Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit ba1d82e6 authored by David S. Miller's avatar David S. Miller
Browse files

Merge branch 'nfp-XDP_TX-optimizations' 



Jakub Kicinski says:

====================
nfp: optimize XDP TX and small fixes

This series optimizes the nfp XDP TX performance a little bit.
I run quick tests on an Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz.
Single core/queue performance for both touch and drop and touch and
forward is above 20Mpps @64B packets, drop being 2Mpps faster.
I think this is max for a single queue on the low power NFPs.

There are also a few minor fixes included for code in net-next.
====================

Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parents 1d11e732 dbf637ff

drivers/net/ethernet/netronome/nfp/nfp_bpf_jit.c

+12 −12
Original line number Diff line number Diff line
@@ -798,7 +798,7 @@ wrp_test_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, 
	const struct bpf_insn *insn = &meta->insn;



	if (insn->off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	wrp_test_reg_one(nfp_prog, insn->dst_reg * 2, alu_op,

			 insn->src_reg * 2, br_mask, insn->off);
@@ -818,7 +818,7 @@ wrp_cmp_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, 
	u32 tmp_reg;



	if (insn->off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog));

	if (!swap)
@@ -847,7 +847,7 @@ wrp_cmp_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta, 
	u8 areg = insn->src_reg * 2, breg = insn->dst_reg * 2;



	if (insn->off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	if (swap) {

		areg ^= breg;
@@ -1132,7 +1132,7 @@ static int mem_ldx4_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 
		emit_alu(nfp_prog, reg_both(meta->insn.dst_reg * 2),

			 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_LEN);

	else

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	return 0;

}
@@ -1143,7 +1143,7 @@ static int mem_ldx4_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 


	if (meta->insn.off != offsetof(struct xdp_md, data) &&

	    meta->insn.off != offsetof(struct xdp_md, data_end))

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	emit_alu(nfp_prog, dst, reg_none(), ALU_OP_NONE, NFP_BPF_ABI_PKT);
@@ -1174,12 +1174,12 @@ static int mem_stx4_skb(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 
	if (meta->insn.off == offsetof(struct sk_buff, mark))

		return wrp_set_mark(nfp_prog, meta->insn.src_reg * 2);



	return -ENOTSUPP;

	return -EOPNOTSUPP;

}



static int mem_stx4_xdp(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)

{

	return -ENOTSUPP;

	return -EOPNOTSUPP;

}



static int mem_stx4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
@@ -1192,7 +1192,7 @@ static int mem_stx4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 
static int jump(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)

{

	if (meta->insn.off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;

	emit_br(nfp_prog, BR_UNC, meta->insn.off, 0);



	return 0;
@@ -1206,7 +1206,7 @@ static int jeq_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 
	u32 tmp_reg;



	if (insn->off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	if (imm & ~0U) {

		tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog));
@@ -1245,7 +1245,7 @@ static int jset_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 
	u32 tmp_reg;



	if (insn->off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	if (!imm) {

		meta->skip = true;
@@ -1276,7 +1276,7 @@ static int jne_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 
	u32 tmp_reg;



	if (insn->off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	if (!imm) {

		emit_alu(nfp_prog, reg_none(), reg_a(insn->dst_reg * 2),
@@ -1302,7 +1302,7 @@ static int jeq_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta) 
	const struct bpf_insn *insn = &meta->insn;



	if (insn->off < 0) /* TODO */

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	emit_alu(nfp_prog, imm_a(nfp_prog), reg_a(insn->dst_reg * 2),

		 ALU_OP_XOR, reg_b(insn->src_reg * 2));

drivers/net/ethernet/netronome/nfp/nfp_net.h

+4 −2
Original line number Diff line number Diff line
@@ -201,6 +201,7 @@ struct nfp_net_tx_buf { 
 * @txds:       Virtual address of TX ring in host memory

 * @dma:        DMA address of the TX ring

 * @size:       Size, in bytes, of the TX ring (needed to free)

 * @is_xdp:	Is this a XDP TX ring?

 */

struct nfp_net_tx_ring {

	struct nfp_net_r_vector *r_vec;
@@ -221,6 +222,7 @@ struct nfp_net_tx_ring { 


	dma_addr_t dma;

	unsigned int size;

	bool is_xdp;

} ____cacheline_aligned;



/* RX and freelist descriptor format */
@@ -468,10 +470,10 @@ struct nfp_net_dp { 
	u8 chained_metadata_format:1;



	u8 rx_dma_dir;

	u8 rx_dma_off;



	u8 rx_offset;



	u32 rx_dma_off;



	u32 ctrl;

	u32 fl_bufsz;

drivers/net/ethernet/netronome/nfp/nfp_net_common.c

+94 −77
Original line number Diff line number Diff line
@@ -478,15 +478,18 @@ static irqreturn_t nfp_net_irq_exn(int irq, void *data) 
 * @tx_ring:  TX ring structure

 * @r_vec:    IRQ vector servicing this ring

 * @idx:      Ring index

 * @is_xdp:   Is this an XDP TX ring?

