Spontaneous reboot with 3c59x kernel module
I have had this problem on two different computers, at least 3 different
3com 59x network cards, and both the slink default kernel-image and my
own compiled 2.2.12 (using kernel-package). In one case I was unable to
install debian (slink) at all until I changed network cards to an ne2k
compatible because the reboot happened before the installation finished.
In the second case I am on cutting-edge potato.
The symptom is that between 1 and 15 minutes after loading the 3c59x
module (either by modprobe, ifconfig, or modconf), the machine reboots
with no warning whatsoever. I have not timed this problem (in all cases
the priority was getting the machine working so I didn't do any real
debugging) so it may be the exact same time delay in each case or it may
vary - but it always happens within a few minutes. In the time between
loading the module and the reboot, the card seems to work - one time I
successfully copied several megabytes of data across the link.
None of the explanations for this problem make sense. It can't be bad
hardware - in all cases win9x or NT were able to use the cards with no
problems. Nobody else in my company uses Debian, but many use Red Hat on
similar cards with no problems. The number of different computers and
situations involved seems to preclude it being some other piece of
hardware or software in my system. But I can't believe that the kernel
developers have missed a spontaneous reboot bug for 2 whole stable
versions of the kernel! That leaves (1) User error by me... but all I
did was select the item from the list in modconf! There are no mandatory
options, but the same thing happened if I tried supplying the optional
ones. Or (2) a debian-specific bug to do with the way kernel-package
compiles things. Could this be a gcc2.7/egcs/gcc2.95 issue? Does
kernel-package use gcc2.95 where available? What is the slink default
kernel-image compiled by?
Does anyone have any similar stories to report, fixes, workarounds, or
suggestions as to where to go to get the bug fixed, if it is a bug? (I'd
say that even if it's user error it's a bug... no bad input should cause
a spontaneous reboot, just an error message).
Thanks in advance,
Stuart.
Reply to: