en_tx.c 20.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
/*
 * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */

#include <asm/page.h>
#include <linux/mlx4/cq.h>
36
#include <linux/slab.h>
37 38 39 40
#include <linux/mlx4/qp.h>
#include <linux/skbuff.h>
#include <linux/if_vlan.h>
#include <linux/vmalloc.h>
41
#include <linux/tcp.h>
42
#include <linux/moduleparam.h>
43 44 45 46 47

#include "mlx4_en.h"

enum {
	MAX_INLINE = 104, /* 128 - 16 - 4 - 4 */
48
	MAX_BF = 256,
49 50 51 52 53
};

static int inline_thold __read_mostly = MAX_INLINE;

module_param_named(inline_thold, inline_thold, int, 0444);
54
MODULE_PARM_DESC(inline_thold, "threshold for using inline data");
55 56

int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv,
57
			   struct mlx4_en_tx_ring *ring, int qpn, u32 size,
58 59 60 61 62 63 64 65 66 67 68 69 70 71
			   u16 stride)
{
	struct mlx4_en_dev *mdev = priv->mdev;
	int tmp;
	int err;

	ring->size = size;
	ring->size_mask = size - 1;
	ring->stride = stride;

	inline_thold = min(inline_thold, MAX_INLINE);

	tmp = size * sizeof(struct mlx4_en_tx_info);
	ring->tx_info = vmalloc(tmp);
72
	if (!ring->tx_info)
73
		return -ENOMEM;
74

75
	en_dbg(DRV, priv, "Allocated tx_info ring at addr:%p size:%d\n",
76 77 78 79 80 81 82 83 84 85 86 87
		 ring->tx_info, tmp);

	ring->bounce_buf = kmalloc(MAX_DESC_SIZE, GFP_KERNEL);
	if (!ring->bounce_buf) {
		err = -ENOMEM;
		goto err_tx;
	}
	ring->buf_size = ALIGN(size * ring->stride, MLX4_EN_PAGE_SIZE);

	err = mlx4_alloc_hwq_res(mdev->dev, &ring->wqres, ring->buf_size,
				 2 * PAGE_SIZE);
	if (err) {
88
		en_err(priv, "Failed allocating hwq resources\n");
89 90 91 92 93
		goto err_bounce;
	}

	err = mlx4_en_map_buffer(&ring->wqres.buf);
	if (err) {
94
		en_err(priv, "Failed to map TX buffer\n");
95 96 97 98 99
		goto err_hwq_res;
	}

	ring->buf = ring->wqres.buf.direct.buf;

100 101 102
	en_dbg(DRV, priv, "Allocated TX ring (addr:%p) - buf:%p size:%d "
	       "buf_size:%d dma:%llx\n", ring, ring->buf, ring->size,
	       ring->buf_size, (unsigned long long) ring->wqres.buf.direct.map);
103

104
	ring->qpn = qpn;
105 106
	err = mlx4_qp_alloc(mdev->dev, ring->qpn, &ring->qp);
	if (err) {
107
		en_err(priv, "Failed allocating qp %d\n", ring->qpn);
108
		goto err_map;
109
	}
110
	ring->qp.event = mlx4_en_sqp_event;
111

112 113 114 115 116 117 118 119 120
	err = mlx4_bf_alloc(mdev->dev, &ring->bf);
	if (err) {
		en_dbg(DRV, priv, "working without blueflame (%d)", err);
		ring->bf.uar = &mdev->priv_uar;
		ring->bf.uar->map = mdev->uar_map;
		ring->bf_enabled = false;
	} else
		ring->bf_enabled = true;

121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
	return 0;

err_map:
	mlx4_en_unmap_buffer(&ring->wqres.buf);
err_hwq_res:
	mlx4_free_hwq_res(mdev->dev, &ring->wqres, ring->buf_size);
err_bounce:
	kfree(ring->bounce_buf);
	ring->bounce_buf = NULL;
err_tx:
	vfree(ring->tx_info);
	ring->tx_info = NULL;
	return err;
}

void mlx4_en_destroy_tx_ring(struct mlx4_en_priv *priv,
			     struct mlx4_en_tx_ring *ring)
{
	struct mlx4_en_dev *mdev = priv->mdev;
140
	en_dbg(DRV, priv, "Destroying tx ring, qpn: %d\n", ring->qpn);
141

142 143
	if (ring->bf_enabled)
		mlx4_bf_free(mdev->dev, &ring->bf);
144 145 146 147 148 149 150 151 152 153 154 155
	mlx4_qp_remove(mdev->dev, &ring->qp);
	mlx4_qp_free(mdev->dev, &ring->qp);
	mlx4_en_unmap_buffer(&ring->wqres.buf);
	mlx4_free_hwq_res(mdev->dev, &ring->wqres, ring->buf_size);
	kfree(ring->bounce_buf);
	ring->bounce_buf = NULL;
	vfree(ring->tx_info);
	ring->tx_info = NULL;
}

int mlx4_en_activate_tx_ring(struct mlx4_en_priv *priv,
			     struct mlx4_en_tx_ring *ring,
156
			     int cq, int user_prio)
157 158 159 160 161 162 163 164 165 166 167 168 169
{
	struct mlx4_en_dev *mdev = priv->mdev;
	int err;

	ring->cqn = cq;
	ring->prod = 0;
	ring->cons = 0xffffffff;
	ring->last_nr_txbb = 1;
	ring->poll_cnt = 0;
	memset(ring->tx_info, 0, ring->size * sizeof(struct mlx4_en_tx_info));
	memset(ring->buf, 0, ring->buf_size);

	ring->qp_state = MLX4_QP_STATE_RST;
170
	ring->doorbell_qpn = ring->qp.qpn << 8;
171 172

	mlx4_en_fill_qp_context(priv, ring->size, ring->stride, 1, 0, ring->qpn,
173
				ring->cqn, user_prio, &ring->context);
174 175
	if (ring->bf_enabled)
		ring->context.usr_page = cpu_to_be32(ring->bf.uar->index);
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209

	err = mlx4_qp_to_ready(mdev->dev, &ring->wqres.mtt, &ring->context,
			       &ring->qp, &ring->qp_state);

	return err;
}

void mlx4_en_deactivate_tx_ring(struct mlx4_en_priv *priv,
				struct mlx4_en_tx_ring *ring)
{
	struct mlx4_en_dev *mdev = priv->mdev;

	mlx4_qp_modify(mdev->dev, NULL, ring->qp_state,
		       MLX4_QP_STATE_RST, NULL, 0, 0, &ring->qp);
}


static u32 mlx4_en_free_tx_desc(struct mlx4_en_priv *priv,
				struct mlx4_en_tx_ring *ring,
				int index, u8 owner)
{
	struct mlx4_en_tx_info *tx_info = &ring->tx_info[index];
	struct mlx4_en_tx_desc *tx_desc = ring->buf + index * TXBB_SIZE;
	struct mlx4_wqe_data_seg *data = (void *) tx_desc + tx_info->data_offset;
	struct sk_buff *skb = tx_info->skb;
	struct skb_frag_struct *frag;
	void *end = ring->buf + ring->buf_size;
	int frags = skb_shinfo(skb)->nr_frags;
	int i;
	__be32 *ptr = (__be32 *)tx_desc;
	__be32 stamp = cpu_to_be32(STAMP_VAL | (!!owner << STAMP_SHIFT));

	/* Optimize the common case when there are no wraparounds */
	if (likely((void *) tx_desc + tx_info->nr_txbb * TXBB_SIZE <= end)) {
210 211
		if (!tx_info->inl) {
			if (tx_info->linear) {
212
				dma_unmap_single(priv->ddev,
213
					(dma_addr_t) be64_to_cpu(data->addr),
214 215
					 be32_to_cpu(data->byte_count),
					 PCI_DMA_TODEVICE);
216 217
				++data;
			}
218

219 220
			for (i = 0; i < frags; i++) {
				frag = &skb_shinfo(skb)->frags[i];
221
				dma_unmap_page(priv->ddev,
222
					(dma_addr_t) be64_to_cpu(data[i].addr),
E
Eric Dumazet 已提交
223
					skb_frag_size(frag), PCI_DMA_TODEVICE);
224
			}
225 226 227 228 229 230 231 232
		}
		/* Stamp the freed descriptor */
		for (i = 0; i < tx_info->nr_txbb * TXBB_SIZE; i += STAMP_STRIDE) {
			*ptr = stamp;
			ptr += STAMP_DWORDS;
		}

	} else {
233 234
		if (!tx_info->inl) {
			if ((void *) data >= end) {
235
				data = ring->buf + ((void *)data - end);
236
			}
237

