fm10k_main.c 53.8 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2
/* Copyright(c) 2013 - 2018 Intel Corporation. */
3 4 5 6 7 8 9

#include <linux/types.h>
#include <linux/module.h>
#include <net/ipv6.h>
#include <net/ip.h>
#include <net/tcp.h>
#include <linux/if_macvlan.h>
10
#include <linux/prefetch.h>
11 12 13

#include "fm10k.h"

J
Jacob Keller 已提交
14
#define DRV_VERSION	"0.23.4-k"
15
#define DRV_SUMMARY	"Intel(R) Ethernet Switch Host Interface Driver"
16 17
const char fm10k_driver_version[] = DRV_VERSION;
char fm10k_driver_name[] = "fm10k";
18
static const char fm10k_driver_string[] = DRV_SUMMARY;
19
static const char fm10k_copyright[] =
J
Jacob Keller 已提交
20
	"Copyright(c) 2013 - 2018 Intel Corporation.";
21 22

MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
23
MODULE_DESCRIPTION(DRV_SUMMARY);
24 25 26
MODULE_LICENSE("GPL");
MODULE_VERSION(DRV_VERSION);

27
/* single workqueue for entire fm10k driver */
28
struct workqueue_struct *fm10k_workqueue;
29

30 31
/**
 * fm10k_init_module - Driver Registration Routine
32 33 34 35 36 37 38 39 40
 *
 * fm10k_init_module is the first routine called when the driver is
 * loaded.  All it does is register with the PCI subsystem.
 **/
static int __init fm10k_init_module(void)
{
	pr_info("%s - version %s\n", fm10k_driver_string, fm10k_driver_version);
	pr_info("%s\n", fm10k_copyright);

41
	/* create driver workqueue */
42 43
	fm10k_workqueue = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0,
					  fm10k_driver_name);
44 45
	if (!fm10k_workqueue)
		return -ENOMEM;
46

A
Alexander Duyck 已提交
47 48
	fm10k_dbg_init();

49 50 51 52 53 54 55 56 57 58 59 60 61
	return fm10k_register_pci_driver();
}
module_init(fm10k_init_module);

/**
 * fm10k_exit_module - Driver Exit Cleanup Routine
 *
 * fm10k_exit_module is called just before the driver is removed
 * from memory.
 **/
static void __exit fm10k_exit_module(void)
{
	fm10k_unregister_pci_driver();
A
Alexander Duyck 已提交
62 63

	fm10k_dbg_exit();
64 65 66

	/* destroy driver workqueue */
	destroy_workqueue(fm10k_workqueue);
67 68
}
module_exit(fm10k_exit_module);
A
Alexander Duyck 已提交
69

70 71 72 73 74 75 76 77 78 79 80
static bool fm10k_alloc_mapped_page(struct fm10k_ring *rx_ring,
				    struct fm10k_rx_buffer *bi)
{
	struct page *page = bi->page;
	dma_addr_t dma;

	/* Only page will be NULL if buffer was consumed */
	if (likely(page))
		return true;

	/* alloc new page for storage */
81
	page = dev_alloc_page();
82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143
	if (unlikely(!page)) {
		rx_ring->rx_stats.alloc_failed++;
		return false;
	}

	/* map page for use */
	dma = dma_map_page(rx_ring->dev, page, 0, PAGE_SIZE, DMA_FROM_DEVICE);

	/* if mapping failed free memory back to system since
	 * there isn't much point in holding memory we can't use
	 */
	if (dma_mapping_error(rx_ring->dev, dma)) {
		__free_page(page);

		rx_ring->rx_stats.alloc_failed++;
		return false;
	}

	bi->dma = dma;
	bi->page = page;
	bi->page_offset = 0;

	return true;
}

/**
 * fm10k_alloc_rx_buffers - Replace used receive buffers
 * @rx_ring: ring to place buffers on
 * @cleaned_count: number of buffers to replace
 **/
void fm10k_alloc_rx_buffers(struct fm10k_ring *rx_ring, u16 cleaned_count)
{
	union fm10k_rx_desc *rx_desc;
	struct fm10k_rx_buffer *bi;
	u16 i = rx_ring->next_to_use;

	/* nothing to do */
	if (!cleaned_count)
		return;

	rx_desc = FM10K_RX_DESC(rx_ring, i);
	bi = &rx_ring->rx_buffer[i];
	i -= rx_ring->count;

	do {
		if (!fm10k_alloc_mapped_page(rx_ring, bi))
			break;

		/* Refresh the desc even if buffer_addrs didn't change
		 * because each write-back erases this info.
		 */
		rx_desc->q.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);

		rx_desc++;
		bi++;
		i++;
		if (unlikely(!i)) {
			rx_desc = FM10K_RX_DESC(rx_ring, 0);
			bi = rx_ring->rx_buffer;
			i -= rx_ring->count;
		}

144 145
		/* clear the status bits for the next_to_use descriptor */
		rx_desc->d.staterr = 0;
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190

		cleaned_count--;
	} while (cleaned_count);

	i += rx_ring->count;

	if (rx_ring->next_to_use != i) {
		/* record the next descriptor to use */
		rx_ring->next_to_use = i;

		/* update next to alloc since we have filled the ring */
		rx_ring->next_to_alloc = i;

		/* Force memory writes to complete before letting h/w
		 * know there are new descriptors to fetch.  (Only
		 * applicable for weak-ordered memory model archs,
		 * such as IA-64).
		 */
		wmb();

		/* notify hardware of new descriptors */
		writel(i, rx_ring->tail);
	}
}

/**
 * fm10k_reuse_rx_page - page flip buffer and store it back on the ring
 * @rx_ring: rx descriptor ring to store buffers on
 * @old_buff: donor buffer to have page reused
 *
 * Synchronizes page for reuse by the interface
 **/
static void fm10k_reuse_rx_page(struct fm10k_ring *rx_ring,
				struct fm10k_rx_buffer *old_buff)
{
	struct fm10k_rx_buffer *new_buff;
	u16 nta = rx_ring->next_to_alloc;

	new_buff = &rx_ring->rx_buffer[nta];

	/* update, and store next to alloc */
	nta++;
	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;

	/* transfer page from old buffer to new buffer */
191
	*new_buff = *old_buff;
192 193 194 195 196 197 198 199

	/* sync the buffer for use by the device */
	dma_sync_single_range_for_device(rx_ring->dev, old_buff->dma,
					 old_buff->page_offset,
					 FM10K_RX_BUFSZ,
					 DMA_FROM_DEVICE);
}

200 201
static inline bool fm10k_page_is_reserved(struct page *page)
{
202
	return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
203 204
}

205 206
static bool fm10k_can_reuse_rx_page(struct fm10k_rx_buffer *rx_buffer,
				    struct page *page,
J
Jeff Kirsher 已提交
207
				    unsigned int __maybe_unused truesize)
208 209
{
	/* avoid re-using remote pages */
210
	if (unlikely(fm10k_page_is_reserved(page)))
211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
		return false;

#if (PAGE_SIZE < 8192)
	/* if we are only owner of page we can reuse it */
	if (unlikely(page_count(page) != 1))
		return false;

	/* flip page offset to other buffer */
	rx_buffer->page_offset ^= FM10K_RX_BUFSZ;
#else
	/* move offset up to the next cache line */
	rx_buffer->page_offset += truesize;

	if (rx_buffer->page_offset > (PAGE_SIZE - FM10K_RX_BUFSZ))
		return false;
#endif

228 229 230
	/* Even if we own the page, we are not allowed to use atomic_set()
	 * This would break get_page_unless_zero() users.
	 */
231
	page_ref_inc(page);
232

233 234 235 236 237 238
	return true;
}

/**
 * fm10k_add_rx_frag - Add contents of Rx buffer to sk_buff
 * @rx_buffer: buffer containing page to add
239
 * @size: packet size from rx_desc
240 241 242 243 244 245 246 247 248 249 250
 * @rx_desc: descriptor containing length of buffer written by hardware
 * @skb: sk_buff to place the data into
 *
 * This function will add the data contained in rx_buffer->page to the skb.
 * This is done either through a direct copy if the data in the buffer is
 * less than the skb header size, otherwise it will just attach the page as
 * a frag to the skb.
 *
 * The function will then update the page offset if necessary and return
 * true if the buffer can be reused by the interface.
 **/
J
Jeff Kirsher 已提交
251
static bool fm10k_add_rx_frag(struct fm10k_rx_buffer *rx_buffer,
252
			      unsigned int size,
253 254 255 256
			      union fm10k_rx_desc *rx_desc,
			      struct sk_buff *skb)
{
	struct page *page = rx_buffer->page;
257
	unsigned char *va = page_address(page) + rx_buffer->page_offset;
258 259 260
#if (PAGE_SIZE < 8192)
	unsigned int truesize = FM10K_RX_BUFSZ;
#else
261
	unsigned int truesize = ALIGN(size, 512);
262
#endif
263
	unsigned int pull_len;
264

265 266
	if (unlikely(skb_is_nonlinear(skb)))
		goto add_tail_frag;
267

268
	if (likely(size <= FM10K_RX_HDR_LEN)) {
269 270
		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));

