On Fri, Mar 08, 2013 at 04:06:30PM +0200, Alexander Shishkin wrote: > Peter Chen <peter.chen@xxxxxxxxxxxxx> writes: > > > On Fri, Mar 08, 2013 at 12:39:04PM +0200, Alexander Shishkin wrote: > >> Peter Chen <peter.chen@xxxxxxxxxxxxx> writes: > >> > >> > Hi David > >> > >> Hi, > >> > >> > I notice at your code for hw_usb_set_address, there is a comment for it > >> > /** > >> > * This function explicitly sets the address, without the "USBADRA" (advance) > >> > * feature, which is not supported by older versions of the controller. > >> > */ > >> > >> That's mine. David's original driver did use USBADRA. I removed it > >> because it's not supported by some versions of chipidea, for example the > >> one that Marvell integrated in their kirkwood SoCs. > >> > >> > If we use USB3.0 host for CV test, we must use this bit, or the host > >> > may send GET_DESCRIPTOR before the controller set address. > >> > >> That's because we don't have the state machine for ep0 currently, > >> USBADRA has nothing to do with it. > > > > Can you explain more? As far as I know, it is the completion of set_address > > cost too much time that the host sends next command before device > > sets address. > > USBADRA doesn't magically speed up anything. It's a convenience thing > that lets you set the controller's address register value straight away > instead of deferring it till after status phase of > SET_ADDRESS. The status phase won't happen any sooner than normally. > Correct, but the problem is after the status phase finished, when the software can set address without using USBADRA? Assuming the busy system, the interrupt latecy, etc? For USB 3.0 host CV test, the host sends GET_DESCRIPTOR very quick (<500us) once it accept the status of SET_ADDRESS > Regards, > -- > Alex > -- Best Regards, Peter Chen -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
