Commit 1e47ee8367babe6a5e8adf44a714c7086657b87e
1 parent
4610476d89
netfilter: nf_conntrack: fix BUG_ON while removing nf_conntrack with netns
canqun zhang reported that we're hitting BUG_ON in the nf_conntrack_destroy path when calling kfree_skb while rmmod'ing the nf_conntrack module. Currently, the nf_ct_destroy hook is being set to NULL in the destroy path of conntrack.init_net. However, this is a problem since init_net may be destroyed before any other existing netns (we cannot assume any specific ordering while releasing existing netns according to what I read in recent emails). Thanks to Gao feng for initial patch to address this issue. Reported-by: canqun zhang <canqunzhang@gmail.com> Acked-by: Gao feng <gaofeng@cn.fujitsu.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Showing 3 changed files with 8 additions and 4 deletions Inline Diff
include/net/netfilter/nf_conntrack_core.h
1 | /* | 1 | /* |
2 | * This header is used to share core functionality between the | 2 | * This header is used to share core functionality between the |
3 | * standalone connection tracking module, and the compatibility layer's use | 3 | * standalone connection tracking module, and the compatibility layer's use |
4 | * of connection tracking. | 4 | * of connection tracking. |
5 | * | 5 | * |
6 | * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp> | 6 | * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp> |
7 | * - generalize L3 protocol dependent part. | 7 | * - generalize L3 protocol dependent part. |
8 | * | 8 | * |
9 | * Derived from include/linux/netfiter_ipv4/ip_conntrack_core.h | 9 | * Derived from include/linux/netfiter_ipv4/ip_conntrack_core.h |
10 | */ | 10 | */ |
11 | 11 | ||
12 | #ifndef _NF_CONNTRACK_CORE_H | 12 | #ifndef _NF_CONNTRACK_CORE_H |
13 | #define _NF_CONNTRACK_CORE_H | 13 | #define _NF_CONNTRACK_CORE_H |
14 | 14 | ||
15 | #include <linux/netfilter.h> | 15 | #include <linux/netfilter.h> |
16 | #include <net/netfilter/nf_conntrack_l3proto.h> | 16 | #include <net/netfilter/nf_conntrack_l3proto.h> |
17 | #include <net/netfilter/nf_conntrack_l4proto.h> | 17 | #include <net/netfilter/nf_conntrack_l4proto.h> |
18 | #include <net/netfilter/nf_conntrack_ecache.h> | 18 | #include <net/netfilter/nf_conntrack_ecache.h> |
19 | 19 | ||
20 | /* This header is used to share core functionality between the | 20 | /* This header is used to share core functionality between the |
21 | standalone connection tracking module, and the compatibility layer's use | 21 | standalone connection tracking module, and the compatibility layer's use |
22 | of connection tracking. */ | 22 | of connection tracking. */ |
23 | extern unsigned int nf_conntrack_in(struct net *net, | 23 | extern unsigned int nf_conntrack_in(struct net *net, |
24 | u_int8_t pf, | 24 | u_int8_t pf, |
25 | unsigned int hooknum, | 25 | unsigned int hooknum, |
26 | struct sk_buff *skb); | 26 | struct sk_buff *skb); |
27 | 27 | ||
28 | extern int nf_conntrack_init(struct net *net); | 28 | extern int nf_conntrack_init(struct net *net); |
29 | extern void nf_conntrack_cleanup(struct net *net); | 29 | extern void nf_conntrack_cleanup(struct net *net); |
30 | 30 | ||
31 | extern int nf_conntrack_proto_init(struct net *net); | 31 | extern int nf_conntrack_proto_init(struct net *net); |
32 | extern void nf_conntrack_proto_fini(struct net *net); | 32 | extern void nf_conntrack_proto_fini(struct net *net); |
33 | 33 | ||
34 | extern void nf_conntrack_cleanup_end(void); | ||
35 | |||
34 | extern bool | 36 | extern bool |
35 | nf_ct_get_tuple(const struct sk_buff *skb, | 37 | nf_ct_get_tuple(const struct sk_buff *skb, |
36 | unsigned int nhoff, | 38 | unsigned int nhoff, |
37 | unsigned int dataoff, | 39 | unsigned int dataoff, |
38 | u_int16_t l3num, | 40 | u_int16_t l3num, |
39 | u_int8_t protonum, | 41 | u_int8_t protonum, |
40 | struct nf_conntrack_tuple *tuple, | 42 | struct nf_conntrack_tuple *tuple, |
41 | const struct nf_conntrack_l3proto *l3proto, | 43 | const struct nf_conntrack_l3proto *l3proto, |
42 | const struct nf_conntrack_l4proto *l4proto); | 44 | const struct nf_conntrack_l4proto *l4proto); |
43 | 45 | ||
44 | extern bool | 46 | extern bool |
45 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, | 47 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
46 | const struct nf_conntrack_tuple *orig, | 48 | const struct nf_conntrack_tuple *orig, |
47 | const struct nf_conntrack_l3proto *l3proto, | 49 | const struct nf_conntrack_l3proto *l3proto, |
48 | const struct nf_conntrack_l4proto *l4proto); | 50 | const struct nf_conntrack_l4proto *l4proto); |
49 | 51 | ||
50 | /* Find a connection corresponding to a tuple. */ | 52 | /* Find a connection corresponding to a tuple. */ |
51 | extern struct nf_conntrack_tuple_hash * | 53 | extern struct nf_conntrack_tuple_hash * |
52 | nf_conntrack_find_get(struct net *net, u16 zone, | 54 | nf_conntrack_find_get(struct net *net, u16 zone, |
53 | const struct nf_conntrack_tuple *tuple); | 55 | const struct nf_conntrack_tuple *tuple); |
54 | 56 | ||
55 | extern int __nf_conntrack_confirm(struct sk_buff *skb); | 57 | extern int __nf_conntrack_confirm(struct sk_buff *skb); |
56 | 58 | ||
57 | /* Confirm a connection: returns NF_DROP if packet must be dropped. */ | 59 | /* Confirm a connection: returns NF_DROP if packet must be dropped. */ |
58 | static inline int nf_conntrack_confirm(struct sk_buff *skb) | 60 | static inline int nf_conntrack_confirm(struct sk_buff *skb) |
59 | { | 61 | { |
60 | struct nf_conn *ct = (struct nf_conn *)skb->nfct; | 62 | struct nf_conn *ct = (struct nf_conn *)skb->nfct; |
61 | int ret = NF_ACCEPT; | 63 | int ret = NF_ACCEPT; |
62 | 64 | ||
63 | if (ct && !nf_ct_is_untracked(ct)) { | 65 | if (ct && !nf_ct_is_untracked(ct)) { |
64 | if (!nf_ct_is_confirmed(ct)) | 66 | if (!nf_ct_is_confirmed(ct)) |
65 | ret = __nf_conntrack_confirm(skb); | 67 | ret = __nf_conntrack_confirm(skb); |
66 | if (likely(ret == NF_ACCEPT)) | 68 | if (likely(ret == NF_ACCEPT)) |
67 | nf_ct_deliver_cached_events(ct); | 69 | nf_ct_deliver_cached_events(ct); |
68 | } | 70 | } |
69 | return ret; | 71 | return ret; |
70 | } | 72 | } |
71 | 73 | ||
72 | int | 74 | int |
73 | print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple, | 75 | print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple, |
74 | const struct nf_conntrack_l3proto *l3proto, | 76 | const struct nf_conntrack_l3proto *l3proto, |
75 | const struct nf_conntrack_l4proto *proto); | 77 | const struct nf_conntrack_l4proto *proto); |
76 | 78 | ||
77 | extern spinlock_t nf_conntrack_lock ; | 79 | extern spinlock_t nf_conntrack_lock ; |
78 | 80 | ||
79 | #endif /* _NF_CONNTRACK_CORE_H */ | 81 | #endif /* _NF_CONNTRACK_CORE_H */ |
80 | 82 |
net/netfilter/nf_conntrack_core.c
1 | /* Connection state tracking for netfilter. This is separated from, | 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 | 2 | but required by, the NAT layer; it can also be used by an iptables |
3 | extension. */ | 3 | extension. */ |
4 | 4 | ||
5 | /* (C) 1999-2001 Paul `Rusty' Russell | 5 | /* (C) 1999-2001 Paul `Rusty' Russell |
6 | * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org> | 6 | * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org> |
7 | * (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org> | 7 | * (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org> |
8 | * | 8 | * |
9 | * This program is free software; you can redistribute it and/or modify | 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 | 10 | * it under the terms of the GNU General Public License version 2 as |
11 | * published by the Free Software Foundation. | 11 | * published by the Free Software Foundation. |
12 | */ | 12 | */ |
13 | 13 | ||
14 | #include <linux/types.h> | 14 | #include <linux/types.h> |
15 | #include <linux/netfilter.h> | 15 | #include <linux/netfilter.h> |
16 | #include <linux/module.h> | 16 | #include <linux/module.h> |
17 | #include <linux/sched.h> | 17 | #include <linux/sched.h> |
18 | #include <linux/skbuff.h> | 18 | #include <linux/skbuff.h> |
19 | #include <linux/proc_fs.h> | 19 | #include <linux/proc_fs.h> |
20 | #include <linux/vmalloc.h> | 20 | #include <linux/vmalloc.h> |
21 | #include <linux/stddef.h> | 21 | #include <linux/stddef.h> |
22 | #include <linux/slab.h> | 22 | #include <linux/slab.h> |
23 | #include <linux/random.h> | 23 | #include <linux/random.h> |
24 | #include <linux/jhash.h> | 24 | #include <linux/jhash.h> |
25 | #include <linux/err.h> | 25 | #include <linux/err.h> |
26 | #include <linux/percpu.h> | 26 | #include <linux/percpu.h> |
27 | #include <linux/moduleparam.h> | 27 | #include <linux/moduleparam.h> |
28 | #include <linux/notifier.h> | 28 | #include <linux/notifier.h> |
29 | #include <linux/kernel.h> | 29 | #include <linux/kernel.h> |
30 | #include <linux/netdevice.h> | 30 | #include <linux/netdevice.h> |
31 | #include <linux/socket.h> | 31 | #include <linux/socket.h> |
32 | #include <linux/mm.h> | 32 | #include <linux/mm.h> |
33 | #include <linux/nsproxy.h> | 33 | #include <linux/nsproxy.h> |
34 | #include <linux/rculist_nulls.h> | 34 | #include <linux/rculist_nulls.h> |
35 | 35 | ||
36 | #include <net/netfilter/nf_conntrack.h> | 36 | #include <net/netfilter/nf_conntrack.h> |
37 | #include <net/netfilter/nf_conntrack_l3proto.h> | 37 | #include <net/netfilter/nf_conntrack_l3proto.h> |
38 | #include <net/netfilter/nf_conntrack_l4proto.h> | 38 | #include <net/netfilter/nf_conntrack_l4proto.h> |
39 | #include <net/netfilter/nf_conntrack_expect.h> | 39 | #include <net/netfilter/nf_conntrack_expect.h> |
40 | #include <net/netfilter/nf_conntrack_helper.h> | 40 | #include <net/netfilter/nf_conntrack_helper.h> |
41 | #include <net/netfilter/nf_conntrack_core.h> | 41 | #include <net/netfilter/nf_conntrack_core.h> |
42 | #include <net/netfilter/nf_conntrack_extend.h> | 42 | #include <net/netfilter/nf_conntrack_extend.h> |
43 | #include <net/netfilter/nf_conntrack_acct.h> | 43 | #include <net/netfilter/nf_conntrack_acct.h> |
44 | #include <net/netfilter/nf_conntrack_ecache.h> | 44 | #include <net/netfilter/nf_conntrack_ecache.h> |
45 | #include <net/netfilter/nf_conntrack_zones.h> | 45 | #include <net/netfilter/nf_conntrack_zones.h> |
46 | #include <net/netfilter/nf_conntrack_timestamp.h> | 46 | #include <net/netfilter/nf_conntrack_timestamp.h> |
47 | #include <net/netfilter/nf_conntrack_timeout.h> | 47 | #include <net/netfilter/nf_conntrack_timeout.h> |
48 | #include <net/netfilter/nf_nat.h> | 48 | #include <net/netfilter/nf_nat.h> |
49 | #include <net/netfilter/nf_nat_core.h> | 49 | #include <net/netfilter/nf_nat_core.h> |
50 | 50 | ||
51 | #define NF_CONNTRACK_VERSION "0.5.0" | 51 | #define NF_CONNTRACK_VERSION "0.5.0" |
52 | 52 | ||
53 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, | 53 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
54 | enum nf_nat_manip_type manip, | 54 | enum nf_nat_manip_type manip, |
55 | const struct nlattr *attr) __read_mostly; | 55 | const struct nlattr *attr) __read_mostly; |
56 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); | 56 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
57 | 57 | ||
58 | int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb, | 58 | int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb, |
59 | struct nf_conn *ct, | 59 | struct nf_conn *ct, |
60 | enum ip_conntrack_info ctinfo, | 60 | enum ip_conntrack_info ctinfo, |
61 | unsigned int protoff); | 61 | unsigned int protoff); |
62 | EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook); | 62 | EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook); |
63 | 63 | ||
64 | DEFINE_SPINLOCK(nf_conntrack_lock); | 64 | DEFINE_SPINLOCK(nf_conntrack_lock); |
65 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); | 65 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
66 | 66 | ||
67 | unsigned int nf_conntrack_htable_size __read_mostly; | 67 | unsigned int nf_conntrack_htable_size __read_mostly; |
68 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); | 68 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
69 | 69 | ||
70 | unsigned int nf_conntrack_max __read_mostly; | 70 | unsigned int nf_conntrack_max __read_mostly; |
71 | EXPORT_SYMBOL_GPL(nf_conntrack_max); | 71 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
72 | 72 | ||
73 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); | 73 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
74 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); | 74 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); |
75 | 75 | ||
76 | unsigned int nf_conntrack_hash_rnd __read_mostly; | 76 | unsigned int nf_conntrack_hash_rnd __read_mostly; |
77 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_rnd); | 77 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_rnd); |
78 | 78 | ||
79 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, u16 zone) | 79 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, u16 zone) |
80 | { | 80 | { |
81 | unsigned int n; | 81 | unsigned int n; |
82 | 82 | ||
83 | /* The direction must be ignored, so we hash everything up to the | 83 | /* The direction must be ignored, so we hash everything up to the |
84 | * destination ports (which is a multiple of 4) and treat the last | 84 | * destination ports (which is a multiple of 4) and treat the last |
85 | * three bytes manually. | 85 | * three bytes manually. |
86 | */ | 86 | */ |
87 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | 87 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); |
88 | return jhash2((u32 *)tuple, n, zone ^ nf_conntrack_hash_rnd ^ | 88 | return jhash2((u32 *)tuple, n, zone ^ nf_conntrack_hash_rnd ^ |
89 | (((__force __u16)tuple->dst.u.all << 16) | | 89 | (((__force __u16)tuple->dst.u.all << 16) | |
90 | tuple->dst.protonum)); | 90 | tuple->dst.protonum)); |
91 | } | 91 | } |
92 | 92 | ||
93 | static u32 __hash_bucket(u32 hash, unsigned int size) | 93 | static u32 __hash_bucket(u32 hash, unsigned int size) |
94 | { | 94 | { |
95 | return ((u64)hash * size) >> 32; | 95 | return ((u64)hash * size) >> 32; |
96 | } | 96 | } |
97 | 97 | ||
98 | static u32 hash_bucket(u32 hash, const struct net *net) | 98 | static u32 hash_bucket(u32 hash, const struct net *net) |
99 | { | 99 | { |
100 | return __hash_bucket(hash, net->ct.htable_size); | 100 | return __hash_bucket(hash, net->ct.htable_size); |
101 | } | 101 | } |
102 | 102 | ||
103 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, | 103 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, |
104 | u16 zone, unsigned int size) | 104 | u16 zone, unsigned int size) |
105 | { | 105 | { |
106 | return __hash_bucket(hash_conntrack_raw(tuple, zone), size); | 106 | return __hash_bucket(hash_conntrack_raw(tuple, zone), size); |
107 | } | 107 | } |
108 | 108 | ||
109 | static inline u_int32_t hash_conntrack(const struct net *net, u16 zone, | 109 | static inline u_int32_t hash_conntrack(const struct net *net, u16 zone, |
110 | const struct nf_conntrack_tuple *tuple) | 110 | const struct nf_conntrack_tuple *tuple) |
111 | { | 111 | { |
112 | return __hash_conntrack(tuple, zone, net->ct.htable_size); | 112 | return __hash_conntrack(tuple, zone, net->ct.htable_size); |
113 | } | 113 | } |
114 | 114 | ||
115 | bool | 115 | bool |
116 | nf_ct_get_tuple(const struct sk_buff *skb, | 116 | nf_ct_get_tuple(const struct sk_buff *skb, |
117 | unsigned int nhoff, | 117 | unsigned int nhoff, |
118 | unsigned int dataoff, | 118 | unsigned int dataoff, |
119 | u_int16_t l3num, | 119 | u_int16_t l3num, |
120 | u_int8_t protonum, | 120 | u_int8_t protonum, |
121 | struct nf_conntrack_tuple *tuple, | 121 | struct nf_conntrack_tuple *tuple, |
122 | const struct nf_conntrack_l3proto *l3proto, | 122 | const struct nf_conntrack_l3proto *l3proto, |
123 | const struct nf_conntrack_l4proto *l4proto) | 123 | const struct nf_conntrack_l4proto *l4proto) |
124 | { | 124 | { |
125 | memset(tuple, 0, sizeof(*tuple)); | 125 | memset(tuple, 0, sizeof(*tuple)); |
126 | 126 | ||
127 | tuple->src.l3num = l3num; | 127 | tuple->src.l3num = l3num; |
128 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | 128 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) |
129 | return false; | 129 | return false; |
130 | 130 | ||
131 | tuple->dst.protonum = protonum; | 131 | tuple->dst.protonum = protonum; |
132 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | 132 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; |
133 | 133 | ||
134 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); | 134 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); |
135 | } | 135 | } |
136 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); | 136 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
137 | 137 | ||
138 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, | 138 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
139 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) | 139 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) |
140 | { | 140 | { |
141 | struct nf_conntrack_l3proto *l3proto; | 141 | struct nf_conntrack_l3proto *l3proto; |
142 | struct nf_conntrack_l4proto *l4proto; | 142 | struct nf_conntrack_l4proto *l4proto; |
143 | unsigned int protoff; | 143 | unsigned int protoff; |
144 | u_int8_t protonum; | 144 | u_int8_t protonum; |
145 | int ret; | 145 | int ret; |
146 | 146 | ||
147 | rcu_read_lock(); | 147 | rcu_read_lock(); |
148 | 148 | ||
149 | l3proto = __nf_ct_l3proto_find(l3num); | 149 | l3proto = __nf_ct_l3proto_find(l3num); |
150 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | 150 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); |
151 | if (ret != NF_ACCEPT) { | 151 | if (ret != NF_ACCEPT) { |
152 | rcu_read_unlock(); | 152 | rcu_read_unlock(); |
153 | return false; | 153 | return false; |
154 | } | 154 | } |
155 | 155 | ||
156 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | 156 | l4proto = __nf_ct_l4proto_find(l3num, protonum); |
157 | 157 | ||
158 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, | 158 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, |
159 | l3proto, l4proto); | 159 | l3proto, l4proto); |
160 | 160 | ||
161 | rcu_read_unlock(); | 161 | rcu_read_unlock(); |
162 | return ret; | 162 | return ret; |
163 | } | 163 | } |
164 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | 164 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); |
165 | 165 | ||
166 | bool | 166 | bool |
167 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, | 167 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
168 | const struct nf_conntrack_tuple *orig, | 168 | const struct nf_conntrack_tuple *orig, |
169 | const struct nf_conntrack_l3proto *l3proto, | 169 | const struct nf_conntrack_l3proto *l3proto, |
170 | const struct nf_conntrack_l4proto *l4proto) | 170 | const struct nf_conntrack_l4proto *l4proto) |
171 | { | 171 | { |
172 | memset(inverse, 0, sizeof(*inverse)); | 172 | memset(inverse, 0, sizeof(*inverse)); |
173 | 173 | ||
174 | inverse->src.l3num = orig->src.l3num; | 174 | inverse->src.l3num = orig->src.l3num; |
175 | if (l3proto->invert_tuple(inverse, orig) == 0) | 175 | if (l3proto->invert_tuple(inverse, orig) == 0) |
176 | return false; | 176 | return false; |
177 | 177 | ||
178 | inverse->dst.dir = !orig->dst.dir; | 178 | inverse->dst.dir = !orig->dst.dir; |
179 | 179 | ||
180 | inverse->dst.protonum = orig->dst.protonum; | 180 | inverse->dst.protonum = orig->dst.protonum; |
