JFS Woes

[Collaborys]

I have a Gentoo 1.4 system configured with a mix of Ext3 (/boot), Ext2 (/tmp) and JFS (everything else) filesystems. I have separate partitions for /boot, /, /usr, /var, /opt, /home, /tmp and of course a swap partition. When I get a power outage (frequent in my part of the world), my gentoo system does not come cleanly. I have narrowed down the cause of this to be the fact that fsck.jfs (needed to clean the JFS filesystems) cannot itself be loaded because it is on a JFS filesystem. I have tried experimenting with /etc/fstab and the 'pass' parameter to force a fs integrity check at boot but fsck.jfs always appears to return an error, even though it reports that the filesystem is clean. My /etc/fstab is currently as follows:

/dev/hda1 /boot ext3 noauto,noatime 1 1
/dev/hda2 / jfs noatime 0 1
/dev/hda3 /usr jfs noatime 0 2
/dev/hda4 /home jfs noatime 0 2
/dev/hdc1 none swap sw 0 0
/dev/hdc2 /var jfs noatime 0 2
/dev/hdc3 /opt jfs noatime 0 2
/dev/hdc4 /tmp ext2 noatime 0 2

On a *normal* boot, I get an error message like:

fsck.jfs version version 1.1.2, 25-Mar-2003
The current device is: /dev/hda3
Block size in bytes: 4096
File system size in blocks: 2000092
Phase 0 - Replay Journal Log
File system is clean.
* Fsck could not correct all errors, manual repair needed [!!]

Give root password for maintenance
(or type Control-D for normal startup):

Typing Control-D leads to a perfectly normal boot.

Just what is happening? fsck.jfs seems to be reporting an error when there is none and thus intefering with the boot sequence.

Can anyone tell me what I'm doing wrong here?

[guitou]

Try running fsck.jfs in your shell...
If it is your root partition, you have to boot from a floppy/cdrom to avoid mounting it.

[mickywicky]

I have had exactly the same problem since I rebuilt my Gentoo installation using JFS, actually.
Running fsck manually return no errors; in fact I can just hit Ctrl+D at boto time, no problem at all.
Funny though my / and /home are JFS, and only my /home gives me this (at every boot): here a snippet of /etc/fstab

/dev/hda1 / jfs noatime 0 1
/dev/hda6 /home jfs noatime 0 2

This machine's kernel knows JFS and ext2/3 but I only use JFS.
Annoying since I use the machine remotely a lot, no way I can reboot unless I am sitting at the console.
Weird???? Maybe

Micky
_________________
l€€t = (1+1) * 3 * 3 * 37 = The prime factorization of The Beast

[Oopsz]

force an fsck on the disk, it may be marked clean but actually dirty.
_________________
Pop-before-SMTP with the Gentoo Virtual Mailhosting Guide

[mickywicky]

>Running fsck manually return no errors
I used to force fsck, but found out it ALWAYS does it regardless... Funny because it used to happen to my root partition, and when I changed the dump/pass options in fstab, all worked for this one... but no luck with /home. maybe I should just mkfs it all over again.
_________________
l€€t = (1+1) * 3 * 3 * 37 = The prime factorization of The Beast

[Collaborys]

I found the answer to this...

fsck.jfs returns the value '1' on an automatically corrected error OR when it replays the journal. So, at start-up when the boot sequence checks the partitions fsck.jfs always returns '1', causing the problems.

The solution is to remove the symbolic link in /sbin for fsck.jfs - it links to jfs_fsck. Replace it with a simple shell script that calls jfs_fsck and returns '0' if the return code is '1'.

[mickywicky]

clever... and why didn't I think about it to begin with??
guess I'll do just that.
or
(bad alternative)
use the old trick that worked on the first ReiserFS-enabled distros, which was to link it to /bin/true

Thanks for the help!!

Micky
_________________
l€€t = (1+1) * 3 * 3 * 37 = The prime factorization of The Beast

[zongsy]

Thank you very much for your advice on how to fix jfs root fs .

I have my /dev/hda5 as my root fs , formated as jfs , when the box shutdown improperly , It can not reboot normally , stops with the following message :

Quote:

* root file system can not be mounted read/write:(

I always boot with LIVECD and run fsck to fix the jfs filesystem.
But it is not efficient

Then I tried as you said :

Quote:

fsck.jfs returns the value '1' on an automatically corrected error OR when it replays the journal. So, at start-up when the boot sequence checks the partitions fsck.jfs always returns '1', causing the problems. The solution is to remove the symbolic link in /sbin for fsck.jfs - it links to jfs_fsck. Replace it with a simple shell script that calls jfs_fsck and returns '0' if the return code is '1'.

I rm /sbin/fsck.jfs symbolic link to /sbin/fs_fsck, replaced with my script:

Quote:

#!/bin/bash FAILURE=1 echo $@ /sbin/jfs_fsck $@ echo "$?" if [ "$?" -eq "$FAILURE" ] then exit 0 fi

however , still it stops at the same position , here is the messages I got in booting process

Quote:

*Remounting root fs read-only (if necessary) *checking root fs -a /dev/hda5 /sbin/jfs_fsck version 1.1.2 the current device is /dev/hda5 Blocksize in bytes 4096 file system size in blocks : 2443880 file system checked read-only file system is clean 1 *remounting root filesystem read/write... *root file system could not be mounted read/write:(

when I gave my root password to manually run fsck with this command
fsck -fpt jfs /dev/hda5
got the following:

Quote:

..... fsck 1.33 version... -fp /dev/hda5 the current device is /dev/hda5 Blocksize in bytes 4096 file system size in blocks : 2443880 phase 1-check blocks files/dirs and dir entries ......|... phase 2-count links phase 3-rescan for duplicate blocks and verify dir tree phase 4-report problems phase 5-check connectivity phase 6-perform approved corrections phase 7-verify file/dir allocation map control information Errors detected in the fileset file/dir allocation map phase 8-verify disk allocation maps incorrect data detected in disk allocation structure/control structure file system checked read-only file system is dirty fs is dirty but marked as clean.in its present state ,the results of accessing /dev/hda5 (except this utility) are undefined .

and I can not boot up still , I have no idea about it .
when root fs is mounted read-only , can fsck.jfs fix errors ?
_________________
gentoo so good

[zongsy]

my /etc/fstab :

[zongsy]

[Milez]

Hey, has anyone else tried this "return 1" fix and gotten it to work? My system has been jfs (except /boot) for a while now, and it used to reboot cleanly every time after a power surge, with fsck.jfs simply replaying the journal for a few seconds after hard reboots.

Now, however, my system won't reboot after a power failure unless I pop in the gentoo cd and do an fsck.jfs from that. The fact that it used to work definitely supports the idea that it is simply a problem with init scripts and stuff - so, like I was sayin, anyone had a success?
_________________
-=Miles Stoudenmire=-

Adopt a post initiative

[petrjanda]

I dont like JFS at all, it's slow, the only advantage is the lowest CPU usage.

[WhackAMole]

I had the problem where it could not mount the root fs as read/write. When I changed my fstab to read "defaults" instead of "noatime and some-other-options-I-was-trying" everything went fine.

I can't say that this would fix the other problem being discussed, but who knows?