[PATCH ath-next v2 2/2] wifi: ath12k: fix MLO peer delete race

Baochen Qiang <[email protected]>
Newsgroups org.infradead.lists.ath12k,org.kernel.vger.linux-wireless
Message-ID <20260629-ath12k-mlo-peer-delete-race-v2-2-362b25590d19@oss.qualcomm.com>
ath12k_peer_mlo_link_peers_delete() sends WMI peer_delete for every
link before waiting for any peer_unmap / peer_delete_resp event. The
shared per-radio completion ar->peer_delete_done could not
disambiguate which peer a response was for: every call to
ath12k_peer_delete_send() did
reinit_completion(&ar->peer_delete_done), so when an event for the
first link arrived between two sends it raised the count to 1 and
the second send promptly cleared it; the wait for the second link
then timed out with

    Timeout in receiving peer delete response

Replace the shared completion with a per-radio waiter list, with
each pending ath12k_peer_delete() caller queueing an
ath12k_peer_delete_wait carrying its (vdev_id, addr) and a private
struct completion. ath12k_peer_delete_resp_event() matches the
response against the list under ar->data_lock and signals the
matching waiter.

Also correct the endian conversion in ath12k_peer_delete_resp_event()
logging, and add the missing \n in some logging.

Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.1.c7-00108-QCAHMTSWPL_V1.0_V2.0_SILICONZ_UPSTREAM-3

Fixes: 8e6f8bc28603 ("wifi: ath12k: Add MLO station state change handling")
Signed-off-by: Baochen Qiang <[email protected]>
---
 drivers/net/wireless/ath/ath12k/core.c |   2 +-
 drivers/net/wireless/ath/ath12k/core.h |   5 +-
 drivers/net/wireless/ath/ath12k/mac.c  |   2 +-
 drivers/net/wireless/ath/ath12k/peer.c | 130 ++++++++++++++++++++++++++-------
 drivers/net/wireless/ath/ath12k/peer.h |  14 +++-
 drivers/net/wireless/ath/ath12k/wmi.c  |  16 ++--
 6 files changed, 131 insertions(+), 38 deletions(-)

diff --git a/drivers/net/wireless/ath/ath12k/core.c b/drivers/net/wireless/ath/ath12k/core.c
index 742d4fd1b598..f71650039292 100644
--- a/drivers/net/wireless/ath/ath12k/core.c
+++ b/drivers/net/wireless/ath/ath12k/core.c
@@ -1524,7 +1524,7 @@ static void ath12k_core_pre_reconfigure_recovery(struct ath12k_base *ab)
 			complete_all(&ar->scan.completed);
 			complete(&ar->scan.on_channel);
 			complete(&ar->peer_assoc_done);
-			complete(&ar->peer_delete_done);
+			ath12k_peer_delete_wait_flush(ar);
 			complete(&ar->install_key_done);
 			complete(&ar->vdev_setup_done);
 			complete(&ar->vdev_delete_done);
diff --git a/drivers/net/wireless/ath/ath12k/core.h b/drivers/net/wireless/ath/ath12k/core.h
index fc5127b5c1a3..1436ff4316e7 100644
--- a/drivers/net/wireless/ath/ath12k/core.h
+++ b/drivers/net/wireless/ath/ath12k/core.h
@@ -665,7 +665,8 @@ struct ath12k {
 
 	/* protects the radio specific data like debug stats, ppdu_stats_info stats,
 	 * vdev_stop_status info, scan data, ath12k_sta info, ath12k_link_vif info,
-	 * channel context data, survey info, test mode data, regd_channel_update_queue.
+	 * channel context data, survey info, test mode data, regd_channel_update_queue,
+	 * peer_delete_waits.
 	 */
 	spinlock_t data_lock;
 
@@ -687,7 +688,7 @@ struct ath12k {
 	u8 radio_idx;
 
 	struct completion peer_assoc_done;
-	struct completion peer_delete_done;
+	struct list_head peer_delete_waits;
 
 	int install_key_status;
 	struct completion install_key_done;
diff --git a/drivers/net/wireless/ath/ath12k/mac.c b/drivers/net/wireless/ath/ath12k/mac.c
index 2e5a075191ae..4c86a8eb5841 100644
--- a/drivers/net/wireless/ath/ath12k/mac.c
+++ b/drivers/net/wireless/ath/ath12k/mac.c
@@ -15040,11 +15040,11 @@ static void ath12k_mac_setup(struct ath12k *ar)
 	spin_lock_init(&ar->dp.ppdu_list_lock);
 	INIT_LIST_HEAD(&ar->arvifs);
 	INIT_LIST_HEAD(&ar->dp.ppdu_stats_info);
+	INIT_LIST_HEAD(&ar->peer_delete_waits);
 
 	init_completion(&ar->vdev_setup_done);
 	init_completion(&ar->vdev_delete_done);
 	init_completion(&ar->peer_assoc_done);
-	init_completion(&ar->peer_delete_done);
 	init_completion(&ar->install_key_done);
 	init_completion(&ar->bss_survey_done);
 	init_completion(&ar->scan.started);
diff --git a/drivers/net/wireless/ath/ath12k/peer.c b/drivers/net/wireless/ath/ath12k/peer.c
index c222bdaa333c..5084e8c42a3f 100644
--- a/drivers/net/wireless/ath/ath12k/peer.c
+++ b/drivers/net/wireless/ath/ath12k/peer.c
@@ -9,6 +9,55 @@
 #include "debug.h"
 #include "debugfs.h"
 
+static void ath12k_peer_delete_wait_register(struct ath12k *ar,
+					     struct ath12k_peer_delete_wait *wait,
+					     u32 vdev_id, const u8 *addr)
+{
+	wait->vdev_id = vdev_id;
+	ether_addr_copy(wait->addr, addr);
+	init_completion(&wait->done);
+
+	spin_lock_bh(&ar->data_lock);
+	list_add(&wait->list, &ar->peer_delete_waits);
+	spin_unlock_bh(&ar->data_lock);
+}
+
+static void ath12k_peer_delete_wait_unregister(struct ath12k *ar,
+					       struct ath12k_peer_delete_wait *wait)
+{
+	spin_lock_bh(&ar->data_lock);
+	list_del(&wait->list);
+	spin_unlock_bh(&ar->data_lock);
+}
+
+void ath12k_peer_delete_resp_signal(struct ath12k *ar, u32 vdev_id, const u8 *addr)
+{
+	struct ath12k_peer_delete_wait *wait;
+
+	guard(spinlock_bh)(&ar->data_lock);
+
+	list_for_each_entry(wait, &ar->peer_delete_waits, list) {
+		if (wait->vdev_id == vdev_id &&
+		    ether_addr_equal(wait->addr, addr)) {
+			complete(&wait->done);
+			return;
+		}
+	}
+
+	ath12k_warn(ar->ab, "failed to find link peer with vdev id %u addr %pM\n",
+		    vdev_id, addr);
+}
+
+void ath12k_peer_delete_wait_flush(struct ath12k *ar)
+{
+	struct ath12k_peer_delete_wait *wait;
+
+	spin_lock_bh(&ar->data_lock);
+	list_for_each_entry(wait, &ar->peer_delete_waits, list)
+		complete(&wait->done);
+	spin_unlock_bh(&ar->data_lock);
+}
+
 static int ath12k_wait_for_dp_link_peer_common(struct ath12k_base *ab, int vdev_id,
 					       const u8 *addr, bool expect_mapped)
 {
@@ -62,20 +111,19 @@ static int ath12k_wait_for_peer_deleted(struct ath12k *ar, int vdev_id, const u8
 	return ath12k_wait_for_dp_link_peer_common(ar->ab, vdev_id, addr, false);
 }
 
-int ath12k_wait_for_peer_delete_done(struct ath12k *ar, u32 vdev_id,
-				     const u8 *addr)
+int ath12k_wait_for_peer_delete_done(struct ath12k *ar,
+				     struct ath12k_peer_delete_wait *wait)
 {
-	int ret;
 	unsigned long time_left;
+	int ret;
 
-	ret = ath12k_wait_for_peer_deleted(ar, vdev_id, addr);
+	ret = ath12k_wait_for_peer_deleted(ar, wait->vdev_id, wait->addr);
 	if (ret) {
-		ath12k_warn(ar->ab, "failed wait for peer deleted");
+		ath12k_warn(ar->ab, "failed wait for peer deleted\n");
 		return ret;
 	}
 
-	time_left = wait_for_completion_timeout(&ar->peer_delete_done,
-						3 * HZ);
+	time_left = wait_for_completion_timeout(&wait->done, 3 * HZ);
 	if (time_left == 0) {
 		ath12k_warn(ar->ab, "Timeout in receiving peer delete response\n");
 		return -ETIMEDOUT;
@@ -91,8 +139,6 @@ static int ath12k_peer_delete_send(struct ath12k *ar, u32 vdev_id, const u8 *add
 
 	lockdep_assert_wiphy(ath12k_ar_to_hw(ar)->wiphy);
 
-	reinit_completion(&ar->peer_delete_done);
-
 	ret = ath12k_wmi_send_peer_delete_cmd(ar, addr, vdev_id);
 	if (ret) {
 		ath12k_warn(ab,
@@ -106,6 +152,7 @@ static int ath12k_peer_delete_send(struct ath12k *ar, u32 vdev_id, const u8 *add
 
 int ath12k_peer_delete(struct ath12k *ar, u32 vdev_id, u8 *addr)
 {
+	struct ath12k_peer_delete_wait wait;
 	int ret;
 
 	lockdep_assert_wiphy(ath12k_ar_to_hw(ar)->wiphy);
@@ -114,17 +161,25 @@ int ath12k_peer_delete(struct ath12k *ar, u32 vdev_id, u8 *addr)
 				     &(ath12k_ar_to_ah(ar)->dp_hw), vdev_id,
 				     addr, ar->hw_link_id);
 
+	/*
+	 * Register the stack waiter before sending so the resp_event for
+	 * this peer cannot arrive while no waiter is queued.
+	 */
+	ath12k_peer_delete_wait_register(ar, &wait, vdev_id, addr);
+
 	ret = ath12k_peer_delete_send(ar, vdev_id, addr);
 	if (ret)
-		return ret;
+		goto out;
 
-	ret = ath12k_wait_for_peer_delete_done(ar, vdev_id, addr);
+	ret = ath12k_wait_for_peer_delete_done(ar, &wait);
 	if (ret)
-		return ret;
+		goto out;
 
 	ar->num_peers--;
 
-	return 0;
+out:
+	ath12k_peer_delete_wait_unregister(ar, &wait);
+	return ret;
 }
 
 static int ath12k_wait_for_peer_created(struct ath12k *ar, int vdev_id, const u8 *addr)
@@ -184,22 +239,26 @@ int ath12k_peer_create(struct ath12k *ar, struct ath12k_link_vif *arvif,
 	peer = ath12k_dp_link_peer_find_by_vdev_and_addr(dp, arg->vdev_id,
 							 arg->peer_addr);
 	if (!peer) {
+		struct ath12k_peer_delete_wait wait;
+
 		spin_unlock_bh(&dp->dp_lock);
 		ath12k_warn(ar->ab, "failed to find peer %pM on vdev %i after creation\n",
 			    arg->peer_addr, arg->vdev_id);
 
-		reinit_completion(&ar->peer_delete_done);
+		ath12k_peer_delete_wait_register(ar, &wait, arg->vdev_id,
+						 arg->peer_addr);
 
 		ret = ath12k_wmi_send_peer_delete_cmd(ar, arg->peer_addr,
 						      arg->vdev_id);
 		if (ret) {
 			ath12k_warn(ar->ab, "failed to delete peer vdev_id %d addr %pM\n",
 				    arg->vdev_id, arg->peer_addr);
+			ath12k_peer_delete_wait_unregister(ar, &wait);
 			return ret;
 		}
 
-		ret = ath12k_wait_for_peer_delete_done(ar, arg->vdev_id,
-						       arg->peer_addr);
+		ret = ath12k_wait_for_peer_delete_done(ar, &wait);
+		ath12k_peer_delete_wait_unregister(ar, &wait);
 		if (ret)
 			return ret;
 
@@ -283,13 +342,14 @@ u16 ath12k_peer_ml_alloc(struct ath12k_hw *ah)
 
 int ath12k_peer_mlo_link_peers_delete(struct ath12k_vif *ahvif, struct ath12k_sta *ahsta)
 {
+	DECLARE_BITMAP(registered, IEEE80211_MLD_MAX_NUM_LINKS);
 	struct ieee80211_sta *sta = ath12k_ahsta_to_sta(ahsta);
 	struct ath12k_hw *ah = ahvif->ah;
 	struct ath12k_link_vif *arvif;
 	struct ath12k_link_sta *arsta;
+	int ret, err_ret = 0;
 	unsigned long links;
 	struct ath12k *ar;
-	int ret, err_ret = 0;
 	u8 link_id;
 
 	lockdep_assert_wiphy(ah->hw->wiphy);
@@ -297,8 +357,19 @@ int ath12k_peer_mlo_link_peers_delete(struct ath12k_vif *ahvif, struct ath12k_st
 	if (!sta->mlo)
 		return -EINVAL;
 
-	/* FW expects delete of all link peers at once before waiting for reception
-	 * of peer unmap or delete responses
+	struct ath12k_peer_delete_wait *waits __free(kfree) =
+				kzalloc_objs(*waits, IEEE80211_MLD_MAX_NUM_LINKS);
+	if (!waits)
+		return -ENOMEM;
+
+	bitmap_zero(registered, IEEE80211_MLD_MAX_NUM_LINKS);
+
+	/*
+	 * Firmware expects delete of all link peers at once before waiting
+	 * for reception of peer unmap or delete responses. Phase 1 registers
+	 * a per-link stack waiter and sends WMI peer delete for every
+	 * link; the resp_event handler matches each response to its
+	 * (vdev_id, addr) waiter on ar->peer_delete_waits.
 	 */
 	links = ahsta->links_map;
 	for_each_set_bit(link_id, &links, IEEE80211_MLD_MAX_NUM_LINKS) {
@@ -318,29 +389,36 @@ int ath12k_peer_mlo_link_peers_delete(struct ath12k_vif *ahvif, struct ath12k_st
 					     arvif->vdev_id, arsta->addr,
 					     ar->hw_link_id);
 
+		ath12k_peer_delete_wait_register(ar, &waits[link_id],
+						 arvif->vdev_id, arsta->addr);
+
 		ret = ath12k_peer_delete_send(ar, arvif->vdev_id, arsta->addr);
 		if (ret) {
 			ath12k_warn(ar->ab,
 				    "failed to delete peer vdev_id %d addr %pM ret %d\n",
 				    arvif->vdev_id, arsta->addr, ret);
 			err_ret = ret;
+			ath12k_peer_delete_wait_unregister(ar, &waits[link_id]);
 			continue;
 		}
+
+		set_bit(link_id, registered);
 	}
 
-	/* Ensure all link peers are deleted and unmapped */
+	/*
+	 * Phase 2: wait for unmap + delete_resp on each registered link
+	 * and tear down the waiter.
+	 */
 	links = ahsta->links_map;
 	for_each_set_bit(link_id, &links, IEEE80211_MLD_MAX_NUM_LINKS) {
-		arvif = wiphy_dereference(ah->hw->wiphy, ahvif->link[link_id]);
-		arsta = wiphy_dereference(ah->hw->wiphy, ahsta->link[link_id]);
-		if (!arvif || !arsta)
+		if (!test_bit(link_id, registered))
 			continue;
 
+		arvif = wiphy_dereference(ah->hw->wiphy, ahvif->link[link_id]);
 		ar = arvif->ar;
-		if (!ar)
-			continue;
 
