Re: [Xen-devel] Tapdisk2 problem

[Daniel Stodden <daniel.stodden@xxxxxxxxxx>]

On Wed, 2010-08-04 at 10:04 -0400, LuÃs Silva wrote:
> Hello,
> 
> I'm using both xen 4.0.1-rc5 and the latest pv_ops dom0 kernel (branch
> xen/stable-2.6.32.x), but I'm having some problems with tapdisk2...
> 
> Basically, if I create a tap device (tapdisk2 -n
> aio:/var/lib/libvirt/images/srv-001-ub1004.img), when I try to
> terminate it by issuing "echo 1 > /sys/class/blktap2/blktap0/remove",
> the last command hangs and tapdisk2 hangs using 100% cpu.
> 
> This happens both on manual device destruction and on virtual machine
> shutdown. I don't know exactly what triggered this behaviour, but I
> suspect kernel changes.
> 
> Can anyone confirm this behaviour?

Yes, with a broken userland. I thought Ian's dot-private patches to
ring.h made it into testing?

changeset:   21707:feee0abed6aa
user:        Keir Fraser <keir.fraser@xxxxxxxxxx>
date:        Fri Jul 02 18:58:02 2010 +0100
summary:     blktap2: make protocol specific usage of shared sring explicit

The problem is that the kernel has those changes. What's going to happen
is that if tapdisk2 anticipates the ring 'message' at offset 17, not 16
as intended, it will fail to find/clear the ring message, while the
kernel will keep it spinning through select(), and that's what you're
seeing.

Keir, okay to just pull that stuff?

Daniel


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