[PATCHv9 09/43] CLK: ti: add support for ti divider-clock

Tero Kristo t-kristo at ti.com
Fri Nov 1 05:54:59 EDT 2013


On 11/01/2013 11:48 AM, Tero Kristo wrote:
> On 10/31/2013 08:02 PM, Nishanth Menon wrote:
>> On 10/25/2013 10:57 AM, Tero Kristo wrote:
>>> This patch adds support for TI divider clock binding, which simply uses
>>> the basic clock divider to provide the features needed.
>>>
>>> Signed-off-by: Tero Kristo <t-kristo at ti.com>
>>> ---
>>>   .../devicetree/bindings/clock/ti/divider.txt       |   86 +++++++
>>>   drivers/clk/ti/Makefile                            |    3 +-
>>>   drivers/clk/ti/divider.c                           |  239
>>> ++++++++++++++++++++
>>>   3 files changed, 327 insertions(+), 1 deletion(-)
>>>   create mode 100644
>>> Documentation/devicetree/bindings/clock/ti/divider.txt
>>>   create mode 100644 drivers/clk/ti/divider.c
>>>
>>> diff --git a/Documentation/devicetree/bindings/clock/ti/divider.txt
>>> b/Documentation/devicetree/bindings/clock/ti/divider.txt
>>> new file mode 100644
>>> index 0000000..65e3dcd
>>> --- /dev/null
>>> +++ b/Documentation/devicetree/bindings/clock/ti/divider.txt
>>> @@ -0,0 +1,86 @@
>>> +Binding for TI divider clock
>>> +
>>> +Binding status: Unstable - ABI compatibility may be broken in the
>>> future
>>> +
>>> +This binding uses the common clock binding[1].  It assumes a
>>> +register-mapped adjustable clock rate divider that does not gate and
>>> has
>>> +only one input clock or parent.  By default the value programmed into
>>> +the register is one less than the actual divisor value.  E.g:
>>> +
>>> +register value        actual divisor value
>>> +0            1
>>> +1            2
>>> +2            3
>>> +
>>> +This assumption may be modified by the following optional properties:
>>> +
>>> +ti,index-starts-at-one - valid divisor values start at 1, not the
>>> default
>>> +of 0.  E.g:
>>> +register value        actual divisor value
>>> +1            1
>>> +2            2
>>> +3            3
>>> +
>>> +ti,index-power-of-two - valid divisor values are powers of two.  E.g:
>>> +register value        actual divisor value
>>> +0            1
>>> +1            2
>>> +2            4
>>> +
>>> +Additionally an array of valid dividers may be supplied like so:
>>> +
>>> +    dividers = <4>, <8>, <0>, <16>;
>> ti,dividers I believe.
>
> True.
>
>>
>>> +
>>> +Which will map the resulting values to a divisor table by their index:
>>> +register value        actual divisor value
>>> +0            4
>>> +1            8
>>> +2            <invalid divisor, skipped>
>>> +3            16
>>> +
>>> +Any zero value in this array means the corresponding bit-value is
>>> invalid
>>> +and must not be used.
>>> +
>>> +The binding must also provide the register to control the divider and
>>> +unless the divider array is provided, min and max dividers. Optionally
>>> +the number of bits to shift that mask, if necessary. If the shift value
>>> +is missing it is the same as supplying a zero shift.
>>> +
>>> +Required properties:
>>> +- compatible : shall be "ti,divider-clock".
>>
>> ti,composite-divider-clock undocumented?
>
> Hmm yea true.
>
>>
>>> +- #clock-cells : from common clock binding; shall be set to 0.
>>> +- clocks : link to phandle of parent clock
>>> +- reg : offset for register controlling adjustable divider
>>> +
>>> +Optional properties:
>>> +- clock-output-names : from common clock binding.
>>> +- ti,dividers : array of integers defining divisors
>>> +- ti,bit-shift : number of bits to shift the divider value, defaults
>>> to 0
>>> +- ti,min-div : min divisor for dividing the input clock rate, only
>>> +  needed if the first divisor is offset from the default value (1)
>>> +- ti,max-div : max divisor for dividing the input clock rate, only
>>> needed
>>> +  if ti,dividers is not defined.
>>> +- ti,index-starts-at-one : valid divisor programming starts at 1,
>>> not zero
>>
>> makes sense only if ti,dividers are not defined, right?
>> CLK_DIVIDER_ONE_BASED is used with !table
