[virt-tools-list] [PATCH 0/2] add USB redirection support in virt-manager

Marc-André Lureau mlureau at redhat.com
Wed Jun 26 16:12:41 UTC 2013



----- Mensaje original -----
> Hi,
> 
> On 06/26/2013 05:16 PM, Guannan Ren wrote:
> > On 06/26/2013 03:51 PM, Hans de Goede wrote:
> >> Hi,
> >>
> >> On 06/25/2013 04:39 PM, Guannan Ren wrote:
> >>> On 06/25/2013 10:18 PM, Cole Robinson wrote:
> >>>> On 06/25/2013 09:58 AM, Guannan Ren wrote:
> >>>>> On 06/25/2013 01:52 AM, Hans de Goede wrote:
> >>>>>> Hi,
> >>>>>>
> >>>>>> On 06/24/2013 12:11 PM, Guannan Ren wrote:
> >>>>>>> This two patch use spicy-client python binding to add USB redirection
> >>>>>>> support in virt-manager if guests use spice viewer.
> >>>>>> Very cool, thanks for working on this. I've done a quick review,
> >>>>>> mostly
> >>>>>> from the usbredir pov rather then a code pov, and there is one thing
> >>>>>> missing.
> >>>>>>
> >>>>>> Now that virt-manager will have a usb-device selection dialog, it
> >>>>>> should probably also enable usb autoredirection by setting the
> >>>>>> auto-connect property of the UsbDeviceManager to true, this will
> >>>>>> enable
> >>>>>> the "If the virt-viewer window has the current focus and I insert a
> >>>>>> USB
> >>>>>> device, it will be automatically redirected to the guest" behavior
> >>>>>> Leonardo is talking about.
> >>>>>>
> >>>>>> I see that Cole is not necessarily a fan of it. Cole, you should give
> >>>>>> this a try, most users love it. In virt-viewer we just always enable
> >>>>>> it without any complaints. Alternatively it could be made an option
> >>>>>> but I would default it to true.
> >>>>>>
> >>>>> I tried it several times, for virt-viewer, auto-usbredir property is
> >>>>> set to
> >>>>> TRUE by default, so
> >>>>> it supports auto-rediret by default. it works well for a single guest.
> >>>>> When there are multiple virt-viewer instance opened for multiple
> >>>>> guests, the
> >>>>> auto usbredirection
> >>>>> becomes a little confusing, sometimes it is redirected to guest A,
> >>>>> sometimes
> >>>>> it goes to guest B
> >>>>> between I re-plugin the usb devices.
> >>>>>
> >>>>> It is a really good feature. for virt-manager, multiple viewers more
> >>>>> often
> >>>>> stay opened for guests
> >>>>> than virt-viewer, so it possibly cause some confusion for user. So I
> >>>>> think
> >>>>> using *disable* to this
> >>>>> feature by default probably is better. we can give an option to enable
> >>>>> it
> >>>>> somehow.
> >>>>>
> >>>> Hmm, maybe we could conditionally enable the property only when the
> >>>> spice
> >>>> widget has focus. So it's only turned on for one console at a time.
> >>>>
> >>>> - Cole
> >>>>
> >>>
> >>>      Currently it works so, the spice widget which has focus gets the usb
> >>>      device finally.
> >>>      Even so,  it is till confusing unless we document it somewhere,
> >>>      otherwise, the user
> >>>      will be curious and try to figure out what is going on about this
> >>>      magic behaviours.
> >>
> >> Right, the has focus check Cole is suggestion is already done in the
> >> spice-gtk widget,
> >> auto-redirecting usb-stick (ie) just because a virt-viewer window is
> >> running minimized
> >> somewhere is not exactly user-friendly.
> >>
> >> Guannan, how is the current behavior confusing to you ? Unless you switch
> >> windows right
> >> at the moment you plug in a usb device, the usb-device will always show up
> >> in the guest
> >> you're working with, since that is the one which window has the keyboard
> >> focus.
> >
> > I added tooltips to it for giving more info and enable auto-redir it by
> > default which I think it is clearer.
> > I encountered an two gtk-client library issues.  I just paste it there. I
> > can file bugs if necessary.
> > First:
> > when I call  SpiceClientGLib.Channel.string_to_type('usbredir').
> > error is reported: Could not locate spice_channel_string_to_type:
> > `spice_channel_string_to_type': /lib64/libspice-client-glib-2.0.so.8:
> > undefined symbol: spice_channel_string_to_type
> > It seems the symbol is local rather than GLOBAL
> > # readelf -s /lib64/libspice-client-glib-2.0.so |grep spice_channel_string
> > 3275: 000000000001e910   159 FUNC    LOCAL  DEFAULT   12
> > spice_channel_string_to_t
> >
> 
> spice-gtk bug, I just send a patch to the spice-devel list for this.

We never explicitely made it public (it's not documented either), but I guess it's time to do it, now that we have the counterpart public.

Btw, why did you need it? Aren't the spice protocol constant in the bindings?




More information about the virt-tools-list mailing list