usb in hvm domU

usb in hvm domU

[Henning Sprang]

Hi,
After asked on the users list first but getting no success or failure
report of anybody using this feature, I just tested the usb access in
hvm as documeted in
http://www.cl.cam.ac.uk/research/srg/netos/xen/readmes/user/user.html#SECTION04344000000000000000

For me, this is quite broken: When I boot a VM with such a parameter,
it takes three times as long to boot, then, I can enter and lsusb
command, and see that the device seems to be there, but from then on,
the VM behaves very strange, not accepting keybord intput correctly
anymore, and I got now 3 times trapped in an vnc window, no way to
exit nicely, all I could do was going to the text console with
ctrl-alt-F1, login and kill vncviewer, then I could use x windows
again.

Similar with SDL - when entering the window by clicking on it, I was
trapped in there, and the system kept booting as very slow as with
vnc. Again, no way to get out, but ctrl-alt-F1, ctrl-alt-F7.

At least for me, this feature does not work very well. It whould
probably be documented as "experimental" in the manual.

Besides that, I was wondering how I could add multiple device with
this configuration syntax. As there are mouse problems in hvm with vnc
graphics, which go away when using usbdevice="tablet",
how can I use for example an usb stick at the same time? I tried
usbdevice='tablet,host:vid:pid', but that doesn't work (with correct
vid and pid values - fir the tests above I used only these and they
lead to me seing the correct device in lsusb, but the problems
described above).
Didn't try yet to use a python list, currently fighting with the other
problems...

Henning

Re: usb in hvm domU

[Henning Sprang]

On 2/12/07, Henning Sprang <henning_sprang@gmx.de> wrote:
> Hi,
> After asked on the users list first but getting no success or failure
> report of anybody using this feature, I just tested the usb access in
> hvm as documeted in
> http://www.cl.cam.ac.uk/research/srg/netos/xen/readmes/user/user.html#SECTION04344000000000000000
>

Hmm, I just read the fine print again, and saw the part "support for
high speed USB 2.0 devices is still under development" - can that be
the reason for the troubles described?
I don't know now how to find out if my memstick is 1.0 or 2.0, it's
not the newest one, but not older than two years, so it can well be a
2.0 stick.

Henning

Re: Re: usb in hvm domU

[Keir Fraser]

On 12/2/07 01:25, "Henning Sprang" <henning_sprang@gmx.de> wrote:

> On 2/12/07, Henning Sprang <henning_sprang@gmx.de> wrote:
>> Hi,
>> After asked on the users list first but getting no success or failure
>> report of anybody using this feature, I just tested the usb access in
>> hvm as documeted in
>> http://www.cl.cam.ac.uk/research/srg/netos/xen/readmes/user/user.html#SECTION
>> 04344000000000000000
>> 
> 
> Hmm, I just read the fine print again, and saw the part "support for
> high speed USB 2.0 devices is still under development" - can that be
> the reason for the troubles described?
> I don't know now how to find out if my memstick is 1.0 or 2.0, it's
> not the newest one, but not older than two years, so it can well be a
> 2.0 stick.

It would make it slower to access, but shouldn't noticeably slow down your
boot time. It sounds more like there are issues with USB pass-through.

 -- Keir

Re: Re: usb in hvm domU

[Henning Sprang]

On 2/12/07, Keir Fraser <Keir.Fraser@cl.cam.ac.uk> wrote:
> It would make it slower to access,

Like when using an usb 1.0 stick on an usb 2.0 port, or the other way round?

> but shouldn't noticeably slow down your
> boot time. It sounds more like there are issues with USB pass-through.

It seems, it's not just a bit "slowed down" but basically unusable.

I'm going to file a bug then. Do you need additional information? (can
take a while until I can further gather this info, I'm  a bit too
loaded currently).

Henning

Re: Re: usb in hvm domU

[Keir Fraser]

On 15/2/07 20:28, "Henning Sprang" <henning_sprang@gmx.de> wrote:

>> It would make it slower to access,
> 
> Like when using an usb 1.0 stick on an usb 2.0 port, or the other way round?

I'm not sure you can tell the difference, but 2.0 device on a 1.0 port I
suppose.

>> but shouldn't noticeably slow down your
>> boot time. It sounds more like there are issues with USB pass-through.
> 
> It seems, it's not just a bit "slowed down" but basically unusable.
> 
> I'm going to file a bug then. Do you need additional information? (can
> take a while until I can further gather this info, I'm  a bit too
> loaded currently).

Just what you tried to do and the resulting problem is probably fine for
now. We'd ask for more info if it is difficult to reproduce.

It may also be worth trying to repro with plain qemu (most recent possible
version). It may well be an issue we have in common with upstream qemu.

 -- Keir

Re: Re: usb in hvm domU

[Henning Sprang]

On 2/16/07, Keir Fraser <Keir.Fraser@cl.cam.ac.uk> wrote:
> It may also be worth trying to repro with plain qemu (most recent possible
> version). It may well be an issue we have in common with upstream qemu.

I don#t come to test this right now.

I added the problem with usb in hvm in genral a bug #906, and the
problem that it's either not possible or not documented how to use
multiple usb devices as bug #907

Henning