Lots of data faults from custom kernel?

[Elliott McCrory]

I have an old 4-260 with a locally-installed token ring and driver.  I
hope that the behaviour we are observing is not the fault of our custom
driver, but it probably is.  The panics never trace back through our
custom driver, and all of the parameters in our custom driver seem to be
in order after the crash.

We are getting "panics" rather regularly (every 1 to 2 days) which usually
all have the following sort of traceback:

% adb -k vmunix.12 vmcore.12
physmem ffd
$c
_panic(0xf813900b,0xf82eeb3c,0xf8138eb0,0x80,0x7c178,0xf8306e98) + 6c
_trap(0x9,0xf82eeb3c,0xf8300fb4,0x80,0x1,0x50) + 1cc
fault(?)
_flush_windows(0x2e,0xf82eeea4,0xf82eeeb8,0x1000,0x1fe3,0xf82eeea4) + 60
_resume(0xf81922ac,0x18,0xf8129960,0xf8129978,0xf82ef000,0xf8129978) + ec
sw_bad(?)
_swtch(0xf81922ac,0xf8307000,0x1,0x1,0xf7fff110,0x7a34) + 80
_sys_rttchk(0xf81922ac,0xf8306f58,0xf8306f58,0xe05,0xf8307000,0xf8304000) + 38

The key is that they all go through "flush_windows."  And the initial
"panic" message thinks that, usually, sendmail is running, although
sometimes I have seen "xterm" and "dump."  I cannot say with 100%
certainty that we are always running X11R4 at the time--in fact, sometimes
nothing is running on the console.

Any help would be appreciated.

  ||    Elliott McCrory, Ph. D.		| mccrory at tomato.fnal.gov (internet)
-/  \-  Fermi National Accelerator Lab	| ALMOND::MCCRORY	 (DECnet)
-\  /-  Batavia, IL  60510		| MCCRORY at FNALAD	 (BITnet)
  ||    708-840-4808			| tomato!linac!mccrory	 (uucp)