Subject: Re: panic in pmap_enter w/ 1.2.1 kernel
To: None <netbsd-users@NetBSD.ORG>
From: Laine Stump <laine@MorningStar.Com>
List: netbsd-users
Date: 05/05/1997 14:07:36
Jason Thorpe writes:
> On Mon, 5 May 1997 10:43:43 -0400 (EDT)
> Laine Stump <laine@MorningStar.Com> wrote:
>
> > I just built 4 new machines all with Intel Providence motherboards and
> > 128MB, and 2 of them with triple striped Seagate Cheetah U/W disks. They
> > seem to be running fine, except that one of the machines with disks is
> > panicking about once a day in pmap_enter in the following place:
> >
> > /*
> > * Page Directory table entry not valid, we need a new PT page
> > */
> > pte = pmap_pte(pmap, va);
> > if (!pte)
> > panic("ptdi %x", pmap->pm_pdir[PTDPTDI]);
>
> ...this was fixed some time ago in NetBSD-current.
(I really *hate* asking this, but....) Since we can't risk putting a
random -current snapshot on these machines, are there any discussions
about when a new release will be made? How about a "stable snapshot" or
something? (Are there known showstoppers in 1.2D?)
> As a stopgap, you
> can change the NKPDE constant in i386/include/pmap.h ... for 128M, a
> value of 24 is appropriate. (24 is the highest you can effectively use.)
I've had that set at 31 for nearly a year, and everything has been
working fine until now. Is the required number of these affected by the
size of swap? I ask because one change on these new machines was that,
due to the rough arithmetic I used for allocating swap partitions on the
3 disks, swap space is now at about 600MB, where it was previously about
500MB.
Alternately, could the fact that I have a number higher than 24 there
have an adverse effect?
> > Also, how do I feed the kernel dump that was collected into gdb? It
> > isn't recognized as a coredump file.
>
> .... gdb -k netbsd.gdb netbsd.core
Thanks! Not much useful info this time ("where" showed an empty stack),
but next time I'll be prepared.