Re: scsi problems, ncr53x8xx and 2+ devices resets bus
On Fri, 19 Dec 1997, Gerard Roudier wrote:
> Here are 2 driver messages from your report and what they let me think:
>
> > Dec 19 12:24:24 siva kernel: ncr53c810-0:0: ERROR (a0:0) (47-67-80) (8/33) @ script (628:0f000000).
> > Dec 19 12:24:24 siva kernel: ncr53c810-0: script cmd = 60000040
>
> DSTAT = 0xa0 -> bit 0x20 means PCI BUS fault condition detected.
> DCMD/DBC = 0x0f000000 CLEAR ACK (current SCRIPTS instruction)
> DSP points to 0x60000040 MOVE WHEN MSG_IN (next SCRIPTS instruction)
>
> The weird thing is that I did'nt find such a sequence of SCSI
> SCRIPTS instructions in the source. Such a sequence exists but with
> instructions in reverse order. The offset in the script (628) is near
> the SCSI reselection step, probably near the read of the IDENTIFY
> message.
>
> Even if your problem seems to be triggered by SCSI device
> configuration and/or load, the above lets me think that it is a
> real PCI bus problem. You are using a very old 810 chip (rev. 1)
> on a recent system. I donnot have access to any errata of so old
> chips, but my guess is that they probably suffer of bunches. Using
> such a old 810 chip with a recent host bridge is IMO not a very safe
> configuration.
thanks for your reply, it explains what i found out after i wrote my
message.
first, i tried another scsi cable - no difference.
second, i installed the SONY tape drive into a 486 box with an ASUS SP3G
motherboard (built-in NCR 53c810 chip). it worked perfectly, backing up
from a Seagate 4GB Hawk drive to the SONY DAT. Took 93 minutes to fill a
90M (2.1GB) tape, with a 'find . -type f | xargs file' running in the
background as a simple stress-test.
I'm reasonably happy with this arrangement because the tape drive was
intended to be a network backup device anyway....it doesn't really matter
which machine it's connected to as long as it works.
I'll still have to get a new scsi card for my Gigabyte HX machine (can't
trust my disks to a scsi adaptor which gets flaky under load, especially
when i'm planning to add at least 4GB of fast scsi disk, get rid of the
IDE drive and make it scsi-only). I had decided to buy an advansys or
buslogic card, but from what you say i should be able to just get a
newer 53c810 card. Good. the 810 cards only cost around $80 here in
Australia, but the others cost at least $300 or $400.
cat /proc/pci on the ASUS machine (axolotl) looks like:
$ cat /proc/pci
PCI devices found:
Bus 0, device 6, function 0:
VGA compatible controller: S3 Inc. Trio32/Trio64 (rev 0).
Medium devsel.
Non-prefetchable 32 bit memory at 0xfb000000.
Bus 0, device 2, function 0:
Non-VGA device: Intel 82378IB (rev 132).
Medium devsel. Master Capable. No bursts.
Bus 0, device 1, function 0:
Non-VGA device: NCR 53c810 (rev 1).
Medium devsel. IRQ 9. Master Capable. Latency=80.
I/O at 0xe800.
Non-prefetchable 32 bit memory at 0xfbfef000.
Bus 0, device 0, function 0:
Non-VGA device: Intel 82424ZX Saturn (rev 4).
Medium devsel. Master Capable. Latency=96.
so the 810 chip on axolotl is also a Rev 1 - not suprising, it was one
of the first boards to come out with built-in 53c810 scsi.
> Telling so, I did agree with you, but I disagree on your statement
> about such a board being 'useless crap'. I have a 3 years old system
> with such a chip (810 rev. 1) that works flawlessly since the day I've
> received it (It is PCI 2.0).
>
> All your recent hardware will be stated as out of age probably in less
> than 3 years, and I am about sure it suffers of bunches of bugs and so
> is probably as crap as the 810 A rev. 1. It is 'usefull crap' today
> and will become 'useless crap' to-morrow.
:-)
> I would be very interested in the result with a recent 810 chip
> version (>=0x12). (In case of you could borrow one).
well, i'll probably be buying one in the near future (maybe not until
january). I'll let you know how it goes.
I'll put my old 810 card into one of my PCI 486 boxes...load it up with my
old archive viper 150MB QIC tape and ancient panasonic scsi cd-rom.
craig
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: