en_rx.c 29.3 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
/*
 * 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.
 *
 */

34
#include <net/busy_poll.h>
35
#include <linux/mlx4/cq.h>
36
#include <linux/slab.h>
37 38
#include <linux/mlx4/qp.h>
#include <linux/skbuff.h>
39
#include <linux/rculist.h>
40 41 42 43 44 45
#include <linux/if_ether.h>
#include <linux/if_vlan.h>
#include <linux/vmalloc.h>

#include "mlx4_en.h"

46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72
static int mlx4_alloc_pages(struct mlx4_en_priv *priv,
			    struct mlx4_en_rx_alloc *page_alloc,
			    const struct mlx4_en_frag_info *frag_info,
			    gfp_t _gfp)
{
	int order;
	struct page *page;
	dma_addr_t dma;

	for (order = MLX4_EN_ALLOC_PREFER_ORDER; ;) {
		gfp_t gfp = _gfp;

		if (order)
			gfp |= __GFP_COMP | __GFP_NOWARN;
		page = alloc_pages(gfp, order);
		if (likely(page))
			break;
		if (--order < 0 ||
		    ((PAGE_SIZE << order) < frag_info->frag_size))
			return -ENOMEM;
	}
	dma = dma_map_page(priv->ddev, page, 0, PAGE_SIZE << order,
			   PCI_DMA_FROMDEVICE);
	if (dma_mapping_error(priv->ddev, dma)) {
		put_page(page);
		return -ENOMEM;
	}
73
	page_alloc->page_size = PAGE_SIZE << order;
74 75
	page_alloc->page = page;
	page_alloc->dma = dma;
76
	page_alloc->page_offset = frag_info->frag_align;
77 78 79
	/* Not doing get_page() for each frag is a big win
	 * on asymetric workloads.
	 */
80 81
	atomic_set(&page->_count,
		   page_alloc->page_size / frag_info->frag_stride);
82 83 84
	return 0;
}

85 86 87
static int mlx4_en_alloc_frags(struct mlx4_en_priv *priv,
			       struct mlx4_en_rx_desc *rx_desc,
			       struct mlx4_en_rx_alloc *frags,
88 89
			       struct mlx4_en_rx_alloc *ring_alloc,
			       gfp_t gfp)
90
{
91
	struct mlx4_en_rx_alloc page_alloc[MLX4_EN_MAX_RX_FRAGS];
92
	const struct mlx4_en_frag_info *frag_info;
93 94
	struct page *page;
	dma_addr_t dma;
95
	int i;
96

97 98
	for (i = 0; i < priv->num_frags; i++) {
		frag_info = &priv->frag_info[i];
99
		page_alloc[i] = ring_alloc[i];
100 101 102 103
		page_alloc[i].page_offset += frag_info->frag_stride;

		if (page_alloc[i].page_offset + frag_info->frag_stride <=
		    ring_alloc[i].page_size)
104
			continue;
105

106 107
		if (mlx4_alloc_pages(priv, &page_alloc[i], frag_info, gfp))
			goto out;
108
	}
109

110 111
	for (i = 0; i < priv->num_frags; i++) {
		frags[i] = ring_alloc[i];
112
		dma = ring_alloc[i].dma + ring_alloc[i].page_offset;
113 114
		ring_alloc[i] = page_alloc[i];
		rx_desc->data[i].addr = cpu_to_be64(dma);
115
	}
116

117
	return 0;
118 119 120 121

out:
	while (i--) {
		frag_info = &priv->frag_info[i];
122
		if (page_alloc[i].page != ring_alloc[i].page) {
123
			dma_unmap_page(priv->ddev, page_alloc[i].dma,
124
				page_alloc[i].page_size, PCI_DMA_FROMDEVICE);
125 126 127 128
			page = page_alloc[i].page;
			atomic_set(&page->_count, 1);
			put_page(page);
		}
129 130 131 132 133 134 135 136
	}
	return -ENOMEM;
}

static void mlx4_en_free_frag(struct mlx4_en_priv *priv,
			      struct mlx4_en_rx_alloc *frags,
			      int i)
{
137
	const struct mlx4_en_frag_info *frag_info = &priv->frag_info[i];
138
	u32 next_frag_end = frags[i].page_offset + 2 * frag_info->frag_stride;
139

140 141

	if (next_frag_end > frags[i].page_size)
142 143
		dma_unmap_page(priv->ddev, frags[i].dma, frags[i].page_size,
			       PCI_DMA_FROMDEVICE);
144

145 146
	if (frags[i].page)
		put_page(frags[i].page);
147 148 149 150 151 152
}

static int mlx4_en_init_allocator(struct mlx4_en_priv *priv,
				  struct mlx4_en_rx_ring *ring)
{
	int i;
153
	struct mlx4_en_rx_alloc *page_alloc;
154 155

	for (i = 0; i < priv->num_frags; i++) {
156
		const struct mlx4_en_frag_info *frag_info = &priv->frag_info[i];
157

158 159
		if (mlx4_alloc_pages(priv, &ring->page_alloc[i],
				     frag_info, GFP_KERNEL))
160
			goto out;
161 162 163 164 165
	}
	return 0;

out:
	while (i--) {
166 167
		struct page *page;

