[alsa-devel] Aureal issues on resume
(added recipients) -------- Original Message -------- Subject: Aureal issues on resume Date: Fri, 02 Sep 2011 13:41:42 +0200 From: Anders Eriksson <aeriksson@fastmail.fm> To: linux-kernel@vger.kernel.org, linux-sound@vger.kernel.org Hi, upon resuming from hibernation, I see this in my log: Sep 2 11:15:20 tippex kernel: [72436.287020] vortex_adb_addroutes: unending route! 0x10 Sep 2 11:15:20 tippex kernel: [72436.287127] vortex_adb_addroutes: unending route! 0x838 Sep 2 11:15:20 tippex kernel: [72436.287277] vortex_adb_addroutes: unending route! 0x11 Sep 2 11:15:20 tippex kernel: [72436.287418] vortex_adb_addroutes: unending route! 0x8b9 Sep 2 11:15:20 tippex kernel: [72436.292037] Vortex: vortex_fifo_setadbctrl fail Sep 2 11:15:55 tippex kernel: [72471.403490] vortex_adb_delroutes: route not found! 0x10 Sep 2 11:15:55 tippex kernel: [72471.403596] vortex_adb_delroutes: route not found! 0x838 Sep 2 11:15:55 tippex kernel: [72471.403710] vortex_adb_delroutes: route not found! 0x11 Sep 2 11:15:55 tippex kernel: [72471.403811] vortex_adb_delroutes: route not found! 0x8b9 And the sound is no more. I need to reboot to get it back. If memory serves, it's been there since 2.6.30 or so. Any ideas what can be triggering it? lspci says it's an 00:04.0 Multimedia audio controller: Aureal Semiconductor Vortex 1 (rev 02) Subsystem: Diamond Multimedia Systems Sonic Impact A3D Flags: bus master, medium devsel, latency 64, IRQ 11 Memory at 50000000 (32-bit, non-prefetchable) [size=128K] I/O ports at 10b0 [size=8] I/O ports at 10b8 [size=8] Capabilities: [dc] Power Management version 1 Kernel driver in use: snd_au8820 Kernel modules: snd-au8820 -Anders
2011/9/2 Clemens Ladisch <clemens@ladisch.de>:
(added recipients)
-------- Original Message -------- Subject: Aureal issues on resume Date: Fri, 02 Sep 2011 13:41:42 +0200 From: Anders Eriksson <aeriksson@fastmail.fm> To: linux-kernel@vger.kernel.org, linux-sound@vger.kernel.org
Hi, upon resuming from hibernation, I see this in my log:
Sep 2 11:15:20 tippex kernel: [72436.287020] vortex_adb_addroutes: unending route! 0x10 Sep 2 11:15:20 tippex kernel: [72436.287127] vortex_adb_addroutes: unending route! 0x838 Sep 2 11:15:20 tippex kernel: [72436.287277] vortex_adb_addroutes: unending route! 0x11 Sep 2 11:15:20 tippex kernel: [72436.287418] vortex_adb_addroutes: unending route! 0x8b9 Sep 2 11:15:20 tippex kernel: [72436.292037] Vortex: vortex_fifo_setadbctrl fail Sep 2 11:15:55 tippex kernel: [72471.403490] vortex_adb_delroutes: route not found! 0x10 Sep 2 11:15:55 tippex kernel: [72471.403596] vortex_adb_delroutes: route not found! 0x838 Sep 2 11:15:55 tippex kernel: [72471.403710] vortex_adb_delroutes: route not found! 0x11 Sep 2 11:15:55 tippex kernel: [72471.403811] vortex_adb_delroutes: route not found! 0x8b9
And the sound is no more. I need to reboot to get it back.
If memory serves, it's been there since 2.6.30 or so. Any ideas what can be triggering it?
lspci says it's an 00:04.0 Multimedia audio controller: Aureal Semiconductor Vortex 1 (rev 02) Subsystem: Diamond Multimedia Systems Sonic Impact A3D Flags: bus master, medium devsel, latency 64, IRQ 11 Memory at 50000000 (32-bit, non-prefetchable) [size=128K] I/O ports at 10b0 [size=8] I/O ports at 10b8 [size=8] Capabilities: [dc] Power Management version 1 Kernel driver in use: snd_au8820 Kernel modules: snd-au8820
-Anders
Those alsa au88x0 drivers does not support suspend/resume As each audio stream is assigned different resources (e.g. SRC, mixer , ...) dynamically and the cards can play 16 mono streams or 8 stereo streams. The possible way is to save the contents of 128K memory mapped I/O ports before suspend to non-volatile memory and restore them after resume Is there any way for alsa driver to allocate non-volatile memory ? au8830 has 256 K mmio
Raymond Yau wrote:
The possible way is to save the contents of 128K memory mapped I/O ports before suspend to non-volatile memory and restore them after resume
Is there any way for alsa driver to allocate non-volatile memory ?
'Normal' memory is saved over suspend. Just use {k|v}malloc. Regards, Clemens
At Sun, 04 Sep 2011 22:32:44 +0200, Clemens Ladisch wrote:
Raymond Yau wrote:
The possible way is to save the contents of 128K memory mapped I/O ports before suspend to non-volatile memory and restore them after resume
Is there any way for alsa driver to allocate non-volatile memory ?
'Normal' memory is saved over suspend. Just use {k|v}malloc.
If it's a 128kB bulk memory, vmalloc() is a more sensible choice. Takashi
At Mon, 12 Sep 2011 09:53:28 +0200, Takashi Iwai wrote:
At Sun, 04 Sep 2011 22:32:44 +0200, Clemens Ladisch wrote:
Raymond Yau wrote:
The possible way is to save the contents of 128K memory mapped I/O ports before suspend to non-volatile memory and restore them after resume
Is there any way for alsa driver to allocate non-volatile memory ?
'Normal' memory is saved over suspend. Just use {k|v}malloc.
If it's a 128kB bulk memory, vmalloc() is a more sensible choice.
BTW, in many cases, you don't have to save/restore the whole mmio region. Instead, stop the stream, save / restore some important registers and do the similar procedure for the normal prepare callback in the resume, and restart the stream again. Takashi
2011/9/12 Takashi Iwai <tiwai@suse.de>:
At Mon, 12 Sep 2011 09:53:28 +0200, Takashi Iwai wrote:
At Sun, 04 Sep 2011 22:32:44 +0200, Clemens Ladisch wrote:
Raymond Yau wrote:
The possible way is to save the contents of 128K memory mapped I/O ports before suspend to non-volatile memory and restore them after resume
Is there any way for alsa driver to allocate non-volatile memory ?
'Normal' memory is saved over suspend. Just use {k|v}malloc.
If it's a 128kB bulk memory, vmalloc() is a more sensible choice.
BTW, in many cases, you don't have to save/restore the whole mmio region. Instead, stop the stream, save / restore some important registers and do the similar procedure for the normal prepare callback in the resume, and restart the stream again.
Takashi
The current au8830 alsa driver seem use less than 192Kbytes out of 256Kbytes mmio http://thread.gmane.org/gmane.linux.kernel/564963/focus=47907 your patch seem initialse the request queue in figure 16 of pat 6167465 and the driver need to restore mixer and equalizer mmio for au8830 in figure 13 04:02.0 Multimedia audio controller: Aureal Semiconductor Vortex 2 (rev fe) Subsystem: Aureal Semiconductor AU8830 Vortex 3D Digital Audio Processor Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx- Status: Cap+ 66MHz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx- Latency: 64 (1000ns min, 3000ns max), Cache Line Size: 32 bytes Interrupt: pin A routed to IRQ 23 Region 0: Memory at ff7c0000 (32-bit, non-prefetchable) [size=256K] Region 1: I/O ports at bc00 [size=8] Region 2: I/O ports at b880 [size=8] Capabilities: [dc] Power Management version 1 Flags: PMEClk- DSI+ D1- D2+ AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-) Status: D0 PME-Enable- DSel=0 DScale=0 PME- Kernel driver in use: au8830 Kernel modules: snd-au8830
At Fri, 16 Sep 2011 10:16:11 +0800, Raymond Yau wrote:
2011/9/12 Takashi Iwai <tiwai@suse.de>:
At Mon, 12 Sep 2011 09:53:28 +0200, Takashi Iwai wrote:
At Sun, 04 Sep 2011 22:32:44 +0200, Clemens Ladisch wrote:
Raymond Yau wrote:
The possible way is to save the contents of 128K memory mapped I/O ports before suspend to non-volatile memory and restore them after resume
Is there any way for alsa driver to allocate non-volatile memory ?
'Normal' memory is saved over suspend. Just use {k|v}malloc.
If it's a 128kB bulk memory, vmalloc() is a more sensible choice.
BTW, in many cases, you don't have to save/restore the whole mmio region. Instead, stop the stream, save / restore some important registers and do the similar procedure for the normal prepare callback in the resume, and restart the stream again.
Takashi
The current au8830 alsa driver seem use less than 192Kbytes out of 256Kbytes mmio
http://thread.gmane.org/gmane.linux.kernel/564963/focus=47907
your patch seem initialse the request queue in figure 16 of pat 6167465
and the driver need to restore mixer and equalizer mmio for au8830 in figure 13
Yeah, that's what I mentioned in the thread. Basically you'll need to restore the values assigned in hw_params() for PCM. The mixer is ac97, so it's restored by ac97_codec.c. The rest are the equalizer and a3d values to be restored separately. Takashi
2011/9/16 Takashi Iwai <tiwai@suse.de>:
At Fri, 16 Sep 2011 10:16:11 +0800, Raymond Yau wrote:
2011/9/12 Takashi Iwai <tiwai@suse.de>:
At Mon, 12 Sep 2011 09:53:28 +0200, Takashi Iwai wrote:
At Sun, 04 Sep 2011 22:32:44 +0200, Clemens Ladisch wrote:
Raymond Yau wrote:
The possible way is to save the contents of 128K memory mapped I/O ports before suspend to non-volatile memory and restore them after resume
Is there any way for alsa driver to allocate non-volatile memory ?
'Normal' memory is saved over suspend. Just use {k|v}malloc.
If it's a 128kB bulk memory, vmalloc() is a more sensible choice.
BTW, in many cases, you don't have to save/restore the whole mmio region. Instead, stop the stream, save / restore some important registers and do the similar procedure for the normal prepare callback in the resume, and restart the stream again.
Takashi
The current au8830 alsa driver seem use less than 192Kbytes out of 256Kbytes mmio
http://thread.gmane.org/gmane.linux.kernel/564963/focus=47907
your patch seem initialse the request queue in figure 16 of pat 6167465
and the driver need to restore mixer and equalizer mmio for au8830 in figure 13
Yeah, that's what I mentioned in the thread.
Basically you'll need to restore the values assigned in hw_params() for PCM.
The error messages are due to snd_vortex_suspend() call snd_vortex_adb_init() There are static routes for the codec, mixin, mixout and equalizer and dynamic routes of playback/capture streams Those VORTEX_ADB_RTBASE and VORTEX_ADB_CHNBASE mmio ports are two linked list, those vortex_route_* function are adding the routes to the tail of those linked lists so the only is to restore the routes since the driver does not keep track of the order of routes of the playback stream or capture stream to the linked list
The mixer is ac97, so it's restored by ac97_codec.c. The rest are the equalizer and a3d values to be restored separately.
I mean the hardware mixer (mixin and mixout) inside au88x0 chip. May be need to initialise mpu401_uart or joystick Is there any way to force the system to suspend, it seem that gnome power management in Fedora 10 does not provide any way to suspend while playing an audio ? ( screen saver set to 1 minutes, put the computer into sleep when inactive for 2 minutes ) Ubuntu 10.04 Live CD , you can force the system to suspend
At Sat, 17 Sep 2011 10:16:19 +0800, Raymond Yau wrote:
2011/9/16 Takashi Iwai <tiwai@suse.de>:
At Fri, 16 Sep 2011 10:16:11 +0800, Raymond Yau wrote:
2011/9/12 Takashi Iwai <tiwai@suse.de>:
At Mon, 12 Sep 2011 09:53:28 +0200, Takashi Iwai wrote:
At Sun, 04 Sep 2011 22:32:44 +0200, Clemens Ladisch wrote:
Raymond Yau wrote: > The possible way is to save the contents of 128K memory mapped I/O > ports before suspend to non-volatile memory and restore them after > resume > > Is there any way for alsa driver to allocate non-volatile memory ?
'Normal' memory is saved over suspend. Just use {k|v}malloc.
If it's a 128kB bulk memory, vmalloc() is a more sensible choice.
BTW, in many cases, you don't have to save/restore the whole mmio region. Instead, stop the stream, save / restore some important registers and do the similar procedure for the normal prepare callback in the resume, and restart the stream again.
Takashi
The current au8830 alsa driver seem use less than 192Kbytes out of 256Kbytes mmio
http://thread.gmane.org/gmane.linux.kernel/564963/focus=47907
your patch seem initialse the request queue in figure 16 of pat 6167465
and the driver need to restore mixer and equalizer mmio for au8830 in figure 13
Yeah, that's what I mentioned in the thread.
Basically you'll need to restore the values assigned in hw_params() for PCM.
The error messages are due to snd_vortex_suspend() call snd_vortex_adb_init()
There are static routes for the codec, mixin, mixout and equalizer and dynamic routes of playback/capture streams
Those VORTEX_ADB_RTBASE and VORTEX_ADB_CHNBASE mmio ports are two linked list, those vortex_route_* function are adding the routes to the tail of those linked lists
so the only is to restore the routes since the driver does not keep track of the order of routes of the playback stream or capture stream to the linked list
The mixer is ac97, so it's restored by ac97_codec.c. The rest are the equalizer and a3d values to be restored separately.
I mean the hardware mixer (mixin and mixout) inside au88x0 chip.
May be need to initialise mpu401_uart or joystick
Is there any way to force the system to suspend, it seem that gnome power management in Fedora 10 does not provide any way to suspend while playing an audio ? ( screen saver set to 1 minutes, put the computer into sleep when inactive for 2 minutes )
See files in Documentation/power directory. You may call s2ram, powersave, or whatever command directly, too. Takashi
2011/9/20 Takashi Iwai <tiwai@suse.de>:
At Sat, 17 Sep 2011 10:16:19 +0800, Raymond Yau wrote:
2011/9/16 Takashi Iwai <tiwai@suse.de>:
At Fri, 16 Sep 2011 10:16:11 +0800, Raymond Yau wrote:
2011/9/12 Takashi Iwai <tiwai@suse.de>:
At Mon, 12 Sep 2011 09:53:28 +0200, Takashi Iwai wrote:
At Sun, 04 Sep 2011 22:32:44 +0200, Clemens Ladisch wrote: > > Raymond Yau wrote: > > The possible way is to save the contents of 128K memory mapped I/O > > ports before suspend to non-volatile memory and restore them after > > resume > > > > Is there any way for alsa driver to allocate non-volatile memory ? > > 'Normal' memory is saved over suspend. Just use {k|v}malloc.
If it's a 128kB bulk memory, vmalloc() is a more sensible choice.
BTW, in many cases, you don't have to save/restore the whole mmio region. Instead, stop the stream, save / restore some important registers and do the similar procedure for the normal prepare callback in the resume, and restart the stream again.
Takashi
The current au8830 alsa driver seem use less than 192Kbytes out of 256Kbytes mmio
http://thread.gmane.org/gmane.linux.kernel/564963/focus=47907
your patch seem initialse the request queue in figure 16 of pat 6167465
and the driver need to restore mixer and equalizer mmio for au8830 in figure 13
Yeah, that's what I mentioned in the thread.
Basically you'll need to restore the values assigned in hw_params() for PCM.
The error messages are due to snd_vortex_suspend() call snd_vortex_adb_init()
There are static routes for the codec, mixin, mixout and equalizer and dynamic routes of playback/capture streams
Those VORTEX_ADB_RTBASE and VORTEX_ADB_CHNBASE mmio ports are two linked list, those vortex_route_* function are adding the routes to the tail of those linked lists
so the only is to restore the routes since the driver does not keep track of the order of routes of the playback stream or capture stream to the linked list
The mixer is ac97, so it's restored by ac97_codec.c. The rest are the equalizer and a3d values to be restored separately.
I mean the hardware mixer (mixin and mixout) inside au88x0 chip.
May be need to initialise mpu401_uart or joystick
Is there any way to force the system to suspend, it seem that gnome power management in Fedora 10 does not provide any way to suspend while playing an audio ? ( screen saver set to 1 minutes, put the computer into sleep when inactive for 2 minutes )
See files in Documentation/power directory. You may call s2ram, powersave, or whatever command directly, too.
