[alsa-devel] [ASoC] list corruption during AC97 registration
Hi Folks,
I hit the following oops (on 2.6.21) when ASoC registers an AC97 codec. Happens both on MIPS and SH, for example the mips oops:
asoc: AC97 HiFi <-> au1xpsc-ac97-1 mapping ok list_add corruption. prev->next should be next (80488f60), but was 865ffad8. (prev=865ffad8). <snip> Call Trace: [<80271290>] __list_add+0x68/0x94 [<802ac0e4>] bus_attach_device+0x48/0x78 [<802aa4c4>] device_add+0x3d8/0x844 [<c00882a8>] snd_soc_register_card+0x1d0/0x2c8 [snd_soc_core] [<c00b3268>] ac97_soc_probe+0x17c/0x1e4 [snd_soc_ac97] [<c0088f48>] soc_probe+0xb0/0x1ac [snd_soc_core] [<802ad410>] really_probe+0x148/0x268 [<802ac1cc>] bus_for_each_drv+0x50/0xa0 [<802ad758>] device_attach+0xe4/0x194 [<802ac0cc>] bus_attach_device+0x30/0x78 [<802aa4c4>] device_add+0x3d8/0x844 [<802aff24>] platform_device_add+0x13c/0x198 [<c00b71c8>] exm1200_mobo_ac97_init+0x1c8/0x1fc [snd_soc_exmmb_ac97] [<801478d0>] sys_init_module+0x1448/0x15b8 [<8010bcc0>] stack_done+0x20/0x3c
System.map says: 80488e60 D ac97_bus_type
As I said, it happens on both MIPS and SH (with machine and PCM drivers written by myself). With CONFIG_DEBUG_LIST disabled the code works and plays sound, but still I'd like to know what causes this list corruption.
Thanks for any hints!
This patch is for ATI RS780 and R600 HDMI support. Signed-off-by: Wolke Liu <wolke.liu <at> amd.com>
diff -r 67cb6a84c18a pci/hda/hda_intel.c --- a/pci/hda/hda_intel.c Fri Apr 20 16:11:43 2007 +0200 +++ b/pci/hda/hda_intel.c Fri Apr 27 04:56:54 2007 -0400 @@ -88,6 +88,8 @@ MODULE_SUPPORTED_DEVICE("{{Intel, ICH6}, "{ATI, SB600}," "{ATI, RS600}," "{ATI, RS690}," + "{ATI, RS780}," + "{ATI, R600}," "{VIA, VT8251}," "{VIA, VT8237A}," "{SiS, SIS966}," @@ -1770,6 +1772,8 @@ static struct pci_device_id azx_ids[] = { 0x1002, 0x4383, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATI }, /* ATI SB600 */ { 0x1002, 0x793b, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI RS600 HDMI */ { 0x1002, 0x7919, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI RS690 HDMI */ + { 0x1002, 0x960c, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI RS780 HDMI */ + { 0x1002, 0xaa00, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI R600 HDMI */ { 0x1106, 0x3288, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_VIA }, /* VIA VT8251/VT8237A */ { 0x1039, 0x7502, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_SIS }, /* SIS966 */ { 0x10b9, 0x5461, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ULI }, /* ULI M5461 */ diff -r 67cb6a84c18a pci/hda/patch_atihdmi.c --- a/pci/hda/patch_atihdmi.c Fri Apr 20 16:11:43 2007 +0200 +++ b/pci/hda/patch_atihdmi.c Fri Apr 27 04:57:29 2007 -0400 @@ -173,6 +173,7 @@ static int patch_atihdmi(struct hda_code */ struct hda_codec_preset snd_hda_preset_atihdmi[] = { { .id = 0x1002793c, .name = "ATI RS600 HDMI", .patch = patch_atihdmi }, - { .id = 0x1002791a, .name = "ATI RS690 HDMI", .patch = patch_atihdmi }, + { .id = 0x1002791a, .name = "ATI RS690/780 HDMI", .patch = patch_atihdmi }, + { .id = 0x1002aa01, .name = "ATI R600 HDMI", .patch = patch_atihdmi }, {} /* terminator */ };
At Fri, 27 Apr 2007 18:04:47 +0800, Wolke Liu wrote:
This patch is for ATI RS780 and R600 HDMI support. Signed-off-by: Wolke Liu <wolke.liu <at> amd.com>
I applied it manually because the embedded patch seems broken by your MUA. Please check the HG tree later whether it works for you.
