soundwire: bus: update Slave status in sdw_clear_slave_status

[bardliao]

sdw_update_slave_status() will call the update_status() callback to
inform Slaves and do the corresponding tasks. We modify Slave status to
SDW_SLAVE_UNATTACHED but not update Slave status in sdw_clear_slave_status()
will lead to Slave status will never be updated elsewhere due to the
status is not changed.

Signed-off-by: Bard Liao yung-chuan.liao@linux.intel.com

[kv2019i]

Patch is good, but the commit message could be improved a bit.

[kv2019i]

Thanks @bardliao !

[bardliao]

Thanks @bardliao !

I would like to let @plbossart has a chance to review next week.

[plbossart]

I am not sure if this works in all cases @bardliao. The intent of the if (!hw_init) test in the resume() stage was to make sure we do not wait for the initialization complete in the case where there are 'ghost' devices exposed in ACPI tables but not physically populated in hardware. Are you sure this doesn't lead to reports of 'timeouts' with e.g. RT700 devices that are copy/pasted in all RVPs configs?

[plbossart]

I am not sure if this works in all cases @bardliao. The intent of the if (!hw_init) test in the resume() stage was to make sure we do not wait for the initialization complete in the case where there are 'ghost' devices exposed in ACPI tables but not physically populated in hardware. Are you sure this doesn't lead to reports of 'timeouts' with e.g. RT700 devices that are copy/pasted in all RVPs configs?

I also don't know if the test for hw_init done after the wait_for_completion(initialization_complete) is actually doing anything. If the attach fails, then hw_init will not be set. Conversely if the device attaches, hw_init will be set already. Testing it in the resume doesn't accomplish anything.

Maybe the right fix is to use

if (!first_hw_init) // instead of if (!hw_init)
   return 0;

wait_for_completion(&initialization_complete);

That way if we resume a device that never attached then we just do nothing.

[bardliao]

I am not sure if this works in all cases @bardliao. The intent of the if (!hw_init) test in the resume() stage was to make sure we do not wait for the initialization complete in the case where there are 'ghost' devices exposed in ACPI tables but not physically populated in hardware. Are you sure this doesn't lead to reports of 'timeouts' with e.g. RT700 devices that are copy/pasted in all RVPs configs?

I also don't know if the test for hw_init done after the wait_for_completion(initialization_complete) is actually doing anything. If the attach fails, then hw_init will not be set. Conversely if the device attaches, hw_init will be set already. Testing it in the resume doesn't accomplish anything.

Maybe the right fix is to use

if (!first_hw_init) // instead of if (!hw_init)
   return 0;

wait_for_completion(&initialization_complete);

That way if we resume a device that never attached then we just do nothing.

Ack

[plbossart]

closing since PR #2941 implements the same fixes.