On Thu, 14 Oct 2010, Jurgen Braam wrote:
> Hi,
>
> > On the other hand, clear_tt_buffer_complete is never called if the
> > driver isn't EHCI. Consequently there's no real need for a NULL check.
> > If you send in a patch removing that check, I'll ACK it. (Provided you
> > also doublecheck all the other in-tree drivers to make sure the pointer
> > entry is present.)
>
> Possibly something like usb-serial.c's set_to_generic_if_null() could be
> used here when (type == EHCI) to prevent future omissions:
>
> #define set_to_generic_if_null(type, function) \
> do { \
> if (!type->function) { \
> type->function = usb_serial_generic_##function; \
> dbg("Had to override the " #function \
> " usb serial operation with the generic one.");\
> } \
> } while (0)
>
> But then in this case perhaps a dev_err() would be more appropriate?
>
> The omission of clear_tt_buffer_complete and endpoint_reset resulted in
> usb-serial devices like ftdi_sio only working correctly the first time
> they were being opened. More info in my msg "[PATCH][RFC] USB: EHCI: add
> missing {endpoint_reset,clear_tt_buffer_complete} operations" posted
> earlier.
If you can figure out how to make this work, go ahead. Bear in mind
that usbcore isn't allowed to modify the contents of an hc_driver
structure, since it is declared "const". Also bear in mind that
usbcore can't contain any pointers to ehci_clear_tt_buffer_complete,
since that function belongs to a lower-level module. Finally, remember
that there is no reliable way for usbcore to tell whether or not a
particular host controller driver is an EHCI driver.
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
