tuntap: correctly linearize skb when zerocopy is used
authorJason Wang <jasowang@redhat.com>
Wed, 10 Jul 2013 05:43:27 +0000 (13:43 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 28 Jul 2013 23:29:58 +0000 (16:29 -0700)
commitd09ec76aebab878e0dc583484152e20a2172a7b4
treeadb0981a16f2b44be0d7b1fcd02039adcbbf42cd
parent6683151aeea6c7aade12dde7ec2008fc380f0728
tuntap: correctly linearize skb when zerocopy is used

[ Upstream commit 3dd5c3308e8b671e8e8882ba972f51cefbe9fd0d ]

Userspace may produce vectors greater than MAX_SKB_FRAGS. When we try to
linearize parts of the skb to let the rest of iov to be fit in
the frags, we need count copylen into linear when calling tun_alloc_skb()
instead of partly counting it into data_len. Since this breaks
zerocopy_sg_from_iovec() since its inner counter assumes nr_frags should
be zero at beginning. This cause nr_frags to be increased wrongly without
setting the correct frags.

This bug were introduced from 0690899b4d4501b3505be069b9a687e68ccbe15b
(tun: experimental zero copy tx support)

Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/net/tun.c