 */

static void

nfp_net_tx_ring_init(struct nfp_net_tx_ring *tx_ring,

		     struct nfp_net_r_vector *r_vec, unsigned int idx)

		     struct nfp_net_r_vector *r_vec, unsigned int idx,

		     bool is_xdp)

{

	struct nfp_net *nn = r_vec->nfp_net;



	tx_ring->idx = idx;

	tx_ring->r_vec = r_vec;

	tx_ring->is_xdp = is_xdp;



	tx_ring->qcidx = tx_ring->idx * nn->stride_tx;

	tx_ring->qcp_q = nn->tx_bar + NFP_QCP_QUEUE_OFF(tx_ring->qcidx);
@@ -924,6 +927,9 @@ static void nfp_net_tx_complete(struct nfp_net_tx_ring *tx_ring) 
	int fidx;

	int idx;



	if (tx_ring->wr_p == tx_ring->rd_p)

		return;



	/* Work out how many descriptors have been transmitted */

	qcp_rd_p = nfp_qcp_rd_ptr_read(tx_ring->qcp_q);
@@ -994,11 +1000,13 @@ static void nfp_net_tx_complete(struct nfp_net_tx_ring *tx_ring) 
static void nfp_net_xdp_complete(struct nfp_net_tx_ring *tx_ring)

{

	struct nfp_net_r_vector *r_vec = tx_ring->r_vec;

	struct nfp_net_dp *dp = &r_vec->nfp_net->dp;

	u32 done_pkts = 0, done_bytes = 0;

	int idx, todo;

	u32 qcp_rd_p;



	if (tx_ring->wr_p == tx_ring->rd_p)

		return;



	/* Work out how many descriptors have been transmitted */

	qcp_rd_p = nfp_qcp_rd_ptr_read(tx_ring->qcp_q);
@@ -1010,22 +1018,12 @@ static void nfp_net_xdp_complete(struct nfp_net_tx_ring *tx_ring) 
	else

		todo = qcp_rd_p + tx_ring->cnt - tx_ring->qcp_rd_p;



	done_pkts = todo;

	while (todo--) {

		idx = tx_ring->rd_p & (tx_ring->cnt - 1);

		tx_ring->rd_p++;



		if (!tx_ring->txbufs[idx].frag)

			continue;



		nfp_net_dma_unmap_rx(dp, tx_ring->txbufs[idx].dma_addr);

		__free_page(virt_to_page(tx_ring->txbufs[idx].frag));



		done_pkts++;

		done_bytes += tx_ring->txbufs[idx].real_len;



		tx_ring->txbufs[idx].dma_addr = 0;

		tx_ring->txbufs[idx].frag = NULL;

		tx_ring->txbufs[idx].fidx = -2;

	}



	tx_ring->qcp_rd_p = qcp_rd_p;
@@ -1050,41 +1048,34 @@ static void nfp_net_xdp_complete(struct nfp_net_tx_ring *tx_ring) 
static void

nfp_net_tx_ring_reset(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring)

{

	struct nfp_net_r_vector *r_vec = tx_ring->r_vec;

	const struct skb_frag_struct *frag;

	struct netdev_queue *nd_q;



	while (tx_ring->rd_p != tx_ring->wr_p) {

	while (!tx_ring->is_xdp && tx_ring->rd_p != tx_ring->wr_p) {

		struct nfp_net_tx_buf *tx_buf;

		int idx;

		struct sk_buff *skb;

		int idx, nr_frags;



		idx = tx_ring->rd_p & (tx_ring->cnt - 1);

		tx_buf = &tx_ring->txbufs[idx];



		if (tx_ring == r_vec->xdp_ring) {

			nfp_net_dma_unmap_rx(dp, tx_buf->dma_addr);

			__free_page(virt_to_page(tx_ring->txbufs[idx].frag));

		} else {

			struct sk_buff *skb = tx_ring->txbufs[idx].skb;

			int nr_frags = skb_shinfo(skb)->nr_frags;

		skb = tx_ring->txbufs[idx].skb;

		nr_frags = skb_shinfo(skb)->nr_frags;



		if (tx_buf->fidx == -1) {

			/* unmap head */

			dma_unmap_single(dp->dev, tx_buf->dma_addr,

						 skb_headlen(skb),

						 DMA_TO_DEVICE);

					 skb_headlen(skb), DMA_TO_DEVICE);

		} else {

			/* unmap fragment */

			frag = &skb_shinfo(skb)->frags[tx_buf->fidx];

			dma_unmap_page(dp->dev, tx_buf->dma_addr,

					       skb_frag_size(frag),

					       DMA_TO_DEVICE);

				       skb_frag_size(frag), DMA_TO_DEVICE);

		}



		/* check for last gather fragment */

		if (tx_buf->fidx == nr_frags - 1)

			dev_kfree_skb_any(skb);

		}



		tx_buf->dma_addr = 0;

		tx_buf->skb = NULL;
@@ -1100,7 +1091,7 @@ nfp_net_tx_ring_reset(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring) 
	tx_ring->qcp_rd_p = 0;

	tx_ring->wr_ptr_add = 0;



	if (tx_ring == r_vec->xdp_ring)

	if (tx_ring->is_xdp)

		return;



	nd_q = netdev_get_tx_queue(dp->netdev, tx_ring->idx);
@@ -1153,16 +1144,13 @@ nfp_net_free_frag(void *frag, bool xdp) 
/**

 * nfp_net_rx_alloc_one() - Allocate and map page frag for RX

 * @dp:		NFP Net data path struct

 * @rx_ring:	RX ring structure of the skb

 * @dma_addr:	Pointer to storage for DMA address (output param)

 *

 * This function will allcate a new page frag, map it for DMA.

 *

 * Return: allocated page frag or NULL on failure.

 */

static void *

nfp_net_rx_alloc_one(struct nfp_net_dp *dp, struct nfp_net_rx_ring *rx_ring,

		     dma_addr_t *dma_addr)

static void *nfp_net_rx_alloc_one(struct nfp_net_dp *dp, dma_addr_t *dma_addr)

{

	void *frag;
@@ -1317,8 +1305,7 @@ nfp_net_rx_ring_bufs_alloc(struct nfp_net_dp *dp, 
	rxbufs = rx_ring->rxbufs;



	for (i = 0; i < rx_ring->cnt - 1; i++) {

		rxbufs[i].frag =

			nfp_net_rx_alloc_one(dp, rx_ring, &rxbufs[i].dma_addr);

		rxbufs[i].frag = nfp_net_rx_alloc_one(dp, &rxbufs[i].dma_addr);

		if (!rxbufs[i].frag) {

			nfp_net_rx_ring_bufs_free(dp, rx_ring);

			return -ENOMEM;
@@ -1496,8 +1483,6 @@ nfp_net_tx_xdp_buf(struct nfp_net_dp *dp, struct nfp_net_rx_ring *rx_ring, 
{

	struct nfp_net_tx_buf *txbuf;

	struct nfp_net_tx_desc *txd;

	dma_addr_t new_dma_addr;

	void *new_frag;

	int wr_idx;



	if (unlikely(nfp_net_tx_full(tx_ring, 1))) {
@@ -1505,17 +1490,13 @@ nfp_net_tx_xdp_buf(struct nfp_net_dp *dp, struct nfp_net_rx_ring *rx_ring, 
		return false;

	}



	new_frag = nfp_net_napi_alloc_one(dp, &new_dma_addr);

	if (unlikely(!new_frag)) {

		nfp_net_rx_drop(dp, rx_ring->r_vec, rx_ring, rxbuf, NULL);

		return false;

	}

	nfp_net_rx_give_one(dp, rx_ring, new_frag, new_dma_addr);



	wr_idx = tx_ring->wr_p & (tx_ring->cnt - 1);



	/* Stash the soft descriptor of the head then initialize it */

	txbuf = &tx_ring->txbufs[wr_idx];



	nfp_net_rx_give_one(dp, rx_ring, txbuf->frag, txbuf->dma_addr);



	txbuf->frag = rxbuf->frag;

	txbuf->dma_addr = rxbuf->dma_addr;

	txbuf->fidx = -1;
@@ -1800,13 +1781,11 @@ static void nfp_net_tx_ring_free(struct nfp_net_tx_ring *tx_ring) 
 * nfp_net_tx_ring_alloc() - Allocate resource for a TX ring

 * @dp:        NFP Net data path struct

 * @tx_ring:   TX Ring structure to allocate

 * @is_xdp:    True if ring will be used for XDP

 *

 * Return: 0 on success, negative errno otherwise.

 */

static int

nfp_net_tx_ring_alloc(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring,

		      bool is_xdp)

nfp_net_tx_ring_alloc(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring)

{

	struct nfp_net_r_vector *r_vec = tx_ring->r_vec;

	int sz;
@@ -1824,7 +1803,7 @@ nfp_net_tx_ring_alloc(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring, 
	if (!tx_ring->txbufs)

		goto err_alloc;



	if (!is_xdp)

	if (!tx_ring->is_xdp)

		netif_set_xps_queue(dp->netdev, &r_vec->affinity_mask,

				    tx_ring->idx);
@@ -1835,6 +1814,45 @@ nfp_net_tx_ring_alloc(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring, 
	return -ENOMEM;

}



static void

nfp_net_tx_ring_bufs_free(struct nfp_net_dp *dp,

			  struct nfp_net_tx_ring *tx_ring)

{

	unsigned int i;



	if (!tx_ring->is_xdp)

		return;



	for (i = 0; i < tx_ring->cnt; i++) {

		if (!tx_ring->txbufs[i].frag)

			return;



		nfp_net_dma_unmap_rx(dp, tx_ring->txbufs[i].dma_addr);

		__free_page(virt_to_page(tx_ring->txbufs[i].frag));

	}

}



static int

nfp_net_tx_ring_bufs_alloc(struct nfp_net_dp *dp,

			   struct nfp_net_tx_ring *tx_ring)

{

	struct nfp_net_tx_buf *txbufs = tx_ring->txbufs;

	unsigned int i;



	if (!tx_ring->is_xdp)

		return 0;



	for (i = 0; i < tx_ring->cnt; i++) {

		txbufs[i].frag = nfp_net_rx_alloc_one(dp, &txbufs[i].dma_addr);

		if (!txbufs[i].frag) {

			nfp_net_tx_ring_bufs_free(dp, tx_ring);

			return -ENOMEM;

		}

	}



	return 0;

}



static int nfp_net_tx_rings_prepare(struct nfp_net *nn, struct nfp_net_dp *dp)

{

	unsigned int r;
@@ -1851,17 +1869,23 @@ static int nfp_net_tx_rings_prepare(struct nfp_net *nn, struct nfp_net_dp *dp) 
			bias = dp->num_stack_tx_rings;



		nfp_net_tx_ring_init(&dp->tx_rings[r], &nn->r_vecs[r - bias],

				     r);

				     r, bias);



		if (nfp_net_tx_ring_alloc(dp, &dp->tx_rings[r], bias))

		if (nfp_net_tx_ring_alloc(dp, &dp->tx_rings[r]))

			goto err_free_prev;



		if (nfp_net_tx_ring_bufs_alloc(dp, &dp->tx_rings[r]))

			goto err_free_ring;

	}



	return 0;



err_free_prev:

	while (r--)

	while (r--) {

		nfp_net_tx_ring_bufs_free(dp, &dp->tx_rings[r]);

err_free_ring:

		nfp_net_tx_ring_free(&dp->tx_rings[r]);

	}

	kfree(dp->tx_rings);

	return -ENOMEM;

}
@@ -1870,8 +1894,10 @@ static void nfp_net_tx_rings_free(struct nfp_net_dp *dp) 
{

	unsigned int r;



	for (r = 0; r < dp->num_tx_rings; r++)

	for (r = 0; r < dp->num_tx_rings; r++) {

		nfp_net_tx_ring_bufs_free(dp, &dp->tx_rings[r]);

		nfp_net_tx_ring_free(&dp->tx_rings[r]);

	}



	kfree(dp->tx_rings);

}
@@ -2369,8 +2395,10 @@ static void nfp_net_close_free_all(struct nfp_net *nn) 
		nfp_net_rx_ring_bufs_free(&nn->dp, &nn->dp.rx_rings[r]);

		nfp_net_rx_ring_free(&nn->dp.rx_rings[r]);

	}

	for (r = 0; r < nn->dp.num_tx_rings; r++)

	for (r = 0; r < nn->dp.num_tx_rings; r++) {

		nfp_net_tx_ring_bufs_free(&nn->dp, &nn->dp.tx_rings[r]);

		nfp_net_tx_ring_free(&nn->dp.tx_rings[r]);

	}

	for (r = 0; r < nn->dp.num_r_vecs; r++)

		nfp_net_cleanup_vector(nn, &nn->r_vecs[r]);
@@ -2651,9 +2679,9 @@ nfp_net_setup_tc(struct net_device *netdev, u32 handle, __be16 proto, 
	struct nfp_net *nn = netdev_priv(netdev);



	if (TC_H_MAJ(handle) != TC_H_MAJ(TC_H_INGRESS))

		return -ENOTSUPP;

		return -EOPNOTSUPP;

	if (proto != htons(ETH_P_ALL))

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	if (tc->type == TC_SETUP_CLSBPF && nfp_net_ebpf_capable(nn)) {

		if (!nn->dp.bpf_offload_xdp)
@@ -2938,11 +2966,7 @@ static int nfp_net_xdp_setup(struct nfp_net *nn, struct bpf_prog *prog) 
	dp->xdp_prog = prog;

	dp->num_tx_rings += prog ? nn->dp.num_rx_rings : -nn->dp.num_rx_rings;

	dp->rx_dma_dir = prog ? DMA_BIDIRECTIONAL : DMA_FROM_DEVICE;

	if (prog)

		dp->rx_dma_off = XDP_PACKET_HEADROOM -

			(nn->dp.rx_offset ?: NFP_NET_MAX_PREPEND);

	else

		dp->rx_dma_off = 0;

	dp->rx_dma_off = prog ? XDP_PACKET_HEADROOM - nn->dp.rx_offset : 0;



	/* We need RX reconfig to remap the buffers (BIDIR vs FROM_DEV) */

	err = nfp_net_ring_reconfig(nn, dp);
@@ -3170,13 +3194,6 @@ int nfp_net_netdev_init(struct net_device *netdev) 
	struct nfp_net *nn = netdev_priv(netdev);

	int err;



	/* XDP calls for 256 byte packet headroom which wouldn't fit in a u8.

	 * We, however, reuse the metadata prepend space for XDP buffers which

	 * is at least 1 byte long and as long as XDP headroom doesn't increase

	 * above 256 the *extra* XDP headroom will fit on 8 bits.

	 */

	BUILD_BUG_ON(XDP_PACKET_HEADROOM > 256);



	nn->dp.chained_metadata_format = nn->fw_ver.major > 3;



	nn->dp.rx_dma_dir = DMA_FROM_DEVICE;

drivers/net/ethernet/netronome/nfp/nfp_net_ethtool.c

+1 −1
Original line number Diff line number Diff line
@@ -791,7 +791,7 @@ static int nfp_net_set_coalesce(struct net_device *netdev, 
	    ec->tx_coalesce_usecs_high ||

	    ec->tx_max_coalesced_frames_high ||

	    ec->rate_sample_interval)

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	/* Compute factor used to convert coalesce '_usecs' parameters to

	 * ME timestamp ticks.  There are 16 ME clock cycles for each timestamp

drivers/net/ethernet/netronome/nfp/nfp_net_offload.c

+6 −6
Original line number Diff line number Diff line
@@ -119,12 +119,12 @@ nfp_net_bpf_get_act(struct nfp_net *nn, struct tc_cls_bpf_offload *cls_bpf) 
		if (tc_no_actions(cls_bpf->exts))

			return NN_ACT_DIRECT;



		return -ENOTSUPP;

		return -EOPNOTSUPP;

	}



	/* TC legacy mode */

	if (!tc_single_action(cls_bpf->exts))

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	tcf_exts_to_list(cls_bpf->exts, &actions);

	list_for_each_entry(a, &actions, list) {
@@ -136,7 +136,7 @@ nfp_net_bpf_get_act(struct nfp_net *nn, struct tc_cls_bpf_offload *cls_bpf) 
			return NN_ACT_TC_REDIR;

	}



	return -ENOTSUPP;

	return -EOPNOTSUPP;

}



static int
@@ -152,7 +152,7 @@ nfp_net_bpf_offload_prepare(struct nfp_net *nn, 
	int ret;



	if (!IS_ENABLED(CONFIG_BPF_SYSCALL))

		return -ENOTSUPP;

		return -EOPNOTSUPP;



	ret = nfp_net_bpf_get_act(nn, cls_bpf);

	if (ret < 0)
@@ -162,7 +162,7 @@ nfp_net_bpf_offload_prepare(struct nfp_net *nn, 
	max_mtu = nn_readb(nn, NFP_NET_CFG_BPF_INL_MTU) * 64 - 32;

	if (max_mtu < nn->dp.netdev->mtu) {

		nn_info(nn, "BPF offload not supported with MTU larger than HW packet split boundary\n");

		return -ENOTSUPP;

		return -EOPNOTSUPP;

	}



	start_off = nn_readw(nn, NFP_NET_CFG_BPF_START);
@@ -289,6 +289,6 @@ int nfp_net_bpf_offload(struct nfp_net *nn, struct tc_cls_bpf_offload *cls_bpf) 
		return nfp_net_bpf_stats_update(nn, cls_bpf);



	default:

		return -ENOTSUPP;

		return -EOPNOTSUPP;

	}

}