Hi Andrei,
* Emeltchenko Andrei <Andrei.Emeltchenko.news@xxxxxxxxx> [2011-01-20 10:38:07 +0200]:
> From: Andrei Emeltchenko <andrei.emeltchenko@xxxxxxxxx>
>
> Crash can happen when tasklet handling connect/disconnect requests
> preempts socket accept. Can be reproduced with "l2test -r" on one
> side and several "l2test -c -b 1000 -i hci0 -P 10 <bdaddr>" on the
> other side.
>
> disable taskets in socket accept and change lock_sock and release_sock
> to bh_lock_sock and bh_unlock_sock since we have to use spinlocks and
> there is no need to mark sock as owned by user.
>
> ...
> [ 3555.897247] Unable to handle kernel NULL pointer dereference at virtual
> address 000000bc
> [ 3555.915039] pgd = cab9c000
> [ 3555.917785] [000000bc] *pgd=8bf3d031, *pte=00000000, *ppte=00000000
> [ 3555.928314] Internal error: Oops: 17 [#1] PREEMPT
> [ 3555.999786] CPU: 0 Not tainted (2.6.32.21-13874-g67918ef #65)
> ...
> [ 3556.005981] PC is at bt_accept_unlink+0x20/0x58 [bluetooth]
> [ 3556.011627] LR is at bt_accept_dequeue+0x3c/0xe8 [bluetooth]
> ...
> [ 3556.161285] [<bf0007fc>] (bt_accept_unlink+0x20/0x58 [bluetooth]) from
> [<bf000870>] (bt_accept_dequeue+0x3c/0xe8 [bluetooth])
> [ 3556.172729] [<bf000870>] (bt_accept_dequeue+0x3c/0xe8 [bluetooth]) from
> [<bf324df8>] (l2cap_sock_accept+0x100/0x15c [l2cap])
> [ 3556.184082] [<bf324df8>] (l2cap_sock_accept+0x100/0x15c [l2cap]) from
> [<c026a0a8>] (sys_accept4+0x120/0x1e0)
> [ 3556.193969] [<c026a0a8>] (sys_accept4+0x120/0x1e0) from [<c002c9a0>]
> (ret_fast_syscall+0x0/0x2c)
> [ 3556.202819] Code: e5813000 e5901164 e580c160 e580c15c (e1d13bbc)
> ...
>
> Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@xxxxxxxxx>
> ---
> net/bluetooth/af_bluetooth.c | 12 ++++++++----
> 1 files changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/net/bluetooth/af_bluetooth.c b/net/bluetooth/af_bluetooth.c
> index a6732b5..f545566 100644
> --- a/net/bluetooth/af_bluetooth.c
> +++ b/net/bluetooth/af_bluetooth.c
> @@ -199,14 +199,15 @@ struct sock *bt_accept_dequeue(struct sock *parent, struct socket *newsock)
>
> BT_DBG("parent %p", parent);
>
> + local_bh_disable();
> list_for_each_safe(p, n, &bt_sk(parent)->accept_q) {
> sk = (struct sock *) list_entry(p, struct bt_sock, accept_q);
>
> - lock_sock(sk);
> + bh_lock_sock(sk);
>
> /* FIXME: Is this check still needed */
> if (sk->sk_state == BT_CLOSED) {
> - release_sock(sk);
> + bh_unlock_sock(sk);
> bt_accept_unlink(sk);
> continue;
> }
> @@ -216,12 +217,15 @@ struct sock *bt_accept_dequeue(struct sock *parent, struct socket *newsock)
> bt_accept_unlink(sk);
> if (newsock)
> sock_graft(sk, newsock);
> - release_sock(sk);
Please add an extra line here, them patch should be ok to upstream.
> + bh_unlock_sock(sk);
> + local_bh_enable();
> return sk;
--
Gustavo F. Padovan
http://profusion.mobi
--
To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
