Re: [PATCH V2 2/2] drivers: input: keyboard

On Wed, Jan 08, 2020 at 02:29:23PM +0800, fengping yu wrote:

From: "fengping.yu" <redacted>

add mtk keypad driver

A bit more details, perhaps?
Also, what about proper English grammar and punctuation?

Change-Id: I20bb26ee4112f51f60476b7ff7d6e42b43f729dc

Are you sure it's ready for upstream?

...

 arch/arm64/configs/defconfig                  |   1 +

Why?

...

Why above in this patch?

...

+obj-$(CONFIG_KEYBOARD_MTK) 		+= mtk-kpd.o

TABs/spaces mix.

...

+#include <linux/atomic.h>
+#include <linux/clk.h>
+#include <linux/debugfs.h>
+#include <linux/delay.h>
+#include <linux/fs.h>
+#include <linux/gpio.h>
+#include <linux/init.h>
+#include <linux/input.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/ioctl.h>
+#include <linux/kernel.h>
+#include <linux/miscdevice.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/pinctrl/consumer.h>
+#include <linux/platform_device.h>
+#include <linux/pm_wakeup.h>
+#include <linux/regmap.h>
+#include <linux/timer.h>
+#include <linux/workqueue.h>

Do you really need all of them?!

...

+#define KP_STA			(0x0000)
+#define KP_MEM1			(0x0004)
+#define KP_MEM2			(0x0008)
+#define KP_MEM3			(0x000c)
+#define KP_MEM4			(0x0010)
+#define KP_MEM5			(0x0014)
+#define KP_DEBOUNCE		(0x0018)
+#define KP_SEL			(0x0020)
+#define KP_EN			(0x0024)

What the purpose of parentheses?

...

+#define KP_COL0_SEL		(1 << 10)
+#define KP_COL1_SEL		(1 << 11)
+#define KP_COL2_SEL		(1 << 12)

BIT()

+#define KPD_DEBOUNCE_MASK	((1U << 14) - 1)
+#define KPD_DOUBLE_KEY_MASK	(1U << 0)

GENMASK() / BIT_MASK() / BIT()

...

+static int kpd_pdrv_probe(struct platform_device *pdev);

Why this forward declaration is present?

...

+static void kpd_get_keymap_state(void __iomem *kp_base, u16 state[])
+{
+	state[0] = readw(kp_base + KP_MEM1);
+	state[1] = readw(kp_base + KP_MEM2);
+	state[2] = readw(kp_base + KP_MEM3);
+	state[3] = readw(kp_base + KP_MEM4);
+	state[4] = readw(kp_base + KP_MEM5);

+	pr_debug("kpd register = %x %x %x %x %x\n",
+		 state[0], state[1], state[2], state[3], state[4]);

Consider print_hex_dump() instead. But do you need this at all?

+}

...

+		for (j = 0; j < 16U; j++) {
+			mask = (u16)1 << j;
+			if (!(change & mask))
+				continue;

for_each_set_bit() ?

+		}

...

+static int kpd_get_dts_info(struct mtk_keypad *keypad,
+			    struct device_node *node)
+{
+	int ret;

Consider pr_*() -> dev_*() in all below code.

+
+	ret = of_property_read_u32(node, "mediatek,key-debounce-ms",
+				   &keypad->key_debounce);
+	if (ret) {
+		pr_debug("read mediatek,key-debounce-ms error.\n");
+		return ret;
+	}
+
+	ret = of_property_read_u32(node, "mediatek,hw-map-num",
+				   &keypad->hw_map_num);
+	if (ret) {
+		pr_debug("read mediatek,hw-map-num error.\n");
+		return ret;
+	}
+
+	if (keypad->hw_map_num > KPD_NUM_KEYS) {
+		pr_debug("hw-map-num error, it cannot bigger than %d.\n",
+			 KPD_NUM_KEYS);

dev_err() ?

+		return -EINVAL;
+	}
+
+	ret = of_property_read_u32_array(node, "mediatek,hw-init-map",
+					 keypad->hw_init_map,
+					 keypad->hw_map_num);
+
+	if (ret) {
+		pr_debug("hw-init-map was not defined in dts.\n");
+		return ret;
+	}
+
+	pr_debug("deb= %d\n", keypad->key_debounce);
+
+	return 0;
+}

It's for ARM64 which might support ACPI. Perhaps you may use device property
API instead of be OF-centric.

...

+	keypad_pinctrl = devm_pinctrl_get(dev);
+	if (IS_ERR(keypad_pinctrl)) {

+		pr_debug("Cannot find keypad_pinctrl!\n");
+

These kind of messages are not for production.

+		return (int)PTR_ERR(keypad_pinctrl);

Why explicit casting? (Same for similar places all over the driver)
Is it really a fatal error?

+	}

...

+	return pinctrl_select_state(keypad_pinctrl,
+				kpd_default);

One line. Perhaps you need to configure your text editor.

...

+	keypad->clk = devm_clk_get(&pdev->dev, "kpd");
+	if (IS_ERR(keypad->clk)) {

+		pr_notice("get kpd-clk fail: %d\n", (int)PTR_ERR(keypad->clk));

Noise as a lot of such messages.

+		return (int)PTR_ERR(keypad->clk);

Casting?!

+	}

...

+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res) {
+		err = -ENODEV;
+		goto err_unprepare_clk;
+	}
+
+	keypad->base = devm_ioremap(&pdev->dev, res->start,
+				    resource_size(res));
+	if (!keypad->base) {
+		pr_notice("KP iomap failed\n");
+		err = -EBUSY;
+		goto err_unprepare_clk;
+	}

What's wrong with devm_platform_ioremap_resource()?

...

+	keypad->irqnr = irq_of_parse_and_map(pdev->dev.of_node, 0);
+	if (!keypad->irqnr) {
+		pr_notice("KP get irqnr failed\n");
+		err = -ENODEV;
+		goto err_unprepare_clk;
+	}

Can't you use platform_get_irq()?
Why not?

...

+	pr_info("kp base: 0x%p, addr:0x%p,  kp irq: %d\n",
+		keypad->base, &keypad->base, keypad->irqnr);

Useless message. Moreover I believe it has wrong specifiers in use.

...

+	memset(keypad->keymap_state, 0xff, sizeof(keypad->keymap_state));

Shouldn't be bitmap_fill()? In such case the variable has to be type of
unsigned long.

...

+	writew((u16)(keypad->key_debounce & KPD_DEBOUNCE_MASK),
+		     keypad->base + KP_DEBOUNCE);

Why explicit casting?

...

+	/* register IRQ */
+	err = request_irq(keypad->irqnr, kpd_irq_handler, IRQF_TRIGGER_NONE,
+			  KPD_NAME, keypad);
+	if (err) {
+		pr_notice("register IRQ failed (%d)\n", err);
+		goto err_irq;
+	}

Why not threaded IRQ handler instead of ugly tasklet?

...

+	pr_info("kpd_probe OK.\n");

Noise.

...

+static const struct of_device_id kpd_of_match[] = {
+	{.compatible = "mediatek,mt6779-keypad"},
+	{.compatible = "mediatek,kp"},

+	{},

Comma is not needed in terminator lines.

+};

-- 
With Best Regards,
Andy Shevchenko