Re: [Xen-users] Xen 4.1.1 DomU Partition table disappearing

[Hans de Bruin <jmdebruin@xxxxxxxxx>]

On 07/05/2011 04:32 AM, Liwei wrote:
> > Message: 4
> > Date: Sun, 03 Jul 2011 23:06:16 +0200
> > From: Hans de Bruin<jmdebruin@xxxxxxxxx>
> > Subject: [Xen-users] Xen 4.1.1 DomU Partition table disappearing
> > To: xen-users@xxxxxxxxxxxxxxxxxxx
> > Message-ID:<4E10D9C8.3060204@xxxxxxxxx>
> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> >
> > I have just installed a new xen server: xen 4.1.1, 64bit
> > xen/stable-2.6.32.x + USB pass throug patch, Atom D510.
> >
> > The first DomU I created is running fine, until I shut it down. At the
> > moment the guest disappears from xentop, the partitiontable gets wiped
> > out. I have checked this by running  watch 'fdisk -l /dev/vg4/temp' on Dom0
> ----snip----
>
> Hello,
>      Don't have a solution here, but I have seen the same problem on
> one of my DomUs. Maybe if we compare our setups we'd find the cause.
>      1. Does your DomU run OpenSuSE (11.4)?
64 bit slackware 1337

>      2. Is your DomU PV or HVM?
My atom is not capable of hvm

>      3. If HVM, are you running PV on HVM drivers?
>      4. Are you using LVM for your VM images?
yes, the volume group is on top of a linux software mirror

>      5. Are you using LVM in your VM?
no

>      6. Have you tried using a separate distribution?
not yet

>      7. Have you tried creating another DomU to see if it exhibits the
> same problem?
not yet

>      8. Have you tried a different Dom0 kernel/xen?
not yet

>      Basically I remember this problem came up with one of my
> xen/Dom0/DomU combinations, but didn't have the chance to find out the
> root cause before moving on to another combination.
I have done some tests using only the setup initrd from my slackware64 
dvd. I filled another lvm block device with a lot of FF's, and put it in 
front in the vm's disk config. so the troubled partition table is on 
block device /dev/xvdb. After shutdown the FF's on the first disk where 
all in tact, the partition table on de second disk was destroyed. So 
much for a workaround.
So I went back to the normal configuration and tried to narrow down the 
exact time of destruction. To see the ammount of damage I filled the 
first 1000 512 byte blocks with FF's and recreated the parition table. 
There are two primary partitions on the disk. 1G xvda1 for swap and 
xvda2 for the rest of the 16GB disk. What clears the partition table is:
mount /dev/xvda2 /mnt

some modification of /mnt like creating a directoy, or accessing a 
directory for the first time that day.
And then the killer: sync

So now there is a question number 9. have I tried differed file systems?

not yet

Some of the damage that is done:

root@luna:/home/hans# hexdump -n 8000 /dev/vg4/temp
0000000 3bc0 9839 0000 0200 0000 2001 0000 0000
0000010 0000 0000 0000 0000 0000 0000 0000 0000
*
0000030 0000 0000 134e 9024 8336 8f46 0000 0000
0000040 0000 0000 0000 0000 0000 0000 0000 0000
*
0001000 ffff ffff ffff ffff ffff ffff ffff ffff
*

root@luna:/home/hans# hexdump -n 8000 /dev/vg4/temp
0000000 3bc0 9839 0000 0200 0000 2501 0000 0000
0000010 0000 0000 0000 0000 0000 0000 0000 0000
*
0000030 0000 0000 134e 6b28 2b2a f6fc 0000 0000
0000040 0000 0000 0000 0000 0000 0000 0000 0000
*
0001000 ffff ffff ffff ffff ffff ffff ffff ffff
*
0001f40

So the first 0x01000 bytes get overwritten. That is the size of one 
memory page. It does not look like a ext4 supper block.
--
Hans






_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users