Commit 08733a0cb7decce40bbbd0331a0449465f13c444
1 parent
26dfab7216
netfilter: handle NF_REPEAT from nf_conntrack_in()
NF_REPEAT is only needed from nf_conntrack_in() under a very specific case required by the TCP protocol tracker, we can handle this case without returning to the core hook path. Handling of NF_REPEAT from the nf_reinject() is left untouched. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Showing 3 changed files with 8 additions and 13 deletions Inline Diff
net/netfilter/core.c
1 | /* netfilter.c: look after the filters for various protocols. | 1 | /* netfilter.c: look after the filters for various protocols. |
2 | * Heavily influenced by the old firewall.c by David Bonn and Alan Cox. | 2 | * Heavily influenced by the old firewall.c by David Bonn and Alan Cox. |
3 | * | 3 | * |
4 | * Thanks to Rob `CmdrTaco' Malda for not influencing this code in any | 4 | * Thanks to Rob `CmdrTaco' Malda for not influencing this code in any |
5 | * way. | 5 | * way. |
6 | * | 6 | * |
7 | * Rusty Russell (C)2000 -- This code is GPL. | 7 | * Rusty Russell (C)2000 -- This code is GPL. |
8 | * Patrick McHardy (c) 2006-2012 | 8 | * Patrick McHardy (c) 2006-2012 |
9 | */ | 9 | */ |
10 | #include <linux/kernel.h> | 10 | #include <linux/kernel.h> |
11 | #include <linux/netfilter.h> | 11 | #include <linux/netfilter.h> |
12 | #include <net/protocol.h> | 12 | #include <net/protocol.h> |
13 | #include <linux/init.h> | 13 | #include <linux/init.h> |
14 | #include <linux/skbuff.h> | 14 | #include <linux/skbuff.h> |
15 | #include <linux/wait.h> | 15 | #include <linux/wait.h> |
16 | #include <linux/module.h> | 16 | #include <linux/module.h> |
17 | #include <linux/interrupt.h> | 17 | #include <linux/interrupt.h> |
18 | #include <linux/if.h> | 18 | #include <linux/if.h> |
19 | #include <linux/netdevice.h> | 19 | #include <linux/netdevice.h> |
20 | #include <linux/netfilter_ipv6.h> | 20 | #include <linux/netfilter_ipv6.h> |
21 | #include <linux/inetdevice.h> | 21 | #include <linux/inetdevice.h> |
22 | #include <linux/proc_fs.h> | 22 | #include <linux/proc_fs.h> |
23 | #include <linux/mutex.h> | 23 | #include <linux/mutex.h> |
24 | #include <linux/slab.h> | 24 | #include <linux/slab.h> |
25 | #include <linux/rcupdate.h> | 25 | #include <linux/rcupdate.h> |
26 | #include <net/net_namespace.h> | 26 | #include <net/net_namespace.h> |
27 | #include <net/sock.h> | 27 | #include <net/sock.h> |
28 | 28 | ||
29 | #include "nf_internals.h" | 29 | #include "nf_internals.h" |
30 | 30 | ||
31 | static DEFINE_MUTEX(afinfo_mutex); | 31 | static DEFINE_MUTEX(afinfo_mutex); |
32 | 32 | ||
33 | const struct nf_afinfo __rcu *nf_afinfo[NFPROTO_NUMPROTO] __read_mostly; | 33 | const struct nf_afinfo __rcu *nf_afinfo[NFPROTO_NUMPROTO] __read_mostly; |
34 | EXPORT_SYMBOL(nf_afinfo); | 34 | EXPORT_SYMBOL(nf_afinfo); |
35 | const struct nf_ipv6_ops __rcu *nf_ipv6_ops __read_mostly; | 35 | const struct nf_ipv6_ops __rcu *nf_ipv6_ops __read_mostly; |
36 | EXPORT_SYMBOL_GPL(nf_ipv6_ops); | 36 | EXPORT_SYMBOL_GPL(nf_ipv6_ops); |
37 | 37 | ||
38 | DEFINE_PER_CPU(bool, nf_skb_duplicated); | 38 | DEFINE_PER_CPU(bool, nf_skb_duplicated); |
39 | EXPORT_SYMBOL_GPL(nf_skb_duplicated); | 39 | EXPORT_SYMBOL_GPL(nf_skb_duplicated); |
40 | 40 | ||
41 | int nf_register_afinfo(const struct nf_afinfo *afinfo) | 41 | int nf_register_afinfo(const struct nf_afinfo *afinfo) |
42 | { | 42 | { |
43 | mutex_lock(&afinfo_mutex); | 43 | mutex_lock(&afinfo_mutex); |
44 | RCU_INIT_POINTER(nf_afinfo[afinfo->family], afinfo); | 44 | RCU_INIT_POINTER(nf_afinfo[afinfo->family], afinfo); |
45 | mutex_unlock(&afinfo_mutex); | 45 | mutex_unlock(&afinfo_mutex); |
46 | return 0; | 46 | return 0; |
47 | } | 47 | } |
48 | EXPORT_SYMBOL_GPL(nf_register_afinfo); | 48 | EXPORT_SYMBOL_GPL(nf_register_afinfo); |
49 | 49 | ||
50 | void nf_unregister_afinfo(const struct nf_afinfo *afinfo) | 50 | void nf_unregister_afinfo(const struct nf_afinfo *afinfo) |
51 | { | 51 | { |
52 | mutex_lock(&afinfo_mutex); | 52 | mutex_lock(&afinfo_mutex); |
53 | RCU_INIT_POINTER(nf_afinfo[afinfo->family], NULL); | 53 | RCU_INIT_POINTER(nf_afinfo[afinfo->family], NULL); |
54 | mutex_unlock(&afinfo_mutex); | 54 | mutex_unlock(&afinfo_mutex); |
55 | synchronize_rcu(); | 55 | synchronize_rcu(); |
56 | } | 56 | } |
57 | EXPORT_SYMBOL_GPL(nf_unregister_afinfo); | 57 | EXPORT_SYMBOL_GPL(nf_unregister_afinfo); |
58 | 58 | ||
59 | #ifdef HAVE_JUMP_LABEL | 59 | #ifdef HAVE_JUMP_LABEL |
60 | struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS]; | 60 | struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS]; |
61 | EXPORT_SYMBOL(nf_hooks_needed); | 61 | EXPORT_SYMBOL(nf_hooks_needed); |
62 | #endif | 62 | #endif |
63 | 63 | ||
64 | static DEFINE_MUTEX(nf_hook_mutex); | 64 | static DEFINE_MUTEX(nf_hook_mutex); |
65 | #define nf_entry_dereference(e) \ | 65 | #define nf_entry_dereference(e) \ |
66 | rcu_dereference_protected(e, lockdep_is_held(&nf_hook_mutex)) | 66 | rcu_dereference_protected(e, lockdep_is_held(&nf_hook_mutex)) |
67 | 67 | ||
68 | static struct nf_hook_entry __rcu **nf_hook_entry_head(struct net *net, const struct nf_hook_ops *reg) | 68 | static struct nf_hook_entry __rcu **nf_hook_entry_head(struct net *net, const struct nf_hook_ops *reg) |
69 | { | 69 | { |
70 | if (reg->pf != NFPROTO_NETDEV) | 70 | if (reg->pf != NFPROTO_NETDEV) |
71 | return net->nf.hooks[reg->pf]+reg->hooknum; | 71 | return net->nf.hooks[reg->pf]+reg->hooknum; |
72 | 72 | ||
73 | #ifdef CONFIG_NETFILTER_INGRESS | 73 | #ifdef CONFIG_NETFILTER_INGRESS |
74 | if (reg->hooknum == NF_NETDEV_INGRESS) { | 74 | if (reg->hooknum == NF_NETDEV_INGRESS) { |
75 | if (reg->dev && dev_net(reg->dev) == net) | 75 | if (reg->dev && dev_net(reg->dev) == net) |
76 | return ®->dev->nf_hooks_ingress; | 76 | return ®->dev->nf_hooks_ingress; |
77 | } | 77 | } |
78 | #endif | 78 | #endif |
79 | return NULL; | 79 | return NULL; |
80 | } | 80 | } |
81 | 81 | ||
82 | int nf_register_net_hook(struct net *net, const struct nf_hook_ops *reg) | 82 | int nf_register_net_hook(struct net *net, const struct nf_hook_ops *reg) |
83 | { | 83 | { |
84 | struct nf_hook_entry __rcu **pp; | 84 | struct nf_hook_entry __rcu **pp; |
85 | struct nf_hook_entry *entry, *p; | 85 | struct nf_hook_entry *entry, *p; |
86 | 86 | ||
87 | if (reg->pf == NFPROTO_NETDEV) { | 87 | if (reg->pf == NFPROTO_NETDEV) { |
88 | #ifndef CONFIG_NETFILTER_INGRESS | 88 | #ifndef CONFIG_NETFILTER_INGRESS |
89 | if (reg->hooknum == NF_NETDEV_INGRESS) | 89 | if (reg->hooknum == NF_NETDEV_INGRESS) |
90 | return -EOPNOTSUPP; | 90 | return -EOPNOTSUPP; |
91 | #endif | 91 | #endif |
92 | if (reg->hooknum != NF_NETDEV_INGRESS || | 92 | if (reg->hooknum != NF_NETDEV_INGRESS || |
93 | !reg->dev || dev_net(reg->dev) != net) | 93 | !reg->dev || dev_net(reg->dev) != net) |
94 | return -EINVAL; | 94 | return -EINVAL; |
95 | } | 95 | } |
96 | 96 | ||
97 | pp = nf_hook_entry_head(net, reg); | 97 | pp = nf_hook_entry_head(net, reg); |
98 | if (!pp) | 98 | if (!pp) |
99 | return -EINVAL; | 99 | return -EINVAL; |
100 | 100 | ||
101 | entry = kmalloc(sizeof(*entry), GFP_KERNEL); | 101 | entry = kmalloc(sizeof(*entry), GFP_KERNEL); |
102 | if (!entry) | 102 | if (!entry) |
103 | return -ENOMEM; | 103 | return -ENOMEM; |
104 | 104 | ||
105 | entry->orig_ops = reg; | 105 | entry->orig_ops = reg; |
106 | entry->ops = *reg; | 106 | entry->ops = *reg; |
107 | entry->next = NULL; | 107 | entry->next = NULL; |
108 | 108 | ||
109 | mutex_lock(&nf_hook_mutex); | 109 | mutex_lock(&nf_hook_mutex); |
110 | 110 | ||
111 | /* Find the spot in the list */ | 111 | /* Find the spot in the list */ |
112 | while ((p = nf_entry_dereference(*pp)) != NULL) { | 112 | while ((p = nf_entry_dereference(*pp)) != NULL) { |
113 | if (reg->priority < p->orig_ops->priority) | 113 | if (reg->priority < p->orig_ops->priority) |
114 | break; | 114 | break; |
115 | pp = &p->next; | 115 | pp = &p->next; |
116 | } | 116 | } |
117 | rcu_assign_pointer(entry->next, p); | 117 | rcu_assign_pointer(entry->next, p); |
118 | rcu_assign_pointer(*pp, entry); | 118 | rcu_assign_pointer(*pp, entry); |
119 | 119 | ||
120 | mutex_unlock(&nf_hook_mutex); | 120 | mutex_unlock(&nf_hook_mutex); |
121 | #ifdef CONFIG_NETFILTER_INGRESS | 121 | #ifdef CONFIG_NETFILTER_INGRESS |
122 | if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS) | 122 | if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS) |
123 | net_inc_ingress_queue(); | 123 | net_inc_ingress_queue(); |
124 | #endif | 124 | #endif |
125 | #ifdef HAVE_JUMP_LABEL | 125 | #ifdef HAVE_JUMP_LABEL |
126 | static_key_slow_inc(&nf_hooks_needed[reg->pf][reg->hooknum]); | 126 | static_key_slow_inc(&nf_hooks_needed[reg->pf][reg->hooknum]); |
127 | #endif | 127 | #endif |
128 | return 0; | 128 | return 0; |
129 | } | 129 | } |
130 | EXPORT_SYMBOL(nf_register_net_hook); | 130 | EXPORT_SYMBOL(nf_register_net_hook); |
131 | 131 | ||
132 | void nf_unregister_net_hook(struct net *net, const struct nf_hook_ops *reg) | 132 | void nf_unregister_net_hook(struct net *net, const struct nf_hook_ops *reg) |
133 | { | 133 | { |
134 | struct nf_hook_entry __rcu **pp; | 134 | struct nf_hook_entry __rcu **pp; |
135 | struct nf_hook_entry *p; | 135 | struct nf_hook_entry *p; |
136 | 136 | ||
137 | pp = nf_hook_entry_head(net, reg); | 137 | pp = nf_hook_entry_head(net, reg); |
138 | if (WARN_ON_ONCE(!pp)) | 138 | if (WARN_ON_ONCE(!pp)) |
139 | return; | 139 | return; |
140 | 140 | ||
141 | mutex_lock(&nf_hook_mutex); | 141 | mutex_lock(&nf_hook_mutex); |
142 | while ((p = nf_entry_dereference(*pp)) != NULL) { | 142 | while ((p = nf_entry_dereference(*pp)) != NULL) { |
143 | if (p->orig_ops == reg) { | 143 | if (p->orig_ops == reg) { |
144 | rcu_assign_pointer(*pp, p->next); | 144 | rcu_assign_pointer(*pp, p->next); |
145 | break; | 145 | break; |
146 | } | 146 | } |
147 | pp = &p->next; | 147 | pp = &p->next; |
148 | } | 148 | } |
149 | mutex_unlock(&nf_hook_mutex); | 149 | mutex_unlock(&nf_hook_mutex); |
150 | if (!p) { | 150 | if (!p) { |
151 | WARN(1, "nf_unregister_net_hook: hook not found!\n"); | 151 | WARN(1, "nf_unregister_net_hook: hook not found!\n"); |
152 | return; | 152 | return; |
153 | } | 153 | } |
154 | #ifdef CONFIG_NETFILTER_INGRESS | 154 | #ifdef CONFIG_NETFILTER_INGRESS |
155 | if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS) | 155 | if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS) |
156 | net_dec_ingress_queue(); | 156 | net_dec_ingress_queue(); |
157 | #endif | 157 | #endif |
158 | #ifdef HAVE_JUMP_LABEL | 158 | #ifdef HAVE_JUMP_LABEL |
159 | static_key_slow_dec(&nf_hooks_needed[reg->pf][reg->hooknum]); | 159 | static_key_slow_dec(&nf_hooks_needed[reg->pf][reg->hooknum]); |
160 | #endif | 160 | #endif |
161 | synchronize_net(); | 161 | synchronize_net(); |
162 | nf_queue_nf_hook_drop(net, p); | 162 | nf_queue_nf_hook_drop(net, p); |
163 | /* other cpu might still process nfqueue verdict that used reg */ | 163 | /* other cpu might still process nfqueue verdict that used reg */ |
164 | synchronize_net(); | 164 | synchronize_net(); |
165 | kfree(p); | 165 | kfree(p); |
166 | } | 166 | } |
167 | EXPORT_SYMBOL(nf_unregister_net_hook); | 167 | EXPORT_SYMBOL(nf_unregister_net_hook); |
168 | 168 | ||
169 | int nf_register_net_hooks(struct net *net, const struct nf_hook_ops *reg, | 169 | int nf_register_net_hooks(struct net *net, const struct nf_hook_ops *reg, |
170 | unsigned int n) | 170 | unsigned int n) |
171 | { | 171 | { |
172 | unsigned int i; | 172 | unsigned int i; |
173 | int err = 0; | 173 | int err = 0; |
174 | 174 | ||
175 | for (i = 0; i < n; i++) { | 175 | for (i = 0; i < n; i++) { |
176 | err = nf_register_net_hook(net, ®[i]); | 176 | err = nf_register_net_hook(net, ®[i]); |
177 | if (err) | 177 | if (err) |
178 | goto err; | 178 | goto err; |
179 | } | 179 | } |
180 | return err; | 180 | return err; |
181 | 181 | ||
182 | err: | 182 | err: |
183 | if (i > 0) | 183 | if (i > 0) |
184 | nf_unregister_net_hooks(net, reg, i); | 184 | nf_unregister_net_hooks(net, reg, i); |
185 | return err; | 185 | return err; |
186 | } | 186 | } |
187 | EXPORT_SYMBOL(nf_register_net_hooks); | 187 | EXPORT_SYMBOL(nf_register_net_hooks); |
188 | 188 | ||
189 | void nf_unregister_net_hooks(struct net *net, const struct nf_hook_ops *reg, | 189 | void nf_unregister_net_hooks(struct net *net, const struct nf_hook_ops *reg, |
190 | unsigned int n) | 190 | unsigned int n) |
191 | { | 191 | { |
192 | while (n-- > 0) | 192 | while (n-- > 0) |
193 | nf_unregister_net_hook(net, ®[n]); | 193 | nf_unregister_net_hook(net, ®[n]); |
194 | } | 194 | } |
195 | EXPORT_SYMBOL(nf_unregister_net_hooks); | 195 | EXPORT_SYMBOL(nf_unregister_net_hooks); |
196 | 196 | ||
197 | static LIST_HEAD(nf_hook_list); | 197 | static LIST_HEAD(nf_hook_list); |
198 | 198 | ||
199 | static int _nf_register_hook(struct nf_hook_ops *reg) | 199 | static int _nf_register_hook(struct nf_hook_ops *reg) |
200 | { | 200 | { |
201 | struct net *net, *last; | 201 | struct net *net, *last; |
202 | int ret; | 202 | int ret; |
203 | 203 | ||
204 | for_each_net(net) { | 204 | for_each_net(net) { |
205 | ret = nf_register_net_hook(net, reg); | 205 | ret = nf_register_net_hook(net, reg); |
206 | if (ret && ret != -ENOENT) | 206 | if (ret && ret != -ENOENT) |
207 | goto rollback; | 207 | goto rollback; |
208 | } | 208 | } |
209 | list_add_tail(®->list, &nf_hook_list); | 209 | list_add_tail(®->list, &nf_hook_list); |
210 | 210 | ||
211 | return 0; | 211 | return 0; |
212 | rollback: | 212 | rollback: |
213 | last = net; | 213 | last = net; |
214 | for_each_net(net) { | 214 | for_each_net(net) { |
215 | if (net == last) | 215 | if (net == last) |
216 | break; | 216 | break; |
217 | nf_unregister_net_hook(net, reg); | 217 | nf_unregister_net_hook(net, reg); |
218 | } | 218 | } |
219 | return ret; | 219 | return ret; |
220 | } | 220 | } |
221 | 221 | ||
222 | int nf_register_hook(struct nf_hook_ops *reg) | 222 | int nf_register_hook(struct nf_hook_ops *reg) |
223 | { | 223 | { |
224 | int ret; | 224 | int ret; |
225 | 225 | ||
226 | rtnl_lock(); | 226 | rtnl_lock(); |
227 | ret = _nf_register_hook(reg); | 227 | ret = _nf_register_hook(reg); |
228 | rtnl_unlock(); | 228 | rtnl_unlock(); |
229 | 229 | ||
230 | return ret; | 230 | return ret; |
231 | } | 231 | } |
232 | EXPORT_SYMBOL(nf_register_hook); | 232 | EXPORT_SYMBOL(nf_register_hook); |
233 | 233 | ||
234 | static void _nf_unregister_hook(struct nf_hook_ops *reg) | 234 | static void _nf_unregister_hook(struct nf_hook_ops *reg) |
235 | { | 235 | { |
236 | struct net *net; | 236 | struct net *net; |
237 | 237 | ||
238 | list_del(®->list); | 238 | list_del(®->list); |
239 | for_each_net(net) | 239 | for_each_net(net) |
240 | nf_unregister_net_hook(net, reg); | 240 | nf_unregister_net_hook(net, reg); |
241 | } | 241 | } |
242 | 242 | ||
243 | void nf_unregister_hook(struct nf_hook_ops *reg) | 243 | void nf_unregister_hook(struct nf_hook_ops *reg) |
244 | { | 244 | { |
245 | rtnl_lock(); | 245 | rtnl_lock(); |
246 | _nf_unregister_hook(reg); | 246 | _nf_unregister_hook(reg); |
247 | rtnl_unlock(); | 247 | rtnl_unlock(); |
248 | } | 248 | } |
249 | EXPORT_SYMBOL(nf_unregister_hook); | 249 | EXPORT_SYMBOL(nf_unregister_hook); |
250 | 250 | ||
251 | int nf_register_hooks(struct nf_hook_ops *reg, unsigned int n) | 251 | int nf_register_hooks(struct nf_hook_ops *reg, unsigned int n) |
252 | { | 252 | { |
253 | unsigned int i; | 253 | unsigned int i; |
254 | int err = 0; | 254 | int err = 0; |
255 | 255 | ||
256 | for (i = 0; i < n; i++) { | 256 | for (i = 0; i < n; i++) { |
257 | err = nf_register_hook(®[i]); | 257 | err = nf_register_hook(®[i]); |
258 | if (err) | 258 | if (err) |
259 | goto err; | 259 | goto err; |
260 | } | 260 | } |
261 | return err; | 261 | return err; |
262 | 262 | ||
263 | err: | 263 | err: |
264 | if (i > 0) | 264 | if (i > 0) |
265 | nf_unregister_hooks(reg, i); | 265 | nf_unregister_hooks(reg, i); |
266 | return err; | 266 | return err; |
267 | } | 267 | } |
268 | EXPORT_SYMBOL(nf_register_hooks); | 268 | EXPORT_SYMBOL(nf_register_hooks); |
269 | 269 | ||
270 | /* Caller MUST take rtnl_lock() */ | 270 | /* Caller MUST take rtnl_lock() */ |
271 | int _nf_register_hooks(struct nf_hook_ops *reg, unsigned int n) | 271 | int _nf_register_hooks(struct nf_hook_ops *reg, unsigned int n) |
272 | { | 272 | { |
273 | unsigned int i; | 273 | unsigned int i; |
274 | int err = 0; | 274 | int err = 0; |
275 | 275 | ||
276 | for (i = 0; i < n; i++) { | 276 | for (i = 0; i < n; i++) { |
277 | err = _nf_register_hook(®[i]); | 277 | err = _nf_register_hook(®[i]); |
278 | if (err) | 278 | if (err) |
279 | goto err; | 279 | goto err; |
280 | } | 280 | } |
281 | return err; | 281 | return err; |
282 | 282 | ||
283 | err: | 283 | err: |
284 | if (i > 0) | 284 | if (i > 0) |
285 | _nf_unregister_hooks(reg, i); | 285 | _nf_unregister_hooks(reg, i); |
286 | return err; | 286 | return err; |
287 | } | 287 | } |
288 | EXPORT_SYMBOL(_nf_register_hooks); | 288 | EXPORT_SYMBOL(_nf_register_hooks); |
289 | 289 | ||
290 | void nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n) | 290 | void nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n) |
291 | { | 291 | { |
292 | while (n-- > 0) | 292 | while (n-- > 0) |
293 | nf_unregister_hook(®[n]); | 293 | nf_unregister_hook(®[n]); |
294 | } | 294 | } |
295 | EXPORT_SYMBOL(nf_unregister_hooks); | 295 | EXPORT_SYMBOL(nf_unregister_hooks); |
296 | 296 | ||
297 | /* Caller MUST take rtnl_lock */ | 297 | /* Caller MUST take rtnl_lock */ |
298 | void _nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n) | 298 | void _nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n) |
299 | { | 299 | { |
300 | while (n-- > 0) | 300 | while (n-- > 0) |
301 | _nf_unregister_hook(®[n]); | 301 | _nf_unregister_hook(®[n]); |
302 | } | 302 | } |
303 | EXPORT_SYMBOL(_nf_unregister_hooks); | 303 | EXPORT_SYMBOL(_nf_unregister_hooks); |
304 | 304 | ||
305 | /* Returns 1 if okfn() needs to be executed by the caller, | 305 | /* Returns 1 if okfn() needs to be executed by the caller, |
306 | * -EPERM for NF_DROP, 0 otherwise. Caller must hold rcu_read_lock. */ | 306 | * -EPERM for NF_DROP, 0 otherwise. Caller must hold rcu_read_lock. */ |
307 | int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state, | 307 | int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state, |
308 | struct nf_hook_entry *entry) | 308 | struct nf_hook_entry *entry) |
309 | { | 309 | { |
310 | unsigned int verdict; | 310 | unsigned int verdict; |
311 | int ret; | 311 | int ret; |
312 | 312 | ||
313 | do { | 313 | do { |
314 | verdict = entry->ops.hook(entry->ops.priv, skb, state); | 314 | verdict = entry->ops.hook(entry->ops.priv, skb, state); |
315 | switch (verdict & NF_VERDICT_MASK) { | 315 | switch (verdict & NF_VERDICT_MASK) { |
316 | case NF_ACCEPT: | 316 | case NF_ACCEPT: |
317 | entry = rcu_dereference(entry->next); | 317 | entry = rcu_dereference(entry->next); |
318 | break; | 318 | break; |
319 | case NF_DROP: | 319 | case NF_DROP: |
320 | kfree_skb(skb); | 320 | kfree_skb(skb); |
321 | ret = NF_DROP_GETERR(verdict); | 321 | ret = NF_DROP_GETERR(verdict); |
322 | if (ret == 0) | 322 | if (ret == 0) |
323 | ret = -EPERM; | 323 | ret = -EPERM; |
324 | return ret; | 324 | return ret; |
325 | case NF_REPEAT: | ||
326 | continue; | ||
327 | case NF_QUEUE: | 325 | case NF_QUEUE: |
328 | ret = nf_queue(skb, state, &entry, verdict); | 326 | ret = nf_queue(skb, state, &entry, verdict); |
329 | if (ret == 1 && entry) | 327 | if (ret == 1 && entry) |
330 | continue; | 328 | continue; |
331 | return ret; | 329 | return ret; |
332 | default: | 330 | default: |
333 | /* Implicit handling for NF_STOLEN, as well as any other | 331 | /* Implicit handling for NF_STOLEN, as well as any other |
334 | * non conventional verdicts. | 332 | * non conventional verdicts. |
335 | */ | 333 | */ |
336 | return 0; | 334 | return 0; |
337 | } | 335 | } |
338 | } while (entry); | 336 | } while (entry); |
339 | 337 | ||
340 | return 1; | 338 | return 1; |
341 | } | 339 | } |
342 | EXPORT_SYMBOL(nf_hook_slow); | 340 | EXPORT_SYMBOL(nf_hook_slow); |
343 | 341 | ||
344 | 342 | ||
345 | int skb_make_writable(struct sk_buff *skb, unsigned int writable_len) | 343 | int skb_make_writable(struct sk_buff *skb, unsigned int writable_len) |
346 | { | 344 | { |
347 | if (writable_len > skb->len) | 345 | if (writable_len > skb->len) |
348 | return 0; | 346 | return 0; |
349 | 347 | ||
350 | /* Not exclusive use of packet? Must copy. */ | 348 | /* Not exclusive use of packet? Must copy. */ |
351 | if (!skb_cloned(skb)) { | 349 | if (!skb_cloned(skb)) { |
352 | if (writable_len <= skb_headlen(skb)) | 350 | if (writable_len <= skb_headlen(skb)) |
353 | return 1; | 351 | return 1; |
354 | } else if (skb_clone_writable(skb, writable_len)) | 352 | } else if (skb_clone_writable(skb, writable_len)) |
355 | return 1; | 353 | return 1; |
356 | 354 | ||
357 | if (writable_len <= skb_headlen(skb)) | 355 | if (writable_len <= skb_headlen(skb)) |
358 | writable_len = 0; | 356 | writable_len = 0; |
359 | else | 357 | else |
360 | writable_len -= skb_headlen(skb); | 358 | writable_len -= skb_headlen(skb); |
361 | 359 | ||
362 | return !!__pskb_pull_tail(skb, writable_len); | 360 | return !!__pskb_pull_tail(skb, writable_len); |
363 | } | 361 | } |
364 | EXPORT_SYMBOL(skb_make_writable); | 362 | EXPORT_SYMBOL(skb_make_writable); |
365 | 363 | ||
366 | /* This needs to be compiled in any case to avoid dependencies between the | 364 | /* This needs to be compiled in any case to avoid dependencies between the |
367 | * nfnetlink_queue code and nf_conntrack. | 365 | * nfnetlink_queue code and nf_conntrack. |
368 | */ | 366 | */ |
369 | struct nfnl_ct_hook __rcu *nfnl_ct_hook __read_mostly; | 367 | struct nfnl_ct_hook __rcu *nfnl_ct_hook __read_mostly; |
370 | EXPORT_SYMBOL_GPL(nfnl_ct_hook); | 368 | EXPORT_SYMBOL_GPL(nfnl_ct_hook); |
371 | 369 | ||
372 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) | 370 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
373 | /* This does not belong here, but locally generated errors need it if connection | 371 | /* This does not belong here, but locally generated errors need it if connection |
374 | tracking in use: without this, connection may not be in hash table, and hence | 372 | tracking in use: without this, connection may not be in hash table, and hence |
375 | manufactured ICMP or RST packets will not be associated with it. */ | 373 | manufactured ICMP or RST packets will not be associated with it. */ |
376 | void (*ip_ct_attach)(struct sk_buff *, const struct sk_buff *) | 374 | void (*ip_ct_attach)(struct sk_buff *, const struct sk_buff *) |
377 | __rcu __read_mostly; | 375 | __rcu __read_mostly; |
378 | EXPORT_SYMBOL(ip_ct_attach); | 376 | EXPORT_SYMBOL(ip_ct_attach); |
379 | 377 | ||
380 | void nf_ct_attach(struct sk_buff *new, const struct sk_buff *skb) | 378 | void nf_ct_attach(struct sk_buff *new, const struct sk_buff *skb) |
381 | { | 379 | { |
382 | void (*attach)(struct sk_buff *, const struct sk_buff *); | 380 | void (*attach)(struct sk_buff *, const struct sk_buff *); |
383 | 381 | ||
384 | if (skb->nfct) { | 382 | if (skb->nfct) { |
385 | rcu_read_lock(); | 383 | rcu_read_lock(); |
386 | attach = rcu_dereference(ip_ct_attach); | 384 | attach = rcu_dereference(ip_ct_attach); |
387 | if (attach) | 385 | if (attach) |
388 | attach(new, skb); | 386 | attach(new, skb); |
389 | rcu_read_unlock(); | 387 | rcu_read_unlock(); |
390 | } | 388 | } |
391 | } | 389 | } |
392 | EXPORT_SYMBOL(nf_ct_attach); | 390 | EXPORT_SYMBOL(nf_ct_attach); |
393 | 391 | ||
394 | void (*nf_ct_destroy)(struct nf_conntrack *) __rcu __read_mostly; | 392 | void (*nf_ct_destroy)(struct nf_conntrack *) __rcu __read_mostly; |
395 | EXPORT_SYMBOL(nf_ct_destroy); | 393 | EXPORT_SYMBOL(nf_ct_destroy); |
396 | 394 | ||
397 | void nf_conntrack_destroy(struct nf_conntrack *nfct) | 395 | void nf_conntrack_destroy(struct nf_conntrack *nfct) |
398 | { | 396 | { |
399 | void (*destroy)(struct nf_conntrack *); | 397 | void (*destroy)(struct nf_conntrack *); |
400 | 398 | ||
401 | rcu_read_lock(); | 399 | rcu_read_lock(); |
402 | destroy = rcu_dereference(nf_ct_destroy); | 400 | destroy = rcu_dereference(nf_ct_destroy); |
403 | BUG_ON(destroy == NULL); | 401 | BUG_ON(destroy == NULL); |
404 | destroy(nfct); | 402 | destroy(nfct); |
405 | rcu_read_unlock(); | 403 | rcu_read_unlock(); |
406 | } | 404 | } |
407 | EXPORT_SYMBOL(nf_conntrack_destroy); | 405 | EXPORT_SYMBOL(nf_conntrack_destroy); |
408 | 406 | ||
409 | /* Built-in default zone used e.g. by modules. */ | 407 | /* Built-in default zone used e.g. by modules. */ |
410 | const struct nf_conntrack_zone nf_ct_zone_dflt = { | 408 | const struct nf_conntrack_zone nf_ct_zone_dflt = { |
411 | .id = NF_CT_DEFAULT_ZONE_ID, | 409 | .id = NF_CT_DEFAULT_ZONE_ID, |
412 | .dir = NF_CT_DEFAULT_ZONE_DIR, | 410 | .dir = NF_CT_DEFAULT_ZONE_DIR, |
413 | }; | 411 | }; |
414 | EXPORT_SYMBOL_GPL(nf_ct_zone_dflt); | 412 | EXPORT_SYMBOL_GPL(nf_ct_zone_dflt); |
415 | #endif /* CONFIG_NF_CONNTRACK */ | 413 | #endif /* CONFIG_NF_CONNTRACK */ |
416 | 414 | ||
417 | #ifdef CONFIG_NF_NAT_NEEDED | 415 | #ifdef CONFIG_NF_NAT_NEEDED |
418 | void (*nf_nat_decode_session_hook)(struct sk_buff *, struct flowi *); | 416 | void (*nf_nat_decode_session_hook)(struct sk_buff *, struct flowi *); |
419 | EXPORT_SYMBOL(nf_nat_decode_session_hook); | 417 | EXPORT_SYMBOL(nf_nat_decode_session_hook); |
420 | #endif | 418 | #endif |
421 | 419 | ||
422 | static int nf_register_hook_list(struct net *net) | 420 | static int nf_register_hook_list(struct net *net) |
423 | { | 421 | { |
424 | struct nf_hook_ops *elem; | 422 | struct nf_hook_ops *elem; |
425 | int ret; | 423 | int ret; |
426 | 424 | ||
427 | rtnl_lock(); | 425 | rtnl_lock(); |
428 | list_for_each_entry(elem, &nf_hook_list, list) { | 426 | list_for_each_entry(elem, &nf_hook_list, list) { |
429 | ret = nf_register_net_hook(net, elem); | 427 | ret = nf_register_net_hook(net, elem); |
430 | if (ret && ret != -ENOENT) | 428 | if (ret && ret != -ENOENT) |
431 | goto out_undo; | 429 | goto out_undo; |
432 | } | 430 | } |
433 | rtnl_unlock(); | 431 | rtnl_unlock(); |
434 | return 0; | 432 | return 0; |
435 | 433 | ||
436 | out_undo: | 434 | out_undo: |
437 | list_for_each_entry_continue_reverse(elem, &nf_hook_list, list) | 435 | list_for_each_entry_continue_reverse(elem, &nf_hook_list, list) |
438 | nf_unregister_net_hook(net, elem); | 436 | nf_unregister_net_hook(net, elem); |
439 | rtnl_unlock(); | 437 | rtnl_unlock(); |
440 | return ret; | 438 | return ret; |
441 | } | 439 | } |
442 | 440 | ||
443 | static void nf_unregister_hook_list(struct net *net) | 441 | static void nf_unregister_hook_list(struct net *net) |
444 | { | 442 | { |
445 | struct nf_hook_ops *elem; | 443 | struct nf_hook_ops *elem; |
446 | 444 | ||
447 | rtnl_lock(); | 445 | rtnl_lock(); |
448 | list_for_each_entry(elem, &nf_hook_list, list) | 446 | list_for_each_entry(elem, &nf_hook_list, list) |
449 | nf_unregister_net_hook(net, elem); | 447 | nf_unregister_net_hook(net, elem); |
450 | rtnl_unlock(); | 448 | rtnl_unlock(); |
451 | } | 449 | } |
452 | 450 | ||
453 | static int __net_init netfilter_net_init(struct net *net) | 451 | static int __net_init netfilter_net_init(struct net *net) |
454 | { | 452 | { |
455 | int i, h, ret; | 453 | int i, h, ret; |
456 | 454 | ||
457 | for (i = 0; i < ARRAY_SIZE(net->nf.hooks); i++) { | 455 | for (i = 0; i < ARRAY_SIZE(net->nf.hooks); i++) { |
458 | for (h = 0; h < NF_MAX_HOOKS; h++) | 456 | for (h = 0; h < NF_MAX_HOOKS; h++) |
459 | RCU_INIT_POINTER(net->nf.hooks[i][h], NULL); | 457 | RCU_INIT_POINTER(net->nf.hooks[i][h], NULL); |
460 | } | 458 | } |
461 | 459 | ||
462 | #ifdef CONFIG_PROC_FS | 460 | #ifdef CONFIG_PROC_FS |
463 | net->nf.proc_netfilter = proc_net_mkdir(net, "netfilter", | 461 | net->nf.proc_netfilter = proc_net_mkdir(net, "netfilter", |
464 | net->proc_net); | 462 | net->proc_net); |
465 | if (!net->nf.proc_netfilter) { | 463 | if (!net->nf.proc_netfilter) { |
466 | if (!net_eq(net, &init_net)) | 464 | if (!net_eq(net, &init_net)) |
467 | pr_err("cannot create netfilter proc entry"); | 465 | pr_err("cannot create netfilter proc entry"); |
468 | 466 | ||
469 | return -ENOMEM; | 467 | return -ENOMEM; |
470 | } | 468 | } |
471 | #endif | 469 | #endif |
472 | ret = nf_register_hook_list(net); | 470 | ret = nf_register_hook_list(net); |
473 | if (ret) | 471 | if (ret) |
474 | remove_proc_entry("netfilter", net->proc_net); | 472 | remove_proc_entry("netfilter", net->proc_net); |
475 | 473 | ||
476 | return ret; | 474 | return ret; |
477 | } | 475 | } |
478 | 476 | ||
479 | static void __net_exit netfilter_net_exit(struct net *net) | 477 | static void __net_exit netfilter_net_exit(struct net *net) |
480 | { | 478 | { |
481 | nf_unregister_hook_list(net); | 479 | nf_unregister_hook_list(net); |
482 | remove_proc_entry("netfilter", net->proc_net); | 480 | remove_proc_entry("netfilter", net->proc_net); |
483 | } | 481 | } |
484 | 482 | ||
485 | static struct pernet_operations netfilter_net_ops = { | 483 | static struct pernet_operations netfilter_net_ops = { |
486 | .init = netfilter_net_init, | 484 | .init = netfilter_net_init, |
487 | .exit = netfilter_net_exit, | 485 | .exit = netfilter_net_exit, |
488 | }; | 486 | }; |
489 | 487 | ||
490 | int __init netfilter_init(void) | 488 | int __init netfilter_init(void) |
491 | { | 489 | { |
492 | int ret; | 490 | int ret; |
493 | 491 | ||
494 | ret = register_pernet_subsys(&netfilter_net_ops); | 492 | ret = register_pernet_subsys(&netfilter_net_ops); |
495 | if (ret < 0) | 493 | if (ret < 0) |
496 | goto err; | 494 | goto err; |
497 | 495 | ||
498 | ret = netfilter_log_init(); | 496 | ret = netfilter_log_init(); |
499 | if (ret < 0) | 497 | if (ret < 0) |
500 | goto err_pernet; | 498 | goto err_pernet; |
501 | 499 | ||
502 | return 0; | 500 | return 0; |
503 | err_pernet: | 501 | err_pernet: |
504 | unregister_pernet_subsys(&netfilter_net_ops); | 502 | unregister_pernet_subsys(&netfilter_net_ops); |
505 | err: | 503 | err: |
506 | return ret; | 504 | return ret; |
507 | } | 505 | } |
508 | 506 |
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 | * (C) 2005-2012 Patrick McHardy <kaber@trash.net> | 8 | * (C) 2005-2012 Patrick McHardy <kaber@trash.net> |
9 | * | 9 | * |
10 | * This program is free software; you can redistribute it and/or modify | 10 | * This program is free software; you can redistribute it and/or modify |
11 | * it under the terms of the GNU General Public License version 2 as | 11 | * it under the terms of the GNU General Public License version 2 as |
12 | * published by the Free Software Foundation. | 12 | * published by the Free Software Foundation. |
13 | */ | 13 | */ |
14 | 14 | ||
15 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | 15 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
16 | 16 | ||
17 | #include <linux/types.h> | 17 | #include <linux/types.h> |
18 | #include <linux/netfilter.h> | 18 | #include <linux/netfilter.h> |
19 | #include <linux/module.h> | 19 | #include <linux/module.h> |
20 | #include <linux/sched.h> | 20 | #include <linux/sched.h> |
21 | #include <linux/skbuff.h> | 21 | #include <linux/skbuff.h> |
22 | #include <linux/proc_fs.h> | 22 | #include <linux/proc_fs.h> |
23 | #include <linux/vmalloc.h> | 23 | #include <linux/vmalloc.h> |
24 | #include <linux/stddef.h> | 24 | #include <linux/stddef.h> |
25 | #include <linux/slab.h> | 25 | #include <linux/slab.h> |
26 | #include <linux/random.h> | 26 | #include <linux/random.h> |
27 | #include <linux/jhash.h> | 27 | #include <linux/jhash.h> |
28 | #include <linux/err.h> | 28 | #include <linux/err.h> |
29 | #include <linux/percpu.h> | 29 | #include <linux/percpu.h> |
30 | #include <linux/moduleparam.h> | 30 | #include <linux/moduleparam.h> |
31 | #include <linux/notifier.h> | 31 | #include <linux/notifier.h> |
32 | #include <linux/kernel.h> | 32 | #include <linux/kernel.h> |
33 | #include <linux/netdevice.h> | 33 | #include <linux/netdevice.h> |
34 | #include <linux/socket.h> | 34 | #include <linux/socket.h> |
35 | #include <linux/mm.h> | 35 | #include <linux/mm.h> |
36 | #include <linux/nsproxy.h> | 36 | #include <linux/nsproxy.h> |
37 | #include <linux/rculist_nulls.h> | 37 | #include <linux/rculist_nulls.h> |
38 | 38 | ||
39 | #include <net/netfilter/nf_conntrack.h> | 39 | #include <net/netfilter/nf_conntrack.h> |
40 | #include <net/netfilter/nf_conntrack_l3proto.h> | 40 | #include <net/netfilter/nf_conntrack_l3proto.h> |
41 | #include <net/netfilter/nf_conntrack_l4proto.h> | 41 | #include <net/netfilter/nf_conntrack_l4proto.h> |
42 | #include <net/netfilter/nf_conntrack_expect.h> | 42 | #include <net/netfilter/nf_conntrack_expect.h> |
43 | #include <net/netfilter/nf_conntrack_helper.h> | 43 | #include <net/netfilter/nf_conntrack_helper.h> |
44 | #include <net/netfilter/nf_conntrack_seqadj.h> | 44 | #include <net/netfilter/nf_conntrack_seqadj.h> |
45 | #include <net/netfilter/nf_conntrack_core.h> | 45 | #include <net/netfilter/nf_conntrack_core.h> |
46 | #include <net/netfilter/nf_conntrack_extend.h> | 46 | #include <net/netfilter/nf_conntrack_extend.h> |
47 | #include <net/netfilter/nf_conntrack_acct.h> | 47 | #include <net/netfilter/nf_conntrack_acct.h> |
48 | #include <net/netfilter/nf_conntrack_ecache.h> | 48 | #include <net/netfilter/nf_conntrack_ecache.h> |
49 | #include <net/netfilter/nf_conntrack_zones.h> | 49 | #include <net/netfilter/nf_conntrack_zones.h> |
50 | #include <net/netfilter/nf_conntrack_timestamp.h> | 50 | #include <net/netfilter/nf_conntrack_timestamp.h> |
51 | #include <net/netfilter/nf_conntrack_timeout.h> | 51 | #include <net/netfilter/nf_conntrack_timeout.h> |
52 | #include <net/netfilter/nf_conntrack_labels.h> | 52 | #include <net/netfilter/nf_conntrack_labels.h> |
53 | #include <net/netfilter/nf_conntrack_synproxy.h> | 53 | #include <net/netfilter/nf_conntrack_synproxy.h> |
54 | #include <net/netfilter/nf_nat.h> | 54 | #include <net/netfilter/nf_nat.h> |
55 | #include <net/netfilter/nf_nat_core.h> | 55 | #include <net/netfilter/nf_nat_core.h> |
56 | #include <net/netfilter/nf_nat_helper.h> | 56 | #include <net/netfilter/nf_nat_helper.h> |
57 | #include <net/netns/hash.h> | 57 | #include <net/netns/hash.h> |
58 | 58 | ||
59 | #define NF_CONNTRACK_VERSION "0.5.0" | 59 | #define NF_CONNTRACK_VERSION "0.5.0" |
60 | 60 | ||
61 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, | 61 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
62 | enum nf_nat_manip_type manip, | 62 | enum nf_nat_manip_type manip, |
63 | const struct nlattr *attr) __read_mostly; | 63 | const struct nlattr *attr) __read_mostly; |
64 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); | 64 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
65 | 65 | ||
66 | __cacheline_aligned_in_smp spinlock_t nf_conntrack_locks[CONNTRACK_LOCKS]; | 66 | __cacheline_aligned_in_smp spinlock_t nf_conntrack_locks[CONNTRACK_LOCKS]; |
67 | EXPORT_SYMBOL_GPL(nf_conntrack_locks); | 67 | EXPORT_SYMBOL_GPL(nf_conntrack_locks); |
68 | 68 | ||
69 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(nf_conntrack_expect_lock); | 69 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(nf_conntrack_expect_lock); |
70 | EXPORT_SYMBOL_GPL(nf_conntrack_expect_lock); | 70 | EXPORT_SYMBOL_GPL(nf_conntrack_expect_lock); |
71 | 71 | ||
72 | struct hlist_nulls_head *nf_conntrack_hash __read_mostly; | 72 | struct hlist_nulls_head *nf_conntrack_hash __read_mostly; |
73 | EXPORT_SYMBOL_GPL(nf_conntrack_hash); | 73 | EXPORT_SYMBOL_GPL(nf_conntrack_hash); |
74 | 74 | ||
75 | struct conntrack_gc_work { | 75 | struct conntrack_gc_work { |
76 | struct delayed_work dwork; | 76 | struct delayed_work dwork; |
77 | u32 last_bucket; | 77 | u32 last_bucket; |
78 | bool exiting; | 78 | bool exiting; |
79 | }; | 79 | }; |
80 | 80 | ||
81 | static __read_mostly struct kmem_cache *nf_conntrack_cachep; | 81 | static __read_mostly struct kmem_cache *nf_conntrack_cachep; |
82 | static __read_mostly spinlock_t nf_conntrack_locks_all_lock; | 82 | static __read_mostly spinlock_t nf_conntrack_locks_all_lock; |
83 | static __read_mostly DEFINE_SPINLOCK(nf_conntrack_locks_all_lock); | 83 | static __read_mostly DEFINE_SPINLOCK(nf_conntrack_locks_all_lock); |
84 | static __read_mostly bool nf_conntrack_locks_all; | 84 | static __read_mostly bool nf_conntrack_locks_all; |
85 | 85 | ||
86 | #define GC_MAX_BUCKETS_DIV 64u | 86 | #define GC_MAX_BUCKETS_DIV 64u |
87 | #define GC_MAX_BUCKETS 8192u | 87 | #define GC_MAX_BUCKETS 8192u |
88 | #define GC_INTERVAL (5 * HZ) | 88 | #define GC_INTERVAL (5 * HZ) |
89 | #define GC_MAX_EVICTS 256u | 89 | #define GC_MAX_EVICTS 256u |
90 | 90 | ||
91 | static struct conntrack_gc_work conntrack_gc_work; | 91 | static struct conntrack_gc_work conntrack_gc_work; |
92 | 92 | ||
93 | void nf_conntrack_lock(spinlock_t *lock) __acquires(lock) | 93 | void nf_conntrack_lock(spinlock_t *lock) __acquires(lock) |
94 | { | 94 | { |
95 | spin_lock(lock); | 95 | spin_lock(lock); |
96 | while (unlikely(nf_conntrack_locks_all)) { | 96 | while (unlikely(nf_conntrack_locks_all)) { |
97 | spin_unlock(lock); | 97 | spin_unlock(lock); |
98 | 98 | ||
99 | /* | 99 | /* |
100 | * Order the 'nf_conntrack_locks_all' load vs. the | 100 | * Order the 'nf_conntrack_locks_all' load vs. the |
101 | * spin_unlock_wait() loads below, to ensure | 101 | * spin_unlock_wait() loads below, to ensure |
102 | * that 'nf_conntrack_locks_all_lock' is indeed held: | 102 | * that 'nf_conntrack_locks_all_lock' is indeed held: |
103 | */ | 103 | */ |
104 | smp_rmb(); /* spin_lock(&nf_conntrack_locks_all_lock) */ | 104 | smp_rmb(); /* spin_lock(&nf_conntrack_locks_all_lock) */ |
105 | spin_unlock_wait(&nf_conntrack_locks_all_lock); | 105 | spin_unlock_wait(&nf_conntrack_locks_all_lock); |
106 | spin_lock(lock); | 106 | spin_lock(lock); |
107 | } | 107 | } |
108 | } | 108 | } |
109 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); | 109 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
110 | 110 | ||
111 | static void nf_conntrack_double_unlock(unsigned int h1, unsigned int h2) | 111 | static void nf_conntrack_double_unlock(unsigned int h1, unsigned int h2) |
112 | { | 112 | { |
113 | h1 %= CONNTRACK_LOCKS; | 113 | h1 %= CONNTRACK_LOCKS; |
114 | h2 %= CONNTRACK_LOCKS; | 114 | h2 %= CONNTRACK_LOCKS; |
115 | spin_unlock(&nf_conntrack_locks[h1]); | 115 | spin_unlock(&nf_conntrack_locks[h1]); |
116 | if (h1 != h2) | 116 | if (h1 != h2) |
117 | spin_unlock(&nf_conntrack_locks[h2]); | 117 | spin_unlock(&nf_conntrack_locks[h2]); |
118 | } | 118 | } |
119 | 119 | ||
120 | /* return true if we need to recompute hashes (in case hash table was resized) */ | 120 | /* return true if we need to recompute hashes (in case hash table was resized) */ |
121 | static bool nf_conntrack_double_lock(struct net *net, unsigned int h1, | 121 | static bool nf_conntrack_double_lock(struct net *net, unsigned int h1, |
122 | unsigned int h2, unsigned int sequence) | 122 | unsigned int h2, unsigned int sequence) |
123 | { | 123 | { |
124 | h1 %= CONNTRACK_LOCKS; | 124 | h1 %= CONNTRACK_LOCKS; |
125 | h2 %= CONNTRACK_LOCKS; | 125 | h2 %= CONNTRACK_LOCKS; |
126 | if (h1 <= h2) { | 126 | if (h1 <= h2) { |
127 | nf_conntrack_lock(&nf_conntrack_locks[h1]); | 127 | nf_conntrack_lock(&nf_conntrack_locks[h1]); |
128 | if (h1 != h2) | 128 | if (h1 != h2) |
129 | spin_lock_nested(&nf_conntrack_locks[h2], | 129 | spin_lock_nested(&nf_conntrack_locks[h2], |
130 | SINGLE_DEPTH_NESTING); | 130 | SINGLE_DEPTH_NESTING); |
131 | } else { | 131 | } else { |
132 | nf_conntrack_lock(&nf_conntrack_locks[h2]); | 132 | nf_conntrack_lock(&nf_conntrack_locks[h2]); |
133 | spin_lock_nested(&nf_conntrack_locks[h1], | 133 | spin_lock_nested(&nf_conntrack_locks[h1], |
134 | SINGLE_DEPTH_NESTING); | 134 | SINGLE_DEPTH_NESTING); |
135 | } | 135 | } |
136 | if (read_seqcount_retry(&nf_conntrack_generation, sequence)) { | 136 | if (read_seqcount_retry(&nf_conntrack_generation, sequence)) { |
137 | nf_conntrack_double_unlock(h1, h2); | 137 | nf_conntrack_double_unlock(h1, h2); |
138 | return true; | 138 | return true; |
139 | } | 139 | } |
140 | return false; | 140 | return false; |
141 | } | 141 | } |
142 | 142 | ||
143 | static void nf_conntrack_all_lock(void) | 143 | static void nf_conntrack_all_lock(void) |
144 | { | 144 | { |
145 | int i; | 145 | int i; |
146 | 146 | ||
147 | spin_lock(&nf_conntrack_locks_all_lock); | 147 | spin_lock(&nf_conntrack_locks_all_lock); |
148 | nf_conntrack_locks_all = true; | 148 | nf_conntrack_locks_all = true; |
149 | 149 | ||
150 | /* | 150 | /* |
151 | * Order the above store of 'nf_conntrack_locks_all' against | 151 | * Order the above store of 'nf_conntrack_locks_all' against |
152 | * the spin_unlock_wait() loads below, such that if | 152 | * the spin_unlock_wait() loads below, such that if |
153 | * nf_conntrack_lock() observes 'nf_conntrack_locks_all' | 153 | * nf_conntrack_lock() observes 'nf_conntrack_locks_all' |
154 | * we must observe nf_conntrack_locks[] held: | 154 | * we must observe nf_conntrack_locks[] held: |
155 | */ | 155 | */ |
156 | smp_mb(); /* spin_lock(&nf_conntrack_locks_all_lock) */ | 156 | smp_mb(); /* spin_lock(&nf_conntrack_locks_all_lock) */ |
157 | 157 | ||
158 | for (i = 0; i < CONNTRACK_LOCKS; i++) { | 158 | for (i = 0; i < CONNTRACK_LOCKS; i++) { |
159 | spin_unlock_wait(&nf_conntrack_locks[i]); | 159 | spin_unlock_wait(&nf_conntrack_locks[i]); |
160 | } | 160 | } |
161 | } | 161 | } |
162 | 162 | ||
163 | static void nf_conntrack_all_unlock(void) | 163 | static void nf_conntrack_all_unlock(void) |
164 | { | 164 | { |
165 | /* | 165 | /* |
166 | * All prior stores must be complete before we clear | 166 | * All prior stores must be complete before we clear |
167 | * 'nf_conntrack_locks_all'. Otherwise nf_conntrack_lock() | 167 | * 'nf_conntrack_locks_all'. Otherwise nf_conntrack_lock() |
168 | * might observe the false value but not the entire | 168 | * might observe the false value but not the entire |
169 | * critical section: | 169 | * critical section: |
170 | */ | 170 | */ |
171 | smp_store_release(&nf_conntrack_locks_all, false); | 171 | smp_store_release(&nf_conntrack_locks_all, false); |
172 | spin_unlock(&nf_conntrack_locks_all_lock); | 172 | spin_unlock(&nf_conntrack_locks_all_lock); |
173 | } | 173 | } |
174 | 174 | ||
175 | unsigned int nf_conntrack_htable_size __read_mostly; | 175 | unsigned int nf_conntrack_htable_size __read_mostly; |
176 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); | 176 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
177 | 177 | ||
178 | unsigned int nf_conntrack_max __read_mostly; | 178 | unsigned int nf_conntrack_max __read_mostly; |
179 | seqcount_t nf_conntrack_generation __read_mostly; | 179 | seqcount_t nf_conntrack_generation __read_mostly; |
180 | 180 | ||
181 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); | 181 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
182 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); | 182 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); |
183 | 183 | ||
184 | static unsigned int nf_conntrack_hash_rnd __read_mostly; | 184 | static unsigned int nf_conntrack_hash_rnd __read_mostly; |
185 | 185 | ||
186 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, | 186 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, |
187 | const struct net *net) | 187 | const struct net *net) |
188 | { | 188 | { |
189 | unsigned int n; | 189 | unsigned int n; |
190 | u32 seed; | 190 | u32 seed; |
191 | 191 | ||
192 | get_random_once(&nf_conntrack_hash_rnd, sizeof(nf_conntrack_hash_rnd)); | 192 | get_random_once(&nf_conntrack_hash_rnd, sizeof(nf_conntrack_hash_rnd)); |
193 | 193 | ||
194 | /* The direction must be ignored, so we hash everything up to the | 194 | /* The direction must be ignored, so we hash everything up to the |
195 | * destination ports (which is a multiple of 4) and treat the last | 195 | * destination ports (which is a multiple of 4) and treat the last |
196 | * three bytes manually. | 196 | * three bytes manually. |
197 | */ | 197 | */ |
198 | seed = nf_conntrack_hash_rnd ^ net_hash_mix(net); | 198 | seed = nf_conntrack_hash_rnd ^ net_hash_mix(net); |
199 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | 199 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); |
200 | return jhash2((u32 *)tuple, n, seed ^ | 200 | return jhash2((u32 *)tuple, n, seed ^ |
201 | (((__force __u16)tuple->dst.u.all << 16) | | 201 | (((__force __u16)tuple->dst.u.all << 16) | |
202 | tuple->dst.protonum)); | 202 | tuple->dst.protonum)); |
203 | } | 203 | } |
204 | 204 | ||
205 | static u32 scale_hash(u32 hash) | 205 | static u32 scale_hash(u32 hash) |
206 | { | 206 | { |
207 | return reciprocal_scale(hash, nf_conntrack_htable_size); | 207 | return reciprocal_scale(hash, nf_conntrack_htable_size); |
208 | } | 208 | } |
209 | 209 | ||
210 | static u32 __hash_conntrack(const struct net *net, | 210 | static u32 __hash_conntrack(const struct net *net, |
211 | const struct nf_conntrack_tuple *tuple, | 211 | const struct nf_conntrack_tuple *tuple, |
212 | unsigned int size) | 212 | unsigned int size) |
213 | { | 213 | { |
214 | return reciprocal_scale(hash_conntrack_raw(tuple, net), size); | 214 | return reciprocal_scale(hash_conntrack_raw(tuple, net), size); |
215 | } | 215 | } |
216 | 216 | ||
217 | static u32 hash_conntrack(const struct net *net, | 217 | static u32 hash_conntrack(const struct net *net, |
218 | const struct nf_conntrack_tuple *tuple) | 218 | const struct nf_conntrack_tuple *tuple) |
219 | { | 219 | { |
220 | return scale_hash(hash_conntrack_raw(tuple, net)); | 220 | return scale_hash(hash_conntrack_raw(tuple, net)); |
221 | } | 221 | } |
222 | 222 | ||
223 | bool | 223 | bool |
224 | nf_ct_get_tuple(const struct sk_buff *skb, | 224 | nf_ct_get_tuple(const struct sk_buff *skb, |
225 | unsigned int nhoff, | 225 | unsigned int nhoff, |
226 | unsigned int dataoff, | 226 | unsigned int dataoff, |
227 | u_int16_t l3num, | 227 | u_int16_t l3num, |
228 | u_int8_t protonum, | 228 | u_int8_t protonum, |
229 | struct net *net, | 229 | struct net *net, |
230 | struct nf_conntrack_tuple *tuple, | 230 | struct nf_conntrack_tuple *tuple, |
231 | const struct nf_conntrack_l3proto *l3proto, | 231 | const struct nf_conntrack_l3proto *l3proto, |
232 | const struct nf_conntrack_l4proto *l4proto) | 232 | const struct nf_conntrack_l4proto *l4proto) |
233 | { | 233 | { |
234 | memset(tuple, 0, sizeof(*tuple)); | 234 | memset(tuple, 0, sizeof(*tuple)); |
235 | 235 | ||
236 | tuple->src.l3num = l3num; | 236 | tuple->src.l3num = l3num; |
237 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | 237 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) |
238 | return false; | 238 | return false; |
239 | 239 | ||
240 | tuple->dst.protonum = protonum; | 240 | tuple->dst.protonum = protonum; |
241 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | 241 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; |
242 | 242 | ||
243 | return l4proto->pkt_to_tuple(skb, dataoff, net, tuple); | 243 | return l4proto->pkt_to_tuple(skb, dataoff, net, tuple); |
244 | } | 244 | } |
245 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); | 245 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
246 | 246 | ||
247 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, | 247 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
248 | u_int16_t l3num, | 248 | u_int16_t l3num, |
249 | struct net *net, struct nf_conntrack_tuple *tuple) | 249 | struct net *net, struct nf_conntrack_tuple *tuple) |
250 | { | 250 | { |
251 | struct nf_conntrack_l3proto *l3proto; | 251 | struct nf_conntrack_l3proto *l3proto; |
252 | struct nf_conntrack_l4proto *l4proto; | 252 | struct nf_conntrack_l4proto *l4proto; |
253 | unsigned int protoff; | 253 | unsigned int protoff; |
254 | u_int8_t protonum; | 254 | u_int8_t protonum; |
255 | int ret; | 255 | int ret; |
256 | 256 | ||
257 | rcu_read_lock(); | 257 | rcu_read_lock(); |
258 | 258 | ||
259 | l3proto = __nf_ct_l3proto_find(l3num); | 259 | l3proto = __nf_ct_l3proto_find(l3num); |
260 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | 260 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); |
261 | if (ret != NF_ACCEPT) { | 261 | if (ret != NF_ACCEPT) { |
262 | rcu_read_unlock(); | 262 | rcu_read_unlock(); |
263 | return false; | 263 | return false; |
264 | } | 264 | } |
265 | 265 | ||
266 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | 266 | l4proto = __nf_ct_l4proto_find(l3num, protonum); |
267 | 267 | ||
268 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, net, tuple, | 268 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, net, tuple, |
269 | l3proto, l4proto); | 269 | l3proto, l4proto); |
270 | 270 | ||
271 | rcu_read_unlock(); | 271 | rcu_read_unlock(); |
272 | return ret; | 272 | return ret; |
273 | } | 273 | } |
274 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | 274 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); |
275 | 275 | ||
276 | bool | 276 | bool |
277 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, | 277 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
278 | const struct nf_conntrack_tuple *orig, | 278 | const struct nf_conntrack_tuple *orig, |
279 | const struct nf_conntrack_l3proto *l3proto, | 279 | const struct nf_conntrack_l3proto *l3proto, |
280 | const struct nf_conntrack_l4proto *l4proto) | 280 | const struct nf_conntrack_l4proto *l4proto) |
281 | { | 281 | { |
282 | memset(inverse, 0, sizeof(*inverse)); | 282 | memset(inverse, 0, sizeof(*inverse)); |
283 | 283 | ||
284 | inverse->src.l3num = orig->src.l3num; | 284 | inverse->src.l3num = orig->src.l3num; |
285 | if (l3proto->invert_tuple(inverse, orig) == 0) | 285 | if (l3proto->invert_tuple(inverse, orig) == 0) |
286 | return false; | 286 | return false; |
287 | 287 | ||
288 | inverse->dst.dir = !orig->dst.dir; | 288 | inverse->dst.dir = !orig->dst.dir; |
289 | 289 | ||
290 | inverse->dst.protonum = orig->dst.protonum; | 290 | inverse->dst.protonum = orig->dst.protonum; |
291 | return l4proto->invert_tuple(inverse, orig); | 291 | return l4proto->invert_tuple(inverse, orig); |
292 | } | 292 | } |
293 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); | 293 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
294 | 294 | ||
295 | static void | 295 | static void |
296 | clean_from_lists(struct nf_conn *ct) | 296 | clean_from_lists(struct nf_conn *ct) |
297 | { | 297 | { |
298 | pr_debug("clean_from_lists(%p)\n", ct); | 298 | pr_debug("clean_from_lists(%p)\n", ct); |
299 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | 299 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
300 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | 300 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); |
301 | 301 | ||
302 | /* Destroy all pending expectations */ | 302 | /* Destroy all pending expectations */ |
303 | nf_ct_remove_expectations(ct); | 303 | nf_ct_remove_expectations(ct); |
304 | } | 304 | } |
305 | 305 | ||
306 | /* must be called with local_bh_disable */ | 306 | /* must be called with local_bh_disable */ |
307 | static void nf_ct_add_to_dying_list(struct nf_conn *ct) | 307 | static void nf_ct_add_to_dying_list(struct nf_conn *ct) |
308 | { | 308 | { |
309 | struct ct_pcpu *pcpu; | 309 | struct ct_pcpu *pcpu; |
310 | 310 | ||
311 | /* add this conntrack to the (per cpu) dying list */ | 311 | /* add this conntrack to the (per cpu) dying list */ |
312 | ct->cpu = smp_processor_id(); | 312 | ct->cpu = smp_processor_id(); |
313 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | 313 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); |
314 | 314 | ||
315 | spin_lock(&pcpu->lock); | 315 | spin_lock(&pcpu->lock); |
316 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | 316 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
317 | &pcpu->dying); | 317 | &pcpu->dying); |
318 | spin_unlock(&pcpu->lock); | 318 | spin_unlock(&pcpu->lock); |
319 | } | 319 | } |
320 | 320 | ||
321 | /* must be called with local_bh_disable */ | 321 | /* must be called with local_bh_disable */ |
322 | static void nf_ct_add_to_unconfirmed_list(struct nf_conn *ct) | 322 | static void nf_ct_add_to_unconfirmed_list(struct nf_conn *ct) |
323 | { | 323 | { |
324 | struct ct_pcpu *pcpu; | 324 | struct ct_pcpu *pcpu; |
325 | 325 | ||
326 | /* add this conntrack to the (per cpu) unconfirmed list */ | 326 | /* add this conntrack to the (per cpu) unconfirmed list */ |
327 | ct->cpu = smp_processor_id(); | 327 | ct->cpu = smp_processor_id(); |
328 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | 328 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); |
329 | 329 | ||
330 | spin_lock(&pcpu->lock); | 330 | spin_lock(&pcpu->lock); |
331 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | 331 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
332 | &pcpu->unconfirmed); | 332 | &pcpu->unconfirmed); |
333 | spin_unlock(&pcpu->lock); | 333 | spin_unlock(&pcpu->lock); |
334 | } | 334 | } |
335 | 335 | ||
336 | /* must be called with local_bh_disable */ | 336 | /* must be called with local_bh_disable */ |
337 | static void nf_ct_del_from_dying_or_unconfirmed_list(struct nf_conn *ct) | 337 | static void nf_ct_del_from_dying_or_unconfirmed_list(struct nf_conn *ct) |
338 | { | 338 | { |
339 | struct ct_pcpu *pcpu; | 339 | struct ct_pcpu *pcpu; |
340 | 340 | ||
341 | /* We overload first tuple to link into unconfirmed or dying list.*/ | 341 | /* We overload first tuple to link into unconfirmed or dying list.*/ |
342 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | 342 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); |
343 | 343 | ||
344 | spin_lock(&pcpu->lock); | 344 | spin_lock(&pcpu->lock); |
345 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); | 345 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); |
346 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | 346 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
347 | spin_unlock(&pcpu->lock); | 347 | spin_unlock(&pcpu->lock); |
348 | } | 348 | } |
349 | 349 | ||
350 | /* Released via destroy_conntrack() */ | 350 | /* Released via destroy_conntrack() */ |
351 | struct nf_conn *nf_ct_tmpl_alloc(struct net *net, | 351 | struct nf_conn *nf_ct_tmpl_alloc(struct net *net, |
352 | const struct nf_conntrack_zone *zone, | 352 | const struct nf_conntrack_zone *zone, |
353 | gfp_t flags) | 353 | gfp_t flags) |
354 | { | 354 | { |
355 | struct nf_conn *tmpl; | 355 | struct nf_conn *tmpl; |
356 | 356 | ||
357 | tmpl = kzalloc(sizeof(*tmpl), flags); | 357 | tmpl = kzalloc(sizeof(*tmpl), flags); |
358 | if (tmpl == NULL) | 358 | if (tmpl == NULL) |
359 | return NULL; | 359 | return NULL; |
360 | 360 | ||
361 | tmpl->status = IPS_TEMPLATE; | 361 | tmpl->status = IPS_TEMPLATE; |
362 | write_pnet(&tmpl->ct_net, net); | 362 | write_pnet(&tmpl->ct_net, net); |
363 | nf_ct_zone_add(tmpl, zone); | 363 | nf_ct_zone_add(tmpl, zone); |
364 | atomic_set(&tmpl->ct_general.use, 0); | 364 | atomic_set(&tmpl->ct_general.use, 0); |
365 | 365 | ||
366 | return tmpl; | 366 | return tmpl; |
367 | } | 367 | } |
368 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_alloc); | 368 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_alloc); |
369 | 369 | ||
370 | void nf_ct_tmpl_free(struct nf_conn *tmpl) | 370 | void nf_ct_tmpl_free(struct nf_conn *tmpl) |
371 | { | 371 | { |
372 | nf_ct_ext_destroy(tmpl); | 372 | nf_ct_ext_destroy(tmpl); |
373 | nf_ct_ext_free(tmpl); | 373 | nf_ct_ext_free(tmpl); |
374 | kfree(tmpl); | 374 | kfree(tmpl); |
375 | } | 375 | } |
376 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_free); | 376 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_free); |
377 | 377 | ||
378 | static void | 378 | static void |
379 | destroy_conntrack(struct nf_conntrack *nfct) | 379 | destroy_conntrack(struct nf_conntrack *nfct) |
380 | { | 380 | { |
381 | struct nf_conn *ct = (struct nf_conn *)nfct; | 381 | struct nf_conn *ct = (struct nf_conn *)nfct; |
382 | struct nf_conntrack_l4proto *l4proto; | 382 | struct nf_conntrack_l4proto *l4proto; |
383 | 383 | ||
384 | pr_debug("destroy_conntrack(%p)\n", ct); | 384 | pr_debug("destroy_conntrack(%p)\n", ct); |
385 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); | 385 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
386 | 386 | ||
387 | if (unlikely(nf_ct_is_template(ct))) { | 387 | if (unlikely(nf_ct_is_template(ct))) { |
388 | nf_ct_tmpl_free(ct); | 388 | nf_ct_tmpl_free(ct); |
389 | return; | 389 | return; |
390 | } | 390 | } |
391 | rcu_read_lock(); | 391 | rcu_read_lock(); |
392 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); | 392 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
393 | if (l4proto->destroy) | 393 | if (l4proto->destroy) |
394 | l4proto->destroy(ct); | 394 | l4proto->destroy(ct); |
395 | 395 | ||
396 | rcu_read_unlock(); | 396 | rcu_read_unlock(); |
397 | 397 | ||
398 | local_bh_disable(); | 398 | local_bh_disable(); |
399 | /* Expectations will have been removed in clean_from_lists, | 399 | /* Expectations will have been removed in clean_from_lists, |
400 | * except TFTP can create an expectation on the first packet, | 400 | * except TFTP can create an expectation on the first packet, |
401 | * before connection is in the list, so we need to clean here, | 401 | * before connection is in the list, so we need to clean here, |
402 | * too. | 402 | * too. |
403 | */ | 403 | */ |
404 | nf_ct_remove_expectations(ct); | 404 | nf_ct_remove_expectations(ct); |
405 | 405 | ||
406 | nf_ct_del_from_dying_or_unconfirmed_list(ct); | 406 | nf_ct_del_from_dying_or_unconfirmed_list(ct); |
407 | 407 | ||
408 | local_bh_enable(); | 408 | local_bh_enable(); |
409 | 409 | ||
410 | if (ct->master) | 410 | if (ct->master) |
411 | nf_ct_put(ct->master); | 411 | nf_ct_put(ct->master); |
412 | 412 | ||
413 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); | 413 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
414 | nf_conntrack_free(ct); | 414 | nf_conntrack_free(ct); |
415 | } | 415 | } |
416 | 416 | ||
417 | static void nf_ct_delete_from_lists(struct nf_conn *ct) | 417 | static void nf_ct_delete_from_lists(struct nf_conn *ct) |
418 | { | 418 | { |
419 | struct net *net = nf_ct_net(ct); | 419 | struct net *net = nf_ct_net(ct); |
420 | unsigned int hash, reply_hash; | 420 | unsigned int hash, reply_hash; |
421 | unsigned int sequence; | 421 | unsigned int sequence; |
422 | 422 | ||
423 | nf_ct_helper_destroy(ct); | 423 | nf_ct_helper_destroy(ct); |
424 | 424 | ||
425 | local_bh_disable(); | 425 | local_bh_disable(); |
426 | do { | 426 | do { |
427 | sequence = read_seqcount_begin(&nf_conntrack_generation); | 427 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
428 | hash = hash_conntrack(net, | 428 | hash = hash_conntrack(net, |
429 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | 429 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
430 | reply_hash = hash_conntrack(net, | 430 | reply_hash = hash_conntrack(net, |
431 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | 431 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
432 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | 432 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); |
433 | 433 | ||
434 | clean_from_lists(ct); | 434 | clean_from_lists(ct); |
435 | nf_conntrack_double_unlock(hash, reply_hash); | 435 | nf_conntrack_double_unlock(hash, reply_hash); |
436 | 436 | ||
437 | nf_ct_add_to_dying_list(ct); | 437 | nf_ct_add_to_dying_list(ct); |
438 | 438 | ||
439 | local_bh_enable(); | 439 | local_bh_enable(); |
440 | } | 440 | } |
441 | 441 | ||
442 | bool nf_ct_delete(struct nf_conn *ct, u32 portid, int report) | 442 | bool nf_ct_delete(struct nf_conn *ct, u32 portid, int report) |
443 | { | 443 | { |
444 | struct nf_conn_tstamp *tstamp; | 444 | struct nf_conn_tstamp *tstamp; |
445 | 445 | ||
446 | if (test_and_set_bit(IPS_DYING_BIT, &ct->status)) | 446 | if (test_and_set_bit(IPS_DYING_BIT, &ct->status)) |
447 | return false; | 447 | return false; |
448 | 448 | ||
449 | tstamp = nf_conn_tstamp_find(ct); | 449 | tstamp = nf_conn_tstamp_find(ct); |
450 | if (tstamp && tstamp->stop == 0) | 450 | if (tstamp && tstamp->stop == 0) |
451 | tstamp->stop = ktime_get_real_ns(); | 451 | tstamp->stop = ktime_get_real_ns(); |
452 | 452 | ||
453 | if (nf_conntrack_event_report(IPCT_DESTROY, ct, | 453 | if (nf_conntrack_event_report(IPCT_DESTROY, ct, |
454 | portid, report) < 0) { | 454 | portid, report) < 0) { |
455 | /* destroy event was not delivered. nf_ct_put will | 455 | /* destroy event was not delivered. nf_ct_put will |
456 | * be done by event cache worker on redelivery. | 456 | * be done by event cache worker on redelivery. |
457 | */ | 457 | */ |
458 | nf_ct_delete_from_lists(ct); | 458 | nf_ct_delete_from_lists(ct); |
459 | nf_conntrack_ecache_delayed_work(nf_ct_net(ct)); | 459 | nf_conntrack_ecache_delayed_work(nf_ct_net(ct)); |
460 | return false; | 460 | return false; |
461 | } | 461 | } |
462 | 462 | ||
463 | nf_conntrack_ecache_work(nf_ct_net(ct)); | 463 | nf_conntrack_ecache_work(nf_ct_net(ct)); |
464 | nf_ct_delete_from_lists(ct); | 464 | nf_ct_delete_from_lists(ct); |
465 | nf_ct_put(ct); | 465 | nf_ct_put(ct); |
466 | return true; | 466 | return true; |
467 | } | 467 | } |
468 | EXPORT_SYMBOL_GPL(nf_ct_delete); | 468 | EXPORT_SYMBOL_GPL(nf_ct_delete); |
469 | 469 | ||
470 | static inline bool | 470 | static inline bool |
471 | nf_ct_key_equal(struct nf_conntrack_tuple_hash *h, | 471 | nf_ct_key_equal(struct nf_conntrack_tuple_hash *h, |
472 | const struct nf_conntrack_tuple *tuple, | 472 | const struct nf_conntrack_tuple *tuple, |
473 | const struct nf_conntrack_zone *zone, | 473 | const struct nf_conntrack_zone *zone, |
474 | const struct net *net) | 474 | const struct net *net) |
475 | { | 475 | { |
476 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | 476 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); |
477 | 477 | ||
478 | /* A conntrack can be recreated with the equal tuple, | 478 | /* A conntrack can be recreated with the equal tuple, |
479 | * so we need to check that the conntrack is confirmed | 479 | * so we need to check that the conntrack is confirmed |
480 | */ | 480 | */ |
481 | return nf_ct_tuple_equal(tuple, &h->tuple) && | 481 | return nf_ct_tuple_equal(tuple, &h->tuple) && |
482 | nf_ct_zone_equal(ct, zone, NF_CT_DIRECTION(h)) && | 482 | nf_ct_zone_equal(ct, zone, NF_CT_DIRECTION(h)) && |
483 | nf_ct_is_confirmed(ct) && | 483 | nf_ct_is_confirmed(ct) && |
484 | net_eq(net, nf_ct_net(ct)); | 484 | net_eq(net, nf_ct_net(ct)); |
485 | } | 485 | } |
486 | 486 | ||
487 | /* caller must hold rcu readlock and none of the nf_conntrack_locks */ | 487 | /* caller must hold rcu readlock and none of the nf_conntrack_locks */ |
488 | static void nf_ct_gc_expired(struct nf_conn *ct) | 488 | static void nf_ct_gc_expired(struct nf_conn *ct) |
489 | { | 489 | { |
490 | if (!atomic_inc_not_zero(&ct->ct_general.use)) | 490 | if (!atomic_inc_not_zero(&ct->ct_general.use)) |
491 | return; | 491 | return; |
492 | 492 | ||
493 | if (nf_ct_should_gc(ct)) | 493 | if (nf_ct_should_gc(ct)) |
494 | nf_ct_kill(ct); | 494 | nf_ct_kill(ct); |
495 | 495 | ||
496 | nf_ct_put(ct); | 496 | nf_ct_put(ct); |
497 | } | 497 | } |
498 | 498 | ||
499 | /* | 499 | /* |
500 | * Warning : | 500 | * Warning : |
501 | * - Caller must take a reference on returned object | 501 | * - Caller must take a reference on returned object |
