Disabling cirrus-vga

Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 384 bytes --]

Hi xen developers,

Does any of you know how to disable the cirrus-vga emulation for hvm in xen?
I have been trying to disable that from the source codes, but even with
that, I still see a VGA controller when I boot up my guest domain. In
particular, I would like to do this because I am passing through a graphics
card to my guest domain using direct pass through.


Thanks,
- billy

[-- Attachment #1.2: Type: text/html, Size: 407 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Jun Koi]

On Fri, Dec 12, 2008 at 9:49 AM, billy lau <billylau@umich.edu> wrote:
> Hi xen developers,
>
> Does any of you know how to disable the cirrus-vga emulation for hvm in xen?
> I have been trying to disable that from the source codes, but even with
> that, I still see a VGA controller when I boot up my guest domain. In
> particular, I would like to do this because I am passing through a graphics
> card to my guest domain using direct pass through.

Sorry that I dont understand your motivation. If your card is
passthrough, the guest would load the appropriate driver for that. So
why do you care about the cirrus-vga driver?

Thanks,
Jun

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 974 bytes --]

On Thu, Dec 11, 2008 at 8:09 PM, Jun Koi <junkoi2004@gmail.com> wrote:

> On Fri, Dec 12, 2008 at 9:49 AM, billy lau <billylau@umich.edu> wrote:
> > Hi xen developers,
> >
> > Does any of you know how to disable the cirrus-vga emulation for hvm in
> xen?
> > I have been trying to disable that from the source codes, but even with
> > that, I still see a VGA controller when I boot up my guest domain. In
> > particular, I would like to do this because I am passing through a
> graphics
> > card to my guest domain using direct pass through.
>
> Sorry that I dont understand your motivation. If your card is
> passthrough, the guest would load the appropriate driver for that. So
> why do you care about the cirrus-vga driver?
>

This is because, even after installing the appropriate driver, there is a
conflict of device. Even after I disabled the cirrus driver, the actual
driver was reported to be "unable to access resources". Does that sound
reasonable then?

- billy

[-- Attachment #1.2: Type: text/html, Size: 1408 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Stefano Stabellini]

billy lau wrote:

> 
> 
> On Thu, Dec 11, 2008 at 8:09 PM, Jun Koi <junkoi2004@gmail.com
> <mailto:junkoi2004@gmail.com>> wrote:
> 
>     On Fri, Dec 12, 2008 at 9:49 AM, billy lau <billylau@umich.edu
>     <mailto:billylau@umich.edu>> wrote:
>     > Hi xen developers,
>     >
>     > Does any of you know how to disable the cirrus-vga emulation for
>     hvm in xen?
>     > I have been trying to disable that from the source codes, but even
>     with
>     > that, I still see a VGA controller when I boot up my guest domain. In
>     > particular, I would like to do this because I am passing through a
>     graphics
>     > card to my guest domain using direct pass through.
> 
>     Sorry that I dont understand your motivation. If your card is
>     passthrough, the guest would load the appropriate driver for that. So
>     why do you care about the cirrus-vga driver?
> 
> 
> This is because, even after installing the appropriate driver, there is
> a conflict of device. Even after I disabled the cirrus driver, the
> actual driver was reported to be "unable to access resources". Does that
> sound reasonable then?


I think at the moment there is no way to do it from the command line.
You have to comment out any call to pci_cirrus_vga_init,
isa_cirrus_vga_init, pci_vmsvga_init, pci_vga_init and isa_vga_init in
hw/pc.c (lines 939-962).

You may also want to disable loadind the vgabios, also in hw/pc.c
(load_image_targphys).

