Repository navigation
[BUG] FW status on IPC timeout in test 07_05_TestKdDmicD0ix16000Hz16b16b2ch #4929
Description
Describe the bug
Tests 07_05_TestKdDmicD0ix16000Hz16b16b2ch and 07_05_TestKdDmicD0ix16000Hz24b32b2ch failed wit error:
ERROR failed to alloc 0x2ea00 bytes zone 0x3 caps 0x8 flags 0x0
To Reproduce
Run test 07_05_TestKdDmicD0ix16000Hz16b16b2ch or 07_05_TestKdDmicD0ix16000Hz24b32b2ch with Diagnostic driver
Reproduction Rate
100%
Environment
- Firmware branch name and commit.
- SOF: Branch: main, Hash: b0cb159
- Name of the platform(s) on which the bug is observed.
- Platform: TGL RVP
Screenshots or console output
[ 996808.255208] ( 10.286458) c0 dma-trace src/trace/trace.c:135 Suppressed 19 similar messages: block %d base 0x%x size %d
[ 996812.994792] ( 4.739583) c0 memory src/lib/alloc.c:958 ERROR failed to alloc 0x2ed00 bytes zone 0x3 caps 0x8 flags 0x0
[ 996821.848958] ( 8.854167) c0 dma-trace src/trace/trace.c:135 Suppressed 19 similar messages: count %d free %d
[ 996828.411458] ( 6.562500) c0 memory src/lib/alloc.c:958 ERROR failed to alloc 0x2ec00 bytes zone 0x3 caps 0x8 flags 0x0
[ 996839.270833] ( 10.859375) c0 memory src/lib/alloc.c:958 ERROR failed to alloc 0x2eb00 bytes zone 0x3 caps 0x8 flags 0x0
[ 996848.750000] ( 9.479167) c0 memory src/lib/alloc.c:958 ERROR failed to alloc 0x2ea00 bytes zone 0x3 caps 0x8 flags 0x0
[ 999899.140625] ( 3050.390625) c0 dma-trace src/trace/trace.c:135 Suppressed 494 similar messages: failed to alloc 0x%x bytes zone 0x%x caps 0x%x flags 0x%x
[ 999909.427083] ( 10.286458) c0 sa src/lib/agent.c:79 WARN validate(), ll drift detected, delta = 47614
Skipped 56724 bytes after the last statement.
Logs:
TGL_07_05_TestKdDmicD0ix16000Hz16b16b2ch.zip
TGL_07_05_TestKdDmicD0ix16000Hz24b32b2ch.zip
Activity
lgirdwood commentedon Oct 27, 2021
@kkarask can you bisect. There has not been any memory updates this week, but a few weeks back there was some updates merged. @keyonjie fyi.
kkarask commentedon Oct 28, 2021
Bisect done.
Last working commit: 5b4ea83
First bad commit: bec163c
keyonjie commentedon Nov 2, 2021
I believe this error message actually doesn't impact the KD functionality?
This is for the KPB history buffer allocation, when fails we will retry with smaller size and the request is from LPSRAM where we have 64KB only on TGL. So I think we actually should always see this but the allocation eventually succeed so no impact in functionality. @mrajwa FYI.
keyonjie commentedon Nov 2, 2021
BTW @marc-hb the filter feature is very annoying here that it kills useful logs for us:
"Suppressed 19 similar messages: block %d base 0x%x size %d"
marc-hb commentedon Nov 2, 2021
We can discuss turning TRACE_FILTERING_ADAPTIVE off by default as long as that change is:
Even before discussing it, anyone can submit a draft PR that turns it off just to get test results and more verbose logs.
bkokoszx commentedon Nov 2, 2021
Hi @keyonjie
I misled you during bugscrub - it looks like we have another problem here (some timeout). @kkarask is checking that. We will make an update later.
[-][BUG] ERROR failed to alloc 0x2ea00 bytes zone 0x3 caps 0x8 flags 0x0 in test 07_05_TestKdDmicD0ix16000Hz16b16b2ch[/-][+][BUG] FW status on IPC timeout in test 07_05_TestKdDmicD0ix16000Hz16b16b2ch[/+]kkarask commentedon Nov 2, 2021
@keyonjie There is timeout on IPC:
Exception TimeoutError occurred at 'C:\work\KK\cavs_scripts-py\utilities\ipc_driver.py':603
FW status on IPC timeout after 5.0 seconds:
0x00000005 0x00000000 0x00004000
when test passed with working commit there is a response from that IPC:
The error: ERROR failed to alloc 0x2ec00 bytes zone 0x3 caps 0x8 flags 0x0 was not the real reason test failed, so I change the title
marc-hb commentedon Nov 2, 2021
Just to be sure: the bisect result is still the same?
kkarask commentedon Nov 3, 2021
@marc-hb yes, bisect result is still the same with reproduction rate 100%.
lgirdwood commentedon Nov 11, 2021
@keyonjie any update ?
keyonjie commentedon Nov 18, 2021
Looks we are quite lack of KWD coverage in our CI now, check-keyword-detection.sh event can't run on any of our 29 daily models, @keqiaozhang @aiChaoSONG @mengdonglin FYI.
Without that, we don't have this PM_GATE IPC validation at all, need to re-enable that and debug what happens here, with the latest main.
wszypelt commentedon Jan 2, 2023
issue does not occur
Tested FW hash: 4943aaa