Re: [Xen-ia64-devel] [PATCH] bug fix new_tlbflush_clock_period()

[Isaku Yamahata <yamahata@xxxxxxxxxxxxx>]

Hi Kouya.
Good catch!
Although this patch already commited and I made the bug,
vti domain also relies on tlb flush lock.
(See flush_vtlb_for_context_switch())

So we should do
  if (!test_bit(_VCPUF_initialize))
        continue
  if (VMX_DOMAIN(v))
        <flush all hash and collision chain of v>
  else
        vcpu_vhpt_flush()

Or

  disable the tlb flush clock usage in flush_vtlb_for_context_switch().

On Fri, Feb 02, 2007 at 03:58:12PM +0900, Kouya SHIMURA wrote:
Content-Description: message body text
> Hi,
> 
> __vhpt_flush() might be called from new_tlbflush_clock_period()
> even if the vcpu doesn't have the per-vcpu-vhpt. 
> It erases 64KB memory content that starts from machine address 0.
> 
> Thanks,
> Kouya
> 
> Signed-off-by: Kouya Shimura <kouya@xxxxxxxxxxxxxx>
> 

> diff -r 0df9dc2f1d03 xen/arch/ia64/xen/flushtlb.c
> --- a/xen/arch/ia64/xen/flushtlb.c    Thu Feb 01 13:54:26 2007 -0700
> +++ b/xen/arch/ia64/xen/flushtlb.c    Fri Feb 02 14:29:45 2007 +0900
> @@ -88,6 +88,9 @@ new_tlbflush_clock_period(void)
>      smp_mb();
>      for_each_domain(d) {
>          for_each_vcpu(d, v) {
> +            if (!test_bit(_VCPUF_initialised, &v->vcpu_flags) ||
> +                VMX_DOMAIN(v))
> +                continue;
>              if (HAS_PERVCPU_VHPT(v->domain))
>                  vcpu_vhpt_flush(v);
>          }

> _______________________________________________
> Xen-ia64-devel mailing list
> Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-ia64-devel

-- 
yamahata

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel