On Wed, 2022-02-23 at 16:16 +0530, Veerendranath Jakkam wrote: > > Two link non-AP MLD representation: > > wlan0 (non-AP MLD) > IFTYPE_STATION (netdev + wdev) > / \ > / \ > link0 link1 > IFTYPE_MLO_LINK (wdev) IFTYPE_MLO_LINK (wdev) > | | > | | > radio(2G) radio(5G) > > In contrast, NL80211_IFTYPE_MLO_LINK can't be used to represent AP MLO > link since an MLD AP must support pre-11be and 11be clients > simultaneously so each AP MLO link affiliated with AP MLD must also act > as independent AP for pre-11be clients so each AP MLO link must be > represented by NL80211_IFTYPE_AP associated with a separate netdev. > > Two link AP MLD representation: > > AP MLD > (netdev + wdev) > / \ > / \ > wlan0 wlan1 > IFTYPE_AP IFTYPE_AP > (netdev + wdev) (netdev + wdev) > | | > | | > radio(2G) radio(5G) So just for posterity's sake - we had more discussions on this out of band, and decided that the "netdev + wdev" on the wlan0/wlan1 will not actually happen - they both should be just "wdev" like in the non-AP MLD. This solves the issue of broadcast (otherwise you'd need AP MLD + wlan0 + wlan1 in a bridge and drop all multicast at wlan0 and wlan1), at the expense of a small amount of flexibility - you cannot consider legacy and MLD clients to be in different networks. However, given the complexities around multicast, you probably cannot consider them to be in different networks _anyway_ because then you'd have to _not_ drop all the multicast on wlan0/wlan1, and then you send multicast twice to the MLD clients, which would be wrong too... So the more reasonable thing to do is to treat this the same way as non- AP MLD with only a single netdev, and effectively behave as if there was an internal kind of bridge inside the AP MLD for the legacy clients. As a consequence, of course this patch will change too. johannes