Subject: Re: Annoying panics...
To: Paul Goyette <firstname.lastname@example.org>
From: Allen Briggs <email@example.com>
Date: 08/27/1997 20:53:55
> > Does anyone know why the following panic would occur?
> > pmap_alloc_pv: pgi_nfree inconsistent
> I used to get that same panic occassionally during reboot, in the part
> where it ran fsck.
I saw this a few times while working on the esp driver. As far as I can
tell, it looks like a bug in the pmap module on the mac68k port (possibly
on other m68k ports, too). I also was having trouble with an occasional
MMU fault panic from pmap_is_modified (called from vm_object_page_clean).
I wasn't able to pin either panic down to a repeatable case, but I think
the second was just a result of the mac's special way of signalling a
phase change on the scsi bus(*).
(*) If you care... These machines have no DMA, but they have a dhip
that handshakes 16-bit words like a DMA controller would. If the
53c96 doesn't request a data transfer within a certain amount of
time, the chip generates a bus error. To handle this, the md part
of the driver sets a flag does a setjmp and the kernel bus error
handler does a longjmp if the flag is set. At the time I saw those
panics, I wasn't screening serial interrupts, so I think I got a
serial interrupt and while I was processing that, I got bus error
and tried to return to the scsi transfer. I think that hosed things
well enough that it crashed soon after. At least, that's my guess
as to what happened--I haven't seen either panic lately.
Allen Briggs - end killing - firstname.lastname@example.org