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 1707 bytes --]

On Fri, Dec 12, 2008 at 5:33 AM, Stefano Stabellini <
stefano.stabellini@eu.citrix.com> wrote:

> billy lau wrote:
>
> >
> >
> > On Thu, Dec 11, 2008 at 8:09 PM, Jun Koi <junkoi2004@gmail.com
> > <mailto:junkoi2004@gmail.com>> wrote:
> >
> >     On Fri, Dec 12, 2008 at 9:49 AM, billy lau <billylau@umich.edu
> >     <mailto:billylau@umich.edu>> wrote:
> >     > Hi xen developers,
> >     >
> >     > Does any of you know how to disable the cirrus-vga emulation for
> >     hvm in xen?
> >     > I have been trying to disable that from the source codes, but even
> >     with
> >     > that, I still see a VGA controller when I boot up my guest domain.
> In
> >     > particular, I would like to do this because I am passing through a
> >     graphics
> >     > card to my guest domain using direct pass through.
> >
> >     Sorry that I dont understand your motivation. If your card is
> >     passthrough, the guest would load the appropriate driver for that. So
> >     why do you care about the cirrus-vga driver?
> >
> >
> > This is because, even after installing the appropriate driver, there is
> > a conflict of device. Even after I disabled the cirrus driver, the
> > actual driver was reported to be "unable to access resources". Does that
> > sound reasonable then?
>
>
> I think at the moment there is no way to do it from the command line.
> You have to comment out any call to pci_cirrus_vga_init,
> isa_cirrus_vga_init, pci_vmsvga_init, pci_vga_init and isa_vga_init in
> hw/pc.c (lines 939-962).
>
> You may also want to disable loadind the vgabios, also in hw/pc.c
> (load_image_targphys).
>

Alright, I will give that a try, and get back to you guys again. Thanks for
the input.

- billy

[-- Attachment #1.2: Type: text/html, Size: 2606 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 4041 bytes --]

Okay, after commenting those lines out, I noticed that my hvmloader doesn't
run anymore. What happens is that when I do a xm list, I see this

Name                                        ID   Mem VCPUs      State
Time(s)
Domain-0                                     0  2950     4     r-----
13.0
vienna-windows                               1  1024     1     ------
1.0

vienna-windows is my hvm guest. Then, looking at the log file,

domid: 1
qemu: the number of cpus is 1
config qemu network with xen bridge for  tap1.0 xenbr0
Watching /local/domain/0/device-model/1/logdirty/next-active
Watching /local/domain/0/device-model/1/command
xs_read(): vncpasswd get error.
/vm/adeb6f31-4656-4a98-032f-14ebeb4b44a9/vncpasswd.
qemu_map_cache_init nr_buckets = 10000 size 3145728
shared page at pfn 3fffe
buffered io page at pfn 3fffc
Time offset set 0
register_real_device: Assigning real physical device 02:00.0 ...
pt_register_regions: IO region registered (size=0x01000000
base_addr=0xfa000000)
pt_register_regions: IO region registered (size=0x10000000
base_addr=0xd0000000)
pt_register_regions: IO region registered (size=0x02000000
base_addr=0xf8000000)
pt_register_regions: IO region registered (size=0x00000080
base_addr=0x0000dc80)
pt_register_regions: Expansion ROM registered (size=0x00020000
base_addr=0xfbd00000)
register_real_device: Real physical device 02:00.0 registered successfuly!
Register xen platform.
Done register platform.
xs_read(/local/domain/0/device-model/1/xen_extended_power_mgmt): read error
medium change watch on `hdc' (index: 1): /home/billy/windowsxp/winxp_sp3.iso
I/O request not ready: 0, ptr: 0, port: 0, data: 0, count: 0, size: 0
pt_iomem_map: e_phys=f0000000 maddr=f8000000 type=0 len=33554432 index=3
first_map=1
pt_iomem_map: e_phys=f3000000 maddr=fa000000 type=0 len=16777216 index=0
first_map=1
pt_iomem_map: e_phys=f4000000 maddr=fbd00000 type=8 len=131072 index=6
first_map=1
pt_ioport_map: e_phys=c200 pio_base=dc80 len=128 index=5 first_map=1

But I cannot connect to my guess (I have been using vncviewer to connect to
it). And it seems that the guest domain just stops, with no status. But
there is no error in the log. Any advice?

thanks,
- billy


On Fri, Dec 12, 2008 at 7:46 AM, billy lau <billylau@umich.edu> wrote:

>
>
> On Fri, Dec 12, 2008 at 5:33 AM, Stefano Stabellini <
> stefano.stabellini@eu.citrix.com> wrote:
>
>> billy lau wrote:
>>
>> >
>> >
>> > On Thu, Dec 11, 2008 at 8:09 PM, Jun Koi <junkoi2004@gmail.com
>> > <mailto:junkoi2004@gmail.com>> wrote:
>> >
>> >     On Fri, Dec 12, 2008 at 9:49 AM, billy lau <billylau@umich.edu
>> >     <mailto:billylau@umich.edu>> wrote:
>> >     > Hi xen developers,
>> >     >
>> >     > Does any of you know how to disable the cirrus-vga emulation for
>> >     hvm in xen?
>> >     > I have been trying to disable that from the source codes, but even
>> >     with
>> >     > that, I still see a VGA controller when I boot up my guest domain.
>> In
>> >     > particular, I would like to do this because I am passing through a
>> >     graphics
>> >     > card to my guest domain using direct pass through.
>> >
>> >     Sorry that I dont understand your motivation. If your card is
>> >     passthrough, the guest would load the appropriate driver for that.
>> So
>> >     why do you care about the cirrus-vga driver?
>> >
>> >
>> > This is because, even after installing the appropriate driver, there is
>> > a conflict of device. Even after I disabled the cirrus driver, the
>> > actual driver was reported to be "unable to access resources". Does that
>> > sound reasonable then?
>>
>>
>> I think at the moment there is no way to do it from the command line.
>> You have to comment out any call to pci_cirrus_vga_init,
>> isa_cirrus_vga_init, pci_vmsvga_init, pci_vga_init and isa_vga_init in
>> hw/pc.c (lines 939-962).
>>
>> You may also want to disable loadind the vgabios, also in hw/pc.c
>> (load_image_targphys).
>>
>
> Alright, I will give that a try, and get back to you guys again. Thanks for
> the input.
>
> - billy
>
>

[-- Attachment #1.2: Type: text/html, Size: 6071 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Jun Koi]

On Sat, Dec 13, 2008 at 9:58 AM, billy lau <billylau@umich.edu> wrote:
> Okay, after commenting those lines out, I noticed that my hvmloader doesn't
> run anymore. What happens is that when I do a xm list, I see this
>
> Name                                        ID   Mem VCPUs      State
> Time(s)
> Domain-0                                     0  2950     4     r-----
> 13.0
> vienna-windows                               1  1024     1     ------
> 1.0
>
> vienna-windows is my hvm guest. Then, looking at the log file,
>
> domid: 1
> qemu: the number of cpus is 1
> config qemu network with xen bridge for  tap1.0 xenbr0
> Watching /local/domain/0/device-model/1/logdirty/next-active
> Watching /local/domain/0/device-model/1/command
> xs_read(): vncpasswd get error.
> /vm/adeb6f31-4656-4a98-032f-14ebeb4b44a9/vncpasswd.

This looks like a bug??


>
> On Fri, Dec 12, 2008 at 7:46 AM, billy lau <billylau@umich.edu> wrote:
>>
>>
>> On Fri, Dec 12, 2008 at 5:33 AM, Stefano Stabellini
>> <stefano.stabellini@eu.citrix.com> wrote:
>>>
>>> billy lau wrote:
>>>
>>> >
>>> >
>>> > On Thu, Dec 11, 2008 at 8:09 PM, Jun Koi <junkoi2004@gmail.com
>>> > <mailto:junkoi2004@gmail.com>> wrote:
>>> >
>>> >     On Fri, Dec 12, 2008 at 9:49 AM, billy lau <billylau@umich.edu
>>> >     <mailto:billylau@umich.edu>> wrote:
>>> >     > Hi xen developers,
>>> >     >
>>> >     > Does any of you know how to disable the cirrus-vga emulation for
>>> >     hvm in xen?
>>> >     > I have been trying to disable that from the source codes, but
>>> > even
>>> >     with
>>> >     > that, I still see a VGA controller when I boot up my guest
>>> > domain. In
>>> >     > particular, I would like to do this because I am passing through
>>> > a
>>> >     graphics
>>> >     > card to my guest domain using direct pass through.
>>> >
>>> >     Sorry that I dont understand your motivation. If your card is
>>> >     passthrough, the guest would load the appropriate driver for that.
>>> > So
>>> >     why do you care about the cirrus-vga driver?
>>> >
>>> >
>>> > This is because, even after installing the appropriate driver, there is
>>> > a conflict of device. Even after I disabled the cirrus driver, the
>>> > actual driver was reported to be "unable to access resources". Does
>>> > that
>>> > sound reasonable then?
>>>
>>>
>>> I think at the moment there is no way to do it from the command line.
>>> You have to comment out any call to pci_cirrus_vga_init,
>>> isa_cirrus_vga_init, pci_vmsvga_init, pci_vga_init and isa_vga_init in
>>> hw/pc.c (lines 939-962).
>>>
>>> You may also want to disable loadind the vgabios, also in hw/pc.c
>>> (load_image_targphys).
>>
>> Alright, I will give that a try, and get back to you guys again. Thanks
>> for the input.
>>
>> - billy
>>
>
>

