On Fri, Apr 22, 2016 at 04:58:19AM -0600, Jan Beulich wrote:
> >>> On 22.04.16 at 12:54, <konrad.w...@oracle.com> wrote:
> > The thing that makes it "safe" (in lieu of your comment about
> > re-ordering) is that the trap code does:
> > 
> > 1289     /* WARN, BUG or ASSERT: decode the filename pointer and line 
> > number. */
> > 1290     filename = bug_ptr(bug);
> > 1291     if ( !is_kernel(filename) && !is_patch(filename) )
> > 1292         goto die;
> > 
> > Which means we first try 'is_kernel'. Since the list addition cannot be
> > part of the payload it will always be part of the hypervisor.
> 
> I'm sorry, but no. The bug hit (and hence the referenced file name)
> may be in some other payload, requiring list traversal.

Ah yes! Let me dig in the rcu list to thwart the compiler from
re-ordering.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Reply via email to