On 2022-02-28 15:46, Mark Brown wrote:
On Mon, Feb 28, 2022 at 03:27:08PM +0000, Robin Murphy wrote:
On 2022-02-28 15:11, Mark Brown wrote:
Well, it could also be that the clock is wired to some other clock which
is always on (which I guess is why the driver allows this in the first
place, there's a lot of sloppy code around stuff like that in the tree).
I wouldn't say the driver "allows" it, so much as it just blindly grabs the
first clock assuming it's SSPCLK per the binding, and thus it will happen to
work out if the underlying physical clock is the same as, or equivalent to,
the APB PCLK. Otherwise, it's already into some degree of not working
properly, by virtue of reading the wrong clock rate.
Ah, the APB clock requirement is inherited from the AMBA implementation
isn't it? We really ought to be extending an AMBA binding here...
Yup, both the "apb_pclk" clock specifier and the "arm,primecell"
compatible technically belong to the common AMBA binding, but I'm not
sure whether schema has the ability to compose at such fine-grained a
level :/
Robin.
