linux-kernel - Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Tue, 7 Apr 2020 15:04:20 -0700
From:   Doug Anderson <dianders@...omium.org>
To:     Ming Lei <ming.lei@...hat.com>
Cc:     Jens Axboe <axboe@...nel.dk>,
        "James E.J. Bottomley" <jejb@...ux.ibm.com>,
        "Martin K. Petersen" <martin.petersen@...cle.com>,
        Paolo Valente <paolo.valente@...aro.org>,
        Salman Qazi <sqazi@...gle.com>,
        linux-block <linux-block@...r.kernel.org>,
        linux-scsi@...r.kernel.org, Guenter Roeck <groeck@...omium.org>,
        Ajay Joshi <ajay.joshi@....com>, Arnd Bergmann <arnd@...db.de>,
        Bart Van Assche <bvanassche@....org>,
        Chaitanya Kulkarni <chaitanya.kulkarni@....com>,
        Damien Le Moal <damien.lemoal@....com>,
        Hou Tao <houtao1@...wei.com>,
        Pavel Begunkov <asml.silence@...il.com>,
        Tejun Heo <tj@...nel.org>, LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 2/2] blk-mq: Rerun dispatching in the case of budget contention

Hi,

On Mon, Apr 6, 2020 at 7:14 PM Ming Lei <ming.lei@...hat.com> wrote:
>
> On Sun, Apr 05, 2020 at 09:26:39AM -0700, Doug Anderson wrote:
> > Hi,
> >
> > On Sun, Apr 5, 2020 at 2:15 AM Ming Lei <ming.lei@...hat.com> wrote:
> > >
> > > @@ -103,6 +104,9 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
> > >                 rq = e->type->ops.dispatch_request(hctx);
> > >                 if (!rq) {
> > >                         blk_mq_put_dispatch_budget(hctx);
> > > +
> > > +                       if (e->type->ops.has_work && e->type->ops.has_work(hctx))
> > > +                               blk_mq_delay_run_hw_queue(hctx, BLK_MQ_BUDGET_DELAY);
> >
> > To really close the race, don't we need to run all the queues
> > associated with the hctx?  I haven't traced it through, but I've been
> > assuming that the multiple "hctx"s associated with the same queue will
> > have the same budget associated with them and thus they can block each
> > other out.
>
> Yeah, we should run all hctxs which share the same budget space.
>
> Also, in theory, we don't have to add the delay, however BFQ may plug the
> dispatch for 9 ms, so looks delay run queue is required.

I have posted up v3.  A few notes:

* Since we should run all "hctxs" I took out the check for has_work()
before kicking the queue.

* As far as I can tell the theoretical race happens for _anyone_ who
puts budget, so I added it to blk_mq_put_dispatch_budget().  Feel free
to shout at me in response to v3 if you believe I'm wrong about that.

Thanks for all your reviews and suggestions so far!

-Doug