ASL kernel panic on Xen guest

breun · **Posted:** Tue Dec 06, 2011 9:58 am

I'm having trouble getting the ASL kernel to start on a particular Xen guest. I get the following on the console:

Quote:

switchroot: mount failed: No such file or directory
Kernel panic - not syncing: Attempted to kill init!
Pid: 1, comm: init Not tainted 2.6.32.43-6.art.x86_64 #1
Call Trace:
[<ffffffff814918d9>] panic+0x7a/0x139
[<ffffffff8105e8ba>] ? exit_ptrace+0xb3/0x133
[<ffffffff81055eed>] do_exit+0x89/0x722
[<ffffffff8100363d>] ? xen_mc_issue+0x2e/0x32
[<ffffffff81056602>] do_group_exit+0x7c/0xa5
[<ffffffff81056642>] sys_exit+0x0/0x17
[<ffffffff81009e2b>] system_call_fastpath+0x16/0x1b
[<ffffffff81009dc9>] ? system_call_after_swapgs+0x19/0x65
[<ffffffff81009dc9>] ? system_call_after_swapgs+0x19/0x65

Any idea what's wrong here?

scott · **Posted:** Tue Dec 06, 2011 4:46 pm

xen is a hot mess. There are at least 3 different forks (that I know of so far) for the paravirt implementation.

The only flavour I could even remotely help you with is the 3tera fork of xen.

breun · **Posted:** Tue Dec 06, 2011 5:26 pm

The strange thing is that we already manage 10 other Xen guests at the same provider that are all running the ASL kernel just fine. I'll ask them if there's anything different about the cluster this one is running on.

scott · **Posted:** Tue Dec 06, 2011 10:02 pm

As I said its an inconsistent mess. The xen flavor/version is going to matter for anything you're doing. For example if its xen + branch (...this part is the mess part, there are several forks) + version less than 4 then its not going to support eventfd() functionality. If you put a kernel in that environment that does, you've got a recipe for difficult to debug failures. Like inotify() will "work" just throw back entirely arcane errors, because the underlying xen does not, nginx will start it just wont serve any pages, things like that.

Also if you get a chance Id love to see how it performs against the tests I set up in the benchmark thread. I ran some against the 3tera systems and it looked like it was doing about 25% of the capacity of the comparable hypervisor image. But I didn't really have the time to do anything extensive.

breun · **Posted:** Fri Dec 09, 2011 7:52 pm

The problem turned out to be a missing module in the initrd. After adding the following lines to /etc/modprobe.conf and regenerating the initrd the ASL kernel booted successfully:

Code:

alias eth0 xennet
alias scsi_hostadapter xenblk

ASL kernel panic on Xen guest

Who is online