14.11.2010 12:44, Josselin Mouette wrote: > Le dimanche 14 novembre 2010 à 07:18 +0300, Michael Tokarev a écrit : >>> I find it a serious hindrance not being able to use a squeeze guest in a >>> squeeze host - after all this should be a very common setup. >> >> This is not realistic at this point to update seabios to >> 0.6.x version (currently used is 0.5), since it is not >> tested at all with 0.12 version of qemu-kvm, and 0.5 is >> what's shipped with 0.12. Virtio block support went into >> seabios in 0.6 series, and there were a few changes needed >> in qemu-kvm side for that too. > > Sorry but I don’t understand why you don’t consider that a serious > regression. Booting on virtio used to work perfectly with lenny guests > on lenny hosts.
Booting on virtio works with lenny and works now. I use virtio all the time, have many production guests with virtio disks, and do a lot of testing too. But even if it were non-bootable -- a regression from what? Kvm version on lenny was completely unreliable. It was nothing more than a development snapshot, a code drop, not even an alpha release. Current kvm in squeeze at least appears to be stable. So in terms of kvm it is definitely not a regression, -- from a non-working system it become something that works at least for most users. Or do you mean regression in squeeze (guest) kernel or parted maybe? So that lenny's parted or kernel were able to boot from virtio and squeeze's does not? Why are you filing a bug against qemu in this case? Or I don't understand something? >> You can use squeeze guest in a squeeze host just fine, >> just install it to ide as you did before, so I'm not >> seeing why it's Severity: important. Care to explain? > > Virtio is the recommended I/O driver in basically all KVM documentation, > and I think there are good reasons for that. Not being able to boot from > it anymore means you have to provide 2 disks to all your guests, which > makes the setup a monstrosity if you have a lot of them. Again, it boots perfectly here, no problems. Maybe squeeze can not _install_ into it -- this is a different matter, and at least it does not look like from your bug description. As far as I understand your problem, you can't _install_ squeeze into a virtio disk, but if you install to ide but later use virtio it works. Is it the right understanding? >> I'm going to downgrade the severity to minor and tag it >> as wontfix, at least for squeeze - booting from virtio >> disk fully supported in 0.13 which is in -experimental. > > How about shipping an alternate Bios image that is not used by default? > At the very least, the documentation should mention that virtio support > is crippled in the package’s current status. I'm not seeing it's cripped at all. All my guests, be it various windows guests or linux - debian lenny, debian squeeze, redhat, -- all are booting from virtio disks (or virtio net for that matter, using pxe remote boot) just fine. Granted, the problem you're desribing - at least if I understand it right - is something I haven't seen so far, this is something ne to me. I'll try to work it out tomorrow, by attempting to install current squeeze into a virtio disk in kvm. As of an alternative bios - it is already shipped in -experimental branch for quite some time. The package's named seabios. This is the version that finally fully supports booting from virtio disks without any additional boot rom kludges. Thanks! /mjt -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
