releases/5.1.3/don-t-jump-to-compute_result-state-from-check_result-state.patch

   1 From 4f4fd7c5798bbdd5a03a60f6269cf1177fbd11ef Mon Sep 17 00:00:00 2001
   2 From: Nigel Croxon <ncroxon@redhat.com>
   3 Date: Fri, 29 Mar 2019 10:46:15 -0700
   4 Subject: Don't jump to compute_result state from check_result state
   5
   6 From: Nigel Croxon <ncroxon@redhat.com>
   7
   8 commit 4f4fd7c5798bbdd5a03a60f6269cf1177fbd11ef upstream.
   9
  10 Changing state from check_state_check_result to
  11 check_state_compute_result not only is unsafe but also doesn't
  12 appear to serve a valid purpose.  A raid6 check should only be
  13 pushing out extra writes if doing repair and a mis-match occurs.
  14 The stripe dev management will already try and do repair writes
  15 for failing sectors.
  16
  17 This patch makes the raid6 check_state_check_result handling
  18 work more like raid5's.  If somehow too many failures for a
  19 check, just quit the check operation for the stripe.  When any
  20 checks pass, don't try and use check_state_compute_result for
  21 a purpose it isn't needed for and is unsafe for.  Just mark the
  22 stripe as in sync for passing its parity checks and let the
  23 stripe dev read/write code and the bad blocks list do their
  24 job handling I/O errors.
  25
  26 Repro steps from Xiao:
  27
  28 These are the steps to reproduce this problem:
  29 1. redefined OPT_MEDIUM_ERR_ADDR to 12000 in scsi_debug.c
  30 2. insmod scsi_debug.ko dev_size_mb=11000  max_luns=1 num_tgts=1
  31 3. mdadm --create /dev/md127 --level=6 --raid-devices=5 /dev/sde1 /dev/sde2 /dev/sde3 /dev/sde5 /dev/sde6
  32 sde is the disk created by scsi_debug
  33 4. echo "2" >/sys/module/scsi_debug/parameters/opts
  34 5. raid-check
  35
  36 It panic:
  37 [ 4854.730899] md: data-check of RAID array md127
  38 [ 4854.857455] sd 5:0:0:0: [sdr] tag#80 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
  39 [ 4854.859246] sd 5:0:0:0: [sdr] tag#80 Sense Key : Medium Error [current]
  40 [ 4854.860694] sd 5:0:0:0: [sdr] tag#80 Add. Sense: Unrecovered read error
  41 [ 4854.862207] sd 5:0:0:0: [sdr] tag#80 CDB: Read(10) 28 00 00 00 2d 88 00 04 00 00
  42 [ 4854.864196] print_req_error: critical medium error, dev sdr, sector 11656 flags 0
  43 [ 4854.867409] sd 5:0:0:0: [sdr] tag#100 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
  44 [ 4854.869469] sd 5:0:0:0: [sdr] tag#100 Sense Key : Medium Error [current]
  45 [ 4854.871206] sd 5:0:0:0: [sdr] tag#100 Add. Sense: Unrecovered read error
  46 [ 4854.872858] sd 5:0:0:0: [sdr] tag#100 CDB: Read(10) 28 00 00 00 2e e0 00 00 08 00
  47 [ 4854.874587] print_req_error: critical medium error, dev sdr, sector 12000 flags 4000
  48 [ 4854.876456] sd 5:0:0:0: [sdr] tag#101 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
  49 [ 4854.878552] sd 5:0:0:0: [sdr] tag#101 Sense Key : Medium Error [current]
  50 [ 4854.880278] sd 5:0:0:0: [sdr] tag#101 Add. Sense: Unrecovered read error
  51 [ 4854.881846] sd 5:0:0:0: [sdr] tag#101 CDB: Read(10) 28 00 00 00 2e e8 00 00 08 00
  52 [ 4854.883691] print_req_error: critical medium error, dev sdr, sector 12008 flags 4000
  53 [ 4854.893927] sd 5:0:0:0: [sdr] tag#166 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
  54 [ 4854.896002] sd 5:0:0:0: [sdr] tag#166 Sense Key : Medium Error [current]
  55 [ 4854.897561] sd 5:0:0:0: [sdr] tag#166 Add. Sense: Unrecovered read error
  56 [ 4854.899110] sd 5:0:0:0: [sdr] tag#166 CDB: Read(10) 28 00 00 00 2e e0 00 00 10 00
  57 [ 4854.900989] print_req_error: critical medium error, dev sdr, sector 12000 flags 0
  58 [ 4854.902757] md/raid:md127: read error NOT corrected!! (sector 9952 on sdr1).
  59 [ 4854.904375] md/raid:md127: read error NOT corrected!! (sector 9960 on sdr1).
  60 [ 4854.906201] ------------[ cut here ]------------
  61 [ 4854.907341] kernel BUG at drivers/md/raid5.c:4190!
  62
  63 raid5.c:4190 above is this BUG_ON:
  64
  65     handle_parity_checks6()
  66         ...
  67         BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
  68
  69 Cc: <stable@vger.kernel.org> # v3.16+
  70 OriginalAuthor: David Jeffery <djeffery@redhat.com>
  71 Cc: Xiao Ni <xni@redhat.com>
  72 Tested-by: David Jeffery <djeffery@redhat.com>
  73 Signed-off-by: David Jeffy <djeffery@redhat.com>
  74 Signed-off-by: Nigel Croxon <ncroxon@redhat.com>
  75 Signed-off-by: Song Liu <songliubraving@fb.com>
  76 Signed-off-by: Jens Axboe <axboe@kernel.dk>
  77 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  78
  79 ---
  80  drivers/md/raid5.c |   19 ++++---------------
  81  1 file changed, 4 insertions(+), 15 deletions(-)
  82
  83 --- a/drivers/md/raid5.c
  84 +++ b/drivers/md/raid5.c
  85 @@ -4223,26 +4223,15 @@ static void handle_parity_checks6(struct
  86         case check_state_check_result:
  87                 sh->check_state = check_state_idle;
  88
  89 +               if (s->failed > 1)
  90 +                       break;
  91                 /* handle a successful check operation, if parity is correct
  92                  * we are done.  Otherwise update the mismatch count and repair
  93                  * parity if !MD_RECOVERY_CHECK
  94                  */
  95                 if (sh->ops.zero_sum_result == 0) {
  96 -                       /* both parities are correct */
  97 -                       if (!s->failed)
  98 -                               set_bit(STRIPE_INSYNC, &sh->state);
  99 -                       else {
 100 -                               /* in contrast to the raid5 case we can validate
 101 -                                * parity, but still have a failure to write
 102 -                                * back
 103 -                                */
 104 -                               sh->check_state = check_state_compute_result;
 105 -                               /* Returning at this point means that we may go
 106 -                                * off and bring p and/or q uptodate again so
 107 -                                * we make sure to check zero_sum_result again
 108 -                                * to verify if p or q need writeback
 109 -                                */
 110 -                       }
 111 +                       /* Any parity checked was correct */
 112 +                       set_bit(STRIPE_INSYNC, &sh->state);
 113                 } else {
 114                         atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
 115                         if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery)) {