You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
PCM read/write error when buffer size < 768 on APL #116
Today I tried to test playback with specified buffer size and found that if the buffer size is less than 768, it will easy introduce I/O error and even with IPC timed out.
How to reproduce:
aplay -D hw:0,0 -r 48000 -c 2 -f S16_LE /dev/zero --buffer-size=384
It will stop with pcm read/write error. dmesg-bufsize.log rmbox-buffsize.log
Don't we have by-design limitations on the buffer/period sizes? If the topology states that the host DMA can work with 48 samples, that's 192 bytes and already larger than the period in the firmware you are using in this case on the host side. I'd think the ALSA period needs to be >= the firmware period?
I checked the buffer/period sizes in topology, the buffer size is fixed as 65536 and the range of period size is (192.16384), there're no issues when playback with buffer-size=65536. I also tried with xrun injection every 10ms and buffer-size < 576, still cannot reproduce this bug. This improvement may related to PR#125.
Activity
lgirdwood commentedon Jul 17, 2018
This looks like another duplicate of the XRUN bug. @keqiaozhang can you check.
plbossart commentedon Jul 17, 2018
Don't we have by-design limitations on the buffer/period sizes? If the topology states that the host DMA can work with 48 samples, that's 192 bytes and already larger than the period in the firmware you are using in this case on the host side. I'd think the ALSA period needs to be >= the firmware period?
keqiaozhang commentedon Jul 18, 2018
I will check this tomorrow.
keqiaozhang commentedon Jul 19, 2018
I checked the buffer/period sizes in topology, the buffer size is fixed as 65536 and the range of period size is (192.16384), there're no issues when playback with buffer-size=65536. I also tried with xrun injection every 10ms and buffer-size < 576, still cannot reproduce this bug. This improvement may related to PR#125.
lgirdwood commentedon Jul 19, 2018
@keqiaozhang ok, please close once you are satisfied.
keqiaozhang commentedon Jul 20, 2018
This issue is fixed by #125
jocelyn-li commentedon Jul 20, 2018
This issue is fixed by #125