[PATCH] wifi: ath11k: don't report firmware-dropped TX frames as failed

Christopher Griffin via B4 Relay devnull+christophergriffin05.gmail.com at kernel.org
Fri Nov 6 06:24:58 PST 2026


From: Christopher Griffin <christophergriffin05 at gmail.com>

When the firmware drops a frame from its TX queue without sending it,
the TX completion carries a TQM release reason other than FRAME_ACKED,
and ath11k_dp_tx_complete_msdu() reports it to mac80211 as an unacked
frame. mac80211 then counts it as a failed transmission: the station's
tx_failed and the mesh path metric's failure average go up although
nothing was lost on the air. With extended TX stats enabled, such a
completion also overwrites the station's TX rate with the empty rate
fields of the completion.

On an IPQ8074 AP, three 30 s UDP downloads offered at 650 Mbit/s into a
40 MHz HE link that carries about 400 Mbit/s raised the station's
tx_failed by 659146, 826963 and 1020775, matching the datagrams iperf
lost, while the firmware's PPDU statistics counted no failed MPDU. Every
one of those completions carried HAL_WBM_TQM_REL_REASON_CMD_REMOVE_MPDU.

Handle the release reasons the way ath12k does since
commit 50971dc6694c ("wifi: ath12k: drop failed transmitted frames from metric calculation.")
and free the frames the firmware removed itself (CMD_REMOVE_MPDU,
CMD_REMOVE_AGED_FRAMES, and CMD_REMOVE_TX of a frame that expected an
ACK) with ieee80211_free_txskb() instead of reporting a TX status,
before the extended TX stats are updated. With the change, three runs
of the same download left tx_failed unchanged; TCP and UDP throughput
stayed within the run-to-run spread of the runs without it.

Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.9.0.1-02146-QCAHKSWPL_SILICONZ-1

Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Assisted-by: Claude:claude-opus-5-5
Signed-off-by: Christopher Griffin <christophergriffin05 at gmail.com>
---
ath12k stopped reporting the firmware's own TX drops as failed
transmissions in 2024; ath11k still does. Found on an Arista C-360
(IPQ8074) while comparing the firmware's per-MCS TX statistics with the
station's tx_failed under an overloaded UDP download: the firmware
counted no failed MPDU while tx_failed matched the datagrams iperf lost.
---
 drivers/net/wireless/ath/ath11k/dp_tx.c | 41 +++++++++++++++++++++++----------
 1 file changed, 29 insertions(+), 12 deletions(-)

diff --git a/drivers/net/wireless/ath/ath11k/dp_tx.c b/drivers/net/wireless/ath/ath11k/dp_tx.c
index cac970c92806a..8f619d89ec100 100644
--- a/drivers/net/wireless/ath/ath11k/dp_tx.c
+++ b/drivers/net/wireless/ath/ath11k/dp_tx.c
@@ -586,22 +586,39 @@ static void ath11k_dp_tx_complete_msdu(struct ath11k *ar,
 	/* skip tx rate update from ieee80211_status*/
 	info->status.rates[0].idx = -1;
 
-	if (ts->status == HAL_WBM_TQM_REL_REASON_FRAME_ACKED &&
-	    !(info->flags & IEEE80211_TX_CTL_NO_ACK)) {
-		info->flags |= IEEE80211_TX_STAT_ACK;
-		info->status.ack_signal = ts->ack_rssi;
+	switch (ts->status) {
+	case HAL_WBM_TQM_REL_REASON_FRAME_ACKED:
+		if (!(info->flags & IEEE80211_TX_CTL_NO_ACK)) {
+			info->flags |= IEEE80211_TX_STAT_ACK;
+			info->status.ack_signal = ts->ack_rssi;
 
-		if (!test_bit(WMI_TLV_SERVICE_HW_DB2DBM_CONVERSION_SUPPORT,
-			      ab->wmi_ab.svc_map))
-			info->status.ack_signal += ATH11K_DEFAULT_NOISE_FLOOR;
+			if (!test_bit(WMI_TLV_SERVICE_HW_DB2DBM_CONVERSION_SUPPORT,
+				      ab->wmi_ab.svc_map))
+				info->status.ack_signal += ATH11K_DEFAULT_NOISE_FLOOR;
 
-		info->status.flags |= IEEE80211_TX_STATUS_ACK_SIGNAL_VALID;
+			info->status.flags |= IEEE80211_TX_STATUS_ACK_SIGNAL_VALID;
+		}
+		break;
+	case HAL_WBM_TQM_REL_REASON_CMD_REMOVE_TX:
+		if (info->flags & IEEE80211_TX_CTL_NO_ACK) {
+			info->flags |= IEEE80211_TX_STAT_NOACK_TRANSMITTED;
+			break;
+		}
+		fallthrough;
+	case HAL_WBM_TQM_REL_REASON_CMD_REMOVE_MPDU:
+	case HAL_WBM_TQM_REL_REASON_CMD_REMOVE_AGED_FRAMES:
+		/* The firmware dropped the frame itself (for example when the
+		 * peer's queue was full); it was not lost on the air, so do
+		 * not report it to mac80211 as a failed transmission.
+		 */
+		ieee80211_free_txskb(ar->hw, msdu);
+		return;
+	default:
+		ath11k_dbg(ab, ATH11K_DBG_DP_TX, "tx frame is not acked status %d\n",
+			   ts->status);
+		break;
 	}
 
-	if (ts->status == HAL_WBM_TQM_REL_REASON_CMD_REMOVE_TX &&
-	    (info->flags & IEEE80211_TX_CTL_NO_ACK))
-		info->flags |= IEEE80211_TX_STAT_NOACK_TRANSMITTED;
-
 	if (unlikely(ath11k_debugfs_is_extd_tx_stats_enabled(ar)) ||
 	    ab->hw_params.single_pdev_only) {
 		if (ts->flags & HAL_TX_STATUS_FLAGS_FIRST_MSDU) {

---
base-commit: 5c8eab13e391aaabd5a1ad58f3ee4bb10ce7573e
change-id: 20261106-ath11k-tqm-internal-drops-4770f6c1e48e

Best regards,
--  
Christopher Griffin <christophergriffin05 at gmail.com>





More information about the ath11k mailing list