[alsa-devel] module snd-atiixp issue with suspend/hibernate/resume
I have a Compaq V2000 laptop with the ATI IXP chipset and make use of the snd-atiixp module. The laptop needs to have the ac97_quirk option set to 7 to enable the mute LED. After setting this option in the /etc/modprobe.d/options.conf file, the mute LED works after a reboot. However, after a resume from suspend or hibernate, the mute LED does not work. The alsa-info.sh script reports that the quirk is still set with 7. At this point a reboot will NOT fix the problem. The only way to fix it is to remove and reinsert the module with modprobe.
This is on a fresh Ubuntu 9.04 install. Any ideas? I wasn't sure if this should be reported here or on the kernel list. I saw a similar issue with the ac97_quirk on the kernel list and they were referred here. I'm a software developer, so I'd be willing to try ideas/possible solutions if you have them.
Thanks, Ryan
On Wed, Apr 29, 2009 at 12:06 AM, Ryan Dunn oryandunn.ml@gmail.com wrote:
I have a Compaq V2000 laptop with the ATI IXP chipset and make use of the snd-atiixp module. The laptop needs to have the ac97_quirk option set to 7 to enable the mute LED. After setting this option in the /etc/modprobe.d/options.conf file, the mute LED works after a reboot. However, after a resume from suspend or hibernate, the mute LED does not work. The alsa-info.sh script reports that the quirk is still set with 7. At this point a reboot will NOT fix the problem. The only way to fix it is to remove and reinsert the module with modprobe.
This is on a fresh Ubuntu 9.04 install. Any ideas? I wasn't sure if this should be reported here or on the kernel list. I saw a similar issue with the ac97_quirk on the kernel list and they were referred here. I'm a software developer, so I'd be willing to try ideas/possible solutions if you have them.
Get the Ubuntu source code for the package that owns that ALSA driver, find the place in the driver's initialization code where ac97_quirk=7 is handled, then check the driver's suspend and resume callbacks and make sure the suspend callback is correctly saving the LED state and that the resume callback is re-initializing the LED from the saved state in the same way the init code does. grep and printk() are your friends ;-)
HTH,
Lee
Thanks Lee,
I grabbed the kernel source from the repos and started perusing the code. It looks like when the module is installed, the call chain starts in atiixp.c and moves into ac97/ac97_codec.c via a call to snd_ac97_tune_hardware() which ultimately twiddles a bit in the AC97_POWERDOWN register. I'm not sure yet if the suspend method properly saves this register on suspend or not; or if the same call chain needs to be repeated on a resume. I appreciate your quick and relevant answer. When I get time, I'll start adding some debug printouts to try this out.
Also, the struct ac97_quirk ac97_quirks[] array has a couple of entries in it currently (in the atiixp.c file). How do I get the subvender/subdevice id's for my system (I didn't seem to have luck with lspci, unless I'm not looking in the right fields)? Would it be possible to edit the driver so that an explict module option would not be needed on my hardware?
Thanks, Ryan
On Wed, Apr 29, 2009 at 10:42 PM, Lee Revell rlrevell@joe-job.com wrote:
On Wed, Apr 29, 2009 at 12:06 AM, Ryan Dunn oryandunn.ml@gmail.com wrote:
I have a Compaq V2000 laptop with the ATI IXP chipset and make use of the snd-atiixp module. The laptop needs to have the ac97_quirk option set to
7
to enable the mute LED. After setting this option in the /etc/modprobe.d/options.conf file, the mute LED works after a reboot. However, after a resume from suspend or hibernate, the mute LED does not work. The alsa-info.sh script reports that the quirk is still set with
At this point a reboot will NOT fix the problem. The only way to fix it
is
to remove and reinsert the module with modprobe.
This is on a fresh Ubuntu 9.04 install. Any ideas? I wasn't sure if
this
should be reported here or on the kernel list. I saw a similar issue
with
the ac97_quirk on the kernel list and they were referred here. I'm a software developer, so I'd be willing to try ideas/possible solutions if
you
have them.
Get the Ubuntu source code for the package that owns that ALSA driver, find the place in the driver's initialization code where ac97_quirk=7 is handled, then check the driver's suspend and resume callbacks and make sure the suspend callback is correctly saving the LED state and that the resume callback is re-initializing the LED from the saved state in the same way the init code does. grep and printk() are your friends ;-)
HTH,
Lee
At Thu, 30 Apr 2009 00:43:02 -0400, Ryan Dunn wrote:
Thanks Lee,
I grabbed the kernel source from the repos and started perusing the code. It looks like when the module is installed, the call chain starts in atiixp.c and moves into ac97/ac97_codec.c via a call to snd_ac97_tune_hardware() which ultimately twiddles a bit in the AC97_POWERDOWN register. I'm not sure yet if the suspend method properly saves this register on suspend or not;
It should. snd_ac97_resume() writes the cached value. Check /proc/asound/card0/codec97#0/ac97#*+regs files before and after suspend. You can change the codec register value directly via proc file, e.g. # echo 0x12 0x1234 > /proc/asound/card0/codec97#0/ac97#0-0+regs (only when you build with the debug option).
The bit 0x8000 of the power-down register should correspond to LED.
Takashi
or if the same call chain needs to be repeated on a resume. I appreciate your quick and relevant answer. When I get time, I'll start adding some debug printouts to try this out.
Also, the struct ac97_quirk ac97_quirks[] array has a couple of entries in it currently (in the atiixp.c file). How do I get the subvender/subdevice id's for my system (I didn't seem to have luck with lspci, unless I'm not looking in the right fields)? Would it be possible to edit the driver so that an explict module option would not be needed on my hardware?
Thanks, Ryan
On Wed, Apr 29, 2009 at 10:42 PM, Lee Revell rlrevell@joe-job.com wrote:
On Wed, Apr 29, 2009 at 12:06 AM, Ryan Dunn <oryandunn.ml@gmail.com> wrote: > I have a Compaq V2000 laptop with the ATI IXP chipset and make use of the > snd-atiixp module. The laptop needs to have the ac97_quirk option set to 7 > to enable the mute LED. After setting this option in the > /etc/modprobe.d/options.conf file, the mute LED works after a reboot. > However, after a resume from suspend or hibernate, the mute LED does not > work. The alsa-info.sh script reports that the quirk is still set with 7. > At this point a reboot will NOT fix the problem. The only way to fix it is > to remove and reinsert the module with modprobe. > > This is on a fresh Ubuntu 9.04 install. Any ideas? I wasn't sure if this > should be reported here or on the kernel list. I saw a similar issue with > the ac97_quirk on the kernel list and they were referred here. I'm a > software developer, so I'd be willing to try ideas/possible solutions if you > have them. Get the Ubuntu source code for the package that owns that ALSA driver, find the place in the driver's initialization code where ac97_quirk=7 is handled, then check the driver's suspend and resume callbacks and make sure the suspend callback is correctly saving the LED state and that the resume callback is re-initializing the LED from the saved state in the same way the init code does. grep and printk() are your friends ;-) HTH, Lee
Thanks Takashi,
I did a bit more testing last night. The LED does work after a reboot (I don't know what I was doing that I thought it didn't). It also retains state after a reboot if I muted it before the reboot. One thing I noticed, if I have it muted when I suspend, on resume, it looks like the mute led flashes back on for a split second, then turns off. Maybe the resume code is overwriting the register? I'll start digging through that code tonight.
Ryan
On Thu, Apr 30, 2009 at 1:53 AM, Takashi Iwai tiwai@suse.de wrote:
At Thu, 30 Apr 2009 00:43:02 -0400, Ryan Dunn wrote:
Thanks Lee,
I grabbed the kernel source from the repos and started perusing the
code. It
looks like when the module is installed, the call chain starts in
atiixp.c and
moves into ac97/ac97_codec.c via a call to snd_ac97_tune_hardware() which ultimately twiddles a bit in the AC97_POWERDOWN register. I'm not sure
yet if
the suspend method properly saves this register on suspend or not;
It should. snd_ac97_resume() writes the cached value. Check /proc/asound/card0/codec97#0/ac97#*+regs files before and after suspend. You can change the codec register value directly via proc file, e.g. # echo 0x12 0x1234 > /proc/asound/card0/codec97#0/ac97#0-0+regs (only when you build with the debug option).
The bit 0x8000 of the power-down register should correspond to LED.
Takashi
or if the same call chain needs to be repeated on a resume. I appreciate your
quick and
relevant answer. When I get time, I'll start adding some debug printouts
to
try this out.
Also, the struct ac97_quirk ac97_quirks[] array has a couple of entries
in it
currently (in the atiixp.c file). How do I get the subvender/subdevice
id's
for my system (I didn't seem to have luck with lspci, unless I'm not
looking
in the right fields)? Would it be possible to edit the driver so that an explict module option would not be needed on my hardware?
Thanks, Ryan
On Wed, Apr 29, 2009 at 10:42 PM, Lee Revell rlrevell@joe-job.com
wrote:
On Wed, Apr 29, 2009 at 12:06 AM, Ryan Dunn <oryandunn.ml@gmail.com> wrote: > I have a Compaq V2000 laptop with the ATI IXP chipset and make useof
the > snd-atiixp module. The laptop needs to have the ac97_quirk optionset
to 7 > to enable the mute LED. After setting this option in the > /etc/modprobe.d/options.conf file, the mute LED works after areboot.
> However, after a resume from suspend or hibernate, the mute LEDdoes not
> work. The alsa-info.sh script reports that the quirk is still setwith
7. > At this point a reboot will NOT fix the problem. The only way tofix it
is > to remove and reinsert the module with modprobe. > > This is on a fresh Ubuntu 9.04 install. Any ideas? I wasn't sureif
this > should be reported here or on the kernel list. I saw a similarissue
with > the ac97_quirk on the kernel list and they were referred here. I'ma
> software developer, so I'd be willing to try ideas/possiblesolutions if
you > have them. Get the Ubuntu source code for the package that owns that ALSAdriver,
find the place in the driver's initialization code where ac97_quirk=7 is handled, then check the driver's suspend and resume callbacks and make sure the suspend callback is correctly saving the LED state and that the resume callback is re-initializing the LED from the saved state in the same way the init code does. grep and printk() are your friends ;-) HTH, Lee
Alsa-devel mailing list Alsa-devel@alsa-project.org http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
Here is a diff of the registers before & after resume: diff regs_before_suspend.txt regs_after_suspend.txt 58c58 < 0:72 = 0000 ---
0:72 = 5000
In the ac97_codec.h file, AC97_POWERDOWN = 0x26. Does this respond to 0:26 in the regs file? If so, it doesn't appear that bit 0x8000 is ever set, but the mute LED works.
On Thu, Apr 30, 2009 at 7:49 AM, Ryan Dunn oryandunn.ml@gmail.com wrote:
Thanks Takashi,
I did a bit more testing last night. The LED does work after a reboot (I don't know what I was doing that I thought it didn't). It also retains state after a reboot if I muted it before the reboot. One thing I noticed, if I have it muted when I suspend, on resume, it looks like the mute led flashes back on for a split second, then turns off. Maybe the resume code is overwriting the register? I'll start digging through that code tonight.
Ryan
On Thu, Apr 30, 2009 at 1:53 AM, Takashi Iwai tiwai@suse.de wrote:
At Thu, 30 Apr 2009 00:43:02 -0400, Ryan Dunn wrote:
Thanks Lee,
I grabbed the kernel source from the repos and started perusing the
code. It
looks like when the module is installed, the call chain starts in
atiixp.c and
moves into ac97/ac97_codec.c via a call to snd_ac97_tune_hardware()
which
ultimately twiddles a bit in the AC97_POWERDOWN register. I'm not sure
yet if
the suspend method properly saves this register on suspend or not;
It should. snd_ac97_resume() writes the cached value. Check /proc/asound/card0/codec97#0/ac97#*+regs files before and after suspend. You can change the codec register value directly via proc file, e.g. # echo 0x12 0x1234 > /proc/asound/card0/codec97#0/ac97#0-0+regs (only when you build with the debug option).
The bit 0x8000 of the power-down register should correspond to LED.
Takashi
or if the same call chain needs to be repeated on a resume. I appreciate your
quick and
relevant answer. When I get time, I'll start adding some debug
printouts to
try this out.
Also, the struct ac97_quirk ac97_quirks[] array has a couple of entries
in it
currently (in the atiixp.c file). How do I get the subvender/subdevice
id's
for my system (I didn't seem to have luck with lspci, unless I'm not
looking
in the right fields)? Would it be possible to edit the driver so that
an
explict module option would not be needed on my hardware?
Thanks, Ryan
On Wed, Apr 29, 2009 at 10:42 PM, Lee Revell rlrevell@joe-job.com
wrote:
On Wed, Apr 29, 2009 at 12:06 AM, Ryan Dunn <oryandunn.ml@gmail.com wrote: > I have a Compaq V2000 laptop with the ATI IXP chipset and make useof
the > snd-atiixp module. The laptop needs to have the ac97_quirk optionset
to 7 > to enable the mute LED. After setting this option in the > /etc/modprobe.d/options.conf file, the mute LED works after areboot.
> However, after a resume from suspend or hibernate, the mute LEDdoes not
> work. The alsa-info.sh script reports that the quirk is still setwith
7. > At this point a reboot will NOT fix the problem. The only way tofix it
is > to remove and reinsert the module with modprobe. > > This is on a fresh Ubuntu 9.04 install. Any ideas? I wasn't sureif
this > should be reported here or on the kernel list. I saw a similarissue
with > the ac97_quirk on the kernel list and they were referred here.I'm a
> software developer, so I'd be willing to try ideas/possiblesolutions if
you > have them. Get the Ubuntu source code for the package that owns that ALSAdriver,
find the place in the driver's initialization code whereac97_quirk=7
is handled, then check the driver's suspend and resume callbacks and make sure the suspend callback is correctly saving the LED state and that the resume callback is re-initializing the LED from the saved state in the same way the init code does. grep and printk() areyour
friends ;-) HTH, Lee
Alsa-devel mailing list Alsa-devel@alsa-project.org http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
I apologize for the list spam. I didn't realize that this bit is only set when the led is on. So with that here are diffs of the regs with and without mute enabled, before and after suspend. Before suspend: diff regs_w-quirk.txt regs_w-quirk+mute.txt 2c2 < 0:02 = 1e1e ---
0:02 = 9e1e
20c20 < 0:26 = 000f ---
0:26 = 800f
After resume: diff regs_w-quirk_after_resume.txt regs_w-quirk_after_resume+mute.txt 2c2 < 0:02 = 1e1e ---
0:02 = 9e1e
20c20 < 0:26 = 000f ---
0:26 = 800f
So it looks as if the registers are all ok after a resume, but the LED doesn't turn back on.
Ryan
On Thu, Apr 30, 2009 at 10:12 PM, Ryan Dunn oryandunn.ml@gmail.com wrote:
Here is a diff of the registers before & after resume: diff regs_before_suspend.txt regs_after_suspend.txt 58c58
< 0:72 = 0000
0:72 = 5000
In the ac97_codec.h file, AC97_POWERDOWN = 0x26. Does this respond to 0:26 in the regs file? If so, it doesn't appear that bit 0x8000 is ever set, but the mute LED works.
