* [PATCH] ALSA: hda - Fix hang caused by race during suspend.
@ 2012-09-28 22:57 Dylan Reid
2012-10-08 8:17 ` Takashi Iwai
0 siblings, 1 reply; 2+ messages in thread
From: Dylan Reid @ 2012-09-28 22:57 UTC (permalink / raw)
To: alsa-devel; +Cc: tiwai, Dylan Reid
There was a race condition when the system suspends while hda_power_work
is running in the work queue. If system suspend (snd_hda_suspend)
happens after the work queue releases power_lock but before it calls
hda_call_codec_suspend, codec_suspend runs with power_on=0, causing the
codec to power up for register reads, and hanging when it calls
cancel_delayed_work_sync from the running work queue.
The call chain from the work queue will look like this:
hda_power_work <<- power_on = 1, unlock, then power_on cleard by suspend
hda_call_codec_suspend
hda_set_power_state
snd_hda_codec_read
codec_exec_verb
snd_hda_power_up
snd_hda_power_save
__snd_hda_power_up
cancel_delayed_work_sync <<-- cancelling executing wq
Fix this by waiting for the work queue to finish before starting suspend
if suspend is not happening on the work queue.
Signed-off-by: Dylan Reid <dgreid@chromium.org>
---
sound/pci/hda/hda_codec.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/sound/pci/hda/hda_codec.c b/sound/pci/hda/hda_codec.c
index 960800b..c0ab72c 100644
--- a/sound/pci/hda/hda_codec.c
+++ b/sound/pci/hda/hda_codec.c
@@ -3612,7 +3612,7 @@ static inline void hda_exec_init_verbs(struct hda_codec *codec) {}
* call suspend and power-down; used both from PM and power-save
* this function returns the power state in the end
*/
-static unsigned int hda_call_codec_suspend(struct hda_codec *codec)
+static unsigned int hda_call_codec_suspend(struct hda_codec *codec, bool in_wq)
{
unsigned int state;
@@ -3620,7 +3620,9 @@ static unsigned int hda_call_codec_suspend(struct hda_codec *codec)
codec->patch_ops.suspend(codec);
hda_cleanup_all_streams(codec);
state = hda_set_power_state(codec, AC_PWRST_D3);
- cancel_delayed_work(&codec->power_work);
+ /* Cancel delayed work if we aren't currently running from it. */
+ if (!in_wq)
+ cancel_delayed_work_sync(&codec->power_work);
spin_lock(&codec->power_lock);
snd_hda_update_power_acct(codec);
trace_hda_power_down(codec);
@@ -4478,7 +4480,7 @@ static void hda_power_work(struct work_struct *work)
}
spin_unlock(&codec->power_lock);
- state = hda_call_codec_suspend(codec);
+ state = hda_call_codec_suspend(codec, true);
codec->pm_down_notified = 0;
if (!bus->power_keep_link_on && (state & AC_PWRST_CLK_STOP_OK)) {
codec->pm_down_notified = 1;
@@ -5127,7 +5129,7 @@ int snd_hda_suspend(struct hda_bus *bus)
list_for_each_entry(codec, &bus->codec_list, list) {
if (hda_codec_is_power_on(codec))
- hda_call_codec_suspend(codec);
+ hda_call_codec_suspend(codec, false);
}
return 0;
}
--
1.7.12.146.g16d26b1
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH] ALSA: hda - Fix hang caused by race during suspend.
2012-09-28 22:57 [PATCH] ALSA: hda - Fix hang caused by race during suspend Dylan Reid
@ 2012-10-08 8:17 ` Takashi Iwai
0 siblings, 0 replies; 2+ messages in thread
From: Takashi Iwai @ 2012-10-08 8:17 UTC (permalink / raw)
To: Dylan Reid; +Cc: alsa-devel
At Fri, 28 Sep 2012 15:57:01 -0700,
Dylan Reid wrote:
>
> There was a race condition when the system suspends while hda_power_work
> is running in the work queue. If system suspend (snd_hda_suspend)
> happens after the work queue releases power_lock but before it calls
> hda_call_codec_suspend, codec_suspend runs with power_on=0, causing the
> codec to power up for register reads, and hanging when it calls
> cancel_delayed_work_sync from the running work queue.
>
> The call chain from the work queue will look like this:
> hda_power_work <<- power_on = 1, unlock, then power_on cleard by suspend
> hda_call_codec_suspend
> hda_set_power_state
> snd_hda_codec_read
> codec_exec_verb
> snd_hda_power_up
> snd_hda_power_save
> __snd_hda_power_up
> cancel_delayed_work_sync <<-- cancelling executing wq
>
> Fix this by waiting for the work queue to finish before starting suspend
> if suspend is not happening on the work queue.
>
> Signed-off-by: Dylan Reid <dgreid@chromium.org>
I queued it for 3.7 kernel with Cc to stable. Thanks!
Takashi
> ---
> sound/pci/hda/hda_codec.c | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/sound/pci/hda/hda_codec.c b/sound/pci/hda/hda_codec.c
> index 960800b..c0ab72c 100644
> --- a/sound/pci/hda/hda_codec.c
> +++ b/sound/pci/hda/hda_codec.c
> @@ -3612,7 +3612,7 @@ static inline void hda_exec_init_verbs(struct hda_codec *codec) {}
> * call suspend and power-down; used both from PM and power-save
> * this function returns the power state in the end
> */
> -static unsigned int hda_call_codec_suspend(struct hda_codec *codec)
> +static unsigned int hda_call_codec_suspend(struct hda_codec *codec, bool in_wq)
> {
> unsigned int state;
>
> @@ -3620,7 +3620,9 @@ static unsigned int hda_call_codec_suspend(struct hda_codec *codec)
> codec->patch_ops.suspend(codec);
> hda_cleanup_all_streams(codec);
> state = hda_set_power_state(codec, AC_PWRST_D3);
> - cancel_delayed_work(&codec->power_work);
> + /* Cancel delayed work if we aren't currently running from it. */
> + if (!in_wq)
> + cancel_delayed_work_sync(&codec->power_work);
> spin_lock(&codec->power_lock);
> snd_hda_update_power_acct(codec);
> trace_hda_power_down(codec);
> @@ -4478,7 +4480,7 @@ static void hda_power_work(struct work_struct *work)
> }
> spin_unlock(&codec->power_lock);
>
> - state = hda_call_codec_suspend(codec);
> + state = hda_call_codec_suspend(codec, true);
> codec->pm_down_notified = 0;
> if (!bus->power_keep_link_on && (state & AC_PWRST_CLK_STOP_OK)) {
> codec->pm_down_notified = 1;
> @@ -5127,7 +5129,7 @@ int snd_hda_suspend(struct hda_bus *bus)
>
> list_for_each_entry(codec, &bus->codec_list, list) {
> if (hda_codec_is_power_on(codec))
> - hda_call_codec_suspend(codec);
> + hda_call_codec_suspend(codec, false);
> }
> return 0;
> }
> --
> 1.7.12.146.g16d26b1
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2012-10-08 8:17 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-09-28 22:57 [PATCH] ALSA: hda - Fix hang caused by race during suspend Dylan Reid
2012-10-08 8:17 ` Takashi Iwai
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).