nfp_net_common.c 98.8 KB
Newer Older
1
/*
2
 * Copyright (C) 2015-2017 Netronome Systems, Inc.
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
 *
 * This software is dual licensed under the GNU General License Version 2,
 * June 1991 as shown in the file COPYING in the top-level directory of this
 * source tree or the BSD 2-Clause License provided below.  You have the
 * option to license this software under the complete terms of either license.
 *
 * The BSD 2-Clause License:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      1. Redistributions of source code must retain the above
 *         copyright notice, this list of conditions and the following
 *         disclaimer.
 *
 *      2. 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.
 */

/*
 * nfp_net_common.c
 * Netronome network device driver: Common functions between PF and VF
 * Authors: Jakub Kicinski <jakub.kicinski@netronome.com>
 *          Jason McMullan <jason.mcmullan@netronome.com>
 *          Rolf Neugebauer <rolf.neugebauer@netronome.com>
 *          Brad Petrus <brad.petrus@netronome.com>
 *          Chris Telfer <chris.telfer@netronome.com>
 */

44
#include <linux/bitfield.h>
45
#include <linux/bpf.h>
46
#include <linux/bpf_trace.h>
47 48 49 50 51 52 53 54 55
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/fs.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/interrupt.h>
#include <linux/ip.h>
#include <linux/ipv6.h>
56
#include <linux/page_ref.h>
57 58 59 60 61 62 63
#include <linux/pci.h>
#include <linux/pci_regs.h>
#include <linux/msi.h>
#include <linux/ethtool.h>
#include <linux/log2.h>
#include <linux/if_vlan.h>
#include <linux/random.h>
64
#include <linux/vmalloc.h>
65 66
#include <linux/ktime.h>

S
Simon Horman 已提交
67
#include <net/switchdev.h>
68 69
#include <net/vxlan.h>

70
#include "nfpcore/nfp_nsp.h"
71
#include "nfp_app.h"
72 73
#include "nfp_net_ctrl.h"
#include "nfp_net.h"
74
#include "nfp_net_sriov.h"
J
Jakub Kicinski 已提交
75
#include "nfp_port.h"
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90

/**
 * nfp_net_get_fw_version() - Read and parse the FW version
 * @fw_ver:	Output fw_version structure to read to
 * @ctrl_bar:	Mapped address of the control BAR
 */
void nfp_net_get_fw_version(struct nfp_net_fw_version *fw_ver,
			    void __iomem *ctrl_bar)
{
	u32 reg;

	reg = readl(ctrl_bar + NFP_NET_CFG_VERSION);
	put_unaligned_le32(reg, fw_ver);
}

91
static dma_addr_t nfp_net_dma_map_rx(struct nfp_net_dp *dp, void *frag)
92
{
93 94 95 96 97 98 99 100 101 102 103
	return dma_map_single_attrs(dp->dev, frag + NFP_NET_RX_BUF_HEADROOM,
				    dp->fl_bufsz - NFP_NET_RX_BUF_NON_DATA,
				    dp->rx_dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
}

static void
nfp_net_dma_sync_dev_rx(const struct nfp_net_dp *dp, dma_addr_t dma_addr)
{
	dma_sync_single_for_device(dp->dev, dma_addr,
				   dp->fl_bufsz - NFP_NET_RX_BUF_NON_DATA,
				   dp->rx_dma_dir);
104 105
}

106
static void nfp_net_dma_unmap_rx(struct nfp_net_dp *dp, dma_addr_t dma_addr)
107
{
108 109 110 111 112 113 114 115 116 117
	dma_unmap_single_attrs(dp->dev, dma_addr,
			       dp->fl_bufsz - NFP_NET_RX_BUF_NON_DATA,
			       dp->rx_dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
}

static void nfp_net_dma_sync_cpu_rx(struct nfp_net_dp *dp, dma_addr_t dma_addr,
				    unsigned int len)
{
	dma_sync_single_for_cpu(dp->dev, dma_addr - NFP_NET_RX_BUF_HEADROOM,
				len, dp->rx_dma_dir);
118 119
}

120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
/* Firmware reconfig
 *
 * Firmware reconfig may take a while so we have two versions of it -
 * synchronous and asynchronous (posted).  All synchronous callers are holding
 * RTNL so we don't have to worry about serializing them.
 */
static void nfp_net_reconfig_start(struct nfp_net *nn, u32 update)
{
	nn_writel(nn, NFP_NET_CFG_UPDATE, update);
	/* ensure update is written before pinging HW */
	nn_pci_flush(nn);
	nfp_qcp_wr_ptr_add(nn->qcp_cfg, 1);
}

/* Pass 0 as update to run posted reconfigs. */
static void nfp_net_reconfig_start_async(struct nfp_net *nn, u32 update)
{
	update |= nn->reconfig_posted;
	nn->reconfig_posted = 0;

	nfp_net_reconfig_start(nn, update);

	nn->reconfig_timer_active = true;
	mod_timer(&nn->reconfig_timer, jiffies + NFP_NET_POLL_TIMEOUT * HZ);
}

static bool nfp_net_reconfig_check_done(struct nfp_net *nn, bool last_check)
{
	u32 reg;

	reg = nn_readl(nn, NFP_NET_CFG_UPDATE);
	if (reg == 0)
		return true;
	if (reg & NFP_NET_CFG_UPDATE_ERR) {
		nn_err(nn, "Reconfig error: 0x%08x\n", reg);
		return true;
	} else if (last_check) {
		nn_err(nn, "Reconfig timeout: 0x%08x\n", reg);
		return true;
	}

	return false;
}

static int nfp_net_reconfig_wait(struct nfp_net *nn, unsigned long deadline)
{
	bool timed_out = false;

	/* Poll update field, waiting for NFP to ack the config */
	while (!nfp_net_reconfig_check_done(nn, timed_out)) {
		msleep(1);
		timed_out = time_is_before_eq_jiffies(deadline);
	}

	if (nn_readl(nn, NFP_NET_CFG_UPDATE) & NFP_NET_CFG_UPDATE_ERR)
		return -EIO;

	return timed_out ? -EIO : 0;
}

180
static void nfp_net_reconfig_timer(struct timer_list *t)
181
{
182
	struct nfp_net *nn = from_timer(nn, t, reconfig_timer);
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229

	spin_lock_bh(&nn->reconfig_lock);

	nn->reconfig_timer_active = false;

	/* If sync caller is present it will take over from us */
	if (nn->reconfig_sync_present)
		goto done;

	/* Read reconfig status and report errors */
	nfp_net_reconfig_check_done(nn, true);

	if (nn->reconfig_posted)
		nfp_net_reconfig_start_async(nn, 0);
done:
	spin_unlock_bh(&nn->reconfig_lock);
}

/**
 * nfp_net_reconfig_post() - Post async reconfig request
 * @nn:      NFP Net device to reconfigure
 * @update:  The value for the update field in the BAR config
 *
 * Record FW reconfiguration request.  Reconfiguration will be kicked off
 * whenever reconfiguration machinery is idle.  Multiple requests can be
 * merged together!
 */
static void nfp_net_reconfig_post(struct nfp_net *nn, u32 update)
{
	spin_lock_bh(&nn->reconfig_lock);

	/* Sync caller will kick off async reconf when it's done, just post */
	if (nn->reconfig_sync_present) {
		nn->reconfig_posted |= update;
		goto done;
	}

	/* Opportunistically check if the previous command is done */
	if (!nn->reconfig_timer_active ||
	    nfp_net_reconfig_check_done(nn, false))
		nfp_net_reconfig_start_async(nn, update);
	else
		nn->reconfig_posted |= update;
done:
	spin_unlock_bh(&nn->reconfig_lock);
}

230 231 232 233 234 235 236 237 238 239 240 241 242
/**
 * nfp_net_reconfig() - Reconfigure the firmware
 * @nn:      NFP Net device to reconfigure
 * @update:  The value for the update field in the BAR config
 *
 * Write the update word to the BAR and ping the reconfig queue.  The
 * poll until the firmware has acknowledged the update by zeroing the
 * update word.
 *
 * Return: Negative errno on error, 0 on success
 */
int nfp_net_reconfig(struct nfp_net *nn, u32 update)
{
243 244 245
	bool cancelled_timer = false;
	u32 pre_posted_requests;
	int ret;
246 247 248

	spin_lock_bh(&nn->reconfig_lock);

249
	nn->reconfig_sync_present = true;
250

251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
	if (nn->reconfig_timer_active) {
		del_timer(&nn->reconfig_timer);
		nn->reconfig_timer_active = false;
		cancelled_timer = true;
	}
	pre_posted_requests = nn->reconfig_posted;
	nn->reconfig_posted = 0;

	spin_unlock_bh(&nn->reconfig_lock);

	if (cancelled_timer)
		nfp_net_reconfig_wait(nn, nn->reconfig_timer.expires);

	/* Run the posted reconfigs which were issued before we started */
	if (pre_posted_requests) {
		nfp_net_reconfig_start(nn, pre_posted_requests);
		nfp_net_reconfig_wait(nn, jiffies + HZ * NFP_NET_POLL_TIMEOUT);
268 269
	}

270 271 272 273 274 275 276 277 278 279
	nfp_net_reconfig_start(nn, update);
	ret = nfp_net_reconfig_wait(nn, jiffies + HZ * NFP_NET_POLL_TIMEOUT);

	spin_lock_bh(&nn->reconfig_lock);

	if (nn->reconfig_posted)
		nfp_net_reconfig_start_async(nn, 0);

	nn->reconfig_sync_present = false;

280
	spin_unlock_bh(&nn->reconfig_lock);
281

282 283 284
	return ret;
}

P
Pablo Cascón 已提交
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308
/**
 * nfp_net_reconfig_mbox() - Reconfigure the firmware via the mailbox
 * @nn:        NFP Net device to reconfigure
 * @mbox_cmd:  The value for the mailbox command
 *
 * Helper function for mailbox updates
 *
 * Return: Negative errno on error, 0 on success
 */
static int nfp_net_reconfig_mbox(struct nfp_net *nn, u32 mbox_cmd)
{
	int ret;

	nn_writeq(nn, NFP_NET_CFG_MBOX_CMD, mbox_cmd);

	ret = nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_MBOX);
	if (ret) {
		nn_err(nn, "Mailbox update error\n");
		return ret;
	}

	return -nn_readl(nn, NFP_NET_CFG_MBOX_RET);
}

309 310 311 312 313 314 315 316
/* Interrupt configuration and handling
 */

/**
 * nfp_net_irq_unmask() - Unmask automasked interrupt
 * @nn:       NFP Network structure
 * @entry_nr: MSI-X table entry
 *
J
Jakub Kicinski 已提交
317
 * Clear the ICR for the IRQ entry.
318 319 320 321 322 323 324 325
 */
static void nfp_net_irq_unmask(struct nfp_net *nn, unsigned int entry_nr)
{
	nn_writeb(nn, NFP_NET_CFG_ICR(entry_nr), NFP_NET_CFG_ICR_UNMASKED);
	nn_pci_flush(nn);
}

/**
326 327 328 329 330
 * nfp_net_irqs_alloc() - allocates MSI-X irqs
 * @pdev:        PCI device structure
 * @irq_entries: Array to be initialized and used to hold the irq entries
 * @min_irqs:    Minimal acceptable number of interrupts
 * @wanted_irqs: Target number of interrupts to allocate
331
 *
332
 * Return: Number of irqs obtained or 0 on error.
333
 */
334 335 336
unsigned int
nfp_net_irqs_alloc(struct pci_dev *pdev, struct msix_entry *irq_entries,
		   unsigned int min_irqs, unsigned int wanted_irqs)
337
{
338 339
	unsigned int i;
	int got_irqs;
340

341 342
	for (i = 0; i < wanted_irqs; i++)
		irq_entries[i].entry = i;
343

344 345 346 347 348
	got_irqs = pci_enable_msix_range(pdev, irq_entries,
					 min_irqs, wanted_irqs);
	if (got_irqs < 0) {
		dev_err(&pdev->dev, "Failed to enable %d-%d MSI-X (err=%d)\n",
			min_irqs, wanted_irqs, got_irqs);
349 350 351
		return 0;
	}

352 353 354 355 356
	if (got_irqs < wanted_irqs)
		dev_warn(&pdev->dev, "Unable to allocate %d IRQs got only %d\n",
			 wanted_irqs, got_irqs);

	return got_irqs;
357 358 359
}

/**
360 361 362 363
 * nfp_net_irqs_assign() - Assign interrupts allocated externally to netdev
 * @nn:		 NFP Network structure
 * @irq_entries: Table of allocated interrupts
 * @n:		 Size of @irq_entries (number of entries to grab)
364
 *
365 366
 * After interrupts are allocated with nfp_net_irqs_alloc() this function
 * should be called to assign them to a specific netdev (port).
367
 */
368 369 370
void
nfp_net_irqs_assign(struct nfp_net *nn, struct msix_entry *irq_entries,
		    unsigned int n)
371
{
372 373
	struct nfp_net_dp *dp = &nn->dp;

374
	nn->max_r_vecs = n - NFP_NET_NON_Q_VECTORS;
375
	dp->num_r_vecs = nn->max_r_vecs;
376

377
	memcpy(nn->irq_entries, irq_entries, sizeof(*irq_entries) * n);
378

379 380
	if (dp->num_rx_rings > dp->num_r_vecs ||
	    dp->num_tx_rings > dp->num_r_vecs)
381 382 383
		dev_warn(nn->dp.dev, "More rings (%d,%d) than vectors (%d).\n",
			 dp->num_rx_rings, dp->num_tx_rings,
			 dp->num_r_vecs);
384

385 386 387
	dp->num_rx_rings = min(dp->num_r_vecs, dp->num_rx_rings);
	dp->num_tx_rings = min(dp->num_r_vecs, dp->num_tx_rings);
	dp->num_stack_tx_rings = dp->num_tx_rings;
388 389 390 391
}

/**
 * nfp_net_irqs_disable() - Disable interrupts
392
 * @pdev:        PCI device structure
393 394 395
 *
 * Undoes what @nfp_net_irqs_alloc() does.
 */
396
void nfp_net_irqs_disable(struct pci_dev *pdev)
397
{
398
	pci_disable_msix(pdev);
399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
}

/**
 * nfp_net_irq_rxtx() - Interrupt service routine for RX/TX rings.
 * @irq:      Interrupt
 * @data:     Opaque data structure
 *
 * Return: Indicate if the interrupt has been handled.
 */
static irqreturn_t nfp_net_irq_rxtx(int irq, void *data)
{
	struct nfp_net_r_vector *r_vec = data;

	napi_schedule_irqoff(&r_vec->napi);

	/* The FW auto-masks any interrupt, either via the MASK bit in
	 * the MSI-X table or via the per entry ICR field.  So there
	 * is no need to disable interrupts here.
	 */
	return IRQ_HANDLED;
}

J
Jakub Kicinski 已提交
421 422 423 424 425 426 427 428 429
static irqreturn_t nfp_ctrl_irq_rxtx(int irq, void *data)
{
	struct nfp_net_r_vector *r_vec = data;

	tasklet_schedule(&r_vec->tasklet);

	return IRQ_HANDLED;
}

430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448
/**
 * nfp_net_read_link_status() - Reread link status from control BAR
 * @nn:       NFP Network structure
 */
static void nfp_net_read_link_status(struct nfp_net *nn)
{
	unsigned long flags;
	bool link_up;
	u32 sts;

	spin_lock_irqsave(&nn->link_status_lock, flags);

	sts = nn_readl(nn, NFP_NET_CFG_STS);
	link_up = !!(sts & NFP_NET_CFG_STS_LINK);

	if (nn->link_up == link_up)
		goto out;

	nn->link_up = link_up;
449 450
	if (nn->port)
		set_bit(NFP_PORT_CHANGED, &nn->port->flags);
451 452

	if (nn->link_up) {
453 454
		netif_carrier_on(nn->dp.netdev);
		netdev_info(nn->dp.netdev, "NIC Link is Up\n");
455
	} else {
456 457
		netif_carrier_off(nn->dp.netdev);
		netdev_info(nn->dp.netdev, "NIC Link is Down\n");
458 459 460 461 462 463 464 465 466 467 468 469 470 471 472
	}
out:
	spin_unlock_irqrestore(&nn->link_status_lock, flags);
}

/**
 * nfp_net_irq_lsc() - Interrupt service routine for link state changes
 * @irq:      Interrupt
 * @data:     Opaque data structure
 *
 * Return: Indicate if the interrupt has been handled.
 */
static irqreturn_t nfp_net_irq_lsc(int irq, void *data)
{
	struct nfp_net *nn = data;
473 474 475
	struct msix_entry *entry;

	entry = &nn->irq_entries[NFP_NET_IRQ_LSC_IDX];
476 477 478

	nfp_net_read_link_status(nn);

479
	nfp_net_irq_unmask(nn, entry->entry);
480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502

	return IRQ_HANDLED;
}

/**
 * nfp_net_irq_exn() - Interrupt service routine for exceptions
 * @irq:      Interrupt
 * @data:     Opaque data structure
 *
 * Return: Indicate if the interrupt has been handled.
 */
static irqreturn_t nfp_net_irq_exn(int irq, void *data)
{
	struct nfp_net *nn = data;

	nn_err(nn, "%s: UNIMPLEMENTED.\n", __func__);
	/* XXX TO BE IMPLEMENTED */
	return IRQ_HANDLED;
}

/**
 * nfp_net_tx_ring_init() - Fill in the boilerplate for a TX ring
 * @tx_ring:  TX ring structure
503 504
 * @r_vec:    IRQ vector servicing this ring
 * @idx:      Ring index
505
 * @is_xdp:   Is this an XDP TX ring?
506
 */
507 508
static void
nfp_net_tx_ring_init(struct nfp_net_tx_ring *tx_ring,
509 510
		     struct nfp_net_r_vector *r_vec, unsigned int idx,
		     bool is_xdp)
511 512 513
{
	struct nfp_net *nn = r_vec->nfp_net;

514 515
	tx_ring->idx = idx;
	tx_ring->r_vec = r_vec;
516
	tx_ring->is_xdp = is_xdp;
517
	u64_stats_init(&tx_ring->r_vec->tx_sync);
518

519 520 521 522 523 524 525
	tx_ring->qcidx = tx_ring->idx * nn->stride_tx;
	tx_ring->qcp_q = nn->tx_bar + NFP_QCP_QUEUE_OFF(tx_ring->qcidx);
}

/**
 * nfp_net_rx_ring_init() - Fill in the boilerplate for a RX ring
 * @rx_ring:  RX ring structure
526 527
 * @r_vec:    IRQ vector servicing this ring
 * @idx:      Ring index
528
 */
529 530 531
static void
nfp_net_rx_ring_init(struct nfp_net_rx_ring *rx_ring,
		     struct nfp_net_r_vector *r_vec, unsigned int idx)
532 533 534
{
	struct nfp_net *nn = r_vec->nfp_net;

535 536
	rx_ring->idx = idx;
	rx_ring->r_vec = r_vec;
537
	u64_stats_init(&rx_ring->r_vec->rx_sync);
538

539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562
	rx_ring->fl_qcidx = rx_ring->idx * nn->stride_rx;
	rx_ring->qcp_fl = nn->rx_bar + NFP_QCP_QUEUE_OFF(rx_ring->fl_qcidx);
}

/**
 * nfp_net_aux_irq_request() - Request an auxiliary interrupt (LSC or EXN)
 * @nn:		NFP Network structure
 * @ctrl_offset: Control BAR offset where IRQ configuration should be written
 * @format:	printf-style format to construct the interrupt name
 * @name:	Pointer to allocated space for interrupt name
 * @name_sz:	Size of space for interrupt name
 * @vector_idx:	Index of MSI-X vector used for this interrupt
 * @handler:	IRQ handler to register for this interrupt
 */
static int
nfp_net_aux_irq_request(struct nfp_net *nn, u32 ctrl_offset,
			const char *format, char *name, size_t name_sz,
			unsigned int vector_idx, irq_handler_t handler)
{
	struct msix_entry *entry;
	int err;

	entry = &nn->irq_entries[vector_idx];

J
Jakub Kicinski 已提交
563
	snprintf(name, name_sz, format, nfp_net_name(nn));
564 565 566 567 568 569
	err = request_irq(entry->vector, handler, 0, name, nn);
	if (err) {
		nn_err(nn, "Failed to request IRQ %d (err=%d).\n",
		       entry->vector, err);
		return err;
	}
570
	nn_writeb(nn, ctrl_offset, entry->entry);
571
	nfp_net_irq_unmask(nn, entry->entry);
572 573 574 575 576 577 578 579 580 581 582 583 584 585

	return 0;
}

/**
 * nfp_net_aux_irq_free() - Free an auxiliary interrupt (LSC or EXN)
 * @nn:		NFP Network structure
 * @ctrl_offset: Control BAR offset where IRQ configuration should be written
 * @vector_idx:	Index of MSI-X vector used for this interrupt
 */
static void nfp_net_aux_irq_free(struct nfp_net *nn, u32 ctrl_offset,
				 unsigned int vector_idx)
{
	nn_writeb(nn, ctrl_offset, 0xff);
586
	nn_pci_flush(nn);
587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
	free_irq(nn->irq_entries[vector_idx].vector, nn);
}

/* Transmit
 *
 * One queue controller peripheral queue is used for transmit.  The
 * driver en-queues packets for transmit by advancing the write
 * pointer.  The device indicates that packets have transmitted by
 * advancing the read pointer.  The driver maintains a local copy of
 * the read and write pointer in @struct nfp_net_tx_ring.  The driver
 * keeps @wr_p in sync with the queue controller write pointer and can
 * determine how many packets have been transmitted by comparing its
 * copy of the read pointer @rd_p with the read pointer maintained by
 * the queue controller peripheral.
 */

/**
 * nfp_net_tx_full() - Check if the TX ring is full
 * @tx_ring: TX ring to check
 * @dcnt:    Number of descriptors that need to be enqueued (must be >= 1)
 *
 * This function checks, based on the *host copy* of read/write
 * pointer if a given TX ring is full.  The real TX queue may have
 * some newly made available slots.
 *
 * Return: True if the ring is full.
 */
614
static int nfp_net_tx_full(struct nfp_net_tx_ring *tx_ring, int dcnt)
615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659
{
	return (tx_ring->wr_p - tx_ring->rd_p) >= (tx_ring->cnt - dcnt);
}

/* Wrappers for deciding when to stop and restart TX queues */
static int nfp_net_tx_ring_should_wake(struct nfp_net_tx_ring *tx_ring)
{
	return !nfp_net_tx_full(tx_ring, MAX_SKB_FRAGS * 4);
}

static int nfp_net_tx_ring_should_stop(struct nfp_net_tx_ring *tx_ring)
{
	return nfp_net_tx_full(tx_ring, MAX_SKB_FRAGS + 1);
}

/**
 * nfp_net_tx_ring_stop() - stop tx ring
 * @nd_q:    netdev queue
 * @tx_ring: driver tx queue structure
 *
 * Safely stop TX ring.  Remember that while we are running .start_xmit()
 * someone else may be cleaning the TX ring completions so we need to be
 * extra careful here.
 */
static void nfp_net_tx_ring_stop(struct netdev_queue *nd_q,
				 struct nfp_net_tx_ring *tx_ring)
{
	netif_tx_stop_queue(nd_q);

	/* We can race with the TX completion out of NAPI so recheck */
	smp_mb();
	if (unlikely(nfp_net_tx_ring_should_wake(tx_ring)))
		netif_tx_start_queue(nd_q);
}

/**
 * nfp_net_tx_tso() - Set up Tx descriptor for LSO
 * @r_vec: per-ring structure
 * @txbuf: Pointer to driver soft TX descriptor
 * @txd: Pointer to HW TX descriptor
 * @skb: Pointer to SKB
 *
 * Set up Tx descriptor for LSO, do nothing for non-LSO skbs.
 * Return error on packet header greater than maximum supported LSO header size.
 */
660
static void nfp_net_tx_tso(struct nfp_net_r_vector *r_vec,
661 662 663 664 665 666 667 668 669
			   struct nfp_net_tx_buf *txbuf,
			   struct nfp_net_tx_desc *txd, struct sk_buff *skb)
{
	u32 hdrlen;
	u16 mss;

	if (!skb_is_gso(skb))
		return;

E
Edwin Peer 已提交
670 671 672
	if (!skb->encapsulation) {
		txd->l3_offset = skb_network_offset(skb);
		txd->l4_offset = skb_transport_offset(skb);
673
		hdrlen = skb_transport_offset(skb) + tcp_hdrlen(skb);
E
Edwin Peer 已提交
674 675 676
	} else {
		txd->l3_offset = skb_inner_network_offset(skb);
		txd->l4_offset = skb_inner_transport_offset(skb);
677 678
		hdrlen = skb_inner_transport_header(skb) - skb->data +
			inner_tcp_hdrlen(skb);
E
Edwin Peer 已提交
679
	}
680 681 682 683 684

	txbuf->pkt_cnt = skb_shinfo(skb)->gso_segs;
	txbuf->real_len += hdrlen * (txbuf->pkt_cnt - 1);

	mss = skb_shinfo(skb)->gso_size & PCIE_DESC_TX_MSS_MASK;
685
	txd->lso_hdrlen = hdrlen;
686 687 688 689 690 691 692 693 694 695
	txd->mss = cpu_to_le16(mss);
	txd->flags |= PCIE_DESC_TX_LSO;

	u64_stats_update_begin(&r_vec->tx_sync);
	r_vec->tx_lso++;
	u64_stats_update_end(&r_vec->tx_sync);
}

/**
 * nfp_net_tx_csum() - Set TX CSUM offload flags in TX descriptor
696
 * @dp:  NFP Net data path struct
697 698 699 700 701 702 703 704
 * @r_vec: per-ring structure
 * @txbuf: Pointer to driver soft TX descriptor
 * @txd: Pointer to TX descriptor
 * @skb: Pointer to SKB
 *
 * This function sets the TX checksum flags in the TX descriptor based
 * on the configuration and the protocol of the packet to be transmitted.
 */
705 706
static void nfp_net_tx_csum(struct nfp_net_dp *dp,
			    struct nfp_net_r_vector *r_vec,
707 708 709 710 711 712 713
			    struct nfp_net_tx_buf *txbuf,
			    struct nfp_net_tx_desc *txd, struct sk_buff *skb)
{
	struct ipv6hdr *ipv6h;
	struct iphdr *iph;
	u8 l4_hdr;

714
	if (!(dp->ctrl & NFP_NET_CFG_CTRL_TXCSUM))
715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732
		return;

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

	txd->flags |= PCIE_DESC_TX_CSUM;
	if (skb->encapsulation)
		txd->flags |= PCIE_DESC_TX_ENCAP;

	iph = skb->encapsulation ? inner_ip_hdr(skb) : ip_hdr(skb);
	ipv6h = skb->encapsulation ? inner_ipv6_hdr(skb) : ipv6_hdr(skb);

	if (iph->version == 4) {
		txd->flags |= PCIE_DESC_TX_IP4_CSUM;
		l4_hdr = iph->protocol;
	} else if (ipv6h->version == 6) {
		l4_hdr = ipv6h->nexthdr;
	} else {
733
		nn_dp_warn(dp, "partial checksum but ipv=%x!\n", iph->version);
734 735 736 737 738 739 740 741 742 743 744
		return;
	}

	switch (l4_hdr) {
	case IPPROTO_TCP:
		txd->flags |= PCIE_DESC_TX_TCP_CSUM;
		break;
	case IPPROTO_UDP:
		txd->flags |= PCIE_DESC_TX_UDP_CSUM;
		break;
	default:
745
		nn_dp_warn(dp, "partial checksum but l4 proto=%x!\n", l4_hdr);
746 747 748 749 750 751 752 753 754 755 756
		return;
	}

	u64_stats_update_begin(&r_vec->tx_sync);
	if (skb->encapsulation)
		r_vec->hw_csum_tx_inner += txbuf->pkt_cnt;
	else
		r_vec->hw_csum_tx += txbuf->pkt_cnt;
	u64_stats_update_end(&r_vec->tx_sync);
}

757 758 759 760 761 762 763
static void nfp_net_tx_xmit_more_flush(struct nfp_net_tx_ring *tx_ring)
{
	wmb();
	nfp_qcp_wr_ptr_add(tx_ring->qcp_q, tx_ring->wr_ptr_add);
	tx_ring->wr_ptr_add = 0;
}

764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783
static int nfp_net_prep_port_id(struct sk_buff *skb)
{
	struct metadata_dst *md_dst = skb_metadata_dst(skb);
	unsigned char *data;

	if (likely(!md_dst))
		return 0;
	if (unlikely(md_dst->type != METADATA_HW_PORT_MUX))
		return 0;

	if (unlikely(skb_cow_head(skb, 8)))
		return -ENOMEM;

	data = skb_push(skb, 8);
	put_unaligned_be32(NFP_NET_META_PORTID, data);
	put_unaligned_be32(md_dst->u.port_info.port_id, data + 4);

	return 8;
}

784 785 786 787 788 789 790 791 792 793 794 795
/**
 * nfp_net_tx() - Main transmit entry point
 * @skb:    SKB to transmit
 * @netdev: netdev structure
 *
 * Return: NETDEV_TX_OK on success.
 */
static int nfp_net_tx(struct sk_buff *skb, struct net_device *netdev)
{
	struct nfp_net *nn = netdev_priv(netdev);
	const struct skb_frag_struct *frag;
	struct nfp_net_tx_desc *txd, txdg;
796
	int f, nr_frags, wr_idx, md_bytes;
797
	struct nfp_net_tx_ring *tx_ring;
798 799
	struct nfp_net_r_vector *r_vec;
	struct nfp_net_tx_buf *txbuf;
800
	struct netdev_queue *nd_q;
801
	struct nfp_net_dp *dp;
802 803 804 805
	dma_addr_t dma_addr;
	unsigned int fsize;
	u16 qidx;

806
	dp = &nn->dp;
807
	qidx = skb_get_queue_mapping(skb);
808
	tx_ring = &dp->tx_rings[qidx];
809
	r_vec = tx_ring->r_vec;
810
	nd_q = netdev_get_tx_queue(dp->netdev, qidx);
811 812 813 814

	nr_frags = skb_shinfo(skb)->nr_frags;

	if (unlikely(nfp_net_tx_full(tx_ring, nr_frags + 1))) {
815 816
		nn_dp_warn(dp, "TX ring %d busy. wrp=%u rdp=%u\n",
			   qidx, tx_ring->wr_p, tx_ring->rd_p);
817
		netif_tx_stop_queue(nd_q);
818
		nfp_net_tx_xmit_more_flush(tx_ring);
819 820 821 822 823 824
		u64_stats_update_begin(&r_vec->tx_sync);
		r_vec->tx_busy++;
		u64_stats_update_end(&r_vec->tx_sync);
		return NETDEV_TX_BUSY;
	}

825 826 827 828 829 830 831
	md_bytes = nfp_net_prep_port_id(skb);
	if (unlikely(md_bytes < 0)) {
		nfp_net_tx_xmit_more_flush(tx_ring);
		dev_kfree_skb_any(skb);
		return NETDEV_TX_OK;
	}

832
	/* Start with the head skbuf */
833
	dma_addr = dma_map_single(dp->dev, skb->data, skb_headlen(skb),
834
				  DMA_TO_DEVICE);
835
	if (dma_mapping_error(dp->dev, dma_addr))
836 837
		goto err_free;

838
	wr_idx = D_IDX(tx_ring, tx_ring->wr_p);
839 840 841 842 843 844 845 846 847 848 849

	/* Stash the soft descriptor of the head then initialize it */
	txbuf = &tx_ring->txbufs[wr_idx];
	txbuf->skb = skb;
	txbuf->dma_addr = dma_addr;
	txbuf->fidx = -1;
	txbuf->pkt_cnt = 1;
	txbuf->real_len = skb->len;

	/* Build TX descriptor */
	txd = &tx_ring->txds[wr_idx];
850
	txd->offset_eop = (nr_frags ? 0 : PCIE_DESC_TX_EOP) | md_bytes;
851 852 853 854 855 856
	txd->dma_len = cpu_to_le16(skb_headlen(skb));
	nfp_desc_set_dma_addr(txd, dma_addr);
	txd->data_len = cpu_to_le16(skb->len);

	txd->flags = 0;
	txd->mss = 0;
857
	txd->lso_hdrlen = 0;
858

E
Edwin Peer 已提交
859
	/* Do not reorder - tso may adjust pkt cnt, vlan may override fields */
860 861 862
	nfp_net_tx_tso(r_vec, txbuf, txd, skb);
	nfp_net_tx_csum(dp, r_vec, txbuf, txd, skb);
	if (skb_vlan_tag_present(skb) && dp->ctrl & NFP_NET_CFG_CTRL_TXVLAN) {
863 864 865 866 867 868 869 870 871 872 873 874 875
		txd->flags |= PCIE_DESC_TX_VLAN;
		txd->vlan = cpu_to_le16(skb_vlan_tag_get(skb));
	}

	/* Gather DMA */
	if (nr_frags > 0) {
		/* all descs must match except for in addr, length and eop */
		txdg = *txd;

		for (f = 0; f < nr_frags; f++) {
			frag = &skb_shinfo(skb)->frags[f];
			fsize = skb_frag_size(frag);

876
			dma_addr = skb_frag_dma_map(dp->dev, frag, 0,
877
						    fsize, DMA_TO_DEVICE);
878
			if (dma_mapping_error(dp->dev, dma_addr))
879 880
				goto err_unmap;

881
			wr_idx = D_IDX(tx_ring, wr_idx + 1);
882 883 884 885 886 887 888 889
			tx_ring->txbufs[wr_idx].skb = skb;
			tx_ring->txbufs[wr_idx].dma_addr = dma_addr;
			tx_ring->txbufs[wr_idx].fidx = f;

			txd = &tx_ring->txds[wr_idx];
			*txd = txdg;
			txd->dma_len = cpu_to_le16(fsize);
			nfp_desc_set_dma_addr(txd, dma_addr);
890
			txd->offset_eop |=
891 892 893 894 895 896 897 898 899 900
				(f == nr_frags - 1) ? PCIE_DESC_TX_EOP : 0;
		}

		u64_stats_update_begin(&r_vec->tx_sync);
		r_vec->tx_gather++;
		u64_stats_update_end(&r_vec->tx_sync);
	}

	netdev_tx_sent_queue(nd_q, txbuf->real_len);

901 902
	skb_tx_timestamp(skb);

903 904 905 906 907
	tx_ring->wr_p += nr_frags + 1;
	if (nfp_net_tx_ring_should_stop(tx_ring))
		nfp_net_tx_ring_stop(nd_q, tx_ring);

	tx_ring->wr_ptr_add += nr_frags + 1;
908 909
	if (!skb->xmit_more || netif_xmit_stopped(nd_q))
		nfp_net_tx_xmit_more_flush(tx_ring);
910 911 912 913

	return NETDEV_TX_OK;

err_unmap:
914
	while (--f >= 0) {
915
		frag = &skb_shinfo(skb)->frags[f];
916
		dma_unmap_page(dp->dev, tx_ring->txbufs[wr_idx].dma_addr,
917 918 919 920 921 922 923 924
			       skb_frag_size(frag), DMA_TO_DEVICE);
		tx_ring->txbufs[wr_idx].skb = NULL;
		tx_ring->txbufs[wr_idx].dma_addr = 0;
		tx_ring->txbufs[wr_idx].fidx = -2;
		wr_idx = wr_idx - 1;
		if (wr_idx < 0)
			wr_idx += tx_ring->cnt;
	}
925
	dma_unmap_single(dp->dev, tx_ring->txbufs[wr_idx].dma_addr,
926 927 928 929 930
			 skb_headlen(skb), DMA_TO_DEVICE);
	tx_ring->txbufs[wr_idx].skb = NULL;
	tx_ring->txbufs[wr_idx].dma_addr = 0;
	tx_ring->txbufs[wr_idx].fidx = -2;
err_free:
931
	nn_dp_warn(dp, "Failed to map DMA TX buffer\n");
932
	nfp_net_tx_xmit_more_flush(tx_ring);
933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948
	u64_stats_update_begin(&r_vec->tx_sync);
	r_vec->tx_errors++;
	u64_stats_update_end(&r_vec->tx_sync);
	dev_kfree_skb_any(skb);
	return NETDEV_TX_OK;
}

/**
 * nfp_net_tx_complete() - Handled completed TX packets
 * @tx_ring:   TX ring structure
 *
 * Return: Number of completed TX descriptors
 */
static void nfp_net_tx_complete(struct nfp_net_tx_ring *tx_ring)
{
	struct nfp_net_r_vector *r_vec = tx_ring->r_vec;
949
	struct nfp_net_dp *dp = &r_vec->nfp_net->dp;
950 951 952 953 954 955 956 957 958
	const struct skb_frag_struct *frag;
	struct netdev_queue *nd_q;
	u32 done_pkts = 0, done_bytes = 0;
	struct sk_buff *skb;
	int todo, nr_frags;
	u32 qcp_rd_p;
	int fidx;
	int idx;

959 960 961
	if (tx_ring->wr_p == tx_ring->rd_p)
		return;

962 963 964 965 966 967
	/* Work out how many descriptors have been transmitted */
	qcp_rd_p = nfp_qcp_rd_ptr_read(tx_ring->qcp_q);

	if (qcp_rd_p == tx_ring->qcp_rd_p)
		return;

968
	todo = D_IDX(tx_ring, qcp_rd_p - tx_ring->qcp_rd_p);
969 970

	while (todo--) {
971
		idx = D_IDX(tx_ring, tx_ring->rd_p++);
972 973 974 975 976 977 978 979 980 981

		skb = tx_ring->txbufs[idx].skb;
		if (!skb)
			continue;

		nr_frags = skb_shinfo(skb)->nr_frags;
		fidx = tx_ring->txbufs[idx].fidx;

		if (fidx == -1) {
			/* unmap head */
982
			dma_unmap_single(dp->dev, tx_ring->txbufs[idx].dma_addr,
983 984 985 986 987 988 989
					 skb_headlen(skb), DMA_TO_DEVICE);

			done_pkts += tx_ring->txbufs[idx].pkt_cnt;
			done_bytes += tx_ring->txbufs[idx].real_len;
		} else {
			/* unmap fragment */
			frag = &skb_shinfo(skb)->frags[fidx];
990
			dma_unmap_page(dp->dev, tx_ring->txbufs[idx].dma_addr,
991 992 993 994 995
				       skb_frag_size(frag), DMA_TO_DEVICE);
		}

		/* check for last gather fragment */
		if (fidx == nr_frags - 1)
J
Jakub Kicinski 已提交
996
			dev_consume_skb_any(skb);
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009

		tx_ring->txbufs[idx].dma_addr = 0;
		tx_ring->txbufs[idx].skb = NULL;
		tx_ring->txbufs[idx].fidx = -2;
	}

	tx_ring->qcp_rd_p = qcp_rd_p;

	u64_stats_update_begin(&r_vec->tx_sync);
	r_vec->tx_bytes += done_bytes;
	r_vec->tx_pkts += done_pkts;
	u64_stats_update_end(&r_vec->tx_sync);

J
Jakub Kicinski 已提交
1010 1011 1012
	if (!dp->netdev)
		return;

1013
	nd_q = netdev_get_tx_queue(dp->netdev, tx_ring->idx);
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
	netdev_tx_completed_queue(nd_q, done_pkts, done_bytes);
	if (nfp_net_tx_ring_should_wake(tx_ring)) {
		/* Make sure TX thread will see updated tx_ring->rd_p */
		smp_mb();

		if (unlikely(netif_tx_queue_stopped(nd_q)))
			netif_tx_wake_queue(nd_q);
	}

	WARN_ONCE(tx_ring->wr_p - tx_ring->rd_p > tx_ring->cnt,
		  "TX ring corruption rd_p=%u wr_p=%u cnt=%u\n",
		  tx_ring->rd_p, tx_ring->wr_p, tx_ring->cnt);
}

1028
static bool nfp_net_xdp_complete(struct nfp_net_tx_ring *tx_ring)
1029 1030 1031
{
	struct nfp_net_r_vector *r_vec = tx_ring->r_vec;
	u32 done_pkts = 0, done_bytes = 0;
1032
	bool done_all;
1033 1034 1035 1036 1037 1038 1039
	int idx, todo;
	u32 qcp_rd_p;

	/* Work out how many descriptors have been transmitted */
	qcp_rd_p = nfp_qcp_rd_ptr_read(tx_ring->qcp_q);

	if (qcp_rd_p == tx_ring->qcp_rd_p)
1040
		return true;
1041

1042
	todo = D_IDX(tx_ring, qcp_rd_p - tx_ring->qcp_rd_p);
1043

1044 1045 1046
	done_all = todo <= NFP_NET_XDP_MAX_COMPLETE;
	todo = min(todo, NFP_NET_XDP_MAX_COMPLETE);

1047
	tx_ring->qcp_rd_p = D_IDX(tx_ring, tx_ring->qcp_rd_p + todo);
1048

1049
	done_pkts = todo;
1050
	while (todo--) {
1051
		idx = D_IDX(tx_ring, tx_ring->rd_p);
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
		tx_ring->rd_p++;

		done_bytes += tx_ring->txbufs[idx].real_len;
	}

	u64_stats_update_begin(&r_vec->tx_sync);
	r_vec->tx_bytes += done_bytes;
	r_vec->tx_pkts += done_pkts;
	u64_stats_update_end(&r_vec->tx_sync);

	WARN_ONCE(tx_ring->wr_p - tx_ring->rd_p > tx_ring->cnt,
1063
		  "XDP TX ring corruption rd_p=%u wr_p=%u cnt=%u\n",
1064
		  tx_ring->rd_p, tx_ring->wr_p, tx_ring->cnt);
1065 1066

	return done_all;
1067 1068
}

1069
/**
1070
 * nfp_net_tx_ring_reset() - Free any untransmitted buffers and reset pointers
1071
 * @dp:		NFP Net data path struct
1072
 * @tx_ring:	TX ring structure
1073 1074 1075
 *
 * Assumes that the device is stopped
 */
1076
static void
1077
nfp_net_tx_ring_reset(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring)
1078 1079
{
	const struct skb_frag_struct *frag;
1080
	struct netdev_queue *nd_q;
1081

1082
	while (!tx_ring->is_xdp && tx_ring->rd_p != tx_ring->wr_p) {
1083
		struct nfp_net_tx_buf *tx_buf;
1084 1085
		struct sk_buff *skb;
		int idx, nr_frags;
1086

1087
		idx = D_IDX(tx_ring, tx_ring->rd_p);
1088
		tx_buf = &tx_ring->txbufs[idx];
1089

1090 1091
		skb = tx_ring->txbufs[idx].skb;
		nr_frags = skb_shinfo(skb)->nr_frags;
1092

1093 1094 1095 1096 1097 1098 1099 1100 1101
		if (tx_buf->fidx == -1) {
			/* unmap head */
			dma_unmap_single(dp->dev, tx_buf->dma_addr,
					 skb_headlen(skb), DMA_TO_DEVICE);
		} else {
			/* unmap fragment */
			frag = &skb_shinfo(skb)->frags[tx_buf->fidx];
			dma_unmap_page(dp->dev, tx_buf->dma_addr,
				       skb_frag_size(frag), DMA_TO_DEVICE);
1102
		}
1103

1104 1105 1106 1107
		/* check for last gather fragment */
		if (tx_buf->fidx == nr_frags - 1)
			dev_kfree_skb_any(skb);

1108 1109 1110
		tx_buf->dma_addr = 0;
		tx_buf->skb = NULL;
		tx_buf->fidx = -2;
1111 1112 1113 1114 1115

		tx_ring->qcp_rd_p++;
		tx_ring->rd_p++;
	}

1116 1117 1118 1119 1120 1121
	memset(tx_ring->txds, 0, sizeof(*tx_ring->txds) * tx_ring->cnt);
	tx_ring->wr_p = 0;
	tx_ring->rd_p = 0;
	tx_ring->qcp_rd_p = 0;
	tx_ring->wr_ptr_add = 0;

J
Jakub Kicinski 已提交
1122
	if (tx_ring->is_xdp || !dp->netdev)
1123 1124
		return;

1125
	nd_q = netdev_get_tx_queue(dp->netdev, tx_ring->idx);
1126 1127 1128 1129 1130 1131 1132 1133
	netdev_tx_reset_queue(nd_q);
}

static void nfp_net_tx_timeout(struct net_device *netdev)
{
	struct nfp_net *nn = netdev_priv(netdev);
	int i;

1134
	for (i = 0; i < nn->dp.netdev->real_num_tx_queues; i++) {
1135 1136 1137 1138 1139 1140 1141 1142 1143
		if (!netif_tx_queue_stopped(netdev_get_tx_queue(netdev, i)))
			continue;
		nn_warn(nn, "TX timeout on ring: %d\n", i);
	}
	nn_warn(nn, "TX watchdog timeout\n");
}

/* Receive processing
 */
1144
static unsigned int
1145
nfp_net_calc_fl_bufsz(struct nfp_net_dp *dp)
1146 1147 1148
{
	unsigned int fl_bufsz;

1149
	fl_bufsz = NFP_NET_RX_BUF_HEADROOM;
1150
	fl_bufsz += dp->rx_dma_off;
1151
	if (dp->rx_offset == NFP_NET_CFG_RX_OFFSET_DYNAMIC)
1152
		fl_bufsz += NFP_NET_MAX_PREPEND;
1153
	else
1154
		fl_bufsz += dp->rx_offset;
1155
	fl_bufsz += ETH_HLEN + VLAN_HLEN * 2 + dp->mtu;
1156

1157 1158 1159
	fl_bufsz = SKB_DATA_ALIGN(fl_bufsz);
	fl_bufsz += SKB_DATA_ALIGN(sizeof(struct skb_shared_info));

1160 1161
	return fl_bufsz;
}
1162

1163 1164 1165 1166 1167 1168 1169 1170 1171
static void
nfp_net_free_frag(void *frag, bool xdp)
{
	if (!xdp)
		skb_free_frag(frag);
	else
		__free_page(virt_to_page(frag));
}

1172
/**
1173
 * nfp_net_rx_alloc_one() - Allocate and map page frag for RX
1174
 * @dp:		NFP Net data path struct
1175 1176
 * @dma_addr:	Pointer to storage for DMA address (output param)
 *
1177
 * This function will allcate a new page frag, map it for DMA.
1178
 *
1179
 * Return: allocated page frag or NULL on failure.
1180
 */
1181
static void *nfp_net_rx_alloc_one(struct nfp_net_dp *dp, dma_addr_t *dma_addr)
1182
{
1183
	void *frag;
1184

1185
	if (!dp->xdp_prog) {
1186
		frag = netdev_alloc_frag(dp->fl_bufsz);
1187 1188 1189
	} else {
		struct page *page;

M
Mel Gorman 已提交
1190
		page = alloc_page(GFP_KERNEL);
1191 1192
		frag = page ? page_address(page) : NULL;
	}
1193
	if (!frag) {
1194
		nn_dp_warn(dp, "Failed to alloc receive page frag\n");
1195 1196 1197
		return NULL;
	}

1198
	*dma_addr = nfp_net_dma_map_rx(dp, frag);
1199
	if (dma_mapping_error(dp->dev, *dma_addr)) {
1200
		nfp_net_free_frag(frag, dp->xdp_prog);
1201
		nn_dp_warn(dp, "Failed to map DMA RX buffer\n");
1202 1203 1204
		return NULL;
	}

1205
	return frag;
1206 1207
}

1208
static void *nfp_net_napi_alloc_one(struct nfp_net_dp *dp, dma_addr_t *dma_addr)
1209 1210 1211
{
	void *frag;

1212
	if (!dp->xdp_prog) {
1213
		frag = napi_alloc_frag(dp->fl_bufsz);
1214 1215
		if (unlikely(!frag))
			return NULL;
1216 1217 1218
	} else {
		struct page *page;

J
Jakub Kicinski 已提交
1219
		page = dev_alloc_page();
1220 1221 1222
		if (unlikely(!page))
			return NULL;
		frag = page_address(page);
1223 1224
	}

1225
	*dma_addr = nfp_net_dma_map_rx(dp, frag);
1226 1227 1228
	if (dma_mapping_error(dp->dev, *dma_addr)) {
		nfp_net_free_frag(frag, dp->xdp_prog);
		nn_dp_warn(dp, "Failed to map DMA RX buffer\n");
1229 1230 1231 1232 1233 1234
		return NULL;
	}

	return frag;
}

1235 1236
/**
 * nfp_net_rx_give_one() - Put mapped skb on the software and hardware rings
1237
 * @dp:		NFP Net data path struct
1238
 * @rx_ring:	RX ring structure
1239
 * @frag:	page fragment buffer
1240 1241
 * @dma_addr:	DMA address of skb mapping
 */
1242 1243
static void nfp_net_rx_give_one(const struct nfp_net_dp *dp,
				struct nfp_net_rx_ring *rx_ring,
1244
				void *frag, dma_addr_t dma_addr)
1245 1246 1247
{
	unsigned int wr_idx;

1248
	wr_idx = D_IDX(rx_ring, rx_ring->wr_p);
1249

1250 1251
	nfp_net_dma_sync_dev_rx(dp, dma_addr);

1252
	/* Stash SKB and DMA address away */
1253
	rx_ring->rxbufs[wr_idx].frag = frag;
1254 1255 1256 1257 1258
	rx_ring->rxbufs[wr_idx].dma_addr = dma_addr;

	/* Fill freelist descriptor */
	rx_ring->rxds[wr_idx].fld.reserved = 0;
	rx_ring->rxds[wr_idx].fld.meta_len_dd = 0;
1259 1260
	nfp_desc_set_dma_addr(&rx_ring->rxds[wr_idx].fld,
			      dma_addr + dp->rx_dma_off);
1261 1262

	rx_ring->wr_p++;
1263
	if (!(rx_ring->wr_p % NFP_NET_FL_BATCH)) {
1264 1265 1266 1267
		/* Update write pointer of the freelist queue. Make
		 * sure all writes are flushed before telling the hardware.
		 */
		wmb();
1268
		nfp_qcp_wr_ptr_add(rx_ring->qcp_fl, NFP_NET_FL_BATCH);
1269 1270 1271 1272
	}
}

/**
1273 1274
 * nfp_net_rx_ring_reset() - Reflect in SW state of freelist after disable
 * @rx_ring:	RX ring structure
1275
 *
1276 1277
 * Warning: Do *not* call if ring buffers were never put on the FW freelist
 *	    (i.e. device was not enabled)!
1278
 */
1279
static void nfp_net_rx_ring_reset(struct nfp_net_rx_ring *rx_ring)
1280
{
1281
	unsigned int wr_idx, last_idx;
1282

1283
	/* Move the empty entry to the end of the list */
1284
	wr_idx = D_IDX(rx_ring, rx_ring->wr_p);
1285 1286
	last_idx = rx_ring->cnt - 1;
	rx_ring->rxbufs[wr_idx].dma_addr = rx_ring->rxbufs[last_idx].dma_addr;
1287
	rx_ring->rxbufs[wr_idx].frag = rx_ring->rxbufs[last_idx].frag;
1288
	rx_ring->rxbufs[last_idx].dma_addr = 0;
1289
	rx_ring->rxbufs[last_idx].frag = NULL;
1290

1291 1292 1293 1294
	memset(rx_ring->rxds, 0, sizeof(*rx_ring->rxds) * rx_ring->cnt);
	rx_ring->wr_p = 0;
	rx_ring->rd_p = 0;
}
1295

1296 1297
/**
 * nfp_net_rx_ring_bufs_free() - Free any buffers currently on the RX ring
1298
 * @dp:		NFP Net data path struct
1299 1300 1301 1302 1303 1304 1305
 * @rx_ring:	RX ring to remove buffers from
 *
 * Assumes that the device is stopped and buffers are in [0, ring->cnt - 1)
 * entries.  After device is disabled nfp_net_rx_ring_reset() must be called
 * to restore required ring geometry.
 */
static void
1306
nfp_net_rx_ring_bufs_free(struct nfp_net_dp *dp,
1307
			  struct nfp_net_rx_ring *rx_ring)
1308 1309
{
	unsigned int i;
1310

1311 1312 1313 1314 1315
	for (i = 0; i < rx_ring->cnt - 1; i++) {
		/* NULL skb can only happen when initial filling of the ring
		 * fails to allocate enough buffers and calls here to free
		 * already allocated ones.
		 */
1316
		if (!rx_ring->rxbufs[i].frag)
1317 1318
			continue;

1319
		nfp_net_dma_unmap_rx(dp, rx_ring->rxbufs[i].dma_addr);
1320
		nfp_net_free_frag(rx_ring->rxbufs[i].frag, dp->xdp_prog);
1321
		rx_ring->rxbufs[i].dma_addr = 0;
1322
		rx_ring->rxbufs[i].frag = NULL;
1323 1324 1325 1326
	}
}

/**
1327
 * nfp_net_rx_ring_bufs_alloc() - Fill RX ring with buffers (don't give to FW)
1328
 * @dp:		NFP Net data path struct
1329
 * @rx_ring:	RX ring to remove buffers from
1330
 */
1331
static int
1332
nfp_net_rx_ring_bufs_alloc(struct nfp_net_dp *dp,
1333
			   struct nfp_net_rx_ring *rx_ring)
1334
{
1335 1336 1337 1338
	struct nfp_net_rx_buf *rxbufs;
	unsigned int i;

	rxbufs = rx_ring->rxbufs;
1339

1340
	for (i = 0; i < rx_ring->cnt - 1; i++) {
1341
		rxbufs[i].frag = nfp_net_rx_alloc_one(dp, &rxbufs[i].dma_addr);
1342
		if (!rxbufs[i].frag) {
1343
			nfp_net_rx_ring_bufs_free(dp, rx_ring);
1344 1345 1346 1347 1348 1349 1350
			return -ENOMEM;
		}
	}

	return 0;
}

1351 1352
/**
 * nfp_net_rx_ring_fill_freelist() - Give buffers from the ring to FW
1353
 * @dp:	     NFP Net data path struct
1354 1355
 * @rx_ring: RX ring to fill
 */
1356 1357 1358
static void
nfp_net_rx_ring_fill_freelist(struct nfp_net_dp *dp,
			      struct nfp_net_rx_ring *rx_ring)
1359 1360 1361 1362
{
	unsigned int i;

	for (i = 0; i < rx_ring->cnt - 1; i++)
1363
		nfp_net_rx_give_one(dp, rx_ring, rx_ring->rxbufs[i].frag,
1364 1365 1366
				    rx_ring->rxbufs[i].dma_addr);
}

1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382
/**
 * nfp_net_rx_csum_has_errors() - group check if rxd has any csum errors
 * @flags: RX descriptor flags field in CPU byte order
 */
static int nfp_net_rx_csum_has_errors(u16 flags)
{
	u16 csum_all_checked, csum_all_ok;

	csum_all_checked = flags & __PCIE_DESC_RX_CSUM_ALL;
	csum_all_ok = flags & __PCIE_DESC_RX_CSUM_ALL_OK;

	return csum_all_checked != (csum_all_ok << PCIE_DESC_RX_CSUM_OK_SHIFT);
}

/**
 * nfp_net_rx_csum() - set SKB checksum field based on RX descriptor flags
1383
 * @dp:  NFP Net data path struct
1384 1385
 * @r_vec: per-ring structure
 * @rxd: Pointer to RX descriptor
1386
 * @meta: Parsed metadata prepend
1387 1388
 * @skb: Pointer to SKB
 */
1389 1390
static void nfp_net_rx_csum(struct nfp_net_dp *dp,
			    struct nfp_net_r_vector *r_vec,
1391 1392
			    struct nfp_net_rx_desc *rxd,
			    struct nfp_meta_parsed *meta, struct sk_buff *skb)
1393 1394 1395
{
	skb_checksum_none_assert(skb);

1396
	if (!(dp->netdev->features & NETIF_F_RXCSUM))
1397 1398
		return;

1399 1400 1401 1402 1403 1404 1405 1406 1407
	if (meta->csum_type) {
		skb->ip_summed = meta->csum_type;
		skb->csum = meta->csum;
		u64_stats_update_begin(&r_vec->rx_sync);
		r_vec->hw_csum_rx_ok++;
		u64_stats_update_end(&r_vec->rx_sync);
		return;
	}

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
	if (nfp_net_rx_csum_has_errors(le16_to_cpu(rxd->rxd.flags))) {
		u64_stats_update_begin(&r_vec->rx_sync);
		r_vec->hw_csum_rx_error++;
		u64_stats_update_end(&r_vec->rx_sync);
		return;
	}

	/* Assume that the firmware will never report inner CSUM_OK unless outer
	 * L4 headers were successfully parsed. FW will always report zero UDP
	 * checksum as CSUM_OK.
	 */
	if (rxd->rxd.flags & PCIE_DESC_RX_TCP_CSUM_OK ||
	    rxd->rxd.flags & PCIE_DESC_RX_UDP_CSUM_OK) {
		__skb_incr_checksum_unnecessary(skb);
		u64_stats_update_begin(&r_vec->rx_sync);
		r_vec->hw_csum_rx_ok++;
		u64_stats_update_end(&r_vec->rx_sync);
	}

	if (rxd->rxd.flags & PCIE_DESC_RX_I_TCP_CSUM_OK ||
	    rxd->rxd.flags & PCIE_DESC_RX_I_UDP_CSUM_OK) {
		__skb_incr_checksum_unnecessary(skb);
		u64_stats_update_begin(&r_vec->rx_sync);
		r_vec->hw_csum_rx_inner_ok++;
		u64_stats_update_end(&r_vec->rx_sync);
	}
}

1436 1437 1438
static void
nfp_net_set_hash(struct net_device *netdev, struct nfp_meta_parsed *meta,
		 unsigned int type, __be32 *hash)
1439
{
1440
	if (!(netdev->features & NETIF_F_RXHASH))
1441 1442
		return;

1443
	switch (type) {
1444 1445 1446
	case NFP_NET_RSS_IPV4:
	case NFP_NET_RSS_IPV6:
	case NFP_NET_RSS_IPV6_EX:
1447
		meta->hash_type = PKT_HASH_TYPE_L3;
1448 1449
		break;
	default:
1450
		meta->hash_type = PKT_HASH_TYPE_L4;
1451 1452
		break;
	}
1453 1454

	meta->hash = get_unaligned_be32(hash);
1455 1456
}

1457
static void
1458
nfp_net_set_hash_desc(struct net_device *netdev, struct nfp_meta_parsed *meta,
1459
		      void *data, struct nfp_net_rx_desc *rxd)
1460
{
1461
	struct nfp_net_rx_hash *rx_hash = data;
1462 1463 1464 1465

	if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
		return;

