[Pkg-xen-devel] Bug#657014: Bug#657014: Bug#657014: Wheezy, Squeeze and Ubuntu server 11.10 has missing boot option to Xen

[Erik Hjelmås]

On 01/24/2012 01:31 PM, Pasi Kärkkäinen wrote:
> On Mon, Jan 23, 2012 at 06:26:29PM +0100, Erik Hjelmås wrote:
>>>> When installing xen-linux-system on a new Dell R810 server, Wheezy,
>>>> Squeeze and Ubuntu server 11.10 all boot fine when booted directly into
>>>> the standard kernel, but when booting through Xen (after installing the
>>>> package xen-linux-system) the boot process hangs as soon as Xen passes
>>>> control to the Dom0 kernel:
>>>>
>>>> "Gave up waiting for root device" etc
>>>>
>>>> and it gives me Busybox, but it is also frozen, so there's no other
>>>> option that reboot.
>>>
>>> This might be the same issue as #649923. But please could you provide
>>> full console logs so we can verify. If you are able to try the patch in
>>> that bug or perhaps a backported 4.1 hypervisor that would also be
>>> potentially interesting.
>>
>> its the same behaviour on both squeeze (4.0) and wheezy (4.1)
>>
>> since everything freezes on boot, I dont have any logs other than the
>> attached screenshot at the time it freezes (or is something logged this
>> early in the boot process by Xen? I can reboot without Xen and inspect
>> other logs)
>>
>
> Please set up a serial console so you can capture the full boot messages
> from both Xen hypervisor and dom0 kernel.
>
> In your case you can also use the SOL (Serial Over LAN),
> provided by the iDRAC management processor.
>
> http://wiki.xen.org/xenwiki/XenSerialConsole

sorry, this wasnt straight forward and Im running out of time, but I 
captured a video showing the full boot process
http://www.ansatt.hig.no/erikh/capture-1.avi

(if you really need the text output I can try setup SOL but that will 
have to be next week)

>>>> After several days of troubleshooting it turns out that adding dom0_mem
>>>> option (e.g. dom0_mem=8192M) to the multiboot line of
>>>> /etc/grub.d/20_linux_xen FIXES THE PROBLEM !
>>>>
>>>> maybe this has to be fixed in the package?
>>>
>>> That file is provided by grub, not the hypervisor but I don't think that
>>> fix will work since a) really it is a workaround not a fix and b) it is
>>> not really possible to determine what is the right number to use for any
>>> given system.
>>
>> I agree, my solution is a workaround and not at fix
>>
>> I first suspected that there was a problem with the initramfs so I added
>> the megasas driver to /etc/initramfs/modules and updated initramfs. This
>> resulted in a "Cannot allocate memory" error (when trying to load the
>> megasas driver) at the same stage in the boot process. And since Busybox
>> is not able to run either, Xen seems to not give any memory available to
>> the Dom0 kernel when it passes control to it in the boot process.
>> Meaning the Dom0 kernel fails immediately since I guess accessing the
>> root devices is one of the first things it tries to do.
>>
>> I will attempt install squeeze and wheezy on a different older server
>> (Dell R900), during the next couple of days to see of the same error
>> pops up, and Ill look for more log data then ,
>>
>
> I assume you have already installed all the latest BIOS/firmware updates etc?

no, I havent updated any firmware, so that maybe something I should try

at least the BIOS seems to be in the most recent version, 2.4.4
(the server is brand new btw)

but I see that there might be firmware for the RAID controller,

/Erik