Upstream commit:

commit b4a5fd95f70b373601abcecdb9d87e4cdb219350
Author: Pavel Hrdina <phrdina>
Date:   Tue Feb 23 17:04:19 2016 +0100

    qemu: introduce vram64 attribute for QXL video device

One more comment about this statement "PLEASE, don't use kilobytes for this value".  This is our representation in XML.  The size in kibibytes is converted to mebibytes for qemu and everything works correctly.

This bug was accidentally moved from POST to MODIFIED via an error in automation, please see mmccune with any questions

Hi,Pavel, I test the patch with libvirt-1.3.3-1.el7.x86_64 , find 'virsh dumpxm' did not show the vram64 setting and the screen resolution still is 2k in linux guest. Would you please help to check that? Thanks a lot.

My test steps:
1.Edit guest with vram64 setting:
  #virsh edit rhel7.2
  ...
   <video>
   <model type='qxl' vram64='131072' heads='1'/>
  </video>
  ...

2.Start the guest:
  #virsh start rhel7.2

3.Dump guest xml and not show vram64 setting in the guest xml:
  #virsh dumpxml rhel7.2 | grep -A5 video
    <video>
      <model type='qxl' ram='65536' vram='65536' vgamem='16384' heads='1' primary='yes'/>         ------not show vram64 setting
      <alias name='video0'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x02' function='0x0'/>
    </video>

4.Login in the guest and check the qxl device and can see the 64b bar:
  #dmesg | grep -i qxl
  [    1.222614] [drm] qxl: 16M of VRAM memory size
[    1.222614] [drm] qxl: 63M of IO pages memory ready (VRAM domain)
[    1.222615] [drm] qxl: 128M of Surface memory size
[    1.227247] fbcon: qxldrmfb (fb0) is primary device
[    1.240532] qxl 0000:00:02.0: fb0: qxldrmfb frame buffer device
[    1.240533] qxl 0000:00:02.0: registered panic notifier
[    1.240537] [drm] Initialized qxl 0.1.0 20120117 for 0000:00:02.0 on minor 0

5.Test with windows2012 guest, the screen resolution can be changed to 2560X1600;

6.Test with linux guest, the screen resolution still only is 2k.

Hi Yan,

Oh, the missing vram64 in XML is a bug, I've just fixed it and send a patch to upstream.  Next rebase will include this patch also for RHEL.

The second part that the linux guest can set only 2k resolution isn't a libvirt issue, since as you've verified, the guest can see that the QXL device has that memory available.

The only bug in libvirt is that we drop the vram64 from live XML.

Upstream patch posted:
https://www.redhat.com/archives/libvir-list/2016-April/msg00670.html

Hi,Pavel, I test the patch with libvirt-1.3.4-1.el7.x86_64, libvirt still drop the vram64 from live XML.

Hi Yen,

I forget to push it after it was acked.  It's pushed now and it will be included in the next rebase.

Hi,Pavel,I test the bug with libvirt-1.3.5-1.el7.x86_64, and find when set vram64 to value more than '1024*2048', the value displayed in the guest is not correct.

steps:
1.Set vram64 to '1024*4096':
   ...
     <model type='qxl' ram='65536' vram='65536' vram64='4194304' vgamem='16384' heads='1' primary='yes'/>
   ...

2.Start the guest:
  #virsh start rhel7.3

3.Check the vram64 value in the qemu cmds:
  #ps aux | grep -i vram64
  ...vram64_size_mb=4096...

4.Login in the guest and check the qxl device and can see the 64b bar:
  ...
[    4.369750] [drm] qxl: 16M of VRAM memory size
[    4.369751] [drm] qxl: 63M of IO pages memory ready (VRAM domain)
***[    4.369752] [drm] qxl: 64M of Surface memory size***
[    4.371639] fbcon: qxldrmfb (fb0) is primary device
[    4.395866] qxl 0000:00:0a.0: fb0: qxldrmfb frame buffer device
[    4.395867] qxl 0000:00:0a.0: registered panic notifier
[    4.395873] [drm] Initialized qxl 0.1.0 20120117 for 0000:00:0a.0 on minor 0

Hi Yan,

If you check qemu command line using "ps ax | grep qemu" you would see that libvirt correctly passes the vram64 argument in MiB to qemu.  This is either qemu or qxl bug.

Verified pass with libvirt-2.0.0-5.el7.x86_64.

Test steps:
1.Edit guest with vram64 setting:
  #virsh edit rhel7.2
  ...
   <video>
   <model type='qxl' vram64='131072' heads='1'/>
  </video>
  ...

2.Start the guest:
  #virsh start rhel7.2

3.Dump guest xml and not show vram64 setting in the guest xml:
  #virsh dumpxml rhel7.2 | grep -A5 video
    <video>
      <model type='qxl' ram='65536' vram='65536' ***vram64='131072'*** vgamem='16384' heads='1' primary='yes'/>         
      <alias name='video0'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x02' function='0x0'/>
    </video>

4.Check the qemu cmd:
#ps aux | grep -i vram64
...vram64_size_mb=128,...


5.Login in the guest and check the qxl device and can see the 64b bar:
  #dmesg | grep -i qxl
[    3.079634] [drm] qxl: 16M of VRAM memory size
[    3.079635] [drm] qxl:63M of IO pages memory ready (VRAM domain)
***[    3.079636] [drm] qxl: 128M of Surface memory size***

6.Set invalid value in the domain xml(<0,>2^32-1,non integer chars):
#virsh edit rhel7.3
...
 <video>
      <model type='qxl' ram='65536' vram='65536' vram64='4294967296' vgamem='16384' heads='1'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x09' function='0x0'/>
    </video

save the xml, failed to validate:
error: XML document failed to validate against schema: Unable to validate doc against /usr/share/libvirt/schemas/domain.rng
Extra element devices in interleave
Element domain failed to validate content

Failed. Try again? [y,n,i,f,?]: 

7.Do migration with vram64 setting:
#virsh migrate rhel7.3 qemu+ssh://10.66.144.8/system --live --verbose
 Migration: [100 %]

8.Do some operations after step6 ,guest works well.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://rhn.redhat.com/errata/RHSA-2016-2577.html