On Tue, Aug 09, 2011 at 05:23:14PM +0800, David Chang wrote: > Hi Greg, > > 2011/6/28 Németh Márton <nm127@xxxxxxxxxxx>: > > From: Márton Németh <nm127@xxxxxxxxxxx> > > > > USBIP v1.0.0 protocol documentation. > > > > Signed-off-by: Márton Németh <nm127@xxxxxxxxxxx> > > --- > > > > Hi, > > > > I tried to document the USBIP protocol as implemented in the Linux kernel 3.0-rc2. > > > > The description is a preliminary draft only, it may contain mistakes. I tried > > to document what I have understand from the source code and from the actual > > captured network traffic when the USBIP is in action. Please review it, correct it, > > point out the missing parts. > > > > During I have documented the protocol a question came into my mind: why do we > > have two different type of package when the URB completition handler is called, > > namely USB_RET_SUBMIT and USBIP_RET_UNLINK? As far as I can see this causes race > > condition because for one URB the completition handler is only called once. If > > the one URB is sent in with USB_CMD_SUBMIT and then unlinked with USB_CMD_UNLINK > > the completition handler is called only once. In the protocol, however, we have > > two different type of packets: USB_RET_SUBMIT and USBIP_RET_UNLINK. The status > > field of these messages may contain anything in this case depending on the timing. > > > > Regards, > > > > Márton Németh > > Sorry, just one question about the usbip protocol document. > I would like to know why didn't you apply this patch. Because it was sent as a "RFC" (i.e. request for comments) and based on the discussion afterward, there were lots of comments :) If someone wants to resend it to me, without that marking, I'll be glad to apply it. thanks, greg k-h _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
