Subject: Re: panic: pmap_zero_page: lock botch
To: Bill Sommerfeld <email@example.com>
From: Chuck Cranor <firstname.lastname@example.org>
Date: 03/26/1999 16:42:05
On Fri, Mar 26, 1999 at 04:17:37PM -0500, Bill Sommerfeld wrote:
> > email@example.com writes:
> > >UH.....
> > >pmap_zero_page() and pmap_copy_page() should be going to splimp()!!
> > is this a new requirement? who is calling these from an interrupt?
> kernel memory allocators?
i believe that most of those allocators call memset() directly.
however, it might be possible that if an interrupt routine
touches a zero-fill area of memory [triggering a page fault?]
that has no page associated with it then pmap_zero_page could
get called through uvm_km_get? [if so, that would be bad since
curproc may very well be null and sometimes the fault routine
likes to sleep]