271 272
		/* page is not reserved, we can reuse buffer as-is */
		if (likely(!fm10k_page_is_reserved(page)))
273 274 275
			return true;

		/* this page cannot be reused so discard it */
276
		__free_page(page);
277 278 279
		return false;
	}

280 281 282 283 284 285 286 287 288 289 290 291 292
	/* we need the header to contain the greater of either ETH_HLEN or
	 * 60 bytes if the skb->len is less than 60 for skb_pad.
	 */
	pull_len = eth_get_headlen(va, FM10K_RX_HDR_LEN);

	/* align pull length to size of long to optimize memcpy performance */
	memcpy(__skb_put(skb, pull_len), va, ALIGN(pull_len, sizeof(long)));

	/* update all of the pointers */
	va += pull_len;
	size -= pull_len;

add_tail_frag:
293
	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
294
			(unsigned long)va & ~PAGE_MASK, size, truesize);
295 296 297 298 299 300 301 302

	return fm10k_can_reuse_rx_page(rx_buffer, page, truesize);
}

static struct sk_buff *fm10k_fetch_rx_buffer(struct fm10k_ring *rx_ring,
					     union fm10k_rx_desc *rx_desc,
					     struct sk_buff *skb)
{
303
	unsigned int size = le16_to_cpu(rx_desc->w.length);
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321
	struct fm10k_rx_buffer *rx_buffer;
	struct page *page;

	rx_buffer = &rx_ring->rx_buffer[rx_ring->next_to_clean];
	page = rx_buffer->page;
	prefetchw(page);

	if (likely(!skb)) {
		void *page_addr = page_address(page) +
				  rx_buffer->page_offset;

		/* prefetch first cache line of first page */
		prefetch(page_addr);
#if L1_CACHE_BYTES < 128
		prefetch(page_addr + L1_CACHE_BYTES);
#endif

		/* allocate a skb to store the frags */
322 323
		skb = napi_alloc_skb(&rx_ring->q_vector->napi,
				     FM10K_RX_HDR_LEN);
324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
		if (unlikely(!skb)) {
			rx_ring->rx_stats.alloc_failed++;
			return NULL;
		}

		/* we will be copying header into skb->data in
		 * pskb_may_pull so it is in our interest to prefetch
		 * it now to avoid a possible cache miss
		 */
		prefetchw(skb->data);
	}

	/* we are reusing so sync this buffer for CPU use */
	dma_sync_single_range_for_cpu(rx_ring->dev,
				      rx_buffer->dma,
				      rx_buffer->page_offset,
340
				      size,
341 342 343
				      DMA_FROM_DEVICE);

	/* pull page into skb */
344
	if (fm10k_add_rx_frag(rx_buffer, size, rx_desc, skb)) {
345 346 347 348 349 350 351 352 353 354 355 356 357 358
		/* hand second half of page back to the ring */
		fm10k_reuse_rx_page(rx_ring, rx_buffer);
	} else {
		/* we are not reusing the buffer so unmap it */
		dma_unmap_page(rx_ring->dev, rx_buffer->dma,
			       PAGE_SIZE, DMA_FROM_DEVICE);
	}

	/* clear contents of rx_buffer */
	rx_buffer->page = NULL;

	return skb;
}

359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
static inline void fm10k_rx_checksum(struct fm10k_ring *ring,
				     union fm10k_rx_desc *rx_desc,
				     struct sk_buff *skb)
{
	skb_checksum_none_assert(skb);

	/* Rx checksum disabled via ethtool */
	if (!(ring->netdev->features & NETIF_F_RXCSUM))
		return;

	/* TCP/UDP checksum error bit is set */
	if (fm10k_test_staterr(rx_desc,
			       FM10K_RXD_STATUS_L4E |
			       FM10K_RXD_STATUS_L4E2 |
			       FM10K_RXD_STATUS_IPE |
			       FM10K_RXD_STATUS_IPE2)) {
		ring->rx_stats.csum_err++;
		return;
	}

	/* It must be a TCP or UDP packet with a valid checksum */
	if (fm10k_test_staterr(rx_desc, FM10K_RXD_STATUS_L4CS2))
		skb->encapsulation = true;
	else if (!fm10k_test_staterr(rx_desc, FM10K_RXD_STATUS_L4CS))
		return;

	skb->ip_summed = CHECKSUM_UNNECESSARY;
386 387

	ring->rx_stats.csum_good++;
388 389 390
}

#define FM10K_RSS_L4_TYPES_MASK \
391 392 393 394
	(BIT(FM10K_RSSTYPE_IPV4_TCP) | \
	 BIT(FM10K_RSSTYPE_IPV4_UDP) | \
	 BIT(FM10K_RSSTYPE_IPV6_TCP) | \
	 BIT(FM10K_RSSTYPE_IPV6_UDP))
395 396 397 398 399 400 401 402 403 404 405 406 407 408 409

static inline void fm10k_rx_hash(struct fm10k_ring *ring,
				 union fm10k_rx_desc *rx_desc,
				 struct sk_buff *skb)
{
	u16 rss_type;

	if (!(ring->netdev->features & NETIF_F_RXHASH))
		return;

	rss_type = le16_to_cpu(rx_desc->w.pkt_info) & FM10K_RXD_RSSTYPE_MASK;
	if (!rss_type)
		return;

	skb_set_hash(skb, le32_to_cpu(rx_desc->d.rss),
410
		     (BIT(rss_type) & FM10K_RSS_L4_TYPES_MASK) ?
411 412 413
		     PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3);
}

414
static void fm10k_type_trans(struct fm10k_ring *rx_ring,
J
Jeff Kirsher 已提交
415
			     union fm10k_rx_desc __maybe_unused *rx_desc,
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
			     struct sk_buff *skb)
{
	struct net_device *dev = rx_ring->netdev;
	struct fm10k_l2_accel *l2_accel = rcu_dereference_bh(rx_ring->l2_accel);

	/* check to see if DGLORT belongs to a MACVLAN */
	if (l2_accel) {
		u16 idx = le16_to_cpu(FM10K_CB(skb)->fi.w.dglort) - 1;

		idx -= l2_accel->dglort;
		if (idx < l2_accel->size && l2_accel->macvlan[idx])
			dev = l2_accel->macvlan[idx];
		else
			l2_accel = NULL;
	}

432
	/* Record Rx queue, or update macvlan statistics */
433
	if (!l2_accel)
434 435 436
		skb_record_rx_queue(skb, rx_ring->queue_index);
	else
		macvlan_count_rx(netdev_priv(dev), skb->len + ETH_HLEN, true,
437 438 439
				 false);

	skb->protocol = eth_type_trans(skb, dev);
440 441
}

442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
/**
 * fm10k_process_skb_fields - Populate skb header fields from Rx descriptor
 * @rx_ring: rx descriptor ring packet is being transacted on
 * @rx_desc: pointer to the EOP Rx descriptor
 * @skb: pointer to current skb being populated
 *
 * This function checks the ring, descriptor, and packet information in
 * order to populate the hash, checksum, VLAN, timestamp, protocol, and
 * other fields within the skb.
 **/
static unsigned int fm10k_process_skb_fields(struct fm10k_ring *rx_ring,
					     union fm10k_rx_desc *rx_desc,
					     struct sk_buff *skb)
{
	unsigned int len = skb->len;

458 459 460 461
	fm10k_rx_hash(rx_ring, rx_desc, skb);

	fm10k_rx_checksum(rx_ring, rx_desc, skb);

462 463
	FM10K_CB(skb)->tstamp = rx_desc->q.timestamp;

464 465 466 467 468 469 470
	FM10K_CB(skb)->fi.w.vlan = rx_desc->w.vlan;

	FM10K_CB(skb)->fi.d.glort = rx_desc->d.glort;

	if (rx_desc->w.vlan) {
		u16 vid = le16_to_cpu(rx_desc->w.vlan);

471
		if ((vid & VLAN_VID_MASK) != rx_ring->vid)
472
			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
473 474 475
		else if (vid & VLAN_PRIO_MASK)
			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
					       vid & VLAN_PRIO_MASK);
476 477
	}

478
	fm10k_type_trans(rx_ring, rx_desc, skb);
479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529

	return len;
}

/**
 * fm10k_is_non_eop - process handling of non-EOP buffers
 * @rx_ring: Rx ring being processed
 * @rx_desc: Rx descriptor for current buffer
 *
 * This function updates next to clean.  If the buffer is an EOP buffer
 * this function exits returning false, otherwise it will place the
 * sk_buff in the next buffer to be chained and return true indicating
 * that this is in fact a non-EOP buffer.
 **/
static bool fm10k_is_non_eop(struct fm10k_ring *rx_ring,
			     union fm10k_rx_desc *rx_desc)
{
	u32 ntc = rx_ring->next_to_clean + 1;

	/* fetch, update, and store next to clean */
	ntc = (ntc < rx_ring->count) ? ntc : 0;
	rx_ring->next_to_clean = ntc;

	prefetch(FM10K_RX_DESC(rx_ring, ntc));

	if (likely(fm10k_test_staterr(rx_desc, FM10K_RXD_STATUS_EOP)))
		return false;

	return true;
}

/**
 * fm10k_cleanup_headers - Correct corrupted or empty headers
 * @rx_ring: rx descriptor ring packet is being transacted on
 * @rx_desc: pointer to the EOP Rx descriptor
 * @skb: pointer to current skb being fixed
 *
 * Address the case where we are pulling data in on pages only
 * and as such no data is present in the skb header.
 *
 * In addition if skb is not at least 60 bytes we need to pad it so that
 * it is large enough to qualify as a valid Ethernet frame.
 *
 * Returns true if an error was encountered and skb was freed.
 **/
static bool fm10k_cleanup_headers(struct fm10k_ring *rx_ring,
				  union fm10k_rx_desc *rx_desc,
				  struct sk_buff *skb)
{
	if (unlikely((fm10k_test_staterr(rx_desc,
					 FM10K_RXD_STATUS_RXE)))) {
530 531 532 533 534 535 536 537 538 539 540 541
#define FM10K_TEST_RXD_BIT(rxd, bit) \
	((rxd)->w.csum_err & cpu_to_le16(bit))
		if (FM10K_TEST_RXD_BIT(rx_desc, FM10K_RXD_ERR_SWITCH_ERROR))
			rx_ring->rx_stats.switch_errors++;
		if (FM10K_TEST_RXD_BIT(rx_desc, FM10K_RXD_ERR_NO_DESCRIPTOR))
			rx_ring->rx_stats.drops++;
		if (FM10K_TEST_RXD_BIT(rx_desc, FM10K_RXD_ERR_PP_ERROR))
			rx_ring->rx_stats.pp_errors++;
		if (FM10K_TEST_RXD_BIT(rx_desc, FM10K_RXD_ERR_SWITCH_READY))
			rx_ring->rx_stats.link_errors++;
		if (FM10K_TEST_RXD_BIT(rx_desc, FM10K_RXD_ERR_TOO_BIG))
			rx_ring->rx_stats.length_errors++;
542 543 544 545 546
		dev_kfree_skb_any(skb);
		rx_ring->rx_stats.errors++;
		return true;
	}

547 548 549
	/* if eth_skb_pad returns an error the skb was freed */
	if (eth_skb_pad(skb))
		return true;
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564

	return false;
}

/**
 * fm10k_receive_skb - helper function to handle rx indications
 * @q_vector: structure containing interrupt and ring information
 * @skb: packet to send up
 **/
static void fm10k_receive_skb(struct fm10k_q_vector *q_vector,
			      struct sk_buff *skb)
{
	napi_gro_receive(&q_vector->napi, skb);
}

565 566 567
static int fm10k_clean_rx_irq(struct fm10k_q_vector *q_vector,
			      struct fm10k_ring *rx_ring,
			      int budget)
568 569 570 571 572
{
	struct sk_buff *skb = rx_ring->skb;
	unsigned int total_bytes = 0, total_packets = 0;
	u16 cleaned_count = fm10k_desc_unused(rx_ring);

573
	while (likely(total_packets < budget)) {
574 575 576 577 578 579 580 581 582 583
		union fm10k_rx_desc *rx_desc;

		/* return some buffers to hardware, one at a time is too slow */
		if (cleaned_count >= FM10K_RX_BUFFER_WRITE) {
			fm10k_alloc_rx_buffers(rx_ring, cleaned_count);
			cleaned_count = 0;
		}

		rx_desc = FM10K_RX_DESC(rx_ring, rx_ring->next_to_clean);

584
		if (!rx_desc->d.staterr)
585 586 587 588
			break;

		/* This memory barrier is needed to keep us from reading
		 * any other fields out of the rx_desc until we know the
589
		 * descriptor has been written back
590
		 */
591
		dma_rmb();
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621

		/* retrieve a buffer from the ring */
		skb = fm10k_fetch_rx_buffer(rx_ring, rx_desc, skb);

		/* exit if we failed to retrieve a buffer */
		if (!skb)
			break;

		cleaned_count++;

		/* fetch next buffer in frame if non-eop */
		if (fm10k_is_non_eop(rx_ring, rx_desc))
			continue;

		/* verify the packet layout is correct */
		if (fm10k_cleanup_headers(rx_ring, rx_desc, skb)) {
			skb = NULL;
			continue;
		}

		/* populate checksum, timestamp, VLAN, and protocol */
		total_bytes += fm10k_process_skb_fields(rx_ring, rx_desc, skb);

		fm10k_receive_skb(q_vector, skb);

		/* reset skb pointer */
		skb = NULL;

		/* update budget accounting */
		total_packets++;
622
	}
623 624 625 626 627 628 629 630 631 632 633

	/* place incomplete frames back on ring for completion */
	rx_ring->skb = skb;

	u64_stats_update_begin(&rx_ring->syncp);
	rx_ring->stats.packets += total_packets;
	rx_ring->stats.bytes += total_bytes;
	u64_stats_update_end(&rx_ring->syncp);
	q_vector->rx.total_packets += total_packets;
	q_vector->rx.total_bytes += total_bytes;

634
	return total_packets;
635 636
}

637 638 639 640
#define VXLAN_HLEN (sizeof(struct udphdr) + 8)
static struct ethhdr *fm10k_port_is_vxlan(struct sk_buff *skb)
{
	struct fm10k_intfc *interface = netdev_priv(skb->dev);
641
	struct fm10k_udp_port *vxlan_port;
642 643 644

	/* we can only offload a vxlan if we recognize it as such */
	vxlan_port = list_first_entry_or_null(&interface->vxlan_port,
645
					      struct fm10k_udp_port, list);
646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686

	if (!vxlan_port)
		return NULL;
	if (vxlan_port->port != udp_hdr(skb)->dest)
		return NULL;

	/* return offset of udp_hdr plus 8 bytes for VXLAN header */
	return (struct ethhdr *)(skb_transport_header(skb) + VXLAN_HLEN);
}

#define FM10K_NVGRE_RESERVED0_FLAGS htons(0x9FFF)
#define NVGRE_TNI htons(0x2000)
struct fm10k_nvgre_hdr {
	__be16 flags;
	__be16 proto;
	__be32 tni;
};

static struct ethhdr *fm10k_gre_is_nvgre(struct sk_buff *skb)
{
	struct fm10k_nvgre_hdr *nvgre_hdr;
	int hlen = ip_hdrlen(skb);

	/* currently only IPv4 is supported due to hlen above */
	if (vlan_get_protocol(skb) != htons(ETH_P_IP))
		return NULL;

	/* our transport header should be NVGRE */
	nvgre_hdr = (struct fm10k_nvgre_hdr *)(skb_network_header(skb) + hlen);

	/* verify all reserved flags are 0 */
	if (nvgre_hdr->flags & FM10K_NVGRE_RESERVED0_FLAGS)
		return NULL;

	/* report start of ethernet header */
	if (nvgre_hdr->flags & NVGRE_TNI)
		return (struct ethhdr *)(nvgre_hdr + 1);

	return (struct ethhdr *)(&nvgre_hdr->tni);
}

687
__be16 fm10k_tx_encap_offload(struct sk_buff *skb)
688
{
689
	u8 l4_hdr = 0, inner_l4_hdr = 0, inner_l4_hlen;
690 691
	struct ethhdr *eth_hdr;

692 693
	if (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
	    skb->inner_protocol != htons(ETH_P_TEB))
694 695
		return 0;

696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722
	switch (vlan_get_protocol(skb)) {
	case htons(ETH_P_IP):
		l4_hdr = ip_hdr(skb)->protocol;
		break;
	case htons(ETH_P_IPV6):
		l4_hdr = ipv6_hdr(skb)->nexthdr;
		break;
	default:
		return 0;
	}

	switch (l4_hdr) {
	case IPPROTO_UDP:
		eth_hdr = fm10k_port_is_vxlan(skb);
		break;
	case IPPROTO_GRE:
		eth_hdr = fm10k_gre_is_nvgre(skb);
		break;
	default:
		return 0;
	}

	if (!eth_hdr)
		return 0;

	switch (eth_hdr->h_proto) {
	case htons(ETH_P_IP):
723 724
		inner_l4_hdr = inner_ip_hdr(skb)->protocol;
		break;
725
	case htons(ETH_P_IPV6):
726
		inner_l4_hdr = inner_ipv6_hdr(skb)->nexthdr;
727 728 729 730 731
		break;
	default:
		return 0;
	}

732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749
	switch (inner_l4_hdr) {
	case IPPROTO_TCP:
		inner_l4_hlen = inner_tcp_hdrlen(skb);
		break;
	case IPPROTO_UDP:
		inner_l4_hlen = 8;
		break;
	default:
		return 0;
	}

	/* The hardware allows tunnel offloads only if the combined inner and
	 * outer header is 184 bytes or less
	 */
	if (skb_inner_transport_header(skb) + inner_l4_hlen -
	    skb_mac_header(skb) > FM10K_TUNNEL_HEADER_LENGTH)
		return 0;

750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
	return eth_hdr->h_proto;
}

static int fm10k_tso(struct fm10k_ring *tx_ring,
		     struct fm10k_tx_buffer *first)
{
	struct sk_buff *skb = first->skb;
	struct fm10k_tx_desc *tx_desc;
	unsigned char *th;
	u8 hdrlen;

	if (skb->ip_summed != CHECKSUM_PARTIAL)
		return 0;

	if (!skb_is_gso(skb))
		return 0;

	/* compute header lengths */
	if (skb->encapsulation) {
		if (!fm10k_tx_encap_offload(skb))
			goto err_vxlan;
		th = skb_inner_transport_header(skb);
	} else {
		th = skb_transport_header(skb);
	}

	/* compute offset from SOF to transport header and add header len */
	hdrlen = (th - skb->data) + (((struct tcphdr *)th)->doff << 2);

	first->tx_flags |= FM10K_TX_FLAGS_CSUM;

	/* update gso size and bytecount with header size */
	first->gso_segs = skb_shinfo(skb)->gso_segs;
	first->bytecount += (first->gso_segs - 1) * hdrlen;

	/* populate Tx descriptor header size and mss */
	tx_desc = FM10K_TX_DESC(tx_ring, tx_ring->next_to_use);
	tx_desc->hdrlen = hdrlen;
	tx_desc->mss = cpu_to_le16(skb_shinfo(skb)->gso_size);

	return 1;
J
Joe Perches 已提交
791

792 793
err_vxlan:
	tx_ring->netdev->features &= ~NETIF_F_GSO_UDP_TUNNEL;
J
Joe Perches 已提交
794
	if (net_ratelimit())
795 796 797 798 799 800 801 802 803 804 805 806 807 808 809
		netdev_err(tx_ring->netdev,
			   "TSO requested for unsupported tunnel, disabling offload\n");
	return -1;
}

static void fm10k_tx_csum(struct fm10k_ring *tx_ring,
			  struct fm10k_tx_buffer *first)
{
	struct sk_buff *skb = first->skb;
	struct fm10k_tx_desc *tx_desc;
	union {
		struct iphdr *ipv4;
		struct ipv6hdr *ipv6;
		u8 *raw;
	} network_hdr;
810 811
	u8 *transport_hdr;
	__be16 frag_off;
812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828
	__be16 protocol;
	u8 l4_hdr = 0;

	if (skb->ip_summed != CHECKSUM_PARTIAL)
		goto no_csum;

	if (skb->encapsulation) {
		protocol = fm10k_tx_encap_offload(skb);
		if (!protocol) {
			if (skb_checksum_help(skb)) {
				dev_warn(tx_ring->dev,
					 "failed to offload encap csum!\n");
				tx_ring->tx_stats.csum_err++;
			}
			goto no_csum;
		}
		network_hdr.raw = skb_inner_network_header(skb);
829
		transport_hdr = skb_inner_transport_header(skb);
830 831 832
	} else {
		protocol = vlan_get_protocol(skb);
		network_hdr.raw = skb_network_header(skb);
833
		transport_hdr = skb_transport_header(skb);
834 835 836 837 838 839 840 841
	}

	switch (protocol) {
	case htons(ETH_P_IP):
		l4_hdr = network_hdr.ipv4->protocol;
		break;
	case htons(ETH_P_IPV6):
		l4_hdr = network_hdr.ipv6->nexthdr;
842 843 844 845 846 847 848 849
		if (likely((transport_hdr - network_hdr.raw) ==
			   sizeof(struct ipv6hdr)))
			break;
		ipv6_skip_exthdr(skb, network_hdr.raw - skb->data +
				      sizeof(struct ipv6hdr),
				 &l4_hdr, &frag_off);
		if (unlikely(frag_off))
			l4_hdr = NEXTHDR_FRAGMENT;
850 851
		break;
	default:
852
		break;
853 854 855 856 857 858 859 860 861
	}

	switch (l4_hdr) {
	case IPPROTO_TCP:
	case IPPROTO_UDP:
		break;
	case IPPROTO_GRE:
		if (skb->encapsulation)
			break;
862
		/* fall through */
863 864 865
	default:
		if (unlikely(net_ratelimit())) {
			dev_warn(tx_ring->dev,
866 867
				 "partial checksum, version=%d l4 proto=%x\n",
				 protocol, l4_hdr);
868
		}
869
		skb_checksum_help(skb);
870 871 872 873 874 875
		tx_ring->tx_stats.csum_err++;
		goto no_csum;
	}

	/* update TX checksum flag */
	first->tx_flags |= FM10K_TX_FLAGS_CSUM;
876
	tx_ring->tx_stats.csum_good++;
877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901

no_csum:
	/* populate Tx descriptor header size and mss */
	tx_desc = FM10K_TX_DESC(tx_ring, tx_ring->next_to_use);
	tx_desc->hdrlen = 0;
	tx_desc->mss = 0;
}

#define FM10K_SET_FLAG(_input, _flag, _result) \
	((_flag <= _result) ? \
	 ((u32)(_input & _flag) * (_result / _flag)) : \
	 ((u32)(_input & _flag) / (_flag / _result)))

static u8 fm10k_tx_desc_flags(struct sk_buff *skb, u32 tx_flags)
{
	/* set type for advanced descriptor with frame checksum insertion */
	u32 desc_flags = 0;

	/* set checksum offload bits */
	desc_flags |= FM10K_SET_FLAG(tx_flags, FM10K_TX_FLAGS_CSUM,
				     FM10K_TXD_FLAG_CSUM);

	return desc_flags;
}

902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918
static bool fm10k_tx_desc_push(struct fm10k_ring *tx_ring,
			       struct fm10k_tx_desc *tx_desc, u16 i,
			       dma_addr_t dma, unsigned int size, u8 desc_flags)
{
	/* set RS and INT for last frame in a cache line */
	if ((++i & (FM10K_TXD_WB_FIFO_SIZE - 1)) == 0)
		desc_flags |= FM10K_TXD_FLAG_RS | FM10K_TXD_FLAG_INT;

	/* record values to descriptor */
	tx_desc->buffer_addr = cpu_to_le64(dma);
	tx_desc->flags = desc_flags;
	tx_desc->buflen = cpu_to_le16(size);

	/* return true if we just wrapped the ring */
	return i == tx_ring->count;
}

919 920 921 922
static int __fm10k_maybe_stop_tx(struct fm10k_ring *tx_ring, u16 size)
{
	netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);

923
	/* Memory barrier before checking head and tail */
924 925
	smp_mb();

926
	/* Check again in a case another CPU has just made room available */
927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942
	if (likely(fm10k_desc_unused(tx_ring) < size))
		return -EBUSY;

	/* A reprieve! - use start_queue because it doesn't call schedule */
	netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index);
	++tx_ring->tx_stats.restart_queue;
	return 0;
}

static inline int fm10k_maybe_stop_tx(struct fm10k_ring *tx_ring, u16 size)
{
	if (likely(fm10k_desc_unused(tx_ring) >= size))
		return 0;
	return __fm10k_maybe_stop_tx(tx_ring, size);
}

943 944 945 946 947 948 949 950 951 952
static void fm10k_tx_map(struct fm10k_ring *tx_ring,
			 struct fm10k_tx_buffer *first)
{
	struct sk_buff *skb = first->skb;
	struct fm10k_tx_buffer *tx_buffer;
	struct fm10k_tx_desc *tx_desc;
	struct skb_frag_struct *frag;
	unsigned char *data;
	dma_addr_t dma;
	unsigned int data_len, size;
953
	u32 tx_flags = first->tx_flags;
954
	u16 i = tx_ring->next_to_use;
955
	u8 flags = fm10k_tx_desc_flags(skb, tx_flags);
956 957 958 959

	tx_desc = FM10K_TX_DESC(tx_ring, i);

	/* add HW VLAN tag */
960 961
	if (skb_vlan_tag_present(skb))
		tx_desc->vlan = cpu_to_le16(skb_vlan_tag_get(skb));
962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
	else
		tx_desc->vlan = 0;

	size = skb_headlen(skb);
	data = skb->data;

	dma = dma_map_single(tx_ring->dev, data, size, DMA_TO_DEVICE);

	data_len = skb->data_len;
	tx_buffer = first;

	for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
		if (dma_mapping_error(tx_ring->dev, dma))
			goto dma_error;

		/* record length, and DMA address */
		dma_unmap_len_set(tx_buffer, len, size);
		dma_unmap_addr_set(tx_buffer, dma, dma);

		while (unlikely(size > FM10K_MAX_DATA_PER_TXD)) {
			if (fm10k_tx_desc_push(tx_ring, tx_desc++, i++, dma,
					       FM10K_MAX_DATA_PER_TXD, flags)) {
				tx_desc = FM10K_TX_DESC(tx_ring, 0);
				i = 0;
			}

			dma += FM10K_MAX_DATA_PER_TXD;
			size -= FM10K_MAX_DATA_PER_TXD;
		}

		if (likely(!data_len))
			break;

		if (fm10k_tx_desc_push(tx_ring, tx_desc++, i++,
				       dma, size, flags)) {
			tx_desc = FM10K_TX_DESC(tx_ring, 0);
			i = 0;
		}

		size = skb_frag_size(frag);
		data_len -= size;

		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
				       DMA_TO_DEVICE);

		tx_buffer = &tx_ring->tx_buffer[i];
	}

	/* write last descriptor with LAST bit set */
	flags |= FM10K_TXD_FLAG_LAST;

	if (fm10k_tx_desc_push(tx_ring, tx_desc, i++, dma, size, flags))
		i = 0;

	/* record bytecount for BQL */
	netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);

	/* record SW timestamp if HW timestamp is not available */
	skb_tx_timestamp(first->skb);

	/* Force memory writes to complete before letting h/w know there
	 * are new descriptors to fetch.  (Only applicable for weak-ordered
	 * memory model archs, such as IA-64).
	 *
	 * We also need this memory barrier to make certain all of the
	 * status bits have been updated before next_to_watch is written.
	 */
	wmb();

	/* set next_to_watch value indicating a packet is present */
	first->next_to_watch = tx_desc;

	tx_ring->next_to_use = i;

1036 1037 1038
	/* Make sure there is space in the ring for the next send. */
	fm10k_maybe_stop_tx(tx_ring, DESC_NEEDED);

1039
	/* notify HW of packet */
1040 1041
	if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
		writel(i, tx_ring->tail);
1042

1043 1044 1045 1046 1047
		/* we need this if more than one processor can write to our tail
		 * at a time, it synchronizes IO on IA64/Altix systems
		 */
		mmiowb();
	}
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069

	return;
dma_error:
	dev_err(tx_ring->dev, "TX DMA map failed\n");

	/* clear dma mappings for failed tx_buffer map */
	for (;;) {
		tx_buffer = &tx_ring->tx_buffer[i];
		fm10k_unmap_and_free_tx_resource(tx_ring, tx_buffer);
		if (tx_buffer == first)
			break;
		if (i == 0)
			i = tx_ring->count;
		i--;
	}

	tx_ring->next_to_use = i;
}

netdev_tx_t fm10k_xmit_frame_ring(struct sk_buff *skb,
				  struct fm10k_ring *tx_ring)
{
1070
	u16 count = TXD_USE_COUNT(skb_headlen(skb));
1071 1072
	struct fm10k_tx_buffer *first;
	unsigned short f;
1073 1074
	u32 tx_flags = 0;
	int tso;
1075 1076 1077 1078 1079 1080 1081 1082

	/* need: 1 descriptor per page * PAGE_SIZE/FM10K_MAX_DATA_PER_TXD,
	 *       + 1 desc for skb_headlen/FM10K_MAX_DATA_PER_TXD,
	 *       + 2 desc gap to keep tail from touching head
	 * otherwise try next time
	 */
	for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
		count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
1083

1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097
	if (fm10k_maybe_stop_tx(tx_ring, count + 3)) {
		tx_ring->tx_stats.tx_busy++;
		return NETDEV_TX_BUSY;
	}

	/* record the location of the first descriptor for this packet */
	first = &tx_ring->tx_buffer[tx_ring->next_to_use];
	first->skb = skb;
	first->bytecount = max_t(unsigned int, skb->len, ETH_ZLEN);
	first->gso_segs = 1;

	/* record initial flags and protocol */
	first->tx_flags = tx_flags;

1098 1099 1100 1101 1102 1103
	tso = fm10k_tso(tx_ring, first);
	if (tso < 0)
		goto out_drop;
	else if (!tso)
		fm10k_tx_csum(tx_ring, first);

1104 1105
	fm10k_tx_map(tx_ring, first);

1106 1107 1108 1109 1110 1111
	return NETDEV_TX_OK;

out_drop:
	dev_kfree_skb_any(first->skb);
	first->skb = NULL;

1112 1113 1114 1115 1116 1117 1118 1119
	return NETDEV_TX_OK;
}

static u64 fm10k_get_tx_completed(struct fm10k_ring *ring)
{
	return ring->stats.packets;
}

1120 1121 1122 1123 1124 1125
/**
 * fm10k_get_tx_pending - how many Tx descriptors not processed
 * @ring: the ring structure
 * @in_sw: is tx_pending being checked in SW or in HW?
 */
u64 fm10k_get_tx_pending(struct fm10k_ring *ring, bool in_sw)
1126
{
1127 1128
	struct fm10k_intfc *interface = ring->q_vector->interface;
	struct fm10k_hw *hw = &interface->hw;
1129
	u32 head, tail;
1130

1131 1132 1133 1134 1135 1136 1137
	if (likely(in_sw)) {
		head = ring->next_to_clean;
		tail = ring->next_to_use;
	} else {
		head = fm10k_read_reg(hw, FM10K_TDH(ring->reg_idx));
		tail = fm10k_read_reg(hw, FM10K_TDT(ring->reg_idx));
	}
1138 1139 1140 1141 1142 1143 1144 1145

	return ((head <= tail) ? tail : tail + ring->count) - head;
}

bool fm10k_check_tx_hang(struct fm10k_ring *tx_ring)
{
	u32 tx_done = fm10k_get_tx_completed(tx_ring);
	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
1146
	u32 tx_pending = fm10k_get_tx_pending(tx_ring, true);
1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161

	clear_check_for_tx_hang(tx_ring);

	/* Check for a hung queue, but be thorough. This verifies
	 * that a transmit has been completed since the previous
	 * check AND there is at least one packet pending. By
	 * requiring this to fail twice we avoid races with
	 * clearing the ARMED bit and conditions where we
	 * run the check_tx_hang logic with a transmit completion
	 * pending but without time to complete it yet.
	 */
	if (!tx_pending || (tx_done_old != tx_done)) {
		/* update completed stats and continue */
		tx_ring->tx_stats.tx_done_old = tx_done;
		/* reset the countdown */
1162
		clear_bit(__FM10K_HANG_CHECK_ARMED, tx_ring->state);
1163 1164 1165 1166 1167

		return false;
	}

	/* make sure it is true for two checks in a row */
1168
	return test_and_set_bit(__FM10K_HANG_CHECK_ARMED, tx_ring->state);
1169 1170 1171 1172 1173 1174 1175 1176 1177
}

/**
 * fm10k_tx_timeout_reset - initiate reset due to Tx timeout
 * @interface: driver private struct
 **/
void fm10k_tx_timeout_reset(struct fm10k_intfc *interface)
{
	/* Do the reset outside of interrupt context */
1178
	if (!test_bit(__FM10K_DOWN, interface->state)) {
1179
		interface->tx_timeout_count++;
1180
		set_bit(FM10K_FLAG_RESET_REQUESTED, interface->flags);
1181 1182 1183 1184 1185 1186 1187 1188
		fm10k_service_event_schedule(interface);
	}
}

/**
 * fm10k_clean_tx_irq - Reclaim resources after transmit completes
 * @q_vector: structure containing interrupt and ring information
 * @tx_ring: tx ring to clean
1189
 * @napi_budget: Used to determine if we are in netpoll
1190 1191
 **/
static bool fm10k_clean_tx_irq(struct fm10k_q_vector *q_vector,
1192
			       struct fm10k_ring *tx_ring, int napi_budget)
1193 1194 1195 1196 1197 1198 1199 1200
{
	struct fm10k_intfc *interface = q_vector->interface;
	struct fm10k_tx_buffer *tx_buffer;
	struct fm10k_tx_desc *tx_desc;
	unsigned int total_bytes = 0, total_packets = 0;
	unsigned int budget = q_vector->tx.work_limit;
	unsigned int i = tx_ring->next_to_clean;

1201
	if (test_bit(__FM10K_DOWN, interface->state))
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
		return true;

	tx_buffer = &tx_ring->tx_buffer[i];
	tx_desc = FM10K_TX_DESC(tx_ring, i);
	i -= tx_ring->count;

	do {
		struct fm10k_tx_desc *eop_desc = tx_buffer->next_to_watch;

		/* if next_to_watch is not set then there is no work pending */
		if (!eop_desc)
			break;

		/* prevent any other reads prior to eop_desc */
1216
		smp_rmb();
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229

		/* if DD is not set pending work has not been completed */
		if (!(eop_desc->flags & FM10K_TXD_FLAG_DONE))
			break;

		/* clear next_to_watch to prevent false hangs */
		tx_buffer->next_to_watch = NULL;

		/* update the statistics for this packet */
		total_bytes += tx_buffer->bytecount;
		total_packets += tx_buffer->gso_segs;

		/* free the skb */
1230
		napi_consume_skb(tx_buffer->skb, napi_budget);
1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330

		/* unmap skb header data */
		dma_unmap_single(tx_ring->dev,
				 dma_unmap_addr(tx_buffer, dma),
				 dma_unmap_len(tx_buffer, len),
				 DMA_TO_DEVICE);

		/* clear tx_buffer data */
		tx_buffer->skb = NULL;
		dma_unmap_len_set(tx_buffer, len, 0);

		/* unmap remaining buffers */
		while (tx_desc != eop_desc) {
			tx_buffer++;
			tx_desc++;
			i++;
			if (unlikely(!i)) {
				i -= tx_ring->count;
				tx_buffer = tx_ring->tx_buffer;
				tx_desc = FM10K_TX_DESC(tx_ring, 0);
			}

			/* unmap any remaining paged data */
			if (dma_unmap_len(tx_buffer, len)) {
				dma_unmap_page(tx_ring->dev,
					       dma_unmap_addr(tx_buffer, dma),
					       dma_unmap_len(tx_buffer, len),
					       DMA_TO_DEVICE);
				dma_unmap_len_set(tx_buffer, len, 0);
			}
		}

