net/mlx5e: Use synchronize_rcu to sync with NAPI

As described in the previous commit, napi_synchronize doesn't quite fit
the purpose when we just need to wait until the currently running NAPI
quits. Its implementation waits until NAPI is not running by polling and
waiting for 1ms in between. In cases where we need to deactivate one
queue (e.g., recovery flows) or where we deactivate them one-by-one
(deactivate channel flow), we may get stuck in napi_synchronize forever
if other queues keep NAPI active, causing a soft lockup. Depending on
kernel configuration (CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC), it may result
in a kernel panic.

To fix the issue, use synchronize_rcu to wait for NAPI to quit, and wrap
the whole NAPI in rcu_read_lock.

Fixes: acc6c5953af1 ("net/mlx5e: Split open/close channels to stages")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

2 files changed, 3 insertions, 14 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
index a33a1f762c70..40db27bf790b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
@@ -31,7 +31,6 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 {
 	struct xdp_buff *xdp = wi->umr.dma_info[page_idx].xsk;
 	u32 cqe_bcnt32 = cqe_bcnt;
-	bool consumed;
 
 	/* Check packet size. Note LRO doesn't use linear SKB */
 	if (unlikely(cqe_bcnt > rq->hw_mtu)) {
@@ -51,10 +50,6 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 	xsk_buff_dma_sync_for_cpu(xdp);
 	prefetch(xdp->data);
 
-	rcu_read_lock();
-	consumed = mlx5e_xdp_handle(rq, NULL, &cqe_bcnt32, xdp);
-	rcu_read_unlock();
-
 	/* Possible flows:
 	 * - XDP_REDIRECT to XSKMAP:
 	 *   The page is owned by the userspace from now.
@@ -70,7 +65,7 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 	 * allocated first from the Reuse Ring, so it has enough space.
 	 */
 
-	if (likely(consumed)) {
+	if (likely(mlx5e_xdp_handle(rq, NULL, &cqe_bcnt32, xdp))) {
 		if (likely(__test_and_clear_bit(MLX5E_RQ_FLAG_XDP_XMIT, rq->flags)))
 			__set_bit(page_idx, wi->xdp_xmit_bitmap); /* non-atomic */
 		return NULL; /* page/packet was consumed by XDP */
@@ -88,7 +83,6 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 					      u32 cqe_bcnt)
 {
 	struct xdp_buff *xdp = wi->di->xsk;
-	bool consumed;
 
 	/* wi->offset is not used in this function, because xdp->data and the
 	 * DMA address point directly to the necessary place. Furthermore, the
@@ -107,11 +101,7 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 		return NULL;
 	}
 
-	rcu_read_lock();
-	consumed = mlx5e_xdp_handle(rq, NULL, &cqe_bcnt, xdp);
-	rcu_read_unlock();
-
-	if (likely(consumed))
+	if (likely(mlx5e_xdp_handle(rq, NULL, &cqe_bcnt, xdp)))
 		return NULL; /* page/packet was consumed by XDP */
 
 	/* XDP_PASS: copy the data from the UMEM to a new SKB. The frame reuse
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/setup.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/setup.c
index dd9df519d383..55e65a438de7 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/setup.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/setup.c
@@ -106,8 +106,7 @@ err_free_cparam:
 void mlx5e_close_xsk(struct mlx5e_channel *c)
 {
 	clear_bit(MLX5E_CHANNEL_STATE_XSK, c->state);
-	napi_synchronize(&c->napi);
-	synchronize_rcu(); /* Sync with the XSK wakeup. */
+	synchronize_rcu(); /* Sync with the XSK wakeup and with NAPI. */
 
 	mlx5e_close_rq(&c->xskrq);
 	mlx5e_close_cq(&c->xskrq.cq);