mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH ath-next] wifi: ath9k_htc: count a frame as queued before it is sent
@ 2026-11-11 21:59 Mehmet Fide
  2026-11-16 16:54 ` Toke Høiland-Jørgensen
  0 siblings, 1 reply; 3+ messages in thread
From: Mehmet Fide @ 2026-11-11 21:59 UTC (permalink / raw)
  To: Toke Høiland-Jørgensen
  Cc: John W . Linville, Sujith Manoharan, linux-wireless,
	linux-kernel, mehmet.fide

From: Mehmet Fide <mehmet.fide@screeningeagle.com>

ath9k_htc_tx() hands the frame to the device and counts it in
tx.queued_cnt afterwards. The TX status of the frame is handled in the
WMI event tasklet, which can run before the sender gets to the counter.
ath9k_htc_tx_process() then takes the counter below zero, warns and
sets it to zero, and the sender's increment leaves it one too high for
good:

WARNING: CPU: 0 PID: 348 at drivers/net/wireless/ath/ath9k/htc_drv_txrx.c:510 ath9k_htc_tx_process+0x400/0x558 [ath9k_htc]
Call trace:
 ath9k_htc_tx_process+0x400/0x558 [ath9k_htc]
 ath9k_htc_txstatus+0x58/0x140 [ath9k_htc]
 ath9k_wmi_event_tasklet+0x16c/0x17c [ath9k_htc]
 tasklet_action_common+0x108/0x150

Count the frame before it is sent and take it back if the send fails.

Fixes: 8e86a54715c4 ("ath9k_htc: Fix TX queue management")
Signed-off-by: Mehmet Fide <mehmet.fide@screeningeagle.com>
---
 drivers/net/wireless/ath/ath9k/htc_drv_main.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/ath/ath9k/htc_drv_main.c b/drivers/net/wireless/ath/ath9k/htc_drv_main.c
index 0d6272ac0dac..ee87352d1241 100644
--- a/drivers/net/wireless/ath/ath9k/htc_drv_main.c
+++ b/drivers/net/wireless/ath/ath9k/htc_drv_main.c
@@ -888,17 +888,21 @@ static void ath9k_htc_tx(struct ieee80211_hw *hw,
 		goto fail_tx;
 	}
 
+	ath9k_htc_check_stop_queues(priv);
+
 	ret = ath9k_htc_tx_start(priv, control->sta, skb, slot, false);
 	if (ret != 0) {
 		ath_dbg(common, XMIT, "Tx failed\n");
 		goto clear_slot;
 	}
 
-	ath9k_htc_check_stop_queues(priv);
-
 	return;
 
 clear_slot:
+	spin_lock_bh(&priv->tx.tx_lock);
+	priv->tx.queued_cnt--;
+	spin_unlock_bh(&priv->tx.tx_lock);
+	ath9k_htc_check_wake_queues(priv);
 	ath9k_htc_tx_clear_slot(priv, slot);
 fail_tx:
 	dev_kfree_skb_any(skb);

base-commit: ff4c7359e310744a7abe53cf66d921fa2969eb19
-- 
2.56.0


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH ath-next] wifi: ath9k_htc: count a frame as queued before it is sent
  2026-11-11 21:59 [PATCH ath-next] wifi: ath9k_htc: count a frame as queued before it is sent Mehmet Fide
@ 2026-11-16 16:54 ` Toke Høiland-Jørgensen
  2026-11-17  7:14   ` Mehmet Fide
  0 siblings, 1 reply; 3+ messages in thread
From: Toke Høiland-Jørgensen @ 2026-11-16 16:54 UTC (permalink / raw)
  To: Mehmet Fide
  Cc: Toke Høiland-Jørgensen, John W . Linville,
	Sujith Manoharan, linux-wireless, linux-kernel, mehmet.fide

On Wed, 11 Nov 2026 22:59:02 +0100, Mehmet Fide <mehmet.fide@gmail.com> wrote:
> diff --git a/drivers/net/wireless/ath/ath9k/htc_drv_main.c b/drivers/net/wireless/ath/ath9k/htc_drv_main.c
> index b46be3682f21..9467b8e512a6 100644
> --- a/drivers/net/wireless/ath/ath9k/htc_drv_main.c
> +++ b/drivers/net/wireless/ath/ath9k/htc_drv_main.c
> @@ -888,17 +888,21 @@ static void ath9k_htc_tx(struct ieee80211_hw *hw,
> [ ... skip 14 lines ... ]
>  
>  clear_slot:
> +	spin_lock_bh(&priv->tx.tx_lock);
> +	priv->tx.queued_cnt--;
> +	spin_unlock_bh(&priv->tx.tx_lock);
> +	ath9k_htc_check_wake_queues(priv);

Hmm, so this results in taking the lock twice, once for decrementing and once
for checking. How about we define a __ath9k_htc_check_wake_queues() function which
takes a boolean "decrement" as the second parameter, and does the decrement
under the lock if set? Then we can turn ath9k_htc_check_wake_queues(priv) into a
wrapper which just calls __ath9k_htc_check_wake_queues(priv, false).

-Toke

-- 


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH ath-next] wifi: ath9k_htc: count a frame as queued before it is sent
  2026-11-16 16:54 ` Toke Høiland-Jørgensen
@ 2026-11-17  7:14   ` Mehmet Fide
  0 siblings, 0 replies; 3+ messages in thread
From: Mehmet Fide @ 2026-11-17  7:14 UTC (permalink / raw)
  To: Toke Høiland-Jørgensen
  Cc: John W . Linville, Sujith Manoharan, linux-wireless,
	linux-kernel, mehmet.fide

Hi Toke,

> How about we define a __ath9k_htc_check_wake_queues() function which
> takes a boolean "decrement" as the second parameter, and does the decrement
> under the lock if set? Then we can turn ath9k_htc_check_wake_queues(priv) into a
> wrapper which just calls __ath9k_htc_check_wake_queues(priv, false).

Sure, will do in v2.

Mehmet

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-11-17  7:15 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-11 21:59 [PATCH ath-next] wifi: ath9k_htc: count a frame as queued before it is sent Mehmet Fide
2026-11-16 16:54 ` Toke Høiland-Jørgensen
2026-11-17  7:14   ` Mehmet Fide

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®