Re: [PATCH v2 3/4] drm: bridge: add it66121 driver

From: Andy Shevchenko
Date: Wed Mar 11 2020 - 09:55:44 EST


On Wed, Mar 11, 2020 at 01:51:34PM +0100, Phong LE wrote:
> This commit is a simple driver for bridge HMDI it66121.
> The input format is RBG and there is no color conversion.
> Audio, HDCP and CEC are not supported yet.

I guess you should have been told in your company how to use get_maintainer.pl
to avoid spamming people.

Hint:
scripts/get_maintainer.pl --git --git-min-percent=67 ...

...

> + * Copyright (C) 2020 BayLibre, SAS
> + * Author: Phong LE <ple@xxxxxxxxxxxx>
> + * Copyright (C) 2018-2019, Artem Mygaiev
> + * Copyright (C) 2017, Fresco Logic, Incorporated.

This too compact to read, perhaps squeeze blank line after your (c) note
followed by description how the rest (c) appear in the file.

> + *

Redundant line.

> + */

...

> +#include <linux/of.h>

This can be replaced with property.h, see comments against ->probe().

> +#include <linux/of_device.h>
> +#include <linux/of_gpio.h>

I didn't find evidence of use of any of those.

...

> +#define IT66121_MASTER_SEL_REG 0x10

> +#define IT66121_AFE_DRV_REG 0x61

> +#define IT66121_INPUT_MODE_REG 0x70

> +#define IT66121_INPUT_CSC_REG 0x72

> +#define IT66121_AFE_XP_REG 0x62

> +#define IT66121_AFE_IP_REG 0x64

> +#define IT66121_AFE_XP_EC1_REG 0x68

> +#define IT66121_SW_RST_REG 0x04

> +#define IT66121_DDC_COMMAND_REG 0x15

> +#define IT66121_HDCP_REG 0x20

> +#define IT66121_INT_STATUS1_REG 0x06

> +#define IT66121_DDC_HEADER_REG 0x11

> +#define IT66121_DDC_OFFSET_REG 0x12
> +#define IT66121_DDC_BYTE_REG 0x13
> +#define IT66121_DDC_SEGMENT_REG 0x14
> +#define IT66121_DDC_RD_FIFO_REG 0x17

> +#define IT66121_CLK_BANK_REG 0x0F

> +#define IT66121_INT_REG 0x05

> +#define IT66121_INT_MASK1_REG 0x09

> +#define IT66121_INT_CLR1_REG 0x0C

> +#define IT66121_AV_MUTE_REG 0xC1

> +#define IT66121_PKT_GEN_CTRL_REG 0xC6

> +#define IT66121_AVIINFO_DB1_REG 0x158
> +#define IT66121_AVIINFO_DB2_REG 0x159
> +#define IT66121_AVIINFO_DB3_REG 0x15A
> +#define IT66121_AVIINFO_DB4_REG 0x15B
> +#define IT66121_AVIINFO_DB5_REG 0x15C
> +#define IT66121_AVIINFO_CSUM_REG 0x15D
> +#define IT66121_AVIINFO_DB6_REG 0x15E
> +#define IT66121_AVIINFO_DB7_REG 0x15F
> +#define IT66121_AVIINFO_DB8_REG 0x160
> +#define IT66121_AVIINFO_DB9_REG 0x161
> +#define IT66121_AVIINFO_DB10_REG 0x162
> +#define IT66121_AVIINFO_DB11_REG 0x163
> +#define IT66121_AVIINFO_DB12_REG 0x164
> +#define IT66121_AVIINFO_DB13_REG 0x165
> +
> +#define IT66121_AVI_INFO_PKT_REG 0xCD

> +#define IT66121_HDMI_MODE_REG 0xC0

> +#define IT66121_SYS_STATUS_REG 0x0E

> +#define IT66121_DDC_STATUS_REG 0x16

It's better to
a) keep register sorted by value (easy to be oriented);
b) have them in fixed width, e.g. 0x0CD.

...

> +#define IT66121_DEVICE_MASK 0x0F

GENMASK() ?

> +#define IT66121_EDID_SLEEP 20000
> +#define IT66121_EDID_TIMEOUT 200000

Care to add units?

> +#define IT66121_AFE_CLK_HIGH 80000

Also, what is the unit of this?

...

> + return regmap_write(ctx->regmap, IT66121_MASTER_SEL_REG,
> + IT66121_MASTER_SEL_HOST);

Indentation?
Same for other similar places.

...

> +static int it66121_configure_afe(struct it66121_ctx *ctx,
> + const struct drm_display_mode *mode)

...like this.

...

> + if (val & (IT66121_DDC_STATUS_NOACK | IT66121_DDC_STATUS_WAIT_BUS |
> + IT66121_DDC_STATUS_ARBI_LOSE))
> + return -EAGAIN;

Perhaps better to

u32 busy = IT66121_DDC_STATUS_NOACK | IT66121_DDC_STATUS_WAIT_BUS |
IT66121_DDC_STATUS_ARBI_LOSE;


if (val & busy)
return -EAGAIN;

?

> +
> + return 0;
> +}

...

> +static int it66121_abort_ddc_ops(struct it66121_ctx *ctx)
> +{
> + int ret;
> + unsigned int swreset, cpdesire;
> +
> + ret = regmap_read(ctx->regmap, IT66121_SW_RST_REG, &swreset);
> + if (ret)
> + return ret;
> +

> + ret = regmap_read(ctx->regmap, IT66121_HDCP_REG, &cpdesire);
> + if (ret)
> + return ret;
> +
> + ret = regmap_write(ctx->regmap, IT66121_HDCP_REG,
> + cpdesire & (~IT66121_HDCP_CPDESIRED & 0xFF));
> + if (ret)
> + return ret;

regmap_update_bits() ?

> +
> + ret = regmap_write(ctx->regmap, IT66121_SW_RST_REG,
> + swreset | IT66121_SW_RST_HDCP);
> + if (ret)
> + return ret;

This should surround the inner update, correct? Otherwise, regmap_update_bits()
can be used as well.

> +}

...

> +static int it66121_get_edid_block(void *context, u8 *buf,
> + unsigned int block, size_t len)
> +{

> + ret = regmap_write(ctx->regmap, IT66121_DDC_COMMAND_REG,
> + IT66121_DDC_COMMAND_EDID_READ);
> + if (ret)
> + return ret;
> +
> + offset += cnt;
> + remain -= cnt;

> + msleep(20);

Should be explained.

> +
> + ret = it66121_wait_ddc_ready(ctx);
> + if (ret)
> + return ret;
> +

> + do {
> + ret = regmap_read(ctx->regmap,
> + IT66121_DDC_RD_FIFO_REG, &val);
> + if (ret)
> + return ret;
> + *(buf++) = val;
> + cnt--;
> + } while (cnt > 0);

I'm wondering if regmap API has a helper for above like cases.

> + }
> +
> + return 0;
> +}

...

> +static int it66121_connector_get_modes(struct drm_connector *connector)
> +{

> + struct it66121_ctx *ctx = container_of(connector, struct it66121_ctx,
> + connector);

Seems like it can be an inline helper and in all places you can use

struct it66121_ctx *ctx = to_it66121_ctx(connector);


> + ret = drm_connector_update_edid_property(connector,
> + ctx->edid);

It quite fits one line, perhaps configure your editor to see 80 limit sharp?
Applies to all similar cases as well.

> + if (ret) {
> + DRM_ERROR("Failed to update EDID property: %d\n", ret);
> + goto unlock;
> + }

> +}

...

> + return (val & IT66121_SYS_STATUS_HPDETECT);

Too many parentheses.

...

> + max_clock = ctx->dual_edge ? 74250 : 148500;

Magic numbers? Also, It seems one definition is enough.

...

> + msleep(50);

Should be explained.

...

> + val = mute ? IT66121_AV_MUTE_ON : (~IT66121_AV_MUTE_ON & 0xFF);

val = mute ? IT66121_AV_MUTE_ON : 0;

> + ret = regmap_write_bits(ctx->regmap, IT66121_AV_MUTE_REG,
> + IT66121_AV_MUTE_ON, val);

...or even

mask = IT66121_AV_MUTE_ON;
val = mute ? mask : 0;

ret = regmap_write_bits(ctx->regmap, IT66121_AV_MUTE_REG, mask, val);

> + if (ret)
> + return ret;

...

> + regmap_write_bits(ctx->regmap, IT66121_CLK_BANK_REG,
> + IT66121_CLK_BANK_PWROFF_TXCLK,
> + ~IT66121_CLK_BANK_PWROFF_TXCLK & 0xFF);

Same idea. What the point of all this operations in the value parameter?
Please, fix over the code.

...


> + if (val & IT66121_SYS_STATUS_ACTIVE_IRQ) {

if (!(val & IT66121_SYS_STATUS_ACTIVE_IRQ))
goto unlock;

> + ret = regmap_read(ctx->regmap, IT66121_INT_STATUS1_REG, &val);
> + if (ret) {
> + dev_err(dev, "Cannot read STATUS1_REG %d\n", ret);
> + } else {
> + if (val & IT66121_INT_STATUS1_DDC_FIFOERR)
> + it66121_clear_ddc_fifo(ctx);

> + if (val & (IT66121_INT_STATUS1_DDC_BUSHANG |
> + IT66121_INT_STATUS1_DDC_NOACK))

Indentation.

> + it66121_abort_ddc_ops(ctx);
> + if (val & IT66121_INT_STATUS1_HPD_STATUS) {
> + regmap_write_bits(ctx->regmap,
> + IT66121_INT_CLR1_REG,
> + IT66121_INT_CLR1_HPD,
> + IT66121_INT_CLR1_HPD);
> +
> + if (!it66121_is_hpd_detect(ctx)) {
> + kfree(ctx->edid);
> + ctx->edid = NULL;
> + }
> + event = true;
> + }
> + }
> +
> + regmap_write_bits(ctx->regmap, IT66121_SYS_STATUS_REG,
> + IT66121_SYS_STATUS_CLEAR_IRQ,
> + IT66121_SYS_STATUS_CLEAR_IRQ);
> + }
> +
> +unlock:
> + mutex_unlock(&ctx->lock);

...

> +static int it66121_probe(struct i2c_client *client,
> + const struct i2c_device_id *id)
> +{

> + u8 ids[4];

Magic, also see below.

> + int i, ret;
> + struct it66121_ctx *ctx;
> + struct device *dev = &client->dev;

> + ctx->conf = (struct it66121_conf *)of_device_get_match_data(dev);

device_get_match_data()

In any case why explicit casting?

> + if (!ctx->conf)
> + return -ENODEV;
> +

> + ctx->dual_edge = of_property_read_bool(dev->of_node, "pclk-dual-edge");

device_property_read_bool()

> + for (i = 0; i < 4; i++) {

Magic.

> + regmap_read(ctx->regmap, i, &ret);
> + ids[i] = ret;
> + }

> +
> + if (ids[0] != IT66121_VENDOR_ID0 ||
> + ids[1] != IT66121_VENDOR_ID1 ||
> + ids[2] != IT66121_DEVICE_ID0 ||
> + ((ids[3] & IT66121_DEVICE_MASK) != IT66121_DEVICE_ID1)) {
> + ite66121_power_off(ctx);
> + return -ENODEV;
> + }
> +
> + ctx->bridge.funcs = &it66121_bridge_funcs;
> + ctx->bridge.of_node = dev->of_node;
> +
> + ret = devm_request_threaded_irq(dev, client->irq, NULL,
> + it66121_irq_threaded_handler,

> + IRQF_SHARED | IRQF_TRIGGER_LOW |
> + IRQF_ONESHOT,

Shouldn't flags come from appropriate resource provider (DT / ACPI / etc)?

> + dev_name(dev),
> + ctx);
> + if (ret < 0) {
> + dev_err(dev, "Failed to request irq %d:%d\n", client->irq, ret);
> + ite66121_power_off(ctx);
> + return ret;
> + }
> +
> + drm_bridge_add(&ctx->bridge);
> +
> + return 0;
> +}

...

> +static const struct of_device_id it66121_dt_match[] = {
> + { .compatible = "ite,it66121",
> + .data = &it66121_conf_simple,
> + },

> + { },

Terminator line doesn't need comma.

> +};

...

> +static const struct i2c_device_id it66121_id[] = {
> + { "it66121", 0 },

> + { },

Ditto.

> +};

--
With Best Regards,
Andy Shevchenko