(cut to tech-kern)

On Thu, Mar 13, 2003 at 05:58:14PM -0800, Brian Buhrow wrote:
> 	Any suggestions as to which commands I should try, or would you like
> the core and kernel files directly?

I'm not an expert on the uvm code (but have done my share of kernel
debugging).

Looking at the genfs_putpages, and considering the fact that there
seem to be problems on completely unrelated pmap systems, my best
guess is that the uvm_object page lists were either corrupted
at the time that genfs_putpages was entered, or that the page it
has it its hand got deleted by another context (would have to be an ISR).

Checking the hash lists is not for mortals to do by hamd!

'pg' itself has to be valid (or you panic in genfs_putpages), but...

Is setting 'endmp.offset = (voff_t)-1;' enough to stop 'endmp' being
used by another thread?


	David

-- 
David Laight: david@l8s.co.uk