On 28/02/2021 10.33, Andy Shevchenko wrote:

On Sun, Feb 28, 2021 at 11:29 AM Andy Shevchenko
[off-list ref] wrote:

quoted

On Sun, Feb 28, 2021 at 11:07 AM Andy Shevchenko
[off-list ref] wrote:

quoted

On Friday, February 26, 2021, Rasmus Villemoes [off-list ref] wrote:

quoted

The only purpose of this driver is to serve as a consumer of the input
clock, to prevent it from being disabled by clk_disable_unused().

We have a clock API to do the same (something like marking it used or so) why do you need a driver?

Example:
https://elixir.bootlin.com/linux/latest/source/drivers/platform/x86/pmc_atom.c#L365

If it's a DT based platform I think you can make it somehow work thru DT.

Okay, briefly looking at the state of affairs [1] seems like you need
to hack it into clock provider.

[1]: https://elixir.bootlin.com/linux/latest/C/ident/CLK_IS_CRITICAL

I did find CLK_IS_CRITICAL and CLK_IGNORE_UNUSED while trying to figure
out how to handle this. However, while CLK_IS_CRITICAL is in principle
settable via DT, the comment above of_clk_detect_critical() seems to
make it clear that adding a call of that from the RTC driver is a total
no-no.

CLK_IGNORE_UNUSED can't be set at all from DT, and wouldn't solve the
problem fully - while we can and do make sure the bootloader sets the
appropriate bit in the RTCs registers, it's more robust if we also
ensure the kernel explicitly enables the clock.

But if there is some way to do this within the clk framework/existing
bindings, I'm all ears - that's the reason I cc'ed the clk list.

Rasmus