Re: [PATCH] [RFC] Fix a small window on CPU online/offline

[Keir Fraser <keir.fraser@eu.citrix.com>]

On 02/04/2010 04:31, "Jiang, Yunhong" <yunhong.jiang@intel.com> wrote:

> Have a short discussion with Kevin, maybe we can sync the state in
> cpu_disable_scheduler if current is idle, and then set a flag so that we will
> not sync again in context siwtch later. If the current is not idle, we can
> leave the context switch to do the sync for us. I will do more investigate to
> see how many changes are needed.

Hm, actually maybe that could work. You might not even need a flag in case
current is non-idle in cpu_disable_scheduler(). It might suffice to force
context_switch() to do full context switch synchronously when the CPU is
going offline (see appended patch). I was thinking there was a race as soon
as the cpu is cleared from cpu_online_map, but actually the problem occurs
only once the vcpu is descheduled, so if we can synchronously sync its state
before calling context_saved(), perhaps we are fine.

This could be a very small patch after all! :-)

 -- Keir

--- a/xen/arch/x86/domain.c    Thu Apr 01 09:55:27 2010 +0100
+++ b/xen/arch/x86/domain.c    Fri Apr 02 08:12:14 2010 +0100
@@ -1442,7 +1442,8 @@
 
     set_current(next);
 
-    if ( (per_cpu(curr_vcpu, cpu) == next) || is_idle_vcpu(next) )
+    if ( (per_cpu(curr_vcpu, cpu) == next) ||
+         (is_idle_vcpu(next) && !cpu_is_offline(cpu)) )
     {
         local_irq_enable();
     }