linux-kernel - Re: [syzbot] possible deadlock in worker

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

Date:   Tue, 22 Feb 2022 08:30:22 -1000
From:   Tejun Heo <tj@...nel.org>
To:     "Fabio M. De Francesco" <fmdefrancesco@...il.com>
Cc:     Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>,
        syzkaller-bugs@...glegroups.com,
        Bart Van Assche <bvanassche@....org>,
        syzbot <syzbot+831661966588c802aae9@...kaller.appspotmail.com>,
        jgg@...pe.ca, linux-kernel@...r.kernel.org,
        linux-rdma@...r.kernel.org, Lai Jiangshan <jiangshanlai@...il.com>
Subject: Re: [syzbot] possible deadlock in worker_thread

Hello,

On Thu, Feb 17, 2022 at 01:27:08PM +0100, Fabio M. De Francesco wrote:
> Just to think and understand... what if the system-wide WQ were allocated as unbound 
> ordered (i.e., as in alloc_ordered_workqueue()) with "max_active" of one?
> 
> 1) Would it solve the locks dependency problem?

It'll actually make deadlocks a lot more prevalent. Some work items take
more than one work to complete (e.g. flushing another work directly or
waiting for something which must be completed by something else which may
involve a system work item) and system wq's max active must be high enough
that all those chains taking place at the same time should require fewer
number of work items than max_active.

> 2) Would it introduce performance penalties (bottlenecks)?

I'd be surprised it wouldn't cause at least notcieable latency increases for
some workloads.

Thanks.

-- 
tejun