Guennadi Liakhovetski

Preventing pipeline parameter negotiation from walking the pipeline backwards to where we came from seems indeed intentional, negotiation is only supposed to check other branches on multi-source or multi-sink components.

As explained above, `.walking` is used in cases like on a mixer, when an audio format is configured on one of branches, we will walk all other branches to check...

**FIXME:** the last commit breaks several things: 1. it adds an unbalanced `notifier_unregister()` 2. it probably breaks (again) the now perfectly correct HDA DMA stop flow All of those must...

I tested this PR locally on a MinnowBoard BYT platform and with it the system locks up hard immediately when "release" is pressed. Debugging

In this version the first patch completely disables PAUSE and RELEASE on legacy ADSPs, the second patch actually replaces the former with the latter, and I let out the HDA...

Taking the https://sof-ci.01.org/sofpr/PR5171/build745/devicetest/?model=ADLP_RVP_NOCODEC_ZEPHYR&testcase=check-pause-resume-playback-10 CI failure as an example. This seems to be the same problem as the one on legacy platforms, which is why the respective commit in this series...

> @lyakh lets figure out the driver changes so that the FW changes are 99% code removal. @lgirdwood so we should block PAUSE / RELEASE at the driver level? But...

> @lyakh can you check for SPARSE usage. Thanks @lgirdwood sorry, maybe @RanderWang has uploaded a wrong version, all the same comments apply, none of them have been addressed, although...

> > @lyakh sorry, there are two commits and I fixed in the second commit. Do I need to fix it in the first commit ? > > Yes please,...

> @keqiaozhang can you turn on Zephyr lock debugging and try again (Zephyr KCONFIG). @lyakh I assume this is correct method to enable Zephyr lock debug ? @keqiaozhang @lgirdwood yes,...