At Mon, 15 Feb 2010 23:35:26 +0100, I wrote:

At Mon, 15 Feb 2010 17:24:54 -0500, Ed Tomlinson wrote:

...

On Monday 15 February 2010 14:21:55 Takashi Iwai wrote:

...

At Mon, 15 Feb 2010 20:20:22 +0100, I wrote:

...

At Sat, 13 Feb 2010 12:17:10 -0500, Ed Tomlinson wrote:

...

Hi,

Inorder to get skype working with linux I have the following in .asoundrc for my normal userid.

---

pcm.pulse { type pulse }

ctl.pulse { type pulse }

pcm.!default { type pulse } ctl.!default { type pulse

---

Try #1 Starting as root if I do: su - <id> aplay sound.wav

<traceback inlined below> aplay sound.wav (silence) killall pulseaudio aplay sound.wav (works as expected now going thru pulseaudio)

Try #2 Starting as root if I do: aplay sound.wav (works as expected using alsa alone) su - <id> aplay sound.wav

<another lockdep traceback> aplay sound.wav (works as expected now going thru pulseaudio)

With .32 there were no tracebacks (lockdep was enabled) and try #1 would work where I now get silence.

Ideas? Ed Tomlinson

aplay -l **** List of PLAYBACK Hardware Devices **** card 0: SB [HDA ATI SB], device 0: ALC1200 Analog [ALC1200 Analog] Subdevices: 0/1 Subdevice #0: subdevice #0 card 0: SB [HDA ATI SB], device 1: ALC1200 Digital [ALC1200 Digital] Subdevices: 1/1 Subdevice #0: subdevice #0 card 1: HDMI [HDA ATI HDMI], device 3: ATI HDMI [ATI HDMI] Subdevices: 1/1 Subdevice #0: subdevice #0

"ALC1200 Analog" is what has speakers connected

The em28xx in the traceback is an input source from a usb hdtv dongle.

pavucontrol setup has been verified.

[ 91.070620] [ 91.070621] ======================================================= [ 91.071378] [ INFO: possible circular locking dependency detected ] [ 91.071378] 2.6.33-rc8-crc #106 [ 91.071378] ------------------------------------------------------- [ 91.071378] pulseaudio/2717 is trying to acquire lock: [ 91.071378] (&dev->lock){+.+.+.}, at: [<ffffffffa02797f1>] snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] [ 91.071378] [ 91.071378] but task is already holding lock: [ 91.071378] (&pcm->open_mutex){+.+.+.}, at: [<ffffffffa010ecd2>] snd_pcm_open+0x182/0x450 [snd_pcm] [ 91.071378] [ 91.071378] which lock already depends on the new lock. [ 91.071378] [ 91.071378] [ 91.071378] the existing dependency chain (in reverse order) is: [ 91.071378] [ 91.071378] -> #3 (&pcm->open_mutex){+.+.+.}: [ 91.071378] [<ffffffff81083215>] __lock_acquire+0xfc5/0x1550 [ 91.071378] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 [ 91.071378] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 [ 91.071378] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 [ 91.071378] [<ffffffffa010ad32>] snd_pcm_release+0x42/0xb0 [snd_pcm] [ 91.071378] [<ffffffff8112115d>] __fput+0x15d/0x290 [ 91.071378] [<ffffffff811212ad>] fput+0x1d/0x30 [ 91.071378] [<ffffffff810fe001>] remove_vma+0x51/0x80 [ 91.071378] [<ffffffff810ffa1d>] do_munmap+0x2fd/0x390 [ 91.071378] [<ffffffff810ffb06>] sys_munmap+0x56/0x80 [ 91.071378] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b [ 91.071378] [ 91.071378] -> #2 (&mm->mmap_sem){++++++}: [ 91.071378] [<ffffffff81083215>] __lock_acquire+0xfc5/0x1550 [ 91.071378] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 [ 91.071378] [<ffffffff810f62c7>] might_fault+0xa7/0xd0 [ 91.071378] [<ffffffff8113102b>] filldir+0x7b/0xe0 [ 91.071378] [<ffffffff8118aa21>] sysfs_readdir+0xf1/0x1b0 [ 91.071378] [<ffffffff81131268>] vfs_readdir+0xb8/0xe0 [ 91.071378] [<ffffffff8113144d>] sys_getdents+0xad/0x110 [ 91.071378] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b [ 91.071378] [ 91.071378] -> #1 (sysfs_mutex){+.+.+.}: [ 91.071378] [<ffffffff81083215>] __lock_acquire+0xfc5/0x1550 [ 91.071378] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 [ 91.071378] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 [ 91.071378] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 [ 91.071378] [<ffffffff8118b082>] sysfs_addrm_start+0x22/0x30 [ 91.071378] [<ffffffff8118bc58>] create_dir+0x58/0xb0 [ 91.514885] [<ffffffff8118bd0a>] sysfs_create_dir+0x5a/0x70 [ 91.514885] [<ffffffff8123ebbd>] kobject_add_internal+0xbd/0x1f0 [ 91.514885] [<ffffffff8123ee08>] kobject_add_varg+0x38/0x60 [ 91.514885] [<ffffffff8123eee4>] kobject_add+0x44/0x70 [ 91.514885] [<ffffffff81392510>] device_add+0xb0/0x5e0 [ 91.514885] [<ffffffff81392a5e>] device_register+0x1e/0x30 [ 91.514885] [<ffffffff8140a50d>] i2c_register_adapter+0x12d/0x250 [ 91.514885] [<ffffffff8140a7f1>] i2c_add_adapter+0xb1/0xd0 [ 91.514885] [<ffffffffa0231ba4>] em28xx_i2c_register+0x104/0x520 [em28xx] [ 91.514885] [<ffffffffa0233daa>] em28xx_usb_probe+0x68a/0xb90 [em28xx] [ 91.514885] [<ffffffff813eb4ab>] usb_probe_interface+0xeb/0x1b0 [ 91.514885] [<ffffffff81395546>] driver_probe_device+0xc6/0x1d0 [ 91.514885] [<ffffffff813956eb>] __driver_attach+0x9b/0xa0 [ 91.514885] [<ffffffff8139492c>] bus_for_each_dev+0x6c/0xa0 [ 91.514885] [<ffffffff8139535e>] driver_attach+0x1e/0x20 [ 91.514885] [<ffffffff81394c11>] bus_add_driver+0xe1/0x280 [ 91.514885] [<ffffffff81395a58>] driver_register+0x98/0x140 [ 91.514885] [<ffffffff813ea06c>] usb_register_driver+0xdc/0x1a0 [ 91.514885] [<ffffffffa0248023>] 0xffffffffa0248023 [ 91.514885] [<ffffffff810001dc>] do_one_initcall+0x3c/0x1d0 [ 91.514885] [<ffffffff81092d75>] sys_init_module+0xe5/0x250 [ 91.514885] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b [ 91.514885] [ 91.514885] -> #0 (&dev->lock){+.+.+.}: [ 91.514885] [<ffffffff81083668>] __lock_acquire+0x1418/0x1550 [ 91.514885] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 [ 91.514885] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 [ 91.514885] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 [ 91.514885] [<ffffffffa02797f1>] snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] [ 91.514885] [<ffffffffa010adee>] snd_pcm_open_substream+0x4e/0x90 [snd_pcm] [ 91.514885] [<ffffffffa010ece9>] snd_pcm_open+0x199/0x450 [snd_pcm] [ 91.514885] [<ffffffffa010efd4>] snd_pcm_capture_open+0x34/0x40 [snd_pcm] [ 91.514885] [<ffffffffa00c5778>] snd_open+0x198/0x4e0 [snd] [ 91.514885] [<ffffffff8112384d>] chrdev_open+0x17d/0x320 [ 91.514885] [<ffffffff8111d358>] __dentry_open+0x1a8/0x400 [ 91.514885] [<ffffffff8111d6b4>] nameidata_to_filp+0x54/0x70 [ 91.514885] [<ffffffff8112e7f1>] do_filp_open+0x841/0xc00 [ 91.514885] [<ffffffff8111e954>] do_sys_open+0xa4/0x180 [ 91.514885] [<ffffffff8111ea70>] sys_open+0x20/0x30 [ 91.514885] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b [ 91.514885] [ 91.514885] other info that might help us debug this: [ 91.514885] [ 91.514885] 1 lock held by pulseaudio/2717: [ 91.514885] #0: (&pcm->open_mutex){+.+.+.}, at: [<ffffffffa010ecd2>] snd_pcm_open+0x182/0x450 [snd_pcm] [ 91.514885] [ 91.514885] stack backtrace: [ 91.514885] Pid: 2717, comm: pulseaudio Not tainted 2.6.33-rc8-crc #106 [ 91.514885] Call Trace: [ 91.514885] [<ffffffff81080fd9>] print_circular_bug+0xe9/0xf0 [ 91.514885] [<ffffffff81083668>] __lock_acquire+0x1418/0x1550 [ 91.514885] [<ffffffff81081dc8>] ? debug_check_no_locks_freed+0xc8/0x150 [ 91.514885] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 [ 91.514885] [<ffffffffa02797f1>] ? snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] [ 91.514885] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 [ 91.514885] [<ffffffffa02797f1>] ? snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] [ 91.514885] [<ffffffffa02797f1>] ? snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] [ 91.514885] [<ffffffffa0109c70>] ? snd_pcm_hw_rule_muldivk+0x0/0xa0 [snd_pcm] [ 91.514885] [<ffffffff8107fd54>] ? lockdep_init_map+0x44/0x130 [ 91.514885] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 [ 91.514885] [<ffffffffa02797f1>] snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] [ 91.514885] [<ffffffffa010adee>] snd_pcm_open_substream+0x4e/0x90 [snd_pcm] [ 91.514885] [<ffffffffa010ece9>] snd_pcm_open+0x199/0x450 [snd_pcm] [ 91.514885] [<ffffffff81081cfd>] ? trace_hardirqs_on+0xd/0x10 [ 91.514885] [<ffffffff810434f0>] ? default_wake_function+0x0/0x20 [ 91.514885] [<ffffffff814b965e>] ? mutex_unlock+0xe/0x10 [ 91.514885] [<ffffffffa010efd4>] snd_pcm_capture_open+0x34/0x40 [snd_pcm] [ 91.514885] [<ffffffffa00c5778>] snd_open+0x198/0x4e0 [snd] [ 91.514885] [<ffffffff8112384d>] chrdev_open+0x17d/0x320 [ 91.514885] [<ffffffff8111d358>] __dentry_open+0x1a8/0x400 [ 91.514885] [<ffffffff811236d0>] ? chrdev_open+0x0/0x320 [ 91.514885] [<ffffffff8111d6b4>] nameidata_to_filp+0x54/0x70 [ 91.514885] [<ffffffff8112e7f1>] do_filp_open+0x841/0xc00 [ 91.514885] [<ffffffff814bf981>] ? sub_preempt_count+0x51/0x60 [ 91.514885] [<ffffffff814bbb7c>] ? _raw_spin_unlock+0x5c/0x70 [ 91.514885] [<ffffffff8111e954>] do_sys_open+0xa4/0x180 [ 91.514885] [<ffffffff8111ea70>] sys_open+0x20/0x30 [ 91.514885] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b [ 92.643423] hda-intel: IRQ timing workaround is activated for card #0. Suggest a bigger bdl_pos_adj.

It looks rather a bug of em28xx driver. Changing dev->lock with an individual one that is only for audio instance should fix the problem.

And, I believe this is no regression but a long-standing problem. It hits just occasionally by an app like pulseauiod which loves tight races.

I guess it depends on how you define a regression. From my POV it worked in all three cases on .32, its failing in .33-rc and in one case is oppsing. In my books thats a regression. There may have been races in .32 but they did not trigger lockdep and never oppsed.

Is the kernel config really identical? Is only the kernel replaced?

I haven't followed em28xx development, so it's possible that the bug came from that side, though. But, in the PCM core, there is no fundamental change that may trigger this bug during 2.6.33 development.

...

In any case, the real question is how do we fix the problem?

As I mentioned, replace the mutex in em28xx code.

I mean, something like below (totally untested).

Takashi

--- diff --git a/drivers/media/video/em28xx/em28xx-audio.c b/drivers/media/video/em28xx/em28xx-audio.c index bd78338..c2632c4 100644 --- a/drivers/media/video/em28xx/em28xx-audio.c +++ b/drivers/media/video/em28xx/em28xx-audio.c @@ -301,9 +301,9 @@ static int snd_em28xx_capture_open(struct snd_pcm_substream *substream) /* Sets volume, mute, etc */

dev->mute = 0; - mutex_lock(&dev->lock); + mutex_lock(&dev->adev.mutex); ret = em28xx_audio_analog_set(dev); - mutex_unlock(&dev->lock); + mutex_unlock(&dev->adev.mutex); if (ret < 0) goto err;

@@ -315,9 +315,9 @@ static int snd_em28xx_capture_open(struct snd_pcm_substream *substream) dprintk("changing alternate number to 7\n"); }

- mutex_lock(&dev->lock); + mutex_lock(&dev->adev.mutex); dev->adev.users++; - mutex_unlock(&dev->lock); + mutex_unlock(&dev->adev.mutex);

snd_pcm_hw_constraint_integer(runtime, SNDRV_PCM_HW_PARAM_PERIODS); dev->adev.capture_pcm_substream = substream; @@ -336,7 +336,7 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) dprintk("closing device\n");

dev->mute = 1; - mutex_lock(&dev->lock); + mutex_lock(&dev->adev.mutex); dev->adev.users--; em28xx_audio_analog_set(dev); if (substream->runtime->dma_area) { @@ -344,7 +344,7 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) vfree(substream->runtime->dma_area); substream->runtime->dma_area = NULL; } - mutex_unlock(&dev->lock); + mutex_unlock(&dev->adev.mutex);

return 0; } @@ -479,6 +479,7 @@ static int em28xx_audio_init(struct em28xx *dev) return err;

spin_lock_init(&adev->slock); + mutex_init(&adev->mutex); err = snd_pcm_new(card, "Em28xx Audio", 0, 0, 1, &pcm); if (err < 0) { snd_card_free(card); diff --git a/drivers/media/video/em28xx/em28xx.h b/drivers/media/video/em28xx/em28xx.h index 80d9b4f..bb1d754 100644 --- a/drivers/media/video/em28xx/em28xx.h +++ b/drivers/media/video/em28xx/em28xx.h @@ -461,6 +461,7 @@ struct em28xx_audio { int users; enum em28xx_stream_state capture_stream; spinlock_t slock; + struct mutex mutex; };

struct em28xx;

At Tue, 16 Feb 2010 08:19:21 -0500, Ed Tomlinson wrote:

On Monday 15 February 2010 18:21:17 Takashi Iwai wrote:

...

At Mon, 15 Feb 2010 23:35:26 +0100, I wrote:

...

At Mon, 15 Feb 2010 17:24:54 -0500, Ed Tomlinson wrote:

...

On Monday 15 February 2010 14:21:55 Takashi Iwai wrote:

...

At Mon, 15 Feb 2010 20:20:22 +0100, I wrote:

...

