Bug 855083

Using virt-manager on a Centos 6.3 host I ran into same issue that F18-alpha-TC6-i386 would not boot using cirrus video.  MY work around was to switch to vga driver.

Proposing as F18 Blocker:
" The release must install and boot successfully as a virtual guest in a situation where the virtual host is running the previous stable Fedora release, using Fedora's current preferred virtualization technology "
https://fedoraproject.org/wiki/Fedora_18_Beta_Release_Criteria

qxl is current default, but cirrus is still used in old VMs. I'm not really sure we want to block on both qxl and cirrus, so this should just start a discussion.

The 'cirrus' driver shouldn't be used any more, it should be using the 'modesetting' driver.

I think we may have sorted this out in the live kickstarts but not in the package set used in the installer images, I guess.

Discussed at 2012-09-26 blocker review meeting: http://meetbot.fedoraproject.org/fedora-qa/2012-09-26/f18-beta-blocker-review-1.2012-09-26-16.03.log.txt . We agreed that qxl has been the default long enough that cirrus bugs have a very high bar to clear to be taken as blocker, and this doesn't really clear it, as it's simple to workaround in many ways (pick a different adapter, or run the install in basic graphics mode or with nomodeset). So this is rejected as a blocker. However, it's accepted as NTH.

Just to clarify and make sure I understand.  Is qxl supposed to be default on qemu/kvm if run from Centos 6?  I don't even get the qxl driver option in centos unless I switch from vnc video to spice video first.  I am thinking the Centos Default still has not caught up with the "Fedora Default".  I can agree though with the NTH as Centos qemu/kvm can do spice/qxl if you know to set it up.  Piece of minutesha for the Docs I guess.

We didn't look into the CentOS case much. Honestly, for Fedora we mostly care about Fedora.

We're not entirely sure why the 'cirrus' driver was still in the environment and used for this test, to be honest. I'll try and find a minute to test with a few different images here, but we also think that:

https://admin.fedoraproject.org/updates/FEDORA-2012-12272/xorg-x11-drv-cirrus-1.5.1-3.fc18

should fix the case where both cirrus and modesetting are present, so that may be enough to resolve this. Needs more testing.

So I just tested this with smoke3. It appears to have both cirrus and modesetting drivers installed, we'd have to look into the compose logs to figure why xorg-x11-drv-cirrus is being pulled in.

The good news, though, is that it looks like cirrus 1.5.1-3 fixes this: anaconda starts up fine, and the X log shows the cirrus driver loading up and then being unloaded because the cirrus kernel module is loaded, and X falls back to modesetting smoothly. So it seems that 1.5.1-3 is good, even with both the cirrus and modesetting drivers present, this will work. The only possibly dangerous scenario would be if we left the modesetting driver out of some compose or other. I'll mark this as fixed (1.5.1-3 is stable).