On Wed, Jun 29, 2011 at 9:06 PM, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > On Wed, 29 Jun 2011, Jassi Brar wrote: > >> >> IMHO Gadget drivers need not be specified at _compile_ time if they >> >> could be used over high speed UDC. The gadget drivers should simply >> >> support HS descriptors too and decide at runtime if they need to use them. >> > >> > Why? >> > >> > If you're building an embedded system with limited memory, and you know >> > the UDC on that system isn't capable of running at high speed, why >> > should the gadget driver include HS descriptors? >> > >> Because >> a) We can(do already?) tag descriptors __initdata > > Not in all the function drivers (I'm thinking especially of > f_sourcesink and f_loopback, but there are others). Let's fix those up > first. Of course. I never said we can simply drop the config symbol. > >> b) Esp when we talk about enabling gadget stack to play nice for embedded linux >> distributions, I doubt if any distribution would compile any gadget >> driver without >> CONFIG_USB_GADGET_DUALSPEED. > > I wasn't referring to distributions but to individual projects. > Call me heretic, but I must admit I am kinda starting to stop fretting about 'limited' embedded system resources :) > When all the gadget and function drivers have been converted to use > __initdata for their descriptors, I'll support getting rid of > CONFIG_USB_GADGET_DUALSPEED. Fair enough. And I am in no hurry :) I meant such things, and maybe some more as we discover, when I suggested ... "All gadget drivers ensure defer preparing descriptors and deciding their own supported feature-set until all info of UDC has been gotten." Thanks, Jassi -- 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
