Re: [RFC PATCH 2/2] wifi: mac80211: Assign tx-stats to the proper link.

[Johannes Berg <johannes@sipsolutions.net>]

On Wed, 2024-08-28 at 08:54 -0700, greearb@candelatech.com wrote:
> From: Ben Greear <greearb@candelatech.com>
> 
> For drivers that can report the tx link-id, account tx
> stats against that link.  If we cannot determine tx link,
> then use deflink.

Strictly speaking, that's not what happens, since the link bits in the
SKB CB might be set on outgoing frames, and then will still be there on
the status.

Also using deflink is totally useless for MLO, so maybe just don't do
anything at all? But might be simpler to just do deflink and document
that the driver must set this? But not sure that really works so well
for drivers now.

> +static struct link_sta_info*
> +ieee80211_get_tx_link_sta(struct sta_info *sta, struct ieee80211_tx_info *info)
> +{
> +	u8 link_id = u32_get_bits(info->control.flags, IEEE80211_TX_CTRL_MLO_LINK);
> +	struct link_sta_info *l_sta_info;

We usually call that 'link_sta' ...

> @@ -48,6 +65,7 @@ static void ieee80211_handle_filtered_frame(struct ieee80211_local *local,
>  	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
>  	struct ieee80211_hdr *hdr = (void *)skb->data;
>  	int ac;
> +	struct link_sta_info *link_sta = ieee80211_get_tx_link_sta(sta, info);

and even you do, please be consistent with existing code and yourself.

> 
> +		link_sta = ieee80211_get_tx_link_sta(sta, info);

Does it really make sense to keep repeating this, rather than passing an
argument?

johannes