On 01.08.2012, at 04:13, Scott Wood wrote: > On 07/31/2012 06:59 PM, Alexander Graf wrote: >> Well, user space always comes in via the ioctl path, which in turn >> does vcpu_load(vcpu), thus should always be running as the vcpu >> context itself. > > Right, it's the timers you're racing with. > >> It boils down to the original thing I was saying back when you >> introduced asynchronous TSR updates. > > Right. > >> Why don't we just make TSR >> updates kvm requests? > > Maybe we should. I prefer to have my code easy to understand. And every time I see cmpxchg and spin locks my readability skills take a serious hit :). The big question is whether we care for the scope of the watchdog patch. Changing it to requests is a purely internal rework. So we could leave it with the (unlikely) race for now, agree on the fact that we won't introduce any other direct TSR accessors and move the code to be request based soon'ish. Alex -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
