error: log_entry_address 0x___ is not in dictionary range #5345

This is a follow-up to #5120.

#5120 was describing a very regular DMA corruption type: every time the end of the ring buffer was reached when it was misaligned.

The frequency of error: log_entry_address 0x____ is not in dictionary range errors has massively gone down now (even before the alignment fix was merged). But it's still not completely zero, this sof-logger warning is still happening in some specific tests on some specific platforms. Note that unlike #5352 this warning is NOT fatal, most of the time the tests PASS with it. It now seems rare enough not make the sof-logger crash (it was taking #5120 very many corruption events to take down the sof-logger eventually) and it does not seem to happen at regular intervals anymore but it still happens.

Looking for instance at daily run ID 10105, I found it happening more than 10 times in these tests (and a few others):

Start Time: 2022-02-10 22:27:29 UTC
Kernel Branch: topic/sof-dev
Kernel Commit: 936898a6
SOF Branch: main
SOF Commit: 443b21d
Zephyr Commit: 5094a6e08cee

Found more than 10 occurences in:
10105?model=CML_RVP_SDW&testcase=volume-basic-test-50
10105?model=CML_SKU0983_SDW&testcase=multiple-pipeline-all-50

It's also happening in suspend/resume and pause/resume tests but I'm guessing it's more or less expected in those? So I ignored them. See thesofproject/linux/issues/3275 for more background.

After excluding all the "expected" cases, it seems to happen only on CML.

While #5120 was most frequent on Zephyr, the warning never appeared with Zephyr in 10105

In 10105?model=CML_RVP_NOCODEC&testcase=check-playback-100times
I found the message 155 times and it seems to be very near src/lib/alloc.c:593 INFO heap: system status every time.

Also seen in

• https://sof-ci.01.org/sofpr/PR5340/build11976/devicetest/?model=TGLU_RVP_SDW&testcase=check-pause-resume-capture-10
• https://sof-ci.01.org/sofpr/PR5340/build11976/devicetest/?model=APL_UP2_NOCODEC&testcase=check-capture-3rounds