On 23.08.24 12:23, Johannes Berg wrote:
On Mon, 2024-08-05 at 21:23 +0200, Felix Fietkau wrote:
When restricting a monitor vif to only operate on a specific set of radios,
filter out rx packets belonging to other radios. This only works if drivers
fill in radio_valid and radio_idx in the rx status.
Why does the driver need to provide the radio, it already provides the
frequency?
But then I wonder if this doesn't go a step too far? This is pretty much
pretending that monitor only exists on a specific sub-radio, but ...
what for? Even userspace could filter on the frequency.
I mean ... I get that you're trying to preserve a notion that you had
that an interface exists on a given PHY and they're all separate, but
they're not really separate any more, get used to it?
Well, there's a performance aspect as well. When only monitoring
specific radios (while operating normally on others), relying on
filtering in user space or even BPF comes at a cost, since mac80211
still has to prepare radiotap headers and potentially clone data packets
received on other radios.
- Felix