168
		page_alloc = &ring->page_alloc[i];
169
		dma_unmap_page(priv->ddev, page_alloc->dma,
170
			       page_alloc->page_size, PCI_DMA_FROMDEVICE);
171 172 173
		page = page_alloc->page;
		atomic_set(&page->_count, 1);
		put_page(page);
174 175 176 177 178 179 180 181 182 183 184 185
		page_alloc->page = NULL;
	}
	return -ENOMEM;
}

static void mlx4_en_destroy_allocator(struct mlx4_en_priv *priv,
				      struct mlx4_en_rx_ring *ring)
{
	struct mlx4_en_rx_alloc *page_alloc;
	int i;

	for (i = 0; i < priv->num_frags; i++) {
186 187
		const struct mlx4_en_frag_info *frag_info = &priv->frag_info[i];

188
		page_alloc = &ring->page_alloc[i];
189 190
		en_dbg(DRV, priv, "Freeing allocator:%d count:%d\n",
		       i, page_count(page_alloc->page));
191

192
		dma_unmap_page(priv->ddev, page_alloc->dma,
193 194 195
				page_alloc->page_size, PCI_DMA_FROMDEVICE);
		while (page_alloc->page_offset + frag_info->frag_stride <
		       page_alloc->page_size) {
196
			put_page(page_alloc->page);
197
			page_alloc->page_offset += frag_info->frag_stride;
198
		}
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
		page_alloc->page = NULL;
	}
}

static void mlx4_en_init_rx_desc(struct mlx4_en_priv *priv,
				 struct mlx4_en_rx_ring *ring, int index)
{
	struct mlx4_en_rx_desc *rx_desc = ring->buf + ring->stride * index;
	int possible_frags;
	int i;

	/* Set size and memtype fields */
	for (i = 0; i < priv->num_frags; i++) {
		rx_desc->data[i].byte_count =
			cpu_to_be32(priv->frag_info[i].frag_size);
		rx_desc->data[i].lkey = cpu_to_be32(priv->mdev->mr.key);
	}

	/* If the number of used fragments does not fill up the ring stride,
	 * remaining (unused) fragments must be padded with null address/size
	 * and a special memory key */
	possible_frags = (ring->stride - sizeof(struct mlx4_en_rx_desc)) / DS_SIZE;
	for (i = priv->num_frags; i < possible_frags; i++) {
		rx_desc->data[i].byte_count = 0;
		rx_desc->data[i].lkey = cpu_to_be32(MLX4_EN_MEMTYPE_PAD);
		rx_desc->data[i].addr = 0;
	}
}

static int mlx4_en_prepare_rx_desc(struct mlx4_en_priv *priv,
229 230
				   struct mlx4_en_rx_ring *ring, int index,
				   gfp_t gfp)
231 232
{
	struct mlx4_en_rx_desc *rx_desc = ring->buf + (index * ring->stride);
233 234
	struct mlx4_en_rx_alloc *frags = ring->rx_info +
					(index << priv->log_rx_info);
235

236
	return mlx4_en_alloc_frags(priv, rx_desc, frags, ring->page_alloc, gfp);
237 238 239 240 241 242 243
}

static inline void mlx4_en_update_rx_prod_db(struct mlx4_en_rx_ring *ring)
{
	*ring->wqres.db.db = cpu_to_be32(ring->prod & 0xffff);
}

244 245 246 247
static void mlx4_en_free_rx_desc(struct mlx4_en_priv *priv,
				 struct mlx4_en_rx_ring *ring,
				 int index)
{
248
	struct mlx4_en_rx_alloc *frags;
249 250
	int nr;

251
	frags = ring->rx_info + (index << priv->log_rx_info);
252
	for (nr = 0; nr < priv->num_frags; nr++) {
253
		en_dbg(DRV, priv, "Freeing fragment:%d\n", nr);
254
		mlx4_en_free_frag(priv, frags, nr);
255 256 257
	}
}

258 259 260 261 262
static int mlx4_en_fill_rx_buffers(struct mlx4_en_priv *priv)
{
	struct mlx4_en_rx_ring *ring;
	int ring_ind;
	int buf_ind;
263
	int new_size;
264 265 266

	for (buf_ind = 0; buf_ind < priv->prof->rx_ring_size; buf_ind++) {
		for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {
267
			ring = priv->rx_ring[ring_ind];
268 269

			if (mlx4_en_prepare_rx_desc(priv, ring,
270 271
						    ring->actual_size,
						    GFP_KERNEL)) {
272
				if (ring->actual_size < MLX4_EN_MIN_RX_SIZE) {
273 274
					en_err(priv, "Failed to allocate "
						     "enough rx buffers\n");
275 276
					return -ENOMEM;
				} else {
277
					new_size = rounddown_pow_of_two(ring->actual_size);
278 279 280
					en_warn(priv, "Only %d buffers allocated "
						      "reducing ring size to %d",
						ring->actual_size, new_size);
281
					goto reduce_rings;
282 283 284 285 286 287
				}
			}
			ring->actual_size++;
			ring->prod++;
		}
	}
288 289 290 291
	return 0;

reduce_rings:
	for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {
292
		ring = priv->rx_ring[ring_ind];
293 294 295 296 297 298 299
		while (ring->actual_size > new_size) {
			ring->actual_size--;
			ring->prod--;
			mlx4_en_free_rx_desc(priv, ring, ring->actual_size);
		}
	}

300 301 302 303 304 305 306 307
	return 0;
}

static void mlx4_en_free_rx_buf(struct mlx4_en_priv *priv,
				struct mlx4_en_rx_ring *ring)
{
	int index;

308 309
	en_dbg(DRV, priv, "Freeing Rx buf - cons:%d prod:%d\n",
	       ring->cons, ring->prod);
310 311

	/* Unmap and free Rx buffers */
312
	BUG_ON((u32) (ring->prod - ring->cons) > ring->actual_size);
313 314
	while (ring->cons != ring->prod) {
		index = ring->cons & ring->size_mask;
315
		en_dbg(DRV, priv, "Processing descriptor:%d\n", index);
316
		mlx4_en_free_rx_desc(priv, ring, index);
317 318 319 320 321
		++ring->cons;
	}
}

int mlx4_en_create_rx_ring(struct mlx4_en_priv *priv,
322 323
			   struct mlx4_en_rx_ring **pring,
			   u32 size, u16 stride)
324 325
{
	struct mlx4_en_dev *mdev = priv->mdev;
326
	struct mlx4_en_rx_ring *ring;
327
	int err = -ENOMEM;
328 329
	int tmp;

330 331 332 333 334 335
	ring = kzalloc(sizeof(*ring), GFP_KERNEL);
	if (!ring) {
		en_err(priv, "Failed to allocate RX ring structure\n");
		return -ENOMEM;
	}

336 337 338 339 340 341
	ring->prod = 0;
	ring->cons = 0;
	ring->size = size;
	ring->size_mask = size - 1;
	ring->stride = stride;
	ring->log_stride = ffs(ring->stride) - 1;
342
	ring->buf_size = ring->size * ring->stride + TXBB_SIZE;
343 344

	tmp = size * roundup_pow_of_two(MLX4_EN_MAX_RX_FRAGS *
345
					sizeof(struct mlx4_en_rx_alloc));
346
	ring->rx_info = vmalloc(tmp);
347 348 349 350
	if (!ring->rx_info) {
		err = -ENOMEM;
		goto err_ring;
	}
351

352
	en_dbg(DRV, priv, "Allocated rx_info ring at addr:%p size:%d\n",
353 354 355 356 357
		 ring->rx_info, tmp);

	err = mlx4_alloc_hwq_res(mdev->dev, &ring->wqres,
				 ring->buf_size, 2 * PAGE_SIZE);
	if (err)
358
		goto err_info;
359 360 361

	err = mlx4_en_map_buffer(&ring->wqres.buf);
	if (err) {
362
		en_err(priv, "Failed to map RX buffer\n");
363 364 365 366
		goto err_hwq;
	}
	ring->buf = ring->wqres.buf.direct.buf;

367 368
	ring->hwtstamp_rx_filter = priv->hwtstamp_config.rx_filter;

369
	*pring = ring;
370 371 372 373
	return 0;

err_hwq:
	mlx4_free_hwq_res(mdev->dev, &ring->wqres, ring->buf_size);
374
err_info:
375 376
	vfree(ring->rx_info);
	ring->rx_info = NULL;
377 378 379 380
err_ring:
	kfree(ring);
	*pring = NULL;

381 382 383 384 385 386 387 388 389 390 391 392 393
	return err;
}

int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv)
{
	struct mlx4_en_rx_ring *ring;
	int i;
	int ring_ind;
	int err;
	int stride = roundup_pow_of_two(sizeof(struct mlx4_en_rx_desc) +
					DS_SIZE * priv->num_frags);

	for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {
394
		ring = priv->rx_ring[ring_ind];
395 396 397 398

		ring->prod = 0;
		ring->cons = 0;
		ring->actual_size = 0;
399
		ring->cqn = priv->rx_cq[ring_ind]->mcq.cqn;
400 401

		ring->stride = stride;
402 403 404
		if (ring->stride <= TXBB_SIZE)
			ring->buf += TXBB_SIZE;

405 406 407 408 409 410
		ring->log_stride = ffs(ring->stride) - 1;
		ring->buf_size = ring->size * ring->stride;

		memset(ring->buf, 0, ring->buf_size);
		mlx4_en_update_rx_prod_db(ring);

411
		/* Initialize all descriptors */
412 413 414 415 416 417
		for (i = 0; i < ring->size; i++)
			mlx4_en_init_rx_desc(priv, ring, i);

		/* Initialize page allocators */
		err = mlx4_en_init_allocator(priv, ring);
		if (err) {
418
			en_err(priv, "Failed initializing ring allocator\n");
419 420
			if (ring->stride <= TXBB_SIZE)
				ring->buf -= TXBB_SIZE;
421 422
			ring_ind--;
			goto err_allocator;
423 424
		}
	}
425 426
	err = mlx4_en_fill_rx_buffers(priv);
	if (err)
427 428 429
		goto err_buffers;

	for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++) {
430
		ring = priv->rx_ring[ring_ind];
431

432
		ring->size_mask = ring->actual_size - 1;
433 434 435 436 437 438 439
		mlx4_en_update_rx_prod_db(ring);
	}

	return 0;

err_buffers:
	for (ring_ind = 0; ring_ind < priv->rx_ring_num; ring_ind++)
440
		mlx4_en_free_rx_buf(priv, priv->rx_ring[ring_ind]);
441 442 443 444

	ring_ind = priv->rx_ring_num - 1;
err_allocator:
	while (ring_ind >= 0) {
445 446 447
		if (priv->rx_ring[ring_ind]->stride <= TXBB_SIZE)
			priv->rx_ring[ring_ind]->buf -= TXBB_SIZE;
		mlx4_en_destroy_allocator(priv, priv->rx_ring[ring_ind]);
448 449 450 451 452 453
		ring_ind--;
	}
	return err;
}

void mlx4_en_destroy_rx_ring(struct mlx4_en_priv *priv,
454 455
			     struct mlx4_en_rx_ring **pring,
			     u32 size, u16 stride)
456 457
{
	struct mlx4_en_dev *mdev = priv->mdev;
458
	struct mlx4_en_rx_ring *ring = *pring;
459 460

	mlx4_en_unmap_buffer(&ring->wqres.buf);
461
	mlx4_free_hwq_res(mdev->dev, &ring->wqres, size * stride + TXBB_SIZE);
462 463
	vfree(ring->rx_info);
	ring->rx_info = NULL;
464 465
	kfree(ring);
	*pring = NULL;
466
#ifdef CONFIG_RFS_ACCEL
467
	mlx4_en_cleanup_filters(priv);
468
#endif
469 470 471 472 473 474
}

void mlx4_en_deactivate_rx_ring(struct mlx4_en_priv *priv,
				struct mlx4_en_rx_ring *ring)
{
	mlx4_en_free_rx_buf(priv, ring);
475 476
	if (ring->stride <= TXBB_SIZE)
		ring->buf -= TXBB_SIZE;
477 478 479 480 481 482
	mlx4_en_destroy_allocator(priv, ring);
}


static int mlx4_en_complete_rx_desc(struct mlx4_en_priv *priv,
				    struct mlx4_en_rx_desc *rx_desc,
483
				    struct mlx4_en_rx_alloc *frags,
484
				    struct sk_buff *skb,
485 486
				    int length)
{
487
	struct skb_frag_struct *skb_frags_rx = skb_shinfo(skb)->frags;
488 489 490 491
	struct mlx4_en_frag_info *frag_info;
	int nr;
	dma_addr_t dma;

492
	/* Collect used fragments while replacing them in the HW descriptors */
493 494 495 496
	for (nr = 0; nr < priv->num_frags; nr++) {
		frag_info = &priv->frag_info[nr];
		if (length <= frag_info->frag_prefix_size)
			break;
497 498
		if (!frags[nr].page)
			goto fail;
499 500

		dma = be64_to_cpu(rx_desc->data[nr].addr);
501 502
		dma_sync_single_for_cpu(priv->ddev, dma, frag_info->frag_size,
					DMA_FROM_DEVICE);
503

504 505 506
		/* Save page reference in skb */
		__skb_frag_set_page(&skb_frags_rx[nr], frags[nr].page);
		skb_frag_size_set(&skb_frags_rx[nr], frag_info->frag_size);
507
		skb_frags_rx[nr].page_offset = frags[nr].page_offset;
508
		skb->truesize += frag_info->frag_stride;
509
		frags[nr].page = NULL;
510 511
	}
	/* Adjust size of last fragment to match actual length */
512
	if (nr > 0)
E
Eric Dumazet 已提交
513 514
		skb_frag_size_set(&skb_frags_rx[nr - 1],
			length - priv->frag_info[nr - 1].frag_prefix_size);
515 516 517 518 519
	return nr;

fail:
	while (nr > 0) {
		nr--;
520
		__skb_frag_unref(&skb_frags_rx[nr]);
521 522 523 524 525 526 527
	}
	return 0;
}


static struct sk_buff *mlx4_en_rx_skb(struct mlx4_en_priv *priv,
				      struct mlx4_en_rx_desc *rx_desc,
528
				      struct mlx4_en_rx_alloc *frags,
529 530 531 532 533 534 535
				      unsigned int length)
{
	struct sk_buff *skb;
	void *va;
	int used_frags;
	dma_addr_t dma;

536
	skb = netdev_alloc_skb(priv->dev, SMALL_PACKET_SIZE + NET_IP_ALIGN);
537
	if (!skb) {
538
		en_dbg(RX_ERR, priv, "Failed allocating skb\n");
539 540 541 542 543 544 545
		return NULL;
	}
	skb_reserve(skb, NET_IP_ALIGN);
	skb->len = length;

	/* Get pointer to first fragment so we could copy the headers into the
	 * (linear part of the) skb */
546
	va = page_address(frags[0].page) + frags[0].page_offset;
547 548 549

	if (length <= SMALL_PACKET_SIZE) {
		/* We are copying all relevant data to the skb - temporarily
550
		 * sync buffers for the copy */
551
		dma = be64_to_cpu(rx_desc->data[0].addr);
552
		dma_sync_single_for_cpu(priv->ddev, dma, length,
553
					DMA_FROM_DEVICE);
554 555 556 557
		skb_copy_to_linear_data(skb, va, length);
		skb->tail += length;
	} else {
		/* Move relevant fragments to skb */
558 559
		used_frags = mlx4_en_complete_rx_desc(priv, rx_desc, frags,
							skb, length);
560 561 562 563
		if (unlikely(!used_frags)) {
			kfree_skb(skb);
			return NULL;
		}
564 565 566 567 568 569 570 571 572 573
		skb_shinfo(skb)->nr_frags = used_frags;

		/* Copy headers into the skb linear buffer */
		memcpy(skb->data, va, HEADER_COPY_SIZE);
		skb->tail += HEADER_COPY_SIZE;

		/* Skip headers in first fragment */
		skb_shinfo(skb)->frags[0].page_offset += HEADER_COPY_SIZE;

		/* Adjust size of first fragment */
E
Eric Dumazet 已提交
574
		skb_frag_size_sub(&skb_shinfo(skb)->frags[0], HEADER_COPY_SIZE);
575 576 577 578 579
		skb->data_len = length - HEADER_COPY_SIZE;
	}
	return skb;
}

