On Wednesday 19 December 2007, Alan Stern wrote: > I don't know how the Freescale controller is supposed to work. > Apparently it doesn't use a companion controller for full-speed > connections. Maybe it has a Transaction Translator in its root hub > instead.
Yes. ARC originally sold that IP design, then it went to TDI, and now another company sells the licences. It doesn't actually need a TT in the root hub, but that's how it looks to the driver. What it does is send out full speed (or low speed) signaling directly from root hub ports, rather than high speed signaling with special split transaction support. > If that's the case, then the TT isn't working right. It could be a > problem with the controller hardware or a problem with the driver; I > have no idea which. Near as I can tell it's a combination of things. The TT scheduler probably doesn't handle that situation correctly. But there seems to be a second problem related to correct-but-suboptimal handling for bandwidth allocation. Basically, quantization: scheduling is split into eight microframes, and a pair of transactions that could fit into seven of them is treated as if they can't because one takes just short of six (counted as six) and the other takes just over one (counted as two) ... which adds up to more than seven, meaning only one of the transactions gets allowed. - Dave - To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
