Re: [PATCH] drm/xe/hwmon: Restrict LONG_MAX overflow check to 32-bit builds

["Mallesh, Koujalagi" <mallesh.koujalagi@intel.com>]

On 23-10-2026 05:28 pm, Jani Nikula wrote:
> On Fri, 23 Oct 2026, Mallesh Koujalagi <mallesh.koujalagi@intel.com> wrote:
>> The temperature decoding code stores the intermediate result in an
>> s64 and checks it against LONG_MAX before assigning it to a long.
>>
>> This check is only needed on 32-bit architectures, where long is
>> narrower than s64. On 64-bit systems, long and s64 are both 64 bits,
>> so the condition can never be true.
>>
>> Wrap the check with BITS_PER_LONG == 32 so it is only compiled where
>> it is needed, preserving the existing overflow protection on 32-bit
>> architectures while eliminating the dead branch on 64-bit builds.
> Why would it matter? Why make the code harder to read with the inline
> #ifdef?

Static analysis tools complain that the code contains dead code on

64-bit systems because the "value > LONG_MAX" check is always false there.


My initial thought was to handle this explicitly for 32 bit systems,

but I agree that the #ifdef makes the code harder to read.


An alternative is to user overflows_type(), which handles both 32-bit and

64-bit cases without conditional code and keeps the overflow check clear

and simple.


Any thoughts? If that sounds reasonable, I'll send a next revision with 
that approach.


Please let me know.


Thanks,

-/Mallesh

>> Fixes: 24ec15d64617 ("drm/xe/hwmon: Decode CRI temperature registers as IEEE-754")
>> Signed-off-by: Mallesh Koujalagi <mallesh.koujalagi@intel.com>
>> ---
>>   drivers/gpu/drm/xe/xe_hwmon.c | 7 ++++++-
>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_hwmon.c b/drivers/gpu/drm/xe/xe_hwmon.c
>> index 42acfeb8818c..2ee51921acf5 100644
>> --- a/drivers/gpu/drm/xe/xe_hwmon.c
>> +++ b/drivers/gpu/drm/xe/xe_hwmon.c
>> @@ -1322,9 +1322,14 @@ static int xe_hwmon_temperature_decode(u32 reg_val, long *temperature)
>>   		value = 0;
>>   	}
>>   
>> -	/* Ensure the magnitude fits in long, which is 32-bit on some arches. */
>> +#if BITS_PER_LONG == 32
>> +	/*
>> +	 * On 32-bit systems, long is smaller than s64, so the value may
>> +	 * still overflow long even if it fits in s64.
>> +	 */
>>   	if (value > LONG_MAX)
>>   		return -ERANGE;
>> +#endif
>>   
>>   	if (reg_val & IEEE754_SIGN)
>>   		value = -value;