Wondering about cirris and stdvga

Wondering about cirris and stdvga

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 2042 bytes --]

Sending again, this time, with Anthony's and xen-devel address spelled
right. Sorry!! :-(
-------------------------------------------
Hello to you, various pseudo-random people,

It's not my field of expertise, so bear with me, at least a little bit
(and, Konrad, you help me, or there will be consequences! :-D)

So, I and Konrad recently discovered --while testing the about to be
released Fedora 25 as a Xen guest-- that the Cirrus emulated graphic
card that we consume from QEMU for HVM guests is broken on Wayland.

We just discovered it because Fedora 25 uses Wayland by default, but it
appears not to be something new:

https://bugzilla.redhat.com/show_bug.cgi?id=1227770

And at least from what we see in that bugreport, not much has happened
so far.

Using "vga='stdvga'" in the config file, or even "vga='qxl'" make
things work again. Disabling Wayland in the guest also works (i.e., if
not using Wayland, Cirrus is ok). And that's what made us think that
it's probably a Wayland issue.

I've tried the same on KVM, and the situation is identical
(Cirrus+Wayland=breaks, whatever-else+Wayland=works,
Cirrus+Xorg=works).

I've also read around that these days, e.g., stdvga is at least as good
as cirrus, performance wise, that cirrus is broken and impossible to
fix (because it is the hardware that it's emulating that was broken),
that stdvga enables better screen resolution in guests, etc.

I'm not sure about these claims, in particular the performance one, is
probably pretty hard to verify. And as I said, it's not my field.

Still I thought it could be worthwhile to at least bring this up:
should we start to consider changing the default from cirrus to stdvga
(or something else)?

Thanks for your time and Regards,
Dario
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[-- Attachment #2: Type: text/plain, Size: 127 bytes --]

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

Re: Wondering about cirris and stdvga

[Pasi Kärkkäinen]

On Fri, Nov 18, 2016 at 07:04:15PM +0100, Dario Faggioli wrote:
> Sending again, this time, with Anthony's and xen-devel address spelled
> right. Sorry!! :-(
> -------------------------------------------
> Hello to you, various pseudo-random people,
> 
> It's not my field of expertise, so bear with me, at least a little bit
> (and, Konrad, you help me, or there will be consequences! :-D)
> 
> So, I and Konrad recently discovered --while testing the about to be
> released Fedora 25 as a Xen guest-- that the Cirrus emulated graphic
> card that we consume from QEMU for HVM guests is broken on Wayland.
> 
> We just discovered it because Fedora 25 uses Wayland by default, but it
> appears not to be something new:
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> 
> And at least from what we see in that bugreport, not much has happened
> so far.
> 
> Using "vga='stdvga'" in the config file, or even "vga='qxl'" make
> things work again. Disabling Wayland in the guest also works (i.e., if
> not using Wayland, Cirrus is ok). And that's what made us think that
> it's probably a Wayland issue.
> 
> I've tried the same on KVM, and the situation is identical
> (Cirrus+Wayland=breaks, whatever-else+Wayland=works,
> Cirrus+Xorg=works).
> 
> I've also read around that these days, e.g., stdvga is at least as good
> as cirrus, performance wise, that cirrus is broken and impossible to
> fix (because it is the hardware that it's emulating that was broken),
> that stdvga enables better screen resolution in guests, etc.
> 
> I'm not sure about these claims, in particular the performance one, is
> probably pretty hard to verify. And as I said, it's not my field.
> 
> Still I thought it could be worthwhile to at least bring this up:
> should we start to consider changing the default from cirrus to stdvga
> (or something else)?
> 

There's multiple things here..

1) Yes, +1, let's change the Xen HVM default to "stdvga".

2) It'd good to create an upstream Wayland bugreport and investigate more about why cirrus is broken with Wayland.

3) It'd be good to have Xen HVM with "qxl" tested by OSStest aswell..


> Thanks for your time and Regards,
> Dario


Thanks,

-- Pasi


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

Re: Wondering about cirris and stdvga

[Paul Durrant]

