Re: Gosh, I can't believe this!!! (was: who's guilty...)
Eloy A. Paris wrote:
>
> well, I have spent my whole Saturday chasing the answer and I think I am
> pretty close now. I think I found the cause: it's that damned Boca Research
> BOCALAN card VL. Today I tried EVERYTHING:
> [...]
> I ran for about 1 and a half years kernel 1.2.13 on this machine and with
> this PCnet32 card. This is why I suspect of kernel 2.0.x.
>
> Uhhmmm... the whole thing is strange. Why a network card problem is causing
> problems at reboot time? Why everything works so well if I turn the computer
> and work forever without reboot the computer with Cntrl-Alt-Del, reboot
> command, or shutdown -r ???
> [...]
> >
> >The options are:
> >
> > w warm boot.
> > c cold boot.
> > b reboot by jumping to the BIOS reboot function.
> > h hard reboot by toggling reset and/or crashing the CPU.
>
> I tried the "w" and it worked in all cases WITH the troubled card
> (BOCALAN-vl). The "c" never worked. Of course, with the Etherworks 3, I
> don't need to pass any reboot=x paremeter to the kernel.
I'll offer this guess: the troublesome card does DMA, and is not being shut
down before the kernel reboots. It writes to its previously programmed
buffer address during the memory test, which of course causes the memory
test to fail. Warm boot works because it does a very cursory memory test
(if any), allowing only a short time window for the card to interfere; cold
boot fails because of a more thorough test.
Perhaps the change between kernels 1.2 and 2.0 was that the driver now takes
advantage of the card's DMA capabilities? Or something isn't getting shut
down properly when switching to runlevel 0? Or the driver doesn't finish
or cancel outstanding operations when it's closed?
--
Bill Roman (roman@songdog.eskimo.com / roman@songdog.uucp) running linux
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-REQUEST@lists.debian.org . Trouble? e-mail to Bruce@Pixar.com
Reply to: