On Tue, 19 Feb 2013, Stefan Tauner wrote: > On Tue, 19 Feb 2013 13:27:49 -0500 (EST) > Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > > > > Adding a completely new API that returns a timestamp associated with > > > the start/end/fixed offset of a frame number sounds like a perfect > > > solution for my problem. I am not sure what you mean with completion > > > (interrupt). AFAICS that term is used mainly/only in relation to > > > (signaling) completed URBs. How does that fit together? What can I do > > > to make it happen? > > > > Yes, completion interrupts are used for signalling when an URB > > completes. That's most of the signalling the CPU gets from a host > > controller; other things like port connect and disconnect are > > comparatively rare. > > Hm but there are also other potential interrupt sources at least for > some HCD types that could be used, e.g. OHCI_INTR_SF at the start of > frames in OHCI or the Frame List Rollover interrupt in EHCI? Yes. (Although the Frame List Rollover interrupt fires at a rather low frequency.) Don't forget -- if this ever does manage to get into the kernel, it will be necessary to modify more than just the EHCI and OHCI drivers. The kernel contains a whole bunch of host controller drivers, and every one would need to be updated. > The drawback of using other interrupt sources like OHCI_INTR_SF would > be of course that they... interrupt, a lot :) > Maybe the gathering through normally unused interrupts could be > enabled/disabled with ioctls (and while that is disabled the value is > only updated on completed URBs)? Sure, if you want to make things even more complicated... > What kind of interface do you propose? I would say a single file in > sysfs with one line frame counter and one line timestamp, but the sysfs > docs clearly don't encourage mixing types like this, if I understand > it right. Of course we need to make sure that the corresponding values > can be retrieved safely in sync... In fact, I think this interface is so unlikely to be used by more than a few people that I propose not adding it at all. :-) For your own purposes, of course, you can do whatever you want. > Do you know about a comparable API I could look at? Nothing other than the ones you're already aware of. 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
