clk: fractional-divider: fix up the fractional clk's jitter
Subsystems (2)
PATCH v3 1/1 clk: fractional-divider: fix up the fractional clk's jitter
Fractional dividers may have special requirements concerning numerator and denominator selection that differ from just getting the best approximation.
For example on Rockchip socs the denominator must be at least 20 times larger than the numerator to generate precise clock frequencies.
Therefore add the ability to provide custom approximation functions, approx = rockchip_fractional_special_approx;
but approx = NULL in the default case.
RK document description:
3.1.9 Fractional divider usage
To get specific frequency, clocks of I2S, SPDIF, UARTcan be generated by fractional divider. Generally you must set that denominator is 20 times larger than numerator to generate precise clock frequency. So the fractional divider applies only to generate low frequency clock like I2S, UART.
--- drivers/clk/clk-fractional-divider.c | 4 ++++ drivers/clk/rockchip/clk.c | 20 ++++++++++++++++++++ include/linux/clk-provider.h | 3 +++ 3 files changed, 27 insertions(+)
drivers/clk/clk-fractional-divider.c +4 -0
@@ -60,6 +60,9 @@ static long clk_fd_round_rate(struct clk_hw *hw, unsigned long rate,if (!rate || rate >= *parent_rate)return *parent_rate; +if (fd->approx) + fd->approx (hw, rate, parent_rate); +/* * Get rate closer to *parent_rate to guarantee there is no overflow * for m and n. In the result it will be the nearest rate left shifted
@@ -145,6 +148,7 @@ struct clk_hw *clk_hw_register_fractional_divider(struct device *dev, fd->nmask =GENMASK (nwidth -1 ,0 ) << nshift; fd->flags = clk_divider_flags; fd->lock = lock; + fd->approx =NULL ; fd->hw.init = &init; hw = &fd->hw;
drivers/clk/rockchip/clk.c +20 -0
@@ -164,6 +164,25 @@ static int rockchip_clk_frac_notifier_cb(struct notifier_block *nb,return notifier_from_errno (ret); } +/** +* fractional divider must set that denominator is 20 times larger than +* numerator to generate precise clock frequency. +*/ +void rockchip_fractional_special_approx (struct clk_hw *hw, +unsigned long rate, +unsigned long *parent_rate) +{ +struct clk_hw *p_parent; +unsigned long p_rate, p_parent_rate; + + p_rate =clk_hw_get_rate (clk_hw_get_parent (hw)); +if ((rate *20 > p_rate) && (p_rate % rate !=0 )) { + p_parent =clk_hw_get_parent (clk_hw_get_parent (hw)); + p_parent_rate =clk_hw_get_rate (p_parent); + *parent_rate = p_parent_rate; + } +} + static struct clk *rockchip_clk_register_frac_branch( struct rockchip_clk_provider *ctx, const char *name, const char *const *parent_names, u8 num_parents,
@@ -210,6 +229,7 @@ static struct clk *rockchip_clk_register_frac_branch( div->nwidth =16 ; div->nmask =GENMASK (div->nwidth -1 ,0 ) << div->nshift; div->lock = lock; + div->approx = rockchip_fractional_special_approx; div_ops = &clk_fractional_divider_ops; clk = clk_register_composite(NULL, name, parent_names, num_parents,
include/linux/clk-provider.h +3 -0
@@ -564,6 +564,9 @@ struct clk_fractional_divider {u8 nwidth;u32 nmask;u8 flags; +void (*approx)(struct clk_hw *hw, +unsigned long rate, +unsigned long *parent_rate);spinlock_t *lock; };
Signature · 1 line
1.9.1
Heiko Stuebner
maintainerHi Elaine,
sorry this took a bit longer, but I didn't manage to find the time to respond properly until now.
Am Freitag, 21. Juli 2017, 16:03:07 CEST schrieb Elaine Zhang:
Quoted 19 lines
Fractional dividers may have special requirements concerning numerator and denominator selection that differ from just getting the best approximation.
For example on Rockchip socs the denominator must be at least 20 times larger than the numerator to generate precise clock frequencies.
Therefore add the ability to provide custom approximation functions, approx =
rockchip_fractional_special_approx;
but approx = NULL in the default case.RK document description:
3.1.9 Fractional divider usage
To get specific frequency, clocks of I2S, SPDIF, UARTcan be generated by fractional divider. Generally you must set that denominator is 20 times larger than numerator to generate precise clock frequency. So the fractional divider applies only to generate low frequency clock like I2S, UART.
Quoted 6 lines
Signed-off-by Elaine Zhang <[email protected]>--- drivers/clk/clk-fractional-divider.c | 4 ++++ drivers/clk/rockchip/clk.c | 20 ++++++++++++++++++++ include/linux/clk-provider.h | 3 +++ 3 files changed, 27 insertions(+)
The code below does go into the right direction, but I'd really like to have the full approximation overrideable. The reason is that while on Rockchip SoCs it's enough to increase the parent rate, other socs might have completely different requirements when determining numerator and denominator.
So I've taken the time and modified your patch into 2 that look like I would envision it - compile-tested only, so please test :-) But essentially it only moves some things around but should be functionally equivalent to your patch.
I'll add the patches as replies to this mail.
Two more things to keep in mind for the future below:
drivers/clk/clk-fractional-divider.c struct clk_hw *clk_hw_register_fractional_divider(struct device *dev, · line 151
Quoted 5 lines
drivers/clk/clk-fractional-divider.c +1
-0@@ -145,6 +148,7 @@ struct clk_hw *clk_hw_register_fractional_divider(struct device *dev, fd->nmask =GENMASK (nwidth -1 ,0 ) << nshift; fd->flags = clk_divider_flags; fd->lock = lock; + fd->approx =NULL ;
not needed, as fd is allocated via kzalloc, so this is already NULL. Also it is nicer to spell approximation full (which I've done in my rework)
[...]
include/linux/clk-provider.h struct clk_fractional_divider { · line 567
Quoted 9 lines
include/linux/clk-provider.h +1
-0@@ -564,6 +564,9 @@ struct clk_fractional_divider {u8 nwidth;u32 nmask;u8 flags; + void (*approx)(struct clk_hw *hw,
Please keep the spacing in line with other elements. Somehow here spaces instead of a tab slipped in between void and (*appro...
Heiko
Heiko Stuebner
maintainerPATCH 1/2 clk: fractional-divider: allow overriding of approximation
From: Elaine Zhang <[email protected]>
Fractional dividers may have special requirements concerning numerator and denominator selection that differ from just getting the best approximation.
For example on Rockchip socs the denominator must be at least 20 times larger than the numerator to generate precise clock frequencies.
Therefore add the ability to provide custom approximation functions.
--- drivers/clk/clk-fractional-divider.c | 28 ++++++++++++++++++++-------- include/linux/clk-provider.h | 3 +++ 2 files changed, 23 insertions(+), 8 deletions(-)
drivers/clk/clk-fractional-divider.c +20 -8
@@ -49,16 +49,12 @@ static unsigned long clk_fd_recalc_rate(struct clk_hw *hw,return ret; } -static long clk_fd_round_rate (struct clk_hw *hw,unsigned long rate, -unsigned long *parent_rate) +static void clk_fd_general_approximation (struct clk_hw *hw,unsigned long rate, +unsigned long *parent_rate, +unsigned long *m,unsigned long *n) {struct clk_fractional_divider *fd =to_clk_fd (hw);unsigned long scale; -unsigned long m, n; -u64 ret; - -if (!rate || rate >= *parent_rate) -return *parent_rate;/* * Get rate closer to *parent_rate to guarantee there is no overflow
@@ -71,7 +67,23 @@ static long clk_fd_round_rate(struct clk_hw *hw, unsigned long rate,rational_best_approximation (rate, *parent_rate,GENMASK (fd->mwidth -1 ,0 ),GENMASK (fd->nwidth -1 ,0 ), - &m, &n); + m, n); +} + +static long clk_fd_round_rate (struct clk_hw *hw,unsigned long rate, +unsigned long *parent_rate) +{ +struct clk_fractional_divider *fd =to_clk_fd (hw); +unsigned long m, n; +u64 ret; + +if (!rate || rate >= *parent_rate) +return *parent_rate; + +if (fd->approximation) + fd->approximation (hw, rate, parent_rate, &m, &n); +else +clk_fd_general_approximation (hw, rate, parent_rate, &m, &n); ret = (u64)*parent_rate * m;do_div (ret, n);
include/linux/clk-provider.h +3 -0
@@ -565,6 +565,9 @@ struct clk_fractional_divider {u8 nwidth;u32 nmask;u8 flags; +void (*approximation)(struct clk_hw *hw, +unsigned long rate,unsigned long *parent_rate, +unsigned long *m,unsigned long *n);spinlock_t *lock; };
Signature · 1 line
2.11.0
Re: [PATCH 1/2] clk: fractional-divider: allow overriding of approximation
hi,heiko:
This a good solution.And I tested it on RK SOCs.It's work well.
Quoted 72 lines · Heiko Stuebner
From: Elaine Zhang <[email protected]>
Fractional dividers may have special requirements concerning numerator and denominator selection that differ from just getting the best approximation.
For example on Rockchip socs the denominator must be at least 20 times larger than the numerator to generate precise clock frequencies.
Therefore add the ability to provide custom approximation functions.
Signed-off-by Elaine Zhang <[email protected]>--- drivers/clk/clk-fractional-divider.c | 28 ++++++++++++++++++++-------- include/linux/clk-provider.h | 3 +++ 2 files changed, 23 insertions(+), 8 deletions(-)drivers/clk/clk-fractional-divider.c +20
-8@@ -49,16 +49,12 @@ static unsigned long clk_fd_recalc_rate(struct clk_hw *hw,return ret; } -static long clk_fd_round_rate (struct clk_hw *hw,unsigned long rate, -unsigned long *parent_rate) +static void clk_fd_general_approximation (struct clk_hw *hw,unsigned long rate, +unsigned long *parent_rate, +unsigned long *m,unsigned long *n) {struct clk_fractional_divider *fd =to_clk_fd (hw);unsigned long scale; -unsigned long m, n; -u64 ret; - -if (!rate || rate >= *parent_rate) -return *parent_rate;/* * Get rate closer to *parent_rate to guarantee there is no overflow @@ -71,7 +67,23 @@ static long clk_fd_round_rate(struct clk_hw *hw, unsigned long rate,rational_best_approximation (rate, *parent_rate,GENMASK (fd->mwidth -1 ,0 ),GENMASK (fd->nwidth -1 ,0 ), - &m, &n); + m, n); +} + +static long clk_fd_round_rate (struct clk_hw *hw,unsigned long rate, +unsigned long *parent_rate) +{ +struct clk_fractional_divider *fd =to_clk_fd (hw); +unsigned long m, n; +u64 ret; + +if (!rate || rate >= *parent_rate) +return *parent_rate; + +if (fd->approximation) + fd->approximation (hw, rate, parent_rate, &m, &n); +else +clk_fd_general_approximation (hw, rate, parent_rate, &m, &n); ret = (u64)*parent_rate * m;do_div (ret, n);include/linux/clk-provider.h +0
-0
include/linux/clk-provider.h struct clk_fractional_divider { · line 573
Quoted 9 lines
include/linux/clk-provider.h +3
-0@@ -565,6 +565,9 @@ struct clk_fractional_divider {u8 nwidth;u32 nmask;u8 flags; +void (*approximation)(struct clk_hw *hw, +unsigned long rate,unsigned long *parent_rate, +unsigned long *m,unsigned long *n);spinlock_t *lock; };
Re: [PATCH 1/2] clk: fractional-divider: allow overriding of approximation
In reply to Heiko StuebnerQuoted 6 lines · Heiko Stuebner
From: Elaine Zhang <[email protected]>
Fractional dividers may have special requirements concerning numerator and denominator selection that differ from just getting the best approximation.
