[dm-devel] [PATCH 6/4 v2] dm-raid: enhance discard_zeroes_data documentation
heinzm at redhat.com
heinzm at redhat.com
Fri May 15 18:35:10 UTC 2015
From: Heinz Mauelshagen <heinzm at redhat.com>
This patch adds documentation explaining background of
discard_zeroes_data issues with raid4/5/6 and the related
devices_handle_discards_safely module parameter.
Signed-of-by: Heinz Mauelshagen <heinzm at redhat.com>
---
Documentation/device-mapper/dm-raid.txt | 21 +++++++++++++++++++++
1 file changed, 21 insertions(+)
diff --git a/Documentation/device-mapper/dm-raid.txt b/Documentation/device-mapper/dm-raid.txt
index bceb865..713739b 100644
--- a/Documentation/device-mapper/dm-raid.txt
+++ b/Documentation/device-mapper/dm-raid.txt
@@ -226,4 +226,25 @@ Version History
1.5.2 'mismatch_cnt' is zero unless [last_]sync_action is "check".
1.6.0 Add devices_handle_discards_safely module parameter to
allow discards on disks anouncing uncertain discard_zeroes_data
+ (see *discard* below for more details)
1.7.0 Add support for MD RAID0 mappings
+
+*discard'
+The MD raid4/5/6 personalities trust discard_zeroes_data ensuring
+to return zeroes as block content after a discard and thus avoid
+rewriting them; the zeroed content is used to reconstruct data
+on a disk failure thus leading to corrupion in any reconstructed
+block derived from discarded ones.
+
+Field reports showed, that an unknown number of disks fail supporting
+discard_zeroes_data properly, hence they may return non-zero block
+content after a discard.
+
+In order to avoid data corruption on recovery of failed RAID device
+content using discarded blocks, we have to avoid using the feature
+by default.
+
+For convenience, the module parameter devices_handle_discards_safely
+allows users knowing for sure that discard_zeroes_data works properly
+on all their raid disks to overwrite the default and make use
+of the discard_zeroes_data optimization.
--
2.1.0
More information about the dm-devel
mailing list