예제 #1
0
파일: en_main.c 프로젝트: Dyoed/ath
static int mlx5e_open_rq(struct mlx5e_channel *c,
			 struct mlx5e_rq_param *param,
			 struct mlx5e_rq *rq)
{
	int err;

	err = mlx5e_create_rq(c, param, rq);
	if (err)
		return err;

	err = mlx5e_enable_rq(rq, param);
	if (err)
		goto err_destroy_rq;

	err = mlx5e_modify_rq(rq, MLX5_RQC_STATE_RST, MLX5_RQC_STATE_RDY);
	if (err)
		goto err_disable_rq;

	set_bit(MLX5E_RQ_STATE_POST_WQES_ENABLE, &rq->state);
	mlx5e_send_nop(&c->sq[0]); /* trigger mlx5e_post_rx_wqes() */

	return 0;

err_disable_rq:
	mlx5e_disable_rq(rq);
err_destroy_rq:
	mlx5e_destroy_rq(rq);

	return err;
}
예제 #2
0
파일: en_main.c 프로젝트: Dyoed/ath
static void mlx5e_close_sq(struct mlx5e_sq *sq)
{
	clear_bit(MLX5E_SQ_STATE_WAKE_TXQ_ENABLE, &sq->state);
	napi_synchronize(&sq->channel->napi); /* prevent netif_tx_wake_queue */
	netif_tx_disable_queue(sq->txq);

	/* ensure hw is notified of all pending wqes */
	if (mlx5e_sq_has_room_for(sq, 1))
		mlx5e_send_nop(sq);

	mlx5e_modify_sq(sq, MLX5_SQC_STATE_RDY, MLX5_SQC_STATE_ERR);
	while (sq->cc != sq->pc) /* wait till sq is empty */
		msleep(20);

	/* avoid destroying sq before mlx5e_poll_tx_cq() is done with it */
	napi_synchronize(&sq->channel->napi);

	mlx5e_disable_sq(sq);
	mlx5e_destroy_sq(sq);
}
예제 #3
0
파일: en_tx.c 프로젝트: DenisLug/mptcp
static netdev_tx_t mlx5e_sq_xmit(struct mlx5e_sq *sq, struct sk_buff *skb)
{
	struct mlx5_wq_cyc       *wq   = &sq->wq;

	u16 pi = sq->pc & wq->sz_m1;
	struct mlx5e_tx_wqe      *wqe  = mlx5_wq_cyc_get_wqe(wq, pi);

	struct mlx5_wqe_ctrl_seg *cseg = &wqe->ctrl;
	struct mlx5_wqe_eth_seg  *eseg = &wqe->eth;
	struct mlx5_wqe_data_seg *dseg;

	u8  opcode = MLX5_OPCODE_SEND;
	dma_addr_t dma_addr = 0;
	bool bf = false;
	u16 headlen;
	u16 ds_cnt;
	u16 ihs;
	int i;

	memset(wqe, 0, sizeof(*wqe));

	if (likely(skb->ip_summed == CHECKSUM_PARTIAL))
		eseg->cs_flags	= MLX5_ETH_WQE_L3_CSUM | MLX5_ETH_WQE_L4_CSUM;
	else
		sq->stats.csum_offload_none++;

	if (sq->cc != sq->prev_cc) {
		sq->prev_cc = sq->cc;
		sq->bf_budget = (sq->cc == sq->pc) ? MLX5E_SQ_BF_BUDGET : 0;
	}

	if (skb_is_gso(skb)) {
		u32 payload_len;

		eseg->mss    = cpu_to_be16(skb_shinfo(skb)->gso_size);
		opcode       = MLX5_OPCODE_LSO;
		ihs          = skb_transport_offset(skb) + tcp_hdrlen(skb);
		payload_len  = skb->len - ihs;
		MLX5E_TX_SKB_CB(skb)->num_bytes = skb->len +
					(skb_shinfo(skb)->gso_segs - 1) * ihs;
		sq->stats.tso_packets++;
		sq->stats.tso_bytes += payload_len;
	} else {
		bf = sq->bf_budget &&
		     !skb->xmit_more &&
		     !skb_shinfo(skb)->nr_frags;
		ihs = mlx5e_get_inline_hdr_size(sq, skb, bf);
		MLX5E_TX_SKB_CB(skb)->num_bytes = max_t(unsigned int, skb->len,
							ETH_ZLEN);
	}

	skb_copy_from_linear_data(skb, eseg->inline_hdr_start, ihs);
	skb_pull_inline(skb, ihs);

	eseg->inline_hdr_sz = cpu_to_be16(ihs);

	ds_cnt  = sizeof(*wqe) / MLX5_SEND_WQE_DS;
	ds_cnt += DIV_ROUND_UP(ihs - sizeof(eseg->inline_hdr_start),
			       MLX5_SEND_WQE_DS);
	dseg    = (struct mlx5_wqe_data_seg *)cseg + ds_cnt;

