[PATCH v2 0/5] wifi: ath11k: fix dest ring-buffer corruption

[PATCH v2 0/5] wifi: ath11k: fix dest ring-buffer corruption

[Johan Hovold]

As a follow up to commits:

	6d037a372f81 ("wifi: ath11k: fix ring-buffer corruption")
	ab52e3e44fe9 ("wifi: ath11k: fix rx completion meta data corruption")

add the remaining missing memory barriers to make sure that destination
ring descriptors are read after the head pointers to avoid using stale
data on weakly ordered architectures like aarch64.

Also switch back to plain accesses for the descriptor fields which is
sufficient after the memory barrier.

New in v2 are two patches that adds the missing barriers also for source
rings and when updating the tail pointer for destination rings.

To avoid leaking ring details from the "hal" (lmac or non-lmac), the
barriers are added to the ath11k_hal_srng_access_end() helper. For
symmetry I therefore moved also the dest ring barriers into
ath11k_hal_srng_access_begin() and made the barrier conditional.

[ Due to this change I did not add Miaoqing's reviewed-by tag. ]

Johan


Changes in v2:
 - add tested-on tags to plain access patches
 - move destination barriers into begin helper
 - fix source ring corruption (new patch)
 - fix dest ring corruption when ring is full (new patch)


Johan Hovold (5):
  wifi: ath11k: fix dest ring-buffer corruption
  wifi: ath11k: use plain access for descriptor length
  wifi: ath11k: use plain accesses for monitor descriptor
  wifi: ath11k: fix source ring-buffer corruption
  wifi: ath11k: fix dest ring-buffer corruption when ring is full

 drivers/net/wireless/ath/ath11k/ce.c    |  3 ---
 drivers/net/wireless/ath/ath11k/dp_rx.c | 25 +++++++-----------
 drivers/net/wireless/ath/ath11k/hal.c   | 35 +++++++++++++++++++++----
 3 files changed, 39 insertions(+), 24 deletions(-)

-- 
2.49.0

[PATCH v2 1/5] wifi: ath11k: fix dest ring-buffer corruption

[Johan Hovold]

Add the missing memory barrier to make sure that destination ring
descriptors are read after the head pointers to avoid using stale data
on weakly ordered architectures like aarch64.

The barrier is added to the ath11k_hal_srng_access_begin() helper for
symmetry with follow-on fixes for source ring buffer corruption which
will add barriers to ath11k_hal_srng_access_end().

Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41

Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Cc: stable@vger.kernel.org	# 5.6
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
---
 drivers/net/wireless/ath/ath11k/ce.c    |  3 ---
 drivers/net/wireless/ath/ath11k/dp_rx.c |  3 ---
 drivers/net/wireless/ath/ath11k/hal.c   | 12 +++++++++++-
 3 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/net/wireless/ath/ath11k/ce.c b/drivers/net/wireless/ath/ath11k/ce.c
index 9d8efec46508..39d9aad33bc6 100644
--- a/drivers/net/wireless/ath/ath11k/ce.c
+++ b/drivers/net/wireless/ath/ath11k/ce.c
@@ -393,9 +393,6 @@ static int ath11k_ce_completed_recv_next(struct ath11k_ce_pipe *pipe,
 		goto err;
 	}
 
-	/* Make sure descriptor is read after the head pointer. */
-	dma_rmb();
-
 	*nbytes = ath11k_hal_ce_dst_status_get_length(desc);
 
 	*skb = pipe->dest_ring->skb[sw_index];
diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
index ea2959305dec..d8dab182a9af 100644
--- a/drivers/net/wireless/ath/ath11k/dp_rx.c
+++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
@@ -2650,9 +2650,6 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
 try_again:
 	ath11k_hal_srng_access_begin(ab, srng);
 