502 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | 502 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) |
503 | */ | 503 | */ |
504 | static struct nf_conntrack_tuple_hash * | 504 | static struct nf_conntrack_tuple_hash * |
505 | ____nf_conntrack_find(struct net *net, const struct nf_conntrack_zone *zone, | 505 | ____nf_conntrack_find(struct net *net, const struct nf_conntrack_zone *zone, |
506 | const struct nf_conntrack_tuple *tuple, u32 hash) | 506 | const struct nf_conntrack_tuple *tuple, u32 hash) |
507 | { | 507 | { |
508 | struct nf_conntrack_tuple_hash *h; | 508 | struct nf_conntrack_tuple_hash *h; |
509 | struct hlist_nulls_head *ct_hash; | 509 | struct hlist_nulls_head *ct_hash; |
510 | struct hlist_nulls_node *n; | 510 | struct hlist_nulls_node *n; |
511 | unsigned int bucket, hsize; | 511 | unsigned int bucket, hsize; |
512 | 512 | ||
513 | begin: | 513 | begin: |
514 | nf_conntrack_get_ht(&ct_hash, &hsize); | 514 | nf_conntrack_get_ht(&ct_hash, &hsize); |
515 | bucket = reciprocal_scale(hash, hsize); | 515 | bucket = reciprocal_scale(hash, hsize); |
516 | 516 | ||
517 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[bucket], hnnode) { | 517 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[bucket], hnnode) { |
518 | struct nf_conn *ct; | 518 | struct nf_conn *ct; |
519 | 519 | ||
520 | ct = nf_ct_tuplehash_to_ctrack(h); | 520 | ct = nf_ct_tuplehash_to_ctrack(h); |
521 | if (nf_ct_is_expired(ct)) { | 521 | if (nf_ct_is_expired(ct)) { |
522 | nf_ct_gc_expired(ct); | 522 | nf_ct_gc_expired(ct); |
523 | continue; | 523 | continue; |
524 | } | 524 | } |
525 | 525 | ||
526 | if (nf_ct_is_dying(ct)) | 526 | if (nf_ct_is_dying(ct)) |
527 | continue; | 527 | continue; |
528 | 528 | ||
529 | if (nf_ct_key_equal(h, tuple, zone, net)) | 529 | if (nf_ct_key_equal(h, tuple, zone, net)) |
530 | return h; | 530 | return h; |
531 | } | 531 | } |
532 | /* | 532 | /* |
533 | * if the nulls value we got at the end of this lookup is | 533 | * if the nulls value we got at the end of this lookup is |
534 | * not the expected one, we must restart lookup. | 534 | * not the expected one, we must restart lookup. |
535 | * We probably met an item that was moved to another chain. | 535 | * We probably met an item that was moved to another chain. |
536 | */ | 536 | */ |
537 | if (get_nulls_value(n) != bucket) { | 537 | if (get_nulls_value(n) != bucket) { |
538 | NF_CT_STAT_INC_ATOMIC(net, search_restart); | 538 | NF_CT_STAT_INC_ATOMIC(net, search_restart); |
539 | goto begin; | 539 | goto begin; |
540 | } | 540 | } |
541 | 541 | ||
542 | return NULL; | 542 | return NULL; |
543 | } | 543 | } |
544 | 544 | ||
545 | /* Find a connection corresponding to a tuple. */ | 545 | /* Find a connection corresponding to a tuple. */ |
546 | static struct nf_conntrack_tuple_hash * | 546 | static struct nf_conntrack_tuple_hash * |
547 | __nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, | 547 | __nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, |
548 | const struct nf_conntrack_tuple *tuple, u32 hash) | 548 | const struct nf_conntrack_tuple *tuple, u32 hash) |
549 | { | 549 | { |
550 | struct nf_conntrack_tuple_hash *h; | 550 | struct nf_conntrack_tuple_hash *h; |
551 | struct nf_conn *ct; | 551 | struct nf_conn *ct; |
552 | 552 | ||
553 | rcu_read_lock(); | 553 | rcu_read_lock(); |
554 | begin: | 554 | begin: |
555 | h = ____nf_conntrack_find(net, zone, tuple, hash); | 555 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
556 | if (h) { | 556 | if (h) { |
557 | ct = nf_ct_tuplehash_to_ctrack(h); | 557 | ct = nf_ct_tuplehash_to_ctrack(h); |
558 | if (unlikely(nf_ct_is_dying(ct) || | 558 | if (unlikely(nf_ct_is_dying(ct) || |
559 | !atomic_inc_not_zero(&ct->ct_general.use))) | 559 | !atomic_inc_not_zero(&ct->ct_general.use))) |
560 | h = NULL; | 560 | h = NULL; |
561 | else { | 561 | else { |
562 | if (unlikely(!nf_ct_key_equal(h, tuple, zone, net))) { | 562 | if (unlikely(!nf_ct_key_equal(h, tuple, zone, net))) { |
563 | nf_ct_put(ct); | 563 | nf_ct_put(ct); |
564 | goto begin; | 564 | goto begin; |
565 | } | 565 | } |
566 | } | 566 | } |
567 | } | 567 | } |
568 | rcu_read_unlock(); | 568 | rcu_read_unlock(); |
569 | 569 | ||
570 | return h; | 570 | return h; |
571 | } | 571 | } |
572 | 572 | ||
573 | struct nf_conntrack_tuple_hash * | 573 | struct nf_conntrack_tuple_hash * |
574 | nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, | 574 | nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, |
575 | const struct nf_conntrack_tuple *tuple) | 575 | const struct nf_conntrack_tuple *tuple) |
576 | { | 576 | { |
577 | return __nf_conntrack_find_get(net, zone, tuple, | 577 | return __nf_conntrack_find_get(net, zone, tuple, |
578 | hash_conntrack_raw(tuple, net)); | 578 | hash_conntrack_raw(tuple, net)); |
579 | } | 579 | } |
580 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); | 580 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
581 | 581 | ||
582 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, | 582 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
583 | unsigned int hash, | 583 | unsigned int hash, |
584 | unsigned int reply_hash) | 584 | unsigned int reply_hash) |
585 | { | 585 | { |
586 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | 586 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
587 | &nf_conntrack_hash[hash]); | 587 | &nf_conntrack_hash[hash]); |
588 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, | 588 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
589 | &nf_conntrack_hash[reply_hash]); | 589 | &nf_conntrack_hash[reply_hash]); |
590 | } | 590 | } |
591 | 591 | ||
592 | int | 592 | int |
593 | nf_conntrack_hash_check_insert(struct nf_conn *ct) | 593 | nf_conntrack_hash_check_insert(struct nf_conn *ct) |
594 | { | 594 | { |
595 | const struct nf_conntrack_zone *zone; | 595 | const struct nf_conntrack_zone *zone; |
596 | struct net *net = nf_ct_net(ct); | 596 | struct net *net = nf_ct_net(ct); |
597 | unsigned int hash, reply_hash; | 597 | unsigned int hash, reply_hash; |
598 | struct nf_conntrack_tuple_hash *h; | 598 | struct nf_conntrack_tuple_hash *h; |
599 | struct hlist_nulls_node *n; | 599 | struct hlist_nulls_node *n; |
600 | unsigned int sequence; | 600 | unsigned int sequence; |
601 | 601 | ||
602 | zone = nf_ct_zone(ct); | 602 | zone = nf_ct_zone(ct); |
603 | 603 | ||
604 | local_bh_disable(); | 604 | local_bh_disable(); |
605 | do { | 605 | do { |
606 | sequence = read_seqcount_begin(&nf_conntrack_generation); | 606 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
607 | hash = hash_conntrack(net, | 607 | hash = hash_conntrack(net, |
608 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | 608 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
609 | reply_hash = hash_conntrack(net, | 609 | reply_hash = hash_conntrack(net, |
610 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | 610 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
611 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | 611 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); |
612 | 612 | ||
613 | /* See if there's one in the list already, including reverse */ | 613 | /* See if there's one in the list already, including reverse */ |
614 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) | 614 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) |
615 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | 615 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
616 | zone, net)) | 616 | zone, net)) |
617 | goto out; | 617 | goto out; |
618 | 618 | ||
619 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) | 619 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) |
620 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, | 620 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
621 | zone, net)) | 621 | zone, net)) |
622 | goto out; | 622 | goto out; |
623 | 623 | ||
624 | smp_wmb(); | 624 | smp_wmb(); |
625 | /* The caller holds a reference to this object */ | 625 | /* The caller holds a reference to this object */ |
626 | atomic_set(&ct->ct_general.use, 2); | 626 | atomic_set(&ct->ct_general.use, 2); |
627 | __nf_conntrack_hash_insert(ct, hash, reply_hash); | 627 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
628 | nf_conntrack_double_unlock(hash, reply_hash); | 628 | nf_conntrack_double_unlock(hash, reply_hash); |
629 | NF_CT_STAT_INC(net, insert); | 629 | NF_CT_STAT_INC(net, insert); |
630 | local_bh_enable(); | 630 | local_bh_enable(); |
631 | return 0; | 631 | return 0; |
632 | 632 | ||
633 | out: | 633 | out: |
634 | nf_conntrack_double_unlock(hash, reply_hash); | 634 | nf_conntrack_double_unlock(hash, reply_hash); |
635 | NF_CT_STAT_INC(net, insert_failed); | 635 | NF_CT_STAT_INC(net, insert_failed); |
636 | local_bh_enable(); | 636 | local_bh_enable(); |
637 | return -EEXIST; | 637 | return -EEXIST; |
638 | } | 638 | } |
639 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); | 639 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); |
640 | 640 | ||
641 | static inline void nf_ct_acct_update(struct nf_conn *ct, | 641 | static inline void nf_ct_acct_update(struct nf_conn *ct, |
642 | enum ip_conntrack_info ctinfo, | 642 | enum ip_conntrack_info ctinfo, |
643 | unsigned int len) | 643 | unsigned int len) |
644 | { | 644 | { |
645 | struct nf_conn_acct *acct; | 645 | struct nf_conn_acct *acct; |
646 | 646 | ||
647 | acct = nf_conn_acct_find(ct); | 647 | acct = nf_conn_acct_find(ct); |
648 | if (acct) { | 648 | if (acct) { |
649 | struct nf_conn_counter *counter = acct->counter; | 649 | struct nf_conn_counter *counter = acct->counter; |
650 | 650 | ||
651 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | 651 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); |
652 | atomic64_add(len, &counter[CTINFO2DIR(ctinfo)].bytes); | 652 | atomic64_add(len, &counter[CTINFO2DIR(ctinfo)].bytes); |
653 | } | 653 | } |
654 | } | 654 | } |
655 | 655 | ||
656 | static void nf_ct_acct_merge(struct nf_conn *ct, enum ip_conntrack_info ctinfo, | 656 | static void nf_ct_acct_merge(struct nf_conn *ct, enum ip_conntrack_info ctinfo, |
657 | const struct nf_conn *loser_ct) | 657 | const struct nf_conn *loser_ct) |
658 | { | 658 | { |
659 | struct nf_conn_acct *acct; | 659 | struct nf_conn_acct *acct; |
660 | 660 | ||
661 | acct = nf_conn_acct_find(loser_ct); | 661 | acct = nf_conn_acct_find(loser_ct); |
662 | if (acct) { | 662 | if (acct) { |
663 | struct nf_conn_counter *counter = acct->counter; | 663 | struct nf_conn_counter *counter = acct->counter; |
664 | unsigned int bytes; | 664 | unsigned int bytes; |
665 | 665 | ||
666 | /* u32 should be fine since we must have seen one packet. */ | 666 | /* u32 should be fine since we must have seen one packet. */ |
667 | bytes = atomic64_read(&counter[CTINFO2DIR(ctinfo)].bytes); | 667 | bytes = atomic64_read(&counter[CTINFO2DIR(ctinfo)].bytes); |
668 | nf_ct_acct_update(ct, ctinfo, bytes); | 668 | nf_ct_acct_update(ct, ctinfo, bytes); |
669 | } | 669 | } |
670 | } | 670 | } |
671 | 671 | ||
672 | /* Resolve race on insertion if this protocol allows this. */ | 672 | /* Resolve race on insertion if this protocol allows this. */ |
673 | static int nf_ct_resolve_clash(struct net *net, struct sk_buff *skb, | 673 | static int nf_ct_resolve_clash(struct net *net, struct sk_buff *skb, |
674 | enum ip_conntrack_info ctinfo, | 674 | enum ip_conntrack_info ctinfo, |
675 | struct nf_conntrack_tuple_hash *h) | 675 | struct nf_conntrack_tuple_hash *h) |
676 | { | 676 | { |
677 | /* This is the conntrack entry already in hashes that won race. */ | 677 | /* This is the conntrack entry already in hashes that won race. */ |
678 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | 678 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); |
679 | struct nf_conntrack_l4proto *l4proto; | 679 | struct nf_conntrack_l4proto *l4proto; |
680 | 680 | ||
681 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); | 681 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
682 | if (l4proto->allow_clash && | 682 | if (l4proto->allow_clash && |
683 | !nfct_nat(ct) && | 683 | !nfct_nat(ct) && |
684 | !nf_ct_is_dying(ct) && | 684 | !nf_ct_is_dying(ct) && |
685 | atomic_inc_not_zero(&ct->ct_general.use)) { | 685 | atomic_inc_not_zero(&ct->ct_general.use)) { |
686 | nf_ct_acct_merge(ct, ctinfo, (struct nf_conn *)skb->nfct); | 686 | nf_ct_acct_merge(ct, ctinfo, (struct nf_conn *)skb->nfct); |
687 | nf_conntrack_put(skb->nfct); | 687 | nf_conntrack_put(skb->nfct); |
688 | /* Assign conntrack already in hashes to this skbuff. Don't | 688 | /* Assign conntrack already in hashes to this skbuff. Don't |
689 | * modify skb->nfctinfo to ensure consistent stateful filtering. | 689 | * modify skb->nfctinfo to ensure consistent stateful filtering. |
690 | */ | 690 | */ |
691 | skb->nfct = &ct->ct_general; | 691 | skb->nfct = &ct->ct_general; |
692 | return NF_ACCEPT; | 692 | return NF_ACCEPT; |
693 | } | 693 | } |
694 | NF_CT_STAT_INC(net, drop); | 694 | NF_CT_STAT_INC(net, drop); |
695 | return NF_DROP; | 695 | return NF_DROP; |
696 | } | 696 | } |
697 | 697 | ||
698 | /* Confirm a connection given skb; places it in hash table */ | 698 | /* Confirm a connection given skb; places it in hash table */ |
699 | int | 699 | int |
700 | __nf_conntrack_confirm(struct sk_buff *skb) | 700 | __nf_conntrack_confirm(struct sk_buff *skb) |
701 | { | 701 | { |
702 | const struct nf_conntrack_zone *zone; | 702 | const struct nf_conntrack_zone *zone; |
703 | unsigned int hash, reply_hash; | 703 | unsigned int hash, reply_hash; |
704 | struct nf_conntrack_tuple_hash *h; | 704 | struct nf_conntrack_tuple_hash *h; |
705 | struct nf_conn *ct; | 705 | struct nf_conn *ct; |
706 | struct nf_conn_help *help; | 706 | struct nf_conn_help *help; |
707 | struct nf_conn_tstamp *tstamp; | 707 | struct nf_conn_tstamp *tstamp; |
708 | struct hlist_nulls_node *n; | 708 | struct hlist_nulls_node *n; |
709 | enum ip_conntrack_info ctinfo; | 709 | enum ip_conntrack_info ctinfo; |
710 | struct net *net; | 710 | struct net *net; |
711 | unsigned int sequence; | 711 | unsigned int sequence; |
712 | int ret = NF_DROP; | 712 | int ret = NF_DROP; |
713 | 713 | ||
714 | ct = nf_ct_get(skb, &ctinfo); | 714 | ct = nf_ct_get(skb, &ctinfo); |
715 | net = nf_ct_net(ct); | 715 | net = nf_ct_net(ct); |
716 | 716 | ||
717 | /* ipt_REJECT uses nf_conntrack_attach to attach related | 717 | /* ipt_REJECT uses nf_conntrack_attach to attach related |
718 | ICMP/TCP RST packets in other direction. Actual packet | 718 | ICMP/TCP RST packets in other direction. Actual packet |
719 | which created connection will be IP_CT_NEW or for an | 719 | which created connection will be IP_CT_NEW or for an |
720 | expected connection, IP_CT_RELATED. */ | 720 | expected connection, IP_CT_RELATED. */ |
721 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | 721 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) |
722 | return NF_ACCEPT; | 722 | return NF_ACCEPT; |
723 | 723 | ||
724 | zone = nf_ct_zone(ct); | 724 | zone = nf_ct_zone(ct); |
725 | local_bh_disable(); | 725 | local_bh_disable(); |
726 | 726 | ||
727 | do { | 727 | do { |
728 | sequence = read_seqcount_begin(&nf_conntrack_generation); | 728 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
729 | /* reuse the hash saved before */ | 729 | /* reuse the hash saved before */ |
730 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; | 730 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; |
731 | hash = scale_hash(hash); | 731 | hash = scale_hash(hash); |
732 | reply_hash = hash_conntrack(net, | 732 | reply_hash = hash_conntrack(net, |
733 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | 733 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
734 | 734 | ||
735 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | 735 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); |
736 | 736 | ||
737 | /* We're not in hash table, and we refuse to set up related | 737 | /* We're not in hash table, and we refuse to set up related |
738 | * connections for unconfirmed conns. But packet copies and | 738 | * connections for unconfirmed conns. But packet copies and |
739 | * REJECT will give spurious warnings here. | 739 | * REJECT will give spurious warnings here. |
740 | */ | 740 | */ |
741 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ | 741 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ |
742 | 742 | ||
743 | /* No external references means no one else could have | 743 | /* No external references means no one else could have |
744 | * confirmed us. | 744 | * confirmed us. |
745 | */ | 745 | */ |
746 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | 746 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
747 | pr_debug("Confirming conntrack %p\n", ct); | 747 | pr_debug("Confirming conntrack %p\n", ct); |
748 | /* We have to check the DYING flag after unlink to prevent | 748 | /* We have to check the DYING flag after unlink to prevent |
749 | * a race against nf_ct_get_next_corpse() possibly called from | 749 | * a race against nf_ct_get_next_corpse() possibly called from |
750 | * user context, else we insert an already 'dead' hash, blocking | 750 | * user context, else we insert an already 'dead' hash, blocking |
751 | * further use of that particular connection -JM. | 751 | * further use of that particular connection -JM. |
752 | */ | 752 | */ |
753 | nf_ct_del_from_dying_or_unconfirmed_list(ct); | 753 | nf_ct_del_from_dying_or_unconfirmed_list(ct); |
754 | 754 | ||
755 | if (unlikely(nf_ct_is_dying(ct))) { | 755 | if (unlikely(nf_ct_is_dying(ct))) { |
756 | nf_ct_add_to_dying_list(ct); | 756 | nf_ct_add_to_dying_list(ct); |
757 | goto dying; | 757 | goto dying; |
758 | } | 758 | } |
759 | 759 | ||
760 | /* See if there's one in the list already, including reverse: | 760 | /* See if there's one in the list already, including reverse: |
761 | NAT could have grabbed it without realizing, since we're | 761 | NAT could have grabbed it without realizing, since we're |
762 | not in the hash. If there is, we lost race. */ | 762 | not in the hash. If there is, we lost race. */ |
763 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) | 763 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) |
764 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | 764 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
765 | zone, net)) | 765 | zone, net)) |
766 | goto out; | 766 | goto out; |
767 | 767 | ||
768 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) | 768 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) |
769 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, | 769 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
770 | zone, net)) | 770 | zone, net)) |
771 | goto out; | 771 | goto out; |
772 | 772 | ||
773 | /* Timer relative to confirmation time, not original | 773 | /* Timer relative to confirmation time, not original |
774 | setting time, otherwise we'd get timer wrap in | 774 | setting time, otherwise we'd get timer wrap in |
775 | weird delay cases. */ | 775 | weird delay cases. */ |
776 | ct->timeout += nfct_time_stamp; | 776 | ct->timeout += nfct_time_stamp; |
777 | atomic_inc(&ct->ct_general.use); | 777 | atomic_inc(&ct->ct_general.use); |
778 | ct->status |= IPS_CONFIRMED; | 778 | ct->status |= IPS_CONFIRMED; |
779 | 779 | ||
780 | /* set conntrack timestamp, if enabled. */ | 780 | /* set conntrack timestamp, if enabled. */ |
781 | tstamp = nf_conn_tstamp_find(ct); | 781 | tstamp = nf_conn_tstamp_find(ct); |
782 | if (tstamp) { | 782 | if (tstamp) { |
783 | if (skb->tstamp.tv64 == 0) | 783 | if (skb->tstamp.tv64 == 0) |
784 | __net_timestamp(skb); | 784 | __net_timestamp(skb); |
785 | 785 | ||
786 | tstamp->start = ktime_to_ns(skb->tstamp); | 786 | tstamp->start = ktime_to_ns(skb->tstamp); |
787 | } | 787 | } |
788 | /* Since the lookup is lockless, hash insertion must be done after | 788 | /* Since the lookup is lockless, hash insertion must be done after |
789 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | 789 | * starting the timer and setting the CONFIRMED bit. The RCU barriers |
790 | * guarantee that no other CPU can find the conntrack before the above | 790 | * guarantee that no other CPU can find the conntrack before the above |
791 | * stores are visible. | 791 | * stores are visible. |
792 | */ | 792 | */ |
793 | __nf_conntrack_hash_insert(ct, hash, reply_hash); | 793 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
794 | nf_conntrack_double_unlock(hash, reply_hash); | 794 | nf_conntrack_double_unlock(hash, reply_hash); |
795 | local_bh_enable(); | 795 | local_bh_enable(); |
796 | 796 | ||
797 | help = nfct_help(ct); | 797 | help = nfct_help(ct); |
798 | if (help && help->helper) | 798 | if (help && help->helper) |
799 | nf_conntrack_event_cache(IPCT_HELPER, ct); | 799 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
800 | 800 | ||
801 | nf_conntrack_event_cache(master_ct(ct) ? | 801 | nf_conntrack_event_cache(master_ct(ct) ? |
802 | IPCT_RELATED : IPCT_NEW, ct); | 802 | IPCT_RELATED : IPCT_NEW, ct); |
803 | return NF_ACCEPT; | 803 | return NF_ACCEPT; |
804 | 804 | ||
805 | out: | 805 | out: |
806 | nf_ct_add_to_dying_list(ct); | 806 | nf_ct_add_to_dying_list(ct); |
807 | ret = nf_ct_resolve_clash(net, skb, ctinfo, h); | 807 | ret = nf_ct_resolve_clash(net, skb, ctinfo, h); |
808 | dying: | 808 | dying: |
809 | nf_conntrack_double_unlock(hash, reply_hash); | 809 | nf_conntrack_double_unlock(hash, reply_hash); |
810 | NF_CT_STAT_INC(net, insert_failed); | 810 | NF_CT_STAT_INC(net, insert_failed); |
811 | local_bh_enable(); | 811 | local_bh_enable(); |
812 | return ret; | 812 | return ret; |
813 | } | 813 | } |
814 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); | 814 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
815 | 815 | ||
816 | /* Returns true if a connection correspondings to the tuple (required | 816 | /* Returns true if a connection correspondings to the tuple (required |
817 | for NAT). */ | 817 | for NAT). */ |
818 | int | 818 | int |
819 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | 819 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, |
820 | const struct nf_conn *ignored_conntrack) | 820 | const struct nf_conn *ignored_conntrack) |
821 | { | 821 | { |
822 | struct net *net = nf_ct_net(ignored_conntrack); | 822 | struct net *net = nf_ct_net(ignored_conntrack); |
823 | const struct nf_conntrack_zone *zone; | 823 | const struct nf_conntrack_zone *zone; |
824 | struct nf_conntrack_tuple_hash *h; | 824 | struct nf_conntrack_tuple_hash *h; |
825 | struct hlist_nulls_head *ct_hash; | 825 | struct hlist_nulls_head *ct_hash; |
826 | unsigned int hash, hsize; | 826 | unsigned int hash, hsize; |
827 | struct hlist_nulls_node *n; | 827 | struct hlist_nulls_node *n; |
828 | struct nf_conn *ct; | 828 | struct nf_conn *ct; |
829 | 829 | ||
830 | zone = nf_ct_zone(ignored_conntrack); | 830 | zone = nf_ct_zone(ignored_conntrack); |
831 | 831 | ||
832 | rcu_read_lock(); | 832 | rcu_read_lock(); |
833 | begin: | 833 | begin: |
834 | nf_conntrack_get_ht(&ct_hash, &hsize); | 834 | nf_conntrack_get_ht(&ct_hash, &hsize); |
835 | hash = __hash_conntrack(net, tuple, hsize); | 835 | hash = __hash_conntrack(net, tuple, hsize); |
836 | 836 | ||
837 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[hash], hnnode) { | 837 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[hash], hnnode) { |
838 | ct = nf_ct_tuplehash_to_ctrack(h); | 838 | ct = nf_ct_tuplehash_to_ctrack(h); |
839 | 839 | ||
840 | if (ct == ignored_conntrack) | 840 | if (ct == ignored_conntrack) |
841 | continue; | 841 | continue; |
842 | 842 | ||
843 | if (nf_ct_is_expired(ct)) { | 843 | if (nf_ct_is_expired(ct)) { |
844 | nf_ct_gc_expired(ct); | 844 | nf_ct_gc_expired(ct); |
845 | continue; | 845 | continue; |
846 | } | 846 | } |
847 | 847 | ||
848 | if (nf_ct_key_equal(h, tuple, zone, net)) { | 848 | if (nf_ct_key_equal(h, tuple, zone, net)) { |
849 | NF_CT_STAT_INC_ATOMIC(net, found); | 849 | NF_CT_STAT_INC_ATOMIC(net, found); |
850 | rcu_read_unlock(); | 850 | rcu_read_unlock(); |
851 | return 1; | 851 | return 1; |
852 | } | 852 | } |
853 | } | 853 | } |
854 | 854 | ||
855 | if (get_nulls_value(n) != hash) { | 855 | if (get_nulls_value(n) != hash) { |
856 | NF_CT_STAT_INC_ATOMIC(net, search_restart); | 856 | NF_CT_STAT_INC_ATOMIC(net, search_restart); |
857 | goto begin; | 857 | goto begin; |
858 | } | 858 | } |
859 | 859 | ||
860 | rcu_read_unlock(); | 860 | rcu_read_unlock(); |
861 | 861 | ||
862 | return 0; | 862 | return 0; |
863 | } | 863 | } |
864 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); | 864 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
865 | 865 | ||
866 | #define NF_CT_EVICTION_RANGE 8 | 866 | #define NF_CT_EVICTION_RANGE 8 |
867 | 867 | ||
868 | /* There's a small race here where we may free a just-assured | 868 | /* There's a small race here where we may free a just-assured |
869 | connection. Too bad: we're in trouble anyway. */ | 869 | connection. Too bad: we're in trouble anyway. */ |
870 | static unsigned int early_drop_list(struct net *net, | 870 | static unsigned int early_drop_list(struct net *net, |
871 | struct hlist_nulls_head *head) | 871 | struct hlist_nulls_head *head) |
872 | { | 872 | { |
873 | struct nf_conntrack_tuple_hash *h; | 873 | struct nf_conntrack_tuple_hash *h; |
874 | struct hlist_nulls_node *n; | 874 | struct hlist_nulls_node *n; |
875 | unsigned int drops = 0; | 875 | unsigned int drops = 0; |
876 | struct nf_conn *tmp; | 876 | struct nf_conn *tmp; |
877 | 877 | ||
878 | hlist_nulls_for_each_entry_rcu(h, n, head, hnnode) { | 878 | hlist_nulls_for_each_entry_rcu(h, n, head, hnnode) { |
879 | tmp = nf_ct_tuplehash_to_ctrack(h); | 879 | tmp = nf_ct_tuplehash_to_ctrack(h); |
880 | 880 | ||
881 | if (nf_ct_is_expired(tmp)) { | 881 | if (nf_ct_is_expired(tmp)) { |
882 | nf_ct_gc_expired(tmp); | 882 | nf_ct_gc_expired(tmp); |
883 | continue; | 883 | continue; |
884 | } | 884 | } |
885 | 885 | ||
886 | if (test_bit(IPS_ASSURED_BIT, &tmp->status) || | 886 | if (test_bit(IPS_ASSURED_BIT, &tmp->status) || |
887 | !net_eq(nf_ct_net(tmp), net) || | 887 | !net_eq(nf_ct_net(tmp), net) || |
888 | nf_ct_is_dying(tmp)) | 888 | nf_ct_is_dying(tmp)) |
889 | continue; | 889 | continue; |
890 | 890 | ||
891 | if (!atomic_inc_not_zero(&tmp->ct_general.use)) | 891 | if (!atomic_inc_not_zero(&tmp->ct_general.use)) |
892 | continue; | 892 | continue; |
893 | 893 | ||
894 | /* kill only if still in same netns -- might have moved due to | 894 | /* kill only if still in same netns -- might have moved due to |
895 | * SLAB_DESTROY_BY_RCU rules. | 895 | * SLAB_DESTROY_BY_RCU rules. |
896 | * | 896 | * |
897 | * We steal the timer reference. If that fails timer has | 897 | * We steal the timer reference. If that fails timer has |
898 | * already fired or someone else deleted it. Just drop ref | 898 | * already fired or someone else deleted it. Just drop ref |
899 | * and move to next entry. | 899 | * and move to next entry. |
900 | */ | 900 | */ |
901 | if (net_eq(nf_ct_net(tmp), net) && | 901 | if (net_eq(nf_ct_net(tmp), net) && |
902 | nf_ct_is_confirmed(tmp) && | 902 | nf_ct_is_confirmed(tmp) && |
903 | nf_ct_delete(tmp, 0, 0)) | 903 | nf_ct_delete(tmp, 0, 0)) |
904 | drops++; | 904 | drops++; |
905 | 905 | ||
906 | nf_ct_put(tmp); | 906 | nf_ct_put(tmp); |
907 | } | 907 | } |
908 | 908 | ||
909 | return drops; | 909 | return drops; |
910 | } | 910 | } |
911 | 911 | ||
912 | static noinline int early_drop(struct net *net, unsigned int _hash) | 912 | static noinline int early_drop(struct net *net, unsigned int _hash) |
913 | { | 913 | { |
914 | unsigned int i; | 914 | unsigned int i; |
915 | 915 | ||
916 | for (i = 0; i < NF_CT_EVICTION_RANGE; i++) { | 916 | for (i = 0; i < NF_CT_EVICTION_RANGE; i++) { |
917 | struct hlist_nulls_head *ct_hash; | 917 | struct hlist_nulls_head *ct_hash; |
918 | unsigned int hash, hsize, drops; | 918 | unsigned int hash, hsize, drops; |
919 | 919 | ||
920 | rcu_read_lock(); | 920 | rcu_read_lock(); |
921 | nf_conntrack_get_ht(&ct_hash, &hsize); | 921 | nf_conntrack_get_ht(&ct_hash, &hsize); |
922 | hash = reciprocal_scale(_hash++, hsize); | 922 | hash = reciprocal_scale(_hash++, hsize); |
923 | 923 | ||
924 | drops = early_drop_list(net, &ct_hash[hash]); | 924 | drops = early_drop_list(net, &ct_hash[hash]); |
925 | rcu_read_unlock(); | 925 | rcu_read_unlock(); |
926 | 926 | ||
927 | if (drops) { | 927 | if (drops) { |
928 | NF_CT_STAT_ADD_ATOMIC(net, early_drop, drops); | 928 | NF_CT_STAT_ADD_ATOMIC(net, early_drop, drops); |
929 | return true; | 929 | return true; |
930 | } | 930 | } |
931 | } | 931 | } |
932 | 932 | ||
933 | return false; | 933 | return false; |
934 | } | 934 | } |
935 | 935 | ||
936 | static void gc_worker(struct work_struct *work) | 936 | static void gc_worker(struct work_struct *work) |
937 | { | 937 | { |
938 | unsigned int i, goal, buckets = 0, expired_count = 0; | 938 | unsigned int i, goal, buckets = 0, expired_count = 0; |
939 | unsigned long next_run = GC_INTERVAL; | 939 | unsigned long next_run = GC_INTERVAL; |
940 | unsigned int ratio, scanned = 0; | 940 | unsigned int ratio, scanned = 0; |
941 | struct conntrack_gc_work *gc_work; | 941 | struct conntrack_gc_work *gc_work; |
942 | 942 | ||
943 | gc_work = container_of(work, struct conntrack_gc_work, dwork.work); | 943 | gc_work = container_of(work, struct conntrack_gc_work, dwork.work); |
944 | 944 | ||
945 | goal = min(nf_conntrack_htable_size / GC_MAX_BUCKETS_DIV, GC_MAX_BUCKETS); | 945 | goal = min(nf_conntrack_htable_size / GC_MAX_BUCKETS_DIV, GC_MAX_BUCKETS); |
946 | i = gc_work->last_bucket; | 946 | i = gc_work->last_bucket; |
947 | 947 | ||
948 | do { | 948 | do { |
949 | struct nf_conntrack_tuple_hash *h; | 949 | struct nf_conntrack_tuple_hash *h; |
950 | struct hlist_nulls_head *ct_hash; | 950 | struct hlist_nulls_head *ct_hash; |
951 | struct hlist_nulls_node *n; | 951 | struct hlist_nulls_node *n; |
952 | unsigned int hashsz; | 952 | unsigned int hashsz; |
953 | struct nf_conn *tmp; | 953 | struct nf_conn *tmp; |
954 | 954 | ||
955 | i++; | 955 | i++; |
956 | rcu_read_lock(); | 956 | rcu_read_lock(); |
957 | 957 | ||
958 | nf_conntrack_get_ht(&ct_hash, &hashsz); | 958 | nf_conntrack_get_ht(&ct_hash, &hashsz); |
959 | if (i >= hashsz) | 959 | if (i >= hashsz) |
960 | i = 0; | 960 | i = 0; |
961 | 961 | ||
962 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[i], hnnode) { | 962 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[i], hnnode) { |
963 | tmp = nf_ct_tuplehash_to_ctrack(h); | 963 | tmp = nf_ct_tuplehash_to_ctrack(h); |
964 | 964 | ||
965 | scanned++; | 965 | scanned++; |
966 | if (nf_ct_is_expired(tmp)) { | 966 | if (nf_ct_is_expired(tmp)) { |
967 | nf_ct_gc_expired(tmp); | 967 | nf_ct_gc_expired(tmp); |
968 | expired_count++; | 968 | expired_count++; |
969 | continue; | 969 | continue; |
970 | } | 970 | } |
971 | } | 971 | } |
972 | 972 | ||
973 | /* could check get_nulls_value() here and restart if ct | 973 | /* could check get_nulls_value() here and restart if ct |
974 | * was moved to another chain. But given gc is best-effort | 974 | * was moved to another chain. But given gc is best-effort |
975 | * we will just continue with next hash slot. | 975 | * we will just continue with next hash slot. |
976 | */ | 976 | */ |
977 | rcu_read_unlock(); | 977 | rcu_read_unlock(); |
978 | cond_resched_rcu_qs(); | 978 | cond_resched_rcu_qs(); |
979 | } while (++buckets < goal && | 979 | } while (++buckets < goal && |
980 | expired_count < GC_MAX_EVICTS); | 980 | expired_count < GC_MAX_EVICTS); |
981 | 981 | ||
982 | if (gc_work->exiting) | 982 | if (gc_work->exiting) |
983 | return; | 983 | return; |
984 | 984 | ||
985 | ratio = scanned ? expired_count * 100 / scanned : 0; | 985 | ratio = scanned ? expired_count * 100 / scanned : 0; |
986 | if (ratio >= 90 || expired_count == GC_MAX_EVICTS) | 986 | if (ratio >= 90 || expired_count == GC_MAX_EVICTS) |
987 | next_run = 0; | 987 | next_run = 0; |
988 | 988 | ||
989 | gc_work->last_bucket = i; | 989 | gc_work->last_bucket = i; |
990 | schedule_delayed_work(&gc_work->dwork, next_run); | 990 | schedule_delayed_work(&gc_work->dwork, next_run); |
991 | } | 991 | } |
992 | 992 | ||
993 | static void conntrack_gc_work_init(struct conntrack_gc_work *gc_work) | 993 | static void conntrack_gc_work_init(struct conntrack_gc_work *gc_work) |
994 | { | 994 | { |
995 | INIT_DELAYED_WORK(&gc_work->dwork, gc_worker); | 995 | INIT_DELAYED_WORK(&gc_work->dwork, gc_worker); |
996 | gc_work->exiting = false; | 996 | gc_work->exiting = false; |
997 | } | 997 | } |
998 | 998 | ||
999 | static struct nf_conn * | 999 | static struct nf_conn * |
1000 | __nf_conntrack_alloc(struct net *net, | 1000 | __nf_conntrack_alloc(struct net *net, |
1001 | const struct nf_conntrack_zone *zone, | 1001 | const struct nf_conntrack_zone *zone, |
1002 | const struct nf_conntrack_tuple *orig, | 1002 | const struct nf_conntrack_tuple *orig, |
1003 | const struct nf_conntrack_tuple *repl, | 1003 | const struct nf_conntrack_tuple *repl, |
1004 | gfp_t gfp, u32 hash) | 1004 | gfp_t gfp, u32 hash) |
1005 | { | 1005 | { |
1006 | struct nf_conn *ct; | 1006 | struct nf_conn *ct; |
1007 | 1007 | ||
1008 | /* We don't want any race condition at early drop stage */ | 1008 | /* We don't want any race condition at early drop stage */ |
1009 | atomic_inc(&net->ct.count); | 1009 | atomic_inc(&net->ct.count); |
1010 | 1010 | ||
1011 | if (nf_conntrack_max && | 1011 | if (nf_conntrack_max && |
1012 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { | 1012 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
1013 | if (!early_drop(net, hash)) { | 1013 | if (!early_drop(net, hash)) { |
1014 | atomic_dec(&net->ct.count); | 1014 | atomic_dec(&net->ct.count); |
1015 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); | 1015 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); |
1016 | return ERR_PTR(-ENOMEM); | 1016 | return ERR_PTR(-ENOMEM); |
1017 | } | 1017 | } |
1018 | } | 1018 | } |
1019 | 1019 | ||
1020 | /* | 1020 | /* |
1021 | * Do not use kmem_cache_zalloc(), as this cache uses | 1021 | * Do not use kmem_cache_zalloc(), as this cache uses |
1022 | * SLAB_DESTROY_BY_RCU. | 1022 | * SLAB_DESTROY_BY_RCU. |
1023 | */ | 1023 | */ |
1024 | ct = kmem_cache_alloc(nf_conntrack_cachep, gfp); | 1024 | ct = kmem_cache_alloc(nf_conntrack_cachep, gfp); |
1025 | if (ct == NULL) | 1025 | if (ct == NULL) |
1026 | goto out; | 1026 | goto out; |
1027 | 1027 | ||
1028 | spin_lock_init(&ct->lock); | 1028 | spin_lock_init(&ct->lock); |
1029 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; | 1029 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
1030 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; | 1030 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
1031 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; | 1031 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
1032 | /* save hash for reusing when confirming */ | 1032 | /* save hash for reusing when confirming */ |
1033 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; | 1033 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; |
1034 | ct->status = 0; | 1034 | ct->status = 0; |
1035 | write_pnet(&ct->ct_net, net); | 1035 | write_pnet(&ct->ct_net, net); |
1036 | memset(&ct->__nfct_init_offset[0], 0, | 1036 | memset(&ct->__nfct_init_offset[0], 0, |
1037 | offsetof(struct nf_conn, proto) - | 1037 | offsetof(struct nf_conn, proto) - |
1038 | offsetof(struct nf_conn, __nfct_init_offset[0])); | 1038 | offsetof(struct nf_conn, __nfct_init_offset[0])); |
1039 | 1039 | ||
1040 | nf_ct_zone_add(ct, zone); | 1040 | nf_ct_zone_add(ct, zone); |
1041 | 1041 | ||
1042 | /* Because we use RCU lookups, we set ct_general.use to zero before | 1042 | /* Because we use RCU lookups, we set ct_general.use to zero before |
1043 | * this is inserted in any list. | 1043 | * this is inserted in any list. |
1044 | */ | 1044 | */ |
1045 | atomic_set(&ct->ct_general.use, 0); | 1045 | atomic_set(&ct->ct_general.use, 0); |
1046 | return ct; | 1046 | return ct; |
1047 | out: | 1047 | out: |
1048 | atomic_dec(&net->ct.count); | 1048 | atomic_dec(&net->ct.count); |
1049 | return ERR_PTR(-ENOMEM); | 1049 | return ERR_PTR(-ENOMEM); |
1050 | } | 1050 | } |
1051 | 1051 | ||
1052 | struct nf_conn *nf_conntrack_alloc(struct net *net, | 1052 | struct nf_conn *nf_conntrack_alloc(struct net *net, |
1053 | const struct nf_conntrack_zone *zone, | 1053 | const struct nf_conntrack_zone *zone, |
1054 | const struct nf_conntrack_tuple *orig, | 1054 | const struct nf_conntrack_tuple *orig, |
1055 | const struct nf_conntrack_tuple *repl, | 1055 | const struct nf_conntrack_tuple *repl, |
1056 | gfp_t gfp) | 1056 | gfp_t gfp) |
1057 | { | 1057 | { |
1058 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); | 1058 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); |
1059 | } | 1059 | } |
1060 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); | 1060 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
1061 | 1061 | ||
1062 | void nf_conntrack_free(struct nf_conn *ct) | 1062 | void nf_conntrack_free(struct nf_conn *ct) |
1063 | { | 1063 | { |
1064 | struct net *net = nf_ct_net(ct); | 1064 | struct net *net = nf_ct_net(ct); |
1065 | 1065 | ||
1066 | /* A freed object has refcnt == 0, that's | 1066 | /* A freed object has refcnt == 0, that's |
1067 | * the golden rule for SLAB_DESTROY_BY_RCU | 1067 | * the golden rule for SLAB_DESTROY_BY_RCU |
1068 | */ | 1068 | */ |
1069 | NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 0); | 1069 | NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 0); |
1070 | 1070 | ||
1071 | nf_ct_ext_destroy(ct); | 1071 | nf_ct_ext_destroy(ct); |
1072 | nf_ct_ext_free(ct); | 1072 | nf_ct_ext_free(ct); |
1073 | kmem_cache_free(nf_conntrack_cachep, ct); | 1073 | kmem_cache_free(nf_conntrack_cachep, ct); |
1074 | smp_mb__before_atomic(); | 1074 | smp_mb__before_atomic(); |
1075 | atomic_dec(&net->ct.count); | 1075 | atomic_dec(&net->ct.count); |
1076 | } | 1076 | } |
1077 | EXPORT_SYMBOL_GPL(nf_conntrack_free); | 1077 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
1078 | 1078 | ||
1079 | 1079 | ||
1080 | /* Allocate a new conntrack: we return -ENOMEM if classification | 1080 | /* Allocate a new conntrack: we return -ENOMEM if classification |
1081 | failed due to stress. Otherwise it really is unclassifiable. */ | 1081 | failed due to stress. Otherwise it really is unclassifiable. */ |
1082 | static struct nf_conntrack_tuple_hash * | 1082 | static struct nf_conntrack_tuple_hash * |
1083 | init_conntrack(struct net *net, struct nf_conn *tmpl, | 1083 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
1084 | const struct nf_conntrack_tuple *tuple, | 1084 | const struct nf_conntrack_tuple *tuple, |
1085 | struct nf_conntrack_l3proto *l3proto, | 1085 | struct nf_conntrack_l3proto *l3proto, |
1086 | struct nf_conntrack_l4proto *l4proto, | 1086 | struct nf_conntrack_l4proto *l4proto, |
1087 | struct sk_buff *skb, | 1087 | struct sk_buff *skb, |
1088 | unsigned int dataoff, u32 hash) | 1088 | unsigned int dataoff, u32 hash) |
1089 | { | 1089 | { |
1090 | struct nf_conn *ct; | 1090 | struct nf_conn *ct; |
1091 | struct nf_conn_help *help; | 1091 | struct nf_conn_help *help; |
1092 | struct nf_conntrack_tuple repl_tuple; | 1092 | struct nf_conntrack_tuple repl_tuple; |
1093 | struct nf_conntrack_ecache *ecache; | 1093 | struct nf_conntrack_ecache *ecache; |
1094 | struct nf_conntrack_expect *exp = NULL; | 1094 | struct nf_conntrack_expect *exp = NULL; |
1095 | const struct nf_conntrack_zone *zone; | 1095 | const struct nf_conntrack_zone *zone; |
1096 | struct nf_conn_timeout *timeout_ext; | 1096 | struct nf_conn_timeout *timeout_ext; |
1097 | struct nf_conntrack_zone tmp; | 1097 | struct nf_conntrack_zone tmp; |
1098 | unsigned int *timeouts; | 1098 | unsigned int *timeouts; |
1099 | 1099 | ||
1100 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { | 1100 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
1101 | pr_debug("Can't invert tuple.\n"); | 1101 | pr_debug("Can't invert tuple.\n"); |
1102 | return NULL; | 1102 | return NULL; |
1103 | } | 1103 | } |
1104 | 1104 | ||
1105 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); | 1105 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); |
1106 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, | 1106 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
1107 | hash); | 1107 | hash); |
1108 | if (IS_ERR(ct)) | 1108 | if (IS_ERR(ct)) |
1109 | return (struct nf_conntrack_tuple_hash *)ct; | 1109 | return (struct nf_conntrack_tuple_hash *)ct; |
1110 | 1110 | ||
1111 | if (!nf_ct_add_synproxy(ct, tmpl)) { | 1111 | if (!nf_ct_add_synproxy(ct, tmpl)) { |
1112 | nf_conntrack_free(ct); | 1112 | nf_conntrack_free(ct); |
1113 | return ERR_PTR(-ENOMEM); | 1113 | return ERR_PTR(-ENOMEM); |
1114 | } | 1114 | } |
1115 | 1115 | ||
1116 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; | 1116 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; |
1117 | if (timeout_ext) { | 1117 | if (timeout_ext) { |
1118 | timeouts = nf_ct_timeout_data(timeout_ext); | 1118 | timeouts = nf_ct_timeout_data(timeout_ext); |
1119 | if (unlikely(!timeouts)) | 1119 | if (unlikely(!timeouts)) |
1120 | timeouts = l4proto->get_timeouts(net); | 1120 | timeouts = l4proto->get_timeouts(net); |
1121 | } else { | 1121 | } else { |
1122 | timeouts = l4proto->get_timeouts(net); | 1122 | timeouts = l4proto->get_timeouts(net); |
1123 | } | 1123 | } |
1124 | 1124 | ||
1125 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { | 1125 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { |
1126 | nf_conntrack_free(ct); | 1126 | nf_conntrack_free(ct); |
1127 | pr_debug("can't track with proto module\n"); | 1127 | pr_debug("can't track with proto module\n"); |
1128 | return NULL; | 1128 | return NULL; |
1129 | } | 1129 | } |
1130 | 1130 | ||
1131 | if (timeout_ext) | 1131 | if (timeout_ext) |
1132 | nf_ct_timeout_ext_add(ct, rcu_dereference(timeout_ext->timeout), | 1132 | nf_ct_timeout_ext_add(ct, rcu_dereference(timeout_ext->timeout), |
1133 | GFP_ATOMIC); | 1133 | GFP_ATOMIC); |
1134 | 1134 | ||
1135 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); | 1135 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
1136 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); | 1136 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
1137 | nf_ct_labels_ext_add(ct); | 1137 | nf_ct_labels_ext_add(ct); |
1138 | 1138 | ||
1139 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; | 1139 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; |
1140 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, | 1140 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, |
1141 | ecache ? ecache->expmask : 0, | 1141 | ecache ? ecache->expmask : 0, |
1142 | GFP_ATOMIC); | 1142 | GFP_ATOMIC); |
1143 | 1143 | ||
1144 | local_bh_disable(); | 1144 | local_bh_disable(); |
1145 | if (net->ct.expect_count) { | 1145 | if (net->ct.expect_count) { |
1146 | spin_lock(&nf_conntrack_expect_lock); | 1146 | spin_lock(&nf_conntrack_expect_lock); |
1147 | exp = nf_ct_find_expectation(net, zone, tuple); | 1147 | exp = nf_ct_find_expectation(net, zone, tuple); |
1148 | if (exp) { | 1148 | if (exp) { |
1149 | pr_debug("expectation arrives ct=%p exp=%p\n", | 1149 | pr_debug("expectation arrives ct=%p exp=%p\n", |
1150 | ct, exp); | 1150 | ct, exp); |
1151 | /* Welcome, Mr. Bond. We've been expecting you... */ | 1151 | /* Welcome, Mr. Bond. We've been expecting you... */ |
1152 | __set_bit(IPS_EXPECTED_BIT, &ct->status); | 1152 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
1153 | /* exp->master safe, refcnt bumped in nf_ct_find_expectation */ | 1153 | /* exp->master safe, refcnt bumped in nf_ct_find_expectation */ |
1154 | ct->master = exp->master; | 1154 | ct->master = exp->master; |
1155 | if (exp->helper) { | 1155 | if (exp->helper) { |
1156 | help = nf_ct_helper_ext_add(ct, exp->helper, | 1156 | help = nf_ct_helper_ext_add(ct, exp->helper, |
1157 | GFP_ATOMIC); | 1157 | GFP_ATOMIC); |
1158 | if (help) | 1158 | if (help) |
1159 | rcu_assign_pointer(help->helper, exp->helper); | 1159 | rcu_assign_pointer(help->helper, exp->helper); |
1160 | } | 1160 | } |
1161 | 1161 | ||
1162 | #ifdef CONFIG_NF_CONNTRACK_MARK | 1162 | #ifdef CONFIG_NF_CONNTRACK_MARK |
1163 | ct->mark = exp->master->mark; | 1163 | ct->mark = exp->master->mark; |
1164 | #endif | 1164 | #endif |
1165 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | 1165 | #ifdef CONFIG_NF_CONNTRACK_SECMARK |
1166 | ct->secmark = exp->master->secmark; | 1166 | ct->secmark = exp->master->secmark; |
1167 | #endif | 1167 | #endif |
1168 | NF_CT_STAT_INC(net, expect_new); | 1168 | NF_CT_STAT_INC(net, expect_new); |
1169 | } | 1169 | } |
1170 | spin_unlock(&nf_conntrack_expect_lock); | 1170 | spin_unlock(&nf_conntrack_expect_lock); |
1171 | } | 1171 | } |
1172 | if (!exp) | 1172 | if (!exp) |
1173 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); | 1173 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
1174 | 1174 | ||
1175 | /* Now it is inserted into the unconfirmed list, bump refcount */ | 1175 | /* Now it is inserted into the unconfirmed list, bump refcount */ |
1176 | nf_conntrack_get(&ct->ct_general); | 1176 | nf_conntrack_get(&ct->ct_general); |
1177 | nf_ct_add_to_unconfirmed_list(ct); | 1177 | nf_ct_add_to_unconfirmed_list(ct); |
1178 | 1178 | ||
1179 | local_bh_enable(); | 1179 | local_bh_enable(); |
1180 | 1180 | ||
1181 | if (exp) { | 1181 | if (exp) { |
1182 | if (exp->expectfn) | 1182 | if (exp->expectfn) |
1183 | exp->expectfn(ct, exp); | 1183 | exp->expectfn(ct, exp); |
1184 | nf_ct_expect_put(exp); | 1184 | nf_ct_expect_put(exp); |
1185 | } | 1185 | } |
1186 | 1186 | ||
1187 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; | 1187 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
1188 | } | 1188 | } |
1189 | 1189 | ||
1190 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | 1190 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ |
1191 | static inline struct nf_conn * | 1191 | static inline struct nf_conn * |
1192 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, | 1192 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
1193 | struct sk_buff *skb, | 1193 | struct sk_buff *skb, |
1194 | unsigned int dataoff, | 1194 | unsigned int dataoff, |
1195 | u_int16_t l3num, | 1195 | u_int16_t l3num, |
1196 | u_int8_t protonum, | 1196 | u_int8_t protonum, |
1197 | struct nf_conntrack_l3proto *l3proto, | 1197 | struct nf_conntrack_l3proto *l3proto, |
1198 | struct nf_conntrack_l4proto *l4proto, | 1198 | struct nf_conntrack_l4proto *l4proto, |
1199 | int *set_reply, | 1199 | int *set_reply, |
1200 | enum ip_conntrack_info *ctinfo) | 1200 | enum ip_conntrack_info *ctinfo) |
1201 | { | 1201 | { |
1202 | const struct nf_conntrack_zone *zone; | 1202 | const struct nf_conntrack_zone *zone; |
1203 | struct nf_conntrack_tuple tuple; | 1203 | struct nf_conntrack_tuple tuple; |
1204 | struct nf_conntrack_tuple_hash *h; | 1204 | struct nf_conntrack_tuple_hash *h; |
1205 | struct nf_conntrack_zone tmp; | 1205 | struct nf_conntrack_zone tmp; |
1206 | struct nf_conn *ct; | 1206 | struct nf_conn *ct; |
1207 | u32 hash; | 1207 | u32 hash; |
1208 | 1208 | ||
1209 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), | 1209 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
1210 | dataoff, l3num, protonum, net, &tuple, l3proto, | 1210 | dataoff, l3num, protonum, net, &tuple, l3proto, |
1211 | l4proto)) { | 1211 | l4proto)) { |
1212 | pr_debug("Can't get tuple\n"); | 1212 | pr_debug("Can't get tuple\n"); |
1213 | return NULL; | 1213 | return NULL; |
1214 | } | 1214 | } |
1215 | 1215 | ||
1216 | /* look for tuple match */ | 1216 | /* look for tuple match */ |
1217 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); | 1217 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); |
1218 | hash = hash_conntrack_raw(&tuple, net); | 1218 | hash = hash_conntrack_raw(&tuple, net); |
1219 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); | 1219 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); |
1220 | if (!h) { | 1220 | if (!h) { |
1221 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, | 1221 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
1222 | skb, dataoff, hash); | 1222 | skb, dataoff, hash); |
1223 | if (!h) | 1223 | if (!h) |
1224 | return NULL; | 1224 | return NULL; |
1225 | if (IS_ERR(h)) | 1225 | if (IS_ERR(h)) |
1226 | return (void *)h; | 1226 | return (void *)h; |
1227 | } | 1227 | } |
1228 | ct = nf_ct_tuplehash_to_ctrack(h); | 1228 | ct = nf_ct_tuplehash_to_ctrack(h); |
1229 | 1229 | ||
1230 | /* It exists; we have (non-exclusive) reference. */ | 1230 | /* It exists; we have (non-exclusive) reference. */ |
1231 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | 1231 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { |
1232 | *ctinfo = IP_CT_ESTABLISHED_REPLY; | 1232 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
1233 | /* Please set reply bit if this packet OK */ | 1233 | /* Please set reply bit if this packet OK */ |
1234 | *set_reply = 1; | 1234 | *set_reply = 1; |
1235 | } else { | 1235 | } else { |
1236 | /* Once we've had two way comms, always ESTABLISHED. */ | 1236 | /* Once we've had two way comms, always ESTABLISHED. */ |
1237 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | 1237 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { |
1238 | pr_debug("normal packet for %p\n", ct); | 1238 | pr_debug("normal packet for %p\n", ct); |
1239 | *ctinfo = IP_CT_ESTABLISHED; | 1239 | *ctinfo = IP_CT_ESTABLISHED; |
1240 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | 1240 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { |
1241 | pr_debug("related packet for %p\n", ct); | 1241 | pr_debug("related packet for %p\n", ct); |
1242 | *ctinfo = IP_CT_RELATED; | 1242 | *ctinfo = IP_CT_RELATED; |
1243 | } else { | 1243 | } else { |
1244 | pr_debug("new packet for %p\n", ct); | 1244 | pr_debug("new packet for %p\n", ct); |
1245 | *ctinfo = IP_CT_NEW; | 1245 | *ctinfo = IP_CT_NEW; |
1246 | } | 1246 | } |
1247 | *set_reply = 0; | 1247 | *set_reply = 0; |
1248 | } | 1248 | } |
1249 | skb->nfct = &ct->ct_general; | 1249 | skb->nfct = &ct->ct_general; |
1250 | skb->nfctinfo = *ctinfo; | 1250 | skb->nfctinfo = *ctinfo; |
1251 | return ct; | 1251 | return ct; |
1252 | } | 1252 | } |
1253 | 1253 | ||
1254 | unsigned int | 1254 | unsigned int |
1255 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, | 1255 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
1256 | struct sk_buff *skb) | 1256 | struct sk_buff *skb) |
1257 | { | 1257 | { |
1258 | struct nf_conn *ct, *tmpl = NULL; | 1258 | struct nf_conn *ct, *tmpl = NULL; |
1259 | enum ip_conntrack_info ctinfo; | 1259 | enum ip_conntrack_info ctinfo; |
1260 | struct nf_conntrack_l3proto *l3proto; | 1260 | struct nf_conntrack_l3proto *l3proto; |
1261 | struct nf_conntrack_l4proto *l4proto; | 1261 | struct nf_conntrack_l4proto *l4proto; |
1262 | unsigned int *timeouts; | 1262 | unsigned int *timeouts; |
1263 | unsigned int dataoff; | 1263 | unsigned int dataoff; |
1264 | u_int8_t protonum; | 1264 | u_int8_t protonum; |
1265 | int set_reply = 0; | 1265 | int set_reply = 0; |
1266 | int ret; | 1266 | int ret; |
1267 | 1267 | ||
1268 | if (skb->nfct) { | 1268 | if (skb->nfct) { |
1269 | /* Previously seen (loopback or untracked)? Ignore. */ | 1269 | /* Previously seen (loopback or untracked)? Ignore. */ |
1270 | tmpl = (struct nf_conn *)skb->nfct; | 1270 | tmpl = (struct nf_conn *)skb->nfct; |
1271 | if (!nf_ct_is_template(tmpl)) { | 1271 | if (!nf_ct_is_template(tmpl)) { |
1272 | NF_CT_STAT_INC_ATOMIC(net, ignore); | 1272 | NF_CT_STAT_INC_ATOMIC(net, ignore); |
1273 | return NF_ACCEPT; | 1273 | return NF_ACCEPT; |
1274 | } | 1274 | } |
1275 | skb->nfct = NULL; | 1275 | skb->nfct = NULL; |
1276 | } | 1276 | } |
1277 | 1277 | ||
1278 | /* rcu_read_lock()ed by nf_hook_thresh */ | 1278 | /* rcu_read_lock()ed by nf_hook_thresh */ |
1279 | l3proto = __nf_ct_l3proto_find(pf); | 1279 | l3proto = __nf_ct_l3proto_find(pf); |
1280 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), | 1280 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
1281 | &dataoff, &protonum); | 1281 | &dataoff, &protonum); |
1282 | if (ret <= 0) { | 1282 | if (ret <= 0) { |
1283 | pr_debug("not prepared to track yet or error occurred\n"); | 1283 | pr_debug("not prepared to track yet or error occurred\n"); |
1284 | NF_CT_STAT_INC_ATOMIC(net, error); | 1284 | NF_CT_STAT_INC_ATOMIC(net, error); |
1285 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 1285 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
1286 | ret = -ret; | 1286 | ret = -ret; |
1287 | goto out; | 1287 | goto out; |
1288 | } | 1288 | } |
1289 | 1289 | ||
1290 | l4proto = __nf_ct_l4proto_find(pf, protonum); | 1290 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
1291 | 1291 | ||
1292 | /* It may be an special packet, error, unclean... | 1292 | /* It may be an special packet, error, unclean... |
1293 | * inverse of the return code tells to the netfilter | 1293 | * inverse of the return code tells to the netfilter |
1294 | * core what to do with the packet. */ | 1294 | * core what to do with the packet. */ |
1295 | if (l4proto->error != NULL) { | 1295 | if (l4proto->error != NULL) { |
1296 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, | 1296 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
1297 | pf, hooknum); | 1297 | pf, hooknum); |
1298 | if (ret <= 0) { | 1298 | if (ret <= 0) { |
1299 | NF_CT_STAT_INC_ATOMIC(net, error); | 1299 | NF_CT_STAT_INC_ATOMIC(net, error); |
1300 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 1300 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
1301 | ret = -ret; | 1301 | ret = -ret; |
1302 | goto out; | 1302 | goto out; |
1303 | } | 1303 | } |
1304 | /* ICMP[v6] protocol trackers may assign one conntrack. */ | 1304 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
1305 | if (skb->nfct) | 1305 | if (skb->nfct) |
1306 | goto out; | 1306 | goto out; |
1307 | } | 1307 | } |
1308 | 1308 | repeat: | |
1309 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, | 1309 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
1310 | l3proto, l4proto, &set_reply, &ctinfo); | 1310 | l3proto, l4proto, &set_reply, &ctinfo); |
1311 | if (!ct) { | 1311 | if (!ct) { |
1312 | /* Not valid part of a connection */ | 1312 | /* Not valid part of a connection */ |
1313 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 1313 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
1314 | ret = NF_ACCEPT; | 1314 | ret = NF_ACCEPT; |
1315 | goto out; | 1315 | goto out; |
1316 | } | 1316 | } |
1317 | 1317 | ||
1318 | if (IS_ERR(ct)) { | 1318 | if (IS_ERR(ct)) { |
1319 | /* Too stressed to deal. */ | 1319 | /* Too stressed to deal. */ |
1320 | NF_CT_STAT_INC_ATOMIC(net, drop); | 1320 | NF_CT_STAT_INC_ATOMIC(net, drop); |
1321 | ret = NF_DROP; | 1321 | ret = NF_DROP; |
1322 | goto out; | 1322 | goto out; |
1323 | } | 1323 | } |
1324 | 1324 | ||
1325 | NF_CT_ASSERT(skb->nfct); | 1325 | NF_CT_ASSERT(skb->nfct); |
1326 | 1326 | ||
1327 | /* Decide what timeout policy we want to apply to this flow. */ | 1327 | /* Decide what timeout policy we want to apply to this flow. */ |
1328 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); | 1328 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); |
1329 | 1329 | ||
1330 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); | 1330 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); |
1331 | if (ret <= 0) { | 1331 | if (ret <= 0) { |
1332 | /* Invalid: inverse of the return code tells | 1332 | /* Invalid: inverse of the return code tells |
1333 | * the netfilter core what to do */ | 1333 | * the netfilter core what to do */ |
1334 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); | 1334 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
1335 | nf_conntrack_put(skb->nfct); | 1335 | nf_conntrack_put(skb->nfct); |
1336 | skb->nfct = NULL; | 1336 | skb->nfct = NULL; |
1337 | NF_CT_STAT_INC_ATOMIC(net, invalid); | 1337 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
1338 | if (ret == -NF_DROP) | 1338 | if (ret == -NF_DROP) |
1339 | NF_CT_STAT_INC_ATOMIC(net, drop); | 1339 | NF_CT_STAT_INC_ATOMIC(net, drop); |
1340 | ret = -ret; | 1340 | ret = -ret; |
1341 | goto out; | 1341 | goto out; |
1342 | } | 1342 | } |
1343 | 1343 | ||
1344 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | 1344 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) |
1345 | nf_conntrack_event_cache(IPCT_REPLY, ct); | 1345 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
1346 | out: | 1346 | out: |
1347 | if (tmpl) { | 1347 | if (tmpl) { |
1348 | /* Special case: we have to repeat this hook, assign the | 1348 | /* Special case: TCP tracker reports an attempt to reopen a |
1349 | * template again to this packet. We assume that this packet | 1349 | * closed/aborted connection. We have to go back and create a |
1350 | * has no conntrack assigned. This is used by nf_ct_tcp. */ | 1350 | * fresh conntrack. |
1351 | */ | ||
1351 | if (ret == NF_REPEAT) | 1352 | if (ret == NF_REPEAT) |
1352 | skb->nfct = (struct nf_conntrack *)tmpl; | 1353 | goto repeat; |
1353 | else | 1354 | else |
1354 | nf_ct_put(tmpl); | 1355 | nf_ct_put(tmpl); |
1355 | } | 1356 | } |
1356 | 1357 | ||
1357 | return ret; | 1358 | return ret; |
1358 | } | 1359 | } |
1359 | EXPORT_SYMBOL_GPL(nf_conntrack_in); | 1360 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
1360 | 1361 | ||
1361 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, | 1362 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
1362 | const struct nf_conntrack_tuple *orig) | 1363 | const struct nf_conntrack_tuple *orig) |
1363 | { | 1364 | { |
1364 | bool ret; | 1365 | bool ret; |
1365 | 1366 | ||
1366 | rcu_read_lock(); | 1367 | rcu_read_lock(); |
1367 | ret = nf_ct_invert_tuple(inverse, orig, | 1368 | ret = nf_ct_invert_tuple(inverse, orig, |
1368 | __nf_ct_l3proto_find(orig->src.l3num), | 1369 | __nf_ct_l3proto_find(orig->src.l3num), |
1369 | __nf_ct_l4proto_find(orig->src.l3num, | 1370 | __nf_ct_l4proto_find(orig->src.l3num, |
1370 | orig->dst.protonum)); | 1371 | orig->dst.protonum)); |
1371 | rcu_read_unlock(); | 1372 | rcu_read_unlock(); |
1372 | return ret; | 1373 | return ret; |
1373 | } | 1374 | } |
1374 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); | 1375 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
1375 | 1376 | ||
1376 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is | 1377 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
1377 | implicitly racy: see __nf_conntrack_confirm */ | 1378 | implicitly racy: see __nf_conntrack_confirm */ |
1378 | void nf_conntrack_alter_reply(struct nf_conn *ct, | 1379 | void nf_conntrack_alter_reply(struct nf_conn *ct, |
1379 | const struct nf_conntrack_tuple *newreply) | 1380 | const struct nf_conntrack_tuple *newreply) |
1380 | { | 1381 | { |
1381 | struct nf_conn_help *help = nfct_help(ct); | 1382 | struct nf_conn_help *help = nfct_help(ct); |
1382 | 1383 | ||
1383 | /* Should be unconfirmed, so not in hash table yet */ | 1384 | /* Should be unconfirmed, so not in hash table yet */ |
1384 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | 1385 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
1385 | 1386 | ||
1386 | pr_debug("Altering reply tuple of %p to ", ct); | 1387 | pr_debug("Altering reply tuple of %p to ", ct); |
1387 | nf_ct_dump_tuple(newreply); | 1388 | nf_ct_dump_tuple(newreply); |
1388 | 1389 | ||
1389 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | 1390 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; |
1390 | if (ct->master || (help && !hlist_empty(&help->expectations))) | 1391 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
1391 | return; | 1392 | return; |
1392 | 1393 | ||
1393 | rcu_read_lock(); | 1394 | rcu_read_lock(); |
1394 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); | 1395 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
1395 | rcu_read_unlock(); | 1396 | rcu_read_unlock(); |
1396 | } | 1397 | } |
1397 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); | 1398 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
1398 | 1399 | ||
1399 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ | 1400 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
1400 | void __nf_ct_refresh_acct(struct nf_conn *ct, | 1401 | void __nf_ct_refresh_acct(struct nf_conn *ct, |
1401 | enum ip_conntrack_info ctinfo, | 1402 | enum ip_conntrack_info ctinfo, |
1402 | const struct sk_buff *skb, | 1403 | const struct sk_buff *skb, |
1403 | unsigned long extra_jiffies, | 1404 | unsigned long extra_jiffies, |
1404 | int do_acct) | 1405 | int do_acct) |
1405 | { | 1406 | { |
1406 | NF_CT_ASSERT(skb); | 1407 | NF_CT_ASSERT(skb); |
1407 | 1408 | ||
1408 | /* Only update if this is not a fixed timeout */ | 1409 | /* Only update if this is not a fixed timeout */ |
1409 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) | 1410 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
1410 | goto acct; | 1411 | goto acct; |
1411 | 1412 | ||
1412 | /* If not in hash table, timer will not be active yet */ | 1413 | /* If not in hash table, timer will not be active yet */ |
1413 | if (nf_ct_is_confirmed(ct)) | 1414 | if (nf_ct_is_confirmed(ct)) |
1414 | extra_jiffies += nfct_time_stamp; | 1415 | extra_jiffies += nfct_time_stamp; |
1415 | 1416 | ||
1416 | ct->timeout = extra_jiffies; | 1417 | ct->timeout = extra_jiffies; |
1417 | acct: | 1418 | acct: |
1418 | if (do_acct) | 1419 | if (do_acct) |
1419 | nf_ct_acct_update(ct, ctinfo, skb->len); | 1420 | nf_ct_acct_update(ct, ctinfo, skb->len); |
1420 | } | 1421 | } |
1421 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); | 1422 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
1422 | 1423 | ||
1423 | bool nf_ct_kill_acct(struct nf_conn *ct, | 1424 | bool nf_ct_kill_acct(struct nf_conn *ct, |
1424 | enum ip_conntrack_info ctinfo, | 1425 | enum ip_conntrack_info ctinfo, |
1425 | const struct sk_buff *skb) | 1426 | const struct sk_buff *skb) |
1426 | { | 1427 | { |
1427 | nf_ct_acct_update(ct, ctinfo, skb->len); | 1428 | nf_ct_acct_update(ct, ctinfo, skb->len); |
1428 | 1429 | ||
1429 | return nf_ct_delete(ct, 0, 0); | 1430 | return nf_ct_delete(ct, 0, 0); |
1430 | } | 1431 | } |
1431 | EXPORT_SYMBOL_GPL(nf_ct_kill_acct); | 1432 | EXPORT_SYMBOL_GPL(nf_ct_kill_acct); |
1432 | 1433 | ||
1433 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) | 1434 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
1434 | 1435 | ||
1435 | #include <linux/netfilter/nfnetlink.h> | 1436 | #include <linux/netfilter/nfnetlink.h> |
1436 | #include <linux/netfilter/nfnetlink_conntrack.h> | 1437 | #include <linux/netfilter/nfnetlink_conntrack.h> |
1437 | #include <linux/mutex.h> | 1438 | #include <linux/mutex.h> |
1438 | 1439 | ||
1439 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be | 1440 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
1440 | * in ip_conntrack_core, since we don't want the protocols to autoload | 1441 | * in ip_conntrack_core, since we don't want the protocols to autoload |
1441 | * or depend on ctnetlink */ | 1442 | * or depend on ctnetlink */ |
1442 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, | 1443 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
1443 | const struct nf_conntrack_tuple *tuple) | 1444 | const struct nf_conntrack_tuple *tuple) |
1444 | { | 1445 | { |
1445 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || | 1446 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || |
1446 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) | 1447 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) |
1447 | goto nla_put_failure; | 1448 | goto nla_put_failure; |
1448 | return 0; | 1449 | return 0; |
1449 | 1450 | ||
1450 | nla_put_failure: | 1451 | nla_put_failure: |
1451 | return -1; | 1452 | return -1; |
1452 | } | 1453 | } |
1453 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); | 1454 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
1454 | 1455 | ||
1455 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { | 1456 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
1456 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | 1457 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, |
1457 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | 1458 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, |
1458 | }; | 1459 | }; |
1459 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); | 1460 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
1460 | 1461 | ||
1461 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], | 1462 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
1462 | struct nf_conntrack_tuple *t) | 1463 | struct nf_conntrack_tuple *t) |
1463 | { | 1464 | { |
1464 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) | 1465 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
1465 | return -EINVAL; | 1466 | return -EINVAL; |
1466 | 1467 | ||
1467 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); | 1468 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
1468 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | 1469 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); |
1469 | 1470 | ||
1470 | return 0; | 1471 | return 0; |
1471 | } | 1472 | } |
1472 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); | 1473 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
1473 | 1474 | ||
1474 | int nf_ct_port_nlattr_tuple_size(void) | 1475 | int nf_ct_port_nlattr_tuple_size(void) |
1475 | { | 1476 | { |
1476 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | 1477 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); |
1477 | } | 1478 | } |
1478 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | 1479 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); |
1479 | #endif | 1480 | #endif |
1480 | 1481 | ||
1481 | /* Used by ipt_REJECT and ip6t_REJECT. */ | 1482 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
1482 | static void nf_conntrack_attach(struct sk_buff *nskb, const struct sk_buff *skb) | 1483 | static void nf_conntrack_attach(struct sk_buff *nskb, const struct sk_buff *skb) |
1483 | { | 1484 | { |
1484 | struct nf_conn *ct; | 1485 | struct nf_conn *ct; |
1485 | enum ip_conntrack_info ctinfo; | 1486 | enum ip_conntrack_info ctinfo; |
1486 | 1487 | ||
1487 | /* This ICMP is in reverse direction to the packet which caused it */ | 1488 | /* This ICMP is in reverse direction to the packet which caused it */ |
1488 | ct = nf_ct_get(skb, &ctinfo); | 1489 | ct = nf_ct_get(skb, &ctinfo); |
1489 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | 1490 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) |
1490 | ctinfo = IP_CT_RELATED_REPLY; | 1491 | ctinfo = IP_CT_RELATED_REPLY; |
1491 | else | 1492 | else |
1492 | ctinfo = IP_CT_RELATED; | 1493 | ctinfo = IP_CT_RELATED; |
1493 | 1494 | ||
1494 | /* Attach to new skbuff, and increment count */ | 1495 | /* Attach to new skbuff, and increment count */ |
1495 | nskb->nfct = &ct->ct_general; | 1496 | nskb->nfct = &ct->ct_general; |
1496 | nskb->nfctinfo = ctinfo; | 1497 | nskb->nfctinfo = ctinfo; |
1497 | nf_conntrack_get(nskb->nfct); | 1498 | nf_conntrack_get(nskb->nfct); |
1498 | } | 1499 | } |
1499 | 1500 | ||
1500 | /* Bring out ya dead! */ | 1501 | /* Bring out ya dead! */ |
1501 | static struct nf_conn * | 1502 | static struct nf_conn * |
1502 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), | 1503 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
1503 | void *data, unsigned int *bucket) | 1504 | void *data, unsigned int *bucket) |
1504 | { | 1505 | { |
1505 | struct nf_conntrack_tuple_hash *h; | 1506 | struct nf_conntrack_tuple_hash *h; |
1506 | struct nf_conn *ct; | 1507 | struct nf_conn *ct; |
1507 | struct hlist_nulls_node *n; | 1508 | struct hlist_nulls_node *n; |
1508 | int cpu; | 1509 | int cpu; |
1509 | spinlock_t *lockp; | 1510 | spinlock_t *lockp; |
1510 | 1511 | ||
1511 | for (; *bucket < nf_conntrack_htable_size; (*bucket)++) { | 1512 | for (; *bucket < nf_conntrack_htable_size; (*bucket)++) { |
1512 | lockp = &nf_conntrack_locks[*bucket % CONNTRACK_LOCKS]; | 1513 | lockp = &nf_conntrack_locks[*bucket % CONNTRACK_LOCKS]; |
1513 | local_bh_disable(); | 1514 | local_bh_disable(); |
1514 | nf_conntrack_lock(lockp); | 1515 | nf_conntrack_lock(lockp); |
1515 | if (*bucket < nf_conntrack_htable_size) { | 1516 | if (*bucket < nf_conntrack_htable_size) { |
1516 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[*bucket], hnnode) { | 1517 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[*bucket], hnnode) { |
1517 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) | 1518 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) |
1518 | continue; | 1519 | continue; |
1519 | ct = nf_ct_tuplehash_to_ctrack(h); | 1520 | ct = nf_ct_tuplehash_to_ctrack(h); |
1520 | if (net_eq(nf_ct_net(ct), net) && | 1521 | if (net_eq(nf_ct_net(ct), net) && |
1521 | iter(ct, data)) | 1522 | iter(ct, data)) |
1522 | goto found; | 1523 | goto found; |
1523 | } | 1524 | } |
1524 | } | 1525 | } |
1525 | spin_unlock(lockp); | 1526 | spin_unlock(lockp); |
1526 | local_bh_enable(); | 1527 | local_bh_enable(); |
1527 | cond_resched(); | 1528 | cond_resched(); |
1528 | } | 1529 | } |
1529 | 1530 | ||
1530 | for_each_possible_cpu(cpu) { | 1531 | for_each_possible_cpu(cpu) { |
1531 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | 1532 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); |
1532 | 1533 | ||
1533 | spin_lock_bh(&pcpu->lock); | 1534 | spin_lock_bh(&pcpu->lock); |
1534 | hlist_nulls_for_each_entry(h, n, &pcpu->unconfirmed, hnnode) { | 1535 | hlist_nulls_for_each_entry(h, n, &pcpu->unconfirmed, hnnode) { |
1535 | ct = nf_ct_tuplehash_to_ctrack(h); | 1536 | ct = nf_ct_tuplehash_to_ctrack(h); |
1536 | if (iter(ct, data)) | 1537 | if (iter(ct, data)) |
1537 | set_bit(IPS_DYING_BIT, &ct->status); | 1538 | set_bit(IPS_DYING_BIT, &ct->status); |
1538 | } | 1539 | } |
1539 | spin_unlock_bh(&pcpu->lock); | 1540 | spin_unlock_bh(&pcpu->lock); |
1540 | cond_resched(); | 1541 | cond_resched(); |
1541 | } | 1542 | } |
1542 | return NULL; | 1543 | return NULL; |
1543 | found: | 1544 | found: |
1544 | atomic_inc(&ct->ct_general.use); | 1545 | atomic_inc(&ct->ct_general.use); |
1545 | spin_unlock(lockp); | 1546 | spin_unlock(lockp); |
1546 | local_bh_enable(); | 1547 | local_bh_enable(); |
1547 | return ct; | 1548 | return ct; |
1548 | } | 1549 | } |
1549 | 1550 | ||
1550 | void nf_ct_iterate_cleanup(struct net *net, | 1551 | void nf_ct_iterate_cleanup(struct net *net, |
1551 | int (*iter)(struct nf_conn *i, void *data), | 1552 | int (*iter)(struct nf_conn *i, void *data), |
1552 | void *data, u32 portid, int report) | 1553 | void *data, u32 portid, int report) |
1553 | { | 1554 | { |
1554 | struct nf_conn *ct; | 1555 | struct nf_conn *ct; |
1555 | unsigned int bucket = 0; | 1556 | unsigned int bucket = 0; |
1556 | 1557 | ||
1557 | might_sleep(); | 1558 | might_sleep(); |
1558 | 1559 | ||
1559 | if (atomic_read(&net->ct.count) == 0) | 1560 | if (atomic_read(&net->ct.count) == 0) |
1560 | return; | 1561 | return; |
1561 | 1562 | ||
1562 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { | 1563 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
1563 | /* Time to push up daises... */ | 1564 | /* Time to push up daises... */ |
1564 | 1565 | ||
1565 | nf_ct_delete(ct, portid, report); | 1566 | nf_ct_delete(ct, portid, report); |
1566 | nf_ct_put(ct); | 1567 | nf_ct_put(ct); |
1567 | cond_resched(); | 1568 | cond_resched(); |
1568 | } | 1569 | } |
1569 | } | 1570 | } |
1570 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); | 1571 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
1571 | 1572 | ||
1572 | static int kill_all(struct nf_conn *i, void *data) | 1573 | static int kill_all(struct nf_conn *i, void *data) |
1573 | { | 1574 | { |
1574 | return 1; | 1575 | return 1; |
1575 | } | 1576 | } |
1576 | 1577 | ||
1577 | void nf_ct_free_hashtable(void *hash, unsigned int size) | 1578 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
1578 | { | 1579 | { |
1579 | if (is_vmalloc_addr(hash)) | 1580 | if (is_vmalloc_addr(hash)) |
1580 | vfree(hash); | 1581 | vfree(hash); |
1581 | else | 1582 | else |
1582 | free_pages((unsigned long)hash, | 1583 | free_pages((unsigned long)hash, |
1583 | get_order(sizeof(struct hlist_head) * size)); | 1584 | get_order(sizeof(struct hlist_head) * size)); |
1584 | } | 1585 | } |
1585 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); | 1586 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
1586 | 1587 | ||
1587 | static int untrack_refs(void) | 1588 | static int untrack_refs(void) |
1588 | { | 1589 | { |
1589 | int cnt = 0, cpu; | 1590 | int cnt = 0, cpu; |
1590 | 1591 | ||
1591 | for_each_possible_cpu(cpu) { | 1592 | for_each_possible_cpu(cpu) { |
1592 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | 1593 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); |
1593 | 1594 | ||
1594 | cnt += atomic_read(&ct->ct_general.use) - 1; | 1595 | cnt += atomic_read(&ct->ct_general.use) - 1; |
1595 | } | 1596 | } |
1596 | return cnt; | 1597 | return cnt; |
1597 | } | 1598 | } |
1598 | 1599 | ||
1599 | void nf_conntrack_cleanup_start(void) | 1600 | void nf_conntrack_cleanup_start(void) |
1600 | { | 1601 | { |
1601 | conntrack_gc_work.exiting = true; | 1602 | conntrack_gc_work.exiting = true; |
1602 | RCU_INIT_POINTER(ip_ct_attach, NULL); | 1603 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
1603 | } | 1604 | } |
1604 | 1605 | ||
1605 | void nf_conntrack_cleanup_end(void) | 1606 | void nf_conntrack_cleanup_end(void) |
1606 | { | 1607 | { |
1607 | RCU_INIT_POINTER(nf_ct_destroy, NULL); | 1608 | RCU_INIT_POINTER(nf_ct_destroy, NULL); |
1608 | while (untrack_refs() > 0) | 1609 | while (untrack_refs() > 0) |
1609 | schedule(); | 1610 | schedule(); |
1610 | 1611 | ||
1611 | cancel_delayed_work_sync(&conntrack_gc_work.dwork); | 1612 | cancel_delayed_work_sync(&conntrack_gc_work.dwork); |
1612 | nf_ct_free_hashtable(nf_conntrack_hash, nf_conntrack_htable_size); | 1613 | nf_ct_free_hashtable(nf_conntrack_hash, nf_conntrack_htable_size); |
1613 | 1614 | ||
1614 | nf_conntrack_proto_fini(); | 1615 | nf_conntrack_proto_fini(); |
1615 | nf_conntrack_seqadj_fini(); | 1616 | nf_conntrack_seqadj_fini(); |
1616 | nf_conntrack_labels_fini(); | 1617 | nf_conntrack_labels_fini(); |
1617 | nf_conntrack_helper_fini(); | 1618 | nf_conntrack_helper_fini(); |
1618 | nf_conntrack_timeout_fini(); | 1619 | nf_conntrack_timeout_fini(); |
1619 | nf_conntrack_ecache_fini(); | 1620 | nf_conntrack_ecache_fini(); |
1620 | nf_conntrack_tstamp_fini(); | 1621 | nf_conntrack_tstamp_fini(); |
1621 | nf_conntrack_acct_fini(); | 1622 | nf_conntrack_acct_fini(); |
1622 | nf_conntrack_expect_fini(); | 1623 | nf_conntrack_expect_fini(); |
1623 | 1624 | ||
1624 | kmem_cache_destroy(nf_conntrack_cachep); | 1625 | kmem_cache_destroy(nf_conntrack_cachep); |
1625 | } | 1626 | } |
1626 | 1627 | ||
1627 | /* | 1628 | /* |
1628 | * Mishearing the voices in his head, our hero wonders how he's | 1629 | * Mishearing the voices in his head, our hero wonders how he's |
1629 | * supposed to kill the mall. | 1630 | * supposed to kill the mall. |
1630 | */ | 1631 | */ |
1631 | void nf_conntrack_cleanup_net(struct net *net) | 1632 | void nf_conntrack_cleanup_net(struct net *net) |
1632 | { | 1633 | { |
1633 | LIST_HEAD(single); | 1634 | LIST_HEAD(single); |
1634 | 1635 | ||
1635 | list_add(&net->exit_list, &single); | 1636 | list_add(&net->exit_list, &single); |
1636 | nf_conntrack_cleanup_net_list(&single); | 1637 | nf_conntrack_cleanup_net_list(&single); |
1637 | } | 1638 | } |
1638 | 1639 | ||
1639 | void nf_conntrack_cleanup_net_list(struct list_head *net_exit_list) | 1640 | void nf_conntrack_cleanup_net_list(struct list_head *net_exit_list) |
1640 | { | 1641 | { |
1641 | int busy; | 1642 | int busy; |
1642 | struct net *net; | 1643 | struct net *net; |
1643 | 1644 | ||
1644 | /* | 1645 | /* |
1645 | * This makes sure all current packets have passed through | 1646 | * This makes sure all current packets have passed through |
1646 | * netfilter framework. Roll on, two-stage module | 1647 | * netfilter framework. Roll on, two-stage module |
1647 | * delete... | 1648 | * delete... |
1648 | */ | 1649 | */ |
1649 | synchronize_net(); | 1650 | synchronize_net(); |
1650 | i_see_dead_people: | 1651 | i_see_dead_people: |
1651 | busy = 0; | 1652 | busy = 0; |
1652 | list_for_each_entry(net, net_exit_list, exit_list) { | 1653 | list_for_each_entry(net, net_exit_list, exit_list) { |
1653 | nf_ct_iterate_cleanup(net, kill_all, NULL, 0, 0); | 1654 | nf_ct_iterate_cleanup(net, kill_all, NULL, 0, 0); |
1654 | if (atomic_read(&net->ct.count) != 0) | 1655 | if (atomic_read(&net->ct.count) != 0) |
1655 | busy = 1; | 1656 | busy = 1; |
1656 | } | 1657 | } |
1657 | if (busy) { | 1658 | if (busy) { |
1658 | schedule(); | 1659 | schedule(); |
1659 | goto i_see_dead_people; | 1660 | goto i_see_dead_people; |
1660 | } | 1661 | } |
1661 | 1662 | ||
1662 | list_for_each_entry(net, net_exit_list, exit_list) { | 1663 | list_for_each_entry(net, net_exit_list, exit_list) { |
1663 | nf_conntrack_proto_pernet_fini(net); | 1664 | nf_conntrack_proto_pernet_fini(net); |
1664 | nf_conntrack_helper_pernet_fini(net); | 1665 | nf_conntrack_helper_pernet_fini(net); |
1665 | nf_conntrack_ecache_pernet_fini(net); | 1666 | nf_conntrack_ecache_pernet_fini(net); |
1666 | nf_conntrack_tstamp_pernet_fini(net); | 1667 | nf_conntrack_tstamp_pernet_fini(net); |
1667 | nf_conntrack_acct_pernet_fini(net); | 1668 | nf_conntrack_acct_pernet_fini(net); |
1668 | nf_conntrack_expect_pernet_fini(net); | 1669 | nf_conntrack_expect_pernet_fini(net); |
1669 | free_percpu(net->ct.stat); | 1670 | free_percpu(net->ct.stat); |
1670 | free_percpu(net->ct.pcpu_lists); | 1671 | free_percpu(net->ct.pcpu_lists); |
1671 | } | 1672 | } |
1672 | } | 1673 | } |
1673 | 1674 | ||
1674 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) | 1675 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
1675 | { | 1676 | { |
1676 | struct hlist_nulls_head *hash; | 1677 | struct hlist_nulls_head *hash; |
1677 | unsigned int nr_slots, i; | 1678 | unsigned int nr_slots, i; |
1678 | size_t sz; | 1679 | size_t sz; |
1679 | 1680 | ||
1680 | if (*sizep > (UINT_MAX / sizeof(struct hlist_nulls_head))) | 1681 | if (*sizep > (UINT_MAX / sizeof(struct hlist_nulls_head))) |
1681 | return NULL; | 1682 | return NULL; |
1682 | 1683 | ||
1683 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); | 1684 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1684 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | 1685 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); |
1685 | 1686 | ||
1686 | if (nr_slots > (UINT_MAX / sizeof(struct hlist_nulls_head))) | 1687 | if (nr_slots > (UINT_MAX / sizeof(struct hlist_nulls_head))) |
1687 | return NULL; | 1688 | return NULL; |
1688 | 1689 | ||
1689 | sz = nr_slots * sizeof(struct hlist_nulls_head); | 1690 | sz = nr_slots * sizeof(struct hlist_nulls_head); |
1690 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | 1691 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, |
1691 | get_order(sz)); | 1692 | get_order(sz)); |
1692 | if (!hash) | 1693 | if (!hash) |
1693 | hash = vzalloc(sz); | 1694 | hash = vzalloc(sz); |
1694 | 1695 | ||
1695 | if (hash && nulls) | 1696 | if (hash && nulls) |
1696 | for (i = 0; i < nr_slots; i++) | 1697 | for (i = 0; i < nr_slots; i++) |
1697 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | 1698 | INIT_HLIST_NULLS_HEAD(&hash[i], i); |
1698 | 1699 | ||
1699 | return hash; | 1700 | return hash; |
1700 | } | 1701 | } |
1701 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); | 1702 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
1702 | 1703 | ||
1703 | int nf_conntrack_hash_resize(unsigned int hashsize) | 1704 | int nf_conntrack_hash_resize(unsigned int hashsize) |
1704 | { | 1705 | { |
1705 | int i, bucket; | 1706 | int i, bucket; |
1706 | unsigned int old_size; | 1707 | unsigned int old_size; |
1707 | struct hlist_nulls_head *hash, *old_hash; | 1708 | struct hlist_nulls_head *hash, *old_hash; |
1708 | struct nf_conntrack_tuple_hash *h; | 1709 | struct nf_conntrack_tuple_hash *h; |
1709 | struct nf_conn *ct; | 1710 | struct nf_conn *ct; |
1710 | 1711 | ||
1711 | if (!hashsize) | 1712 | if (!hashsize) |
1712 | return -EINVAL; | 1713 | return -EINVAL; |
1713 | 1714 | ||
1714 | hash = nf_ct_alloc_hashtable(&hashsize, 1); | 1715 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
1715 | if (!hash) | 1716 | if (!hash) |
1716 | return -ENOMEM; | 1717 | return -ENOMEM; |
1717 | 1718 | ||
1718 | old_size = nf_conntrack_htable_size; | 1719 | old_size = nf_conntrack_htable_size; |
1719 | if (old_size == hashsize) { | 1720 | if (old_size == hashsize) { |
1720 | nf_ct_free_hashtable(hash, hashsize); | 1721 | nf_ct_free_hashtable(hash, hashsize); |
1721 | return 0; | 1722 | return 0; |
1722 | } | 1723 | } |
1723 | 1724 | ||
1724 | local_bh_disable(); | 1725 | local_bh_disable(); |
1725 | nf_conntrack_all_lock(); | 1726 | nf_conntrack_all_lock(); |
1726 | write_seqcount_begin(&nf_conntrack_generation); | 1727 | write_seqcount_begin(&nf_conntrack_generation); |
1727 | 1728 | ||
1728 | /* Lookups in the old hash might happen in parallel, which means we | 1729 | /* Lookups in the old hash might happen in parallel, which means we |
1729 | * might get false negatives during connection lookup. New connections | 1730 | * might get false negatives during connection lookup. New connections |
1730 | * created because of a false negative won't make it into the hash | 1731 | * created because of a false negative won't make it into the hash |
1731 | * though since that required taking the locks. | 1732 | * though since that required taking the locks. |
1732 | */ | 1733 | */ |
1733 | 1734 | ||
1734 | for (i = 0; i < nf_conntrack_htable_size; i++) { | 1735 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
1735 | while (!hlist_nulls_empty(&nf_conntrack_hash[i])) { | 1736 | while (!hlist_nulls_empty(&nf_conntrack_hash[i])) { |
1736 | h = hlist_nulls_entry(nf_conntrack_hash[i].first, | 1737 | h = hlist_nulls_entry(nf_conntrack_hash[i].first, |
1737 | struct nf_conntrack_tuple_hash, hnnode); | 1738 | struct nf_conntrack_tuple_hash, hnnode); |
1738 | ct = nf_ct_tuplehash_to_ctrack(h); | 1739 | ct = nf_ct_tuplehash_to_ctrack(h); |
1739 | hlist_nulls_del_rcu(&h->hnnode); | 1740 | hlist_nulls_del_rcu(&h->hnnode); |
1740 | bucket = __hash_conntrack(nf_ct_net(ct), | 1741 | bucket = __hash_conntrack(nf_ct_net(ct), |
1741 | &h->tuple, hashsize); | 1742 | &h->tuple, hashsize); |
1742 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); | 1743 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
1743 | } | 1744 | } |
1744 | } | 1745 | } |
1745 | old_size = nf_conntrack_htable_size; | 1746 | old_size = nf_conntrack_htable_size; |
1746 | old_hash = nf_conntrack_hash; | 1747 | old_hash = nf_conntrack_hash; |
1747 | 1748 | ||
1748 | nf_conntrack_hash = hash; | 1749 | nf_conntrack_hash = hash; |
1749 | nf_conntrack_htable_size = hashsize; | 1750 | nf_conntrack_htable_size = hashsize; |
1750 | 1751 | ||
1751 | write_seqcount_end(&nf_conntrack_generation); | 1752 | write_seqcount_end(&nf_conntrack_generation); |
1752 | nf_conntrack_all_unlock(); | 1753 | nf_conntrack_all_unlock(); |
1753 | local_bh_enable(); | 1754 | local_bh_enable(); |
1754 | 1755 | ||
1755 | synchronize_net(); | 1756 | synchronize_net(); |
1756 | nf_ct_free_hashtable(old_hash, old_size); | 1757 | nf_ct_free_hashtable(old_hash, old_size); |
1757 | return 0; | 1758 | return 0; |
1758 | } | 1759 | } |
1759 | 1760 | ||
1760 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) | 1761 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
1761 | { | 1762 | { |
1762 | unsigned int hashsize; | 1763 | unsigned int hashsize; |
1763 | int rc; | 1764 | int rc; |
1764 | 1765 | ||
1765 | if (current->nsproxy->net_ns != &init_net) | 1766 | if (current->nsproxy->net_ns != &init_net) |
1766 | return -EOPNOTSUPP; | 1767 | return -EOPNOTSUPP; |
1767 | 1768 | ||
1768 | /* On boot, we can set this without any fancy locking. */ | 1769 | /* On boot, we can set this without any fancy locking. */ |
1769 | if (!nf_conntrack_htable_size) | 1770 | if (!nf_conntrack_htable_size) |
1770 | return param_set_uint(val, kp); | 1771 | return param_set_uint(val, kp); |
1771 | 1772 | ||
1772 | rc = kstrtouint(val, 0, &hashsize); | 1773 | rc = kstrtouint(val, 0, &hashsize); |
1773 | if (rc) | 1774 | if (rc) |
1774 | return rc; | 1775 | return rc; |
1775 | 1776 | ||
1776 | return nf_conntrack_hash_resize(hashsize); | 1777 | return nf_conntrack_hash_resize(hashsize); |
1777 | } | 1778 | } |
1778 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); | 1779 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
1779 | 1780 | ||
1780 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, | 1781 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
1781 | &nf_conntrack_htable_size, 0600); | 1782 | &nf_conntrack_htable_size, 0600); |
1782 | 1783 | ||
1783 | void nf_ct_untracked_status_or(unsigned long bits) | 1784 | void nf_ct_untracked_status_or(unsigned long bits) |
1784 | { | 1785 | { |
1785 | int cpu; | 1786 | int cpu; |
1786 | 1787 | ||
1787 | for_each_possible_cpu(cpu) | 1788 | for_each_possible_cpu(cpu) |
1788 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; | 1789 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; |
1789 | } | 1790 | } |
1790 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); | 1791 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); |
1791 | 1792 | ||
1792 | int nf_conntrack_init_start(void) | 1793 | int nf_conntrack_init_start(void) |
1793 | { | 1794 | { |
1794 | int max_factor = 8; | 1795 | int max_factor = 8; |
1795 | int ret = -ENOMEM; | 1796 | int ret = -ENOMEM; |
1796 | int i, cpu; | 1797 | int i, cpu; |
1797 | 1798 | ||
1798 | seqcount_init(&nf_conntrack_generation); | 1799 | seqcount_init(&nf_conntrack_generation); |
1799 | 1800 | ||
1800 | for (i = 0; i < CONNTRACK_LOCKS; i++) | 1801 | for (i = 0; i < CONNTRACK_LOCKS; i++) |
1801 | spin_lock_init(&nf_conntrack_locks[i]); | 1802 | spin_lock_init(&nf_conntrack_locks[i]); |
1802 | 1803 | ||
1803 | if (!nf_conntrack_htable_size) { | 1804 | if (!nf_conntrack_htable_size) { |
1804 | /* Idea from tcp.c: use 1/16384 of memory. | 1805 | /* Idea from tcp.c: use 1/16384 of memory. |
1805 | * On i386: 32MB machine has 512 buckets. | 1806 | * On i386: 32MB machine has 512 buckets. |
1806 | * >= 1GB machines have 16384 buckets. | 1807 | * >= 1GB machines have 16384 buckets. |
1807 | * >= 4GB machines have 65536 buckets. | 1808 | * >= 4GB machines have 65536 buckets. |
1808 | */ | 1809 | */ |
1809 | nf_conntrack_htable_size | 1810 | nf_conntrack_htable_size |
1810 | = (((totalram_pages << PAGE_SHIFT) / 16384) | 1811 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
1811 | / sizeof(struct hlist_head)); | 1812 | / sizeof(struct hlist_head)); |
1812 | if (totalram_pages > (4 * (1024 * 1024 * 1024 / PAGE_SIZE))) | 1813 | if (totalram_pages > (4 * (1024 * 1024 * 1024 / PAGE_SIZE))) |
1813 | nf_conntrack_htable_size = 65536; | 1814 | nf_conntrack_htable_size = 65536; |
1814 | else if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) | 1815 | else if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
1815 | nf_conntrack_htable_size = 16384; | 1816 | nf_conntrack_htable_size = 16384; |
1816 | if (nf_conntrack_htable_size < 32) | 1817 | if (nf_conntrack_htable_size < 32) |
1817 | nf_conntrack_htable_size = 32; | 1818 | nf_conntrack_htable_size = 32; |
1818 | 1819 | ||
1819 | /* Use a max. factor of four by default to get the same max as | 1820 | /* Use a max. factor of four by default to get the same max as |
1820 | * with the old struct list_heads. When a table size is given | 1821 | * with the old struct list_heads. When a table size is given |
1821 | * we use the old value of 8 to avoid reducing the max. | 1822 | * we use the old value of 8 to avoid reducing the max. |
1822 | * entries. */ | 1823 | * entries. */ |
1823 | max_factor = 4; | 1824 | max_factor = 4; |
1824 | } | 1825 | } |
1825 | 1826 | ||
1826 | nf_conntrack_hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size, 1); | 1827 | nf_conntrack_hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size, 1); |
1827 | if (!nf_conntrack_hash) | 1828 | if (!nf_conntrack_hash) |
1828 | return -ENOMEM; | 1829 | return -ENOMEM; |
1829 | 1830 | ||
1830 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; | 1831 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
1831 | 1832 | ||
1832 | nf_conntrack_cachep = kmem_cache_create("nf_conntrack", | 1833 | nf_conntrack_cachep = kmem_cache_create("nf_conntrack", |
1833 | sizeof(struct nf_conn), 0, | 1834 | sizeof(struct nf_conn), 0, |
1834 | SLAB_DESTROY_BY_RCU | SLAB_HWCACHE_ALIGN, NULL); | 1835 | SLAB_DESTROY_BY_RCU | SLAB_HWCACHE_ALIGN, NULL); |
1835 | if (!nf_conntrack_cachep) | 1836 | if (!nf_conntrack_cachep) |
1836 | goto err_cachep; | 1837 | goto err_cachep; |
1837 | 1838 | ||
1838 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", | 1839 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
1839 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, | 1840 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
1840 | nf_conntrack_max); | 1841 | nf_conntrack_max); |
1841 | 1842 | ||
1842 | ret = nf_conntrack_expect_init(); | 1843 | ret = nf_conntrack_expect_init(); |
1843 | if (ret < 0) | 1844 | if (ret < 0) |
1844 | goto err_expect; | 1845 | goto err_expect; |
1845 | 1846 | ||
1846 | ret = nf_conntrack_acct_init(); | 1847 | ret = nf_conntrack_acct_init(); |
1847 | if (ret < 0) | 1848 | if (ret < 0) |
1848 | goto err_acct; | 1849 | goto err_acct; |
1849 | 1850 | ||
1850 | ret = nf_conntrack_tstamp_init(); | 1851 | ret = nf_conntrack_tstamp_init(); |
1851 | if (ret < 0) | 1852 | if (ret < 0) |
1852 | goto err_tstamp; | 1853 | goto err_tstamp; |
1853 | 1854 | ||
1854 | ret = nf_conntrack_ecache_init(); | 1855 | ret = nf_conntrack_ecache_init(); |
1855 | if (ret < 0) | 1856 | if (ret < 0) |
1856 | goto err_ecache; | 1857 | goto err_ecache; |
1857 | 1858 | ||
1858 | ret = nf_conntrack_timeout_init(); | 1859 | ret = nf_conntrack_timeout_init(); |
1859 | if (ret < 0) | 1860 | if (ret < 0) |
1860 | goto err_timeout; | 1861 | goto err_timeout; |
1861 | 1862 | ||
1862 | ret = nf_conntrack_helper_init(); | 1863 | ret = nf_conntrack_helper_init(); |
1863 | if (ret < 0) | 1864 | if (ret < 0) |
1864 | goto err_helper; | 1865 | goto err_helper; |
1865 | 1866 | ||
1866 | ret = nf_conntrack_labels_init(); | 1867 | ret = nf_conntrack_labels_init(); |
1867 | if (ret < 0) | 1868 | if (ret < 0) |
1868 | goto err_labels; | 1869 | goto err_labels; |
1869 | 1870 | ||
1870 | ret = nf_conntrack_seqadj_init(); | 1871 | ret = nf_conntrack_seqadj_init(); |
1871 | if (ret < 0) | 1872 | if (ret < 0) |
1872 | goto err_seqadj; | 1873 | goto err_seqadj; |
1873 | 1874 | ||
1874 | ret = nf_conntrack_proto_init(); | 1875 | ret = nf_conntrack_proto_init(); |
1875 | if (ret < 0) | 1876 | if (ret < 0) |
1876 | goto err_proto; | 1877 | goto err_proto; |
1877 | 1878 | ||
1878 | /* Set up fake conntrack: to never be deleted, not in any hashes */ | 1879 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
1879 | for_each_possible_cpu(cpu) { | 1880 | for_each_possible_cpu(cpu) { |
1880 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | 1881 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); |
1881 | write_pnet(&ct->ct_net, &init_net); | 1882 | write_pnet(&ct->ct_net, &init_net); |
1882 | atomic_set(&ct->ct_general.use, 1); | 1883 | atomic_set(&ct->ct_general.use, 1); |
1883 | } | 1884 | } |
1884 | /* - and look it like as a confirmed connection */ | 1885 | /* - and look it like as a confirmed connection */ |
1885 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); | 1886 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
1886 | 1887 | ||
1887 | conntrack_gc_work_init(&conntrack_gc_work); | 1888 | conntrack_gc_work_init(&conntrack_gc_work); |
1888 | schedule_delayed_work(&conntrack_gc_work.dwork, GC_INTERVAL); | 1889 | schedule_delayed_work(&conntrack_gc_work.dwork, GC_INTERVAL); |
1889 | 1890 | ||
1890 | return 0; | 1891 | return 0; |
1891 | 1892 | ||
1892 | err_proto: | 1893 | err_proto: |
1893 | nf_conntrack_seqadj_fini(); | 1894 | nf_conntrack_seqadj_fini(); |
1894 | err_seqadj: | 1895 | err_seqadj: |
1895 | nf_conntrack_labels_fini(); | 1896 | nf_conntrack_labels_fini(); |
1896 | err_labels: | 1897 | err_labels: |
1897 | nf_conntrack_helper_fini(); | 1898 | nf_conntrack_helper_fini(); |
1898 | err_helper: | 1899 | err_helper: |
1899 | nf_conntrack_timeout_fini(); | 1900 | nf_conntrack_timeout_fini(); |
1900 | err_timeout: | 1901 | err_timeout: |
1901 | nf_conntrack_ecache_fini(); | 1902 | nf_conntrack_ecache_fini(); |
1902 | err_ecache: | 1903 | err_ecache: |
1903 | nf_conntrack_tstamp_fini(); | 1904 | nf_conntrack_tstamp_fini(); |
1904 | err_tstamp: | 1905 | err_tstamp: |
1905 | nf_conntrack_acct_fini(); | 1906 | nf_conntrack_acct_fini(); |
1906 | err_acct: | 1907 | err_acct: |
1907 | nf_conntrack_expect_fini(); | 1908 | nf_conntrack_expect_fini(); |
1908 | err_expect: | 1909 | err_expect: |
1909 | kmem_cache_destroy(nf_conntrack_cachep); | 1910 | kmem_cache_destroy(nf_conntrack_cachep); |
1910 | err_cachep: | 1911 | err_cachep: |
1911 | nf_ct_free_hashtable(nf_conntrack_hash, nf_conntrack_htable_size); | 1912 | nf_ct_free_hashtable(nf_conntrack_hash, nf_conntrack_htable_size); |
1912 | return ret; | 1913 | return ret; |
1913 | } | 1914 | } |
1914 | 1915 | ||
1915 | void nf_conntrack_init_end(void) | 1916 | void nf_conntrack_init_end(void) |
1916 | { | 1917 | { |
1917 | /* For use by REJECT target */ | 1918 | /* For use by REJECT target */ |
1918 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); | 1919 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); |
1919 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); | 1920 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); |
1920 | } | 1921 | } |
1921 | 1922 | ||
1922 | /* | 1923 | /* |
1923 | * We need to use special "null" values, not used in hash table | 1924 | * We need to use special "null" values, not used in hash table |
1924 | */ | 1925 | */ |
1925 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | 1926 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) |
1926 | #define DYING_NULLS_VAL ((1<<30)+1) | 1927 | #define DYING_NULLS_VAL ((1<<30)+1) |
1927 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) | 1928 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) |
1928 | 1929 | ||
1929 | int nf_conntrack_init_net(struct net *net) | 1930 | int nf_conntrack_init_net(struct net *net) |
1930 | { | 1931 | { |
1931 | int ret = -ENOMEM; | 1932 | int ret = -ENOMEM; |
1932 | int cpu; | 1933 | int cpu; |
1933 | 1934 | ||
1934 | atomic_set(&net->ct.count, 0); | 1935 | atomic_set(&net->ct.count, 0); |
1935 | 1936 | ||
1936 | net->ct.pcpu_lists = alloc_percpu(struct ct_pcpu); | 1937 | net->ct.pcpu_lists = alloc_percpu(struct ct_pcpu); |
1937 | if (!net->ct.pcpu_lists) | 1938 | if (!net->ct.pcpu_lists) |
1938 | goto err_stat; | 1939 | goto err_stat; |
1939 | 1940 | ||
1940 | for_each_possible_cpu(cpu) { | 1941 | for_each_possible_cpu(cpu) { |
1941 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | 1942 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); |
1942 | 1943 | ||
1943 | spin_lock_init(&pcpu->lock); | 1944 | spin_lock_init(&pcpu->lock); |
1944 | INIT_HLIST_NULLS_HEAD(&pcpu->unconfirmed, UNCONFIRMED_NULLS_VAL); | 1945 | INIT_HLIST_NULLS_HEAD(&pcpu->unconfirmed, UNCONFIRMED_NULLS_VAL); |
1945 | INIT_HLIST_NULLS_HEAD(&pcpu->dying, DYING_NULLS_VAL); | 1946 | INIT_HLIST_NULLS_HEAD(&pcpu->dying, DYING_NULLS_VAL); |
1946 | } | 1947 | } |
1947 | 1948 | ||
1948 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); | 1949 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1949 | if (!net->ct.stat) | 1950 | if (!net->ct.stat) |
1950 | goto err_pcpu_lists; | 1951 | goto err_pcpu_lists; |
1951 | 1952 | ||
1952 | ret = nf_conntrack_expect_pernet_init(net); | 1953 | ret = nf_conntrack_expect_pernet_init(net); |
1953 | if (ret < 0) | 1954 | if (ret < 0) |
1954 | goto err_expect; | 1955 | goto err_expect; |
1955 | ret = nf_conntrack_acct_pernet_init(net); | 1956 | ret = nf_conntrack_acct_pernet_init(net); |
1956 | if (ret < 0) | 1957 | if (ret < 0) |
1957 | goto err_acct; | 1958 | goto err_acct; |
1958 | ret = nf_conntrack_tstamp_pernet_init(net); | 1959 | ret = nf_conntrack_tstamp_pernet_init(net); |
1959 | if (ret < 0) | 1960 | if (ret < 0) |
1960 | goto err_tstamp; | 1961 | goto err_tstamp; |
1961 | ret = nf_conntrack_ecache_pernet_init(net); | 1962 | ret = nf_conntrack_ecache_pernet_init(net); |
1962 | if (ret < 0) | 1963 | if (ret < 0) |
1963 | goto err_ecache; | 1964 | goto err_ecache; |
1964 | ret = nf_conntrack_helper_pernet_init(net); | 1965 | ret = nf_conntrack_helper_pernet_init(net); |
1965 | if (ret < 0) | 1966 | if (ret < 0) |
1966 | goto err_helper; | 1967 | goto err_helper; |
1967 | ret = nf_conntrack_proto_pernet_init(net); | 1968 | ret = nf_conntrack_proto_pernet_init(net); |
1968 | if (ret < 0) | 1969 | if (ret < 0) |
1969 | goto err_proto; | 1970 | goto err_proto; |
1970 | return 0; | 1971 | return 0; |
1971 | 1972 | ||
1972 | err_proto: | 1973 | err_proto: |
1973 | nf_conntrack_helper_pernet_fini(net); | 1974 | nf_conntrack_helper_pernet_fini(net); |
1974 | err_helper: | 1975 | err_helper: |
1975 | nf_conntrack_ecache_pernet_fini(net); | 1976 | nf_conntrack_ecache_pernet_fini(net); |
1976 | err_ecache: | 1977 | err_ecache: |
1977 | nf_conntrack_tstamp_pernet_fini(net); | 1978 | nf_conntrack_tstamp_pernet_fini(net); |
1978 | err_tstamp: | 1979 | err_tstamp: |
1979 | nf_conntrack_acct_pernet_fini(net); | 1980 | nf_conntrack_acct_pernet_fini(net); |
1980 | err_acct: | 1981 | err_acct: |
1981 | nf_conntrack_expect_pernet_fini(net); | 1982 | nf_conntrack_expect_pernet_fini(net); |
1982 | err_expect: | 1983 | err_expect: |
1983 | free_percpu(net->ct.stat); | 1984 | free_percpu(net->ct.stat); |
1984 | err_pcpu_lists: | 1985 | err_pcpu_lists: |
1985 | free_percpu(net->ct.pcpu_lists); | 1986 | free_percpu(net->ct.pcpu_lists); |
1986 | err_stat: | 1987 | err_stat: |
1987 | return ret; | 1988 | return ret; |
1988 | } | 1989 | } |
1989 | 1990 |
net/openvswitch/conntrack.c
1 | /* | 1 | /* |
2 | * Copyright (c) 2015 Nicira, Inc. | 2 | * Copyright (c) 2015 Nicira, Inc. |
3 | * | 3 | * |
4 | * This program is free software; you can redistribute it and/or | 4 | * This program is free software; you can redistribute it and/or |
5 | * modify it under the terms of version 2 of the GNU General Public | 5 | * modify it under the terms of version 2 of the GNU General Public |
6 | * License as published by the Free Software Foundation. | 6 | * License as published by the Free Software Foundation. |
7 | * | 7 | * |
8 | * This program is distributed in the hope that it will be useful, but | 8 | * This program is distributed in the hope that it will be useful, but |
9 | * WITHOUT ANY WARRANTY; without even the implied warranty of | 9 | * WITHOUT ANY WARRANTY; without even the implied warranty of |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | 10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
11 | * General Public License for more details. | 11 | * General Public License for more details. |
12 | */ | 12 | */ |
13 | 13 | ||
14 | #include <linux/module.h> | 14 | #include <linux/module.h> |
15 | #include <linux/openvswitch.h> | 15 | #include <linux/openvswitch.h> |
16 | #include <linux/tcp.h> | 16 | #include <linux/tcp.h> |
17 | #include <linux/udp.h> | 17 | #include <linux/udp.h> |
18 | #include <linux/sctp.h> | 18 | #include <linux/sctp.h> |
19 | #include <net/ip.h> | 19 | #include <net/ip.h> |
20 | #include <net/netfilter/nf_conntrack_core.h> | 20 | #include <net/netfilter/nf_conntrack_core.h> |
21 | #include <net/netfilter/nf_conntrack_helper.h> | 21 | #include <net/netfilter/nf_conntrack_helper.h> |
22 | #include <net/netfilter/nf_conntrack_labels.h> | 22 | #include <net/netfilter/nf_conntrack_labels.h> |
23 | #include <net/netfilter/nf_conntrack_seqadj.h> | 23 | #include <net/netfilter/nf_conntrack_seqadj.h> |
24 | #include <net/netfilter/nf_conntrack_zones.h> | 24 | #include <net/netfilter/nf_conntrack_zones.h> |
25 | #include <net/netfilter/ipv6/nf_defrag_ipv6.h> | 25 | #include <net/netfilter/ipv6/nf_defrag_ipv6.h> |
26 | 26 | ||
27 | #ifdef CONFIG_NF_NAT_NEEDED | 27 | #ifdef CONFIG_NF_NAT_NEEDED |
28 | #include <linux/netfilter/nf_nat.h> | 28 | #include <linux/netfilter/nf_nat.h> |
29 | #include <net/netfilter/nf_nat_core.h> | 29 | #include <net/netfilter/nf_nat_core.h> |
30 | #include <net/netfilter/nf_nat_l3proto.h> | 30 | #include <net/netfilter/nf_nat_l3proto.h> |
31 | #endif | 31 | #endif |
32 | 32 | ||
33 | #include "datapath.h" | 33 | #include "datapath.h" |
34 | #include "conntrack.h" | 34 | #include "conntrack.h" |
35 | #include "flow.h" | 35 | #include "flow.h" |
36 | #include "flow_netlink.h" | 36 | #include "flow_netlink.h" |
37 | 37 | ||
38 | struct ovs_ct_len_tbl { | 38 | struct ovs_ct_len_tbl { |
39 | int maxlen; | 39 | int maxlen; |
40 | int minlen; | 40 | int minlen; |
41 | }; | 41 | }; |
42 | 42 | ||
43 | /* Metadata mark for masked write to conntrack mark */ | 43 | /* Metadata mark for masked write to conntrack mark */ |
44 | struct md_mark { | 44 | struct md_mark { |
45 | u32 value; | 45 | u32 value; |
46 | u32 mask; | 46 | u32 mask; |
47 | }; | 47 | }; |
48 | 48 | ||
49 | /* Metadata label for masked write to conntrack label. */ | 49 | /* Metadata label for masked write to conntrack label. */ |
50 | struct md_labels { | 50 | struct md_labels { |
51 | struct ovs_key_ct_labels value; | 51 | struct ovs_key_ct_labels value; |
52 | struct ovs_key_ct_labels mask; | 52 | struct ovs_key_ct_labels mask; |
53 | }; | 53 | }; |
54 | 54 | ||
55 | enum ovs_ct_nat { | 55 | enum ovs_ct_nat { |
56 | OVS_CT_NAT = 1 << 0, /* NAT for committed connections only. */ | 56 | OVS_CT_NAT = 1 << 0, /* NAT for committed connections only. */ |
57 | OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */ | 57 | OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */ |
58 | OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */ | 58 | OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */ |
59 | }; | 59 | }; |
60 | 60 | ||
61 | /* Conntrack action context for execution. */ | 61 | /* Conntrack action context for execution. */ |
62 | struct ovs_conntrack_info { | 62 | struct ovs_conntrack_info { |
63 | struct nf_conntrack_helper *helper; | 63 | struct nf_conntrack_helper *helper; |
64 | struct nf_conntrack_zone zone; | 64 | struct nf_conntrack_zone zone; |
65 | struct nf_conn *ct; | 65 | struct nf_conn *ct; |
66 | u8 commit : 1; | 66 | u8 commit : 1; |
67 | u8 nat : 3; /* enum ovs_ct_nat */ | 67 | u8 nat : 3; /* enum ovs_ct_nat */ |
68 | u16 family; | 68 | u16 family; |
69 | struct md_mark mark; | 69 | struct md_mark mark; |
70 | struct md_labels labels; | 70 | struct md_labels labels; |
71 | #ifdef CONFIG_NF_NAT_NEEDED | 71 | #ifdef CONFIG_NF_NAT_NEEDED |
72 | struct nf_nat_range range; /* Only present for SRC NAT and DST NAT. */ | 72 | struct nf_nat_range range; /* Only present for SRC NAT and DST NAT. */ |
73 | #endif | 73 | #endif |
74 | }; | 74 | }; |
75 | 75 | ||
76 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info); | 76 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info); |
77 | 77 | ||
78 | static u16 key_to_nfproto(const struct sw_flow_key *key) | 78 | static u16 key_to_nfproto(const struct sw_flow_key *key) |
79 | { | 79 | { |
80 | switch (ntohs(key->eth.type)) { | 80 | switch (ntohs(key->eth.type)) { |
81 | case ETH_P_IP: | 81 | case ETH_P_IP: |
82 | return NFPROTO_IPV4; | 82 | return NFPROTO_IPV4; |
83 | case ETH_P_IPV6: | 83 | case ETH_P_IPV6: |
84 | return NFPROTO_IPV6; | 84 | return NFPROTO_IPV6; |
85 | default: | 85 | default: |
86 | return NFPROTO_UNSPEC; | 86 | return NFPROTO_UNSPEC; |
87 | } | 87 | } |
88 | } | 88 | } |
89 | 89 | ||
90 | /* Map SKB connection state into the values used by flow definition. */ | 90 | /* Map SKB connection state into the values used by flow definition. */ |
91 | static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo) | 91 | static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo) |
92 | { | 92 | { |
93 | u8 ct_state = OVS_CS_F_TRACKED; | 93 | u8 ct_state = OVS_CS_F_TRACKED; |
94 | 94 | ||
95 | switch (ctinfo) { | 95 | switch (ctinfo) { |
96 | case IP_CT_ESTABLISHED_REPLY: | 96 | case IP_CT_ESTABLISHED_REPLY: |
97 | case IP_CT_RELATED_REPLY: | 97 | case IP_CT_RELATED_REPLY: |
98 | ct_state |= OVS_CS_F_REPLY_DIR; | 98 | ct_state |= OVS_CS_F_REPLY_DIR; |
99 | break; | 99 | break; |
100 | default: | 100 | default: |
101 | break; | 101 | break; |
102 | } | 102 | } |
103 | 103 | ||
104 | switch (ctinfo) { | 104 | switch (ctinfo) { |
105 | case IP_CT_ESTABLISHED: | 105 | case IP_CT_ESTABLISHED: |
106 | case IP_CT_ESTABLISHED_REPLY: | 106 | case IP_CT_ESTABLISHED_REPLY: |
107 | ct_state |= OVS_CS_F_ESTABLISHED; | 107 | ct_state |= OVS_CS_F_ESTABLISHED; |
108 | break; | 108 | break; |
109 | case IP_CT_RELATED: | 109 | case IP_CT_RELATED: |
110 | case IP_CT_RELATED_REPLY: | 110 | case IP_CT_RELATED_REPLY: |
111 | ct_state |= OVS_CS_F_RELATED; | 111 | ct_state |= OVS_CS_F_RELATED; |
112 | break; | 112 | break; |
113 | case IP_CT_NEW: | 113 | case IP_CT_NEW: |
114 | ct_state |= OVS_CS_F_NEW; | 114 | ct_state |= OVS_CS_F_NEW; |
115 | break; | 115 | break; |
116 | default: | 116 | default: |
117 | break; | 117 | break; |
118 | } | 118 | } |
119 | 119 | ||
120 | return ct_state; | 120 | return ct_state; |
121 | } | 121 | } |
122 | 122 | ||
123 | static u32 ovs_ct_get_mark(const struct nf_conn *ct) | 123 | static u32 ovs_ct_get_mark(const struct nf_conn *ct) |
124 | { | 124 | { |
125 | #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) | 125 | #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) |
126 | return ct ? ct->mark : 0; | 126 | return ct ? ct->mark : 0; |
127 | #else | 127 | #else |
128 | return 0; | 128 | return 0; |
129 | #endif | 129 | #endif |
130 | } | 130 | } |
131 | 131 | ||
132 | static void ovs_ct_get_labels(const struct nf_conn *ct, | 132 | static void ovs_ct_get_labels(const struct nf_conn *ct, |
133 | struct ovs_key_ct_labels *labels) | 133 | struct ovs_key_ct_labels *labels) |
134 | { | 134 | { |
135 | struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL; | 135 | struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL; |
136 | 136 | ||
137 | if (cl) { | 137 | if (cl) { |
138 | size_t len = sizeof(cl->bits); | 138 | size_t len = sizeof(cl->bits); |
139 | 139 | ||
140 | if (len > OVS_CT_LABELS_LEN) | 140 | if (len > OVS_CT_LABELS_LEN) |
141 | len = OVS_CT_LABELS_LEN; | 141 | len = OVS_CT_LABELS_LEN; |
142 | else if (len < OVS_CT_LABELS_LEN) | 142 | else if (len < OVS_CT_LABELS_LEN) |
143 | memset(labels, 0, OVS_CT_LABELS_LEN); | 143 | memset(labels, 0, OVS_CT_LABELS_LEN); |
144 | memcpy(labels, cl->bits, len); | 144 | memcpy(labels, cl->bits, len); |
145 | } else { | 145 | } else { |
146 | memset(labels, 0, OVS_CT_LABELS_LEN); | 146 | memset(labels, 0, OVS_CT_LABELS_LEN); |
147 | } | 147 | } |
148 | } | 148 | } |
149 | 149 | ||
150 | static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state, | 150 | static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state, |
151 | const struct nf_conntrack_zone *zone, | 151 | const struct nf_conntrack_zone *zone, |
152 | const struct nf_conn *ct) | 152 | const struct nf_conn *ct) |
153 | { | 153 | { |
154 | key->ct.state = state; | 154 | key->ct.state = state; |
155 | key->ct.zone = zone->id; | 155 | key->ct.zone = zone->id; |
156 | key->ct.mark = ovs_ct_get_mark(ct); | 156 | key->ct.mark = ovs_ct_get_mark(ct); |
157 | ovs_ct_get_labels(ct, &key->ct.labels); | 157 | ovs_ct_get_labels(ct, &key->ct.labels); |
158 | } | 158 | } |
159 | 159 | ||
160 | /* Update 'key' based on skb->nfct. If 'post_ct' is true, then OVS has | 160 | /* Update 'key' based on skb->nfct. If 'post_ct' is true, then OVS has |
161 | * previously sent the packet to conntrack via the ct action. If | 161 | * previously sent the packet to conntrack via the ct action. If |
162 | * 'keep_nat_flags' is true, the existing NAT flags retained, else they are | 162 | * 'keep_nat_flags' is true, the existing NAT flags retained, else they are |
163 | * initialized from the connection status. | 163 | * initialized from the connection status. |
164 | */ | 164 | */ |
165 | static void ovs_ct_update_key(const struct sk_buff *skb, | 165 | static void ovs_ct_update_key(const struct sk_buff *skb, |
166 | const struct ovs_conntrack_info *info, | 166 | const struct ovs_conntrack_info *info, |
167 | struct sw_flow_key *key, bool post_ct, | 167 | struct sw_flow_key *key, bool post_ct, |
168 | bool keep_nat_flags) | 168 | bool keep_nat_flags) |
169 | { | 169 | { |
170 | const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt; | 170 | const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt; |
171 | enum ip_conntrack_info ctinfo; | 171 | enum ip_conntrack_info ctinfo; |
172 | struct nf_conn *ct; | 172 | struct nf_conn *ct; |
173 | u8 state = 0; | 173 | u8 state = 0; |
174 | 174 | ||
175 | ct = nf_ct_get(skb, &ctinfo); | 175 | ct = nf_ct_get(skb, &ctinfo); |
176 | if (ct) { | 176 | if (ct) { |
177 | state = ovs_ct_get_state(ctinfo); | 177 | state = ovs_ct_get_state(ctinfo); |
178 | /* All unconfirmed entries are NEW connections. */ | 178 | /* All unconfirmed entries are NEW connections. */ |
179 | if (!nf_ct_is_confirmed(ct)) | 179 | if (!nf_ct_is_confirmed(ct)) |
180 | state |= OVS_CS_F_NEW; | 180 | state |= OVS_CS_F_NEW; |
181 | /* OVS persists the related flag for the duration of the | 181 | /* OVS persists the related flag for the duration of the |
182 | * connection. | 182 | * connection. |
183 | */ | 183 | */ |
184 | if (ct->master) | 184 | if (ct->master) |
185 | state |= OVS_CS_F_RELATED; | 185 | state |= OVS_CS_F_RELATED; |
186 | if (keep_nat_flags) { | 186 | if (keep_nat_flags) { |
187 | state |= key->ct.state & OVS_CS_F_NAT_MASK; | 187 | state |= key->ct.state & OVS_CS_F_NAT_MASK; |
188 | } else { | 188 | } else { |
189 | if (ct->status & IPS_SRC_NAT) | 189 | if (ct->status & IPS_SRC_NAT) |
190 | state |= OVS_CS_F_SRC_NAT; | 190 | state |= OVS_CS_F_SRC_NAT; |
191 | if (ct->status & IPS_DST_NAT) | 191 | if (ct->status & IPS_DST_NAT) |
192 | state |= OVS_CS_F_DST_NAT; | 192 | state |= OVS_CS_F_DST_NAT; |
193 | } | 193 | } |
194 | zone = nf_ct_zone(ct); | 194 | zone = nf_ct_zone(ct); |
195 | } else if (post_ct) { | 195 | } else if (post_ct) { |
196 | state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID; | 196 | state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID; |
197 | if (info) | 197 | if (info) |
198 | zone = &info->zone; | 198 | zone = &info->zone; |
199 | } | 199 | } |
200 | __ovs_ct_update_key(key, state, zone, ct); | 200 | __ovs_ct_update_key(key, state, zone, ct); |
201 | } | 201 | } |
202 | 202 | ||
203 | /* This is called to initialize CT key fields possibly coming in from the local | 203 | /* This is called to initialize CT key fields possibly coming in from the local |
204 | * stack. | 204 | * stack. |
205 | */ | 205 | */ |
206 | void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key) | 206 | void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key) |
207 | { | 207 | { |
208 | ovs_ct_update_key(skb, NULL, key, false, false); | 208 | ovs_ct_update_key(skb, NULL, key, false, false); |
209 | } | 209 | } |
210 | 210 | ||
211 | int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb) | 211 | int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb) |
212 | { | 212 | { |
213 | if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state)) | 213 | if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state)) |
214 | return -EMSGSIZE; | 214 | return -EMSGSIZE; |
215 | 215 | ||
216 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | 216 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && |
217 | nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone)) | 217 | nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone)) |
218 | return -EMSGSIZE; | 218 | return -EMSGSIZE; |
219 | 219 | ||
220 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && | 220 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && |
221 | nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark)) | 221 | nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark)) |
222 | return -EMSGSIZE; | 222 | return -EMSGSIZE; |
223 | 223 | ||
224 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && | 224 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
225 | nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels), | 225 | nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels), |
226 | &key->ct.labels)) | 226 | &key->ct.labels)) |
227 | return -EMSGSIZE; | 227 | return -EMSGSIZE; |
228 | 228 | ||
229 | return 0; | 229 | return 0; |
230 | } | 230 | } |
231 | 231 | ||
232 | static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key, | 232 | static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key, |
233 | u32 ct_mark, u32 mask) | 233 | u32 ct_mark, u32 mask) |
234 | { | 234 | { |
235 | #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) | 235 | #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) |
236 | enum ip_conntrack_info ctinfo; | 236 | enum ip_conntrack_info ctinfo; |
237 | struct nf_conn *ct; | 237 | struct nf_conn *ct; |
238 | u32 new_mark; | 238 | u32 new_mark; |
239 | 239 | ||
240 | /* The connection could be invalid, in which case set_mark is no-op. */ | 240 | /* The connection could be invalid, in which case set_mark is no-op. */ |
241 | ct = nf_ct_get(skb, &ctinfo); | 241 | ct = nf_ct_get(skb, &ctinfo); |
242 | if (!ct) | 242 | if (!ct) |
243 | return 0; | 243 | return 0; |
244 | 244 | ||
245 | new_mark = ct_mark | (ct->mark & ~(mask)); | 245 | new_mark = ct_mark | (ct->mark & ~(mask)); |
246 | if (ct->mark != new_mark) { | 246 | if (ct->mark != new_mark) { |
247 | ct->mark = new_mark; | 247 | ct->mark = new_mark; |
248 | nf_conntrack_event_cache(IPCT_MARK, ct); | 248 | nf_conntrack_event_cache(IPCT_MARK, ct); |
249 | key->ct.mark = new_mark; | 249 | key->ct.mark = new_mark; |
250 | } | 250 | } |
251 | 251 | ||
252 | return 0; | 252 | return 0; |
253 | #else | 253 | #else |
254 | return -ENOTSUPP; | 254 | return -ENOTSUPP; |
255 | #endif | 255 | #endif |
256 | } | 256 | } |
257 | 257 | ||
258 | static int ovs_ct_set_labels(struct sk_buff *skb, struct sw_flow_key *key, | 258 | static int ovs_ct_set_labels(struct sk_buff *skb, struct sw_flow_key *key, |
259 | const struct ovs_key_ct_labels *labels, | 259 | const struct ovs_key_ct_labels *labels, |
260 | const struct ovs_key_ct_labels *mask) | 260 | const struct ovs_key_ct_labels *mask) |
261 | { | 261 | { |
262 | enum ip_conntrack_info ctinfo; | 262 | enum ip_conntrack_info ctinfo; |
263 | struct nf_conn_labels *cl; | 263 | struct nf_conn_labels *cl; |
264 | struct nf_conn *ct; | 264 | struct nf_conn *ct; |
265 | int err; | 265 | int err; |
266 | 266 | ||
267 | /* The connection could be invalid, in which case set_label is no-op.*/ | 267 | /* The connection could be invalid, in which case set_label is no-op.*/ |
268 | ct = nf_ct_get(skb, &ctinfo); | 268 | ct = nf_ct_get(skb, &ctinfo); |
269 | if (!ct) | 269 | if (!ct) |
270 | return 0; | 270 | return 0; |
271 | 271 | ||
272 | cl = nf_ct_labels_find(ct); | 272 | cl = nf_ct_labels_find(ct); |
273 | if (!cl) { | 273 | if (!cl) { |
274 | nf_ct_labels_ext_add(ct); | 274 | nf_ct_labels_ext_add(ct); |
275 | cl = nf_ct_labels_find(ct); | 275 | cl = nf_ct_labels_find(ct); |
276 | } | 276 | } |
277 | if (!cl || sizeof(cl->bits) < OVS_CT_LABELS_LEN) | 277 | if (!cl || sizeof(cl->bits) < OVS_CT_LABELS_LEN) |
278 | return -ENOSPC; | 278 | return -ENOSPC; |
279 | 279 | ||
280 | err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask, | 280 | err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask, |
281 | OVS_CT_LABELS_LEN / sizeof(u32)); | 281 | OVS_CT_LABELS_LEN / sizeof(u32)); |
282 | if (err) | 282 | if (err) |
283 | return err; | 283 | return err; |
284 | 284 | ||
285 | ovs_ct_get_labels(ct, &key->ct.labels); | 285 | ovs_ct_get_labels(ct, &key->ct.labels); |
286 | return 0; | 286 | return 0; |
287 | } | 287 | } |
288 | 288 | ||
289 | /* 'skb' should already be pulled to nh_ofs. */ | 289 | /* 'skb' should already be pulled to nh_ofs. */ |
290 | static int ovs_ct_helper(struct sk_buff *skb, u16 proto) | 290 | static int ovs_ct_helper(struct sk_buff *skb, u16 proto) |
291 | { | 291 | { |
292 | const struct nf_conntrack_helper *helper; | 292 | const struct nf_conntrack_helper *helper; |
293 | const struct nf_conn_help *help; | 293 | const struct nf_conn_help *help; |
294 | enum ip_conntrack_info ctinfo; | 294 | enum ip_conntrack_info ctinfo; |
295 | unsigned int protoff; | 295 | unsigned int protoff; |
296 | struct nf_conn *ct; | 296 | struct nf_conn *ct; |
297 | int err; | 297 | int err; |
298 | 298 | ||
299 | ct = nf_ct_get(skb, &ctinfo); | 299 | ct = nf_ct_get(skb, &ctinfo); |
300 | if (!ct || ctinfo == IP_CT_RELATED_REPLY) | 300 | if (!ct || ctinfo == IP_CT_RELATED_REPLY) |
301 | return NF_ACCEPT; | 301 | return NF_ACCEPT; |
302 | 302 | ||
303 | help = nfct_help(ct); | 303 | help = nfct_help(ct); |
304 | if (!help) | 304 | if (!help) |
305 | return NF_ACCEPT; | 305 | return NF_ACCEPT; |
306 | 306 | ||
307 | helper = rcu_dereference(help->helper); | 307 | helper = rcu_dereference(help->helper); |
308 | if (!helper) | 308 | if (!helper) |
309 | return NF_ACCEPT; | 309 | return NF_ACCEPT; |
310 | 310 | ||
311 | switch (proto) { | 311 | switch (proto) { |
312 | case NFPROTO_IPV4: | 312 | case NFPROTO_IPV4: |
313 | protoff = ip_hdrlen(skb); | 313 | protoff = ip_hdrlen(skb); |
314 | break; | 314 | break; |
315 | case NFPROTO_IPV6: { | 315 | case NFPROTO_IPV6: { |
316 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; | 316 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; |
317 | __be16 frag_off; | 317 | __be16 frag_off; |
318 | int ofs; | 318 | int ofs; |
319 | 319 | ||
320 | ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr, | 320 | ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr, |
321 | &frag_off); | 321 | &frag_off); |
322 | if (ofs < 0 || (frag_off & htons(~0x7)) != 0) { | 322 | if (ofs < 0 || (frag_off & htons(~0x7)) != 0) { |
323 | pr_debug("proto header not found\n"); | 323 | pr_debug("proto header not found\n"); |
324 | return NF_ACCEPT; | 324 | return NF_ACCEPT; |
325 | } | 325 | } |
326 | protoff = ofs; | 326 | protoff = ofs; |
327 | break; | 327 | break; |
328 | } | 328 | } |
329 | default: | 329 | default: |
330 | WARN_ONCE(1, "helper invoked on non-IP family!"); | 330 | WARN_ONCE(1, "helper invoked on non-IP family!"); |
331 | return NF_DROP; | 331 | return NF_DROP; |
332 | } | 332 | } |
333 | 333 | ||
334 | err = helper->help(skb, protoff, ct, ctinfo); | 334 | err = helper->help(skb, protoff, ct, ctinfo); |
335 | if (err != NF_ACCEPT) | 335 | if (err != NF_ACCEPT) |
336 | return err; | 336 | return err; |
337 | 337 | ||
338 | /* Adjust seqs after helper. This is needed due to some helpers (e.g., | 338 | /* Adjust seqs after helper. This is needed due to some helpers (e.g., |
339 | * FTP with NAT) adusting the TCP payload size when mangling IP | 339 | * FTP with NAT) adusting the TCP payload size when mangling IP |
340 | * addresses and/or port numbers in the text-based control connection. | 340 | * addresses and/or port numbers in the text-based control connection. |
341 | */ | 341 | */ |
342 | if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) && | 342 | if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) && |
343 | !nf_ct_seq_adjust(skb, ct, ctinfo, protoff)) | 343 | !nf_ct_seq_adjust(skb, ct, ctinfo, protoff)) |
344 | return NF_DROP; | 344 | return NF_DROP; |
345 | return NF_ACCEPT; | 345 | return NF_ACCEPT; |
346 | } | 346 | } |
347 | 347 | ||
348 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero | 348 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero |
349 | * value if 'skb' is freed. | 349 | * value if 'skb' is freed. |
350 | */ | 350 | */ |
351 | static int handle_fragments(struct net *net, struct sw_flow_key *key, | 351 | static int handle_fragments(struct net *net, struct sw_flow_key *key, |
352 | u16 zone, struct sk_buff *skb) | 352 | u16 zone, struct sk_buff *skb) |
353 | { | 353 | { |
354 | struct ovs_skb_cb ovs_cb = *OVS_CB(skb); | 354 | struct ovs_skb_cb ovs_cb = *OVS_CB(skb); |
355 | int err; | 355 | int err; |
356 | 356 | ||
357 | if (key->eth.type == htons(ETH_P_IP)) { | 357 | if (key->eth.type == htons(ETH_P_IP)) { |
358 | enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone; | 358 | enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone; |
359 | 359 | ||
360 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); | 360 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); |
361 | err = ip_defrag(net, skb, user); | 361 | err = ip_defrag(net, skb, user); |
362 | if (err) | 362 | if (err) |
363 | return err; | 363 | return err; |
364 | 364 | ||
365 | ovs_cb.mru = IPCB(skb)->frag_max_size; | 365 | ovs_cb.mru = IPCB(skb)->frag_max_size; |
366 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) | 366 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) |
367 | } else if (key->eth.type == htons(ETH_P_IPV6)) { | 367 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
368 | enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone; | 368 | enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone; |
369 | 369 | ||
370 | skb_orphan(skb); | 370 | skb_orphan(skb); |
371 | memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); | 371 | memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); |
372 | err = nf_ct_frag6_gather(net, skb, user); | 372 | err = nf_ct_frag6_gather(net, skb, user); |
373 | if (err) | 373 | if (err) |
374 | return err; | 374 | return err; |
375 | 375 | ||
376 | key->ip.proto = ipv6_hdr(skb)->nexthdr; | 376 | key->ip.proto = ipv6_hdr(skb)->nexthdr; |
377 | ovs_cb.mru = IP6CB(skb)->frag_max_size; | 377 | ovs_cb.mru = IP6CB(skb)->frag_max_size; |
378 | #endif | 378 | #endif |
379 | } else { | 379 | } else { |
380 | kfree_skb(skb); | 380 | kfree_skb(skb); |
381 | return -EPFNOSUPPORT; | 381 | return -EPFNOSUPPORT; |
382 | } | 382 | } |
383 | 383 | ||
384 | key->ip.frag = OVS_FRAG_TYPE_NONE; | 384 | key->ip.frag = OVS_FRAG_TYPE_NONE; |
385 | skb_clear_hash(skb); | 385 | skb_clear_hash(skb); |
386 | skb->ignore_df = 1; | 386 | skb->ignore_df = 1; |
387 | *OVS_CB(skb) = ovs_cb; | 387 | *OVS_CB(skb) = ovs_cb; |
388 | 388 | ||
389 | return 0; | 389 | return 0; |
390 | } | 390 | } |
391 | 391 | ||
392 | static struct nf_conntrack_expect * | 392 | static struct nf_conntrack_expect * |
393 | ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone, | 393 | ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone, |
394 | u16 proto, const struct sk_buff *skb) | 394 | u16 proto, const struct sk_buff *skb) |
395 | { | 395 | { |
396 | struct nf_conntrack_tuple tuple; | 396 | struct nf_conntrack_tuple tuple; |
397 | 397 | ||
398 | if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple)) | 398 | if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple)) |
399 | return NULL; | 399 | return NULL; |
400 | return __nf_ct_expect_find(net, zone, &tuple); | 400 | return __nf_ct_expect_find(net, zone, &tuple); |
401 | } | 401 | } |
402 | 402 | ||
403 | /* This replicates logic from nf_conntrack_core.c that is not exported. */ | 403 | /* This replicates logic from nf_conntrack_core.c that is not exported. */ |
404 | static enum ip_conntrack_info | 404 | static enum ip_conntrack_info |
405 | ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h) | 405 | ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h) |
406 | { | 406 | { |
407 | const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | 407 | const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); |
408 | 408 | ||
409 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) | 409 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) |
410 | return IP_CT_ESTABLISHED_REPLY; | 410 | return IP_CT_ESTABLISHED_REPLY; |
411 | /* Once we've had two way comms, always ESTABLISHED. */ | 411 | /* Once we've had two way comms, always ESTABLISHED. */ |
412 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | 412 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) |
413 | return IP_CT_ESTABLISHED; | 413 | return IP_CT_ESTABLISHED; |
414 | if (test_bit(IPS_EXPECTED_BIT, &ct->status)) | 414 | if (test_bit(IPS_EXPECTED_BIT, &ct->status)) |
415 | return IP_CT_RELATED; | 415 | return IP_CT_RELATED; |
416 | return IP_CT_NEW; | 416 | return IP_CT_NEW; |
417 | } | 417 | } |
418 | 418 | ||
419 | /* Find an existing connection which this packet belongs to without | 419 | /* Find an existing connection which this packet belongs to without |
420 | * re-attributing statistics or modifying the connection state. This allows an | 420 | * re-attributing statistics or modifying the connection state. This allows an |
421 | * skb->nfct lost due to an upcall to be recovered during actions execution. | 421 | * skb->nfct lost due to an upcall to be recovered during actions execution. |
422 | * | 422 | * |
423 | * Must be called with rcu_read_lock. | 423 | * Must be called with rcu_read_lock. |
424 | * | 424 | * |
425 | * On success, populates skb->nfct and skb->nfctinfo, and returns the | 425 | * On success, populates skb->nfct and skb->nfctinfo, and returns the |
426 | * connection. Returns NULL if there is no existing entry. | 426 | * connection. Returns NULL if there is no existing entry. |
427 | */ | 427 | */ |
428 | static struct nf_conn * | 428 | static struct nf_conn * |
429 | ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone, | 429 | ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone, |
430 | u8 l3num, struct sk_buff *skb) | 430 | u8 l3num, struct sk_buff *skb) |
431 | { | 431 | { |
432 | struct nf_conntrack_l3proto *l3proto; | 432 | struct nf_conntrack_l3proto *l3proto; |
433 | struct nf_conntrack_l4proto *l4proto; | 433 | struct nf_conntrack_l4proto *l4proto; |
434 | struct nf_conntrack_tuple tuple; | 434 | struct nf_conntrack_tuple tuple; |
435 | struct nf_conntrack_tuple_hash *h; | 435 | struct nf_conntrack_tuple_hash *h; |
436 | struct nf_conn *ct; | 436 | struct nf_conn *ct; |
437 | unsigned int dataoff; | 437 | unsigned int dataoff; |
438 | u8 protonum; | 438 | u8 protonum; |
439 | 439 | ||
440 | l3proto = __nf_ct_l3proto_find(l3num); | 440 | l3proto = __nf_ct_l3proto_find(l3num); |
441 | if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff, | 441 | if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff, |
442 | &protonum) <= 0) { | 442 | &protonum) <= 0) { |
443 | pr_debug("ovs_ct_find_existing: Can't get protonum\n"); | 443 | pr_debug("ovs_ct_find_existing: Can't get protonum\n"); |
444 | return NULL; | 444 | return NULL; |
445 | } | 445 | } |
446 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | 446 | l4proto = __nf_ct_l4proto_find(l3num, protonum); |
447 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num, | 447 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num, |
448 | protonum, net, &tuple, l3proto, l4proto)) { | 448 | protonum, net, &tuple, l3proto, l4proto)) { |
449 | pr_debug("ovs_ct_find_existing: Can't get tuple\n"); | 449 | pr_debug("ovs_ct_find_existing: Can't get tuple\n"); |
450 | return NULL; | 450 | return NULL; |
451 | } | 451 | } |
452 | 452 | ||
453 | /* look for tuple match */ | 453 | /* look for tuple match */ |
454 | h = nf_conntrack_find_get(net, zone, &tuple); | 454 | h = nf_conntrack_find_get(net, zone, &tuple); |
455 | if (!h) | 455 | if (!h) |
456 | return NULL; /* Not found. */ | 456 | return NULL; /* Not found. */ |
457 | 457 | ||
458 | ct = nf_ct_tuplehash_to_ctrack(h); | 458 | ct = nf_ct_tuplehash_to_ctrack(h); |
459 | 459 | ||
460 | skb->nfct = &ct->ct_general; | 460 | skb->nfct = &ct->ct_general; |
461 | skb->nfctinfo = ovs_ct_get_info(h); | 461 | skb->nfctinfo = ovs_ct_get_info(h); |
462 | return ct; | 462 | return ct; |
463 | } | 463 | } |
464 | 464 | ||
465 | /* Determine whether skb->nfct is equal to the result of conntrack lookup. */ | 465 | /* Determine whether skb->nfct is equal to the result of conntrack lookup. */ |
466 | static bool skb_nfct_cached(struct net *net, | 466 | static bool skb_nfct_cached(struct net *net, |
467 | const struct sw_flow_key *key, | 467 | const struct sw_flow_key *key, |
468 | const struct ovs_conntrack_info *info, | 468 | const struct ovs_conntrack_info *info, |
469 | struct sk_buff *skb) | 469 | struct sk_buff *skb) |
470 | { | 470 | { |
471 | enum ip_conntrack_info ctinfo; | 471 | enum ip_conntrack_info ctinfo; |
472 | struct nf_conn *ct; | 472 | struct nf_conn *ct; |
473 | 473 | ||
474 | ct = nf_ct_get(skb, &ctinfo); | 474 | ct = nf_ct_get(skb, &ctinfo); |
475 | /* If no ct, check if we have evidence that an existing conntrack entry | 475 | /* If no ct, check if we have evidence that an existing conntrack entry |
476 | * might be found for this skb. This happens when we lose a skb->nfct | 476 | * might be found for this skb. This happens when we lose a skb->nfct |
477 | * due to an upcall. If the connection was not confirmed, it is not | 477 | * due to an upcall. If the connection was not confirmed, it is not |
478 | * cached and needs to be run through conntrack again. | 478 | * cached and needs to be run through conntrack again. |
479 | */ | 479 | */ |
480 | if (!ct && key->ct.state & OVS_CS_F_TRACKED && | 480 | if (!ct && key->ct.state & OVS_CS_F_TRACKED && |
481 | !(key->ct.state & OVS_CS_F_INVALID) && | 481 | !(key->ct.state & OVS_CS_F_INVALID) && |
482 | key->ct.zone == info->zone.id) | 482 | key->ct.zone == info->zone.id) |
483 | ct = ovs_ct_find_existing(net, &info->zone, info->family, skb); | 483 | ct = ovs_ct_find_existing(net, &info->zone, info->family, skb); |
484 | if (!ct) | 484 | if (!ct) |
485 | return false; | 485 | return false; |
486 | if (!net_eq(net, read_pnet(&ct->ct_net))) | 486 | if (!net_eq(net, read_pnet(&ct->ct_net))) |
487 | return false; | 487 | return false; |
488 | if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct))) | 488 | if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct))) |
489 | return false; | 489 | return false; |
490 | if (info->helper) { | 490 | if (info->helper) { |
491 | struct nf_conn_help *help; | 491 | struct nf_conn_help *help; |
492 | 492 | ||
493 | help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER); | 493 | help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER); |
494 | if (help && rcu_access_pointer(help->helper) != info->helper) | 494 | if (help && rcu_access_pointer(help->helper) != info->helper) |
495 | return false; | 495 | return false; |
496 | } | 496 | } |
497 | 497 | ||
498 | return true; | 498 | return true; |
499 | } | 499 | } |
500 | 500 | ||
501 | #ifdef CONFIG_NF_NAT_NEEDED | 501 | #ifdef CONFIG_NF_NAT_NEEDED |
502 | /* Modelled after nf_nat_ipv[46]_fn(). | 502 | /* Modelled after nf_nat_ipv[46]_fn(). |
503 | * range is only used for new, uninitialized NAT state. | 503 | * range is only used for new, uninitialized NAT state. |
504 | * Returns either NF_ACCEPT or NF_DROP. | 504 | * Returns either NF_ACCEPT or NF_DROP. |
505 | */ | 505 | */ |
506 | static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct, | 506 | static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct, |
507 | enum ip_conntrack_info ctinfo, | 507 | enum ip_conntrack_info ctinfo, |
508 | const struct nf_nat_range *range, | 508 | const struct nf_nat_range *range, |
509 | enum nf_nat_manip_type maniptype) | 509 | enum nf_nat_manip_type maniptype) |
510 | { | 510 | { |
511 | int hooknum, nh_off, err = NF_ACCEPT; | 511 | int hooknum, nh_off, err = NF_ACCEPT; |
512 | 512 | ||
513 | nh_off = skb_network_offset(skb); | 513 | nh_off = skb_network_offset(skb); |
514 | skb_pull(skb, nh_off); | 514 | skb_pull(skb, nh_off); |
515 | 515 | ||
516 | /* See HOOK2MANIP(). */ | 516 | /* See HOOK2MANIP(). */ |
517 | if (maniptype == NF_NAT_MANIP_SRC) | 517 | if (maniptype == NF_NAT_MANIP_SRC) |
518 | hooknum = NF_INET_LOCAL_IN; /* Source NAT */ | 518 | hooknum = NF_INET_LOCAL_IN; /* Source NAT */ |
519 | else | 519 | else |
520 | hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */ | 520 | hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */ |
521 | 521 | ||
522 | switch (ctinfo) { | 522 | switch (ctinfo) { |
523 | case IP_CT_RELATED: | 523 | case IP_CT_RELATED: |
524 | case IP_CT_RELATED_REPLY: | 524 | case IP_CT_RELATED_REPLY: |
525 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && | 525 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
526 | skb->protocol == htons(ETH_P_IP) && | 526 | skb->protocol == htons(ETH_P_IP) && |
527 | ip_hdr(skb)->protocol == IPPROTO_ICMP) { | 527 | ip_hdr(skb)->protocol == IPPROTO_ICMP) { |
528 | if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo, | 528 | if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo, |
529 | hooknum)) | 529 | hooknum)) |
530 | err = NF_DROP; | 530 | err = NF_DROP; |
531 | goto push; | 531 | goto push; |
532 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && | 532 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
533 | skb->protocol == htons(ETH_P_IPV6)) { | 533 | skb->protocol == htons(ETH_P_IPV6)) { |
534 | __be16 frag_off; | 534 | __be16 frag_off; |
535 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; | 535 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; |
536 | int hdrlen = ipv6_skip_exthdr(skb, | 536 | int hdrlen = ipv6_skip_exthdr(skb, |
537 | sizeof(struct ipv6hdr), | 537 | sizeof(struct ipv6hdr), |
538 | &nexthdr, &frag_off); | 538 | &nexthdr, &frag_off); |
539 | 539 | ||
540 | if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) { | 540 | if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) { |
541 | if (!nf_nat_icmpv6_reply_translation(skb, ct, | 541 | if (!nf_nat_icmpv6_reply_translation(skb, ct, |
542 | ctinfo, | 542 | ctinfo, |
543 | hooknum, | 543 | hooknum, |
544 | hdrlen)) | 544 | hdrlen)) |
545 | err = NF_DROP; | 545 | err = NF_DROP; |
546 | goto push; | 546 | goto push; |
547 | } | 547 | } |
548 | } | 548 | } |
549 | /* Non-ICMP, fall thru to initialize if needed. */ | 549 | /* Non-ICMP, fall thru to initialize if needed. */ |
550 | case IP_CT_NEW: | 550 | case IP_CT_NEW: |
551 | /* Seen it before? This can happen for loopback, retrans, | 551 | /* Seen it before? This can happen for loopback, retrans, |
552 | * or local packets. | 552 | * or local packets. |
553 | */ | 553 | */ |
554 | if (!nf_nat_initialized(ct, maniptype)) { | 554 | if (!nf_nat_initialized(ct, maniptype)) { |
555 | /* Initialize according to the NAT action. */ | 555 | /* Initialize according to the NAT action. */ |
556 | err = (range && range->flags & NF_NAT_RANGE_MAP_IPS) | 556 | err = (range && range->flags & NF_NAT_RANGE_MAP_IPS) |
557 | /* Action is set up to establish a new | 557 | /* Action is set up to establish a new |
558 | * mapping. | 558 | * mapping. |
559 | */ | 559 | */ |
560 | ? nf_nat_setup_info(ct, range, maniptype) | 560 | ? nf_nat_setup_info(ct, range, maniptype) |
561 | : nf_nat_alloc_null_binding(ct, hooknum); | 561 | : nf_nat_alloc_null_binding(ct, hooknum); |
562 | if (err != NF_ACCEPT) | 562 | if (err != NF_ACCEPT) |
563 | goto push; | 563 | goto push; |
564 | } | 564 | } |
565 | break; | 565 | break; |
566 | 566 | ||
567 | case IP_CT_ESTABLISHED: | 567 | case IP_CT_ESTABLISHED: |
568 | case IP_CT_ESTABLISHED_REPLY: | 568 | case IP_CT_ESTABLISHED_REPLY: |
569 | break; | 569 | break; |
570 | 570 | ||
571 | default: | 571 | default: |
572 | err = NF_DROP; | 572 | err = NF_DROP; |
573 | goto push; | 573 | goto push; |
574 | } | 574 | } |
575 | 575 | ||
576 | err = nf_nat_packet(ct, ctinfo, hooknum, skb); | 576 | err = nf_nat_packet(ct, ctinfo, hooknum, skb); |
577 | push: | 577 | push: |
578 | skb_push(skb, nh_off); | 578 | skb_push(skb, nh_off); |
579 | 579 | ||
580 | return err; | 580 | return err; |
581 | } | 581 | } |
582 | 582 | ||
583 | static void ovs_nat_update_key(struct sw_flow_key *key, | 583 | static void ovs_nat_update_key(struct sw_flow_key *key, |
584 | const struct sk_buff *skb, | 584 | const struct sk_buff *skb, |
585 | enum nf_nat_manip_type maniptype) | 585 | enum nf_nat_manip_type maniptype) |
586 | { | 586 | { |
587 | if (maniptype == NF_NAT_MANIP_SRC) { | 587 | if (maniptype == NF_NAT_MANIP_SRC) { |
588 | __be16 src; | 588 | __be16 src; |
589 | 589 | ||
590 | key->ct.state |= OVS_CS_F_SRC_NAT; | 590 | key->ct.state |= OVS_CS_F_SRC_NAT; |
591 | if (key->eth.type == htons(ETH_P_IP)) | 591 | if (key->eth.type == htons(ETH_P_IP)) |
592 | key->ipv4.addr.src = ip_hdr(skb)->saddr; | 592 | key->ipv4.addr.src = ip_hdr(skb)->saddr; |
593 | else if (key->eth.type == htons(ETH_P_IPV6)) | 593 | else if (key->eth.type == htons(ETH_P_IPV6)) |
594 | memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr, | 594 | memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr, |
595 | sizeof(key->ipv6.addr.src)); | 595 | sizeof(key->ipv6.addr.src)); |
596 | else | 596 | else |
597 | return; | 597 | return; |
598 | 598 | ||
599 | if (key->ip.proto == IPPROTO_UDP) | 599 | if (key->ip.proto == IPPROTO_UDP) |
600 | src = udp_hdr(skb)->source; | 600 | src = udp_hdr(skb)->source; |
601 | else if (key->ip.proto == IPPROTO_TCP) | 601 | else if (key->ip.proto == IPPROTO_TCP) |
602 | src = tcp_hdr(skb)->source; | 602 | src = tcp_hdr(skb)->source; |
603 | else if (key->ip.proto == IPPROTO_SCTP) | 603 | else if (key->ip.proto == IPPROTO_SCTP) |
604 | src = sctp_hdr(skb)->source; | 604 | src = sctp_hdr(skb)->source; |
605 | else | 605 | else |
606 | return; | 606 | return; |
607 | 607 | ||
608 | key->tp.src = src; | 608 | key->tp.src = src; |
609 | } else { | 609 | } else { |
610 | __be16 dst; | 610 | __be16 dst; |
611 | 611 | ||
612 | key->ct.state |= OVS_CS_F_DST_NAT; | 612 | key->ct.state |= OVS_CS_F_DST_NAT; |
613 | if (key->eth.type == htons(ETH_P_IP)) | 613 | if (key->eth.type == htons(ETH_P_IP)) |
614 | key->ipv4.addr.dst = ip_hdr(skb)->daddr; | 614 | key->ipv4.addr.dst = ip_hdr(skb)->daddr; |
615 | else if (key->eth.type == htons(ETH_P_IPV6)) | 615 | else if (key->eth.type == htons(ETH_P_IPV6)) |
616 | memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr, | 616 | memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr, |
617 | sizeof(key->ipv6.addr.dst)); | 617 | sizeof(key->ipv6.addr.dst)); |
618 | else | 618 | else |
619 | return; | 619 | return; |
620 | 620 | ||
621 | if (key->ip.proto == IPPROTO_UDP) | 621 | if (key->ip.proto == IPPROTO_UDP) |
622 | dst = udp_hdr(skb)->dest; | 622 | dst = udp_hdr(skb)->dest; |
623 | else if (key->ip.proto == IPPROTO_TCP) | 623 | else if (key->ip.proto == IPPROTO_TCP) |
624 | dst = tcp_hdr(skb)->dest; | 624 | dst = tcp_hdr(skb)->dest; |
625 | else if (key->ip.proto == IPPROTO_SCTP) | 625 | else if (key->ip.proto == IPPROTO_SCTP) |
626 | dst = sctp_hdr(skb)->dest; | 626 | dst = sctp_hdr(skb)->dest; |
627 | else | 627 | else |
628 | return; | 628 | return; |
629 | 629 | ||
630 | key->tp.dst = dst; | 630 | key->tp.dst = dst; |
631 | } | 631 | } |
632 | } | 632 | } |
633 | 633 | ||
634 | /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */ | 634 | /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */ |
635 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | 635 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, |
636 | const struct ovs_conntrack_info *info, | 636 | const struct ovs_conntrack_info *info, |
637 | struct sk_buff *skb, struct nf_conn *ct, | 637 | struct sk_buff *skb, struct nf_conn *ct, |
638 | enum ip_conntrack_info ctinfo) | 638 | enum ip_conntrack_info ctinfo) |
639 | { | 639 | { |
640 | enum nf_nat_manip_type maniptype; | 640 | enum nf_nat_manip_type maniptype; |
641 | int err; | 641 | int err; |
642 | 642 | ||
643 | if (nf_ct_is_untracked(ct)) { | 643 | if (nf_ct_is_untracked(ct)) { |
644 | /* A NAT action may only be performed on tracked packets. */ | 644 | /* A NAT action may only be performed on tracked packets. */ |
645 | return NF_ACCEPT; | 645 | return NF_ACCEPT; |
646 | } | 646 | } |
647 | 647 | ||
648 | /* Add NAT extension if not confirmed yet. */ | 648 | /* Add NAT extension if not confirmed yet. */ |
649 | if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct)) | 649 | if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct)) |
650 | return NF_ACCEPT; /* Can't NAT. */ | 650 | return NF_ACCEPT; /* Can't NAT. */ |
651 | 651 | ||
652 | /* Determine NAT type. | 652 | /* Determine NAT type. |
653 | * Check if the NAT type can be deduced from the tracked connection. | 653 | * Check if the NAT type can be deduced from the tracked connection. |
654 | * Make sure new expected connections (IP_CT_RELATED) are NATted only | 654 | * Make sure new expected connections (IP_CT_RELATED) are NATted only |
655 | * when committing. | 655 | * when committing. |
656 | */ | 656 | */ |
657 | if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW && | 657 | if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW && |
658 | ct->status & IPS_NAT_MASK && | 658 | ct->status & IPS_NAT_MASK && |
659 | (ctinfo != IP_CT_RELATED || info->commit)) { | 659 | (ctinfo != IP_CT_RELATED || info->commit)) { |
660 | /* NAT an established or related connection like before. */ | 660 | /* NAT an established or related connection like before. */ |
661 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) | 661 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) |
662 | /* This is the REPLY direction for a connection | 662 | /* This is the REPLY direction for a connection |
663 | * for which NAT was applied in the forward | 663 | * for which NAT was applied in the forward |
664 | * direction. Do the reverse NAT. | 664 | * direction. Do the reverse NAT. |
665 | */ | 665 | */ |
666 | maniptype = ct->status & IPS_SRC_NAT | 666 | maniptype = ct->status & IPS_SRC_NAT |
667 | ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC; | 667 | ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC; |
668 | else | 668 | else |
669 | maniptype = ct->status & IPS_SRC_NAT | 669 | maniptype = ct->status & IPS_SRC_NAT |
670 | ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST; | 670 | ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST; |
671 | } else if (info->nat & OVS_CT_SRC_NAT) { | 671 | } else if (info->nat & OVS_CT_SRC_NAT) { |
672 | maniptype = NF_NAT_MANIP_SRC; | 672 | maniptype = NF_NAT_MANIP_SRC; |
673 | } else if (info->nat & OVS_CT_DST_NAT) { | 673 | } else if (info->nat & OVS_CT_DST_NAT) { |
674 | maniptype = NF_NAT_MANIP_DST; | 674 | maniptype = NF_NAT_MANIP_DST; |
675 | } else { | 675 | } else { |
676 | return NF_ACCEPT; /* Connection is not NATed. */ | 676 | return NF_ACCEPT; /* Connection is not NATed. */ |
677 | } | 677 | } |
678 | err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype); | 678 | err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype); |
679 | 679 | ||
680 | /* Mark NAT done if successful and update the flow key. */ | 680 | /* Mark NAT done if successful and update the flow key. */ |
681 | if (err == NF_ACCEPT) | 681 | if (err == NF_ACCEPT) |
682 | ovs_nat_update_key(key, skb, maniptype); | 682 | ovs_nat_update_key(key, skb, maniptype); |
683 | 683 | ||
684 | return err; | 684 | return err; |
685 | } | 685 | } |
686 | #else /* !CONFIG_NF_NAT_NEEDED */ | 686 | #else /* !CONFIG_NF_NAT_NEEDED */ |
687 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | 687 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, |
688 | const struct ovs_conntrack_info *info, | 688 | const struct ovs_conntrack_info *info, |
689 | struct sk_buff *skb, struct nf_conn *ct, | 689 | struct sk_buff *skb, struct nf_conn *ct, |
690 | enum ip_conntrack_info ctinfo) | 690 | enum ip_conntrack_info ctinfo) |
691 | { | 691 | { |
692 | return NF_ACCEPT; | 692 | return NF_ACCEPT; |
693 | } | 693 | } |
694 | #endif | 694 | #endif |
695 | 695 | ||
696 | /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if | 696 | /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if |
697 | * not done already. Update key with new CT state after passing the packet | 697 | * not done already. Update key with new CT state after passing the packet |
698 | * through conntrack. | 698 | * through conntrack. |
699 | * Note that if the packet is deemed invalid by conntrack, skb->nfct will be | 699 | * Note that if the packet is deemed invalid by conntrack, skb->nfct will be |
700 | * set to NULL and 0 will be returned. | 700 | * set to NULL and 0 will be returned. |
701 | */ | 701 | */ |
702 | static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key, | 702 | static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key, |
703 | const struct ovs_conntrack_info *info, | 703 | const struct ovs_conntrack_info *info, |
704 | struct sk_buff *skb) | 704 | struct sk_buff *skb) |
705 | { | 705 | { |
706 | /* If we are recirculating packets to match on conntrack fields and | 706 | /* If we are recirculating packets to match on conntrack fields and |
707 | * committing with a separate conntrack action, then we don't need to | 707 | * committing with a separate conntrack action, then we don't need to |
708 | * actually run the packet through conntrack twice unless it's for a | 708 | * actually run the packet through conntrack twice unless it's for a |
709 | * different zone. | 709 | * different zone. |
710 | */ | 710 | */ |
711 | bool cached = skb_nfct_cached(net, key, info, skb); | 711 | bool cached = skb_nfct_cached(net, key, info, skb); |
712 | enum ip_conntrack_info ctinfo; | 712 | enum ip_conntrack_info ctinfo; |
713 | struct nf_conn *ct; | 713 | struct nf_conn *ct; |
714 | 714 | ||
715 | if (!cached) { | 715 | if (!cached) { |
716 | struct nf_conn *tmpl = info->ct; | 716 | struct nf_conn *tmpl = info->ct; |
717 | int err; | 717 | int err; |
718 | 718 | ||
719 | /* Associate skb with specified zone. */ | 719 | /* Associate skb with specified zone. */ |
720 | if (tmpl) { | 720 | if (tmpl) { |
721 | if (skb->nfct) | 721 | if (skb->nfct) |
722 | nf_conntrack_put(skb->nfct); | 722 | nf_conntrack_put(skb->nfct); |
723 | nf_conntrack_get(&tmpl->ct_general); | 723 | nf_conntrack_get(&tmpl->ct_general); |
724 | skb->nfct = &tmpl->ct_general; | 724 | skb->nfct = &tmpl->ct_general; |
725 | skb->nfctinfo = IP_CT_NEW; | 725 | skb->nfctinfo = IP_CT_NEW; |
726 | } | 726 | } |
727 | 727 | ||
728 | /* Repeat if requested, see nf_iterate(). */ | 728 | err = nf_conntrack_in(net, info->family, |
729 | do { | 729 | NF_INET_PRE_ROUTING, skb); |
730 | err = nf_conntrack_in(net, info->family, | ||
731 | NF_INET_PRE_ROUTING, skb); | ||
732 | } while (err == NF_REPEAT); | ||
733 | |||
734 | if (err != NF_ACCEPT) | 730 | if (err != NF_ACCEPT) |
735 | return -ENOENT; | 731 | return -ENOENT; |
736 | 732 | ||
737 | /* Clear CT state NAT flags to mark that we have not yet done | 733 | /* Clear CT state NAT flags to mark that we have not yet done |
738 | * NAT after the nf_conntrack_in() call. We can actually clear | 734 | * NAT after the nf_conntrack_in() call. We can actually clear |
739 | * the whole state, as it will be re-initialized below. | 735 | * the whole state, as it will be re-initialized below. |
740 | */ | 736 | */ |
741 | key->ct.state = 0; | 737 | key->ct.state = 0; |
742 | 738 | ||
743 | /* Update the key, but keep the NAT flags. */ | 739 | /* Update the key, but keep the NAT flags. */ |
744 | ovs_ct_update_key(skb, info, key, true, true); | 740 | ovs_ct_update_key(skb, info, key, true, true); |
745 | } | 741 | } |
746 | 742 | ||
747 | ct = nf_ct_get(skb, &ctinfo); | 743 | ct = nf_ct_get(skb, &ctinfo); |
748 | if (ct) { | 744 | if (ct) { |
749 | /* Packets starting a new connection must be NATted before the | 745 | /* Packets starting a new connection must be NATted before the |
750 | * helper, so that the helper knows about the NAT. We enforce | 746 | * helper, so that the helper knows about the NAT. We enforce |
751 | * this by delaying both NAT and helper calls for unconfirmed | 747 | * this by delaying both NAT and helper calls for unconfirmed |
752 | * connections until the committing CT action. For later | 748 | * connections until the committing CT action. For later |
753 | * packets NAT and Helper may be called in either order. | 749 | * packets NAT and Helper may be called in either order. |
754 | * | 750 | * |
755 | * NAT will be done only if the CT action has NAT, and only | 751 | * NAT will be done only if the CT action has NAT, and only |
756 | * once per packet (per zone), as guarded by the NAT bits in | 752 | * once per packet (per zone), as guarded by the NAT bits in |
757 | * the key->ct.state. | 753 | * the key->ct.state. |
758 | */ | 754 | */ |
759 | if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) && | 755 | if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) && |
760 | (nf_ct_is_confirmed(ct) || info->commit) && | 756 | (nf_ct_is_confirmed(ct) || info->commit) && |
761 | ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) { | 757 | ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) { |
762 | return -EINVAL; | 758 | return -EINVAL; |
763 | } | 759 | } |
764 | 760 | ||
765 | /* Userspace may decide to perform a ct lookup without a helper | 761 | /* Userspace may decide to perform a ct lookup without a helper |
766 | * specified followed by a (recirculate and) commit with one. | 762 | * specified followed by a (recirculate and) commit with one. |
767 | * Therefore, for unconfirmed connections which we will commit, | 763 | * Therefore, for unconfirmed connections which we will commit, |
768 | * we need to attach the helper here. | 764 | * we need to attach the helper here. |
769 | */ | 765 | */ |
770 | if (!nf_ct_is_confirmed(ct) && info->commit && | 766 | if (!nf_ct_is_confirmed(ct) && info->commit && |
771 | info->helper && !nfct_help(ct)) { | 767 | info->helper && !nfct_help(ct)) { |
772 | int err = __nf_ct_try_assign_helper(ct, info->ct, | 768 | int err = __nf_ct_try_assign_helper(ct, info->ct, |
773 | GFP_ATOMIC); | 769 | GFP_ATOMIC); |
774 | if (err) | 770 | if (err) |
775 | return err; | 771 | return err; |
776 | } | 772 | } |
777 | 773 | ||
778 | /* Call the helper only if: | 774 | /* Call the helper only if: |
779 | * - nf_conntrack_in() was executed above ("!cached") for a | 775 | * - nf_conntrack_in() was executed above ("!cached") for a |
780 | * confirmed connection, or | 776 | * confirmed connection, or |
781 | * - When committing an unconfirmed connection. | 777 | * - When committing an unconfirmed connection. |
782 | */ | 778 | */ |
783 | if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) && | 779 | if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) && |
784 | ovs_ct_helper(skb, info->family) != NF_ACCEPT) { | 780 | ovs_ct_helper(skb, info->family) != NF_ACCEPT) { |
785 | return -EINVAL; | 781 | return -EINVAL; |
786 | } | 782 | } |
787 | } | 783 | } |
788 | 784 | ||
789 | return 0; | 785 | return 0; |
790 | } | 786 | } |
791 | 787 | ||
792 | /* Lookup connection and read fields into key. */ | 788 | /* Lookup connection and read fields into key. */ |
793 | static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key, | 789 | static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key, |
794 | const struct ovs_conntrack_info *info, | 790 | const struct ovs_conntrack_info *info, |
795 | struct sk_buff *skb) | 791 | struct sk_buff *skb) |
796 | { | 792 | { |
797 | struct nf_conntrack_expect *exp; | 793 | struct nf_conntrack_expect *exp; |
798 | 794 | ||
799 | /* If we pass an expected packet through nf_conntrack_in() the | 795 | /* If we pass an expected packet through nf_conntrack_in() the |
800 | * expectation is typically removed, but the packet could still be | 796 | * expectation is typically removed, but the packet could still be |
801 | * lost in upcall processing. To prevent this from happening we | 797 | * lost in upcall processing. To prevent this from happening we |
802 | * perform an explicit expectation lookup. Expected connections are | 798 | * perform an explicit expectation lookup. Expected connections are |
803 | * always new, and will be passed through conntrack only when they are | 799 | * always new, and will be passed through conntrack only when they are |
804 | * committed, as it is OK to remove the expectation at that time. | 800 | * committed, as it is OK to remove the expectation at that time. |
805 | */ | 801 | */ |
806 | exp = ovs_ct_expect_find(net, &info->zone, info->family, skb); | 802 | exp = ovs_ct_expect_find(net, &info->zone, info->family, skb); |
807 | if (exp) { | 803 | if (exp) { |
808 | u8 state; | 804 | u8 state; |
809 | 805 | ||
810 | /* NOTE: New connections are NATted and Helped only when | 806 | /* NOTE: New connections are NATted and Helped only when |
811 | * committed, so we are not calling into NAT here. | 807 | * committed, so we are not calling into NAT here. |
812 | */ | 808 | */ |
813 | state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED; | 809 | state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED; |
814 | __ovs_ct_update_key(key, state, &info->zone, exp->master); | 810 | __ovs_ct_update_key(key, state, &info->zone, exp->master); |
815 | } else { | 811 | } else { |
816 | struct nf_conn *ct; | 812 | struct nf_conn *ct; |
817 | int err; | 813 | int err; |
818 | 814 | ||
819 | err = __ovs_ct_lookup(net, key, info, skb); | 815 | err = __ovs_ct_lookup(net, key, info, skb); |
820 | if (err) | 816 | if (err) |
821 | return err; | 817 | return err; |
822 | 818 | ||
823 | ct = (struct nf_conn *)skb->nfct; | 819 | ct = (struct nf_conn *)skb->nfct; |
824 | if (ct) | 820 | if (ct) |
825 | nf_ct_deliver_cached_events(ct); | 821 | nf_ct_deliver_cached_events(ct); |
826 | } | 822 | } |
827 | 823 | ||
828 | return 0; | 824 | return 0; |
829 | } | 825 | } |
830 | 826 | ||
831 | static bool labels_nonzero(const struct ovs_key_ct_labels *labels) | 827 | static bool labels_nonzero(const struct ovs_key_ct_labels *labels) |
832 | { | 828 | { |
833 | size_t i; | 829 | size_t i; |
834 | 830 | ||
835 | for (i = 0; i < sizeof(*labels); i++) | 831 | for (i = 0; i < sizeof(*labels); i++) |
836 | if (labels->ct_labels[i]) | 832 | if (labels->ct_labels[i]) |
837 | return true; | 833 | return true; |
838 | 834 | ||
839 | return false; | 835 | return false; |
840 | } | 836 | } |
841 | 837 | ||
842 | /* Lookup connection and confirm if unconfirmed. */ | 838 | /* Lookup connection and confirm if unconfirmed. */ |
843 | static int ovs_ct_commit(struct net *net, struct sw_flow_key *key, | 839 | static int ovs_ct_commit(struct net *net, struct sw_flow_key *key, |
844 | const struct ovs_conntrack_info *info, | 840 | const struct ovs_conntrack_info *info, |
845 | struct sk_buff *skb) | 841 | struct sk_buff *skb) |
846 | { | 842 | { |
847 | int err; | 843 | int err; |
848 | 844 | ||
849 | err = __ovs_ct_lookup(net, key, info, skb); | 845 | err = __ovs_ct_lookup(net, key, info, skb); |
850 | if (err) | 846 | if (err) |
851 | return err; | 847 | return err; |
852 | 848 | ||
853 | /* Apply changes before confirming the connection so that the initial | 849 | /* Apply changes before confirming the connection so that the initial |
854 | * conntrack NEW netlink event carries the values given in the CT | 850 | * conntrack NEW netlink event carries the values given in the CT |
855 | * action. | 851 | * action. |
856 | */ | 852 | */ |
857 | if (info->mark.mask) { | 853 | if (info->mark.mask) { |
858 | err = ovs_ct_set_mark(skb, key, info->mark.value, | 854 | err = ovs_ct_set_mark(skb, key, info->mark.value, |
859 | info->mark.mask); | 855 | info->mark.mask); |
860 | if (err) | 856 | if (err) |
861 | return err; | 857 | return err; |
862 | } | 858 | } |
863 | if (labels_nonzero(&info->labels.mask)) { | 859 | if (labels_nonzero(&info->labels.mask)) { |
864 | err = ovs_ct_set_labels(skb, key, &info->labels.value, | 860 | err = ovs_ct_set_labels(skb, key, &info->labels.value, |
865 | &info->labels.mask); | 861 | &info->labels.mask); |
866 | if (err) | 862 | if (err) |
867 | return err; | 863 | return err; |
868 | } | 864 | } |
869 | /* This will take care of sending queued events even if the connection | 865 | /* This will take care of sending queued events even if the connection |
870 | * is already confirmed. | 866 | * is already confirmed. |
871 | */ | 867 | */ |
872 | if (nf_conntrack_confirm(skb) != NF_ACCEPT) | 868 | if (nf_conntrack_confirm(skb) != NF_ACCEPT) |
873 | return -EINVAL; | 869 | return -EINVAL; |
874 | 870 | ||
875 | return 0; | 871 | return 0; |
876 | } | 872 | } |
877 | 873 | ||
878 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero | 874 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero |
879 | * value if 'skb' is freed. | 875 | * value if 'skb' is freed. |
880 | */ | 876 | */ |
881 | int ovs_ct_execute(struct net *net, struct sk_buff *skb, | 877 | int ovs_ct_execute(struct net *net, struct sk_buff *skb, |
882 | struct sw_flow_key *key, | 878 | struct sw_flow_key *key, |
883 | const struct ovs_conntrack_info *info) | 879 | const struct ovs_conntrack_info *info) |
884 | { | 880 | { |
885 | int nh_ofs; | 881 | int nh_ofs; |
886 | int err; | 882 | int err; |
887 | 883 | ||
888 | /* The conntrack module expects to be working at L3. */ | 884 | /* The conntrack module expects to be working at L3. */ |
889 | nh_ofs = skb_network_offset(skb); | 885 | nh_ofs = skb_network_offset(skb); |
890 | skb_pull(skb, nh_ofs); | 886 | skb_pull(skb, nh_ofs); |
891 | 887 | ||
892 | if (key->ip.frag != OVS_FRAG_TYPE_NONE) { | 888 | if (key->ip.frag != OVS_FRAG_TYPE_NONE) { |
893 | err = handle_fragments(net, key, info->zone.id, skb); | 889 | err = handle_fragments(net, key, info->zone.id, skb); |
894 | if (err) | 890 | if (err) |
895 | return err; | 891 | return err; |
896 | } | 892 | } |
897 | 893 | ||
898 | if (info->commit) | 894 | if (info->commit) |
899 | err = ovs_ct_commit(net, key, info, skb); | 895 | err = ovs_ct_commit(net, key, info, skb); |
900 | else | 896 | else |
901 | err = ovs_ct_lookup(net, key, info, skb); | 897 | err = ovs_ct_lookup(net, key, info, skb); |
902 | 898 | ||
903 | skb_push(skb, nh_ofs); | 899 | skb_push(skb, nh_ofs); |
904 | if (err) | 900 | if (err) |
905 | kfree_skb(skb); | 901 | kfree_skb(skb); |
906 | return err; | 902 | return err; |
907 | } | 903 | } |
908 | 904 | ||
909 | static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name, | 905 | static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name, |
910 | const struct sw_flow_key *key, bool log) | 906 | const struct sw_flow_key *key, bool log) |
911 | { | 907 | { |
912 | struct nf_conntrack_helper *helper; | 908 | struct nf_conntrack_helper *helper; |
913 | struct nf_conn_help *help; | 909 | struct nf_conn_help *help; |
914 | 910 | ||
915 | helper = nf_conntrack_helper_try_module_get(name, info->family, | 911 | helper = nf_conntrack_helper_try_module_get(name, info->family, |
916 | key->ip.proto); | 912 | key->ip.proto); |
917 | if (!helper) { | 913 | if (!helper) { |
918 | OVS_NLERR(log, "Unknown helper \"%s\"", name); | 914 | OVS_NLERR(log, "Unknown helper \"%s\"", name); |
919 | return -EINVAL; | 915 | return -EINVAL; |
920 | } | 916 | } |
921 | 917 | ||
922 | help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL); | 918 | help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL); |
923 | if (!help) { | 919 | if (!help) { |
924 | module_put(helper->me); | 920 | module_put(helper->me); |
925 | return -ENOMEM; | 921 | return -ENOMEM; |
926 | } | 922 | } |
927 | 923 | ||
928 | rcu_assign_pointer(help->helper, helper); | 924 | rcu_assign_pointer(help->helper, helper); |
929 | info->helper = helper; | 925 | info->helper = helper; |
930 | return 0; | 926 | return 0; |
931 | } | 927 | } |
932 | 928 | ||
933 | #ifdef CONFIG_NF_NAT_NEEDED | 929 | #ifdef CONFIG_NF_NAT_NEEDED |
934 | static int parse_nat(const struct nlattr *attr, | 930 | static int parse_nat(const struct nlattr *attr, |
935 | struct ovs_conntrack_info *info, bool log) | 931 | struct ovs_conntrack_info *info, bool log) |
936 | { | 932 | { |
937 | struct nlattr *a; | 933 | struct nlattr *a; |
938 | int rem; | 934 | int rem; |
939 | bool have_ip_max = false; | 935 | bool have_ip_max = false; |
940 | bool have_proto_max = false; | 936 | bool have_proto_max = false; |
941 | bool ip_vers = (info->family == NFPROTO_IPV6); | 937 | bool ip_vers = (info->family == NFPROTO_IPV6); |
942 | 938 | ||
943 | nla_for_each_nested(a, attr, rem) { | 939 | nla_for_each_nested(a, attr, rem) { |
944 | static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = { | 940 | static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = { |
945 | [OVS_NAT_ATTR_SRC] = {0, 0}, | 941 | [OVS_NAT_ATTR_SRC] = {0, 0}, |
946 | [OVS_NAT_ATTR_DST] = {0, 0}, | 942 | [OVS_NAT_ATTR_DST] = {0, 0}, |
947 | [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr), | 943 | [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr), |
948 | sizeof(struct in6_addr)}, | 944 | sizeof(struct in6_addr)}, |
949 | [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr), | 945 | [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr), |
950 | sizeof(struct in6_addr)}, | 946 | sizeof(struct in6_addr)}, |
951 | [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)}, | 947 | [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)}, |
952 | [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)}, | 948 | [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)}, |
953 | [OVS_NAT_ATTR_PERSISTENT] = {0, 0}, | 949 | [OVS_NAT_ATTR_PERSISTENT] = {0, 0}, |
954 | [OVS_NAT_ATTR_PROTO_HASH] = {0, 0}, | 950 | [OVS_NAT_ATTR_PROTO_HASH] = {0, 0}, |
955 | [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0}, | 951 | [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0}, |
956 | }; | 952 | }; |
957 | int type = nla_type(a); | 953 | int type = nla_type(a); |
958 | 954 | ||
959 | if (type > OVS_NAT_ATTR_MAX) { | 955 | if (type > OVS_NAT_ATTR_MAX) { |
960 | OVS_NLERR(log, | 956 | OVS_NLERR(log, |
961 | "Unknown NAT attribute (type=%d, max=%d).\n", | 957 | "Unknown NAT attribute (type=%d, max=%d).\n", |
962 | type, OVS_NAT_ATTR_MAX); | 958 | type, OVS_NAT_ATTR_MAX); |
963 | return -EINVAL; | 959 | return -EINVAL; |
964 | } | 960 | } |
965 | 961 | ||
966 | if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) { | 962 | if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) { |
967 | OVS_NLERR(log, | 963 | OVS_NLERR(log, |
968 | "NAT attribute type %d has unexpected length (%d != %d).\n", | 964 | "NAT attribute type %d has unexpected length (%d != %d).\n", |
969 | type, nla_len(a), | 965 | type, nla_len(a), |
970 | ovs_nat_attr_lens[type][ip_vers]); | 966 | ovs_nat_attr_lens[type][ip_vers]); |
971 | return -EINVAL; | 967 | return -EINVAL; |
972 | } | 968 | } |
973 | 969 | ||
974 | switch (type) { | 970 | switch (type) { |
975 | case OVS_NAT_ATTR_SRC: | 971 | case OVS_NAT_ATTR_SRC: |
976 | case OVS_NAT_ATTR_DST: | 972 | case OVS_NAT_ATTR_DST: |
977 | if (info->nat) { | 973 | if (info->nat) { |
978 | OVS_NLERR(log, | 974 | OVS_NLERR(log, |
979 | "Only one type of NAT may be specified.\n" | 975 | "Only one type of NAT may be specified.\n" |
980 | ); | 976 | ); |
981 | return -ERANGE; | 977 | return -ERANGE; |
982 | } | 978 | } |
983 | info->nat |= OVS_CT_NAT; | 979 | info->nat |= OVS_CT_NAT; |
984 | info->nat |= ((type == OVS_NAT_ATTR_SRC) | 980 | info->nat |= ((type == OVS_NAT_ATTR_SRC) |
985 | ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT); | 981 | ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT); |
986 | break; | 982 | break; |
987 | 983 | ||
988 | case OVS_NAT_ATTR_IP_MIN: | 984 | case OVS_NAT_ATTR_IP_MIN: |
989 | nla_memcpy(&info->range.min_addr, a, | 985 | nla_memcpy(&info->range.min_addr, a, |
990 | sizeof(info->range.min_addr)); | 986 | sizeof(info->range.min_addr)); |
991 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; | 987 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; |
992 | break; | 988 | break; |
993 | 989 | ||
994 | case OVS_NAT_ATTR_IP_MAX: | 990 | case OVS_NAT_ATTR_IP_MAX: |
995 | have_ip_max = true; | 991 | have_ip_max = true; |
996 | nla_memcpy(&info->range.max_addr, a, | 992 | nla_memcpy(&info->range.max_addr, a, |
997 | sizeof(info->range.max_addr)); | 993 | sizeof(info->range.max_addr)); |
998 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; | 994 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; |
999 | break; | 995 | break; |
1000 | 996 | ||
1001 | case OVS_NAT_ATTR_PROTO_MIN: | 997 | case OVS_NAT_ATTR_PROTO_MIN: |
1002 | info->range.min_proto.all = htons(nla_get_u16(a)); | 998 | info->range.min_proto.all = htons(nla_get_u16(a)); |
1003 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | 999 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; |
1004 | break; | 1000 | break; |
1005 | 1001 | ||
1006 | case OVS_NAT_ATTR_PROTO_MAX: | 1002 | case OVS_NAT_ATTR_PROTO_MAX: |
1007 | have_proto_max = true; | 1003 | have_proto_max = true; |
1008 | info->range.max_proto.all = htons(nla_get_u16(a)); | 1004 | info->range.max_proto.all = htons(nla_get_u16(a)); |
1009 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | 1005 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; |
1010 | break; | 1006 | break; |
1011 | 1007 | ||
1012 | case OVS_NAT_ATTR_PERSISTENT: | 1008 | case OVS_NAT_ATTR_PERSISTENT: |
1013 | info->range.flags |= NF_NAT_RANGE_PERSISTENT; | 1009 | info->range.flags |= NF_NAT_RANGE_PERSISTENT; |
1014 | break; | 1010 | break; |
1015 | 1011 | ||
1016 | case OVS_NAT_ATTR_PROTO_HASH: | 1012 | case OVS_NAT_ATTR_PROTO_HASH: |
1017 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM; | 1013 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM; |
1018 | break; | 1014 | break; |
1019 | 1015 | ||
1020 | case OVS_NAT_ATTR_PROTO_RANDOM: | 1016 | case OVS_NAT_ATTR_PROTO_RANDOM: |
1021 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY; | 1017 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY; |
1022 | break; | 1018 | break; |
1023 | 1019 | ||
1024 | default: | 1020 | default: |
1025 | OVS_NLERR(log, "Unknown nat attribute (%d).\n", type); | 1021 | OVS_NLERR(log, "Unknown nat attribute (%d).\n", type); |
1026 | return -EINVAL; | 1022 | return -EINVAL; |
1027 | } | 1023 | } |
1028 | } | 1024 | } |
1029 | 1025 | ||
1030 | if (rem > 0) { | 1026 | if (rem > 0) { |
1031 | OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem); | 1027 | OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem); |
1032 | return -EINVAL; | 1028 | return -EINVAL; |
1033 | } | 1029 | } |
1034 | if (!info->nat) { | 1030 | if (!info->nat) { |
1035 | /* Do not allow flags if no type is given. */ | 1031 | /* Do not allow flags if no type is given. */ |
1036 | if (info->range.flags) { | 1032 | if (info->range.flags) { |
1037 | OVS_NLERR(log, | 1033 | OVS_NLERR(log, |
1038 | "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n" | 1034 | "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n" |
1039 | ); | 1035 | ); |
1040 | return -EINVAL; | 1036 | return -EINVAL; |
1041 | } | 1037 | } |
1042 | info->nat = OVS_CT_NAT; /* NAT existing connections. */ | 1038 | info->nat = OVS_CT_NAT; /* NAT existing connections. */ |
1043 | } else if (!info->commit) { | 1039 | } else if (!info->commit) { |
1044 | OVS_NLERR(log, | 1040 | OVS_NLERR(log, |
1045 | "NAT attributes may be specified only when CT COMMIT flag is also specified.\n" | 1041 | "NAT attributes may be specified only when CT COMMIT flag is also specified.\n" |
1046 | ); | 1042 | ); |
1047 | return -EINVAL; | 1043 | return -EINVAL; |
1048 | } | 1044 | } |
1049 | /* Allow missing IP_MAX. */ | 1045 | /* Allow missing IP_MAX. */ |
1050 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) { | 1046 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) { |
1051 | memcpy(&info->range.max_addr, &info->range.min_addr, | 1047 | memcpy(&info->range.max_addr, &info->range.min_addr, |
1052 | sizeof(info->range.max_addr)); | 1048 | sizeof(info->range.max_addr)); |
1053 | } | 1049 | } |
1054 | /* Allow missing PROTO_MAX. */ | 1050 | /* Allow missing PROTO_MAX. */ |
1055 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | 1051 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && |
1056 | !have_proto_max) { | 1052 | !have_proto_max) { |
1057 | info->range.max_proto.all = info->range.min_proto.all; | 1053 | info->range.max_proto.all = info->range.min_proto.all; |
1058 | } | 1054 | } |
1059 | return 0; | 1055 | return 0; |
1060 | } | 1056 | } |
1061 | #endif | 1057 | #endif |
1062 | 1058 | ||
1063 | static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = { | 1059 | static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = { |
1064 | [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 }, | 1060 | [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 }, |
1065 | [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16), | 1061 | [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16), |
1066 | .maxlen = sizeof(u16) }, | 1062 | .maxlen = sizeof(u16) }, |
1067 | [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark), | 1063 | [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark), |
1068 | .maxlen = sizeof(struct md_mark) }, | 1064 | .maxlen = sizeof(struct md_mark) }, |
1069 | [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels), | 1065 | [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels), |
1070 | .maxlen = sizeof(struct md_labels) }, | 1066 | .maxlen = sizeof(struct md_labels) }, |
1071 | [OVS_CT_ATTR_HELPER] = { .minlen = 1, | 1067 | [OVS_CT_ATTR_HELPER] = { .minlen = 1, |
1072 | .maxlen = NF_CT_HELPER_NAME_LEN }, | 1068 | .maxlen = NF_CT_HELPER_NAME_LEN }, |
1073 | #ifdef CONFIG_NF_NAT_NEEDED | 1069 | #ifdef CONFIG_NF_NAT_NEEDED |
1074 | /* NAT length is checked when parsing the nested attributes. */ | 1070 | /* NAT length is checked when parsing the nested attributes. */ |
1075 | [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX }, | 1071 | [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX }, |
1076 | #endif | 1072 | #endif |
1077 | }; | 1073 | }; |
1078 | 1074 | ||
1079 | static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info, | 1075 | static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info, |
1080 | const char **helper, bool log) | 1076 | const char **helper, bool log) |
1081 | { | 1077 | { |
1082 | struct nlattr *a; | 1078 | struct nlattr *a; |
1083 | int rem; | 1079 | int rem; |
1084 | 1080 | ||
1085 | nla_for_each_nested(a, attr, rem) { | 1081 | nla_for_each_nested(a, attr, rem) { |
1086 | int type = nla_type(a); | 1082 | int type = nla_type(a); |
1087 | int maxlen = ovs_ct_attr_lens[type].maxlen; | 1083 | int maxlen = ovs_ct_attr_lens[type].maxlen; |
1088 | int minlen = ovs_ct_attr_lens[type].minlen; | 1084 | int minlen = ovs_ct_attr_lens[type].minlen; |
1089 | 1085 | ||
1090 | if (type > OVS_CT_ATTR_MAX) { | 1086 | if (type > OVS_CT_ATTR_MAX) { |
1091 | OVS_NLERR(log, | 1087 | OVS_NLERR(log, |
1092 | "Unknown conntrack attr (type=%d, max=%d)", | 1088 | "Unknown conntrack attr (type=%d, max=%d)", |
1093 | type, OVS_CT_ATTR_MAX); | 1089 | type, OVS_CT_ATTR_MAX); |
1094 | return -EINVAL; | 1090 | return -EINVAL; |
1095 | } | 1091 | } |
1096 | if (nla_len(a) < minlen || nla_len(a) > maxlen) { | 1092 | if (nla_len(a) < minlen || nla_len(a) > maxlen) { |
1097 | OVS_NLERR(log, | 1093 | OVS_NLERR(log, |
1098 | "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)", | 1094 | "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)", |
1099 | type, nla_len(a), maxlen); | 1095 | type, nla_len(a), maxlen); |
1100 | return -EINVAL; | 1096 | return -EINVAL; |
1101 | } | 1097 | } |
1102 | 1098 | ||
1103 | switch (type) { | 1099 | switch (type) { |
1104 | case OVS_CT_ATTR_COMMIT: | 1100 | case OVS_CT_ATTR_COMMIT: |
1105 | info->commit = true; | 1101 | info->commit = true; |
1106 | break; | 1102 | break; |
1107 | #ifdef CONFIG_NF_CONNTRACK_ZONES | 1103 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1108 | case OVS_CT_ATTR_ZONE: | 1104 | case OVS_CT_ATTR_ZONE: |
1109 | info->zone.id = nla_get_u16(a); | 1105 | info->zone.id = nla_get_u16(a); |
1110 | break; | 1106 | break; |
1111 | #endif | 1107 | #endif |
1112 | #ifdef CONFIG_NF_CONNTRACK_MARK | 1108 | #ifdef CONFIG_NF_CONNTRACK_MARK |
1113 | case OVS_CT_ATTR_MARK: { | 1109 | case OVS_CT_ATTR_MARK: { |
1114 | struct md_mark *mark = nla_data(a); | 1110 | struct md_mark *mark = nla_data(a); |
1115 | 1111 | ||
1116 | if (!mark->mask) { | 1112 | if (!mark->mask) { |
1117 | OVS_NLERR(log, "ct_mark mask cannot be 0"); | 1113 | OVS_NLERR(log, "ct_mark mask cannot be 0"); |
1118 | return -EINVAL; | 1114 | return -EINVAL; |
1119 | } | 1115 | } |
1120 | info->mark = *mark; | 1116 | info->mark = *mark; |
1121 | break; | 1117 | break; |
1122 | } | 1118 | } |
1123 | #endif | 1119 | #endif |
1124 | #ifdef CONFIG_NF_CONNTRACK_LABELS | 1120 | #ifdef CONFIG_NF_CONNTRACK_LABELS |
1125 | case OVS_CT_ATTR_LABELS: { | 1121 | case OVS_CT_ATTR_LABELS: { |
1126 | struct md_labels *labels = nla_data(a); | 1122 | struct md_labels *labels = nla_data(a); |
1127 | 1123 | ||
1128 | if (!labels_nonzero(&labels->mask)) { | 1124 | if (!labels_nonzero(&labels->mask)) { |
1129 | OVS_NLERR(log, "ct_labels mask cannot be 0"); | 1125 | OVS_NLERR(log, "ct_labels mask cannot be 0"); |
1130 | return -EINVAL; | 1126 | return -EINVAL; |
1131 | } | 1127 | } |
1132 | info->labels = *labels; | 1128 | info->labels = *labels; |
1133 | break; | 1129 | break; |
1134 | } | 1130 | } |
1135 | #endif | 1131 | #endif |
1136 | case OVS_CT_ATTR_HELPER: | 1132 | case OVS_CT_ATTR_HELPER: |
1137 | *helper = nla_data(a); | 1133 | *helper = nla_data(a); |
1138 | if (!memchr(*helper, '\0', nla_len(a))) { | 1134 | if (!memchr(*helper, '\0', nla_len(a))) { |
1139 | OVS_NLERR(log, "Invalid conntrack helper"); | 1135 | OVS_NLERR(log, "Invalid conntrack helper"); |
1140 | return -EINVAL; | 1136 | return -EINVAL; |
1141 | } | 1137 | } |
1142 | break; | 1138 | break; |
1143 | #ifdef CONFIG_NF_NAT_NEEDED | 1139 | #ifdef CONFIG_NF_NAT_NEEDED |
1144 | case OVS_CT_ATTR_NAT: { | 1140 | case OVS_CT_ATTR_NAT: { |
1145 | int err = parse_nat(a, info, log); | 1141 | int err = parse_nat(a, info, log); |
1146 | 1142 | ||
1147 | if (err) | 1143 | if (err) |
1148 | return err; | 1144 | return err; |
1149 | break; | 1145 | break; |
1150 | } | 1146 | } |
1151 | #endif | 1147 | #endif |
1152 | default: | 1148 | default: |
1153 | OVS_NLERR(log, "Unknown conntrack attr (%d)", | 1149 | OVS_NLERR(log, "Unknown conntrack attr (%d)", |
1154 | type); | 1150 | type); |
1155 | return -EINVAL; | 1151 | return -EINVAL; |
1156 | } | 1152 | } |
1157 | } | 1153 | } |
1158 | 1154 | ||
1159 | #ifdef CONFIG_NF_CONNTRACK_MARK | 1155 | #ifdef CONFIG_NF_CONNTRACK_MARK |
1160 | if (!info->commit && info->mark.mask) { | 1156 | if (!info->commit && info->mark.mask) { |
1161 | OVS_NLERR(log, | 1157 | OVS_NLERR(log, |
1162 | "Setting conntrack mark requires 'commit' flag."); | 1158 | "Setting conntrack mark requires 'commit' flag."); |
1163 | return -EINVAL; | 1159 | return -EINVAL; |
1164 | } | 1160 | } |
1165 | #endif | 1161 | #endif |
1166 | #ifdef CONFIG_NF_CONNTRACK_LABELS | 1162 | #ifdef CONFIG_NF_CONNTRACK_LABELS |
1167 | if (!info->commit && labels_nonzero(&info->labels.mask)) { | 1163 | if (!info->commit && labels_nonzero(&info->labels.mask)) { |
1168 | OVS_NLERR(log, | 1164 | OVS_NLERR(log, |
1169 | "Setting conntrack labels requires 'commit' flag."); | 1165 | "Setting conntrack labels requires 'commit' flag."); |
1170 | return -EINVAL; | 1166 | return -EINVAL; |
1171 | } | 1167 | } |
1172 | #endif | 1168 | #endif |
1173 | if (rem > 0) { | 1169 | if (rem > 0) { |
1174 | OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem); | 1170 | OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem); |
1175 | return -EINVAL; | 1171 | return -EINVAL; |
1176 | } | 1172 | } |
1177 | 1173 | ||
1178 | return 0; | 1174 | return 0; |
1179 | } | 1175 | } |
1180 | 1176 | ||
1181 | bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr) | 1177 | bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr) |
1182 | { | 1178 | { |
1183 | if (attr == OVS_KEY_ATTR_CT_STATE) | 1179 | if (attr == OVS_KEY_ATTR_CT_STATE) |
1184 | return true; | 1180 | return true; |
1185 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | 1181 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && |
1186 | attr == OVS_KEY_ATTR_CT_ZONE) | 1182 | attr == OVS_KEY_ATTR_CT_ZONE) |
1187 | return true; | 1183 | return true; |
1188 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && | 1184 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && |
1189 | attr == OVS_KEY_ATTR_CT_MARK) | 1185 | attr == OVS_KEY_ATTR_CT_MARK) |
1190 | return true; | 1186 | return true; |
1191 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && | 1187 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
1192 | attr == OVS_KEY_ATTR_CT_LABELS) { | 1188 | attr == OVS_KEY_ATTR_CT_LABELS) { |
1193 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); | 1189 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1194 | 1190 | ||
1195 | return ovs_net->xt_label; | 1191 | return ovs_net->xt_label; |
1196 | } | 1192 | } |
1197 | 1193 | ||
1198 | return false; | 1194 | return false; |
1199 | } | 1195 | } |
1200 | 1196 | ||
1201 | int ovs_ct_copy_action(struct net *net, const struct nlattr *attr, | 1197 | int ovs_ct_copy_action(struct net *net, const struct nlattr *attr, |
1202 | const struct sw_flow_key *key, | 1198 | const struct sw_flow_key *key, |
1203 | struct sw_flow_actions **sfa, bool log) | 1199 | struct sw_flow_actions **sfa, bool log) |
1204 | { | 1200 | { |
1205 | struct ovs_conntrack_info ct_info; | 1201 | struct ovs_conntrack_info ct_info; |
1206 | const char *helper = NULL; | 1202 | const char *helper = NULL; |
1207 | u16 family; | 1203 | u16 family; |
1208 | int err; | 1204 | int err; |
1209 | 1205 | ||
1210 | family = key_to_nfproto(key); | 1206 | family = key_to_nfproto(key); |
1211 | if (family == NFPROTO_UNSPEC) { | 1207 | if (family == NFPROTO_UNSPEC) { |
1212 | OVS_NLERR(log, "ct family unspecified"); | 1208 | OVS_NLERR(log, "ct family unspecified"); |
1213 | return -EINVAL; | 1209 | return -EINVAL; |
1214 | } | 1210 | } |
1215 | 1211 | ||
1216 | memset(&ct_info, 0, sizeof(ct_info)); | 1212 | memset(&ct_info, 0, sizeof(ct_info)); |
1217 | ct_info.family = family; | 1213 | ct_info.family = family; |
1218 | 1214 | ||
1219 | nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID, | 1215 | nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID, |
1220 | NF_CT_DEFAULT_ZONE_DIR, 0); | 1216 | NF_CT_DEFAULT_ZONE_DIR, 0); |
1221 | 1217 | ||
1222 | err = parse_ct(attr, &ct_info, &helper, log); | 1218 | err = parse_ct(attr, &ct_info, &helper, log); |
1223 | if (err) | 1219 | if (err) |
1224 | return err; | 1220 | return err; |
1225 | 1221 | ||
1226 | /* Set up template for tracking connections in specific zones. */ | 1222 | /* Set up template for tracking connections in specific zones. */ |
1227 | ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL); | 1223 | ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL); |
1228 | if (!ct_info.ct) { | 1224 | if (!ct_info.ct) { |
1229 | OVS_NLERR(log, "Failed to allocate conntrack template"); | 1225 | OVS_NLERR(log, "Failed to allocate conntrack template"); |
1230 | return -ENOMEM; | 1226 | return -ENOMEM; |
1231 | } | 1227 | } |
1232 | 1228 | ||
1233 | __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status); | 1229 | __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status); |
1234 | nf_conntrack_get(&ct_info.ct->ct_general); | 1230 | nf_conntrack_get(&ct_info.ct->ct_general); |
1235 | 1231 | ||
1236 | if (helper) { | 1232 | if (helper) { |
1237 | err = ovs_ct_add_helper(&ct_info, helper, key, log); | 1233 | err = ovs_ct_add_helper(&ct_info, helper, key, log); |
1238 | if (err) | 1234 | if (err) |
1239 | goto err_free_ct; | 1235 | goto err_free_ct; |
1240 | } | 1236 | } |
1241 | 1237 | ||
1242 | err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info, | 1238 | err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info, |
1243 | sizeof(ct_info), log); | 1239 | sizeof(ct_info), log); |
1244 | if (err) | 1240 | if (err) |
1245 | goto err_free_ct; | 1241 | goto err_free_ct; |
1246 | 1242 | ||
1247 | return 0; | 1243 | return 0; |
1248 | err_free_ct: | 1244 | err_free_ct: |
1249 | __ovs_ct_free_action(&ct_info); | 1245 | __ovs_ct_free_action(&ct_info); |
1250 | return err; | 1246 | return err; |
1251 | } | 1247 | } |
1252 | 1248 | ||
1253 | #ifdef CONFIG_NF_NAT_NEEDED | 1249 | #ifdef CONFIG_NF_NAT_NEEDED |
1254 | static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info, | 1250 | static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info, |
1255 | struct sk_buff *skb) | 1251 | struct sk_buff *skb) |
1256 | { | 1252 | { |
1257 | struct nlattr *start; | 1253 | struct nlattr *start; |
1258 | 1254 | ||
1259 | start = nla_nest_start(skb, OVS_CT_ATTR_NAT); | 1255 | start = nla_nest_start(skb, OVS_CT_ATTR_NAT); |
1260 | if (!start) | 1256 | if (!start) |
1261 | return false; | 1257 | return false; |
1262 | 1258 | ||
1263 | if (info->nat & OVS_CT_SRC_NAT) { | 1259 | if (info->nat & OVS_CT_SRC_NAT) { |
1264 | if (nla_put_flag(skb, OVS_NAT_ATTR_SRC)) | 1260 | if (nla_put_flag(skb, OVS_NAT_ATTR_SRC)) |
1265 | return false; | 1261 | return false; |
1266 | } else if (info->nat & OVS_CT_DST_NAT) { | 1262 | } else if (info->nat & OVS_CT_DST_NAT) { |
1267 | if (nla_put_flag(skb, OVS_NAT_ATTR_DST)) | 1263 | if (nla_put_flag(skb, OVS_NAT_ATTR_DST)) |
1268 | return false; | 1264 | return false; |
1269 | } else { | 1265 | } else { |
1270 | goto out; | 1266 | goto out; |
1271 | } | 1267 | } |
1272 | 1268 | ||
1273 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS) { | 1269 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS) { |
1274 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && | 1270 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
1275 | info->family == NFPROTO_IPV4) { | 1271 | info->family == NFPROTO_IPV4) { |
1276 | if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN, | 1272 | if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1277 | info->range.min_addr.ip) || | 1273 | info->range.min_addr.ip) || |
1278 | (info->range.max_addr.ip | 1274 | (info->range.max_addr.ip |
1279 | != info->range.min_addr.ip && | 1275 | != info->range.min_addr.ip && |
1280 | (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX, | 1276 | (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX, |
1281 | info->range.max_addr.ip)))) | 1277 | info->range.max_addr.ip)))) |
1282 | return false; | 1278 | return false; |
1283 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && | 1279 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
1284 | info->family == NFPROTO_IPV6) { | 1280 | info->family == NFPROTO_IPV6) { |
1285 | if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN, | 1281 | if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1286 | &info->range.min_addr.in6) || | 1282 | &info->range.min_addr.in6) || |
1287 | (memcmp(&info->range.max_addr.in6, | 1283 | (memcmp(&info->range.max_addr.in6, |
1288 | &info->range.min_addr.in6, | 1284 | &info->range.min_addr.in6, |
1289 | sizeof(info->range.max_addr.in6)) && | 1285 | sizeof(info->range.max_addr.in6)) && |
1290 | (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX, | 1286 | (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX, |
1291 | &info->range.max_addr.in6)))) | 1287 | &info->range.max_addr.in6)))) |
1292 | return false; | 1288 | return false; |
1293 | } else { | 1289 | } else { |
1294 | return false; | 1290 | return false; |
1295 | } | 1291 | } |
1296 | } | 1292 | } |
1297 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | 1293 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && |
1298 | (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN, | 1294 | (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN, |
1299 | ntohs(info->range.min_proto.all)) || | 1295 | ntohs(info->range.min_proto.all)) || |
1300 | (info->range.max_proto.all != info->range.min_proto.all && | 1296 | (info->range.max_proto.all != info->range.min_proto.all && |
1301 | nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX, | 1297 | nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX, |
1302 | ntohs(info->range.max_proto.all))))) | 1298 | ntohs(info->range.max_proto.all))))) |
1303 | return false; | 1299 | return false; |
1304 | 1300 | ||
1305 | if (info->range.flags & NF_NAT_RANGE_PERSISTENT && | 1301 | if (info->range.flags & NF_NAT_RANGE_PERSISTENT && |
1306 | nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT)) | 1302 | nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT)) |
1307 | return false; | 1303 | return false; |
1308 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM && | 1304 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM && |
1309 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH)) | 1305 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH)) |
1310 | return false; | 1306 | return false; |
1311 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY && | 1307 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY && |
1312 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM)) | 1308 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM)) |
1313 | return false; | 1309 | return false; |
1314 | out: | 1310 | out: |
1315 | nla_nest_end(skb, start); | 1311 | nla_nest_end(skb, start); |
1316 | 1312 | ||
1317 | return true; | 1313 | return true; |
1318 | } | 1314 | } |
1319 | #endif | 1315 | #endif |
1320 | 1316 | ||
1321 | int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info, | 1317 | int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info, |
1322 | struct sk_buff *skb) | 1318 | struct sk_buff *skb) |
1323 | { | 1319 | { |
1324 | struct nlattr *start; | 1320 | struct nlattr *start; |
1325 | 1321 | ||
1326 | start = nla_nest_start(skb, OVS_ACTION_ATTR_CT); | 1322 | start = nla_nest_start(skb, OVS_ACTION_ATTR_CT); |
1327 | if (!start) | 1323 | if (!start) |
1328 | return -EMSGSIZE; | 1324 | return -EMSGSIZE; |
1329 | 1325 | ||
1330 | if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT)) | 1326 | if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT)) |
1331 | return -EMSGSIZE; | 1327 | return -EMSGSIZE; |
1332 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | 1328 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && |
1333 | nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id)) | 1329 | nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id)) |
1334 | return -EMSGSIZE; | 1330 | return -EMSGSIZE; |
1335 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask && | 1331 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask && |
1336 | nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark), | 1332 | nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark), |
1337 | &ct_info->mark)) | 1333 | &ct_info->mark)) |
1338 | return -EMSGSIZE; | 1334 | return -EMSGSIZE; |
1339 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && | 1335 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
1340 | labels_nonzero(&ct_info->labels.mask) && | 1336 | labels_nonzero(&ct_info->labels.mask) && |
1341 | nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels), | 1337 | nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels), |
1342 | &ct_info->labels)) | 1338 | &ct_info->labels)) |
1343 | return -EMSGSIZE; | 1339 | return -EMSGSIZE; |
1344 | if (ct_info->helper) { | 1340 | if (ct_info->helper) { |
1345 | if (nla_put_string(skb, OVS_CT_ATTR_HELPER, | 1341 | if (nla_put_string(skb, OVS_CT_ATTR_HELPER, |
1346 | ct_info->helper->name)) | 1342 | ct_info->helper->name)) |
1347 | return -EMSGSIZE; | 1343 | return -EMSGSIZE; |
1348 | } | 1344 | } |
1349 | #ifdef CONFIG_NF_NAT_NEEDED | 1345 | #ifdef CONFIG_NF_NAT_NEEDED |
1350 | if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb)) | 1346 | if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb)) |
1351 | return -EMSGSIZE; | 1347 | return -EMSGSIZE; |
1352 | #endif | 1348 | #endif |
1353 | nla_nest_end(skb, start); | 1349 | nla_nest_end(skb, start); |
1354 | 1350 | ||
1355 | return 0; | 1351 | return 0; |
1356 | } | 1352 | } |
1357 | 1353 | ||
1358 | void ovs_ct_free_action(const struct nlattr *a) | 1354 | void ovs_ct_free_action(const struct nlattr *a) |
1359 | { | 1355 | { |
1360 | struct ovs_conntrack_info *ct_info = nla_data(a); | 1356 | struct ovs_conntrack_info *ct_info = nla_data(a); |
1361 | 1357 | ||
1362 | __ovs_ct_free_action(ct_info); | 1358 | __ovs_ct_free_action(ct_info); |
1363 | } | 1359 | } |
1364 | 1360 | ||
1365 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info) | 1361 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info) |
1366 | { | 1362 | { |
1367 | if (ct_info->helper) | 1363 | if (ct_info->helper) |
1368 | module_put(ct_info->helper->me); | 1364 | module_put(ct_info->helper->me); |
1369 | if (ct_info->ct) | 1365 | if (ct_info->ct) |
1370 | nf_ct_tmpl_free(ct_info->ct); | 1366 | nf_ct_tmpl_free(ct_info->ct); |
1371 | } | 1367 | } |
1372 | 1368 | ||
1373 | void ovs_ct_init(struct net *net) | 1369 | void ovs_ct_init(struct net *net) |
1374 | { | 1370 | { |
1375 | unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE; | 1371 | unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE; |
1376 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); | 1372 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1377 | 1373 | ||
1378 | if (nf_connlabels_get(net, n_bits - 1)) { | 1374 | if (nf_connlabels_get(net, n_bits - 1)) { |
1379 | ovs_net->xt_label = false; | 1375 | ovs_net->xt_label = false; |
1380 | OVS_NLERR(true, "Failed to set connlabel length"); | 1376 | OVS_NLERR(true, "Failed to set connlabel length"); |
1381 | } else { | 1377 | } else { |
1382 | ovs_net->xt_label = true; | 1378 | ovs_net->xt_label = true; |
1383 | } | 1379 | } |
1384 | } | 1380 | } |
1385 | 1381 | ||
1386 | void ovs_ct_exit(struct net *net) | 1382 | void ovs_ct_exit(struct net *net) |
1387 | { | 1383 | { |
1388 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); | 1384 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1389 | 1385 | ||
1390 | if (ovs_net->xt_label) | 1386 | if (ovs_net->xt_label) |
1391 | nf_connlabels_put(net); | 1387 | nf_connlabels_put(net); |
1392 | } | 1388 | } |
1393 | 1389 |