On Fri, Sep 11, 2020 at 2:28 PM Alexandre Belloni <alexandre.bell...@bootlin.com> wrote: > > On 07/09/2020 11:34:59+0200, Bartosz Golaszewski wrote: > > On Fri, Sep 4, 2020 at 5:41 PM Alexandre Belloni > > <alexandre.bell...@bootlin.com> wrote: > > > > > > On 04/09/2020 17:21:15+0200, Bartosz Golaszewski wrote: > > > > From: Bartosz Golaszewski <bgolaszew...@baylibre.com> > > > > > > > > Align the arguments passed to devm_rtc_device_register() with the upper > > > > line. > > > > > > > > Signed-off-by: Bartosz Golaszewski <bgolaszew...@baylibre.com> > > > > --- > > > > drivers/rtc/rtc-rx8010.c | 2 +- > > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > > > diff --git a/drivers/rtc/rtc-rx8010.c b/drivers/rtc/rtc-rx8010.c > > > > index 181fc21cefa8..ed8ba38b4991 100644 > > > > --- a/drivers/rtc/rtc-rx8010.c > > > > +++ b/drivers/rtc/rtc-rx8010.c > > > > @@ -450,7 +450,7 @@ static int rx8010_probe(struct i2c_client *client, > > > > } > > > > > > > > rx8010->rtc = devm_rtc_device_register(&client->dev, client->name, > > > > - &rx8010_rtc_ops, THIS_MODULE); > > > > + &rx8010_rtc_ops, > > > > THIS_MODULE); > > > > > > > > > > You have bonus points if you replace that patch by switching from > > > devm_rtc_device_register to devm_rtc_allocate_device and > > > rtc_register_device. > > > > > > More bonus points if you also set range_min and range_max and then get > > > rid of the range checking in set_time. > > > > > > > Hi Alexandre! > > > > I've just looked at the code and wondered why there's no devm > > counterpart for rtc_register_device(). Then I noticed that the release > > callback for devm_rtc_allocate_device() takes care of unregistering > > the device. This looks like serious devres abuse to me. In general the > > idea is for the release callback to only undo whatever the devres > > function did and this should be opaque to the concerned resources. > > > > In this case I believe there's no need for the 'registered' field in > > struct rtc_device - this structure should *not* care about this - and > > there should be devm_rtc_register_device() whose release callback > > would take care of the unregistering. Since this function would be > > called after devm_rtc_allocate_device(), it would be released before > > so the ordering should be fine. > > > > Note that the input subsystem is also doing it that way which is > probably not a good reason alone to do it like that. But, IIRC, there
I'm seeing this pattern elsewhere in the kernel too and I just recently fixed this for MDIO. I think it's just a matter of people copy-pasting a bad implementation. > was an actual reason this was done this way and it was the ordering of > the rtc_nvmem_register/rtc_nvmem_unregister with rtc_device_unregister. > I'm not sure this is still necessary though. > To me - each of these should have their own 'unregister' function and appropriate devres helpers *OR* RTC-related nvmem structures could be set up and assigned to struct rtc_device after devm_rtc_allocate_device() and picked up by the registration function (and also undone by rtc_unregister_device()). I'll try to allocate some time to look into this but it's not like it's urgent or anything - it's just a potential improvement. Bartosz
