Re: [Qemu-devel] Regression: virtio-pci: convert to ioeventfd callbacks

[Cornelia Huck]

On Tue, 28 Jun 2016 08:22:05 +0200
Peter Lieven <address@hidden> wrote:

> Am 27.06.2016 um 17:09 schrieb Cornelia Huck:
> > On Mon, 27 Jun 2016 11:44:47 +0200
> > Peter Lieven <address@hidden> wrote:
> >
> >> Hi, with the above patch applied:
> >>
> >> commit 9f06e71a567ba5ee8b727e65a2d5347fd331d2aa
> >> Author: Cornelia Huck <address@hidden>
> >> Date:   Fri Jun 10 11:04:12 2016 +0200
> >>
> >>       virtio-pci: convert to ioeventfd callbacks
> >>
> >> a Ubuntu 14.04 VM freezes at startup when blk-mq is set up - even if there 
> >> is only one queue.
> > OK, I played around a bit with Stefan's block branch rebased upon
> > current master (and the conflict fixed up)... and I can't seem to hit
> > any error with virtio-ccw (will see if I find time to play around with
> > virtio-pci; but unless I've messed up, this should be the same code
> > path). VIRTIO_BLK_F_MQ is correctly negotiated if num_queues > 1, and I
> > can use the disk.
> >
> > So, two questions:
> > - Which branch are you using? Any chance of a mis-merge there?
> > - Can you share the interesting portion of your command line?
> >
> 
> Hi Cornelia,
> 
> sorry I should have been a little more precise. I am using current master for 
> git.qemu.org.
> 
> I tried around and I can only reproduce the regression with iSCSI *and* 
> dataplane enabled.
> Here is the reproducing commandline for me.
> 
> ./x86_64-softmmu/qemu-system-x86_64 -enable-kvm -nodefaults -drive 
> if=none,format=raw,file=iscsi://172.21.200.56/iqn.2001-05.com.equallogic:0-8a0906-98f384e0a-7d2004ee0a85767a-00lieven-test/0,cache=none,aio=native,id=disk0
>  -object iothread,id=iothread0 
> -device virtio-blk-pci,drive=disk0,iothread=iothread0 -vga vmware -monitor 
> stdio

Nothing really exciting there, maybe a timing issue triggered by the
iscsi backing?

> 
> Have you the possiblity to install Ubuntu 14.04 on an iSCSI target?

No, I don't have any iscsi target available.

> 
> Let me know if you need further information. Maybe the regression is an 
> incompatiblity between the iSCSI driver and ioeventfd?

Running in gdb and looking at the backchains may be helpful. Or
checking in a guest dump whether there's any hang there instead. (Does
this happen with guests other than Ubuntu 14.04, maybe with a similar
kernel level?)