dev/virtio stuff

[Bryan Venteicher]

Hi,

----- Original Message -----
> From: "Sean Bruno" <seanbru at yahoo-inc.com>
> To: "Bryan Venteicher" <bryanv at daemoninthecloset.org>
> Cc: sbruno at freebsd.org, freebsd-virtualization at freebsd.org
> Sent: Wednesday, October 31, 2012 12:33:11 PM
> Subject: Re: dev/virtio stuff
> 
> See below.
> 
> On Mon, 2012-10-29 at 22:51 -0700, Bryan Venteicher wrote:
> > Hi,
> > 
> > ----- Original Message -----
> > > From: "Sean Bruno" <seanbru at yahoo-inc.com>
> > > To: freebsd-virtualization at freebsd.org
> > > Sent: Monday, October 29, 2012 4:49:42 PM
> > > Subject: dev/virtio stuff
> > > 
> > > So ... I see we have virtio nowish.  I've started my hackery to
> > > make
> > > some man(4) pages for this, but have some questions.  Doing some
> > > raw
> > > QEMU things, I can't quite see how to get the virtio blk and net
> > > devices working.
> > > 
> > 
> > There are already virtio(4) man pages, but it looks like they
> > didn't
> > get MFC to 9.
> 
> Want me to shovel them back?
> 

Sure. That would be great.

> > 
> > > I've loaded the virtio drivers as modules.  Is this not going to
> > > work in
> > > this case?  I don't see the virtio versions of disk devices in
> > > /dev
> > > so I'm not sure what the problem here is.
> > > 
> > > http://people.freebsd.org/~sbruno/virtio.txt
> > 
> > >> FreeBSD 7.5-YAHOO-20121028 #0 ybsd_9 at 324991: Sun Oct 28 22:08:46
> > >> UTC 2012
> > >>    kernel at yahoo-inc.com:/sys/amd64/compile/YAHOO amd64
> > 
> > I assume this is 9-STABLE'ish?
> > 
> Yah, its pretty close to stable/9 +- 1 or to commits.
> 
> > >> virtio_pci0: <VirtIO PCI Block adapter> port 0xc200-0xc23f irq
> > >> 11 at device 4.0 on pci0
> > >> vtblk0: <VirtIO Block Adapter> on virtio_pci0
> > >> virtio_pci0: host features: 0x41000014
> > >> <NotifyOnEmpty,DiskGeometry,MaxNumSegs>
> > >> virtio_pci0: negotiated features: 0x14 <DiskGeometry,MaxNumSegs>
> > >> vtblk0: 4096MB (8388608 512 byte sectors)
> > 
> > The VirtIO block gets probed and attached correctly, and right
> > after the
> > last printf, hands off to disk_create(9) to create the GEOM
> > plumbing and
> > device. VirtIO block devices get created as vtbdX, not as ad, ada,
> > or da.
> > 
> > Can you see if g_disk_create() gets called (with flag==EV_CANCEL)?
> > What does a verbose boot show?
> > 
> > Bryan
> 
> Grabbed a verbose dmesg for you here:
> 
> http://people.freebsd.org/~sbruno/viritio_dmesg.txt
> 

>> GEOM: new disk vtbd0

So it seems the VirtIO block device got created just fine.

> Sean
> 
>