On Sun, Aug 18, 2013 at 10:44 PM, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > On Sun, 18 Aug 2013, Ming Lei wrote: > >> > As far as I can see, these don't need to disable interrupts. All they >> > protect against is the code in usb_sg_wait() and usb_sg_cancel(), which >> > both run in process context. >> >> Yes. >> >> > But will lockdep complain if they don't disable interrupts? >> >> Looks lockdep won't complain because the lock can't be held in >> another hardirq context. > > Don't be so sure. Suppose you have two mass-storage devices, one > connected by EHCI and one connected by UHCI. The one using UHCI _will_ > invoke the completion handler in hardirq context, because uhci-hcd > doesn't support tasklets. > > Have you tested this? It can't be triggered since we don't enable local interrupts yet before calling completion handler. Also uhci-hcd/ohci-hcd should support tasklet later, even for xhci-hcd, there is only little performance loss with tasklet. > >> As I mentioned in 00/50, the patchset is basically a mechanical >> change, so one patch can be dropped if anyone reviews and >> concludes it isn't needed. > > I'm afraid that it might be needed to keep lockdep happy, not to > prevent real problems. Right, so the patch should be kept. Thanks, -- Ming Lei -- 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
