net/mlx5e: Add XSK zero-copy support

This commit adds support for AF_XDP zero-copy RX and TX.

We create a dedicated XSK RQ inside the channel, it means that two
RQs are running simultaneously: one for non-XSK traffic and the other
for XSK traffic. The regular and XSK RQs use a single ID namespace split
into two halves: the lower half is regular RQs, and the upper half is
XSK RQs. When any zero-copy AF_XDP socket is active, changing the number
of channels is not allowed, because it would break to mapping between
XSK RQ IDs and channels.

XSK requires different page allocation and release routines. Such
functions as mlx5e_{alloc,free}_rx_mpwqe and mlx5e_{get,put}_rx_frag are
generic enough to be used for both regular and XSK RQs, and they use the
mlx5e_page_{alloc,release} wrappers around the real allocation
functions. Function pointers are not used to avoid losing the
performance with retpolines. Wherever it's certain that the regular
(non-XSK) page release function should be used, it's called directly.

Only the stats that could be meaningful for XSK are exposed to the
userspace. Those that don't take part in the XSK flow are not
considered.

Note that we don't wait for WQEs on the XSK RQ (unlike the regular RQ),
because the newer xdpsock sample doesn't provide any Fill Ring entries
at the setup stage.

We create a dedicated XSK SQ in the channel. This separation has its
advantages:

1. When the UMEM is closed, the XSK SQ can also be closed and stop
receiving completions. If an existing SQ was used for XSK, it would
continue receiving completions for the packets of the closed socket. If
a new UMEM was opened at that point, it would start getting completions
that don't belong to it.

2. Calculating statistics separately.

When the userspace kicks the TX, the driver triggers a hardware
interrupt by posting a NOP to a dedicated XSK ICO (internal control
operations) SQ, in order to trigger NAPI on the right CPU core. This XSK
ICO SQ is protected by a spinlock, as the userspace application may kick
the TX from any core.

Store the pointers to the UMEMs in the net device private context,
independently from the kernel. This way the driver can distinguish
between the zero-copy and non-zero-copy UMEMs. The kernel function
xdp_get_umem_from_qid does not care about this difference, but the
driver is only interested in zero-copy UMEMs, particularly, on the
cleanup it determines whether to close the XSK RQ and SQ or not by
looking at the presence of the UMEM. Use state_lock to protect the
access to this area of UMEM pointers.

LRO isn't compatible with XDP, but there may be active UMEMs while
XDP is off. If this is the case, don't allow LRO to ensure XDP can
be reenabled at any time.

The validation of XSK parameters typically happens when XSK queues
open. However, when the interface is down or the XDP program isn't
set, it's still possible to have active AF_XDP sockets and even to
open new, but the XSK queues will be closed. To cover these cases,
perform the validation also in these flows:

1. A new UMEM is registered, but the XSK queues aren't going to be
created due to missing XDP program or interface being down.

2. MTU changes while there are UMEMs registered.

Having this early check prevents mlx5e_open_channels from failing
at a later stage, where recovery is impossible and the application
has no chance to handle the error, because it got the successful
return value for an MTU change or XSK open operation.

The performance testing was performed on a machine with the following
configuration:

- 24 cores of Intel Xeon E5-2620 v3 @ 2.40 GHz
- Mellanox ConnectX-5 Ex with 100 Gbit/s link

The results with retpoline disabled, single stream:

txonly: 33.3 Mpps (21.5 Mpps with queue and app pinned to the same CPU)
rxdrop: 12.2 Mpps
l2fwd: 9.4 Mpps

The results with retpoline enabled, single stream:

txonly: 21.3 Mpps (14.1 Mpps with queue and app pinned to the same CPU)
rxdrop: 9.9 Mpps
l2fwd: 6.8 Mpps

Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

1 files changed, 369 insertions, 158 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index f6c603572bd6..67b562c7f8a1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -38,6 +38,7 @@
 #include <linux/bpf.h>
 #include <linux/if_bridge.h>
 #include <net/page_pool.h>
+#include <net/xdp_sock.h>
 #include "eswitch.h"
 #include "en.h"
 #include "en_tc.h"
@@ -56,6 +57,10 @@
 #include "en/monitor_stats.h"
 #include "en/reporter.h"
 #include "en/params.h"
+#include "en/xsk/umem.h"
+#include "en/xsk/setup.h"
+#include "en/xsk/rx.h"
+#include "en/xsk/tx.h"
 
 bool mlx5e_check_fragmented_striding_rq_cap(struct mlx5_core_dev *mdev)
 {
@@ -85,18 +90,31 @@ void mlx5e_init_rq_type_params(struct mlx5_core_dev *mdev,
 	mlx5_core_info(mdev, "MLX5E: StrdRq(%d) RqSz(%ld) StrdSz(%ld) RxCqeCmprss(%d)\n",
 		       params->rq_wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ,
 		       params->rq_wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ ?
-		       BIT(mlx5e_mpwqe_get_log_rq_size(params)) :
+		       BIT(mlx5e_mpwqe_get_log_rq_size(params, NULL)) :
 		       BIT(params->log_rq_mtu_frames),
-		       BIT(mlx5e_mpwqe_get_log_stride_size(mdev, params)),
+		       BIT(mlx5e_mpwqe_get_log_stride_size(mdev, params, NULL)),
 		       MLX5E_GET_PFLAG(params, MLX5E_PFLAG_RX_CQE_COMPRESS));
 }
 
 bool mlx5e_striding_rq_possible(struct mlx5_core_dev *mdev,
 				struct mlx5e_params *params)
 {
-	return mlx5e_check_fragmented_striding_rq_cap(mdev) &&
-		!MLX5_IPSEC_DEV(mdev) &&
-		!(params->xdp_prog && !mlx5e_rx_mpwqe_is_linear_skb(mdev, params));
+	if (!mlx5e_check_fragmented_striding_rq_cap(mdev))
+		return false;
+
+	if (MLX5_IPSEC_DEV(mdev))
+		return false;
+
+	if (params->xdp_prog) {
+		/* XSK params are not considered here. If striding RQ is in use,
+		 * and an XSK is being opened, mlx5e_rx_mpwqe_is_linear_skb will
+		 * be called with the known XSK params.
+		 */
+		if (!mlx5e_rx_mpwqe_is_linear_skb(mdev, params, NULL))
+			return false;
+	}
+
+	return true;
 }
 
 void mlx5e_set_rq_type(struct mlx5_core_dev *mdev, struct mlx5e_params *params)
@@ -365,6 +383,8 @@ static void mlx5e_free_di_list(struct mlx5e_rq *rq)
 
 static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 			  struct mlx5e_params *params,
+			  struct mlx5e_xsk_param *xsk,
+			  struct xdp_umem *umem,
 			  struct mlx5e_rq_param *rqp,
 			  struct mlx5e_rq *rq)
 {
@@ -372,6 +392,8 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 	struct mlx5_core_dev *mdev = c->mdev;
 	void *rqc = rqp->rqc;
 	void *rqc_wq = MLX5_ADDR_OF(rqc, rqc, wq);
+	u32 num_xsk_frames = 0;
+	u32 rq_xdp_ix;
 	u32 pool_size;
 	int wq_sz;
 	int err;
@@ -388,8 +410,13 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 	rq->ix      = c->ix;
 	rq->mdev    = mdev;
 	rq->hw_mtu  = MLX5E_SW2HW_MTU(params, params->sw_mtu);
-	rq->stats   = &c->priv->channel_stats[c->ix].rq;
 	rq->xdpsq   = &c->rq_xdpsq;
+	rq->umem    = umem;
+
+	if (rq->umem)
+		rq->stats = &c->priv->channel_stats[c->ix].xskrq;
+	else
+		rq->stats = &c->priv->channel_stats[c->ix].rq;
 
 	rq->xdp_prog = params->xdp_prog ? bpf_prog_inc(params->xdp_prog) : NULL;
 	if (IS_ERR(rq->xdp_prog)) {
@@ -398,12 +425,16 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 		goto err_rq_wq_destroy;
 	}
 
-	err = xdp_rxq_info_reg(&rq->xdp_rxq, rq->netdev, rq->ix);
+	rq_xdp_ix = rq->ix;
+	if (xsk)
+		rq_xdp_ix += params->num_channels * MLX5E_RQ_GROUP_XSK;
+	err = xdp_rxq_info_reg(&rq->xdp_rxq, rq->netdev, rq_xdp_ix);
 	if (err < 0)
 		goto err_rq_wq_destroy;
 
 	rq->buff.map_dir = rq->xdp_prog ? DMA_BIDIRECTIONAL : DMA_FROM_DEVICE;
-	rq->buff.headroom = mlx5e_get_rq_headroom(mdev, params);
+	rq->buff.headroom = mlx5e_get_rq_headroom(mdev, params, xsk);
+	rq->buff.umem_headroom = xsk ? xsk->headroom : 0;
 	pool_size = 1 << params->log_rq_mtu_frames;
 
 	switch (rq->wq_type) {
@@ -417,7 +448,12 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 
 		wq_sz = mlx5_wq_ll_get_size(&rq->mpwqe.wq);
 
-		pool_size = MLX5_MPWRQ_PAGES_PER_WQE << mlx5e_mpwqe_get_log_rq_size(params);
+		if (xsk)
+			num_xsk_frames = wq_sz <<
+				mlx5e_mpwqe_get_log_num_strides(mdev, params, xsk);
+
+		pool_size = MLX5_MPWRQ_PAGES_PER_WQE <<
+			mlx5e_mpwqe_get_log_rq_size(params, xsk);
 
 		rq->post_wqes = mlx5e_post_rx_mpwqes;
 		rq->dealloc_wqe = mlx5e_dealloc_rx_mpwqe;
@@ -436,12 +472,15 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 			goto err_rq_wq_destroy;
 		}
 
-		rq->mpwqe.skb_from_cqe_mpwrq =
-			mlx5e_rx_mpwqe_is_linear_skb(mdev, params) ?
-			mlx5e_skb_from_cqe_mpwrq_linear :
-			mlx5e_skb_from_cqe_mpwrq_nonlinear;
-		rq->mpwqe.log_stride_sz = mlx5e_mpwqe_get_log_stride_size(mdev, params);
-		rq->mpwqe.num_strides = BIT(mlx5e_mpwqe_get_log_num_strides(mdev, params));
+		rq->mpwqe.skb_from_cqe_mpwrq = xsk ?
+			mlx5e_xsk_skb_from_cqe_mpwrq_linear :
+			mlx5e_rx_mpwqe_is_linear_skb(mdev, params, NULL) ?
+				mlx5e_skb_from_cqe_mpwrq_linear :
+				mlx5e_skb_from_cqe_mpwrq_nonlinear;
+
+		rq->mpwqe.log_stride_sz = mlx5e_mpwqe_get_log_stride_size(mdev, params, xsk);
+		rq->mpwqe.num_strides =
+			BIT(mlx5e_mpwqe_get_log_num_strides(mdev, params, xsk));
 
 		err = mlx5e_create_rq_umr_mkey(mdev, rq);
 		if (err)
@@ -462,6 +501,9 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 
 		wq_sz = mlx5_wq_cyc_get_size(&rq->wqe.wq);
 
+		if (xsk)
+			num_xsk_frames = wq_sz << rq->wqe.info.log_num_frags;
+
 		rq->wqe.info = rqp->frags_info;
 		rq->wqe.frags =
 			kvzalloc_node(array_size(sizeof(*rq->wqe.frags),
@@ -475,6 +517,7 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 		err = mlx5e_init_di_list(rq, wq_sz, c->cpu);
 		if (err)
 			goto err_free;
+
 		rq->post_wqes = mlx5e_post_rx_wqes;
 		rq->dealloc_wqe = mlx5e_dealloc_rx_wqe;
 
@@ -490,37 +533,53 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 			goto err_free;
 		}
 
-		rq->wqe.skb_from_cqe = mlx5e_rx_is_linear_skb(params) ?
-			mlx5e_skb_from_cqe_linear :
-			mlx5e_skb_from_cqe_nonlinear;
+		rq->wqe.skb_from_cqe = xsk ?
+			mlx5e_xsk_skb_from_cqe_linear :
+			mlx5e_rx_is_linear_skb(params, NULL) ?
+				mlx5e_skb_from_cqe_linear :
+				mlx5e_skb_from_cqe_nonlinear;
 		rq->mkey_be = c->mkey_be;
 	}
 
-	/* Create a page_pool and register it with rxq */
-	pp_params.order     = 0;
-	pp_params.flags     = 0; /* No-internal DMA mapping in page_pool */
-	pp_params.pool_size = pool_size;
-	pp_params.nid       = cpu_to_node(c->cpu);
-	pp_params.dev       = c->pdev;
-	pp_params.dma_dir   = rq->buff.map_dir;
-
-	/* page_pool can be used even when there is no rq->xdp_prog,
-	 * given page_pool does not handle DMA mapping there is no
-	 * required state to clear. And page_pool gracefully handle
-	 * elevated refcnt.
-	 */
-	rq->page_pool = page_pool_create(&pp_params);
-	if (IS_ERR(rq->page_pool)) {
-		err = PTR_ERR(rq->page_pool);
-		rq->page_pool = NULL;
-		goto err_free;
+	if (xsk) {
+		err = mlx5e_xsk_resize_reuseq(umem, num_xsk_frames);
+		if (unlikely(err)) {
+			mlx5_core_err(mdev, "Unable to allocate the Reuse Ring for %u frames\n",
+				      num_xsk_frames);
+			goto err_free;
+		}
+
+		rq->zca.free = mlx5e_xsk_zca_free;
+		err = xdp_rxq_info_reg_mem_model(&rq->xdp_rxq,
+						 MEM_TYPE_ZERO_COPY,
+						 &rq->zca);
+	} else {
+		/* Create a page_pool and register it with rxq */
+		pp_params.order     = 0;
+		pp_params.flags     = 0; /* No-internal DMA mapping in page_pool */
+		pp_params.pool_size = pool_size;
+		pp_params.nid       = cpu_to_node(c->cpu);
+		pp_params.dev       = c->pdev;
+		pp_params.dma_dir   = rq->buff.map_dir;
+
+		/* page_pool can be used even when there is no rq->xdp_prog,
+		 * given page_pool does not handle DMA mapping there is no
+		 * required state to clear. And page_pool gracefully handle
+		 * elevated refcnt.
+		 */
+		rq->page_pool = page_pool_create(&pp_params);
+		if (IS_ERR(rq->page_pool)) {
+			err = PTR_ERR(rq->page_pool);
+			rq->page_pool = NULL;
+			goto err_free;
+		}
+		err = xdp_rxq_info_reg_mem_model(&rq->xdp_rxq,
+						 MEM_TYPE_PAGE_POOL, rq->page_pool);
+		if (err)
+			page_pool_free(rq->page_pool);
 	}
-	err = xdp_rxq_info_reg_mem_model(&rq->xdp_rxq,
-					 MEM_TYPE_PAGE_POOL, rq->page_pool);
-	if (err) {
-		page_pool_free(rq->page_pool);
+	if (err)
 		goto err_free;
-	}
 
 	for (i = 0; i < wq_sz; i++) {
 		if (rq->wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ) {
@@ -611,7 +670,11 @@ static void mlx5e_free_rq(struct mlx5e_rq *rq)
 	     i = (i + 1) & (MLX5E_CACHE_SIZE - 1)) {
 		struct mlx5e_dma_info *dma_info = &rq->page_cache.page_cache[i];
 
-		mlx5e_page_release(rq, dma_info, false);
+		/* With AF_XDP, page_cache is not used, so this loop is not
+		 * entered, and it's safe to call mlx5e_page_release_dynamic
+		 * directly.
+		 */
+		mlx5e_page_release_dynamic(rq, dma_info, false);
 	}
 
 	xdp_rxq_info_unreg(&rq->xdp_rxq);
@@ -748,7 +811,7 @@ static void mlx5e_destroy_rq(struct mlx5e_rq *rq)
 	mlx5_core_destroy_rq(rq->mdev, rq->rqn);
 }
 
-static int mlx5e_wait_for_min_rx_wqes(struct mlx5e_rq *rq, int wait_time)
+int mlx5e_wait_for_min_rx_wqes(struct mlx5e_rq *rq, int wait_time)
 {
 	unsigned long exp_time = jiffies + msecs_to_jiffies(wait_time);
 	struct mlx5e_channel *c = rq->channel;
@@ -806,14 +869,13 @@ static void mlx5e_free_rx_descs(struct mlx5e_rq *rq)
 
 }
 
-static int mlx5e_open_rq(struct mlx5e_channel *c,
-			 struct mlx5e_params *params,
-			 struct mlx5e_rq_param *param,
-			 struct mlx5e_rq *rq)
+int mlx5e_open_rq(struct mlx5e_channel *c, struct mlx5e_params *params,
+		  struct mlx5e_rq_param *param, struct mlx5e_xsk_param *xsk,
+		  struct xdp_umem *umem, struct mlx5e_rq *rq)
 {
 	int err;
 
-	err = mlx5e_alloc_rq(c, params, param, rq);
+	err = mlx5e_alloc_rq(c, params, xsk, umem, param, rq);
 	if (err)
 		return err;
 
@@ -851,13 +913,13 @@ static void mlx5e_activate_rq(struct mlx5e_rq *rq)
 	mlx5e_trigger_irq(&rq->channel->icosq);
 }
 
-static void mlx5e_deactivate_rq(struct mlx5e_rq *rq)
+void mlx5e_deactivate_rq(struct mlx5e_rq *rq)
 {
 	clear_bit(MLX5E_RQ_STATE_ENABLED, &rq->state);
 	napi_synchronize(&rq->channel->napi); /* prevent mlx5e_post_rx_wqes */
 }
 
-static void mlx5e_close_rq(struct mlx5e_rq *rq)
+void mlx5e_close_rq(struct mlx5e_rq *rq)
 {
 	cancel_work_sync(&rq->dim.work);
 	mlx5e_destroy_rq(rq);
@@ -910,6 +972,7 @@ static int mlx5e_alloc_xdpsq_db(struct mlx5e_xdpsq *sq, int numa)
 
 static int mlx5e_alloc_xdpsq(struct mlx5e_channel *c,
 			     struct mlx5e_params *params,
+			     struct xdp_umem *umem,
 			     struct mlx5e_sq_param *param,
 			     struct mlx5e_xdpsq *sq,
 			     bool is_redirect)
@@ -925,9 +988,13 @@ static int mlx5e_alloc_xdpsq(struct mlx5e_channel *c,
 	sq->uar_map   = mdev->mlx5e_res.bfreg.map;
 	sq->min_inline_mode = params->tx_min_inline_mode;
 	sq->hw_mtu    = MLX5E_SW2HW_MTU(params, params->sw_mtu);
-	sq->stats     = is_redirect ?
-		&c->priv->channel_stats[c->ix].xdpsq :
-		&c->priv->channel_stats[c->ix].rq_xdpsq;
+	sq->umem      = umem;
+
+	sq->stats = sq->umem ?
+		&c->priv->channel_stats[c->ix].xsksq :
+		is_redirect ?
+			&c->priv->channel_stats[c->ix].xdpsq :
+			&c->priv->channel_stats[c->ix].rq_xdpsq;
 
 	param->wq.db_numa_node = cpu_to_node(c->cpu);
 	err = mlx5_wq_cyc_create(mdev, &param->wq, sqc_wq, wq, &sq->wq_ctrl);
@@ -1307,10 +1374,8 @@ static void mlx5e_tx_err_cqe_work(struct work_struct *recover_work)
 	mlx5e_tx_reporter_err_cqe(sq);
 }
 
-static int mlx5e_open_icosq(struct mlx5e_channel *c,
-			    struct mlx5e_params *params,
-			    struct mlx5e_sq_param *param,
-			    struct mlx5e_icosq *sq)
+int mlx5e_open_icosq(struct mlx5e_channel *c, struct mlx5e_params *params,
+		     struct mlx5e_sq_param *param, struct mlx5e_icosq *sq)
 {
 	struct mlx5e_create_sq_param csp = {};
 	int err;
@@ -1336,7 +1401,7 @@ err_free_icosq:
 	return err;
 }
 
-static void mlx5e_close_icosq(struct mlx5e_icosq *sq)
+void mlx5e_close_icosq(struct mlx5e_icosq *sq)
 {
 	struct mlx5e_channel *c = sq->channel;
 
@@ -1347,16 +1412,14 @@ static void mlx5e_close_icosq(struct mlx5e_icosq *sq)
 	mlx5e_free_icosq(sq);
 }
 
-static int mlx5e_open_xdpsq(struct mlx5e_channel *c,
-			    struct mlx5e_params *params,
-			    struct mlx5e_sq_param *param,
-			    struct mlx5e_xdpsq *sq,
-			    bool is_redirect)
+int mlx5e_open_xdpsq(struct mlx5e_channel *c, struct mlx5e_params *params,
+		     struct mlx5e_sq_param *param, struct xdp_umem *umem,
+		     struct mlx5e_xdpsq *sq, bool is_redirect)
 {
 	struct mlx5e_create_sq_param csp = {};
 	int err;
 
-	err = mlx5e_alloc_xdpsq(c, params, param, sq, is_redirect);
+	err = mlx5e_alloc_xdpsq(c, params, umem, param, sq, is_redirect);
 	if (err)
 		return err;
 
@@ -1410,7 +1473,7 @@ err_free_xdpsq:
 	return err;
 }
 
-static void mlx5e_close_xdpsq(struct mlx5e_xdpsq *sq)
+void mlx5e_close_xdpsq(struct mlx5e_xdpsq *sq)
 {
 	struct mlx5e_channel *c = sq->channel;
 
@@ -1539,10 +1602,8 @@ static void mlx5e_destroy_cq(struct mlx5e_cq *cq)
 	mlx5_core_destroy_cq(cq->mdev, &cq->mcq);
 }
 
-static int mlx5e_open_cq(struct mlx5e_channel *c,
-			 struct net_dim_cq_moder moder,
-			 struct mlx5e_cq_param *param,
-			 struct mlx5e_cq *cq)
+int mlx5e_open_cq(struct mlx5e_channel *c, struct net_dim_cq_moder moder,
+		  struct mlx5e_cq_param *param, struct mlx5e_cq *cq)
 {
 	struct mlx5_core_dev *mdev = c->mdev;
 	int err;
@@ -1565,7 +1626,7 @@ err_free_cq:
 	return err;
 }
 
-static void mlx5e_close_cq(struct mlx5e_cq *cq)
+void mlx5e_close_cq(struct mlx5e_cq *cq)
 {
 	mlx5e_destroy_cq(cq);
 	mlx5e_free_cq(cq);
@@ -1779,17 +1840,17 @@ static int mlx5e_open_queues(struct mlx5e_channel *c,
 		goto err_close_icosq;
 
 	if (c->xdp) {
-		err = mlx5e_open_xdpsq(c, params, &cparam->xdp_sq,
+		err = mlx5e_open_xdpsq(c, params, &cparam->xdp_sq, NULL,
 				       &c->rq_xdpsq, false);
 		if (err)
 			goto err_close_sqs;
 	}
 
-	err = mlx5e_open_rq(c, params, &cparam->rq, &c->rq);
+	err = mlx5e_open_rq(c, params, &cparam->rq, NULL, NULL, &c->rq);
 	if (err)
 		goto err_close_xdp_sq;
 
-	err = mlx5e_open_xdpsq(c, params, &cparam->xdp_sq, &c->xdpsq, true);
+	err = mlx5e_open_xdpsq(c, params, &cparam->xdp_sq, NULL, &c->xdpsq, true);
 	if (err)
 		goto err_close_rq;
 
@@ -1849,10 +1910,12 @@ static void mlx5e_close_queues(struct mlx5e_channel *c)
 static int mlx5e_open_channel(struct mlx5e_priv *priv, int ix,
 			      struct mlx5e_params *params,
 			      struct mlx5e_channel_param *cparam,
+			      struct xdp_umem *umem,
 			      struct mlx5e_channel **cp)
 {
 	int cpu = cpumask_first(mlx5_comp_irq_get_affinity_mask(priv->mdev, ix));
 	struct net_device *netdev = priv->netdev;
+	struct mlx5e_xsk_param xsk;
 	struct mlx5e_channel *c;
 	unsigned int irq;
 	int err;
@@ -1889,10 +1952,20 @@ static int mlx5e_open_channel(struct mlx5e_priv *priv, int ix,
 	if (unlikely(err))
 		goto err_napi_del;
 
+	if (umem) {
+		mlx5e_build_xsk_param(umem, &xsk);
+		err = mlx5e_open_xsk(priv, params, &xsk, umem, c);
+		if (unlikely(err))
+			goto err_close_queues;
+	}
+
 	*cp = c;
 
 	return 0;
 
+err_close_queues:
+	mlx5e_close_queues(c);
+
 err_napi_del:
 	netif_napi_del(&c->napi);
 	mlx5e_free_xps_cpumask(c);
@@ -1911,12 +1984,18 @@ static void mlx5e_activate_channel(struct mlx5e_channel *c)
 		mlx5e_activate_txqsq(&c->sq[tc]);
 	mlx5e_activate_rq(&c->rq);
 	netif_set_xps_queue(c->netdev, c->xps_cpumask, c->ix);
+
+	if (test_bit(MLX5E_CHANNEL_STATE_XSK, c->state))
+		mlx5e_activate_xsk(c);
 }
 
 static void mlx5e_deactivate_channel(struct mlx5e_channel *c)
 {
 	int tc;
 
+	if (test_bit(MLX5E_CHANNEL_STATE_XSK, c->state))
+		mlx5e_deactivate_xsk(c);
+
 	mlx5e_deactivate_rq(&c->rq);
 	for (tc = 0; tc < c->num_tc; tc++)
 		mlx5e_deactivate_txqsq(&c->sq[tc]);
@@ -1924,6 +2003,8 @@ static void mlx5e_deactivate_channel(struct mlx5e_channel *c)
 
 static void mlx5e_close_channel(struct mlx5e_channel *c)
 {
+	if (test_bit(MLX5E_CHANNEL_STATE_XSK, c->state))
+		mlx5e_close_xsk(c);
 	mlx5e_close_queues(c);
 	netif_napi_del(&c->napi);
 	mlx5e_free_xps_cpumask(c);
@@ -1935,6 +2016,7 @@ static void mlx5e_close_channel(struct mlx5e_channel *c)
 
 static void mlx5e_build_rq_frags_info(struct mlx5_core_dev *mdev,
 				      struct mlx5e_params *params,
+				      struct mlx5e_xsk_param *xsk,
 				      struct mlx5e_rq_frags_info *info)
 {
 	u32 byte_count = MLX5E_SW2HW_MTU(params, params->sw_mtu);
@@ -1947,10 +2029,10 @@ static void mlx5e_build_rq_frags_info(struct mlx5_core_dev *mdev,
 		byte_count += MLX5E_METADATA_ETHER_LEN;
 #endif
 
-	if (mlx5e_rx_is_linear_skb(params)) {
+	if (mlx5e_rx_is_linear_skb(params, xsk)) {
 		int frag_stride;
 
-		frag_stride = mlx5e_rx_get_linear_frag_sz(params, NULL);
+		frag_stride = mlx5e_rx_get_linear_frag_sz(params, xsk);
 		frag_stride = roundup_pow_of_two(frag_stride);
 
 		info->arr[0].frag_size = byte_count;
@@ -2008,9 +2090,10 @@ static u8 mlx5e_get_rq_log_wq_sz(void *rqc)
 	return MLX5_GET(wq, wq, log_wq_sz);
 }
 
-static void mlx5e_build_rq_param(struct mlx5e_priv *priv,
-				 struct mlx5e_params *params,
-				 struct mlx5e_rq_param *param)
+void mlx5e_build_rq_param(struct mlx5e_priv *priv,
+			  struct mlx5e_params *params,
+			  struct mlx5e_xsk_param *xsk,
+			  struct mlx5e_rq_param *param)
 {
 	struct mlx5_core_dev *mdev = priv->mdev;
 	void *rqc = param->rqc;
@@ -2020,16 +2103,16 @@ static void mlx5e_build_rq_param(struct mlx5e_priv *priv,
 	switch (params->rq_wq_type) {
 	case MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ:
 		MLX5_SET(wq, wq, log_wqe_num_of_strides,
-			 mlx5e_mpwqe_get_log_num_strides(mdev, params) -
+			 mlx5e_mpwqe_get_log_num_strides(mdev, params, xsk) -
 			 MLX5_MPWQE_LOG_NUM_STRIDES_BASE);
 		MLX5_SET(wq, wq, log_wqe_stride_size,
-			 mlx5e_mpwqe_get_log_stride_size(mdev, params) -
+			 mlx5e_mpwqe_get_log_stride_size(mdev, params, xsk) -
 			 MLX5_MPWQE_LOG_STRIDE_SZ_BASE);
-		MLX5_SET(wq, wq, log_wq_sz, mlx5e_mpwqe_get_log_rq_size(params));
+		MLX5_SET(wq, wq, log_wq_sz, mlx5e_mpwqe_get_log_rq_size(params, xsk));
 		break;
 	default: /* MLX5_WQ_TYPE_CYCLIC */
 		MLX5_SET(wq, wq, log_wq_sz, params->log_rq_mtu_frames);
-		mlx5e_build_rq_frags_info(mdev, params, &param->frags_info);
+		mlx5e_build_rq_frags_info(mdev, params, xsk, &param->frags_info);
 		ndsegs = param->frags_info.num_frags;
 	}
 
@@ -2060,8 +2143,8 @@ static void mlx5e_build_drop_rq_param(struct mlx5e_priv *priv,
 	param->wq.buf_numa_node = dev_to_node(mdev->device);
 }
 
-static void mlx5e_build_sq_param_common(struct mlx5e_priv *priv,
-					struct mlx5e_sq_param *param)
+void mlx5e_build_sq_param_common(struct mlx5e_priv *priv,
+				 struct mlx5e_sq_param *param)
 {
 	void *sqc = param->sqc;
 	void *wq = MLX5_ADDR_OF(sqc, sqc, wq);
@@ -2097,9 +2180,10 @@ static void mlx5e_build_common_cq_param(struct mlx5e_priv *priv,
 		MLX5_SET(cqc, cqc, cqe_sz, CQE_STRIDE_128_PAD);
 }
 
-static void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv,
-				    struct mlx5e_params *params,
-				    struct mlx5e_cq_param *param)
+void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv,
+			     struct mlx5e_params *params,
+			     struct mlx5e_xsk_param *xsk,
+			     struct mlx5e_cq_param *param)
 {
 	struct mlx5_core_dev *mdev = priv->mdev;
 	void *cqc = param->cqc;
@@ -2107,8 +2191,8 @@ static void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv,
 
 	switch (params->rq_wq_type) {
 	case MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ:
-		log_cq_size = mlx5e_mpwqe_get_log_rq_size(params) +
-			mlx5e_mpwqe_get_log_num_strides(mdev, params);
+		log_cq_size = mlx5e_mpwqe_get_log_rq_size(params, xsk) +
+			mlx5e_mpwqe_get_log_num_strides(mdev, params, xsk);
 		break;
 	default: /* MLX5_WQ_TYPE_CYCLIC */
 		log_cq_size = params->log_rq_mtu_frames;
@@ -2124,9 +2208,9 @@ static void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv,
 	param->cq_period_mode = params->rx_cq_moderation.cq_period_mode;
 }
 
-static void mlx5e_build_tx_cq_param(struct mlx5e_priv *priv,
-				    struct mlx5e_params *params,
-				    struct mlx5e_cq_param *param)
+void mlx5e_build_tx_cq_param(struct mlx5e_priv *priv,
+			     struct mlx5e_params *params,
+			     struct mlx5e_cq_param *param)
 {
 	void *cqc = param->cqc;
 
@@ -2136,9 +2220,9 @@ static void mlx5e_build_tx_cq_param(struct mlx5e_priv *priv,
 	param->cq_period_mode = params->tx_cq_moderation.cq_period_mode;
 }
 
-static void mlx5e_build_ico_cq_param(struct mlx5e_priv *priv,
-				     u8 log_wq_size,
-				     struct mlx5e_cq_param *param)
+void mlx5e_build_ico_cq_param(struct mlx5e_priv *priv,
+			      u8 log_wq_size,
+			      struct mlx5e_cq_param *param)
 {
 	void *cqc = param->cqc;
 
@@ -2149,9 +2233,9 @@ static void mlx5e_build_ico_cq_param(struct mlx5e_priv *priv,
 	param->cq_period_mode = NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE;
 }
 
-static void mlx5e_build_icosq_param(struct mlx5e_priv *priv,
-				    u8 log_wq_size,
-				    struct mlx5e_sq_param *param)
+void mlx5e_build_icosq_param(struct mlx5e_priv *priv,
+			     u8 log_wq_size,
+			     struct mlx5e_sq_param *param)
 {
 	void *sqc = param->sqc;
 	void *wq = MLX5_ADDR_OF(sqc, sqc, wq);
@@ -2162,9 +2246,9 @@ static void mlx5e_build_icosq_param(struct mlx5e_priv *priv,
 	MLX5_SET(sqc, sqc, reg_umr, MLX5_CAP_ETH(priv->mdev, reg_umr_sq));
 }
 
-static void mlx5e_build_xdpsq_param(struct mlx5e_priv *priv,
-				    struct mlx5e_params *params,
-				    struct mlx5e_sq_param *param)
+void mlx5e_build_xdpsq_param(struct mlx5e_priv *priv,
+			     struct mlx5e_params *params,
+			     struct mlx5e_sq_param *param)
 {
 	void *sqc = param->sqc;
 	void *wq = MLX5_ADDR_OF(sqc, sqc, wq);
@@ -2192,14 +2276,14 @@ static void mlx5e_build_channel_param(struct mlx5e_priv *priv,
 {
 	u8 icosq_log_wq_sz;
 
-	mlx5e_build_rq_param(priv, params, &cparam->rq);
+	mlx5e_build_rq_param(priv, params, NULL, &cparam->rq);
 
 	icosq_log_wq_sz = mlx5e_build_icosq_log_wq_sz(params, &cparam->rq);
 
 	mlx5e_build_sq_param(priv, params, &cparam->sq);
 	mlx5e_build_xdpsq_param(priv, params, &cparam->xdp_sq);
 	mlx5e_build_icosq_param(priv, icosq_log_wq_sz, &cparam->icosq);
-	mlx5e_build_rx_cq_param(priv, params, &cparam->rx_cq);
+	mlx5e_build_rx_cq_param(priv, params, NULL, &cparam->rx_cq);
 	mlx5e_build_tx_cq_param(priv, params, &cparam->tx_cq);
 	mlx5e_build_ico_cq_param(priv, icosq_log_wq_sz, &cparam->icosq_cq);
 }
@@ -2220,7 +2304,12 @@ int mlx5e_open_channels(struct mlx5e_priv *priv,
 
 	mlx5e_build_channel_param(priv, &chs->params, cparam);
 	for (i = 0; i < chs->num; i++) {
-		err = mlx5e_open_channel(priv, i, &chs->params, cparam, &chs->c[i]);
+		struct xdp_umem *umem = NULL;
+
+		if (chs->params.xdp_prog)
+			umem = mlx5e_xsk_get_umem(&chs->params, chs->params.xsk, i);
+
+		err = mlx5e_open_channel(priv, i, &chs->params, cparam, umem, &chs->c[i]);
 		if (err)
 			goto err_close_channels;
 	}
@@ -2262,6 +2351,10 @@ static int mlx5e_wait_channels_min_rx_wqes(struct mlx5e_channels *chs)
 		int timeout = err ? 0 : MLX5E_RQ_WQES_TIMEOUT;
 
 		err |= mlx5e_wait_for_min_rx_wqes(&chs->c[i]->rq, timeout);
+
+		/* Don't wait on the XSK RQ, because the newer xdpsock sample
+		 * doesn't provide any Fill Ring entries at the setup stage.
+		 */
 	}
 
 	return err ? -ETIMEDOUT : 0;
@@ -2334,35 +2427,35 @@ int mlx5e_create_indirect_rqt(struct mlx5e_priv *priv)
 	return err;
 }
 
-int mlx5e_create_direct_rqts(struct mlx5e_priv *priv)
+int mlx5e_create_direct_rqts(struct mlx5e_priv *priv, struct mlx5e_tir *tirs)
 {
-	struct mlx5e_rqt *rqt;
+	const int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 	int err;
 	int ix;
 
-	for (ix = 0; ix < mlx5e_get_netdev_max_channels(priv->netdev); ix++) {
-		rqt = &priv->direct_tir[ix].rqt;
-		err = mlx5e_create_rqt(priv, 1 /*size */, rqt);
-		if (err)
+	for (ix = 0; ix < max_nch; ix++) {
+		err = mlx5e_create_rqt(priv, 1 /*size */, &tirs[ix].rqt);
+		if (unlikely(err))
 			goto err_destroy_rqts;
 	}
 
 	return 0;
 
 err_destroy_rqts:
-	mlx5_core_warn(priv->mdev, "create direct rqts failed, %d\n", err);
+	mlx5_core_warn(priv->mdev, "create rqts failed, %d\n", err);
 	for (ix--; ix >= 0; ix--)
-		mlx5e_destroy_rqt(priv, &priv->direct_tir[ix].rqt);
+		mlx5e_destroy_rqt(priv, &tirs[ix].rqt);
 
 	return err;
 }
 
-void mlx5e_destroy_direct_rqts(struct mlx5e_priv *priv)
+void mlx5e_destroy_direct_rqts(struct mlx5e_priv *priv, struct mlx5e_tir *tirs)
 {
+	const int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 	int i;
 
-	for (i = 0; i < mlx5e_get_netdev_max_channels(priv->netdev); i++)
-		mlx5e_destroy_rqt(priv, &priv->direct_tir[i].rqt);
+	for (i = 0; i < max_nch; i++)
+		mlx5e_destroy_rqt(priv, &tirs[i].rqt);
 }
 
 static int mlx5e_rx_hash_fn(int hfunc)
@@ -2782,11 +2875,12 @@ static void mlx5e_build_tx2sq_maps(struct mlx5e_priv *priv)
 void mlx5e_activate_priv_channels(struct mlx5e_priv *priv)
 {
 	int num_txqs = priv->channels.num * priv->channels.params.num_tc;
+	int num_rxqs = priv->channels.num * MLX5E_NUM_RQ_GROUPS;
 	struct net_device *netdev = priv->netdev;
 
 	mlx5e_netdev_set_tcs(netdev);
 	netif_set_real_num_tx_queues(netdev, num_txqs);
-	netif_set_real_num_rx_queues(netdev, priv->channels.num);
+	netif_set_real_num_rx_queues(netdev, num_rxqs);
 
 	mlx5e_build_tx2sq_maps(priv);
 	mlx5e_activate_channels(&priv->channels);
@@ -2798,10 +2892,14 @@ void mlx5e_activate_priv_channels(struct mlx5e_priv *priv)
 
 	mlx5e_wait_channels_min_rx_wqes(&priv->channels);
 	mlx5e_redirect_rqts_to_channels(priv, &priv->channels);
+
+	mlx5e_xsk_redirect_rqts_to_channels(priv, &priv->channels);
 }
 
 void mlx5e_deactivate_priv_channels(struct mlx5e_priv *priv)
 {
+	mlx5e_xsk_redirect_rqts_to_drop(priv, &priv->channels);
+
 	mlx5e_redirect_rqts_to_drop(priv);
 
 	if (mlx5e_is_vport_rep(priv))
@@ -2880,9 +2978,12 @@ void mlx5e_timestamp_init(struct mlx5e_priv *priv)
 int mlx5e_open_locked(struct net_device *netdev)
 {
 	struct mlx5e_priv *priv = netdev_priv(netdev);
+	bool is_xdp = priv->channels.params.xdp_prog;
 	int err;
 
 	set_bit(MLX5E_STATE_OPENED, &priv->state);
+	if (is_xdp)
+		mlx5e_xdp_set_open(priv);
 
 	err = mlx5e_open_channels(priv, &priv->channels);
 	if (err)
@@ -2897,6 +2998,8 @@ int mlx5e_open_locked(struct net_device *netdev)
 	return 0;
 
 err_clear_state_opened_flag:
+	if (is_xdp)
+		mlx5e_xdp_set_closed(priv);
 	clear_bit(MLX5E_STATE_OPENED, &priv->state);
 	return err;
 }
@@ -2928,6 +3031,8 @@ int mlx5e_close_locked(struct net_device *netdev)
 	if (!test_bit(MLX5E_STATE_OPENED, &priv->state))
 		return 0;
 
+	if (priv->channels.params.xdp_prog)
+		mlx5e_xdp_set_closed(priv);
 	clear_bit(MLX5E_STATE_OPENED, &priv->state);
 
 	netif_carrier_off(priv->netdev);
@@ -3184,13 +3289,13 @@ err_destroy_inner_tirs:
 	return err;
 }
 
-int mlx5e_create_direct_tirs(struct mlx5e_priv *priv)
+int mlx5e_create_direct_tirs(struct mlx5e_priv *priv, struct mlx5e_tir *tirs)
 {
-	int nch = mlx5e_get_netdev_max_channels(priv->netdev);
+	const int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 	struct mlx5e_tir *tir;
 	void *tirc;
 	int inlen;
-	int err;
+	int err = 0;
 	u32 *in;
 	int ix;
 
@@ -3199,25 +3304,24 @@ int mlx5e_create_direct_tirs(struct mlx5e_priv *priv)
 	if (!in)
 		return -ENOMEM;
 
-	for (ix = 0; ix < nch; ix++) {
+	for (ix = 0; ix < max_nch; ix++) {
 		memset(in, 0, inlen);
-		tir = &priv->direct_tir[ix];
+		tir = &tirs[ix];
 		tirc = MLX5_ADDR_OF(create_tir_in, in, ctx);
-		mlx5e_build_direct_tir_ctx(priv, priv->direct_tir[ix].rqt.rqtn, tirc);
+		mlx5e_build_direct_tir_ctx(priv, tir->rqt.rqtn, tirc);
 		err = mlx5e_create_tir(priv->mdev, tir, in, inlen);
-		if (err)
+		if (unlikely(err))
 			goto err_destroy_ch_tirs;
 	}
 
-	kvfree(in);
-
-	return 0;
+	goto out;
 
 err_destroy_ch_tirs:
-	mlx5_core_warn(priv->mdev, "create direct tirs failed, %d\n", err);
+	mlx5_core_warn(priv->mdev, "create tirs failed, %d\n", err);
 	for (ix--; ix >= 0; ix--)
-		mlx5e_destroy_tir(priv->mdev, &priv->direct_tir[ix]);
+		mlx5e_destroy_tir(priv->mdev, &tirs[ix]);
 
+out:
 	kvfree(in);
 
 	return err;
@@ -3237,13 +3341,13 @@ void mlx5e_destroy_indirect_tirs(struct mlx5e_priv *priv, bool inner_ttc)
 		mlx5e_destroy_tir(priv->mdev, &priv->inner_indir_tir[i]);
 }
 
-void mlx5e_destroy_direct_tirs(struct mlx5e_priv *priv)
+void mlx5e_destroy_direct_tirs(struct mlx5e_priv *priv, struct mlx5e_tir *tirs)
 {
-	int nch = mlx5e_get_netdev_max_channels(priv->netdev);
+	const int max_nch = mlx5e_get_netdev_max_channels(priv->netdev);
 	int i;
 
-	for (i = 0; i < nch; i++)
-		mlx5e_destroy_tir(priv->mdev, &priv->direct_tir[i]);
+	for (i = 0; i < max_nch; i++)
+		mlx5e_destroy_tir(priv->mdev, &tirs[i]);
 }
 
 static int mlx5e_modify_channels_scatter_fcs(struct mlx5e_channels *chs, bool enable)
@@ -3385,11 +3489,12 @@ void mlx5e_fold_sw_stats64(struct mlx5e_priv *priv, struct rtnl_link_stats64 *s)
 
 	for (i = 0; i < mlx5e_get_netdev_max_channels(priv->netdev); i++) {
 		struct mlx5e_channel_stats *channel_stats = &priv->channel_stats[i];
+		struct mlx5e_rq_stats *xskrq_stats = &channel_stats->xskrq;
 		struct mlx5e_rq_stats *rq_stats = &channel_stats->rq;
 		int j;
 
-		s->rx_packets   += rq_stats->packets;
-		s->rx_bytes     += rq_stats->bytes;
+		s->rx_packets   += rq_stats->packets + xskrq_stats->packets;
+		s->rx_bytes     += rq_stats->bytes + xskrq_stats->bytes;
 
 		for (j = 0; j < priv->max_opened_tc; j++) {
 			struct mlx5e_sq_stats *sq_stats = &channel_stats->sq[j];
@@ -3488,6 +3593,13 @@ static int set_feature_lro(struct net_device *netdev, bool enable)
 
 	mutex_lock(&priv->state_lock);
 
+	if (enable && priv->xsk.refcnt) {
+		netdev_warn(netdev, "LRO is incompatible with AF_XDP (%hu XSKs are active)\n",
+			    priv->xsk.refcnt);
+		err = -EINVAL;
+		goto out;
+	}
+
 	old_params = &priv->channels.params;
 	if (enable && !MLX5E_GET_PFLAG(old_params, MLX5E_PFLAG_RX_STRIDING_RQ)) {
 		netdev_warn(netdev, "can't set LRO with legacy RQ\n");
@@ -3501,8 +3613,8 @@ static int set_feature_lro(struct net_device *netdev, bool enable)
 	new_channels.params.lro_en = enable;
 
 	if (old_params->rq_wq_type != MLX5_WQ_TYPE_CYCLIC) {
-		if (mlx5e_rx_mpwqe_is_linear_skb(mdev, old_params) ==
-		    mlx5e_rx_mpwqe_is_linear_skb(mdev, &new_channels.params))
+		if (mlx5e_rx_mpwqe_is_linear_skb(mdev, old_params, NULL) ==
+		    mlx5e_rx_mpwqe_is_linear_skb(mdev, &new_channels.params, NULL))
 			reset = false;
 	}
 
@@ -3692,6 +3804,43 @@ static netdev_features_t mlx5e_fix_features(struct net_device *netdev,
 	return features;
 }
 
+static bool mlx5e_xsk_validate_mtu(struct net_device *netdev,
+				   struct mlx5e_channels *chs,
+				   struct mlx5e_params *new_params,
+				   struct mlx5_core_dev *mdev)
+{
+	u16 ix;
+
+	for (ix = 0; ix < chs->params.num_channels; ix++) {
+		struct xdp_umem *umem = mlx5e_xsk_get_umem(&chs->params, chs->params.xsk, ix);
+		struct mlx5e_xsk_param xsk;
+
+		if (!umem)
+			continue;
+
+		mlx5e_build_xsk_param(umem, &xsk);
+
+		if (!mlx5e_validate_xsk_param(new_params, &xsk, mdev)) {
+			u32 hr = mlx5e_get_linear_rq_headroom(new_params, &xsk);
+			int max_mtu_frame, max_mtu_page, max_mtu;
+
+			/* Two criteria must be met:
+			 * 1. HW MTU + all headrooms <= XSK frame size.
+			 * 2. Size of SKBs allocated on XDP_PASS <= PAGE_SIZE.
+			 */
+			max_mtu_frame = MLX5E_HW2SW_MTU(new_params, xsk.chunk_size - hr);
+			max_mtu_page = mlx5e_xdp_max_mtu(new_params, &xsk);
+			max_mtu = min(max_mtu_frame, max_mtu_page);
+
+			netdev_err(netdev, "MTU %d is too big for an XSK running on channel %hu. Try MTU <= %d\n",
+				   new_params->sw_mtu, ix, max_mtu);
+			return false;
+		}
+	}
+
+	return true;
+}
+
 int mlx5e_change_mtu(struct net_device *netdev, int new_mtu,
 		     change_hw_mtu_cb set_mtu_cb)
 {
@@ -3712,18 +3861,31 @@ int mlx5e_change_mtu(struct net_device *netdev, int new_mtu,
 	new_channels.params.sw_mtu = new_mtu;
 
 	if (params->xdp_prog &&
-	    !mlx5e_rx_is_linear_skb(&new_channels.params)) {
+	    !mlx5e_rx_is_linear_skb(&new_channels.params, NULL)) {
 		netdev_err(netdev, "MTU(%d) > %d is not allowed while XDP enabled\n",
 			   new_mtu, mlx5e_xdp_max_mtu(params, NULL));
 		err = -EINVAL;
 		goto out;
 	}
 
+	if (priv->xsk.refcnt &&
+	    !mlx5e_xsk_validate_mtu(netdev, &priv->channels,
+				    &new_channels.params, priv->mdev)) {
+		err = -EINVAL;
+		goto out;
+	}
+
 	if (params->rq_wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ) {
-		bool is_linear = mlx5e_rx_mpwqe_is_linear_skb(priv->mdev, &new_channels.params);
-		u8 ppw_old = mlx5e_mpwqe_log_pkts_per_wqe(params);
-		u8 ppw_new = mlx5e_mpwqe_log_pkts_per_wqe(&new_channels.params);
+		bool is_linear = mlx5e_rx_mpwqe_is_linear_skb(priv->mdev,
+							      &new_channels.params,
+							      NULL);
+		u8 ppw_old = mlx5e_mpwqe_log_pkts_per_wqe(params, NULL);
+		u8 ppw_new = mlx5e_mpwqe_log_pkts_per_wqe(&new_channels.params, NULL);
+
+		/* If XSK is active, XSK RQs are linear. */
+		is_linear |= priv->xsk.refcnt;
 
+		/* Always reset in linear mode - hw_mtu is used in data path. */
 		reset = reset && (is_linear || (ppw_old != ppw_new));
 	}
 
@@ -4156,7 +4318,10 @@ static int mlx5e_xdp_allowed(struct mlx5e_priv *priv, struct bpf_prog *prog)
 	new_channels.params = priv->channels.params;
 	new_channels.params.xdp_prog = prog;
 
-	if (!mlx5e_rx_is_linear_skb(&new_channels.params)) {
+	/* No XSK params: AF_XDP can't be enabled yet at the point of setting
+	 * the XDP program.
+	 */
+	if (!mlx5e_rx_is_linear_skb(&new_channels.params, NULL)) {
 		netdev_warn(netdev, "XDP is not allowed with MTU(%d) > %d\n",
 			    new_channels.params.sw_mtu,
 			    mlx5e_xdp_max_mtu(&new_channels.params, NULL));
@@ -4166,6 +4331,16 @@ static int mlx5e_xdp_allowed(struct mlx5e_priv *priv, struct bpf_prog *prog)
 	return 0;
 }
 
+static int mlx5e_xdp_update_state(struct mlx5e_priv *priv)
+{
+	if (priv->channels.params.xdp_prog)
+		mlx5e_xdp_set_open(priv);
+	else
+		mlx5e_xdp_set_closed(priv);
+
+	return 0;
+}
+
 static int mlx5e_xdp_set(struct net_device *netdev, struct bpf_prog *prog)
 {
 	struct mlx5e_priv *priv = netdev_priv(netdev);
@@ -4205,7 +4380,7 @@ static int mlx5e_xdp_set(struct net_device *netdev, struct bpf_prog *prog)
 		mlx5e_set_rq_type(priv->mdev, &new_channels.params);
 		old_prog = priv->channels.params.xdp_prog;
 
-		err = mlx5e_safe_switch_channels(priv, &new_channels, NULL);
+		err = mlx5e_safe_switch_channels(priv, &new_channels, mlx5e_xdp_update_state);
 		if (err)
 			goto unlock;
 	} else {
@@ -4229,19 +4404,29 @@ static int mlx5e_xdp_set(struct net_device *netdev, struct bpf_prog *prog)
 	 */
 	for (i = 0; i < priv->channels.num; i++) {
 		struct mlx5e_channel *c = priv->channels.c[i];
+		bool xsk_open = test_bit(MLX5E_CHANNEL_STATE_XSK, c->state);
 
 		clear_bit(MLX5E_RQ_STATE_ENABLED, &c->rq.state);
+		if (xsk_open)
+			clear_bit(MLX5E_RQ_STATE_ENABLED, &c->xskrq.state);
 		napi_synchronize(&c->napi);
 		/* prevent mlx5e_poll_rx_cq from accessing rq->xdp_prog */
 
 		old_prog = xchg(&c->rq.xdp_prog, prog);
+		if (old_prog)
+			bpf_prog_put(old_prog);
+
+		if (xsk_open) {
+			old_prog = xchg(&c->xskrq.xdp_prog, prog);
+			if (old_prog)
+				bpf_prog_put(old_prog);
+		}
 
 		set_bit(MLX5E_RQ_STATE_ENABLED, &c->rq.state);
+		if (xsk_open)
+			set_bit(MLX5E_RQ_STATE_ENABLED, &c->xskrq.state);
 		/* napi_schedule in case we have missed anything */
 		napi_schedule(&c->napi);
-
-		if (old_prog)
-			bpf_prog_put(old_prog);
 	}
 
 unlock:
@@ -4272,6 +4457,9 @@ static int mlx5e_xdp(struct net_device *dev, struct netdev_bpf *xdp)
 	case XDP_QUERY_PROG:
 		xdp->prog_id = mlx5e_xdp_query(dev);
 		return 0;
+	case XDP_SETUP_XSK_UMEM:
+		return mlx5e_xsk_setup_umem(dev, xdp->xsk.umem,
+					    xdp->xsk.queue_id);
 	default:
 		return -EINVAL;
 	}
@@ -4354,6 +4542,7 @@ const struct net_device_ops mlx5e_netdev_ops = {
 	.ndo_tx_timeout          = mlx5e_tx_timeout,
 	.ndo_bpf		 = mlx5e_xdp,
 	.ndo_xdp_xmit            = mlx5e_xdp_xmit,
+	.ndo_xsk_async_xmit      = mlx5e_xsk_async_xmit,
 #ifdef CONFIG_MLX5_EN_ARFS
 	.ndo_rx_flow_steer	 = mlx5e_rx_flow_steer,
 #endif
@@ -4505,11 +4694,13 @@ void mlx5e_build_rq_params(struct mlx5_core_dev *mdev,
 	 * - Striding RQ configuration is not possible/supported.
 	 * - Slow PCI heuristic.
 	 * - Legacy RQ would use linear SKB while Striding RQ would use non-linear.
+	 *
+	 * No XSK params: checking the availability of striding RQ in general.
 	 */
 	if (!slow_pci_heuristic(mdev) &&
 	    mlx5e_striding_rq_possible(mdev, params) &&
-	    (mlx5e_rx_mpwqe_is_linear_skb(mdev, params) ||
-	     !mlx5e_rx_is_linear_skb(params)))
+	    (mlx5e_rx_mpwqe_is_linear_skb(mdev, params, NULL) ||
+	     !mlx5e_rx_is_linear_skb(params, NULL)))
 		MLX5E_SET_PFLAG(params, MLX5E_PFLAG_RX_STRIDING_RQ, true);
 	mlx5e_set_rq_type(mdev, params);
 	mlx5e_init_rq_type_params(mdev, params);
@@ -4531,6 +4722,7 @@ void mlx5e_build_rss_params(struct mlx5e_rss_params *rss_params,
 }
 
 void mlx5e_build_nic_params(struct mlx5_core_dev *mdev,
+			    struct mlx5e_xsk *xsk,
 			    struct mlx5e_rss_params *rss_params,
 			    struct mlx5e_params *params,
 			    u16 max_channels, u16 mtu)
@@ -4566,9 +4758,11 @@ void mlx5e_build_nic_params(struct mlx5_core_dev *mdev,
 	/* HW LRO */
 
 	/* TODO: && MLX5_CAP_ETH(mdev, lro_cap) */
-	if (params->rq_wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ)
-		if (!mlx5e_rx_mpwqe_is_linear_skb(mdev, params))
+	if (params->rq_wq_type == MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ) {
+		/* No XSK params: checking the availability of striding RQ in general. */
+		if (!mlx5e_rx_mpwqe_is_linear_skb(mdev, params, NULL))
 			params->lro_en = !slow_pci_heuristic(mdev);
+	}
 	params->lro_timeout = mlx5e_choose_lro_timeout(mdev, MLX5E_DEFAULT_LRO_TIMEOUT);
 
 	/* CQ moderation params */
@@ -4587,6 +4781,9 @@ void mlx5e_build_nic_params(struct mlx5_core_dev *mdev,
 	mlx5e_build_rss_params(rss_params, params->num_channels);
 	params->tunneled_offload_en =
 		mlx5e_tunnel_inner_ft_supported(mdev);
+
+	/* AF_XDP */
+	params->xsk = xsk;
 }
 
 static void mlx5e_set_netdev_dev_addr(struct net_device *netdev)
@@ -4759,7 +4956,7 @@ static int mlx5e_nic_init(struct mlx5_core_dev *mdev,
 	if (err)
 		return err;
 
-	mlx5e_build_nic_params(mdev, rss, &priv->channels.params,
+	mlx5e_build_nic_params(mdev, &priv->xsk, rss, &priv->channels.params,
 			       mlx5e_get_netdev_max_channels(netdev),
 			       netdev->mtu);
 
@@ -4801,7 +4998,7 @@ static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
 	if (err)
 		goto err_close_drop_rq;
 
-	err = mlx5e_create_direct_rqts(priv);
+	err = mlx5e_create_direct_rqts(priv, priv->direct_tir);
 	if (err)
 		goto err_destroy_indirect_rqts;
 
@@ -4809,14 +5006,22 @@ static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
 	if (err)
 		goto err_destroy_direct_rqts;
 
-	err = mlx5e_create_direct_tirs(priv);
+	err = mlx5e_create_direct_tirs(priv, priv->direct_tir);
 	if (err)
 		goto err_destroy_indirect_tirs;
 
+	err = mlx5e_create_direct_rqts(priv, priv->xsk_tir);
+	if (unlikely(err))
+		goto err_destroy_direct_tirs;
+
+	err = mlx5e_create_direct_tirs(priv, priv->xsk_tir);
+	if (unlikely(err))
+		goto err_destroy_xsk_rqts;
+
 	err = mlx5e_create_flow_steering(priv);
 	if (err) {
 		mlx5_core_warn(mdev, "create flow steering failed, %d\n", err);
-		goto err_destroy_direct_tirs;
+		goto err_destroy_xsk_tirs;
 	}
 
 	err = mlx5e_tc_nic_init(priv);
@@ -4827,12 +5032,16 @@ static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
 
 err_destroy_flow_steering:
 	mlx5e_destroy_flow_steering(priv);
+err_destroy_xsk_tirs:
+	mlx5e_destroy_direct_tirs(priv, priv->xsk_tir);
+err_destroy_xsk_rqts:
+	mlx5e_destroy_direct_rqts(priv, priv->xsk_tir);
 err_destroy_direct_tirs:
-	mlx5e_destroy_direct_tirs(priv);
+	mlx5e_destroy_direct_tirs(priv, priv->direct_tir);
 err_destroy_indirect_tirs:
 	mlx5e_destroy_indirect_tirs(priv, true);
 err_destroy_direct_rqts:
-	mlx5e_destroy_direct_rqts(priv);
+	mlx5e_destroy_direct_rqts(priv, priv->direct_tir);
 err_destroy_indirect_rqts:
 	mlx5e_destroy_rqt(priv, &priv->indir_rqt);
 err_close_drop_rq:
@@ -4846,9 +5055,11 @@ static void mlx5e_cleanup_nic_rx(struct mlx5e_priv *priv)
 {
 	mlx5e_tc_nic_cleanup(priv);
 	mlx5e_destroy_flow_steering(priv);
-	mlx5e_destroy_direct_tirs(priv);
+	mlx5e_destroy_direct_tirs(priv, priv->xsk_tir);
+	mlx5e_destroy_direct_rqts(priv, priv->xsk_tir);
+	mlx5e_destroy_direct_tirs(priv, priv->direct_tir);
 	mlx5e_destroy_indirect_tirs(priv, true);
-	mlx5e_destroy_direct_rqts(priv);
+	mlx5e_destroy_direct_rqts(priv, priv->direct_tir);
 	mlx5e_destroy_rqt(priv, &priv->indir_rqt);
 	mlx5e_close_drop_rq(&priv->drop_rq);
 	mlx5e_destroy_q_counters(priv);
@@ -4998,7 +5209,7 @@ struct net_device *mlx5e_create_netdev(struct mlx5_core_dev *mdev,
 
 	netdev = alloc_etherdev_mqs(sizeof(struct mlx5e_priv),
 				    nch * profile->max_tc,
-				    nch);
+				    nch * MLX5E_NUM_RQ_GROUPS);
 	if (!netdev) {
 		mlx5_core_err(mdev, "alloc_etherdev_mqs() failed\n");
 		return NULL;