Re: [RFC 0/4] alternative changes for NL80211_CMD_SET_BSS support

++

On Wed, 2025-06-04 at 10:55 +0200, Arend van Spriel wrote:
> Here a series with my idea on how to solve the fact that drivers may not
> support or handle all parameters contained in NL80211_CMD_SET_BSS message
> as sent by user-space. The driver can announce in the wiphy info which
> parameters it supports and user-space can indicate that the kernel may
> enforce strict checking for the NL80211_CMD_SET_BSS.

So since we have the two alternatives now and there was no discussion, I
guess you're looking to me to decide ;-)

I like the way you here expose to userspace what's supported, and it
also seems slightly easier for drivers to just set the supported bitmap
rather than checking what was given in the callback?

With your approach the userspace check is opt-in, though I'd probably
structure it so that for *new* attributes, if any are ever added, we'd
always check, regardless of the strict setting.

I have some quibbles here, in particular I don't think
WIPHY_BSS_PARAM_MASK should exist, and mac80211 should explicitly list
out the ones it supports (since new ones could be added that it doesn't
have), but otherwise looks fine.

I guess in summary I tend to prefer your changes for better handling of
userspace backward compatibility and less driver code.

johannes