From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jens Axboe Subject: Re: [PATCH] BUG: ll_merge_requests_fn() updates req->nr_phys_segments wrongly Date: Thu, 25 Sep 2008 15:55:26 +0200 Message-ID: <20080925135526.GO2677@kernel.dk> References: <200809232358.02810.knikanth@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <200809232358.02810.knikanth@suse.de> Sender: linux-kernel-owner@vger.kernel.org To: Nikanth Karthikesan Cc: linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org, FUJITA Tomonori List-Id: linux-scsi@vger.kernel.org On Tue, Sep 23 2008, Nikanth Karthikesan wrote: > > From: Nikanth Karthikesan > > ll_merge_requests_fn() decreases the req->nr_phys_segments if > blk_phys_contig_segment() returns true, but it is perfectly possible that > blk_hw_contig_segment() is false. A new hw_segment implies a new phys_segment. > So decrementing nr_phys_segments wrongly here triggers the BUG_ON() in > scsi_init_sgtable(), as blk_rq_map_sg() would map 1 phys_segment more than > req->nr_phys_segment. This is easily reproducible. Other callers of > blk_rq_map_sg() should also be affected by this bug. > > Signed-off-by: Nikanth Karthikesan > --- > > block/blk-merge.c | 10 ++++------ > 1 files changed, 4 insertions(+), 6 deletions(-) > > diff --git a/block/blk-merge.c b/block/blk-merge.c > index 5efc9e7..6e6d04b 100644 > --- a/block/blk-merge.c > +++ b/block/blk-merge.c > @@ -392,11 +392,6 @@ static int ll_merge_requests_fn(struct request_queue *q, > struct request *req, > return 0; > > total_phys_segments = req->nr_phys_segments + next->nr_phys_segments; > - if (blk_phys_contig_segment(q, req->biotail, next->bio)) > - total_phys_segments--; This is totally broken, I gather you are thinking the bug is fixed since the counts always match up. Well that's mainly because you know don't do ANY merging. > - if (total_phys_segments > q->max_phys_segments) > - return 0; Hmm? I think you should try and describe the problem you are seeing instead, then we can perhaps find the real issue. -- Jens Axboe