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 1064 bytes --]

On Fri, Dec 12, 2008 at 9:58 PM, Jun Koi <junkoi2004@gmail.com> wrote:

> On Sat, Dec 13, 2008 at 9:58 AM, billy lau <billylau@umich.edu> wrote:
> > Okay, after commenting those lines out, I noticed that my hvmloader
> doesn't
> > run anymore. What happens is that when I do a xm list, I see this
> >
> > Name                                        ID   Mem VCPUs      State
> > Time(s)
> > Domain-0                                     0  2950     4     r-----
> > 13.0
> > vienna-windows                               1  1024     1     ------
> > 1.0
> >
> > vienna-windows is my hvm guest. Then, looking at the log file,
> >
> > domid: 1
> > qemu: the number of cpus is 1
> > config qemu network with xen bridge for  tap1.0 xenbr0
> > Watching /local/domain/0/device-model/1/logdirty/next-active
> > Watching /local/domain/0/device-model/1/command
> > xs_read(): vncpasswd get error.
> > /vm/adeb6f31-4656-4a98-032f-14ebeb4b44a9/vncpasswd.
>
> This looks like a bug??


I don't really know. It is there all the time, even before I did the
modification.

- billy

[-- Attachment #1.2: Type: text/html, Size: 1862 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 103 bytes --]

So, does anyone else have any thought on the original problem of disabling
cirrus-vga in xen?

- billy

[-- Attachment #1.2: Type: text/html, Size: 113 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Stefano Stabellini]

billy lau wrote:

> So, does anyone else have any thought on the original problem of
> disabling cirrus-vga in xen?
> 

Can you post the patch you used to disable cirrus?
Without it, it is very difficult to tell what is wrong...

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 722 bytes --]

Basically, I have just commented out all calls to pci_cirrus_vga_init,
isa_cirrus_vga_init, pci_vmsvga_init, pci_vga_init and isa_vga_init in
hw/pc.c, and also load_image_targphys() for loading vgabios, like what you
have suggested.

On Mon, Dec 15, 2008 at 5:46 AM, Stefano Stabellini <
stefano.stabellini@eu.citrix.com> wrote:

> billy lau wrote:
>
> > So, does anyone else have any thought on the original problem of
> > disabling cirrus-vga in xen?
> >
>
> Can you post the patch you used to disable cirrus?
> Without it, it is very difficult to tell what is wrong...
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
>

[-- Attachment #1.2: Type: text/html, Size: 1172 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Stefano Stabellini]

billy lau wrote:

> Basically, I have just commented out all calls to pci_cirrus_vga_init,
> isa_cirrus_vga_init, pci_vmsvga_init, pci_vga_init and isa_vga_init in
> hw/pc.c, and also load_image_targphys() for loading vgabios, like what
> you have suggested.



It works well for me, I tested it with a Linux HVM guest using the
serial port.
I used the following patch:

---

diff -r 0ea6bd53cfb6 hw/pc.c
--- a/hw/pc.c	Thu Oct 23 10:26:02 2008 +0100
+++ b/hw/pc.c	Mon Dec 15 12:05:54 2008 +0000
@@ -854,7 +854,7 @@
         goto vga_bios_error;
     vga_bios_offset = qemu_ram_alloc(65536);
 
-    ret = load_image_targphys(buf, vga_bios_offset, vga_bios_size);
+    ret = vga_bios_size;
     if (ret != vga_bios_size) {
     vga_bios_error:
         fprintf(stderr, "qemu: could not load VGA BIOS '%s'\n", buf);
@@ -930,31 +930,6 @@
     register_ioport_write(0x80, 1, 1, ioport80_write, NULL);
 
     register_ioport_write(0xf0, 1, 1, ioportF0_write, NULL);
-
-    if (cirrus_vga_enabled) {
-        if (pci_enabled) {
-            pci_cirrus_vga_init(pci_bus,
-                                ds, phys_ram_base + vga_ram_addr,
-                                vga_ram_addr, vga_ram_size);
-        } else {
-            isa_cirrus_vga_init(ds, phys_ram_base + vga_ram_addr,
-                                vga_ram_addr, vga_ram_size);
-        }
-    } else if (vmsvga_enabled) {
-        if (pci_enabled)
-            pci_vmsvga_init(pci_bus, ds, phys_ram_base + vga_ram_addr,
-                            vga_ram_addr, vga_ram_size);
-        else
-            fprintf(stderr, "%s: vmware_vga: no PCI bus\n", __FUNCTION__);
-    } else {
-        if (pci_enabled) {
-            pci_vga_init(pci_bus, ds, phys_ram_base + vga_ram_addr,
-                         vga_ram_addr, vga_ram_size, 0, 0);
-        } else {
-            isa_vga_init(ds, phys_ram_base + vga_ram_addr,
-                         vga_ram_addr, vga_ram_size);
-        }
-    }
 
 #ifdef CONFIG_PASSTHROUGH
     /* Pass-through Initialization

Re: Disabling cirrus-vga

[Trolle Selander]

[-- Attachment #1.1: Type: text/plain, Size: 427 bytes --]

Maybe it would be a good idea if the "nographic" option actually removed all
virtual graphics cards? Is anyone out there actually running with the
nographic option and still depending on the presence of a virtual VGA card
in the VM? Until this thread, I actually thought the nographic option did
just this, and it still strikes me as a more "logical" behavior, but maybe
changing it now would break existing setups.

-- Trolle

[-- Attachment #1.2: Type: text/html, Size: 458 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Stefano Stabellini]

Trolle Selander wrote:

> Maybe it would be a good idea if the "nographic" option actually removed
> all virtual graphics cards? Is anyone out there actually running with
> the nographic option and still depending on the presence of a virtual
> VGA card in the VM? Until this thread, I actually thought the nographic
> option did just this, and it still strikes me as a more "logical"
> behavior, but maybe changing it now would break existing setups.
> 

nographic is an old qemu cmd line option (since 2003), I doubt people
would be very happy if we change the meaning of it.
However we could add a new one.

Re: Disabling cirrus-vga

[Daniel P. Berrange]

On Tue, Dec 16, 2008 at 11:03:40AM +0000, Stefano Stabellini wrote:
> Trolle Selander wrote:
> 
> > Maybe it would be a good idea if the "nographic" option actually removed
> > all virtual graphics cards? Is anyone out there actually running with
> > the nographic option and still depending on the presence of a virtual
> > VGA card in the VM? Until this thread, I actually thought the nographic
> > option did just this, and it still strikes me as a more "logical"
> > behavior, but maybe changing it now would break existing setups.
> > 
> 
> nographic is an old qemu cmd line option (since 2003), I doubt people
> would be very happy if we change the meaning of it.

Absolutely not !

> However we could add a new one.

Latest upstream QEMU now has a sane arg for specifying video device
type

         -vga [std|cirrus|vmware]

If we were to follow existing practice for -serial/-net/etc disabling then
the answer would be to support '-vga none'  Of course this should be
suggested on the qemu-devel list really.

Regards,
Daniel
-- 
|: Red Hat, Engineering, London   -o-   http://people.redhat.com/berrange/ :|
|: http://libvirt.org  -o-  http://virt-manager.org  -o-  http://ovirt.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505  -o-  F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

Re: Disabling cirrus-vga

[Stefano Stabellini]

Daniel P. Berrange wrote:

> Latest upstream QEMU now has a sane arg for specifying video device
> type
> 
>          -vga [std|cirrus|vmware]
> 
> If we were to follow existing practice for -serial/-net/etc disabling then
> the answer would be to support '-vga none'  Of course this should be
> suggested on the qemu-devel list really.
> 

I think this is a good idea.

Re: Disabling cirrus-vga

[Jun Koi]

On Tue, Dec 16, 2008 at 8:25 PM, Daniel P. Berrange <berrange@redhat.com> wrote:
> On Tue, Dec 16, 2008 at 11:03:40AM +0000, Stefano Stabellini wrote:
>> Trolle Selander wrote:
>>
>> > Maybe it would be a good idea if the "nographic" option actually removed
>> > all virtual graphics cards? Is anyone out there actually running with
>> > the nographic option and still depending on the presence of a virtual
>> > VGA card in the VM? Until this thread, I actually thought the nographic
>> > option did just this, and it still strikes me as a more "logical"
>> > behavior, but maybe changing it now would break existing setups.
>> >
>>
>> nographic is an old qemu cmd line option (since 2003), I doubt people
>> would be very happy if we change the meaning of it.
>
> Absolutely not !
>
>> However we could add a new one.
>
> Latest upstream QEMU now has a sane arg for specifying video device
> type
>
>         -vga [std|cirrus|vmware]
>

I am wondering what is the advantage of vmware over std/cirrus? In
which case we should use vmware??

Thanks,
Jun

Re: Disabling cirrus-vga

[Daniel P. Berrange]

On Thu, Dec 18, 2008 at 10:45:29AM +0900, Jun Koi wrote:
> On Tue, Dec 16, 2008 at 8:25 PM, Daniel P. Berrange <berrange@redhat.com> wrote:
> > On Tue, Dec 16, 2008 at 11:03:40AM +0000, Stefano Stabellini wrote:
> >> Trolle Selander wrote:
> >>
> >> > Maybe it would be a good idea if the "nographic" option actually removed
> >> > all virtual graphics cards? Is anyone out there actually running with
> >> > the nographic option and still depending on the presence of a virtual
> >> > VGA card in the VM? Until this thread, I actually thought the nographic
> >> > option did just this, and it still strikes me as a more "logical"
> >> > behavior, but maybe changing it now would break existing setups.
> >> >
> >>
> >> nographic is an old qemu cmd line option (since 2003), I doubt people
> >> would be very happy if we change the meaning of it.
> >
> > Absolutely not !
> >
> >> However we could add a new one.
> >
> > Latest upstream QEMU now has a sane arg for specifying video device
> > type
> >
> >         -vga [std|cirrus|vmware]
> >
> 
> I am wondering what is the advantage of vmware over std/cirrus? In
> which case we should use vmware??

I believe it allows much larger screen resolutions. The downside of
course is that you need special drivers in the guest.

Daniel
-- 
|: Red Hat, Engineering, London   -o-   http://people.redhat.com/berrange/ :|
|: http://libvirt.org  -o-  http://virt-manager.org  -o-  http://ovirt.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505  -o-  F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 318 bytes --]

Let me ask a simple question. If i use the option of stdvga, and yet I
passthrough my primary graphics card to the guets, am I supposed to be able
to access the graphics card from the guest? I still could not solve the
issue of the driver in hvm guest not being able to access resources.

Thanks for any help,
- billy

[-- Attachment #1.2: Type: text/html, Size: 357 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Stefano Stabellini]

billy lau wrote:

> Let me ask a simple question. If i use the option of stdvga, and yet I
> passthrough my primary graphics card to the guets, am I supposed to be
> able to access the graphics card from the guest? I still could not solve
> the issue of the driver in hvm guest not being able to access resources.
> 

I think until you figure out what exactly is causing the problem you are
having with passthrough, it is best if you try without any other
emulated graphic card.

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 1029 bytes --]

On Thu, Dec 18, 2008 at 6:38 AM, Stefano Stabellini <
stefano.stabellini@eu.citrix.com> wrote:

> billy lau wrote:
>
> > Let me ask a simple question. If i use the option of stdvga, and yet I
> > passthrough my primary graphics card to the guets, am I supposed to be
> > able to access the graphics card from the guest? I still could not solve
> > the issue of the driver in hvm guest not being able to access resources.
> >
>
> I think until you figure out what exactly is causing the problem you are
> having with passthrough, it is best if you try without any other
> emulated graphic card.
>

I mean, that's exactly what I wanted to do. I want to disable all emulated
graphics card, including cirrus-vga so that my guest can directly access the
graphics card. That was when the error with the device model failure, which
I was not able to identify why.

- billy

- billy


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

[-- Attachment #1.2: Type: text/html, Size: 1780 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[billy lau]

[-- Attachment #1.1: Type: text/plain, Size: 4449 bytes --]

Okay, I have rechecked my codes, it is similar to what is done with your
patch. This time, I tried a linux hvm guest as well. And it happens that
this is in my xen log when i do a xm log:

[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for devices vif.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072)
XendDomainInfo.handleShutdownWatch
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for 0.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback
/local/domain/0/backend/vif/1/0/hotplug-status.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback 1.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for devices vscsi.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for devices vbd.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for 768.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback
/local/domain/0/backend/vbd/1/768/hotplug-status.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback 1.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for 5632.
[2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback
/local/domain/0/backend/vbd/1/5632/hotplug-status.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) hotplugStatusCallback
/local/domain/0/backend/vbd/1/5632/hotplug-status.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) hotplugStatusCallback 1.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices irq.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices vkbd.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices vfb.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices
console.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for 0.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices pci.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for 0.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices
ioports.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices tap.
[2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices vtpm.
[2008-12-15 20:25:48 2588] INFO (__init__:1072) Domain linux-test (1)
unpaused.
[2008-12-15 20:25:55 2588] WARNING (__init__:1072) domain linux-test: device
model failure: pid 2999: died due to signal 7; see
/var/log/xen/qemu-dm-linux-test.log


But, again, looking at qemu-dm-linux-test.log, there is no error message:
domid: 1
qemu: the number of cpus is 1
config qemu network with xen bridge for  tap1.0 xenbr0
Watching /local/domain/0/device-model/1/logdirty/next-active
Watching /local/domain/0/device-model/1/command
xs_read(): vncpasswd get error.
/vm/e9ccff9f-dc55-89e3-612f-5c4cad69cd87/vncpasswd.
qemu_map_cache_init nr_buckets = 10000 size 3145728
shared page at pfn 3fffe
buffered io page at pfn 3fffc
Time offset set 0
register_real_device: Assigning real physical device 02:00.0 ...
pt_register_regions: IO region registered (size=0x01000000
base_addr=0xfa000000)
pt_register_regions: IO region registered (size=0x10000000
base_addr=0xd0000000)
pt_register_regions: IO region registered (size=0x02000000
base_addr=0xf8000000)
pt_register_regions: IO region registered (size=0x00000080
base_addr=0x0000dc80)
pt_register_regions: Expansion ROM registered (size=0x00020000
base_addr=0xfbd00000)
register_real_device: Real physical device 02:00.0 registered successfuly!
Register xen platform.
Done register platform.
xs_read(/local/domain/0/device-model/1/xen_extended_power_mgmt): read error
medium change watch on `hdc' (index: 1):
/home/billy/Desktop/debian-40r5-i386-netinst.iso
I/O request not ready: 0, ptr: 0, port: 0, data: 0, count: 0, size: 0
pt_iomem_map: e_phys=f0000000 maddr=f8000000 type=0 len=33554432 index=3
first_map=1
pt_iomem_map: e_phys=f3000000 maddr=fa000000 type=0 len=16777216 index=0
first_map=1
pt_iomem_map: e_phys=f4000000 maddr=fbd00000 type=8 len=131072 index=6
first_map=1
pt_ioport_map: e_phys=c200 pio_base=dc80 len=128 index=5 first_map=1

and it just ends there. xm list shows me this, which is the same as before,
without state status:
Name                                        ID   Mem VCPUs      State
Time(s)
Domain-0                                    0  2951     4     r-----
15.9
linux-test                                     1  1024     1     ------
5.5

So, I guess the question is, what does it mean by device model failure due
to signal 7?

