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@xxxxxxxxxxxxxxx # 5.6
> Signed-off-by: Johan Hovold <johan+linaro@xxxxxxxxxx>
> ---
> 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)
