aq_ring.c 12.2 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
2 3
/*
 * aQuantia Corporation Network Driver
4
 * Copyright (C) 2014-2019 aQuantia Corporation. All rights reserved
5 6 7 8 9 10 11
 */

/* File aq_ring.c: Definition of functions for Rx/Tx rings. */

#include "aq_ring.h"
#include "aq_nic.h"
#include "aq_hw.h"
12
#include "aq_hw_utils.h"
13
#include "aq_ptp.h"
14 15 16 17

#include <linux/netdevice.h>
#include <linux/etherdevice.h>

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
static inline void aq_free_rxpage(struct aq_rxpage *rxpage, struct device *dev)
{
	unsigned int len = PAGE_SIZE << rxpage->order;

	dma_unmap_page(dev, rxpage->daddr, len, DMA_FROM_DEVICE);

	/* Drop the ref for being in the ring. */
	__free_pages(rxpage->page, rxpage->order);
	rxpage->page = NULL;
}

static int aq_get_rxpage(struct aq_rxpage *rxpage, unsigned int order,
			 struct device *dev)
{
	struct page *page;
	int ret = -ENOMEM;
N
Nikita Danilov 已提交
34
	dma_addr_t daddr;
35 36 37 38 39 40 41 42 43 44 45 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 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95

	page = dev_alloc_pages(order);
	if (unlikely(!page))
		goto err_exit;

	daddr = dma_map_page(dev, page, 0, PAGE_SIZE << order,
			     DMA_FROM_DEVICE);

	if (unlikely(dma_mapping_error(dev, daddr)))
		goto free_page;

	rxpage->page = page;
	rxpage->daddr = daddr;
	rxpage->order = order;
	rxpage->pg_off = 0;

	return 0;

free_page:
	__free_pages(page, order);

err_exit:
	return ret;
}

static int aq_get_rxpages(struct aq_ring_s *self, struct aq_ring_buff_s *rxbuf,
			  int order)
{
	int ret;

	if (rxbuf->rxdata.page) {
		/* One means ring is the only user and can reuse */
		if (page_ref_count(rxbuf->rxdata.page) > 1) {
			/* Try reuse buffer */
			rxbuf->rxdata.pg_off += AQ_CFG_RX_FRAME_MAX;
			if (rxbuf->rxdata.pg_off + AQ_CFG_RX_FRAME_MAX <=
				(PAGE_SIZE << order)) {
				self->stats.rx.pg_flips++;
			} else {
				/* Buffer exhausted. We have other users and
				 * should release this page and realloc
				 */
				aq_free_rxpage(&rxbuf->rxdata,
					       aq_nic_get_dev(self->aq_nic));
				self->stats.rx.pg_losts++;
			}
		} else {
			rxbuf->rxdata.pg_off = 0;
			self->stats.rx.pg_reuses++;
		}
	}

	if (!rxbuf->rxdata.page) {
		ret = aq_get_rxpage(&rxbuf->rxdata, order,
				    aq_nic_get_dev(self->aq_nic));
		return ret;
	}

	return 0;
}

96 97 98 99 100 101 102 103 104 105 106 107 108
static struct aq_ring_s *aq_ring_alloc(struct aq_ring_s *self,
				       struct aq_nic_s *aq_nic)
{
	int err = 0;

	self->buff_ring =
		kcalloc(self->size, sizeof(struct aq_ring_buff_s), GFP_KERNEL);

	if (!self->buff_ring) {
		err = -ENOMEM;
		goto err_exit;
	}
	self->dx_ring = dma_alloc_coherent(aq_nic_get_dev(aq_nic),
109 110
					   self->size * self->dx_size,
					   &self->dx_ring_pa, GFP_KERNEL);
111 112 113 114 115 116 117 118 119 120
	if (!self->dx_ring) {
		err = -ENOMEM;
		goto err_exit;
	}

err_exit:
	if (err < 0) {
		aq_ring_free(self);
		self = NULL;
	}
N
Nikita Danilov 已提交
121

122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
	return self;
}

struct aq_ring_s *aq_ring_tx_alloc(struct aq_ring_s *self,
				   struct aq_nic_s *aq_nic,
				   unsigned int idx,
				   struct aq_nic_cfg_s *aq_nic_cfg)
{
	int err = 0;

	self->aq_nic = aq_nic;
	self->idx = idx;
	self->size = aq_nic_cfg->txds;
	self->dx_size = aq_nic_cfg->aq_hw_caps->txd_size;

	self = aq_ring_alloc(self, aq_nic);
	if (!self) {
		err = -ENOMEM;
		goto err_exit;
	}

err_exit:
	if (err < 0) {
		aq_ring_free(self);
		self = NULL;
	}
N
Nikita Danilov 已提交
148

149 150 151 152 153 154 155 156 157 158 159 160 161 162
	return self;
}

struct aq_ring_s *aq_ring_rx_alloc(struct aq_ring_s *self,
				   struct aq_nic_s *aq_nic,
				   unsigned int idx,
				   struct aq_nic_cfg_s *aq_nic_cfg)
{
	int err = 0;

	self->aq_nic = aq_nic;
	self->idx = idx;
	self->size = aq_nic_cfg->rxds;
	self->dx_size = aq_nic_cfg->aq_hw_caps->rxd_size;
163 164 165 166 167
	self->page_order = fls(AQ_CFG_RX_FRAME_MAX / PAGE_SIZE +
			       (AQ_CFG_RX_FRAME_MAX % PAGE_SIZE ? 1 : 0)) - 1;

	if (aq_nic_cfg->rxpageorder > self->page_order)
		self->page_order = aq_nic_cfg->rxpageorder;
168 169 170 171 172 173 174 175 176 177 178 179

	self = aq_ring_alloc(self, aq_nic);
	if (!self) {
		err = -ENOMEM;
		goto err_exit;
	}

err_exit:
	if (err < 0) {
		aq_ring_free(self);
		self = NULL;
	}
N
Nikita Danilov 已提交
180

181 182 183
	return self;
}

184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
struct aq_ring_s *
aq_ring_hwts_rx_alloc(struct aq_ring_s *self, struct aq_nic_s *aq_nic,
		      unsigned int idx, unsigned int size, unsigned int dx_size)
{
	struct device *dev = aq_nic_get_dev(aq_nic);
	size_t sz = size * dx_size + AQ_CFG_RXDS_DEF;

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

	self->aq_nic = aq_nic;
	self->idx = idx;
	self->size = size;
	self->dx_size = dx_size;

	self->dx_ring = dma_alloc_coherent(dev, sz, &self->dx_ring_pa,
					   GFP_KERNEL);
	if (!self->dx_ring) {
		aq_ring_free(self);
		return NULL;
	}

	return self;
}

208 209 210 211 212
int aq_ring_init(struct aq_ring_s *self)
{
	self->hw_head = 0;
	self->sw_head = 0;
	self->sw_tail = 0;
N
Nikita Danilov 已提交
213

214 215 216
	return 0;
}

P
Pavel Belous 已提交
217 218 219 220 221 222
static inline bool aq_ring_dx_in_range(unsigned int h, unsigned int i,
				       unsigned int t)
{
	return (h < t) ? ((h < i) && (i < t)) : ((h < i) || (i < t));
}

I
Igor Russkikh 已提交
223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
void aq_ring_update_queue_state(struct aq_ring_s *ring)
{
	if (aq_ring_avail_dx(ring) <= AQ_CFG_SKB_FRAGS_MAX)
		aq_ring_queue_stop(ring);
	else if (aq_ring_avail_dx(ring) > AQ_CFG_RESTART_DESC_THRES)
		aq_ring_queue_wake(ring);
}

void aq_ring_queue_wake(struct aq_ring_s *ring)
{
	struct net_device *ndev = aq_nic_get_ndev(ring->aq_nic);

	if (__netif_subqueue_stopped(ndev, ring->idx)) {
		netif_wake_subqueue(ndev, ring->idx);
		ring->stats.tx.queue_restarts++;
	}
}

void aq_ring_queue_stop(struct aq_ring_s *ring)
{
	struct net_device *ndev = aq_nic_get_ndev(ring->aq_nic);

	if (!__netif_subqueue_stopped(ndev, ring->idx))
		netif_stop_subqueue(ndev, ring->idx);
}

249
bool aq_ring_tx_clean(struct aq_ring_s *self)
250 251
{
	struct device *dev = aq_nic_get_dev(self->aq_nic);
252
	unsigned int budget;
253

254 255
	for (budget = AQ_CFG_TX_CLEAN_BUDGET;
	     budget && self->sw_head != self->hw_head; budget--) {
256 257 258
		struct aq_ring_buff_s *buff = &self->buff_ring[self->sw_head];

		if (likely(buff->is_mapped)) {
P
Pavel Belous 已提交
259 260 261 262 263 264 265 266
			if (unlikely(buff->is_sop)) {
				if (!buff->is_eop &&
				    buff->eop_index != 0xffffU &&
				    (!aq_ring_dx_in_range(self->sw_head,
						buff->eop_index,
						self->hw_head)))
					break;

267 268
				dma_unmap_single(dev, buff->pa, buff->len,
						 DMA_TO_DEVICE);
P
Pavel Belous 已提交
269
			} else {
270 271
				dma_unmap_page(dev, buff->pa, buff->len,
					       DMA_TO_DEVICE);
P
Pavel Belous 已提交
272
			}
273 274
		}

275 276 277
		if (unlikely(buff->is_eop)) {
			++self->stats.rx.packets;
			self->stats.tx.bytes += buff->skb->len;
278

279 280
			dev_kfree_skb_any(buff->skb);
		}
P
Pavel Belous 已提交
281 282
		buff->pa = 0U;
		buff->eop_index = 0xffffU;
283
		self->sw_head = aq_ring_next_dx(self, self->sw_head);
P
Pavel Belous 已提交
284
	}
285 286

	return !!budget;
287 288
}

289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
static void aq_rx_checksum(struct aq_ring_s *self,
			   struct aq_ring_buff_s *buff,
			   struct sk_buff *skb)
{
	if (!(self->aq_nic->ndev->features & NETIF_F_RXCSUM))
		return;

	if (unlikely(buff->is_cso_err)) {
		++self->stats.rx.errors;
		skb->ip_summed = CHECKSUM_NONE;
		return;
	}
	if (buff->is_ip_cso) {
		__skb_incr_checksum_unnecessary(skb);
	} else {
		skb->ip_summed = CHECKSUM_NONE;
	}
306 307 308

	if (buff->is_udp_cso || buff->is_tcp_cso)
		__skb_incr_checksum_unnecessary(skb);
309 310
}

311
#define AQ_SKB_ALIGN SKB_DATA_ALIGN(sizeof(struct skb_shared_info))
312 313 314 315
int aq_ring_rx_clean(struct aq_ring_s *self,
		     struct napi_struct *napi,
		     int *work_done,
		     int budget)
316 317 318
{
	struct net_device *ndev = aq_nic_get_ndev(self->aq_nic);
	bool is_rsc_completed = true;
319
	int err = 0;
320 321 322 323 324

	for (; (self->sw_head != self->hw_head) && budget;
		self->sw_head = aq_ring_next_dx(self, self->sw_head),
		--budget, ++(*work_done)) {
		struct aq_ring_buff_s *buff = &self->buff_ring[self->sw_head];
325
		bool is_ptp_ring = aq_ptp_ring(self->aq_nic, self);
326
		struct aq_ring_buff_s *buff_ = NULL;
327 328 329
		struct sk_buff *skb = NULL;
		unsigned int next_ = 0U;
		unsigned int i = 0U;
330
		u16 hdr_len;
331 332 333 334 335

		if (buff->is_cleaned)
			continue;

		if (!buff->is_eop) {
336 337 338 339
			buff_ = buff;
			do {
				next_ = buff_->next,
				buff_ = &self->buff_ring[next_];
340 341 342 343 344
				is_rsc_completed =
					aq_ring_dx_in_range(self->sw_head,
							    next_,
							    self->hw_head);

345
				if (unlikely(!is_rsc_completed))
346 347
					break;

348
				buff->is_error |= buff_->is_error;
349
				buff->is_cso_err |= buff_->is_cso_err;
350 351

			} while (!buff_->is_eop);
352 353 354 355 356

			if (!is_rsc_completed) {
				err = 0;
				goto err_exit;
			}
357 358
			if (buff->is_error ||
			    (buff->is_lro && buff->is_cso_err)) {
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374
				buff_ = buff;
				do {
					next_ = buff_->next,
					buff_ = &self->buff_ring[next_];

					buff_->is_cleaned = true;
				} while (!buff_->is_eop);

				++self->stats.rx.errors;
				continue;
			}
		}

		if (buff->is_error) {
			++self->stats.rx.errors;
			continue;
375 376
		}

377 378 379 380 381
		dma_sync_single_range_for_cpu(aq_nic_get_dev(self->aq_nic),
					      buff->rxdata.daddr,
					      buff->rxdata.pg_off,
					      buff->len, DMA_FROM_DEVICE);

382
		/* for single fragment packets use build_skb() */
383 384
		if (buff->is_eop &&
		    buff->len <= AQ_CFG_RX_FRAME_MAX - AQ_SKB_ALIGN) {
385
			skb = build_skb(aq_buf_vaddr(&buff->rxdata),
386
					AQ_CFG_RX_FRAME_MAX);
387 388 389 390
			if (unlikely(!skb)) {
				err = -ENOMEM;
				goto err_exit;
			}
391 392 393 394 395
			if (is_ptp_ring)
				buff->len -=
					aq_ptp_extract_ts(self->aq_nic, skb,
						aq_buf_vaddr(&buff->rxdata),
						buff->len);
396
			skb_put(skb, buff->len);
397
			page_ref_inc(buff->rxdata.page);
398
		} else {
399
			skb = napi_alloc_skb(napi, AQ_CFG_RX_HDR_SIZE);
400 401 402 403
			if (unlikely(!skb)) {
				err = -ENOMEM;
				goto err_exit;
			}
404 405 406 407 408
			if (is_ptp_ring)
				buff->len -=
					aq_ptp_extract_ts(self->aq_nic, skb,
						aq_buf_vaddr(&buff->rxdata),
						buff->len);
409

410 411
			hdr_len = buff->len;
			if (hdr_len > AQ_CFG_RX_HDR_SIZE)
412 413
				hdr_len = eth_get_headlen(skb->dev,
							  aq_buf_vaddr(&buff->rxdata),
414 415
							  AQ_CFG_RX_HDR_SIZE);

416
			memcpy(__skb_put(skb, hdr_len), aq_buf_vaddr(&buff->rxdata),
417 418 419
			       ALIGN(hdr_len, sizeof(long)));

			if (buff->len - hdr_len > 0) {
420 421
				skb_add_rx_frag(skb, 0, buff->rxdata.page,
						buff->rxdata.pg_off + hdr_len,
422
						buff->len - hdr_len,
423 424
						AQ_CFG_RX_FRAME_MAX);
				page_ref_inc(buff->rxdata.page);
425
			}
426

427
			if (!buff->is_eop) {
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442
				buff_ = buff;
				i = 1U;
				do {
					next_ = buff_->next,
					buff_ = &self->buff_ring[next_];

					dma_sync_single_range_for_cpu(
							aq_nic_get_dev(self->aq_nic),
							buff_->rxdata.daddr,
							buff_->rxdata.pg_off,
							buff_->len,
							DMA_FROM_DEVICE);
					skb_add_rx_frag(skb, i++,
							buff_->rxdata.page,
							buff_->rxdata.pg_off,
443
							buff_->len,
444 445
							AQ_CFG_RX_FRAME_MAX);
					page_ref_inc(buff_->rxdata.page);
446
					buff_->is_cleaned = 1;
447 448 449 450 451 452

					buff->is_ip_cso &= buff_->is_ip_cso;
					buff->is_udp_cso &= buff_->is_udp_cso;
					buff->is_tcp_cso &= buff_->is_tcp_cso;
					buff->is_cso_err |= buff_->is_cso_err;

453
				} while (!buff_->is_eop);
454 455 456
			}
		}

457 458 459 460
		if (buff->is_vlan)
			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
					       buff->vlan_rx_tag);

461
		skb->protocol = eth_type_trans(skb, ndev);
462 463

		aq_rx_checksum(self, buff, skb);
464 465 466 467

		skb_set_hash(skb, buff->rss_hash,
			     buff->is_hash_l4 ? PKT_HASH_TYPE_L4 :
			     PKT_HASH_TYPE_NONE);
468 469
		/* Send all PTP traffic to 0 queue */
		skb_record_rx_queue(skb, is_ptp_ring ? 0 : self->idx);
470 471 472

		++self->stats.rx.packets;
		self->stats.rx.bytes += skb->len;
473 474

		napi_gro_receive(napi, skb);
475 476 477 478 479 480
	}

err_exit:
	return err;
}

481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
void aq_ring_hwts_rx_clean(struct aq_ring_s *self, struct aq_nic_s *aq_nic)
{
	while (self->sw_head != self->hw_head) {
		u64 ns;

		aq_nic->aq_hw_ops->extract_hwts(aq_nic->aq_hw,
						self->dx_ring +
						(self->sw_head * self->dx_size),
						self->dx_size, &ns);
		aq_ptp_tx_hwtstamp(aq_nic, ns);

		self->sw_head = aq_ring_next_dx(self, self->sw_head);
	}
}

496 497
int aq_ring_rx_fill(struct aq_ring_s *self)
{
498
	unsigned int page_order = self->page_order;
499 500 501 502
	struct aq_ring_buff_s *buff = NULL;
	int err = 0;
	int i = 0;

503 504 505 506
	if (aq_ring_avail_dx(self) < min_t(unsigned int, AQ_CFG_RX_REFILL_THRES,
					   self->size / 2))
		return err;

507 508 509 510 511 512 513
	for (i = aq_ring_avail_dx(self); i--;
		self->sw_tail = aq_ring_next_dx(self, self->sw_tail)) {
		buff = &self->buff_ring[self->sw_tail];

		buff->flags = 0U;
		buff->len = AQ_CFG_RX_FRAME_MAX;

514 515
		err = aq_get_rxpages(self, buff, page_order);
		if (err)
516 517
			goto err_exit;

518
		buff->pa = aq_buf_daddr(&buff->rxdata);
519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
		buff = NULL;
	}

err_exit:
	return err;
}

void aq_ring_rx_deinit(struct aq_ring_s *self)
{
	if (!self)
		goto err_exit;

	for (; self->sw_head != self->sw_tail;
		self->sw_head = aq_ring_next_dx(self, self->sw_head)) {
		struct aq_ring_buff_s *buff = &self->buff_ring[self->sw_head];

535
		aq_free_rxpage(&buff->rxdata, aq_nic_get_dev(self->aq_nic));
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
	}

err_exit:;
}

void aq_ring_free(struct aq_ring_s *self)
{
	if (!self)
		goto err_exit;

	kfree(self->buff_ring);

	if (self->dx_ring)
		dma_free_coherent(aq_nic_get_dev(self->aq_nic),
				  self->size * self->dx_size, self->dx_ring,
				  self->dx_ring_pa);

err_exit:;
}