After losing my file system to corruption four times with default 6.2 installs I found the update RHBA-2000:013-01. I installed the fix and tested it by banging the hell out of my system. It seemed to work. Over the next day or so I rebooted many times and did a number of compiles, etc that made the system page. I rebooted it and left it running while I left town for a couple of days and came back to find a bunch of EXT2-fs errors on my console. Rebooting is like before. It brings you up into single user mode and when you run fsck on the fs you get millions of errors and most of your files end up in lost+found. I am running a full 6.2 install with all updates on a dual PIII system with 256mb of RAM and an IDE hard drive. One common thing I noticed in the log files is that the errors start cropping up at 4:02am as cron.daily is running. There are a couple of scripts in there that couple be pretty i/o intensive. Here is the contents of my /var/log/messages: Apr 23 04:02:01 neuromancer anacron[843]: Updated timestamp for job `cron.daily' to 2000-04-23 Apr 23 04:02:02 neuromancer syslogd 1.3-3: restart. Apr 23 04:02:02 neuromancer syslogd 1.3-3: restart. Apr 23 04:02:03 neuromancer syslogd 1.3-3: restart. Apr 23 04:02:03 neuromancer syslogd 1.3-3: restart. Apr 23 04:02:03 neuromancer syslogd 1.3-3: restart. Apr 23 04:02:29 neuromancer kernel: locknr=-871099916 sector=1621135264 size=4096 count=1 Apr 23 04:02:29 neuromancer kernel: EXT2-fs error (device ide0(3,5)): ext2_readdir: directory #356225 contains a hole at offset 443846656 Apr 23 04:02:29 neuromancer kernel: attempt to access beyond end of device Apr 23 04:02:29 neuromancer kernel: 03:05: rw=0, want=810567636, limit=10490413 Apr 23 04:02:29 neuromancer kernel: dev 03:05 blksize=4096 blocknr=-871099916 sector=1621135264 size=4096 count=1 Apr 23 04:02:29 neuromancer kernel: EXT2-fs error (device ide0(3,5)): ext2_readdir: directory #356225 contains a hole at offset 443850752 Apr 23 04:02:29 neuromancer kernel: attempt to access beyond end of device Apr 23 04:02:29 neuromancer kernel: 03:05: rw=0, want=810567636, limit=10490413 Apr 23 04:02:29 neuromancer kernel: dev 03:05 blksize=4096 blocknr=-871099916 sector=1621135264 size=4096 count=1 Apr 23 04:02:29 neuromancer kernel: EXT2-fs error (device ide0(3,5)): ext2_readdir: directory #356225 contains a hole at offset 443854848 ... ... Apr 23 04:03:11 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 405262 Apr 23 04:03:11 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 404857 Apr 23 04:03:11 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 404858 Apr 23 04:03:11 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 404867 Apr 23 04:03:15 neuromancer kernel: EXT2-fs warning (deviceide0(3,5)): ext2_free_inode: bit already cleared for inode 404801 Apr 23 04:03:15 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 404871 Apr 23 04:03:15 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 404958 Apr 23 04:03:15 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 404963 Apr 23 04:03:15 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 404973 Apr 23 04:03:15 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 405052 ... ... Apr 24 04:02:01 neuromancer anacron[6542]: Updated timestamp for job `cron.daily' to 2000-04-24 Apr 24 04:02:13 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 405583 Apr 24 04:02:13 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 405280 Apr 24 04:02:13 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 405060 Apr 24 04:02:13 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 405294 Apr 24 04:02:13 neuromancer kernel: EXT2-fs warning (device ide0(3,5)): ext2_free_inode: bit already cleared for inode 405297 I cat'd everything in /proc before shutting down the system and have the full text of these logs if needed.
What brand harddrive? How new? and what EIDE/IDE card do you have?(I have nothing to do with RedHat staff I have had similar problems on some fairly new drives) -Stan Bubrouski
Drive was a Maxtor 27gb 5400 rpm IDE (I don't have the model number handy). The controller is the onboard controller on a Tyan Tiger 100 (S1832DL) motherboard.
Have you tried 2.2.16-3 from the errata? fsck the partition to make sure it is sane, make sure you do not have interrupts enabled for that hard drive, and try the latest errata kernel.
These are more consistant with memory corruption than I/O problems in all honesty. It looks like hw though. Does the box pass memtest86
I'm having the same kond of problems : from /var/log/messages: Sep 11 04:02:19 proteus kernel: EXT2-fs warning (device ide1(22,1)): ext2_free_blocks: bit already cleared for block 4724586 Sep 11 04:02:19 proteus kernel: EXT2-fs warning (device ide1(22,1)): ext2_free_inode: bit already cleared for inode 2351256 It's running on a dual PIII 450, 256 Mb Ram, 6gb system hdd & 30 Gb data drive w/ kernel-smp-2.2.16-3 These problems killed the system a couple of times, after witch I had to fsck the mounts & repair the errors that showed up. I had to go back to a single-processor 2.2.14-15 kernel to have these errors dissapear
Closed - no answer so assuming it failed memtest86