On Thu, Apr 30, 2009 at 7:49 AM, Ryan Dunn oryandunn.ml@gmail.com wrote:
Thanks Takashi,
I did a bit more testing last night. The LED does work after a reboot (I don't know what I was doing that I thought it didn't). It also retains state after a reboot if I muted it before the reboot. One thing I noticed, if I have it muted when I suspend, on resume, it looks like the mute led flashes back on for a split second, then turns off. Maybe the resume code is overwriting the register? I'll start digging through that code tonight.
Ryan
On Thu, Apr 30, 2009 at 1:53 AM, Takashi Iwai tiwai@suse.de wrote:
At Thu, 30 Apr 2009 00:43:02 -0400, Ryan Dunn wrote:
Thanks Lee,
I grabbed the kernel source from the repos and started perusing the
code. It
looks like when the module is installed, the call chain starts in
atiixp.c and
moves into ac97/ac97_codec.c via a call to snd_ac97_tune_hardware()
which
ultimately twiddles a bit in the AC97_POWERDOWN register. I'm not sure
yet if
the suspend method properly saves this register on suspend or not;
It should. snd_ac97_resume() writes the cached value. Check /proc/asound/card0/codec97#0/ac97#*+regs files before and after suspend. You can change the codec register value directly via proc file, e.g. # echo 0x12 0x1234 > /proc/asound/card0/codec97#0/ac97#0-0+regs (only when you build with the debug option).
The bit 0x8000 of the power-down register should correspond to LED.
Takashi
or if the same call chain needs to be repeated on a resume. I appreciate your
quick and
relevant answer. When I get time, I'll start adding some debug
printouts to
try this out.
Also, the struct ac97_quirk ac97_quirks[] array has a couple of entries
in it
currently (in the atiixp.c file). How do I get the subvender/subdevice
id's
for my system (I didn't seem to have luck with lspci, unless I'm not
looking
in the right fields)? Would it be possible to edit the driver so that
an
explict module option would not be needed on my hardware?
Thanks, Ryan
On Wed, Apr 29, 2009 at 10:42 PM, Lee Revell rlrevell@joe-job.com
wrote:
On Wed, Apr 29, 2009 at 12:06 AM, Ryan Dunn <oryandunn.ml@gmail.com>
wrote: > I have a Compaq V2000 laptop with the ATI IXP chipset and makeuse of
the > snd-atiixp module. The laptop needs to have the ac97_quirkoption set
to 7 > to enable the mute LED. After setting this option in the > /etc/modprobe.d/options.conf file, the mute LED works after areboot.
> However, after a resume from suspend or hibernate, the mute LEDdoes not
> work. The alsa-info.sh script reports that the quirk is stillset with
7. > At this point a reboot will NOT fix the problem. The only way tofix it
is > to remove and reinsert the module with modprobe. > > This is on a fresh Ubuntu 9.04 install. Any ideas? I wasn'tsure if
this > should be reported here or on the kernel list. I saw a similarissue
with > the ac97_quirk on the kernel list and they were referred here.I'm a
> software developer, so I'd be willing to try ideas/possiblesolutions if
you > have them. Get the Ubuntu source code for the package that owns that ALSAdriver,
find the place in the driver's initialization code whereac97_quirk=7
is handled, then check the driver's suspend and resume callbacksand
make sure the suspend callback is correctly saving the LED stateand
that the resume callback is re-initializing the LED from the saved state in the same way the init code does. grep and printk() areyour
friends ;-) HTH, Lee
Alsa-devel mailing list Alsa-devel@alsa-project.org http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
At Thu, 30 Apr 2009 22:24:35 -0400, Ryan Dunn wrote:
I apologize for the list spam. I didn't realize that this bit is only set when the led is on. So with that here are diffs of the regs with and without mute enabled, before and after suspend. Before suspend: diff regs_w-quirk.txt regs_w-quirk+mute.txt 2c2
< 0:02 = 1e1e
0:02 = 9e1e
20c20
< 0:26 = 000f
0:26 = 800f
After resume: diff regs_w-quirk_after_resume.txt regs_w-quirk_after_resume+mute.txt 2c2
< 0:02 = 1e1e
0:02 = 9e1e
20c20
< 0:26 = 000f
0:26 = 800f
So it looks as if the registers are all ok after a resume, but the LED doesn't turn back on.
