Re: nl80211: missing minimum TX power attribute causes misleading userspace behavior

[Ben Greear <greearb@candelatech.com>]

On 4/9/26 01:05, Steffen May wrote:
> 
> This issue was discovered during the analysis of two documented OpenWrt bugs. Bug 1 is a type comparison error in LuCI wireless.js that causes 0 dBm to be 
> displayed incorrectly. Bug 2 is in ucode mac80211.sh where the value 0 is treated as falsy, causing the router to transmit at maximum power instead. Both bugs 
> are proven and reported.
> 
> During verification of Bug 2 on five devices with three different chipsets, it became apparent that the hardware floor is device-dependent and completely 
> unknown to the stack. This is not a bug but a missing feature.
> 
> Because the floor is unknown, iwinfo generates selection lists containing values that have no real effect on the actual output power. Userspace accepts 
> configurations such as 0 dBm even though the hardware cannot apply them. The system reports success while the hardware remains at its minimum supported level. 
> This creates false assumptions.
> 
> Measurements
> 
> The deviation between the requested transmit power and the actual hardware floor is not constant but depends on the hardware:
> 
> Device                Chipset            Requested     Actual floor Difference
> Buffalo WZR-600DHP    Atheros AR7161     1 dBm         3 dBm          +2 dB
> TP-Link Archer C7     Qualcomm QCA9558   1 dBm         5 dBm          +4 dB
> OpenWrt One           MTK Filogic        1 dBm         1 dBm           0 dB
> Cudy WR3000           MTK Filogic        1 dBm         1 dBm           0 dB
> GL.iNet GL-MT6000     MTK Filogic        1 dBm         1 dBm           0 dB

If you set that tplink to 1dbm, and then read the reported power, does it properly show 5dbm?

If so, that seems good enough?

If not, then we should fix the reporting, but having a floor reported doesn't seem helpful
to me.  There are lots of things that can affect actual txpower.  User-space can at best
offer its suggestion of preferred txpower.  Kernel/driver/firmware/hardware then makes final
decision.

Thanks,
Ben

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com