Quoted 7 lines
For example on Rockchip socs the denominator must be at least 20 times larger than the numerator to generate precise clock frequencies.
Therefore add the ability to provide custom approximation functions.
Signed-off-by Elaine Zhang <[email protected]>---
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
Heiko Stuebner
maintainerRe: [PATCH 1/2] clk: fractional-divider: allow overriding of approximation
In reply to Heiko StuebnerAm Dienstag, 1. August 2017, 18:21:22 CEST schrieb Heiko Stuebner:
Quoted 6 lines
From: Elaine Zhang <[email protected]>
Fractional dividers may have special requirements concerning numerator and denominator selection that differ from just getting the best approximation.
Quoted 6 lines
For example on Rockchip socs the denominator must be at least 20 times larger than the numerator to generate precise clock frequencies.
Therefore add the ability to provide custom approximation functions.
Signed-off-by Elaine Zhang <[email protected]>
applied for 4.14 with Stephen's Ack
Heiko
Heiko Stuebner
maintainerPATCH 2/2 clk: rockchip: add special approximation to fix up fractional clk's jitter
In reply to Heiko StuebnerFrom: Elaine Zhang <[email protected]>
Quoted 1 line
From Rockchips fractional divider description:
3.1.9 Fractional divider usage
To get specific frequency, clocks of I2S, SPDIF, UARTcan be generated by fractional divider. Generally you must set that denominator is 20 times larger than numerator to generate precise clock frequency. So the fractional divider applies only to generate low frequency clock like I2S, UART.
Therefore add a special approximation function that handles this special requirement.
--- drivers/clk/rockchip/clk.c | 36 ++++++++++++++++++++++++++++++++++++ 1 file changed, 36 insertions(+)
drivers/clk/rockchip/clk.c +36 -0
@@ -29,6 +29,7 @@#include <linux/mfd/syscon.h> #include <linux/regmap.h> #include <linux/reboot.h> +#include <linux/rational.h> #include "clk.h" /**
@@ -164,6 +165,40 @@ static int rockchip_clk_frac_notifier_cb(struct notifier_block *nb,return notifier_from_errno (ret); } +/** +* fractional divider must set that denominator is 20 times larger than +* numerator to generate precise clock frequency. +*/ +void rockchip_fractional_approximation (struct clk_hw *hw, +unsigned long rate,unsigned long *parent_rate, +unsigned long *m,unsigned long *n) +{ +struct clk_fractional_divider *fd =to_clk_fd (hw); +unsigned long p_rate, p_parent_rate; +struct clk_hw *p_parent; +unsigned long scale; + + p_rate =clk_hw_get_rate (clk_hw_get_parent (hw)); +if ((rate *20 > p_rate) && (p_rate % rate !=0 )) { + p_parent =clk_hw_get_parent (clk_hw_get_parent (hw)); + p_parent_rate =clk_hw_get_rate (p_parent); + *parent_rate = p_parent_rate; + } + +/* +* Get rate closer to *parent_rate to guarantee there is no overflow +* for m and n. In the result it will be the nearest rate left shifted +* by (scale - fd->nwidth) bits. +*/ + scale =fls_long (*parent_rate / rate -1 ); +if (scale > fd->nwidth) + rate <<= scale - fd->nwidth; + +rational_best_approximation (rate, *parent_rate, +GENMASK (fd->mwidth -1 ,0 ),GENMASK (fd->nwidth -1 ,0 ), + m, n); +} + static struct clk *rockchip_clk_register_frac_branch( struct rockchip_clk_provider *ctx, const char *name, const char *const *parent_names, u8 num_parents,
@@ -210,6 +245,7 @@ static struct clk *rockchip_clk_register_frac_branch( div->nwidth =16 ; div->nmask =GENMASK (div->nwidth -1 ,0 ) << div->nshift; div->lock = lock; + div->approximation = rockchip_fractional_approximation; div_ops = &clk_fractional_divider_ops; clk = clk_register_composite(NULL, name, parent_names, num_parents,
Signature · 1 line
2.11.0
Re: [PATCH 2/2] clk: rockchip: add special approximation to fix up fractional clk's jitter
Quoted 73 lines · Heiko Stuebner
From: Elaine Zhang <[email protected]>
Quoted 1 line
From Rockchips fractional divider description:
3.1.9 Fractional divider usage
To get specific frequency, clocks of I2S, SPDIF, UARTcan be generated by fractional divider. Generally you must set that denominator is 20 times larger than numerator to generate precise clock frequency. So the fractional divider applies only to generate low frequency clock like I2S, UART.Therefore add a special approximation function that handles this special requirement.
Signed-off-by Elaine Zhang <[email protected]>--- drivers/clk/rockchip/clk.c | 36 ++++++++++++++++++++++++++++++++++++ 1 file changed, 36 insertions(+)drivers/clk/rockchip/clk.c +35
-0@@ -29,6 +29,7 @@#include <linux/mfd/syscon.h> #include <linux/regmap.h> #include <linux/reboot.h> +#include <linux/rational.h> #include "clk.h" /** @@ -164,6 +165,40 @@ static int rockchip_clk_frac_notifier_cb(struct notifier_block *nb,return notifier_from_errno (ret); } +/** +* fractional divider must set that denominator is 20 times larger than +* numerator to generate precise clock frequency. +*/ +void rockchip_fractional_approximation (struct clk_hw *hw, +unsigned long rate,unsigned long *parent_rate, +unsigned long *m,unsigned long *n) +{ +struct clk_fractional_divider *fd =to_clk_fd (hw); +unsigned long p_rate, p_parent_rate; +struct clk_hw *p_parent; +unsigned long scale; + + p_rate =clk_hw_get_rate (clk_hw_get_parent (hw)); +if ((rate *20 > p_rate) && (p_rate % rate !=0 )) { + p_parent =clk_hw_get_parent (clk_hw_get_parent (hw)); + p_parent_rate =clk_hw_get_rate (p_parent); + *parent_rate = p_parent_rate; + } + +/* +* Get rate closer to *parent_rate to guarantee there is no overflow +* for m and n. In the result it will be the nearest rate left shifted +* by (scale - fd->nwidth) bits. +*/ + scale =fls_long (*parent_rate / rate -1 ); +if (scale > fd->nwidth) + rate <<= scale - fd->nwidth; + +rational_best_approximation (rate, *parent_rate, +GENMASK (fd->mwidth -1 ,0 ),GENMASK (fd->nwidth -1 ,0 ), + m, n); +} + static struct clk *rockchip_clk_register_frac_branch( struct rockchip_clk_provider *ctx, const char *name, const char *const *parent_names, u8 num_parents,
drivers/clk/rockchip/clk.c static struct clk *rockchip_clk_register_frac_branch( · line 251
Quoted 7 lines
drivers/clk/rockchip/clk.c +1
-0@@ -210,6 +245,7 @@ static struct clk *rockchip_clk_register_frac_branch( div->nwidth =16 ; div->nmask =GENMASK (div->nwidth -1 ,0 ) << div->nshift; div->lock = lock; + div->approximation = rockchip_fractional_approximation; div_ops = &clk_fractional_divider_ops; clk = clk_register_composite(NULL, name, parent_names, num_parents,
Heiko Stuebner
maintainerRe: [PATCH 2/2] clk: rockchip: add special approximation to fix up fractional clk's jitter
In reply to Heiko StuebnerAm Dienstag, 1. August 2017, 18:22:24 CEST schrieb Heiko Stuebner:
Quoted 10 lines
From: Elaine Zhang <[email protected]>
From Rockchips fractional divider description:
3.1.9 Fractional divider usage
To get specific frequency, clocks of I2S, SPDIF, UARTcan be generated by fractional divider. Generally you must set that denominator is 20 times larger than numerator to generate precise clock frequency. So the fractional divider applies only to generate low frequency clock like I2S, UART.
Quoted 4 lines
Therefore add a special approximation function that handles this special requirement.
Signed-off-by Elaine Zhang <[email protected]>
applied for 4.14
Heiko