linux-kernel - Re: [PATCH] mmc: block: use REQ_HIPRI flag to complete request directly in own complete workqueue

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]

Message-ID: <20210205162221.GA27114@lst.de>
Date:   Fri, 5 Feb 2021 17:22:21 +0100
From:   Christoph Hellwig <hch@....de>
To:     Ulf Hansson <ulf.hansson@...aro.org>
Cc:     Liu Xiang <liu.xiang@...ngsmart.com>,
        "linux-mmc@...r.kernel.org" <linux-mmc@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        liuxiang_1999@....com, Christoph Hellwig <hch@....de>,
        Adrian Hunter <adrian.hunter@...el.com>
Subject: Re: [PATCH] mmc: block: use REQ_HIPRI flag to complete request
 directly in own complete workqueue

On Fri, Feb 05, 2021 at 03:24:06PM +0100, Ulf Hansson wrote:
> On Thu, 21 Jan 2021 at 09:13, Liu Xiang <liu.xiang@...ngsmart.com> wrote:
> >
> > After commit "40d09b53bfc557af7481b9d80f060a7ac9c7d314", request is
> > completed in softirq. This may cause the system to suffer bad preemptoff
> > time.
> > The mmc driver has its own complete workqueue, but it can not work
> > well now.
> > The REQ_HIPRI flag can be used to complete request directly in its own
> > complete workqueue and the preemptoff problem could be avoided.
> 
> I am trying to understand all of the problem, but I don't quite get
> it, sorry. Would it be possible for you to extend the description in
> the commit message a bit?

Yes, the message sounds weird.  The mentioned commit should obviously
not make any difference for drivers not using it.

> More exactly, what will happen if we tag a request with REQ_HIPRI
> before completing it? Apologize for my ignorance, but I am currently a
> bit overwhelmed with work, so I didn't have the time to really look it
> up myself.

Drivers must never set REQ_HIPRI!  This is a flag that is set by
the submitter, and actually cleared for most drivers that don't support
it by the block layer.