580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
static void validate_loopback(struct mlx4_en_priv *priv, struct sk_buff *skb)
{
	int i;
	int offset = ETH_HLEN;

	for (i = 0; i < MLX4_LOOPBACK_TEST_PAYLOAD; i++, offset++) {
		if (*(skb->data + offset) != (unsigned char) (i & 0xff))
			goto out_loopback;
	}
	/* Loopback found */
	priv->loopback_ok = 1;

out_loopback:
	dev_kfree_skb_any(skb);
}
595

596 597 598 599 600 601
static void mlx4_en_refill_rx_buffers(struct mlx4_en_priv *priv,
				     struct mlx4_en_rx_ring *ring)
{
	int index = ring->prod & ring->size_mask;

	while ((u32) (ring->prod - ring->cons) < ring->actual_size) {
602
		if (mlx4_en_prepare_rx_desc(priv, ring, index, GFP_ATOMIC))
603 604 605 606 607 608
			break;
		ring->prod++;
		index = ring->prod & ring->size_mask;
	}
}

609 610 611
int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int budget)
{
	struct mlx4_en_priv *priv = netdev_priv(dev);
612
	struct mlx4_en_dev *mdev = priv->mdev;
613
	struct mlx4_cqe *cqe;
614
	struct mlx4_en_rx_ring *ring = priv->rx_ring[cq->ring];
615
	struct mlx4_en_rx_alloc *frags;
616 617 618 619 620 621 622
	struct mlx4_en_rx_desc *rx_desc;
	struct sk_buff *skb;
	int index;
	int nr;
	unsigned int length;
	int polled = 0;
	int ip_summed;
O
Or Gerlitz 已提交
623
	int factor = priv->cqe_factor;
624
	u64 timestamp;
625 626 627 628 629 630 631 632

	if (!priv->port_up)
		return 0;

	/* We assume a 1:1 mapping between CQEs and Rx descriptors, so Rx
	 * descriptor offset can be deduced from the CQE index instead of
	 * reading 'cqe->index' */
	index = cq->mcq.cons_index & ring->size_mask;
O
Or Gerlitz 已提交
633
	cqe = &cq->buf[(index << factor) + factor];
634 635 636 637 638

	/* Process all completed CQEs */
	while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
		    cq->mcq.cons_index & cq->size)) {

639
		frags = ring->rx_info + (index << priv->log_rx_info);
640 641 642 643 644 645 646 647 648 649
		rx_desc = ring->buf + (index << ring->log_stride);

		/*
		 * make sure we read the CQE after we read the ownership bit
		 */
		rmb();

		/* Drop packet on bad receive or bad checksum */
		if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
						MLX4_CQE_OPCODE_ERROR)) {
650
			en_err(priv, "CQE completed in error - vendor "
651 652 653 654 655 656
				  "syndrom:%d syndrom:%d\n",
				  ((struct mlx4_err_cqe *) cqe)->vendor_err_syndrome,
				  ((struct mlx4_err_cqe *) cqe)->syndrome);
			goto next;
		}
		if (unlikely(cqe->badfcs_enc & MLX4_CQE_BAD_FCS)) {
657
			en_dbg(RX_ERR, priv, "Accepted frame with bad FCS\n");
658 659 660
			goto next;
		}

661 662 663 664 665 666 667 668 669 670 671 672 673
		/* Check if we need to drop the packet if SRIOV is not enabled
		 * and not performing the selftest or flb disabled
		 */
		if (priv->flags & MLX4_EN_FLAG_RX_FILTER_NEEDED) {
			struct ethhdr *ethh;
			dma_addr_t dma;
			/* Get pointer to first fragment since we haven't
			 * skb yet and cast it to ethhdr struct
			 */
			dma = be64_to_cpu(rx_desc->data[0].addr);
			dma_sync_single_for_cpu(priv->ddev, dma, sizeof(*ethh),
						DMA_FROM_DEVICE);
			ethh = (struct ethhdr *)(page_address(frags[0].page) +
674
						 frags[0].page_offset);
675

676 677 678 679 680 681 682 683 684
			if (is_multicast_ether_addr(ethh->h_dest)) {
				struct mlx4_mac_entry *entry;
				struct hlist_head *bucket;
				unsigned int mac_hash;

				/* Drop the packet, since HW loopback-ed it */
				mac_hash = ethh->h_source[MLX4_EN_MAC_HASH_IDX];
				bucket = &priv->mac_hash[mac_hash];
				rcu_read_lock();
685
				hlist_for_each_entry_rcu(entry, bucket, hlist) {
686 687 688 689 690 691 692 693
					if (ether_addr_equal_64bits(entry->mac,
								    ethh->h_source)) {
						rcu_read_unlock();
						goto next;
					}
				}
				rcu_read_unlock();
			}
694
		}
695

696 697 698 699
		/*
		 * Packet is OK - process it.
		 */
		length = be32_to_cpu(cqe->byte_cnt);
700
		length -= ring->fcs_del;
701 702 703
		ring->bytes += length;
		ring->packets++;

704
		if (likely(dev->features & NETIF_F_RXCSUM)) {
705 706
			if ((cqe->status & cpu_to_be16(MLX4_CQE_STATUS_IPOK)) &&
			    (cqe->checksum == cpu_to_be16(0xffff))) {
707
				ring->csum_ok++;
708
				/* This packet is eligible for GRO if it is:
709 710 711
				 * - DIX Ethernet (type interpretation)
				 * - TCP/IP (v4)
				 * - without IP options
712 713 714 715 716
				 * - not an IP fragment
				 * - no LLS polling in progress
				 */
				if (!mlx4_en_cq_ll_polling(cq) &&
				    (dev->features & NETIF_F_GRO)) {
717
					struct sk_buff *gro_skb = napi_get_frags(&cq->napi);
718 719
					if (!gro_skb)
						goto next;
720

721 722 723
					nr = mlx4_en_complete_rx_desc(priv,
						rx_desc, frags, gro_skb,
						length);
724 725 726
					if (!nr)
						goto next;

727 728 729 730 731
					skb_shinfo(gro_skb)->nr_frags = nr;
					gro_skb->len = length;
					gro_skb->data_len = length;
					gro_skb->ip_summed = CHECKSUM_UNNECESSARY;

732 733 734
					if ((cqe->vlan_my_qpn &
					    cpu_to_be32(MLX4_CQE_VLAN_PRESENT_MASK)) &&
					    (dev->features & NETIF_F_HW_VLAN_CTAG_RX)) {
J
Jiri Pirko 已提交
735 736
						u16 vid = be16_to_cpu(cqe->sl_vid);

737
						__vlan_hwaccel_put_tag(gro_skb, htons(ETH_P_8021Q), vid);
J
Jiri Pirko 已提交
738 739
					}

Y
Yevgeny Petrilin 已提交
740 741 742
					if (dev->features & NETIF_F_RXHASH)
						gro_skb->rxhash = be32_to_cpu(cqe->immed_rss_invalid);

743
					skb_record_rx_queue(gro_skb, cq->ring);
744

745 746 747 748 749 750 751 752
					if (ring->hwtstamp_rx_filter == HWTSTAMP_FILTER_ALL) {
						timestamp = mlx4_en_get_cqe_ts(cqe);
						mlx4_en_fill_hwtstamps(mdev,
								       skb_hwtstamps(gro_skb),
								       timestamp);
					}

					napi_gro_frags(&cq->napi);
753 754 755
					goto next;
				}

756
				/* GRO not possible, complete processing here */
757 758 759
				ip_summed = CHECKSUM_UNNECESSARY;
			} else {
				ip_summed = CHECKSUM_NONE;
760
				ring->csum_none++;
761 762 763
			}
		} else {
			ip_summed = CHECKSUM_NONE;
764
			ring->csum_none++;
765 766
		}

767
		skb = mlx4_en_rx_skb(priv, rx_desc, frags, length);
768 769 770 771 772
		if (!skb) {
			priv->stats.rx_dropped++;
			goto next;
		}

773 774 775 776 777
                if (unlikely(priv->validate_loopback)) {
			validate_loopback(priv, skb);
			goto next;
		}

778 779
		skb->ip_summed = ip_summed;
		skb->protocol = eth_type_trans(skb, dev);
780
		skb_record_rx_queue(skb, cq->ring);
781

Y
Yevgeny Petrilin 已提交
782 783 784
		if (dev->features & NETIF_F_RXHASH)
			skb->rxhash = be32_to_cpu(cqe->immed_rss_invalid);

785 786 787
		if ((be32_to_cpu(cqe->vlan_my_qpn) &
		    MLX4_CQE_VLAN_PRESENT_MASK) &&
		    (dev->features & NETIF_F_HW_VLAN_CTAG_RX))
788
			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), be16_to_cpu(cqe->sl_vid));
J
Jiri Pirko 已提交
789

790 791 792 793 794 795
		if (ring->hwtstamp_rx_filter == HWTSTAMP_FILTER_ALL) {
			timestamp = mlx4_en_get_cqe_ts(cqe);
			mlx4_en_fill_hwtstamps(mdev, skb_hwtstamps(skb),
					       timestamp);
		}

796
		skb_mark_napi_id(skb, &cq->napi);
797

798
		/* Push it up the stack */
J
Jiri Pirko 已提交
799
		netif_receive_skb(skb);
800 801

next:
802 803 804
		for (nr = 0; nr < priv->num_frags; nr++)
			mlx4_en_free_frag(priv, frags, nr);

805 806
		++cq->mcq.cons_index;
		index = (cq->mcq.cons_index) & ring->size_mask;
O
Or Gerlitz 已提交
807
		cqe = &cq->buf[(index << factor) + factor];
808
		if (++polled == budget)
809 810 811 812 813 814 815 816
			goto out;
	}

out:
	AVG_PERF_COUNTER(priv->pstats.rx_coal_avg, polled);
	mlx4_cq_set_ci(&cq->mcq);
	wmb(); /* ensure HW sees CQ consumer before we post new buffers */
	ring->cons = cq->mcq.cons_index;
817
	mlx4_en_refill_rx_buffers(priv, ring);
818 819 820 821 822 823 824 825 826 827 828
	mlx4_en_update_rx_prod_db(ring);
	return polled;
}


void mlx4_en_rx_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);

	if (priv->port_up)
829
		napi_schedule(&cq->napi);
830 831 832 833 834 835 836 837 838 839 840 841
	else
		mlx4_en_arm_cq(priv, cq);
}

/* Rx CQ polling - called by NAPI */
int mlx4_en_poll_rx_cq(struct napi_struct *napi, int budget)
{
	struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
	struct net_device *dev = cq->dev;
	struct mlx4_en_priv *priv = netdev_priv(dev);
	int done;

842 843 844
	if (!mlx4_en_cq_lock_napi(cq))
		return budget;

845 846
	done = mlx4_en_process_rx_cq(dev, cq, budget);

847 848
	mlx4_en_cq_unlock_napi(cq);

849 850 851 852 853
	/* If we used up all the quota - we're probably not done yet... */
	if (done == budget)
		INC_PERF_COUNTER(priv->pstats.napi_quota);
	else {
		/* Done for now */
854
		napi_complete(napi);
855 856 857 858 859
		mlx4_en_arm_cq(priv, cq);
	}
	return done;
}

860
static const int frag_sizes[] = {
861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893
	FRAG_SZ0,
	FRAG_SZ1,
	FRAG_SZ2,
	FRAG_SZ3
};

void mlx4_en_calc_rx_buf(struct net_device *dev)
{
	struct mlx4_en_priv *priv = netdev_priv(dev);
	int eff_mtu = dev->mtu + ETH_HLEN + VLAN_HLEN + ETH_LLC_SNAP_SIZE;
	int buf_size = 0;
	int i = 0;

	while (buf_size < eff_mtu) {
		priv->frag_info[i].frag_size =
			(eff_mtu > buf_size + frag_sizes[i]) ?
				frag_sizes[i] : eff_mtu - buf_size;
		priv->frag_info[i].frag_prefix_size = buf_size;
		if (!i)	{
			priv->frag_info[i].frag_align = NET_IP_ALIGN;
			priv->frag_info[i].frag_stride =
				ALIGN(frag_sizes[i] + NET_IP_ALIGN, SMP_CACHE_BYTES);
		} else {
			priv->frag_info[i].frag_align = 0;
			priv->frag_info[i].frag_stride =
				ALIGN(frag_sizes[i], SMP_CACHE_BYTES);
		}
		buf_size += priv->frag_info[i].frag_size;
		i++;
	}

	priv->num_frags = i;
	priv->rx_skb_size = eff_mtu;
894
	priv->log_rx_info = ROUNDUP_LOG2(i * sizeof(struct mlx4_en_rx_alloc));
895

896
	en_dbg(DRV, priv, "Rx buffer scatter-list (effective-mtu:%d "
897 898
		  "num_frags:%d):\n", eff_mtu, priv->num_frags);
	for (i = 0; i < priv->num_frags; i++) {
899 900 901 902 903 904 905
		en_err(priv,
		       "  frag:%d - size:%d prefix:%d align:%d stride:%d\n",
		       i,
		       priv->frag_info[i].frag_size,
		       priv->frag_info[i].frag_prefix_size,
		       priv->frag_info[i].frag_align,
		       priv->frag_info[i].frag_stride);
906 907 908 909 910
	}
}

/* RSS related functions */

911 912
static int mlx4_en_config_rss_qp(struct mlx4_en_priv *priv, int qpn,
				 struct mlx4_en_rx_ring *ring,
913 914 915 916 917 918 919
				 enum mlx4_qp_state *state,
				 struct mlx4_qp *qp)
{
	struct mlx4_en_dev *mdev = priv->mdev;
	struct mlx4_qp_context *context;
	int err = 0;

920 921
	context = kmalloc(sizeof(*context), GFP_KERNEL);
	if (!context)
922 923 924 925
		return -ENOMEM;

	err = mlx4_qp_alloc(mdev->dev, qpn, qp);
	if (err) {
926
		en_err(priv, "Failed to allocate qp #%x\n", qpn);
927 928 929 930 931
		goto out;
	}
	qp->event = mlx4_en_sqp_event;

	memset(context, 0, sizeof *context);
932
	mlx4_en_fill_qp_context(priv, ring->actual_size, ring->stride, 0, 0,
933
				qpn, ring->cqn, -1, context);
934
	context->db_rec_addr = cpu_to_be64(ring->wqres.db.dma);
935

936
	/* Cancel FCS removal if FW allows */
937
	if (mdev->dev->caps.flags & MLX4_DEV_CAP_FLAG_FCS_KEEP) {
938
		context->param3 |= cpu_to_be32(1 << 29);
939 940 941
		ring->fcs_del = ETH_FCS_LEN;
	} else
		ring->fcs_del = 0;
942

943
	err = mlx4_qp_to_ready(mdev->dev, &ring->wqres.mtt, context, qp, state);
944 945 946 947
	if (err) {
		mlx4_qp_remove(mdev->dev, qp);
		mlx4_qp_free(mdev->dev, qp);
	}
948
	mlx4_en_update_rx_prod_db(ring);
949 950 951 952 953
out:
	kfree(context);
	return err;
}

954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
int mlx4_en_create_drop_qp(struct mlx4_en_priv *priv)
{
	int err;
	u32 qpn;

	err = mlx4_qp_reserve_range(priv->mdev->dev, 1, 1, &qpn);
	if (err) {
		en_err(priv, "Failed reserving drop qpn\n");
		return err;
	}
	err = mlx4_qp_alloc(priv->mdev->dev, qpn, &priv->drop_qp);
	if (err) {
		en_err(priv, "Failed allocating drop qp\n");
		mlx4_qp_release_range(priv->mdev->dev, qpn, 1);
		return err;
	}

	return 0;
}

void mlx4_en_destroy_drop_qp(struct mlx4_en_priv *priv)
{
	u32 qpn;

	qpn = priv->drop_qp.qpn;
	mlx4_qp_remove(priv->mdev->dev, &priv->drop_qp);
	mlx4_qp_free(priv->mdev->dev, &priv->drop_qp);
	mlx4_qp_release_range(priv->mdev->dev, qpn, 1);
}

984 985 986 987 988 989
/* Allocate rx qp's and configure them according to rss map */
int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
{
	struct mlx4_en_dev *mdev = priv->mdev;
	struct mlx4_en_rss_map *rss_map = &priv->rss_map;
	struct mlx4_qp_context context;
990
	struct mlx4_rss_context *rss_context;
991
	int rss_rings;
992
	void *ptr;
993
	u8 rss_mask = (MLX4_RSS_IPV4 | MLX4_RSS_TCP_IPV4 | MLX4_RSS_IPV6 |
994
			MLX4_RSS_TCP_IPV6);
995
	int i, qpn;
996 997
	int err = 0;
	int good_qps = 0;
Y
Yevgeny Petrilin 已提交
998 999 1000
	static const u32 rsskey[10] = { 0xD181C62C, 0xF7F4DB5B, 0x1983A2FC,
				0x943E1ADB, 0xD9389E6B, 0xD1039C2C, 0xA74499AD,
				0x593D56D9, 0xF3253C06, 0x2ADC1FFC};
1001

1002
	en_dbg(DRV, priv, "Configuring rss steering\n");
1003 1004 1005
	err = mlx4_qp_reserve_range(mdev->dev, priv->rx_ring_num,
				    priv->rx_ring_num,
				    &rss_map->base_qpn);
1006
	if (err) {
1007
		en_err(priv, "Failed reserving %d qps\n", priv->rx_ring_num);
1008 1009 1010
		return err;
	}

1011
	for (i = 0; i < priv->rx_ring_num; i++) {
1012
		qpn = rss_map->base_qpn + i;
1013
		err = mlx4_en_config_rss_qp(priv, qpn, priv->rx_ring[i],
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
					    &rss_map->state[i],
					    &rss_map->qps[i]);
		if (err)
			goto rss_err;

		++good_qps;
	}

	/* Configure RSS indirection qp */
	err = mlx4_qp_alloc(mdev->dev, priv->base_qpn, &rss_map->indir_qp);
	if (err) {
1025
		en_err(priv, "Failed to allocate RSS indirection QP\n");
1026
		goto rss_err;
1027 1028 1029
	}
	rss_map->indir_qp.event = mlx4_en_sqp_event;
	mlx4_en_fill_qp_context(priv, 0, 0, 0, 1, priv->base_qpn,
1030
				priv->rx_ring[0]->cqn, -1, &context);
1031

1032 1033 1034 1035 1036
	if (!priv->prof->rss_rings || priv->prof->rss_rings > priv->rx_ring_num)
		rss_rings = priv->rx_ring_num;
	else
		rss_rings = priv->prof->rss_rings;

1037 1038
	ptr = ((void *) &context) + offsetof(struct mlx4_qp_context, pri_path)
					+ MLX4_RSS_OFFSET_IN_QPC_PRI_PATH;
1039
	rss_context = ptr;
1040
	rss_context->base_qpn = cpu_to_be32(ilog2(rss_rings) << 24 |
1041
					    (rss_map->base_qpn));
1042
	rss_context->default_qpn = cpu_to_be32(rss_map->base_qpn);
1043 1044 1045 1046
	if (priv->mdev->profile.udp_rss) {
		rss_mask |=  MLX4_RSS_UDP_IPV4 | MLX4_RSS_UDP_IPV6;
		rss_context->base_qpn_udp = rss_context->default_qpn;
	}
Y
Yevgeny Petrilin 已提交
1047
	rss_context->flags = rss_mask;
1048
	rss_context->hash_fn = MLX4_RSS_HASH_TOP;
Y
Yevgeny Petrilin 已提交
1049
	for (i = 0; i < 10; i++)
1050
		rss_context->rss_key[i] = cpu_to_be32(rsskey[i]);
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070

	err = mlx4_qp_to_ready(mdev->dev, &priv->res.mtt, &context,
			       &rss_map->indir_qp, &rss_map->indir_state);
	if (err)
		goto indir_err;

	return 0;

indir_err:
	mlx4_qp_modify(mdev->dev, NULL, rss_map->indir_state,
		       MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->indir_qp);
	mlx4_qp_remove(mdev->dev, &rss_map->indir_qp);
	mlx4_qp_free(mdev->dev, &rss_map->indir_qp);
rss_err:
	for (i = 0; i < good_qps; i++) {
		mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i],
			       MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->qps[i]);
		mlx4_qp_remove(mdev->dev, &rss_map->qps[i]);
		mlx4_qp_free(mdev->dev, &rss_map->qps[i]);
	}
1071
	mlx4_qp_release_range(mdev->dev, rss_map->base_qpn, priv->rx_ring_num);
1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
	return err;
}

void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv)
{
	struct mlx4_en_dev *mdev = priv->mdev;
	struct mlx4_en_rss_map *rss_map = &priv->rss_map;
	int i;

	mlx4_qp_modify(mdev->dev, NULL, rss_map->indir_state,
		       MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->indir_qp);
	mlx4_qp_remove(mdev->dev, &rss_map->indir_qp);
	mlx4_qp_free(mdev->dev, &rss_map->indir_qp);

1086
	for (i = 0; i < priv->rx_ring_num; i++) {
1087 1088 1089 1090 1091
		mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i],
			       MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->qps[i]);
		mlx4_qp_remove(mdev->dev, &rss_map->qps[i]);
		mlx4_qp_free(mdev->dev, &rss_map->qps[i]);
	}
1092
	mlx4_qp_release_range(mdev->dev, rss_map->base_qpn, priv->rx_ring_num);
1093
}