[keyonjie]

No description provided.

[ranj063]

closing this. rolled into #4702

[ranj063]

sorry closed the wrong one. meant to close mine :)

[lyakh]

no, this breaks atomicity. The original code made sure that no IRQ could interrupt execution between line 353 and the call to dw_dma_stop(). After that patch this atomicity is broken. Nested calls to irq_local_disable() or irq_local_enable() are fine as long as they're done with different flags. Those names are a bit misleading, in the kernel they're called local_irq_save() and local_irq_restore(). But the important point is, that if you call irq_local_disable() twice with different flags, if interrupts are enabled during the first call, it will save the CPU interrupt state with enabled interrupts into flags and disable interrupts. The next call to irq_local_disable() will save already disabled interrupts into a different copy of flags. Then the first call to irq_local_enable() with the second copy of the flags will restore that CPU state, which still had interrupts disabled. And only the second call to irq_local_enable() with the first copy of flags will re-enable the interrupts.
It is a different question though whether atomicity is important here.

[gkbldcig]

Can one of the admins verify this patch?

[lgirdwood]

@keyonjie what's the ETA to merge this ?

[keyonjie]

this was only for test, closing it now.