At Mon, 27 Apr 2009 17:11:39 +0200, I wrote:
At Mon, 27 Apr 2009 17:09:27 +0200 (CEST), Jaroslav Kysela wrote:
On Mon, 27 Apr 2009, Jon Smirl wrote:
I have fixed all of the other bugs. If I take out the position estimation code it fails.
Jaroslav hasn't commented on this yet....
Appearently, FIFO is too big. I'm working on some changes to take account fifo_size to the jiffies based check. Please, set fifo_size correctly in your driver. USB drivers with big "URB FIFOs" should be changed as well, too.
Ah good, that's what I was thinking to fix up later :)
After thinking twice, I found that we basically don't need much jiffies testing for that kind of hardware because it can't report bad values anyway.
A test patch is below. It suppose that such hardware should have SNDRV_PCM_INFO_BATCH in runtime->hw.info (although most drivers don't set it yet).
Comments?
Takashi
--- diff --git a/sound/core/pcm_lib.c b/sound/core/pcm_lib.c index 63d088f..a2a792c 100644 --- a/sound/core/pcm_lib.c +++ b/sound/core/pcm_lib.c @@ -249,6 +249,12 @@ static int snd_pcm_update_hw_ptr_interrupt(struct snd_pcm_substream *substream) new_hw_ptr = hw_base + pos; } } + /* Skip the jiffies check for hardwares with BATCH flag. + * Such hardware usually just increases the position at each IRQ, + * thus it can't give any strange position. + */ + if (runtime->hw.info & SNDRV_PCM_INFO_BATCH) + goto no_jiffies_check; hdelta = new_hw_ptr - old_hw_ptr; jdelta = jiffies - runtime->hw_ptr_jiffies; if (((hdelta * HZ) / runtime->rate) > jdelta + HZ/100) { @@ -272,6 +278,7 @@ static int snd_pcm_update_hw_ptr_interrupt(struct snd_pcm_substream *substream) hw_base -= hw_base % runtime->buffer_size; delta = 0; } + no_jiffies_check: if (delta > runtime->period_size + runtime->period_size / 2) { hw_ptr_error(substream, "Lost interrupts? "