linux-kernel - Re: [PATCH] floppy: convert to delayed work and single-thread wq

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:	Thu, 17 May 2012 07:55:19 -0700
From:	Tejun Heo <tj@...nel.org>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	Jiri Kosina <jkosina@...e.cz>,
	Stephen Hemminger <shemminger@...tta.com>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Jens Axboe <axboe@...nel.dk>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] floppy: convert to delayed work and single-thread wq

Hello, Linus.

On Wed, May 16, 2012 at 01:42:36PM -0700, Linus Torvalds wrote:
> On Wed, May 16, 2012 at 1:29 PM, Tejun Heo <tj@...nel.org> wrote:
> >
> > Ummm... still confused.  flush_work_sync() is fine too.  If you have
> > two, two calls to flush_work_sync() are equivalent to flushing the
> > workqueue in effect.  You just need to avoid flush_workqueue() because
> > system workqueues may be hosting work items which can run arbitrarily
> > long.
> 
> Umm. If there are abritrarily long things and these are serialized,
> then that workqueue is not good for putting floppy work on it either,
> is it? I don't think you can have it both ways.

They're not being serialized.  system_nrt_wq like any other system wqs
has 256 as max_active - so upto 256 work items per cpu can be
executing (ie. assigned to an active worker) at any given time and
because it's a large shared pool, some of them are allowed to take
long time.

> Either it's "good enough" for putting floppy_work, fd_timeout and
> fd_timer on, or it's not. If it's good enough, then flush_workqueue()
> should damn well be timely enough. And if flush_workqueue() isn't
> timely enough, then it doesn't sound like system_nrt_wq is the wrong
> choice.

So, per-work item, it's timely enough.  It shouldn't be different from
using a dedicated workqueue.  Different work items don't interact with
each other differently whether they belong to the same workqueue or
different ones as long as the workqueue's max_active limit isn't
reached.  flush_workqueue() is a different story as it forces all work
items belonging to the workqueue to be related.  That's why one of the
valid reasons to create a workqueue is to have a separate flush
(workqueue) domain - e.g. when the caller doesn't know which work
items need to be flushed because they're dynamically allocated and
freed on work item execution.

Thanks.

-- 
tejun
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/