2.6.33-git7: serious filesystem errors

Anyone else testing Quirky 008 getting any messages at bootup (or by running 'dmesg') that look like this?:

EXT2-fs (loop1): error: ext2_lookup: deleted inode referenced: 12334

I've been getting lots of them. A problem with Aufs? It was ok with the 2.6.33-rc8 kernel.

There are files in /var/run that are supposed to be deleted. These are files that dhcpcd uses, which breaks dhcpcd.

This problem is occurring during normal operation, not after power failure or any other abnormal condition.

I did a fsck on the 'quirkysave' file and that fixed it. For example:

/mnt/home/quirkysave-4MAR2010.2fs: Entry 'dhcpcd.sock' in /var/run (12301) has deleted/unused inode 12334. CLEARED.

This is not good.


Posted on 7 Mar 2010, 22:08


Comments:

Posted on 7 Mar 2010, 23:11 by tronkel
Filesystem errors
Just had a look at dmesg output.
Seems OK as far as file system is concerned. Nothing about deleted inodes at any rate.

Lost the Alsa sound earlier today for no apparent reason. Seemed to need re-configuring using the Alsa wizard. Although the wizard didn't pick up all of the installed sound cards I configured it for the one it did find. Then went to Retrovol and re-enabled the correct channels. Sound then worked again.

This sort of loss of data my well be as a result of a dodgy file system though.

On the other hand the problem my be unique to specific hardware.


Posted on 7 Mar 2010, 23:39 by pemasu2
Filesystem errors
I have got same errors with ext2 savefile. And Wireless didnt connect. Also X crashed and I had to boot another Puppy to write this. e2fsck didnt resolve X problem.


Posted on 7 Mar 2010, 23:41 by Lobster
dmesg
getting this in dmesg if any use . . .

EXT3-fs (sda5): warning: maximal mount count reached, running e2fsck is recommended

EXT3-fs (sda5): using internal journal
EXT3-fs (sda5): mounted filesystem with writeback data mode
ISO 9660 Extensions: RRIP_1991A
usb 1-1: USB disconnect, address 2
kjournald starting. Commit interval 5 seconds
EXT3-fs (sda5): warning: maximal mount count reached, running e2fsck is recommended
EXT3-fs (sda5): using internal journal
EXT3-fs (sda5): mounted filesystem with writeback data mode




Posted on 8 Mar 2010, 6:35 by X
Q8
About 8 boots haven't noticed anything unusual except lobsters root login on tty1


Posted on 8 Mar 2010, 8:21 by jim1911
Pupsave file enhancements
Please take a look at TazOC's latest Lighthouse. He has increased personal storage file (pupsave) size choices ranging up to the limit for a FAT32 partition (nearly 4GB). Resize pupsave utility updated with option to resize by 1024M as well as smaller values.

Also, and more significant, the pupsave file system is created and mounted ext3 on an ext3/ext4/reiserfs host partition incorporating tune2fs -o journal_data_ordered for the 3fs pupsave. The proper pupsave type is automatically selected. This is probably the way to go with the later kernels.

Many of us are using 3fs save file and modifying the initrd.gz file so that it is mounted as a journaling file system on ext3/4 partitions. I have also done this with Quirky 008 and have found it to be very stable on my hardware.
Thanks,
Jim







Posted on 8 Mar 2010, 21:01 by BarryK
ext2/3 drivers dropped
This might be significant. In the 2.6.33 release notes is this statement:

Use ext4 file system driver for ext2/ext3 file system mounts

This would mean, I think, that the ext2 and ext3 drivers have been dropped, as the ext4 driver is supposed to handle those filesystems.



Posted on 9 Mar 2010, 23:12 by jemimah
ext4 driver
I've been using the ext4 driver for ext2 filesystems with this kernel in Puppeee for the past few days ant it's working fine. It did complain about some errors in my save file, but I presume those were already there and it seems happy now that I've fscked it.






Posted on 10 Mar 2010, 5:47 by pemasu
Filesystem errors
EXT2-fs (loop1): error: ext2_lookup: deleted inode referenced: 18478
EXT2-fs (loop1): error: ext2_free_blocks: bit already cleared for block 71177

Sometimes X works.. Sometimes wireless connects to net but sometimes it does not. Interesting behaviour. What is sure is filesystem errors and my ext3 partition has even wrong date when I e2fsck it after using 008.



Posted on 11 Mar 2010, 7:49 by Jemimah
filesystem
I'm definitely seeing some filesystem issues now in save files created with the new kernel. I wonder if it has something to do with the save file not being unmounted cleanly at shutdown. I guess I'll see if my users have issues with the next Puppeee beta.


Posted on 13 Mar 2010, 16:30 by BarryK
Now it's behaving
You know the saying "a watched pot never boils", well now I can't get any filesystem errors. I have been watching ever since I started this blog thread.



Posted on 13 Mar 2010, 23:36 by aarf
file system errors
something strange is indeed foot. just had to replace the .sfs in all of quirky6, quirky8 and upup458 because they all refused to boot i.e. the kernel panic-not syncing routine. dpup484 which is on the same FAT SDcard as the others was not affected however when i closed dpup458 it had the errors mentioned at the beginning of barry's initial post flash by after the last pupsaving dialog. usually in the past it was the pupsave causing the non-boot issues, this time the pupsaves were all ok and didn't need changing.

except however:
# e2fsck quirkysave-q6.2fs
e2fsck 1.41.9 (22-Aug-2009)
Superblock last mount time (Sun Mar 14 05:34:59 2010,
now = Sat Mar 13 22:00:44 2010) is in the future.
Fix<y>? yes

Superblock last write time (Sun Mar 14 05:56:35 2010,
now = Sat Mar 13 22:00:44 2010) is in the future.
Fix<y>? yes

quirkysave-q6.2fs was not cleanly unmounted, check forced.
Pass 1: Checking inodes, blocks, and sizes
Pass 2: Checking directory structure
Pass 3: Checking directory connectivity
Pass 4: Checking reference counts
Pass 5: Checking group summary information

quirkysave-q6.2fs: ***** FILE SYSTEM WAS MODIFIED *****
quirkysave-q6.2fs: 854/4776 files (2.6% non-contiguous), 7580/19072 blocks
#
quirky6 with pupsave still failed to boot: x failed to start and still refused after recommended coaxing. also had "root login on 'UNKNOWN'" in the dialog.
next boot of dpup458 didn't have the previous error messages on close but another message "unmounting stray file systems."


quirky8 has again died and i am not altogether certain that my opera portable 10.10 was not the cause of the problem. by locking up the FAT file system into read-only mode with corrupted files.


Posted on 17 Mar 2010, 11:48 by Jemimah
fs corruption
I definitely had orphaned inodes every time I rebooted and ran an fsck with aufs and 2.6.33-git7. I switched to unionfs and the problem went away. A few users had obvious trouble too.

Also you might want to test to see if openoffice writer crashes the kernel on quirky 08 as it did on Puppeee. I think it triggers a rather serious scheduler or process table bug - but the problem is fixed in 2.6.33.1.


Posted on 19 Mar 2010, 9:52 by f00-
fs errors (p431s10/q008)
On attempted reboot of p431 (stardust 10 remaster, liveCD boot + usb savefile), bootmssg:

EXT2-fs error (device loop1): ext2_check_page: bad entry in directory #10283: :
unaligned directory entry - offset=0, inode=4167987830, rec_len=60450, name_len=99

..and then the "Kernel panic - not syncing: Attempted to kill init!" (but graceful, no flashing kybd leds or violent signs of distress). My q008 is on a multisession liveCD (but the remaster build files are on that same usb stick albeit a few levels deep and in another partition from the p431s10 savefile). Curious, it is - the p431s10 savefile (or bootCD) may have some anomalies with /lib/modules/2.6.30.5/initrd/kernel/drivers/usb/core/usbcore.ko and ../hid/usbhid.ko as well (seems to be an issue lately with using savefile on usb stick for me). Perhaps q010 will help!