[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <YhSa30p1cPm4CVCg@infradead.org>
Date: Tue, 22 Feb 2022 00:12:15 -0800
From: Christoph Hellwig <hch@...radead.org>
To: Wang Jianchao <jianchao.wan9@...il.com>
Cc: Christoph Hellwig <hch@...radead.org>,
Jens Axboe <axboe@...nel.dk>, Josef Bacik <jbacik@...com>,
Tejun Heo <tj@...nel.org>,
Bart Van Assche <bvanassche@....org>,
linux-block@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC V4 2/6] blk-wbt: make wbt pluggable
On Fri, Feb 18, 2022 at 11:41:11AM +0800, Wang Jianchao wrote:
>
>
> On 2022/2/17 4:50 下午, Christoph Hellwig wrote:
> >> +struct rq_qos *wbt_rq_qos(struct request_queue *q);
> >> int wbt_init(struct request_queue *);
> >
> >
> > Please move the wb_lat sysfs attribute into blk-wbt.c as well, which
> > removes the need to expose these two functions.
> >
>
> Given this patchset:
> (1) Do we need to reserve the wb_lat sysfs when we turn off the wbt ?
> (2) Do we need to disable wbt automatically when switch io scheduler
> to bfq ? Or just tell the user turn off the wbt by themselves ?
I don't think this needs any changes to what is done there today,
i.e. keep the sysfs attribute around.
Powered by blists - more mailing lists