	MLX5E_TX_SKB_CB(skb)->num_dma = 0;

	headlen = skb_headlen(skb);
	if (headlen) {
		dma_addr = dma_map_single(sq->pdev, skb->data, headlen,
					  DMA_TO_DEVICE);
		if (unlikely(dma_mapping_error(sq->pdev, dma_addr)))
			goto dma_unmap_wqe_err;

		dseg->addr       = cpu_to_be64(dma_addr);
		dseg->lkey       = sq->mkey_be;
		dseg->byte_count = cpu_to_be32(headlen);

		mlx5e_dma_push(sq, dma_addr, headlen);
		MLX5E_TX_SKB_CB(skb)->num_dma++;

		dseg++;
	}

	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
		int fsz = skb_frag_size(frag);

		dma_addr = skb_frag_dma_map(sq->pdev, frag, 0, fsz,
					    DMA_TO_DEVICE);
		if (unlikely(dma_mapping_error(sq->pdev, dma_addr)))
			goto dma_unmap_wqe_err;

		dseg->addr       = cpu_to_be64(dma_addr);
		dseg->lkey       = sq->mkey_be;
		dseg->byte_count = cpu_to_be32(fsz);

		mlx5e_dma_push(sq, dma_addr, fsz);
		MLX5E_TX_SKB_CB(skb)->num_dma++;

		dseg++;
	}

	ds_cnt += MLX5E_TX_SKB_CB(skb)->num_dma;

	cseg->opmod_idx_opcode = cpu_to_be32((sq->pc << 8) | opcode);
	cseg->qpn_ds           = cpu_to_be32((sq->sqn << 8) | ds_cnt);

	sq->skb[pi] = skb;

	MLX5E_TX_SKB_CB(skb)->num_wqebbs = DIV_ROUND_UP(ds_cnt,
							MLX5_SEND_WQEBB_NUM_DS);
	sq->pc += MLX5E_TX_SKB_CB(skb)->num_wqebbs;

	netdev_tx_sent_queue(sq->txq, MLX5E_TX_SKB_CB(skb)->num_bytes);

	if (unlikely(!mlx5e_sq_has_room_for(sq, MLX5E_SQ_STOP_ROOM))) {
		netif_tx_stop_queue(sq->txq);
		sq->stats.stopped++;
	}

	if (!skb->xmit_more || netif_xmit_stopped(sq->txq)) {
		int bf_sz = 0;

		if (bf && sq->uar_bf_map)
			bf_sz = MLX5E_TX_SKB_CB(skb)->num_wqebbs << 3;

		cseg->fm_ce_se = MLX5_WQE_CTRL_CQ_UPDATE;
		mlx5e_tx_notify_hw(sq, wqe, bf_sz);
	}

	/* fill sq edge with nops to avoid wqe wrap around */
	while ((sq->pc & wq->sz_m1) > sq->edge)
		mlx5e_send_nop(sq, false);

	sq->bf_budget = bf ? sq->bf_budget - 1 : 0;

	sq->stats.packets++;
	return NETDEV_TX_OK;

dma_unmap_wqe_err:
	sq->stats.dropped++;
	mlx5e_dma_unmap_wqe_err(sq, skb);

	dev_kfree_skb_any(skb);

	return NETDEV_TX_OK;
}
예제 #4
0
static netdev_tx_t mlx5e_sq_xmit(struct mlx5e_sq *sq, struct sk_buff *skb)
{
	struct mlx5_wq_cyc       *wq   = &sq->wq;

	u16 pi = sq->pc & wq->sz_m1;
	struct mlx5e_tx_wqe      *wqe  = mlx5_wq_cyc_get_wqe(wq, pi);
	struct mlx5e_tx_wqe_info *wi   = &sq->wqe_info[pi];

	struct mlx5_wqe_ctrl_seg *cseg = &wqe->ctrl;
	struct mlx5_wqe_eth_seg  *eseg = &wqe->eth;
	struct mlx5_wqe_data_seg *dseg;

	unsigned char *skb_data = skb->data;
	unsigned int skb_len = skb->len;
	u8  opcode = MLX5_OPCODE_SEND;
	dma_addr_t dma_addr = 0;
	unsigned int num_bytes;
	bool bf = false;
	u16 headlen;
	u16 ds_cnt = sizeof(*wqe) / MLX5_SEND_WQE_DS;
	u16 ihs;
	int i;

	memset(wqe, 0, sizeof(*wqe));

	if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
		eseg->cs_flags	= MLX5_ETH_WQE_L3_CSUM;
		if (skb->encapsulation) {
			eseg->cs_flags |= MLX5_ETH_WQE_L3_INNER_CSUM |
					  MLX5_ETH_WQE_L4_INNER_CSUM;
			sq->stats.csum_partial_inner++;
		} else {
			eseg->cs_flags |= MLX5_ETH_WQE_L4_CSUM;
			sq->stats.csum_partial++;
		}
	} else
		sq->stats.csum_none++;

	if (sq->cc != sq->prev_cc) {
		sq->prev_cc = sq->cc;
		sq->bf_budget = (sq->cc == sq->pc) ? MLX5E_SQ_BF_BUDGET : 0;
	}

