net: skb->dst accessors
[safe/jmp/linux-2.6] / net / xfrm / xfrm_input.c
index c58a6f0..e0009c1 100644 (file)
@@ -4,21 +4,23 @@
  * Changes:
  *     YOSHIFUJI Hideaki @USAGI
  *             Split up af-specific portion
- *     
+ *
  */
 
 #include <linux/slab.h>
 #include <linux/module.h>
+#include <linux/netdevice.h>
+#include <net/dst.h>
 #include <net/ip.h>
 #include <net/xfrm.h>
 
-static kmem_cache_t *secpath_cachep;
+static struct kmem_cache *secpath_cachep __read_mostly;
 
 void __secpath_destroy(struct sec_path *sp)
 {
        int i;
        for (i = 0; i < sp->len; i++)
-               xfrm_state_put(sp->x[i].xvec);
+               xfrm_state_put(sp->xvec[i]);
        kmem_cache_free(secpath_cachep, sp);
 }
 EXPORT_SYMBOL(__secpath_destroy);
@@ -27,7 +29,7 @@ struct sec_path *secpath_dup(struct sec_path *src)
 {
        struct sec_path *sp;
 
-       sp = kmem_cache_alloc(secpath_cachep, SLAB_ATOMIC);
+       sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
        if (!sp)
                return NULL;
 
@@ -37,7 +39,7 @@ struct sec_path *secpath_dup(struct sec_path *src)
 
                memcpy(sp, src, sizeof(*sp));
                for (i = 0; i < sp->len; i++)
-                       xfrm_state_hold(sp->x[i].xvec);
+                       xfrm_state_hold(sp->xvec[i]);
        }
        atomic_set(&sp->refcnt, 1);
        return sp;
@@ -46,44 +48,234 @@ EXPORT_SYMBOL(secpath_dup);
 
 /* Fetch spi and seq from ipsec header */
 
-int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, u32 *spi, u32 *seq)
+int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
 {
        int offset, offset_seq;
+       int hlen;
 
        switch (nexthdr) {
        case IPPROTO_AH:
+               hlen = sizeof(struct ip_auth_hdr);
                offset = offsetof(struct ip_auth_hdr, spi);
                offset_seq = offsetof(struct ip_auth_hdr, seq_no);
                break;
        case IPPROTO_ESP:
+               hlen = sizeof(struct ip_esp_hdr);
                offset = offsetof(struct ip_esp_hdr, spi);
                offset_seq = offsetof(struct ip_esp_hdr, seq_no);
                break;
        case IPPROTO_COMP:
                if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
                        return -EINVAL;
-               *spi = ntohl(ntohs(*(u16*)(skb->h.raw + 2)));
+               *spi = htonl(ntohs(*(__be16*)(skb_transport_header(skb) + 2)));
                *seq = 0;
                return 0;
        default:
                return 1;
        }
 
-       if (!pskb_may_pull(skb, 16))
+       if (!pskb_may_pull(skb, hlen))
                return -EINVAL;
 