> -----Original Message-----
> From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> Sent: 19 November 2016 10:57
> To: Dario Faggioli <dario.faggioli@citrix.com>
> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>; Paul Durrant
> <Paul.Durrant@citrix.com>; Anthony Perard <anthony.perard@citrix.com>;
> xen-devel <xen-devel@lists.xenproject.org>; Stefano Stabellini
> <sstabellini@kernel.org>; Ian Jackson <Ian.Jackson@citrix.com>; Roger Pau
> Monne <roger.pau@citrix.com>
> Subject: Re: [Xen-devel] Wondering about cirris and stdvga
> 
> On Fri, Nov 18, 2016 at 07:04:15PM +0100, Dario Faggioli wrote:
> > Sending again, this time, with Anthony's and xen-devel address spelled
> > right. Sorry!! :-(
> > -------------------------------------------
> > Hello to you, various pseudo-random people,
> >
> > It's not my field of expertise, so bear with me, at least a little bit
> > (and, Konrad, you help me, or there will be consequences! :-D)
> >
> > So, I and Konrad recently discovered --while testing the about to be
> > released Fedora 25 as a Xen guest-- that the Cirrus emulated graphic
> > card that we consume from QEMU for HVM guests is broken on Wayland.
> >
> > We just discovered it because Fedora 25 uses Wayland by default, but it
> > appears not to be something new:
> >
> > https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> >
> > And at least from what we see in that bugreport, not much has happened
> > so far.
> >
> > Using "vga='stdvga'" in the config file, or even "vga='qxl'" make
> > things work again. Disabling Wayland in the guest also works (i.e., if
> > not using Wayland, Cirrus is ok). And that's what made us think that
> > it's probably a Wayland issue.
> >
> > I've tried the same on KVM, and the situation is identical
> > (Cirrus+Wayland=breaks, whatever-else+Wayland=works,
> > Cirrus+Xorg=works).
> >
> > I've also read around that these days, e.g., stdvga is at least as good
> > as cirrus, performance wise, that cirrus is broken and impossible to
> > fix (because it is the hardware that it's emulating that was broken),
> > that stdvga enables better screen resolution in guests, etc.
> >
> > I'm not sure about these claims, in particular the performance one, is
> > probably pretty hard to verify. And as I said, it's not my field.
> >
> > Still I thought it could be worthwhile to at least bring this up:
> > should we start to consider changing the default from cirrus to stdvga
> > (or something else)?
> >
> 
> There's multiple things here..
> 
> 1) Yes, +1, let's change the Xen HVM default to "stdvga".
> 

In general std-vga also gets a +1 from me, but I have recently found that a Windows Server 2008 guest does not boot (ar at least the display freezes on boot) when using std-vga with QEMU trad but everything is fine with Cirrus... so probably worth making the default dependent on which QEMU is being used. Never had any issues with std-vga on upstream QEMU.

  Paul

> 2) It'd good to create an upstream Wayland bugreport and investigate more
> about why cirrus is broken with Wayland.
> 
> 3) It'd be good to have Xen HVM with "qxl" tested by OSStest aswell..
> 
> 
> > Thanks for your time and Regards,
> > Dario
> 
> 
> Thanks,
> 
> -- Pasi


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

Re: Wondering about cirris and stdvga

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 1524 bytes --]

On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> There's multiple things here..
> 
> 1) Yes, +1, let's change the Xen HVM default to "stdvga".
> 
:-)

> 2) It'd good to create an upstream Wayland bugreport and investigate
> more about why cirrus is broken with Wayland.
> 
Sure, I can do that.

> 3) It'd be good to have Xen HVM with "qxl" tested by OSStest aswell..
> 
I've had a look at that too, and I'm happy to continue trying, although
I don't know much on the subject, so any help is appreciated.

I've found out that, if you install spice dev libraries, (upstream)
QEMU gets built with qxl enabled, and I can specity "vga=qxl" in guest
config, and that works.

I see the boot, I see X / Wayland, etc.

What I haven't understood yet is how to actually connect with a "spice
client" (or whatever the correct terminology is) and get the
acceleration and all the benefits. I've tried just connecting with
spicy or virt-viewer, but couldn't get that to work.

I think I've got to install/enable stuff inside the guest too, but I'm
not sure what yet.

I'll try to fetch docs online, and fish back one of those old thread we
had here about spice support, to see if there is some more info.

Thanks and Regards,
Dario
--
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[-- Attachment #2: Type: text/plain, Size: 127 bytes --]

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

Re: Wondering about cirris and stdvga

[Stefano Stabellini]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 3230 bytes --]

On Mon, 21 Nov 2016, Paul Durrant wrote:
> > -----Original Message-----
> > From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> > Sent: 19 November 2016 10:57
> > To: Dario Faggioli <dario.faggioli@citrix.com>
> > Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>; Paul Durrant
> > <Paul.Durrant@citrix.com>; Anthony Perard <anthony.perard@citrix.com>;
> > xen-devel <xen-devel@lists.xenproject.org>; Stefano Stabellini
> > <sstabellini@kernel.org>; Ian Jackson <Ian.Jackson@citrix.com>; Roger Pau
> > Monne <roger.pau@citrix.com>
> > Subject: Re: [Xen-devel] Wondering about cirris and stdvga
> > 
> > On Fri, Nov 18, 2016 at 07:04:15PM +0100, Dario Faggioli wrote:
> > > Sending again, this time, with Anthony's and xen-devel address spelled
> > > right. Sorry!! :-(
> > > -------------------------------------------
> > > Hello to you, various pseudo-random people,
> > >
> > > It's not my field of expertise, so bear with me, at least a little bit
> > > (and, Konrad, you help me, or there will be consequences! :-D)
> > >
> > > So, I and Konrad recently discovered --while testing the about to be
> > > released Fedora 25 as a Xen guest-- that the Cirrus emulated graphic
> > > card that we consume from QEMU for HVM guests is broken on Wayland.
> > >
> > > We just discovered it because Fedora 25 uses Wayland by default, but it
> > > appears not to be something new:
> > >
> > > https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> > >
> > > And at least from what we see in that bugreport, not much has happened
> > > so far.
> > >
> > > Using "vga='stdvga'" in the config file, or even "vga='qxl'" make
> > > things work again. Disabling Wayland in the guest also works (i.e., if
> > > not using Wayland, Cirrus is ok). And that's what made us think that
> > > it's probably a Wayland issue.
> > >
> > > I've tried the same on KVM, and the situation is identical
> > > (Cirrus+Wayland=breaks, whatever-else+Wayland=works,
> > > Cirrus+Xorg=works).
> > >
> > > I've also read around that these days, e.g., stdvga is at least as good
> > > as cirrus, performance wise, that cirrus is broken and impossible to
> > > fix (because it is the hardware that it's emulating that was broken),
> > > that stdvga enables better screen resolution in guests, etc.
> > >
> > > I'm not sure about these claims, in particular the performance one, is
> > > probably pretty hard to verify. And as I said, it's not my field.
> > >
> > > Still I thought it could be worthwhile to at least bring this up:
> > > should we start to consider changing the default from cirrus to stdvga
> > > (or something else)?
> > >
> > 
> > There's multiple things here..
> > 
> > 1) Yes, +1, let's change the Xen HVM default to "stdvga".
> > 
> 
> In general std-vga also gets a +1 from me, but I have recently found that a Windows Server 2008 guest does not boot (ar at least the display freezes on boot) when using std-vga with QEMU trad but everything is fine with Cirrus... so probably worth making the default dependent on which QEMU is being used. Never had any issues with std-vga on upstream QEMU.

In all fairness, I think it's worth mentioning that the original reason
for defaulting to Cirrus was that Windows XP wasn't able to boot on
stdvga either.

[-- Attachment #2: Type: text/plain, Size: 127 bytes --]

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

Re: Wondering about cirris and stdvga

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 2015 bytes --]

On Mon, 2016-11-21 at 10:04 +0100, Dario Faggioli wrote:
> On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> > 2) It'd good to create an upstream Wayland bugreport and
> > investigate
> > more about why cirrus is broken with Wayland.
> > 
> Sure, I can do that.
> 
An update.

The discussion here has gone on a bit:
https://bugzilla.redhat.com/show_bug.cgi?id=1227770

The conclusion seems to be that:
"cirrus (virtual) hardware is simply to old to run wayland."

And so this is (and will very likely remain) a 'WONTFIX' for cirrus, at
least on Fedora.

I've also opened a thread on wayland-devel mailing list:
https://lists.freedesktop.org/archives/wayland-devel/2016-November/0318
56.html

There, I learned that Wayland is not the component to blame, as Wayland
is the protocol. So, in our case, the 'bug' is most likely in
gnome-shell / Mutter.

That's not a good thing, though. In fact, just to cite a few sentences
from the thread:

"Packed 24bpp is going to be pain, not least because I don't know of
any clients which render in packed-24"

"The 24bpp paths in pretty much everything are also badly untested, so
that's asking for trouble."

"you will need to test and fix every single Wayland compositor out
there."

"I really think you'd be far far better off trying to figure out how to
move off the legacy Cirrus emulation as soon as you can."

So, we can try seeing if I manage to get some logs out of Mutter to
figure out the actual bug more precisely _but_, considering all that
people have said both here and in the other forums, I think it would be
better to spend that time figuring out how to switch (and document this
for 4.8 and previous version, of course).

Regards,
Dario

-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[-- Attachment #2: Type: text/plain, Size: 127 bytes --]

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

Re: Wondering about cirris and stdvga

[Konrad Rzeszutek Wilk]

On Fri, Nov 25, 2016 at 07:17:31PM +0100, Dario Faggioli wrote:
> On Mon, 2016-11-21 at 10:04 +0100, Dario Faggioli wrote:
> > On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> > > 2) It'd good to create an upstream Wayland bugreport and
> > > investigate
> > > more about why cirrus is broken with Wayland.
> > > 
> > Sure, I can do that.
> > 
> An update.
> 
> The discussion here has gone on a bit:
> https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> 
> The conclusion seems to be that:
> "cirrus (virtual) hardware is simply to old to run wayland."
> 
> And so this is (and will very likely remain) a 'WONTFIX' for cirrus, at
> least on Fedora.
> 
> I've also opened a thread on wayland-devel mailing list:
> https://lists.freedesktop.org/archives/wayland-devel/2016-November/0318
> 56.html
> 
> There, I learned that Wayland is not the component to blame, as Wayland
> is the protocol. So, in our case, the 'bug' is most likely in
> gnome-shell / Mutter.
> 
> That's not a good thing, though. In fact, just to cite a few sentences
> from the thread:
> 
> "Packed 24bpp is going to be pain, not least because I don't know of
> any clients which render in packed-24"
> 
> "The 24bpp paths in pretty much everything are also badly untested, so
> that's asking for trouble."
> 
> "you will need to test and fix every single Wayland compositor out
> there."
> 
> "I really think you'd be far far better off trying to figure out how to
> move off the legacy Cirrus emulation as soon as you can."
> 
> So, we can try seeing if I manage to get some logs out of Mutter to
> figure out the actual bug more precisely _but_, considering all that
> people have said both here and in the other forums, I think it would be
> better to spend that time figuring out how to switch (and document this
> for 4.8 and previous version, of course).

+1

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

Re: Wondering about cirris and stdvga

[Konrad Rzeszutek Wilk]

On Fri, Nov 25, 2016 at 07:17:31PM +0100, Dario Faggioli wrote:
> On Mon, 2016-11-21 at 10:04 +0100, Dario Faggioli wrote:
> > On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> > > 2) It'd good to create an upstream Wayland bugreport and
> > > investigate
> > > more about why cirrus is broken with Wayland.
> > > 
> > Sure, I can do that.
> > 
> An update.
> 
> The discussion here has gone on a bit:
> https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> 
> The conclusion seems to be that:
> "cirrus (virtual) hardware is simply to old to run wayland."
> 
> And so this is (and will very likely remain) a 'WONTFIX' for cirrus, at
> least on Fedora.
> 
> I've also opened a thread on wayland-devel mailing list:
> https://lists.freedesktop.org/archives/wayland-devel/2016-November/0318
> 56.html
> 
> There, I learned that Wayland is not the component to blame, as Wayland
> is the protocol. So, in our case, the 'bug' is most likely in
> gnome-shell / Mutter.
> 
> That's not a good thing, though. In fact, just to cite a few sentences
> from the thread:
> 
> "Packed 24bpp is going to be pain, not least because I don't know of
> any clients which render in packed-24"
> 
> "The 24bpp paths in pretty much everything are also badly untested, so
> that's asking for trouble."
> 
> "you will need to test and fix every single Wayland compositor out
> there."
> 
> "I really think you'd be far far better off trying to figure out how to
> move off the legacy Cirrus emulation as soon as you can."
> 
> So, we can try seeing if I manage to get some logs out of Mutter to
> figure out the actual bug more precisely _but_, considering all that
> people have said both here and in the other forums, I think it would be
> better to spend that time figuring out how to switch (and document this
> for 4.8 and previous version, of course).


Yes. Also as there does not seem to be any supported OS that 
_needs_ the old Cirrus OS to boot and function.


Here is what I am thinking to propose for Xen 4.9:

diff --git a/docs/man/xl.cfg.pod.5.in b/docs/man/xl.cfg.pod.5.in
index 52802d5..f77c7ee 100644
--- a/docs/man/xl.cfg.pod.5.in
+++ b/docs/man/xl.cfg.pod.5.in
@@ -1669,16 +1669,16 @@ If B<videoram> is set less than 128MB, an error will be triggered.
 =item B<stdvga=BOOLEAN>
 
 Select a standard VGA card with VBE (VESA BIOS Extensions) as the
-emulated graphics device. The default is false (0) which means to emulate
-a Cirrus Logic GD5446 VGA card. If your guest supports VBE 2.0 or
-later (e.g. Windows XP onwards) then you should enable this.
+emulated graphics device. The default is true (1) which means to emulate
+a standard VGA card. If your guest does not support VBE 2.0 or is older than
+Windows XP then should should disable this and use Cirrus Logic GD5446 VGA card.
 stdvga supports more video ram and bigger resolutions than Cirrus.
 This option is deprecated, use vga="stdvga" instead.
 
 =item B<vga="STRING">
 
 Selects the emulated video card (none|stdvga|cirrus|qxl).
-The default is cirrus.
+The default is stdvga.
 
