On 5/21/2025 12:49 PM, Johannes Berg wrote:
> On Tue, 2025-04-29 at 09:30 +0530, Roopni Devanathan wrote:
>
>> +++ b/include/net/mac80211.h
>> @@ -4569,7 +4569,8 @@ struct ieee80211_ops {
>> struct ieee80211_key_conf *key,
>> struct ieee80211_key_seq *seq);
>> int (*set_frag_threshold)(struct ieee80211_hw *hw, u32 value);
>> - int (*set_rts_threshold)(struct ieee80211_hw *hw, u32 value);
>> + int (*set_rts_threshold)(struct ieee80211_hw *hw, u8 radio_id,
>> + u32 value);
>
> That should probably have documentation updates. And passing a negative
> value to a u8 seems awkward? Maybe that should just be 'int'? For a
> value that's likely passed in a register, u8 will probably require more
> (machine) code anyway.
>
Using 'int' leads to adding NLA_U32 policy. I think using 'int' is
costlier than using 's8'. Can we just revert back to using 's8' instead?
There will still be a default value of -1 and the radio indices will not
require u32 value anyway.
If this is okay, I'll send out the next version of patch series with
other comments addressed and s8 used instead of u8 (and int).
> johannes
