Re: [PATCH] hwmon: (gpd-fan) Reject EC PWM value 0 as invalid

From: Guenter Roeck

Date: Tue Jun 09 2026 - 23:58:13 EST

On 6/9/26 18:10, Pei Xiao wrote:


在 2026/6/9 23:41, Cryolitia PukNgae 写道:

On 6/9/26 07:41, Pei Xiao wrote:
The EC firmware is expected to return values in [1, pwm_max]. A read of 0
is illegal and would cause underflow in the conversion formula. Explicitly
check for 0 and return -EIO.

Signed-off-by: Pei Xiao <xiaopei01@xxxxxxxxxx>
---
  drivers/hwmon/gpd-fan.c | 6 +++++-
  1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/hwmon/gpd-fan.c b/drivers/hwmon/gpd-fan.c
index d1993cd645cb..decb61936b95 100644
--- a/drivers/hwmon/gpd-fan.c
+++ b/drivers/hwmon/gpd-fan.c
@@ -341,7 +341,11 @@ static int gpd_wm2_read_pwm(struct gpd_fan_data *data)
        gpd_ecram_read(drvdata, drvdata->pwm_write, &var);
  -    // Match gpd_generic_write_pwm(u8) below
+    /* EC PWM register valid range is 1 ~ pwm_max; 0 is an invalid
state */
+    if (unlikely(!var))
+        return -EIO;
+
+    /* Match gpd_generic_write_pwm() below */
      return DIV_ROUND_CLOSEST((var - 1) * 255, (drvdata->pwm_max - 1));
  }

Have you ever tested the behavior on a real device? When it returns 0,
what state is the device in?

Hi Cryolitia,

Thanks for your  messages!
Yes, this is not an issue, but this problem was reported during the Sashiko
AI review.

I disagree: it is an issue. We can not assume that the EC never returns 0,
and the case must be handled, either by returning an error as implemented above
or by returning 0.

Guenter

 The maintainer suggested I take a look. After checking, I think it might
be better to add a check:
if it's 0, just return directly. In case some EC returns 0 someday, causing
an underflow,
returning an error code would be more intuitive than returning an incorrect
value.

https://sashiko.dev/#/patchset/cover.1780880972.git.xiaopei01@xxxxxxxxxx?part=1

Pei.

thanks,

Cryolitia PukNgae