linux-kernel - Re: linux-next/linux memleak after IO on dax mountpoint

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:	Fri, 3 Jun 2016 14:35:14 +0800
From:	Xiong Zhou <xzhou@...hat.com>
To:	David Drysdale <drysdale@...gle.com>, hch@....de
Cc:	Xiong Zhou <xzhou@...hat.com>,
	Stephen Rothwell <sfr@...b.auug.org.au>, axboe@...com,
	linux-next@...r.kernel.org, linux-nvdimm@...1.01.org,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: linux-next/linux memleak after IO on dax mountpoint

On Thu, Jun 02, 2016 at 04:22:37PM +0100, David Drysdale wrote:
> On Sat, May 28, 2016 at 5:05 AM, Xiong Zhou <xzhou@...hat.com> wrote:
> > On Fri, May 27, 2016 at 04:46:17PM +0800, Xiong Zhou wrote:
> > ...
> >> Still working on to id which commit in this merge causes this issuer,
> >
> > Narrowed down to:
> >
> > 37e5823 block: add offset in blk_add_request_payload()
...
> > 0bf77e9 nvme: switch to RCU freeing the namespace
> > 9082e87 block: remove struct bio_batch
> 
> FWIW, I'm also seeing kmemleak report a leak with v4.7-rc1, in
> a different scenario (just normal desktop use).  Not done much
> digging so far, but this commit (9082e87bf) looks like it might be
> relevant

<snip..>

Thanks for the tip, accelerated my searching in the block merge.

On top of v4.7-rc1 , in order to revert this commit cleanly:

	9082e87 block: remove struct bio_batch

, have to revert these two:

	bbd848e0f block: reinstate early return of -EOPNOTSUPP from blkdev_issue_discard
	38f2525 block: add __blkdev_issue_discard

, then have to revert these two dependances:

202bae5 dm thin: unroll issue_discard() to create longer discard bio chains
3dba53a9 dm thin: use __blkdev_issue_discard for async discard support

With all these five commits reverted, NO memleak happens.

Reverting other four commits while not reverting 9082e87, memleak
happens.

Thanks,
Xiong