[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v2] xen/blkfront: convert to blk-mq APIs

On 07/12/2015 02:14 AM, Jens Axboe wrote:
> On 07/11/2015 07:30 AM, Bob Liu wrote:
>> Note: This patch is based on original work of Arianna's internship for
>> GNOME's Outreach Program for Women.
> Great to see this finally get prepped to go in!
>> Only one hardware queue is used now, so there is no performance change.
> I would hope that the blk-mq path, even with one queue, is a perf win over 
> the old interface. So I'm not sure that is correct. But the bigger win will 
> be with more queues, of course.

Right, but there are memory consumption and migration issues while using more 
hardware queues.
So I separated this patch from that big patchset and hope can be merged first.

>> The legacy non-mq code is deleted completely which is the same as other
>> drivers like virtio, mtip, and nvme.
>> Also dropped one unnecessary holding of info->io_lock when calling
>> blk_mq_stop_hw_queues().
>> Changes in v2:
>>   - Reorganized blk_mq_queue_rq()
>>   - Restored most io_locks in place
> Looks good to me. The most common error case is the busy-out not stopping 
> queues, or not restarting them at completion. But that all looks fine.
> I would, however, rename blk_mq_queue_rq(). It sounds like a core function. 
> blkif_queue_rq() would be more appropriate.

Will send v3.

>> Signed-off-by: Arianna Avanzini <avanzini.arianna@xxxxxxxxx>
>> Signed-off-by: Bob Liu <bob.liu@xxxxxxxxxx>
> Acked-by: Jens Axboe <axboe@xxxxxx>

Thank you!


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.