Re: [PATCH] hwmon: (ibmaem) clean up sensors if version sysfs file fails

[Guenter Roeck]

On 6/14/26 23:46, Pengpeng Hou wrote:
> aem_register_sensors() creates the dynamic sensor files and the name
> sysfs file before creating the version file.
>
> Errors from the dynamic sensor files and from the name file already jump
> to aem_remove_sensors(), but an error from creating the version file is
> returned directly.  That leaves the earlier sysfs files published even
> though the AEM instance probe fails and the backing state is torn down.
>
> Route the version-file failure through the same cleanup path as the
> earlier sysfs creation failures.
>
> Signed-off-by: Pengpeng Hou <pengpeng@xxxxxxxxxxx>
> ---
>   drivers/hwmon/ibmaem.c | 3 +++
>   1 file changed, 3 insertions(+)
>
> diff --git a/drivers/hwmon/ibmaem.c b/drivers/hwmon/ibmaem.c
> index 0a9c3a29e6f2..305c41dd15bb 100644
> --- a/drivers/hwmon/ibmaem.c
> +++ b/drivers/hwmon/ibmaem.c
> @@ -945,6 +945,9 @@ static int aem_register_sensors(struct aem_data *data,
>   	if (err)
>   		goto error;
>   	err = device_create_file(dev, &sensor_dev_attr_version.dev_attr);
> +	if (err)
> +		goto error;
> +
>   	return err;
>   
>   error:


The main problem with this patch is that the hardware supported by the
driver ("x3350, x3550, x3650, x3655, x3755, x3850 M2, x3950 M2, and
certain HC10/HS2x/LS2x/QS2x blades") is no longer supported by Linux
distributions. On top of that, the driver uses a long since obsolete
API to register with the hwmon subsystem, triggering a kernel warning
whenever the driver is loaded.

That is a general problem with this and similar patches: They fix a
minor and unlikely to be seen problem while missing all the real issues
(as can be seen from Sashiko's feedback).

So, no, I am not going to apply this patch without evidence that the
driver is actually used.

Guenter