On Fri, 30 Sep 2011, Felipe Balbi wrote: > Hi, > > On Fri, Sep 30, 2011 at 10:24:20PM +0300, Felipe Balbi wrote: > > This is actually a good point. We can't simply roundup request length on > > UDC driver because we don't know if gadget driver has allocated enough > > space should host send more data. > > > > OTOH, if we don't roundup request->lenght, driver won't work *always* on > > SuperSpeed. > > > > From DWC3 point this patch should be acceptable, but there could be > > another controller with a different quirk where DMA wouldn't complete > > unless we pass correct size, dunno. > > > > Alan ? I don't have any particular objection beyond the fact that Paul's patch isn't complete (compare it to the one I just posted). As for other controller quirks, that's a little more complex. If we ever do encounter hardware that doesn't work when the actual packet size is smaller than the requested size, we'd have to take some pretty drastic action to support it. A quirk like that is liable to affect every gadget driver. Somehow the gadget drivers would have to tell the controller driver whether or not it's permissible to fail a transfer merely because the packet size was smaller than expected. If not, the transfer would have to use PIO rather than DMA. > I just remembered, though, that MSC has no short packets, that's why we > set short_not_ok flag. So, is this really necessary ? I'll run some > extra tests on monday. Remember, Paul wasn't talking about short packets -- he was talking about packets that aren't a multiple of the maxpacket value. It's not the same thing. Alan Stern -- 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
