Xen devel - Apr 2010 - Xen-4.0.0 RC9 Test Report. Xen: #21087 & Dom0: #4ebd13...

Hi All,
	Here is the test report for Xen-4.0.0-rc9(Xen C/S 21087). Our testing contains
old bug verification and regression testing. There is no new bug exposed in this
testing. The vcpu pin issue is fixed. VT-d/SR-IOV and TXT basic function can
work on our platform.
	We use Pv_ops(xen/master, 2.6.31.12) as Dom0 in our testing.

Status Summary
===========================================================================Feature
Result
------------------------------------------------------
VT-x					PASS
VT-x (EPT+VPID)		PASS
VT-x (Real Mode)		PASS
RAS 				Buggy
VT-d					PASS
SR-IOV				Buggy
TXT					PASS
PowerMgmt			PASS
Other				PASS

Xen Info
===========================================================================xen-changeset:
21087:e1af8db30335
pvops git: commit 4ebd1335d19117929e939878a2e38b6c92a0ffb7
ioemu git: commit f1d909f0f854194f5a40d850886d1413fb8b63c2

Old Open Bugs (3)
===========================================================================1.
System hang after many times cpu online/offline
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1596

2. [RAS] CPUs are not in the correct NUMA node after hot-add memory
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1573

3. [SR-IOV] Qemu report pci_msix_writel error while assigning VF to guest
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1575

Fixed Bugs(2):
===========================================================================1.
vcpu pin can''t work on PAE host
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1595

2. CPU may panic when running cpu online/offline for more than 100 times
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1594


Best Regards,
Jiajun

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

On 06/04/2010 09:47, "Xu, Jiajun" <jiajun.xu@intel.com> wrote:
> Old Open Bugs (3)
>
===========================================================================>
1. System hang after many times cpu online/offline
> http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1596
I have a potential fix for this which I hope Yunhong Jiang will comment on
soon. However it is probably too late for 4.0.0 now anyway, which we expect
to release tomorrow.

Thanks for all your testing efforts! And, of course, more broadly, Intel
team''s bug fixing.

 -- Keir



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

>-----Original Message-----
>From: Keir Fraser [mailto:keir.fraser@eu.citrix.com]
>Sent: Tuesday, April 06, 2010 5:45 PM
>To: Xu, Jiajun; xen-devel@lists.xensource.com
>Cc: Jiang, Yunhong
>Subject: Re: [Xen-devel] Xen-4.0.0 RC9 Test Report. Xen: #21087 & Dom0:
>#4ebd13...
>
>On 06/04/2010 09:47, "Xu, Jiajun" <jiajun.xu@intel.com>
wrote:
>
>> Old Open Bugs (3)
>>
>===================================================================>=======>>
1. System hang after many times cpu online/offline
>> http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1596
>
>I have a potential fix for this which I hope Yunhong Jiang will comment on
>soon. However it is probably too late for 4.0.0 now anyway, which we expect
>to release tomorrow.
Keir, really thanks for your patch very much. I should not take leave in last
Friday and this monday :( In fact, although your patch fixed the issue in my
mail, but there still another bug in PM side for CPU online/offline, which will
cause panic sometimes, so anyway, CPU online/offline can''t pass our
stress test in xen 4.0.
I''m testing the patch. Seems it at least passed loop count 500, o*line
all APs, leaves only BSP online.
A potential issue in the patch is, in following change, it may trigger the
assert of __sync_lazy_execstate(), which assume current is idle_vcpu, however,
at this time, we can''t gurrante this. A check for current vcpu is
needed.

+	/*
+	 * If we are running the idle vcpu, sync last non-idle vcpu''s state
+	 * before changing cpu_online_map. If we are running non-idle vcpu,
+         * we will synchronously sync the state in context_switch() later.
+	 */
+	__sync_lazy_execstate();

Again, really thanks for your help on this.

--jyh
>
>Thanks for all your testing efforts! And, of course, more broadly, Intel
>team''s bug fixing.
>
> -- Keir

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

On 06/04/2010 11:16, "Jiang, Yunhong" <yunhong.jiang@intel.com>
wrote:
> Keir, really thanks for your patch very much. I should not take leave in
last
> Friday and this monday :( In fact, although your patch fixed the issue in
my
> mail, but there still another bug in PM side for CPU online/offline, which
> will cause panic sometimes, so anyway, CPU online/offline can''t
pass our
> stress test in xen 4.0.
Okay, well, seems we will get there for 4.0.1 instead. Thanks!
> I''m testing the patch. Seems it at least passed loop count 500,
o*line all
> APs, leaves only BSP online.
> A potential issue in the patch is, in following change, it may trigger the
> assert of __sync_lazy_execstate(), which assume current is idle_vcpu,
however,
> at this time, we can''t gurrante this. A check for current vcpu is
needed.
Oh yes, obviously I didn''t test my own patch. ;-) Well I will make that
change and apply the patch to xen-unstable. We can wait to backport to
4.0-testing until we have a complete set of patches which pass all your
tests.
> Again, really thanks for your help on this.
Thank *you*!

 -- Keir



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

On 06/04/2010 11:16, "Jiang, Yunhong" <yunhong.jiang@intel.com>
wrote:
> Keir, really thanks for your patch very much. I should not take leave in
last
> Friday and this monday :( In fact, although your patch fixed the issue in
my
> mail, but there still another bug in PM side for CPU online/offline, which
> will cause panic sometimes, so anyway, CPU online/offline can''t
pass our
> stress test in xen 4.0.
> I''m testing the patch. Seems it at least passed loop count 500,
o*line all
> APs, leaves only BSP online.
> A potential issue in the patch is, in following change, it may trigger the
> assert of __sync_lazy_execstate(), which assume current is idle_vcpu,
however,
> at this time, we can''t gurrante this. A check for current vcpu is
needed.
I looked at the code again, and are you sure about this? As in, have you
seen the assertion trigger? The check that current is the idle_vcpu is only
made ''if(switch_required)'', and that can only be the case if
we are running
the idle_vcpu! So I think my patch is good as it is, would you agree?

 -- Keir



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

>-----Original Message-----
>From: Keir Fraser [mailto:keir.fraser@eu.citrix.com]
>Sent: Wednesday, April 07, 2010 3:04 PM
>To: Jiang, Yunhong; Xu, Jiajun; xen-devel@lists.xensource.com
>Subject: Re: [Xen-devel] Xen-4.0.0 RC9 Test Report. Xen: #21087 & Dom0:
>#4ebd13...
>
>On 06/04/2010 11:16, "Jiang, Yunhong"
<yunhong.jiang@intel.com> wrote:
>
>> Keir, really thanks for your patch very much. I should not take leave
in last
>> Friday and this monday :( In fact, although your patch fixed the issue
in my
>> mail, but there still another bug in PM side for CPU online/offline,
which
>> will cause panic sometimes, so anyway, CPU online/offline
can''t pass our
>> stress test in xen 4.0.
>> I''m testing the patch. Seems it at least passed loop count
500, o*line all
>> APs, leaves only BSP online.
>> A potential issue in the patch is, in following change, it may trigger
the
>> assert of __sync_lazy_execstate(), which assume current is idle_vcpu,
however,
>> at this time, we can''t gurrante this. A check for current vcpu
is needed.
>
>I looked at the code again, and are you sure about this? As in, have you
>seen the assertion trigger? The check that current is the idle_vcpu is only
>made ''if(switch_required)'', and that can only be the case
if we are running
>the idle_vcpu! So I think my patch is good as it is, would you agree?
Aha, yes, you are right, the patch is correct.
I tested your patch in my first round (I added the _redudant_ check in the
second round:$ ) and didn''t trigger the assertion, the first round runs
for about 900 round before triger another bug. So, yes, it''s a wrong
alarm.

--jyh
>
> -- Keir
>

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

On 07/04/2010 08:24, "Jiang, Yunhong" <yunhong.jiang@intel.com>
wrote:
>> I looked at the code again, and are you sure about this? As in, have
you
>> seen the assertion trigger? The check that current is the idle_vcpu is
only
>> made ''if(switch_required)'', and that can only be the
case if we are running
>> the idle_vcpu! So I think my patch is good as it is, would you agree?
> 
> Aha, yes, you are right, the patch is correct.
> I tested your patch in my first round (I added the _redudant_ check in the
> second round:$ ) and didn''t trigger the assertion, the first round
runs for
> about 900 round before triger another bug. So, yes, it''s a wrong
alarm.
I applied the patch as xen-unstable:21109. It actually includes a further
change, to add an extra BUG()-check to cpu_exit_clear(). I think it should
work fine.

 Thanks,
 Keir



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

>-----Original Message-----
>From: Keir Fraser [mailto:keir.fraser@eu.citrix.com]
>Sent: Wednesday, April 07, 2010 3:45 PM
>To: Jiang, Yunhong; Xu, Jiajun; xen-devel@lists.xensource.com
>Subject: Re: [Xen-devel] Xen-4.0.0 RC9 Test Report. Xen: #21087 & Dom0:
>#4ebd13...
>
>On 07/04/2010 08:24, "Jiang, Yunhong"
<yunhong.jiang@intel.com> wrote:
>
>>> I looked at the code again, and are you sure about this? As in,
have you
>>> seen the assertion trigger? The check that current is the idle_vcpu
is only
>>> made ''if(switch_required)'', and that can only be
the case if we are running
>>> the idle_vcpu! So I think my patch is good as it is, would you
agree?
>>
>> Aha, yes, you are right, the patch is correct.
>> I tested your patch in my first round (I added the _redudant_ check in
the
>> second round:$ ) and didn''t trigger the assertion, the first
round runs for
>> about 900 round before triger another bug. So, yes, it''s a
wrong alarm.
>
>I applied the patch as xen-unstable:21109. It actually includes a further
>change, to add an extra BUG()-check to cpu_exit_clear(). I think it should
>work fine.
Really thanks. I will test it later.

--jyh
>
> Thanks,
> Keir
>

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