#56 - Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]

Owner: Julien Grall <julien.grall@arm.com>

Date: Wed Dec 28 15:00:02 2016

Last Update: Wed Dec 28 15:00:02 2016

Severity: blocker

Affects:

State: Open

[ Retrieve as mbox ]


From: Andrew Cooper <andrew.cooper3@citrix.com>
To: xen-devel@lists.xen.org, "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>
Cc: Julien Grall <julien.grall@arm.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Fri, 23 Dec 2016 17:00:31 +0000
Message-ID: <a182bc22-6954-5adb-8872-56bf531eaadc@citrix.com>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
> Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version have the
> same problem.
>
> My cpu is AMD A10-7860K

This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS 
record truncated" during domain restore").

I submitted patches to fix it ("Fix issues with zero-length records in 
migration v2" in July even), which made no progress.  As the author of 
migration v2, and unfortunately of this bug, I stand by v1 of my fix 
without any further modification.

You can find the patches here: 
https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=refs/heads/tools-fix-zero-length-records

Julien: As 4.9 RM, please mark this as a release blocker.  It is very 
poor that we as a community had this reported and fixed 6 months ago, 
and yet it still didn't make it into 4.8

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

From: Andrew Cooper <andrew.cooper3@citrix.com>
To: xen-devel@lists.xen.org, "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>
Cc: Julien Grall <julien.grall@arm.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Fri, 23 Dec 2016 17:16:32 +0000
Message-ID: <a0a173e8-fc15-68e9-8e71-46135c8dad6a@citrix.com>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

On 23/12/16 17:00, Andrew Cooper wrote:
> On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
>> Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version 
>> have the
>> same problem.
>>
>> My cpu is AMD A10-7860K
>
> This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS 
> record truncated" during domain restore").
>
> I submitted patches to fix it ("Fix issues with zero-length records in 
> migration v2" in July even), which made no progress. As the author of 
> migration v2, and unfortunately of this bug, I stand by v1 of my fix 
> without any further modification.
>
> You can find the patches here: 
> https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=refs/heads/tools-fix-zero-length-records

If you don't want to take patches, you should be able to work around the 
issue by booting Xen with

cpuid_mask_ext_ecx=fbffffff

Specifically, you are looking to hide the DBEXT feature from Xen so it 
doesn't choose to insert them into the migration stream to start with.  
This migration bug only manifests when the hardware is capable, but the 
VM isn't using the feature.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

From: "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>
To: Andrew Cooper <andrew.cooper3@citrix.com>
Cc: xen-devel@lists.xen.org, Julien Grall <julien.grall@arm.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Fri, 23 Dec 2016 19:56:19 -0300
Message-ID: <11067138.3bJAmV3OH1@rick>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

El Vie 23 Dic 2016 17:16:32 Andrew Cooper escribió:
> On 23/12/16 17:00, Andrew Cooper wrote:
> > On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
> >> Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version
> >> have the
> >> same problem.
> >> 
> >> My cpu is AMD A10-7860K
> > 
> > This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS
> > record truncated" during domain restore").
> > 
> > I submitted patches to fix it ("Fix issues with zero-length records in
> > migration v2" in July even), which made no progress. As the author of
> > migration v2, and unfortunately of this bug, I stand by v1 of my fix
> > without any further modification.
> > 
> > You can find the patches here:
> > https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=r
> > efs/heads/tools-fix-zero-length-records
> If you don't want to take patches, you should be able to work around the
> issue by booting Xen with
> 
> cpuid_mask_ext_ecx=fbffffff
> 
> Specifically, you are looking to hide the DBEXT feature from Xen so it
> doesn't choose to insert them into the migration stream to start with.
> This migration bug only manifests when the hardware is capable, but the
> VM isn't using the feature.
> 
> ~Andrew

Hi Andrew, first i try boot xen with cpuid_mask_ext_ecx=fbffffff, restore and 
migration works well, then try apply the patchs 

tools/python: Adjust migration v2 library to warn about...
tools/libxc: Avoid generating inappropriate zero-length...	
tools/libxc: Tolerate zero-length records in migration...

remove cpuid_mask, recompile and install xen-4.7.1, reboot and the problem 
persist. I'm missing something?

Regards


 




_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

From: Andrew Cooper <andrew.cooper3@citrix.com>
To: "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>
Cc: xen-devel@lists.xen.org, Julien Grall <julien.grall@arm.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Fri, 23 Dec 2016 23:22:17 +0000
Message-ID: <1488cbde-b327-82c8-fc5c-4ba717f9fcbb@citrix.com>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

On 23/12/16 22:56, Ing. Ricardo Brisighelli wrote:
> El Vie 23 Dic 2016 17:16:32 Andrew Cooper escribió:
>> On 23/12/16 17:00, Andrew Cooper wrote:
>>> On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
>>>> Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version
>>>> have the
>>>> same problem.
>>>>
>>>> My cpu is AMD A10-7860K
>>> This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS
>>> record truncated" during domain restore").
>>>
>>> I submitted patches to fix it ("Fix issues with zero-length records in
>>> migration v2" in July even), which made no progress. As the author of
>>> migration v2, and unfortunately of this bug, I stand by v1 of my fix
>>> without any further modification.
>>>
>>> You can find the patches here:
>>> https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=r
>>> efs/heads/tools-fix-zero-length-records
>> If you don't want to take patches, you should be able to work around the
>> issue by booting Xen with
>>
>> cpuid_mask_ext_ecx=fbffffff
>>
>> Specifically, you are looking to hide the DBEXT feature from Xen so it
>> doesn't choose to insert them into the migration stream to start with.
>> This migration bug only manifests when the hardware is capable, but the
>> VM isn't using the feature.
>>
>> ~Andrew
> Hi Andrew, first i try boot xen with cpuid_mask_ext_ecx=fbffffff, restore and
> migration works well, then try apply the patchs
>
> tools/python: Adjust migration v2 library to warn about...
> tools/libxc: Avoid generating inappropriate zero-length...	
> tools/libxc: Tolerate zero-length records in migration...
>
> remove cpuid_mask, recompile and install xen-4.7.1, reboot and the problem
> persist. I'm missing something?

Does Gentoo split the various parts of Xen apart into sub-packages? This 
needs to be the dom0 tools build, not the hypervisor build.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

From: "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>
To: xen-devel@lists.xen.org
Cc: Julien Grall <julien.grall@arm.com>, Andrew Cooper <andrew.cooper3@citrix.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Sat, 24 Dec 2016 10:24:30 -0300
Message-ID: <11963615.l4C2HeH9aP@rick>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

El Vie 23 Dic 2016 23:22:17 Andrew Cooper escribió:
> On 23/12/16 22:56, Ing. Ricardo Brisighelli wrote:
> > El Vie 23 Dic 2016 17:16:32 Andrew Cooper escribió:
> >> On 23/12/16 17:00, Andrew Cooper wrote:
> >>> On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
> >>>> Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version
> >>>> have the
> >>>> same problem.
> >>>> 
> >>>> My cpu is AMD A10-7860K
> >>> 
> >>> This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS
> >>> record truncated" during domain restore").
> >>> 
> >>> I submitted patches to fix it ("Fix issues with zero-length records in
> >>> migration v2" in July even), which made no progress. As the author of
> >>> migration v2, and unfortunately of this bug, I stand by v1 of my fix
> >>> without any further modification.
> >>> 
> >>> You can find the patches here:
> >>> https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h
> >>> =r
> >>> efs/heads/tools-fix-zero-length-records
> >> 
> >> If you don't want to take patches, you should be able to work around the
> >> issue by booting Xen with
> >> 
> >> cpuid_mask_ext_ecx=fbffffff
> >> 
> >> Specifically, you are looking to hide the DBEXT feature from Xen so it
> >> doesn't choose to insert them into the migration stream to start with.
> >> This migration bug only manifests when the hardware is capable, but the
> >> VM isn't using the feature.
> >> 
> >> ~Andrew
> > 
> > Hi Andrew, first i try boot xen with cpuid_mask_ext_ecx=fbffffff, restore
> > and migration works well, then try apply the patchs
> > 
> > tools/python: Adjust migration v2 library to warn about...
> > tools/libxc: Avoid generating inappropriate zero-length...
> > tools/libxc: Tolerate zero-length records in migration...
> > 
> > remove cpuid_mask, recompile and install xen-4.7.1, reboot and the problem
> > persist. I'm missing something?
> 
> Does Gentoo split the various parts of Xen apart into sub-packages? This
> needs to be the dom0 tools build, not the hypervisor build.
> 
> ~Andrew
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> https://lists.xen.org/xen-devel
> 

I recompile xen-tools and all works well.

Thanks very much Andrew

Regards






_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

From: Doug Goldstein <cardoe@cardoe.com>
To: "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>, xen-devel@lists.xen.org
Cc: Julien Grall <julien.grall@arm.com>, Andrew Cooper <andrew.cooper3@citrix.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Mon, 26 Dec 2016 14:43:03 -0600
Message-ID: <076f39eb-9a89-8b67-b682-95f640e17d70@cardoe.com>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

[Part 1 (text/plain, inline)]
>>>>> This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS
>>>>> record truncated" during domain restore").
>>>>>
>>>>> I submitted patches to fix it ("Fix issues with zero-length records in
>>>>> migration v2" in July even), which made no progress. As the author of
>>>>> migration v2, and unfortunately of this bug, I stand by v1 of my fix
>>>>> without any further modification.
>>>>>
>>>>> You can find the patches here:
>>>>> https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h
>>>>> =r
>>>>> efs/heads/tools-fix-zero-length-records
>>>>

