Skip to contentSkip to conversation

hwmon: add fan/pwm driver for corsair h100i platinum

PATCH Superseded jaap aarts 2 participants 4 messages Aug 16, 2020

Subsystems (2)

v5 of 7 Superseded Latest is v7 →

lore
Share & tools
lore

jaap aarts

author
#0 lore raw

PATCH v5 1/1 hwmon: add fan/pwm driver for corsair h100i platinum

Adds fan/pwm support for H100i platinum.
Custom temp/fan curves are not supported.

v5:
- fixed #defines not being caps
- check amount of written bytes + fixed expecting wrong amount
- moved to sizeof(var) instead of sizeof(type)
- fixed bounds checks
- merged cases and simplefied switch statements
- more error checking
- destroy the mutex

---
 drivers/hwmon/Kconfig               |   7 +
 drivers/hwmon/Makefile              |   1 +
 drivers/hwmon/corsair_hydro_i_pro.c | 701 ++++++++++++++++++++++++++++
 3 files changed, 709 insertions(+)
 create mode 100644 drivers/hwmon/corsair_hydro_i_pro.c
drivers/hwmon/Kconfig +7 -0
index 288ae9f63588..f466b72d0f67 100644
@@ -378,6 +378,13 @@ config SENSORS_ARM_SCPI
 	  and power sensors available on ARM Ltd's SCP based platforms. The
 	  actual number and type of sensors exported depend on the platform.
 
+config SENSORS_CORSAIR_HYDRO_I_PRO
+	tristate "Corsair hydro HXXXi pro driver"
+	depends on USB
+	help
+	  If you say yes here you get support for the corsair hydro HXXXi pro
+	  range of devices.
+
 config SENSORS_ASB100
 	tristate "Asus ASB100 Bach"
 	depends on X86 && I2C
drivers/hwmon/Makefile +1 -0
index 3e32c21f5efe..ec63294b3ef1 100644
@@ -20,6 +20,7 @@ obj-$(CONFIG_SENSORS_W83793)	+= w83793.o
 obj-$(CONFIG_SENSORS_W83795)	+= w83795.o
 obj-$(CONFIG_SENSORS_W83781D)	+= w83781d.o
 obj-$(CONFIG_SENSORS_W83791D)	+= w83791d.o
+obj-$(CONFIG_SENSORS_CORSAIR_HYDRO_I_PRO)	+= corsair_hydro_i_pro.o
 
 obj-$(CONFIG_SENSORS_AB8500)	+= abx500.o ab8500.o
 obj-$(CONFIG_SENSORS_ABITUGURU)	+= abituguru.o
drivers/hwmon/corsair_hydro_i_pro.c new file +700 -0

700 lines not on this page · raw message

Signature · 1 line
2.28.0
#1 lore raw

Hello,

I added a couple notes about the code inline.

drivers/hwmon/corsair_hydro_i_pro.c line 61

Quoted 7 lines

[...]

drivers/hwmon/corsair_hydro_i_pro.c new file +6 -0
+#define MAX_FAN_COUNT 2
+#define MAX_PWM_CHANNEL_COUNT MAX_FAN_COUNT
+
+struct hwmon_data {
+	struct hydro_i_pro_device *hdev;
+	int channel_count;

This is a nitpick, but the 'channel_count' value comes from the 'fancount' of an entry of 'config_table', which has type 'u8', so I don't see any need for it to be an 'int'.

drivers/hwmon/corsair_hydro_i_pro.c line 80

Quoted 4 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +3 -0
+	void *channel_data[MAX_PWM_CHANNEL_COUNT];
+};
[...]
+struct curve_point quiet_curve[FAN_CURVE_COUNT] = {

This should be 'static' (and possibly 'const'?).

