Loss of SCSI tape reliability with 2.4 kernel

[Bill Freeman]

	Does anyone know of issues with SCSI and/or st with 2.4
kernels, and, hopefully, what to do about it?

	I've got a box set up to back up the machines on our network
to DAT using Amanda.  It's a RH6.2 system (from an GNHLUG labeled CD).

	But this box is also my workstation, and I have RH7.1
installed in other partitions for normal daily life.  Rather than boot
into 6.2 to do backups, I have configured Amanda on the 7.1
partitions, and after fixing security problems (on 6.2 the server
identifies to the client on the same box as localhost... where in the
7.1 world it uses the configured machine and domain, so .amandahosts
had to be different), Amanda shows signs of life.

	In fact, if I mess up and put the wrong tape in the drive,
Amanda happily stuffs backups on my tape buffer partition (so called
degraded mode).  But when it runs the tape drive, whether as part of
amdump of as part of a subsequent amflush (copies the degraded mode
disk copies of the dump to the tape), in a reasonably short time, the
machine hangs totally (reset required to recover).

	I was thinking that maybe my SCSI controller was failing and
wedging the bus during DMA, or some such, but things work fine if I
boot back into the 6.2 system (2.2 kernel), including doing the amflush
for the degraded mode dumps made under 7.1 (2.4 kernel).

	The problem occurs with both the 2.4.9-34 and 2.4.9-12 i686
kernels from RH (the two that I've tried).  The -34 is the latest
update.

	Any thoughts?

							Bill