FWIW, this is fixed in Gentoo now with xen-4.8.0-r1 or xen-4.7.1-r5 and
newer.

https://github.com/gentoo/gentoo/commit/ea2b45715c75943192933e14b58a348027927f2a

-- 
Doug Goldstein
[signature.asc (application/pgp-signature, attachment)]
[Part 3 (text/plain, inline)]

From: Julien Grall <julien.grall@arm.com>
To: Andrew Cooper <andrew.cooper3@citrix.com>, "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>, xen-devel@lists.xen.org
Cc: Wei Liu <wei.liu2@citrix.com>, Ian Jackson <Ian.Jackson@eu.citrix.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Wed, 28 Dec 2016 14:50:07 +0000
Message-ID: <daef8bbe-4bb2-03c6-4802-492b3b995ec3@arm.com>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

(Adding Ian and Wei)

Hi Andrew,

On 23/12/16 17:00, Andrew Cooper wrote:
> On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
>> Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version
>> have the
>> same problem.
>>
>> My cpu is AMD A10-7860K
>
> This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS
> record truncated" during domain restore").
>
> I submitted patches to fix it ("Fix issues with zero-length records in
> migration v2" in July even), which made no progress.  As the author of
> migration v2, and unfortunately of this bug, I stand by v1 of my fix
> without any further modification.
>
> You can find the patches here:
> https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=refs/heads/tools-fix-zero-length-records
>
>
> Julien: As 4.9 RM, please mark this as a release blocker.  It is very
> poor that we as a community had this reported and fixed 6 months ago,
> and yet it still didn't make it into 4.8

Sounds sensible. I will create an entry in the bug tracker.

Cheers,

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

From: Julien Grall <julien.grall@arm.com>
To: Andrew Cooper <andrew.cooper3@citrix.com>, xen-devel@lists.xen.org, "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Wed, 28 Dec 2016 14:58:12 +0000
Message-ID: <f61005fa-891a-0ce9-be40-371d4e9fe817@arm.com>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

create ^
severity it blocker
affects it 4.6, 4.7, 4.8

thanks

On 23/12/16 17:00, Andrew Cooper wrote:
> On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
>> Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version
>> have the
>> same problem.
>>
>> My cpu is AMD A10-7860K
>
> This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS
> record truncated" during domain restore").
>
> I submitted patches to fix it ("Fix issues with zero-length records in
> migration v2" in July even), which made no progress.  As the author of
> migration v2, and unfortunately of this bug, I stand by v1 of my fix
> without any further modification.
>
> You can find the patches here:
> https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=refs/heads/tools-fix-zero-length-records
>
>
> Julien: As 4.9 RM, please mark this as a release blocker.  It is very
> poor that we as a community had this reported and fixed 6 months ago,
> and yet it still didn't make it into 4.8
>
> ~Andrew

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel


From: Wei Liu <wei.liu2@citrix.com>
To: Julien Grall <julien.grall@arm.com>
Cc: Ian Jackson <Ian.Jackson@eu.citrix.com>, "Ing. Ricardo Brisighelli" <ricardo@estilo.com.ar>, xen-devel@lists.xen.org, Wei Liu <wei.liu2@citrix.com>, Andrew Cooper <andrew.cooper3@citrix.com>
Subject: Re: [Xen-devel] [xl restore and migrating problem - hardware compatibility]
Date: Wed, 4 Jan 2017 11:00:13 +0000
Message-ID: <20170104110013.GF13806@citrix.com>

[ Reply to this message; Retrieve Raw Message; Archives: marc.info, gmane ]

On Wed, Dec 28, 2016 at 02:50:07PM +0000, Julien Grall wrote:
> (Adding Ian and Wei)
> 
> Hi Andrew,
> 
> On 23/12/16 17:00, Andrew Cooper wrote:
> >On 23/12/16 16:32, Ing. Ricardo Brisighelli wrote:
> >>Hi, i'm gentoo user and try with xen 4.6.3 and 4.7.1 in both version
> >>have the
> >>same problem.
> >>
> >>My cpu is AMD A10-7860K
> >
> >This issue has been reported before ("[Xen-devel] "X86_PV_VCPU_MSRS
> >record truncated" during domain restore").
> >
> >I submitted patches to fix it ("Fix issues with zero-length records in
> >migration v2" in July even), which made no progress.  As the author of
> >migration v2, and unfortunately of this bug, I stand by v1 of my fix
> >without any further modification.
> >
> >You can find the patches here:
> >https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=refs/heads/tools-fix-zero-length-records
> >
> >
> >Julien: As 4.9 RM, please mark this as a release blocker.  It is very
> >poor that we as a community had this reported and fixed 6 months ago,
> >and yet it still didn't make it into 4.8
> 
> Sounds sensible. I will create an entry in the bug tracker.
> 

Ian made some comments on the old thread but Andrew didn't reply back.

We should resurrect the discussion there.

Wei.

> Cheers,
> 
> -- 
> Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel