -0700 | [diff] [blame] | 597 | break; |
Changli Gao | 5ae27aa | 2009-11-05 14:51:31 +0100 | [diff] [blame] | 598 | |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 599 | hash = (hash + 1) % net->ct.htable_size; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 600 | } |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 601 | rcu_read_unlock(); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 602 | |
| 603 | if (!ct) |
| 604 | return dropped; |
| 605 | |
| 606 | if (del_timer(&ct->timeout)) { |
| 607 | death_by_timeout((unsigned long)ct); |
| 608 | dropped = 1; |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 609 | NF_CT_STAT_INC_ATOMIC(net, early_drop); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 610 | } |
| 611 | nf_ct_put(ct); |
| 612 | return dropped; |
| 613 | } |
| 614 | |
Changli Gao | f682cef | 2011-01-05 04:23:23 +0000 | [diff] [blame] | 615 | void init_nf_conntrack_hash_rnd(void) |
| 616 | { |
| 617 | unsigned int rand; |
| 618 | |
| 619 | /* |
| 620 | * Why not initialize nf_conntrack_rnd in a "init()" function ? |
| 621 | * Because there isn't enough entropy when system initializing, |
| 622 | * and we initialize it as late as possible. |
| 623 | */ |
| 624 | do { |
| 625 | get_random_bytes(&rand, sizeof(rand)); |
| 626 | } while (!rand); |
| 627 | cmpxchg(&nf_conntrack_hash_rnd, 0, rand); |
| 628 | } |
| 629 | |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 630 | static struct nf_conn * |
| 631 | __nf_conntrack_alloc(struct net *net, u16 zone, |
| 632 | const struct nf_conntrack_tuple *orig, |
| 633 | const struct nf_conntrack_tuple *repl, |
| 634 | gfp_t gfp, u32 hash) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 635 | { |
Julia Lawall | cd7fcbf | 2009-01-12 00:06:08 +0000 | [diff] [blame] | 636 | struct nf_conn *ct; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 637 | |
Changli Gao | b239096 | 2010-09-16 19:55:03 +0200 | [diff] [blame] | 638 | if (unlikely(!nf_conntrack_hash_rnd)) { |
Changli Gao | f682cef | 2011-01-05 04:23:23 +0000 | [diff] [blame] | 639 | init_nf_conntrack_hash_rnd(); |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 640 | /* recompute the hash as nf_conntrack_hash_rnd is initialized */ |
| 641 | hash = hash_conntrack_raw(orig, zone); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 642 | } |
| 643 | |
Pablo Neira Ayuso | 5251e2d | 2006-09-20 12:01:06 -0700 | [diff] [blame] | 644 | /* We don't want any race condition at early drop stage */ |
Alexey Dobriyan | 49ac871 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 645 | atomic_inc(&net->ct.count); |
Pablo Neira Ayuso | 5251e2d | 2006-09-20 12:01:06 -0700 | [diff] [blame] | 646 | |
Patrick McHardy | 76eb946 | 2008-01-31 04:41:44 -0800 | [diff] [blame] | 647 | if (nf_conntrack_max && |
Alexey Dobriyan | 49ac871 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 648 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 649 | if (!early_drop(net, hash_bucket(hash, net))) { |
Alexey Dobriyan | 49ac871 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 650 | atomic_dec(&net->ct.count); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 651 | if (net_ratelimit()) |
| 652 | printk(KERN_WARNING |
| 653 | "nf_conntrack: table full, dropping" |
| 654 | " packet.\n"); |
| 655 | return ERR_PTR(-ENOMEM); |
| 656 | } |
| 657 | } |
| 658 | |
Eric Dumazet | 941297f | 2009-07-16 14:03:40 +0200 | [diff] [blame] | 659 | /* |
| 660 | * Do not use kmem_cache_zalloc(), as this cache uses |
| 661 | * SLAB_DESTROY_BY_RCU. |
| 662 | */ |
Eric Dumazet | 5b3501f | 2010-02-08 11:16:56 -0800 | [diff] [blame] | 663 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 664 | if (ct == NULL) { |
Alexey Dobriyan | 49ac871 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 665 | atomic_dec(&net->ct.count); |
Yasuyuki Kozakai | dacd2a1 | 2007-07-07 22:25:51 -0700 | [diff] [blame] | 666 | return ERR_PTR(-ENOMEM); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 667 | } |
Eric Dumazet | 941297f | 2009-07-16 14:03:40 +0200 | [diff] [blame] | 668 | /* |
| 669 | * Let ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.next |
| 670 | * and ct->tuplehash[IP_CT_DIR_REPLY].hnnode.next unchanged. |
| 671 | */ |
| 672 | memset(&ct->tuplehash[IP_CT_DIR_MAX], 0, |
Changli Gao | e5fc9e7 | 2010-11-12 17:33:17 +0100 | [diff] [blame] | 673 | offsetof(struct nf_conn, proto) - |
| 674 | offsetof(struct nf_conn, tuplehash[IP_CT_DIR_MAX])); |
Patrick McHardy | 440f0d5 | 2009-06-10 14:32:47 +0200 | [diff] [blame] | 675 | spin_lock_init(&ct->lock); |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 676 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
Eric Dumazet | 941297f | 2009-07-16 14:03:40 +0200 | [diff] [blame] | 677 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 678 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 679 | /* save hash for reusing when confirming */ |
| 680 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 681 | /* Don't set timer yet: wait for confirmation */ |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 682 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
Eric Dumazet | c2d9ba9 | 2010-06-01 06:51:19 +0000 | [diff] [blame] | 683 | write_pnet(&ct->ct_net, net); |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 684 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
| 685 | if (zone) { |
| 686 | struct nf_conntrack_zone *nf_ct_zone; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 687 | |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 688 | nf_ct_zone = nf_ct_ext_add(ct, NF_CT_EXT_ZONE, GFP_ATOMIC); |
| 689 | if (!nf_ct_zone) |
| 690 | goto out_free; |
| 691 | nf_ct_zone->id = zone; |
| 692 | } |
| 693 | #endif |
Eric Dumazet | 941297f | 2009-07-16 14:03:40 +0200 | [diff] [blame] | 694 | /* |
| 695 | * changes to lookup keys must be done before setting refcnt to 1 |
| 696 | */ |
| 697 | smp_wmb(); |
| 698 | atomic_set(&ct->ct_general.use, 1); |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 699 | return ct; |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 700 | |
| 701 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
| 702 | out_free: |
| 703 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
| 704 | return ERR_PTR(-ENOMEM); |
| 705 | #endif |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 706 | } |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 707 | |
| 708 | struct nf_conn *nf_conntrack_alloc(struct net *net, u16 zone, |
| 709 | const struct nf_conntrack_tuple *orig, |
| 710 | const struct nf_conntrack_tuple *repl, |
| 711 | gfp_t gfp) |
| 712 | { |
| 713 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); |
| 714 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 715 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 716 | |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 717 | void nf_conntrack_free(struct nf_conn *ct) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 718 | { |
Eric Dumazet | 1d45209 | 2009-03-24 14:26:50 +0100 | [diff] [blame] | 719 | struct net *net = nf_ct_net(ct); |
| 720 | |
Patrick McHardy | ceeff75 | 2008-06-11 17:51:10 -0700 | [diff] [blame] | 721 | nf_ct_ext_destroy(ct); |
Eric Dumazet | 1d45209 | 2009-03-24 14:26:50 +0100 | [diff] [blame] | 722 | atomic_dec(&net->ct.count); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 723 | nf_ct_ext_free(ct); |
Eric Dumazet | 5b3501f | 2010-02-08 11:16:56 -0800 | [diff] [blame] | 724 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 725 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 726 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 727 | |