-       *spi = *(u32*)(skb->h.raw + offset);
-       *seq = *(u32*)(skb->h.raw + offset_seq);
+       *spi = *(__be32*)(skb_transport_header(skb) + offset);
+       *seq = *(__be32*)(skb_transport_header(skb) + offset_seq);
        return 0;
 }
-EXPORT_SYMBOL(xfrm_parse_spi);
+
+int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
+{
+       struct xfrm_mode *inner_mode = x->inner_mode;
+       int err;
+
+       err = x->outer_mode->afinfo->extract_input(x, skb);
+       if (err)
+               return err;
+
+       if (x->sel.family == AF_UNSPEC) {
+               inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
+               if (inner_mode == NULL)
+                       return -EAFNOSUPPORT;
+       }
+
+       skb->protocol = inner_mode->afinfo->eth_proto;
+       return inner_mode->input2(x, skb);
+}
+EXPORT_SYMBOL(xfrm_prepare_input);
+
+int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
+{
+       struct net *net = dev_net(skb->dev);
+       int err;
+       __be32 seq;
+       struct xfrm_state *x;
+       xfrm_address_t *daddr;
+       struct xfrm_mode *inner_mode;
+       unsigned int family;
+       int decaps = 0;
+       int async = 0;
+
+       /* A negative encap_type indicates async resumption. */
+       if (encap_type < 0) {
+               async = 1;
+               x = xfrm_input_state(skb);
+               seq = XFRM_SKB_CB(skb)->seq.input;
+               goto resume;
+       }
+
+       /* Allocate new secpath or COW existing one. */
+       if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
+               struct sec_path *sp;
+
+               sp = secpath_dup(skb->sp);
+               if (!sp) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
+                       goto drop;
+               }
+               if (skb->sp)
+                       secpath_put(skb->sp);
+               skb->sp = sp;
+       }
+
+       daddr = (xfrm_address_t *)(skb_network_header(skb) +
+                                  XFRM_SPI_SKB_CB(skb)->daddroff);
+       family = XFRM_SPI_SKB_CB(skb)->family;
+
+       seq = 0;
+       if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
+               XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
+               goto drop;
+       }
+
+       do {
+               if (skb->sp->len == XFRM_MAX_DEPTH) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
+                       goto drop;
+               }
+
+               x = xfrm_state_lookup(net, daddr, spi, nexthdr, family);
+               if (x == NULL) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
+                       xfrm_audit_state_notfound(skb, family, spi, seq);
+                       goto drop;
+               }
+
+               skb->sp->xvec[skb->sp->len++] = x;
+
+               spin_lock(&x->lock);
+               if (unlikely(x->km.state != XFRM_STATE_VALID)) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEINVALID);
+                       goto drop_unlock;
+               }
+
+               if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
+                       goto drop_unlock;
+               }
+
+               if (x->props.replay_window && xfrm_replay_check(x, skb, seq)) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
+                       goto drop_unlock;
+               }
+
+               if (xfrm_state_check_expire(x)) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
+                       goto drop_unlock;
+               }
+
+               spin_unlock(&x->lock);
+
+               XFRM_SKB_CB(skb)->seq.input = seq;
+
+               nexthdr = x->type->input(x, skb);
+
+               if (nexthdr == -EINPROGRESS)
+                       return 0;
+
+resume:
+               spin_lock(&x->lock);
+               if (nexthdr <= 0) {
+                       if (nexthdr == -EBADMSG) {
+                               xfrm_audit_state_icvfail(x, skb,
+                                                        x->type->proto);
+                               x->stats.integrity_failed++;
+                       }
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
+                       goto drop_unlock;
+               }
+
+               /* only the first xfrm gets the encap type */
+               encap_type = 0;
+
+               if (x->props.replay_window)
+                       xfrm_replay_advance(x, seq);
+
+               x->curlft.bytes += skb->len;
+               x->curlft.packets++;
+
+               spin_unlock(&x->lock);
+
+               XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
+
+               inner_mode = x->inner_mode;
+
+               if (x->sel.family == AF_UNSPEC) {
+                       inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
+                       if (inner_mode == NULL)
+                               goto drop;
+               }
+
+               if (inner_mode->input(x, skb)) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
+                       goto drop;
+               }
+
+               if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
+                       decaps = 1;
+                       break;
+               }
+
+               /*
+                * We need the inner address.  However, we only get here for
+                * transport mode so the outer address is identical.
+                */
+               daddr = &x->id.daddr;
+               family = x->outer_mode->afinfo->family;
+
+               err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
+               if (err < 0) {
+                       XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
+                       goto drop;
+               }
+       } while (!err);
+
+       nf_reset(skb);
+
+       if (decaps) {
+               skb_dst_drop(skb);
+               netif_rx(skb);
+               return 0;
+       } else {
+               return x->inner_mode->afinfo->transport_finish(skb, async);
+       }
+
+drop_unlock:
+       spin_unlock(&x->lock);
+drop:
+       kfree_skb(skb);
+       return 0;
+}
+EXPORT_SYMBOL(xfrm_input);
+
+int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
+{
+       return xfrm_input(skb, nexthdr, 0, -1);
+}
+EXPORT_SYMBOL(xfrm_input_resume);
 
 void __init xfrm_input_init(void)
 {
        secpath_cachep = kmem_cache_create("secpath_cache",
                                           sizeof(struct sec_path),
-                                          0, SLAB_HWCACHE_ALIGN,
-                                          NULL, NULL);
-       if (!secpath_cachep)
-               panic("XFRM: failed to allocate secpath_cache\n");
+                                          0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
+                                          NULL);
 }