linux-kernel - Re: [PATCH 4/5] block: make blkdev_get/put() handle exclusive access

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: <4CD9239E.2080209@kernel.org>
Date:	Tue, 09 Nov 2010 11:34:06 +0100
From:	Tejun Heo <tj@...nel.org>
To:	Christoph Hellwig <hch@...radead.org>
CC:	axboe@...nel.dk, linux-kernel@...r.kernel.org, petero2@...ia.com,
	schwidefsky@...ibm.com, heiko.carstens@...ibm.com, jack@...e.cz,
	akpm@...ux-foundation.org, adilger.kernel@...ger.ca, tytso@....edu,
	mfasheh@...e.com, joel.becker@...cle.com, aelder@....com,
	dm-devel@...hat.com, drbd-dev@...ts.linbit.com, neilb@...e.de,
	leochen@...adcom.com, sbranden@...adcom.com,
	chris.mason@...cle.com, swhiteho@...hat.com,
	shaggy@...ux.vnet.ibm.com, joern@...fs.org,
	konishi.ryusuke@....ntt.co.jp, reiserfs-devel@...r.kernel.org,
	viro@...iv.linux.org.uk
Subject: Re: [PATCH 4/5] block: make blkdev_get/put() handle exclusive access

Hello,

On 11/03/2010 05:10 PM, Christoph Hellwig wrote:
> On Mon, Nov 01, 2010 at 05:15:28PM +0100, Tejun Heo wrote:
>> * blkdev_get() is extended to include exclusive access management.
>>   @holder argument is added and, if is @FMODE_EXCL specified, it will
>>   gain exclusive access atomically w.r.t. other exclusive accesses.
>>
>> * blkdev_put() is similarly extended.  It now takes @mode argument and
>>   if @FMODE_EXCL is set, it releases an exclusive access.  Also, when
>>   the last exclusive claim is released, the holder/slave symlinks are
>>   removed automatically.
> 
> Could we get rid of FMODE_EXCL and just make a non-NULL holder field
> mean to open it exlusively (and pass a holder to the blkdev_put to
> release it)?

Yeah, I agree it's a bit awkward.  I'd really like to force one way or
the other tho.  ie. if non-NULL holder is gonna be required, I'll add
WARN_ON_ONCE(mode & FMODE_EXCL).  There are several issues to
consider.

* As Jan suggested, @mode in blkdev_put() isn't too useful.  I decided
  to keep it and use FMODE_EXCL for exclusive releases as that it is
  at least useful for something.  If we're gonna add @holder to
  blkdev_put(), it would make more sense to drop @mode.  It's not like
  there's a way to enforce restrictions according to open @mode during
  device access if there are mixed r/w opens.

* Some users don't keep @holder easily accessible until blkdev_put()
  is called, so the conversion will take a bit more effort.  No big
  deal in itself.

* What if @holder on blkdev_put() mismatches the current holder?
  Probably WARN_ON_ONCE() is the only recourse.  At that point, it's a
  bit silly to have to keep @holder around till blkdev_put().  Holders
  and opners counting already provide meaningful warning mechanism
  against spurious or missing exclusive releases.  Maybe we can have
  blkdev_put() and blkdev_put_exclusive()?  Or make it take boolean
  @excl?

So, after the above points, I decided to keep @mode.  It is a bit
awkward but the other way didn't seem too hip either.  Any better
ideas?

>> * bd_link_disk_holder() remains the same but bd_unlink_disk_holder()
>>   is no longer necessary and removed.
> 
> That's a rather asymetric interface.  What about having
> blkdev_get_stacked that require a gendisk as holder and set up the
> links underneath?

That will make the number of functions multiplied by two -
blkdev_get_by_path_stacked() and blkdev_get_by_dev_stacked().  The
symlinking for stacked drivers is an oddball feature which is and will
be only used by md and dm.  So, I think it's better to keep it
separate and oddball.

>> open_bdev_exclusive() and open_by_devnum() can use further cleanup -
>> rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop
>> special features.  Well, let's leave them for another day.
> 
> s/blkdev_get_by_devt/blkdev_get_by_dev/
> 
> And yes, that rename is a good idea and should go in ASAP after this
> patch.

Alright, will do it.

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/