On Wed, 29 Jan 2020 at 11:53, Daniel Lezcano <daniel.lezcano@xxxxxxxxxx> wrote: > > On 29/01/2020 12:41, Joel Stanley wrote: > > On Wed, 29 Jan 2020 at 11:25, Daniel Lezcano <daniel.lezcano@xxxxxxxxxx> wrote: > >> > >> On 07/11/2019 10:42, Joel Stanley wrote: > >>> In preparation for supporting the ast2600 which uses a different method > >>> to clear bits in the control register, use a callback for performing the > >>> shutdown sequence. > >>> > >>> Reviewed-by: Cédric Le Goater <clg@xxxxxxxx> > >>> Reviewed-by: Linus Walleij <linus.walleij@xxxxxxxxxx> > >>> Signed-off-by: Joel Stanley <joel@xxxxxxxxx> > >> > >> It will be cleaner if you create a struct of_device_id array where you > >> store the different variant data. > > > > I agree, and that's the path I would have taken when writing a normal > > driver. However as the timer drivers probe with the > > TIMER_OF_DECLARE/timer_probe infrastructure, we can't register our own > > .data pointer (TIMER_OF_DECLARE uses .data to store the _init > > function). > > > > Unless I'm missing something? > > I was suggesting to add the array like: > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/clocksource/timer-atmel-tcb.c#n351 > > TIMER_OF_DECLARE is used and the probe functions relies on the match array: > > match = of_match_node(atmel_tcb_of_match, node->parent); I see. So we don't use this table for probing the driver, just getting the associated data. I'm not convinced that's an improvement over what we have. If you have a strong preference I can try it out.
