[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4751eb9b-755f-f890-cdda-03abff15e6f4@acm.org>
Date: Wed, 23 Aug 2023 07:39:39 -0700
From: Bart Van Assche <bvanassche@....org>
To: chengming.zhou@...ux.dev, axboe@...nel.dk, ming.lei@...hat.com,
hch@....de
Cc: linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
zhouchengming@...edance.com
Subject: Re: [PATCH v2 3/6] blk-mq-tag: remove bt_for_each()
On 8/23/23 02:04, chengming.zhou@...ux.dev wrote:
> Change the only user of bt_for_each() to use the introduced function
> __blk_mq_tagset_busy_iter() to specify queue filter when iterating.
>
> Since blk_mq_queue_tag_busy_iter() is only used to iterate over started
> requests, __blk_mq_tagset_busy_iter() already have BT_TAG_ITER_STARTED
> filter to iterate over started requests only, there should be no
> problem.
>
> Only one potential disadvantage I can see is that we lost the
> blk_mq_hw_queue_mapped() filter, which maybe not happen for now.
> Unmapped hctx was used to dynamically map or unmap when CPU hotplug,
> but we don't do this anymore, we always map all possible CPUs now.
> So it seems unmapped hctx may only happen if something wrong with
> driver's tagset map settings.
Reviewed-by: Bart Van Assche <bvanassche@....org>
Powered by blists - more mailing lists