1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Rusty Russell (C)2000 -- This code is GPL.
4 * Patrick McHardy (c) 2006-2012
5 */
6
7 #include <linux/kernel.h>
8 #include <linux/slab.h>
9 #include <linux/init.h>
10 #include <linux/module.h>
11 #include <linux/proc_fs.h>
12 #include <linux/skbuff.h>
13 #include <linux/netfilter.h>
14 #include <linux/netfilter_ipv4.h>
15 #include <linux/netfilter_ipv6.h>
16 #include <linux/netfilter_bridge.h>
17 #include <linux/seq_file.h>
18 #include <linux/rcupdate.h>
19 #include <net/protocol.h>
20 #include <net/netfilter/nf_queue.h>
21 #include <net/dst.h>
22
23 #include "nf_internals.h"
24
25 static const struct nf_queue_handler __rcu *nf_queue_handler;
26
27 /*
28 * Hook for nfnetlink_queue to register its queue handler.
29 * We do this so that most of the NFQUEUE code can be modular.
30 *
31 * Once the queue is registered it must reinject all packets it
32 * receives, no matter what.
33 */
34
nf_register_queue_handler(const struct nf_queue_handler * qh)35 void nf_register_queue_handler(const struct nf_queue_handler *qh)
36 {
37 /* should never happen, we only have one queueing backend in kernel */
38 WARN_ON(rcu_access_pointer(nf_queue_handler));
39 rcu_assign_pointer(nf_queue_handler, qh);
40 }
41 EXPORT_SYMBOL(nf_register_queue_handler);
42
43 /* The caller must flush their queue before this */
nf_unregister_queue_handler(void)44 void nf_unregister_queue_handler(void)
45 {
46 RCU_INIT_POINTER(nf_queue_handler, NULL);
47 }
48 EXPORT_SYMBOL(nf_unregister_queue_handler);
49
nf_queue_sock_put(struct sock * sk)50 static void nf_queue_sock_put(struct sock *sk)
51 {
52 #ifdef CONFIG_INET
53 sock_gen_put(sk);
54 #else
55 sock_put(sk);
56 #endif
57 }
58
nf_queue_entry_release_refs(struct nf_queue_entry * entry)59 static void nf_queue_entry_release_refs(struct nf_queue_entry *entry)
60 {
61 struct nf_hook_state *state = &entry->state;
62
63 /* Release those devices we held, or Alexey will kill me. */
64 dev_put(entry->skb_dev);
65 dev_put(state->in);
66 dev_put(state->out);
67 if (state->sk)
68 nf_queue_sock_put(state->sk);
69
70 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
71 dev_put(entry->bridge_dev);
72 dev_put(entry->physin);
73 dev_put(entry->physout);
74 #endif
75 }
76
nf_queue_entry_free(struct nf_queue_entry * entry)77 void nf_queue_entry_free(struct nf_queue_entry *entry)
78 {
79 nf_queue_entry_release_refs(entry);
80 kfree(entry);
81 }
82 EXPORT_SYMBOL_GPL(nf_queue_entry_free);
83
__nf_queue_entry_init_physdevs(struct nf_queue_entry * entry)84 static void __nf_queue_entry_init_physdevs(struct nf_queue_entry *entry)
85 {
86 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
87 const struct sk_buff *skb = entry->skb;
88 struct dst_entry *dst = skb_dst(skb);
89 struct net_device *dev = NULL;
90
91 if (nf_bridge_info_exists(skb)) {
92 entry->physin = nf_bridge_get_physindev(skb, entry->state.net);
93 entry->physout = nf_bridge_get_physoutdev(skb);
94 } else {
95 entry->physin = NULL;
96 entry->physout = NULL;
97 }
98
99 if (entry->state.pf == NFPROTO_BRIDGE &&
100 dst && (dst->flags & DST_FAKE_RTABLE))
101 dev = dst_dev_rcu(dst);
102
103 /* Must hold a reference on the bridge device: dst_hold() protects
104 * the dst itself, but the fake rtable is embedded in bridge-private
105 * storage that netdevice teardown can free independently.
106 */
107 entry->bridge_dev = dev;
108 #endif
109 }
110
111 /* Bump dev refs so they don't vanish while packet is out */
nf_queue_entry_get_refs(struct nf_queue_entry * entry)112 bool nf_queue_entry_get_refs(struct nf_queue_entry *entry)
113 {
114 struct nf_hook_state *state = &entry->state;
115
116 if (state->sk && !refcount_inc_not_zero(&state->sk->sk_refcnt))
117 return false;
118
119 dev_hold(entry->skb_dev);
120 dev_hold(state->in);
121 dev_hold(state->out);
122
123 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
124 dev_hold(entry->bridge_dev);
125 dev_hold(entry->physin);
126 dev_hold(entry->physout);
127 #endif
128 return true;
129 }
130 EXPORT_SYMBOL_GPL(nf_queue_entry_get_refs);
131
nf_queue_nf_hook_drop(struct net * net)132 void nf_queue_nf_hook_drop(struct net *net)
133 {
134 const struct nf_queue_handler *qh;
135
136 rcu_read_lock();
137 qh = rcu_dereference(nf_queue_handler);
138 if (qh)
139 qh->nf_hook_drop(net);
140 rcu_read_unlock();
141 }
142 EXPORT_SYMBOL_GPL(nf_queue_nf_hook_drop);
143
nf_ip_saveroute(const struct sk_buff * skb,struct nf_queue_entry * entry)144 static void nf_ip_saveroute(const struct sk_buff *skb,
145 struct nf_queue_entry *entry)
146 {
147 struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
148
149 if (entry->state.hook == NF_INET_LOCAL_OUT) {
150 const struct iphdr *iph = ip_hdr(skb);
151
152 rt_info->tos = iph->tos;
153 rt_info->daddr = iph->daddr;
154 rt_info->saddr = iph->saddr;
155 rt_info->mark = skb->mark;
156 }
157 }
158
nf_ip6_saveroute(const struct sk_buff * skb,struct nf_queue_entry * entry)159 static void nf_ip6_saveroute(const struct sk_buff *skb,
160 struct nf_queue_entry *entry)
161 {
162 struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
163
164 if (entry->state.hook == NF_INET_LOCAL_OUT) {
165 const struct ipv6hdr *iph = ipv6_hdr(skb);
166
167 rt_info->daddr = iph->daddr;
168 rt_info->saddr = iph->saddr;
169 rt_info->mark = skb->mark;
170 }
171 }
172
__nf_queue(struct sk_buff * skb,const struct nf_hook_state * state,unsigned int index,unsigned int queuenum)173 static int __nf_queue(struct sk_buff *skb, const struct nf_hook_state *state,
174 unsigned int index, unsigned int queuenum)
175 {
176 struct nf_queue_entry *entry = NULL;
177 const struct nf_queue_handler *qh;
178 unsigned int route_key_size;
179 int status;
180
181 /* QUEUE == DROP if no one is waiting, to be safe. */
182 qh = rcu_dereference(nf_queue_handler);
183 if (!qh)
184 return -ESRCH;
185
186 switch (state->pf) {
187 case AF_INET:
188 route_key_size = sizeof(struct ip_rt_info);
189 break;
190 case AF_INET6:
191 route_key_size = sizeof(struct ip6_rt_info);
192 break;
193 default:
194 route_key_size = 0;
195 break;
196 }
197
198 if (skb_sk_is_prefetched(skb)) {
199 struct sock *sk = skb->sk;
200
201 if (!sk_is_refcounted(sk)) {
202 if (!refcount_inc_not_zero(&sk->sk_refcnt))
203 return -ENOTCONN;
204
205 /* drop refcount on skb_orphan */
206 skb->destructor = sock_edemux;
207 }
208 }
209
210 entry = kmalloc(sizeof(*entry) + route_key_size, GFP_ATOMIC);
211 if (!entry)
212 return -ENOMEM;
213
214 if (skb_dst(skb) && !skb_dst_force(skb)) {
215 kfree(entry);
216 return -ENETDOWN;
217 }
218
219 *entry = (struct nf_queue_entry) {
220 .skb = skb,
221 .skb_dev = skb->dev,
222 .state = *state,
223 .hook_index = index,
224 .size = sizeof(*entry) + route_key_size,
225 };
226 __nf_queue_entry_init_physdevs(entry);
227
228 if (!nf_queue_entry_get_refs(entry)) {
229 kfree(entry);
230 return -ENOTCONN;
231 }
232
233 switch (entry->state.pf) {
234 case AF_INET:
235 nf_ip_saveroute(skb, entry);
236 break;
237 case AF_INET6:
238 nf_ip6_saveroute(skb, entry);
239 break;
240 }
241
242 status = qh->outfn(entry, queuenum);
243 if (status < 0) {
244 nf_queue_entry_free(entry);
245 return status;
246 }
247
248 return 0;
249 }
250
251 /* Packets leaving via this function must come back through nf_reinject(). */
nf_queue(struct sk_buff * skb,struct nf_hook_state * state,unsigned int index,unsigned int verdict)252 int nf_queue(struct sk_buff *skb, struct nf_hook_state *state,
253 unsigned int index, unsigned int verdict)
254 {
255 int ret;
256
257 ret = __nf_queue(skb, state, index, verdict >> NF_VERDICT_QBITS);
258 if (ret < 0) {
259 if (ret == -ESRCH &&
260 (verdict & NF_VERDICT_FLAG_QUEUE_BYPASS))
261 return 1;
262 kfree_skb(skb);
263 }
264
265 return 0;
266 }
267 EXPORT_SYMBOL_GPL(nf_queue);
268