On Thu, Feb 15, 2024 at 02:29:32PM +0100, Peter Krempa wrote:
> On Wed, Feb 14, 2024 at 18:11:15 +0100, Andrea Bolognani wrote:
> > This provides coverage for the scenario in which an attempt is
> > made to use a USB device together with the legacy USB controller.
> >
> > Note that, while the test case passes, the configuration it
> > produces doesn't actually work:
> >
> > $ qemu-system-x86_64 -M pc -usb -device usb-tablet,bus=usb.0
> > Bus 'usb.0' not found
> >
> > This is because the bus created by the legacy USB controller is
> > automatically assigned ID "usb-bus.0", but libvirt doesn't take
> > this into consideration when assigning addresses to USB devices.
> >
> > In other words, the legacy USB controller will only work as long
> > as no attempt is made to attach devices to it, which arguably
> > doesn't make for a very useful controller.
>
> Looking at the history the change of the bus name is documented in
> commit f9618633a86cc74b33f178f05154d4edbc08c0fc :
>
> usb: update docs for bus name change
>
> At some point the default usb bus name changed from 'usb.0' to
> 'usb-bus.0' (probably as part of the qom conversion). Update
> the usb documentation accordingly.
>
> Released in qemu-1.5.0. The most likely commit which did the change was
> 0d936928ef87ca1bb7b41b5b89c400c699a7691c ("convert busses to QOM")
Good spelunking. I didn't go that far back, I just checked that QEMU
4.2.0 also behaved like this.
> Thus arguably we can remove support for '-usb' at least for x86_64, but
> almost certainly for anything else too.
Spoiler: that's exactly what happens a few additional patches in ;)
--
Andrea Bolognani / Red Hat / Virtualization
_______________________________________________
Devel mailing list -- devel@xxxxxxxxxxxxxxxxx
To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxx
