[TCP]: tcp_memory_pressure and tcp_socket are__read_mostly candidates
[linux-2.6] / net / ipv4 / xfrm4_input.c
1 /*
2  * xfrm4_input.c
3  *
4  * Changes:
5  *      YOSHIFUJI Hideaki @USAGI
6  *              Split up af-specific portion
7  *      Derek Atkins <derek@ihtfp.com>
8  *              Add Encapsulation support
9  *
10  */
11
12 #include <linux/module.h>
13 #include <linux/string.h>
14 #include <linux/netfilter.h>
15 #include <linux/netfilter_ipv4.h>
16 #include <net/ip.h>
17 #include <net/xfrm.h>
18
19 int xfrm4_rcv(struct sk_buff *skb)
20 {
21         return xfrm4_rcv_encap(skb, 0);
22 }
23
24 EXPORT_SYMBOL(xfrm4_rcv);
25
26 static int xfrm4_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
27 {
28         switch (nexthdr) {
29         case IPPROTO_IPIP:
30         case IPPROTO_IPV6:
31                 *spi = ip_hdr(skb)->saddr;
32                 *seq = 0;
33                 return 0;
34         }
35
36         return xfrm_parse_spi(skb, nexthdr, spi, seq);
37 }
38
39 #ifdef CONFIG_NETFILTER
40 static inline int xfrm4_rcv_encap_finish(struct sk_buff *skb)
41 {
42         if (skb->dst == NULL) {
43                 const struct iphdr *iph = ip_hdr(skb);
44
45                 if (ip_route_input(skb, iph->daddr, iph->saddr, iph->tos,
46                                    skb->dev))
47                         goto drop;
48         }
49         return dst_input(skb);
50 drop:
51         kfree_skb(skb);
52         return NET_RX_DROP;
53 }
54 #endif
55
56 int xfrm4_rcv_encap(struct sk_buff *skb, __u16 encap_type)
57 {
58         __be32 spi, seq;
59         struct xfrm_state *xfrm_vec[XFRM_MAX_DEPTH];
60         struct xfrm_state *x;
61         int xfrm_nr = 0;
62         int decaps = 0;
63         int err = xfrm4_parse_spi(skb, ip_hdr(skb)->protocol, &spi, &seq);
64
65         if (err != 0)
66                 goto drop;
67
68         do {
69                 const struct iphdr *iph = ip_hdr(skb);
70
71                 if (xfrm_nr == XFRM_MAX_DEPTH)
72                         goto drop;
73
74                 x = xfrm_state_lookup((xfrm_address_t *)&iph->daddr, spi,
75                                 iph->protocol != IPPROTO_IPV6 ? iph->protocol : IPPROTO_IPIP, AF_INET);
76                 if (x == NULL)
77                         goto drop;
78
79                 spin_lock(&x->lock);
80                 if (unlikely(x->km.state != XFRM_STATE_VALID))
81                         goto drop_unlock;
82
83                 if ((x->encap ? x->encap->encap_type : 0) != encap_type)
84                         goto drop_unlock;
85
86                 if (x->props.replay_window && xfrm_replay_check(x, seq))
87                         goto drop_unlock;
88
89                 if (xfrm_state_check_expire(x))
90                         goto drop_unlock;
91
92                 if (x->type->input(x, skb))
93                         goto drop_unlock;
94
95                 /* only the first xfrm gets the encap type */
96                 encap_type = 0;
97
98                 if (x->props.replay_window)
99                         xfrm_replay_advance(x, seq);
100
101                 x->curlft.bytes += skb->len;
102                 x->curlft.packets++;
103
104                 spin_unlock(&x->lock);
105
106                 xfrm_vec[xfrm_nr++] = x;
107
108                 if (x->mode->input(x, skb))
109                         goto drop;
110
111                 if (x->props.mode == XFRM_MODE_TUNNEL) {
112                         decaps = 1;
113                         break;
114                 }
115
116                 err = xfrm_parse_spi(skb, ip_hdr(skb)->protocol, &spi, &seq);
117                 if (err < 0)
118                         goto drop;
119         } while (!err);
120
121         /* Allocate new secpath or COW existing one. */
122
123         if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
124                 struct sec_path *sp;
125                 sp = secpath_dup(skb->sp);
126                 if (!sp)
127                         goto drop;
128                 if (skb->sp)
129                         secpath_put(skb->sp);
130                 skb->sp = sp;
131         }
132         if (xfrm_nr + skb->sp->len > XFRM_MAX_DEPTH)
133                 goto drop;
134
135         memcpy(skb->sp->xvec + skb->sp->len, xfrm_vec,
136                xfrm_nr * sizeof(xfrm_vec[0]));
137         skb->sp->len += xfrm_nr;
138
139         nf_reset(skb);
140
141         if (decaps) {
142                 if (!(skb->dev->flags&IFF_LOOPBACK)) {
143                         dst_release(skb->dst);
144                         skb->dst = NULL;
145                 }
146                 netif_rx(skb);
147                 return 0;
148         } else {
149 #ifdef CONFIG_NETFILTER
150                 __skb_push(skb, skb->data - skb_network_header(skb));
151                 ip_hdr(skb)->tot_len = htons(skb->len);
152                 ip_send_check(ip_hdr(skb));
153
154                 NF_HOOK(PF_INET, NF_IP_PRE_ROUTING, skb, skb->dev, NULL,
155                         xfrm4_rcv_encap_finish);
156                 return 0;
157 #else
158                 return -ip_hdr(skb)->protocol;
159 #endif
160         }
161
162 drop_unlock:
163         spin_unlock(&x->lock);
164         xfrm_state_put(x);
165 drop:
166         while (--xfrm_nr >= 0)
167                 xfrm_state_put(xfrm_vec[xfrm_nr]);
168
169         kfree_skb(skb);
170         return 0;
171 }