| 728 | /* Allocate a new conntrack: we return -ENOMEM if classification |
| 729 | failed due to stress. Otherwise it really is unclassifiable. */ |
| 730 | static struct nf_conntrack_tuple_hash * |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 731 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
Alexey Dobriyan | 5a1fb39 | 2008-10-08 11:35:02 +0200 | [diff] [blame] | 732 | const struct nf_conntrack_tuple *tuple, |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 733 | struct nf_conntrack_l3proto *l3proto, |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 734 | struct nf_conntrack_l4proto *l4proto, |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 735 | struct sk_buff *skb, |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 736 | unsigned int dataoff, u32 hash) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 737 | { |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 738 | struct nf_conn *ct; |
Patrick McHarrdy | 3c158f7 | 2007-06-05 12:55:27 -0700 | [diff] [blame] | 739 | struct nf_conn_help *help; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 740 | struct nf_conntrack_tuple repl_tuple; |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 741 | struct nf_conntrack_ecache *ecache; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 742 | struct nf_conntrack_expect *exp; |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 743 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 744 | |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 745 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 746 | pr_debug("Can't invert tuple.\n"); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 747 | return NULL; |
| 748 | } |
| 749 | |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 750 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
| 751 | hash); |
Joe Perches | 0a9ee81 | 2011-08-29 14:17:25 -0700 | [diff] [blame] | 752 | if (IS_ERR(ct)) |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 753 | return (struct nf_conntrack_tuple_hash *)ct; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 754 | |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 755 | if (!l4proto->new(ct, skb, dataoff)) { |
| 756 | nf_conntrack_free(ct); |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 757 | pr_debug("init conntrack: can't track with proto module\n"); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 758 | return NULL; |
| 759 | } |
| 760 | |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 761 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 762 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 763 | |
| 764 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; |
| 765 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, |
| 766 | ecache ? ecache->expmask : 0, |
| 767 | GFP_ATOMIC); |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 768 | |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 769 | spin_lock_bh(&nf_conntrack_lock); |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 770 | exp = nf_ct_find_expectation(net, zone, tuple); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 771 | if (exp) { |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 772 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 773 | ct, exp); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 774 | /* Welcome, Mr. Bond. We've been expecting you... */ |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 775 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
| 776 | ct->master = exp->master; |
Yasuyuki Kozakai | ceceae1 | 2007-07-07 22:23:42 -0700 | [diff] [blame] | 777 | if (exp->helper) { |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 778 | help = nf_ct_helper_ext_add(ct, GFP_ATOMIC); |
Yasuyuki Kozakai | ceceae1 | 2007-07-07 22:23:42 -0700 | [diff] [blame] | 779 | if (help) |
Eric Dumazet | cf778b0 | 2012-01-12 04:41:32 +0000 | [diff] [blame] | 780 | rcu_assign_pointer(help->helper, exp->helper); |
Yasuyuki Kozakai | ceceae1 | 2007-07-07 22:23:42 -0700 | [diff] [blame] | 781 | } |
| 782 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 783 | #ifdef CONFIG_NF_CONNTRACK_MARK |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 784 | ct->mark = exp->master->mark; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 785 | #endif |
James Morris | 7c9728c | 2006-06-09 00:31:46 -0700 | [diff] [blame] | 786 | #ifdef CONFIG_NF_CONNTRACK_SECMARK |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 787 | ct->secmark = exp->master->secmark; |
James Morris | 7c9728c | 2006-06-09 00:31:46 -0700 | [diff] [blame] | 788 | #endif |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 789 | nf_conntrack_get(&ct->master->ct_general); |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 790 | NF_CT_STAT_INC(net, expect_new); |
Yasuyuki Kozakai | 22e7410 | 2006-11-27 10:25:59 -0800 | [diff] [blame] | 791 | } else { |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 792 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 793 | NF_CT_STAT_INC(net, new); |
Yasuyuki Kozakai | 22e7410 | 2006-11-27 10:25:59 -0800 | [diff] [blame] | 794 | } |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 795 | |
| 796 | /* Overload tuple linked list to put us in unconfirmed list. */ |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 797 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
Alexey Dobriyan | 63c9a26 | 2008-10-08 11:35:04 +0200 | [diff] [blame] | 798 | &net->ct.unconfirmed); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 799 | |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 800 | spin_unlock_bh(&nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 801 | |
| 802 | if (exp) { |
| 803 | if (exp->expectfn) |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 804 | exp->expectfn(ct, exp); |
Patrick McHardy | 6823645 | 2007-07-07 22:30:49 -0700 | [diff] [blame] | 805 | nf_ct_expect_put(exp); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 806 | } |
| 807 | |
Patrick McHardy | c88130b | 2008-01-31 04:42:11 -0800 | [diff] [blame] | 808 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 809 | } |
| 810 | |
| 811 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ |
| 812 | static inline struct nf_conn * |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 813 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
Alexey Dobriyan | a702a65 | 2008-10-08 11:35:04 +0200 | [diff] [blame] | 814 | struct sk_buff *skb, |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 815 | unsigned int dataoff, |
| 816 | u_int16_t l3num, |
| 817 | u_int8_t protonum, |
| 818 | struct nf_conntrack_l3proto *l3proto, |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 819 | struct nf_conntrack_l4proto *l4proto, |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 820 | int *set_reply, |
| 821 | enum ip_conntrack_info *ctinfo) |
| 822 | { |
| 823 | struct nf_conntrack_tuple tuple; |
| 824 | struct nf_conntrack_tuple_hash *h; |
| 825 | struct nf_conn *ct; |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 826 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 827 | u32 hash; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 828 | |
Arnaldo Carvalho de Melo | bbe735e | 2007-03-10 22:16:10 -0300 | [diff] [blame] | 829 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 830 | dataoff, l3num, protonum, &tuple, l3proto, |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 831 | l4proto)) { |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 832 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 833 | return NULL; |
| 834 | } |
| 835 | |
