From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753814AbZCIJir (ORCPT ); Mon, 9 Mar 2009 05:38:47 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752630AbZCIJij (ORCPT ); Mon, 9 Mar 2009 05:38:39 -0400 Received: from brick.kernel.dk ([93.163.65.50]:52162 "EHLO kernel.dk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752594AbZCIJii (ORCPT ); Mon, 9 Mar 2009 05:38:38 -0400 Date: Mon, 9 Mar 2009 10:38:36 +0100 From: Jens Axboe To: Li Zefan Cc: LKML , martin.petersen@oracle.com Subject: Re: [PATCH] block: fix memory leak in bio_clone() Message-ID: <20090309093836.GJ11787@kernel.dk> References: <49B4DD9C.5030902@cn.fujitsu.com> <20090309092407.GI11787@kernel.dk> <49B4E2BA.3030809@cn.fujitsu.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <49B4E2BA.3030809@cn.fujitsu.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Mar 09 2009, Li Zefan wrote: > Jens Axboe wrote: > > On Mon, Mar 09 2009, Li Zefan wrote: > >> If bio_integrity_clone() fails, bio_clone() returns NULL without freeing > >> the newly allocated bio. > >> > >> Signed-off-by: Li Zefan > >> --- > >> fs/bio.c | 4 +++- > >> 1 files changed, 3 insertions(+), 1 deletions(-) > >> > >> diff --git a/fs/bio.c b/fs/bio.c > >> index 124b95c..896330e 100644 > >> --- a/fs/bio.c > >> +++ b/fs/bio.c > >> @@ -465,8 +465,10 @@ struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask) > >> > >> ret = bio_integrity_clone(b, bio, fs_bio_set); > >> > >> - if (ret < 0) > >> + if (ret < 0) { > >> + bio_put(bio); > >> return NULL; > >> + } > >> } > >> > >> return b; > >> -- 1.5.4.rc3 > > > > Good spotting. But it looks like there are actually several problems > > there. bio_integrity_clone() is mempool backed. Currently that ret < 0 > > can never trigger, since bio_integrity_clone() has hard-wired __GFP_WAIT > > Do you mean GFP_NOIO? The __GFP_WAIT part of GFP_NOIO is the important bit :-) > > as the mempool mask. So the leak will not occur, but it does mean that > > it isn't honoring the gfp_mask passed in to bio_clone(), which is the > > I noticed there was a patch to do this, and seems you planed to merge > it into .29? > > "[PATCH] Add gfp_mask to bio_integrity_clone()" > > http://lkml.org/lkml/2008/10/30/11 Hmm strange, apparently that never got queued up and I had forgotten all about that issue until your email. I'll add it asap. > > first bug. The second bug is that it should be using its own bioset, as > > it is illegal to do multiple __GFP_WAIT allocations on a single mempool > > and always expect progress. That one still stands. -- Jens Axboe