On 12/05/2021 22:02:17+0200, Christophe JAILLET wrote:

quoted

The only error path that will not print a message by default (it is
dev_dbg) is when rtc-ops is NULL which I don't expect would regress
anyway.

A better way to remove the dead code would be to switch to
devm_rtc_allocate_device/devm_rtc_register_device.

I don't follow you here.
Isn't devm_rtc_device_register = devm_rtc_allocate_device +
devm_rtc_register_device?

What would be the benefit for switch to the latter?

The immediate benefit is that this solve a possible but very unlikely
race condition around the character device removal when probe ultimately
fails. The other benefit is that I won't have to do it later to handle
the modern features.

quoted

And even better would
be to take that opportunity to set range_min and range_max ;)

Maybe, but this goes beyond my knowledge.
I'll let someone else propose a patch for it.

CJ

-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com