On Wed, Oct 5, 2016 at 10:46 PM, Bart Van Assche
[off-list ref] wrote:

On 10/04/16 21:32, Ming Lei wrote:

quoted

On Wed, Oct 5, 2016 at 12:16 PM, Bart Van Assche
[off-list ref] wrote:

quoted

On 10/01/16 15:56, Ming Lei wrote:

quoted

If we just call the rcu/srcu read lock(or the mutex) around .queue_rq(),
the above code needn't to be duplicated any more.

Can you have a look at the attached patch? That patch uses an srcu read
lock
for all queue types, whether or not the BLK_MQ_F_BLOCKING flag has been
set.

That is much cleaner now.

quoted

Additionally, I have dropped the QUEUE_FLAG_QUIESCING flag. Just like
previous versions, this patch has been tested.

I think the flag of QUEUE_FLAG_QUIESCING is still needed because we
have to set this flag to prevent new coming .queue_rq() from being run,
and synchronize_srcu() won't wait for completion of that at all (see
section of 'Update-Side Primitives' in [1]).

[1] https://lwn.net/Articles/202847/

Hello Ming,

How about using the existing flag BLK_MQ_S_STOPPED instead of introducing a
new QUEUE_FLAG_QUIESCING flag? From the comment above blk_mq_quiesce_queue()

That looks fine, and we need to stop direct issue first after hw queue
becomes BLK_MQ_S_STOPPED.

in the patch that was attached to my previous e-mail: "Additionally, it is
not prevented that new queue_rq() calls occur unless the queue has been
stopped first."

Thanks,

Bart.

-- 
Ming Lei