On Tue, 11 Jan 2000, Sven Dietrich wrote:

> Ok, here is the updated info on the XP1000 SCSI problems.
> 
> When I tried gunzip'ing a file, this occurred:
> 
> Jan 6 23:26:07 host /netbsd: stray 6600 irq 47
> Jan 6 23:26:07 host /netbsd: stray 6600 irq 47
> This message repeated three times
> Jan 6 23:26:08 host /netbsd: stray 6600 irq 47; stopped logging
> isp0: mbox cmd 15 failed with error 0
...

  Hmmm - from my experience with stray interrupts with a 3COM 3C905
board, when the kernel reports "stopped logging", it actually has disabled
that interrupt.  That might be why all the following errors occurred.

...
> isp0: command for fast posting handle 0x59 not found
> <cycles through address 0x68>
> isp0: Ultra Mode Capable
> ...
> isp0: command for fast posting handle 0x6 not found
> <cycles through address 0xa>
> 
> 
> Eventually gets back to the "stray 6600 irq 47" and repeats

  If the error recover process re-enables that interrupt, this might
explain why it repeats.  The elinkxl driver did not re-enable the
interrupts, so that board was dead once it reached the "stopped logging"
state.

  When I was trying to figure out what my 3COM board was doing, I
increased the limit on stray interrupts to get a better idea of how
often the stray interrupts were occuring.  If the stray interrupt is
caused by some condition not handled by the driver, they would keep
occuring until the interrupt was disabled (which was the case with the
elinkxl driver).  Since my 3COM board was the only device using that
interrupt, I modified the driver to display status information when if the
driver didn't think the device was interrupting.

--
Michael L. Hitch			mhitch@montana.edu
Computer Consultant
Information Technology Center
Montana State University	Bozeman, MT	USA