Linux Hardware Monitor development
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Benedikt Niedermayr" <benedikt.niedermayr@siemens.com>
Cc: linux-hwmon@vger.kernel.org
Subject: Re: [PATCH 1/1] hwmon: simatic-ipc-ec-sensors: add BX-54A EC driver
Date: Mon, 16 Nov 2026 16:50:46 +0000	[thread overview]
Message-ID: <sashiko-outbox-217386@kernel.org> (raw)
In-Reply-To: <20261116163257.1303207-2-benedikt.niedermayr@siemens.com>

Thank you for your contribution! Sashiko AI review found 8 potential
issues to consider:
- [High] hwmon: simatic-ipc-ec-sensors: temp limits exposed as *_alarm
  attributes
- [High] hwmon: pointer-sign mismatch passing s8 * to ec_read() in temp
  read callback
- [Medium] hwmon: premature s8 cast before clamp() in
  simatic_ipc_ec_temp_write()
- [Medium] hwmon: u8 truncation in simatic_ipc_ec_fan_write() bypasses
  range check
- [Medium] hwmon: uninitialized reads on ec_read() error in temp and fan
  read callbacks
- [Medium] hwmon: devm buffer stored in global struct in
  simatic_ipc_ec_hwmon_probe()
- [Low] hwmon: unsanitized and unlowercased chip_name in
  simatic_ipc_ec_fan_probe()
- [Low] hwmon: simatic-ipc-ec-sensors: missing Documentation/hwmon/
  driver doc
--

commit 830c0088242b1976cce59ca2195a3aaabae749d2
Author: Benedikt Niedermayr <benedikt.niedermayr@siemens.com>

hwmon: simatic-ipc-ec-sensors: add BX-54A EC driver

This commit adds a hardware monitoring driver for temperature sensors and
fan control on the embedded controller of Siemens SIMATIC IPC BX-54A
systems.

> diff --git a/MAINTAINERS b/MAINTAINERS
> index ccbcdc0799b9f..8e567b60f6487 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -25763,6 +25763,12 @@ F:	drivers/media/mmc/siano/
>  F:	drivers/media/usb/siano/
>  F:	drivers/media/usb/siano/
>  
> +SIEMENS IPC EC SENSORS DRIVER
> +M:	Benedikt Niedermayr <benedikt.niedermayr@siemens.com>
> +L:	linux-hwmon@vger.kernel.org
> +S:	Maintained
> +F:	drivers/hwmon/simatic-ipc-ec-sensors.c

[Severity: Low]
New hwmon drivers require a documentation file under
Documentation/hwmon/simatic-ipc-ec-sensors.rst per
Documentation/hwmon/submitting-patches.rst.

[ ... ]

> diff --git a/drivers/hwmon/simatic-ipc-ec-sensors.c b/drivers/hwmon/simatic-ipc-ec-sensors.c
> new file mode 100644
> index 0000000000000..690da5dde2bf8
> --- /dev/null
> +++ b/drivers/hwmon/simatic-ipc-ec-sensors.c
> @@ -0,0 +1,568 @@

[ ... ]

> +static umode_t simatic_ipc_ec_temp_is_visible(const struct simatic_ipc_ec_temp *temp,
> +					      u32 attr, int channel)
> +{
> +	switch (attr) {
> +	case hwmon_temp_input:
> +		return 0444;
> +	case hwmon_temp_max_alarm:
> +	case hwmon_temp_min_alarm:
> +	case hwmon_temp_crit:
> +		return 0644;
> +	default:
> +		break;
> +	}
> +
> +	return 0;
> +}
> +
> +static int simatic_ipc_ec_temp_read(const struct simatic_ipc_ec_temp *temp, u32 attr,
> +				    int channel, long *val)
> +{
> +	s8 value;
> +	int ret;
> +	u8 reg;
> +
> +	switch (attr) {
> +	case hwmon_temp_input:
> +		reg = temp_reg(temp->base, channel, TEMP_REG_VALUE);
> +		break;
> +	case hwmon_temp_max_alarm:
> +		reg = temp_reg(temp->base, channel, TEMP_REG_ALARM_HIGHLIMIT);
> +		break;
> +	case hwmon_temp_min_alarm:
> +		reg = temp_reg(temp->base, channel, TEMP_REG_ALARM_LOWLIMIT);
> +		break;
> +	case hwmon_temp_crit:
> +		reg = temp_reg(temp->base, channel, TEMP_REG_CRITICAL);
> +		break;
> +	default:
> +		return -EOPNOTSUPP;
> +	}

[Severity: High]
Should the low and high temperature thresholds use hwmon_temp_min and
hwmon_temp_max (with HWMON_T_MIN and HWMON_T_MAX in
simatic_ipc_ec_hwmon_probe()) instead of hwmon_temp_min_alarm and
hwmon_temp_max_alarm?

Under the hwmon ABI, tempX_min_alarm and tempX_max_alarm are read-only
boolean flags (0 or 1), so exposing writable millidegree Celsius limits
there causes tools like libsensors to treat any non-zero limit as an
active alarm while missing the min and max thresholds.

> +
> +	ret = ec_read(reg, &value);

[Severity: High]
Does passing &value (s8 *) to ec_read(u8, u8 *) trigger a -Wpointer-sign
warning (and build failure under CONFIG_WERROR)?

> +
> +	/* Degree Celsius to millidegree Celsius */
> +	*val = value * 1000;
> +
> +	return ret;
> +}
> +
> +static int simatic_ipc_ec_temp_write(const struct simatic_ipc_ec_temp *temp, u32 attr,
> +				     int channel, long val)
> +{
> +	/*
> +	 * Millidegree Celsius to degree Celsius
> +	 */
> +	s8 value = (s8)clamp((s8)DIV_ROUND_CLOSEST(val, 1000), (s8)-128, (s8)127);

[Severity: Medium]
Does casting the result of DIV_ROUND_CLOSEST(val, 1000) to s8 before
clamp() make clamp() a no-op?

