linux-kernel - Re: [GIT PULL] Queue free fix (was Re: [PATCH] block: Free queue resources at blk_release

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]

Message-ID: <20110928180905.GB21628@redhat.com>
Date:	Wed, 28 Sep 2011 14:09:05 -0400
From:	Vivek Goyal <vgoyal@...hat.com>
To:	Christoph Hellwig <hch@...radead.org>
Cc:	James Bottomley <James.Bottomley@...senPartnership.com>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Jens Axboe <axboe@...nel.dk>, Hannes Reinecke <hare@...e.de>,
	James Bottomley <James.Bottomley@...allels.com>,
	"linux-scsi@...r.kernel.org" <linux-scsi@...r.kernel.org>,
	Linux Kernel <linux-kernel@...r.kernel.org>
Subject: Re: [GIT PULL] Queue free fix (was Re: [PATCH] block: Free queue
 resources at blk_release_queue())

On Wed, Sep 28, 2011 at 01:53:04PM -0400, Christoph Hellwig wrote:
> On Wed, Sep 28, 2011 at 01:48:59PM -0400, Vivek Goyal wrote:
> > I am wondering if we should retain blk_throtl_exit() in blk_cleanup_queue()
> > before lock swap and just move elevator cleanup in blk_release_queue().
> > 
> > A note to myself, I should probably enhance blk_throtl_exit() to look for any
> > queued throttled bio and single their completion with error (-ENODEV) or
> > something like that.
> 
> The root of this evil is how queue_lock is implemented and (ab)used.
> Instead of letting the driver assign a pointer to make the core use
> its locks we really need to make the queue_lock a lock embedded directly
> into the queue, and drivers may or may not use that lock for their
> internal data structures.  For high performance drivers they preferable
> should use their own locks as queue_lock is far too contended already
> for any high IOPS device.   The same applies to throtteling btw -
> instead of overloading an already highly contended lock it really
> should have its own.

I had thought of implementing a separate lock for throttling. Then I
noticed few operations like checking for queue flags where I would
be required to hold queue locks.

So I could do lock nesting.

hold throttling_lock;
   hold queue_lock;

IIRC, I also had noticed some operations where queue might want to call
into throttling with queue lock held and that would have led to lock
order problems.

So I had given up on the idea and continued to use queue lock for
throttling. Thought it could still probably be done if one could
justify additional complexity.

Thanks
Vivek
--
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/