| 836 | /* look for tuple match */ |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 837 | hash = hash_conntrack_raw(&tuple, zone); |
| 838 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 839 | if (!h) { |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 840 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 841 | skb, dataoff, hash); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 842 | if (!h) |
| 843 | return NULL; |
| 844 | if (IS_ERR(h)) |
| 845 | return (void *)h; |
| 846 | } |
| 847 | ct = nf_ct_tuplehash_to_ctrack(h); |
| 848 | |
| 849 | /* It exists; we have (non-exclusive) reference. */ |
| 850 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { |
Eric Dumazet | fb04883 | 2011-05-19 15:44:27 +0200 | [diff] [blame] | 851 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 852 | /* Please set reply bit if this packet OK */ |
| 853 | *set_reply = 1; |
| 854 | } else { |
| 855 | /* Once we've had two way comms, always ESTABLISHED. */ |
| 856 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 857 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 858 | *ctinfo = IP_CT_ESTABLISHED; |
| 859 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 860 | pr_debug("nf_conntrack_in: related packet for %p\n", |
| 861 | ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 862 | *ctinfo = IP_CT_RELATED; |
| 863 | } else { |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 864 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 865 | *ctinfo = IP_CT_NEW; |
| 866 | } |
| 867 | *set_reply = 0; |
| 868 | } |
| 869 | skb->nfct = &ct->ct_general; |
| 870 | skb->nfctinfo = *ctinfo; |
| 871 | return ct; |
| 872 | } |
| 873 | |
| 874 | unsigned int |
Alexey Dobriyan | a702a65 | 2008-10-08 11:35:04 +0200 | [diff] [blame] | 875 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
| 876 | struct sk_buff *skb) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 877 | { |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 878 | struct nf_conn *ct, *tmpl = NULL; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 879 | enum ip_conntrack_info ctinfo; |
| 880 | struct nf_conntrack_l3proto *l3proto; |
Martin Josefsson | 605dcad | 2006-11-29 02:35:06 +0100 | [diff] [blame] | 881 | struct nf_conntrack_l4proto *l4proto; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 882 | unsigned int dataoff; |
| 883 | u_int8_t protonum; |
| 884 | int set_reply = 0; |
| 885 | int ret; |
| 886 | |
Herbert Xu | 3db05fe | 2007-10-15 00:53:15 -0700 | [diff] [blame] | 887 | if (skb->nfct) { |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 888 | /* Previously seen (loopback or untracked)? Ignore. */ |
| 889 | tmpl = (struct nf_conn *)skb->nfct; |
| 890 | if (!nf_ct_is_template(tmpl)) { |
| 891 | NF_CT_STAT_INC_ATOMIC(net, ignore); |
| 892 | return NF_ACCEPT; |
| 893 | } |
| 894 | skb->nfct = NULL; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 895 | } |
| 896 | |
Patrick McHardy | 923f490 | 2007-02-12 11:12:57 -0800 | [diff] [blame] | 897 | /* rcu_read_lock()ed by nf_hook_slow */ |
Jan Engelhardt | 76108ce | 2008-10-08 11:35:00 +0200 | [diff] [blame] | 898 | l3proto = __nf_ct_l3proto_find(pf); |
Herbert Xu | 3db05fe | 2007-10-15 00:53:15 -0700 | [diff] [blame] | 899 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
Yasuyuki Kozakai | ffc3069 | 2007-07-14 20:44:50 -0700 | [diff] [blame] | 900 | &dataoff, &protonum); |
| 901 | if (ret <= 0) { |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 902 | pr_debug("not prepared to track yet or error occurred\n"); |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 903 | NF_CT_STAT_INC_ATOMIC(net, error); |
| 904 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 905 | ret = -ret; |
| 906 | goto out; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 907 | } |
| 908 | |
Jan Engelhardt | 76108ce | 2008-10-08 11:35:00 +0200 | [diff] [blame] | 909 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 910 | |
| 911 | /* It may be an special packet, error, unclean... |
| 912 | * inverse of the return code tells to the netfilter |
| 913 | * core what to do with the packet. */ |
Alexey Dobriyan | 74c51a1 | 2008-10-08 11:35:05 +0200 | [diff] [blame] | 914 | if (l4proto->error != NULL) { |
Patrick McHardy | 8fea97e | 2010-02-15 17:45:08 +0100 | [diff] [blame] | 915 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
| 916 | pf, hooknum); |
Alexey Dobriyan | 74c51a1 | 2008-10-08 11:35:05 +0200 | [diff] [blame] | 917 | if (ret <= 0) { |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 918 | NF_CT_STAT_INC_ATOMIC(net, error); |
| 919 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 920 | ret = -ret; |
| 921 | goto out; |
Alexey Dobriyan | 74c51a1 | 2008-10-08 11:35:05 +0200 | [diff] [blame] | 922 | } |
Pablo Neira Ayuso | 88ed01d | 2011-06-02 15:08:45 +0200 | [diff] [blame] | 923 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
| 924 | if (skb->nfct) |
| 925 | goto out; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 926 | } |
| 927 | |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 928 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
Alexey Dobriyan | a702a65 | 2008-10-08 11:35:04 +0200 | [diff] [blame] | 929 | l3proto, l4proto, &set_reply, &ctinfo); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 930 | if (!ct) { |
| 931 | /* Not valid part of a connection */ |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 932 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 933 | ret = NF_ACCEPT; |
| 934 | goto out; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 935 | } |
| 936 | |
| 937 | if (IS_ERR(ct)) { |
| 938 | /* Too stressed to deal. */ |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 939 | NF_CT_STAT_INC_ATOMIC(net, drop); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 940 | ret = NF_DROP; |
| 941 | goto out; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 942 | } |
| 943 | |
Herbert Xu | 3db05fe | 2007-10-15 00:53:15 -0700 | [diff] [blame] | 944 | NF_CT_ASSERT(skb->nfct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 945 | |
Herbert Xu | 3db05fe | 2007-10-15 00:53:15 -0700 | [diff] [blame] | 946 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum); |
Christoph Paasch | ec8d540 | 2009-03-16 15:51:29 +0100 | [diff] [blame] | 947 | if (ret <= 0) { |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 948 | /* Invalid: inverse of the return code tells |
| 949 | * the netfilter core what to do */ |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 950 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
Herbert Xu | 3db05fe | 2007-10-15 00:53:15 -0700 | [diff] [blame] | 951 | nf_conntrack_put(skb->nfct); |
| 952 | skb->nfct = NULL; |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 953 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
Pablo Neira Ayuso | 7d1e045 | 2009-02-24 14:48:01 +0100 | [diff] [blame] | 954 | if (ret == -NF_DROP) |
| 955 | NF_CT_STAT_INC_ATOMIC(net, drop); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 956 | ret = -ret; |
| 957 | goto out; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 958 | } |
| 959 | |
| 960 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) |
Patrick McHardy | 858b3133 | 2010-02-03 13:48:53 +0100 | [diff] [blame] | 961 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 962 | out: |
Pablo Neira Ayuso | c317428 | 2011-02-09 08:08:20 +0100 | [diff] [blame] | 963 | if (tmpl) { |
| 964 | /* Special case: we have to repeat this hook, assign the |
| 965 | * template again to this packet. We assume that this packet |
| 966 | * has no conntrack assigned. This is used by nf_ct_tcp. */ |
| 967 | if (ret == NF_REPEAT) |
| 968 | skb->nfct = (struct nf_conntrack *)tmpl; |
| 969 | else |
| 970 | nf_ct_put(tmpl); |
| 971 | } |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 972 | |
| 973 | return ret; |
| 974 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 975 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 976 | |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 977 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
| 978 | const struct nf_conntrack_tuple *orig) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 979 | { |
Jan Engelhardt | 5f2b4c9 | 2008-04-14 11:15:53 +0200 | [diff] [blame] | 980 | bool ret; |
Patrick McHardy | 923f490 | 2007-02-12 11:12:57 -0800 | [diff] [blame] | 981 | |
| 982 | rcu_read_lock(); |
| 983 | ret = nf_ct_invert_tuple(inverse, orig, |
| 984 | __nf_ct_l3proto_find(orig->src.l3num), |
| 985 | __nf_ct_l4proto_find(orig->src.l3num, |
| 986 | orig->dst.protonum)); |
| 987 | rcu_read_unlock(); |
| 988 | return ret; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 989 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 990 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 991 | |
Jozsef Kadlecsik | 5b1158e | 2006-12-02 22:07:13 -0800 | [diff] [blame] | 992 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
| 993 | implicitly racy: see __nf_conntrack_confirm */ |
| 994 | void nf_conntrack_alter_reply(struct nf_conn *ct, |
| 995 | const struct nf_conntrack_tuple *newreply) |
| 996 | { |
| 997 | struct nf_conn_help *help = nfct_help(ct); |
| 998 | |
Jozsef Kadlecsik | 5b1158e | 2006-12-02 22:07:13 -0800 | [diff] [blame] | 999 | /* Should be unconfirmed, so not in hash table yet */ |
| 1000 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
| 1001 | |
Patrick McHardy | 0d53778 | 2007-07-07 22:39:38 -0700 | [diff] [blame] | 1002 | pr_debug("Altering reply tuple of %p to ", ct); |
Jan Engelhardt | 3c9fba6 | 2008-04-14 11:15:54 +0200 | [diff] [blame] | 1003 | nf_ct_dump_tuple(newreply); |
Jozsef Kadlecsik | 5b1158e | 2006-12-02 22:07:13 -0800 | [diff] [blame] | 1004 | |
| 1005 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; |
Patrick McHardy | ef1a5a5 | 2008-04-14 11:21:01 +0200 | [diff] [blame] | 1006 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
Patrick McHardy | c52fbb4 | 2008-01-31 04:37:36 -0800 | [diff] [blame] | 1007 | return; |
Yasuyuki Kozakai | ceceae1 | 2007-07-07 22:23:42 -0700 | [diff] [blame] | 1008 | |
Patrick McHardy | c52fbb4 | 2008-01-31 04:37:36 -0800 | [diff] [blame] | 1009 | rcu_read_lock(); |
Patrick McHardy | b2a15a6 | 2010-02-03 14:13:03 +0100 | [diff] [blame] | 1010 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
Patrick McHardy | c52fbb4 | 2008-01-31 04:37:36 -0800 | [diff] [blame] | 1011 | rcu_read_unlock(); |
Jozsef Kadlecsik | 5b1158e | 2006-12-02 22:07:13 -0800 | [diff] [blame] | 1012 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 1013 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
Jozsef Kadlecsik | 5b1158e | 2006-12-02 22:07:13 -0800 | [diff] [blame] | 1014 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1015 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
| 1016 | void __nf_ct_refresh_acct(struct nf_conn *ct, |
| 1017 | enum ip_conntrack_info ctinfo, |
| 1018 | const struct sk_buff *skb, |
| 1019 | unsigned long extra_jiffies, |
| 1020 | int do_acct) |
| 1021 | { |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1022 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
| 1023 | NF_CT_ASSERT(skb); |
| 1024 | |
Eric Leblond | 997ae83 | 2006-05-29 18:24:20 -0700 | [diff] [blame] | 1025 | /* Only update if this is not a fixed timeout */ |
Patrick McHardy | 47d9504 | 2008-01-31 04:36:31 -0800 | [diff] [blame] | 1026 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
| 1027 | goto acct; |
Eric Leblond | 997ae83 | 2006-05-29 18:24:20 -0700 | [diff] [blame] | 1028 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1029 | /* If not in hash table, timer will not be active yet */ |
| 1030 | if (!nf_ct_is_confirmed(ct)) { |
| 1031 | ct->timeout.expires = extra_jiffies; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1032 | } else { |
Martin Josefsson | be00c8e | 2006-11-29 02:35:12 +0100 | [diff] [blame] | 1033 | unsigned long newtime = jiffies + extra_jiffies; |
| 1034 | |
| 1035 | /* Only update the timeout if the new timeout is at least |
| 1036 | HZ jiffies from the old timeout. Need del_timer for race |
| 1037 | avoidance (may already be dying). */ |
Patrick McHardy | 65cb9fd | 2009-06-13 12:21:49 +0200 | [diff] [blame] | 1038 | if (newtime - ct->timeout.expires >= HZ) |
| 1039 | mod_timer_pending(&ct->timeout, newtime); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1040 | } |
| 1041 | |
Patrick McHardy | 47d9504 | 2008-01-31 04:36:31 -0800 | [diff] [blame] | 1042 | acct: |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1043 | if (do_acct) { |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1044 | struct nf_conn_counter *acct; |
Martin Josefsson | 3ffd5ee | 2006-11-29 02:35:10 +0100 | [diff] [blame] | 1045 | |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1046 | acct = nf_conn_acct_find(ct); |
| 1047 | if (acct) { |
Eric Dumazet | b3e0bfa | 2011-12-14 14:45:20 +0100 | [diff] [blame] | 1048 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); |
| 1049 | atomic64_add(skb->len, &acct[CTINFO2DIR(ctinfo)].bytes); |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1050 | } |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1051 | } |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1052 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 1053 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1054 | |
David S. Miller | 4c88949 | 2008-07-14 20:22:38 -0700 | [diff] [blame] | 1055 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
| 1056 | enum ip_conntrack_info ctinfo, |
| 1057 | const struct sk_buff *skb, |
| 1058 | int do_acct) |
Patrick McHardy | 5109176 | 2008-06-09 15:59:06 -0700 | [diff] [blame] | 1059 | { |
Fabian Hugelshofer | 718d4ad | 2008-06-09 15:59:40 -0700 | [diff] [blame] | 1060 | if (do_acct) { |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1061 | struct nf_conn_counter *acct; |
| 1062 | |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1063 | acct = nf_conn_acct_find(ct); |
| 1064 | if (acct) { |
Eric Dumazet | b3e0bfa | 2011-12-14 14:45:20 +0100 | [diff] [blame] | 1065 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); |
| 1066 | atomic64_add(skb->len - skb_network_offset(skb), |
| 1067 | &acct[CTINFO2DIR(ctinfo)].bytes); |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1068 | } |
Fabian Hugelshofer | 718d4ad | 2008-06-09 15:59:40 -0700 | [diff] [blame] | 1069 | } |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1070 | |
David S. Miller | 4c88949 | 2008-07-14 20:22:38 -0700 | [diff] [blame] | 1071 | if (del_timer(&ct->timeout)) { |
Patrick McHardy | 5109176 | 2008-06-09 15:59:06 -0700 | [diff] [blame] | 1072 | ct->timeout.function((unsigned long)ct); |
David S. Miller | 4c88949 | 2008-07-14 20:22:38 -0700 | [diff] [blame] | 1073 | return true; |
| 1074 | } |
| 1075 | return false; |
Patrick McHardy | 5109176 | 2008-06-09 15:59:06 -0700 | [diff] [blame] | 1076 | } |
Fabian Hugelshofer | 718d4ad | 2008-06-09 15:59:40 -0700 | [diff] [blame] | 1077 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
Patrick McHardy | 5109176 | 2008-06-09 15:59:06 -0700 | [diff] [blame] | 1078 | |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 1079 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
| 1080 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { |
| 1081 | .len = sizeof(struct nf_conntrack_zone), |
