xen-netback: fix fragment detection in checksum setup
authorPaul Durrant <Paul.Durrant@citrix.com>
Tue, 3 Dec 2013 17:39:29 +0000 (17:39 +0000)
committerDavid S. Miller <davem@davemloft.net>
Fri, 6 Dec 2013 01:31:40 +0000 (20:31 -0500)
The code to detect fragments in checksum_setup() was missing for IPv4 and
too eager for IPv6. (It transpires that Windows seems to send IPv6 packets
with a fragment header even if they are not a fragment - i.e. offset is zero,
and M bit is not set).

This patch also incorporates a fix to callers of maybe_pull_tail() where
skb->network_header was being erroneously added to the length argument.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: David Vrabel <david.vrabel@citrix.com>
cc: David Miller <davem@davemloft.net>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/xen-netback/netback.c
include/linux/ipv6.h
include/net/ipv6.h

index 64f0e0d18b8188c1729f525e26a2ca8b88724d8f..acf13920e6d1804c94c672560f45154545ec9c54 100644 (file)
@@ -1149,49 +1149,72 @@ static int xenvif_set_skb_gso(struct xenvif *vif,
        return 0;
 }
 
-static inline void maybe_pull_tail(struct sk_buff *skb, unsigned int len)
+static inline int maybe_pull_tail(struct sk_buff *skb, unsigned int len,
+                                 unsigned int max)
 {
-       if (skb_is_nonlinear(skb) && skb_headlen(skb) < len) {
-               /* If we need to pullup then pullup to the max, so we
-                * won't need to do it again.
-                */
-               int target = min_t(int, skb->len, MAX_TCP_HEADER);
-               __pskb_pull_tail(skb, target - skb_headlen(skb));
-       }
+       if (skb_headlen(skb) >= len)
+               return 0;
+
+       /* If we need to pullup then pullup to the max, so we
+        * won't need to do it again.
+        */
+       if (max > skb->len)
+               max = skb->len;
+
+       if (__pskb_pull_tail(skb, max - skb_headlen(skb)) == NULL)
+               return -ENOMEM;
+
+       if (skb_headlen(skb) < len)
+               return -EPROTO;
+
+       return 0;
 }
 
+/* This value should be large enough to cover a tagged ethernet header plus
+ * maximally sized IP and TCP or UDP headers.
+ */
+#define MAX_IP_HDR_LEN 128
+
 static int checksum_setup_ip(struct xenvif *vif, struct sk_buff *skb,
                             int recalculate_partial_csum)
 {
-       struct iphdr *iph = (void *)skb->data;
-       unsigned int header_size;
        unsigned int off;
-       int err = -EPROTO;
+       bool fragment;
+       int err;
+
+       fragment = false;
+
+       err = maybe_pull_tail(skb,
+                             sizeof(struct iphdr),
+                             MAX_IP_HDR_LEN);
+       if (err < 0)
+               goto out;
 
-       off = sizeof(struct iphdr);
+       if (ip_hdr(skb)->frag_off & htons(IP_OFFSET | IP_MF))
+               fragment = true;
 
-       header_size = skb->network_header + off + MAX_IPOPTLEN;
-       maybe_pull_tail(skb, header_size);
+       off = ip_hdrlen(skb);
 
-       off = iph->ihl * 4;
+       err = -EPROTO;
 
-       switch (iph->protocol) {
+       switch (ip_hdr(skb)->protocol) {
        case IPPROTO_TCP:
                if (!skb_partial_csum_set(skb, off,
                                          offsetof(struct tcphdr, check)))
                        goto out;
 
                if (recalculate_partial_csum) {
-                       struct tcphdr *tcph = tcp_hdr(skb);
-
-                       header_size = skb->network_header +
-                               off +
-                               sizeof(struct tcphdr);
-                       maybe_pull_tail(skb, header_size);
-
-                       tcph->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
-                                                        skb->len - off,
-                                                        IPPROTO_TCP, 0);
+                       err = maybe_pull_tail(skb,
+                                             off + sizeof(struct tcphdr),
+                                             MAX_IP_HDR_LEN);
+                       if (err < 0)
+                               goto out;
+
+                       tcp_hdr(skb)->check =
+                               ~csum_tcpudp_magic(ip_hdr(skb)->saddr,
+                                                  ip_hdr(skb)->daddr,
+                                                  skb->len - off,
+                                                  IPPROTO_TCP, 0);
                }
                break;
        case IPPROTO_UDP:
@@ -1200,24 +1223,20 @@ static int checksum_setup_ip(struct xenvif *vif, struct sk_buff *skb,
                        goto out;
 
