On Mon, Feb 23, 2009 at 12:21:18PM +0200, Takashi Iwai wrote:
At Mon, 23 Feb 2009 17:53:51 +0800, Wu Fengguang wrote:
On Mon, Feb 23, 2009 at 10:54:36AM +0200, Takashi Iwai wrote:
At Sat, 21 Feb 2009 17:28:21 +0100, I wrote:
Regarding the default pin configurations. I made a series of patches to add the interface to change the default pincfg values dynamically via hwdep sysfs. The patches are found in test/hda-pincfg branch of sound git tree git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6.git
and included in the very latest alsa-driver-unstable snapshot ftp://ftp.kernel.org/pub/linux/kernel/people/tiwai/snapshot/alsa-driver-unstable-snapshot.tar.gz
Basically, you can get / set the default pincfg values on the fly, and trigger the codec reconfiguration after that.
The init_pin_configs file shows the initial default pincfgs set by BIOS.
# cat /sys/class/sound/hwC0D0/init_pin_configs 0x14 0x9993013f 0x15 0x01014c10 0x16 0x99030120 0x18 0x01a19c30 0x19 0x02a11c31 0x1a 0x01813c32 0x1b 0x02211c1f 0x1c 0x99330133 0x1d 0x411111f0 0x1e 0x411111f0 0x1f 0x411111f0
Then, you can override the default value by writing to override_pin_configs sysfs file.
# cat /sys/class/sound/hwC0D0/override_pin_configs # echo 0x1d 0x02a11c30 > /sys/class/sound/hwC0D0/override_pin_configs # cat /sys/class/sound/hwC0D0/override_pin_configs 0x1d 0x02a11c30
Note that this value won't appear in init_pin_configs.
Also, there is another file, cur_pin_configs. This contains the pincfgs that are set by the driver explicitly. For example, the static pincfg table in patch_sigmatel.c appears here.
Then you can trigger the codec reconfiguration by
# echo 1 > /sys/class/sound/hwC0D0/reconfig
This will re-setup the driver based on the given pin configs.
Cool! This interface will greatly ease the developer-user interactions.
Without knowing this feature, I had to write a patch for Paulo, and compile/verify it before sending it to him. Paulo also have to pull the kernel source, apply the patch, compile and run it...
Now you turned that tedious process into two simple echo commands :-)
Unfortunately, in the current scenario, it can't override the static pincfg table in the driver like in patch_sigmatel.c. It's just for fixing and BIOS setup.
Unfortunately patch_sigmatel has a long list of static pincfgs, i.e.:
static unsigned int ref927x_pin_configs[14] = { 0x02214020, 0x02a19080, 0x0181304e, 0x01014010, 0x01a19040, 0x01011012, 0x01016011, 0x0101201f, 0x183301f0, 0x18a001f0, 0x18a001f0, 0x01442070, 0x01c42190, 0x40000100, };This effectively prohibits itself from being dynamically reconfigured via the override_pin_configs interface...
Right. Considering the implementation again, I think it must be relatively easy to fix that. Just changing the order of the evaluation of pincfg, and add the check not to overwrite the value.
I'll work on it in the next week.
Fixed now on sound-unstable tree, together with renaming of override_pin and cur_pin to user_pin and driver_pin.
It'd be appreciated if someone can test it with a device with Sigmatel/IDT (as I have no such hardware)...
Hi Takashi,
Here are the good behaviors:
- the driver_pin_configs matches the numbers defined in source code
- the init_pin_configs/user_pin_configs will be exactly restored between module reloading
- setting user_pin_configs and doing reconfig did changed the 0x0c pin from output to input.
However,
- after reconfig, Surround, Center and LFE channels are muted and have to be turned on again in alsamixer.
That's a feature. You can restore via "alsactl restore"
Ah OK, I'd take it as a workaround ;-)
- if reconfig while playing audio, it will trigger a panic:
Ah, that's bad. I know this can be a problem (and thus is marked as EXPERIMENTAL :)
We need some mutex and/or flag to disable the reconfig during the critical operation.
It won't be a trivial work. The reset/rebuild is a long and complex process. For the moment I'd suggest to simply fail reconfig if there are active streams, so as to let this useful feature go upstream ASAP.
Thanks, Fengguang