Re: [PATCH v5 09/12] clk: actions: Add fixed factor clock support
From: Manivannan Sadhasivam <hidden>
Date: 2018-03-20 09:04:13
Also in:
linux-arm-kernel, linux-clk, lkml
Hi Stephen, On Mon, Mar 19, 2018 at 06:10:03PM -0700, Stephen Boyd wrote:
Quoting Manivannan Sadhasivam (2018-03-17 03:09:49)quoted
Add support for Actions Semi fixed factor clock together with helper functions to be used in composite clock. Signed-off-by: Manivannan Sadhasivam <redacted> --- drivers/clk/actions/Makefile | 1 + drivers/clk/actions/owl-fixed-factor.c | 81 ++++++++++++++++++++++++++++++++++ drivers/clk/actions/owl-fixed-factor.h | 62 ++++++++++++++++++++++++++ 3 files changed, 144 insertions(+) create mode 100644 drivers/clk/actions/owl-fixed-factor.c create mode 100644 drivers/clk/actions/owl-fixed-factor.hdiff --git a/drivers/clk/actions/Makefile b/drivers/clk/actions/Makefile index 994357fa560b..b618696ba54e 100644 --- a/drivers/clk/actions/Makefile +++ b/drivers/clk/actions/Makefile@@ -5,3 +5,4 @@ clk-owl-y += owl-gate.o clk-owl-y += owl-mux.o clk-owl-y += owl-divider.o clk-owl-y += owl-factor.o +clk-owl-y += owl-fixed-factor.odiff --git a/drivers/clk/actions/owl-fixed-factor.c b/drivers/clk/actions/owl-fixed-factor.c new file mode 100644 index 000000000000..f1281565129c --- /dev/null +++ b/drivers/clk/actions/owl-fixed-factor.c@@ -0,0 +1,81 @@ +// SPDX-License-Identifier: GPL-2.0+ +// +// OWL fixed factor clock driver +// +// Copyright (c) 2014 Actions Semi Inc. +// Author: David Liu <liuwei@actions-semi.com> +// +// Copyright (c) 2018 Linaro Ltd. +// Author: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> + +#include <linux/clk-provider.h> +#include <linux/regmap.h> +#include <linux/slab.h> + +#include "owl-fixed-factor.h" + +long owl_fix_fact_helper_round_rate(struct owl_clk_common *common, + const struct owl_fix_fact_hw *fix_fact_hw, + unsigned long rate, + unsigned long *parent_rate) +{ + if (clk_hw_get_flags(&common->hw) & CLK_SET_RATE_PARENT) { + unsigned long best_parent; + + best_parent = (rate / fix_fact_hw->mul) * fix_fact_hw->div; + *parent_rate = clk_hw_round_rate(clk_hw_get_parent(&common->hw), + best_parent); + } + + return (*parent_rate / fix_fact_hw->div) * fix_fact_hw->mul; +} + +static long owl_fix_fact_round_rate(struct clk_hw *hw, unsigned long rate, + unsigned long *parent_rate) +{ + struct owl_fix_fact *fix_fact = hw_to_owl_fix_fact(hw); + struct owl_fix_fact_hw *fix_fact_hw = &fix_fact->fix_fact_hw; + + return owl_fix_fact_helper_round_rate(&fix_fact->common, fix_fact_hw, + rate, parent_rate); +} + +unsigned long owl_fix_fact_helper_recalc_rate(struct owl_clk_common *common, + const struct owl_fix_fact_hw *fix_fact_hw, + unsigned long parent_rate) +{ + unsigned long long int rate; + + rate = (unsigned long long int)parent_rate * fix_fact_hw->mul; + do_div(rate, fix_fact_hw->div); + + return (unsigned long)rate;You can drop the cast.
Okay.
quoted
+} + +static unsigned long owl_fix_fact_recalc_rate(struct clk_hw *hw, + unsigned long parent_rate) +{ + struct owl_fix_fact *fix_fact = hw_to_owl_fix_fact(hw); + struct owl_fix_fact_hw *fix_fact_hw = &fix_fact->fix_fact_hw; + + return owl_fix_fact_helper_recalc_rate(&fix_fact->common, fix_fact_hw, + parent_rate); +} + +static int owl_fix_fact_set_rate(struct clk_hw *hw, unsigned long rate, + unsigned long parent_rate) +{ + /* + * We must report success but we can do so unconditionally because + * clk_fix_fact_round_rate returns values that ensure this call is aWhat function is that?
It should be owl_fix_fact_round_rate.
quoted
+ * nop. + */ + + return 0; +} + +const struct clk_ops owl_fix_fact_ops = { + .round_rate = owl_fix_fact_round_rate, + .recalc_rate = owl_fix_fact_recalc_rate, + .set_rate = owl_fix_fact_set_rate, +};Why can't you use the regular fixed factor clk code and ops?
That's going to be really messy. Since I'm having the clk_hw embedded inside owl_clk_common and using it for registering all the clocks, using generic fixed factor functions will be a different approach _only_ for this clock and it won't look good I guess. Also, it may become complicated with composite clocks. If you still want to use the generic fixed factor code, I can do that in next revision. Your views? Thanks, Mani