| 1082 | .align = __alignof__(struct nf_conntrack_zone), |
| 1083 | .id = NF_CT_EXT_ZONE, |
| 1084 | }; |
| 1085 | #endif |
| 1086 | |
Igor Maravić | c0cd115 | 2011-12-12 02:58:24 +0000 | [diff] [blame] | 1087 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1088 | |
| 1089 | #include <linux/netfilter/nfnetlink.h> |
| 1090 | #include <linux/netfilter/nfnetlink_conntrack.h> |
Ingo Molnar | 57b47a5 | 2006-03-20 22:35:41 -0800 | [diff] [blame] | 1091 | #include <linux/mutex.h> |
| 1092 | |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1093 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
| 1094 | * in ip_conntrack_core, since we don't want the protocols to autoload |
| 1095 | * or depend on ctnetlink */ |
Patrick McHardy | fdf7083 | 2007-09-28 14:37:41 -0700 | [diff] [blame] | 1096 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1097 | const struct nf_conntrack_tuple *tuple) |
| 1098 | { |
Patrick McHardy | 77236b6 | 2007-12-17 22:29:45 -0800 | [diff] [blame] | 1099 | NLA_PUT_BE16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port); |
| 1100 | NLA_PUT_BE16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1101 | return 0; |
| 1102 | |
Patrick McHardy | df6fb86 | 2007-09-28 14:37:03 -0700 | [diff] [blame] | 1103 | nla_put_failure: |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1104 | return -1; |
| 1105 | } |
Patrick McHardy | fdf7083 | 2007-09-28 14:37:41 -0700 | [diff] [blame] | 1106 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1107 | |
Patrick McHardy | f73e924 | 2007-09-28 14:39:55 -0700 | [diff] [blame] | 1108 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
| 1109 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, |
| 1110 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1111 | }; |
Patrick McHardy | f73e924 | 2007-09-28 14:39:55 -0700 | [diff] [blame] | 1112 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1113 | |
Patrick McHardy | fdf7083 | 2007-09-28 14:37:41 -0700 | [diff] [blame] | 1114 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1115 | struct nf_conntrack_tuple *t) |
| 1116 | { |
Patrick McHardy | df6fb86 | 2007-09-28 14:37:03 -0700 | [diff] [blame] | 1117 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1118 | return -EINVAL; |
| 1119 | |
Patrick McHardy | 77236b6 | 2007-12-17 22:29:45 -0800 | [diff] [blame] | 1120 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
| 1121 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1122 | |
| 1123 | return 0; |
| 1124 | } |
Patrick McHardy | fdf7083 | 2007-09-28 14:37:41 -0700 | [diff] [blame] | 1125 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
Holger Eitzenberger | 5c0de29 | 2009-03-25 21:52:17 +0100 | [diff] [blame] | 1126 | |
| 1127 | int nf_ct_port_nlattr_tuple_size(void) |
| 1128 | { |
| 1129 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); |
| 1130 | } |
| 1131 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1132 | #endif |
| 1133 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1134 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
Patrick McHardy | b334aad | 2008-01-14 23:48:57 -0800 | [diff] [blame] | 1135 | static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1136 | { |
| 1137 | struct nf_conn *ct; |
| 1138 | enum ip_conntrack_info ctinfo; |
| 1139 | |
| 1140 | /* This ICMP is in reverse direction to the packet which caused it */ |
| 1141 | ct = nf_ct_get(skb, &ctinfo); |
| 1142 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) |
Eric Dumazet | fb04883 | 2011-05-19 15:44:27 +0200 | [diff] [blame] | 1143 | ctinfo = IP_CT_RELATED_REPLY; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1144 | else |
| 1145 | ctinfo = IP_CT_RELATED; |
| 1146 | |
| 1147 | /* Attach to new skbuff, and increment count */ |
| 1148 | nskb->nfct = &ct->ct_general; |
| 1149 | nskb->nfctinfo = ctinfo; |
| 1150 | nf_conntrack_get(nskb->nfct); |
| 1151 | } |
| 1152 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1153 | /* Bring out ya dead! */ |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1154 | static struct nf_conn * |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1155 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1156 | void *data, unsigned int *bucket) |
| 1157 | { |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1158 | struct nf_conntrack_tuple_hash *h; |
| 1159 | struct nf_conn *ct; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1160 | struct hlist_nulls_node *n; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1161 | |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 1162 | spin_lock_bh(&nf_conntrack_lock); |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 1163 | for (; *bucket < net->ct.htable_size; (*bucket)++) { |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1164 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1165 | ct = nf_ct_tuplehash_to_ctrack(h); |
| 1166 | if (iter(ct, data)) |
| 1167 | goto found; |
| 1168 | } |
YOSHIFUJI Hideaki | 601e68e | 2007-02-12 11:15:49 -0800 | [diff] [blame] | 1169 | } |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1170 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1171 | ct = nf_ct_tuplehash_to_ctrack(h); |
| 1172 | if (iter(ct, data)) |
Patrick McHardy | ec68e97 | 2007-03-04 15:57:01 -0800 | [diff] [blame] | 1173 | set_bit(IPS_DYING_BIT, &ct->status); |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1174 | } |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 1175 | spin_unlock_bh(&nf_conntrack_lock); |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1176 | return NULL; |
| 1177 | found: |
Martin Josefsson | c073e3f | 2006-10-30 15:13:58 -0800 | [diff] [blame] | 1178 | atomic_inc(&ct->ct_general.use); |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 1179 | spin_unlock_bh(&nf_conntrack_lock); |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1180 | return ct; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1181 | } |
| 1182 | |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1183 | void nf_ct_iterate_cleanup(struct net *net, |
| 1184 | int (*iter)(struct nf_conn *i, void *data), |
| 1185 | void *data) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1186 | { |
Patrick McHardy | df0933d | 2006-09-20 11:57:53 -0700 | [diff] [blame] | 1187 | struct nf_conn *ct; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1188 | unsigned int bucket = 0; |
| 1189 | |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1190 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1191 | /* Time to push up daises... */ |
| 1192 | if (del_timer(&ct->timeout)) |
| 1193 | death_by_timeout((unsigned long)ct); |
| 1194 | /* ... else the timer will get him soon. */ |
| 1195 | |
| 1196 | nf_ct_put(ct); |
| 1197 | } |
| 1198 | } |
Patrick McHardy | 13b1833 | 2006-12-02 22:11:25 -0800 | [diff] [blame] | 1199 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1200 | |
Pablo Neira Ayuso | 19abb7b | 2008-11-18 11:56:20 +0100 | [diff] [blame] | 1201 | struct __nf_ct_flush_report { |
| 1202 | u32 pid; |
| 1203 | int report; |
| 1204 | }; |
| 1205 | |
Pablo Neira Ayuso | 274d383 | 2009-06-02 20:08:38 +0200 | [diff] [blame] | 1206 | static int kill_report(struct nf_conn *i, void *data) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1207 | { |
Pablo Neira Ayuso | 19abb7b | 2008-11-18 11:56:20 +0100 | [diff] [blame] | 1208 | struct __nf_ct_flush_report *fr = (struct __nf_ct_flush_report *)data; |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 1209 | struct nf_conn_tstamp *tstamp; |
| 1210 | |
| 1211 | tstamp = nf_conn_tstamp_find(i); |
| 1212 | if (tstamp && tstamp->stop == 0) |
| 1213 | tstamp->stop = ktime_to_ns(ktime_get_real()); |