Thanks!
Takashi
diff -r 67cb6a84c18a pci/hda/hda_intel.c --- a/pci/hda/hda_intel.c Fri Apr 20 16:11:43 2007 +0200 +++ b/pci/hda/hda_intel.c Fri Apr 27 04:56:54 2007 -0400 @@ -88,6 +88,8 @@ MODULE_SUPPORTED_DEVICE("{{Intel, ICH6}, "{ATI, SB600}," "{ATI, RS600}," "{ATI, RS690},"
"{ATI, RS780},""{ATI, R600}," "{VIA, VT8251}," "{VIA, VT8237A}," "{SiS, SIS966},"@@ -1770,6 +1772,8 @@ static struct pci_device_id azx_ids[] = { 0x1002, 0x4383, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATI }, /* ATI SB600 */ { 0x1002, 0x793b, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI RS600 HDMI */ { 0x1002, 0x7919, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI RS690 HDMI */
{ 0x1002, 0x960c, PCI_ANY_ID, PCI_ANY_ID, 0, 0,AZX_DRIVER_ATIHDMI }, /* ATI RS780 HDMI */
{ 0x1002, 0xaa00, PCI_ANY_ID, PCI_ANY_ID, 0, 0,AZX_DRIVER_ATIHDMI }, /* ATI R600 HDMI */ { 0x1106, 0x3288, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_VIA }, /* VIA VT8251/VT8237A */ { 0x1039, 0x7502, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_SIS }, /* SIS966 */ { 0x10b9, 0x5461, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ULI }, /* ULI M5461 */ diff -r 67cb6a84c18a pci/hda/patch_atihdmi.c --- a/pci/hda/patch_atihdmi.c Fri Apr 20 16:11:43 2007 +0200 +++ b/pci/hda/patch_atihdmi.c Fri Apr 27 04:57:29 2007 -0400 @@ -173,6 +173,7 @@ static int patch_atihdmi(struct hda_code */ struct hda_codec_preset snd_hda_preset_atihdmi[] = { { .id = 0x1002793c, .name = "ATI RS600 HDMI", .patch = patch_atihdmi },
{ .id = 0x1002791a, .name = "ATI RS690 HDMI", .patch =patch_atihdmi },
{ .id = 0x1002791a, .name = "ATI RS690/780 HDMI", .patch =patch_atihdmi },
{ .id = 0x1002aa01, .name = "ATI R600 HDMI", .patch =patch_atihdmi }, {} /* terminator */ };
Thank you!
Regards, Wolke -----Original Message----- From: alsa-devel-bounces@alsa-project.org [mailto:alsa-devel-bounces@alsa-project.org] On Behalf Of Takashi Iwai Sent: Friday, April 27, 2007 6:22 PM To: Wolke Liu Cc: alsa-devel@alsa-project.org; Yufei Ma Subject: Re: [alsa-devel] [patch]Add ATI RS780,R600 HDMI audio support
At Fri, 27 Apr 2007 18:04:47 +0800, Wolke Liu wrote:
This patch is for ATI RS780 and R600 HDMI support. Signed-off-by: Wolke Liu <wolke.liu <at> amd.com>
I applied it manually because the embedded patch seems broken by your MUA. Please check the HG tree later whether it works for you.
Thanks!
Takashi
diff -r 67cb6a84c18a pci/hda/hda_intel.c --- a/pci/hda/hda_intel.c Fri Apr 20 16:11:43 2007 +0200 +++ b/pci/hda/hda_intel.c Fri Apr 27 04:56:54 2007 -0400 @@ -88,6 +88,8 @@ MODULE_SUPPORTED_DEVICE("{{Intel, ICH6}, "{ATI, SB600}," "{ATI, RS600}," "{ATI, RS690},"
"{ATI, RS780},""{ATI, R600}," "{VIA, VT8251}," "{VIA, VT8237A}," "{SiS, SIS966},"@@ -1770,6 +1772,8 @@ static struct pci_device_id azx_ids[] = { 0x1002, 0x4383, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATI }, /* ATI SB600 */ { 0x1002, 0x793b, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI RS600 HDMI */ { 0x1002, 0x7919, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ATIHDMI }, /* ATI RS690 HDMI */
{ 0x1002, 0x960c, PCI_ANY_ID, PCI_ANY_ID, 0, 0,AZX_DRIVER_ATIHDMI }, /* ATI RS780 HDMI */
{ 0x1002, 0xaa00, PCI_ANY_ID, PCI_ANY_ID, 0, 0,AZX_DRIVER_ATIHDMI }, /* ATI R600 HDMI */ { 0x1106, 0x3288, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_VIA }, /* VIA VT8251/VT8237A */ { 0x1039, 0x7502, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_SIS }, /* SIS966 */ { 0x10b9, 0x5461, PCI_ANY_ID, PCI_ANY_ID, 0, 0, AZX_DRIVER_ULI }, /* ULI M5461 */ diff -r 67cb6a84c18a pci/hda/patch_atihdmi.c --- a/pci/hda/patch_atihdmi.c Fri Apr 20 16:11:43 2007 +0200 +++ b/pci/hda/patch_atihdmi.c Fri Apr 27 04:57:29 2007 -0400 @@ -173,6 +173,7 @@ static int patch_atihdmi(struct hda_code */ struct hda_codec_preset snd_hda_preset_atihdmi[] = { { .id = 0x1002793c, .name = "ATI RS600 HDMI", .patch = patch_atihdmi },
{ .id = 0x1002791a, .name = "ATI RS690 HDMI", .patch =patch_atihdmi },
{ .id = 0x1002791a, .name = "ATI RS690/780 HDMI", .patch =patch_atihdmi },
{ .id = 0x1002aa01, .name = "ATI R600 HDMI", .patch =patch_atihdmi }, {} /* terminator */ };
_______________________________________________ Alsa-devel mailing list Alsa-devel@alsa-project.org http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
On Thu, 2007-04-26 at 19:16 +0200, Manuel Lauss wrote:
Hi Folks,
I hit the following oops (on 2.6.21) when ASoC registers an AC97 codec. Happens both on MIPS and SH, for example the mips oops:
asoc: AC97 HiFi <-> au1xpsc-ac97-1 mapping ok list_add corruption. prev->next should be next (80488f60), but was 865ffad8. (prev=865ffad8).
<snip> Call Trace: [<80271290>] __list_add+0x68/0x94 [<802ac0e4>] bus_attach_device+0x48/0x78 [<802aa4c4>] device_add+0x3d8/0x844 [<c00882a8>] snd_soc_register_card+0x1d0/0x2c8 [snd_soc_core] [<c00b3268>] ac97_soc_probe+0x17c/0x1e4 [snd_soc_ac97] [<c0088f48>] soc_probe+0xb0/0x1ac [snd_soc_core] [<802ad410>] really_probe+0x148/0x268 [<802ac1cc>] bus_for_each_drv+0x50/0xa0 [<802ad758>] device_attach+0xe4/0x194 [<802ac0cc>] bus_attach_device+0x30/0x78 [<802aa4c4>] device_add+0x3d8/0x844 [<802aff24>] platform_device_add+0x13c/0x198 [<c00b71c8>] exm1200_mobo_ac97_init+0x1c8/0x1fc [snd_soc_exmmb_ac97] [<801478d0>] sys_init_module+0x1448/0x15b8 [<8010bcc0>] stack_done+0x20/0x3c
System.map says: 80488e60 D ac97_bus_type
As I said, it happens on both MIPS and SH (with machine and PCM drivers written by myself). With CONFIG_DEBUG_LIST disabled the code works and plays sound, but still I'd like to know what causes this list corruption.
Thanks for any hints!
Thanks for reporting this. I'll add to ALSA bugtracker (after some new categories have been added) and fix on early next week when I have access to my AC97 hardware.
Fwiw, I'd start by looking at soc_ac97_dev_register()
HTH
Liam
Hi Liam,
On Fri, Apr 27, 2007 at 11:21:18AM +0100, Liam Girdwood wrote:
[<802ac0e4>] bus_attach_device+0x48/0x78 [<802aa4c4>] device_add+0x3d8/0x844 [<c00882a8>] snd_soc_register_card+0x1d0/0x2c8 [snd_soc_core]
System.map says: 80488e60 D ac97_bus_type
Fwiw, I'd start by looking at soc_ac97_dev_register()
Do I need call a special ASoC function to create an AC97 bus, like the snd_soc_new_ac97_codec() function, from board code?
Thanks,
On Fri, 2007-04-27 at 12:47 +0200, Manuel Lauss wrote:
Hi Liam,
On Fri, Apr 27, 2007 at 11:21:18AM +0100, Liam Girdwood wrote:
[<802ac0e4>] bus_attach_device+0x48/0x78 [<802aa4c4>] device_add+0x3d8/0x844 [<c00882a8>] snd_soc_register_card+0x1d0/0x2c8 [snd_soc_core]
System.map says: 80488e60 D ac97_bus_type
Fwiw, I'd start by looking at soc_ac97_dev_register()
Do I need call a special ASoC function to create an AC97 bus, like the snd_soc_new_ac97_codec() function, from board code?
ac97_bus_type is declared in ac97_bus.c, so there is no function to create it. The generic AC97 driver (codecs/ac97.c) does all the ac97 init for you in ac97_soc_probe (e.g. create PCM's, register with AC97 bus, create new AC97 mixer, register card), although I've never turned on the list debug here.
Btw, is there anyway you can post your code ? It should make debugging a little easier and I can host it on my dev branch.
Liam
Liam Girdwood wrote:
On Fri, 2007-04-27 at 12:47 +0200, Manuel Lauss wrote:
Hi Liam,
On Fri, Apr 27, 2007 at 11:21:18AM +0100, Liam Girdwood wrote:
[<802ac0e4>] bus_attach_device+0x48/0x78 [<802aa4c4>] device_add+0x3d8/0x844 [<c00882a8>] snd_soc_register_card+0x1d0/0x2c8 [snd_soc_core]
System.map says: 80488e60 D ac97_bus_type
Fwiw, I'd start by looking at soc_ac97_dev_register()
Do I need call a special ASoC function to create an AC97 bus, like the snd_soc_new_ac97_codec() function, from board code?
ac97_bus_type is declared in ac97_bus.c, so there is no function to create it. The generic AC97 driver (codecs/ac97.c) does all the ac97 init for you in ac97_soc_probe (e.g. create PCM's, register with AC97 bus, create new AC97 mixer, register card), although I've never turned on the list debug here.
the WM9712 code calls the above mentioned function, I concluded the generic ac97 driver should do so too?
Btw, is there anyway you can post your code ? It should make debugging a little easier and I can host it on my dev branch.
Sure, I'd be very happy if you could take a closer look: for au1xxx: http://mlau.at/files/asoc-au1xpsc-v1.patch http://mlau.at/files/au1x-dbdma-audio-features.patch
for the SH7760: http://mlau.at/files/asoc-sh7760-v1.patch http://mlau.at/files/dmabrg.patch
I recorded the oopses here: http://mlau.at/files/asoc-ac97-list-corruption.txt
Thanks!
participants (4)
-
Liam Girdwood -
Manuel Lauss -
Takashi Iwai -
Wolke Liu