Created attachment 53012 [details] kernel .config Hello! After each system boot this message pops: kernel: EXT4-fs (sda10): error count: 5 kernel: EXT4-fs (sda10): initial error at 1301325048: htree_dirblock_to_tree:586: inode 260099: block 1056737 kernel: EXT4-fs (sda10): last error at 1301325083: htree_dirblock_to_tree:586: inode 260099: block 1056737 Strange thing is that the message pops 5 minutes right after everything settle. It is clearly visible in the attached syslog. The machine is a Fujitsu-Siemens Lifebook C1020, the OS is Debian Testing. sda10 is the last ~7GB partition on the 120GB disk. This is almost a fresh install, up since a week or such and the message was there right since with the install, with the default kernel 2.6.32.5. Tried to boot with acpi disabled, without 'lapic' (as it needs anyway to enable it) and removing most of the modules right after booting, tried in recovery mode, single mode, but no luck. The message just comes after 5 mins. But what is waiting for 5 mins then tries to tamper like from 'outside' with a mounted file system? The thing should be kernel (and/or Bios) related as it seems, maybe something with the well reputed Via chipset? (Irony) Any help would be much appreciated.
Created attachment 53022 [details] dmesg
Created attachment 53032 [details] lsmod right after booting
Created attachment 53042 [details] part of syslog with the error message
Forgot to tell that Memtest86+ 4.20 passes, however i didn't leave it on for a night to run or so. I will let it a go tonight.
ext4 now has the helpful feature of reminding you about errors it encountered in the past, and doing so on a regular basis. Isn't that a helpful message? ;) I think it gets truncated. I think it came out of: 584 if (ext4_check_dir_entry(dir, NULL, de, bh, 585 (block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb)) 586 + ((char *)de - bh->b_data))) { have you tried running e2fsck -f on the filesystem? Looks like you have a problem in a directory. -Eric
Thanks for your input Eric; yes, fsck is happy all the time (also with -fcc flags). Find -xdev -inum 260099 shows nothing either, but in the past there was possibly something there, so it makes sense regarding 'past stuff reminder'. Anyway it would be *a bit more helpful* if it would tell us something about it's *cool* behaviour, or at least some hint about some action I'd be able to pop to clean things up. De ja vu in the aspect of SMART, because n-2 of my n drives suffer from lovely errors like 'uncorrectable' (because of bad cabling what got solved of course) and 'write_dma' (right after I pulled the laptop out of the factory sealed box) and such. However this drive is one out of the ones without any error, and the long self-test ran on it shows zero faults. Memtest also passes all the time. At night. Anything I can do to 'reset' this thing, or maybe some other recommendation I should give a go?
Oh well, just got a SIGSEGV from aptitude.. Reboot and everything is fine again. Something can be incompatible with something in this laptop I feel..
This message will get reset if you use the e2fsck from e2fsprogs 1.41.13 or later. I recommend using e2fsprogs 1.41.14. You can translate the "initial error" and "last error" time stamps in the kernel messages here: kernel: EXT4-fs (sda10): error count: 5 kernel: EXT4-fs (sda10): initial error at 1301325048: htree_dirblock_to_tree:586: inode 260099: block 1056737 kernel: EXT4-fs (sda10): last error at 1301325083: htree_dirblock_to_tree:586: inode 260099: block 1056737 via the date command: % date --date=@1301325048 Mon Mar 28 11:10:48 EDT 2011 % date --date=@1301325083 Mon Mar 28 11:11:23 EDT 2011 (We use seconds since the epoch because we don't want to put time zone conversion routines in the kernel.) In any case, it's not a big deal, and you can fix this easily by upgrading to a newer version of e2fsprogs. - Ted
Thanks Ted, it was purged by e2fsprogs 1.41.14-1 from Debian experimental. The rare stuff 'just did read a file but it contains something else now' is still up, but this is another thing, however I'm sure now it was the cause of all the fs errors. I'll try and tweak with hdparm and module options (if any) to get to some more compatible controlling level. Case closed, thanx a lot!
OK, so I'll close this bug as a hardware error. If you can get something reproducible that you think points this at being a file system bug, please reopen this bug. Thanks!!