181 | return l4proto->invert_tuple(inverse, orig); | 181 | return l4proto->invert_tuple(inverse, orig); |
182 | } | 182 | } |
183 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); | 183 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
184 | 184 | ||
185 | static void | 185 | static void |
186 | clean_from_lists(struct nf_conn *ct) | 186 | clean_from_lists(struct nf_conn *ct) |
187 | { | 187 | { |
188 | pr_debug("clean_from_lists(%p)\n", ct); | 188 | pr_debug("clean_from_lists(%p)\n", ct); |
189 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | 189 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
190 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | 190 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); |
191 | 191 | ||
192 | /* Destroy all pending expectations */ | 192 | /* Destroy all pending expectations */ |
193 | nf_ct_remove_expectations(ct); | 193 | nf_ct_remove_expectations(ct); |
194 | } | 194 | } |
195 | 195 | ||
196 | static void | 196 | static void |
197 | destroy_conntrack(struct nf_conntrack *nfct) | 197 | destroy_conntrack(struct nf_conntrack *nfct) |
198 | { | 198 | { |
199 | struct nf_conn *ct = (struct nf_conn *)nfct; | 199 | struct nf_conn *ct = (struct nf_conn *)nfct; |
200 | struct net *net = nf_ct_net(ct); | 200 | struct net *net = nf_ct_net(ct); |
201 | struct nf_conntrack_l4proto *l4proto; | 201 | struct nf_conntrack_l4proto *l4proto; |
202 | 202 | ||
203 | pr_debug("destroy_conntrack(%p)\n", ct); | 203 | pr_debug("destroy_conntrack(%p)\n", ct); |
204 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); | 204 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
205 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | 205 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); |
206 | 206 | ||
207 | /* To make sure we don't get any weird locking issues here: | 207 | /* To make sure we don't get any weird locking issues here: |
208 | * destroy_conntrack() MUST NOT be called with a write lock | 208 | * destroy_conntrack() MUST NOT be called with a write lock |
209 | * to nf_conntrack_lock!!! -HW */ | 209 | * to nf_conntrack_lock!!! -HW */ |
210 | rcu_read_lock(); | 210 | rcu_read_lock(); |
211 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); | 211 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
212 | if (l4proto && l4proto->destroy) | 212 | if (l4proto && l4proto->destroy) |
213 | l4proto->destroy(ct); | 213 | l4proto->destroy(ct); |
214 | 214 | ||
215 | rcu_read_unlock(); | 215 | rcu_read_unlock(); |
216 | 216 | ||
217 | spin_lock_bh(&nf_conntrack_lock); | 217 | spin_lock_bh(&nf_conntrack_lock); |
218 | /* Expectations will have been removed in clean_from_lists, | 218 | /* Expectations will have been removed in clean_from_lists, |
219 | * except TFTP can create an expectation on the first packet, | 219 | * except TFTP can create an expectation on the first packet, |
220 | * before connection is in the list, so we need to clean here, | 220 | * before connection is in the list, so we need to clean here, |
221 | * too. */ | 221 | * too. */ |
222 | nf_ct_remove_expectations(ct); | 222 | nf_ct_remove_expectations(ct); |
223 | 223 | ||
224 | /* We overload first tuple to link into unconfirmed or dying list.*/ | 224 | /* We overload first tuple to link into unconfirmed or dying list.*/ |
225 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); | 225 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); |
226 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | 226 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
227 | 227 | ||
228 | NF_CT_STAT_INC(net, delete); | 228 | NF_CT_STAT_INC(net, delete); |
229 | spin_unlock_bh(&nf_conntrack_lock); | 229 | spin_unlock_bh(&nf_conntrack_lock); |
230 | 230 | ||
231 | if (ct->master) | 231 | if (ct->master) |
232 | nf_ct_put(ct->master); | 232 | nf_ct_put(ct->master); |
233 | 233 | ||
234 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); | 234 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
235 | nf_conntrack_free(ct); | 235 | nf_conntrack_free(ct); |
236 | } | 236 | } |
237 | 237 | ||
238 | void nf_ct_delete_from_lists(struct nf_conn *ct) | 238 | void nf_ct_delete_from_lists(struct nf_conn *ct) |
239 | { | 239 | { |
240 | struct net *net = nf_ct_net(ct); | 240 | struct net *net = nf_ct_net(ct); |
241 | 241 | ||
242 | nf_ct_helper_destroy(ct); | 242 | nf_ct_helper_destroy(ct); |
243 | spin_lock_bh(&nf_conntrack_lock); | 243 | spin_lock_bh(&nf_conntrack_lock); |
244 | /* Inside lock so preempt is disabled on module removal path. | 244 | /* Inside lock so preempt is disabled on module removal path. |
245 | * Otherwise we can get spurious warnings. */ | 245 | * Otherwise we can get spurious warnings. */ |
246 | NF_CT_STAT_INC(net, delete_list); | 246 | NF_CT_STAT_INC(net, delete_list); |
247 | clean_from_lists(ct); | 247 | clean_from_lists(ct); |
248 | /* add this conntrack to the dying list */ | 248 | /* add this conntrack to the dying list */ |
249 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | 249 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
250 | &net->ct.dying); | 250 | &net->ct.dying); |
251 | spin_unlock_bh(&nf_conntrack_lock); | 251 | spin_unlock_bh(&nf_conntrack_lock); |
252 | } | 252 | } |
253 | EXPORT_SYMBOL_GPL(nf_ct_delete_from_lists); | 253 | EXPORT_SYMBOL_GPL(nf_ct_delete_from_lists); |
254 | 254 | ||
255 | static void death_by_event(unsigned long ul_conntrack) | 255 | static void death_by_event(unsigned long ul_conntrack) |
256 | { | 256 | { |
257 | struct nf_conn *ct = (void *)ul_conntrack; | 257 | struct nf_conn *ct = (void *)ul_conntrack; |
258 | struct net *net = nf_ct_net(ct); | 258 | struct net *net = nf_ct_net(ct); |
259 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); | 259 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
260 | 260 | ||
261 | BUG_ON(ecache == NULL); | 261 | BUG_ON(ecache == NULL); |
262 | 262 | ||
263 | if (nf_conntrack_event(IPCT_DESTROY, ct) < 0) { | 263 | if (nf_conntrack_event(IPCT_DESTROY, ct) < 0) { |
264 | /* bad luck, let's retry again */ | 264 | /* bad luck, let's retry again */ |
265 | ecache->timeout.expires = jiffies + | 265 | ecache->timeout.expires = jiffies + |
266 | (random32() % net->ct.sysctl_events_retry_timeout); | 266 | (random32() % net->ct.sysctl_events_retry_timeout); |
267 | add_timer(&ecache->timeout); | 267 | add_timer(&ecache->timeout); |
268 | return; | 268 | return; |
269 | } | 269 | } |
270 | /* we've got the event delivered, now it's dying */ | 270 | /* we've got the event delivered, now it's dying */ |
271 | set_bit(IPS_DYING_BIT, &ct->status); | 271 | set_bit(IPS_DYING_BIT, &ct->status); |
272 | nf_ct_put(ct); | 272 | nf_ct_put(ct); |
273 | } | 273 | } |
274 | 274 | ||
275 | void nf_ct_dying_timeout(struct nf_conn *ct) | 275 | void nf_ct_dying_timeout(struct nf_conn *ct) |
276 | { | 276 | { |
277 | struct net *net = nf_ct_net(ct); | 277 | struct net *net = nf_ct_net(ct); |
278 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); | 278 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
279 | 279 | ||
280 | BUG_ON(ecache == NULL); | 280 | BUG_ON(ecache == NULL); |
281 | 281 | ||
282 | /* set a new timer to retry event delivery */ | 282 | /* set a new timer to retry event delivery */ |
283 | setup_timer(&ecache->timeout, death_by_event, (unsigned long)ct); | 283 | setup_timer(&ecache->timeout, death_by_event, (unsigned long)ct); |
284 | ecache->timeout.expires = jiffies + | 284 | ecache->timeout.expires = jiffies + |
285 | (random32() % net->ct.sysctl_events_retry_timeout); | 285 | (random32() % net->ct.sysctl_events_retry_timeout); |
286 | add_timer(&ecache->timeout); | 286 | add_timer(&ecache->timeout); |
287 | } | 287 | } |
288 | EXPORT_SYMBOL_GPL(nf_ct_dying_timeout); | 288 | EXPORT_SYMBOL_GPL(nf_ct_dying_timeout); |
289 | 289 | ||
290 | static void death_by_timeout(unsigned long ul_conntrack) | 290 | static void death_by_timeout(unsigned long ul_conntrack) |
291 | { | 291 | { |
292 | struct nf_conn *ct = (void *)ul_conntrack; | 292 | struct nf_conn *ct = (void *)ul_conntrack; |
293 | struct nf_conn_tstamp *tstamp; | 293 | struct nf_conn_tstamp *tstamp; |
294 | 294 | ||
295 | tstamp = nf_conn_tstamp_find(ct); | 295 | tstamp = nf_conn_tstamp_find(ct); |
296 | if (tstamp && tstamp->stop == 0) | 296 | if (tstamp && tstamp->stop == 0) |
297 | tstamp->stop = ktime_to_ns(ktime_get_real()); | 297 | tstamp->stop = ktime_to_ns(ktime_get_real()); |
298 | 298 | ||
299 | if (!test_bit(IPS_DYING_BIT, &ct->status) && | 299 | if (!test_bit(IPS_DYING_BIT, &ct->status) && |
300 | unlikely(nf_conntrack_event(IPCT_DESTROY, ct) < 0)) { | 300 | unlikely(nf_conntrack_event(IPCT_DESTROY, ct) < 0)) { |
301 | /* destroy event was not delivered */ | 301 | /* destroy event was not delivered */ |
302 | nf_ct_delete_from_lists(ct); | 302 | nf_ct_delete_from_lists(ct); |
303 | nf_ct_dying_timeout(ct); | 303 | nf_ct_dying_timeout(ct); |
304 | return; | 304 | return; |
305 | } | 305 | } |
306 | set_bit(IPS_DYING_BIT, &ct->status); | 306 | set_bit(IPS_DYING_BIT, &ct->status); |
307 | nf_ct_delete_from_lists(ct); | 307 | nf_ct_delete_from_lists(ct); |
308 | nf_ct_put(ct); | 308 | nf_ct_put(ct); |
309 | } | 309 | } |
310 | 310 | ||
311 | /* | 311 | /* |
312 | * Warning : | 312 | * Warning : |
313 | * - Caller must take a reference on returned object | 313 | * - Caller must take a reference on returned object |
314 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | 314 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) |
315 | * OR | 315 | * OR |
316 | * - Caller must lock nf_conntrack_lock before calling this function | 316 | * - Caller must lock nf_conntrack_lock before calling this function |
317 | */ | 317 | */ |
318 | static struct nf_conntrack_tuple_hash * | 318 | static struct nf_conntrack_tuple_hash * |
319 | ____nf_conntrack_find(struct net *net, u16 zone, | 319 | ____nf_conntrack_find(struct net *net, u16 zone, |
320 | const struct nf_conntrack_tuple *tuple, u32 hash) | 320 | const struct nf_conntrack_tuple *tuple, u32 hash) |
321 | { | 321 | { |
322 | struct nf_conntrack_tuple_hash *h; | 322 | struct nf_conntrack_tuple_hash *h; |
323 | struct hlist_nulls_node *n; | 323 | struct hlist_nulls_node *n; |
324 | unsigned int bucket = hash_bucket(hash, net); | 324 | unsigned int bucket = hash_bucket(hash, net); |
325 | 325 | ||
326 | /* Disable BHs the entire time since we normally need to disable them | 326 | /* Disable BHs the entire time since we normally need to disable them |
327 | * at least once for the stats anyway. | 327 | * at least once for the stats anyway. |
328 | */ | 328 | */ |
329 | local_bh_disable(); | 329 | local_bh_disable(); |
330 | begin: | 330 | begin: |
331 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[bucket], hnnode) { | 331 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[bucket], hnnode) { |
332 | if (nf_ct_tuple_equal(tuple, &h->tuple) && | 332 | if (nf_ct_tuple_equal(tuple, &h->tuple) && |
333 | nf_ct_zone(nf_ct_tuplehash_to_ctrack(h)) == zone) { | 333 | nf_ct_zone(nf_ct_tuplehash_to_ctrack(h)) == zone) { |
334 | NF_CT_STAT_INC(net, found); | 334 | NF_CT_STAT_INC(net, found); |
335 | local_bh_enable(); | 335 | local_bh_enable(); |
336 | return h; | 336 | return h; |
337 | } | 337 | } |
338 | NF_CT_STAT_INC(net, searched); | 338 | NF_CT_STAT_INC(net, searched); |
339 | } | 339 | } |
340 | /* | 340 | /* |
341 | * if the nulls value we got at the end of this lookup is | 341 | * if the nulls value we got at the end of this lookup is |
342 | * not the expected one, we must restart lookup. | 342 | * not the expected one, we must restart lookup. |
343 | * We probably met an item that was moved to another chain. | 343 | * We probably met an item that was moved to another chain. |
344 | */ | 344 | */ |
345 | if (get_nulls_value(n) != bucket) { | 345 | if (get_nulls_value(n) != bucket) { |
346 | NF_CT_STAT_INC(net, search_restart); | 346 | NF_CT_STAT_INC(net, search_restart); |
347 | goto begin; | 347 | goto begin; |
348 | } | 348 | } |
349 | local_bh_enable(); | 349 | local_bh_enable(); |
350 | 350 | ||
351 | return NULL; | 351 | return NULL; |
352 | } | 352 | } |
353 | 353 | ||
354 | struct nf_conntrack_tuple_hash * | 354 | struct nf_conntrack_tuple_hash * |
355 | __nf_conntrack_find(struct net *net, u16 zone, | 355 | __nf_conntrack_find(struct net *net, u16 zone, |
356 | const struct nf_conntrack_tuple *tuple) | 356 | const struct nf_conntrack_tuple *tuple) |
357 | { | 357 | { |
358 | return ____nf_conntrack_find(net, zone, tuple, | 358 | return ____nf_conntrack_find(net, zone, tuple, |
359 | hash_conntrack_raw(tuple, zone)); | 359 | hash_conntrack_raw(tuple, zone)); |
360 | } | 360 | } |
361 | EXPORT_SYMBOL_GPL(__nf_conntrack_find); | 361 | EXPORT_SYMBOL_GPL(__nf_conntrack_find); |
362 | 362 | ||
363 | /* Find a connection corresponding to a tuple. */ | 363 | /* Find a connection corresponding to a tuple. */ |
364 | static struct nf_conntrack_tuple_hash * | 364 | static struct nf_conntrack_tuple_hash * |
365 | __nf_conntrack_find_get(struct net *net, u16 zone, | 365 | __nf_conntrack_find_get(struct net *net, u16 zone, |
366 | const struct nf_conntrack_tuple *tuple, u32 hash) | 366 | const struct nf_conntrack_tuple *tuple, u32 hash) |
367 | { | 367 | { |
368 | struct nf_conntrack_tuple_hash *h; | 368 | struct nf_conntrack_tuple_hash *h; |
369 | struct nf_conn *ct; | 369 | struct nf_conn *ct; |
370 | 370 | ||
371 | rcu_read_lock(); | 371 | rcu_read_lock(); |
372 | begin: | 372 | begin: |
373 | h = ____nf_conntrack_find(net, zone, tuple, hash); | 373 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
374 | if (h) { | 374 | if (h) { |
375 | ct = nf_ct_tuplehash_to_ctrack(h); | 375 | ct = nf_ct_tuplehash_to_ctrack(h); |
376 | if (unlikely(nf_ct_is_dying(ct) || | 376 | if (unlikely(nf_ct_is_dying(ct) || |
377 | !atomic_inc_not_zero(&ct->ct_general.use))) | 377 | !atomic_inc_not_zero(&ct->ct_general.use))) |
378 | h = NULL; | 378 | h = NULL; |
379 | else { | 379 | else { |
380 | if (unlikely(!nf_ct_tuple_equal(tuple, &h->tuple) || | 380 | if (unlikely(!nf_ct_tuple_equal(tuple, &h->tuple) || |
381 | nf_ct_zone(ct) != zone)) { | 381 | nf_ct_zone(ct) != zone)) { |
382 | nf_ct_put(ct); | 382 | nf_ct_put(ct); |
383 | goto begin; | 383 | goto begin; |
384 | } | 384 | } |
385 | } | 385 | } |
386 | } | 386 | } |
387 | rcu_read_unlock(); | 387 | rcu_read_unlock(); |
388 | 388 | ||
389 | return h; | 389 | return h; |
390 | } | 390 | } |
391 | 391 | ||
392 | struct nf_conntrack_tuple_hash * | 392 | struct nf_conntrack_tuple_hash * |
393 | nf_conntrack_find_get(struct net *net, u16 zone, | 393 | nf_conntrack_find_get(struct net *net, u16 zone, |
394 | const struct nf_conntrack_tuple *tuple) | 394 | const struct nf_conntrack_tuple *tuple) |
395 | { | 395 | { |
396 | return __nf_conntrack_find_get(net, zone, tuple, | 396 | return __nf_conntrack_find_get(net, zone, tuple, |
397 | hash_conntrack_raw(tuple, zone)); | 397 | hash_conntrack_raw(tuple, zone)); |
398 | } | 398 | } |
399 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); | 399 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
400 | 400 | ||
401 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, | 401 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
402 | unsigned int hash, | 402 | unsigned int hash, |
403 | unsigned int repl_hash) | 403 | unsigned int repl_hash) |
404 | { | 404 | { |
405 | struct net *net = nf_ct_net(ct); | 405 | struct net *net = nf_ct_net(ct); |
406 | 406 | ||
407 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | 407 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
408 | &net->ct.hash[hash]); | 408 | &net->ct.hash[hash]); |
409 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, | 409 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
410 | &net->ct.hash[repl_hash]); | 410 | &net->ct.hash[repl_hash]); |
411 | } | 411 | } |
412 | 412 | ||
413 | int | 413 | int |
414 | nf_conntrack_hash_check_insert(struct nf_conn *ct) | 414 | nf_conntrack_hash_check_insert(struct nf_conn *ct) |
415 | { | 415 | { |
416 | struct net *net = nf_ct_net(ct); | 416 | struct net *net = nf_ct_net(ct); |
417 | unsigned int hash, repl_hash; | 417 | unsigned int hash, repl_hash; |
418 | struct nf_conntrack_tuple_hash *h; | 418 | struct nf_conntrack_tuple_hash *h; |
419 | struct hlist_nulls_node *n; | 419 | struct hlist_nulls_node *n; |
420 | u16 zone; | 420 | u16 zone; |
421 | 421 | ||
422 | zone = nf_ct_zone(ct); | 422 | zone = nf_ct_zone(ct); |
423 | hash = hash_conntrack(net, zone, | 423 | hash = hash_conntrack(net, zone, |
424 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | 424 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
425 | repl_hash = hash_conntrack(net, zone, | 425 | repl_hash = hash_conntrack(net, zone, |
426 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | 426 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
427 | 427 | ||
428 | spin_lock_bh(&nf_conntrack_lock); | 428 | spin_lock_bh(&nf_conntrack_lock); |
429 | 429 | ||
430 | /* See if there's one in the list already, including reverse */ | 430 | /* See if there's one in the list already, including reverse */ |
431 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) | 431 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
432 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | 432 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
433 | &h->tuple) && | 433 | &h->tuple) && |
434 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | 434 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) |
435 | goto out; | 435 | goto out; |
436 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) | 436 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) |
437 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, | 437 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
438 | &h->tuple) && | 438 | &h->tuple) && |
439 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | 439 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) |
440 | goto out; | 440 | goto out; |
441 | 441 | ||
442 | add_timer(&ct->timeout); | 442 | add_timer(&ct->timeout); |
443 | nf_conntrack_get(&ct->ct_general); | 443 | nf_conntrack_get(&ct->ct_general); |
444 | __nf_conntrack_hash_insert(ct, hash, repl_hash); | 444 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
445 | NF_CT_STAT_INC(net, insert); | 445 | NF_CT_STAT_INC(net, insert); |
446 | spin_unlock_bh(&nf_conntrack_lock); | 446 | spin_unlock_bh(&nf_conntrack_lock); |
447 | 447 | ||
448 | return 0; | 448 | return 0; |
449 | 449 | ||
450 | out: | 450 | out: |
451 | NF_CT_STAT_INC(net, insert_failed); | 451 | NF_CT_STAT_INC(net, insert_failed); |
452 | spin_unlock_bh(&nf_conntrack_lock); | 452 | spin_unlock_bh(&nf_conntrack_lock); |
453 | return -EEXIST; | 453 | return -EEXIST; |
454 | } | 454 | } |
455 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); | 455 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); |
456 | 456 | ||
457 | /* Confirm a connection given skb; places it in hash table */ | 457 | /* Confirm a connection given skb; places it in hash table */ |
458 | int | 458 | int |
459 | __nf_conntrack_confirm(struct sk_buff *skb) | 459 | __nf_conntrack_confirm(struct sk_buff *skb) |
460 | { | 460 | { |
461 | unsigned int hash, repl_hash; | 461 | unsigned int hash, repl_hash; |
462 | struct nf_conntrack_tuple_hash *h; | 462 | struct nf_conntrack_tuple_hash *h; |
463 | struct nf_conn *ct; | 463 | struct nf_conn *ct; |
464 | struct nf_conn_help *help; | 464 | struct nf_conn_help *help; |
465 | struct nf_conn_tstamp *tstamp; | 465 | struct nf_conn_tstamp *tstamp; |
466 | struct hlist_nulls_node *n; | 466 | struct hlist_nulls_node *n; |
467 | enum ip_conntrack_info ctinfo; | 467 | enum ip_conntrack_info ctinfo; |
468 | struct net *net; | 468 | struct net *net; |
469 | u16 zone; | 469 | u16 zone; |
470 | 470 | ||
471 | ct = nf_ct_get(skb, &ctinfo); | 471 | ct = nf_ct_get(skb, &ctinfo); |
472 | net = nf_ct_net(ct); | 472 | net = nf_ct_net(ct); |
473 | 473 | ||
474 | /* ipt_REJECT uses nf_conntrack_attach to attach related | 474 | /* ipt_REJECT uses nf_conntrack_attach to attach related |
475 | ICMP/TCP RST packets in other direction. Actual packet | 475 | ICMP/TCP RST packets in other direction. Actual packet |
476 | which created connection will be IP_CT_NEW or for an | 476 | which created connection will be IP_CT_NEW or for an |
477 | expected connection, IP_CT_RELATED. */ | 477 | expected connection, IP_CT_RELATED. */ |
478 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | 478 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) |
479 | return NF_ACCEPT; | 479 | return NF_ACCEPT; |
480 | 480 | ||
481 | zone = nf_ct_zone(ct); | 481 | zone = nf_ct_zone(ct); |
482 | /* reuse the hash saved before */ | 482 | /* reuse the hash saved before */ |
483 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; | 483 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; |
484 | hash = hash_bucket(hash, net); | 484 | hash = hash_bucket(hash, net); |
485 | repl_hash = hash_conntrack(net, zone, | 485 | repl_hash = hash_conntrack(net, zone, |
486 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | 486 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
487 | 487 | ||
488 | /* We're not in hash table, and we refuse to set up related | 488 | /* We're not in hash table, and we refuse to set up related |
489 | connections for unconfirmed conns. But packet copies and | 489 | connections for unconfirmed conns. But packet copies and |
490 | REJECT will give spurious warnings here. */ | 490 | REJECT will give spurious warnings here. */ |
491 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ | 491 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ |
492 | 492 | ||
493 | /* No external references means no one else could have | 493 | /* No external references means no one else could have |
494 | confirmed us. */ | 494 | confirmed us. */ |
495 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | 495 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
496 | pr_debug("Confirming conntrack %p\n", ct); | 496 | pr_debug("Confirming conntrack %p\n", ct); |
497 | 497 | ||
498 | spin_lock_bh(&nf_conntrack_lock); | 498 | spin_lock_bh(&nf_conntrack_lock); |
499 | 499 | ||
500 | /* We have to check the DYING flag inside the lock to prevent | 500 | /* We have to check the DYING flag inside the lock to prevent |
501 | a race against nf_ct_get_next_corpse() possibly called from | 501 | a race against nf_ct_get_next_corpse() possibly called from |
502 | user context, else we insert an already 'dead' hash, blocking | 502 | user context, else we insert an already 'dead' hash, blocking |
503 | further use of that particular connection -JM */ | 503 | further use of that particular connection -JM */ |
504 | 504 | ||
505 | if (unlikely(nf_ct_is_dying(ct))) { | 505 | if (unlikely(nf_ct_is_dying(ct))) { |
506 | spin_unlock_bh(&nf_conntrack_lock); | 506 | spin_unlock_bh(&nf_conntrack_lock); |
507 | return NF_ACCEPT; | 507 | return NF_ACCEPT; |
508 | } | 508 | } |
509 | 509 | ||
510 | /* See if there's one in the list already, including reverse: | 510 | /* See if there's one in the list already, including reverse: |
511 | NAT could have grabbed it without realizing, since we're | 511 | NAT could have grabbed it without realizing, since we're |
512 | not in the hash. If there is, we lost race. */ | 512 | not in the hash. If there is, we lost race. */ |
513 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) | 513 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
514 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | 514 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
515 | &h->tuple) && | 515 | &h->tuple) && |
516 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | 516 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) |
517 | goto out; | 517 | goto out; |
518 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) | 518 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) |
519 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, | 519 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
520 | &h->tuple) && | 520 | &h->tuple) && |
521 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | 521 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) |
522 | goto out; | 522 | goto out; |
523 | 523 | ||
524 | /* Remove from unconfirmed list */ | 524 | /* Remove from unconfirmed list */ |
525 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | 525 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
526 | 526 | ||
527 | /* Timer relative to confirmation time, not original | 527 | /* Timer relative to confirmation time, not original |
528 | setting time, otherwise we'd get timer wrap in | 528 | setting time, otherwise we'd get timer wrap in |
529 | weird delay cases. */ | 529 | weird delay cases. */ |
530 | ct->timeout.expires += jiffies; | 530 | ct->timeout.expires += jiffies; |
531 | add_timer(&ct->timeout); | 531 | add_timer(&ct->timeout); |
532 | atomic_inc(&ct->ct_general.use); | 532 | atomic_inc(&ct->ct_general.use); |
533 | ct->status |= IPS_CONFIRMED; | 533 | ct->status |= IPS_CONFIRMED; |
534 | 534 | ||
535 | /* set conntrack timestamp, if enabled. */ | 535 | /* set conntrack timestamp, if enabled. */ |
536 | tstamp = nf_conn_tstamp_find(ct); | 536 | tstamp = nf_conn_tstamp_find(ct); |
537 | if (tstamp) { | 537 | if (tstamp) { |
538 | if (skb->tstamp.tv64 == 0) | 538 | if (skb->tstamp.tv64 == 0) |
539 | __net_timestamp(skb); | 539 | __net_timestamp(skb); |
540 | 540 | ||
541 | tstamp->start = ktime_to_ns(skb->tstamp); | 541 | tstamp->start = ktime_to_ns(skb->tstamp); |
542 | } | 542 | } |
543 | /* Since the lookup is lockless, hash insertion must be done after | 543 | /* Since the lookup is lockless, hash insertion must be done after |
544 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | 544 | * starting the timer and setting the CONFIRMED bit. The RCU barriers |
545 | * guarantee that no other CPU can find the conntrack before the above | 545 | * guarantee that no other CPU can find the conntrack before the above |
546 | * stores are visible. | 546 | * stores are visible. |
547 | */ | 547 | */ |
548 | __nf_conntrack_hash_insert(ct, hash, repl_hash); | 548 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
549 | NF_CT_STAT_INC(net, insert); | 549 | NF_CT_STAT_INC(net, insert); |
550 | spin_unlock_bh(&nf_conntrack_lock); | 550 | spin_unlock_bh(&nf_conntrack_lock); |
551 | 551 | ||
552 | help = nfct_help(ct); | 552 | help = nfct_help(ct); |
553 | if (help && help->helper) | 553 | if (help && help->helper) |
554 | nf_conntrack_event_cache(IPCT_HELPER, ct); | 554 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
555 | 555 | ||
556 | nf_conntrack_event_cache(master_ct(ct) ? | 556 | nf_conntrack_event_cache(master_ct(ct) ? |
557 | IPCT_RELATED : IPCT_NEW, ct); | 557 | IPCT_RELATED : IPCT_NEW, ct); |
558 | return NF_ACCEPT; | 558 | return NF_ACCEPT; |
559 | 559 | ||
560 | out: | 560 | out: |
561 | NF_CT_STAT_INC(net, insert_failed); | 561 | NF_CT_STAT_INC(net, insert_failed); |
562 | spin_unlock_bh(&nf_conntrack_lock); | 562 | spin_unlock_bh(&nf_conntrack_lock); |
563 | return NF_DROP; | 563 | return NF_DROP; |
564 | } | 564 | } |
565 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); | 565 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
566 | 566 | ||
567 | /* Returns true if a connection correspondings to the tuple (required | 567 | /* Returns true if a connection correspondings to the tuple (required |
568 | for NAT). */ | 568 | for NAT). */ |
569 | int | 569 | int |
570 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | 570 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, |
571 | const struct nf_conn *ignored_conntrack) | 571 | const struct nf_conn *ignored_conntrack) |
572 | { | 572 | { |
573 | struct net *net = nf_ct_net(ignored_conntrack); | 573 | struct net *net = nf_ct_net(ignored_conntrack); |
574 | struct nf_conntrack_tuple_hash *h; | 574 | struct nf_conntrack_tuple_hash *h; |
575 | struct hlist_nulls_node *n; | 575 | struct hlist_nulls_node *n; |
576 | struct nf_conn *ct; | 576 | struct nf_conn *ct; |
577 | u16 zone = nf_ct_zone(ignored_conntrack); | 577 | u16 zone = nf_ct_zone(ignored_conntrack); |
578 | unsigned int hash = hash_conntrack(net, zone, tuple); | 578 | unsigned int hash = hash_conntrack(net, zone, tuple); |
579 | 579 | ||
580 | /* Disable BHs the entire time since we need to disable them at | 580 | /* Disable BHs the entire time since we need to disable them at |
581 | * least once for the stats anyway. | 581 | * least once for the stats anyway. |
582 | */ | 582 | */ |
583 | rcu_read_lock_bh(); | 583 | rcu_read_lock_bh(); |
584 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { | 584 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
585 | ct = nf_ct_tuplehash_to_ctrack(h); | 585 | ct = nf_ct_tuplehash_to_ctrack(h); |
586 | if (ct != ignored_conntrack && | 586 | if (ct != ignored_conntrack && |
587 | nf_ct_tuple_equal(tuple, &h->tuple) && | 587 | nf_ct_tuple_equal(tuple, &h->tuple) && |
588 | nf_ct_zone(ct) == zone) { | 588 | nf_ct_zone(ct) == zone) { |
589 | NF_CT_STAT_INC(net, found); | 589 | NF_CT_STAT_INC(net, found); |
590 | rcu_read_unlock_bh(); | 590 | rcu_read_unlock_bh(); |
591 | return 1; | 591 | return 1; |
592 | } | 592 | } |
593 | NF_CT_STAT_INC(net, searched); | 593 | NF_CT_STAT_INC(net, searched); |
594 | } | 594 | } |
595 | rcu_read_unlock_bh(); | 595 | rcu_read_unlock_bh(); |
596 | 596 | ||
597 | return 0; | 597 | return 0; |
598 | } | 598 | } |
599 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); | 599 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
600 | 600 | ||
601 | #define NF_CT_EVICTION_RANGE 8 | 601 | #define NF_CT_EVICTION_RANGE 8 |
602 | 602 | ||
603 | /* There's a small race here where we may free a just-assured | 603 | /* There's a small race here where we may free a just-assured |
604 | connection. Too bad: we're in trouble anyway. */ | 604 | connection. Too bad: we're in trouble anyway. */ |
605 | static noinline int early_drop(struct net *net, unsigned int hash) | 605 | static noinline int early_drop(struct net *net, unsigned int hash) |
606 | { | 606 | { |
607 | /* Use oldest entry, which is roughly LRU */ | 607 | /* Use oldest entry, which is roughly LRU */ |
608 | struct nf_conntrack_tuple_hash *h; | 608 | struct nf_conntrack_tuple_hash *h; |
609 | struct nf_conn *ct = NULL, *tmp; | 609 | struct nf_conn *ct = NULL, *tmp; |
610 | struct hlist_nulls_node *n; | 610 | struct hlist_nulls_node *n; |
611 | unsigned int i, cnt = 0; | 611 | unsigned int i, cnt = 0; |
612 | int dropped = 0; | 612 | int dropped = 0; |
613 | 613 | ||
614 | rcu_read_lock(); | 614 | rcu_read_lock(); |
615 | for (i = 0; i < net->ct.htable_size; i++) { | 615 | for (i = 0; i < net->ct.htable_size; i++) { |
616 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], | 616 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
617 | hnnode) { | 617 | hnnode) { |
618 | tmp = nf_ct_tuplehash_to_ctrack(h); | 618 | tmp = nf_ct_tuplehash_to_ctrack(h); |
619 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) | 619 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) |
620 | ct = tmp; | 620 | ct = tmp; |
621 | cnt++; | 621 | cnt++; |
622 | } | 622 | } |
623 | 623 | ||
624 | if (ct != NULL) { | 624 | if (ct != NULL) { |
625 | if (likely(!nf_ct_is_dying(ct) && | 625 | if (likely(!nf_ct_is_dying(ct) && |
626 | atomic_inc_not_zero(&ct->ct_general.use))) | 626 | atomic_inc_not_zero(&ct->ct_general.use))) |
627 | break; | 627 | break; |
628 | else | 628 | else |
629 | ct = NULL; | 629 | ct = NULL; |
630 | } | 630 | } |
631 | 631 | ||
632 | if (cnt >= NF_CT_EVICTION_RANGE) | 632 | if (cnt >= NF_CT_EVICTION_RANGE) |
633 | break; | 633 | break; |
634 | 634 | ||
635 | hash = (hash + 1) % net->ct.htable_size; | 635 | hash = (hash + 1) % net->ct.htable_size; |
636 | } | 636 | } |
637 | rcu_read_unlock(); | 637 | rcu_read_unlock(); |
638 | 638 | ||
639 | if (!ct) | 639 | if (!ct) |
640 | return dropped; | 640 | return dropped; |
641 | 641 | ||
642 | if (del_timer(&ct->timeout)) { | 642 | if (del_timer(&ct->timeout)) { |
643 | death_by_timeout((unsigned long)ct); | 643 | death_by_timeout((unsigned long)ct); |
644 | /* Check if we indeed killed this entry. Reliable event | 644 | /* Check if we indeed killed this entry. Reliable event |
645 | delivery may have inserted it into the dying list. */ | 645 | delivery may have inserted it into the dying list. */ |
646 | if (test_bit(IPS_DYING_BIT, &ct->status)) { | 646 | if (test_bit(IPS_DYING_BIT, &ct->status)) { |
647 | dropped = 1; | 647 | dropped = 1; |
648 | NF_CT_STAT_INC_ATOMIC(net, early_drop); | 648 | NF_CT_STAT_INC_ATOMIC(net, early_drop); |
649 | } | 649 | } |
650 | } | 650 | } |
651 | nf_ct_put(ct); | 651 | nf_ct_put(ct); |
652 | return dropped; | 652 | return dropped; |
653 | } | 653 | } |
654 | 654 | ||
655 | void init_nf_conntrack_hash_rnd(void) | 655 | void init_nf_conntrack_hash_rnd(void) |
656 | { | 656 | { |
657 | unsigned int rand; | 657 | unsigned int rand; |
658 | 658 | ||
659 | /* | 659 | /* |
660 | * Why not initialize nf_conntrack_rnd in a "init()" function ? | 660 | * Why not initialize nf_conntrack_rnd in a "init()" function ? |
661 | * Because there isn't enough entropy when system initializing, | 661 | * Because there isn't enough entropy when system initializing, |
662 | * and we initialize it as late as possible. | 662 | * and we initialize it as late as possible. |
663 | */ | 663 | */ |
664 | do { | 664 | do { |
665 | get_random_bytes(&rand, sizeof(rand)); | 665 | get_random_bytes(&rand, sizeof(rand)); |
666 | } while (!rand); | 666 | } while (!rand); |
667 | cmpxchg(&nf_conntrack_hash_rnd, 0, rand); | 667 | cmpxchg(&nf_conntrack_hash_rnd, 0, rand); |
668 | } | 668 | } |
669 | 669 | ||
670 | static struct nf_conn * | 670 | static struct nf_conn * |
671 | __nf_conntrack_alloc(struct net *net, u16 zone, | 671 | __nf_conntrack_alloc(struct net *net, u16 zone, |
672 | const struct nf_conntrack_tuple *orig, | 672 | const struct nf_conntrack_tuple *orig, |
673 | const struct nf_conntrack_tuple *repl, | 673 | const struct nf_conntrack_tuple *repl, |
674 | gfp_t gfp, u32 hash) | 674 | gfp_t gfp, u32 hash) |
675 | { | 675 | { |
676 | struct nf_conn *ct; | 676 | struct nf_conn *ct; |
677 | 677 | ||
678 | if (unlikely(!nf_conntrack_hash_rnd)) { | 678 | if (unlikely(!nf_conntrack_hash_rnd)) { |
679 | init_nf_conntrack_hash_rnd(); | 679 | init_nf_conntrack_hash_rnd(); |
680 | /* recompute the hash as nf_conntrack_hash_rnd is initialized */ | 680 | /* recompute the hash as nf_conntrack_hash_rnd is initialized */ |
681 | hash = hash_conntrack_raw(orig, zone); | 681 | hash = hash_conntrack_raw(orig, zone); |
682 | } | 682 | } |
683 | 683 | ||
684 | /* We don't want any race condition at early drop stage */ | 684 | /* We don't want any race condition at early drop stage */ |
685 | atomic_inc(&net->ct.count); | 685 | atomic_inc(&net->ct.count); |
686 | 686 | ||
687 | if (nf_conntrack_max && | 687 | if (nf_conntrack_max && |
688 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { | 688 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
689 | if (!early_drop(net, hash_bucket(hash, net))) { | 689 | if (!early_drop(net, hash_bucket(hash, net))) { |
690 | atomic_dec(&net->ct.count); | 690 | atomic_dec(&net->ct.count); |
691 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); | 691 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); |
692 | return ERR_PTR(-ENOMEM); | 692 | return ERR_PTR(-ENOMEM); |
693 | } | 693 | } |
694 | } | 694 | } |
695 | 695 | ||
696 | /* | 696 | /* |
697 | * Do not use kmem_cache_zalloc(), as this cache uses | 697 | * Do not use kmem_cache_zalloc(), as this cache uses |
698 | * SLAB_DESTROY_BY_RCU. | 698 | * SLAB_DESTROY_BY_RCU. |
699 | */ | 699 | */ |
700 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); | 700 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); |
701 | if (ct == NULL) { | 701 | if (ct == NULL) { |
702 | atomic_dec(&net->ct.count); | 702 | atomic_dec(&net->ct.count); |
703 | return ERR_PTR(-ENOMEM); | 703 | return ERR_PTR(-ENOMEM); |
704 | } | 704 | } |
705 | /* | 705 | /* |
706 | * Let ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.next | 706 | * Let ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.next |
707 | * and ct->tuplehash[IP_CT_DIR_REPLY].hnnode.next unchanged. | 707 | * and ct->tuplehash[IP_CT_DIR_REPLY].hnnode.next unchanged. |
708 | */ | 708 | */ |
709 | memset(&ct->tuplehash[IP_CT_DIR_MAX], 0, | 709 | memset(&ct->tuplehash[IP_CT_DIR_MAX], 0, |
710 | offsetof(struct nf_conn, proto) - | 710 | offsetof(struct nf_conn, proto) - |
711 | offsetof(struct nf_conn, tuplehash[IP_CT_DIR_MAX])); | 711 | offsetof(struct nf_conn, tuplehash[IP_CT_DIR_MAX])); |
712 | spin_lock_init(&ct->lock); | 712 | spin_lock_init(&ct->lock); |
713 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; | 713 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
714 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; | 714 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
715 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; | 715 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
716 | /* save hash for reusing when confirming */ | 716 | /* save hash for reusing when confirming */ |
717 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; | 717 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; |
718 | /* Don't set timer yet: wait for confirmation */ | 718 | /* Don't set timer yet: wait for confirmation */ |
719 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); | 719 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
720 | write_pnet(&ct->ct_net, net); | 720 | write_pnet(&ct->ct_net, net); |
721 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 721 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
722 | if (zone) { | 722 | if (zone) { |
723 | struct nf_conntrack_zone *nf_ct_zone; | 723 | struct nf_conntrack_zone *nf_ct_zone; |
724 | 724 | ||
725 | nf_ct_zone = nf_ct_ext_add(ct, NF_CT_EXT_ZONE, GFP_ATOMIC); | 725 | nf_ct_zone = nf_ct_ext_add(ct, NF_CT_EXT_ZONE, GFP_ATOMIC); |
726 | if (!nf_ct_zone) | 726 | if (!nf_ct_zone) |
727 | goto out_free; | 727 | goto out_free; |
728 | nf_ct_zone->id = zone; | 728 | nf_ct_zone->id = zone; |
729 | } | 729 | } |
730 | #endif | 730 | #endif |
731 | /* | 731 | /* |
732 | * changes to lookup keys must be done before setting refcnt to 1 | 732 | * changes to lookup keys must be done before setting refcnt to 1 |
733 | */ | 733 | */ |
734 | smp_wmb(); | 734 | smp_wmb(); |
735 | atomic_set(&ct->ct_general.use, 1); | 735 | atomic_set(&ct->ct_general.use, 1); |
