Skip to contentSkip to conversation

intel8x0 ACPI suspend/resume bug

DISCUSSION christophe barbe 2 participants 7 messages Jun 28, 2005 – Jun 29, 2005

lore
Share & tools
lore
#0 lore raw

Hi,

On my IBM X40, the sound outputs become crappy after a suspend/resume cycle. I am convinced that it's a gain value that goes wrong (but I could be wrong) as most of the time one side is not outputing sound anymore (only statics) but I can sometimes get this channel back by lowering its volume (so I think it's a kind of hardware protection that cuts the channel off if the gain is too high).

Anyway I wrote every details in my bug report:
https://bugtrack.alsa-project.org/alsa-bug/view.php?id=917
and I would really appreciate if someone could help me hunt this bug. I tried everything that came to my mind and don't know what to try now.

NOTE: The most recent alsa driver I tried is
alsa-driver-1.0.9rc4a.tar.bz2.

Thanks,
Christophe

Signature · 10 lines
Christophe Barb� <[email protected]>
GnuPG FingerPrint: E0F6 FADF 2A5C F072 6AF8  F67A 8F45 2F1E D72C B41E



-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
#1 lore raw

At Tue, 28 Jun 2005 11:06:39 -0400,

Quoted 11 lines · christophe barbe

Hi,

On my IBM X40, the sound outputs become crappy after a suspend/resume cycle. I am convinced that it's a gain value that goes wrong (but I could be wrong) as most of the time one side is not outputing sound anymore (only statics) but I can sometimes get this channel back by lowering its volume (so I think it's a kind of hardware protection that cuts the channel off if the gain is too high).

Quoted 7 lines

Anyway I wrote every details in my bug report:

   https://bugtrack.alsa-project.org/alsa-bug/view.php?id=917

and I would really appreciate if someone could help me hunt this bug. I tried everything that came to my mind and don't know what to try now.

NOTE: The most recent alsa driver I tried is
alsa-driver-1.0.9rc4a.tar.bz2.

First of all, use alsa-1.0.9b, just to be sure.

Then get the dump of /proc/asound/card0/codec97#*/* files before and after the suspend, and compare them.

Takashi

-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies from IBM. Find simple to follow Roadmaps, straightforward articles, informative Webcasts and more! Get everything you need to get up to speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click

#2 lore raw
Quoted 1 line · Takashi Iwai

First of all, use alsa-1.0.9b, just to be sure.

I will try it asap.

Quoted 2 lines

Then get the dump of /proc/asound/card0/codec97#*/* files before and after the suspend, and compare them.

I did this and the result is in the bug report. Let me quote myself:

I saved all proc files before and after a suspend/resume cycle and look at the diff:

In /proc/asound/card0/intel8x0
I have the following diff:
before: Global control : 0xc0000002
after: Global control : 0x00000002

/proc/asound/card0/pcm0p/sub0/[status,sw_params,hw_params] contains "closed" after while they have more content before but I suspect it's normal.
And the file /proc/asound/card0/pcm0p/sub0/info contains the same in both case except the value of subdevices_avail which is 0 before and 1 before.

Finally I have the following change in the registers:

before:
0:10 = 8989
0:12 = 9f9f
0:14 = 0000
0:16 = 9f9f
0:18 = 1010
0:1a = 0000
0:1c = 8080
...
0:76 = 2010

after:
0:10 = 8909
0:12 = 9f1f
0:14 = 0000
0:16 = 9f1f
0:18 = 1010
0:1a = 0000
0:1c = 8000
...
0:76 = 0010

I will try to figure out the meaning of the global control and the registers. The sound output was perfect before and crappy after (symmetric for once, but the more I increase the pcm volume the crappier it becomes until a certain point where I get only sound from the left channel).

For the control word, the meaning of the most significant c seems to be related to the s/dif output which is strange especially since I don't have a digital output AFAIK.

#define ICH_REG_GLOB_CNT 0x2c /* dword - global control */ #define ICH_PCM_SPDIF_MASK 0xc0000000 /* s/pdif pcm slot mask (ICH4) */
...
#define ICH_PCM_SPDIF_78 0x40000000 /* s/pdif pcm on slots 7&8 */ #define ICH_PCM_SPDIF_69 0x80000000 /* s/pdif pcm on slots 6&9 */

Thanks,
Christophe

Quoted 2 lines

Takashi

-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies from IBM. Find simple to follow Roadmaps, straightforward articles, informative Webcasts and more! Get everything you need to get up to speed, fast. http://ads.osdn.com/?ad_idt77&alloc_id492&op=click

#3 lore raw
Quoted 4 lines · christophe barbe

Le mercredi 29 juin 2005 ?? 11:31 +0200, Takashi Iwai a ??crit :

Quoted 1 line

First of all, use alsa-1.0.9b, just to be sure.

I will try it asap.

Done with same results.

Quoted 8 lines

I saved all proc files before and after a suspend/resume cycle and look at the diff:

In /proc/asound/card0/intel8x0
I have the following diff:
before: Global control : 0xc0000002
after: Global control : 0x00000002

Quoted 8 lines

/proc/asound/card0/pcm0p/sub0/[status,sw_params,hw_params] contains "closed" after while they have more content before but I suspect it's normal.
And the file /proc/asound/card0/pcm0p/sub0/info contains the same in both case except the value of subdevices_avail which is 0 before and 1 before.

Finally I have the following change in the registers:

I downloaded the datasheet for Intel 82801DB which I think is what I have and found what I believe to be the registers of the file codec97#0/ac97#0-0+regs and found the following

Quoted 11 lines

before:
0:10 = 8989
0:12 = 9f9f
0:14 = 0000
0:16 = 9f9f
0:18 = 1010
0:1a = 0000
0:1c = 8080
...
0:76 = 2010

Quoted 2 lines

after:
0:10 = 8909

Line in volume

Quoted 1 line

0:12 = 9f1f

CD volume

Quoted 2 lines

0:14 = 0000
0:16 = 9f1f

Aux in volume

Quoted 3 lines

0:18 = 1010
0:1a = 0000
0:1c = 8000

Record gain

Quoted 2 lines

...
0:76 = 0010

This one seems to be undocumented.

I have not yet find how the volume is encoded but I suspect it's two 8-bit words one for the left and one for the right which seems to match the fact that I usually loose the sound on one side. But all registers seems to be unrelated to mp3-to-pcm output.

Any idea for the register 76 ?

Christophe

Signature · 10 lines
Christophe Barbe <[email protected]>
GnuPG FingerPrint: E0F6 FADF 2A5C F072 6AF8  F67A 8F45 2F1E D72C B41E



-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
#4 lore raw
Quoted 1 line · christophe barbe

I have not yet find how the volume is encoded but I suspect it's two 8-bit words one for the left and one for the right which seems to match the fact that I usually loose the sound on one side. But all registers seems to be unrelated to mp3-to-pcm output.

I found the description of the registers in the AC'97 spec and it says that the bit7 of the registers 10,12,16 and 1c should read 0 (as they are marked X in the table).
Is this correct?

Interestingly this bits read 1 before a suspend/resume cycle but 0 after. This is really strange.

Christophe

Signature · 10 lines
Christophe Barb� <[email protected]>
GnuPG FingerPrint: E0F6 FADF 2A5C F072 6AF8  F67A 8F45 2F1E D72C B41E



-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
#5 lore raw
In reply to Christophe Barbe
Quoted 7 lines · christophe barbe

I saved all proc files before and after a suspend/resume cycle and look at the diff:

In /proc/asound/card0/intel8x0
I have the following diff:
before: Global control : 0xc0000002
after: Global control : 0x00000002

I think that the real problem is here as the intel datasheet says for the bits 31:30 :

"S/PDIF Slot Map (SSM) -- R/W. If the run/pause bus master bit (bit 0 of offset 2Bh) is set, then the value in these bits indicate which slots S/PDIF data is transmitted on. Software must ensure that the programming here does not conflict with the PCM channels being used. If there is a conflict, unpredictable behavior will result - the hardware will not check for a conflict.

      00 = Reserved
      01 = Slots 7 and 8
      10 = Slots 6 and 9
      11 = Slots 10 and 11"

And after a suspend/resume cycle, I have 11 which does not make sense since I don't have a S/PDIF output.

Is there an easy way to set this bits to 00?

Christophe

Signature · 10 lines
Christophe Barb� <[email protected]>
GnuPG FingerPrint: E0F6 FADF 2A5C F072 6AF8  F67A 8F45 2F1E D72C B41E



-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
#6 lore raw
Quoted 23 lines · christophe barbe
Quoted 7 lines · christophe barbe

I saved all proc files before and after a suspend/resume cycle and look at the diff:

In /proc/asound/card0/intel8x0
I have the following diff:
before: Global control : 0xc0000002
after: Global control : 0x00000002

I think that the real problem is here as the intel datasheet says for the bits 31:30 :

"S/PDIF Slot Map (SSM) -- R/W. If the run/pause bus master bit (bit 0 of offset 2Bh) is set, then the value in these bits indicate which slots S/PDIF data is transmitted on. Software must ensure that the programming here does not conflict with the PCM channels being used. If there is a conflict, unpredictable behavior will result - the hardware will not check for a conflict.

      00 = Reserved
      01 = Slots 7 and 8
      10 = Slots 6 and 9
      11 = Slots 10 and 11"
Quoted 4 lines

And after a suspend/resume cycle, I have 11 which does not make sense since I don't have a S/PDIF output.

Is there an easy way to set this bits to 00?

Obviously I confused myself, the bits should be 11 and after the suspend/resume cycle are 00 which is not good.

Christophe

Signature · 10 lines
Christophe Barbe <[email protected]>
GnuPG FingerPrint: E0F6 FADF 2A5C F072 6AF8  F67A 8F45 2F1E D72C B41E



-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click