Repository navigation
[BUG] [TGL] PCM 99 capture fails with the IGO integrated firmware with Google Assistant "OK Google" feature #4568
Description
Describe the bug
Turn on Google Assistant "OK Google" feature,
arecord -Dhw:0,99 try.wav -c4 -r48000 -fs32_le -d60
localhost ~/dhara # Recording WAVE 'try.wav' : Signed 32 bit Little Endian, Rate 48000 Hz, Channels 4
arecord: set_params:1405: Unable to install hw params:
ACCESS: RW_INTERLEAVED
FORMAT: S32_LE
SUBFORMAT: STD
SAMPLE_BITS: 32
FRAME_BITS: 128
CHANNELS: 4
RATE: 48000
PERIOD_TIME: (21333 21334)
PERIOD_SIZE: 1024
PERIOD_BYTES: 16384
PERIODS: 4
BUFFER_TIME: (85333 85334)
BUFFER_SIZE: 4096
BUFFER_BYTES: 65536
TICK_TIME: 0
While the capture works after I turn off "OK Google" feature.
Reproduction Rate
always
Expected behavior
no hw params failure
Impact
What impact does this issue have on your progress (e.g., annoyance, showstopper)
blocking
tgl-13 branch with IGO patches & IGO binary
dmesg logs:
57.017745] x86/PAT: flashrom:4591 conflicting memory types fe000000-100000000 uncached-minus<->uncached
[ 57.020237] x86/PAT: reserve_memtype failed [mem 0xfe000000-0xffffffff], track uncached-minus, req uncached-minus
[ 71.223828] sof-audio-pci 0000:00:1f.3: error: ipc error for 0x60010000 size 20
[ 71.224352] sof-audio-pci 0000:00:1f.3: error: hw params ipc failed for stream 3
[ 71.224867] sof-audio-pci 0000:00:1f.3: ASoC: error at snd_soc_pcm_component_hw_params on 0000:00:1f.3: -12
[ 71.225532] DMIC: ASoC: hw_params FE failed -12
localhost ~ #
Activity
keyonjie commentedon Jul 29, 2021
looks the buffer heap is used up here, it is understandable as both KPB history buffer and this IGO are requires of quite big buffer, need to debug and fine tune this part:
fuyuntsuo commentedon Jul 30, 2021
Add more information:
igois using the Buffer Zone as its working buffer (<40KB IIRC. code here), butkpbis using the Run-time Zone (49KB in the log. code here)The size of these two zones are supposed to be determined in linking time. These two components are not sharing the same heap.
keyonjie commentedon Jul 30, 2021
the 'kpb' history buffer could requires up to e.g. 384KB history buffer at buffer zone.
keyonjie commentedon Jul 30, 2021
@fuyuntsuo @RDharageswari please help to check if the PR #4575 help for this issue.
fuyuntsuo commentedon Jul 30, 2021
@keyonjie
I've applied the PR to private repo
tgl-013-drop-stablebut it fails during compiling due to memory map change.log.txt
Actually I have tried to build the IGO with Github Main branch but fails neither. Seems like after the commit dcd4f34 we don't have enough memory for IGO any more.
dcd4f3.txt
0001-Disable-GWD-DSM-and-BT.zip
yongzhi1 commentedon Jul 30, 2021
Hi @keyonjie, thanks to PR #4575, the reported issue should have been resolved. That said, I have not done extensive testing, only verified the reported scenario, ie. WoV + DMIC recording. The FW was based on tigerlake_chrome_igonr.config.
lgirdwood commentedon Jul 30, 2021
@yongzhi1 @RDharageswari who will be able to do full test (and when) ? @mwasko fyi.
@keyonjie @slawblauciak good efforts in debugging this issue !
cujomalainey commentedon Sep 29, 2021
Ping on this bug, has this issue been fixed?
bkokoszx commentedon Oct 1, 2021
@sathya-nujella
I think we can close it now?
cujomalainey commentedon Oct 1, 2021
No activity for 3m, I am going to close this, feel free to reopen if i am incorrect.