Repository navigation
[BUG] Memory leak detected #4175
Description
Describe the bug
Memory leak detected in tests: 05_10_TestDmicMemLeak and 07_06_TestKdMemLeak
Topology for test 05_10_TestDmicMemLeak :
+----------+ +------------------+
| OS |-->| AUDIO DEVICE |-------------+
+----------+ +------------------+ |
|
+--------------------+ |
+----------+ | +------+ +-----+ | +------+ |
| Host cap |<---|Buffer|<--| Dai |<---| DMIC |<-+
+----------+ | +------+ +-----+ | +------+
+--------------------+
Topology for test 07_06_TestKdMemLeak
Topology:
pipe_plb
+------------------+
+----+ | +---+ +-------+ |
|Host|-->|Buf|->|SSP Dai|----------------------------------+
+----+ | +---+ +-------+ | |
+----------------- + |
|
+-------------------------+ |
+--------+ | +---+ +-----+ +---+ | |
|KD Dummy|<->|Buf|<->|ChSel|<->|Buf|<->+ |
+--------+ | +---+ +-----+ +---+ | | pipe_cap |
+-------------------------+ | |
| |
+-----------------+ |
+--------------------|-----------------------------------+ |
|+----+ +------+ +---+ +---+ +---+ +---+ +-------+ | |
||Host|<-|Buffer|<-|Kpb|<-|Buf|<-|Vol|<-|Buf|<-|SSP Dai|---+
|+----+ +------+ +---+ +---+ +---+ +---+ +-------+ |
+--------------------------------------------------------+
To Reproduce
Run tests 05_10_TestDmicMemLeak and 07_06_TestKdMemLeak with Diagnostic driver
Reproduction Rate
100/100
Problem doesn't exist before: ea5d9be
Environment
-
Firmware branch name and commit
- SOF: Branch: main, Hash: ea5d9be
-
Name of the platform(s) on which the bug is observed.
- Platform: TGL RVP, APL RVP, WHL RVP, ICL RVP
-
Test environment:
Branch: master, Hash: b7ed915d911acda055bbb822800b76557e9fe7de
Logs:
TGL:
07_06_TestKdMemLeak.zip
05_10_TestDmicMemLeak.zip
WHL:
05_10_TestDmicMemLeakWHL.zip
07_06_TestKdMemLeakWHL.zip
Activity
lgirdwood commentedon May 12, 2021
This could be two leaks, one in KWD and the other in DMIC. @singalsu can you check the dmic and @mrajwa the KWD. Thanks
mrajwa commentedon May 12, 2021
@slawblauciak will check KWD
singalsu commentedon May 12, 2021
Thanks, I'll test here 05_10_TestDmicMemLeak.
singalsu commentedon May 18, 2021
@kkarask There's three DMIC driver related fixed merged #4196, #4195, and #4159. Could you please re-run 05_10_TestDmicMemLeak since I'm not able to run the test with Linux as host OS.
kkarask commentedon May 18, 2021
@singalsu Issue still exists after marged above PR. Problem with memory leak is present at that Hash: f510435
singalsu commentedon May 25, 2021
I could not repeat this issue with SOF in single and overlapping capture cases with arecord scripting. Maybe this should be bisected if it's a regression if there would be a good version found from the past?
lgirdwood commentedon May 25, 2021
@singalsu I assume we are using uncache regions for DMIC config now like SSP ? Please see @keyonjie recent patches.
kkarask commentedon May 26, 2021
@singalsu Last stable version: 47d223c First memory leak detected: ea5d9be
kkarask commentedon Jun 8, 2021
@singalsu Good news! Memory leak is not visible on none of the platforms.
singalsu commentedon Jun 8, 2021
@kkarask That's great news, thanks for update! Can this issue be closed?