Return-Path: Date: Mon, 4 Jun 2012 10:14:19 -0700 (PDT) From: Mat Martineau To: Gustavo Padovan cc: linux-bluetooth@vger.kernel.org Subject: Re: [PATCH 2/2] Bluetooth: Release all locks before sleep In-Reply-To: <1338492344-1474-2-git-send-email-gustavo@padovan.org> Message-ID: References: <1338492344-1474-1-git-send-email-gustavo@padovan.org> <1338492344-1474-2-git-send-email-gustavo@padovan.org> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; format=flowed; charset=US-ASCII Sender: linux-bluetooth-owner@vger.kernel.org List-ID: Hi Gustavo - On Thu, 31 May 2012, Gustavo Padovan wrote: > From: Gustavo Padovan > > To avoid deadlock we need to release locks while waiting for a ack to > arrive. > > Signed-off-by: Gustavo Padovan > --- > net/bluetooth/l2cap_core.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/net/bluetooth/l2cap_core.c b/net/bluetooth/l2cap_core.c > index 1cb3ca0..94273dc 100644 > --- a/net/bluetooth/l2cap_core.c > +++ b/net/bluetooth/l2cap_core.c > @@ -1553,13 +1553,14 @@ done: > int __l2cap_wait_ack(struct sock *sk) > { > struct l2cap_chan *chan = l2cap_pi(sk)->chan; > + struct l2cap_conn *conn = chan->conn; I don't think you want to store the l2cap_conn pointer on the stack, the structure can be freed when the locks are released. > DECLARE_WAITQUEUE(wait, current); > int err = 0; > int timeo = HZ/5; > > add_wait_queue(sk_sleep(sk), &wait); > set_current_state(TASK_INTERRUPTIBLE); > - while (chan->unacked_frames > 0 && chan->conn) { > + while (chan->unacked_frames > 0 && conn) { > if (!timeo) > timeo = HZ/5; > > @@ -1570,7 +1571,9 @@ int __l2cap_wait_ack(struct sock *sk) > > release_sock(sk); > l2cap_chan_unlock(chan); > + mutex_unlock(&conn->chan_lock); > timeo = schedule_timeout(timeo); > + mutex_lock(&conn->chan_lock); > l2cap_chan_lock(chan); > lock_sock(sk); > set_current_state(TASK_INTERRUPTIBLE); > -- > 1.7.10.2 I think it would be better to only acquire the chan_lock mutex when calling l2cap_chan_close in l2cap_sock_shutdown. However, you will have to be careful to avoid deadlocks. I think it's only safe to acquire the conn->chan_lock when both the l2cap_chan and the socket are unlocked. Can you collapse all the locking changes in to one patch? Since __l2cap_wait_ack is only called from l2cap_sock.c, would it make sense to move the function to l2cap_sock.c? -- Mat Martineau Employee of Qualcomm Innovation Center, Inc. Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum