From: Takashi Iwai tiwai@suse.de
commit d21b96c8ed2aea7e6b7bf4735e1d2503cfbf4072 upstream.
The code change for switching to non-atomic mode brought the unexpected mutex deadlock in get_msg(). It converted the spinlock with the existing mutex, but there were calls with the already holding the mutex. Since the only place that needs the extra lock is the code path from snd_mixart_send_msg(), remove the mutex lock in get_msg() and apply in the caller side for fixing the mutex deadlock.
Fixes: 8d3a8b5cb57d ("ALSA: mixart: Use nonatomic PCM ops") Reported-by: Dan Carpenter dan.carpenter@oracle.com Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20201119121440.18945-1-tiwai@suse.de Signed-off-by: Takashi Iwai tiwai@suse.de Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
--- sound/pci/mixart/mixart_core.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
--- a/sound/pci/mixart/mixart_core.c +++ b/sound/pci/mixart/mixart_core.c @@ -83,7 +83,6 @@ static int get_msg(struct mixart_mgr *mg unsigned int i; #endif
- mutex_lock(&mgr->msg_lock); err = 0;
/* copy message descriptor from miXart to driver */ @@ -132,8 +131,6 @@ static int get_msg(struct mixart_mgr *mg writel_be(headptr, MIXART_MEM(mgr, MSG_OUTBOUND_FREE_HEAD));
_clean_exit: - mutex_unlock(&mgr->msg_lock); - return err; }
@@ -271,7 +268,9 @@ int snd_mixart_send_msg(struct mixart_mg resp.data = resp_data; resp.size = max_resp_size;
+ mutex_lock(&mgr->msg_lock); err = get_msg(mgr, &resp, msg_frame); + mutex_unlock(&mgr->msg_lock);
if( request->message_id != resp.message_id ) dev_err(&mgr->pci->dev, "RESPONSE ERROR!\n");