At Sat, 13 Feb 2010 12:17:10 -0500, Ed Tomlinson wrote: > > Hi, > > Inorder to get skype working with linux I have the following in .asoundrc for my normal userid. > > --- > pcm.pulse { > type pulse > } > > ctl.pulse { > type pulse > } > > pcm.!default { > type pulse > } > ctl.!default { > type pulse > --- > > Try #1 > Starting as root if I do: > su - <id> > aplay sound.wav > <traceback inlined below> > aplay sound.wav > (silence) > killall pulseaudio > aplay sound.wav > (works as expected now going thru pulseaudio) > > Try #2 > Starting as root if I do: > aplay sound.wav > (works as expected using alsa alone) > su - <id> > aplay sound.wav > <another lockdep traceback> > aplay sound.wav > (works as expected now going thru pulseaudio) > > With .32 there were no tracebacks (lockdep was enabled) and try #1 would work where I now get silence. > > Ideas? > Ed Tomlinson > > aplay -l > **** List of PLAYBACK Hardware Devices **** > card 0: SB [HDA ATI SB], device 0: ALC1200 Analog [ALC1200 Analog] > Subdevices: 0/1 > Subdevice #0: subdevice #0 > card 0: SB [HDA ATI SB], device 1: ALC1200 Digital [ALC1200 Digital] > Subdevices: 1/1 > Subdevice #0: subdevice #0 > card 1: HDMI [HDA ATI HDMI], device 3: ATI HDMI [ATI HDMI] > Subdevices: 1/1 > Subdevice #0: subdevice #0 > > "ALC1200 Analog" is what has speakers connected > > The em28xx in the traceback is an input source from a usb hdtv dongle. > > pavucontrol setup has been verified. > > [ 91.070620] > [ 91.070621] ======================================================= > [ 91.071378] [ INFO: possible circular locking dependency detected ] > [ 91.071378] 2.6.33-rc8-crc #106 > [ 91.071378] ------------------------------------------------------- > [ 91.071378] pulseaudio/2717 is trying to acquire lock: > [ 91.071378] (&dev->lock){+.+.+.}, at: [<ffffffffa02797f1>] snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] > [ 91.071378] > [ 91.071378] but task is already holding lock: > [ 91.071378] (&pcm->open_mutex){+.+.+.}, at: [<ffffffffa010ecd2>] snd_pcm_open+0x182/0x450 [snd_pcm] > [ 91.071378] > [ 91.071378] which lock already depends on the new lock. > [ 91.071378] > [ 91.071378] > [ 91.071378] the existing dependency chain (in reverse order) is: > [ 91.071378] > [ 91.071378] -> #3 (&pcm->open_mutex){+.+.+.}: > [ 91.071378] [<ffffffff81083215>] __lock_acquire+0xfc5/0x1550 > [ 91.071378] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 > [ 91.071378] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 > [ 91.071378] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 > [ 91.071378] [<ffffffffa010ad32>] snd_pcm_release+0x42/0xb0 [snd_pcm] > [ 91.071378] [<ffffffff8112115d>] __fput+0x15d/0x290 > [ 91.071378] [<ffffffff811212ad>] fput+0x1d/0x30 > [ 91.071378] [<ffffffff810fe001>] remove_vma+0x51/0x80 > [ 91.071378] [<ffffffff810ffa1d>] do_munmap+0x2fd/0x390 > [ 91.071378] [<ffffffff810ffb06>] sys_munmap+0x56/0x80 > [ 91.071378] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b > [ 91.071378] > [ 91.071378] -> #2 (&mm->mmap_sem){++++++}: > [ 91.071378] [<ffffffff81083215>] __lock_acquire+0xfc5/0x1550 > [ 91.071378] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 > [ 91.071378] [<ffffffff810f62c7>] might_fault+0xa7/0xd0 > [ 91.071378] [<ffffffff8113102b>] filldir+0x7b/0xe0 > [ 91.071378] [<ffffffff8118aa21>] sysfs_readdir+0xf1/0x1b0 > [ 91.071378] [<ffffffff81131268>] vfs_readdir+0xb8/0xe0 > [ 91.071378] [<ffffffff8113144d>] sys_getdents+0xad/0x110 > [ 91.071378] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b > [ 91.071378] > [ 91.071378] -> #1 (sysfs_mutex){+.+.+.}: > [ 91.071378] [<ffffffff81083215>] __lock_acquire+0xfc5/0x1550 > [ 91.071378] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 > [ 91.071378] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 > [ 91.071378] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 > [ 91.071378] [<ffffffff8118b082>] sysfs_addrm_start+0x22/0x30 > [ 91.071378] [<ffffffff8118bc58>] create_dir+0x58/0xb0 > [ 91.514885] [<ffffffff8118bd0a>] sysfs_create_dir+0x5a/0x70 > [ 91.514885] [<ffffffff8123ebbd>] kobject_add_internal+0xbd/0x1f0 > [ 91.514885] [<ffffffff8123ee08>] kobject_add_varg+0x38/0x60 > [ 91.514885] [<ffffffff8123eee4>] kobject_add+0x44/0x70 > [ 91.514885] [<ffffffff81392510>] device_add+0xb0/0x5e0 > [ 91.514885] [<ffffffff81392a5e>] device_register+0x1e/0x30 > [ 91.514885] [<ffffffff8140a50d>] i2c_register_adapter+0x12d/0x250 > [ 91.514885] [<ffffffff8140a7f1>] i2c_add_adapter+0xb1/0xd0 > [ 91.514885] [<ffffffffa0231ba4>] em28xx_i2c_register+0x104/0x520 [em28xx] > [ 91.514885] [<ffffffffa0233daa>] em28xx_usb_probe+0x68a/0xb90 [em28xx] > [ 91.514885] [<ffffffff813eb4ab>] usb_probe_interface+0xeb/0x1b0 > [ 91.514885] [<ffffffff81395546>] driver_probe_device+0xc6/0x1d0 > [ 91.514885] [<ffffffff813956eb>] __driver_attach+0x9b/0xa0 > [ 91.514885] [<ffffffff8139492c>] bus_for_each_dev+0x6c/0xa0 > [ 91.514885] [<ffffffff8139535e>] driver_attach+0x1e/0x20 > [ 91.514885] [<ffffffff81394c11>] bus_add_driver+0xe1/0x280 > [ 91.514885] [<ffffffff81395a58>] driver_register+0x98/0x140 > [ 91.514885] [<ffffffff813ea06c>] usb_register_driver+0xdc/0x1a0 > [ 91.514885] [<ffffffffa0248023>] 0xffffffffa0248023 > [ 91.514885] [<ffffffff810001dc>] do_one_initcall+0x3c/0x1d0 > [ 91.514885] [<ffffffff81092d75>] sys_init_module+0xe5/0x250 > [ 91.514885] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b > [ 91.514885] > [ 91.514885] -> #0 (&dev->lock){+.+.+.}: > [ 91.514885] [<ffffffff81083668>] __lock_acquire+0x1418/0x1550 > [ 91.514885] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 > [ 91.514885] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 > [ 91.514885] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 > [ 91.514885] [<ffffffffa02797f1>] snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] > [ 91.514885] [<ffffffffa010adee>] snd_pcm_open_substream+0x4e/0x90 [snd_pcm] > [ 91.514885] [<ffffffffa010ece9>] snd_pcm_open+0x199/0x450 [snd_pcm] > [ 91.514885] [<ffffffffa010efd4>] snd_pcm_capture_open+0x34/0x40 [snd_pcm] > [ 91.514885] [<ffffffffa00c5778>] snd_open+0x198/0x4e0 [snd] > [ 91.514885] [<ffffffff8112384d>] chrdev_open+0x17d/0x320 > [ 91.514885] [<ffffffff8111d358>] __dentry_open+0x1a8/0x400 > [ 91.514885] [<ffffffff8111d6b4>] nameidata_to_filp+0x54/0x70 > [ 91.514885] [<ffffffff8112e7f1>] do_filp_open+0x841/0xc00 > [ 91.514885] [<ffffffff8111e954>] do_sys_open+0xa4/0x180 > [ 91.514885] [<ffffffff8111ea70>] sys_open+0x20/0x30 > [ 91.514885] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b > [ 91.514885] > [ 91.514885] other info that might help us debug this: > [ 91.514885] > [ 91.514885] 1 lock held by pulseaudio/2717: > [ 91.514885] #0: (&pcm->open_mutex){+.+.+.}, at: [<ffffffffa010ecd2>] snd_pcm_open+0x182/0x450 [snd_pcm] > [ 91.514885] > [ 91.514885] stack backtrace: > [ 91.514885] Pid: 2717, comm: pulseaudio Not tainted 2.6.33-rc8-crc #106 > [ 91.514885] Call Trace: > [ 91.514885] [<ffffffff81080fd9>] print_circular_bug+0xe9/0xf0 > [ 91.514885] [<ffffffff81083668>] __lock_acquire+0x1418/0x1550 > [ 91.514885] [<ffffffff81081dc8>] ? debug_check_no_locks_freed+0xc8/0x150 > [ 91.514885] [<ffffffff8108383c>] lock_acquire+0x9c/0x140 > [ 91.514885] [<ffffffffa02797f1>] ? snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] > [ 91.514885] [<ffffffff814b96c1>] __mutex_lock_common+0x61/0x610 > [ 91.514885] [<ffffffffa02797f1>] ? snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] > [ 91.514885] [<ffffffffa02797f1>] ? snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] > [ 91.514885] [<ffffffffa0109c70>] ? snd_pcm_hw_rule_muldivk+0x0/0xa0 [snd_pcm] > [ 91.514885] [<ffffffff8107fd54>] ? lockdep_init_map+0x44/0x130 > [ 91.514885] [<ffffffff814b9d53>] mutex_lock_nested+0x43/0x50 > [ 91.514885] [<ffffffffa02797f1>] snd_em28xx_capture_open+0x61/0x190 [em28xx_alsa] > [ 91.514885] [<ffffffffa010adee>] snd_pcm_open_substream+0x4e/0x90 [snd_pcm] > [ 91.514885] [<ffffffffa010ece9>] snd_pcm_open+0x199/0x450 [snd_pcm] > [ 91.514885] [<ffffffff81081cfd>] ? trace_hardirqs_on+0xd/0x10 > [ 91.514885] [<ffffffff810434f0>] ? default_wake_function+0x0/0x20 > [ 91.514885] [<ffffffff814b965e>] ? mutex_unlock+0xe/0x10 > [ 91.514885] [<ffffffffa010efd4>] snd_pcm_capture_open+0x34/0x40 [snd_pcm] > [ 91.514885] [<ffffffffa00c5778>] snd_open+0x198/0x4e0 [snd] > [ 91.514885] [<ffffffff8112384d>] chrdev_open+0x17d/0x320 > [ 91.514885] [<ffffffff8111d358>] __dentry_open+0x1a8/0x400 > [ 91.514885] [<ffffffff811236d0>] ? chrdev_open+0x0/0x320 > [ 91.514885] [<ffffffff8111d6b4>] nameidata_to_filp+0x54/0x70 > [ 91.514885] [<ffffffff8112e7f1>] do_filp_open+0x841/0xc00 > [ 91.514885] [<ffffffff814bf981>] ? sub_preempt_count+0x51/0x60 > [ 91.514885] [<ffffffff814bbb7c>] ? _raw_spin_unlock+0x5c/0x70 > [ 91.514885] [<ffffffff8111e954>] do_sys_open+0xa4/0x180 > [ 91.514885] [<ffffffff8111ea70>] sys_open+0x20/0x30 > [ 91.514885] [<ffffffff81002adb>] system_call_fastpath+0x16/0x1b > [ 92.643423] hda-intel: IRQ timing workaround is activated for card #0. Suggest a bigger bdl_pos_adj.

