On Sat, 23 Jun 2018, Benjamin Herrenschmidt wrote: > Hi folks ! > > I noticed some UDC drivers setup a "desc" pointer for ep0.desc for the > driver, some don't. > > Is it officially needed ? No, it isn't. The only "core" routine that requires a desc pointer in the endpoint structure is usb_ep_enable, and that routine explicitly says it may not be called for ep0. > Additionally some UCDs NULL out the desc pointer in ep_disable, is that > also a requirement ? No. A disabled endpoint doesn't have any requirements on its descriptor or lack thereof. > I somewhat fear that if somebody calls stuff like usb_ep_align() on ep0 > it will crash without a valid desc pointer... That would be a pretty strange thing to do. I'd say it's not worth worrying about until somebody makes that mistake. > That leads me to wonder, should we sprinkle null checks (and maybe > WARN_ON_ONCE) on some of those accessors to catch those cases ? I think segmentation faults will do a good job of catching them. :-) Alan Stern > I notices a few reports (and I'm about to debug one with my own UDC, > aspeed vhub when I'm back at work on Monday, with HID) where we get a > crash with a NULL dereference and while I don't know for sure yet > that's the problem, the fact that EP0 has a NULL desc with most UDCs > and that we clear the desc on EP disablement somewhat tickles me the > wrong way... > > Cheers, > Ben. > > > -- 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
