On 21/08/2009 13:03, "Stephen Smalley" <sds@xxxxxxxxxxxxx> wrote:
> On Thu, 2009年08月20日 at 17:02 +0100, Keir Fraser wrote:
>> Looks like a lock normally taken with irqs disabled may be being taken with
>> irqs enabled. Which wouldn't be allowed but could be easy to fix.
>
> This appears to have been introduced by CS 20073, x86: Implement
> per-cpu vector for xen hypervisor. It introduces mixed usage of
> spin_lock() and spin_lock_irqsave() on the vector_lock, whereas
> previously it was always spin_lock_irqsave().
Xiantao, Can you make a patch for this? Any acquisition of the lock where
irqs may not already be disabled must do spin_lock_irq or spin_lock_irqsave.
 -- Keir
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel