"Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

"Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Luke Ross]

Hi,

Last month I submitted a patch adding support for non-standard sample 
rates to the snd-usb-audio driver [1], which was added to the ALSA GIT 
tree.

This works fine with 2.6.16 on i386 on one of my machines, but causes a 
kernel panic on opening the sound device on my other machine (dual AMD 
Opterton with x86_64 kernel).

I need to go and hook up a serial cable to get the full panic, but the 
gist of it is:

Unable to handle kernel paging request

snd_pcm_hw_refine+658
snd_pcm_common_ioctl+423
snd_pcm_hw_constraint_list+31
snd_usb_pcm_open+976
snd_pcm_hw_constraint_minmax+39

RIP: snd_pcm_hw_params_any

I've had this on both 2.6.17 with only the above patch (which seems to 
compile cleanly), and also on 2.6.18-rc7 with the MM patches whih 
includes a git-alsa pull.

Before I dig out the serial cable to get a full panic trace, is there 
anything which I ought to be looking at as a first port of call, which 
would explain why this happens on x86_64 bit not i386?

Thanks,

Luke

[1] http://www.spinics.net/lists/alsa-devel/msg03536.html

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Takashi Iwai]

At Wed, 20 Sep 2006 23:51:28 +0100,
Luke Ross wrote:
> 
> Hi,
> 
> Last month I submitted a patch adding support for non-standard sample 
> rates to the snd-usb-audio driver [1], which was added to the ALSA GIT 
> tree.
> 
> This works fine with 2.6.16 on i386 on one of my machines, but causes a 
> kernel panic on opening the sound device on my other machine (dual AMD 
> Opterton with x86_64 kernel).
> 
> I need to go and hook up a serial cable to get the full panic, but the 
> gist of it is:
> 
> Unable to handle kernel paging request
> 
> snd_pcm_hw_refine+658
> snd_pcm_common_ioctl+423
> snd_pcm_hw_constraint_list+31
> snd_usb_pcm_open+976
> snd_pcm_hw_constraint_minmax+39
> 
> RIP: snd_pcm_hw_params_any
> 
> I've had this on both 2.6.17 with only the above patch (which seems to 
> compile cleanly), and also on 2.6.18-rc7 with the MM patches whih 
> includes a git-alsa pull.
> 
> Before I dig out the serial cable to get a full panic trace, is there 
> anything which I ought to be looking at as a first port of call, which 
> would explain why this happens on x86_64 bit not i386?

Try to build with debug option.  Maybe it comes from OSS emulation?

Or are you using dmix?


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Luke Ross]

Hi,

On Thu, Sep 21, 2006 at 08:42:24PM +0200, Takashi Iwai wrote:
> At Wed, 20 Sep 2006 23:51:28 +0100,
> Luke Ross wrote:
> > 
> > Last month I submitted a patch adding support for non-standard sample 
> > rates to the snd-usb-audio driver [1], which was added to the ALSA GIT 
> > tree.
> > 
> > This works fine with 2.6.16 on i386 on one of my machines, but causes a 
> > kernel panic on opening the sound device on my other machine (dual AMD 
> > Opterton with x86_64 kernel).
> 
> Try to build with debug option.  Maybe it comes from OSS emulation?
> 
> Or are you using dmix?

At the moment I'm using neither. I'm using an alsa-enabled mpg123:

mpg123 -a hw:1,0 sound.mp3

I've built the kernel with CONFIG_SND_DEBUG, which I assume is what you 
mean, but I didn't get any further output. Do you need to enable the 
debug at runtime as well?

Luke

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Takashi Iwai]

At Fri, 22 Sep 2006 11:48:50 +0100,
Luke Ross wrote:
> 
> Hi,
> 
> On Thu, Sep 21, 2006 at 08:42:24PM +0200, Takashi Iwai wrote:
> > At Wed, 20 Sep 2006 23:51:28 +0100,
> > Luke Ross wrote:
> > > 
> > > Last month I submitted a patch adding support for non-standard sample 
> > > rates to the snd-usb-audio driver [1], which was added to the ALSA GIT 
> > > tree.
> > > 
> > > This works fine with 2.6.16 on i386 on one of my machines, but causes a 
> > > kernel panic on opening the sound device on my other machine (dual AMD 
> > > Opterton with x86_64 kernel).
> > 
> > Try to build with debug option.  Maybe it comes from OSS emulation?
> > 
> > Or are you using dmix?
> 
> At the moment I'm using neither. I'm using an alsa-enabled mpg123:
> 
> mpg123 -a hw:1,0 sound.mp3

Is there any other known program to cause this?
Also, does it happen with the very first time you use the sound
driver, too?

I still couldn't find a proper way to reproduce this bug...


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Luke Ross]

Hi,

On Mon, Sep 25, 2006 at 06:29:01PM +0200, Takashi Iwai wrote:
>
> > > Or are you using dmix?
> > 
> > At the moment I'm using neither. I'm using an alsa-enabled mpg123:
> > 
> > mpg123 -a hw:1,0 sound.mp3
> 
> Is there any other known program to cause this?
> Also, does it happen with the very first time you use the sound
> driver, too?

Yes, it happens with all programs I've tried so far, including 
gstream-properties, and it always happens on the first time the channel 
is PCM opened (the mixer is fine).

> I still couldn't find a proper way to reproduce this bug...

So I can compare, which kernel version are you using? As touched on 
before, it only happens to me on x86_64, and it may be SMP dependent 
(my i386 machine is uniprocessor, my x86_64 one is SMP).

I'm still awaiting the laplink cable to do a full panic log :-(

Thanks,

Luke
-- 
``Never attribute to malice what can be explained by incompetence.'' (Hanlon)

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Takashi Iwai]

At Mon, 25 Sep 2006 00:48:58 +0100,
Luke Ross wrote:
> 
> Hi,
> 
> On Mon, Sep 25, 2006 at 06:29:01PM +0200, Takashi Iwai wrote:
> >
> > > > Or are you using dmix?
> > > 
> > > At the moment I'm using neither. I'm using an alsa-enabled mpg123:
> > > 
> > > mpg123 -a hw:1,0 sound.mp3
> > 
> > Is there any other known program to cause this?
> > Also, does it happen with the very first time you use the sound
> > driver, too?
> 
> Yes, it happens with all programs I've tried so far, including 
> gstream-properties, and it always happens on the first time the channel 
> is PCM opened (the mixer is fine).

I have no mpg123 but only mpg321.  And I don't use gstreamer...
Anyway, then something else than I thought of.

> > I still couldn't find a proper way to reproduce this bug...
> 
> So I can compare, which kernel version are you using? As touched on 
> before, it only happens to me on x86_64, and it may be SMP dependent 
> (my i386 machine is uniprocessor, my x86_64 one is SMP).

My test machine is an Intel x86-64 single core running (almost)
vanilla 2.6.18 SMP kernel with HT.  ALSA 1.0.13rc3 is now built on
that, and seems working with mpg321, at least.
The whole ALSA stuff is built as modules.


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Luke Ross]

Hi,

On Tue, Sep 26, 2006 at 06:37:54PM +0200, Takashi Iwai wrote:
> 
> > > I still couldn't find a proper way to reproduce this bug...
> > 
> > So I can compare, which kernel version are you using? As touched on 
> > before, it only happens to me on x86_64, and it may be SMP dependent 
> > (my i386 machine is uniprocessor, my x86_64 one is SMP).
> 
> My test machine is an Intel x86-64 single core running (almost)
> vanilla 2.6.18 SMP kernel with HT.  ALSA 1.0.13rc3 is now built on
> that, and seems working with mpg321, at least.
> The whole ALSA stuff is built as modules.

Sorry if I've jumped to an incorrect conclusion here, but you are 
including the patch and/or Hg pull? Neither 2.6.18 nor 1.0.13rc3 
incorporate the patch in question, as it went in after both of these 
were released. The patch is in Hg and also the latest -mm, although I 
can also reproduce the problem by taking a stock kernel and applying the 
patch as posted to the mailing list[1]

Regards,

Luke

[1] 
http://hg-mirror.alsa-project.org/alsa-kernel?cmd=changeset;node=d55e9f78b1c4fbff1cd5d4c6462ec111aaea9267;style=raw
-- 
``Avoid employing unlucky people - throw half of the pile of CVs in the bin
  without reading them'' (David Brent)

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Takashi Iwai]

At Tue, 26 Sep 2006 11:13:15 +0100,
Luke Ross wrote:
> 
> Hi,
> 
> On Tue, Sep 26, 2006 at 06:37:54PM +0200, Takashi Iwai wrote:
> > 
> > > > I still couldn't find a proper way to reproduce this bug...
> > > 
> > > So I can compare, which kernel version are you using? As touched on 
> > > before, it only happens to me on x86_64, and it may be SMP dependent 
> > > (my i386 machine is uniprocessor, my x86_64 one is SMP).
> > 
> > My test machine is an Intel x86-64 single core running (almost)
> > vanilla 2.6.18 SMP kernel with HT.  ALSA 1.0.13rc3 is now built on
> > that, and seems working with mpg321, at least.
> > The whole ALSA stuff is built as modules.
> 
> Sorry if I've jumped to an incorrect conclusion here, but you are 
> including the patch and/or Hg pull? Neither 2.6.18 nor 1.0.13rc3 
> incorporate the patch in question, as it went in after both of these 
> were released.

Hm?  ALSA 1.0.13rc3 includes _all_ HG patches...


Takashi

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Luke Ross]

Hi,

On Mon, Sep 25, 2006 at 06:29:01PM +0200, Takashi Iwai wrote:
> > On Thu, Sep 21, 2006 at 08:42:24PM +0200, Takashi Iwai wrote:
> > > At Wed, 20 Sep 2006 23:51:28 +0100,
> > > Luke Ross wrote:
> > > > 
> > > > Last month I submitted a patch adding support for non-standard sample 
> > > > rates to the snd-usb-audio driver [1], which was added to the ALSA GIT 
> > > > tree.
> > > > 
> > > > This works fine with 2.6.16 on i386 on one of my machines, but causes a 
> > > > kernel panic on opening the sound device on my other machine (dual AMD 
> > > > Opterton with x86_64 kernel).
> > > 
> > > Try to build with debug option.  Maybe it comes from OSS emulation?
> > > 
> > > Or are you using dmix?
> > 
> > At the moment I'm using neither. I'm using an alsa-enabled mpg123:
> > 
> > mpg123 -a hw:1,0 sound.mp3

Here's the panic log, in the hope it makes things clearer.

Cheers,

Luke

Fedora Core release 5 (Bordeaux)
Kernel 2.6.17-1.2187_FC5 on an x86_64

hintertux login: Unable to handle kernel paging request at 0000000067f53f32 RIP: 
<ffffffff881514a8>{:snd_pcm:snd_interval_list+91}
PGD 3d184067 PUD 0 
Oops: 0000 [1] SMP 
last sysfs file: /devices/pci0000:00/0000:00:07.3/i2c-0/0-002e/fan4_min
CPU 1 
Modules linked in: ipv6 autofs4 w83627hf eeprom lm75 lm85 hwmon_vid hwmon i2c_isa hidp l2cap bluetooth ipt_REJECT xt_tcpudp xt_state ip_conntrack nfnetlink iptable_filter ip_tables x_tables dm_mirror dm_multipath dm_mod video button battery acpi_memhotplug ac lp parport_pc parport usb_storage snd_usb_audio snd_usb_lib snd_rawmidi snd_hwdep snd_intel8x0 snd_ac97_codec snd_ac97_bus snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq ohci1394 snd_seq_device ieee1394 snd_pcm_oss snd_mixer_oss ohci_hcd sg tg3 snd_pcm snd_timer snd soundcore snd_page_alloc hw_random i2c_amd8111 i2c_amd756 serio_raw i2c_core raid1 ext3 jbd sata_sil libata sd_mod scsi_mod
Pid: 3027, comm: mpg123 Not tainted 2.6.17-1.2187_FC5 #1
RIP: 0010:[<ffffffff881514a8>] <ffffffff881514a8>{:snd_pcm:snd_interval_list+91}
RSP: 0018:ffff810020389b50  EFLAGS: 00010206
RAX: 0000000037f8096b RBX: ffff810037f80c00 RCX: 0000000037f8096b
RDX: ffffffff88151986 RSI: 0000000037f80978 RDI: ffff810037f80d28
RBP: ffff810020389cb8 R08: 0000000000000fa0 R09: 0000000000000001
R10: ffff810020389cb3 R11: 00000000ffffffff R12: ffff810038a86450
R13: 0000000000000017 R14: 0000000000000019 R15: 0000000000000017
FS:  00002aaaaaae1380(0000) GS:ffff810037e91640(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000067f53f32 CR3: 000000002006d000 CR4: 00000000000006e0
Process mpg123 (pid: 3027, threadinfo ffff810020388000, task ffff810020353860)
Stack: ffffffff8814f43b ffff81001d15a960 0000000300000002 0000000500000004 
       0000000700000006 0000000900000008 0000000b0000000a 0000000d0000000c 
       0000000f0000000e 0000001100000010 
Call Trace: <ffffffff8814f43b>{:snd_pcm:snd_pcm_hw_refine+658}
       <ffffffff8814fb4a>{:snd_pcm:snd_pcm_common_ioctl1+447}
       <ffffffff88154338>{:snd_pcm:snd_pcm_hw_constraint_list+31}
       <ffffffff883246fe>{:snd_usb_audio:snd_usb_pcm_open+976}
       <ffffffff88151986>{:snd_pcm:snd_pcm_hw_constraint_minmax+39}
       <ffffffff80269d1c>{_read_unlock_irq+9} <ffffffff802075ed>{find_get_page+71}
       <ffffffff80213e89>{filemap_nopage+390} <ffffffff802089be>{__handle_mm_fault+1303}
       <ffffffff80269d8e>{_spin_lock_irqsave+9} <ffffffff80222cba>{__up_read+25}
       <ffffffff88150b93>{:snd_pcm:snd_pcm_playback_ioctl1+569}
       <ffffffff8020dd02>{do_mmap_pgoff+1543} <ffffffff881513e9>{:snd_pcm:snd_pcm_playback_ioctl+38}
       <ffffffff80244ada>{do_ioctl+33} <ffffffff80231e88>{vfs_ioctl+598}
       <ffffffff8024fb4d>{sys_ioctl+89} <ffffffff80262b41>{tracesys+209}

Code: 8b 04 82 39 47 04 75 08 f6 47 08 02 75 04 eb 18 77 09 ff c9 
RIP <ffffffff881514a8>{:snd_pcm:snd_interval_list+91} RSP <ffff810020389b50>
CR2: 0000000067f53f32
 <0>Kernel panic - not syncing: Oops
 

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Takashi Iwai]

At Wed, 4 Oct 2006 21:13:41 +0100,
Luke Ross wrote:
> 
> Hi,
> 
> On Mon, Sep 25, 2006 at 06:29:01PM +0200, Takashi Iwai wrote:
> > > On Thu, Sep 21, 2006 at 08:42:24PM +0200, Takashi Iwai wrote:
> > > > At Wed, 20 Sep 2006 23:51:28 +0100,
> > > > Luke Ross wrote:
> > > > > 
> > > > > Last month I submitted a patch adding support for non-standard sample 
> > > > > rates to the snd-usb-audio driver [1], which was added to the ALSA GIT 
> > > > > tree.
> > > > > 
> > > > > This works fine with 2.6.16 on i386 on one of my machines, but causes a 
> > > > > kernel panic on opening the sound device on my other machine (dual AMD 
> > > > > Opterton with x86_64 kernel).
> > > > 
> > > > Try to build with debug option.  Maybe it comes from OSS emulation?
> > > > 
> > > > Or are you using dmix?
> > > 
> > > At the moment I'm using neither. I'm using an alsa-enabled mpg123:
> > > 
> > > mpg123 -a hw:1,0 sound.mp3
> 
> Here's the panic log, in the hope it makes things clearer.

According to the stack trace, the problem looks rather like a problem
of usb-audio driver than the PCM core stuff I've suspected.

Could you try the patch below?  It's to the latest HG but should be
applicable to other recent versions, too.


Takashi

diff -r 04575c70b975 usb/usbaudio.c
--- a/usb/usbaudio.c	Thu Oct 05 16:21:19 2006 +0200
+++ b/usb/usbaudio.c	Fri Oct 06 15:31:19 2006 +0200
@@ -2457,6 +2457,7 @@ static int parse_audio_format_rates(stru
 		return -1;
 	}
 
+	fp->needs_knot = 0;
 	if (nr_rates) {
 		/*
 		 * build the rate table and bitmap flags
@@ -2473,7 +2474,6 @@ static int parse_audio_format_rates(stru
 			return -1;
 		}
 
-		fp->needs_knot = 0;
 		fp->nr_rates = nr_rates;
 		fp->rate_min = fp->rate_max = combine_triple(&fmt[8]);
 		for (r = 0, idx = offset + 1; r < nr_rates; r++, idx += 3) {

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: "Unable to handle kernel paging request" panic in snd-usb-audio on x86_64

[Luke Ross]

Hi,

On Fri, Oct 06, 2006 at 03:31:59PM +0200, Takashi Iwai wrote:
>
> According to the stack trace, the problem looks rather like a problem
> of usb-audio driver than the PCM core stuff I've suspected.
> 
> Could you try the patch below?  It's to the latest HG but should be
> applicable to other recent versions, too.

I gave your patch a try, but the problem is still there.

Just for the record, the device in question supports only one rate, 
which is a non-ALSA-standard one. Therefore, needs_knot should be set 
and hence the call to snd_pcm_hw_constraint_list.

As a basic sanity check I did some printk just before the call and 
neither runtime nor fp are null, and fp->nr_rates is 1 as expected. I'm 
happy to try some more printk to try and work out what's going on, but 
still don't quite see why the code/soundcard combination works on i386 
but not x86_64 - could it be arch dependent?

Luke

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV