Re: [Xen-users] Trouble booting FreeBSD i386 PV DomU

[Cherry G. Mathew <cherry@xxxxxx>]

>>>>> "Roger" == Roger Pau Monnà <roger.pau@xxxxxxxxxx> writes:

    Roger> On 31/03/13 19:45, Cherry G.Mathew wrote:
    >> 
    >> On 21/02/13 10:18, tech mailinglists wrote:
    >>> Hello all,
    >>> 
    >>> I have created a FreeBSD PV DomU image formatted with ZFS. I
    >>> compiled FreeBSD with KERNCONF=XEN for the kernel and the normal
    >>> world and distribution target. Then I transfered it to a Debian
    >>> Dom0 with Xen 4.2.1.
    >>> 
    >>> I tried to boot the image befor I migrate it to LVM and I now
    >>> get teh following output:
    >>> 
    >> 
    >> [...]
    >> 
    >>> run_interrupt_driven_hooks: still waiting after 300 seconds for
    >>> xenbus_free_evtchn panic: run_interrupt_driven_config_hooks:
    >>> waited too long cpuid = 0 KDB: enter: panic [ thread pid 0 tid
    >>> 100000 ] Stopped at kdb_enter+0x3a: movl $0,kdb_why
    db> 
    >> 
    >> 
    >> Try see if patch below should help the boot a bit along.

    Roger> The problem is already solved, it was crashing because
    Roger> blkback in Dom0 was not correctly working -- nothing to do
    Roger> with FreeBSD. The 9.x releases work correctly, the problem is
    Roger> with HEAD and the backtrace I've posted. 

I'd be keen to know why this is not an issue with other OSs (I run a
NetBSD-pv on a xen linux dom0. For FreeBSD amd64 PV, I seem to be
getting xenstore events, and the panic is caused by interrupt boot time
hooks not being unhooked by the xenbus driver, because it waits for
child devices on xenbus to finish their "attach". This is a bit
redundant afaict, because the hooks are called together in the order of
registration - the registrations themselves being in the autoconf device
tree traversal order - and for the xen drivers, I don't see any
interrupt hook specific dependencies in the children attaching to xenbus
(I may be wrong).

Anyway, that's what the patch does.


    Roger> BTW, if you wish to try HEAD you will need the ELFNOTE patch
    Roger> I've posted and boot with "hw.mca.enabled=0" to get to the
    Roger> panic I've posted earlier.

Thanks, it's one of the first things I had to fix for amd64 PV on
clang. There's a clock related issue with time offset calculation that
is clang specific and is a bit bothersome too.

Cheers,
-- 
Cherry

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users