src/patches/suse-2.6.27.39/patches.fixes/block-failfast-merge-fix

   1 From ab0fd1debe730ec9998678a0c53caefbd121ed10 Mon Sep 17 00:00:00 2001
   2 From: Tejun Heo <tj@kernel.org>
   3 Date: Fri, 3 Jul 2009 12:56:18 +0200
   4 Subject: [PATCH] block: don't merge requests of different failfast settings
   5 References: bnc#519111
   6
   7 Block layer used to merge requests and bios with different failfast
   8 settings.  This caused regular IOs to fail prematurely when they were
   9 merged into failfast requests for readahead.
  10
  11 Niel Lambrechts could trigger the problem semi-reliably on ext4 when
  12 resuming from STR.  ext4 uses readahead when reading inodes and
  13 combined with the deterministic extra SATA PHY exception cycle during
  14 resume on the specific configuration, non-readahead inode read would
  15 fail causing ext4 errors.  Please read the following thread for
  16 details.
  17
  18   http://lkml.org/lkml/2009/5/23/21
  19
  20 This patch makes block layer reject merging if the failfast settings
  21 don't match.  This is correct but likely to lower IO performance by
  22 preventing regular IOs from mingling into surrounding readahead
  23 requests.  Changes to allow such mixed merges and handle errors
  24 correctly will be added later.
  25
  26 Signed-off-by: Tejun Heo <tj@kernel.org>
  27 Reported-by: Niel Lambrechts <niel.lambrechts@gmail.com>
  28 Cc: Theodore Tso <tytso@mit.edu>
  29 Signed-off-by: Jens Axboe <axboe@carl.(none)>
  30 Signed-off-by: Tejun Heo <teheo@suse.de>
  31 ---
  32  block/blk-merge.c |    6 ++++++
  33  block/elevator.c  |    8 ++++++++
  34  2 files changed, 14 insertions(+)
  35
  36 --- a/block/blk-merge.c
  37 +++ b/block/blk-merge.c
  38 @@ -376,6 +376,12 @@ static int attempt_merge(struct request_
  39         if (blk_integrity_rq(req) != blk_integrity_rq(next))
  40                 return 0;
  41
  42 +       /* don't merge requests of different failfast settings */
  43 +       if (blk_failfast_dev(req)       != blk_failfast_dev(next)       ||
  44 +           blk_failfast_transport(req) != blk_failfast_transport(next) ||
  45 +           blk_failfast_driver(req)    != blk_failfast_driver(next))
  46 +               return 0;
  47 +
  48         /*
  49          * If we are allowed to merge, then append bio list
  50          * from next to rq and release next. merge_requests_fn
  51 --- a/block/elevator.c
  52 +++ b/block/elevator.c
  53 @@ -97,6 +97,14 @@ int elv_rq_merge_ok(struct request *rq,
  54         if (bio_integrity(bio) != blk_integrity_rq(rq))
  55                 return 0;
  56
  57 +       /*
  58 +        * Don't merge if failfast settings don't match
  59 +        */
  60 +       if (!bio_failfast_dev(bio)       != !blk_failfast_dev(rq)       ||
  61 +           !bio_failfast_transport(bio) != !blk_failfast_transport(rq) ||
  62 +           !bio_failfast_driver(bio)    != !blk_failfast_driver(rq))
  63 +               return 0;
  64 +
  65         if (!elv_iosched_allow_merge(rq, bio))
  66                 return 0;
  67