]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commitdiff
ipv4: Avoid reading user iov twice after raw_probe_proto_opt
authorHerbert Xu <herbert@gondor.apana.org.au>
Fri, 7 Nov 2014 13:27:09 +0000 (21:27 +0800)
committerDavid S. Miller <davem@davemloft.net>
Mon, 10 Nov 2014 19:25:35 +0000 (14:25 -0500)
Ever since raw_probe_proto_opt was added it had the problem of
causing the user iov to be read twice, once during the probe for
the protocol header and once again in ip_append_data.

This is a potential security problem since it means that whatever
we're probing may be invalid.  This patch plugs the hole by
firstly advancing the iov so we don't read the same spot again,
and secondly saving what we read the first time around for use
by ip_append_data.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/raw.c

index 9be905057b2420551048c2d17579e1d23af9539d..43385a9fa44196bd54be66416413c5c7eb03057a 100644 (file)
 #include <linux/netfilter.h>
 #include <linux/netfilter_ipv4.h>
 #include <linux/compat.h>
+#include <linux/uio.h>
+
+struct raw_frag_vec {
+       struct iovec *iov;
+       union {
+               struct icmphdr icmph;
+               char c[1];
+       } hdr;
+       int hlen;
+};
 
 static struct raw_hashinfo raw_v4_hashinfo = {
        .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
@@ -420,25 +430,57 @@ error:
        return err;
 }
 
-static int raw_probe_proto_opt(struct flowi4 *fl4, struct msghdr *msg)
+static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
 {
-       struct icmphdr icmph;
        int err;
 
        if (fl4->flowi4_proto != IPPROTO_ICMP)
                return 0;
 
        /* We only need the first two bytes. */
-       err = memcpy_fromiovecend((void *)&icmph, msg->msg_iov, 0, 2);
+       rfv->hlen = 2;
+
+       err = memcpy_fromiovec(rfv->hdr.c, rfv->iov, rfv->hlen);
        if (err)
                return err;
 
-       fl4->fl4_icmp_type = icmph.type;
-       fl4->fl4_icmp_code = icmph.code;
+       fl4->fl4_icmp_type = rfv->hdr.icmph.type;
+       fl4->fl4_icmp_code = rfv->hdr.icmph.code;
 
        return 0;
 }
 
+static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
+                      struct sk_buff *skb)
+{
+       struct raw_frag_vec *rfv = from;
+
+       if (offset < rfv->hlen) {
+               int copy = min(rfv->hlen - offset, len);
+
+               if (skb->ip_summed == CHECKSUM_PARTIAL)
+                       memcpy(to, rfv->hdr.c + offset, copy);
+               else
+                       skb->csum = csum_block_add(
+                               skb->csum,
+                               csum_partial_copy_nocheck(rfv->hdr.c + offset,
+                                                         to, copy, 0),
+                               odd);
+
+               odd = 0;
+               offset += copy;
+               to += copy;
+               len -= copy;
+
+               if (!len)
+                       return 0;
+       }
+
+       offset -= rfv->hlen;
+
+       return ip_generic_getfrag(rfv->iov, to, offset, len, odd, skb);
+}
+
 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
                       size_t len)
 {
@@ -452,6 +494,7 @@ static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
        u8  tos;
        int err;
        struct ip_options_data opt_copy;
+       struct raw_frag_vec rfv;
 
        err = -EMSGSIZE;
        if (len > 0xFFFF)
@@ -557,7 +600,10 @@ static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
                           daddr, saddr, 0, 0);
 
        if (!inet->hdrincl) {
-               err = raw_probe_proto_opt(&fl4, msg);
+               rfv.iov = msg->msg_iov;
+               rfv.hlen = 0;
+
+               err = raw_probe_proto_opt(&rfv, &fl4);
                if (err)
                        goto done;
        }
@@ -588,8 +634,8 @@ back_from_confirm:
                if (!ipc.addr)
                        ipc.addr = fl4.daddr;
                lock_sock(sk);
-               err = ip_append_data(sk, &fl4, ip_generic_getfrag,
-                                    msg->msg_iov, len, 0,
+               err = ip_append_data(sk, &fl4, raw_getfrag,
+                                    &rfv, len, 0,
                                     &ipc, &rt, msg->msg_flags);
                if (err)
                        ip_flush_pending_frames(sk);