>
> Yeah, ignored if table is present.
>
>>
>>> +- ti,index-power-of-two : valid divisor programming must be a power
>>> of two
>>
>> makes sense only if ti,dividers are not defined, right?
>> CLK_DIVIDER_POWER_OF_TWO is used with !table
>
> Yea.
>
>>
>>> +- ti,autoidle-shift : bit shift of the autoidle enable bit for the
>>> clock
>>> +- ti,invert-autoidle-bit : autoidle is enabled by setting the bit to 0
>>
>> These are part of auto idle driver bindings, so maybe give a link to
>> that?
>
> Yea can do.
>
>>
>>> +- ti,set-rate-parent : clk_set_rate is propagated to parent
>>
>> yeah - this is one of those properties that should probably become
>> generic at a later point in time.
>>
>>> +
>>> +Examples:
>>> +dpll_usb_m2_ck: dpll_usb_m2_ck at 4a008190 {
>>> +    #clock-cells = <0>;
>>> +    compatible = "ti,divider-clock";
>>> +    clocks = <&dpll_usb_ck>;
>>> +    ti,max-div = <127>;
>>> +    reg = <0x190>;
>>> +    ti,index-starts-at-one;
>>> +};
>>> +
>>> +aess_fclk: aess_fclk at 4a004528 {
>>> +    #clock-cells = <0>;
>>> +    compatible = "ti,divider-clock";
>>> +    clocks = <&abe_clk>;
>>> +    ti,bit-shift = <24>;
>>> +    reg = <0x528>;
>>> +    ti,max-div = <2>;
>>> +};
>>
>> an example of ti,dividers will be useful as well.
>
> Ok.
>
>>
>>> diff --git a/drivers/clk/ti/Makefile b/drivers/clk/ti/Makefile
>>> index a4a7595..640ebf9 100644
>>> --- a/drivers/clk/ti/Makefile
>>> +++ b/drivers/clk/ti/Makefile
>>> @@ -1,3 +1,4 @@
>>>   ifneq ($(CONFIG_OF),)
>>> -obj-y                    += clk.o dpll.o autoidle.o composite.o
>>> +obj-y                    += clk.o dpll.o autoidle.o divider.o \
>>> +                       composite.o
>>>   endif
>>> diff --git a/drivers/clk/ti/divider.c b/drivers/clk/ti/divider.c
>>> new file mode 100644
>>> index 0000000..787bc8f
>>> --- /dev/null
>>> +++ b/drivers/clk/ti/divider.c
>>> @@ -0,0 +1,239 @@
>>> +/*
>>> + * TI Divider Clock
>>> + *
>>> + * Copyright (C) 2013 Texas Instruments, Inc.
>>> + *
>>> + * Tero Kristo <t-kristo at ti.com>
>>> + *
>>> + * This program is free software; you can redistribute it and/or modify
>>> + * it under the terms of the GNU General Public License version 2 as
>>> + * published by the Free Software Foundation.
>>> + *
>>> + * This program is distributed "as is" WITHOUT ANY WARRANTY of any
>>> + * kind, whether express or implied; without even the implied warranty
>>> + * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>>> + * GNU General Public License for more details.
>>> + */
>>> +
>>> +#include <linux/clk-provider.h>
>>> +#include <linux/slab.h>
>>> +#include <linux/err.h>
>>> +#include <linux/of.h>
>>> +#include <linux/of_address.h>
>>> +#include <linux/clk/ti.h>
>>> +
>>> +static struct clk_div_table
>>> +__init *ti_clk_get_div_table(struct device_node *node)
>>> +{
>>> +    struct clk_div_table *table;
>>> +    const __be32 *divspec;
>>> +    u32 val;
>>> +    u32 num_div;
>>> +    u32 valid_div;
>>> +    int i;
>>> +
>>> +    divspec = of_get_property(node, "ti,dividers", &num_div);
>>> +
>>> +    if (!divspec)
>>> +        return NULL;
>>> +
>>> +    num_div /= 4;
>>> +
>>> +    valid_div = 0;
>>> +
>>> +    /* Determine required size for divider table */
>>> +    for (i = 0; i < num_div; i++) {
>>> +        of_property_read_u32_index(node, "ti,dividers", i, &val);
>>> +        if (val)
>>> +            valid_div++;
>>> +    }
>>> +
>>> +    if (!valid_div) {
>>> +        pr_err("%s: no valid dividers for %s table\n", __func__,
>>> +               node->name);
>>> +        return ERR_PTR(-EINVAL);
>>> +    }
>>> +
>>> +    table = kzalloc(sizeof(*table) * (valid_div + 1), GFP_KERNEL);
>>> +
>>> +    if (!table) {
>>> +        pr_err("%s: unable to allocate memory for %s table\n",
>>> __func__,
>>> +               node->name);
>>
>> you could drop this.
>
> True.
>
>>
>>> +        return ERR_PTR(-ENOMEM);
>>> +    }
>>> +
>>> +    valid_div = 0;
>>> +
>>> +    for (i = 0; i < num_div; i++) {
>>> +        of_property_read_u32_index(node, "ti,dividers", i, &val);
>>> +        if (val) {
>>> +            table[valid_div].div = val;
>>> +            table[valid_div].val = i;
>>> +            valid_div++;
>>> +        }
>>> +    }
>>> +
>>> +    return table;
>>> +}
>>> +
>>> +static int _get_divider_width(struct device_node *node,
>>> +                  const struct clk_div_table *table,
>>> +                  u8 flags)
>>> +{
>>> +    u32 min_div;
>>> +    u32 max_div;
>>> +    u32 val = 0;
>>> +    u32 div;
>>> +
>>> +    if (!table) {
>>> +        /* Clk divider table not provided, determine min/max divs */
>>> +        if (of_property_read_u32(node, "ti,min-div", &min_div)) {
>>> +            pr_debug("%s: no min-div for %s, default=1\n",
>>> +                 __func__, node->name);
>>> +            min_div = 1;
>>> +        }
>>> +
>>> +        if (of_property_read_u32(node, "ti,max-div", &max_div)) {
>>> +            pr_err("%s: no max-div for %s!\n", __func__,
>>> +                   node->name);
>>> +            return -EINVAL;
>>
>> will be nice to get warning if ti,divider is defined and any of the
>> non-operational properties are defined as well.
>
> Hmm ok, can add checks for those.
>
>>
>>> +        }
>>> +
>>> +        /* Determine bit width for the field */
>>> +        if (flags & CLK_DIVIDER_ONE_BASED)
>>> +            val = 1;
>>> +
>>> +        div = min_div;
>>> +
>>> +        while (div < max_div) {
>>> +            if (flags & CLK_DIVIDER_POWER_OF_TWO)
>>> +                div <<= 1;
>>> +            else
>>> +                div++;
>>> +            val++;
>>> +        }
>>> +    } else {
>>> +        div = 0;
>>> +
>>> +        while (table[div].div) {
>>> +            val = table[div].val;
>>> +            div++;
>>> +        }
>>> +    }
>>> +
>>> +    return fls(val);
>>> +}
>>> +
>>> +/**
>>> + * of_ti_divider_clk_setup() - Setup function for simple div rate clock
>>> + */
>>> +static int __init of_ti_divider_clk_setup(struct device_node *node,
>>> +                      struct regmap *regmap)
>>> +{
>>> +    struct clk *clk;
>>> +    const char *clk_name = node->name;
>>> +    void __iomem *reg;
>>> +    const char *parent_name;
>>> +    u8 clk_divider_flags = 0;
>>> +    u8 width = 0;
>>> +    u8 shift = 0;
>>> +    struct clk_div_table *table = NULL;
>>> +    u32 val = 0;
>>> +    u32 flags = 0;
>>> +    int ret = 0;
>>> +
>>> +    parent_name = of_clk_get_parent_name(node, 0);
>>> +
>>> +    if (of_property_read_u32(node, "reg", &val)) {
>>> +        pr_err("%s: %s must have reg\n", __func__, clk_name);
>>> +        return -EINVAL;
>>> +    }
>>> +
>>> +    reg = (void *)val;
>>> +
>>> +    if (!of_property_read_u32(node, "ti,bit-shift", &val))
>>> +        shift = val;
>>> +
>>> +    if (of_property_read_bool(node, "ti,index-starts-at-one"))
>>> +        clk_divider_flags |= CLK_DIVIDER_ONE_BASED;
>>> +
>>> +    if (of_property_read_bool(node, "ti,index-power-of-two"))
>>> +        clk_divider_flags |= CLK_DIVIDER_POWER_OF_TWO;
>>> +
>>> +    if (of_property_read_bool(node, "ti,set-rate-parent"))
>>> +        flags |= CLK_SET_RATE_PARENT;
>>> +
>>> +    table = ti_clk_get_div_table(node);
>>> +
>>> +    if (IS_ERR(table))
>>> +        return PTR_ERR(table);
>>> +
>>> +    ret = _get_divider_width(node, table, clk_divider_flags);
>>> +    if (ret < 0)
>>> +        goto cleanup;
>>> +
>>> +    width = ret;
>>> +
>>> +    clk = clk_register_divider_table_regmap(NULL, clk_name,
>>> parent_name,
>>> +                        flags, reg, regmap, shift,
>>> +                        width, clk_divider_flags, table,
>>> +                        NULL);
>>> +
>>> +    if (!IS_ERR(clk)) {
>>> +        of_clk_add_provider(node, of_clk_src_simple_get, clk);
>>> +        ret = of_ti_autoidle_setup(node, regmap);
>>
>> if this fails, table will be freed though, we have added provider and
>> registerd table_regmap, no?
>
> Provider and regmap are shared for the whole IP block (CM/PRM whatever.)
> Those are only initialized once.

Some minor confusion here in my comment, sorry about that.

Regmap is shared for the whole IP block and initialized only once, we 
can't remove it here as it is not owned by this clock. For 
clock-provider / unregister part, I don't want to clean those up as 1) 
unregister doesn't currently do anything 2) autoidle setup is not 
critical failure, it just breaks PM. I can add error print for it though.

-Tero

>
>>
>>> +    } else {
>>> +        ret = PTR_ERR(clk);
>>> +    }
>>> +
>>> +    if (!ret)
>>> +        return 0;
>>> +cleanup:
>>> +    kfree(table);
>>> +    return ret;
>>> +}
>>> +CLK_OF_DECLARE(divider_clk, "ti,divider-clock",
>>> of_ti_divider_clk_setup);
>>> +
>>> +static int __init of_ti_composite_divider_clk_setup(struct
>>> device_node *node,
>>> +                             struct regmap *regmap)
>>> +{
>>> +    struct clk_divider *div;
>>> +    u32 val;
>>> +    int ret;
>>> +
>>> +    div = kzalloc(sizeof(*div), GFP_KERNEL);
>>> +    if (!div)
>>> +        return -ENOMEM;
>>> +
>>> +    of_property_read_u32(node, "reg", &val);
>>
>> reg is a mandatory property, no? error checks?
>
> Ok can add. :P
>
>>
>>> +    div->reg = (void *)val;
>>> +    div->regmap = regmap;
>>> +
>>> +    div->table = ti_clk_get_div_table(node);
>>> +
>>> +    if (of_property_read_bool(node, "ti,index-starts-at-one"))
>>> +        div->flags |= CLK_DIVIDER_ONE_BASED;
>>
>> alright, this does not really map back to ti,divder-clock style
>> handling.. aren't they supposed to be consistent?
>> how about "ti,index-power-of-two", bit shift etc?
>
> I can add non-used property checks yes.
>
>>> +
>>> +    ret = _get_divider_width(node, div->table, div->flags);
>>> +    if (ret < 0)
>>> +        goto cleanup;
>>> +
>>> +    div->width = ret;
>>> +
>>> +    if (of_property_read_u32(node, "ti,bit-shift", &val)) {
>>> +        pr_debug("%s: missing bit-shift property for %s, default=0\n",
>>> +             __func__, node->name);
>>> +        val = 0;
>>> +    }
>>> +    div->shift = val;
>>> +
>>> +    ret = ti_clk_add_component(node, &div->hw,
>>> CLK_COMPONENT_TYPE_DIVIDER);
>>> +    if (!ret)
>>> +        return 0;
>>> +
>>> +cleanup:
>>> +    kfree(div);
>>> +    return ret;
>>> +}
>>> +CLK_OF_DECLARE(ti_composite_divider_clk, "ti,composite-divider-clock",
>>> +           of_ti_composite_divider_clk_setup);
>>>
>>
>>
>




More information about the linux-arm-kernel mailing list