[BUG] etrace sometimes empty on BDW and BYT #5385
Description
The shared memory "etrace" is sometimes empty on BDW and BYT
This makes the sof-logger test fails like this:
[REMOTE_ERROR] Initial FW ABI banner not found in sof-test/logs/check-sof-logger/2021-___/logger.etrace.txt
Initializing the DMA trace earlier reproduces this 100% of the time: #4334. Plenty of examples there.
In theory the DMA trace and etrace should be independent from each other. But in practice, theory and practice are different.
While much much rarer, it can also happen WITHOUT the #4334 code change, examples:
https://sof-ci.01.org/sofpr/PR5019/build11260/devicetest/?model=BDW_WSB_RT286&testcase=check-sof-logger
https://sof-ci.01.org/sofpr/PR4808/build10525/devicetest/?model=BDW_WSB_RT286&testcase=check-sof-logger
https://sof-ci.01.org/sofpr/PR4787/build10401/devicetest/?model=BDW_WSB_RT286&testcase=check-sof-logger
https://sof-ci.01.org/sofpr/PR4777/build10384/devicetest/?model=BDW_WSB_RT286&testcase=check-sof-logger
https://sof-ci.01.org/sofpr/PR4724/build10447/devicetest/?model=BDW_WSB_RT286&testcase=check-sof-logger
In daily run 7315?model=BDW_WSB_RT286&testcase=check-sof-logger
From vague memories of the code the DMA hardware seems very different on BDW and BYT.
References:
-
The difference between etrace vs DMA trace is documented in this comment:
[BUG] error trace debugfs entry (etrace) is always empty #3265 (comment) -
Other issue where the DMA trace gets stuck: [BUG]Empty or stuck DMA trace - Initial FW ABI banner message not found in logger.data.txt #4333
-
sof-test fix that was confusingly reporting an empty DMA when the etrace was actually empty:
check-sof-logger: fix FAIL message to point at the correct file sof-test#798
Only the final, red message was wrong, the dumped files were always correct.