[Qemu-devel] About usb passthru and speed mismatch

[Qemu-devel] About usb passthru and speed mismatch

[Lin Ma]

Hi Gerd,

Based on the current implementation of 'usb_host_open' in 
hw/usb/host-libusb.c,

When user performs usb_add, The usb device will be detached from kernel 
first, Then be checked for speed mismatch second.
If it found speed mismatch, The usb device isn't attached to guest, But 
the usb device can't be reattached to kernel either.

I'd like to write patch to add 'usb_check_attach' before detaching the 
device from kernel  _or_  add 'usb_host_attach_kernel' under 'fail:' of 
usb_host_open, (the latter doesn't make sense I think).
May I have your thoughts?


BTW, Have you missed a patch: "usb: Use usb_bus_find(-1) instead of 
usb_enabled() in usb_device_add/usb_device_del" which sending on June 
4th ? May I have your thoughts about that patch as well?


Thanks,
Lin

Re: [Qemu-devel] About usb passthru and speed mismatch

[Gerd Hoffmann]

On Do, 2015-06-11 at 23:47 +0800, Lin Ma wrote:
> Hi Gerd,
> 
> Based on the current implementation of 'usb_host_open' in 
> hw/usb/host-libusb.c,
> 
> When user performs usb_add, The usb device will be detached from kernel 
> first, Then be checked for speed mismatch second.
> If it found speed mismatch, The usb device isn't attached to guest, But 
> the usb device can't be reattached to kernel either.
> 
> I'd like to write patch to add 'usb_check_attach' before detaching the 
> device from kernel  _or_  add 'usb_host_attach_kernel' under 'fail:' of 
> usb_host_open, (the latter doesn't make sense I think).

Re-attach to the kernel when open fails (for whatever reasons) makes
perfect sense to me.

cheers,
  Gerd