On Sun, Nov 18, 2007 at 09:39:57PM +0000, Andrew Doran wrote:

> On Sun, Nov 18, 2007 at 09:19:15PM +0100, Manuel Bouyer wrote:
> 
> > I've tested a kernel built from HEAD (which I didn't do for some time) and
> > seeing panics when starting xend (this dom0 has only 64Mb RAM allocated).
> > It's always a uvm_fault() on a kernel address, but the fault instruction
> > varies. Here's a sample of a panic:
> 
> I think it's because the int/trap stubs aren't clearing the direction flag.
> i386/amd64 have the changes but (as usual) I kept forgetting to update xen.
> I'll fix it shortly.

Acutally, it already has all the necessary changes. Is there a point where
the kernel is entered with Xen that the direction flag might not be cleared?
I can't see one. It is cleared for the copyout() in the traceback that you
posted.

> I've also seen it in the pool code. It was always handling a trap after a
> copyin or copyout though.

I did many hours of low-memory stress testing on the updated pool code
before checking it in, so I don't believe that there is an (obvious) problem
there. It could be perhaps be related to the removal of the _CPU options.

Andrew