Quoted 24 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +24 -0
+	{
+		.temp = 0x1F,
+		.pwm = 0x15,
+	},
+	{
+		.temp = 0x21,
+		.pwm = 0x1E,
+	},
+	{
+		.temp = 0x24,
+		.pwm = 0x25,
+	},
+	{
+		.temp = 0x27,
+		.pwm = 0x2D,
+	},
+	{
+		.temp = 0x29,
+		.pwm = 0x38,
+	},
+	{
+		.temp = 0x2C,
+		.pwm = 0x4A,
+	},

drivers/hwmon/corsair_hydro_i_pro.c line 125

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +9 -0
+	{
+		.temp = 0x2F,
+		.pwm = 0x64,
+	},
+};
[...]
+static const struct device_config config_table[] = {
+	{
+		.vendor_id = 0x1b1c,
+		.product_id = 0x0c15,

If I see it correctly, you never use 'vendor_id', nor 'product_id', right?

Quoted 4 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +4 -0
+		.fancount = 2,
+		.rpm_min = 200,
+		.rpm_max = 3000,
+		.pwm_max = 100,

drivers/hwmon/corsair_hydro_i_pro.c line 147

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +9 -0
+		.name = "corsair_H100i_pro",
+		.hwmon_info = dual_fan,
+	},
+};
[...]
+
+#define SUCCES_LENGTH 3
+#define SUCCES_CODE 0x12, 0x34
+
+static bool check_succes(enum opcodes command, char ret[SUCCES_LENGTH])

I missed this it seems, but my suggestion about the 'static' before the size applies here as well.

Quoted 5 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +5 -0
+{
+	char success[SUCCES_LENGTH] = { command, SUCCES_CODE };
+
+	return memcmp(ret, success, SUCCES_LENGTH) == 0;
+}

drivers/hwmon/corsair_hydro_i_pro.c line 162

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+
+static int acquire_lock(struct hydro_i_pro_device *hdev)
+{
+	int retval = usb_autopm_get_interface(hdev->interface);
+
+	if (retval)
+		return retval;
+
+	mutex_lock(&hdev->io_mutex);
+	return 0;

I think this should be

return mutex_lock_interruptible(...);

as per
https://www.kernel.org/doc/html/latest/kernel-hacking/locking.html#locking-only-in-user-context

drivers/hwmon/corsair_hydro_i_pro.c line 299

Quoted 11 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+}
[...]
+static int set_fan_target_pwm(struct hydro_i_pro_device *hdev,
+			      struct hwmon_fan_data *fan_data, u8 val)
+{
+	int retval;
+	unsigned char *send_buf = hdev->bulk_out_buffer;
+	unsigned char *recv_buf = hdev->bulk_in_buffer;
+
+	fan_data->fan_target = 0;
+	fan_data->fan_pwm_target = val;

Personally, I'd place these two lines at the end of the function so that they are not modified if the USB transaction fails.
(same applies to set_fan_pwm_curve() and set_fan_target_rpm())

drivers/hwmon/corsair_hydro_i_pro.c line 306

Quoted 7 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +7 -0
+
+	send_buf[0] = PWM_FAN_TARGET_CMD;
+	send_buf[1] = fan_data->fan_channel;
+	send_buf[2] = fan_data->fan_pwm_target;
+	dev_info(&hdev->udev->dev, "debug:%d,%d,%d", send_buf[0], send_buf[1],
+		 send_buf[2]);
+	dev_info(&hdev->udev->dev, "val:%d", fan_data->fan_pwm_target);

This should be dev_dbg() in my opinion if you intend to have such messages.

Quoted 47 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +46 -0
+
+	retval = transfer_usb(hdev, send_buf, recv_buf, 3, 3);
+	if (retval)
+		return retval;
+
+	if (!check_succes(send_buf[0], recv_buf)) {
+		dev_warn(
+			&hdev->udev->dev,
+			"failed setting fan pwm for %s on channel %d %d,%d,%d\n",
+			hdev->config->name, fan_data->fan_channel, recv_buf[0],
+			recv_buf[1], recv_buf[2]);
+		return -EINVAL;
+	}
+	return 0;
+}
[...]
+static int hwmon_write(struct device *dev, enum hwmon_sensor_types type,
+		       u32 attr, int channel, long val)
+{
+	struct hwmon_data *data = dev_get_drvdata(dev);
+	struct hydro_i_pro_device *hdev = data->hdev;
+	struct hwmon_fan_data *fan_data;
+	int retval = 0;
+
+	switch (type) {
+	case hwmon_fan:
+		switch (attr) {
+		case hwmon_fan_target:
+			fan_data = data->channel_data[channel];
+			if (fan_data->mode != 1)
+				return -EINVAL;
+
+			val = clamp_val(val, hdev->config->rpm_min,
+					hdev->config->rpm_max);
+
+			retval = acquire_lock(hdev);
+			if (retval)
+				goto exit;
+
+			retval = set_fan_target_rpm(hdev, fan_data, val);
+			if (retval)
+				goto cleanup;
+
+			break;
+		default:
+			return -EINVAL;
+		}

drivers/hwmon/corsair_hydro_i_pro.c line 387

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+		break;
+	case hwmon_pwm:
+		switch (attr) {
+		case hwmon_pwm_input:
+			fan_data = data->channel_data[channel];
+			if (fan_data->mode != 1)
+				return -EINVAL;
+
+			val = clamp_val(val, 0, hdev->config->pwm_max);
+			dev_info(&hdev->udev->dev, "val:%ld", val);

Same here regarding dev_dbg().

Quoted 14 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +14 -0
+			retval = acquire_lock(hdev);
+			if (retval)
+				goto exit;
+
+			retval = set_fan_target_pwm(hdev, fan_data, val);
+			if (retval)
+				goto cleanup;
+
+			break;
+		case hwmon_pwm_enable:
+			fan_data = data->channel_data[channel];
+
+			switch (val) {
+			case 2:

drivers/hwmon/corsair_hydro_i_pro.c line 411

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+			case 0:
+				retval = acquire_lock(hdev);
+				if (retval)
+					goto exit;
+
+				retval = set_fan_pwm_curve(hdev, fan_data,
+							   DEFAULT_CURVE);
+				if (retval)
+					goto cleanup;
+				fan_data->mode = 0;

This could be 'fan_data->mode = val', no?

Quoted 62 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +62 -0
+				break;
+			case 1:
+				retval = acquire_lock(hdev);
+				if (retval)
+					goto exit;
+
+				if (fan_data->fan_target != 0) {
+					retval = set_fan_target_rpm(
+						hdev, fan_data,
+						fan_data->fan_target);
+					if (retval)
+						goto cleanup;
+				} else if (fan_data->fan_pwm_target != 0) {
+					retval = set_fan_target_pwm(
+						hdev, fan_data,
+						fan_data->fan_pwm_target);
+					if (retval)
+						goto cleanup;
+				}
+				fan_data->mode = 1;
+				break;
+			default:
+				return -EINVAL;
+			}
+			break;
+		default:
+			return -EINVAL;
+		}
+		break;
+	default:
+		return -EINVAL;
+	}
+
+cleanup:
+	mutex_unlock(&hdev->io_mutex);
+	usb_autopm_put_interface(hdev->interface);
+exit:
+	return retval;
+}
+
+static int hwmon_read(struct device *dev, enum hwmon_sensor_types type,
+		      u32 attr, int channel, long *val)
+{
+	struct hwmon_data *data = dev_get_drvdata(dev);
+	struct hydro_i_pro_device *hdev = data->hdev;
+	struct hwmon_fan_data *fan_data;
+	int retval = 0;
+
+	switch (type) {
+	case hwmon_fan:
+		switch (attr) {
+		case hwmon_fan_input:
+			fan_data = data->channel_data[channel];
+
+			retval = acquire_lock(hdev);
+			if (retval)
+				goto exit;
+
+			retval = get_fan_current_rpm(hdev, fan_data, val);
+			break;
+		case hwmon_fan_target:
+			fan_data = data->channel_data[channel];

drivers/hwmon/corsair_hydro_i_pro.c line 483

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+			if (fan_data->mode != 1) {
+				*val = 0;
+				goto exit;
+			}
+			*val = fan_data->fan_target;
+			goto exit;
+		case hwmon_fan_min:
+			*val = hdev->config->rpm_min;
+			goto exit;
+

Any reason why you don't handle 'hwmon_fan_max' as well?

Quoted 32 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +32 -0
+		default:
+			return -EINVAL;
+		}
+		break;
+	case hwmon_pwm:
+		switch (attr) {
+		case hwmon_pwm_enable:
+			fan_data = data->channel_data[channel];
+			*val = fan_data->mode;
+			goto exit;
+		default:
+			return -EINVAL;
+		}
+		goto exit;
+
+	default:
+		return -EINVAL;
+	}
+
+	mutex_unlock(&hdev->io_mutex);
+	usb_autopm_put_interface(hdev->interface);
+exit:
+	return retval;
+}
+
+static const struct hwmon_ops i_pro_ops = {
+	.is_visible = hwmon_is_visible,
+	.read = hwmon_read,
+	.write = hwmon_write,
+};
+
+static int hwmon_init(struct hydro_i_pro_device *hdev)

drivers/hwmon/corsair_hydro_i_pro.c line 525

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+{
+	u8 fan_id;
+	struct device *hwmon_dev;
+	struct hwmon_fan_data *fan;
+	struct hwmon_data *data;
+	struct hwmon_chip_info *hwmon_info;
+
+	data = devm_kzalloc(&hdev->udev->dev, sizeof(*data), GFP_KERNEL);
+	hwmon_info = devm_kzalloc(&hdev->udev->dev,
+				  sizeof(struct hwmon_chip_info), GFP_KERNEL);

I think you missed this 'sizeof' when coverting to 'sizeof(*ptr)'.

drivers/hwmon/corsair_hydro_i_pro.c line 535

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+
+	/* You did something bad!! Either adjust  MAX_FAN_COUNT or the fancount for the config!*/
+	WARN_ON(hdev->config->fancount >= MAX_PWM_CHANNEL_COUNT);
+	data->channel_count =
+		clamp_val(hdev->config->fancount, 0, MAX_PWM_CHANNEL_COUNT);
+
+	/* For each fan create a data channel a fan config entry and a pwm config entry */
+	for (fan_id = 0; fan_id < data->channel_count; fan_id++) {
+		fan = devm_kzalloc(&hdev->udev->dev,
+				   sizeof(struct hwmon_fan_data), GFP_KERNEL);

And this one as well.

Quoted 7 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +7 -0
+		fan->fan_channel = fan_id;
+		fan->mode = 0;
+		data->channel_data[fan_id] = fan;
+	}
+
+	hwmon_info->ops = &i_pro_ops;
+	hwmon_info->info = hdev->config->hwmon_info;

drivers/hwmon/corsair_hydro_i_pro.c line 552

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+
+	data->hdev = hdev;
+	hwmon_dev = devm_hwmon_device_register_with_info(
+		&hdev->udev->dev, hdev->config->name, data, hwmon_info, NULL);
+	if (IS_ERR(hwmon_dev))
+		return PTR_ERR(hwmon_dev);
+
+	dev_info(&hdev->udev->dev, "setup hwmon for %s\n", hdev->config->name);
+	return 0;
+}

It is still possible that hwmon_init() leaks memory on failure.

drivers/hwmon/corsair_hydro_i_pro.c line 613

Quoted 12 lines

[...]

drivers/hwmon/corsair_hydro_i_pro.c new file +11 -0
+static int astk_probe(struct usb_interface *interface,
+		      const struct usb_device_id *id)
+{
+	struct hydro_i_pro_device *hdev;
+	int retval;
+	struct usb_endpoint_descriptor *bulk_in, *bulk_out;
+
+	hdev = kzalloc(sizeof(*hdev), GFP_KERNEL);
+
+	if (!hdev) {
+		kfree(hdev);

If this 'if' is taken, then hdev == NULL, so no need to free it.

Quoted 17 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +17 -0
+		retval = -ENOMEM;
+		goto exit;
+	}
+
+	hdev->config = (const struct device_config *)id->driver_info;
+	/* You should set the driver_info to a pointer to the correct configuration!!*/
+	WARN_ON(!hdev->config);
+
+	retval = usb_find_common_endpoints(interface->cur_altsetting, &bulk_in,
+					   &bulk_out, NULL, NULL);
+	if (retval) {
+		kfree(hdev);
+		goto exit;
+	}
+
+	hdev->udev = usb_get_dev(interface_to_usbdev(interface));
+	hdev->interface = usb_get_intf(interface);

drivers/hwmon/corsair_hydro_i_pro.c line 640

Quoted 10 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+
+	/*
+	 * set up the endpoint information
+	 * use only the first bulk-in and bulk-out endpoints
+	 */
+	hdev->bulk_in_size = usb_endpoint_maxp(bulk_in);
+	hdev->bulk_in_buffer = kmalloc(hdev->bulk_in_size, GFP_KERNEL);
+	hdev->bulk_in_endpointAddr = bulk_in->bEndpointAddress;
+	hdev->bulk_out_size = usb_endpoint_maxp(bulk_out);
+	hdev->bulk_out_buffer = kmalloc(hdev->bulk_out_size, GFP_KERNEL);

I think both allocations should be checked.

drivers/hwmon/corsair_hydro_i_pro.c line 653

Quoted 13 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +13 -0
+	hdev->bulk_out_endpointAddr = bulk_out->bEndpointAddress;
+
+	retval = init_device(hdev->udev);
+	if (retval) {
+		dev_err(&interface->dev, "failed initialising %s:%d\n",
+			hdev->config->name, retval);
+		kfree(hdev);
+		goto exit;
+	}
+
+	retval = hwmon_init(hdev);
+	if (retval) {
+		dev_err(&interface->dev, "failed initialising hwmon%s:%d\n",
                                                                  ^^^^^^

I think you need more spaces here.

drivers/hwmon/corsair_hydro_i_pro.c line 663

Quoted 11 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +10 -0
+			hdev->config->name, retval);
+		kfree(hdev);
+		goto exit;
+	}
+
+	usb_set_intfdata(interface, hdev);
+	mutex_init(&hdev->io_mutex);
+exit:
+	return retval;
+}
[...]

Barnabás Pőcze

author
#2 lore raw
Quoted 6 lines · Barnabás Pőcze

Hello,

I added a couple notes about the code inline.

Quoted 11 lines

drivers/hwmon/corsair_hydro_i_pro.c line 61

Quoted 7 lines

[...]

drivers/hwmon/corsair_hydro_i_pro.c new file +6 -0
+#define MAX_FAN_COUNT 2
+#define MAX_PWM_CHANNEL_COUNT MAX_FAN_COUNT
+
+struct hwmon_data {
+     struct hydro_i_pro_device *hdev;
+     int channel_count;

This is a nitpick, but the 'channel_count' value comes from the 'fancount' of an entry of 'config_table', which has type 'u8', so I don't see any need for it to be an 'int'.

In practice this will indeed never go above u8, even with a pump, didn't catch that one

Quoted 7 lines

[...]

drivers/hwmon/corsair_hydro_i_pro.c line 125

Quoted 4 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +4 -0
+static const struct device_config config_table[] = {
+     {
+             .vendor_id = 0x1b1c,
+             .product_id = 0x0c15,

If I see it correctly, you never use 'vendor_id', nor 'product_id', right?

This was a really sloppy patch apparently. Those were only there for finding them, so they should indeed go now.

Quoted 10 lines

[...]

drivers/hwmon/corsair_hydro_i_pro.c line 306

Quoted 7 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +7 -0
+
+     send_buf[0] = PWM_FAN_TARGET_CMD;
+     send_buf[1] = fan_data->fan_channel;
+     send_buf[2] = fan_data->fan_pwm_target;
+     dev_info(&hdev->udev->dev, "debug:%d,%d,%d", send_buf[0], send_buf[1],
+              send_buf[2]);
+     dev_info(&hdev->udev->dev, "val:%d", fan_data->fan_pwm_target);

This should be dev_dbg() in my opinion if you intend to have such messages.

Those weren't even supposed to go in :(

Quoted 19 lines

[...]

Quoted 18 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +18 -0
+                     retval = acquire_lock(hdev);
+                     if (retval)
+                             goto exit;
+
+                     retval = set_fan_target_pwm(hdev, fan_data, val);
+                     if (retval)
+                             goto cleanup;
+
+                     break;
+             case hwmon_pwm_enable:
+                     fan_data = data->channel_data[channel];
+
+                     switch (val) {
+                     case 2:
+                     case 0:
+                             retval = acquire_lock(hdev);
+                             if (retval)
+                                     goto exit;
Quoted 8 lines

drivers/hwmon/corsair_hydro_i_pro.c line 411

Quoted 6 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +6 -0
+
+                             retval = set_fan_pwm_curve(hdev, fan_data,
+                                                        DEFAULT_CURVE);
+                             if (retval)
+                                     goto cleanup;
+                             fan_data->mode = 0;

This could be 'fan_data->mode = val', no?

I didnt think about that, the mode can be read from userspace and when userspace sets it to 2 they probably expect to read 2. I usually try to use constants as much as possible instead of using variables when the value is known, but I missed this when merging the statements.

Quoted 14 lines

[...]

drivers/hwmon/corsair_hydro_i_pro.c line 525

Quoted 11 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +11 -0
+static int hwmon_init(struct hydro_i_pro_device *hdev)
+{
+     u8 fan_id;
+     struct device *hwmon_dev;
+     struct hwmon_fan_data *fan;
+     struct hwmon_data *data;
+     struct hwmon_chip_info *hwmon_info;
+
+     data = devm_kzalloc(&hdev->udev->dev, sizeof(*data), GFP_KERNEL);
+     hwmon_info = devm_kzalloc(&hdev->udev->dev,
+                               sizeof(struct hwmon_chip_info), GFP_KERNEL);

I think you missed this 'sizeof' when coverting to 'sizeof(*ptr)'.

Yes, I did indeed mist that one :/

Quoted 12 lines

[...]

Quoted 11 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +11 -0
+             fan->fan_channel = fan_id;
+             fan->mode = 0;
+             data->channel_data[fan_id] = fan;
+     }
+
+     hwmon_info->ops = &i_pro_ops;
+     hwmon_info->info = hdev->config->hwmon_info;
+
+     data->hdev = hdev;
+     hwmon_dev = devm_hwmon_device_register_with_info(
+             &hdev->udev->dev, hdev->config->name, data, hwmon_info, NULL);
Quoted 8 lines

drivers/hwmon/corsair_hydro_i_pro.c line 552

Quoted 6 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +6 -0
+     if (IS_ERR(hwmon_dev))
+             return PTR_ERR(hwmon_dev);
+
+     dev_info(&hdev->udev->dev, "setup hwmon for %s\n", hdev->config->name);
+     return 0;
+}

It is still possible that hwmon_init() leaks memory on failure.

you mean the allocated memory doesn't get deallocated until the driver detaches? I can't free the memory myself because when the driver does detach it will double-free.
I have looked at how other drivers handle this, none of them deallocate any memory.
If you mean error check the allocations, I added those now.

Quoted 1 line

[...]

Kind regards,

Jaap Aarts

#3 lore raw
Quoted 20 lines
Quoted 11 lines
Quoted 11 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +11 -0
+             fan->fan_channel = fan_id;
+             fan->mode = 0;
+             data->channel_data[fan_id] = fan;
+     }
+
+     hwmon_info->ops = &i_pro_ops;
+     hwmon_info->info = hdev->config->hwmon_info;
+
+     data->hdev = hdev;
+     hwmon_dev = devm_hwmon_device_register_with_info(
+             &hdev->udev->dev, hdev->config->name, data, hwmon_info, NULL);
Quoted 8 lines

drivers/hwmon/corsair_hydro_i_pro.c line 552

Quoted 6 lines
drivers/hwmon/corsair_hydro_i_pro.c new file +6 -0
+     if (IS_ERR(hwmon_dev))
+             return PTR_ERR(hwmon_dev);
+
+     dev_info(&hdev->udev->dev, "setup hwmon for %s\n", hdev->config->name);
+     return 0;
+}

It is still possible that hwmon_init() leaks memory on failure.

Quoted 7 lines

you mean the allocated memory doesn't get deallocated until the driver detaches? I can't free the memory myself because when the driver does detach it will double-free.
I have looked at how other drivers handle this, none of them deallocate any memory.
If you mean error check the allocations, I added those now.

Sorry, my bad, the devm_ part went over my head when reading. I also forgot to mention that in astk_probe() you use kmalloc() to allocate the IO buffers, but those are not freed if init_device() or hwmon_init() fails.

Barnabás Pőcze