[Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andreas Kinzler]

While AMD Ryzen 2700X was working perfectly in my tests with Windows 10, 
the new 3700X does not even boot a Windows HVM. With viridian=1 you get 
BSOD HAL_MEMORY_ALLOCATION and with viridian=0 you get "multiprocessor 
config not supported".

xl dmesg says:
(XEN) d1v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8065c06bf88 bf8
(XEN) d2v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8035b049f88 bf8

Linux domUs with PV and PVH seem to work so far.

Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is unchanged 
from 2700X (working) to 3700X (crashing).

Is it a known problem? Did someone test the new EPYCs?

Regards Andreas


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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andrew Cooper]

On 20/08/2019 19:03, Andreas Kinzler wrote:
> While AMD Ryzen 2700X was working perfectly in my tests with Windows
> 10, the new 3700X does not even boot a Windows HVM. With viridian=1
> you get BSOD HAL_MEMORY_ALLOCATION and with viridian=0 you get
> "multiprocessor config not supported".
>
> xl dmesg says:
> (XEN) d1v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8065c06bf88 bf8
> (XEN) d2v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8035b049f88 bf8
>
> Linux domUs with PV and PVH seem to work so far.
>
> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is unchanged
> from 2700X (working) to 3700X (crashing).

So you've done a Zen v1 => Zen v2 CPU upgrade and an existing system?

> Is it a known problem? Did someone test the new EPYCs?

This looks familiar, and is still somewhere on my TODO list.

Does booting with a single vCPU work?

~Andrew

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andrew Cooper]

On 20/08/2019 19:03, Andreas Kinzler wrote:
> While AMD Ryzen 2700X was working perfectly in my tests with Windows
> 10, the new 3700X does not even boot a Windows HVM. With viridian=1
> you get BSOD HAL_MEMORY_ALLOCATION and with viridian=0 you get
> "multiprocessor config not supported".
>
> xl dmesg says:
> (XEN) d1v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8065c06bf88 bf8
> (XEN) d2v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8035b049f88 bf8
>
> Linux domUs with PV and PVH seem to work so far.
>
> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is unchanged
> from 2700X (working) to 3700X (crashing).

So you've done a Zen v1 => Zen v2 CPU upgrade and an existing system?

>
> Is it a known problem? Did someone test the new EPYCs?

This looks familiar, and is still somewhere on my TODO list.

Does booting with a single vCPU work?

~Andrew

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andreas Kinzler]

On 20.08.2019 20:12, Andrew Cooper wrote:
>> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is unchanged
>> from 2700X (working) to 3700X (crashing).
> So you've done a Zen v1 => Zen v2 CPU upgrade and an existing system?

With "existing system" you mean the Windows installation? Yes, but it is 
not relevant. The same BSODs happen if you boot the HVM with just the 
iso installation medium and no disks.

>> Is it a known problem? Did someone test the new EPYCs?
> This looks familiar, and is still somewhere on my TODO list.

Do you already know the reason or is that still to investigate?

> Does booting with a single vCPU work?

Number of vCPUs make no difference

Regards Andreas

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andrew Cooper]

On 20/08/2019 21:36, Andreas Kinzler wrote:
> On 20.08.2019 20:12, Andrew Cooper wrote:
>>> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is unchanged
>>> from 2700X (working) to 3700X (crashing).
>> So you've done a Zen v1 => Zen v2 CPU upgrade and an existing system?
>
> With "existing system" you mean the Windows installation?

I meant same computer, not same VM.

> Yes, but it is not relevant. The same BSODs happen if you boot the HVM
> with just the iso installation medium and no disks.

That's a useful datapoint.  I wouldn't expect this to be relevant, given
how Window's HAL works.

>
>>> Is it a known problem? Did someone test the new EPYCs?
>> This looks familiar, and is still somewhere on my TODO list.
>
> Do you already know the reason or is that still to investigate?
>
>> Does booting with a single vCPU work?
>
> Number of vCPUs make no difference

Hmm - perhaps its not the same issue then.  Either way, its firmly in
the "still to investigate" phase.

