abi: define debug abi version for user space dbg interfaces

[mmaka1]

The sof-logger and potentially other debug API clients perform ABI
compatibility check using the single FW ABI version. The same one is used
by the primary FW client which is the kernel driver. If there is a change
made to the debug API, the main ABI has to be updated to protect integrity
of the debug tools while such a change may not affect the kernel driver
at all.

This patch introduces new debug ABI version to be increased when changing
user space debug interfaces while the the main ABI is not affected.
Recompilation and installation of the new driver every time the tunneled
debug protocol is upgraded may be avoided.

Signed-off-by: Marcin Maka marcin.maka@linux.intel.com

[mmaka1]

Implements #2374

[paulstelian97]

I feel like the separating out of this function can be its own patch.

[jajanusz]

If it were just refactor, then yea. But here you'd have 1 diff to have it changed in another.

[lgirdwood]

Do we need corresponding kernel PR for this one ?

[mmaka1]

Do we need corresponding kernel PR for this one ?

@lgirdwood I tested with unchanged kernel and seems to work fine. Another fw_ready extension does not break compatibility, therefore I increased 'minor' component only in the main ABI before aligning initial debug ABI version with the main one.

[lgirdwood]

Do we need corresponding kernel PR for this one ?

@lgirdwood I tested with unchanged kernel and seems to work fine. Another fw_ready extension does not break compatibility, therefore I increased 'minor' component only in the main ABI before aligning initial debug ABI version with the main one.

Ok great, @plbossart good for you ? or would you need a patch that printk() the ABI version for userspace debug ?

[lgirdwood]

SOFCI TEST

[lgirdwood]

@zrombel@xiulipan CI is showing a pass afaict but we are seeing pending/failed ?

[zrombel]

Quickbuild changes are causing some problems, sorry for that. When it will be fixed I will rebuild this PR.

[zrombel]

QB is back on track. Build and tests passed.

[jajanusz]

Is 'debug abi' the best name for this? We have folders that are named by-abi-user (kernelspace or userspace), but ABI version symbols are named by-feature (generic vs debug).

[jajanusz]

Nvm, I see it's going to be more granular for user, versions are in sof_ipc_user_abi_version

[jajanusz]

sof-ci/jenkins/pr-tools-build Pending — Tools Building shows as pending, but it's CI reporting error, cos results there are skipped/ok

[lgirdwood]

@mmaka1 could you copy the rimage patches to the external rimage repo to as another PR.
@fredoh9@xiulipan any idea why tools build is timing out ? Server OK ?

[jajanusz]

@mmaka1 can u fix checkpatch issue?

[xiulipan]

@lgirdwood As we are trying to update the GitHub behind a proyx, there is always some chance that the status update is missed. Will try to add some more strict check to see if we can use retry or other method to avoid such issue.

[mmaka1]

@jajanusz I fixed the alignment issue. Another one be fixed is the memcpy issue however this function is used in many places in that source file, so there is a general replacement, done as another patch, required.

[lgirdwood]

@zrombel we have a DMIC CI failure on ICL, but it should be unrelated to this patch ?

[wwittbrx]

@lgirdwood Aside from SSP KD tests failing due to regression on master (should be fixed by #2440) this looks good to be merged.

[lgirdwood]

SOFCI TEST

[lgirdwood]

@wwittbrx running CI again now that Jenkins has had some fixes too.