KVM: Recalculate mmu pages needed for every memory region change
[linux-2.6] / net / xfrm / xfrm_input.c
1 /*
2  * xfrm_input.c
3  *
4  * Changes:
5  *      YOSHIFUJI Hideaki @USAGI
6  *              Split up af-specific portion
7  *
8  */
9
10 #include <linux/slab.h>
11 #include <linux/module.h>
12 #include <linux/netdevice.h>
13 #include <net/dst.h>
14 #include <net/ip.h>
15 #include <net/xfrm.h>
16
17 static struct kmem_cache *secpath_cachep __read_mostly;
18
19 void __secpath_destroy(struct sec_path *sp)
20 {
21         int i;
22         for (i = 0; i < sp->len; i++)
23                 xfrm_state_put(sp->xvec[i]);
24         kmem_cache_free(secpath_cachep, sp);
25 }
26 EXPORT_SYMBOL(__secpath_destroy);
27
28 struct sec_path *secpath_dup(struct sec_path *src)
29 {
30         struct sec_path *sp;
31
32         sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
33         if (!sp)
34                 return NULL;
35
36         sp->len = 0;
37         if (src) {
38                 int i;
39
40                 memcpy(sp, src, sizeof(*sp));
41                 for (i = 0; i < sp->len; i++)
42                         xfrm_state_hold(sp->xvec[i]);
43         }
44         atomic_set(&sp->refcnt, 1);
45         return sp;
46 }
47 EXPORT_SYMBOL(secpath_dup);
48
49 /* Fetch spi and seq from ipsec header */
50
51 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
52 {
53         int offset, offset_seq;
54         int hlen;
55
56         switch (nexthdr) {
57         case IPPROTO_AH:
58                 hlen = sizeof(struct ip_auth_hdr);
59                 offset = offsetof(struct ip_auth_hdr, spi);
60                 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
61                 break;
62         case IPPROTO_ESP:
63                 hlen = sizeof(struct ip_esp_hdr);
64                 offset = offsetof(struct ip_esp_hdr, spi);
65                 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
66                 break;
67         case IPPROTO_COMP:
68                 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
69                         return -EINVAL;
70                 *spi = htonl(ntohs(*(__be16*)(skb_transport_header(skb) + 2)));
71                 *seq = 0;
72                 return 0;
73         default:
74                 return 1;
75         }
76
77         if (!pskb_may_pull(skb, hlen))
78                 return -EINVAL;
79
80         *spi = *(__be32*)(skb_transport_header(skb) + offset);
81         *seq = *(__be32*)(skb_transport_header(skb) + offset_seq);
82         return 0;
83 }
84 EXPORT_SYMBOL(xfrm_parse_spi);
85
86 int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
87 {
88         int err;
89
90         err = x->outer_mode->afinfo->extract_input(x, skb);
91         if (err)
92                 return err;
93
94         skb->protocol = x->inner_mode->afinfo->eth_proto;
95         return x->inner_mode->input2(x, skb);
96 }
97 EXPORT_SYMBOL(xfrm_prepare_input);
98
99 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
100 {
101         int err;
102         __be32 seq;
103         struct xfrm_state *x;
104         xfrm_address_t *daddr;
105         unsigned int family;
106         int decaps = 0;
107         int async = 0;
108
109         /* A negative encap_type indicates async resumption. */
110         if (encap_type < 0) {
111                 async = 1;
112                 x = xfrm_input_state(skb);
113                 seq = XFRM_SKB_CB(skb)->seq;
114                 goto resume;
115         }
116
117         /* Allocate new secpath or COW existing one. */
118         if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
119                 struct sec_path *sp;
120
121                 sp = secpath_dup(skb->sp);
122                 if (!sp) {
123                         XFRM_INC_STATS(LINUX_MIB_XFRMINERROR);
124                         goto drop;
125                 }
126                 if (skb->sp)
127                         secpath_put(skb->sp);
128                 skb->sp = sp;
129         }
130
131         daddr = (xfrm_address_t *)(skb_network_header(skb) +
132                                    XFRM_SPI_SKB_CB(skb)->daddroff);
133         family = XFRM_SPI_SKB_CB(skb)->family;
134
135         seq = 0;
136         if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
137                 XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
138                 goto drop;
139         }
140
141         do {
142                 if (skb->sp->len == XFRM_MAX_DEPTH) {
143                         XFRM_INC_STATS(LINUX_MIB_XFRMINBUFFERERROR);
144                         goto drop;
145                 }
146
147                 x = xfrm_state_lookup(daddr, spi, nexthdr, family);
148                 if (x == NULL) {
149                         XFRM_INC_STATS(LINUX_MIB_XFRMINNOSTATES);
150                         xfrm_audit_state_notfound(skb, family, spi, seq);
151                         goto drop;
152                 }
153
154                 skb->sp->xvec[skb->sp->len++] = x;
155
156                 spin_lock(&x->lock);
157                 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
158                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEINVALID);
159                         goto drop_unlock;
160                 }
161
162                 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
163                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEINVALID);
164                         goto drop_unlock;
165                 }
166
167                 if (x->props.replay_window && xfrm_replay_check(x, skb, seq)) {
168                         XFRM_INC_STATS(LINUX_MIB_XFRMINSEQOUTOFWINDOW);
169                         goto drop_unlock;
170                 }
171
172                 if (xfrm_state_check_expire(x)) {
173                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEEXPIRED);
174                         goto drop_unlock;
175                 }
176
177                 spin_unlock(&x->lock);
178
179                 XFRM_SKB_CB(skb)->seq = seq;
180
181                 nexthdr = x->type->input(x, skb);
182
183                 if (nexthdr == -EINPROGRESS)
184                         return 0;
185
186 resume:
187                 spin_lock(&x->lock);
188                 if (nexthdr <= 0) {
189                         if (nexthdr == -EBADMSG) {
190                                 xfrm_audit_state_icvfail(x, skb,
191                                                          x->type->proto);
192                                 x->stats.integrity_failed++;
193                         }
194                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEPROTOERROR);
195                         goto drop_unlock;
196                 }
197
198                 /* only the first xfrm gets the encap type */
199                 encap_type = 0;
200
201                 if (x->props.replay_window)
202                         xfrm_replay_advance(x, seq);
203
204                 x->curlft.bytes += skb->len;
205                 x->curlft.packets++;
206
207                 spin_unlock(&x->lock);
208
209                 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
210
211                 if (x->inner_mode->input(x, skb)) {
212                         XFRM_INC_STATS(LINUX_MIB_XFRMINSTATEMODEERROR);
213                         goto drop;
214                 }
215
216                 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
217                         decaps = 1;
218                         break;
219                 }
220
221                 /*
222                  * We need the inner address.  However, we only get here for
223                  * transport mode so the outer address is identical.
224                  */
225                 daddr = &x->id.daddr;
226                 family = x->outer_mode->afinfo->family;
227
228                 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
229                 if (err < 0) {
230                         XFRM_INC_STATS(LINUX_MIB_XFRMINHDRERROR);
231                         goto drop;
232                 }
233         } while (!err);
234
235         nf_reset(skb);
236
237         if (decaps) {
238                 dst_release(skb->dst);
239                 skb->dst = NULL;
240                 netif_rx(skb);
241                 return 0;
242         } else {
243                 return x->inner_mode->afinfo->transport_finish(skb, async);
244         }
245
246 drop_unlock:
247         spin_unlock(&x->lock);
248 drop:
249         kfree_skb(skb);
250         return 0;
251 }
252 EXPORT_SYMBOL(xfrm_input);
253
254 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
255 {
256         return xfrm_input(skb, nexthdr, 0, -1);
257 }
258 EXPORT_SYMBOL(xfrm_input_resume);
259
260 void __init xfrm_input_init(void)
261 {
262         secpath_cachep = kmem_cache_create("secpath_cache",
263                                            sizeof(struct sec_path),
264                                            0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
265                                            NULL);
266 }