~Andrew

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andreas Kinzler]

On 20.08.2019 22:38, Andrew Cooper wrote:
> On 20/08/2019 21:36, Andreas Kinzler wrote:
>> On 20.08.2019 20:12, Andrew Cooper wrote:
>>>> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is unchanged
>>>> from 2700X (working) to 3700X (crashing).
>>> So you've done a Zen v1 => Zen v2 CPU upgrade and an existing system?
>> With "existing system" you mean the Windows installation?
> I meant same computer, not same VM.

Tried with 2 mainboards: Asrock X370 Pro4 and AsrockRack X470D4U.
You need to flash the BIOS for Zen2. X470D4U BIOS 3.1 works with 2700X 
but not with 3700X. X370 Pro4 with somewhat older BIOS worked for 2700X 
and does not work with current (6.00) BIOS and 3700X.

>> Yes, but it is not relevant. The same BSODs happen if you boot the HVM
>> with just the iso installation medium and no disks.
> That's a useful datapoint.  I wouldn't expect this to be relevant, given
> how Window's HAL works.

It should make debugging for you quite "simple" because it can be 
reproduced very easily.

Regards Andreas

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Steven Haigh]

On 2019-08-21 06:57, Andreas Kinzler wrote:
> On 20.08.2019 22:38, Andrew Cooper wrote:
>> On 20/08/2019 21:36, Andreas Kinzler wrote:
>>> On 20.08.2019 20:12, Andrew Cooper wrote:
>>>>> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is 
>>>>> unchanged
>>>>> from 2700X (working) to 3700X (crashing).
>>>> So you've done a Zen v1 => Zen v2 CPU upgrade and an existing 
>>>> system?
>>> With "existing system" you mean the Windows installation?
>> I meant same computer, not same VM.
> 
> Tried with 2 mainboards: Asrock X370 Pro4 and AsrockRack X470D4U.
> You need to flash the BIOS for Zen2. X470D4U BIOS 3.1 works with 2700X
> but not with 3700X. X370 Pro4 with somewhat older BIOS worked for
> 2700X and does not work with current (6.00) BIOS and 3700X.
> 
>>> Yes, but it is not relevant. The same BSODs happen if you boot the 
>>> HVM
>>> with just the iso installation medium and no disks.
>> That's a useful datapoint.  I wouldn't expect this to be relevant, 
>> given
>> how Window's HAL works.
> 
> It should make debugging for you quite "simple" because it can be
> reproduced very easily.

Just to add a data point to this - I also see this problem on a Ryzen 9 
3900x.

xl dmesg shows:
(XEN) d2v0 VIRIDIAN CRASH: ac 0 a0a0 fffff80293254750 aea
(XEN) d3v0 VIRIDIAN CRASH: ac 0 a0a0 fffff80093a40750 aea
(XEN) d5v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8028e422350 aea
(XEN) d6v0 VIRIDIAN CRASH: ac 0 a0a0 fffff80309431750 aea
(XEN) d10v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8012823e750 aea
(XEN) d11v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8032e657350 aea

Windows usually has a stopcode of "HAL MEMORY ALLOCATION" when it blue 
screens.

 From xl info:
hw_caps                : 
178bf3ff:f6d8320b:2e500800:244037ff:0000000f:219c91a9:00400004:00000500
virt_caps              : hvm hvm_directio
xen_version            : 4.11.2
xen_caps               : xen-3.0-x86_64 xen-3.0-x86_32p hvm-3.0-x86_32 
hvm-3.0-x86_32p hvm-3.0-x86_64

Is there any further info that can be provided? Not being able to 
virtualise Windows is a bit of a PITA...

-- 
Steven Haigh

? netwiz@crc.id.au     ? http://www.crc.id.au
? +61 (3) 9001 6090    ? 0412 935 897

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Steven Haigh]

On 2019-09-02 11:42, Steven Haigh wrote:
> On 2019-08-21 06:57, Andreas Kinzler wrote:
>> On 20.08.2019 22:38, Andrew Cooper wrote:
>>> On 20/08/2019 21:36, Andreas Kinzler wrote:
>>>> On 20.08.2019 20:12, Andrew Cooper wrote:
>>>>>> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is 
>>>>>> unchanged
>>>>>> from 2700X (working) to 3700X (crashing).
>>>>> So you've done a Zen v1 => Zen v2 CPU upgrade and an existing 
>>>>> system?
>>>> With "existing system" you mean the Windows installation?
>>> I meant same computer, not same VM.
>> 
>> Tried with 2 mainboards: Asrock X370 Pro4 and AsrockRack X470D4U.
>> You need to flash the BIOS for Zen2. X470D4U BIOS 3.1 works with 2700X
>> but not with 3700X. X370 Pro4 with somewhat older BIOS worked for
>> 2700X and does not work with current (6.00) BIOS and 3700X.
>> 
>>>> Yes, but it is not relevant. The same BSODs happen if you boot the 
>>>> HVM
>>>> with just the iso installation medium and no disks.
>>> That's a useful datapoint.  I wouldn't expect this to be relevant, 
>>> given
>>> how Window's HAL works.
>> 
>> It should make debugging for you quite "simple" because it can be
>> reproduced very easily.
> 
> Just to add a data point to this - I also see this problem on a Ryzen 9 
> 3900x.
> 
> xl dmesg shows:
> (XEN) d2v0 VIRIDIAN CRASH: ac 0 a0a0 fffff80293254750 aea
> (XEN) d3v0 VIRIDIAN CRASH: ac 0 a0a0 fffff80093a40750 aea
> (XEN) d5v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8028e422350 aea
> (XEN) d6v0 VIRIDIAN CRASH: ac 0 a0a0 fffff80309431750 aea
> (XEN) d10v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8012823e750 aea
> (XEN) d11v0 VIRIDIAN CRASH: ac 0 a0a0 fffff8032e657350 aea
> 
> Windows usually has a stopcode of "HAL MEMORY ALLOCATION" when it blue 
> screens.
> 
> From xl info:
> hw_caps                :
> 178bf3ff:f6d8320b:2e500800:244037ff:0000000f:219c91a9:00400004:00000500
> virt_caps              : hvm hvm_directio
> xen_version            : 4.11.2
> xen_caps               : xen-3.0-x86_64 xen-3.0-x86_32p hvm-3.0-x86_32
> hvm-3.0-x86_32p hvm-3.0-x86_64
> 
> Is there any further info that can be provided? Not being able to
> virtualise Windows is a bit of a PITA...

Further to the above, I did some experimentation. The following is a 
list of attempted boot configurations and their outcomes:

viridian=1
vcpus=4
STOPCODE: HAL MEMORY ALLOCATION

viridian=0
vcpus=4
STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED

viridian=0
vcpus=1
Boot OK - get to Windows Server 2016 login etc

As such, it looks like its not a completely fatal problem - but running 
Windows on a single vcpu is.... unpleasant ;)

-- 
Steven Haigh

? netwiz@crc.id.au     ? http://www.crc.id.au
? +61 (3) 9001 6090    ? 0412 935 897

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Paul Durrant]

> -----Original Message-----
> 
> Further to the above, I did some experimentation. The following is a
> list of attempted boot configurations and their outcomes:
> 
> viridian=1
> vcpus=4
> STOPCODE: HAL MEMORY ALLOCATION
> 
> viridian=0
> vcpus=4
> STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED
> 
> viridian=0
> vcpus=1
> Boot OK - get to Windows Server 2016 login etc
> 

And to complete the set, how about viridian=1 vcpus=1?

  Paul
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Steven Haigh]

On 2019-09-02 18:04, Paul Durrant wrote:
>> -----Original Message-----
>> Further to the above, I did some experimentation. The following is a
>> list of attempted boot configurations and their outcomes:
>> 
>> viridian=1
>> vcpus=4
>> STOPCODE: HAL MEMORY ALLOCATION
>> 
>> viridian=0
>> vcpus=4
>> STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED
>> 
>> viridian=0
>> vcpus=1
>> Boot OK - get to Windows Server 2016 login etc
>> 
> 
> And to complete the set, how about viridian=1 vcpus=1?

Any vcpus value where viridian=1 is used creates a HAL MEMORY ALLOCATION 
stopcode when trying to boot Windows.

-- 
Steven Haigh

? netwiz@crc.id.au     ? http://www.crc.id.au
? +61 (3) 9001 6090    ? 0412 935 897

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Paul Durrant]

> -----Original Message-----
> From: Steven Haigh <netwiz@crc.id.au>
> Sent: 02 September 2019 09:09
> To: Paul Durrant <Paul.Durrant@citrix.com>
> Cc: Andreas Kinzler <hfp@posteo.de>; Andrew Cooper <Andrew.Cooper3@citrix.com>; xen-
> devel@lists.xenproject.org
> Subject: Re: Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)
> 
> On 2019-09-02 18:04, Paul Durrant wrote:
> >> -----Original Message-----
> >> Further to the above, I did some experimentation. The following is a
> >> list of attempted boot configurations and their outcomes:
> >>
> >> viridian=1
> >> vcpus=4
> >> STOPCODE: HAL MEMORY ALLOCATION
> >>
> >> viridian=0
> >> vcpus=4
> >> STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED
> >>
> >> viridian=0
> >> vcpus=1
> >> Boot OK - get to Windows Server 2016 login etc
> >>
> >
> > And to complete the set, how about viridian=1 vcpus=1?
> 
> Any vcpus value where viridian=1 is used creates a HAL MEMORY ALLOCATION
> stopcode when trying to boot Windows.

Ok, so I guess that issue hits first and, only if you get beyond that do you hit the multiprocessor problem.

The viridian option is not actually a boolean any more (that interpretation is just for compat) so it would be a good datapoint to know which of the enlightenments causes the change in behaviour. Could you try viridian=['base'] to see if that's sufficient to cause the problem? (I'm guessing it probably is but it would be good to know).

  Paul

> 
> --
> Steven Haigh
> 
> ? netwiz@crc.id.au     ? http://www.crc.id.au
> ? +61 (3) 9001 6090    ? 0412 935 897
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Steven Haigh]

On 2019-09-02 18:20, Paul Durrant wrote:
>> -----Original Message-----
>> From: Steven Haigh <netwiz@crc.id.au>
>> Sent: 02 September 2019 09:09
>> To: Paul Durrant <Paul.Durrant@citrix.com>
>> Cc: Andreas Kinzler <hfp@posteo.de>; Andrew Cooper 
>> <Andrew.Cooper3@citrix.com>; xen-
>> devel@lists.xenproject.org
>> Subject: Re: Windows HVM no longer boots with AMD Ryzen 3700X (and 
>> 3900X)
>> 
>> On 2019-09-02 18:04, Paul Durrant wrote:
>> >> -----Original Message-----
>> >> Further to the above, I did some experimentation. The following is a
>> >> list of attempted boot configurations and their outcomes:
>> >>
>> >> viridian=1
>> >> vcpus=4
>> >> STOPCODE: HAL MEMORY ALLOCATION
>> >>
>> >> viridian=0
>> >> vcpus=4
>> >> STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED
>> >>
>> >> viridian=0
>> >> vcpus=1
>> >> Boot OK - get to Windows Server 2016 login etc
>> >>
>> >
>> > And to complete the set, how about viridian=1 vcpus=1?
>> 
>> Any vcpus value where viridian=1 is used creates a HAL MEMORY 
>> ALLOCATION
>> stopcode when trying to boot Windows.
> 
> Ok, so I guess that issue hits first and, only if you get beyond that
> do you hit the multiprocessor problem.
> 
> The viridian option is not actually a boolean any more (that
> interpretation is just for compat) so it would be a good datapoint to
> know which of the enlightenments causes the change in behaviour. Could
> you try viridian=['base'] to see if that's sufficient to cause the
> problem? (I'm guessing it probably is but it would be good to know).


Hi Paul,

I can confirm that viridian=['base'] crashes with the same HAL MEMORY 
ALLOCATION stopcode - even on 1 vcpu.

-- 
Steven Haigh

? netwiz@crc.id.au     ? http://www.crc.id.au
? +61 (3) 9001 6090    ? 0412 935 897

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Steven Haigh]

