Re: [Qemu-devel] [PATCH] Add USB sys file-system support (v2)

[TJ <linux@tjworld.net>]

On Fri, 2008-09-05 at 14:19 -0500, Jason Wessel wrote:
> > What version of kvm, or what was the commit HEAD hash?
> >   
> 
> Since you posed this to the qemu-devel mailing list, I tried it on the
> QEMU development head.

I've posted to qemu-devel and kvm-devel, working primary with the kvm
release packaging it for Ubuntu Gutsy/Hardy/Intrepid.

> 
> > What does the patch report as the method of access to the USB
> > file-system? 
> husb: using /dev/bus/usb
> 
> > Does the same device work without the patch? 
> 
> Of course not, because the host does not have /proc/bus/usb.

If you're using Ubuntu you can easily test with /proc/bus/usb by editing
the comments out to enable the 'magic lines' in 
/etc/init.d/mountdevsubfs.sh or issue them manually:

#
# Magic to make /proc/bus/usb work
#
#mkdir -p /dev/bus/usb/.usbfs
#domount usbfs "" /dev/bus/usb/.usbfs -obusmode=0700,devmode=0600,listmode=0644
#ln -s .usbfs/devices /dev/bus/usb/devices
#mount --rbind /dev/bus/usb /proc/bus/usb


> What might be a defect is that it did not work with the
> |host:vendor_id:product_id and immediately failed.

Yes, I've been looking at that. Although the code-path goes through the
patch - usb_host_scan() - it simply passed the structures and function
pointer on without touching them. I'm currently analysing it with some
dprintf() entries to see why usb_host_find_device() doesn't get the
device.

> I am guessing it is unique to this class of device or perhaps even this
> particular device.

I agree, maybe it is the serial character-based devices. As I can
reproduce it with a different VID:PID of the same class it seems more
likely.