-		ret = ath12k_wait_for_peer_delete_done(ar, arvif->vdev_id, arsta->addr);
+		ret = ath12k_wait_for_peer_delete_done(ar, &waits[link_id]);
+		ath12k_peer_delete_wait_unregister(ar, &waits[link_id]);
 		if (ret) {
 			err_ret = ret;
 			continue;
diff --git a/drivers/net/wireless/ath/ath12k/peer.h b/drivers/net/wireless/ath/ath12k/peer.h
index 49d89796bc46..9343944c2b8e 100644
--- a/drivers/net/wireless/ath/ath12k/peer.h
+++ b/drivers/net/wireless/ath/ath12k/peer.h
@@ -9,13 +9,23 @@
 
 #include "dp_peer.h"
 
+struct ath12k_peer_delete_wait {
+	struct list_head list;
+	u32 vdev_id;
+	u8 addr[ETH_ALEN];
+	struct completion done;
+};
+
+void ath12k_peer_delete_resp_signal(struct ath12k *ar, u32 vdev_id, const u8 *addr);
+void ath12k_peer_delete_wait_flush(struct ath12k *ar);
+
 void ath12k_peer_cleanup(struct ath12k *ar, u32 vdev_id);
 int ath12k_peer_delete(struct ath12k *ar, u32 vdev_id, u8 *addr);
 int ath12k_peer_create(struct ath12k *ar, struct ath12k_link_vif *arvif,
 		       struct ieee80211_sta *sta,
 		       struct ath12k_wmi_peer_create_arg *arg);
-int ath12k_wait_for_peer_delete_done(struct ath12k *ar, u32 vdev_id,
-				     const u8 *addr);
+int ath12k_wait_for_peer_delete_done(struct ath12k *ar,
+				     struct ath12k_peer_delete_wait *wait);
 int ath12k_peer_mlo_link_peers_delete(struct ath12k_vif *ahvif, struct ath12k_sta *ahsta);
 struct ath12k_ml_peer *ath12k_peer_ml_find(struct ath12k_hw *ah,
 					   const u8 *addr);
diff --git a/drivers/net/wireless/ath/ath12k/wmi.c b/drivers/net/wireless/ath/ath12k/wmi.c
index 84a31b953db8..6066ca8d9fc4 100644
--- a/drivers/net/wireless/ath/ath12k/wmi.c
+++ b/drivers/net/wireless/ath/ath12k/wmi.c
@@ -7072,25 +7072,29 @@ static void ath12k_peer_delete_resp_event(struct ath12k_base *ab, struct sk_buff
 {
 	struct wmi_peer_delete_resp_event peer_del_resp;
 	struct ath12k *ar;
+	u32 vdev_id;
 
 	if (ath12k_pull_peer_del_resp_ev(ab, skb, &peer_del_resp) != 0) {
-		ath12k_warn(ab, "failed to extract peer delete resp");
+		ath12k_warn(ab, "failed to extract peer delete resp\n");
 		return;
 	}
 
+	vdev_id = le32_to_cpu(peer_del_resp.vdev_id);
+
 	rcu_read_lock();
-	ar = ath12k_mac_get_ar_by_vdev_id(ab, le32_to_cpu(peer_del_resp.vdev_id));
+	ar = ath12k_mac_get_ar_by_vdev_id(ab, vdev_id);
 	if (!ar) {
-		ath12k_warn(ab, "invalid vdev id in peer delete resp ev %d",
-			    peer_del_resp.vdev_id);
+		ath12k_warn(ab, "invalid vdev id in peer delete resp ev %d\n",
+			    vdev_id);
 		rcu_read_unlock();
 		return;
 	}
 
-	complete(&ar->peer_delete_done);
+	ath12k_peer_delete_resp_signal(ar, vdev_id,
+				       peer_del_resp.peer_macaddr.addr);
 	rcu_read_unlock();
 	ath12k_dbg(ab, ATH12K_DBG_WMI, "peer delete resp for vdev id %d addr %pM\n",
-		   peer_del_resp.vdev_id, peer_del_resp.peer_macaddr.addr);
+		   vdev_id, peer_del_resp.peer_macaddr.addr);
 }
 
 static void ath12k_vdev_delete_resp_event(struct ath12k_base *ab,

-- 
2.25.1
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.