It looks rather a bug of em28xx driver. Changing dev->lock with an individual one that is only for audio instance should fix the problem.

And, I believe this is no regression but a long-standing problem. It hits just occasionally by an app like pulseauiod which loves tight races.

I guess it depends on how you define a regression. From my POV it worked in all three cases on .32, its failing in .33-rc and in one case is oppsing. In my books thats a regression. There may have been races in .32 but they did not trigger lockdep and never oppsed.

Is the kernel config really identical? Is only the kernel replaced?

I haven't followed em28xx development, so it's possible that the bug came from that side, though. But, in the PCM core, there is no fundamental change that may trigger this bug during 2.6.33 development.

...

In any case, the real question is how do we fix the problem?

As I mentioned, replace the mutex in em28xx code.

I mean, something like below (totally untested).

Takashi

---

diff --git a/drivers/media/video/em28xx/em28xx-audio.c b/drivers/media/video/em28xx/em28xx-audio.c index bd78338..c2632c4 100644 --- a/drivers/media/video/em28xx/em28xx-audio.c +++ b/drivers/media/video/em28xx/em28xx-audio.c @@ -301,9 +301,9 @@ static int snd_em28xx_capture_open(struct snd_pcm_substream *substream) /* Sets volume, mute, etc */

dev->mute = 0;

• mutex_lock(&dev->lock);

• mutex_lock(&dev->adev.mutex); ret = em28xx_audio_analog_set(dev);

• mutex_unlock(&dev->lock);

• mutex_unlock(&dev->adev.mutex); if (ret < 0) goto err;

@@ -315,9 +315,9 @@ static int snd_em28xx_capture_open(struct snd_pcm_substream *substream) dprintk("changing alternate number to 7\n"); }

• mutex_lock(&dev->lock);

• mutex_lock(&dev->adev.mutex); dev->adev.users++;

• mutex_unlock(&dev->lock);

• mutex_unlock(&dev->adev.mutex);

  snd_pcm_hw_constraint_integer(runtime, SNDRV_PCM_HW_PARAM_PERIODS); dev->adev.capture_pcm_substream = substream;

@@ -336,7 +336,7 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) dprintk("closing device\n");

dev->mute = 1;

• mutex_lock(&dev->lock);

• mutex_lock(&dev->adev.mutex); dev->adev.users--; em28xx_audio_analog_set(dev); if (substream->runtime->dma_area) {

@@ -344,7 +344,7 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) vfree(substream->runtime->dma_area); substream->runtime->dma_area = NULL; }

• mutex_unlock(&dev->lock);

• mutex_unlock(&dev->adev.mutex);

  return 0;

} @@ -479,6 +479,7 @@ static int em28xx_audio_init(struct em28xx *dev) return err;

spin_lock_init(&adev->slock);

• mutex_init(&adev->mutex); err = snd_pcm_new(card, "Em28xx Audio", 0, 0, 1, &pcm); if (err < 0) { snd_card_free(card);

diff --git a/drivers/media/video/em28xx/em28xx.h b/drivers/media/video/em28xx/em28xx.h index 80d9b4f..bb1d754 100644 --- a/drivers/media/video/em28xx/em28xx.h +++ b/drivers/media/video/em28xx/em28xx.h @@ -461,6 +461,7 @@ struct em28xx_audio { int users; enum em28xx_stream_state capture_stream; spinlock_t slock;

• struct mutex mutex;

};

struct em28xx;

Thanks for the patch. It helps in that it eliminates the opps but lockdep still triggers and aplay still fails. Here is the new traceback.

Hmm, fixing this isn't so trivial. The same problem occurs on other subsystems like NFS over years. And it's still there, AFAIK. The mmap mutex appears suddenly in the strange code path at close.

The patch below might fix, but I'm not 100% sure whether this has no side effect.

Anyway, I doubt very much it being a regression. There is no change in ALSA core side, and also in V4L em28xx code. Maybe the lockdep wasn't triggered by some reason. And, this lockdep warning is almost harmless...

thanks,

Takashi

--- diff --git a/drivers/media/video/em28xx/em28xx-audio.c b/drivers/media/video/em28xx/em28xx-audio.c index bd78338..1759e1f 100644 --- a/drivers/media/video/em28xx/em28xx-audio.c +++ b/drivers/media/video/em28xx/em28xx-audio.c @@ -301,23 +301,12 @@ static int snd_em28xx_capture_open(struct snd_pcm_substream *substream) /* Sets volume, mute, etc */

dev->mute = 0; - mutex_lock(&dev->lock); ret = em28xx_audio_analog_set(dev); - mutex_unlock(&dev->lock); if (ret < 0) goto err;

runtime->hw = snd_em28xx_hw_capture; - if (dev->alt == 0 && dev->adev.users == 0) { - int errCode; - dev->alt = 7; - errCode = usb_set_interface(dev->udev, 0, 7); - dprintk("changing alternate number to 7\n"); - } - - mutex_lock(&dev->lock); dev->adev.users++; - mutex_unlock(&dev->lock);

snd_pcm_hw_constraint_integer(runtime, SNDRV_PCM_HW_PARAM_PERIODS); dev->adev.capture_pcm_substream = substream; @@ -336,7 +325,6 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) dprintk("closing device\n");

dev->mute = 1; - mutex_lock(&dev->lock); dev->adev.users--; em28xx_audio_analog_set(dev); if (substream->runtime->dma_area) { @@ -344,7 +332,6 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) vfree(substream->runtime->dma_area); substream->runtime->dma_area = NULL; } - mutex_unlock(&dev->lock);

return 0; } @@ -366,6 +353,14 @@ static int snd_em28xx_hw_capture_params(struct snd_pcm_substream *substream, /* TODO: set up em28xx audio chip to deliver the correct audio format, current default is 48000hz multiplexed => 96000hz mono which shouldn't matter since analogue TV only supports mono */ + + if (dev->alt == 0 && dev->adev.users == 1) { + int errCode; + dev->alt = 7; + errCode = usb_set_interface(dev->udev, 0, 7); + dprintk("changing alternate number to 7\n"); + } + return 0; }

At Wed, 17 Feb 2010 08:16:32 -0500, Ed Tomlinson wrote:

On Tuesday 16 February 2010 08:37:46 Takashi Iwai wrote:

...

...

Thanks for the patch. It helps in that it eliminates the opps but lockdep still triggers and aplay still fails. Here is the new traceback.

Hmm, fixing this isn't so trivial. The same problem occurs on other subsystems like NFS over years. And it's still there, AFAIK. The mmap mutex appears suddenly in the strange code path at close.

The patch below might fix, but I'm not 100% sure whether this has no side effect.

Anyway, I doubt very much it being a regression. There is no change in ALSA core side, and also in V4L em28xx code. Maybe the lockdep wasn't triggered by some reason. And, this lockdep warning is almost harmless...

Takashi,

The second patch eliminating the lock causes oppes every time (one follows just in case its helpful).

Are you sure? The patch should causes a compile error, so you must have patched manually in a wrong place ;)

Try the patch below (again untested, though).

thanks,

Takashi

--- diff --git a/drivers/media/video/em28xx/em28xx-audio.c b/drivers/media/video/em28xx/em28xx-audio.c index bd78338..c51e623 100644 --- a/drivers/media/video/em28xx/em28xx-audio.c +++ b/drivers/media/video/em28xx/em28xx-audio.c @@ -301,23 +301,12 @@ static int snd_em28xx_capture_open(struct snd_pcm_substream *substream) /* Sets volume, mute, etc */

dev->mute = 0; - mutex_lock(&dev->lock); ret = em28xx_audio_analog_set(dev); - mutex_unlock(&dev->lock); if (ret < 0) goto err;

runtime->hw = snd_em28xx_hw_capture; - if (dev->alt == 0 && dev->adev.users == 0) { - int errCode; - dev->alt = 7; - errCode = usb_set_interface(dev->udev, 0, 7); - dprintk("changing alternate number to 7\n"); - } - - mutex_lock(&dev->lock); dev->adev.users++; - mutex_unlock(&dev->lock);

snd_pcm_hw_constraint_integer(runtime, SNDRV_PCM_HW_PARAM_PERIODS); dev->adev.capture_pcm_substream = substream; @@ -336,7 +325,6 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) dprintk("closing device\n");

dev->mute = 1; - mutex_lock(&dev->lock); dev->adev.users--; em28xx_audio_analog_set(dev); if (substream->runtime->dma_area) { @@ -344,7 +332,6 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) vfree(substream->runtime->dma_area); substream->runtime->dma_area = NULL; } - mutex_unlock(&dev->lock);

return 0; } @@ -352,6 +339,7 @@ static int snd_em28xx_pcm_close(struct snd_pcm_substream *substream) static int snd_em28xx_hw_capture_params(struct snd_pcm_substream *substream, struct snd_pcm_hw_params *hw_params) { + struct em28xx *dev = snd_pcm_substream_chip(substream); unsigned int channels, rate, format; int ret;

@@ -366,6 +354,14 @@ static int snd_em28xx_hw_capture_params(struct snd_pcm_substream *substream, /* TODO: set up em28xx audio chip to deliver the correct audio format, current default is 48000hz multiplexed => 96000hz mono which shouldn't matter since analogue TV only supports mono */ + + if (dev->alt == 0 && dev->adev.users == 1) { + int errCode; + dev->alt = 7; + errCode = usb_set_interface(dev->udev, 0, 7); + dprintk("changing alternate number to 7\n"); + } + return 0; }