Guennadi Liakhovetski

checkpatch complaining about `memcpy()` which is just copied by this PR from existing code, only arguments have been changed

@wszypelt re-run QB?

ADLP_RVP_NOCODEC wasn't avail for the test, we could re-run, but in fact this PR has no run-time changes - only simple compile-time changes - changing type-casts and sparse attributes. So...

I modified the code a bit: ``` case SOF_DAI_INTEL_ALH: /* SDW HW FIFO always requires 32bit MSB aligned sample data for * all formats, such as 8/16/24/32 bits. */ dev->ipc_config.frame_fmt...

> Was this discovered while debugging recent regressions #6075 and/or #6091? Just trying to connect some dots, sorry if unrelated. @marc-hb #6075 , yes

Further, the same problem also appears on TGL, where the code was modified to add that NULL-dereference, also when built with crosstool gcc and with the Zephyr toolchain. Apparently that...

Update: just checked, that doing the same - adding a low address (e.g. 0x10) dereference in `main()` behaves in one of two ways: (1) causes an exception when reading from...

Update Nr.2: on BYT neither writes to nor reads from low addresses cause exceptions. Update Nr.3: actually using 0 instead of a low address has the same effect

> @lyakh what happens when > > 1. your pointer is volatile ? > OR > > 2. You write back the cache after the write ? @lgirdwood sorry, I...

@mmaka1 any idea about this?