Re: [PATCH V3 2/2] drivers: input: keyboard: add mtk keypad driver
From: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Date: 2020-01-14 16:36:15
Also in:
linux-arm-kernel, linux-devicetree, linux-mediatek, lkml
On Tue, Jan 14, 2020 at 08:27:41PM +0800, fengping.yu wrote:
Signed-off-by: fengping.yu <redacted> Add matrix keypad driver for MTK SoC.
Something wrong with ordering. ...
+obj-$(CONFIG_KEYBOARD_MTK) += mtk-kpd.o
TABs / space mix. I think you ignored my comment. ...
quoted hunk
+#include <linux/clk.h> +#include <linux/fs.h> +#include <linux/gpio.h> +#include <linux/init.h> +#include <linux/input/matrix_keypad.h> +#include <linux/interrupt.h> +#include <linux/io.h> +#include <linux/module.h> +#include <linux/of_address.h> +#include <linux/of_irq.h> +#include <linux/pinctrl/consumer.h> +#include <linux/platform_device.h> +#include <linux/regmap.h>
...
+#define KPD_DEBOUNCE_MASK GENMASK_ULL(13, 0)
ULL? Why? ...
+static void kpd_get_keymap_state(void __iomem *kp_base, u32 state[])
u32 *state
+{+ memset_io(state, 0xff, KPD_NUM_MEMS);
This is strange. What is this for and why is it needed?
+ memcpy_fromio(state, kp_base, KPD_NUM_MEMS); +}
...
quoted hunk
+static irqreturn_t kpd_irq_handler(int irq, void *dev_id) +{ + /* use _nosync to avoid deadlock */ + struct mtk_keypad *keypad = dev_id; + unsigned short *keycode = keypad->input_dev->keycode; + u32 new_state[KPD_NUM_MEMS], mask; + unsigned long change = 0; + int bit_nr; + int pressed; + u32 row_shift = get_count_order(keypad->n_cols); + unsigned short code; + int i; + + disable_irq_nosync(keypad->irqnr); + + kpd_get_keymap_state(keypad->base, new_state); + + for (i = 0; i < KPD_NUM_MEMS; i++) { + change = new_state[i] ^ keypad->keymap_state[i]; + if (!change) + continue; + + for_each_set_bit(bit_nr, &change, 32) {
+ mask = 1 << bit_nr; + pressed = (new_state[i] & mask) == 0U;
test_bit() ? Can't new_state be unsigned long? In this case you may get rid of outer loop I suppose.
quoted hunk
+ dev_dbg(&keypad->input_dev->dev, + "%s", pressed ? "pressed" : "released"); + + code = keycode[MATRIX_SCAN_CODE(i, bit_nr, row_shift)]; + + input_report_key(keypad->input_dev, code, pressed); + input_sync(keypad->input_dev); + + dev_dbg(&keypad->input_dev->dev, + "report Linux keycode = %d\n", code); + } + } + + memcpy(keypad->keymap_state, new_state, sizeof(new_state)); + + enable_irq(keypad->irqnr); + + return IRQ_HANDLED; +}
...
quoted hunk
+static int kpd_get_dts_info(struct mtk_keypad *keypad) +{ + int ret; + struct device *dev = keypad->input_dev->dev.parent; + struct device_node *node = dev->of_node; + + ret = matrix_keypad_parse_properties(dev, &keypad->n_rows, + &keypad->n_cols);
+
Redundant line.
quoted hunk
+ if (ret) { + dev_err(dev, "failed to parse keypad params.\n"); + return ret; + } + + ret = of_property_read_u32(node, "mediatek,debounce-us", + &keypad->key_debounce); + if (ret) {
+ pr_debug("read mediatek,debounce-us error.\n");dev_dbg() ?
quoted hunk
+ return ret; + } + + keypad->wakeup = of_property_read_bool(node, "wakeup-source"); + + dev_dbg(dev, "n_row=%d n_col=%d debounce=%d.\n", + keypad->n_rows, keypad->n_cols, + keypad->key_debounce); + + return 0; +}
...
quoted hunk
+static int kpd_gpio_init(struct device *dev) +{ + struct pinctrl *keypad_pinctrl; + struct pinctrl_state *kpd_default; + + keypad_pinctrl = devm_pinctrl_get(dev); + if (IS_ERR(keypad_pinctrl)) { + dev_err(dev, "Cannot find keypad_pinctrl!\n");
+
Redundant line
quoted hunk
+ return PTR_ERR(keypad_pinctrl); + } + + kpd_default = pinctrl_lookup_state(keypad_pinctrl, "default"); + if (IS_ERR(kpd_default)) { + dev_err(dev, "Cannot find ecall_state!\n");
+
Ditto.
quoted hunk
+ return PTR_ERR(kpd_default); + } + + return pinctrl_select_state(keypad_pinctrl, kpd_default); +}
...
+ int err;
err or ret? Be consistent with the name of variable that keeps returned code.
quoted hunk
+ if (!pdev->dev.of_node) { + dev_err(&pdev->dev, "No device tree data present.\n"); + return -EINVAL; + }
I think this is redundant. below calls will fail when property is not found.
quoted hunk
+ keypad = devm_kzalloc(&pdev->dev, sizeof(*keypad), GFP_KERNEL); + if (!keypad) + return -ENOMEM; +
+ memset(keypad->keymap_state, 0xff, sizeof(keypad->keymap_state));
bitmap_fill() ?
quoted hunk
+ err = matrix_keypad_build_keymap(NULL, NULL, + keypad->n_rows, + keypad->n_cols, + NULL, + keypad->input_dev);
quoted hunk
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0); + if (!res) { + dev_err(&pdev->dev, "get IO memory resource failed.\n"); + return -ENODEV; + } + + keypad->base = devm_ioremap_resource(&pdev->dev, res);
devm_platform_ioremap_resource()
quoted hunk
+ if (IS_ERR(keypad->base)) { + dev_err(&pdev->dev, "KP iomap failed\n"); + return PTR_ERR(keypad->base); + } + + writew((keypad->key_debounce * 32 / 1000 & KPD_DEBOUNCE_MASK),
Useless parentheses. It's not lisp!
+ keypad->base + KP_DEBOUNCE);
quoted hunk
+ err = clk_prepare_enable(keypad->clk); + if (err) { + dev_err(&pdev->dev, "kpd-clk prepare enable failed.\n"); + return err; + } + + err = kpd_gpio_init(&pdev->dev); + if (err) { + dev_err(&pdev->dev, "gpio init failed\n");
+ return err;
Starting from here down the clocks left in a wrong state.
quoted hunk
+ } + + keypad->irqnr = platform_get_irq(pdev, 0);
+ if (!keypad->irqnr) {This is wrong...
+ dev_err(&pdev->dev, "KP get irqnr failed\n");
+ return -EINVAL;
...and so does this.
+ }
-- With Best Regards, Andy Shevchenko