-	/* Make sure descriptor is read after the head pointer. */
-	dma_rmb();
-
 	while (likely(desc =
 	      (struct hal_reo_dest_ring *)ath11k_hal_srng_dst_get_next_entry(ab,
 									     srng))) {
diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
index 8cb1505a5a0c..921114686ba3 100644
--- a/drivers/net/wireless/ath/ath11k/hal.c
+++ b/drivers/net/wireless/ath/ath11k/hal.c
@@ -823,13 +823,23 @@ u32 *ath11k_hal_srng_src_peek(struct ath11k_base *ab, struct hal_srng *srng)
 
 void ath11k_hal_srng_access_begin(struct ath11k_base *ab, struct hal_srng *srng)
 {
+	u32 hp;
+
 	lockdep_assert_held(&srng->lock);
 
 	if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
 		srng->u.src_ring.cached_tp =
 			*(volatile u32 *)srng->u.src_ring.tp_addr;
 	} else {
-		srng->u.dst_ring.cached_hp = READ_ONCE(*srng->u.dst_ring.hp_addr);
+		hp = READ_ONCE(*srng->u.dst_ring.hp_addr);
+
+		if (hp != srng->u.dst_ring.cached_hp) {
+			srng->u.dst_ring.cached_hp = hp;
+			/* Make sure descriptor is read after the head
+			 * pointer.
+			 */
+			dma_rmb();
+		}
 
 		/* Try to prefetch the next descriptor in the ring */
 		if (srng->flags & HAL_SRNG_FLAGS_CACHED)
-- 
2.49.0

[PATCH v2 2/5] wifi: ath11k: use plain access for descriptor length

[Johan Hovold]

The read memory barrier added by commit 6d037a372f81 ("wifi: ath11k: fix
ring-buffer corruption") is enough to guarantee ordering also for plain
descriptor accesses if the length helper is ever inlined so drop the
unnecessary READ_ONCE().

Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41

Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
---
 drivers/net/wireless/ath/ath11k/hal.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
index 921114686ba3..b1f5a927dddd 100644
--- a/drivers/net/wireless/ath/ath11k/hal.c
+++ b/drivers/net/wireless/ath/ath11k/hal.c
@@ -599,7 +599,7 @@ u32 ath11k_hal_ce_dst_status_get_length(void *buf)
 	struct hal_ce_srng_dst_status_desc *desc = buf;
 	u32 len;
 
-	len = FIELD_GET(HAL_CE_DST_STATUS_DESC_FLAGS_LEN, READ_ONCE(desc->flags));
+	len = FIELD_GET(HAL_CE_DST_STATUS_DESC_FLAGS_LEN, desc->flags);
 	desc->flags &= ~HAL_CE_DST_STATUS_DESC_FLAGS_LEN;
 
 	return len;
-- 
2.49.0

[PATCH v2 3/5] wifi: ath11k: use plain accesses for monitor descriptor

[Johan Hovold]

The read memory barrier added by commit ab52e3e44fe9 ("wifi: ath11k: fix
rx completion meta data corruption") is enough to guarantee ordering
also for plain descriptor accesses so drop the unnecessary READ_ONCE().

Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41

Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
---
 drivers/net/wireless/ath/ath11k/dp_rx.c | 22 +++++++++-------------
 1 file changed, 9 insertions(+), 13 deletions(-)

diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
index d8dab182a9af..218ab41c0f3c 100644
--- a/drivers/net/wireless/ath/ath11k/dp_rx.c
+++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
@@ -2637,7 +2637,7 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
 	struct ath11k *ar;
 	struct hal_reo_dest_ring *desc;
 	enum hal_reo_dest_ring_push_reason push_reason;
-	u32 cookie, info0, rx_msdu_info0, rx_mpdu_info0;
+	u32 cookie;
 	int i;
 
 	for (i = 0; i < MAX_RADIOS; i++)
@@ -2654,7 +2654,7 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
 	      (struct hal_reo_dest_ring *)ath11k_hal_srng_dst_get_next_entry(ab,
 									     srng))) {
 		cookie = FIELD_GET(BUFFER_ADDR_INFO1_SW_COOKIE,
-				   READ_ONCE(desc->buf_addr_info.info1));
+				   desc->buf_addr_info.info1);
 		buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
 				   cookie);
 		mac_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_PDEV_ID, cookie);
@@ -2683,9 +2683,8 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
 
 		num_buffs_reaped[mac_id]++;
 
-		info0 = READ_ONCE(desc->info0);
 		push_reason = FIELD_GET(HAL_REO_DEST_RING_INFO0_PUSH_REASON,
-					info0);
+					desc->info0);
 		if (unlikely(push_reason !=
 			     HAL_REO_DEST_RING_PUSH_REASON_ROUTING_INSTRUCTION)) {
 			dev_kfree_skb_any(msdu);
@@ -2693,21 +2692,18 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
 			continue;
 		}
 
-		rx_msdu_info0 = READ_ONCE(desc->rx_msdu_info.info0);
-		rx_mpdu_info0 = READ_ONCE(desc->rx_mpdu_info.info0);
-
-		rxcb->is_first_msdu = !!(rx_msdu_info0 &
+		rxcb->is_first_msdu = !!(desc->rx_msdu_info.info0 &
 					 RX_MSDU_DESC_INFO0_FIRST_MSDU_IN_MPDU);
-		rxcb->is_last_msdu = !!(rx_msdu_info0 &
+		rxcb->is_last_msdu = !!(desc->rx_msdu_info.info0 &
 					RX_MSDU_DESC_INFO0_LAST_MSDU_IN_MPDU);
-		rxcb->is_continuation = !!(rx_msdu_info0 &
+		rxcb->is_continuation = !!(desc->rx_msdu_info.info0 &
 					   RX_MSDU_DESC_INFO0_MSDU_CONTINUATION);
 		rxcb->peer_id = FIELD_GET(RX_MPDU_DESC_META_DATA_PEER_ID,
-					  READ_ONCE(desc->rx_mpdu_info.meta_data));
+					  desc->rx_mpdu_info.meta_data);
 		rxcb->seq_no = FIELD_GET(RX_MPDU_DESC_INFO0_SEQ_NUM,
-					 rx_mpdu_info0);
+					 desc->rx_mpdu_info.info0);
 		rxcb->tid = FIELD_GET(HAL_REO_DEST_RING_INFO0_RX_QUEUE_NUM,
-				      info0);
+				      desc->info0);
 
 		rxcb->mac_id = mac_id;
 		__skb_queue_tail(&msdu_list[mac_id], msdu);
-- 
2.49.0

[PATCH v2 4/5] wifi: ath11k: fix source ring-buffer corruption

[Johan Hovold]

Add the missing memory barrier to make sure that LMAC source ring
descriptors are written before updating the head pointer to avoid
passing stale data to the firmware on weakly ordered architectures like
aarch64.

Note that non-LMAC rings use MMIO write accessors which have the
required write memory barrier.

Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41

Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Cc: stable@vger.kernel.org      # 5.6
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
---
 drivers/net/wireless/ath/ath11k/hal.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
index b1f5a927dddd..927ed2bc3fbf 100644
--- a/drivers/net/wireless/ath/ath11k/hal.c
+++ b/drivers/net/wireless/ath/ath11k/hal.c
@@ -862,7 +862,11 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
 		if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
 			srng->u.src_ring.last_tp =
 				*(volatile u32 *)srng->u.src_ring.tp_addr;
-			*srng->u.src_ring.hp_addr = srng->u.src_ring.hp;
+			/* Make sure descriptor is written before updating the
+			 * head pointer.
+			 */
+			dma_wmb();
+			WRITE_ONCE(*srng->u.src_ring.hp_addr, srng->u.src_ring.hp);
 		} else {
 			srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
 			*srng->u.dst_ring.tp_addr = srng->u.dst_ring.tp;
@@ -871,6 +875,10 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
 		if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
 			srng->u.src_ring.last_tp =
 				*(volatile u32 *)srng->u.src_ring.tp_addr;
+			/* Assume implementation use an MMIO write accessor
+			 * which has the required wmb() so that the descriptor
+			 * is written before the updating the head pointer.
+			 */
 			ath11k_hif_write32(ab,
 					   (unsigned long)srng->u.src_ring.hp_addr -
 					   (unsigned long)ab->mem,
-- 
2.49.0

[PATCH v2 5/5] wifi: ath11k: fix dest ring-buffer corruption when ring is full

[Johan Hovold]

Add the missing memory barriers to make sure that destination ring
descriptors are read before updating the tail pointer (and passing
ownership to the device) to avoid memory corruption on weakly ordered
architectures like aarch64 when the ring is full.

Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41

Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Cc: stable@vger.kernel.org      # 5.6
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
---
 drivers/net/wireless/ath/ath11k/hal.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
index 927ed2bc3fbf..7eeffb36899e 100644
--- a/drivers/net/wireless/ath/ath11k/hal.c
+++ b/drivers/net/wireless/ath/ath11k/hal.c
@@ -854,7 +854,6 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
 {
 	lockdep_assert_held(&srng->lock);
 
-	/* TODO: See if we need a write memory barrier here */
 	if (srng->flags & HAL_SRNG_FLAGS_LMAC_RING) {
 		/* For LMAC rings, ring pointer updates are done through FW and
 		 * hence written to a shared memory location that is read by FW
@@ -869,7 +868,11 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
 			WRITE_ONCE(*srng->u.src_ring.hp_addr, srng->u.src_ring.hp);
 		} else {
 			srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
-			*srng->u.dst_ring.tp_addr = srng->u.dst_ring.tp;
+			/* Make sure descriptor is read before updating the
+			 * tail pointer.
+			 */
+			dma_mb();
+			WRITE_ONCE(*srng->u.dst_ring.tp_addr, srng->u.dst_ring.tp);
 		}
 	} else {
 		if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
@@ -885,6 +888,10 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
 					   srng->u.src_ring.hp);
 		} else {
 			srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
+			/* Make sure descriptor is read before updating the
+			 * tail pointer.
+			 */
+			mb();
 			ath11k_hif_write32(ab,
 					   (unsigned long)srng->u.dst_ring.tp_addr -
 					   (unsigned long)ab->mem,
-- 
2.49.0

Re: [PATCH v2 1/5] wifi: ath11k: fix dest ring-buffer corruption

[Baochen Qiang]

On 6/4/2025 10:34 PM, Johan Hovold wrote:
> Add the missing memory barrier to make sure that destination ring
> descriptors are read after the head pointers to avoid using stale data
> on weakly ordered architectures like aarch64.
> 
> The barrier is added to the ath11k_hal_srng_access_begin() helper for
> symmetry with follow-on fixes for source ring buffer corruption which
> will add barriers to ath11k_hal_srng_access_end().
> 
> Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41
> 
> Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
> Cc: stable@vger.kernel.org	# 5.6
> Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
> ---
>  drivers/net/wireless/ath/ath11k/ce.c    |  3 ---
>  drivers/net/wireless/ath/ath11k/dp_rx.c |  3 ---
>  drivers/net/wireless/ath/ath11k/hal.c   | 12 +++++++++++-
>  3 files changed, 11 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/ce.c b/drivers/net/wireless/ath/ath11k/ce.c
> index 9d8efec46508..39d9aad33bc6 100644
> --- a/drivers/net/wireless/ath/ath11k/ce.c
> +++ b/drivers/net/wireless/ath/ath11k/ce.c
> @@ -393,9 +393,6 @@ static int ath11k_ce_completed_recv_next(struct ath11k_ce_pipe *pipe,
>  		goto err;
>  	}
>  
> -	/* Make sure descriptor is read after the head pointer. */
> -	dma_rmb();
> -
>  	*nbytes = ath11k_hal_ce_dst_status_get_length(desc);
>  
>  	*skb = pipe->dest_ring->skb[sw_index];
> diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
> index ea2959305dec..d8dab182a9af 100644
> --- a/drivers/net/wireless/ath/ath11k/dp_rx.c
> +++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
> @@ -2650,9 +2650,6 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
>  try_again:
>  	ath11k_hal_srng_access_begin(ab, srng);
>  
> -	/* Make sure descriptor is read after the head pointer. */
> -	dma_rmb();
> -
>  	while (likely(desc =
>  	      (struct hal_reo_dest_ring *)ath11k_hal_srng_dst_get_next_entry(ab,
>  									     srng))) {
> diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
> index 8cb1505a5a0c..921114686ba3 100644
> --- a/drivers/net/wireless/ath/ath11k/hal.c
> +++ b/drivers/net/wireless/ath/ath11k/hal.c
> @@ -823,13 +823,23 @@ u32 *ath11k_hal_srng_src_peek(struct ath11k_base *ab, struct hal_srng *srng)
>  
>  void ath11k_hal_srng_access_begin(struct ath11k_base *ab, struct hal_srng *srng)
>  {
> +	u32 hp;
> +
>  	lockdep_assert_held(&srng->lock);
>  
>  	if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
>  		srng->u.src_ring.cached_tp =
>  			*(volatile u32 *)srng->u.src_ring.tp_addr;
>  	} else {
> -		srng->u.dst_ring.cached_hp = READ_ONCE(*srng->u.dst_ring.hp_addr);
> +		hp = READ_ONCE(*srng->u.dst_ring.hp_addr);
> +
> +		if (hp != srng->u.dst_ring.cached_hp) {

My ath12k comments apply here: this consumes more CPU cycles

> +			srng->u.dst_ring.cached_hp = hp;
> +			/* Make sure descriptor is read after the head
> +			 * pointer.
> +			 */
> +			dma_rmb();
> +		}
>  
>  		/* Try to prefetch the next descriptor in the ring */
>  		if (srng->flags & HAL_SRNG_FLAGS_CACHED)

Re: [PATCH v2 1/5] wifi: ath11k: fix dest ring-buffer corruption

[Baochen Qiang]

On 6/4/2025 10:34 PM, Johan Hovold wrote:
> Add the missing memory barrier to make sure that destination ring
> descriptors are read after the head pointers to avoid using stale data
> on weakly ordered architectures like aarch64.
> 
> The barrier is added to the ath11k_hal_srng_access_begin() helper for
> symmetry with follow-on fixes for source ring buffer corruption which
> will add barriers to ath11k_hal_srng_access_end().
> 
> Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41
> 
> Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
> Cc: stable@vger.kernel.org	# 5.6
> Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
> ---
>  drivers/net/wireless/ath/ath11k/ce.c    |  3 ---
>  drivers/net/wireless/ath/ath11k/dp_rx.c |  3 ---
>  drivers/net/wireless/ath/ath11k/hal.c   | 12 +++++++++++-
>  3 files changed, 11 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/ce.c b/drivers/net/wireless/ath/ath11k/ce.c
> index 9d8efec46508..39d9aad33bc6 100644
> --- a/drivers/net/wireless/ath/ath11k/ce.c
> +++ b/drivers/net/wireless/ath/ath11k/ce.c
> @@ -393,9 +393,6 @@ static int ath11k_ce_completed_recv_next(struct ath11k_ce_pipe *pipe,
>  		goto err;
>  	}
>  
> -	/* Make sure descriptor is read after the head pointer. */
> -	dma_rmb();
> -
>  	*nbytes = ath11k_hal_ce_dst_status_get_length(desc);
>  
>  	*skb = pipe->dest_ring->skb[sw_index];
> diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
> index ea2959305dec..d8dab182a9af 100644
> --- a/drivers/net/wireless/ath/ath11k/dp_rx.c
> +++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
> @@ -2650,9 +2650,6 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
>  try_again:
>  	ath11k_hal_srng_access_begin(ab, srng);
>  
> -	/* Make sure descriptor is read after the head pointer. */
> -	dma_rmb();
> -
>  	while (likely(desc =
>  	      (struct hal_reo_dest_ring *)ath11k_hal_srng_dst_get_next_entry(ab,
>  									     srng))) {
> diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
> index 8cb1505a5a0c..921114686ba3 100644
> --- a/drivers/net/wireless/ath/ath11k/hal.c
> +++ b/drivers/net/wireless/ath/ath11k/hal.c
> @@ -823,13 +823,23 @@ u32 *ath11k_hal_srng_src_peek(struct ath11k_base *ab, struct hal_srng *srng)
>  
>  void ath11k_hal_srng_access_begin(struct ath11k_base *ab, struct hal_srng *srng)
>  {
> +	u32 hp;
> +
>  	lockdep_assert_held(&srng->lock);
>  
>  	if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
>  		srng->u.src_ring.cached_tp =
>  			*(volatile u32 *)srng->u.src_ring.tp_addr;
>  	} else {
> -		srng->u.dst_ring.cached_hp = READ_ONCE(*srng->u.dst_ring.hp_addr);
> +		hp = READ_ONCE(*srng->u.dst_ring.hp_addr);
> +
> +		if (hp != srng->u.dst_ring.cached_hp) {
> +			srng->u.dst_ring.cached_hp = hp;
> +			/* Make sure descriptor is read after the head
> +			 * pointer.
> +			 */
> +			dma_rmb();
> +		}
>  
>  		/* Try to prefetch the next descriptor in the ring */
>  		if (srng->flags & HAL_SRNG_FLAGS_CACHED)

Reviewed-by: Baochen Qiang <quic_bqiang@quicinc.com>

Re: [PATCH v2 2/5] wifi: ath11k: use plain access for descriptor length

[Baochen Qiang]

On 6/4/2025 10:34 PM, Johan Hovold wrote:
> The read memory barrier added by commit 6d037a372f81 ("wifi: ath11k: fix
> ring-buffer corruption") is enough to guarantee ordering also for plain
> descriptor accesses if the length helper is ever inlined so drop the
> unnecessary READ_ONCE().
> 
> Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41
> 
> Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
> ---
>  drivers/net/wireless/ath/ath11k/hal.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
> index 921114686ba3..b1f5a927dddd 100644
> --- a/drivers/net/wireless/ath/ath11k/hal.c
> +++ b/drivers/net/wireless/ath/ath11k/hal.c
> @@ -599,7 +599,7 @@ u32 ath11k_hal_ce_dst_status_get_length(void *buf)
>  	struct hal_ce_srng_dst_status_desc *desc = buf;
>  	u32 len;
>  
> -	len = FIELD_GET(HAL_CE_DST_STATUS_DESC_FLAGS_LEN, READ_ONCE(desc->flags));
> +	len = FIELD_GET(HAL_CE_DST_STATUS_DESC_FLAGS_LEN, desc->flags);
>  	desc->flags &= ~HAL_CE_DST_STATUS_DESC_FLAGS_LEN;
>  
>  	return len;

Reviewed-by: Baochen Qiang <quic_bqiang@quicinc.com>

Re: [PATCH v2 3/5] wifi: ath11k: use plain accesses for monitor descriptor

[Baochen Qiang]

On 6/4/2025 10:34 PM, Johan Hovold wrote:
> The read memory barrier added by commit ab52e3e44fe9 ("wifi: ath11k: fix
> rx completion meta data corruption") is enough to guarantee ordering
> also for plain descriptor accesses so drop the unnecessary READ_ONCE().
> 
> Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41
> 
> Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
> ---
>  drivers/net/wireless/ath/ath11k/dp_rx.c | 22 +++++++++-------------
>  1 file changed, 9 insertions(+), 13 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
> index d8dab182a9af..218ab41c0f3c 100644
> --- a/drivers/net/wireless/ath/ath11k/dp_rx.c
> +++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
> @@ -2637,7 +2637,7 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
>  	struct ath11k *ar;
>  	struct hal_reo_dest_ring *desc;
>  	enum hal_reo_dest_ring_push_reason push_reason;
> -	u32 cookie, info0, rx_msdu_info0, rx_mpdu_info0;
> +	u32 cookie;
>  	int i;
>  
>  	for (i = 0; i < MAX_RADIOS; i++)
> @@ -2654,7 +2654,7 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
>  	      (struct hal_reo_dest_ring *)ath11k_hal_srng_dst_get_next_entry(ab,
>  									     srng))) {
>  		cookie = FIELD_GET(BUFFER_ADDR_INFO1_SW_COOKIE,
> -				   READ_ONCE(desc->buf_addr_info.info1));
> +				   desc->buf_addr_info.info1);
>  		buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
>  				   cookie);
>  		mac_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_PDEV_ID, cookie);
> @@ -2683,9 +2683,8 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
>  
>  		num_buffs_reaped[mac_id]++;
>  
> -		info0 = READ_ONCE(desc->info0);
>  		push_reason = FIELD_GET(HAL_REO_DEST_RING_INFO0_PUSH_REASON,
> -					info0);
> +					desc->info0);
>  		if (unlikely(push_reason !=
>  			     HAL_REO_DEST_RING_PUSH_REASON_ROUTING_INSTRUCTION)) {
>  			dev_kfree_skb_any(msdu);
> @@ -2693,21 +2692,18 @@ int ath11k_dp_process_rx(struct ath11k_base *ab, int ring_id,
>  			continue;
>  		}
>  
> -		rx_msdu_info0 = READ_ONCE(desc->rx_msdu_info.info0);
> -		rx_mpdu_info0 = READ_ONCE(desc->rx_mpdu_info.info0);
> -
> -		rxcb->is_first_msdu = !!(rx_msdu_info0 &
> +		rxcb->is_first_msdu = !!(desc->rx_msdu_info.info0 &
>  					 RX_MSDU_DESC_INFO0_FIRST_MSDU_IN_MPDU);
> -		rxcb->is_last_msdu = !!(rx_msdu_info0 &
> +		rxcb->is_last_msdu = !!(desc->rx_msdu_info.info0 &
>  					RX_MSDU_DESC_INFO0_LAST_MSDU_IN_MPDU);
> -		rxcb->is_continuation = !!(rx_msdu_info0 &
> +		rxcb->is_continuation = !!(desc->rx_msdu_info.info0 &
>  					   RX_MSDU_DESC_INFO0_MSDU_CONTINUATION);
>  		rxcb->peer_id = FIELD_GET(RX_MPDU_DESC_META_DATA_PEER_ID,
> -					  READ_ONCE(desc->rx_mpdu_info.meta_data));
> +					  desc->rx_mpdu_info.meta_data);
>  		rxcb->seq_no = FIELD_GET(RX_MPDU_DESC_INFO0_SEQ_NUM,
> -					 rx_mpdu_info0);
> +					 desc->rx_mpdu_info.info0);
>  		rxcb->tid = FIELD_GET(HAL_REO_DEST_RING_INFO0_RX_QUEUE_NUM,
> -				      info0);
> +				      desc->info0);
>  
>  		rxcb->mac_id = mac_id;
>  		__skb_queue_tail(&msdu_list[mac_id], msdu);

Reviewed-by: Baochen Qiang <quic_bqiang@quicinc.com>

Re: [PATCH v2 4/5] wifi: ath11k: fix source ring-buffer corruption

[Baochen Qiang]

On 6/4/2025 10:34 PM, Johan Hovold wrote:
> Add the missing memory barrier to make sure that LMAC source ring
> descriptors are written before updating the head pointer to avoid
> passing stale data to the firmware on weakly ordered architectures like
> aarch64.
> 
> Note that non-LMAC rings use MMIO write accessors which have the
> required write memory barrier.
> 
> Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41
> 
> Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
> Cc: stable@vger.kernel.org      # 5.6
> Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
> ---
>  drivers/net/wireless/ath/ath11k/hal.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
> index b1f5a927dddd..927ed2bc3fbf 100644
> --- a/drivers/net/wireless/ath/ath11k/hal.c
> +++ b/drivers/net/wireless/ath/ath11k/hal.c
> @@ -862,7 +862,11 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
>  		if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
>  			srng->u.src_ring.last_tp =
>  				*(volatile u32 *)srng->u.src_ring.tp_addr;
> -			*srng->u.src_ring.hp_addr = srng->u.src_ring.hp;
> +			/* Make sure descriptor is written before updating the
> +			 * head pointer.
> +			 */
> +			dma_wmb();
> +			WRITE_ONCE(*srng->u.src_ring.hp_addr, srng->u.src_ring.hp);
>  		} else {
>  			srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
>  			*srng->u.dst_ring.tp_addr = srng->u.dst_ring.tp;
> @@ -871,6 +875,10 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
>  		if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
>  			srng->u.src_ring.last_tp =
>  				*(volatile u32 *)srng->u.src_ring.tp_addr;
> +			/* Assume implementation use an MMIO write accessor
> +			 * which has the required wmb() so that the descriptor
> +			 * is written before the updating the head pointer.
> +			 */
>  			ath11k_hif_write32(ab,
>  					   (unsigned long)srng->u.src_ring.hp_addr -
>  					   (unsigned long)ab->mem,

Reviewed-by: Baochen Qiang <quic_bqiang@quicinc.com>

Re: [PATCH v2 5/5] wifi: ath11k: fix dest ring-buffer corruption when ring is full

[Baochen Qiang]

On 6/4/2025 10:34 PM, Johan Hovold wrote:
> Add the missing memory barriers to make sure that destination ring
> descriptors are read before updating the tail pointer (and passing
> ownership to the device) to avoid memory corruption on weakly ordered
> architectures like aarch64 when the ring is full.
> 
> Tested-on: WCN6855 hw2.1 WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.41
> 
> Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
> Cc: stable@vger.kernel.org      # 5.6
> Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
> ---
>  drivers/net/wireless/ath/ath11k/hal.c | 11 +++++++++--
>  1 file changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/hal.c b/drivers/net/wireless/ath/ath11k/hal.c
> index 927ed2bc3fbf..7eeffb36899e 100644
> --- a/drivers/net/wireless/ath/ath11k/hal.c
> +++ b/drivers/net/wireless/ath/ath11k/hal.c
> @@ -854,7 +854,6 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
>  {
>  	lockdep_assert_held(&srng->lock);
>  
> -	/* TODO: See if we need a write memory barrier here */
>  	if (srng->flags & HAL_SRNG_FLAGS_LMAC_RING) {
>  		/* For LMAC rings, ring pointer updates are done through FW and
>  		 * hence written to a shared memory location that is read by FW
> @@ -869,7 +868,11 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
>  			WRITE_ONCE(*srng->u.src_ring.hp_addr, srng->u.src_ring.hp);
>  		} else {
>  			srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
> -			*srng->u.dst_ring.tp_addr = srng->u.dst_ring.tp;
> +			/* Make sure descriptor is read before updating the
> +			 * tail pointer.
> +			 */
> +			dma_mb();
> +			WRITE_ONCE(*srng->u.dst_ring.tp_addr, srng->u.dst_ring.tp);
>  		}
>  	} else {
>  		if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
> @@ -885,6 +888,10 @@ void ath11k_hal_srng_access_end(struct ath11k_base *ab, struct hal_srng *srng)
>  					   srng->u.src_ring.hp);
>  		} else {
>  			srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
> +			/* Make sure descriptor is read before updating the
> +			 * tail pointer.
> +			 */
> +			mb();
>  			ath11k_hif_write32(ab,
>  					   (unsigned long)srng->u.dst_ring.tp_addr -
>  					   (unsigned long)ab->mem,

Reviewed-by: Baochen Qiang <quic_bqiang@quicinc.com>

Re: [PATCH v2 0/5] wifi: ath11k: fix dest ring-buffer corruption

[Jeff Johnson]

On Wed, 04 Jun 2025 16:34:52 +0200, Johan Hovold wrote:
> As a follow up to commits:
> 
> 	6d037a372f81 ("wifi: ath11k: fix ring-buffer corruption")
> 	ab52e3e44fe9 ("wifi: ath11k: fix rx completion meta data corruption")
> 
> add the remaining missing memory barriers to make sure that destination
> ring descriptors are read after the head pointers to avoid using stale
> data on weakly ordered architectures like aarch64.
> 
> [...]

Applied, thanks!

[1/5] wifi: ath11k: fix dest ring-buffer corruption
      commit: 8c1ba5091fa9a2d1478da63173b16a701bdf86bb
[2/5] wifi: ath11k: use plain access for descriptor length
      commit: 3a690e9091ec07f70cedd9a7dcc074c8554f1bed
[3/5] wifi: ath11k: use plain accesses for monitor descriptor
      commit: 4aba95fb1faed7fe7f6e1edfb60c333e0275dfc1
[4/5] wifi: ath11k: fix source ring-buffer corruption
      commit: 6efa0df54022c6c9fd4d294b87622c7fcdc418c8
[5/5] wifi: ath11k: fix dest ring-buffer corruption when ring is full
      commit: aa6956150f820e6a6deba44be325ddfcb5b10f88

Best regards,
-- 
Jeff Johnson <jeff.johnson@oss.qualcomm.com>