sof: Define possible UUID entries usage in IPC during declaration

[ktrzcinx]

A few UUID entries, related with components from topology,
should be loaded into SRAM to allow runtime comparison with
value in IPC message during component creation. UUID of the
components whose are not created by IPC message are not needed
in runtime, so there should not be there to reduce memory
consumtion.

Signed-off-by: Karol Trzcinski karolx.trzcinski@linux.intel.com

It's a part of #2920
@keyonjie please check if I choose 'RT` UUID for all needed components.

[paulstelian97]

We'll wait until we're sure this won't cause any Out Of Memory issues in DSP SRAM.

[mmaka1]

I'd keep the DECLARE_SOF_UUID()/SOF_UUID() for logging-only case unchanged. And introduce DECLARE_SOF_RT_UUID()/SOF_RT_UUID() for run-time identification.

[ktrzcinx]

From CI:

0.13s$ ./scripts/host-testbench.sh
==========================================================
test volume with ./volume_run.sh 16 16 48000 zeros_in.raw volume_out.raw
volume test failed!

@singalsu maybe you have a suggestion what is going on ?

[singalsu]

From CI:

0.13s$ ./scripts/host-testbench.sh
==========================================================
test volume with ./volume_run.sh 16 16 48000 zeros_in.raw volume_out.raw
volume test failed!

@singalsu maybe you have a suggestion what is going on ?

OK, I'll see what causes that.

[singalsu]

@ktrzcinx It segfaults to this line

0x00007ffff7e4e8bd in comp_new (comp=comp@entry=0x7fffffffd4a0) at /home/singalsu/work/current/sof/sof/src/audio/component.c:79

tr_info(&comp_tr, "comp new %s type %d id %d.%d", drv->tctx->uuid_p, comp->type, comp->pipeline_id, comp->id);
drv->tctx is NULL

Where is it set normally? Probably when updating this it has been forgotten to add to testbench runtime too. The trace output in testbench has been somewhat broken for a while already due to left without updates but this is fist time that it segfaults.

[ktrzcinx]

@singalsu Thanks, comp_file_host and comp_file_dai didn't have assigned any pointer to .tctx, which one I start to use in comp_new(). So I added one more commit and now should be ok, I guess.

[singalsu]

@singalsu Thanks, comp_file_host and comp_file_dai didn't have assigned any pointer to .tctx, which one I start to use in comp_new(). So I added one more commit and now should be ok, I guess.

Superb, thanks!! Yes I just ran the process_test and everything worked again.

[mmaka1]

Implements and closes #769

[lgirdwood]

Jenkins DUT failure.