]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
tun: wake up waitqueues after IFF_UP is set
authorFei Li <lifei.shirley@bytedance.com>
Mon, 17 Jun 2019 13:26:36 +0000 (21:26 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 3 Jul 2019 11:13:42 +0000 (13:13 +0200)
commit9590d1d1b033cb3c1211a90c66b928a141d6b129
treed290aedff3510ba4075a7be9e84ac634b3c13491
parenta54c0c1d392138f2012c1f644c434cbeed4cac77
tun: wake up waitqueues after IFF_UP is set

[ Upstream commit 72b319dc08b4924a29f5e2560ef6d966fa54c429 ]

Currently after setting tap0 link up, the tun code wakes tx/rx waited
queues up in tun_net_open() when .ndo_open() is called, however the
IFF_UP flag has not been set yet. If there's already a wait queue, it
would fail to transmit when checking the IFF_UP flag in tun_sendmsg().
Then the saving vhost_poll_start() will add the wq into wqh until it
is waken up again. Although this works when IFF_UP flag has been set
when tun_chr_poll detects; this is not true if IFF_UP flag has not
been set at that time. Sadly the latter case is a fatal error, as
the wq will never be waken up in future unless later manually
setting link up on purpose.

Fix this by moving the wakeup process into the NETDEV_UP event
notifying process, this makes sure IFF_UP has been set before all
waited queues been waken up.

Signed-off-by: Fei Li <lifei.shirley@bytedance.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/net/tun.c