Re: [PATCH v2 1/3] clk: rockchip: add support for gate link

From: Sebastian Reichel
Date: Tue Aug 01 2023 - 13:43:58 EST


Hi,

On Tue, Aug 01, 2023 at 03:43:55PM +0800, Elaine Zhang wrote:
> Recent Rockchip SoCs have a new hardware block called Native Interface
> Unit (NIU), which gates clocks to devices behind them. These effectively
> need two parent clocks.
> Use GATE_LINK to handle this.
>
> Signed-off-by: Elaine Zhang <zhangqing@xxxxxxxxxxxxxx>
> ---

This looks better than the current upstream solution of just
marking all linked clocks as criticial. I do have a few comments,
though.

> drivers/clk/rockchip/Makefile | 1 +
> drivers/clk/rockchip/clk-gate-link.c | 189 +++++++++++++++++++++++++++
> drivers/clk/rockchip/clk.c | 7 +
> drivers/clk/rockchip/clk.h | 22 ++++
> 4 files changed, 219 insertions(+)
> create mode 100644 drivers/clk/rockchip/clk-gate-link.c
>
> diff --git a/drivers/clk/rockchip/Makefile b/drivers/clk/rockchip/Makefile
> index 36894f6a7022..87cc39d54f72 100644
> --- a/drivers/clk/rockchip/Makefile
> +++ b/drivers/clk/rockchip/Makefile
> @@ -13,6 +13,7 @@ clk-rockchip-y += clk-inverter.o
> clk-rockchip-y += clk-mmc-phase.o
> clk-rockchip-y += clk-muxgrf.o
> clk-rockchip-y += clk-ddr.o
> +clk-rockchip-y += clk-gate-link.o
> clk-rockchip-$(CONFIG_RESET_CONTROLLER) += softrst.o
>
> obj-$(CONFIG_CLK_PX30) += clk-px30.o
> diff --git a/drivers/clk/rockchip/clk-gate-link.c b/drivers/clk/rockchip/clk-gate-link.c
> new file mode 100644
> index 000000000000..06d25f00ec0c
> --- /dev/null
> +++ b/drivers/clk/rockchip/clk-gate-link.c
> @@ -0,0 +1,189 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (c) 2023 Fuzhou Rockchip Electronics Co., Ltd
> + */
> +#include <linux/clk.h>
> +#include <linux/clk-provider.h>
> +#include <linux/module.h>
> +#include <linux/slab.h>
> +#include <linux/io.h>
> +#include <linux/err.h>
> +#include <linux/string.h>
> +#include "clk.h"
> +
> +/**
> + * struct clk_gate_link - gating link clock
> + *
> + * @hw: handle between common and hardware-specific interfaces
> + * @reg: register controlling gate
> + * @bit_idx: single bit controlling gate
> + * @flags: hardware-specific flags
> + * @lock: register lock
> + * @link: links clk
> + */
> +struct clk_gate_link {
> + struct clk_hw hw;
> + void __iomem *reg;
> + u8 bit_idx;
> + u8 flags;
> + spinlock_t *lock;
> + struct clk *link;
> +};
> +
> +#define to_clk_gate_link(_hw) container_of(_hw, struct clk_gate_link, hw)
> +
> +static inline u32 clk_gate_readl(struct clk_gate_link *gate)
> +{
> + if (gate->flags & CLK_GATE_BIG_ENDIAN)
> + return ioread32be(gate->reg);
> +
> + return readl(gate->reg);
> +}
> +
> +static inline void clk_gate_writel(struct clk_gate_link *gate, u32 val)
> +{
> + if (gate->flags & CLK_GATE_BIG_ENDIAN)
> + iowrite32be(val, gate->reg);
> + else
> + writel(val, gate->reg);
> +}

Rockchip platform has no clocks using CLK_GATE_BIG_ENDIAN flag, so
you can just use readl/writel. But by reusing clk-gate's function
(see next comment) you don't need to directly read/write registers
anyways.

> +
> +/*
> + * It works on following logic:
> + *
> + * For enabling clock, enable = 1
> + * set2dis = 1 -> clear bit -> set = 0
> + * set2dis = 0 -> set bit -> set = 1
> + *
> + * For disabling clock, enable = 0
> + * set2dis = 1 -> set bit -> set = 1
> + * set2dis = 0 -> clear bit -> set = 0
> + *
> + * So, result is always: enable xor set2dis.
> + */
> +static void clk_gate_endisable(struct clk_hw *hw, int enable)
> +{
> + struct clk_gate_link *gate = to_clk_gate_link(hw);
> + int set = gate->flags & CLK_GATE_SET_TO_DISABLE ? 1 : 0;
> + unsigned long flags;
> + u32 reg;
> +
> + set ^= enable;
> +
> + if (gate->lock)
> + spin_lock_irqsave(gate->lock, flags);
> + else
> + __acquire(gate->lock);
> +
> + if (gate->flags & CLK_GATE_HIWORD_MASK) {
> + reg = BIT(gate->bit_idx + 16);
> + if (set)
> + reg |= BIT(gate->bit_idx);
> + } else {
> + reg = clk_gate_readl(gate);
> +
> + if (set)
> + reg |= BIT(gate->bit_idx);
> + else
> + reg &= ~BIT(gate->bit_idx);
> + }
> +
> + clk_gate_writel(gate, reg);
> +
> + if (gate->lock)
> + spin_unlock_irqrestore(gate->lock, flags);
> + else
> + __release(gate->lock);
> +}

By changing "struct clk_gate_link" to look like this:

struct clk_gate_link {
struct clk_gate gate;
struct clk *link;
}

It can be casted to be a normal struct clk_gate, which
means you can reuse clk_gate_endisable() (needs to be
exported) and clk_gate_is_enabled().

> +static int clk_gate_link_enable(struct clk_hw *hw)
> +{
> + struct clk_gate_link *gate = to_clk_gate_link(hw);
> +
> + clk_gate_endisable(hw, 1);
> + clk_enable(gate->link);
> +
> + return 0;
> +}
> +
> +static void clk_gate_link_disable(struct clk_hw *hw)
> +{
> + struct clk_gate_link *gate = to_clk_gate_link(hw);
> +
> + clk_gate_endisable(hw, 0);
> + clk_disable(gate->link);
> +}
> +
> +static int clk_gate_link_is_enabled(struct clk_hw *hw)
> +{
> + u32 reg;
> + struct clk_gate_link *gate = to_clk_gate_link(hw);
> +
> + reg = clk_gate_readl(gate);
> +
> + /* if a set bit disables this clk, flip it before masking */
> + if (gate->flags & CLK_GATE_SET_TO_DISABLE)
> + reg ^= BIT(gate->bit_idx);
> +
> + reg &= BIT(gate->bit_idx);
> +
> + return reg ? 1 : 0;
> +}
> +
> +const struct clk_ops clk_gate_link_ops = {
> + .enable = clk_gate_link_enable,
> + .disable = clk_gate_link_disable,
> + .is_enabled = clk_gate_link_is_enabled,
> +};
> +EXPORT_SYMBOL_GPL(clk_gate_link_ops);

I don't think they need to be exported?

> +
> +struct clk *rockchip_clk_register_gate_link(struct rockchip_clk_provider *ctx,
> + const char *name, const char *parent_name,
> + unsigned int link_id, u8 flags,
> + void __iomem *gate_offset, u8 gate_shift,
> + u8 gate_flags, spinlock_t *lock)
> +{
> + struct clk_gate_link *gate;
> + struct clk_init_data init = {};
> + struct clk **clks;
> + struct clk *clk_link;
> +
> + if (gate_flags & CLK_GATE_HIWORD_MASK) {
> + if (gate_shift > 15) {
> + pr_err("gate bit exceeds LOWORD field\n");
> + return ERR_PTR(-ENOMEM);
> + }
> + }
> +
> + /* allocate the gate */
> + gate = kzalloc(sizeof(*gate), GFP_KERNEL);
> + if (!gate)
> + return ERR_PTR(-ENOMEM);
> +
> + clks = ctx->clk_data.clks;
> + gate->link = clks[link_id];
> +
> + init.name = name;
> + init.ops = &clk_gate_link_ops;
> + init.flags = flags;
> + init.parent_names = parent_name ? &parent_name : NULL;
> + init.num_parents = 1;
> +
> + /* struct clk_gate assignments */
> + gate->reg = gate_offset;
> + gate->bit_idx = gate_shift;
> + gate->flags = gate_flags;
> + gate->lock = lock;
> + gate->hw.init = &init;
> +
> + clk_link = clk_register(NULL, &gate->hw);
> + if (IS_ERR(clk_link)) {
> + kfree(gate);
> + pr_err("%s clk_register field\n", name);
> + return ERR_CAST(clk_link);
> + }
> + clk_prepare(gate->link);