		/* move us one more past the eop_desc for start of next pkt */
		tx_buffer++;
		tx_desc++;
		i++;
		if (unlikely(!i)) {
			i -= tx_ring->count;
			tx_buffer = tx_ring->tx_buffer;
			tx_desc = FM10K_TX_DESC(tx_ring, 0);
		}

		/* issue prefetch for next Tx descriptor */
		prefetch(tx_desc);

		/* update budget accounting */
		budget--;
	} while (likely(budget));

	i += tx_ring->count;
	tx_ring->next_to_clean = i;
	u64_stats_update_begin(&tx_ring->syncp);
	tx_ring->stats.bytes += total_bytes;
	tx_ring->stats.packets += total_packets;
	u64_stats_update_end(&tx_ring->syncp);
	q_vector->tx.total_bytes += total_bytes;
	q_vector->tx.total_packets += total_packets;

	if (check_for_tx_hang(tx_ring) && fm10k_check_tx_hang(tx_ring)) {
		/* schedule immediate reset if we believe we hung */
		struct fm10k_hw *hw = &interface->hw;

		netif_err(interface, drv, tx_ring->netdev,
			  "Detected Tx Unit Hang\n"
			  "  Tx Queue             <%d>\n"
			  "  TDH, TDT             <%x>, <%x>\n"
			  "  next_to_use          <%x>\n"
			  "  next_to_clean        <%x>\n",
			  tx_ring->queue_index,
			  fm10k_read_reg(hw, FM10K_TDH(tx_ring->reg_idx)),
			  fm10k_read_reg(hw, FM10K_TDT(tx_ring->reg_idx)),
			  tx_ring->next_to_use, i);

		netif_stop_subqueue(tx_ring->netdev,
				    tx_ring->queue_index);

		netif_info(interface, probe, tx_ring->netdev,
			   "tx hang %d detected on queue %d, resetting interface\n",
			   interface->tx_timeout_count + 1,
			   tx_ring->queue_index);

		fm10k_tx_timeout_reset(interface);

		/* the netdev is about to reset, no point in enabling stuff */
		return true;
	}

	/* notify netdev of completed buffers */
	netdev_tx_completed_queue(txring_txq(tx_ring),
				  total_packets, total_bytes);

#define TX_WAKE_THRESHOLD min_t(u16, FM10K_MIN_TXD - 1, DESC_NEEDED * 2)
	if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) &&
		     (fm10k_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) {
		/* Make sure that anybody stopping the queue after this
		 * sees the new next_to_clean.
		 */
		smp_mb();
		if (__netif_subqueue_stopped(tx_ring->netdev,
					     tx_ring->queue_index) &&
1331
		    !test_bit(__FM10K_DOWN, interface->state)) {
1332 1333 1334 1335 1336 1337 1338 1339 1340
			netif_wake_subqueue(tx_ring->netdev,
					    tx_ring->queue_index);
			++tx_ring->tx_stats.restart_queue;
		}
	}

	return !!budget;
}

A
Alexander Duyck 已提交
1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352
/**
 * fm10k_update_itr - update the dynamic ITR value based on packet size
 *
 *      Stores a new ITR value based on strictly on packet size.  The
 *      divisors and thresholds used by this function were determined based
 *      on theoretical maximum wire speed and testing data, in order to
 *      minimize response time while increasing bulk throughput.
 *
 * @ring_container: Container for rings to have ITR updated
 **/
static void fm10k_update_itr(struct fm10k_ring_container *ring_container)
{
1353
	unsigned int avg_wire_size, packets, itr_round;
A
Alexander Duyck 已提交
1354 1355

	/* Only update ITR if we are using adaptive setting */
1356
	if (!ITR_IS_ADAPTIVE(ring_container->itr))
A
Alexander Duyck 已提交
1357 1358 1359 1360 1361 1362 1363 1364
		goto clear_counts;

	packets = ring_container->total_packets;
	if (!packets)
		goto clear_counts;

	avg_wire_size = ring_container->total_bytes / packets;

1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394
	/* The following is a crude approximation of:
	 *  wmem_default / (size + overhead) = desired_pkts_per_int
	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
	 *
	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
	 * formula down to
	 *
	 *  (34 * (size + 24)) / (size + 640) = ITR
	 *
	 * We first do some math on the packet size and then finally bitshift
	 * by 8 after rounding up. We also have to account for PCIe link speed
	 * difference as ITR scales based on this.
	 */
	if (avg_wire_size <= 360) {
		/* Start at 250K ints/sec and gradually drop to 77K ints/sec */
		avg_wire_size *= 8;
		avg_wire_size += 376;
	} else if (avg_wire_size <= 1152) {
		/* 77K ints/sec to 45K ints/sec */
		avg_wire_size *= 3;
		avg_wire_size += 2176;
	} else if (avg_wire_size <= 1920) {
		/* 45K ints/sec to 38K ints/sec */
		avg_wire_size += 4480;
	} else {
		/* plateau at a limit of 38K ints/sec */
		avg_wire_size = 6656;
	}
A
Alexander Duyck 已提交
1395

1396 1397 1398 1399
	/* Perform final bitshift for division after rounding up to ensure
	 * that the calculation will never get below a 1. The bit shift
	 * accounts for changes in the ITR due to PCIe link speed.
	 */
1400
	itr_round = READ_ONCE(ring_container->itr_scale) + 8;
1401
	avg_wire_size += BIT(itr_round) - 1;
1402
	avg_wire_size >>= itr_round;
A
Alexander Duyck 已提交
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436

	/* write back value and retain adaptive flag */
	ring_container->itr = avg_wire_size | FM10K_ITR_ADAPTIVE;

clear_counts:
	ring_container->total_bytes = 0;
	ring_container->total_packets = 0;
}

static void fm10k_qv_enable(struct fm10k_q_vector *q_vector)
{
	/* Enable auto-mask and clear the current mask */
	u32 itr = FM10K_ITR_ENABLE;

	/* Update Tx ITR */
	fm10k_update_itr(&q_vector->tx);

	/* Update Rx ITR */
	fm10k_update_itr(&q_vector->rx);

	/* Store Tx itr in timer slot 0 */
	itr |= (q_vector->tx.itr & FM10K_ITR_MAX);

	/* Shift Rx itr to timer slot 1 */
	itr |= (q_vector->rx.itr & FM10K_ITR_MAX) << FM10K_ITR_INTERVAL1_SHIFT;

	/* Write the final value to the ITR register */
	writel(itr, q_vector->itr);
}

static int fm10k_poll(struct napi_struct *napi, int budget)
{
	struct fm10k_q_vector *q_vector =
			       container_of(napi, struct fm10k_q_vector, napi);
1437
	struct fm10k_ring *ring;
1438
	int per_ring_budget, work_done = 0;
1439 1440
	bool clean_complete = true;

1441 1442 1443 1444
	fm10k_for_each_ring(ring, q_vector->tx) {
		if (!fm10k_clean_tx_irq(q_vector, ring, budget))
			clean_complete = false;
	}
1445

1446 1447 1448 1449
	/* Handle case where we are called by netpoll with a budget of 0 */
	if (budget <= 0)
		return budget;

1450 1451 1452 1453
	/* attempt to distribute budget to each queue fairly, but don't
	 * allow the budget to go below 1 because we'll exit polling
	 */
	if (q_vector->rx.count > 1)
B
Bruce Allan 已提交
1454
		per_ring_budget = max(budget / q_vector->rx.count, 1);
1455 1456 1457
	else
		per_ring_budget = budget;

1458 1459 1460 1461
	fm10k_for_each_ring(ring, q_vector->rx) {
		int work = fm10k_clean_rx_irq(q_vector, ring, per_ring_budget);

		work_done += work;
1462 1463
		if (work >= per_ring_budget)
			clean_complete = false;
1464
	}
1465 1466 1467 1468

	/* If all work not completed, return budget and keep polling */
	if (!clean_complete)
		return budget;
A
Alexander Duyck 已提交
1469 1470

	/* all work done, exit the polling mode */
1471
	napi_complete_done(napi, work_done);
A
Alexander Duyck 已提交
1472 1473 1474 1475

	/* re-enable the q_vector */
	fm10k_qv_enable(q_vector);

1476
	return min(work_done, budget - 1);
A
Alexander Duyck 已提交
1477 1478
}

1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
/**
 * fm10k_set_qos_queues: Allocate queues for a QOS-enabled device
 * @interface: board private structure to initialize
 *
 * When QoS (Quality of Service) is enabled, allocate queues for
 * each traffic class.  If multiqueue isn't available,then abort QoS
 * initialization.
 *
 * This function handles all combinations of Qos and RSS.
 *
 **/
static bool fm10k_set_qos_queues(struct fm10k_intfc *interface)
{
	struct net_device *dev = interface->netdev;
	struct fm10k_ring_feature *f;
	int rss_i, i;
	int pcs;

	/* Map queue offset and counts onto allocated tx queues */
	pcs = netdev_get_num_tc(dev);

	if (pcs <= 1)
		return false;

	/* set QoS mask and indices */
	f = &interface->ring_feature[RING_F_QOS];
	f->indices = pcs;
1506
	f->mask = BIT(fls(pcs - 1)) - 1;
1507 1508 1509

	/* determine the upper limit for our current DCB mode */
	rss_i = interface->hw.mac.max_queues / pcs;
1510
	rss_i = BIT(fls(rss_i) - 1);
1511 1512 1513 1514 1515

	/* set RSS mask and indices */
	f = &interface->ring_feature[RING_F_RSS];
	rss_i = min_t(u16, rss_i, f->limit);
	f->indices = rss_i;
1516
	f->mask = BIT(fls(rss_i - 1)) - 1;
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545

	/* configure pause class to queue mapping */
	for (i = 0; i < pcs; i++)
		netdev_set_tc_queue(dev, i, rss_i, rss_i * i);

	interface->num_rx_queues = rss_i * pcs;
	interface->num_tx_queues = rss_i * pcs;

	return true;
}

/**
 * fm10k_set_rss_queues: Allocate queues for RSS
 * @interface: board private structure to initialize
 *
 * This is our "base" multiqueue mode.  RSS (Receive Side Scaling) will try
 * to allocate one Rx queue per CPU, and if available, one Tx queue per CPU.
 *
 **/
static bool fm10k_set_rss_queues(struct fm10k_intfc *interface)
{
	struct fm10k_ring_feature *f;
	u16 rss_i;

	f = &interface->ring_feature[RING_F_RSS];
	rss_i = min_t(u16, interface->hw.mac.max_queues, f->limit);

	/* record indices and power of 2 mask for RSS */
	f->indices = rss_i;
1546
	f->mask = BIT(fls(rss_i - 1)) - 1;
1547 1548 1549 1550 1551 1552 1553

	interface->num_rx_queues = rss_i;
	interface->num_tx_queues = rss_i;

	return true;
}

A
Alexander Duyck 已提交
1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
/**
 * fm10k_set_num_queues: Allocate queues for device, feature dependent
 * @interface: board private structure to initialize
 *
 * This is the top level queue allocation routine.  The order here is very
 * important, starting with the "most" number of features turned on at once,
 * and ending with the smallest set of features.  This way large combinations
 * can be allocated if they're turned on, and smaller combinations are the
 * fallthrough conditions.
 *
 **/
static void fm10k_set_num_queues(struct fm10k_intfc *interface)
{
1567
	/* Attempt to setup QoS and RSS first */
1568 1569 1570
	if (fm10k_set_qos_queues(interface))
		return;

1571
	/* If we don't have QoS, just fallback to only RSS. */
1572
	fm10k_set_rss_queues(interface);
A
Alexander Duyck 已提交
1573 1574
}

1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
/**
 * fm10k_reset_num_queues - Reset the number of queues to zero
 * @interface: board private structure
 *
 * This function should be called whenever we need to reset the number of
 * queues after an error condition.
 */
static void fm10k_reset_num_queues(struct fm10k_intfc *interface)
{
	interface->num_tx_queues = 0;
	interface->num_rx_queues = 0;
	interface->num_q_vectors = 0;
}

A
Alexander Duyck 已提交
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606
/**
 * fm10k_alloc_q_vector - Allocate memory for a single interrupt vector
 * @interface: board private structure to initialize
 * @v_count: q_vectors allocated on interface, used for ring interleaving
 * @v_idx: index of vector in interface struct
 * @txr_count: total number of Tx rings to allocate
 * @txr_idx: index of first Tx ring to allocate
 * @rxr_count: total number of Rx rings to allocate
 * @rxr_idx: index of first Rx ring to allocate
 *
 * We allocate one q_vector.  If allocation fails we return -ENOMEM.
 **/
static int fm10k_alloc_q_vector(struct fm10k_intfc *interface,
				unsigned int v_count, unsigned int v_idx,
				unsigned int txr_count, unsigned int txr_idx,
				unsigned int rxr_count, unsigned int rxr_idx)
{
	struct fm10k_q_vector *q_vector;
1607
	struct fm10k_ring *ring;
A
Alexander Duyck 已提交
1608 1609 1610
	int ring_count, size;

	ring_count = txr_count + rxr_count;
1611 1612
	size = sizeof(struct fm10k_q_vector) +
	       (sizeof(struct fm10k_ring) * ring_count);
A
Alexander Duyck 已提交
1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627

	/* allocate q_vector and rings */
	q_vector = kzalloc(size, GFP_KERNEL);
	if (!q_vector)
		return -ENOMEM;

	/* initialize NAPI */
	netif_napi_add(interface->netdev, &q_vector->napi,
		       fm10k_poll, NAPI_POLL_WEIGHT);

	/* tie q_vector and interface together */
	interface->q_vector[v_idx] = q_vector;
	q_vector->interface = interface;
	q_vector->v_idx = v_idx;

1628 1629 1630
	/* initialize pointer to rings */
	ring = q_vector->ring;

A
Alexander Duyck 已提交
1631
	/* save Tx ring container info */
1632 1633
	q_vector->tx.ring = ring;
	q_vector->tx.work_limit = FM10K_DEFAULT_TX_WORK;
A
Alexander Duyck 已提交
1634
	q_vector->tx.itr = interface->tx_itr;
1635
	q_vector->tx.itr_scale = interface->hw.mac.itr_scale;
A
Alexander Duyck 已提交
1636 1637
	q_vector->tx.count = txr_count;

1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
	while (txr_count) {
		/* assign generic ring traits */
		ring->dev = &interface->pdev->dev;
		ring->netdev = interface->netdev;

		/* configure backlink on ring */
		ring->q_vector = q_vector;

		/* apply Tx specific ring traits */
		ring->count = interface->tx_ring_count;
		ring->queue_index = txr_idx;

		/* assign ring to interface */
		interface->tx_ring[txr_idx] = ring;

		/* update count and index */
		txr_count--;
		txr_idx += v_count;

		/* push pointer to next ring */
		ring++;
	}

A
Alexander Duyck 已提交
1661
	/* save Rx ring container info */
1662
	q_vector->rx.ring = ring;
A
Alexander Duyck 已提交
1663
	q_vector->rx.itr = interface->rx_itr;
1664
	q_vector->rx.itr_scale = interface->hw.mac.itr_scale;
A
Alexander Duyck 已提交
1665 1666
	q_vector->rx.count = rxr_count;

1667 1668 1669 1670
	while (rxr_count) {
		/* assign generic ring traits */
		ring->dev = &interface->pdev->dev;
		ring->netdev = interface->netdev;
1671
		rcu_assign_pointer(ring->l2_accel, interface->l2_accel);
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690

		/* configure backlink on ring */
		ring->q_vector = q_vector;

		/* apply Rx specific ring traits */
		ring->count = interface->rx_ring_count;
		ring->queue_index = rxr_idx;

		/* assign ring to interface */
		interface->rx_ring[rxr_idx] = ring;

		/* update count and index */
		rxr_count--;
		rxr_idx += v_count;

		/* push pointer to next ring */
		ring++;
	}

A
Alexander Duyck 已提交
1691 1692
	fm10k_dbg_q_vector_init(q_vector);

A
Alexander Duyck 已提交
1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707
	return 0;
}

/**
 * fm10k_free_q_vector - Free memory allocated for specific interrupt vector
 * @interface: board private structure to initialize
 * @v_idx: Index of vector to be freed
 *
 * This function frees the memory allocated to the q_vector.  In addition if
 * NAPI is enabled it will delete any references to the NAPI struct prior
 * to freeing the q_vector.
 **/
static void fm10k_free_q_vector(struct fm10k_intfc *interface, int v_idx)
{
	struct fm10k_q_vector *q_vector = interface->q_vector[v_idx];
1708 1709
	struct fm10k_ring *ring;

A
Alexander Duyck 已提交
1710 1711
	fm10k_dbg_q_vector_exit(q_vector);

1712 1713 1714 1715 1716
	fm10k_for_each_ring(ring, q_vector->tx)
		interface->tx_ring[ring->queue_index] = NULL;

	fm10k_for_each_ring(ring, q_vector->rx)
		interface->rx_ring[ring->queue_index] = NULL;
A
Alexander Duyck 已提交
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771

	interface->q_vector[v_idx] = NULL;
	netif_napi_del(&q_vector->napi);
	kfree_rcu(q_vector, rcu);
}

/**
 * fm10k_alloc_q_vectors - Allocate memory for interrupt vectors
 * @interface: board private structure to initialize
 *
 * We allocate one q_vector per queue interrupt.  If allocation fails we
 * return -ENOMEM.
 **/
static int fm10k_alloc_q_vectors(struct fm10k_intfc *interface)
{
	unsigned int q_vectors = interface->num_q_vectors;
	unsigned int rxr_remaining = interface->num_rx_queues;
	unsigned int txr_remaining = interface->num_tx_queues;
	unsigned int rxr_idx = 0, txr_idx = 0, v_idx = 0;
	int err;

	if (q_vectors >= (rxr_remaining + txr_remaining)) {
		for (; rxr_remaining; v_idx++) {
			err = fm10k_alloc_q_vector(interface, q_vectors, v_idx,
						   0, 0, 1, rxr_idx);
			if (err)
				goto err_out;

			/* update counts and index */
			rxr_remaining--;
			rxr_idx++;
		}
	}

	for (; v_idx < q_vectors; v_idx++) {
		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);

		err = fm10k_alloc_q_vector(interface, q_vectors, v_idx,
					   tqpv, txr_idx,
					   rqpv, rxr_idx);

		if (err)
			goto err_out;

