[Qemu-devel] CONFIG_FB_CIRRUS for Linux guests

[Qemu-devel] CONFIG_FB_CIRRUS for Linux guests

[Jan Kiszka]

Hi,

just received some Linux guest kernel that refuses to properly use its
framebuffer console under QEMU (probably any version, tried down to some
0.10.x). It turned out that it has CONFIG_FB_CIRRUS enabled and fails to
initialize the display when this is actually used. Anyone came across
this before or has some idea what goes wrong?

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Qemu-devel] CONFIG_FB_CIRRUS for Linux guests

[Alexander Graf]

On 26.01.2010, at 15:44, Jan Kiszka wrote:

> Hi,
> 
> just received some Linux guest kernel that refuses to properly use its
> framebuffer console under QEMU (probably any version, tried down to some
> 0.10.x). It turned out that it has CONFIG_FB_CIRRUS enabled and fails to
> initialize the display when this is actually used. Anyone came across
> this before or has some idea what goes wrong?

The only thing I've seen that might be a similar issue is that the Cirrus driver on Windows NT 4 just shows a black screen with KVM.

Alex

Re: [Qemu-devel] CONFIG_FB_CIRRUS for Linux guests

[Anthony Liguori]

On 01/26/2010 08:44 AM, Jan Kiszka wrote:
> Hi,
>
> just received some Linux guest kernel that refuses to properly use its
> framebuffer console under QEMU (probably any version, tried down to some
> 0.10.x). It turned out that it has CONFIG_FB_CIRRUS enabled and fails to
> initialize the display when this is actually used. Anyone came across
> this before or has some idea what goes wrong?
>    

Looks like it worked at some point:

commit ca896ef389fc44c7a673d1bd2b94823e28da37a1
Author: aurel32 <aurel32@c046a42c-6fe2-441c-8c8c-71466251a162>
Date:   Thu May 8 12:21:27 2008 +0000

     cirrusfb: proper "Attribute Controller Toggle Readback" register 
behaviour

     (Marcelo Tosatti)


     git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4386 
c046a42c-6fe2-441c

The cirrusfb Linux driver uses the toggle readback register to determine
whether to flip 0xC30 to data/address mode.

Without proper implementation the VGA mode will be erroneously set to
blank resulting in crashes when starting BITBLT.

3d4h index 24h (R):  Attribute Controller Toggle Readback Register
bit    7  If set the Attribute Controller will next access a data register, if
           clear it'll access the index register


Signed-off-by: Marcelo Tosatti<mtosatti@redhat.com>

Regards,

Anthony Liguori

> Jan
>
>    

Re: [Qemu-devel] CONFIG_FB_CIRRUS for Linux guests

[René Rebe]

Hi,

I saw that some pretty long time ago, but didn't care enough to investigate, just disabled the cirrus fb  in T2 (http://t2-project.org):

rene@8way:t2-trunk-clean$ svn log -r29093
------------------------------------------------------------------------
r29093 | rene | 2008-05-27 09:37:57 +0200 (Tue, 27 May 2008) | 4 lines

        * updated linux26 module blacklist, do not list ide-generic (that's
          handled these days), but add cirrusfb, which causes a pure black
          screen in Qemu

------------------------------------------------------------------------
rene@8way:t2-trunk-clean$ svn di -c r29093
Index: package/base/module-init-tools/modprobe.conf.data
===================================================================
--- package/base/module-init-tools/modprobe.conf.data   (revision 29092)
+++ package/base/module-init-tools/modprobe.conf.data   (revision 29093)
@@ -203,5 +203,4 @@
 alias char-major-89-* i2c-dev
 alias net-pf-24 pppoe
 
-# do not auto-load (e.g. on boot)
-blacklist ide-generic
+blacklist cirrusfb # plain black screen in Qemu

On 26.01.2010, at 17:02, Alexander Graf wrote:

> 
> On 26.01.2010, at 15:44, Jan Kiszka wrote:
> 
>> Hi,
>> 
>> just received some Linux guest kernel that refuses to properly use its
>> framebuffer console under QEMU (probably any version, tried down to some
>> 0.10.x). It turned out that it has CONFIG_FB_CIRRUS enabled and fails to
>> initialize the display when this is actually used. Anyone came across
>> this before or has some idea what goes wrong?
> 
> The only thing I've seen that might be a similar issue is that the Cirrus driver on Windows NT 4 just shows a black screen with KVM.
> 
> Alex
> 

-- 
  René Rebe, ExactCODE GmbH, Jaegerstr. 67, DE-10117 Berlin
  http://exactcode.com | http://t2-project.org | http://rene.rebe.name

Re: [Qemu-devel] CONFIG_FB_CIRRUS for Linux guests

[Jan Kiszka]

Anthony Liguori wrote:
> On 01/26/2010 08:44 AM, Jan Kiszka wrote:
>> Hi,
>>
>> just received some Linux guest kernel that refuses to properly use its
>> framebuffer console under QEMU (probably any version, tried down to some
>> 0.10.x). It turned out that it has CONFIG_FB_CIRRUS enabled and fails to
>> initialize the display when this is actually used. Anyone came across
>> this before or has some idea what goes wrong?
>>    
> 
> Looks like it worked at some point:
> 
> commit ca896ef389fc44c7a673d1bd2b94823e28da37a1
> Author: aurel32 <aurel32@c046a42c-6fe2-441c-8c8c-71466251a162>
> Date:   Thu May 8 12:21:27 2008 +0000
> 
>      cirrusfb: proper "Attribute Controller Toggle Readback" register 
> behaviour
> 
>      (Marcelo Tosatti)
> 
> 
>      git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4386 
> c046a42c-6fe2-441c
> 
> The cirrusfb Linux driver uses the toggle readback register to determine
> whether to flip 0xC30 to data/address mode.
> 
> Without proper implementation the VGA mode will be erroneously set to
> blank resulting in crashes when starting BITBLT.
> 
> 3d4h index 24h (R):  Attribute Controller Toggle Readback Register
> bit    7  If set the Attribute Controller will next access a data register, if
>            clear it'll access the index register
> 
> 
> Signed-off-by: Marcelo Tosatti<mtosatti@redhat.com>
> 

Yep, and 2bec46dc97571a3c34b18fe4ca198e7bfbdca41f broke it again.

Unless Glauber immediately confesses the crime and removes the bug he
left behind, I will try to understand the reason later.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux