[PATCH] hwmon: Relax name attribute validation for new APIs

[PATCH] hwmon: Relax name attribute validation for new APIs

[Guenter Roeck]

While invalid name attributes are really not desirable and do mess up
libsensors, enforcing valid names has the detrimental effect of driving
users away from using the new hardware monitoring API, especially those
registering name attributes violating the ABI restrictions. Another
undesirable side effect is that this violation and the resulting error
may only be discovered some time after a conversion to the new API,
which in turn may trigger a revert of that conversion.

To solve the problem, relax validation and only issue a warning instead
of returning an error if a name attribute violating the ABI is provided.
This lets callers continue to provide invalid name attributes while
notifying them about it.

Many thanks are due to Dmitry Torokhov for the idea.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
---
 drivers/hwmon/hwmon.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/hwmon/hwmon.c b/drivers/hwmon/hwmon.c
index affff8195fff..53c54a81f7ad 100644
--- a/drivers/hwmon/hwmon.c
+++ b/drivers/hwmon/hwmon.c
@@ -544,9 +544,10 @@ __hwmon_device_register(struct device *dev, const char *name, void *drvdata,
 	struct device *hdev;
 	int i, j, err, id;
 
-	/* Do not accept invalid characters in hwmon name attribute */
+	/* Complain about invalid characters in hwmon name attribute */
 	if (name && (!strlen(name) || strpbrk(name, "-* \t\n")))
-		return ERR_PTR(-EINVAL);
+		dev_warn(dev, "hwmon: '%s' is not a valid name attribute\n",
+			 name);
 
 	id = ida_simple_get(&hwmon_ida, 0, 0, GFP_KERNEL);
 	if (id < 0)
-- 
2.7.4

Re: [PATCH] hwmon: Relax name attribute validation for new APIs

[Jean Delvare]

Hi Guenter, Dmitry,

On Fri, 27 Jan 2017 19:49:49 -0800, Guenter Roeck wrote:
> While invalid name attributes are really not desirable and do mess up
> libsensors, enforcing valid names has the detrimental effect of driving
> users away from using the new hardware monitoring API, especially those
> registering name attributes violating the ABI restrictions. Another
> undesirable side effect is that this violation and the resulting error
> may only be discovered some time after a conversion to the new API,
> which in turn may trigger a revert of that conversion.
> 
> To solve the problem, relax validation and only issue a warning instead
> of returning an error if a name attribute violating the ABI is provided.
> This lets callers continue to provide invalid name attributes while
> notifying them about it.
> 
> Many thanks are due to Dmitry Torokhov for the idea.
> 
> Signed-off-by: Guenter Roeck <linux@roeck-us.net>
> ---
>  drivers/hwmon/hwmon.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/hwmon/hwmon.c b/drivers/hwmon/hwmon.c
> index affff8195fff..53c54a81f7ad 100644
> --- a/drivers/hwmon/hwmon.c
> +++ b/drivers/hwmon/hwmon.c
> @@ -544,9 +544,10 @@ __hwmon_device_register(struct device *dev, const char *name, void *drvdata,
>  	struct device *hdev;
>  	int i, j, err, id;
>  
> -	/* Do not accept invalid characters in hwmon name attribute */
> +	/* Complain about invalid characters in hwmon name attribute */
>  	if (name && (!strlen(name) || strpbrk(name, "-* \t\n")))
> -		return ERR_PTR(-EINVAL);
> +		dev_warn(dev, "hwmon: '%s' is not a valid name attribute\n",
> +			 name);

May I suggest adding ", please fix"?

>  
>  	id = ida_simple_get(&hwmon_ida, 0, 0, GFP_KERNEL);
>  	if (id < 0)

Reviewed-by: Jean Delvare <jdelvare@suse.de>

Do I understand correctly that in the long run we will make it a fatal
error again?

-- 
Jean Delvare
SUSE L3 Support

Re: [PATCH] hwmon: Relax name attribute validation for new APIs

[Guenter Roeck]

On 01/31/2017 01:07 AM, Jean Delvare wrote:
> Hi Guenter, Dmitry,
>
> On Fri, 27 Jan 2017 19:49:49 -0800, Guenter Roeck wrote:
>> While invalid name attributes are really not desirable and do mess up
>> libsensors, enforcing valid names has the detrimental effect of driving
>> users away from using the new hardware monitoring API, especially those
>> registering name attributes violating the ABI restrictions. Another
>> undesirable side effect is that this violation and the resulting error
>> may only be discovered some time after a conversion to the new API,
>> which in turn may trigger a revert of that conversion.
>>
>> To solve the problem, relax validation and only issue a warning instead
>> of returning an error if a name attribute violating the ABI is provided.
>> This lets callers continue to provide invalid name attributes while
>> notifying them about it.
>>
>> Many thanks are due to Dmitry Torokhov for the idea.
>>
>> Signed-off-by: Guenter Roeck <linux@roeck-us.net>
>> ---
>>  drivers/hwmon/hwmon.c | 5 +++--
>>  1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/hwmon/hwmon.c b/drivers/hwmon/hwmon.c
>> index affff8195fff..53c54a81f7ad 100644
>> --- a/drivers/hwmon/hwmon.c
>> +++ b/drivers/hwmon/hwmon.c
>> @@ -544,9 +544,10 @@ __hwmon_device_register(struct device *dev, const char *name, void *drvdata,
>>  	struct device *hdev;
>>  	int i, j, err, id;
>>
>> -	/* Do not accept invalid characters in hwmon name attribute */
>> +	/* Complain about invalid characters in hwmon name attribute */
>>  	if (name && (!strlen(name) || strpbrk(name, "-* \t\n")))
>> -		return ERR_PTR(-EINVAL);
>> +		dev_warn(dev, "hwmon: '%s' is not a valid name attribute\n",
>> +			 name);
>
> May I suggest adding ", please fix"?
>

Ok, will do.

>>
>>  	id = ida_simple_get(&hwmon_ida, 0, 0, GFP_KERNEL);
>>  	if (id < 0)
>
> Reviewed-by: Jean Delvare <jdelvare@suse.de>
>
> Do I understand correctly that in the long run we will make it a fatal
> error again?
>

Hopefully, after there are no more drivers to convert, and after all converted
drivers fixed the problem.

Thanks,
Guenter