mapping devices on boot

idella4 · Posted: Wed May 19, 2010 7:37 am Post subject: mapping devices on boot

I've got a few examples of booting going awry.] These are vms booted in xen. I hope that doesn't scare everyone off.
Once created, the booting process in vms seems to follow regular patterns. This is a case of kernel procedures /processes not working correctly.

I submit this on the basis that it could be a regular bootup doing the same thing In fact, I have had this happen and got around it accordingly.
This is a vm losing its way;

Hu · Administrator Joined: 06 Mar 2007 Posts: 23091

It is a bit distracting having most of your post in blue.

What problem do you want help solving? Your first code sample shows some problems with the initramfs failing to mount a filesystem off the Xen virtual hard disk node, but you also said that you "got around it accordingly."

It looks like your root problem with the initramfs is that the device node xvda1 exists, but no driver claims the associated device number, so you get ENODEV. It does not help that your initramfs then goes on to try actions which can only succeed if mounting root failed. It should have stopped after the first failure, to avoid cluttering the screen. I see no evidence that the initramfs started a dynamic device manager, so the device node at /dev/xvda1 is probably a static node that was inserted into the initramfs when it was built. Thus, the device node can exist regardless of whether a driver is registered to serve requests on that device.

Similarly, the failure to insert the kernel modules returns ENODEV, not ENOENT. It is not immediately clear from the kernel source what would trigger ENODEV on module insertion.

The rootfs that you successfully remounted read-write is the root of the initramfs, not the root of the Xen virtual hard drive.

In the final example, what is the uname of the affected system?

idella4 · Posted: Thu May 20, 2010 7:55 am Post subject:

Hu,

thanks, that's very good. Looks is if you can help me pin this down. And for both of us I shall revert to standard black
.
When I got around it accordingly, that's perhaps a bit off the mark. I used a gentoo kernel to run a debian lenny system
The lenny xen kernel misfired and could not boot into xen. I used the gentoo xen kernel In preparing it, a number of times on attempting to boot in, it reported on the screen could not find /lib/modules/2.6.31-xen-r10-AMD64/ modules.dep.
I looked and looked and re-checked, and was sure that the name of the folder was indeed 2.6.31-xen-r10-AMD64, or something close to that which I used, and there was a modules.dep file. I suppose I didn't get around it, I just persevered compiling, and finally it found it, and booted in.
That was for a host system.
This is for a guest. That form of resolution that I can see doesn't apply.

Hu · Administrator Joined: 06 Mar 2007 Posts: 23091

So the same kernel and initramfs work when booting a Debian Lenny system, but fail to boot Gentoo? Given that the failure occurs while still in the initramfs, that would implicate a configuration error in the Xen configuration file for the Gentoo guest.

idella4 · Posted: Fri May 21, 2010 6:31 am Post subject:

thanks Hu,

yes. I have a few kernels now that I use for booting guests. The gentoo kernel I made boots the host, but doesn't manage to boot guests. Ubuntu made a 2.6.31 xen kernel for guests. It managed to boot up guests that I installed where my gentoo kernel didn't get there. Then I made a gentoo 64, which is not really crucial to this process, but these example I an doing in a 64 gentoo. I just made a 64 version of the Ubuntu guest kernel. I have a couple of versions, some not effective so for this ofcourse only the effective one is cited.

So, this kernel 2.6.31.13-xen-domU works on a guest lenny, but not on the two in the examples. It is a 64 version of the Ubuntu kernel. I can use the ubuntu original kernel in a 32 system to boot 32 bit guests which my 32 gentoo kernel doesn't. The examples cited are a squeeze and a centos5.