thanks,
- billy

[-- Attachment #1.2: Type: text/html, Size: 5469 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Disabling cirrus-vga

[Stefano Stabellini]

billy lau wrote:

> Okay, I have rechecked my codes, it is similar to what is done with your
> patch. This time, I tried a linux hvm guest as well. And it happens that
> this is in my xen log when i do a xm log:
> 
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for devices vif.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072)
> XendDomainInfo.handleShutdownWatch
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for 0.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback
> /local/domain/0/backend/vif/1/0/hotplug-status.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback 1.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for devices vscsi.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for devices vbd.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for 768.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback
> /local/domain/0/backend/vbd/1/768/hotplug-status.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback 1.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) Waiting for 5632.
> [2008-12-15 20:25:47 2588] DEBUG (__init__:1072) hotplugStatusCallback
> /local/domain/0/backend/vbd/1/5632/hotplug-status.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) hotplugStatusCallback
> /local/domain/0/backend/vbd/1/5632/hotplug-status.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) hotplugStatusCallback 1.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices irq.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices vkbd.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices vfb.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices
> console.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for 0.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices pci.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for 0.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices
> ioports.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices tap.
> [2008-12-15 20:25:48 2588] DEBUG (__init__:1072) Waiting for devices vtpm.
> [2008-12-15 20:25:48 2588] INFO (__init__:1072) Domain linux-test (1)
> unpaused.
> [2008-12-15 20:25:55 2588] WARNING (__init__:1072) domain linux-test:
> device model failure: pid 2999: died due to signal 7; see
> /var/log/xen/qemu-dm-linux-test.log
> 
> 
> But, again, looking at qemu-dm-linux-test.log, there is no error message:
> domid: 1
> qemu: the number of cpus is 1
> config qemu network with xen bridge for  tap1.0 xenbr0
> Watching /local/domain/0/device-model/1/logdirty/next-active
> Watching /local/domain/0/device-model/1/command
> xs_read(): vncpasswd get error.
> /vm/e9ccff9f-dc55-89e3-612f-5c4cad69cd87/vncpasswd.
> qemu_map_cache_init nr_buckets = 10000 size 3145728
> shared page at pfn 3fffe
> buffered io page at pfn 3fffc
> Time offset set 0
> register_real_device: Assigning real physical device 02:00.0 ...
> pt_register_regions: IO region registered (size=0x01000000
> base_addr=0xfa000000)
> pt_register_regions: IO region registered (size=0x10000000
> base_addr=0xd0000000)
> pt_register_regions: IO region registered (size=0x02000000
> base_addr=0xf8000000)
> pt_register_regions: IO region registered (size=0x00000080
> base_addr=0x0000dc80)
> pt_register_regions: Expansion ROM registered (size=0x00020000
> base_addr=0xfbd00000)
> register_real_device: Real physical device 02:00.0 registered successfuly!
> Register xen platform.
> Done register platform.
> xs_read(/local/domain/0/device-model/1/xen_extended_power_mgmt): read error
> medium change watch on `hdc' (index: 1):
> /home/billy/Desktop/debian-40r5-i386-netinst.iso
> I/O request not ready: 0, ptr: 0, port: 0, data: 0, count: 0, size: 0
> pt_iomem_map: e_phys=f0000000 maddr=f8000000 type=0 len=33554432 index=3
> first_map=1
> pt_iomem_map: e_phys=f3000000 maddr=fa000000 type=0 len=16777216 index=0
> first_map=1
> pt_iomem_map: e_phys=f4000000 maddr=fbd00000 type=8 len=131072 index=6
> first_map=1
> pt_ioport_map: e_phys=c200 pio_base=dc80 len=128 index=5 first_map=1
> 
> and it just ends there. xm list shows me this, which is the same as
> before, without state status:
> Name                                        ID   Mem VCPUs      State  
> Time(s)
> Domain-0                                    0  2951     4     r-----    
> 15.9
> linux-test                                     1  1024     1    
> ------      5.5
> 
> So, I guess the question is, what does it mean by device model failure
> due to signal 7?

It means qemu died because it received a SIGBUS signal; the problem
seems to be related to the ioport or memory mapping of passthrough device.