                if (recalculate_partial_csum) {
-                       struct udphdr *udph = udp_hdr(skb);
-
-                       header_size = skb->network_header +
-                               off +
-                               sizeof(struct udphdr);
-                       maybe_pull_tail(skb, header_size);
-
-                       udph->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
-                                                        skb->len - off,
-                                                        IPPROTO_UDP, 0);
+                       err = maybe_pull_tail(skb,
+                                             off + sizeof(struct udphdr),
+                                             MAX_IP_HDR_LEN);
+                       if (err < 0)
+                               goto out;
+
+                       udp_hdr(skb)->check =
+                               ~csum_tcpudp_magic(ip_hdr(skb)->saddr,
+                                                  ip_hdr(skb)->daddr,
+                                                  skb->len - off,
+                                                  IPPROTO_UDP, 0);
                }
                break;
        default:
-               if (net_ratelimit())
-                       netdev_err(vif->dev,
-                                  "Attempting to checksum a non-TCP/UDP packet, "
-                                  "dropping a protocol %d packet\n",
-                                  iph->protocol);
                goto out;
        }
 
@@ -1227,75 +1246,99 @@ out:
        return err;
 }
 
+/* This value should be large enough to cover a tagged ethernet header plus
+ * an IPv6 header, all options, and a maximal TCP or UDP header.
+ */
+#define MAX_IPV6_HDR_LEN 256
+
+#define OPT_HDR(type, skb, off) \
+       (type *)(skb_network_header(skb) + (off))
+
 static int checksum_setup_ipv6(struct xenvif *vif, struct sk_buff *skb,
                               int recalculate_partial_csum)
 {
-       int err = -EPROTO;
-       struct ipv6hdr *ipv6h = (void *)skb->data;
+       int err;
        u8 nexthdr;
-       unsigned int header_size;
        unsigned int off;
+       unsigned int len;
        bool fragment;
        bool done;
 
+       fragment = false;
        done = false;
 
        off = sizeof(struct ipv6hdr);
 
-       header_size = skb->network_header + off;
-       maybe_pull_tail(skb, header_size);
+       err = maybe_pull_tail(skb, off, MAX_IPV6_HDR_LEN);
+       if (err < 0)
+               goto out;
 
-       nexthdr = ipv6h->nexthdr;
+       nexthdr = ipv6_hdr(skb)->nexthdr;
 
-       while ((off <= sizeof(struct ipv6hdr) + ntohs(ipv6h->payload_len)) &&
-              !done) {
+       len = sizeof(struct ipv6hdr) + ntohs(ipv6_hdr(skb)->payload_len);
+       while (off <= len && !done) {
                switch (nexthdr) {
                case IPPROTO_DSTOPTS:
                case IPPROTO_HOPOPTS:
                case IPPROTO_ROUTING: {
-                       struct ipv6_opt_hdr *hp = (void *)(skb->data + off);
+                       struct ipv6_opt_hdr *hp;
 
-                       header_size = skb->network_header +
-                               off +
-                               sizeof(struct ipv6_opt_hdr);
-                       maybe_pull_tail(skb, header_size);
+                       err = maybe_pull_tail(skb,
+                                             off +
+                                             sizeof(struct ipv6_opt_hdr),
+                                             MAX_IPV6_HDR_LEN);
+                       if (err < 0)
+                               goto out;
 
+                       hp = OPT_HDR(struct ipv6_opt_hdr, skb, off);
                        nexthdr = hp->nexthdr;
                        off += ipv6_optlen(hp);
                        break;
                }
                case IPPROTO_AH: {
-                       struct ip_auth_hdr *hp = (void *)(skb->data + off);
+                       struct ip_auth_hdr *hp;
+
+                       err = maybe_pull_tail(skb,
+                                             off +
+                                             sizeof(struct ip_auth_hdr),
+                                             MAX_IPV6_HDR_LEN);
+                       if (err < 0)
+                               goto out;
+
+                       hp = OPT_HDR(struct ip_auth_hdr, skb, off);
+                       nexthdr = hp->nexthdr;
+                       off += ipv6_authlen(hp);
+                       break;
+               }
+               case IPPROTO_FRAGMENT: {
+                       struct frag_hdr *hp;
 
-                       header_size = skb->network_header +
-                               off +
-                               sizeof(struct ip_auth_hdr);
-                       maybe_pull_tail(skb, header_size);
+                       err = maybe_pull_tail(skb,
+                                             off +
+                                             sizeof(struct frag_hdr),
+                                             MAX_IPV6_HDR_LEN);
+                       if (err < 0)
+                               goto out;
+
+                       hp = OPT_HDR(struct frag_hdr, skb, off);
+
+                       if (hp->frag_off & htons(IP6_OFFSET | IP6_MF))
+                               fragment = true;
 
                        nexthdr = hp->nexthdr;
-                       off += (hp->hdrlen+2)<<2;
+                       off += sizeof(struct frag_hdr);
                        break;
                }
-               case IPPROTO_FRAGMENT:
-                       fragment = true;
-                       /* fall through */
                default:
                        done = true;
                        break;
                }
        }
 
-       if (!done) {
-               if (net_ratelimit())
-                       netdev_err(vif->dev, "Failed to parse packet header\n");
-               goto out;
-       }
+       err = -EPROTO;
 
-       if (fragment) {
-               if (net_ratelimit())
-                       netdev_err(vif->dev, "Packet is a fragment!\n");
+       if (!done || fragment)
                goto out;
-       }
 
        switch (nexthdr) {
        case IPPROTO_TCP:
@@ -1304,17 +1347,17 @@ static int checksum_setup_ipv6(struct xenvif *vif, struct sk_buff *skb,
                        goto out;
 
                if (recalculate_partial_csum) {
-                       struct tcphdr *tcph = tcp_hdr(skb);
-
-                       header_size = skb->network_header +
-                               off +
-                               sizeof(struct tcphdr);
-                       maybe_pull_tail(skb, header_size);
-
-                       tcph->check = ~csum_ipv6_magic(&ipv6h->saddr,
-                                                      &ipv6h->daddr,
-                                                      skb->len - off,
-                                                      IPPROTO_TCP, 0);
+                       err = maybe_pull_tail(skb,
+                                             off + sizeof(struct tcphdr),
+                                             MAX_IPV6_HDR_LEN);
+                       if (err < 0)
+                               goto out;
+
+                       tcp_hdr(skb)->check =
+                               ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
+                                                &ipv6_hdr(skb)->daddr,
+                                                skb->len - off,
+                                                IPPROTO_TCP, 0);
                }
                break;
        case IPPROTO_UDP:
@@ -1323,25 +1366,20 @@ static int checksum_setup_ipv6(struct xenvif *vif, struct sk_buff *skb,
                        goto out;
 
                if (recalculate_partial_csum) {
-                       struct udphdr *udph = udp_hdr(skb);
-
-                       header_size = skb->network_header +
-                               off +
-                               sizeof(struct udphdr);
-                       maybe_pull_tail(skb, header_size);
-
-                       udph->check = ~csum_ipv6_magic(&ipv6h->saddr,
-                                                      &ipv6h->daddr,
-                                                      skb->len - off,
-                                                      IPPROTO_UDP, 0);
+                       err = maybe_pull_tail(skb,
+                                             off + sizeof(struct udphdr),
+                                             MAX_IPV6_HDR_LEN);
+                       if (err < 0)
+                               goto out;
+
+                       udp_hdr(skb)->check =
+                               ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
+                                                &ipv6_hdr(skb)->daddr,
+                                                skb->len - off,
+                                                IPPROTO_UDP, 0);
                }
                break;
        default:
-               if (net_ratelimit())
-                       netdev_err(vif->dev,
-                                  "Attempting to checksum a non-TCP/UDP packet, "
-                                  "dropping a protocol %d packet\n",
-                                  nexthdr);
                goto out;
        }
 
index 5d89d1b808a6cc896f22332c477126c3d2f2b563..c56c350324e4b4670ef8d31397a1a777b9c08578 100644 (file)
@@ -4,6 +4,7 @@
 #include <uapi/linux/ipv6.h>
 
 #define ipv6_optlen(p)  (((p)->hdrlen+1) << 3)
+#define ipv6_authlen(p) (((p)->hdrlen+2) << 2)
 /*
  * This structure contains configuration options per IPv6 link.
  */
index eb198acaac1d5f42b53ac97838511563fe7e874b..488316e339a124fa55fdeac686a4fc6d28c0dd2c 100644 (file)
@@ -110,7 +110,8 @@ struct frag_hdr {
        __be32  identification;
 };
 
-#define        IP6_MF  0x0001
+#define        IP6_MF          0x0001
+#define        IP6_OFFSET      0xFFF8
 
 #include <net/sock.h>