On Thu, Sep 15, 2011 at 03:38:09PM -0700, Adam Williamson wrote:
On Thu, 2011-09-15 at 10:10 -0500, W. Michael Petullo wrote:
There are known bugs in FC15/FC16 that have been filled some time ago that folks will sadly run into: 728775, 658387 and 668063
Fortunatly the bugs have patches attached and the files to be modified are shell scripts.
Yep, links here:
The Fedora update system reports this is fixed in grub2-1.99-6.fc16.
Peter Jones submitted a new grubby package yesterday. This seems to fix bug #658387 (i.e., new-kernel-pkg creates a Dom0-style grub.cfg entry if /etc/sysconfig/kernel contains "HYPERVISOR=/boot/xen.gz").
I have not yet tested this on Fedora 16. However, I did test on Fedora 15. In this case, bug #668063 is still in effect. That is, grubby creates most of a GRUB record, but the "module initramfs-..." entry is missing.
Has anyone yet tested this new grubby package on Fedora 16 yet? Does using GRUB 2 makes #668063 irrelevant?
I believe so (the irrelevant part). But the functionality part of grubby picking up /etc/sysconfig/kernel and making that kernel the default is not in Grub2 - so not sure how that can be addressed.
I just added some more description to this bug.
So, there's a meta-point here: we currently 'require' Beta releases to boot as guests on Xen hosts:
"The release must boot successfully as a virtual guest in a situation where the virtual host is running a supported Xen implementation"
I really don't have much knowledge of Xen and haven't followed this discussion closely, but do any currently-known bugs prevent this? If so, please flag them up so they can be considered as Beta blockers...thanks!
Somehow the xen-kbdfront driver is not included in the initrd image (I think) - and we end with anaconda but can't type anything. I've been trying to figure out how to inject said module in the install initrd to see if that is really the problem but running in roadblocks (like xz 5.1.1alpha or 5.0.3 complains about corrupt image, or I've no idea how to make driver disks).
Any ideas?
On 09/16/2011 01:43 AM, Konrad Rzeszutek Wilk wrote:
On Thu, Sep 15, 2011 at 03:38:09PM -0700, Adam Williamson wrote:
On Thu, 2011-09-15 at 10:10 -0500, W. Michael Petullo wrote:
There are known bugs in FC15/FC16 that have been filled some time ago that folks will sadly run into: 728775, 658387 and 668063
Fortunatly the bugs have patches attached and the files to be modified are shell scripts.
Yep, links here:
The Fedora update system reports this is fixed in grub2-1.99-6.fc16.
Peter Jones submitted a new grubby package yesterday. This seems to fix bug #658387 (i.e., new-kernel-pkg creates a Dom0-style grub.cfg entry if /etc/sysconfig/kernel contains "HYPERVISOR=/boot/xen.gz").
I have not yet tested this on Fedora 16. However, I did test on Fedora 15. In this case, bug #668063 is still in effect. That is, grubby creates most of a GRUB record, but the "module initramfs-..." entry is missing.
Has anyone yet tested this new grubby package on Fedora 16 yet? Does using GRUB 2 makes #668063 irrelevant?
I believe so (the irrelevant part). But the functionality part of grubby picking up /etc/sysconfig/kernel and making that kernel the default is not in Grub2 - so not sure how that can be addressed.
I just added some more description to this bug.
So, there's a meta-point here: we currently 'require' Beta releases to boot as guests on Xen hosts:
"The release must boot successfully as a virtual guest in a situation where the virtual host is running a supported Xen implementation"
I really don't have much knowledge of Xen and haven't followed this discussion closely, but do any currently-known bugs prevent this? If so, please flag them up so they can be considered as Beta blockers...thanks!
Somehow the xen-kbdfront driver is not included in the initrd image (I think) - and we end with anaconda but can't type anything. I've been trying to figure out how to inject said module in the install initrd to see if that is really the problem but running in roadblocks (like xz 5.1.1alpha or 5.0.3 complains about corrupt image, or I've no idea how to make driver disks).
I saw the same thing and worked around it by booting with "vnc lang=en_US.UTF-8 keymap=us" which eliminiates the need to enter anything before it starts X on a vnc server. But it doesn't really address the original problem.
I spent last Friday trying to get it to do an HVM install, but that seemed to be very much a Xen problem. It wouldn't accept keyboard input or find its emulated devices properly until I added "acpi=off", or set "acpi=0" in the Xen config. But even then it refused to see my HD image, even though the BIOS could list it.
I got stuck at that point, and am now (successfully) doing a PV install (with the above workaround).
J
So, there's a meta-point here: we currently 'require' Beta releases to boot as guests on Xen hosts:
"The release must boot successfully as a virtual guest in a situation where the virtual host is running a supported Xen implementation"
I really don't have much knowledge of Xen and haven't followed this discussion closely, but do any currently-known bugs prevent this? If so, please flag them up so they can be considered as Beta blockers...thanks!
I filled Bug 740378 - F16: Can't use keyboard when installing F16-Alpha under Xen (regression) as guest which is pretty descripting what is below.
Also adding in Jeremy's workaround in it.
Besides that, there is also 738085 - Patch to reduce spurious Xen entries in grub menu
which has a patch to fix the grub2 menu-thingy..
Somehow the xen-kbdfront driver is not included in the initrd image (I think) - and we end with anaconda but can't type anything. I've been trying to figure out how to inject said module in the install initrd to see if that is really the problem but running in roadblocks (like xz 5.1.1alpha or 5.0.3 complains about corrupt image, or I've no idea how to make driver disks).
I saw the same thing and worked around it by booting with "vnc lang=en_US.UTF-8 keymap=us" which eliminiates the need to enter anything before it starts X on a vnc server. But it doesn't really address the original problem.
I spent last Friday trying to get it to do an HVM install, but that seemed to be very much a Xen problem. It wouldn't accept keyboard input or find its emulated devices properly until I added "acpi=off", or set "acpi=0" in the Xen config. But even then it refused to see my HD image, even though the BIOS could list it.
I got stuck at that point, and am now (successfully) doing a PV install (with the above workaround).
On 09/21/2011 01:03 PM, Konrad Rzeszutek Wilk wrote:
So, there's a meta-point here: we currently 'require' Beta releases to boot as guests on Xen hosts:
"The release must boot successfully as a virtual guest in a situation where the virtual host is running a supported Xen implementation"
I really don't have much knowledge of Xen and haven't followed this discussion closely, but do any currently-known bugs prevent this? If so, please flag them up so they can be considered as Beta blockers...thanks!
I filled Bug 740378 - F16: Can't use keyboard when installing F16-Alpha under Xen (regression) as guest which is pretty descripting what is below.
Also adding in Jeremy's workaround in it.
Though I couldn't repro the general problems I was having - I later did a clean F14 install with no problems. So I don't really know what's going on here; I might try another F16 hvm install to see how it goes.
But there is a bona-fide bug that F16 doesn't include xen-platform-pci by default in its initramfs, so it ends up unplugging its emulated devices without discovering the PV ones to replace them...
Besides that, there is also 738085 - Patch to reduce spurious Xen entries in grub menu
which has a patch to fix the grub2 menu-thingy..
I'd noticed that - though at present I haven't managed to get F16 Xen to boot in an hvm domain (it hangs when setting up interrupts in the nested dom0).
J
On Wed, Sep 21, 2011 at 03:46:31PM -0700, Jeremy Fitzhardinge wrote:
On 09/21/2011 01:03 PM, Konrad Rzeszutek Wilk wrote:
So, there's a meta-point here: we currently 'require' Beta releases to boot as guests on Xen hosts:
"The release must boot successfully as a virtual guest in a situation where the virtual host is running a supported Xen implementation"
I really don't have much knowledge of Xen and haven't followed this discussion closely, but do any currently-known bugs prevent this? If so, please flag them up so they can be considered as Beta blockers...thanks!
I filled Bug 740378 - F16: Can't use keyboard when installing F16-Alpha under Xen (regression) as guest which is pretty descripting what is below.
Also adding in Jeremy's workaround in it.
Though I couldn't repro the general problems I was having - I later did a clean F14 install with no problems. So I don't really know what's going on here; I might try another F16 hvm install to see how it goes.
But there is a bona-fide bug that F16 doesn't include xen-platform-pci by default in its initramfs, so it ends up unplugging its emulated devices without discovering the PV ones to replace them...
Can you open a BZ at bugzilla.redhat.com please?
It might need to be assigned to the kernel team.
Besides that, there is also 738085 - Patch to reduce spurious Xen entries in grub menu
which has a patch to fix the grub2 menu-thingy..
I'd noticed that - though at present I haven't managed to get F16 Xen to boot in an hvm domain (it hangs when setting up interrupts in the nested dom0).
J
Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Wed, Sep 21, 2011 at 06:59:57PM -0400, Konrad Rzeszutek Wilk wrote:
On Wed, Sep 21, 2011 at 03:46:31PM -0700, Jeremy Fitzhardinge wrote:
On 09/21/2011 01:03 PM, Konrad Rzeszutek Wilk wrote:
So, there's a meta-point here: we currently 'require' Beta releases to boot as guests on Xen hosts:
"The release must boot successfully as a virtual guest in a situation where the virtual host is running a supported Xen implementation"
I really don't have much knowledge of Xen and haven't followed this discussion closely, but do any currently-known bugs prevent this? If so, please flag them up so they can be considered as Beta blockers...thanks!
I filled Bug 740378 - F16: Can't use keyboard when installing F16-Alpha under Xen (regression) as guest which is pretty descripting what is below.
Also adding in Jeremy's workaround in it.
Though I couldn't repro the general problems I was having - I later did a clean F14 install with no problems. So I don't really know what's going on here; I might try another F16 hvm install to see how it goes.
But there is a bona-fide bug that F16 doesn't include xen-platform-pci by default in its initramfs, so it ends up unplugging its emulated devices without discovering the PV ones to replace them...
Can you open a BZ at bugzilla.redhat.com please?
Also should we change the default xen-platform-pci to =y in upstream Linux .config to avoid having problems with every distro ?
-- Pasi
It might need to be assigned to the kernel team.
Besides that, there is also 738085 - Patch to reduce spurious Xen entries in grub menu
which has a patch to fix the grub2 menu-thingy..
I'd noticed that - though at present I haven't managed to get F16 Xen to boot in an hvm domain (it hangs when setting up interrupts in the nested dom0).
J
Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Thu, Sep 22, 2011 at 12:47:57PM +0300, Pasi Kärkkäinen wrote:
On Wed, Sep 21, 2011 at 06:59:57PM -0400, Konrad Rzeszutek Wilk wrote:
On Wed, Sep 21, 2011 at 03:46:31PM -0700, Jeremy Fitzhardinge wrote:
On 09/21/2011 01:03 PM, Konrad Rzeszutek Wilk wrote:
> So, there's a meta-point here: we currently 'require' Beta releases to > boot as guests on Xen hosts: > > "The release must boot successfully as a virtual guest in a situation > where the virtual host is running a supported Xen implementation" > > I really don't have much knowledge of Xen and haven't followed this > discussion closely, but do any currently-known bugs prevent this? If so, > please flag them up so they can be considered as Beta blockers...thanks!
I filled Bug 740378 - F16: Can't use keyboard when installing F16-Alpha under Xen (regression) as guest which is pretty descripting what is below.
Also adding in Jeremy's workaround in it.
Though I couldn't repro the general problems I was having - I later did a clean F14 install with no problems. So I don't really know what's going on here; I might try another F16 hvm install to see how it goes.
But there is a bona-fide bug that F16 doesn't include xen-platform-pci by default in its initramfs, so it ends up unplugging its emulated devices without discovering the PV ones to replace them...
Can you open a BZ at bugzilla.redhat.com please?
Also should we change the default xen-platform-pci to =y in upstream Linux .config to avoid having problems with every distro ?
Or some form of it. Stefano is working to provide a patch that will latch on CONFIG_PVONHVM and make that work.
On Thu, Sep 22, 2011 at 03:08:41PM -0400, Konrad Rzeszutek Wilk wrote:
On Thu, Sep 22, 2011 at 12:47:57PM +0300, Pasi Kärkkäinen wrote:
On Wed, Sep 21, 2011 at 06:59:57PM -0400, Konrad Rzeszutek Wilk wrote:
On Wed, Sep 21, 2011 at 03:46:31PM -0700, Jeremy Fitzhardinge wrote:
On 09/21/2011 01:03 PM, Konrad Rzeszutek Wilk wrote:
>> So, there's a meta-point here: we currently 'require' Beta releases to >> boot as guests on Xen hosts: >> >> "The release must boot successfully as a virtual guest in a situation >> where the virtual host is running a supported Xen implementation" >> >> I really don't have much knowledge of Xen and haven't followed this >> discussion closely, but do any currently-known bugs prevent this? If so, >> please flag them up so they can be considered as Beta blockers...thanks!
I filled Bug 740378 - F16: Can't use keyboard when installing F16-Alpha under Xen (regression) as guest which is pretty descripting what is below.
Also adding in Jeremy's workaround in it.
Though I couldn't repro the general problems I was having - I later did a clean F14 install with no problems. So I don't really know what's going on here; I might try another F16 hvm install to see how it goes.
But there is a bona-fide bug that F16 doesn't include xen-platform-pci by default in its initramfs, so it ends up unplugging its emulated devices without discovering the PV ones to replace them...
Can you open a BZ at bugzilla.redhat.com please?
Also should we change the default xen-platform-pci to =y in upstream Linux .config to avoid having problems with every distro ?
Or some form of it. Stefano is working to provide a patch that will latch on CONFIG_PVONHVM and make that work.
Ok, great!
-- Pasi
On Wed, Sep 21, 2011 at 04:03:49PM -0400, Konrad Rzeszutek Wilk wrote:
So, there's a meta-point here: we currently 'require' Beta releases to boot as guests on Xen hosts:
"The release must boot successfully as a virtual guest in a situation where the virtual host is running a supported Xen implementation"
I really don't have much knowledge of Xen and haven't followed this discussion closely, but do any currently-known bugs prevent this? If so, please flag them up so they can be considered as Beta blockers...thanks!
I filled Bug 740378 - F16: Can't use keyboard when installing F16-Alpha under Xen (regression) as guest which is pretty descripting what is below.
Also adding in Jeremy's workaround in it.
Besides that, there is also 738085 - Patch to reduce spurious Xen entries in grub menu
which has a patch to fix the grub2 menu-thingy..
I also filed the F16 Xen pvfb problem on rhel5 dom0:
"Fedora 16 beta Xen pvfb graphical console does not work on rhel5 dom0": https://bugzilla.redhat.com/show_bug.cgi?id=740657
-- Pasi
xen@lists.stg.fedoraproject.org