fsck -y leaves problems behind

[Kevin O'Gorman]

Fixing an ext4 file system required multiple passes. This is counter-intuitive in a program that will sometimes say something to the effect that "problems remain" but did not do so this time.

ProblemType: Bug
DistroRelease: Ubuntu 11.04
Package: e2fsprogs 1.41.14-1ubuntu3
ProcVersionSignature: Ubuntu 2.6.38-11.50-generic-pae 2.6.38.8
Uname: Linux 2.6.38-11-generic-pae i686
Architecture: i386
Date: Sat Oct 1 14:33:30 2011
InstallationMedia: Ubuntu 11.04 "Natty Narwhal" - Release i386 (20110427.1)
ProcEnviron:
 LANGUAGE=en_US:en
 PATH=(custom, user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: e2fsprogs
UpgradeStatus: No upgrade log present (probably fresh install)

[Kevin O'Gorman (kogorman-pacbell)]

The attachments are the three runs required to verify a clean filesystem. I would expect the first run to do it and the second to verify it. I did not expect the second run to find remaining or created problems.

Each of the three includes commands that verify that the partition was not mounted.

[Dave Gilbert (ubuntu-treblig)]

Well in the 1st run it does warn you that something odd happened:

WARNING: PROGRAMMING BUG IN E2FSCK!
 OR SOME BONEHEAD (YOU) IS CHECKING A MOUNTED (LIVE) FILESYSTEM.
inode_link_info[23300069] is 2, inode.i_links_count is 1. They should be the same!
Inode 23300069 ref count is 1, should be 1. Fix? yes

so I think after that warning all bets are off - and I can see from your df -h that your /dev/sdg1 isn't mounted,
so it does look like an e2fsck bug it got to that point.

Can I ask what did you do to your fs to get it into this state where the fsck started off with those
errors and can you describe the hardware.

Dave

[Theodore Ts'o (tytso)]

I suppose I should change that message to be "programming bug in e2fsck, checking a live file system, dodgy hardware/RAID array out of sync".

In this case, I'm going to put my money on the last one, and specifically, that you have a hardware problem. The errors that showed up in the second fsck run are localized to a specific inode table block that holds inodes 2330067, 2330069, 23300074, and 2330076. In the first fsck run, fsck complained about a programming error or that the file system was mounted read-write. Basically, that happens when the value that fsck had cached changed out from under it. This normally happens when the file system is mount and it's being modified while fsck runs, but it can also happen when the hardware is dodgy and returns one set of values at one time, and another set of values at another. (One cause of this is a RAID1 array which is out of sync, so you get the value from one side of the mirror, and then later you get the value on the other side of the mirror.)

In this case, we specifically modified the inode table block containing 23300074 during the first check, only to have a different set of complains on the second pass. That indicates to me that the disk is dodgy, and the contents which we wrote during the first pass was different from the set of problems you had on the second pass.

[Kevin O'Gorman (kogorman-pacbell)]

I'm not sure what I did. The drive is an "internal" 2 TB drive that I use in a
drive dock, only for backups. It had not been checked in a long time and was
seriously hosed. That is why I ran fsck with -y: to get to a sane state where I
would not damage things further by doing file operations. I wound up with over
100 things in lost+found, almost all empty files and directories.

I have about 8 of these in various sizes. Very handy for off-site backups (keep
one in the house of a friend in the next town, rotate, repeat). The systems
they run on are on a reliable UPS, but that does not help much when Linux
crashes. Since these migrate, they are not automounted and not subject to
boot-time fsck scrutiny. I'll have to do something about that.

 Kevin O'Gorman, Ph.D. mailto:<email address hidden>
Home Page: http://www.kosmanor.com/~kevin

________________________________
From: Dave Gilbert <email address hidden>
To: <email address hidden>
Sent: Sat, October 1, 2011 3:57:43 PM
Subject: [Bug 864344] Re: fsck -y leaves problems behind

Well in the 1st run it does warn you that something odd happened:

WARNING: PROGRAMMING BUG IN E2FSCK!
    OR SOME BONEHEAD (YOU) IS CHECKING A MOUNTED (LIVE) FILESYSTEM.
inode_link_info[23300069] is 2, inode.i_links_count is 1. They should be the
same!
Inode 23300069 ref count is 1, should be 1. Fix? yes

so I think after that warning all bets are off - and I can see from your df -h
that your /dev/sdg1 isn't mounted,
so it does look like an e2fsck bug it got to that point.

Can I ask what did you do to your fs to get it into this state where the fsck
started off with those
errors and can you describe the hardware.

Dave

--
You received this bug notification because you are subscribed to the bug
report.
https://bugs.launchpad.net/bugs/864344

Title:
  fsck -y leaves problems behind

Status in “e2fsprogs” package in Ubuntu:
  New

Bug description:
  Fixing an ext4 file system required multiple passes. This is
  counter-intuitive in a program that will sometimes say something to
  the effect that "problems remain" but did not do so this time.

  ProblemType: Bug
  DistroRelease: Ubuntu 11.04
  Package: e2fsprogs 1.41.14-1ubuntu3
  ProcVersionSignature: Ubuntu 2.6.38-11.50-generic-pae 2.6.38.8
  Uname: Linux 2.6.38-11-generic-pae i686
  Architecture: i386
  Date: Sat Oct 1 14:33:30 2011
  InstallationMedia: Ubuntu 11.04 "Natty Narwhal" - Release i386 (20110427.1)
  ProcEnviron:
   LANGUAGE=en_US:en
   PATH=(custom, user)
   LANG=en_US.UTF-8
   SHELL=/bin/bash
  SourcePackage: e2fsprogs
  UpgradeStatus: No upgrade log present (probably fresh install)

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/e2fsprogs/+bug/864344/+subscriptions