Because the first argument is already truncated to s8, writing a threshold
outside [-128000, 127499] wraps modulo 256 instead of saturating (for
example, 130000 wraps to -126 C), and values near LONG_MAX or LONG_MIN can
also overflow DIV_ROUND_CLOSEST(val, 1000) unless val is clamped before
division.

[ ... ]

> +static int simatic_ipc_ec_fan_read(struct simatic_ipc_ec_fan *fan, enum hwmon_sensor_types type,
> +				   u32 attr, int channel, long *val)
> +{
> +	u16 value_u16;
> +	u8 value_u8;
> +	int ret;
> +	u8 reg;
> +
> +	switch (type) {
> +	case hwmon_fan:
> +		switch (attr) {
> +		case hwmon_fan_input:
> +			reg = fan_reg(fan->base, FAN_REG_SPEED);
> +			ret = simatic_ipc_ec_read_block(reg, &value_u16, sizeof(value_u16));
> +			*val = value_u16;
> +			return ret;
> +		default:
> +			break;
> +		}
> +		break;
> +	case hwmon_pwm:
> +		switch (attr) {
> +		case hwmon_pwm_input:
> +			reg = fan_reg(fan->base, FAN_REG_DUTY_LEVEL);
> +			ret = ec_read(reg, &value_u8);
> +			*val = value_u8;
> +			return ret;
> +		case hwmon_pwm_enable:
> +			reg = fan_reg(fan->base, FAN_REG_MODE);
> +			ret = ec_read(reg, &value_u8);
> +			*val = CONTROL_MODE_GET(value_u8);
> +			return ret;

[Severity: Medium]
When ec_read() or simatic_ipc_ec_read_block() fails here and in
simatic_ipc_ec_temp_read(), does assigning *val read uninitialized stack
variables (value_u16, value_u8, and value)?  ec_read() leaves its output
parameter untouched on error, which triggers KMSAN uninitialized-memory
warnings when *val is computed before checking ret.

[ ... ]

> +static int simatic_ipc_ec_fan_set_control_mode(struct simatic_ipc_ec_fan *fan, u8 ctrl_mode)
> +{
> +	int ret;
> +	u8 mode;
> +
> +	/* 0: full speed, 1: mannual control, 2: auto-control */
> +	if (ctrl_mode > 2)
> +		return -EINVAL;

[ ... ]

> +static int simatic_ipc_ec_fan_write(struct simatic_ipc_ec_fan *fan, u32 attr, int channel,
> +				    long val)
> +{
> +	switch (attr) {
> +	case hwmon_pwm_input:
> +		if (val < 0 || val > 0xFF)
> +			return -EINVAL;
> +		return ec_write(fan_reg(fan->base, FAN_REG_DUTY_LEVEL), (u8)val);
> +	case hwmon_pwm_enable:
> +		return simatic_ipc_ec_fan_set_control_mode(fan, (u8)val);

[Severity: Medium]
Does casting val to u8 before calling simatic_ipc_ec_fan_set_control_mode()
allow out-of-range values such as 256 or -255 to bypass the ctrl_mode > 2
check and change the EC fan control mode?

[ ... ]

> +static struct hwmon_channel_info simatic_ipc_ec_temp_ch_info = {
> +	.type = hwmon_temp,
> +	.config = NULL
> +};

[ ... ]

> +static int simatic_ipc_ec_hwmon_probe(struct device *dev, struct simatic_ipc_ec_sensor *data)
> +{

[ ... ]

> +	temp_config = devm_kzalloc(dev, (temp_count + 1) * sizeof(u32), GFP_KERNEL);
> +	if (!temp_config)
> +		return -ENOMEM;
> +
> +	simatic_ipc_ec_temp_ch_info.config = temp_config;

[Severity: Medium]
Can storing the per-device devm-allocated temp_config into the static
global simatic_ipc_ec_temp_ch_info.config leave a dangling pointer on probe
failure or unbind, or overwrite another instance's config if more than one
110A5004 device is probed?

[ ... ]

> +static int simatic_ipc_ec_fan_probe(struct device *dev, struct simatic_ipc_ec_sensor *data)
> +{

[ ... ]

> +	if (fan_name) {
> +		char *p;
> +
> +		snprintf(fan->chip_name, CHIP_NAME_MAX_LEN, "simatic_ipc_ec_%s", fan_name);
> +		for (p = fan->chip_name; *p; p++)
> +			*p = tolower(*p);
> +	} else {
> +		snprintf(fan->chip_name, CHIP_NAME_MAX_LEN, "simatic_ipc_ec_%s", label);
> +	}

[Severity: Low]
What happens if label contains uppercase characters when fan_name is
absent, or if either string contains characters disallowed in hwmon chip
names such as '-' or spaces?

The else branch skips the tolower() loop, and neither branch sanitizes
characters rejected by __hwmon_device_register() (or via
devm_hwmon_sanitize_name()).

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261116163257.1303207-1-benedikt.niedermayr@siemens.com?part=1

  reply	other threads:[~2026-11-16 16:50 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-16 16:32 [PATCH 0/1] hwmon: simatic-ipc-ec-sensors: add BX-54A EC driver Benedikt Niedermayr
2026-11-16 16:32 ` [PATCH 1/1] " Benedikt Niedermayr
2026-11-16 16:50   ` sashiko-bot [this message]
2026-11-16 20:38     ` Guenter Roeck

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=sashiko-outbox-217386@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=benedikt.niedermayr@siemens.com \
    --cc=linux-hwmon@vger.kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox