Messages in this thread |  | | Date | Tue, 31 Dec 1996 04:37:31 -0500 | From | "Theodore Y. Ts'o" <> | Subject | Re: Serious ext2fs problem |
| |
From: Chris Adams <cadams@ro.com> Date: Mon, 30 Dec 1996 14:04:26 -0600 (CST)
I think there is a bug in ext2fs. I have a news server that has a DPT 3224 RAID adapter with a 10G RAID 0 array attached. I am running RedHat 4.0, Linux 2.0.27, and INN 1.4unoff4. I also have applied the "noatime" patch to the kernel, and mount the news spool with the noatime option. >From time to time, INN will think the disk is full because it gets an error trying to write. The kernel log has lines like:
EXT2-fs error (device 08:31): ext2_new_block: Free blocks count corrupted for block group 546
in it when this happens (device 8:31 is /dev/sdd1, which is the RAID disk). I have shutdown INN, unmounted the disk, and run fsck on it, but that never finds any errors. I also re-made the filesystem on it once, and that seemed to help for a while, but now I am getting errors again, and fsck won't fix them.
Please read the e2fsck man page about "reporting bugs", where it describes how to report a coherent bug report so that I have a chance of trying to help you. (Silly things like exactly which version of e2fsck you are using, etc....)
Also, you *did* run fsck using the -f flag, to force checking the filesystem, *right*? :-)
Fsck most certainly does check to make sure the free block count is correct. So either there's some sort of kernel bug, or hardware bug, assuming that e2fsck really had a chance to do its thing. (If the clean bit was set, even though the filesystem had errors, using the -f flag is important since it forces e2fsck to check the filesystem even if the clean bit is set.)
- Ted
|  |