 In general, QXL should work with the Spice remote display protocol
 for acceleration, and QXL driver is necessary in guest in this case.
diff --git a/tools/libxl/libxl_create.c b/tools/libxl/libxl_create.c
index 25389e1..ded5ebe 100644
--- a/tools/libxl/libxl_create.c
+++ b/tools/libxl/libxl_create.c
@@ -225,7 +225,7 @@ int libxl__domain_build_info_setdefault(libxl__gc *gc,
             if (b_info->device_model_version == LIBXL_DEVICE_MODEL_VERSION_NONE)
                 b_info->u.hvm.vga.kind = LIBXL_VGA_INTERFACE_TYPE_NONE;
             else
-                b_info->u.hvm.vga.kind = LIBXL_VGA_INTERFACE_TYPE_CIRRUS;
+                b_info->u.hvm.vga.kind = LIBXL_VGA_INTERFACE_TYPE_STD;
         }
 
         if (!b_info->u.hvm.hdtype)
> 
> Regards,
> Dario
> 
> -- 
> <<This happens because I choose it to happen!>> (Raistlin Majere)
> -----------------------------------------------------------------
> Dario Faggioli, Ph.D, http://about.me/dario.faggioli
> Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)



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

Re: Wondering about cirris and stdvga

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 1266 bytes --]

On Fri, 2017-03-17 at 10:19 -0400, Konrad Rzeszutek Wilk wrote:
> --- a/tools/libxl/libxl_create.c
> +++ b/tools/libxl/libxl_create.c
> @@ -225,7 +225,7 @@ int libxl__domain_build_info_setdefault(libxl__gc
> *gc,
>              if (b_info->device_model_version ==
> LIBXL_DEVICE_MODEL_VERSION_NONE)
>                  b_info->u.hvm.vga.kind =
> LIBXL_VGA_INTERFACE_TYPE_NONE;
>              else
> -                b_info->u.hvm.vga.kind =
> LIBXL_VGA_INTERFACE_TYPE_CIRRUS;
> +                b_info->u.hvm.vga.kind =
> LIBXL_VGA_INTERFACE_TYPE_STD;
>          }
>  
For what is worth, I'd agree.

If I remember correct, though, there was the concern of what happens
when migrating a domain from Xen 4.8 --which would have been build with
cirrus by default-- to Xen 4.9 --which, with this, would build the
receiver domain with stdvga... Or is this not an issue / does not
happen?

Regards,
Dario
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[-- Attachment #2: Type: text/plain, Size: 127 bytes --]

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

Re: Wondering about cirris and stdvga

[Konrad Rzeszutek Wilk]

On March 17, 2017 11:51:03 AM EDT, Dario Faggioli <dario.faggioli@citrix.com> wrote:
>On Fri, 2017-03-17 at 10:19 -0400, Konrad Rzeszutek Wilk wrote:
>> --- a/tools/libxl/libxl_create.c
>> +++ b/tools/libxl/libxl_create.c
>> @@ -225,7 +225,7 @@ int libxl__domain_build_info_setdefault(libxl__gc
>> *gc,
>>              if (b_info->device_model_version ==
>> LIBXL_DEVICE_MODEL_VERSION_NONE)
>>                  b_info->u.hvm.vga.kind =
>> LIBXL_VGA_INTERFACE_TYPE_NONE;
>>              else
>> -                b_info->u.hvm.vga.kind =
>> LIBXL_VGA_INTERFACE_TYPE_CIRRUS;
>> +                b_info->u.hvm.vga.kind =
>> LIBXL_VGA_INTERFACE_TYPE_STD;
>>          }
>>  
>For what is worth, I'd agree.
>
>If I remember correct, though, there was the concern of what happens
>when migrating a domain from Xen 4.8 --which would have been build with
>cirrus by default-- to Xen 4.9 --which, with this, would build the
>receiver domain with stdvga... Or is this not an issue / does not
>happen?
>


No clue, let me try it out.

>Regards,
>Dario


Thanks!

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

Re: Wondering about cirris and stdvga

[Roger Pau Monné]

On Fri, Mar 17, 2017 at 10:19:47AM -0400, Konrad Rzeszutek Wilk wrote:
> On Fri, Nov 25, 2016 at 07:17:31PM +0100, Dario Faggioli wrote:
> > On Mon, 2016-11-21 at 10:04 +0100, Dario Faggioli wrote:
> > > On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> > > > 2) It'd good to create an upstream Wayland bugreport and
> > > > investigate
> > > > more about why cirrus is broken with Wayland.
> > > > 
> > > Sure, I can do that.
> > > 
> > An update.
> > 
> > The discussion here has gone on a bit:
> > https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> > 
> > The conclusion seems to be that:
> > "cirrus (virtual) hardware is simply to old to run wayland."
> > 
> > And so this is (and will very likely remain) a 'WONTFIX' for cirrus, at
> > least on Fedora.
> > 
> > I've also opened a thread on wayland-devel mailing list:
> > https://lists.freedesktop.org/archives/wayland-devel/2016-November/0318
> > 56.html
> > 
> > There, I learned that Wayland is not the component to blame, as Wayland
> > is the protocol. So, in our case, the 'bug' is most likely in
> > gnome-shell / Mutter.
> > 
> > That's not a good thing, though. In fact, just to cite a few sentences
> > from the thread:
> > 
> > "Packed 24bpp is going to be pain, not least because I don't know of
> > any clients which render in packed-24"
> > 
> > "The 24bpp paths in pretty much everything are also badly untested, so
> > that's asking for trouble."
> > 
> > "you will need to test and fix every single Wayland compositor out
> > there."
> > 
> > "I really think you'd be far far better off trying to figure out how to
> > move off the legacy Cirrus emulation as soon as you can."
> > 
> > So, we can try seeing if I manage to get some logs out of Mutter to
> > figure out the actual bug more precisely _but_, considering all that
> > people have said both here and in the other forums, I think it would be
> > better to spend that time figuring out how to switch (and document this
> > for 4.8 and previous version, of course).
> 
> 
> Yes. Also as there does not seem to be any supported OS that 
> _needs_ the old Cirrus OS to boot and function.

Not that I oppose to change to stdvga, but what would happen to Windows VMs
that suddenly change from cirrus to stdvga, is that going to trigger the
license invalidation stuff? AFAIK this happens when you change hardware, but
maybe a VGA change doesn't trigger that because it's common for people to
upgrade VGA cards?

Roger.

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

Re: Wondering about cirris and stdvga

[Paul Durrant]

> -----Original Message-----
> From: Roger Pau Monne
> Sent: 20 March 2017 14:14
> To: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> Cc: Dario Faggioli <dario.faggioli@citrix.com>; Ian Jackson
> <Ian.Jackson@citrix.com>; Pasi Kärkkäinen <pasik@iki.fi>; Stefano Stabellini
> <sstabellini@kernel.org>; Paul Durrant <Paul.Durrant@citrix.com>; Anthony
> Perard <anthony.perard@citrix.com>; xen-devel <xen-
> devel@lists.xenproject.org>; Roger Pau Monne <roger.pau@citrix.com>;
> ajax <ajax@redhat.com>
> Subject: Re: [Xen-devel] Wondering about cirris and stdvga
> 
> On Fri, Mar 17, 2017 at 10:19:47AM -0400, Konrad Rzeszutek Wilk wrote:
> > On Fri, Nov 25, 2016 at 07:17:31PM +0100, Dario Faggioli wrote:
> > > On Mon, 2016-11-21 at 10:04 +0100, Dario Faggioli wrote:
> > > > On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> > > > > 2) It'd good to create an upstream Wayland bugreport and
> > > > > investigate
> > > > > more about why cirrus is broken with Wayland.
> > > > >
> > > > Sure, I can do that.
> > > >
> > > An update.
> > >
> > > The discussion here has gone on a bit:
> > > https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> > >
> > > The conclusion seems to be that:
> > > "cirrus (virtual) hardware is simply to old to run wayland."
> > >
> > > And so this is (and will very likely remain) a 'WONTFIX' for cirrus, at
> > > least on Fedora.
> > >
> > > I've also opened a thread on wayland-devel mailing list:
> > > https://lists.freedesktop.org/archives/wayland-devel/2016-
> November/0318
> > > 56.html
> > >
> > > There, I learned that Wayland is not the component to blame, as Wayland
> > > is the protocol. So, in our case, the 'bug' is most likely in
> > > gnome-shell / Mutter.
> > >
> > > That's not a good thing, though. In fact, just to cite a few sentences
> > > from the thread:
> > >
> > > "Packed 24bpp is going to be pain, not least because I don't know of
> > > any clients which render in packed-24"
> > >
> > > "The 24bpp paths in pretty much everything are also badly untested, so
> > > that's asking for trouble."
> > >
> > > "you will need to test and fix every single Wayland compositor out
> > > there."
> > >
> > > "I really think you'd be far far better off trying to figure out how to
> > > move off the legacy Cirrus emulation as soon as you can."
> > >
> > > So, we can try seeing if I manage to get some logs out of Mutter to
> > > figure out the actual bug more precisely _but_, considering all that
> > > people have said both here and in the other forums, I think it would be
> > > better to spend that time figuring out how to switch (and document this
> > > for 4.8 and previous version, of course).
> >
> >
> > Yes. Also as there does not seem to be any supported OS that
> > _needs_ the old Cirrus OS to boot and function.
> 
> Not that I oppose to change to stdvga, but what would happen to Windows
> VMs
> that suddenly change from cirrus to stdvga, is that going to trigger the
> license invalidation stuff? AFAIK this happens when you change hardware,
> but
> maybe a VGA change doesn't trigger that because it's common for people to
> upgrade VGA cards?
> 

Changing anything that Windows considers part of the core system will invalidate a license but, as you say, the graphics device may not be core. There is another issue (which I just verified myself) which as that at least some versions of Windows (Server 2008 in my case) won't boot when using stdvga with qemu trad.

  Paul

> Roger.

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

Re: Wondering about cirris and stdvga

[Roger Pau Monne]

On Mon, Mar 20, 2017 at 02:21:50PM +0000, Paul Durrant wrote:
> > -----Original Message-----
> > From: Roger Pau Monne
> > Sent: 20 March 2017 14:14
> > To: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> > Cc: Dario Faggioli <dario.faggioli@citrix.com>; Ian Jackson
> > <Ian.Jackson@citrix.com>; Pasi Kärkkäinen <pasik@iki.fi>; Stefano Stabellini
> > <sstabellini@kernel.org>; Paul Durrant <Paul.Durrant@citrix.com>; Anthony
> > Perard <anthony.perard@citrix.com>; xen-devel <xen-
> > devel@lists.xenproject.org>; Roger Pau Monne <roger.pau@citrix.com>;
> > ajax <ajax@redhat.com>
> > Subject: Re: [Xen-devel] Wondering about cirris and stdvga
> > 
> > On Fri, Mar 17, 2017 at 10:19:47AM -0400, Konrad Rzeszutek Wilk wrote:
> > > On Fri, Nov 25, 2016 at 07:17:31PM +0100, Dario Faggioli wrote:
> > > > On Mon, 2016-11-21 at 10:04 +0100, Dario Faggioli wrote:
> > > > > On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> > > > > > 2) It'd good to create an upstream Wayland bugreport and
> > > > > > investigate
> > > > > > more about why cirrus is broken with Wayland.
> > > > > >
> > > > > Sure, I can do that.
> > > > >
> > > > An update.
> > > >
> > > > The discussion here has gone on a bit:
> > > > https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> > > >
> > > > The conclusion seems to be that:
> > > > "cirrus (virtual) hardware is simply to old to run wayland."
> > > >
> > > > And so this is (and will very likely remain) a 'WONTFIX' for cirrus, at
> > > > least on Fedora.
> > > >
> > > > I've also opened a thread on wayland-devel mailing list:
> > > > https://lists.freedesktop.org/archives/wayland-devel/2016-
> > November/0318
> > > > 56.html
> > > >
> > > > There, I learned that Wayland is not the component to blame, as Wayland
> > > > is the protocol. So, in our case, the 'bug' is most likely in
> > > > gnome-shell / Mutter.
> > > >
> > > > That's not a good thing, though. In fact, just to cite a few sentences
> > > > from the thread:
> > > >
> > > > "Packed 24bpp is going to be pain, not least because I don't know of
> > > > any clients which render in packed-24"
> > > >
> > > > "The 24bpp paths in pretty much everything are also badly untested, so
> > > > that's asking for trouble."
> > > >
> > > > "you will need to test and fix every single Wayland compositor out
> > > > there."
> > > >
> > > > "I really think you'd be far far better off trying to figure out how to
> > > > move off the legacy Cirrus emulation as soon as you can."
> > > >
> > > > So, we can try seeing if I manage to get some logs out of Mutter to
> > > > figure out the actual bug more precisely _but_, considering all that
> > > > people have said both here and in the other forums, I think it would be
> > > > better to spend that time figuring out how to switch (and document this
> > > > for 4.8 and previous version, of course).
> > >
> > >
> > > Yes. Also as there does not seem to be any supported OS that
> > > _needs_ the old Cirrus OS to boot and function.
> > 
> > Not that I oppose to change to stdvga, but what would happen to Windows
> > VMs
> > that suddenly change from cirrus to stdvga, is that going to trigger the
> > license invalidation stuff? AFAIK this happens when you change hardware,
> > but
> > maybe a VGA change doesn't trigger that because it's common for people to
> > upgrade VGA cards?
> > 
> 
> Changing anything that Windows considers part of the core system will invalidate a license but, as you say, the graphics device may not be core. There is another issue (which I just verified myself) which as that at least some versions of Windows (Server 2008 in my case) won't boot when using stdvga with qemu trad.

IMHO, I would leave qemu-trad alone and just change the default gfx card for
qemu-xen. I don't see much point in touching anything in qemu-trad.

Roger.

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

Re: Wondering about cirris and stdvga

[Konrad Rzeszutek Wilk]

On Mon, Mar 20, 2017 at 02:25:16PM +0000, Roger Pau Monne wrote:
> On Mon, Mar 20, 2017 at 02:21:50PM +0000, Paul Durrant wrote:
> > > -----Original Message-----
> > > From: Roger Pau Monne
> > > Sent: 20 March 2017 14:14
> > > To: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> > > Cc: Dario Faggioli <dario.faggioli@citrix.com>; Ian Jackson
> > > <Ian.Jackson@citrix.com>; Pasi Kärkkäinen <pasik@iki.fi>; Stefano Stabellini
> > > <sstabellini@kernel.org>; Paul Durrant <Paul.Durrant@citrix.com>; Anthony
> > > Perard <anthony.perard@citrix.com>; xen-devel <xen-
> > > devel@lists.xenproject.org>; Roger Pau Monne <roger.pau@citrix.com>;
> > > ajax <ajax@redhat.com>
> > > Subject: Re: [Xen-devel] Wondering about cirris and stdvga
> > > 
> > > On Fri, Mar 17, 2017 at 10:19:47AM -0400, Konrad Rzeszutek Wilk wrote:
> > > > On Fri, Nov 25, 2016 at 07:17:31PM +0100, Dario Faggioli wrote:
> > > > > On Mon, 2016-11-21 at 10:04 +0100, Dario Faggioli wrote:
> > > > > > On Sat, 2016-11-19 at 12:56 +0200, Pasi Kärkkäinen wrote:
> > > > > > > 2) It'd good to create an upstream Wayland bugreport and
> > > > > > > investigate
> > > > > > > more about why cirrus is broken with Wayland.
> > > > > > >
> > > > > > Sure, I can do that.
> > > > > >
> > > > > An update.
> > > > >
> > > > > The discussion here has gone on a bit:
> > > > > https://bugzilla.redhat.com/show_bug.cgi?id=1227770
> > > > >
> > > > > The conclusion seems to be that:
> > > > > "cirrus (virtual) hardware is simply to old to run wayland."
> > > > >
> > > > > And so this is (and will very likely remain) a 'WONTFIX' for cirrus, at
> > > > > least on Fedora.
> > > > >
> > > > > I've also opened a thread on wayland-devel mailing list:
> > > > > https://lists.freedesktop.org/archives/wayland-devel/2016-
> > > November/0318
> > > > > 56.html
> > > > >
> > > > > There, I learned that Wayland is not the component to blame, as Wayland
> > > > > is the protocol. So, in our case, the 'bug' is most likely in
> > > > > gnome-shell / Mutter.
> > > > >
> > > > > That's not a good thing, though. In fact, just to cite a few sentences
> > > > > from the thread:
> > > > >
> > > > > "Packed 24bpp is going to be pain, not least because I don't know of
> > > > > any clients which render in packed-24"
> > > > >
> > > > > "The 24bpp paths in pretty much everything are also badly untested, so
> > > > > that's asking for trouble."
> > > > >
> > > > > "you will need to test and fix every single Wayland compositor out
> > > > > there."
> > > > >
> > > > > "I really think you'd be far far better off trying to figure out how to
> > > > > move off the legacy Cirrus emulation as soon as you can."
> > > > >
> > > > > So, we can try seeing if I manage to get some logs out of Mutter to
> > > > > figure out the actual bug more precisely _but_, considering all that
> > > > > people have said both here and in the other forums, I think it would be
> > > > > better to spend that time figuring out how to switch (and document this
> > > > > for 4.8 and previous version, of course).
> > > >
> > > >
> > > > Yes. Also as there does not seem to be any supported OS that
> > > > _needs_ the old Cirrus OS to boot and function.
> > > 
> > > Not that I oppose to change to stdvga, but what would happen to Windows
> > > VMs
> > > that suddenly change from cirrus to stdvga, is that going to trigger the
> > > license invalidation stuff? AFAIK this happens when you change hardware,
> > > but
> > > maybe a VGA change doesn't trigger that because it's common for people to
> > > upgrade VGA cards?
> > > 
> > 
> > Changing anything that Windows considers part of the core system will invalidate a license but, as you say, the graphics device may not be core. There is another issue (which I just verified myself) which as that at least some versions of Windows (Server 2008 in my case) won't boot when using stdvga with qemu trad.
> 
> IMHO, I would leave qemu-trad alone and just change the default gfx card for
> qemu-xen. I don't see much point in touching anything in qemu-trad.

Exactly. The patch I had was only for qemu-xen.

> 
> Roger.

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