Repository navigation
[BUG]: IPC timeout observed for topology with mux components #7185
Description
with PR: #7099 observing IPC timeout with TDM mode (Mux component) on AMD platforms.
Attaching the error snippet.
{{ message }}
with PR: #7099 observing IPC timeout with TDM mode (Mux component) on AMD platforms.
Attaching the error snippet.
Activity
ipc: ipc-common: Fix the logic for setting the pipeline source/sink c…
ranj063 commentedon Mar 1, 2023
@bhiregoudar can you please paste the topology graph that's failing with this patch?
maruthi-machani commentedon Mar 2, 2023
lgirdwood commentedon Mar 6, 2023
@ujfalusi @lyakh any inputs here ?
@maruthi-machani do you have FW logs prior to timeout ? i.e. any logs from shared memory on your platform.
ujfalusi commentedon Mar 6, 2023
@maruthi-machani, a description on how this failure can be triggered would help
maruthi-machani commentedon Mar 7, 2023
Issue observed during initial topology loading. by reverting the changes in ipc-common.c, the topology load is successful.
lgirdwood commentedon Mar 8, 2023
@maruthi-machani not seeing this on Intel HW atm, are you able to add some debug, it looks like an exception (I'm not sure why the kernel would say the otherwise). I can only think we are trying to deref a mux buffer that does not or no longer exists ?
kv2019i commentedon Mar 13, 2023
@maruthi-machani Do you have time to debug this for v2.5? We can't really see this on any Intel platform, so we'd need help on the debug. We are targetting for a v2.5-rc1 end of this week.
maruthi-machani commentedon Mar 14, 2023
@kv2019i we need sometime to debug this. seems this is specific to AMD platform.
maruthi-machani commentedon Mar 19, 2023
@kv2019i we are observing for pipeline_29 buffer_get_comp() is waiting to acquire the buffer (buffer_acquire()).
Is this happen due to improper termination of pipeline_29 (ACPHS_VIRTUAL Dai). Can you suggest any reference topology for proper termination of mux component having two PCM playbacks.
maruthi-machani commentedon Mar 20, 2023
@kv2019i with the echo-ref dai topology, the issue of waiting to acquire the buffer got resolved. Also IPC timeout not observed.
kv2019i commentedon Mar 21, 2023
Thanks @maruthi-machani ! So I gather this is at least not cause to block SOF2.5 release?
maruthi-machani commentedon Mar 23, 2023
with AMD topology specific changes the issue is resolved, no blocker.
lgirdwood commentedon Mar 24, 2023
@maruthi-machani Have these updates been merged, can we close ?
maruthi-machani commentedon Mar 27, 2023
updates merged #7313, please close the issue.
maruthi-machani commentedon Mar 28, 2023
updates merged #7313, closing the issue.