[PATCH 1/6] nvme: unquiesce io queues when removing namespaces

[Sagi Grimberg]

>>> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
>>> index 47d7ba2827ff..98fa8315bc65 100644
>>> --- a/drivers/nvme/host/core.c
>>> +++ b/drivers/nvme/host/core.c
>>> @@ -3903,6 +3903,12 @@ void nvme_remove_namespaces(struct nvme_ctrl *ctrl)
>>>   	 */
>>>   	nvme_mpath_clear_ctrl_paths(ctrl);
>>>   
>>> +	/*
>>> +	 * Unquiesce io queues so any pending IO won't hang, especially
>>> +	 * those submitted from scan work
>>> +	 */
>>> +	nvme_unquiesce_io_queues(ctrl);
>>
>> What quiesce does this pair with?
> 
> The one done during teardown of error recovery, then removal comes and breaks
> the in-progress error recovery.

unquiesce is also safe because of NVME_CTRL_STOPPED flag. All the
drivers depend on it being this way as we unquiesce unconditionally
after reset/(re)connect.

We don't track enough state here to know if the io queues were quiesced
before or not (in the happy delete they weren't and in the unhappy
delete they were).