736 | return ct; | 736 | return ct; |
737 | 737 | ||
738 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 738 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
739 | out_free: | 739 | out_free: |
740 | atomic_dec(&net->ct.count); | 740 | atomic_dec(&net->ct.count); |
741 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); | 741 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
742 | return ERR_PTR(-ENOMEM); | 742 | return ERR_PTR(-ENOMEM); |
743 | #endif | 743 | #endif |
744 | } | 744 | } |
745 | 745 | ||
746 | struct nf_conn *nf_conntrack_alloc(struct net *net, u16 zone, | 746 | struct nf_conn *nf_conntrack_alloc(struct net *net, u16 zone, |
747 | const struct nf_conntrack_tuple *orig, | 747 | const struct nf_conntrack_tuple *orig, |
748 | const struct nf_conntrack_tuple *repl, | 748 | const struct nf_conntrack_tuple *repl, |
749 | gfp_t gfp) | 749 | gfp_t gfp) |
750 | { | 750 | { |
751 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); | 751 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); |
752 | } | 752 | } |
753 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); | 753 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
754 | 754 | ||
755 | void nf_conntrack_free(struct nf_conn *ct) | 755 | void nf_conntrack_free(struct nf_conn *ct) |
756 | { | 756 | { |
757 | struct net *net = nf_ct_net(ct); | 757 | struct net *net = nf_ct_net(ct); |
758 | 758 | ||
759 | nf_ct_ext_destroy(ct); | 759 | nf_ct_ext_destroy(ct); |
760 | atomic_dec(&net->ct.count); | 760 | atomic_dec(&net->ct.count); |
761 | nf_ct_ext_free(ct); | 761 | nf_ct_ext_free(ct); |
762 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); | 762 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
763 | } | 763 | } |
764 | EXPORT_SYMBOL_GPL(nf_conntrack_free); | 764 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
765 | 765 | ||
766 | /* Allocate a new conntrack: we return -ENOMEM if classification | 766 | /* Allocate a new conntrack: we return -ENOMEM if classification |
767 | failed due to stress. Otherwise it really is unclassifiable. */ | 767 | failed due to stress. Otherwise it really is unclassifiable. */ |
768 | static struct nf_conntrack_tuple_hash * | 768 | static struct nf_conntrack_tuple_hash * |
769 | init_conntrack(struct net *net, struct nf_conn *tmpl, | 769 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
770 | const struct nf_conntrack_tuple *tuple, | 770 | const struct nf_conntrack_tuple *tuple, |
771 | struct nf_conntrack_l3proto *l3proto, | 771 | struct nf_conntrack_l3proto *l3proto, |
772 | struct nf_conntrack_l4proto *l4proto, | 772 | struct nf_conntrack_l4proto *l4proto, |
773 | struct sk_buff *skb, | 773 | struct sk_buff *skb, |
774 | unsigned int dataoff, u32 hash) | 774 | unsigned int dataoff, u32 hash) |
775 | { | 775 | { |
776 | struct nf_conn *ct; | 776 | struct nf_conn *ct; |
777 | struct nf_conn_help *help; | 777 | struct nf_conn_help *help; |
778 | struct nf_conntrack_tuple repl_tuple; | 778 | struct nf_conntrack_tuple repl_tuple; |
779 | struct nf_conntrack_ecache *ecache; | 779 | struct nf_conntrack_ecache *ecache; |
780 | struct nf_conntrack_expect *exp; | 780 | struct nf_conntrack_expect *exp; |
781 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; | 781 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
782 | struct nf_conn_timeout *timeout_ext; | 782 | struct nf_conn_timeout *timeout_ext; |
783 | unsigned int *timeouts; | 783 | unsigned int *timeouts; |
784 | 784 | ||
785 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { | 785 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
786 | pr_debug("Can't invert tuple.\n"); | 786 | pr_debug("Can't invert tuple.\n"); |
787 | return NULL; | 787 | return NULL; |
788 | } | 788 | } |
789 | 789 | ||
790 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, | 790 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
791 | hash); | 791 | hash); |
792 | if (IS_ERR(ct)) | 792 | if (IS_ERR(ct)) |
793 | return (struct nf_conntrack_tuple_hash *)ct; | 793 | return (struct nf_conntrack_tuple_hash *)ct; |
794 | 794 | ||
795 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; | 795 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; |
796 | if (timeout_ext) | 796 | if (timeout_ext) |
797 | timeouts = NF_CT_TIMEOUT_EXT_DATA(timeout_ext); | 797 | timeouts = NF_CT_TIMEOUT_EXT_DATA(timeout_ext); |
798 | else | 798 | else |
799 | timeouts = l4proto->get_timeouts(net); | 799 | timeouts = l4proto->get_timeouts(net); |
800 | 800 | ||
801 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { | 801 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { |
802 | nf_conntrack_free(ct); | 802 | nf_conntrack_free(ct); |
803 | pr_debug("init conntrack: can't track with proto module\n"); | 803 | pr_debug("init conntrack: can't track with proto module\n"); |
804 | return NULL; | 804 | return NULL; |
805 | } | 805 | } |
806 | 806 | ||
807 | if (timeout_ext) | 807 | if (timeout_ext) |
808 | nf_ct_timeout_ext_add(ct, timeout_ext->timeout, GFP_ATOMIC); | 808 | nf_ct_timeout_ext_add(ct, timeout_ext->timeout, GFP_ATOMIC); |
809 | 809 | ||
810 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); | 810 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
811 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); | 811 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
812 | 812 | ||
813 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; | 813 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; |
814 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, | 814 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, |
815 | ecache ? ecache->expmask : 0, | 815 | ecache ? ecache->expmask : 0, |
816 | GFP_ATOMIC); | 816 | GFP_ATOMIC); |
817 | 817 | ||
818 | spin_lock_bh(&nf_conntrack_lock); | 818 | spin_lock_bh(&nf_conntrack_lock); |
819 | exp = nf_ct_find_expectation(net, zone, tuple); | 819 | exp = nf_ct_find_expectation(net, zone, tuple); |
820 | if (exp) { | 820 | if (exp) { |
821 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", | 821 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
822 | ct, exp); | 822 | ct, exp); |
823 | /* Welcome, Mr. Bond. We've been expecting you... */ | 823 | /* Welcome, Mr. Bond. We've been expecting you... */ |
824 | __set_bit(IPS_EXPECTED_BIT, &ct->status); | 824 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
825 | ct->master = exp->master; | 825 | ct->master = exp->master; |
826 | if (exp->helper) { | 826 | if (exp->helper) { |
827 | help = nf_ct_helper_ext_add(ct, exp->helper, | 827 | help = nf_ct_helper_ext_add(ct, exp->helper, |
828 | GFP_ATOMIC); | 828 | GFP_ATOMIC); |
829 | if (help) | 829 | if (help) |
830 | rcu_assign_pointer(help->helper, exp->helper); | 830 | rcu_assign_pointer(help->helper, exp->helper); |
831 | } | 831 | } |
832 | 832 | ||
833 | #ifdef CONFIG_NF_CONNTRACK_MARK | 833 | #ifdef CONFIG_NF_CONNTRACK_MARK |
834 | ct->mark = exp->master->mark; | 834 | ct->mark = exp->master->mark; |
835 | #endif | 835 | #endif |
836 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | 836 | #ifdef CONFIG_NF_CONNTRACK_SECMARK |
837 | ct->secmark = exp->master->secmark; | 837 | ct->secmark = exp->master->secmark; |
838 | #endif | 838 | #endif |
839 | nf_conntrack_get(&ct->master->ct_general); | 839 | nf_conntrack_get(&ct->master->ct_general); |
840 | NF_CT_STAT_INC(net, expect_new); | 840 | NF_CT_STAT_INC(net, expect_new); |
841 | } else { | 841 | } else { |
842 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); | 842 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
843 | NF_CT_STAT_INC(net, new); | 843 | NF_CT_STAT_INC(net, new); |
844 | } | 844 | } |
845 | 845 | ||
846 | /* Overload tuple linked list to put us in unconfirmed list. */ | 846 | /* Overload tuple linked list to put us in unconfirmed list. */ |
847 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | 847 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
848 | &net->ct.unconfirmed); | 848 | &net->ct.unconfirmed); |
849 | 849 | ||
850 | spin_unlock_bh(&nf_conntrack_lock); | 850 | spin_unlock_bh(&nf_conntrack_lock); |
851 | 851 | ||
852 | if (exp) { | 852 | if (exp) { |
853 | if (exp->expectfn) | 853 | if (exp->expectfn) |
854 | exp->expectfn(ct, exp); | 854 | exp->expectfn(ct, exp); |
855 | nf_ct_expect_put(exp); | 855 | nf_ct_expect_put(exp); |
856 | } | 856 | } |
857 | 857 | ||
858 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; | 858 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
859 | } | 859 | } |
860 | 860 | ||
861 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | 861 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ |
862 | static inline struct nf_conn * | 862 | static inline struct nf_conn * |
863 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, | 863 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
864 | struct sk_buff *skb, | 864 | struct sk_buff *skb, |
865 | unsigned int dataoff, | 865 | unsigned int dataoff, |
866 | u_int16_t l3num, | 866 | u_int16_t l3num, |
867 | u_int8_t protonum, | 867 | u_int8_t protonum, |
868 | struct nf_conntrack_l3proto *l3proto, | 868 | struct nf_conntrack_l3proto *l3proto, |
869 | struct nf_conntrack_l4proto *l4proto, | 869 | struct nf_conntrack_l4proto *l4proto, |
870 | int *set_reply, | 870 | int *set_reply, |
871 | enum ip_conntrack_info *ctinfo) | 871 | enum ip_conntrack_info *ctinfo) |
872 | { | 872 | { |
873 | struct nf_conntrack_tuple tuple; | 873 | struct nf_conntrack_tuple tuple; |
874 | struct nf_conntrack_tuple_hash *h; | 874 | struct nf_conntrack_tuple_hash *h; |
875 | struct nf_conn *ct; | 875 | struct nf_conn *ct; |
876 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; | 876 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
877 | u32 hash; | 877 | u32 hash; |
878 | 878 | ||
879 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), | 879 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
880 | dataoff, l3num, protonum, &tuple, l3proto, | 880 | dataoff, l3num, protonum, &tuple, l3proto, |
881 | l4proto)) { | 881 | l4proto)) { |
882 | pr_debug("resolve_normal_ct: Can't get tuple\n"); | 882 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
883 | return NULL; | 883 | return NULL; |
884 | } | 884 | } |
885 | 885 | ||
886 | /* look for tuple match */ | 886 | /* look for tuple match */ |
887 | hash = hash_conntrack_raw(&tuple, zone); | 887 | hash = hash_conntrack_raw(&tuple, zone); |
888 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); | 888 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); |
889 | if (!h) { | 889 | if (!h) { |
890 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, | 890 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
891 | skb, dataoff, hash); | 891 | skb, dataoff, hash); |
892 | if (!h) | 892 | if (!h) |
893 | return NULL; | 893 | return NULL; |
894 | if (IS_ERR(h)) | 894 | if (IS_ERR(h)) |
895 | return (void *)h; | 895 | return (void *)h; |
896 | } | 896 | } |
897 | ct = nf_ct_tuplehash_to_ctrack(h); | 897 | ct = nf_ct_tuplehash_to_ctrack(h); |
898 | 898 | ||
899 | /* It exists; we have (non-exclusive) reference. */ | 899 | /* It exists; we have (non-exclusive) reference. */ |
900 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | 900 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { |
901 | *ctinfo = IP_CT_ESTABLISHED_REPLY; | 901 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
902 | /* Please set reply bit if this packet OK */ | 902 | /* Please set reply bit if this packet OK */ |
903 | *set_reply = 1; | 903 | *set_reply = 1; |
904 | } else { | 904 | } else { |
905 | /* Once we've had two way comms, always ESTABLISHED. */ | 905 | /* Once we've had two way comms, always ESTABLISHED. */ |
906 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | 906 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { |
907 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); | 907 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
908 | *ctinfo = IP_CT_ESTABLISHED; | 908 | *ctinfo = IP_CT_ESTABLISHED; |
909 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | 909 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { |
910 | pr_debug("nf_conntrack_in: related packet for %p\n", | 910 | pr_debug("nf_conntrack_in: related packet for %p\n", |
911 | ct); | 911 | ct); |
912 | *ctinfo = IP_CT_RELATED; | 912 | *ctinfo = IP_CT_RELATED; |
913 | } else { | 913 | } else { |
914 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); | 914 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
915 | *ctinfo = IP_CT_NEW; | 915 | *ctinfo = IP_CT_NEW; |
916 | } | 916 | } |
917 | *set_reply = 0; | 917 | *set_reply = 0; |
918 | } | 918 | } |
919 | skb->nfct = &ct->ct_general; | 919 | skb->nfct = &ct->ct_general; |
920 | skb->nfctinfo = *ctinfo; | 920 | skb->nfctinfo = *ctinfo; |
921 | return ct; | 921 | return ct; |
922 | } | 922 | } |
923 | 923 | ||
924 | unsigned int | 924 | unsigned int |
925 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, | 925 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
926 | struct sk_buff *skb) | 926 | struct sk_buff *skb) |
927 | { | 927 | { |
928 | struct nf_conn *ct, *tmpl = NULL; | 928 | struct nf_conn *ct, *tmpl = NULL; |
929 | enum ip_conntrack_info ctinfo; | 929 | enum ip_conntrack_info ctinfo; |
930 | struct nf_conntrack_l3proto *l3proto; | 930 | struct nf_conntrack_l3proto *l3proto; |
931 | struct nf_conntrack_l4proto *l4proto; | 931 | struct nf_conntrack_l4proto *l4proto; |
932 | unsigned int *timeouts; | 932 | unsigned int *timeouts; |
933 | unsigned int dataoff; | 933 | unsigned int dataoff; |
934 | u_int8_t protonum; | 934 | u_int8_t protonum; |
935 | int set_reply = 0; | 935 | int set_reply = 0; |
936 | int ret; | 936 | int ret; |
937 | 937 | ||
938 | if (skb->nfct) { | 938 | if (skb->nfct) { |
939 | /* Previously seen (loopback or untracked)? Ignore. */ | 939 | /* Previously seen (loopback or untracked)? Ignore. */ |
940 | tmpl = (struct nf_conn *)skb->nfct; | 940 | tmpl = (struct nf_conn *)skb->nfct; |
941 | if (!nf_ct_is_template(tmpl)) { | 941 | if (!nf_ct_is_template(tmpl)) { |
942 | NF_CT_STAT_INC_ATOMIC(net, ignore); | 942 | NF_CT_STAT_INC_ATOMIC(net, ignore); |
943 | return NF_ACCEPT; | 943 | return NF_ACCEPT; |
944 | } | 944 | } |
945 | skb->nfct = NULL; | 945 | skb->nfct = NULL; |
946 | } | 946 | } |
947 | 947 | ||
948 | /* rcu_read_lock()ed by nf_hook_slow */ | 948 | /* rcu_read_lock()ed by nf_hook_slow */ |
949 | l3proto = __nf_ct_l3proto_find(pf); | 949 | l3proto = __nf_ct_l3proto_find(pf); |
950 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), | 950 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
951 | &dataoff, &protonum); | 951 | &dataoff, &protonum); |
952 | if (ret <= 0) { | 952 | if (ret <= 0) { |
953 | pr_debug("not prepared to track yet or error occurred\n"); | 953 | pr_debug("not prepared to track yet or error occurred\n"); |
954 | NF_CT_STAT_INC_ATOMIC(net, error); | 954 | NF_CT_STAT_INC_ATOMIC(net, error); |
955 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 955 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
956 | ret = -ret; | 956 | ret = -ret; |
957 | goto out; | 957 | goto out; |
958 | } | 958 | } |
959 | 959 | ||
960 | l4proto = __nf_ct_l4proto_find(pf, protonum); | 960 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
961 | 961 | ||
962 | /* It may be an special packet, error, unclean... | 962 | /* It may be an special packet, error, unclean... |
963 | * inverse of the return code tells to the netfilter | 963 | * inverse of the return code tells to the netfilter |
964 | * core what to do with the packet. */ | 964 | * core what to do with the packet. */ |
965 | if (l4proto->error != NULL) { | 965 | if (l4proto->error != NULL) { |
966 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, | 966 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
967 | pf, hooknum); | 967 | pf, hooknum); |
968 | if (ret <= 0) { | 968 | if (ret <= 0) { |
969 | NF_CT_STAT_INC_ATOMIC(net, error); | 969 | NF_CT_STAT_INC_ATOMIC(net, error); |
970 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 970 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
971 | ret = -ret; | 971 | ret = -ret; |
972 | goto out; | 972 | goto out; |
973 | } | 973 | } |
974 | /* ICMP[v6] protocol trackers may assign one conntrack. */ | 974 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
975 | if (skb->nfct) | 975 | if (skb->nfct) |
976 | goto out; | 976 | goto out; |
977 | } | 977 | } |
978 | 978 | ||
979 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, | 979 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
980 | l3proto, l4proto, &set_reply, &ctinfo); | 980 | l3proto, l4proto, &set_reply, &ctinfo); |
981 | if (!ct) { | 981 | if (!ct) { |
982 | /* Not valid part of a connection */ | 982 | /* Not valid part of a connection */ |
983 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 983 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
984 | ret = NF_ACCEPT; | 984 | ret = NF_ACCEPT; |
985 | goto out; | 985 | goto out; |
986 | } | 986 | } |
987 | 987 | ||
988 | if (IS_ERR(ct)) { | 988 | if (IS_ERR(ct)) { |
989 | /* Too stressed to deal. */ | 989 | /* Too stressed to deal. */ |
990 | NF_CT_STAT_INC_ATOMIC(net, drop); | 990 | NF_CT_STAT_INC_ATOMIC(net, drop); |
991 | ret = NF_DROP; | 991 | ret = NF_DROP; |
992 | goto out; | 992 | goto out; |
993 | } | 993 | } |
994 | 994 | ||
995 | NF_CT_ASSERT(skb->nfct); | 995 | NF_CT_ASSERT(skb->nfct); |
996 | 996 | ||
997 | /* Decide what timeout policy we want to apply to this flow. */ | 997 | /* Decide what timeout policy we want to apply to this flow. */ |
998 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); | 998 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); |
999 | 999 | ||
1000 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); | 1000 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); |
1001 | if (ret <= 0) { | 1001 | if (ret <= 0) { |
1002 | /* Invalid: inverse of the return code tells | 1002 | /* Invalid: inverse of the return code tells |
1003 | * the netfilter core what to do */ | 1003 | * the netfilter core what to do */ |
1004 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); | 1004 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
1005 | nf_conntrack_put(skb->nfct); | 1005 | nf_conntrack_put(skb->nfct); |
1006 | skb->nfct = NULL; | 1006 | skb->nfct = NULL; |
1007 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 1007 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
1008 | if (ret == -NF_DROP) | 1008 | if (ret == -NF_DROP) |
1009 | NF_CT_STAT_INC_ATOMIC(net, drop); | 1009 | NF_CT_STAT_INC_ATOMIC(net, drop); |
1010 | ret = -ret; | 1010 | ret = -ret; |
1011 | goto out; | 1011 | goto out; |
1012 | } | 1012 | } |
1013 | 1013 | ||
1014 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | 1014 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) |
1015 | nf_conntrack_event_cache(IPCT_REPLY, ct); | 1015 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
1016 | out: | 1016 | out: |
1017 | if (tmpl) { | 1017 | if (tmpl) { |
1018 | /* Special case: we have to repeat this hook, assign the | 1018 | /* Special case: we have to repeat this hook, assign the |
1019 | * template again to this packet. We assume that this packet | 1019 | * template again to this packet. We assume that this packet |
1020 | * has no conntrack assigned. This is used by nf_ct_tcp. */ | 1020 | * has no conntrack assigned. This is used by nf_ct_tcp. */ |