On 2019-09-02 18:25, Steven Haigh wrote:
> On 2019-09-02 18:20, Paul Durrant wrote:
>>> -----Original Message-----
>>> From: Steven Haigh <netwiz@crc.id.au>
>>> Sent: 02 September 2019 09:09
>>> To: Paul Durrant <Paul.Durrant@citrix.com>
>>> Cc: Andreas Kinzler <hfp@posteo.de>; Andrew Cooper 
>>> <Andrew.Cooper3@citrix.com>; xen-
>>> devel@lists.xenproject.org
>>> Subject: Re: Windows HVM no longer boots with AMD Ryzen 3700X (and 
>>> 3900X)
>>> 
>>> On 2019-09-02 18:04, Paul Durrant wrote:
>>> >> -----Original Message-----
>>> >> Further to the above, I did some experimentation. The following is a
>>> >> list of attempted boot configurations and their outcomes:
>>> >>
>>> >> viridian=1
>>> >> vcpus=4
>>> >> STOPCODE: HAL MEMORY ALLOCATION
>>> >>
>>> >> viridian=0
>>> >> vcpus=4
>>> >> STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED
>>> >>
>>> >> viridian=0
>>> >> vcpus=1
>>> >> Boot OK - get to Windows Server 2016 login etc
>>> >>
>>> >
>>> > And to complete the set, how about viridian=1 vcpus=1?
>>> 
>>> Any vcpus value where viridian=1 is used creates a HAL MEMORY 
>>> ALLOCATION
>>> stopcode when trying to boot Windows.
>> 
>> Ok, so I guess that issue hits first and, only if you get beyond that
>> do you hit the multiprocessor problem.
>> 
>> The viridian option is not actually a boolean any more (that
>> interpretation is just for compat) so it would be a good datapoint to
>> know which of the enlightenments causes the change in behaviour. Could
>> you try viridian=['base'] to see if that's sufficient to cause the
>> problem? (I'm guessing it probably is but it would be good to know).
> 
> 
> Hi Paul,
> 
> I can confirm that viridian=['base'] crashes with the same HAL MEMORY
> ALLOCATION stopcode - even on 1 vcpu.

Also, just wondering, we're using 8.2.0 of the Windows PV drivers on 
this VM.

Does this matter? Has there been any changes that would affect this in 
8.2.1 or 8.2.2?

-- 
Steven Haigh

? netwiz@crc.id.au     ? http://www.crc.id.au
? +61 (3) 9001 6090    ? 0412 935 897

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Paul Durrant]

> -----Original Message-----
> From: Steven Haigh <netwiz@crc.id.au>
> Sent: 02 September 2019 09:32
> To: Paul Durrant <Paul.Durrant@citrix.com>
> Cc: Andrew Cooper <Andrew.Cooper3@citrix.com>; Andreas Kinzler <hfp@posteo.de>; xen-
> devel@lists.xenproject.org
> Subject: Re: Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)
> 
> On 2019-09-02 18:25, Steven Haigh wrote:
> > On 2019-09-02 18:20, Paul Durrant wrote:
> >>> -----Original Message-----
> >>> From: Steven Haigh <netwiz@crc.id.au>
> >>> Sent: 02 September 2019 09:09
> >>> To: Paul Durrant <Paul.Durrant@citrix.com>
> >>> Cc: Andreas Kinzler <hfp@posteo.de>; Andrew Cooper
> >>> <Andrew.Cooper3@citrix.com>; xen-
> >>> devel@lists.xenproject.org
> >>> Subject: Re: Windows HVM no longer boots with AMD Ryzen 3700X (and
> >>> 3900X)
> >>>
> >>> On 2019-09-02 18:04, Paul Durrant wrote:
> >>> >> -----Original Message-----
> >>> >> Further to the above, I did some experimentation. The following is a
> >>> >> list of attempted boot configurations and their outcomes:
> >>> >>
> >>> >> viridian=1
> >>> >> vcpus=4
> >>> >> STOPCODE: HAL MEMORY ALLOCATION
> >>> >>
> >>> >> viridian=0
> >>> >> vcpus=4
> >>> >> STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED
> >>> >>
> >>> >> viridian=0
> >>> >> vcpus=1
> >>> >> Boot OK - get to Windows Server 2016 login etc
> >>> >>
> >>> >
> >>> > And to complete the set, how about viridian=1 vcpus=1?
> >>>
> >>> Any vcpus value where viridian=1 is used creates a HAL MEMORY
> >>> ALLOCATION
> >>> stopcode when trying to boot Windows.
> >>
> >> Ok, so I guess that issue hits first and, only if you get beyond that
> >> do you hit the multiprocessor problem.
> >>
> >> The viridian option is not actually a boolean any more (that
> >> interpretation is just for compat) so it would be a good datapoint to
> >> know which of the enlightenments causes the change in behaviour. Could
> >> you try viridian=['base'] to see if that's sufficient to cause the
> >> problem? (I'm guessing it probably is but it would be good to know).
> >
> >
> > Hi Paul,
> >
> > I can confirm that viridian=['base'] crashes with the same HAL MEMORY
> > ALLOCATION stopcode - even on 1 vcpu.
> 
> Also, just wondering, we're using 8.2.0 of the Windows PV drivers on
> this VM.
> 
> Does this matter? Has there been any changes that would affect this in
> 8.2.1 or 8.2.2?
> 

From what you describe, I think this is happening way before any PV driver code is entered. I guess it would be prudent to make sure by trying it on a fresh VM (but didn't you say before that this happens when booting from the installation media?)

  Paul
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X (and 3900X)

[Steven Haigh]

On Mon, Sep 2, 2019 at 6:34 PM, Paul Durrant <Paul.Durrant@citrix.com> 
wrote:
>>  -----Original Message-----
>>  From: Steven Haigh <netwiz@crc.id.au>
>>  Sent: 02 September 2019 09:32
>>  To: Paul Durrant <Paul.Durrant@citrix.com>
>>  Cc: Andrew Cooper <Andrew.Cooper3@citrix.com>; Andreas Kinzler 
>> <hfp@posteo.de>; xen-
>>  devel@lists.xenproject.org
>>  Subject: Re: Windows HVM no longer boots with AMD Ryzen 3700X (and 
>> 3900X)
>> 
>>  On 2019-09-02 18:25, Steven Haigh wrote:
>>  > On 2019-09-02 18:20, Paul Durrant wrote:
>>  >>> -----Original Message-----
>>  >>> From: Steven Haigh <netwiz@crc.id.au>
>>  >>> Sent: 02 September 2019 09:09
>>  >>> To: Paul Durrant <Paul.Durrant@citrix.com>
>>  >>> Cc: Andreas Kinzler <hfp@posteo.de>; Andrew Cooper
>>  >>> <Andrew.Cooper3@citrix.com>; xen-
>>  >>> devel@lists.xenproject.org
>>  >>> Subject: Re: Windows HVM no longer boots with AMD Ryzen 3700X 
>> (and
>>  >>> 3900X)
>>  >>>
>>  >>> On 2019-09-02 18:04, Paul Durrant wrote:
>>  >>> >> -----Original Message-----
>>  >>> >> Further to the above, I did some experimentation. The 
>> following is a
>>  >>> >> list of attempted boot configurations and their outcomes:
>>  >>> >>
>>  >>> >> viridian=1
>>  >>> >> vcpus=4
>>  >>> >> STOPCODE: HAL MEMORY ALLOCATION
>>  >>> >>
>>  >>> >> viridian=0
>>  >>> >> vcpus=4
>>  >>> >> STOPCODE: MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED
>>  >>> >>
>>  >>> >> viridian=0
>>  >>> >> vcpus=1
>>  >>> >> Boot OK - get to Windows Server 2016 login etc
>>  >>> >>
>>  >>> >
>>  >>> > And to complete the set, how about viridian=1 vcpus=1?
>>  >>>
>>  >>> Any vcpus value where viridian=1 is used creates a HAL MEMORY
>>  >>> ALLOCATION
>>  >>> stopcode when trying to boot Windows.
>>  >>
>>  >> Ok, so I guess that issue hits first and, only if you get beyond 
>> that
>>  >> do you hit the multiprocessor problem.
>>  >>
>>  >> The viridian option is not actually a boolean any more (that
>>  >> interpretation is just for compat) so it would be a good 
>> datapoint to
>>  >> know which of the enlightenments causes the change in behaviour. 
>> Could
>>  >> you try viridian=['base'] to see if that's sufficient to cause 
>> the
>>  >> problem? (I'm guessing it probably is but it would be good to 
>> know).
>>  >
>>  >
>>  > Hi Paul,
>>  >
>>  > I can confirm that viridian=['base'] crashes with the same HAL 
>> MEMORY
>>  > ALLOCATION stopcode - even on 1 vcpu.
>> 
>>  Also, just wondering, we're using 8.2.0 of the Windows PV drivers on
>>  this VM.
>> 
>>  Does this matter? Has there been any changes that would affect this 
>> in
>>  8.2.1 or 8.2.2?
>> 
> 
> From what you describe, I think this is happening way before any PV 
> driver code is entered. I guess it would be prudent to make sure by 
> trying it on a fresh VM (but didn't you say before that this happens 
> when booting from the installation media?)


The original poster mentioned the problem with the install media. To be 
honest, I haven't tried that as yet.

My case / tests are from a working Windows Server 2016 install imaged 
on a different machine (that works and boots fine there).





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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andreas Kinzler]

On 20.08.2019 22:38, Andrew Cooper wrote:
> On 20/08/2019 21:36, Andreas Kinzler wrote:
>>>> Is it a known problem? Did someone test the new EPYCs?
>>> This looks familiar, and is still somewhere on my TODO list.
>> Do you already know the reason or is that still to investigate?
>>> Does booting with a single vCPU work?
> Hmm - perhaps its not the same issue then.  Either way, its firmly in
> the "still to investigate" phase.

Any update on the topic?

Regards Andreas

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Joost Heijne]

Hi,

I ran into this list looking for a solution for the exact same problem
as described here. I'm trying to migrate an existing Windows
Server 2016 image to a Ryzen 3900x and encountered the same
difficulties. I disabled the processor's SMT feature in the BIOS and
now it works perfectly.

Regards,

Joost Heijne

-----Original Message-----
From: Xen-devel [mailto:xen-devel-bounces@lists.xenproject.org] On Behalf Of Andreas Kinzler
Sent: Wednesday, October 2, 2019 12:05
To: Andrew Cooper; xen-devel@lists.xenproject.org; Paul.Durrant@citrix.com
Subject: Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

On 20.08.2019 22:38, Andrew Cooper wrote:
> On 20/08/2019 21:36, Andreas Kinzler wrote:
>>>> Is it a known problem? Did someone test the new EPYCs?
>>> This looks familiar, and is still somewhere on my TODO list.
>> Do you already know the reason or is that still to investigate?
>>> Does booting with a single vCPU work?
> Hmm - perhaps its not the same issue then.  Either way, its firmly in
> the "still to investigate" phase.

Any update on the topic?

Regards Andreas

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

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andrew Cooper]

On 31/01/2020 12:43, Joost Heijne wrote:
> Hi,
>
> I ran into this list looking for a solution for the exact same problem
> as described here. I'm trying to migrate an existing Windows
> Server 2016 image to a Ryzen 3900x and encountered the same
> difficulties. I disabled the processor's SMT feature in the BIOS and
> now it works perfectly.

That will work, but we also got a fix for this issue into Xen 4.13.

Are you saying that, despite that fix, you were still seeing issues?

~Andrew

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Joost Heijne]

Ah, I didn't check for that. I run Xen from the package distributed by Ubuntu, which is 4.9.2 currently.

Joost

-----Original Message-----
From: Andrew Cooper [mailto:andrew.cooper3@citrix.com] 
Sent: Friday, January 31, 2020 14:06
To: Joost Heijne; Andreas Kinzler; xen-devel@lists.xenproject.org; Paul.Durrant@citrix.com
Subject: Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

On 31/01/2020 12:43, Joost Heijne wrote:
> Hi,
>
> I ran into this list looking for a solution for the exact same problem
> as described here. I'm trying to migrate an existing Windows
> Server 2016 image to a Ryzen 3900x and encountered the same
> difficulties. I disabled the processor's SMT feature in the BIOS and
> now it works perfectly.

That will work, but we also got a fix for this issue into Xen 4.13.

Are you saying that, despite that fix, you were still seeing issues?

~Andrew
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Jan Beulich]

On 20.08.2019 22:36, Andreas Kinzler wrote:
> On 20.08.2019 20:12, Andrew Cooper wrote:
>>> Xen version 4.10.2. dom0 kernel 4.13.16. The BIOS version is unchanged
>>> from 2700X (working) to 3700X (crashing).
>> So you've done a Zen v1 => Zen v2 CPU upgrade and an existing system?
> 
> With "existing system" you mean the Windows installation? Yes, but it is 
> not relevant. The same BSODs happen if you boot the HVM with just the 
> iso installation medium and no disks.
> 
>>> Is it a known problem? Did someone test the new EPYCs?
>> This looks familiar, and is still somewhere on my TODO list.
> 
> Do you already know the reason or is that still to investigate?
> 
>> Does booting with a single vCPU work?
> 
> Number of vCPUs make no difference

Well, according to Steven it does, with viridian=0. Could you
re-check this?

