----- "Gerd Hoffmann" <kra...@redhat.com> wrote: > On 10/10/10 13:12, Alon Levy wrote: > > > > ----- "Anthony Liguori"<anth...@codemonkey.ws> wrote: > > > >> How is this different than usb_del? Is it that it detaches it but > >> does > >> not delete the device? > > > > yes. There is no usb_attach command because it was harder to write > (can't > > use the bus.addr since a detached device doesn't have them) and I > didn't > > need it right now, my device attaches itself based on a external > event. > > Which points out a problem with this patch: It should better not use > > bus.addr. addr isn't fixed and even can be uninitialized. Yes, > usb_del > uses it (for historical reasons). But we better should not use it in > > new code. Better use the device id (like device_del). Which will > work > for usb_attach too. > > Next question: What is the use case? attach/detach is used by > devices > internally. usb-host does attach/detach when devices get plugged-in > and > -out on the host. The ccid device does simliar things on vsclient > connect/disconnect. So toggeling the attach state via monitor easily >
debugging. naturally when developing the ccid I had cases where I'd rather detach the device then bring down qemu. since there is no way currently to add/remove chardev's from monitor, removing/adding a device is not enough to reset a device state to the state right after start. > could have unwanted side effects ... > > cheers, > Gerd