1021 | if (ret == NF_REPEAT) | 1021 | if (ret == NF_REPEAT) |
1022 | skb->nfct = (struct nf_conntrack *)tmpl; | 1022 | skb->nfct = (struct nf_conntrack *)tmpl; |
1023 | else | 1023 | else |
1024 | nf_ct_put(tmpl); | 1024 | nf_ct_put(tmpl); |
1025 | } | 1025 | } |
1026 | 1026 | ||
1027 | return ret; | 1027 | return ret; |
1028 | } | 1028 | } |
1029 | EXPORT_SYMBOL_GPL(nf_conntrack_in); | 1029 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
1030 | 1030 | ||
1031 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, | 1031 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
1032 | const struct nf_conntrack_tuple *orig) | 1032 | const struct nf_conntrack_tuple *orig) |
1033 | { | 1033 | { |
1034 | bool ret; | 1034 | bool ret; |
1035 | 1035 | ||
1036 | rcu_read_lock(); | 1036 | rcu_read_lock(); |
1037 | ret = nf_ct_invert_tuple(inverse, orig, | 1037 | ret = nf_ct_invert_tuple(inverse, orig, |
1038 | __nf_ct_l3proto_find(orig->src.l3num), | 1038 | __nf_ct_l3proto_find(orig->src.l3num), |
1039 | __nf_ct_l4proto_find(orig->src.l3num, | 1039 | __nf_ct_l4proto_find(orig->src.l3num, |
1040 | orig->dst.protonum)); | 1040 | orig->dst.protonum)); |
1041 | rcu_read_unlock(); | 1041 | rcu_read_unlock(); |
1042 | return ret; | 1042 | return ret; |
1043 | } | 1043 | } |
1044 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); | 1044 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
1045 | 1045 | ||
1046 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is | 1046 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
1047 | implicitly racy: see __nf_conntrack_confirm */ | 1047 | implicitly racy: see __nf_conntrack_confirm */ |
1048 | void nf_conntrack_alter_reply(struct nf_conn *ct, | 1048 | void nf_conntrack_alter_reply(struct nf_conn *ct, |
1049 | const struct nf_conntrack_tuple *newreply) | 1049 | const struct nf_conntrack_tuple *newreply) |
1050 | { | 1050 | { |
1051 | struct nf_conn_help *help = nfct_help(ct); | 1051 | struct nf_conn_help *help = nfct_help(ct); |
1052 | 1052 | ||
1053 | /* Should be unconfirmed, so not in hash table yet */ | 1053 | /* Should be unconfirmed, so not in hash table yet */ |
1054 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | 1054 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
1055 | 1055 | ||
1056 | pr_debug("Altering reply tuple of %p to ", ct); | 1056 | pr_debug("Altering reply tuple of %p to ", ct); |
1057 | nf_ct_dump_tuple(newreply); | 1057 | nf_ct_dump_tuple(newreply); |
1058 | 1058 | ||
1059 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | 1059 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; |
1060 | if (ct->master || (help && !hlist_empty(&help->expectations))) | 1060 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
1061 | return; | 1061 | return; |
1062 | 1062 | ||
1063 | rcu_read_lock(); | 1063 | rcu_read_lock(); |
1064 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); | 1064 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
1065 | rcu_read_unlock(); | 1065 | rcu_read_unlock(); |
1066 | } | 1066 | } |
1067 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); | 1067 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
1068 | 1068 | ||
1069 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ | 1069 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
1070 | void __nf_ct_refresh_acct(struct nf_conn *ct, | 1070 | void __nf_ct_refresh_acct(struct nf_conn *ct, |
1071 | enum ip_conntrack_info ctinfo, | 1071 | enum ip_conntrack_info ctinfo, |
1072 | const struct sk_buff *skb, | 1072 | const struct sk_buff *skb, |
1073 | unsigned long extra_jiffies, | 1073 | unsigned long extra_jiffies, |
1074 | int do_acct) | 1074 | int do_acct) |
1075 | { | 1075 | { |
1076 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); | 1076 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
1077 | NF_CT_ASSERT(skb); | 1077 | NF_CT_ASSERT(skb); |
1078 | 1078 | ||
1079 | /* Only update if this is not a fixed timeout */ | 1079 | /* Only update if this is not a fixed timeout */ |
1080 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) | 1080 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
1081 | goto acct; | 1081 | goto acct; |
1082 | 1082 | ||
1083 | /* If not in hash table, timer will not be active yet */ | 1083 | /* If not in hash table, timer will not be active yet */ |
1084 | if (!nf_ct_is_confirmed(ct)) { | 1084 | if (!nf_ct_is_confirmed(ct)) { |
1085 | ct->timeout.expires = extra_jiffies; | 1085 | ct->timeout.expires = extra_jiffies; |
1086 | } else { | 1086 | } else { |
1087 | unsigned long newtime = jiffies + extra_jiffies; | 1087 | unsigned long newtime = jiffies + extra_jiffies; |
1088 | 1088 | ||
1089 | /* Only update the timeout if the new timeout is at least | 1089 | /* Only update the timeout if the new timeout is at least |
1090 | HZ jiffies from the old timeout. Need del_timer for race | 1090 | HZ jiffies from the old timeout. Need del_timer for race |
1091 | avoidance (may already be dying). */ | 1091 | avoidance (may already be dying). */ |
1092 | if (newtime - ct->timeout.expires >= HZ) | 1092 | if (newtime - ct->timeout.expires >= HZ) |
1093 | mod_timer_pending(&ct->timeout, newtime); | 1093 | mod_timer_pending(&ct->timeout, newtime); |
1094 | } | 1094 | } |
1095 | 1095 | ||
1096 | acct: | 1096 | acct: |
1097 | if (do_acct) { | 1097 | if (do_acct) { |
1098 | struct nf_conn_counter *acct; | 1098 | struct nf_conn_counter *acct; |
1099 | 1099 | ||
1100 | acct = nf_conn_acct_find(ct); | 1100 | acct = nf_conn_acct_find(ct); |
1101 | if (acct) { | 1101 | if (acct) { |
1102 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); | 1102 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); |
1103 | atomic64_add(skb->len, &acct[CTINFO2DIR(ctinfo)].bytes); | 1103 | atomic64_add(skb->len, &acct[CTINFO2DIR(ctinfo)].bytes); |
1104 | } | 1104 | } |
1105 | } | 1105 | } |
1106 | } | 1106 | } |
1107 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); | 1107 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
1108 | 1108 | ||
1109 | bool __nf_ct_kill_acct(struct nf_conn *ct, | 1109 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
1110 | enum ip_conntrack_info ctinfo, | 1110 | enum ip_conntrack_info ctinfo, |
1111 | const struct sk_buff *skb, | 1111 | const struct sk_buff *skb, |
1112 | int do_acct) | 1112 | int do_acct) |
1113 | { | 1113 | { |
1114 | if (do_acct) { | 1114 | if (do_acct) { |
1115 | struct nf_conn_counter *acct; | 1115 | struct nf_conn_counter *acct; |
1116 | 1116 | ||
1117 | acct = nf_conn_acct_find(ct); | 1117 | acct = nf_conn_acct_find(ct); |
1118 | if (acct) { | 1118 | if (acct) { |
1119 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); | 1119 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); |
1120 | atomic64_add(skb->len - skb_network_offset(skb), | 1120 | atomic64_add(skb->len - skb_network_offset(skb), |
1121 | &acct[CTINFO2DIR(ctinfo)].bytes); | 1121 | &acct[CTINFO2DIR(ctinfo)].bytes); |
1122 | } | 1122 | } |
1123 | } | 1123 | } |
1124 | 1124 | ||
1125 | if (del_timer(&ct->timeout)) { | 1125 | if (del_timer(&ct->timeout)) { |
1126 | ct->timeout.function((unsigned long)ct); | 1126 | ct->timeout.function((unsigned long)ct); |
1127 | return true; | 1127 | return true; |
1128 | } | 1128 | } |
1129 | return false; | 1129 | return false; |
1130 | } | 1130 | } |
1131 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); | 1131 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
1132 | 1132 | ||
1133 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 1133 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1134 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { | 1134 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { |
1135 | .len = sizeof(struct nf_conntrack_zone), | 1135 | .len = sizeof(struct nf_conntrack_zone), |
1136 | .align = __alignof__(struct nf_conntrack_zone), | 1136 | .align = __alignof__(struct nf_conntrack_zone), |
1137 | .id = NF_CT_EXT_ZONE, | 1137 | .id = NF_CT_EXT_ZONE, |
1138 | }; | 1138 | }; |
1139 | #endif | 1139 | #endif |
1140 | 1140 | ||
1141 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) | 1141 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
1142 | 1142 | ||
1143 | #include <linux/netfilter/nfnetlink.h> | 1143 | #include <linux/netfilter/nfnetlink.h> |
1144 | #include <linux/netfilter/nfnetlink_conntrack.h> | 1144 | #include <linux/netfilter/nfnetlink_conntrack.h> |
1145 | #include <linux/mutex.h> | 1145 | #include <linux/mutex.h> |
1146 | 1146 | ||
1147 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be | 1147 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
1148 | * in ip_conntrack_core, since we don't want the protocols to autoload | 1148 | * in ip_conntrack_core, since we don't want the protocols to autoload |
1149 | * or depend on ctnetlink */ | 1149 | * or depend on ctnetlink */ |
1150 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, | 1150 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
1151 | const struct nf_conntrack_tuple *tuple) | 1151 | const struct nf_conntrack_tuple *tuple) |
1152 | { | 1152 | { |
1153 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || | 1153 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || |
1154 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) | 1154 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) |
1155 | goto nla_put_failure; | 1155 | goto nla_put_failure; |
1156 | return 0; | 1156 | return 0; |
1157 | 1157 | ||
1158 | nla_put_failure: | 1158 | nla_put_failure: |
1159 | return -1; | 1159 | return -1; |
1160 | } | 1160 | } |
1161 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); | 1161 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
1162 | 1162 | ||
1163 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { | 1163 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
1164 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | 1164 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, |
1165 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | 1165 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, |
1166 | }; | 1166 | }; |
1167 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); | 1167 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
1168 | 1168 | ||
1169 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], | 1169 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
1170 | struct nf_conntrack_tuple *t) | 1170 | struct nf_conntrack_tuple *t) |
1171 | { | 1171 | { |
1172 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) | 1172 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
1173 | return -EINVAL; | 1173 | return -EINVAL; |
1174 | 1174 | ||
1175 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); | 1175 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
1176 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | 1176 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); |
1177 | 1177 | ||
1178 | return 0; | 1178 | return 0; |
1179 | } | 1179 | } |
1180 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); | 1180 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
1181 | 1181 | ||
1182 | int nf_ct_port_nlattr_tuple_size(void) | 1182 | int nf_ct_port_nlattr_tuple_size(void) |
1183 | { | 1183 | { |
1184 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | 1184 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); |
1185 | } | 1185 | } |
1186 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | 1186 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); |
1187 | #endif | 1187 | #endif |
1188 | 1188 | ||
1189 | /* Used by ipt_REJECT and ip6t_REJECT. */ | 1189 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
1190 | static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb) | 1190 | static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb) |
1191 | { | 1191 | { |
1192 | struct nf_conn *ct; | 1192 | struct nf_conn *ct; |
1193 | enum ip_conntrack_info ctinfo; | 1193 | enum ip_conntrack_info ctinfo; |
1194 | 1194 | ||
1195 | /* This ICMP is in reverse direction to the packet which caused it */ | 1195 | /* This ICMP is in reverse direction to the packet which caused it */ |
1196 | ct = nf_ct_get(skb, &ctinfo); | 1196 | ct = nf_ct_get(skb, &ctinfo); |
1197 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | 1197 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) |
1198 | ctinfo = IP_CT_RELATED_REPLY; | 1198 | ctinfo = IP_CT_RELATED_REPLY; |
1199 | else | 1199 | else |
1200 | ctinfo = IP_CT_RELATED; | 1200 | ctinfo = IP_CT_RELATED; |
1201 | 1201 | ||
1202 | /* Attach to new skbuff, and increment count */ | 1202 | /* Attach to new skbuff, and increment count */ |
1203 | nskb->nfct = &ct->ct_general; | 1203 | nskb->nfct = &ct->ct_general; |
1204 | nskb->nfctinfo = ctinfo; | 1204 | nskb->nfctinfo = ctinfo; |
1205 | nf_conntrack_get(nskb->nfct); | 1205 | nf_conntrack_get(nskb->nfct); |
1206 | } | 1206 | } |
1207 | 1207 | ||
1208 | /* Bring out ya dead! */ | 1208 | /* Bring out ya dead! */ |
1209 | static struct nf_conn * | 1209 | static struct nf_conn * |
1210 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), | 1210 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
1211 | void *data, unsigned int *bucket) | 1211 | void *data, unsigned int *bucket) |
1212 | { | 1212 | { |
1213 | struct nf_conntrack_tuple_hash *h; | 1213 | struct nf_conntrack_tuple_hash *h; |
1214 | struct nf_conn *ct; | 1214 | struct nf_conn *ct; |
1215 | struct hlist_nulls_node *n; | 1215 | struct hlist_nulls_node *n; |
1216 | 1216 | ||
1217 | spin_lock_bh(&nf_conntrack_lock); | 1217 | spin_lock_bh(&nf_conntrack_lock); |
1218 | for (; *bucket < net->ct.htable_size; (*bucket)++) { | 1218 | for (; *bucket < net->ct.htable_size; (*bucket)++) { |
1219 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { | 1219 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
1220 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) | 1220 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) |
1221 | continue; | 1221 | continue; |
1222 | ct = nf_ct_tuplehash_to_ctrack(h); | 1222 | ct = nf_ct_tuplehash_to_ctrack(h); |
1223 | if (iter(ct, data)) | 1223 | if (iter(ct, data)) |
1224 | goto found; | 1224 | goto found; |
1225 | } | 1225 | } |
1226 | } | 1226 | } |
1227 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { | 1227 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { |
1228 | ct = nf_ct_tuplehash_to_ctrack(h); | 1228 | ct = nf_ct_tuplehash_to_ctrack(h); |
1229 | if (iter(ct, data)) | 1229 | if (iter(ct, data)) |
1230 | set_bit(IPS_DYING_BIT, &ct->status); | 1230 | set_bit(IPS_DYING_BIT, &ct->status); |
1231 | } | 1231 | } |
1232 | spin_unlock_bh(&nf_conntrack_lock); | 1232 | spin_unlock_bh(&nf_conntrack_lock); |
1233 | return NULL; | 1233 | return NULL; |
1234 | found: | 1234 | found: |
1235 | atomic_inc(&ct->ct_general.use); | 1235 | atomic_inc(&ct->ct_general.use); |
1236 | spin_unlock_bh(&nf_conntrack_lock); | 1236 | spin_unlock_bh(&nf_conntrack_lock); |
1237 | return ct; | 1237 | return ct; |
1238 | } | 1238 | } |
1239 | 1239 | ||
1240 | void nf_ct_iterate_cleanup(struct net *net, | 1240 | void nf_ct_iterate_cleanup(struct net *net, |
1241 | int (*iter)(struct nf_conn *i, void *data), | 1241 | int (*iter)(struct nf_conn *i, void *data), |
1242 | void *data) | 1242 | void *data) |
1243 | { | 1243 | { |
1244 | struct nf_conn *ct; | 1244 | struct nf_conn *ct; |
1245 | unsigned int bucket = 0; | 1245 | unsigned int bucket = 0; |
1246 | 1246 | ||
1247 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { | 1247 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
1248 | /* Time to push up daises... */ | 1248 | /* Time to push up daises... */ |
1249 | if (del_timer(&ct->timeout)) | 1249 | if (del_timer(&ct->timeout)) |
1250 | death_by_timeout((unsigned long)ct); | 1250 | death_by_timeout((unsigned long)ct); |
1251 | /* ... else the timer will get him soon. */ | 1251 | /* ... else the timer will get him soon. */ |
1252 | 1252 | ||
1253 | nf_ct_put(ct); | 1253 | nf_ct_put(ct); |
1254 | } | 1254 | } |
1255 | } | 1255 | } |
1256 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); | 1256 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
1257 | 1257 | ||
1258 | struct __nf_ct_flush_report { | 1258 | struct __nf_ct_flush_report { |
1259 | u32 pid; | 1259 | u32 pid; |
1260 | int report; | 1260 | int report; |
1261 | }; | 1261 | }; |
1262 | 1262 | ||
1263 | static int kill_report(struct nf_conn *i, void *data) | 1263 | static int kill_report(struct nf_conn *i, void *data) |
1264 | { | 1264 | { |
1265 | struct __nf_ct_flush_report *fr = (struct __nf_ct_flush_report *)data; | 1265 | struct __nf_ct_flush_report *fr = (struct __nf_ct_flush_report *)data; |
1266 | struct nf_conn_tstamp *tstamp; | 1266 | struct nf_conn_tstamp *tstamp; |
1267 | 1267 | ||
1268 | tstamp = nf_conn_tstamp_find(i); | 1268 | tstamp = nf_conn_tstamp_find(i); |
1269 | if (tstamp && tstamp->stop == 0) | 1269 | if (tstamp && tstamp->stop == 0) |
1270 | tstamp->stop = ktime_to_ns(ktime_get_real()); | 1270 | tstamp->stop = ktime_to_ns(ktime_get_real()); |
1271 | 1271 | ||
1272 | /* If we fail to deliver the event, death_by_timeout() will retry */ | 1272 | /* If we fail to deliver the event, death_by_timeout() will retry */ |
1273 | if (nf_conntrack_event_report(IPCT_DESTROY, i, | 1273 | if (nf_conntrack_event_report(IPCT_DESTROY, i, |
1274 | fr->pid, fr->report) < 0) | 1274 | fr->pid, fr->report) < 0) |
1275 | return 1; | 1275 | return 1; |
1276 | 1276 | ||
1277 | /* Avoid the delivery of the destroy event in death_by_timeout(). */ | 1277 | /* Avoid the delivery of the destroy event in death_by_timeout(). */ |
1278 | set_bit(IPS_DYING_BIT, &i->status); | 1278 | set_bit(IPS_DYING_BIT, &i->status); |
1279 | return 1; | 1279 | return 1; |
1280 | } | 1280 | } |
1281 | 1281 | ||
1282 | static int kill_all(struct nf_conn *i, void *data) | 1282 | static int kill_all(struct nf_conn *i, void *data) |
1283 | { | 1283 | { |
1284 | return 1; | 1284 | return 1; |
1285 | } | 1285 | } |
1286 | 1286 | ||
1287 | void nf_ct_free_hashtable(void *hash, unsigned int size) | 1287 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
1288 | { | 1288 | { |
1289 | if (is_vmalloc_addr(hash)) | 1289 | if (is_vmalloc_addr(hash)) |
1290 | vfree(hash); | 1290 | vfree(hash); |
1291 | else | 1291 | else |
1292 | free_pages((unsigned long)hash, | 1292 | free_pages((unsigned long)hash, |
1293 | get_order(sizeof(struct hlist_head) * size)); | 1293 | get_order(sizeof(struct hlist_head) * size)); |
1294 | } | 1294 | } |
1295 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); | 1295 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
1296 | 1296 | ||
1297 | void nf_conntrack_flush_report(struct net *net, u32 pid, int report) | 1297 | void nf_conntrack_flush_report(struct net *net, u32 pid, int report) |
1298 | { | 1298 | { |
1299 | struct __nf_ct_flush_report fr = { | 1299 | struct __nf_ct_flush_report fr = { |
1300 | .pid = pid, | 1300 | .pid = pid, |
1301 | .report = report, | 1301 | .report = report, |
1302 | }; | 1302 | }; |
1303 | nf_ct_iterate_cleanup(net, kill_report, &fr); | 1303 | nf_ct_iterate_cleanup(net, kill_report, &fr); |
1304 | } | 1304 | } |
1305 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); | 1305 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
1306 | 1306 | ||
1307 | static void nf_ct_release_dying_list(struct net *net) | 1307 | static void nf_ct_release_dying_list(struct net *net) |
1308 | { | 1308 | { |
1309 | struct nf_conntrack_tuple_hash *h; | 1309 | struct nf_conntrack_tuple_hash *h; |
1310 | struct nf_conn *ct; | 1310 | struct nf_conn *ct; |
1311 | struct hlist_nulls_node *n; | 1311 | struct hlist_nulls_node *n; |
1312 | 1312 | ||
1313 | spin_lock_bh(&nf_conntrack_lock); | 1313 | spin_lock_bh(&nf_conntrack_lock); |
1314 | hlist_nulls_for_each_entry(h, n, &net->ct.dying, hnnode) { | 1314 | hlist_nulls_for_each_entry(h, n, &net->ct.dying, hnnode) { |
1315 | ct = nf_ct_tuplehash_to_ctrack(h); | 1315 | ct = nf_ct_tuplehash_to_ctrack(h); |
1316 | /* never fails to remove them, no listeners at this point */ | 1316 | /* never fails to remove them, no listeners at this point */ |
1317 | nf_ct_kill(ct); | 1317 | nf_ct_kill(ct); |
1318 | } | 1318 | } |
1319 | spin_unlock_bh(&nf_conntrack_lock); | 1319 | spin_unlock_bh(&nf_conntrack_lock); |
1320 | } | 1320 | } |
1321 | 1321 | ||
1322 | static int untrack_refs(void) | 1322 | static int untrack_refs(void) |
1323 | { | 1323 | { |
1324 | int cnt = 0, cpu; | 1324 | int cnt = 0, cpu; |
1325 | 1325 | ||
1326 | for_each_possible_cpu(cpu) { | 1326 | for_each_possible_cpu(cpu) { |
1327 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | 1327 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); |
1328 | 1328 | ||
1329 | cnt += atomic_read(&ct->ct_general.use) - 1; | 1329 | cnt += atomic_read(&ct->ct_general.use) - 1; |
1330 | } | 1330 | } |
1331 | return cnt; | 1331 | return cnt; |
1332 | } | 1332 | } |
1333 | 1333 | ||
1334 | static void nf_conntrack_cleanup_init_net(void) | 1334 | static void nf_conntrack_cleanup_init_net(void) |
1335 | { | 1335 | { |
1336 | while (untrack_refs() > 0) | 1336 | while (untrack_refs() > 0) |
1337 | schedule(); | 1337 | schedule(); |
1338 | 1338 | ||
1339 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 1339 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1340 | nf_ct_extend_unregister(&nf_ct_zone_extend); | 1340 | nf_ct_extend_unregister(&nf_ct_zone_extend); |
1341 | #endif | 1341 | #endif |
1342 | } | 1342 | } |
1343 | 1343 | ||
1344 | static void nf_conntrack_cleanup_net(struct net *net) | 1344 | static void nf_conntrack_cleanup_net(struct net *net) |
1345 | { | 1345 | { |
1346 | i_see_dead_people: | 1346 | i_see_dead_people: |
1347 | nf_ct_iterate_cleanup(net, kill_all, NULL); | 1347 | nf_ct_iterate_cleanup(net, kill_all, NULL); |
1348 | nf_ct_release_dying_list(net); | 1348 | nf_ct_release_dying_list(net); |
1349 | if (atomic_read(&net->ct.count) != 0) { | 1349 | if (atomic_read(&net->ct.count) != 0) { |
1350 | schedule(); | 1350 | schedule(); |
1351 | goto i_see_dead_people; | 1351 | goto i_see_dead_people; |
1352 | } | 1352 | } |
1353 | 1353 | ||
1354 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); | 1354 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
1355 | nf_conntrack_helper_fini(net); | 1355 | nf_conntrack_helper_fini(net); |
1356 | nf_conntrack_timeout_fini(net); | 1356 | nf_conntrack_timeout_fini(net); |
1357 | nf_conntrack_ecache_fini(net); | 1357 | nf_conntrack_ecache_fini(net); |
1358 | nf_conntrack_tstamp_fini(net); | 1358 | nf_conntrack_tstamp_fini(net); |
1359 | nf_conntrack_acct_fini(net); | 1359 | nf_conntrack_acct_fini(net); |
1360 | nf_conntrack_expect_fini(net); | 1360 | nf_conntrack_expect_fini(net); |
1361 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); | 1361 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1362 | kfree(net->ct.slabname); | 1362 | kfree(net->ct.slabname); |
1363 | free_percpu(net->ct.stat); | 1363 | free_percpu(net->ct.stat); |
1364 | } | 1364 | } |
1365 | 1365 | ||
1366 | /* Mishearing the voices in his head, our hero wonders how he's | 1366 | /* Mishearing the voices in his head, our hero wonders how he's |
1367 | supposed to kill the mall. */ | 1367 | supposed to kill the mall. */ |
1368 | void nf_conntrack_cleanup(struct net *net) | 1368 | void nf_conntrack_cleanup(struct net *net) |
1369 | { | 1369 | { |
1370 | if (net_eq(net, &init_net)) | 1370 | if (net_eq(net, &init_net)) |
1371 | RCU_INIT_POINTER(ip_ct_attach, NULL); | 1371 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
1372 | 1372 | ||
1373 | /* This makes sure all current packets have passed through | 1373 | /* This makes sure all current packets have passed through |
1374 | netfilter framework. Roll on, two-stage module | 1374 | netfilter framework. Roll on, two-stage module |
1375 | delete... */ | 1375 | delete... */ |
1376 | synchronize_net(); | 1376 | synchronize_net(); |
1377 | nf_conntrack_proto_fini(net); | 1377 | nf_conntrack_proto_fini(net); |
1378 | nf_conntrack_cleanup_net(net); | 1378 | nf_conntrack_cleanup_net(net); |
1379 | } | ||
1379 | 1380 | ||
1380 | if (net_eq(net, &init_net)) { | 1381 | void nf_conntrack_cleanup_end(void) |
1381 | RCU_INIT_POINTER(nf_ct_destroy, NULL); | 1382 | { |
1382 | nf_conntrack_cleanup_init_net(); | 1383 | RCU_INIT_POINTER(nf_ct_destroy, NULL); |
1383 | } | 1384 | nf_conntrack_cleanup_init_net(); |
1384 | } | 1385 | } |
1385 | 1386 | ||
1386 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) | 1387 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
1387 | { | 1388 | { |
1388 | struct hlist_nulls_head *hash; | 1389 | struct hlist_nulls_head *hash; |
1389 | unsigned int nr_slots, i; | 1390 | unsigned int nr_slots, i; |
1390 | size_t sz; | 1391 | size_t sz; |
1391 | 1392 | ||
1392 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); | 1393 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1393 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | 1394 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); |
1394 | sz = nr_slots * sizeof(struct hlist_nulls_head); | 1395 | sz = nr_slots * sizeof(struct hlist_nulls_head); |
1395 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | 1396 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, |
1396 | get_order(sz)); | 1397 | get_order(sz)); |
1397 | if (!hash) { | 1398 | if (!hash) { |
1398 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); | 1399 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); |
1399 | hash = vzalloc(sz); | 1400 | hash = vzalloc(sz); |
1400 | } | 1401 | } |
1401 | 1402 | ||
1402 | if (hash && nulls) | 1403 | if (hash && nulls) |
1403 | for (i = 0; i < nr_slots; i++) | 1404 | for (i = 0; i < nr_slots; i++) |
1404 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | 1405 | INIT_HLIST_NULLS_HEAD(&hash[i], i); |
1405 | 1406 | ||
1406 | return hash; | 1407 | return hash; |
1407 | } | 1408 | } |
1408 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); | 1409 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
1409 | 1410 | ||
1410 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) | 1411 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
1411 | { | 1412 | { |
1412 | int i, bucket, rc; | 1413 | int i, bucket, rc; |
1413 | unsigned int hashsize, old_size; | 1414 | unsigned int hashsize, old_size; |
1414 | struct hlist_nulls_head *hash, *old_hash; | 1415 | struct hlist_nulls_head *hash, *old_hash; |
1415 | struct nf_conntrack_tuple_hash *h; | 1416 | struct nf_conntrack_tuple_hash *h; |
1416 | struct nf_conn *ct; | 1417 | struct nf_conn *ct; |
1417 | 1418 | ||
1418 | if (current->nsproxy->net_ns != &init_net) | 1419 | if (current->nsproxy->net_ns != &init_net) |
1419 | return -EOPNOTSUPP; | 1420 | return -EOPNOTSUPP; |
1420 | 1421 | ||
1421 | /* On boot, we can set this without any fancy locking. */ | 1422 | /* On boot, we can set this without any fancy locking. */ |
1422 | if (!nf_conntrack_htable_size) | 1423 | if (!nf_conntrack_htable_size) |
1423 | return param_set_uint(val, kp); | 1424 | return param_set_uint(val, kp); |
1424 | 1425 | ||
1425 | rc = kstrtouint(val, 0, &hashsize); | 1426 | rc = kstrtouint(val, 0, &hashsize); |
1426 | if (rc) | 1427 | if (rc) |
1427 | return rc; | 1428 | return rc; |
1428 | if (!hashsize) | 1429 | if (!hashsize) |
1429 | return -EINVAL; | 1430 | return -EINVAL; |
1430 | 1431 | ||
1431 | hash = nf_ct_alloc_hashtable(&hashsize, 1); | 1432 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
1432 | if (!hash) | 1433 | if (!hash) |
1433 | return -ENOMEM; | 1434 | return -ENOMEM; |
1434 | 1435 | ||
1435 | /* Lookups in the old hash might happen in parallel, which means we | 1436 | /* Lookups in the old hash might happen in parallel, which means we |
1436 | * might get false negatives during connection lookup. New connections | 1437 | * might get false negatives during connection lookup. New connections |
1437 | * created because of a false negative won't make it into the hash | 1438 | * created because of a false negative won't make it into the hash |
1438 | * though since that required taking the lock. | 1439 | * though since that required taking the lock. |
1439 | */ | 1440 | */ |
1440 | spin_lock_bh(&nf_conntrack_lock); | 1441 | spin_lock_bh(&nf_conntrack_lock); |
1441 | for (i = 0; i < init_net.ct.htable_size; i++) { | 1442 | for (i = 0; i < init_net.ct.htable_size; i++) { |
1442 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { | 1443 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1443 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | 1444 | h = hlist_nulls_entry(init_net.ct.hash[i].first, |
1444 | struct nf_conntrack_tuple_hash, hnnode); | 1445 | struct nf_conntrack_tuple_hash, hnnode); |
1445 | ct = nf_ct_tuplehash_to_ctrack(h); | 1446 | ct = nf_ct_tuplehash_to_ctrack(h); |
1446 | hlist_nulls_del_rcu(&h->hnnode); | 1447 | hlist_nulls_del_rcu(&h->hnnode); |
1447 | bucket = __hash_conntrack(&h->tuple, nf_ct_zone(ct), | 1448 | bucket = __hash_conntrack(&h->tuple, nf_ct_zone(ct), |
1448 | hashsize); | 1449 | hashsize); |
1449 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); | 1450 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
1450 | } | 1451 | } |
1451 | } | 1452 | } |
1452 | old_size = init_net.ct.htable_size; | 1453 | old_size = init_net.ct.htable_size; |
1453 | old_hash = init_net.ct.hash; | 1454 | old_hash = init_net.ct.hash; |
1454 | 1455 | ||
1455 | init_net.ct.htable_size = nf_conntrack_htable_size = hashsize; | 1456 | init_net.ct.htable_size = nf_conntrack_htable_size = hashsize; |
1456 | init_net.ct.hash = hash; | 1457 | init_net.ct.hash = hash; |
1457 | spin_unlock_bh(&nf_conntrack_lock); | 1458 | spin_unlock_bh(&nf_conntrack_lock); |
1458 | 1459 | ||
1459 | nf_ct_free_hashtable(old_hash, old_size); | 1460 | nf_ct_free_hashtable(old_hash, old_size); |
1460 | return 0; | 1461 | return 0; |
1461 | } | 1462 | } |
1462 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); | 1463 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
1463 | 1464 | ||
1464 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, | 1465 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
1465 | &nf_conntrack_htable_size, 0600); | 1466 | &nf_conntrack_htable_size, 0600); |
1466 | 1467 | ||
1467 | void nf_ct_untracked_status_or(unsigned long bits) | 1468 | void nf_ct_untracked_status_or(unsigned long bits) |
1468 | { | 1469 | { |
1469 | int cpu; | 1470 | int cpu; |
1470 | 1471 | ||
1471 | for_each_possible_cpu(cpu) | 1472 | for_each_possible_cpu(cpu) |
1472 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; | 1473 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; |
1473 | } | 1474 | } |
1474 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); | 1475 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); |
1475 | 1476 | ||
1476 | static int nf_conntrack_init_init_net(void) | 1477 | static int nf_conntrack_init_init_net(void) |
1477 | { | 1478 | { |
1478 | int max_factor = 8; | 1479 | int max_factor = 8; |
1479 | int ret, cpu; | 1480 | int ret, cpu; |
1480 | 1481 | ||
1481 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB | 1482 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB |
1482 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ | 1483 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
1483 | if (!nf_conntrack_htable_size) { | 1484 | if (!nf_conntrack_htable_size) { |
1484 | nf_conntrack_htable_size | 1485 | nf_conntrack_htable_size |
1485 | = (((totalram_pages << PAGE_SHIFT) / 16384) | 1486 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
1486 | / sizeof(struct hlist_head)); | 1487 | / sizeof(struct hlist_head)); |
1487 | if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) | 1488 | if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
1488 | nf_conntrack_htable_size = 16384; | 1489 | nf_conntrack_htable_size = 16384; |
1489 | if (nf_conntrack_htable_size < 32) | 1490 | if (nf_conntrack_htable_size < 32) |
1490 | nf_conntrack_htable_size = 32; | 1491 | nf_conntrack_htable_size = 32; |
1491 | 1492 | ||
1492 | /* Use a max. factor of four by default to get the same max as | 1493 | /* Use a max. factor of four by default to get the same max as |
1493 | * with the old struct list_heads. When a table size is given | 1494 | * with the old struct list_heads. When a table size is given |
1494 | * we use the old value of 8 to avoid reducing the max. | 1495 | * we use the old value of 8 to avoid reducing the max. |
1495 | * entries. */ | 1496 | * entries. */ |
1496 | max_factor = 4; | 1497 | max_factor = 4; |
1497 | } | 1498 | } |
1498 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; | 1499 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
1499 | 1500 | ||
1500 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", | 1501 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
1501 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, | 1502 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
1502 | nf_conntrack_max); | 1503 | nf_conntrack_max); |
1503 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 1504 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1504 | ret = nf_ct_extend_register(&nf_ct_zone_extend); | 1505 | ret = nf_ct_extend_register(&nf_ct_zone_extend); |
1505 | if (ret < 0) | 1506 | if (ret < 0) |
1506 | goto err_extend; | 1507 | goto err_extend; |
1507 | #endif | 1508 | #endif |
1508 | /* Set up fake conntrack: to never be deleted, not in any hashes */ | 1509 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
1509 | for_each_possible_cpu(cpu) { | 1510 | for_each_possible_cpu(cpu) { |
1510 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | 1511 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); |
1511 | write_pnet(&ct->ct_net, &init_net); | 1512 | write_pnet(&ct->ct_net, &init_net); |
1512 | atomic_set(&ct->ct_general.use, 1); | 1513 | atomic_set(&ct->ct_general.use, 1); |
1513 | } | 1514 | } |
1514 | /* - and look it like as a confirmed connection */ | 1515 | /* - and look it like as a confirmed connection */ |
1515 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); | 1516 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
1516 | return 0; | 1517 | return 0; |
1517 | 1518 | ||
1518 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 1519 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1519 | err_extend: | 1520 | err_extend: |
1520 | #endif | 1521 | #endif |
1521 | return ret; | 1522 | return ret; |
1522 | } | 1523 | } |
1523 | 1524 | ||
1524 | /* | 1525 | /* |
1525 | * We need to use special "null" values, not used in hash table | 1526 | * We need to use special "null" values, not used in hash table |
1526 | */ | 1527 | */ |
1527 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | 1528 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) |
1528 | #define DYING_NULLS_VAL ((1<<30)+1) | 1529 | #define DYING_NULLS_VAL ((1<<30)+1) |
1529 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) | 1530 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) |
1530 | 1531 | ||
1531 | static int nf_conntrack_init_net(struct net *net) | 1532 | static int nf_conntrack_init_net(struct net *net) |
1532 | { | 1533 | { |
1533 | int ret; | 1534 | int ret; |
1534 | 1535 | ||
1535 | atomic_set(&net->ct.count, 0); | 1536 | atomic_set(&net->ct.count, 0); |
1536 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, UNCONFIRMED_NULLS_VAL); | 1537 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, UNCONFIRMED_NULLS_VAL); |
1537 | INIT_HLIST_NULLS_HEAD(&net->ct.dying, DYING_NULLS_VAL); | 1538 | INIT_HLIST_NULLS_HEAD(&net->ct.dying, DYING_NULLS_VAL); |
1538 | INIT_HLIST_NULLS_HEAD(&net->ct.tmpl, TEMPLATE_NULLS_VAL); | 1539 | INIT_HLIST_NULLS_HEAD(&net->ct.tmpl, TEMPLATE_NULLS_VAL); |
1539 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); | 1540 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1540 | if (!net->ct.stat) { | 1541 | if (!net->ct.stat) { |
1541 | ret = -ENOMEM; | 1542 | ret = -ENOMEM; |
1542 | goto err_stat; | 1543 | goto err_stat; |
1543 | } | 1544 | } |
1544 | 1545 | ||
1545 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); | 1546 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); |
1546 | if (!net->ct.slabname) { | 1547 | if (!net->ct.slabname) { |
1547 | ret = -ENOMEM; | 1548 | ret = -ENOMEM; |
1548 | goto err_slabname; | 1549 | goto err_slabname; |
1549 | } | 1550 | } |
1550 | 1551 | ||
1551 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, | 1552 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, |
1552 | sizeof(struct nf_conn), 0, | 1553 | sizeof(struct nf_conn), 0, |
1553 | SLAB_DESTROY_BY_RCU, NULL); | 1554 | SLAB_DESTROY_BY_RCU, NULL); |
1554 | if (!net->ct.nf_conntrack_cachep) { | 1555 | if (!net->ct.nf_conntrack_cachep) { |
1555 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); | 1556 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); |
1556 | ret = -ENOMEM; | 1557 | ret = -ENOMEM; |
1557 | goto err_cache; | 1558 | goto err_cache; |
1558 | } | 1559 | } |
1559 | 1560 | ||
1560 | net->ct.htable_size = nf_conntrack_htable_size; | 1561 | net->ct.htable_size = nf_conntrack_htable_size; |
1561 | net->ct.hash = nf_ct_alloc_hashtable(&net->ct.htable_size, 1); | 1562 | net->ct.hash = nf_ct_alloc_hashtable(&net->ct.htable_size, 1); |
1562 | if (!net->ct.hash) { | 1563 | if (!net->ct.hash) { |
1563 | ret = -ENOMEM; | 1564 | ret = -ENOMEM; |
1564 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); | 1565 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); |
1565 | goto err_hash; | 1566 | goto err_hash; |
1566 | } | 1567 | } |
1567 | ret = nf_conntrack_expect_init(net); | 1568 | ret = nf_conntrack_expect_init(net); |
1568 | if (ret < 0) | 1569 | if (ret < 0) |
1569 | goto err_expect; | 1570 | goto err_expect; |
1570 | ret = nf_conntrack_acct_init(net); | 1571 | ret = nf_conntrack_acct_init(net); |
1571 | if (ret < 0) | 1572 | if (ret < 0) |
1572 | goto err_acct; | 1573 | goto err_acct; |
1573 | ret = nf_conntrack_tstamp_init(net); | 1574 | ret = nf_conntrack_tstamp_init(net); |
1574 | if (ret < 0) | 1575 | if (ret < 0) |
1575 | goto err_tstamp; | 1576 | goto err_tstamp; |
1576 | ret = nf_conntrack_ecache_init(net); | 1577 | ret = nf_conntrack_ecache_init(net); |
1577 | if (ret < 0) | 1578 | if (ret < 0) |
1578 | goto err_ecache; | 1579 | goto err_ecache; |
1579 | ret = nf_conntrack_timeout_init(net); | 1580 | ret = nf_conntrack_timeout_init(net); |
1580 | if (ret < 0) | 1581 | if (ret < 0) |
1581 | goto err_timeout; | 1582 | goto err_timeout; |
1582 | ret = nf_conntrack_helper_init(net); | 1583 | ret = nf_conntrack_helper_init(net); |
1583 | if (ret < 0) | 1584 | if (ret < 0) |
1584 | goto err_helper; | 1585 | goto err_helper; |
1585 | return 0; | 1586 | return 0; |
1586 | err_helper: | 1587 | err_helper: |
1587 | nf_conntrack_timeout_fini(net); | 1588 | nf_conntrack_timeout_fini(net); |
1588 | err_timeout: | 1589 | err_timeout: |
1589 | nf_conntrack_ecache_fini(net); | 1590 | nf_conntrack_ecache_fini(net); |
1590 | err_ecache: | 1591 | err_ecache: |
1591 | nf_conntrack_tstamp_fini(net); | 1592 | nf_conntrack_tstamp_fini(net); |
1592 | err_tstamp: | 1593 | err_tstamp: |
1593 | nf_conntrack_acct_fini(net); | 1594 | nf_conntrack_acct_fini(net); |
1594 | err_acct: | 1595 | err_acct: |
1595 | nf_conntrack_expect_fini(net); | 1596 | nf_conntrack_expect_fini(net); |
1596 | err_expect: | 1597 | err_expect: |
1597 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); | 1598 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
1598 | err_hash: | 1599 | err_hash: |
1599 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); | 1600 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1600 | err_cache: | 1601 | err_cache: |
1601 | kfree(net->ct.slabname); | 1602 | kfree(net->ct.slabname); |
1602 | err_slabname: | 1603 | err_slabname: |
1603 | free_percpu(net->ct.stat); | 1604 | free_percpu(net->ct.stat); |
1604 | err_stat: | 1605 | err_stat: |
1605 | return ret; | 1606 | return ret; |
1606 | } | 1607 | } |
1607 | 1608 | ||
1608 | s16 (*nf_ct_nat_offset)(const struct nf_conn *ct, | 1609 | s16 (*nf_ct_nat_offset)(const struct nf_conn *ct, |
1609 | enum ip_conntrack_dir dir, | 1610 | enum ip_conntrack_dir dir, |
1610 | u32 seq); | 1611 | u32 seq); |
1611 | EXPORT_SYMBOL_GPL(nf_ct_nat_offset); | 1612 | EXPORT_SYMBOL_GPL(nf_ct_nat_offset); |
1612 | 1613 | ||
1613 | int nf_conntrack_init(struct net *net) | 1614 | int nf_conntrack_init(struct net *net) |
1614 | { | 1615 | { |
1615 | int ret; | 1616 | int ret; |
1616 | 1617 | ||
1617 | if (net_eq(net, &init_net)) { | 1618 | if (net_eq(net, &init_net)) { |
1618 | ret = nf_conntrack_init_init_net(); | 1619 | ret = nf_conntrack_init_init_net(); |
1619 | if (ret < 0) | 1620 | if (ret < 0) |
1620 | goto out_init_net; | 1621 | goto out_init_net; |
1621 | } | 1622 | } |
1622 | ret = nf_conntrack_proto_init(net); | 1623 | ret = nf_conntrack_proto_init(net); |
1623 | if (ret < 0) | 1624 | if (ret < 0) |
1624 | goto out_proto; | 1625 | goto out_proto; |
1625 | ret = nf_conntrack_init_net(net); | 1626 | ret = nf_conntrack_init_net(net); |
1626 | if (ret < 0) | 1627 | if (ret < 0) |
1627 | goto out_net; | 1628 | goto out_net; |
1628 | 1629 | ||
1629 | if (net_eq(net, &init_net)) { | 1630 | if (net_eq(net, &init_net)) { |
1630 | /* For use by REJECT target */ | 1631 | /* For use by REJECT target */ |
1631 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); | 1632 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); |
1632 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); | 1633 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); |
1633 | 1634 | ||
1634 | /* Howto get NAT offsets */ | 1635 | /* Howto get NAT offsets */ |
1635 | RCU_INIT_POINTER(nf_ct_nat_offset, NULL); | 1636 | RCU_INIT_POINTER(nf_ct_nat_offset, NULL); |
1636 | } | 1637 | } |
1637 | return 0; | 1638 | return 0; |
1638 | 1639 | ||
1639 | out_net: | 1640 | out_net: |
1640 | nf_conntrack_proto_fini(net); | 1641 | nf_conntrack_proto_fini(net); |
1641 | out_proto: | 1642 | out_proto: |
1642 | if (net_eq(net, &init_net)) | 1643 | if (net_eq(net, &init_net)) |
1643 | nf_conntrack_cleanup_init_net(); | 1644 | nf_conntrack_cleanup_init_net(); |
1644 | out_init_net: | 1645 | out_init_net: |
1645 | return ret; | 1646 | return ret; |
1646 | } | 1647 | } |
1647 | 1648 |
net/netfilter/nf_conntrack_standalone.c
1 | /* (C) 1999-2001 Paul `Rusty' Russell | 1 | /* (C) 1999-2001 Paul `Rusty' Russell |
2 | * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org> | 2 | * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org> |
3 | * | 3 | * |
4 | * This program is free software; you can redistribute it and/or modify | 4 | * This program is free software; you can redistribute it and/or modify |
5 | * it under the terms of the GNU General Public License version 2 as | 5 | * it under the terms of the GNU General Public License version 2 as |
6 | * published by the Free Software Foundation. | 6 | * published by the Free Software Foundation. |
7 | */ | 7 | */ |
8 | 8 | ||
9 | #include <linux/types.h> | 9 | #include <linux/types.h> |
10 | #include <linux/netfilter.h> | 10 | #include <linux/netfilter.h> |
11 | #include <linux/slab.h> | 11 | #include <linux/slab.h> |
12 | #include <linux/module.h> | 12 | #include <linux/module.h> |
13 | #include <linux/skbuff.h> | 13 | #include <linux/skbuff.h> |
14 | #include <linux/proc_fs.h> | 14 | #include <linux/proc_fs.h> |
15 | #include <linux/seq_file.h> | 15 | #include <linux/seq_file.h> |
16 | #include <linux/percpu.h> | 16 | #include <linux/percpu.h> |
17 | #include <linux/netdevice.h> | 17 | #include <linux/netdevice.h> |
18 | #include <linux/security.h> | 18 | #include <linux/security.h> |
19 | #include <net/net_namespace.h> | 19 | #include <net/net_namespace.h> |
20 | #ifdef CONFIG_SYSCTL | 20 | #ifdef CONFIG_SYSCTL |
21 | #include <linux/sysctl.h> | 21 | #include <linux/sysctl.h> |
22 | #endif | 22 | #endif |
23 | 23 | ||
24 | #include <net/netfilter/nf_conntrack.h> | 24 | #include <net/netfilter/nf_conntrack.h> |
25 | #include <net/netfilter/nf_conntrack_core.h> | 25 | #include <net/netfilter/nf_conntrack_core.h> |
26 | #include <net/netfilter/nf_conntrack_l3proto.h> | 26 | #include <net/netfilter/nf_conntrack_l3proto.h> |
27 | #include <net/netfilter/nf_conntrack_l4proto.h> | 27 | #include <net/netfilter/nf_conntrack_l4proto.h> |
28 | #include <net/netfilter/nf_conntrack_expect.h> | 28 | #include <net/netfilter/nf_conntrack_expect.h> |
29 | #include <net/netfilter/nf_conntrack_helper.h> | 29 | #include <net/netfilter/nf_conntrack_helper.h> |
30 | #include <net/netfilter/nf_conntrack_acct.h> | 30 | #include <net/netfilter/nf_conntrack_acct.h> |
31 | #include <net/netfilter/nf_conntrack_zones.h> | 31 | #include <net/netfilter/nf_conntrack_zones.h> |
32 | #include <net/netfilter/nf_conntrack_timestamp.h> | 32 | #include <net/netfilter/nf_conntrack_timestamp.h> |
33 | #include <linux/rculist_nulls.h> | 33 | #include <linux/rculist_nulls.h> |
34 | 34 | ||
35 | MODULE_LICENSE("GPL"); | 35 | MODULE_LICENSE("GPL"); |
36 | 36 | ||
37 | #ifdef CONFIG_NF_CONNTRACK_PROCFS | 37 | #ifdef CONFIG_NF_CONNTRACK_PROCFS |
38 | int | 38 | int |
39 | print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple, | 39 | print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple, |
40 | const struct nf_conntrack_l3proto *l3proto, | 40 | const struct nf_conntrack_l3proto *l3proto, |
41 | const struct nf_conntrack_l4proto *l4proto) | 41 | const struct nf_conntrack_l4proto *l4proto) |
42 | { | 42 | { |
43 | return l3proto->print_tuple(s, tuple) || l4proto->print_tuple(s, tuple); | 43 | return l3proto->print_tuple(s, tuple) || l4proto->print_tuple(s, tuple); |
44 | } | 44 | } |
45 | EXPORT_SYMBOL_GPL(print_tuple); | 45 | EXPORT_SYMBOL_GPL(print_tuple); |
46 | 46 | ||
47 | struct ct_iter_state { | 47 | struct ct_iter_state { |
48 | struct seq_net_private p; | 48 | struct seq_net_private p; |
49 | unsigned int bucket; | 49 | unsigned int bucket; |
50 | u_int64_t time_now; | 50 | u_int64_t time_now; |
51 | }; | 51 | }; |
52 | 52 | ||
53 | static struct hlist_nulls_node *ct_get_first(struct seq_file *seq) | 53 | static struct hlist_nulls_node *ct_get_first(struct seq_file *seq) |
54 | { | 54 | { |
55 | struct net *net = seq_file_net(seq); | 55 | struct net *net = seq_file_net(seq); |
56 | struct ct_iter_state *st = seq->private; | 56 | struct ct_iter_state *st = seq->private; |
57 | struct hlist_nulls_node *n; | 57 | struct hlist_nulls_node *n; |
58 | 58 | ||
59 | for (st->bucket = 0; | 59 | for (st->bucket = 0; |
60 | st->bucket < net->ct.htable_size; | 60 | st->bucket < net->ct.htable_size; |
61 | st->bucket++) { | 61 | st->bucket++) { |
62 | n = rcu_dereference(hlist_nulls_first_rcu(&net->ct.hash[st->bucket])); | 62 | n = rcu_dereference(hlist_nulls_first_rcu(&net->ct.hash[st->bucket])); |
63 | if (!is_a_nulls(n)) | 63 | if (!is_a_nulls(n)) |
64 | return n; | 64 | return n; |
65 | } | 65 | } |
66 | return NULL; | 66 | return NULL; |
67 | } | 67 | } |
68 | 68 | ||
69 | static struct hlist_nulls_node *ct_get_next(struct seq_file *seq, | 69 | static struct hlist_nulls_node *ct_get_next(struct seq_file *seq, |
70 | struct hlist_nulls_node *head) | 70 | struct hlist_nulls_node *head) |
71 | { | 71 | { |
72 | struct net *net = seq_file_net(seq); | 72 | struct net *net = seq_file_net(seq); |
73 | struct ct_iter_state *st = seq->private; | 73 | struct ct_iter_state *st = seq->private; |
74 | 74 | ||
75 | head = rcu_dereference(hlist_nulls_next_rcu(head)); | 75 | head = rcu_dereference(hlist_nulls_next_rcu(head)); |
76 | while (is_a_nulls(head)) { | 76 | while (is_a_nulls(head)) { |
77 | if (likely(get_nulls_value(head) == st->bucket)) { | 77 | if (likely(get_nulls_value(head) == st->bucket)) { |
78 | if (++st->bucket >= net->ct.htable_size) | 78 | if (++st->bucket >= net->ct.htable_size) |
79 | return NULL; | 79 | return NULL; |
80 | } | 80 | } |
81 | head = rcu_dereference( | 81 | head = rcu_dereference( |
82 | hlist_nulls_first_rcu( | 82 | hlist_nulls_first_rcu( |
83 | &net->ct.hash[st->bucket])); | 83 | &net->ct.hash[st->bucket])); |
84 | } | 84 | } |
85 | return head; | 85 | return head; |
86 | } | 86 | } |
87 | 87 | ||
88 | static struct hlist_nulls_node *ct_get_idx(struct seq_file *seq, loff_t pos) | 88 | static struct hlist_nulls_node *ct_get_idx(struct seq_file *seq, loff_t pos) |
89 | { | 89 | { |
90 | struct hlist_nulls_node *head = ct_get_first(seq); | 90 | struct hlist_nulls_node *head = ct_get_first(seq); |
91 | 91 | ||
92 | if (head) | 92 | if (head) |
93 | while (pos && (head = ct_get_next(seq, head))) | 93 | while (pos && (head = ct_get_next(seq, head))) |
94 | pos--; | 94 | pos--; |
95 | return pos ? NULL : head; | 95 | return pos ? NULL : head; |
96 | } | 96 | } |
97 | 97 | ||
98 | static void *ct_seq_start(struct seq_file *seq, loff_t *pos) | 98 | static void *ct_seq_start(struct seq_file *seq, loff_t *pos) |
99 | __acquires(RCU) | 99 | __acquires(RCU) |
100 | { | 100 | { |
101 | struct ct_iter_state *st = seq->private; | 101 | struct ct_iter_state *st = seq->private; |
102 | 102 | ||
103 | st->time_now = ktime_to_ns(ktime_get_real()); | 103 | st->time_now = ktime_to_ns(ktime_get_real()); |
104 | rcu_read_lock(); | 104 | rcu_read_lock(); |
105 | return ct_get_idx(seq, *pos); | 105 | return ct_get_idx(seq, *pos); |
106 | } | 106 | } |
107 | 107 | ||
108 | static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos) | 108 | static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos) |
109 | { | 109 | { |
110 | (*pos)++; | 110 | (*pos)++; |
111 | return ct_get_next(s, v); | 111 | return ct_get_next(s, v); |
112 | } | 112 | } |
113 | 113 | ||
114 | static void ct_seq_stop(struct seq_file *s, void *v) | 114 | static void ct_seq_stop(struct seq_file *s, void *v) |
115 | __releases(RCU) | 115 | __releases(RCU) |
116 | { | 116 | { |
117 | rcu_read_unlock(); | 117 | rcu_read_unlock(); |
118 | } | 118 | } |
119 | 119 | ||
120 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | 120 | #ifdef CONFIG_NF_CONNTRACK_SECMARK |
121 | static int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct) | 121 | static int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct) |
122 | { | 122 | { |
123 | int ret; | 123 | int ret; |
124 | u32 len; | 124 | u32 len; |
125 | char *secctx; | 125 | char *secctx; |
126 | 126 | ||
127 | ret = security_secid_to_secctx(ct->secmark, &secctx, &len); | 127 | ret = security_secid_to_secctx(ct->secmark, &secctx, &len); |
128 | if (ret) | 128 | if (ret) |
129 | return 0; | 129 | return 0; |
130 | 130 | ||
131 | ret = seq_printf(s, "secctx=%s ", secctx); | 131 | ret = seq_printf(s, "secctx=%s ", secctx); |
132 | 132 | ||
133 | security_release_secctx(secctx, len); | 133 | security_release_secctx(secctx, len); |
134 | return ret; | 134 | return ret; |
135 | } | 135 | } |
136 | #else | 136 | #else |
137 | static inline int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct) | 137 | static inline int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct) |
138 | { | 138 | { |
139 | return 0; | 139 | return 0; |
140 | } | 140 | } |
141 | #endif | 141 | #endif |
142 | 142 | ||
143 | #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP | 143 | #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP |
144 | static int ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct) | 144 | static int ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct) |
145 | { | 145 | { |
146 | struct ct_iter_state *st = s->private; | 146 | struct ct_iter_state *st = s->private; |
147 | struct nf_conn_tstamp *tstamp; | 147 | struct nf_conn_tstamp *tstamp; |
148 | s64 delta_time; | 148 | s64 delta_time; |
149 | 149 | ||
150 | tstamp = nf_conn_tstamp_find(ct); | 150 | tstamp = nf_conn_tstamp_find(ct); |
151 | if (tstamp) { | 151 | if (tstamp) { |
152 | delta_time = st->time_now - tstamp->start; | 152 | delta_time = st->time_now - tstamp->start; |
153 | if (delta_time > 0) | 153 | if (delta_time > 0) |
154 | delta_time = div_s64(delta_time, NSEC_PER_SEC); | 154 | delta_time = div_s64(delta_time, NSEC_PER_SEC); |
155 | else | 155 | else |
156 | delta_time = 0; | 156 | delta_time = 0; |
157 | 157 | ||
158 | return seq_printf(s, "delta-time=%llu ", | 158 | return seq_printf(s, "delta-time=%llu ", |
159 | (unsigned long long)delta_time); | 159 | (unsigned long long)delta_time); |
160 | } | 160 | } |
161 | return 0; | 161 | return 0; |
162 | } | 162 | } |
163 | #else | 163 | #else |
164 | static inline int | 164 | static inline int |
165 | ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct) | 165 | ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct) |
166 | { | 166 | { |
167 | return 0; | 167 | return 0; |
168 | } | 168 | } |
169 | #endif | 169 | #endif |
170 | 170 | ||
171 | /* return 0 on success, 1 in case of error */ | 171 | /* return 0 on success, 1 in case of error */ |
172 | static int ct_seq_show(struct seq_file *s, void *v) | 172 | static int ct_seq_show(struct seq_file *s, void *v) |
173 | { | 173 | { |
174 | struct nf_conntrack_tuple_hash *hash = v; | 174 | struct nf_conntrack_tuple_hash *hash = v; |
175 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash); | 175 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash); |
176 | const struct nf_conntrack_l3proto *l3proto; | 176 | const struct nf_conntrack_l3proto *l3proto; |
177 | const struct nf_conntrack_l4proto *l4proto; | 177 | const struct nf_conntrack_l4proto *l4proto; |
178 | int ret = 0; | 178 | int ret = 0; |
179 | 179 | ||
180 | NF_CT_ASSERT(ct); | 180 | NF_CT_ASSERT(ct); |
181 | if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use))) | 181 | if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use))) |
182 | return 0; | 182 | return 0; |
183 | 183 | ||
184 | /* we only want to print DIR_ORIGINAL */ | 184 | /* we only want to print DIR_ORIGINAL */ |
185 | if (NF_CT_DIRECTION(hash)) | 185 | if (NF_CT_DIRECTION(hash)) |
186 | goto release; | 186 | goto release; |
187 | 187 | ||
188 | l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct)); | 188 | l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct)); |
189 | NF_CT_ASSERT(l3proto); | 189 | NF_CT_ASSERT(l3proto); |
190 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); | 190 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
191 | NF_CT_ASSERT(l4proto); | 191 | NF_CT_ASSERT(l4proto); |
192 | 192 | ||
193 | ret = -ENOSPC; | 193 | ret = -ENOSPC; |
194 | if (seq_printf(s, "%-8s %u %-8s %u %ld ", | 194 | if (seq_printf(s, "%-8s %u %-8s %u %ld ", |
195 | l3proto->name, nf_ct_l3num(ct), | 195 | l3proto->name, nf_ct_l3num(ct), |
196 | l4proto->name, nf_ct_protonum(ct), | 196 | l4proto->name, nf_ct_protonum(ct), |
197 | timer_pending(&ct->timeout) | 197 | timer_pending(&ct->timeout) |
198 | ? (long)(ct->timeout.expires - jiffies)/HZ : 0) != 0) | 198 | ? (long)(ct->timeout.expires - jiffies)/HZ : 0) != 0) |
199 | goto release; | 199 | goto release; |
200 | 200 | ||
201 | if (l4proto->print_conntrack && l4proto->print_conntrack(s, ct)) | 201 | if (l4proto->print_conntrack && l4proto->print_conntrack(s, ct)) |
202 | goto release; | 202 | goto release; |
203 | 203 | ||
204 | if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | 204 | if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
205 | l3proto, l4proto)) | 205 | l3proto, l4proto)) |
206 | goto release; | 206 | goto release; |
207 | 207 | ||
208 | if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL)) | 208 | if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL)) |
209 | goto release; | 209 | goto release; |
210 | 210 | ||
211 | if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status))) | 211 | if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status))) |
212 | if (seq_printf(s, "[UNREPLIED] ")) | 212 | if (seq_printf(s, "[UNREPLIED] ")) |
213 | goto release; | 213 | goto release; |
214 | 214 | ||
215 | if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, | 215 | if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
216 | l3proto, l4proto)) | 216 | l3proto, l4proto)) |
217 | goto release; | 217 | goto release; |
218 | 218 | ||
219 | if (seq_print_acct(s, ct, IP_CT_DIR_REPLY)) | 219 | if (seq_print_acct(s, ct, IP_CT_DIR_REPLY)) |
220 | goto release; | 220 | goto release; |
221 | 221 | ||
222 | if (test_bit(IPS_ASSURED_BIT, &ct->status)) | 222 | if (test_bit(IPS_ASSURED_BIT, &ct->status)) |
223 | if (seq_printf(s, "[ASSURED] ")) | 223 | if (seq_printf(s, "[ASSURED] ")) |
224 | goto release; | 224 | goto release; |
225 | 225 | ||
226 | #if defined(CONFIG_NF_CONNTRACK_MARK) | 226 | #if defined(CONFIG_NF_CONNTRACK_MARK) |
227 | if (seq_printf(s, "mark=%u ", ct->mark)) | 227 | if (seq_printf(s, "mark=%u ", ct->mark)) |
228 | goto release; | 228 | goto release; |
229 | #endif | 229 | #endif |
230 | 230 | ||
231 | if (ct_show_secctx(s, ct)) | 231 | if (ct_show_secctx(s, ct)) |
232 | goto release; | 232 | goto release; |
233 | 233 | ||
234 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 234 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
235 | if (seq_printf(s, "zone=%u ", nf_ct_zone(ct))) | 235 | if (seq_printf(s, "zone=%u ", nf_ct_zone(ct))) |
236 | goto release; | 236 | goto release; |
237 | #endif | 237 | #endif |
238 | 238 | ||
239 | if (ct_show_delta_time(s, ct)) | 239 | if (ct_show_delta_time(s, ct)) |
240 | goto release; | 240 | goto release; |
241 | 241 | ||
242 | if (seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use))) | 242 | if (seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use))) |
243 | goto release; | 243 | goto release; |
244 | 244 | ||
245 | ret = 0; | 245 | ret = 0; |
246 | release: | 246 | release: |
247 | nf_ct_put(ct); | 247 | nf_ct_put(ct); |
248 | return ret; | 248 | return ret; |
249 | } | 249 | } |
250 | 250 | ||
251 | static const struct seq_operations ct_seq_ops = { | 251 | static const struct seq_operations ct_seq_ops = { |
252 | .start = ct_seq_start, | 252 | .start = ct_seq_start, |
253 | .next = ct_seq_next, | 253 | .next = ct_seq_next, |
254 | .stop = ct_seq_stop, | 254 | .stop = ct_seq_stop, |
255 | .show = ct_seq_show | 255 | .show = ct_seq_show |
256 | }; | 256 | }; |
257 | 257 | ||
258 | static int ct_open(struct inode *inode, struct file *file) | 258 | static int ct_open(struct inode *inode, struct file *file) |
259 | { | 259 | { |
260 | return seq_open_net(inode, file, &ct_seq_ops, | 260 | return seq_open_net(inode, file, &ct_seq_ops, |
261 | sizeof(struct ct_iter_state)); | 261 | sizeof(struct ct_iter_state)); |
262 | } | 262 | } |
263 | 263 | ||
264 | static const struct file_operations ct_file_ops = { | 264 | static const struct file_operations ct_file_ops = { |
265 | .owner = THIS_MODULE, | 265 | .owner = THIS_MODULE, |
266 | .open = ct_open, | 266 | .open = ct_open, |
267 | .read = seq_read, | 267 | .read = seq_read, |
268 | .llseek = seq_lseek, | 268 | .llseek = seq_lseek, |
269 | .release = seq_release_net, | 269 | .release = seq_release_net, |
270 | }; | 270 | }; |
271 | 271 | ||
272 | static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos) | 272 | static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos) |
273 | { | 273 | { |
274 | struct net *net = seq_file_net(seq); | 274 | struct net *net = seq_file_net(seq); |
275 | int cpu; | 275 | int cpu; |
276 | 276 | ||
277 | if (*pos == 0) | 277 | if (*pos == 0) |
278 | return SEQ_START_TOKEN; | 278 | return SEQ_START_TOKEN; |
279 | 279 | ||
280 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { | 280 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
281 | if (!cpu_possible(cpu)) | 281 | if (!cpu_possible(cpu)) |
282 | continue; | 282 | continue; |
283 | *pos = cpu + 1; | 283 | *pos = cpu + 1; |
284 | return per_cpu_ptr(net->ct.stat, cpu); | 284 | return per_cpu_ptr(net->ct.stat, cpu); |
285 | } | 285 | } |
286 | 286 | ||
287 | return NULL; | 287 | return NULL; |
288 | } | 288 | } |
289 | 289 | ||
290 | static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | 290 | static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
291 | { | 291 | { |
292 | struct net *net = seq_file_net(seq); | 292 | struct net *net = seq_file_net(seq); |
293 | int cpu; | 293 | int cpu; |
294 | 294 | ||
295 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { | 295 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
296 | if (!cpu_possible(cpu)) | 296 | if (!cpu_possible(cpu)) |
297 | continue; | 297 | continue; |
298 | *pos = cpu + 1; | 298 | *pos = cpu + 1; |
299 | return per_cpu_ptr(net->ct.stat, cpu); | 299 | return per_cpu_ptr(net->ct.stat, cpu); |
300 | } | 300 | } |
301 | 301 | ||
302 | return NULL; | 302 | return NULL; |
303 | } | 303 | } |
304 | 304 | ||
305 | static void ct_cpu_seq_stop(struct seq_file *seq, void *v) | 305 | static void ct_cpu_seq_stop(struct seq_file *seq, void *v) |
306 | { | 306 | { |
307 | } | 307 | } |
308 | 308 | ||
309 | static int ct_cpu_seq_show(struct seq_file *seq, void *v) | 309 | static int ct_cpu_seq_show(struct seq_file *seq, void *v) |
310 | { | 310 | { |
311 | struct net *net = seq_file_net(seq); | 311 | struct net *net = seq_file_net(seq); |
312 | unsigned int nr_conntracks = atomic_read(&net->ct.count); | 312 | unsigned int nr_conntracks = atomic_read(&net->ct.count); |
313 | const struct ip_conntrack_stat *st = v; | 313 | const struct ip_conntrack_stat *st = v; |
314 | 314 | ||
315 | if (v == SEQ_START_TOKEN) { | 315 | if (v == SEQ_START_TOKEN) { |
316 | seq_printf(seq, "entries searched found new invalid ignore delete delete_list insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n"); | 316 | seq_printf(seq, "entries searched found new invalid ignore delete delete_list insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n"); |
317 | return 0; | 317 | return 0; |
318 | } | 318 | } |
319 | 319 | ||
320 | seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x " | 320 | seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x " |
321 | "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n", | 321 | "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n", |
322 | nr_conntracks, | 322 | nr_conntracks, |
323 | st->searched, | 323 | st->searched, |
324 | st->found, | 324 | st->found, |
325 | st->new, | 325 | st->new, |
326 | st->invalid, | 326 | st->invalid, |
327 | st->ignore, | 327 | st->ignore, |
328 | st->delete, | 328 | st->delete, |
329 | st->delete_list, | 329 | st->delete_list, |
330 | st->insert, | 330 | st->insert, |
331 | st->insert_failed, | 331 | st->insert_failed, |
332 | st->drop, | 332 | st->drop, |
333 | st->early_drop, | 333 | st->early_drop, |
334 | st->error, | 334 | st->error, |
335 | 335 | ||
336 | st->expect_new, | 336 | st->expect_new, |
337 | st->expect_create, | 337 | st->expect_create, |
338 | st->expect_delete, | 338 | st->expect_delete, |
339 | st->search_restart | 339 | st->search_restart |
340 | ); | 340 | ); |
341 | return 0; | 341 | return 0; |
342 | } | 342 | } |
343 | 343 | ||
344 | static const struct seq_operations ct_cpu_seq_ops = { | 344 | static const struct seq_operations ct_cpu_seq_ops = { |
345 | .start = ct_cpu_seq_start, | 345 | .start = ct_cpu_seq_start, |
346 | .next = ct_cpu_seq_next, | 346 | .next = ct_cpu_seq_next, |
347 | .stop = ct_cpu_seq_stop, | 347 | .stop = ct_cpu_seq_stop, |
348 | .show = ct_cpu_seq_show, | 348 | .show = ct_cpu_seq_show, |
349 | }; | 349 | }; |
350 | 350 | ||
351 | static int ct_cpu_seq_open(struct inode *inode, struct file *file) | 351 | static int ct_cpu_seq_open(struct inode *inode, struct file *file) |
352 | { | 352 | { |
353 | return seq_open_net(inode, file, &ct_cpu_seq_ops, | 353 | return seq_open_net(inode, file, &ct_cpu_seq_ops, |
354 | sizeof(struct seq_net_private)); | 354 | sizeof(struct seq_net_private)); |
355 | } | 355 | } |
356 | 356 | ||
357 | static const struct file_operations ct_cpu_seq_fops = { | 357 | static const struct file_operations ct_cpu_seq_fops = { |
358 | .owner = THIS_MODULE, | 358 | .owner = THIS_MODULE, |
359 | .open = ct_cpu_seq_open, | 359 | .open = ct_cpu_seq_open, |
360 | .read = seq_read, | 360 | .read = seq_read, |
361 | .llseek = seq_lseek, | 361 | .llseek = seq_lseek, |
362 | .release = seq_release_net, | 362 | .release = seq_release_net, |
363 | }; | 363 | }; |
364 | 364 | ||
365 | static int nf_conntrack_standalone_init_proc(struct net *net) | 365 | static int nf_conntrack_standalone_init_proc(struct net *net) |
366 | { | 366 | { |
367 | struct proc_dir_entry *pde; | 367 | struct proc_dir_entry *pde; |
368 | 368 | ||
369 | pde = proc_net_fops_create(net, "nf_conntrack", 0440, &ct_file_ops); | 369 | pde = proc_net_fops_create(net, "nf_conntrack", 0440, &ct_file_ops); |
370 | if (!pde) | 370 | if (!pde) |
371 | goto out_nf_conntrack; | 371 | goto out_nf_conntrack; |
372 | 372 | ||
373 | pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat, | 373 | pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat, |
374 | &ct_cpu_seq_fops); | 374 | &ct_cpu_seq_fops); |
375 | if (!pde) | 375 | if (!pde) |
376 | goto out_stat_nf_conntrack; | 376 | goto out_stat_nf_conntrack; |
377 | return 0; | 377 | return 0; |
378 | 378 | ||
379 | out_stat_nf_conntrack: | 379 | out_stat_nf_conntrack: |
380 | proc_net_remove(net, "nf_conntrack"); | 380 | proc_net_remove(net, "nf_conntrack"); |
381 | out_nf_conntrack: | 381 | out_nf_conntrack: |
382 | return -ENOMEM; | 382 | return -ENOMEM; |
383 | } | 383 | } |
384 | 384 | ||
385 | static void nf_conntrack_standalone_fini_proc(struct net *net) | 385 | static void nf_conntrack_standalone_fini_proc(struct net *net) |
386 | { | 386 | { |
387 | remove_proc_entry("nf_conntrack", net->proc_net_stat); | 387 | remove_proc_entry("nf_conntrack", net->proc_net_stat); |
388 | proc_net_remove(net, "nf_conntrack"); | 388 | proc_net_remove(net, "nf_conntrack"); |
389 | } | 389 | } |
390 | #else | 390 | #else |
391 | static int nf_conntrack_standalone_init_proc(struct net *net) | 391 | static int nf_conntrack_standalone_init_proc(struct net *net) |
392 | { | 392 | { |
393 | return 0; | 393 | return 0; |
394 | } | 394 | } |
395 | 395 | ||
396 | static void nf_conntrack_standalone_fini_proc(struct net *net) | 396 | static void nf_conntrack_standalone_fini_proc(struct net *net) |
397 | { | 397 | { |
398 | } | 398 | } |
399 | #endif /* CONFIG_NF_CONNTRACK_PROCFS */ | 399 | #endif /* CONFIG_NF_CONNTRACK_PROCFS */ |
400 | 400 | ||
401 | /* Sysctl support */ | 401 | /* Sysctl support */ |
402 | 402 | ||
403 | #ifdef CONFIG_SYSCTL | 403 | #ifdef CONFIG_SYSCTL |
404 | /* Log invalid packets of a given protocol */ | 404 | /* Log invalid packets of a given protocol */ |
405 | static int log_invalid_proto_min = 0; | 405 | static int log_invalid_proto_min = 0; |
406 | static int log_invalid_proto_max = 255; | 406 | static int log_invalid_proto_max = 255; |
407 | 407 | ||
408 | static struct ctl_table_header *nf_ct_netfilter_header; | 408 | static struct ctl_table_header *nf_ct_netfilter_header; |
409 | 409 | ||
410 | static ctl_table nf_ct_sysctl_table[] = { | 410 | static ctl_table nf_ct_sysctl_table[] = { |
411 | { | 411 | { |
412 | .procname = "nf_conntrack_max", | 412 | .procname = "nf_conntrack_max", |
413 | .data = &nf_conntrack_max, | 413 | .data = &nf_conntrack_max, |
414 | .maxlen = sizeof(int), | 414 | .maxlen = sizeof(int), |
415 | .mode = 0644, | 415 | .mode = 0644, |
416 | .proc_handler = proc_dointvec, | 416 | .proc_handler = proc_dointvec, |
417 | }, | 417 | }, |
418 | { | 418 | { |
419 | .procname = "nf_conntrack_count", | 419 | .procname = "nf_conntrack_count", |
420 | .data = &init_net.ct.count, | 420 | .data = &init_net.ct.count, |
421 | .maxlen = sizeof(int), | 421 | .maxlen = sizeof(int), |
422 | .mode = 0444, | 422 | .mode = 0444, |
423 | .proc_handler = proc_dointvec, | 423 | .proc_handler = proc_dointvec, |
424 | }, | 424 | }, |
425 | { | 425 | { |
426 | .procname = "nf_conntrack_buckets", | 426 | .procname = "nf_conntrack_buckets", |
427 | .data = &init_net.ct.htable_size, | 427 | .data = &init_net.ct.htable_size, |
428 | .maxlen = sizeof(unsigned int), | 428 | .maxlen = sizeof(unsigned int), |
429 | .mode = 0444, | 429 | .mode = 0444, |
430 | .proc_handler = proc_dointvec, | 430 | .proc_handler = proc_dointvec, |
431 | }, | 431 | }, |
432 | { | 432 | { |
433 | .procname = "nf_conntrack_checksum", | 433 | .procname = "nf_conntrack_checksum", |
434 | .data = &init_net.ct.sysctl_checksum, | 434 | .data = &init_net.ct.sysctl_checksum, |
435 | .maxlen = sizeof(unsigned int), | 435 | .maxlen = sizeof(unsigned int), |
436 | .mode = 0644, | 436 | .mode = 0644, |
437 | .proc_handler = proc_dointvec, | 437 | .proc_handler = proc_dointvec, |
438 | }, | 438 | }, |
439 | { | 439 | { |
440 | .procname = "nf_conntrack_log_invalid", | 440 | .procname = "nf_conntrack_log_invalid", |
441 | .data = &init_net.ct.sysctl_log_invalid, | 441 | .data = &init_net.ct.sysctl_log_invalid, |
442 | .maxlen = sizeof(unsigned int), | 442 | .maxlen = sizeof(unsigned int), |
443 | .mode = 0644, | 443 | .mode = 0644, |
444 | .proc_handler = proc_dointvec_minmax, | 444 | .proc_handler = proc_dointvec_minmax, |
445 | .extra1 = &log_invalid_proto_min, | 445 | .extra1 = &log_invalid_proto_min, |
446 | .extra2 = &log_invalid_proto_max, | 446 | .extra2 = &log_invalid_proto_max, |
447 | }, | 447 | }, |
448 | { | 448 | { |
449 | .procname = "nf_conntrack_expect_max", | 449 | .procname = "nf_conntrack_expect_max", |
450 | .data = &nf_ct_expect_max, | 450 | .data = &nf_ct_expect_max, |
451 | .maxlen = sizeof(int), | 451 | .maxlen = sizeof(int), |
452 | .mode = 0644, | 452 | .mode = 0644, |
453 | .proc_handler = proc_dointvec, | 453 | .proc_handler = proc_dointvec, |
454 | }, | 454 | }, |
455 | { } | 455 | { } |
456 | }; | 456 | }; |
457 | 457 | ||
458 | #define NET_NF_CONNTRACK_MAX 2089 | 458 | #define NET_NF_CONNTRACK_MAX 2089 |
459 | 459 | ||
460 | static ctl_table nf_ct_netfilter_table[] = { | 460 | static ctl_table nf_ct_netfilter_table[] = { |
461 | { | 461 | { |
462 | .procname = "nf_conntrack_max", | 462 | .procname = "nf_conntrack_max", |
463 | .data = &nf_conntrack_max, | 463 | .data = &nf_conntrack_max, |
464 | .maxlen = sizeof(int), | 464 | .maxlen = sizeof(int), |
465 | .mode = 0644, | 465 | .mode = 0644, |
466 | .proc_handler = proc_dointvec, | 466 | .proc_handler = proc_dointvec, |
467 | }, | 467 | }, |
468 | { } | 468 | { } |
469 | }; | 469 | }; |
470 | 470 | ||
471 | static int nf_conntrack_standalone_init_sysctl(struct net *net) | 471 | static int nf_conntrack_standalone_init_sysctl(struct net *net) |
472 | { | 472 | { |
473 | struct ctl_table *table; | 473 | struct ctl_table *table; |
474 | 474 | ||
475 | if (net_eq(net, &init_net)) { | 475 | if (net_eq(net, &init_net)) { |
476 | nf_ct_netfilter_header = | 476 | nf_ct_netfilter_header = |
477 | register_net_sysctl(&init_net, "net", nf_ct_netfilter_table); | 477 | register_net_sysctl(&init_net, "net", nf_ct_netfilter_table); |
478 | if (!nf_ct_netfilter_header) | 478 | if (!nf_ct_netfilter_header) |
479 | goto out; | 479 | goto out; |
480 | } | 480 | } |
481 | 481 | ||
482 | table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table), | 482 | table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table), |
483 | GFP_KERNEL); | 483 | GFP_KERNEL); |
484 | if (!table) | 484 | if (!table) |
485 | goto out_kmemdup; | 485 | goto out_kmemdup; |
486 | 486 | ||
487 | table[1].data = &net->ct.count; | 487 | table[1].data = &net->ct.count; |
488 | table[2].data = &net->ct.htable_size; | 488 | table[2].data = &net->ct.htable_size; |
489 | table[3].data = &net->ct.sysctl_checksum; | 489 | table[3].data = &net->ct.sysctl_checksum; |
490 | table[4].data = &net->ct.sysctl_log_invalid; | 490 | table[4].data = &net->ct.sysctl_log_invalid; |
491 | 491 | ||
492 | /* Don't export sysctls to unprivileged users */ | 492 | /* Don't export sysctls to unprivileged users */ |
493 | if (net->user_ns != &init_user_ns) | 493 | if (net->user_ns != &init_user_ns) |
494 | table[0].procname = NULL; | 494 | table[0].procname = NULL; |
495 | 495 | ||
496 | net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table); | 496 | net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table); |
497 | if (!net->ct.sysctl_header) | 497 | if (!net->ct.sysctl_header) |
498 | goto out_unregister_netfilter; | 498 | goto out_unregister_netfilter; |
499 | 499 | ||
500 | return 0; | 500 | return 0; |
501 | 501 | ||
502 | out_unregister_netfilter: | 502 | out_unregister_netfilter: |
503 | kfree(table); | 503 | kfree(table); |
504 | out_kmemdup: | 504 | out_kmemdup: |
505 | if (net_eq(net, &init_net)) | 505 | if (net_eq(net, &init_net)) |
506 | unregister_net_sysctl_table(nf_ct_netfilter_header); | 506 | unregister_net_sysctl_table(nf_ct_netfilter_header); |
507 | out: | 507 | out: |
508 | printk(KERN_ERR "nf_conntrack: can't register to sysctl.\n"); | 508 | printk(KERN_ERR "nf_conntrack: can't register to sysctl.\n"); |
509 | return -ENOMEM; | 509 | return -ENOMEM; |
510 | } | 510 | } |
511 | 511 | ||
512 | static void nf_conntrack_standalone_fini_sysctl(struct net *net) | 512 | static void nf_conntrack_standalone_fini_sysctl(struct net *net) |
513 | { | 513 | { |
514 | struct ctl_table *table; | 514 | struct ctl_table *table; |
515 | 515 | ||
516 | if (net_eq(net, &init_net)) | 516 | if (net_eq(net, &init_net)) |
517 | unregister_net_sysctl_table(nf_ct_netfilter_header); | 517 | unregister_net_sysctl_table(nf_ct_netfilter_header); |
518 | table = net->ct.sysctl_header->ctl_table_arg; | 518 | table = net->ct.sysctl_header->ctl_table_arg; |
519 | unregister_net_sysctl_table(net->ct.sysctl_header); | 519 | unregister_net_sysctl_table(net->ct.sysctl_header); |
520 | kfree(table); | 520 | kfree(table); |
521 | } | 521 | } |
522 | #else | 522 | #else |
523 | static int nf_conntrack_standalone_init_sysctl(struct net *net) | 523 | static int nf_conntrack_standalone_init_sysctl(struct net *net) |
524 | { | 524 | { |
525 | return 0; | 525 | return 0; |
526 | } | 526 | } |
527 | 527 | ||
528 | static void nf_conntrack_standalone_fini_sysctl(struct net *net) | 528 | static void nf_conntrack_standalone_fini_sysctl(struct net *net) |
529 | { | 529 | { |
530 | } | 530 | } |
531 | #endif /* CONFIG_SYSCTL */ | 531 | #endif /* CONFIG_SYSCTL */ |
532 | 532 | ||
533 | static int nf_conntrack_net_init(struct net *net) | 533 | static int nf_conntrack_net_init(struct net *net) |
534 | { | 534 | { |
535 | int ret; | 535 | int ret; |
536 | 536 | ||
537 | ret = nf_conntrack_init(net); | 537 | ret = nf_conntrack_init(net); |
538 | if (ret < 0) | 538 | if (ret < 0) |
539 | goto out_init; | 539 | goto out_init; |
540 | ret = nf_conntrack_standalone_init_proc(net); | 540 | ret = nf_conntrack_standalone_init_proc(net); |
541 | if (ret < 0) | 541 | if (ret < 0) |
542 | goto out_proc; | 542 | goto out_proc; |
543 | net->ct.sysctl_checksum = 1; | 543 | net->ct.sysctl_checksum = 1; |
544 | net->ct.sysctl_log_invalid = 0; | 544 | net->ct.sysctl_log_invalid = 0; |
545 | ret = nf_conntrack_standalone_init_sysctl(net); | 545 | ret = nf_conntrack_standalone_init_sysctl(net); |
546 | if (ret < 0) | 546 | if (ret < 0) |
547 | goto out_sysctl; | 547 | goto out_sysctl; |
548 | return 0; | 548 | return 0; |
549 | 549 | ||
550 | out_sysctl: | 550 | out_sysctl: |
551 | nf_conntrack_standalone_fini_proc(net); | 551 | nf_conntrack_standalone_fini_proc(net); |
552 | out_proc: | 552 | out_proc: |
553 | nf_conntrack_cleanup(net); | 553 | nf_conntrack_cleanup(net); |
554 | out_init: | 554 | out_init: |
555 | return ret; | 555 | return ret; |
556 | } | 556 | } |
557 | 557 | ||
558 | static void nf_conntrack_net_exit(struct net *net) | 558 | static void nf_conntrack_net_exit(struct net *net) |
559 | { | 559 | { |
560 | nf_conntrack_standalone_fini_sysctl(net); | 560 | nf_conntrack_standalone_fini_sysctl(net); |
561 | nf_conntrack_standalone_fini_proc(net); | 561 | nf_conntrack_standalone_fini_proc(net); |
562 | nf_conntrack_cleanup(net); | 562 | nf_conntrack_cleanup(net); |
563 | } | 563 | } |
564 | 564 | ||
565 | static struct pernet_operations nf_conntrack_net_ops = { | 565 | static struct pernet_operations nf_conntrack_net_ops = { |
566 | .init = nf_conntrack_net_init, | 566 | .init = nf_conntrack_net_init, |
567 | .exit = nf_conntrack_net_exit, | 567 | .exit = nf_conntrack_net_exit, |
568 | }; | 568 | }; |
569 | 569 | ||
570 | static int __init nf_conntrack_standalone_init(void) | 570 | static int __init nf_conntrack_standalone_init(void) |
571 | { | 571 | { |
572 | return register_pernet_subsys(&nf_conntrack_net_ops); | 572 | return register_pernet_subsys(&nf_conntrack_net_ops); |
573 | } | 573 | } |
574 | 574 | ||
575 | static void __exit nf_conntrack_standalone_fini(void) | 575 | static void __exit nf_conntrack_standalone_fini(void) |
576 | { | 576 | { |
577 | unregister_pernet_subsys(&nf_conntrack_net_ops); | 577 | unregister_pernet_subsys(&nf_conntrack_net_ops); |
578 | nf_conntrack_cleanup_end(); | ||
578 | } | 579 | } |
579 | 580 | ||
580 | module_init(nf_conntrack_standalone_init); | 581 | module_init(nf_conntrack_standalone_init); |
581 | module_exit(nf_conntrack_standalone_fini); | 582 | module_exit(nf_conntrack_standalone_fini); |
582 | 583 | ||
583 | /* Some modules need us, but don't depend directly on any symbol. | 584 | /* Some modules need us, but don't depend directly on any symbol. |
584 | They should call this. */ | 585 | They should call this. */ |
585 | void need_conntrack(void) | 586 | void need_conntrack(void) |
586 | { | 587 | { |
587 | } | 588 | } |
588 | EXPORT_SYMBOL_GPL(need_conntrack); | 589 | EXPORT_SYMBOL_GPL(need_conntrack); |
589 | 590 |