From: rpeterso@redhat.com <rpeterso@redhat.com>
To: cluster-devel.redhat.com
Subject: [Cluster-devel] [PATCH 11/42] fsck.gfs2: Special case '..' when processing bad formal inode number
Date: Mon, 8 Apr 2013 07:40:43 -0700 [thread overview]
Message-ID: <1365432074-17615-12-git-send-email-rpeterso@redhat.com> (raw)
In-Reply-To: <1365432074-17615-1-git-send-email-rpeterso@redhat.com>
From: Bob Peterson <rpeterso@redhat.com>
In a recent patch to fsck.gfs2, we added the ability to make sure the
formal inode number in each directory entry matches the formal inode
number in the dinode. If it doesn't match, fsck tries to fix it up.
We can't do much for regular files, but we can fix up directories.
If the directory linkage is intact, it just fixes the formal inode number.
But to check if the directory linkage is intact, we were checking to make
sure the child directory points to the parent with its "..". For example,
suppose we have gfs2 mounted as /mnt/gfs2, and at the root, we have
directory "a", and within "a" we have a subdirectory "b". In other words:
/mnt/gfs2/a/b/...
Now suppose fsck.gfs2 finds a formal inode number mismatch between the
dirent inside "a" which points to "b" and the inode "b" itself. Since
both "a" and "b" are directories, it tries to determine if the directory
linkage is intact by testing whether b's ".." dirent actually points
back to "a". And if it's good, we can just fix the formal inode number
so that they match.
That's all well and good, and works for the most part. However, if
the dirent found to be wrong isn't "b" but ".." we've got a problem.
Today's algorithm would look up the ".." of ".." which won't be
pointingi back to what we want.
For this patch, I'm special-casing ".." and making it just delete the
correct directory entry. However, we have to do it in such a way that
it doesn't decrement di_entries, since the entry is invalid.
rhbz#902920
---
gfs2/fsck/pass2.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/gfs2/fsck/pass2.c b/gfs2/fsck/pass2.c
index a71be4b..5d8c2b6 100644
--- a/gfs2/fsck/pass2.c
+++ b/gfs2/fsck/pass2.c
@@ -260,7 +260,7 @@ static int bad_formal_ino(struct gfs2_inode *ip, struct gfs2_dirent *dent,
(unsigned long long)entry.no_formal_ino,
(unsigned long long)ii->di_num.no_formal_ino,
(unsigned long long)ii->di_num.no_formal_ino);
- if (q != gfs2_inode_dir) {
+ if (q != gfs2_inode_dir || !strcmp("..", tmp_name)) {
if (query( _("Remove the corrupt directory entry? (y/n) ")))
return 1;
log_err( _("Corrupt directory entry not removed.\n"));
--
1.7.11.7
next prev parent reply other threads:[~2013-04-08 14:40 UTC|newest]
Thread overview: 43+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-04-08 14:40 [Cluster-devel] [PATCH 00/42] fsck.gfs2 fixes and improvements rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 01/42] libgfs2: externalize dir_split_leaf rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 02/42] libgfs2: allow dir_split_leaf to receive a leaf buffer rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 03/42] libgfs2: let dir_split_leaf receive a "broken" lindex rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 04/42] fsck.gfs2: Move function find_free_blk to util.c rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 05/42] fsck.gfs2: Split out function to make sure lost+found exists rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 06/42] fsck.gfs2: Check for formal inode mismatch when adding to lost+found rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 07/42] fsck.gfs2: shorten some debug messages in lost+found rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 08/42] fsck.gfs2: Move basic directory entry checks to separate function rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 09/42] fsck.gfs2: Add formal inode check to basic dirent checks rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 10/42] fsck.gfs2: Add new function to check dir hash tables rpeterso
2013-04-08 14:40 ` rpeterso [this message]
2013-04-08 14:40 ` [Cluster-devel] [PATCH 12/42] fsck.gfs2: Move function to read directory hash table to util.c rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 13/42] fsck.gfs2: Misc cleanups rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 14/42] fsck.gfs2: Verify dirent hash values correspond to proper leaf block rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 15/42] fsck.gfs2: re-read hash table if directory height or depth changes rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 16/42] fsck.gfs2: fix leaf blocks, don't try to patch the hash table rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 17/42] fsck.gfs2: check leaf depth when validating leaf blocks rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 18/42] fsck.gfs2: small cleanups rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 19/42] fsck.gfs2: reprocess inodes when blocks are added rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 20/42] fsck.gfs2: Remove redundant leaf depth check rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 21/42] fsck.gfs2: link dinodes that only have extended attribute problems rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 22/42] fsck.gfs2: Add clarifying message to duplicate processing rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 23/42] fsck.gfs2: separate function to calculate metadata block header size rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 24/42] fsck.gfs2: Rework the "undo" functions rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 25/42] fsck.gfs2: Check for interrupt when resolving duplicates rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 26/42] fsck.gfs2: Consistent naming of struct duptree variables rpeterso
2013-04-08 14:40 ` [Cluster-devel] [PATCH 27/42] fsck.gfs2: Keep proper counts when duplicates are found rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 28/42] fsck.gfs2: print metadata block reference on data errors rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 29/42] fsck.gfs2: print block count values when fixing them rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 30/42] fsck.gfs2: Do not invalidate metablocks of dinodes with invalid mode rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 31/42] fsck.gfs2: Log when unrecoverable data block errors are encountered rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 32/42] fsck.gfs2: don't remove buffers from the list when errors are found rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 33/42] fsck.gfs2: Don't flag GFS1 non-dinode blocks as duplicates rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 34/42] fsck.gfs2: externalize check_leaf rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 35/42] fsck.gfs2 pass2: check leaf blocks when fixing hash table rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 36/42] fsck.gfs2: standardize check_metatree return codes rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 37/42] fsck.gfs2: don't invalidate files with duplicate data block refs rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 38/42] fsck.gfs2: check for duplicate first references rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 39/42] fsck.gfs2: When flagging a duplicate reference, show valid or invalid rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 40/42] fsck.gfs2: major duplicate reference reform rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 41/42] fsck.gfs2: Remove all bad eattr blocks rpeterso
2013-04-08 14:41 ` [Cluster-devel] [PATCH 42/42] fsck.gfs2: Remove unused variable rpeterso
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1365432074-17615-12-git-send-email-rpeterso@redhat.com \
--to=rpeterso@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).