kernel-small_fixes.patch

   1 --- linux-2.6.33/scripts/mod/modpost.c~ 2010-02-24 19:52:17.000000000 +0100
   2 +++ linux-2.6.33/scripts/mod/modpost.c  2010-03-07 14:26:47.242168558 +0100
   3 @@ -15,7 +15,8 @@
   4  #include <stdio.h>
   5  #include <ctype.h>
   6  #include "modpost.h"
   7 -#include "../../include/generated/autoconf.h"
   8 +// PLD architectures don't use CONFIG_SYMBOL_PREFIX
   9 +//#include "../../include/generated/autoconf.h"
  10  #include "../../include/linux/license.h"
  11
  12  /* Some toolchains use a `_' prefix for all user symbols. */
  13
  14 --- linux-3.0/scripts/kconfig/lxdialog/check-lxdialog.sh~       2011-07-22 04:17:23.000000000 +0200
  15 +++ linux-3.0/scripts/kconfig/lxdialog/check-lxdialog.sh        2011-08-25 21:26:04.799150642 +0200
  16 @@ -9,6 +9,12 @@
  17                         $cc -print-file-name=lib${lib}.${ext} | grep -q /
  18                         if [ $? -eq 0 ]; then
  19                                 echo "-l${lib}"
  20 +                               for libt in tinfow tinfo ; do
  21 +                                       $cc -print-file-name=lib${libt}.${ext} | grep -q /
  22 +                                       if [ $? -eq 0 ]; then
  23 +                                               echo "-l${libt}"
  24 +                                       fi
  25 +                               done
  26                                 exit
  27                         fi
  28                 done
  29
  30
  31 Hi all,
  32  There is a risk of data loss with md/raid6 arrays running on Linux since
  33  2.6.32.
  34  If:
  35    - the array is doubly degraded
  36    - one or both failed devices are being recovered, and
  37    - the array is written to
  38
  39  then it is possible for data on the array to be lost.  The patch below fixes
  40  the problem.  If you apply the patch to an older kernel which has separate
  41  handle_stripe5() and handle_stripe6() functions, be sure that patch changes
  42  handle_stripe6().
  43
  44  There is no risk to an optimal array or a singly-degraded array.  There is
  45  also no risk on a doubly-degraded array which is not recovering a device or
  46  is not receiving write requests.
  47
  48  If you have data on a RAID6 array, please consider how to avoid corruption,
  49  possibly by applying the patch, possibly by removing any hot spares so
  50  recovery does not automatically start.
  51
  52  This patch will be sent upstream shortly and will subsequently appear in
  53  future "-stable" kernels.
  54
  55 NeilBrown
  56
  57 From f94e37dce722ec7b6666fd04be357f422daa02b5 Mon Sep 17 00:00:00 2001
  58 From: NeilBrown <neilb@suse.de>
  59 Date: Wed, 13 Aug 2014 09:57:07 +1000
  60 Subject: [PATCH] md/raid6: avoid data corruption during recovery of
  61  double-degraded RAID6
  62
  63 During recovery of a double-degraded RAID6 it is possible for
  64 some blocks not to be recovered properly, leading to corruption.
  65
  66 If a write happens to one block in a stripe that would be written to a
  67 missing device, and at the same time that stripe is recovering data
  68 to the other missing device, then that recovered data may not be written.
  69
  70 This patch skips, in the double-degraded case, an optimisation that is
  71 only safe for single-degraded arrays.
  72
  73 Bug was introduced in 2.6.32 and fix is suitable for any kernel since
  74 then.  In an older kernel with separate handle_stripe5() and
  75 handle_stripe6() functions that patch must change handle_stripe6().
  76
  77 Cc: stable@vger.kernel.org (2.6.32+)
  78 Fixes: 6c0069c0ae9659e3a91b68eaed06a5c6c37f45c8
  79 Cc: Yuri Tikhonov <yur@emcraft.com>
  80 Cc: Dan Williams <dan.j.williams@intel.com>
  81 Reported-by: "Manibalan P" <pmanibalan@amiindia.co.in>
  82 Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1090423
  83 Signed-off-by: NeilBrown <neilb@suse.de>
  84
  85 diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
  86 index 6b2d615d1094..183588b11fc1 100644
  87 --- a/drivers/md/raid5.c
  88 +++ b/drivers/md/raid5.c
  89 @@ -3817,6 +3817,8 @@ static void handle_stripe(struct stripe_head *sh)
  90                                 set_bit(R5_Wantwrite, &dev->flags);
  91                                 if (prexor)
  92                                         continue;
  93 +                               if (s.failed > 1)
  94 +                                       continue;
  95                                 if (!test_bit(R5_Insync, &dev->flags) ||
  96                                     ((i == sh->pd_idx || i == sh->qd_idx)  &&
  97                                      s.failed == 0))
  98
  99