	if (skb_is_gso(skb)) {
		eseg->mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
		opcode    = MLX5_OPCODE_LSO;

		if (skb->encapsulation) {
			ihs = skb_inner_transport_header(skb) - skb->data +
			      inner_tcp_hdrlen(skb);
			sq->stats.tso_inner_packets++;
			sq->stats.tso_inner_bytes += skb->len - ihs;
		} else {
			ihs = skb_transport_offset(skb) + tcp_hdrlen(skb);
			sq->stats.tso_packets++;
			sq->stats.tso_bytes += skb->len - ihs;
		}

		num_bytes = skb->len + (skb_shinfo(skb)->gso_segs - 1) * ihs;
	} else {
		bf = sq->bf_budget   &&
		     !skb->xmit_more &&
		     !skb_shinfo(skb)->nr_frags;
		ihs = mlx5e_get_inline_hdr_size(sq, skb, bf);
		num_bytes = max_t(unsigned int, skb->len, ETH_ZLEN);
	}

	wi->num_bytes = num_bytes;

	if (skb_vlan_tag_present(skb)) {
		mlx5e_insert_vlan(eseg->inline_hdr_start, skb, ihs, &skb_data,
				  &skb_len);
		ihs += VLAN_HLEN;
	} else {
		memcpy(eseg->inline_hdr_start, skb_data, ihs);
		mlx5e_tx_skb_pull_inline(&skb_data, &skb_len, ihs);
	}

	eseg->inline_hdr_sz = cpu_to_be16(ihs);

	ds_cnt += DIV_ROUND_UP(ihs - sizeof(eseg->inline_hdr_start),
			       MLX5_SEND_WQE_DS);

	dseg = (struct mlx5_wqe_data_seg *)cseg + ds_cnt;

	wi->num_dma = 0;

	headlen = skb_len - skb->data_len;
	if (headlen) {
		dma_addr = dma_map_single(sq->pdev, skb_data, headlen,
					  DMA_TO_DEVICE);
		if (unlikely(dma_mapping_error(sq->pdev, dma_addr)))
			goto dma_unmap_wqe_err;

		dseg->addr       = cpu_to_be64(dma_addr);
		dseg->lkey       = sq->mkey_be;
		dseg->byte_count = cpu_to_be32(headlen);

		mlx5e_dma_push(sq, dma_addr, headlen, MLX5E_DMA_MAP_SINGLE);
		wi->num_dma++;

		dseg++;
	}

	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
		int fsz = skb_frag_size(frag);

		dma_addr = skb_frag_dma_map(sq->pdev, frag, 0, fsz,
					    DMA_TO_DEVICE);
		if (unlikely(dma_mapping_error(sq->pdev, dma_addr)))
			goto dma_unmap_wqe_err;

		dseg->addr       = cpu_to_be64(dma_addr);
		dseg->lkey       = sq->mkey_be;
		dseg->byte_count = cpu_to_be32(fsz);

		mlx5e_dma_push(sq, dma_addr, fsz, MLX5E_DMA_MAP_PAGE);
		wi->num_dma++;

		dseg++;
	}

	ds_cnt += wi->num_dma;

	cseg->opmod_idx_opcode = cpu_to_be32((sq->pc << 8) | opcode);
	cseg->qpn_ds           = cpu_to_be32((sq->sqn << 8) | ds_cnt);

	sq->skb[pi] = skb;

	wi->num_wqebbs = DIV_ROUND_UP(ds_cnt, MLX5_SEND_WQEBB_NUM_DS);
	sq->pc += wi->num_wqebbs;

	if (unlikely(MLX5E_TX_HW_STAMP(sq->channel->priv, skb)))
		skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;

	netdev_tx_sent_queue(sq->txq, wi->num_bytes);

	if (unlikely(!mlx5e_sq_has_room_for(sq, MLX5E_SQ_STOP_ROOM))) {
		netif_tx_stop_queue(sq->txq);
		sq->stats.queue_stopped++;
	}

	if (!skb->xmit_more || netif_xmit_stopped(sq->txq)) {
		int bf_sz = 0;

		if (bf && sq->uar_bf_map)
			bf_sz = wi->num_wqebbs << 3;

		cseg->fm_ce_se = MLX5_WQE_CTRL_CQ_UPDATE;
		mlx5e_tx_notify_hw(sq, &wqe->ctrl, bf_sz);
	}

	sq->bf_budget = bf ? sq->bf_budget - 1 : 0;

	/* fill sq edge with nops to avoid wqe wrap around */
	while ((sq->pc & wq->sz_m1) > sq->edge)
		mlx5e_send_nop(sq, false);

	sq->stats.packets++;
	sq->stats.bytes += num_bytes;
	return NETDEV_TX_OK;

dma_unmap_wqe_err:
	sq->stats.queue_dropped++;
	mlx5e_dma_unmap_wqe_err(sq, wi->num_dma);

	dev_kfree_skb_any(skb);

	return NETDEV_TX_OK;
}