I wanted to create a backup of my VS 4k90 root disk, and I ran into a
whole lot of SCSI weirdness like I've never seen before. I got three IBM
DNES-318350 disks out of storage in th hope that at least one would
work, but that seems not to be the case.
The issues start with SCSI timeouts before the disks are even attached:
[ 3.539508] (asc0:2:0): selection failed; 11 left in FIFO [intr 18, stat 93, step 3]
[ 13.539441] probe(asc0:0:2:0): asc0: timed out [ecb 0x87ec1fa8 (flags 0x1, dleft 10, stat 0)], <state 2, nexus 0x87ec1fa8, phase(l 13, c 3, p 3), resid 0, msg(q 20,o 80) >
[ 15.549440] probe(asc0:0:2:0): asc0: timed out [ecb 0x87ec1fa8 (flags 0x41, dleft 10, stat 0)], <state 2, nexus 0x87ec1fa8, phase(l 13, c 3, p 3), resid 0, msg(q 20,o 80) > AGAIN
[ 15.579450] (asc0:2:0): selection failed; 11 left in FIFO [intr 18, stat 93, step 3]
[ 16.579446] probe(asc0:0:2:0): asc0: timed out [ecb 0x87ec1fa8 (flags 0x1, dleft 4000, stat 0)], <state 2, nexus 0x87ec1fa8, phase(l 13, c 3, p 3), resid 0, msg(q 20,o 80) >
[ 18.589440] probe(asc0:0:2:0): asc0: timed out [ecb 0x87ec1fa8 (flags 0x41, dleft 4000, stat 0)], <state 2, nexus 0x87ec1fa8, phase(l 13, c 3, p 3), resid 0, msg(q 20,o 80) > AGAIN
[ 18.605122] probe(asc0:0:2:0): async, 8-bit transfers
But eventually, the disk attaches:
[ 18.612879] sd1 at scsibus0 target 2 lun 0: <IBM, DNES-318350, SA30> disk fixed
[ 19.499438] sd1: 17501 MB, 11474 cyl, 10 head, 312 sec, 512 bytes/sect x 35843670 sectors
[ 19.519502] sd1: sync (160.00ns offset 15), 8-bit (6.250MB/s) transfers, tagged queueing
From then on, I can label it, newfs it, and transfer the contents of my
old disks by running "dump | restore". Here the oddity continues, after
transferring /usr (approx. 10GB), the system still works, but if left
sitting like this for a few hours it'll grind to a halt. It can still be
pinged, but it otherwise it appears dead. Continuing transferring data
from other filesystems sees transfer rates dropping continuously until
it all grinds to a halt, too.
I restart the system, and everything seems to work. I can compare the
transferred data and see everything that copied was copied correctly.
Then I install boot blocks:
# installboot /dev/rsd1c /usr/mdec/sdboot
but booting the system from the new disk, I get only this:
?06 HLT INST
PC= 00002AD6 PSL= 041F0000
I'm not sure this is related to the SCSI issues I'm seeing. Has anyone
tried doing a fresh install NetBSD/vax on a SCSI VAX recently? Is it
possible our bootblocks suffer from bitrot?
I've seen the same and some more weirdness in my 2nd 4k90 a few months
ago, but I figured it might be an issue with the system board. Now I
think it's more likely an issue with the IBM disks, but then it's 3 of
them behaving oddly in exactly the same ways. Note that I never noticed
any SCSI related problems with the disk that's been in this box for the
past two decades (a Seagate ST318418N).
Does anyone else have a VS 4k90 (or a different model VS 4k), seeing
similar issues?
Thanks,
Hans