At Fri, 27 Jul 2007 00:55:14 -0700, Paul Vojta wrote:
On Fri, Jul 06, 2007 at 11:45:00AM +0200, Takashi Iwai wrote:
0x0b = CS4231_TEST_INIT, and 0x20 = CS4231_CALIB_IN_PROGRESS. At least, mce_down() should show messages if the calibration gets timeout. Did you see any related messages? Or didn't you build with --with-debug=full?
I have now recompiled the alsa drivers from alsa-driver-hg20070718 and with --debug=full.
I found that the bug would also occur the first time playing a .wav file after "aplay -t au file.au".
The difference between opl3sa2 and 4232 occurred because of the lines
if ((timeout & CS4231_MCE) == 0 || !(chip->hardware & (CS4231_HW_CS4231_MASK | CS4231_HW_CS4232_MASK))) { return; }
in mce_down() in cs4231_lib.c. With opl3sa2 it was always returning at this step, hence it didn't finish recalibrating before trying to play the sound.
If I changed it so that the second test is always false for opl3sa2, then sound would mostly work, but intermittently I would get messages about "auto calibration time out" or "serious init problem - codec still busy". If I put a printk() statement in the beginning of mce_down to help debug this, then the problem would go away, so I put in udelay(100) and that also made the problems go away.
Therefore, I submit the following patch.
Synopsis: Fix bugs in mode change/recalibration for opl3sa2 driver.
Signed-off by: Paul Vojta vojta@math.berkeley.edu
Great, finally you spotted out. I applied it to ALSA HG tree now.
thanks,
Takashi