Then I'd say it's rather a BIOS problem.
Takashi
Thanks Takashi, I appreciate your help so far. I would normally agree with you that it is a BIOS issue, however, the machine is a dualboot with WinXP, where the mute LED works properly after a resume. Do you have any other suggestions on where the issue may be? I hate to keep bugging the list for such a small issue, but I'm not familiar with the code at all. When I get more free time, I'll try to trace through the resume path to see if I can find anything. How does the driver actually toggle the LED? Is the register a physical register on the card that the hardware reads? Or does the code do something else to toggle the LED?
Ryan
On Mon, May 4, 2009 at 3:02 AM, Takashi Iwai tiwai@suse.de wrote:
At Thu, 30 Apr 2009 22:24:35 -0400, Ryan Dunn wrote:
I apologize for the list spam. I didn't realize that this bit is only
set
when the led is on. So with that here are diffs of the regs with and
without
mute enabled, before and after suspend. Before suspend: diff regs_w-quirk.txt regs_w-quirk+mute.txt 2c2
< 0:02 = 1e1e
0:02 = 9e1e
20c20
< 0:26 = 000f
0:26 = 800f
After resume: diff regs_w-quirk_after_resume.txt regs_w-quirk_after_resume+mute.txt 2c2
< 0:02 = 1e1e
0:02 = 9e1e
20c20
< 0:26 = 000f
0:26 = 800f
So it looks as if the registers are all ok after a resume, but the LED
doesn't
turn back on.
Then I'd say it's rather a BIOS problem.
Takashi
At Mon, 4 May 2009 20:21:24 -0400, Ryan Dunn wrote:
Thanks Takashi, I appreciate your help so far. I would normally agree with you that it is a BIOS issue, however, the machine is a dualboot with WinXP, where the mute LED works properly after a resume. Do you have any other suggestions on where the issue may be? I hate to keep bugging the list for such a small issue, but I'm not familiar with the code at all. When I get more free time, I'll try to trace through the resume path to see if I can find anything. How does the driver actually toggle the LED? Is the register a physical register on the card that the hardware reads? Or does the code do something else to toggle the LED?
To be sure, you can toggle the bit of register 0x26 manually by writing the proc file. If it doesn't work, it's a BIOS issue indeed.
However, there might be some fix / workaround in the suspend part. So, after confirming that the register change doesn't work, report it on either LKML or kernel bugzilla.
Takashi
Ryan
On Mon, May 4, 2009 at 3:02 AM, Takashi Iwai tiwai@suse.de wrote:
At Thu, 30 Apr 2009 22:24:35 -0400, Ryan Dunn wrote: > > I apologize for the list spam. I didn't realize that this bit is only set > when the led is on. So with that here are diffs of the regs with and without > mute enabled, before and after suspend. > Before suspend: > diff regs_w-quirk.txt regs_w-quirk+mute.txt > 2c2 > < 0:02 = 1e1e > --- > > 0:02 = 9e1e > 20c20 > < 0:26 = 000f > --- > > 0:26 = 800f > > After resume: > diff regs_w-quirk_after_resume.txt regs_w-quirk_after_resume+mute.txt > 2c2 > < 0:02 = 1e1e > --- > > 0:02 = 9e1e > 20c20 > < 0:26 = 000f > --- > > 0:26 = 800f > > So it looks as if the registers are all ok after a resume, but the LED doesn't > turn back on. Then I'd say it's rather a BIOS problem. Takashi
participants (3)
-
Lee Revell -
Ryan Dunn -
Takashi Iwai