So the linked clock will always be prepared? Can't we just extend clk_gate_link_ops with

.prepare = clk_gate_link_prepare,
.unprepare = clk_gate_link_unprepare,

int clk_gate_link_prepare(clk_gate_link_prepare) {
struct clk_gate_link *gate = to_clk_gate_link(hw);
return clk_prepare(gate->link);
}

void clk_gate_link_unprepare(clk_gate_link_prepare) {
struct clk_gate_link *gate = to_clk_gate_link(hw);
clk_unprepare(gate->link);
}

Greetings,

-- Sebastian

> +
> + return clk_link;
> +}
> +EXPORT_SYMBOL_GPL(rockchip_clk_register_gate_link);
> diff --git a/drivers/clk/rockchip/clk.c b/drivers/clk/rockchip/clk.c
> index 4059d9365ae6..d981ef6c5487 100644
> --- a/drivers/clk/rockchip/clk.c
> +++ b/drivers/clk/rockchip/clk.c
> @@ -509,6 +509,13 @@ void rockchip_clk_register_branches(struct rockchip_clk_provider *ctx,
> ctx->reg_base + list->gate_offset,
> list->gate_shift, list->gate_flags, &ctx->lock);
> break;
> +
> + case branch_gate_link:
> + clk = rockchip_clk_register_gate_link(ctx, list->name,
> + list->parent_names[0], list->link_id, flags,
> + ctx->reg_base + list->gate_offset,
> + list->gate_shift, list->gate_flags, &ctx->lock);
> + break;
> case branch_composite:
> clk = rockchip_clk_register_branch(list->name,
> list->parent_names, list->num_parents,
> diff --git a/drivers/clk/rockchip/clk.h b/drivers/clk/rockchip/clk.h
> index 758ebaf2236b..b49e3cede33a 100644
> --- a/drivers/clk/rockchip/clk.h
> +++ b/drivers/clk/rockchip/clk.h
> @@ -517,6 +517,7 @@ enum rockchip_clk_branch_type {
> branch_divider,
> branch_fraction_divider,
> branch_gate,
> + branch_gate_link,
> branch_mmc,
> branch_inverter,
> branch_factor,
> @@ -529,6 +530,7 @@ struct rockchip_clk_branch {
> enum rockchip_clk_branch_type branch_type;
> const char *name;
> const char *const *parent_names;
> + unsigned int link_id;
> u8 num_parents;
> unsigned long flags;
> int muxdiv_offset;
> @@ -842,6 +844,20 @@ struct rockchip_clk_branch {
> .gate_flags = gf, \
> }
>
> +#define GATE_LINK(_id, cname, pname, _linkid, f, o, b, gf) \
> + { \
> + .id = _id, \
> + .branch_type = branch_gate_link, \
> + .name = cname, \
> + .parent_names = (const char *[]){ pname }, \
> + .num_parents = 1, \
> + .link_id = _linkid, \
> + .flags = f, \
> + .gate_offset = o, \
> + .gate_shift = b, \
> + .gate_flags = gf, \
> + }
> +
> #define MMC(_id, cname, pname, offset, shift) \
> { \
> .id = _id, \
> @@ -1002,6 +1018,12 @@ struct clk *rockchip_clk_register_halfdiv(const char *name,
> unsigned long flags,
> spinlock_t *lock);
>
> +struct clk *rockchip_clk_register_gate_link(struct rockchip_clk_provider *ctx,
> + const char *name, const char *parent_name,
> + unsigned int link_id, u8 flags,
> + void __iomem *gate_offset, u8 gate_shift,
> + u8 gate_flags, spinlock_t *lock);
> +
> #ifdef CONFIG_RESET_CONTROLLER
> void rockchip_register_softrst_lut(struct device_node *np,
> const int *lookup_table,
> --
> 2.17.1
>

Attachment: signature.asc
Description: PGP signature