238
			if (tx_info->linear) {
239
				dma_unmap_single(priv->ddev,
240
					(dma_addr_t) be64_to_cpu(data->addr),
241 242
					 be32_to_cpu(data->byte_count),
					 PCI_DMA_TODEVICE);
243 244
				++data;
			}
245

246 247 248
			for (i = 0; i < frags; i++) {
				/* Check for wraparound before unmapping */
				if ((void *) data >= end)
249
					data = ring->buf;
250
				frag = &skb_shinfo(skb)->frags[i];
251
				dma_unmap_page(priv->ddev,
252
					(dma_addr_t) be64_to_cpu(data->addr),
E
Eric Dumazet 已提交
253
					 skb_frag_size(frag), PCI_DMA_TODEVICE);
254
				++data;
255
			}
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
		}
		/* Stamp the freed descriptor */
		for (i = 0; i < tx_info->nr_txbb * TXBB_SIZE; i += STAMP_STRIDE) {
			*ptr = stamp;
			ptr += STAMP_DWORDS;
			if ((void *) ptr >= end) {
				ptr = ring->buf;
				stamp ^= cpu_to_be32(0x80000000);
			}
		}

	}
	dev_kfree_skb_any(skb);
	return tx_info->nr_txbb;
}


int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring)
{
	struct mlx4_en_priv *priv = netdev_priv(dev);
	int cnt = 0;

	/* Skip last polled descriptor */
	ring->cons += ring->last_nr_txbb;
280
	en_dbg(DRV, priv, "Freeing Tx buf - cons:0x%x prod:0x%x\n",
281 282 283 284
		 ring->cons, ring->prod);

	if ((u32) (ring->prod - ring->cons) > ring->size) {
		if (netif_msg_tx_err(priv))
285
			en_warn(priv, "Tx consumer passed producer!\n");
286 287 288 289 290 291 292 293 294 295 296 297
		return 0;
	}

	while (ring->cons != ring->prod) {
		ring->last_nr_txbb = mlx4_en_free_tx_desc(priv, ring,
						ring->cons & ring->size_mask,
						!!(ring->cons & ring->size));
		ring->cons += ring->last_nr_txbb;
		cnt++;
	}

	if (cnt)
298
		en_dbg(DRV, priv, "Freed %d uncompleted tx descriptors\n", cnt);
299 300 301 302 303 304 305 306 307

	return cnt;
}

static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
{
	struct mlx4_en_priv *priv = netdev_priv(dev);
	struct mlx4_cq *mcq = &cq->mcq;
	struct mlx4_en_tx_ring *ring = &priv->tx_ring[cq->ring];
308
	struct mlx4_cqe *cqe;
309
	u16 index;
310
	u16 new_index, ring_index;
311
	u32 txbbs_skipped = 0;
312 313 314 315
	u32 cons_index = mcq->cons_index;
	int size = cq->size;
	u32 size_mask = ring->size_mask;
	struct mlx4_cqe *buf = cq->buf;
316 317
	u32 packets = 0;
	u32 bytes = 0;
O
Or Gerlitz 已提交
318
	int factor = priv->cqe_factor;
319 320 321 322

	if (!priv->port_up)
		return;

323
	index = cons_index & size_mask;
O
Or Gerlitz 已提交
324
	cqe = &buf[(index << factor) + factor];
325 326 327 328 329 330 331 332 333 334 335 336 337 338
	ring_index = ring->cons & size_mask;

	/* Process all completed CQEs */
	while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
			cons_index & size)) {
		/*
		 * make sure we read the CQE after we read the
		 * ownership bit
		 */
		rmb();

		/* Skip over last polled CQE */
		new_index = be16_to_cpu(cqe->wqe_index) & size_mask;

339 340
		do {
			txbbs_skipped += ring->last_nr_txbb;
341 342
			ring_index = (ring_index + ring->last_nr_txbb) & size_mask;
			/* free next descriptor */
343
			ring->last_nr_txbb = mlx4_en_free_tx_desc(
344 345 346
					priv, ring, ring_index,
					!!((ring->cons + txbbs_skipped) &
							ring->size));
347 348
			packets++;
			bytes += ring->tx_info[ring_index].nr_bytes;
349 350 351 352
		} while (ring_index != new_index);

		++cons_index;
		index = cons_index & size_mask;
O
Or Gerlitz 已提交
353
		cqe = &buf[(index << factor) + factor];
354
	}
355 356 357 358 359 360


