Re: [PATCH] clk: add gpio gated clock
From: Mike Turquette
Date: Fri Sep 26 2014 - 19:54:24 EST
Quoting Jyri Sarha (2014-09-11 01:44:24)
> On 09/10/2014 01:14 AM, Mike Turquette wrote:
> > Quoting Jyri Sarha (2014-09-05 05:21:34)
> >> The added gpio-gate-clock is a basic clock that can be enabled and
> >> disabled trough a gpio output. The DT binding document for the clock
> >> is also added. For EPROBE_DEFER handling the registering of the clock
> >> has to be delayed until of_clk_get() call time.
> >> Signed-off-by: Jyri Sarha <jsarha@xxxxxx>
> >> ---
> >> This is my final attempt to get this generic gpio controlled basic
> >> clock into mainline. Of course I gladly fix any issues that the patch
> >> may have. However, if there is no response, I give up and move it to TI
> >> specific clocks.
> > I searched through my archives and found a post from January. You Cc'd
> > me as "<mturquette@xxxxxxxxxx>". Note that the address is wrapped in
> > chevrons but there is no name string (e.g. "Mike Turquette").
> > My mailer doesn't parse this well it was not flagged as to:me in my
> > filters. Maybe other mailers handle this better? If you leave out the
> > name string in the future then it would probably be best to drop the
> > chevrons.
> Then git send-email adds the chevrons, but in the future I'll put the
> name string there too.
> >> I've been sending this patch as a part of Beaglebone-Black HDMI audio
> >> patch series since last autumn. Since the previous version I have done
> >> some minor cleanups and changed the clock's compatible property from
> >> "gpio-clock" to "gpio-gate-clock". All the file names, comments,
> >> etc. have also been changed accordingly.
> > Is your platform the only one to take advantage of this clock type so
> > far? I feel that it is esoteric enough that it shouldn't be made
> > generic.
> > The main reason is that all of the generic clock types needs to be
> > overhauled at some point. E.g. the clk-gate should have its
> > machine-specific logic separated from its machine-independent logic. If
> > the gate clock were to populate .enable and .disable callbacks and then
> > leave the actual register banging, or regmap'ing, or gpio'ing up to your
> > backend driver then that would be a big improvement and would avoid the
> > need to create this new clock type outright.
> > So that's on my todo list, but it's not done yet. For your patch I think
> > that putting this code into drivers/clk/ti would probably be best,
> > unless other folks could use it as-is. Even if others could use it today
> > I would want to remove it eventually for the reasons stated in the
> > paragraph above.
> Ok, I see. I do not know of anybody else needing a gpio gate clock at
> the moment. I'll put the driver under drivers/clk/ti unless someone
> comes forward soon.
Well nobody came forward but after thinking about it I've seen this
design elsewhere, so it should probably be generic. And the underlying
machine-specific ops are less relevant to this type since most of that
is abstracted away behind the GPIO api.
Applied to clk-next. Let me know if that causes a problem for you if you
have merged this into the TI clk stuff.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/