		/* update counts and index */
		rxr_remaining -= rqpv;
		txr_remaining -= tqpv;
		rxr_idx++;
		txr_idx++;
	}

	return 0;

err_out:
1772
	fm10k_reset_num_queues(interface);
A
Alexander Duyck 已提交
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791

	while (v_idx--)
		fm10k_free_q_vector(interface, v_idx);

	return -ENOMEM;
}

/**
 * fm10k_free_q_vectors - Free memory allocated for interrupt vectors
 * @interface: board private structure to initialize
 *
 * This function frees the memory allocated to the q_vectors.  In addition if
 * NAPI is enabled it will delete any references to the NAPI struct prior
 * to freeing the q_vector.
 **/
static void fm10k_free_q_vectors(struct fm10k_intfc *interface)
{
	int v_idx = interface->num_q_vectors;

1792
	fm10k_reset_num_queues(interface);
A
Alexander Duyck 已提交
1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860

	while (v_idx--)
		fm10k_free_q_vector(interface, v_idx);
}

/**
 * f10k_reset_msix_capability - reset MSI-X capability
 * @interface: board private structure to initialize
 *
 * Reset the MSI-X capability back to its starting state
 **/
static void fm10k_reset_msix_capability(struct fm10k_intfc *interface)
{
	pci_disable_msix(interface->pdev);
	kfree(interface->msix_entries);
	interface->msix_entries = NULL;
}

/**
 * f10k_init_msix_capability - configure MSI-X capability
 * @interface: board private structure to initialize
 *
 * Attempt to configure the interrupts using the best available
 * capabilities of the hardware and the kernel.
 **/
static int fm10k_init_msix_capability(struct fm10k_intfc *interface)
{
	struct fm10k_hw *hw = &interface->hw;
	int v_budget, vector;

	/* It's easy to be greedy for MSI-X vectors, but it really
	 * doesn't do us much good if we have a lot more vectors
	 * than CPU's.  So let's be conservative and only ask for
	 * (roughly) the same number of vectors as there are CPU's.
	 * the default is to use pairs of vectors
	 */
	v_budget = max(interface->num_rx_queues, interface->num_tx_queues);
	v_budget = min_t(u16, v_budget, num_online_cpus());

	/* account for vectors not related to queues */
	v_budget += NON_Q_VECTORS(hw);

	/* At the same time, hardware can only support a maximum of
	 * hw.mac->max_msix_vectors vectors.  With features
	 * such as RSS and VMDq, we can easily surpass the number of Rx and Tx
	 * descriptor queues supported by our device.  Thus, we cap it off in
	 * those rare cases where the cpu count also exceeds our vector limit.
	 */
	v_budget = min_t(int, v_budget, hw->mac.max_msix_vectors);

	/* A failure in MSI-X entry allocation is fatal. */
	interface->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
					  GFP_KERNEL);
	if (!interface->msix_entries)
		return -ENOMEM;

	/* populate entry values */
	for (vector = 0; vector < v_budget; vector++)
		interface->msix_entries[vector].entry = vector;

	/* Attempt to enable MSI-X with requested value */
	v_budget = pci_enable_msix_range(interface->pdev,
					 interface->msix_entries,
					 MIN_MSIX_COUNT(hw),
					 v_budget);
	if (v_budget < 0) {
		kfree(interface->msix_entries);
		interface->msix_entries = NULL;
1861
		return v_budget;
A
Alexander Duyck 已提交
1862 1863 1864 1865 1866 1867 1868 1869
	}

	/* record the number of queues available for q_vectors */
	interface->num_q_vectors = v_budget - NON_Q_VECTORS(hw);

	return 0;
}

1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934
/**
 * fm10k_cache_ring_qos - Descriptor ring to register mapping for QoS
 * @interface: Interface structure continaining rings and devices
 *
 * Cache the descriptor ring offsets for Qos
 **/
static bool fm10k_cache_ring_qos(struct fm10k_intfc *interface)
{
	struct net_device *dev = interface->netdev;
	int pc, offset, rss_i, i, q_idx;
	u16 pc_stride = interface->ring_feature[RING_F_QOS].mask + 1;
	u8 num_pcs = netdev_get_num_tc(dev);

	if (num_pcs <= 1)
		return false;

	rss_i = interface->ring_feature[RING_F_RSS].indices;

	for (pc = 0, offset = 0; pc < num_pcs; pc++, offset += rss_i) {
		q_idx = pc;
		for (i = 0; i < rss_i; i++) {
			interface->tx_ring[offset + i]->reg_idx = q_idx;
			interface->tx_ring[offset + i]->qos_pc = pc;
			interface->rx_ring[offset + i]->reg_idx = q_idx;
			interface->rx_ring[offset + i]->qos_pc = pc;
			q_idx += pc_stride;
		}
	}

	return true;
}

/**
 * fm10k_cache_ring_rss - Descriptor ring to register mapping for RSS
 * @interface: Interface structure continaining rings and devices
 *
 * Cache the descriptor ring offsets for RSS
 **/
static void fm10k_cache_ring_rss(struct fm10k_intfc *interface)
{
	int i;

	for (i = 0; i < interface->num_rx_queues; i++)
		interface->rx_ring[i]->reg_idx = i;

	for (i = 0; i < interface->num_tx_queues; i++)
		interface->tx_ring[i]->reg_idx = i;
}

/**
 * fm10k_assign_rings - Map rings to network devices
 * @interface: Interface structure containing rings and devices
 *
 * This function is meant to go though and configure both the network
 * devices so that they contain rings, and configure the rings so that
 * they function with their network devices.
 **/
static void fm10k_assign_rings(struct fm10k_intfc *interface)
{
	if (fm10k_cache_ring_qos(interface))
		return;

	fm10k_cache_ring_rss(interface);
}

A
Alexander Duyck 已提交
1935 1936 1937
static void fm10k_init_reta(struct fm10k_intfc *interface)
{
	u16 i, rss_i = interface->ring_feature[RING_F_RSS].indices;
1938
	u32 reta;
A
Alexander Duyck 已提交
1939

1940 1941 1942 1943
	/* If the Rx flow indirection table has been configured manually, we
	 * need to maintain it when possible.
	 */
	if (netif_is_rxfh_configured(interface->netdev)) {
A
Alexander Duyck 已提交
1944 1945 1946 1947 1948 1949 1950
		for (i = FM10K_RETA_SIZE; i--;) {
			reta = interface->reta[i];
			if ((((reta << 24) >> 24) < rss_i) &&
			    (((reta << 16) >> 24) < rss_i) &&
			    (((reta <<  8) >> 24) < rss_i) &&
			    (((reta)       >> 24) < rss_i))
				continue;
1951 1952 1953 1954

			/* this should never happen */
			dev_err(&interface->pdev->dev,
				"RSS indirection table assigned flows out of queue bounds. Reconfiguring.\n");
A
Alexander Duyck 已提交
1955 1956 1957 1958 1959 1960 1961 1962
			goto repopulate_reta;
		}

		/* do nothing if all of the elements are in bounds */
		return;
	}

repopulate_reta:
1963
	fm10k_write_reta(interface, NULL);
A
Alexander Duyck 已提交
1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985
}

/**
 * fm10k_init_queueing_scheme - Determine proper queueing scheme
 * @interface: board private structure to initialize
 *
 * We determine which queueing scheme to use based on...
 * - Hardware queue count (num_*_queues)
 *   - defined by miscellaneous hardware support/features (RSS, etc.)
 **/
int fm10k_init_queueing_scheme(struct fm10k_intfc *interface)
{
	int err;

	/* Number of supported queues */
	fm10k_set_num_queues(interface);

	/* Configure MSI-X capability */
	err = fm10k_init_msix_capability(interface);
	if (err) {
		dev_err(&interface->pdev->dev,
			"Unable to initialize MSI-X capability\n");
1986
		goto err_init_msix;
A
Alexander Duyck 已提交
1987 1988 1989 1990
	}

	/* Allocate memory for queues */
	err = fm10k_alloc_q_vectors(interface);
1991
	if (err) {
1992 1993 1994
		dev_err(&interface->pdev->dev,
			"Unable to allocate queue vectors\n");
		goto err_alloc_q_vectors;
1995
	}
A
Alexander Duyck 已提交
1996

1997 1998 1999
	/* Map rings to devices, and map devices to physical queues */
	fm10k_assign_rings(interface);

A
Alexander Duyck 已提交
2000 2001 2002 2003
	/* Initialize RSS redirection table */
	fm10k_init_reta(interface);

	return 0;
2004 2005 2006 2007 2008 2009

err_alloc_q_vectors:
	fm10k_reset_msix_capability(interface);
err_init_msix:
	fm10k_reset_num_queues(interface);
	return err;
A
Alexander Duyck 已提交
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
}

/**
 * fm10k_clear_queueing_scheme - Clear the current queueing scheme settings
 * @interface: board private structure to clear queueing scheme on
 *
 * We go through and clear queueing specific resources and reset the structure
 * to pre-load conditions
 **/
void fm10k_clear_queueing_scheme(struct fm10k_intfc *interface)
{
	fm10k_free_q_vectors(interface);
	fm10k_reset_msix_capability(interface);
}