Controller trashes FS (what's new)

Richard B. Johnson (root@analogic.com)
Sun, 26 Jan 1997 14:29:10 -0500 (EST)


I have been having trouble with SCSI Controllers.
I got this Pentium System with an AHA-2940 Controller. The aic7xxx driver
is apparently not being maintained so no bugs were getting fixed. These
bugs were not killers, just little things like:

o System crashes when you execute ls on a MS-DOS file system if
the partition size is greater than 500 megabytes. These things
were reported by myself and others during the past year.

o System crashes when I attempt to use an EXABYTE tape drive. This
was explained as a fault of the controller to properly handle
disconnect (EXABYTE doesn't run sync and doesn't disconnect).

Anyway..............

So therefore, I spent a lot of money and bought a Bus Logic BT-958
SCSI Controller. This had been recommended because, amongst other things
"It had the best written Linux driver"....

Last night it trashed 3 file-systems, two were not even mounted. The
Linux root partition was not recognizable as a file-system of any known
type. This is an ext2 file-system. My "/home/user" partition was not
repairable using e2fsck although I was able to copy scrambled files off
to a network file-server. Most are 'C' sources and can be repaired.

Two file-systems that were not even mounted was my MS-DOS boot partition
on a completely different drive, and a unnamed 32-bit operating system
that I designed a few years ago. These things should not ever have been
touched! They were completely "unknown" to Linux, they were on different
physical devices than anything Linux uses.

I have not been able to do backups since my tape drive didn't work with
the ADAPTEC controller. Now that I had a new controller installed, and
now that I was able to read and write to the Tape Drive, I confidently
told crond to make a backup...

#
# Backup procedure
#
LOG="/root/logs/backup.log"
TAPE="/dev/st0"
cd /
/bin/date>${LOG}
/bin/tar -cvlf ${TAPE} . >>${LOG}
#/bin/mt offline /dev/st3
/bin/date>>/${LOG}
exit 0

This kicked off last night at 2 AM. I came in this morning to retrieve
my first backup tape that I would have ever been able to make on this
machine. Instead of a backup, the system was trashed.

I have been backing up Linux on another machine. It is only a 486
and it uses a slow-speed ADAPTEC controller.

The only way I can backup this machine is to tar everything off to the
Network. This is why it has now taken 11 hours for me to get this machine
back up. I will never be able to restore anything that I had on the
32-bit operating system partition because it did not have a file-system
that was recognized by Linux. Therefore I was never able to back it up.
To recover this, will require about 20 hours of feeding in floppies.
This is not fun.