linux-kernel - Re: [PATCH v3 1/3] lockdep: Make LOCKDEP_CROSSRELEASE configs all part of PROVE

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:   Wed, 23 Aug 2017 12:46:48 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Byungchul Park <byungchul.park@....com>
Cc:     Dave Chinner <david@...morbit.com>, mingo@...nel.org,
        linux-kernel@...r.kernel.org, kernel-team@....com,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Tejun Heo <tj@...nel.org>, Boqun Feng <boqun.feng@...il.com>
Subject: Re: [PATCH v3 1/3] lockdep: Make LOCKDEP_CROSSRELEASE configs all
 part of PROVE_LOCKING

On Wed, Aug 23, 2017 at 11:31:18AM +0900, Byungchul Park wrote:
> On Tue, Aug 22, 2017 at 11:06:03AM +0200, Peter Zijlstra wrote:

> Currently, we do the following in process_one_work(),
> 
> lockdep_map_acquire for a workqueue
> lockdep_map_acquire for a work
> 
> But IMHO it should be,
> 
> lockdep_map_acquire for a pair of workqueue and work.
> 
> Right?

No, it is right. We need the two 'locks'.

The work one is for flush_work(), the workqueue one is for
flush_workqueue().

Just like how flush_work() must not depend on any lock taken inside the
work, flush_workqueue() callers must not hold any lock acquired inside
any work ran inside the workqueue. This cannot be done with a single
'lock'.

The reason flush_work() also depends on the wq 'lock' is because doing
flush_work() from inside work is a possible problem for single threaded
workqueues and workqueues with a rescuer.

> > Agreed. The interaction with workqueues is buggered.
> 
> I think original uses of lockdep_map were already wrong. I mean it's
> not a problem of newly introduced code.

Not wrong per-se, the new code certainly places more constraints on it.