[2.6.27] filesystem (ext3) corruption (access beyond end)
Ok I ran several disk-intensive jobs (git clone, kernel build, apt-get dist-upgrade) but none of those could trigger any bugs.
BUT, running my http://vanheusden.com/pyk/ script, I definately can trigger the problem:
[85202.195563] rtc0: alarms up to one month, y3k, hpet irqs [85204.035802] journal_bmap: journal block not found at offset 2060 on dm-0 [85204.035818] Aborting journal on device dm-0. [85311.242093] ext3_abort called. [85311.242120] EXT3-fs error (device dm-0): ext3_journal_start_sb: Detected aborted journal [85311.242154] Remounting filesystem read-only [86331.285847] EXT3-fs error (device dm-0): htree_dirblock_to_tree: bad entry in directory #24568: rec_len % 4 != 0 - offset=0, inode=4098364138, rec_len=59542, name_len=76 [89931.353863] EXT3-fs error (device dm-0): htree_dirblock_to_tree: bad entry in directory #24568: rec_len % 4 != 0 - offset=0, inode=4098364138, rec_len=59542, name_len=76
[92499.263276] attempt to access beyond end of device [92499.263296] dm-2: rw=17, want=4177066240, limit=6004736 [92499.263311] Buffer I/O error on device dm-2, logical block 522133279 [92499.263333] lost page write due to I/O error on dm-2 [92499.263341] Aborting journal on device dm-2. [92499.263504] ext3_abort called. [92499.263515] EXT3-fs error (device dm-2): ext3_journal_start_sb: Detected aborted journal [92499.263543] Remounting filesystem read-only
[93531.419902] EXT3-fs error (device dm-0): htree_dirblock_to_tree: bad entry in directory #24568: rec_len % 4 != 0 - offset=0, inode=4098364138, rec_len=59542, name_len=76
Very odd since the script only rmmod/insmods modules and trying to rmmod for example ext3-module shouldn';t be a problem since it is in use.
Quoted 19 lines · Randolf Pohl
I think, stable and gregkh should be CCed on this, and yes, I guess this deserves some investigation.
Regards,
Randolf
http://lkml.org/lkml/2008/12/4/246
Quoted 8 lines · Folkert van Heusden
Hi,
With 2.6.26 and 2.6.27 a system of mine (a P4 with HT and IDE disk) gets massive disk-corruption: in dmesg warnings for access beyond drive end.
Now this bug did not happen in 2.6.18 (debian kernel) and it seems to be gone in 2.6.28-rc* as well.
My question now is: should it still be investigated?
Folkert van Heusden
Signature · 6 lines
Multitail est un outil permettant la visualisation de fichiers de journalisation et/ou le suivi de l'exécution de commandes. Filtrage, mise en couleur de mot-clé, fusions, visualisation de différences (diff-view), etc. http://www.vanheusden.com/multitail/ ---------------------------------------------------------------------- Phone: +31-6-41278122, PGP-key: 1F28D8AE, www.vanheusden.com