Bug 7343

Summary: another kernel fsync/fdatasync bug
Product: Mageia Reporter: AL13N <alien>
Component: RPM PackagesAssignee: Thomas Backlund <tmb>
Status: RESOLVED FIXED QA Contact:
Severity: critical    
Priority: release_blocker CC: thierry.vignaud, tmb
Version: Cauldron   
Target Milestone: ---   
Hardware: All   
OS: Linux   
Whiteboard: MGA1TOO MGA2TOO
Source RPM: kernel CVE:
Status comment:

Description AL13N 2012-09-04 14:29:04 CEST
see http://www.spinics.net/lists/linux-ext4/msg33716.html & http://www.spinics.net/lists/linux-ext4/msg33717.html

mga1 & mga2 too...

this caused for some mariadb data corruption on power loss, due to some change not being synced to disk (ie: the crash handler thinks it's synced, but it isn't...)
AL13N 2012-09-04 14:30:13 CEST

Priority: Normal => release_blocker
CC: (none) => tmb
Whiteboard: (none) => MGA1TOO MGA2TOO

Comment 1 Thierry Vignaud 2012-09-05 07:56:11 CEST
That could explain why I recently lost my .bash_history on reboot

CC: (none) => thierry.vignaud
Assignee: bugsquad => tmb

Comment 2 AL13N 2012-09-05 08:31:20 CEST
maybe not, because this one appears to be a corner case, ie: when the inode (size) has changed, but not the blocks. (iiuc)

for mariadb, it meant that the recovery crash handler has thought that a certain block was synced, while it wasn't, and thus recovered it wrongly, making it corrupt...
Comment 3 Manuel Hiebel 2012-11-24 18:53:03 CET
is this bug still valid with last kernel ? (kernel 3.4 for mga2, kernel 3.6 for cauldron)

Keywords: (none) => NEEDINFO

Comment 4 AL13N 2012-11-24 20:11:39 CET
well, since then, i've seen some more reports of ext4 fsync/fdatasync bugs...

it would be nice to know what is going on now (Thomas will likely know)
Comment 5 Thomas Backlund 2012-11-24 21:14:13 CET
Cauldron is ok.

mga2: ext3 bug squashed in 3.4.11, ext4 one in 3.4.14.

3.4.20 will be pushed to QA as soon as it is released.
Comment 6 Manuel Hiebel 2012-11-24 21:22:25 CET
ok thanks so need anymore of this one specific one.

Keywords: NEEDINFO => (none)
Status: NEW => RESOLVED
Resolution: (none) => FIXED