While, according to AMD's processor specs page, the 3700X is just an
8-core chip, I wonder whether
https://lists.xenproject.org/archives/html/xen-devel/2019-09/msg01954.html
still affects this configuration as well. Could you give this a try in
at least the viridian=0 case? As to Linux, did you check that PVH
(or HVM, which you don't mention) guests actually start all their vCPU-s
successfully?

Jan

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andreas Kinzler]

[-- Attachment #1: Type: text/plain, Size: 511 bytes --]

On 23.09.2019 10:17, Jan Beulich wrote:
> While, according to AMD's processor specs page, the 3700X is just an
> 8-core chip, I wonder whether
> https://lists.xenproject.org/archives/html/xen-devel/2019-09/msg01954.html
> still affects this configuration as well. Could you give this a try in
> at least the viridian=0 case? As to Linux, did you check that PVH

As a first input for the Xen developers I used the tool from 
http://www.etallen.com/cpuid.html to dump complete cpuid information.

Regards Andreas

[-- Attachment #2: cpuid-3700x.tar.xz --]
[-- Type: application/octet-stream, Size: 7364 bytes --]

[-- Attachment #3: Type: text/plain, Size: 157 bytes --]

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Steven Haigh]

If this is helpful, I can probably provide the same from:
	* Ryzen 1700x
	* Ryzen 2700x
	* Ryzen 3900x

I'll leave it to those in the know as to what is useful or not...
Steven Haigh

📧 netwiz@crc.id.au     💻 https://www.crc.id.au
📞 +613 9001 6090       📱 +614 1293 5897


On Tue, Sep 24, 2019 at 11:56, Andreas Kinzler <hfp@posteo.de> wrote:
> On 23.09.2019 10:17, Jan Beulich wrote:
>> While, according to AMD's processor specs page, the 3700X is just an
>> 8-core chip, I wonder whether
>> https://lists.xenproject.org/archives/html/xen-devel/2019-09/msg01954.html
>> still affects this configuration as well. Could you give this a try 
>> in
>> at least the viridian=0 case? As to Linux, did you check that PVH
> 
> As a first input for the Xen developers I used the tool from 
> http://www.etallen.com/cpuid.html to dump complete cpuid information.
> 
> Regards Andreas
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xenproject.org
> https://lists.xenproject.org/mailman/listinfo/xen-devel



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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Andreas Kinzler]

[-- Attachment #1: Type: text/plain, Size: 947 bytes --]

On 23.09.2019 10:17, Jan Beulich wrote:
>>> Does booting with a single vCPU work?
>> Number of vCPUs make no difference
> Well, according to Steven it does, with viridian=0. Could you
> re-check this?

I can confirm that viridian=0 AND vcpus=1 makes the system bootable 
(with long delay though)

> at least the viridian=0 case? As to Linux, did you check that PVH
> (or HVM, which you don't mention) guests actually start all their vCPU-s
> successfully?

I just tried PVH and HVM with 8 vcpus. Everything works, tested with 
make -j9 on a kernel tree.

 > 8-core chip, I wonder whether
 > 
https://lists.xenproject.org/archives/html/xen-devel/2019-09/msg01954.html
 > still affects this configuration as well. Could you give this a try in

Does it still make sense to try the patch given the cpuid I posted?
Also I have an AMD 7302P in my lab (cpuid dump attached). No change in 
behavior from 3700X - cpuid is nearly the same.

Regards Andreas

[-- Attachment #2: cpuid-amd7302p.tar.xz --]
[-- Type: application/octet-stream, Size: 8176 bytes --]

[-- Attachment #3: Type: text/plain, Size: 157 bytes --]

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

Re: [Xen-devel] Windows HVM no longer boots with AMD Ryzen 3700X

[Jan Beulich]

On 24.09.2019 18:58, Andreas Kinzler wrote:
> Does it still make sense to try the patch given the cpuid I posted?

The patch shouldn't make any difference there, afaict.

Jan

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