Re: [Xen-devel] [PATCH] schedule: mask out XEN_RUNSTATE_UPDATE from state entry time

2019-07-18 Thread Andrii Anisov
Hello Jan, On 18.07.19 18:04, Jan Beulich wrote: On 18.07.2019 16:12, Andrii Anisov wrote: The scenario is quite trivial: some vcpu uptdating its runstate values (e.g. context switching) while those values are being read by a guest using other vcpu. Well, that's (afaia) not an intended usage

Re: [Xen-devel] [PATCH] schedule: mask out XEN_RUNSTATE_UPDATE from state entry time

2019-07-18 Thread Jan Beulich
On 18.07.2019 16:12, Andrii Anisov wrote: > On 18.07.19 14:10, Jan Beulich wrote: >> A concrete scenario where update_runstate_area() and vcpu_runstate_change() >> can actually race would be very worthwhile to point out here. In particular >> on Arm I'm not (yet?) seeing how this could happen. > >

Re: [Xen-devel] [PATCH] schedule: mask out XEN_RUNSTATE_UPDATE from state entry time

2019-07-18 Thread Andrii Anisov
Hello Jan, On 18.07.19 14:10, Jan Beulich wrote: Considering the value of XEN_RUNSTATE_UPDATE it must be a rather rare race anyway, I would guess. I'm not sure about the exact rate of the race, but with following prints: diff --git a/xen/common/schedule.c b/xen/common/schedule.c index 25f6ab3

Re: [Xen-devel] [PATCH] schedule: mask out XEN_RUNSTATE_UPDATE from state entry time

2019-07-18 Thread Andrii Anisov
Hello Jan, On 18.07.19 14:10, Jan Beulich wrote: A concrete scenario where update_runstate_area() and vcpu_runstate_change() can actually race would be very worthwhile to point out here. In particular on Arm I'm not (yet?) seeing how this could happen. The scenario is quite trivial: some vcpu

Re: [Xen-devel] [PATCH] schedule: mask out XEN_RUNSTATE_UPDATE from state entry time

2019-07-18 Thread Jan Beulich
On 18.07.2019 11:54, Andrii Anisov wrote: > Using XEN_RUNSTATE_UPDATE mask during the process of copying runstate > values to guest causes runstate entry time to be eventually considered > negative on a calculation of the time delta. So the XEN_RUNSTATE_UPDATE > should be masked out during the calc

Re: [Xen-devel] [PATCH] schedule: mask out XEN_RUNSTATE_UPDATE from state entry time

2019-07-18 Thread Andrii Anisov
Hello George, The patch was also CC'ed to you, but I got: "554 esa1.hc3370-68.iphmx.com You are being rejected because your senderbase score is below our accepted policy." from your server. ps. I hope I'm not bothering you too much. On 18.07.19 12:54, Andrii Anisov wrote: From: Andrii Aniso

[Xen-devel] [PATCH] schedule: mask out XEN_RUNSTATE_UPDATE from state entry time

2019-07-18 Thread Andrii Anisov
From: Andrii Anisov Using XEN_RUNSTATE_UPDATE mask during the process of copying runstate values to guest causes runstate entry time to be eventually considered negative on a calculation of the time delta. So the XEN_RUNSTATE_UPDATE should be masked out during the calculation of the time spent in