Hi Stephen,
On 10/5/2019 4:50 AM, Stephen Boyd wrote:
Quoting Taniya Das (2019-10-04 10:39:31)
Hi Stephen,
On 10/3/2019 9:31 PM, Stephen Boyd wrote:
Quoting Taniya Das (2019-10-03 03:31:15)
Hi Stephen,
On 10/1/2019 8:08 PM, Stephen Boyd wrote:
Why do you want to keep them critical and registered? I'm suggesting
that any clk that is marked critical and doesn't have a parent should
instead become a register write in probe to turn the clk on.
Sure, let me do a one-time enable from probe for the clocks which
doesn't have a parent.
But I would now have to educate the clients of these clocks to remove
using them.
If anyone is using these clks we can return NULL from the provider for
the specifier so that we indicate there isn't support for them in the
kernel. At least I hope that code path still works given all the recent
changes to clk_get().
Could you please confirm if you are referring to update the below?
I wasn't suggesting that explicitly but sure. Something like this would
be necessary to make clk_get() pass back a NULL pointer to the caller.
Does everything keep working with this change?
Even if I pass back NULL, I don't see it working. Please suggest how to
take it forward.
--
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation.
--
