[NOTE: the regression below is found only in 3.2-3.4 stable trees, so
there is no upstream commit corresponding to this patch]
The recent fix for the race at disconnection of usb-audio devices
(upstream commit 978520b7) triggers Oops when a device is unplugged
while playing on 3.2 and 3.4 kernels. The culprit is that the
shutdown flag check was wrongly added around the urb deactivation code
snippet. The urb deactivation code has to be performed even after the
device disconnected. Otherwise it remains undead and pokes the wild
access in the end.
The regression fix is simply reverting the shutdown flag check in that
code.
Reported-and-tested-by: Chris J Arges <[email protected]>
Cc: <[email protected]> [v3.2..v3.4]
Signed-off-by: Takashi Iwai <[email protected]>
---
Looking at linux-stable git tree, 3.2.y and 3.4.y are affected at
least. Ben, Greg, please apply this.
I'm not sure whether other trees (3.3.y, 3.5.y) already included the
commit in question (upstream commit 978520b7). Even the tree contains
its backport, it's not necessarily broken, e.g. 3.7.y doesn't need
this patch. This patch is only for the broken backport to older
kernels.
Thanks!
Takashi
sound/usb/endpoint.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/sound/usb/endpoint.c b/sound/usb/endpoint.c
index 24c5114..9ab2b3e 100644
--- a/sound/usb/endpoint.c
+++ b/sound/usb/endpoint.c
@@ -148,10 +148,8 @@ void snd_usb_release_substream_urbs(struct
snd_usb_substream *subs, int force)
int i;
/* stop urbs (to be sure) */
- if (!subs->stream->chip->shutdown) {
- deactivate_urbs(subs, force, 1);
- wait_clear_urbs(subs);
- }
+ deactivate_urbs(subs, force, 1);
+ wait_clear_urbs(subs);
for (i = 0; i < MAX_URBS; i++)
release_urb_ctx(&subs->dataurb[i]);
--
1.8.1.1
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