Takashi
1)Make sure that your computer can suspend and resume normally without au88x0 sound card 2)Backup all your important data before you try the patch with your au88x0 sound card 3)Only tested with au8830 , so it may need some modification for au8820 and au8810 my au8830 ,with two streams playing at different sampling rate at the same time, can suspend with the following script and resume as expected by pressing power button with the attached patch sync echo 1 > /sys/power/pm_trace echo mem > /sys/power/state It is strange that it can resume without uncomment those SNDRV_PCM_INFO_RESUME in au88x0_pcm.c static struct snd_pcm_hardware snd_vortex_playback_hw_adb = { .info = (SNDRV_PCM_INFO_MMAP | /* SNDRV_PCM_INFO_RESUME | */ SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_INTERLEAVED | ALSA au88x0.c:131: vortex suspending... snd_au8830 0000:04:02.0: PCI INT A disabled vortex suspended ... ALSA au88x0.c:248: vortex resuming... snd_au8830 0000:04:02.0: restoring config space at offset 0x1 (was 0x2900007, writing 0x2900003) snd_au8830 0000:04:02.0: PCI INT A -> GSI 23 (level, low) -> IRQ 23 ALSA au88x0.c:350: vortex resumed
At Thu, 20 Oct 2011 17:42:40 +0800, Raymond Yau wrote:
2011/9/20 Takashi Iwai <tiwai@suse.de>:
At Sat, 17 Sep 2011 10:16:19 +0800, Raymond Yau wrote:
2011/9/16 Takashi Iwai <tiwai@suse.de>:
At Fri, 16 Sep 2011 10:16:11 +0800, Raymond Yau wrote:
2011/9/12 Takashi Iwai <tiwai@suse.de>:
At Mon, 12 Sep 2011 09:53:28 +0200, Takashi Iwai wrote: > > At Sun, 04 Sep 2011 22:32:44 +0200, > Clemens Ladisch wrote: > > > > Raymond Yau wrote: > > > The possible way is to save the contents of 128K memory mapped I/O > > > ports before suspend to non-volatile memory and restore them after > > > resume > > > > > > Is there any way for alsa driver to allocate non-volatile memory ? > > > > 'Normal' memory is saved over suspend. Just use {k|v}malloc. > > If it's a 128kB bulk memory, vmalloc() is a more sensible choice.
BTW, in many cases, you don't have to save/restore the whole mmio region. Instead, stop the stream, save / restore some important registers and do the similar procedure for the normal prepare callback in the resume, and restart the stream again.
Takashi
The current au8830 alsa driver seem use less than 192Kbytes out of 256Kbytes mmio
http://thread.gmane.org/gmane.linux.kernel/564963/focus=47907
your patch seem initialse the request queue in figure 16 of pat 6167465
and the driver need to restore mixer and equalizer mmio for au8830 in figure 13
Yeah, that's what I mentioned in the thread.
Basically you'll need to restore the values assigned in hw_params() for PCM.
The error messages are due to snd_vortex_suspend() call snd_vortex_adb_init()
There are static routes for the codec, mixin, mixout and equalizer and dynamic routes of playback/capture streams
Those VORTEX_ADB_RTBASE and VORTEX_ADB_CHNBASE mmio ports are two linked list, those vortex_route_* function are adding the routes to the tail of those linked lists
so the only is to restore the routes since the driver does not keep track of the order of routes of the playback stream or capture stream to the linked list
The mixer is ac97, so it's restored by ac97_codec.c. The rest are the equalizer and a3d values to be restored separately.
I mean the hardware mixer (mixin and mixout) inside au88x0 chip.
May be need to initialise mpu401_uart or joystick
Is there any way to force the system to suspend, it seem that gnome power management in Fedora 10 does not provide any way to suspend while playing an audio ? ( screen saver set to 1 minutes, put the computer into sleep when inactive for 2 minutes )
See files in Documentation/power directory. You may call s2ram, powersave, or whatever command directly, too.
Takashi
1)Make sure that your computer can suspend and resume normally without au88x0 sound card 2)Backup all your important data before you try the patch with your au88x0 sound card 3)Only tested with au8830 , so it may need some modification for au8820 and au8810
my au8830 ,with two streams playing at different sampling rate at the same time, can suspend with the following script and resume as expected by pressing power button with the attached patch
sync echo 1 > /sys/power/pm_trace echo mem > /sys/power/state
It is strange that it can resume without uncomment those SNDRV_PCM_INFO_RESUME in au88x0_pcm.c
static struct snd_pcm_hardware snd_vortex_playback_hw_adb = { .info = (SNDRV_PCM_INFO_MMAP | /* SNDRV_PCM_INFO_RESUME | */ SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_INTERLEAVED |
This is normal. SNDRV_PCM_INFO_RESUME means that the driver supports the full resume, that is, the driver recovers the everything including the stream-restart in the resume. Most of drivers don't have it, and let the user-space re-trigger the stream upon resume. So, I believe the situation will be improved with your patch, at least. Could you brush it up to be merged to the upstream with the usual things (a proper commit log and your sign-off)? thanks, Takashi
participants (3)
-
Clemens Ladisch -
Raymond Yau -
Takashi Iwai