Pablo Neira Ayuso | 19abb7b | 2008-11-18 11:56:20 +0100 | [diff] [blame] | 1214 | |
Pablo Neira Ayuso | dd7669a | 2009-06-13 12:30:52 +0200 | [diff] [blame] | 1215 | /* If we fail to deliver the event, death_by_timeout() will retry */ |
| 1216 | if (nf_conntrack_event_report(IPCT_DESTROY, i, |
| 1217 | fr->pid, fr->report) < 0) |
| 1218 | return 1; |
| 1219 | |
| 1220 | /* Avoid the delivery of the destroy event in death_by_timeout(). */ |
| 1221 | set_bit(IPS_DYING_BIT, &i->status); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1222 | return 1; |
| 1223 | } |
| 1224 | |
Pablo Neira Ayuso | 274d383 | 2009-06-02 20:08:38 +0200 | [diff] [blame] | 1225 | static int kill_all(struct nf_conn *i, void *data) |
| 1226 | { |
| 1227 | return 1; |
| 1228 | } |
| 1229 | |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1230 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1231 | { |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1232 | if (is_vmalloc_addr(hash)) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1233 | vfree(hash); |
| 1234 | else |
YOSHIFUJI Hideaki | 601e68e | 2007-02-12 11:15:49 -0800 | [diff] [blame] | 1235 | free_pages((unsigned long)hash, |
Patrick McHardy | f205c5e | 2007-07-07 22:28:14 -0700 | [diff] [blame] | 1236 | get_order(sizeof(struct hlist_head) * size)); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1237 | } |
Patrick McHardy | ac565e5 | 2007-07-07 22:30:08 -0700 | [diff] [blame] | 1238 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1239 | |
Pablo Neira Ayuso | 274d383 | 2009-06-02 20:08:38 +0200 | [diff] [blame] | 1240 | void nf_conntrack_flush_report(struct net *net, u32 pid, int report) |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1241 | { |
Pablo Neira Ayuso | 19abb7b | 2008-11-18 11:56:20 +0100 | [diff] [blame] | 1242 | struct __nf_ct_flush_report fr = { |
| 1243 | .pid = pid, |
| 1244 | .report = report, |
| 1245 | }; |
Pablo Neira Ayuso | 274d383 | 2009-06-02 20:08:38 +0200 | [diff] [blame] | 1246 | nf_ct_iterate_cleanup(net, kill_report, &fr); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1247 | } |
Pablo Neira Ayuso | 274d383 | 2009-06-02 20:08:38 +0200 | [diff] [blame] | 1248 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
Pablo Neira Ayuso | c1d10ad | 2006-01-05 12:19:05 -0800 | [diff] [blame] | 1249 | |
Alexey Dobriyan | ee254fa | 2009-08-31 14:23:15 +0200 | [diff] [blame] | 1250 | static void nf_ct_release_dying_list(struct net *net) |
Pablo Neira Ayuso | dd7669a | 2009-06-13 12:30:52 +0200 | [diff] [blame] | 1251 | { |
| 1252 | struct nf_conntrack_tuple_hash *h; |
| 1253 | struct nf_conn *ct; |
| 1254 | struct hlist_nulls_node *n; |
| 1255 | |
| 1256 | spin_lock_bh(&nf_conntrack_lock); |
Alexey Dobriyan | ee254fa | 2009-08-31 14:23:15 +0200 | [diff] [blame] | 1257 | hlist_nulls_for_each_entry(h, n, &net->ct.dying, hnnode) { |
Pablo Neira Ayuso | dd7669a | 2009-06-13 12:30:52 +0200 | [diff] [blame] | 1258 | ct = nf_ct_tuplehash_to_ctrack(h); |
| 1259 | /* never fails to remove them, no listeners at this point */ |
| 1260 | nf_ct_kill(ct); |
| 1261 | } |
| 1262 | spin_unlock_bh(&nf_conntrack_lock); |
| 1263 | } |
| 1264 | |
Eric Dumazet | b3c5163 | 2010-06-09 14:43:38 +0200 | [diff] [blame] | 1265 | static int untrack_refs(void) |
| 1266 | { |
| 1267 | int cnt = 0, cpu; |
| 1268 | |
| 1269 | for_each_possible_cpu(cpu) { |
| 1270 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); |
| 1271 | |
| 1272 | cnt += atomic_read(&ct->ct_general.use) - 1; |
| 1273 | } |
| 1274 | return cnt; |
| 1275 | } |
| 1276 | |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1277 | static void nf_conntrack_cleanup_init_net(void) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1278 | { |
Eric Dumazet | b3c5163 | 2010-06-09 14:43:38 +0200 | [diff] [blame] | 1279 | while (untrack_refs() > 0) |
Patrick McHardy | 9edd7ca | 2010-02-08 11:16:26 -0800 | [diff] [blame] | 1280 | schedule(); |
| 1281 | |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1282 | nf_conntrack_helper_fini(); |
| 1283 | nf_conntrack_proto_fini(); |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 1284 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
| 1285 | nf_ct_extend_unregister(&nf_ct_zone_extend); |
| 1286 | #endif |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1287 | } |
Yasuyuki Kozakai | 7d3cdc6 | 2006-02-15 15:22:21 -0800 | [diff] [blame] | 1288 | |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1289 | static void nf_conntrack_cleanup_net(struct net *net) |
| 1290 | { |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1291 | i_see_dead_people: |
Pablo Neira Ayuso | 274d383 | 2009-06-02 20:08:38 +0200 | [diff] [blame] | 1292 | nf_ct_iterate_cleanup(net, kill_all, NULL); |
Alexey Dobriyan | ee254fa | 2009-08-31 14:23:15 +0200 | [diff] [blame] | 1293 | nf_ct_release_dying_list(net); |
Alexey Dobriyan | 49ac871 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1294 | if (atomic_read(&net->ct.count) != 0) { |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1295 | schedule(); |
| 1296 | goto i_see_dead_people; |
| 1297 | } |
| 1298 | |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1299 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
Pablo Neira Ayuso | a0891aa | 2009-06-13 12:26:29 +0200 | [diff] [blame] | 1300 | nf_conntrack_ecache_fini(net); |
Stephen Hemminger | fe8f661 | 2011-03-14 19:20:44 +0100 | [diff] [blame] | 1301 | nf_conntrack_tstamp_fini(net); |
Alexey Dobriyan | d716a4d | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1302 | nf_conntrack_acct_fini(net); |
Alexey Dobriyan | 9b03f38 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1303 | nf_conntrack_expect_fini(net); |
Eric Dumazet | 5b3501f | 2010-02-08 11:16:56 -0800 | [diff] [blame] | 1304 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
| 1305 | kfree(net->ct.slabname); |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 1306 | free_percpu(net->ct.stat); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1307 | } |
| 1308 | |
| 1309 | /* Mishearing the voices in his head, our hero wonders how he's |
| 1310 | supposed to kill the mall. */ |
| 1311 | void nf_conntrack_cleanup(struct net *net) |
| 1312 | { |
| 1313 | if (net_eq(net, &init_net)) |
Stephen Hemminger | a9b3cd7 | 2011-08-01 16:19:00 +0000 | [diff] [blame] | 1314 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1315 | |
| 1316 | /* This makes sure all current packets have passed through |
| 1317 | netfilter framework. Roll on, two-stage module |
| 1318 | delete... */ |
| 1319 | synchronize_net(); |
| 1320 | |
| 1321 | nf_conntrack_cleanup_net(net); |
| 1322 | |
| 1323 | if (net_eq(net, &init_net)) { |
Stephen Hemminger | a9b3cd7 | 2011-08-01 16:19:00 +0000 | [diff] [blame] | 1324 | RCU_INIT_POINTER(nf_ct_destroy, NULL); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1325 | nf_conntrack_cleanup_init_net(); |
| 1326 | } |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1327 | } |
| 1328 | |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1329 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1330 | { |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1331 | struct hlist_nulls_head *hash; |
| 1332 | unsigned int nr_slots, i; |
| 1333 | size_t sz; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1334 | |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1335 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
| 1336 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); |
| 1337 | sz = nr_slots * sizeof(struct hlist_nulls_head); |
| 1338 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, |
| 1339 | get_order(sz)); |
YOSHIFUJI Hideaki | 601e68e | 2007-02-12 11:15:49 -0800 | [diff] [blame] | 1340 | if (!hash) { |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1341 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); |
Eric Dumazet | 966567b | 2011-12-19 16:01:38 -0500 | [diff] [blame] | 1342 | hash = vzalloc(sz); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1343 | } |
| 1344 | |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1345 | if (hash && nulls) |
| 1346 | for (i = 0; i < nr_slots; i++) |
| 1347 | INIT_HLIST_NULLS_HEAD(&hash[i], i); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1348 | |
| 1349 | return hash; |
| 1350 | } |
Patrick McHardy | ac565e5 | 2007-07-07 22:30:08 -0700 | [diff] [blame] | 1351 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1352 | |
Patrick McHardy | fae718d | 2007-12-24 21:09:10 -0800 | [diff] [blame] | 1353 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1354 | { |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1355 | int i, bucket; |
Stephen Hemminger | 96eb24d | 2008-01-31 04:07:29 -0800 | [diff] [blame] | 1356 | unsigned int hashsize, old_size; |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1357 | struct hlist_nulls_head *hash, *old_hash; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1358 | struct nf_conntrack_tuple_hash *h; |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 1359 | struct nf_conn *ct; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1360 | |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 1361 | if (current->nsproxy->net_ns != &init_net) |
| 1362 | return -EOPNOTSUPP; |
| 1363 | |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1364 | /* On boot, we can set this without any fancy locking. */ |
| 1365 | if (!nf_conntrack_htable_size) |
| 1366 | return param_set_uint(val, kp); |
| 1367 | |
Stephen Hemminger | 96eb24d | 2008-01-31 04:07:29 -0800 | [diff] [blame] | 1368 | hashsize = simple_strtoul(val, NULL, 0); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1369 | if (!hashsize) |
| 1370 | return -EINVAL; |
| 1371 | |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1372 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1373 | if (!hash) |
| 1374 | return -ENOMEM; |
| 1375 | |
Patrick McHardy | 76507f6 | 2008-01-31 04:38:38 -0800 | [diff] [blame] | 1376 | /* Lookups in the old hash might happen in parallel, which means we |
| 1377 | * might get false negatives during connection lookup. New connections |
| 1378 | * created because of a false negative won't make it into the hash |
| 1379 | * though since that required taking the lock. |
| 1380 | */ |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 1381 | spin_lock_bh(&nf_conntrack_lock); |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 1382 | for (i = 0; i < init_net.ct.htable_size; i++) { |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1383 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
| 1384 | h = hlist_nulls_entry(init_net.ct.hash[i].first, |
| 1385 | struct nf_conntrack_tuple_hash, hnnode); |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 1386 | ct = nf_ct_tuplehash_to_ctrack(h); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1387 | hlist_nulls_del_rcu(&h->hnnode); |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 1388 | bucket = __hash_conntrack(&h->tuple, nf_ct_zone(ct), |
Changli Gao | 99f07e9 | 2010-09-21 17:49:20 +0200 | [diff] [blame] | 1389 | hashsize); |
Eric Dumazet | ea781f1 | 2009-03-25 21:05:46 +0100 | [diff] [blame] | 1390 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1391 | } |
| 1392 | } |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 1393 | old_size = init_net.ct.htable_size; |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1394 | old_hash = init_net.ct.hash; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1395 | |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 1396 | init_net.ct.htable_size = nf_conntrack_htable_size = hashsize; |
Alexey Dobriyan | 400dad3 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1397 | init_net.ct.hash = hash; |
Patrick McHardy | f8ba1af | 2008-01-31 04:38:58 -0800 | [diff] [blame] | 1398 | spin_unlock_bh(&nf_conntrack_lock); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1399 | |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1400 | nf_ct_free_hashtable(old_hash, old_size); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1401 | return 0; |
| 1402 | } |
Patrick McHardy | fae718d | 2007-12-24 21:09:10 -0800 | [diff] [blame] | 1403 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1404 | |
Patrick McHardy | fae718d | 2007-12-24 21:09:10 -0800 | [diff] [blame] | 1405 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1406 | &nf_conntrack_htable_size, 0600); |
| 1407 | |
Eric Dumazet | 5bfddbd | 2010-06-08 16:09:52 +0200 | [diff] [blame] | 1408 | void nf_ct_untracked_status_or(unsigned long bits) |
| 1409 | { |
Eric Dumazet | b3c5163 | 2010-06-09 14:43:38 +0200 | [diff] [blame] | 1410 | int cpu; |
| 1411 | |
| 1412 | for_each_possible_cpu(cpu) |
| 1413 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; |
Eric Dumazet | 5bfddbd | 2010-06-08 16:09:52 +0200 | [diff] [blame] | 1414 | } |
| 1415 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); |
| 1416 | |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1417 | static int nf_conntrack_init_init_net(void) |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1418 | { |
Patrick McHardy | f205c5e | 2007-07-07 22:28:14 -0700 | [diff] [blame] | 1419 | int max_factor = 8; |
Eric Dumazet | b3c5163 | 2010-06-09 14:43:38 +0200 | [diff] [blame] | 1420 | int ret, cpu; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1421 | |
| 1422 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB |
Patrick McHardy | f205c5e | 2007-07-07 22:28:14 -0700 | [diff] [blame] | 1423 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1424 | if (!nf_conntrack_htable_size) { |
| 1425 | nf_conntrack_htable_size |
Jan Beulich | 4481374 | 2009-09-21 17:03:05 -0700 | [diff] [blame] | 1426 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
Patrick McHardy | f205c5e | 2007-07-07 22:28:14 -0700 | [diff] [blame] | 1427 | / sizeof(struct hlist_head)); |
Jan Beulich | 4481374 | 2009-09-21 17:03:05 -0700 | [diff] [blame] | 1428 | if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
Patrick McHardy | f205c5e | 2007-07-07 22:28:14 -0700 | [diff] [blame] | 1429 | nf_conntrack_htable_size = 16384; |
| 1430 | if (nf_conntrack_htable_size < 32) |
| 1431 | nf_conntrack_htable_size = 32; |
| 1432 | |
| 1433 | /* Use a max. factor of four by default to get the same max as |
| 1434 | * with the old struct list_heads. When a table size is given |
| 1435 | * we use the old value of 8 to avoid reducing the max. |
| 1436 | * entries. */ |
| 1437 | max_factor = 4; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1438 | } |
Patrick McHardy | f205c5e | 2007-07-07 22:28:14 -0700 | [diff] [blame] | 1439 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
Patrick McHardy | 8e5105a | 2007-07-07 22:27:33 -0700 | [diff] [blame] | 1440 | |
Stephen Hemminger | 654d0fb | 2010-05-13 15:02:08 +0200 | [diff] [blame] | 1441 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
Patrick McHardy | 8e5105a | 2007-07-07 22:27:33 -0700 | [diff] [blame] | 1442 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
| 1443 | nf_conntrack_max); |
| 1444 | |
Patrick McHardy | ac5357e | 2007-03-14 16:38:25 -0700 | [diff] [blame] | 1445 | ret = nf_conntrack_proto_init(); |
Patrick McHardy | 933a41e | 2006-11-29 02:35:18 +0100 | [diff] [blame] | 1446 | if (ret < 0) |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1447 | goto err_proto; |
Patrick McHardy | 933a41e | 2006-11-29 02:35:18 +0100 | [diff] [blame] | 1448 | |
Yasuyuki Kozakai | ceceae1 | 2007-07-07 22:23:42 -0700 | [diff] [blame] | 1449 | ret = nf_conntrack_helper_init(); |
| 1450 | if (ret < 0) |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1451 | goto err_helper; |
Yasuyuki Kozakai | ceceae1 | 2007-07-07 22:23:42 -0700 | [diff] [blame] | 1452 | |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 1453 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
| 1454 | ret = nf_ct_extend_register(&nf_ct_zone_extend); |
| 1455 | if (ret < 0) |
| 1456 | goto err_extend; |
| 1457 | #endif |
Patrick McHardy | 9edd7ca | 2010-02-08 11:16:26 -0800 | [diff] [blame] | 1458 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
Eric Dumazet | b3c5163 | 2010-06-09 14:43:38 +0200 | [diff] [blame] | 1459 | for_each_possible_cpu(cpu) { |
| 1460 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); |
Eric Dumazet | b3c5163 | 2010-06-09 14:43:38 +0200 | [diff] [blame] | 1461 | write_pnet(&ct->ct_net, &init_net); |
| 1462 | atomic_set(&ct->ct_general.use, 1); |
| 1463 | } |
Patrick McHardy | 9edd7ca | 2010-02-08 11:16:26 -0800 | [diff] [blame] | 1464 | /* - and look it like as a confirmed connection */ |
Eric Dumazet | 5bfddbd | 2010-06-08 16:09:52 +0200 | [diff] [blame] | 1465 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1466 | return 0; |
| 1467 | |
Patrick McHardy | 5d0aa2c | 2010-02-15 18:13:33 +0100 | [diff] [blame] | 1468 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
| 1469 | err_extend: |
| 1470 | nf_conntrack_helper_fini(); |
| 1471 | #endif |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1472 | err_helper: |
| 1473 | nf_conntrack_proto_fini(); |
| 1474 | err_proto: |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1475 | return ret; |
| 1476 | } |
| 1477 | |
Eric Dumazet | 8cc2019 | 2009-06-22 14:13:55 +0200 | [diff] [blame] | 1478 | /* |
| 1479 | * We need to use special "null" values, not used in hash table |
| 1480 | */ |
| 1481 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) |
| 1482 | #define DYING_NULLS_VAL ((1<<30)+1) |
| 1483 | |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1484 | static int nf_conntrack_init_net(struct net *net) |
| 1485 | { |
| 1486 | int ret; |
| 1487 | |
| 1488 | atomic_set(&net->ct.count, 0); |
Eric Dumazet | 8cc2019 | 2009-06-22 14:13:55 +0200 | [diff] [blame] | 1489 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, UNCONFIRMED_NULLS_VAL); |
| 1490 | INIT_HLIST_NULLS_HEAD(&net->ct.dying, DYING_NULLS_VAL); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1491 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
| 1492 | if (!net->ct.stat) { |
| 1493 | ret = -ENOMEM; |
| 1494 | goto err_stat; |
| 1495 | } |
Eric Dumazet | 5b3501f | 2010-02-08 11:16:56 -0800 | [diff] [blame] | 1496 | |
| 1497 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); |
| 1498 | if (!net->ct.slabname) { |
| 1499 | ret = -ENOMEM; |
| 1500 | goto err_slabname; |
| 1501 | } |
| 1502 | |
| 1503 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, |
| 1504 | sizeof(struct nf_conn), 0, |
| 1505 | SLAB_DESTROY_BY_RCU, NULL); |
| 1506 | if (!net->ct.nf_conntrack_cachep) { |
| 1507 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); |
| 1508 | ret = -ENOMEM; |
| 1509 | goto err_cache; |
| 1510 | } |
Patrick McHardy | d696c7b | 2010-02-08 11:18:07 -0800 | [diff] [blame] | 1511 | |
| 1512 | net->ct.htable_size = nf_conntrack_htable_size; |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1513 | net->ct.hash = nf_ct_alloc_hashtable(&net->ct.htable_size, 1); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1514 | if (!net->ct.hash) { |
| 1515 | ret = -ENOMEM; |
| 1516 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); |
| 1517 | goto err_hash; |
| 1518 | } |
| 1519 | ret = nf_conntrack_expect_init(net); |
| 1520 | if (ret < 0) |
| 1521 | goto err_expect; |
Alexey Dobriyan | d716a4d | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1522 | ret = nf_conntrack_acct_init(net); |
Krzysztof Piotr Oledzki | 5840157 | 2008-07-21 10:01:34 -0700 | [diff] [blame] | 1523 | if (ret < 0) |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1524 | goto err_acct; |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 1525 | ret = nf_conntrack_tstamp_init(net); |
| 1526 | if (ret < 0) |
| 1527 | goto err_tstamp; |
Pablo Neira Ayuso | a0891aa | 2009-06-13 12:26:29 +0200 | [diff] [blame] | 1528 | ret = nf_conntrack_ecache_init(net); |
| 1529 | if (ret < 0) |
| 1530 | goto err_ecache; |
Yasuyuki Kozakai | 7d3cdc6 | 2006-02-15 15:22:21 -0800 | [diff] [blame] | 1531 | |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1532 | return 0; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1533 | |
Pablo Neira Ayuso | a0891aa | 2009-06-13 12:26:29 +0200 | [diff] [blame] | 1534 | err_ecache: |
Pablo Neira Ayuso | a992ca2 | 2011-01-19 16:00:07 +0100 | [diff] [blame] | 1535 | nf_conntrack_tstamp_fini(net); |
| 1536 | err_tstamp: |
Pablo Neira Ayuso | a0891aa | 2009-06-13 12:26:29 +0200 | [diff] [blame] | 1537 | nf_conntrack_acct_fini(net); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1538 | err_acct: |
Alexey Dobriyan | 9b03f38 | 2008-10-08 11:35:03 +0200 | [diff] [blame] | 1539 | nf_conntrack_expect_fini(net); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1540 | err_expect: |
Patrick McHardy | d862a66 | 2011-01-14 15:45:56 +0100 | [diff] [blame] | 1541 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
Alexey Dobriyan | 6058fa6 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 1542 | err_hash: |
Eric Dumazet | 5b3501f | 2010-02-08 11:16:56 -0800 | [diff] [blame] | 1543 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
| 1544 | err_cache: |
| 1545 | kfree(net->ct.slabname); |
| 1546 | err_slabname: |
Alexey Dobriyan | 0d55af8 | 2008-10-08 11:35:07 +0200 | [diff] [blame] | 1547 | free_percpu(net->ct.stat); |
| 1548 | err_stat: |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1549 | return ret; |
| 1550 | } |
| 1551 | |
Jozsef Kadlecsik | f9dd09c | 2009-11-06 00:43:42 -0800 | [diff] [blame] | 1552 | s16 (*nf_ct_nat_offset)(const struct nf_conn *ct, |
| 1553 | enum ip_conntrack_dir dir, |
| 1554 | u32 seq); |
| 1555 | EXPORT_SYMBOL_GPL(nf_ct_nat_offset); |
| 1556 | |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1557 | int nf_conntrack_init(struct net *net) |
| 1558 | { |
| 1559 | int ret; |
| 1560 | |
| 1561 | if (net_eq(net, &init_net)) { |
| 1562 | ret = nf_conntrack_init_init_net(); |
| 1563 | if (ret < 0) |
| 1564 | goto out_init_net; |
| 1565 | } |
| 1566 | ret = nf_conntrack_init_net(net); |
| 1567 | if (ret < 0) |
| 1568 | goto out_net; |
| 1569 | |
| 1570 | if (net_eq(net, &init_net)) { |
| 1571 | /* For use by REJECT target */ |
Stephen Hemminger | a9b3cd7 | 2011-08-01 16:19:00 +0000 | [diff] [blame] | 1572 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); |
| 1573 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); |
Jozsef Kadlecsik | f9dd09c | 2009-11-06 00:43:42 -0800 | [diff] [blame] | 1574 | |
| 1575 | /* Howto get NAT offsets */ |
Stephen Hemminger | a9b3cd7 | 2011-08-01 16:19:00 +0000 | [diff] [blame] | 1576 | RCU_INIT_POINTER(nf_ct_nat_offset, NULL); |
Alexey Dobriyan | 08f6547 | 2008-10-08 11:35:09 +0200 | [diff] [blame] | 1577 | } |
| 1578 | return 0; |
| 1579 | |
| 1580 | out_net: |
| 1581 | if (net_eq(net, &init_net)) |
| 1582 | nf_conntrack_cleanup_init_net(); |
| 1583 | out_init_net: |
| 1584 | return ret; |
Yasuyuki Kozakai | 9fb9cbb | 2005-11-09 16:38:16 -0800 | [diff] [blame] | 1585 | } |