Subject: Re: Reason for pagefault trap
To: None <port-i386@NetBSD.ORG>
From: Kent Polk <kent@tiamat.goathill.org>
List: port-i386
Date: 06/07/2000 04:34:13
On 6 Jun 2000 20:20:01 -0500, David Maxwell wrote:
>On Tue, Jun 06, 2000 at 07:11:33PM -0400, Bill Sommerfeld wrote:
>> > It would have to know how to initialize enough of the hardware to
>> > gets the test done. Certainly it wouldn't need drive support, but
>> > it would need keyboard, display, and some subset of bus support.
>> 
>> It could use the same standalone library which biosboot uses.
>> 
>> > (And people with USB keyboards... tough luck?)
>> 
>> biosboot works with USB keyboards using the bios keyboard driver,
>> right?
>
>I don't know the answer - but I did find the answer to something else -
>under Linux, Memtest86 can be run from LILO. So I'd imagine it includes
>everything to make basic keyboard/display combos usable. It may not
>handle USB of course.
>
>I took a (brief) shot at building it, but it's not happy with the NetBSD's
>gnu as. Has anyone listening gotten it to compile on NetBSD before?

Though there is no executable in the linux version, the windows
version has one. I just dd'ed the file to floppy, booted from floppy
and it worked for me. The reason for this exercise is that I am
having similar problems to the original poster:

> uvm_fault (0xc4a53d10, 0x1000,0,1) -> 1
> kernel: pagefault trap, code=0
> stopped in cc1 at _pmap_page_remove + 0xac: movl 0x4(%ebx),%ecx

The box is an old 233MHz Pentium with 40M of SIMMs that I am trying
to set up as a firewall. The box is fairly stable with minimal
processes loaded. It appears to be pagefaulting primarily when
processes are being spawned. As more processes are loaded, the
pagefaults dramatically increase. They usually occur when I attempt
to login and if I run ipf, the box pagefaults pretty much every
hour:

uvm_fault (0xc36beb00, 0x1000,0,1) -> 1
kernel: pagefault trap, code=0
stopped in sh at _uvm_fault + 0x12b: testl %edx, \
  0xffffffff(%edi, %edi, 8)

Since it gets worse with more processes, I thought possibly one of
the SIMM sticks in higher memory was bad, but I just ran memtest86
for about 6 hours and not a single failure showed up.

Any suggestions?
Thanks