2.0.33: SB16+VIA-VP2/97+WD1.6Gig + hdparm -u 1 --> disk coruption?

Mark Gray (markgray@iago.nac.net)
26 Mar 1998 07:52:24 -0500


The SB16 is a real Creative Labs SB16 (early pnp that works without
isapnp and has an high DMA channel for 16bit), the VIA VP2/97 is on a
FIC PA-2011 ATX Motherboard (200MHz AMD-K6) and the Western Digital
1.6Gig hardisk is a very early one. For quite some time I have been
getting the following error in my logs:

kernel: Sound: Fatal error. Audio interrupt after freeing buffers.

This is while using playmidi constantly and the past two weeks of logs
(up 24/7) gives 87 messages. Yesterday after using glint to upgrade
RedHat-5.0 to glibc-2.0.7-4 I started getting the following disk errors:

Mar 25 11:38:58 k6 kernel: hdb: read_intr: status=0x59 { DriveReady SeekComplete DataRequest Error }
Mar 25 11:38:58 k6 kernel: hdb: read_intr: error=0x40 { UncorrectableError }, LBAsect=2466509, sector=672268
Mar 25 11:38:58 k6 kernel: end_request: I/O error, dev 03:44, sector 672268
Mar 25 11:38:58 k6 kernel: EXT2-fs error (device 03:44): ext2_readdir: directory #83314 contains a hole at offset 0

Along with errors for a bunch of other sector numbers, but all
completely contained within the /usr/share/zoneinfo/ directory
installed by glibc (in spite of considerable other disk usage before
noticing the error messages.) (I nfs export /usr ro to my 386sx16 in
case that matters.) The only way I could manage to salvage the system
was to reboot to a rescue patition on another disk,

rm -rf /mnt/hdb4/usr/share/zoneinfo/

then tar the rest of the partition to a spare partition, and then

dd if=/dev/zero of=/dev/hdb4 bs=1024

to recover the sectors destroyed (hard hardware errors otherwise),
then mke2fs the partition, reinstall from the tar, replace
/usr/share/zoneinfo/ from another computer, remove the hdparm -u1
from my startup scripts (I used to use:)

/sbin/hdparm -c 1 -X 12 -A 1 -m 8 -u 1 /dev/hdb

Now:

/sbin/hdparm -c 1 -X 12 -A 1 -m 8 /dev/hdb

And everthing has been running perfectly ever since. I would let this
pass without notice as either resulting from the overconfidence
(stupidity) of upgrading glibc while running X, or an aging early
model Western Digital harddisk _except_ that since rebooting without
the "-u 1" 16 hours ago I have yet to get any of the:

kernel: Sound: Fatal error. Audio interrupt after freeing buffers.

errors in my logs. So anyway --- anyone want more info?

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu