[alsa-devel] [PATCH - 1/1] pcm: direct: Round down of slave_hw_ptr when buffer size is two period size

[Channaiah Vanitha (RBEI/ECF3)]

Hello Takashi-san,

This patch is regarding the fix for rounding down/up of slave pointers.
For buffer_size >= 2*period_size, round down of slave pointers and
for buffer_size < 2*period_size, round up of slave pointers will avoid xruns.
which otherwise causes snd_pcm_wait() to block for more than expected snd_pcm_period_elapsed() which leads to xruns.

We had similar discussion for same issue in below link:
https://mailman.alsa-project.org/pipermail/alsa-devel/2019-June/151169.html

Regards,
Vanitha


-----Original Message-----
From: Channaiah Vanitha (RBEI/ECF3) <Vanitha.Channaiah at in.bosch.com> 
Sent: Thursday, February 6, 2020 1:43 PM
To: Channaiah Vanitha (RBEI/ECF3) <Vanitha.Channaiah at in.bosch.com>; patch at alsa-project.org
Cc: alsa-devel at alsa-project.org; Wischer Timo (ADITG/ESS) <twischer at de.adit-jv.com>
Subject: [PATCH - 1/1] pcm: direct: Round down of slave_hw_ptr when buffer size is two period size

From: Vanitha Channaiah <vanitha.channaiah at in.bosch.com>

For buffer size equal to two period size, the start position of slave_hw_ptr is rounded down in order to avoid xruns

For e.g.:
Considering below parameters and its values:
Period size = 96 (0x60)
Buffer size = 192 (0xC0)
Timer ticks = 1
avail_min = 0x60
slave_hw_ptr = unaligned value during dmix_start()

Issue:
- Initial, app_ptr = hw_ptr = 0
- Application fills buffer size of data. so app_ptr = 0xC0, hw_ptr = 0
- During dmix_start(), current slave_hw_ptr is not rounded down. current slave_hw_ptr would be 0x66
- slave_hw_ptr is keep on updating at the hardware 0x67, 0x68, 0x69
- The diff calculation between old_slave_hw_ptr(0x66) and new_slave_hw_ptr(0x69)
  results in avail = 0x6
- After 1 snd_pcm_period_elapsed(), slave_hw_ptr = 0xC0
- The result of avail = 0x5A which is less than avail_min(0x60)
- Here, xruns will be observed

Fix:
- Initial, app_ptr = hw_ptr = 0
- Application fills buffer size of data. so app_ptr = 0xC0, hw_ptr = 0
- Round down of slave_hw_ptr during dmix_start() leads to below calculation:
- During dmix_start(), slave_hw_ptr rounded to 0x60 (old slave_hw_ptr)
- The diff calculation between old_slave_hw_ptr(0x60) and new_slave_hw_ptr(0x69)
  results in avail = 0x9
- After 1 snd_pcm_period_elapsed(), slave_hw_ptr = 0xC0
- The result of avail = 0x60 which is avail_min(0x60)
- Here, xruns can be avoided

Signed-off-by: Vanitha Channaiah <vanitha.channaiah at in.bosch.com>

diff --git a/src/pcm/pcm_direct.c b/src/pcm/pcm_direct.c index 54d9900..a201fa3 100644
--- a/src/pcm/pcm_direct.c
+++ b/src/pcm/pcm_direct.c
@@ -2043,10 +2043,14 @@ int snd_pcm_direct_parse_open_conf(snd_config_t *root, snd_config_t *conf,
 
 void snd_pcm_direct_reset_slave_ptr(snd_pcm_t *pcm, snd_pcm_direct_t *dmix)  {
+    /*
+     * For buffer size equal to two period size, slave_hw_ptr is rounded down
+     * to avoid xruns
+     */
 
 	if (dmix->hw_ptr_alignment == SND_PCM_HW_PTR_ALIGNMENT_ROUNDUP ||
 		(dmix->hw_ptr_alignment == SND_PCM_HW_PTR_ALIGNMENT_AUTO &&
-		pcm->buffer_size <= pcm->period_size * 2))
+		pcm->buffer_size < pcm->period_size * 2))
 		dmix->slave_appl_ptr =
 			((dmix->slave_appl_ptr + dmix->slave_period_size - 1) /
 			dmix->slave_period_size) * dmix->slave_period_size;
--
2.7.4