All of lore.kernel.org
 help / color / mirror / Atom feed
From: Sean Anderson <seanga2@gmail.com>
To: Weijie Gao <weijie.gao@mediatek.com>, u-boot@lists.denx.de
Cc: GSS_MTK_Uboot_upstream <GSS_MTK_Uboot_upstream@mediatek.com>,
	Lukasz Majewski <lukma@denx.de>
Subject: Re: [PATCH v2 03/14] clk: mtmips: add clock driver for MediaTek MT7621 SoC
Date: Wed, 15 Dec 2021 11:11:56 -0500	[thread overview]
Message-ID: <eb7342eb-9fee-a0be-49ab-32dce9f4550e@gmail.com> (raw)
In-Reply-To: <0a795ee34bbeb7d8e58cf8f60c35d5048f9bda95.camel@mediatek.com>

Hi Weijie,

(sorry for the delayed response)

On 12/3/21 5:06 AM, Weijie Gao wrote:
> On Fri, 2021-11-26 at 12:44 -0500, Sean Anderson wrote:
>> On 11/18/21 8:35 PM, Weijie Gao wrote:
>>> This patch adds a clock driver for MediaTek MT7621 SoC.
>>> This driver provides clock gate control as well as getting clock
>>> frequency
>>> for CPU/SYS/XTAL and some peripherals.
>>>
>>> Signed-off-by: Weijie Gao <weijie.gao@mediatek.com>
>>> ---
>>> v2 changes: none
>>> ---
>>>    drivers/clk/mtmips/Makefile            |   1 +
>>>    drivers/clk/mtmips/clk-mt7621.c        | 260
>>> +++++++++++++++++++++++++
>>>    include/dt-bindings/clock/mt7621-clk.h |  42 ++++
>>>    3 files changed, 303 insertions(+)
>>>    create mode 100644 drivers/clk/mtmips/clk-mt7621.c
>>>    create mode 100644 include/dt-bindings/clock/mt7621-clk.h
>>>
>>> diff --git a/drivers/clk/mtmips/Makefile
>>> b/drivers/clk/mtmips/Makefile
>>> index 732e7f2545..ee8b5afe87 100644
>>> --- a/drivers/clk/mtmips/Makefile
>>> +++ b/drivers/clk/mtmips/Makefile
>>> @@ -1,4 +1,5 @@
>>>    # SPDX-License-Identifier: GPL-2.0
>>>    
>>>    obj-$(CONFIG_SOC_MT7620) += clk-mt7620.o
>>> +obj-$(CONFIG_SOC_MT7621) += clk-mt7621.o
>>>    obj-$(CONFIG_SOC_MT7628) += clk-mt7628.o
>>> diff --git a/drivers/clk/mtmips/clk-mt7621.c
>>> b/drivers/clk/mtmips/clk-mt7621.c
>>> new file mode 100644
>>> index 0000000000..3799d1806a
>>> --- /dev/null
>>> +++ b/drivers/clk/mtmips/clk-mt7621.c
>>> @@ -0,0 +1,260 @@
>>> +// SPDX-License-Identifier: GPL-2.0
>>> +/*
>>> + * Copyright (C) 2021 MediaTek Inc. All Rights Reserved.
>>> + *
>>> + * Author: Weijie Gao <weijie.gao@mediatek.com>
>>> + */
>>> +
>>> +#include <common.h>
>>> +#include <clk-uclass.h>
>>> +#include <dm.h>
>>> +#include <dm/device_compat.h>
>>> +#include <regmap.h>
>>> +#include <syscon.h>
>>> +#include <dt-bindings/clock/mt7621-clk.h>
>>> +#include <linux/bitops.h>
>>> +#include <linux/io.h>
>>> +
>>> +#define SYSC_MAP_SIZE			0x100
>>> +#define MEMC_MAP_SIZE			0x1000
>>> +
>>> +/* SYSC */
>>> +#define SYSCFG0_REG			0x10
>>> +#define XTAL_MODE_SEL_S			6
>>> +#define XTAL_MODE_SEL_M			0x1c0
>>
>> Please use genmask to define this:
>>
>> #define XTAL_MODE_SEL_M			GENMASK(8, 6)
>>
>> and SEL_S is unnecessary, see commentary below.
>>
>>> +
>>> +#define CLKCFG0_REG			0x2c
>>> +#define CPU_CLK_SEL_S			30
>>> +#define CPU_CLK_SEL_M			0xc0000000
>>> +#define PERI_CLK_SEL			0x10
>>> +
>>> +#define CLKCFG1_REG			0x30
>>> +
>>> +#define CUR_CLK_STS_REG			0x44
>>> +#define CUR_CPU_FDIV_S			8
>>> +#define CUR_CPU_FDIV_M			0x1f00
>>> +#define CUR_CPU_FFRAC_S			0
>>> +#define CUR_CPU_FFRAC_M			0x1f
>>> +
>>> +/* MEMC */
>>> +#define MEMPLL1_REG			0x0604
>>> +#define RG_MEPL_DIV2_SEL_S		1
>>> +#define RG_MEPL_DIV2_SEL_M		0x06
>>> +
>>> +#define MEMPLL6_REG			0x0618
>>> +#define MEMPLL18_REG			0x0648
>>> +#define RG_MEPL_PREDIV_S		12
>>> +#define RG_MEPL_PREDIV_M		0x3000
>>> +#define RG_MEPL_FBDIV_S			4
>>> +#define RG_MEPL_FBDIV_M			0x7f0
>>> +
>>> +/* Clock sources */
>>> +#define CLK_SRC_CPU			-1
>>> +#define CLK_SRC_CPU_D2			-2
>>> +#define CLK_SRC_DDR			-3
>>> +#define CLK_SRC_SYS			-4
>>> +#define CLK_SRC_XTAL			-5
>>> +#define CLK_SRC_PERI			-6
>>
>> Please use an enum. And why are these negative?
> 
> I'll rewrite this
> 
>>
>>> +/* EPLL clock */
>>> +#define EPLL_CLK			50000000
>>> +
>>> +struct mt7621_clk_priv {
>>> +	void __iomem *sysc_base;
>>> +	void __iomem *memc_base;
>>> +	int cpu_clk;
>>> +	int ddr_clk;
>>> +	int sys_clk;
>>> +	int xtal_clk;
>>> +};
>>> +
>>> +static const int mt7621_clks[] = {
>>> +	[CLK_SYS] = CLK_SRC_SYS,
>>> +	[CLK_DDR] = CLK_SRC_DDR,
>>> +	[CLK_CPU] = CLK_SRC_CPU,
>>> +	[CLK_XTAL] = CLK_SRC_XTAL,
>>> +	[CLK_MIPS_CNT] = CLK_SRC_CPU_D2,
>>> +	[CLK_UART3] = CLK_SRC_PERI,
>>> +	[CLK_UART2] = CLK_SRC_PERI,
>>> +	[CLK_UART1] = CLK_SRC_PERI,
>>> +	[CLK_SPI] = CLK_SRC_SYS,
>>> +	[CLK_I2C] = CLK_SRC_PERI,
>>> +	[CLK_TIMER] = CLK_SRC_PERI,
>>> +};
>>> +
>>> +static ulong mt7621_clk_get_rate(struct clk *clk)
>>> +{
>>> +	struct mt7621_clk_priv *priv = dev_get_priv(clk->dev);
>>> +	u32 val;
>>> +
>>> +	if (clk->id >= ARRAY_SIZE(mt7621_clks))
>>> +		return 0;
>>> +
>>> +	switch (mt7621_clks[clk->id]) {
>>> +	case CLK_SRC_CPU:
>>> +		return priv->cpu_clk;
>>> +	case CLK_SRC_CPU_D2:
>>> +		return priv->cpu_clk / 2;
>>> +	case CLK_SRC_DDR:
>>> +		return priv->ddr_clk;
>>> +	case CLK_SRC_SYS:
>>> +		return priv->sys_clk;
>>> +	case CLK_SRC_XTAL:
>>> +		return priv->xtal_clk;
>>> +	case CLK_SRC_PERI:
>>> +		val = readl(priv->sysc_base + CLKCFG0_REG);
>>> +		if (val & PERI_CLK_SEL)
>>> +			return priv->xtal_clk;
>>> +		else
>>> +			return EPLL_CLK;
>>> +	default:
>>> +		return 0;
>>
>> -ENOSYS
>>
>>> +	}
>>> +}
>>> +
>>> +static int mt7621_clk_enable(struct clk *clk)
>>> +{
>>> +	struct mt7621_clk_priv *priv = dev_get_priv(clk->dev);
>>> +
>>> +	if (clk->id > 31)
>>
>> Please compare with a symbol.
> 
> OK. actually the clock gate register is 32-bit, and 31 is the MSB.

see below

>>
>>> +		return -1;
>>
>> -ENOSYS
>>
>>> +
>>> +	setbits_32(priv->sysc_base + CLKCFG1_REG, BIT(clk->id));
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +static int mt7621_clk_disable(struct clk *clk)
>>> +{
>>> +	struct mt7621_clk_priv *priv = dev_get_priv(clk->dev);
>>> +
>>> +	if (clk->id > 31)
>>> +		return -1;
>>> +
>>> +	clrbits_32(priv->sysc_base + CLKCFG1_REG, BIT(clk->id));
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +const struct clk_ops mt7621_clk_ops = {
>>> +	.enable = mt7621_clk_enable,
>>> +	.disable = mt7621_clk_disable,
>>> +	.get_rate = mt7621_clk_get_rate,
>>> +};
>>> +
>>> +static void mt7621_get_clocks(struct mt7621_clk_priv *priv)
>>> +{
>>> +	u32 bs, xtal_sel, clkcfg0, cur_clk, mempll, dividx, fb;
>>> +	u32 xtal_clk, xtal_div, ffiv, ffrac, cpu_clk, ddr_clk;
>>> +	static const u32 xtal_div_tbl[] = {0, 1, 2, 2};
>>> +
>>> +	bs = readl(priv->sysc_base + SYSCFG0_REG);
>>> +	clkcfg0 = readl(priv->sysc_base + CLKCFG0_REG);
>>> +	cur_clk = readl(priv->sysc_base + CUR_CLK_STS_REG);
>>> +
>>> +	xtal_sel = (bs & XTAL_MODE_SEL_M) >> XTAL_MODE_SEL_S;
>>
>> xtal_sel = FIELD_GET(XTAL_MODE_SEL_M, bs);
> 
> got it.
> 
>>
>>> +
>>> +	if (xtal_sel <= 2)
>>> +		xtal_clk = 20 * 1000 * 1000;
>>> +	else if (xtal_sel <= 5)
>>> +		xtal_clk = 40 * 1000 * 1000;
>>> +	else
>>> +		xtal_clk = 25 * 1000 * 1000;
>>> +
>>> +	switch ((clkcfg0 & CPU_CLK_SEL_M) >> CPU_CLK_SEL_S) {
>>
>> ditto
>>
>>> +	case 0:
>>> +		cpu_clk = 500 * 1000 * 1000;
>>> +		break;
>>> +	case 1:
>>> +		mempll = readl(priv->memc_base + MEMPLL18_REG);
>>> +		dividx = (mempll & RG_MEPL_PREDIV_M) >>
>>> RG_MEPL_PREDIV_S;
>>> +		fb = (mempll & RG_MEPL_FBDIV_M) >>
>>> RG_MEPL_FBDIV_S;
>>
>> ditto
>>
>>> +		xtal_div = 1 << xtal_div_tbl[dividx];
>>> +		cpu_clk = (fb + 1) * xtal_clk / xtal_div;
>>> +		break;
>>> +	default:
>>> +		cpu_clk = xtal_clk;
>>> +	}
>>> +
>>> +	ffiv = (cur_clk & CUR_CPU_FDIV_M) >> CUR_CPU_FDIV_S;
>>> +	ffrac = (cur_clk & CUR_CPU_FFRAC_M) >> CUR_CPU_FFRAC_S;
>>
>> ditto
>>
>>> +	cpu_clk = cpu_clk / ffiv * ffrac;
>>> +
>>> +	mempll = readl(priv->memc_base + MEMPLL6_REG);
>>> +	dividx = (mempll & RG_MEPL_PREDIV_M) >> RG_MEPL_PREDIV_S;
>>> +	fb = (mempll & RG_MEPL_FBDIV_M) >> RG_MEPL_FBDIV_S;
>>
>> ditto
>>
>>> +	xtal_div = 1 << xtal_div_tbl[dividx];
>>> +	ddr_clk = fb * xtal_clk / xtal_div;
>>> +
>>> +	bs = readl(priv->memc_base + MEMPLL1_REG);
>>> +	if (((bs & RG_MEPL_DIV2_SEL_M) >> RG_MEPL_DIV2_SEL_S) ==
>>> 0)
>>
>> ditto
>>
>> and you can just use
>>
>> 	if (!cond)
>>
>>> +		ddr_clk *= 2;
>>> +
>>> +	priv->cpu_clk = cpu_clk;
>>> +	priv->sys_clk = cpu_clk / 4;
>>> +	priv->ddr_clk = ddr_clk;
>>> +	priv->xtal_clk = xtal_clk;
>>
>> Please implement the above logic in get_rate(). For example:
>>
>> static ulong do_mt7621_clk_get_rate()
>> {
>> 	...
>> 	switch (clk->id) {
>> 		case CLK_SYS:
>> 			cpu_clk = do_mt7621_clk_get_rate(priv,
>> CLK_SYS);
>> 			return IS_ERROR_VALUE(cpu_clk) ? cpu_clk :
>> cpu_clk / 4;
>> 		...
>> 	}
>> }
>>
>> static ulong mt7621_clk_get_rate()
>> {
>> 	struct mt7621_clk_priv *priv = dev_get_priv(clk->dev);
>>
