> On Sun, Jan 21, 2007 at 08:37:51PM +1030, Brett Lymn wrote:
> > 
> > OK, I have changed things around in uvm_io_done() so that the veriexec
> > check is done without locks held but before the pages are unlocked.
> > This seems to work ok for me with a kernel that has both LOCK_DEBUG
> > and DIAGNOSTIC enabled on a 2 processor machine (well, a 2 processor
> > qemu machine).  Diff is attached (again, sorry for the horrible diff -
> > I cannot seem to get cvs to give a better/clearer one)
> > 
> 
> Folks, I have not had any responses to this change - are there any
> objections if I commit it?
> 
> -- 
> Brett Lymn

i tend to object, because it seems too complicate to me.
sorry, i don't have enough time to investigate it further right now.

YAMAMOTO Takashi