Re: current-amd64 page fault trap

[Andrew Doran <ad%netbsd.org@localhost>]

On Fri, May 02, 2008 at 07:01:42PM +0200, Lars Nordlund wrote:

> (Perhaps related to the "overnight crash on current-amd64" posted 
> yesterday?)
> 
> The last couple of weeks I have had problems with my NetBSD-current-amd64.
> The computer has been un-wakeable in the morning. Often related to high 
> load during the night. CVS updates, pkgsrc-builds or things like that. 
> Since I dont have a serial console, use a USB keyboard and always leave 
> the computer in X to get the DPMS sleep thingy going.. It is hard to 
> reach the debugger.
> 
> Anyway, now I know how to reproduce the crash. Or at least produce a 
> crash.. While writing a DVDRW and doing heavy disk I/O, the trap may
> happen. Here is one I just caught:
> 
> (typed from screen..)
> kernel: page fault trap, code=0
> Stopped in pid 0.51 (system) at netbsd:pool_put+0x223:   movq  %r13,
> 0x8(%rax) db{1}> bt
> pool_put() at netbsd:pool_put+0x233
> pool_cache_invalidate_groups() at
> netbsd:pool_cache_invalidate_groups+0x46 pool_cache_invalidate() at
> netbsd:pool_cache_invalidate+0x95 pool_reclaim() at
> netbsd:pool_reclaim+0x5b pool_drain_end() at netbsd:pool_drain_end+0x28
> uvm_pageout() at netbsd:uvm_pageout+0x570

Could you try running memtest86+ or similar on this machine, for a few
hours? That's an unusual place for it to crash.

Andrew