>> 	return do_mt7621_clk_get_rate(priv, clk->id);
>> }
> 
> ok
> 
>>
>>> +}
>>> +
>>> +static int mt7621_clk_probe(struct udevice *dev)
>>> +{
>>> +	struct mt7621_clk_priv *priv = dev_get_priv(dev);
>>> +	struct ofnode_phandle_args args;
>>> +	struct regmap *regmap;
>>> +	void __iomem *base;
>>> +	int ret;
>>> +
>>> +	/* get corresponding sysc phandle */
>>> +	ret = dev_read_phandle_with_args(dev, "mediatek,sysc",
>>> NULL, 0, 0,
>>> +					 &args);
>>> +	if (ret)
>>> +		return ret;
>>> +
>>> +	regmap = syscon_node_to_regmap(args.node);
>>
>> According to the Linux binding for this node, it is supposed to live
>> under the syscon already. So you can do
>>
>> 	syscon_node_to_regmap(dev_ofnode(dev_get_parent(dev)));
>>
>> and skip the phandle.
> 
> I'll try this
> 
>>
>>> +	if (IS_ERR(regmap))
>>> +		return PTR_ERR(regmap);
>>> +
>>> +	base = regmap_get_range(regmap, 0);
>>> +	if (!base) {
>>> +		dev_err(dev, "Unable to find sysc\n");
>>
>> dev_dbg (see doc/develop/driver-model/design.rst)
> 
> ok
> 
>>
>>> +		return -ENODEV;
>>> +	}
>>> +
>>> +	priv->sysc_base = ioremap_nocache((phys_addr_t)base,
>>> SYSC_MAP_SIZE);
>>> +
>>> +	/* get corresponding memc phandle */
>>> +	ret = dev_read_phandle_with_args(dev, "mediatek,memc",
>>> NULL, 0, 0,
>>
>> should be "ralink,memctl".
> 
> ok
> 
>>
>>> +					 &args);
>>> +	if (ret)
>>> +		return ret;
>>> +
>>> +	regmap = syscon_node_to_regmap(args.node);
>>> +	if (IS_ERR(regmap))
>>> +		return PTR_ERR(regmap);
>>
>> These two steps can be compined with syscon_regmap_lookup_by_phandle.
>>
>>> +	base = regmap_get_range(regmap, 0);
>>> +	if (!base) {
>>> +		dev_err(dev, "Unable to find memc\n");
>>
>> dev_dbg
>>
>>> +		return -ENODEV;
>>> +	}
>>> +
>>> +	priv->memc_base = ioremap_nocache((phys_addr_t)base,
>>> MEMC_MAP_SIZE);
>>> +
>>> +	mt7621_get_clocks(priv);
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +static const struct udevice_id mt7621_clk_ids[] = {
>>> +	{ .compatible = "mediatek,mt7621-clk" },
>>> +	{ }
>>> +};
>>> +
>>> +U_BOOT_DRIVER(mt7621_clk) = {
>>> +	.name = "mt7621-clk",
>>> +	.id = UCLASS_CLK,
>>> +	.of_match = mt7621_clk_ids,
>>> +	.probe = mt7621_clk_probe,
>>> +	.priv_auto = sizeof(struct mt7621_clk_priv),
>>> +	.ops = &mt7621_clk_ops,
>>> +};
>>> diff --git a/include/dt-bindings/clock/mt7621-clk.h b/include/dt-
>>> bindings/clock/mt7621-clk.h
>>> new file mode 100644
>>> index 0000000000..b24aef351c
>>> --- /dev/null
>>> +++ b/include/dt-bindings/clock/mt7621-clk.h
>>> @@ -0,0 +1,42 @@
>>> +/* SPDX-License-Identifier: GPL-2.0 */
>>> +/*
>>> + * Copyright (C) 2021 MediaTek Inc. All rights reserved.
>>> + *
>>> + * Author:  Weijie Gao <weijie.gao@mediatek.com>
>>> + */
>>> +
>>> +#ifndef _DT_BINDINGS_MT7621_CLK_H_
>>> +#define _DT_BINDINGS_MT7621_CLK_H_
>>> +
>>> +/* Base clocks */
>>> +#define CLK_MIPS_CNT			36
>>> +#define CLK_SYS				35
>>> +#define CLK_DDR				34
>>> +#define CLK_CPU				33
>>> +#define CLK_XTAL			32
>>
>> Why is there a gap?
> 
> 0~31 values are bits in clock gate register, and bit 31 is unused.
> values above 31 represent clock sources not defined in the gate
> register.

OK, but above your check is for clock IDs less than 31. So shouldn't it
be something like

	if (clk->id > CLK_SDXC)
		return -EINVAL;

>>
>>> +/* Peripheral clocks */
>>> +#define CLK_SDXC			30
>>> +#define CLK_CRYPTO			29
>>> +#define CLK_PCIE2			26
>>> +#define CLK_PCIE1			25
>>> +#define CLK_PCIE0			24
>>> +#define CLK_GMAC			23
>>> +#define CLK_UART3			21
>>> +#define CLK_UART2			20
>>> +#define CLK_UART1			19
>>> +#define CLK_SPI				18
>>> +#define CLK_I2S				17
>>> +#define CLK_I2C				16
>>> +#define CLK_NFI				15
>>> +#define CLK_GDMA			14
>>> +#define CLK_PIO				13
>>> +#define CLK_PCM				11
>>> +#define CLK_MC				10
>>> +#define CLK_INTC			9
>>> +#define CLK_TIMER			8
>>> +#define CLK_SPDIFTX			7
>>> +#define CLK_FE				6
>>> +#define CLK_HSDMA			5
>>> +
>>> +#endif /* _DT_BINDINGS_MT7621_CLK_H_ */
>>
>> This file looks very different from
>> include/dt-bindings/clock/mt7621-clk.h in Linux. In particular, it is
>> backwards, the IDs are different (HSDMA is 8 in Linux but 5 here),
> 
> 5 directly represents the bit in clock gate register, which means a
> mapping must be done for the include/dt-bindings/clock/mt7621-clk.h
> (i.e. subtract by 3) in kernel.
> 
> btw, the file in kernel is not submitted by mediatek.

I think aligning the defines with the names in the datasheet is a good
idea. Can you send a patch to Linux to update the names of the defines?

>> some
>> of the IDs are named differently (SP_DIVTX vs SPDIFTX), and there is
>> no
>> MT7621 prefix. Can you comment on these? Are they deliberate?
> 
> The name SPDIFTX comes from the MT7621 programming guide of mediatek.
> Adding MT7621 seems better.
>> Note that
>> in general, numerical IDs should be kept the same between Linux and
>> U-Boot so we can use the same device tree. If you need to map between
>> logical clock ID and a position in a register, I suggest something
>> like
>>
>> struct {
>> 	u8 gate_bit;
>> } clocks {
>> 	[CLK_HSDMA] = { .gate_bit = 5 },
>> };
>>
> 
> This is a driver dedicated for u-boot, and actually only the SYS clock
> is used. I believe using correct gate bit number is clearer.

It is fine to implement only the necessary functionality, but it should
be done in a way which is easy to extend in the future, and which won't
cause us compatibility problems.

Generally, I would like to preserve both source and binary compatibility
with Linux where possible. I am not sure whether they are hard
requirements, so I made a post regarding that question [1]. For now,
addressing my above comments will be fine.

--Sean

[1] https://lore.kernel.org/u-boot/c670a4cc-b234-03d4-adfb-e6a8560c2d86@gmail.com/T/#u

  reply	other threads:[~2021-12-15 16:12 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-19  1:35 [PATCH v2 00/14] Add support for MediaTek MT7621 SoC Weijie Gao
2021-11-19  1:35 ` [PATCH v2 01/14] mips: mtmips: add " Weijie Gao
2021-11-19  1:35 ` [PATCH v2 02/14] mips: mtmips: add two reference boards for mt7621 Weijie Gao
2021-11-19  1:35 ` [PATCH v2 03/14] clk: mtmips: add clock driver for MediaTek MT7621 SoC Weijie Gao
2021-11-26 17:44   ` Sean Anderson
2021-12-03 10:06     ` Weijie Gao
2021-12-15 16:11       ` Sean Anderson [this message]
2021-12-27  8:06         ` Weijie Gao
2021-12-31  2:30           ` Sean Anderson
2021-11-19  1:36 ` [PATCH v2 04/14] reset: mtmips: add reset controller support " Weijie Gao
2021-11-19  1:36 ` [PATCH v2 05/14] pinctrl: mtmips: add " Weijie Gao
2021-11-19  1:36 ` [PATCH v2 06/14] nand: raw: " Weijie Gao
2021-11-19  1:36 ` [PATCH v2 07/14] usb: xhci-mtk: " Weijie Gao
2021-11-19  1:36 ` [PATCH v2 08/14] phy: mtk-tphy: " Weijie Gao
2021-11-19  1:36 ` [PATCH v2 09/14] spi: " Weijie Gao
2021-11-19  1:36 ` [PATCH v2 10/14] gpio: " Weijie Gao
2021-11-19  1:36 ` [PATCH v2 11/14] watchdog: " Weijie Gao
2021-11-19  1:37 ` [PATCH v2 12/14] mmc: mediatek: " Weijie Gao
2021-11-22 23:41   ` Jaehoon Chung
2021-11-19  1:37 ` [PATCH v2 13/14] net: " Weijie Gao
2021-11-21 19:14   ` Ramon Fried
2021-11-22  8:09     ` Weijie Gao
2021-11-19  1:37 ` [PATCH v2 14/14] MAINTAINERS: update maintainer for MediaTek MIPS platform Weijie Gao

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=eb7342eb-9fee-a0be-49ab-32dce9f4550e@gmail.com \
    --to=seanga2@gmail.com \
    --cc=GSS_MTK_Uboot_upstream@mediatek.com \
    --cc=lukma@denx.de \
    --cc=u-boot@lists.denx.de \
    --cc=weijie.gao@mediatek.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.