On Mon, Jul 02, 2012 at 08:42:53PM -0400, Alan Stern wrote: > On Mon, 2 Jul 2012, Sarah Sharp wrote: > > > Hmm, so this is for an Etron host controller. We found this issue on > > the Fresco Logic host controller, and I'm not sure if the behavior is > > the same, since they're from different vendors. > > > > What devices trigger the warning? Do you notice anything odd with those > > devices? > > > > Do you have a USB headset with a speaker to test with? If you notice > > robotic-sounding audio when you record from the headset, then we need to > > trust the untransferred length, not the completion status. > > Maybe I asked this before, I can't remember... > > Is there any good reason not to trust the untransferred length always? > Isn't that just as valid as using the completion status (and evidently > more reliable in several cases)? I'm not sure if I can always trust the untransferred length. Early xHCI prototypes used to get the length wrong, but the status right. In particular, the untransferred length could be some very large number for a babble transfer, because the hardware attempted to subtract the number of received bytes from the buffer size. That did get fixed eventually, but in the meantime I reworked the ring code to trust the completion code more. So I'm just trying to be cautious here. If it turns out to be common to a third host controller type, then yes, I will consider just trusting the untransferred length. Sarah Sharp -- 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
