On Fri, 2025-05-16 at 16:15 +0800, Chen-Yu Tsai wrote: > The conversions don't always produce round numbers, and since we can't let > the system exceed the _actual_ limit, the values can only be rounded down. > For example, 500 mW converts to roughly 26.9897 dBm. We can't use 27 dBm, > since if the system outputs at 26.999 dBm, it would be in violation of > the rules. While technically true, I'd think it's probably hard to even measure that accurately, 27 dBm is 501.187 mW? > Or, if we understand 3 dBm reduction to be halving the power, That's fairly precise, a factor of 0.501. > we could > fix up any rules that "have their base limit in mW and were converted to > dBm to apply the reduction" be rewritten in mW with the numbers halved. > That would require someone to go through the entries though. But if > the rules are already written in mW, and what you observe is the kernel > rounding down the numbers, then perhaps the kernel may need to support > both units. I don't think _drivers_ (or firmware) would support both units, so that'd just shift the rounding issue elsewhere? Though I think in some places we use "mBm" instead of dBm? But then we also just convert to dBm in many places so that'd likely not make it any better. johannes
