Re: [PATCH 4/5] wifi: ath12k: fix MLO beacon handling using per-link addressing

[Baochen Qiang <baochen.qiang@oss.qualcomm.com>]

On 7/28/2026 12:27 AM, Jose Ignacio Tornos Martinez wrote:
> ath12k_mac_handle_beacon_iter() uses ahvif->deflink for both BSSID
> matching and connection_loss_work cancellation. In MLO, deflink is
> only the first link created for the MLD VIF and does not represent
> the other links. This causes two problems:
> 
> 1. Beacon BSSID matching only checks deflink's BSS config
>    (vif->bss_conf.bssid), so beacons received on non-deflink links
>    never match and never cancel connection_loss_work.
> 
> 2. Only deflink's connection_loss_work is cancelled, leaving
>    non-deflink connection_loss_work timers running even when beacons
>    are being received normally.
> 
> When the firmware reports a beacon miss event on any link,
> ath12k_mac_handle_beacon_miss() queues connection_loss_work on that
> link with a 3-second timeout. If a beacon is received before the
> timeout, ath12k_mac_handle_beacon_iter() should cancel it. But the
> deflink-only handling means that for non-deflink MLO links, beacons
> are never matched and connection_loss_work is never cancelled through
> this path.
> 
> Fix by handling non-MLO and MLO cases separately. For non-MLO, keep
> the existing deflink behavior. For MLO, iterate all active links to
> match the beacon BSSID against each link's BSS config. When a match
> is found, cancel connection_loss_work on all links because the
> work callback calls per-VIF ieee80211_connection_loss() regardless
> of which link queued it.

is it legitimate to change mac80211 to support per-link connection loss? if not, better to
move connection_loss_work from ath12k_link_vif to ath12k_vif ?

> 
> Tested on WCN7850 with MLO (Wi-Fi 7).
> 
> Signed-off-by: Jose Ignacio Tornos Martinez <jtornosm@redhat.com>
> ---
>  drivers/net/wireless/ath/ath12k/mac.c | 36 ++++++++++++++++++++++++---
>  1 file changed, 32 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath12k/mac.c b/drivers/net/wireless/ath/ath12k/mac.c
> index e36a37852fab..ebc35636b4ef 100644
> --- a/drivers/net/wireless/ath/ath12k/mac.c
> +++ b/drivers/net/wireless/ath/ath12k/mac.c
> @@ -1946,15 +1946,43 @@ static void ath12k_mac_handle_beacon_iter(void *data, u8 *mac,
>  	struct sk_buff *skb = data;
>  	struct ieee80211_mgmt *mgmt = (void *)skb->data;
>  	struct ath12k_vif *ahvif = ath12k_vif_to_ahvif(vif);
> -	struct ath12k_link_vif *arvif = &ahvif->deflink;
> +	struct ieee80211_bss_conf *link_conf;
> +	struct ath12k_link_vif *arvif;
> +	unsigned long links;
> +	u8 link_id;
>  
> -	if (vif->type != NL80211_IFTYPE_STATION || !arvif->is_created)
> +	if (vif->type != NL80211_IFTYPE_STATION)
>  		return;
>  
> -	if (!ether_addr_equal(mgmt->bssid, vif->bss_conf.bssid))
> +	if (!ieee80211_vif_is_mld(vif)) {
> +		arvif = &ahvif->deflink;
> +		if (arvif->is_created &&
> +		    ether_addr_equal(mgmt->bssid, vif->bss_conf.bssid))
> +			cancel_delayed_work(&arvif->connection_loss_work);
>  		return;
> +	}
>  
> -	cancel_delayed_work(&arvif->connection_loss_work);
> +	/* For MLO, each link has a different AP BSSID. Check the beacon
> +	 * against all link BSS configs. If any matches, cancel
> +	 * connection_loss_work on all links since it calls per-VIF
> +	 * ieee80211_connection_loss() regardless of which link queued it.
> +	 */
> +	links = ahvif->links_map;
> +	for_each_set_bit(link_id, &links, IEEE80211_MLD_MAX_NUM_LINKS) {
> +		link_conf = rcu_dereference(vif->link_conf[link_id]);
> +		if (link_conf &&
> +		    ether_addr_equal(mgmt->bssid, link_conf->bssid))
> +			goto found;
> +	}
> +
> +	return;
> +
> +found:
> +	for_each_set_bit(link_id, &links, IEEE80211_MLD_MAX_NUM_LINKS) {
> +		arvif = rcu_dereference(ahvif->link[link_id]);
> +		if (arvif && arvif->is_created)
> +			cancel_delayed_work(&arvif->connection_loss_work);
> +	}
>  }
>  
>  void ath12k_mac_handle_beacon(struct ath12k *ar, struct sk_buff *skb)