	/*
	 * To prevent CQ overflow we first update CQ consumer and only then
	 * the ring consumer.
	 */
361
	mcq->cons_index = cons_index;
362 363 364
	mlx4_cq_set_ci(mcq);
	wmb();
	ring->cons += txbbs_skipped;
365
	netdev_tx_completed_queue(ring->tx_queue, packets, bytes);
366

367 368 369 370 371 372 373
	/*
	 * Wakeup Tx queue if this stopped, and at least 1 packet
	 * was completed
	 */
	if (netif_tx_queue_stopped(ring->tx_queue) && txbbs_skipped > 0) {
		netif_tx_wake_queue(ring->tx_queue);
		priv->port_stats.wake_queue++;
374 375 376 377 378 379 380 381 382
	}
}

void mlx4_en_tx_irq(struct mlx4_cq *mcq)
{
	struct mlx4_en_cq *cq = container_of(mcq, struct mlx4_en_cq, mcq);
	struct mlx4_en_priv *priv = netdev_priv(cq->dev);

	mlx4_en_process_tx_cq(cq->dev, cq);
383
	mlx4_en_arm_cq(priv, cq);
384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
}


static struct mlx4_en_tx_desc *mlx4_en_bounce_to_desc(struct mlx4_en_priv *priv,
						      struct mlx4_en_tx_ring *ring,
						      u32 index,
						      unsigned int desc_size)
{
	u32 copy = (ring->size - index) * TXBB_SIZE;
	int i;

	for (i = desc_size - copy - 4; i >= 0; i -= 4) {
		if ((i & (TXBB_SIZE - 1)) == 0)
			wmb();

		*((u32 *) (ring->buf + i)) =
			*((u32 *) (ring->bounce_buf + copy + i));
	}

	for (i = copy - 4; i >= 4 ; i -= 4) {
		if ((i & (TXBB_SIZE - 1)) == 0)
			wmb();

		*((u32 *) (ring->buf + index * TXBB_SIZE + i)) =
			*((u32 *) (ring->bounce_buf + i));
	}

	/* Return real descriptor location */
	return ring->buf + index * TXBB_SIZE;
}

static int is_inline(struct sk_buff *skb, void **pfrag)
{
	void *ptr;

	if (inline_thold && !skb_is_gso(skb) && skb->len <= inline_thold) {
		if (skb_shinfo(skb)->nr_frags == 1) {
421
			ptr = skb_frag_address_safe(&skb_shinfo(skb)->frags[0]);
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465
			if (unlikely(!ptr))
				return 0;

			if (pfrag)
				*pfrag = ptr;

			return 1;
		} else if (unlikely(skb_shinfo(skb)->nr_frags))
			return 0;
		else
			return 1;
	}

	return 0;
}

static int inline_size(struct sk_buff *skb)
{
	if (skb->len + CTRL_SIZE + sizeof(struct mlx4_wqe_inline_seg)
	    <= MLX4_INLINE_ALIGN)
		return ALIGN(skb->len + CTRL_SIZE +
			     sizeof(struct mlx4_wqe_inline_seg), 16);
	else
		return ALIGN(skb->len + CTRL_SIZE + 2 *
			     sizeof(struct mlx4_wqe_inline_seg), 16);
}

static int get_real_size(struct sk_buff *skb, struct net_device *dev,
			 int *lso_header_size)
{
	struct mlx4_en_priv *priv = netdev_priv(dev);
	int real_size;

	if (skb_is_gso(skb)) {
		*lso_header_size = skb_transport_offset(skb) + tcp_hdrlen(skb);
		real_size = CTRL_SIZE + skb_shinfo(skb)->nr_frags * DS_SIZE +
			ALIGN(*lso_header_size + 4, DS_SIZE);
		if (unlikely(*lso_header_size != skb_headlen(skb))) {
			/* We add a segment for the skb linear buffer only if
			 * it contains data */
			if (*lso_header_size < skb_headlen(skb))
				real_size += DS_SIZE;
			else {
				if (netif_msg_tx_err(priv))
466
					en_warn(priv, "Non-linear headers\n");
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491
				return 0;
			}
		}
	} else {
		*lso_header_size = 0;
		if (!is_inline(skb, NULL))
			real_size = CTRL_SIZE + (skb_shinfo(skb)->nr_frags + 1) * DS_SIZE;
		else
			real_size = inline_size(skb);
	}

	return real_size;
}

static void build_inline_wqe(struct mlx4_en_tx_desc *tx_desc, struct sk_buff *skb,
			     int real_size, u16 *vlan_tag, int tx_ind, void *fragptr)
{
	struct mlx4_wqe_inline_seg *inl = &tx_desc->inl;
	int spc = MLX4_INLINE_ALIGN - CTRL_SIZE - sizeof *inl;

	if (skb->len <= spc) {
		inl->byte_count = cpu_to_be32(1 << 31 | skb->len);
		skb_copy_from_linear_data(skb, inl + 1, skb_headlen(skb));
		if (skb_shinfo(skb)->nr_frags)
			memcpy(((void *)(inl + 1)) + skb_headlen(skb), fragptr,
E
Eric Dumazet 已提交
492
			       skb_frag_size(&skb_shinfo(skb)->frags[0]));
493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511

	} else {
		inl->byte_count = cpu_to_be32(1 << 31 | spc);
		if (skb_headlen(skb) <= spc) {
			skb_copy_from_linear_data(skb, inl + 1, skb_headlen(skb));
			if (skb_headlen(skb) < spc) {
				memcpy(((void *)(inl + 1)) + skb_headlen(skb),
					fragptr, spc - skb_headlen(skb));
				fragptr +=  spc - skb_headlen(skb);
			}
			inl = (void *) (inl + 1) + spc;
			memcpy(((void *)(inl + 1)), fragptr, skb->len - spc);
		} else {
			skb_copy_from_linear_data(skb, inl + 1, spc);
			inl = (void *) (inl + 1) + spc;
			skb_copy_from_linear_data_offset(skb, spc, inl + 1,
					skb_headlen(skb) - spc);
			if (skb_shinfo(skb)->nr_frags)
				memcpy(((void *)(inl + 1)) + skb_headlen(skb) - spc,
E
Eric Dumazet 已提交
512
					fragptr, skb_frag_size(&skb_shinfo(skb)->frags[0]));
513 514 515 516 517 518
		}

		wmb();
		inl->byte_count = cpu_to_be32(1 << 31 | (skb->len - spc));
	}
	tx_desc->ctrl.vlan_tag = cpu_to_be16(*vlan_tag);
519 520
	tx_desc->ctrl.ins_vlan = MLX4_WQE_CTRL_INS_VLAN *
		(!!vlan_tx_tag_present(skb));
521 522 523
	tx_desc->ctrl.fence_size = (real_size / 16) & 0x3f;
}

Y
Yevgeny Petrilin 已提交
524
u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb)
525
{
526
	struct mlx4_en_priv *priv = netdev_priv(dev);
527
	u16 rings_p_up = priv->num_tx_rings_p_up;
528
	u8 up = 0;
529

530 531 532 533 534
	if (dev->num_tc)
		return skb_tx_hash(dev, skb);

	if (vlan_tx_tag_present(skb))
		up = vlan_tx_tag_get(skb) >> VLAN_PRIO_SHIFT;
Y
Yevgeny Petrilin 已提交
535

536
	return __skb_tx_hash(dev, skb, rings_p_up) + up * rings_p_up;
537 538
}

539
static void mlx4_bf_copy(void __iomem *dst, unsigned long *src, unsigned bytecnt)
540 541 542 543
{
	__iowrite64_copy(dst, src, bytecnt / 8);
}

544
netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
545 546 547 548 549 550 551 552
{
	struct mlx4_en_priv *priv = netdev_priv(dev);
	struct mlx4_en_dev *mdev = priv->mdev;
	struct mlx4_en_tx_ring *ring;
	struct mlx4_en_tx_desc *tx_desc;
	struct mlx4_wqe_data_seg *data;
	struct skb_frag_struct *frag;
	struct mlx4_en_tx_info *tx_info;
553
	struct ethhdr *ethh;
554 555 556 557 558
	int tx_ind = 0;
	int nr_txbb;
	int desc_size;
	int real_size;
	dma_addr_t dma;
559
	u32 index, bf_index;
560
	__be32 op_own;
Y
Yevgeny Petrilin 已提交
561
	u16 vlan_tag = 0;
562 563 564
	int i;
	int lso_header_size;
	void *fragptr;
565
	bool bounce = false;
566

567 568 569
	if (!priv->port_up)
		goto tx_drop;

570 571
	real_size = get_real_size(skb, dev, &lso_header_size);
	if (unlikely(!real_size))
572
		goto tx_drop;
573

L
Lucas De Marchi 已提交
574
	/* Align descriptor to TXBB size */
575 576 577 578
	desc_size = ALIGN(real_size, TXBB_SIZE);
	nr_txbb = desc_size / TXBB_SIZE;
	if (unlikely(nr_txbb > MAX_DESC_TXBBS)) {
		if (netif_msg_tx_err(priv))
579
			en_warn(priv, "Oversized header or SG list\n");
580
		goto tx_drop;
581 582
	}

Y
Yevgeny Petrilin 已提交
583
	tx_ind = skb->queue_mapping;
584
	ring = &priv->tx_ring[tx_ind];
585
	if (vlan_tx_tag_present(skb))
Y
Yevgeny Petrilin 已提交
586
		vlan_tag = vlan_tx_tag_get(skb);
587 588 589 590

	/* Check available TXBBs And 2K spare for prefetch */
	if (unlikely(((int)(ring->prod - ring->cons)) >
		     ring->size - HEADROOM - MAX_DESC_TXBBS)) {
Y
Yevgeny Petrilin 已提交
591
		/* every full Tx ring stops queue */
592
		netif_tx_stop_queue(ring->tx_queue);
593 594 595 596 597 598 599 600 601 602 603
		priv->port_stats.queue_stopped++;

		return NETDEV_TX_BUSY;
	}

	/* Track current inflight packets for performance analysis */
	AVG_PERF_COUNTER(priv->pstats.inflight_avg,
			 (u32) (ring->prod - ring->cons - 1));

	/* Packet is good - grab an index and transmit it */
	index = ring->prod & ring->size_mask;
604
	bf_index = ring->prod;
605 606 607 608 609

	/* See if we have enough space for whole descriptor TXBB for setting
	 * SW ownership on next descriptor; if not, use a bounce buffer. */
	if (likely(index + nr_txbb <= ring->size))
		tx_desc = ring->buf + index * TXBB_SIZE;
610
	else {
611
		tx_desc = (struct mlx4_en_tx_desc *) ring->bounce_buf;
612 613
		bounce = true;
	}
614 615 616 617 618 619 620 621 622

	/* Save skb in tx_info ring */
	tx_info = &ring->tx_info[index];
	tx_info->skb = skb;
	tx_info->nr_txbb = nr_txbb;

	/* Prepare ctrl segement apart opcode+ownership, which depends on
	 * whether LSO is used */
	tx_desc->ctrl.vlan_tag = cpu_to_be16(vlan_tag);
623 624
	tx_desc->ctrl.ins_vlan = MLX4_WQE_CTRL_INS_VLAN *
		!!vlan_tx_tag_present(skb);
625
	tx_desc->ctrl.fence_size = (real_size / 16) & 0x3f;
A
Amir Vadai 已提交
626
	tx_desc->ctrl.srcrb_flags = priv->ctrl_flags;
627 628 629
	if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
		tx_desc->ctrl.srcrb_flags |= cpu_to_be32(MLX4_WQE_CTRL_IP_CSUM |
							 MLX4_WQE_CTRL_TCP_UDP_CSUM);
630
		ring->tx_csum++;
631 632
	}

633
	/* Copy dst mac address to wqe */
634
	ethh = (struct ethhdr *)skb->data;
E
Eric Dumazet 已提交
635 636
	tx_desc->ctrl.srcrb_flags16[0] = get_unaligned((__be16 *)ethh->h_dest);
	tx_desc->ctrl.imm = get_unaligned((__be32 *)(ethh->h_dest + 2));
637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
	/* Handle LSO (TSO) packets */
	if (lso_header_size) {
		/* Mark opcode as LSO */
		op_own = cpu_to_be32(MLX4_OPCODE_LSO | (1 << 6)) |
			((ring->prod & ring->size) ?
				cpu_to_be32(MLX4_EN_BIT_DESC_OWN) : 0);

		/* Fill in the LSO prefix */
		tx_desc->lso.mss_hdr_size = cpu_to_be32(
			skb_shinfo(skb)->gso_size << 16 | lso_header_size);

		/* Copy headers;
		 * note that we already verified that it is linear */
		memcpy(tx_desc->lso.header, skb->data, lso_header_size);
		data = ((void *) &tx_desc->lso +
			ALIGN(lso_header_size + 4, DS_SIZE));

		priv->port_stats.tso_packets++;
		i = ((skb->len - lso_header_size) / skb_shinfo(skb)->gso_size) +
			!!((skb->len - lso_header_size) % skb_shinfo(skb)->gso_size);
657
		tx_info->nr_bytes = skb->len + (i - 1) * lso_header_size;
658 659 660 661 662 663 664
		ring->packets += i;
	} else {
		/* Normal (Non LSO) packet */
		op_own = cpu_to_be32(MLX4_OPCODE_SEND) |
			((ring->prod & ring->size) ?
			 cpu_to_be32(MLX4_EN_BIT_DESC_OWN) : 0);
		data = &tx_desc->data;
665
		tx_info->nr_bytes = max_t(unsigned int, skb->len, ETH_ZLEN);
666 667 668
		ring->packets++;

	}
669 670
	ring->bytes += tx_info->nr_bytes;
	netdev_tx_sent_queue(ring->tx_queue, tx_info->nr_bytes);
671 672 673 674 675 676 677 678 679 680 681 682 683
	AVG_PERF_COUNTER(priv->pstats.tx_pktsz_avg, skb->len);


	/* valid only for none inline segments */
	tx_info->data_offset = (void *) data - (void *) tx_desc;

	tx_info->linear = (lso_header_size < skb_headlen(skb) && !is_inline(skb, NULL)) ? 1 : 0;
	data += skb_shinfo(skb)->nr_frags + tx_info->linear - 1;

	if (!is_inline(skb, &fragptr)) {
		/* Map fragments */
		for (i = skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) {
			frag = &skb_shinfo(skb)->frags[i];
684
			dma = skb_frag_dma_map(priv->ddev, frag,
685 686
					       0, skb_frag_size(frag),
					       DMA_TO_DEVICE);
687 688 689
			data->addr = cpu_to_be64(dma);
			data->lkey = cpu_to_be32(mdev->mr.key);
			wmb();
E
Eric Dumazet 已提交
690
			data->byte_count = cpu_to_be32(skb_frag_size(frag));
691 692 693 694 695
			--data;
		}

		/* Map linear part */
		if (tx_info->linear) {
696
			dma = dma_map_single(priv->ddev, skb->data + lso_header_size,
697 698 699 700 701 702
					     skb_headlen(skb) - lso_header_size, PCI_DMA_TODEVICE);
			data->addr = cpu_to_be64(dma);
			data->lkey = cpu_to_be32(mdev->mr.key);
			wmb();
			data->byte_count = cpu_to_be32(skb_headlen(skb) - lso_header_size);
		}
703 704
		tx_info->inl = 0;
	} else {
705
		build_inline_wqe(tx_desc, skb, real_size, &vlan_tag, tx_ind, fragptr);
706 707
		tx_info->inl = 1;
	}
708 709 710 711

	ring->prod += nr_txbb;

	/* If we used a bounce buffer then copy descriptor back into place */
712
	if (bounce)
713 714
		tx_desc = mlx4_en_bounce_to_desc(priv, ring, index, desc_size);

715
	if (ring->bf_enabled && desc_size <= MAX_BF && !bounce && !vlan_tx_tag_present(skb)) {
716
		*(__be32 *) (&tx_desc->ctrl.vlan_tag) |= cpu_to_be32(ring->doorbell_qpn);
717 718 719 720 721
		op_own |= htonl((bf_index & 0xffff) << 8);
		/* Ensure new descirptor hits memory
		* before setting ownership of this descriptor to HW */
		wmb();
		tx_desc->ctrl.owner_opcode = op_own;
722

723 724 725 726 727 728 729 730 731 732 733 734 735 736
		wmb();

		mlx4_bf_copy(ring->bf.reg + ring->bf.offset, (unsigned long *) &tx_desc->ctrl,
		     desc_size);

		wmb();

		ring->bf.offset ^= ring->bf.buf_size;
	} else {
		/* Ensure new descirptor hits memory
		* before setting ownership of this descriptor to HW */
		wmb();
		tx_desc->ctrl.owner_opcode = op_own;
		wmb();
737
		iowrite32be(ring->doorbell_qpn, ring->bf.uar->map + MLX4_SEND_DOORBELL);
738
	}
739

740
	return NETDEV_TX_OK;
741 742 743 744 745

tx_drop:
	dev_kfree_skb_any(skb);
	priv->stats.tx_dropped++;
	return NETDEV_TX_OK;
746 747
}