On Thu, Apr 14, 2022 at 09:07:29AM +0200, Nuno Sá wrote: > On Wed, 2022-04-13 at 19:58 +0300, Andy Shevchenko wrote: > > On Wed, Apr 13, 2022 at 03:38:47PM +0000, Sa, Nuno wrote: > > > > From: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> > > > > Sent: Wednesday, April 13, 2022 4:41 PM > > > > > > The extended clocks are optional and may not be present for some > > > > SoCs > > > > supported by this driver. Nevertheless, in case the clock is > > > > provided > > > > but some error happens during its getting, that error should be > > > > handled > > > > properly. Use devm_clk_get_optional() API for that. Also report > > > > possible > > > > errors using dev_err_probe() to handle properly -EPROBE_DEFER > > > > error. > > > > > This is a nice cleanup patch... But the subject might be a bit > > > misleading as it says "Fix". So I would expect a Fixes tag which > > > I'm not sure it's really worth it here. Yes, the code was pretty > > > much > > > doing clk_get_optional() "by hand" but I think it was still > > > functional. > > > So to me, this is more an improvement rather than a fix... > > > > Actually it is a fix, but not critical since no-one complains aloud > > so far. > > The problematic part is logs exhausting if repetitive deferred probe > > happens. > > Still not really agree with it... In the commit message you state that > errors are not properly handled and so let's use > 'devm_clk_get_optional()'. I don't think that is true because If im not > missing nothing there's no fundamental change between the previous code > and using 'devm_clk_get_optional()'. So to me this is an enhancement > because we were doing something "by hand" when we have an API for it. > > That said, introducing dev_err_probe() indeed stops possibly annoying > error messages for EPROBE_DEFER (and that could be seen as a fix, not > really devm_clk_get_optional()). I honestly still don't see it as fix > but we are also not adding a Fixes tag so I don't really care :). > > (But I still think the commit message is a bit misleading) Yes, the commit message can be amended. I won't split these two since the fix part is not critical (nobody so far complained aloud about it). That's why I prefer to set the main point of the change as switching to devm_clk_get_optional() than deferred probe messages. -- With Best Regards, Andy Shevchenko