1466
	nfp_net_set_hash(netdev, meta, get_unaligned_be32(&rx_hash->hash_type),
1467 1468 1469 1470
			 &rx_hash->hash);
}

static void *
1471
nfp_net_parse_meta(struct net_device *netdev, struct nfp_meta_parsed *meta,
1472
		   void *data, int meta_len)
1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
{
	u32 meta_info;

	meta_info = get_unaligned_be32(data);
	data += 4;

	while (meta_info) {
		switch (meta_info & NFP_NET_META_FIELD_MASK) {
		case NFP_NET_META_HASH:
			meta_info >>= NFP_NET_META_FIELD_SIZE;
1483
			nfp_net_set_hash(netdev, meta,
1484 1485 1486 1487 1488
					 meta_info & NFP_NET_META_FIELD_MASK,
					 (__be32 *)data);
			data += 4;
			break;
		case NFP_NET_META_MARK:
1489
			meta->mark = get_unaligned_be32(data);
1490 1491
			data += 4;
			break;
1492 1493 1494 1495
		case NFP_NET_META_PORTID:
			meta->portid = get_unaligned_be32(data);
			data += 4;
			break;
1496 1497 1498 1499 1500 1501
		case NFP_NET_META_CSUM:
			meta->csum_type = CHECKSUM_COMPLETE;
			meta->csum =
				(__force __wsum)__get_unaligned_cpu32(data);
			data += 4;
			break;
1502 1503 1504 1505 1506 1507 1508 1509 1510 1511
		default:
			return NULL;
		}

		meta_info >>= NFP_NET_META_FIELD_SIZE;
	}

	return data;
}

1512
static void
1513 1514 1515
nfp_net_rx_drop(const struct nfp_net_dp *dp, struct nfp_net_r_vector *r_vec,
		struct nfp_net_rx_ring *rx_ring, struct nfp_net_rx_buf *rxbuf,
		struct sk_buff *skb)
1516 1517 1518
{
	u64_stats_update_begin(&r_vec->rx_sync);
	r_vec->rx_drops++;
1519 1520 1521 1522 1523
	/* If we have both skb and rxbuf the replacement buffer allocation
	 * must have failed, count this as an alloc failure.
	 */
	if (skb && rxbuf)
		r_vec->rx_replace_buf_alloc_fail++;
1524 1525
	u64_stats_update_end(&r_vec->rx_sync);

1526 1527 1528 1529 1530
	/* skb is build based on the frag, free_skb() would free the frag
	 * so to be able to reuse it we need an extra ref.
	 */
	if (skb && rxbuf && skb->head == rxbuf->frag)
		page_ref_inc(virt_to_head_page(rxbuf->frag));
1531
	if (rxbuf)
1532
		nfp_net_rx_give_one(dp, rx_ring, rxbuf->frag, rxbuf->dma_addr);
1533 1534 1535 1536
	if (skb)
		dev_kfree_skb_any(skb);
}

1537
static bool
1538
nfp_net_tx_xdp_buf(struct nfp_net_dp *dp, struct nfp_net_rx_ring *rx_ring,
1539
		   struct nfp_net_tx_ring *tx_ring,
1540
		   struct nfp_net_rx_buf *rxbuf, unsigned int dma_off,
1541
		   unsigned int pkt_len, bool *completed)
1542 1543 1544 1545 1546 1547
{
	struct nfp_net_tx_buf *txbuf;
	struct nfp_net_tx_desc *txd;
	int wr_idx;

	if (unlikely(nfp_net_tx_full(tx_ring, 1))) {
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557
		if (!*completed) {
			nfp_net_xdp_complete(tx_ring);
			*completed = true;
		}

		if (unlikely(nfp_net_tx_full(tx_ring, 1))) {
			nfp_net_rx_drop(dp, rx_ring->r_vec, rx_ring, rxbuf,
					NULL);
			return false;
		}
1558 1559
	}

1560
	wr_idx = D_IDX(tx_ring, tx_ring->wr_p);
1561 1562 1563

	/* Stash the soft descriptor of the head then initialize it */
	txbuf = &tx_ring->txbufs[wr_idx];
1564 1565 1566

	nfp_net_rx_give_one(dp, rx_ring, txbuf->frag, txbuf->dma_addr);

1567 1568 1569 1570 1571 1572
	txbuf->frag = rxbuf->frag;
	txbuf->dma_addr = rxbuf->dma_addr;
	txbuf->fidx = -1;
	txbuf->pkt_cnt = 1;
	txbuf->real_len = pkt_len;

1573
	dma_sync_single_for_device(dp->dev, rxbuf->dma_addr + dma_off,
1574
				   pkt_len, DMA_BIDIRECTIONAL);
1575 1576 1577 1578 1579

	/* Build TX descriptor */
	txd = &tx_ring->txds[wr_idx];
	txd->offset_eop = PCIE_DESC_TX_EOP;
	txd->dma_len = cpu_to_le16(pkt_len);
1580
	nfp_desc_set_dma_addr(txd, rxbuf->dma_addr + dma_off);
1581 1582 1583 1584
	txd->data_len = cpu_to_le16(pkt_len);

	txd->flags = 0;
	txd->mss = 0;
1585
	txd->lso_hdrlen = 0;
1586 1587 1588

	tx_ring->wr_p++;
	tx_ring->wr_ptr_add++;
1589
	return true;
1590 1591
}

1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605
/**
 * nfp_net_rx() - receive up to @budget packets on @rx_ring
 * @rx_ring:   RX ring to receive from
 * @budget:    NAPI budget
 *
 * Note, this function is separated out from the napi poll function to
 * more cleanly separate packet receive code from other bookkeeping
 * functions performed in the napi poll function.
 *
 * Return: Number of packets received.
 */
static int nfp_net_rx(struct nfp_net_rx_ring *rx_ring, int budget)
{
	struct nfp_net_r_vector *r_vec = rx_ring->r_vec;
1606
	struct nfp_net_dp *dp = &r_vec->nfp_net->dp;
1607 1608
	struct nfp_net_tx_ring *tx_ring;
	struct bpf_prog *xdp_prog;
1609
	bool xdp_tx_cmpl = false;
1610
	unsigned int true_bufsz;
1611
	struct sk_buff *skb;
J
Jakub Kicinski 已提交
1612
	int pkts_polled = 0;
J
Jesper Dangaard Brouer 已提交
1613
	struct xdp_buff xdp;
1614 1615
	int idx;

1616
	rcu_read_lock();
1617 1618
	xdp_prog = READ_ONCE(dp->xdp_prog);
	true_bufsz = xdp_prog ? PAGE_SIZE : dp->fl_bufsz;
J
Jesper Dangaard Brouer 已提交
1619
	xdp.rxq = &rx_ring->xdp_rxq;
1620 1621
	tx_ring = r_vec->xdp_ring;

J
Jakub Kicinski 已提交
1622
	while (pkts_polled < budget) {
1623
		unsigned int meta_len, data_len, meta_off, pkt_len, pkt_off;
1624 1625
		struct nfp_net_rx_buf *rxbuf;
		struct nfp_net_rx_desc *rxd;
1626
		struct nfp_meta_parsed meta;
1627
		struct net_device *netdev;
1628
		dma_addr_t new_dma_addr;
1629
		u32 meta_len_xdp = 0;
1630 1631
		void *new_frag;

1632
		idx = D_IDX(rx_ring, rx_ring->rd_p);
1633 1634

		rxd = &rx_ring->rxds[idx];
J
Jakub Kicinski 已提交
1635
		if (!(rxd->rxd.meta_len_dd & PCIE_DESC_RX_DD))
1636
			break;
J
Jakub Kicinski 已提交
1637

1638 1639 1640 1641 1642
		/* Memory barrier to ensure that we won't do other reads
		 * before the DD bit.
		 */
		dma_rmb();

1643 1644
		memset(&meta, 0, sizeof(meta));

1645 1646 1647
		rx_ring->rd_p++;
		pkts_polled++;

1648
		rxbuf =	&rx_ring->rxbufs[idx];
1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
		/*         < meta_len >
		 *  <-- [rx_offset] -->
		 *  ---------------------------------------------------------
		 * | [XX] |  metadata  |             packet           | XXXX |
		 *  ---------------------------------------------------------
		 *         <---------------- data_len --------------->
		 *
		 * The rx_offset is fixed for all packets, the meta_len can vary
		 * on a packet by packet basis. If rx_offset is set to zero
		 * (_RX_OFFSET_DYNAMIC) metadata starts at the beginning of the
		 * buffer and is immediately followed by the packet (no [XX]).
		 */
1661 1662
		meta_len = rxd->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK;
		data_len = le16_to_cpu(rxd->rxd.data_len);
1663
		pkt_len = data_len - meta_len;
1664

1665
		pkt_off = NFP_NET_RX_BUF_HEADROOM + dp->rx_dma_off;
1666
		if (dp->rx_offset == NFP_NET_CFG_RX_OFFSET_DYNAMIC)
1667
			pkt_off += meta_len;
1668
		else
1669 1670
			pkt_off += dp->rx_offset;
		meta_off = pkt_off - meta_len;
1671 1672 1673 1674

		/* Stats update */
		u64_stats_update_begin(&r_vec->rx_sync);
		r_vec->rx_pkts++;
1675
		r_vec->rx_bytes += pkt_len;
1676 1677
		u64_stats_update_end(&r_vec->rx_sync);

1678 1679 1680 1681
		if (unlikely(meta_len > NFP_NET_MAX_PREPEND ||
			     (dp->rx_offset && meta_len > dp->rx_offset))) {
			nn_dp_warn(dp, "oversized RX packet metadata %u\n",
				   meta_len);
1682
			nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf, NULL);
1683 1684 1685
			continue;
		}

1686 1687 1688
		nfp_net_dma_sync_cpu_rx(dp, rxbuf->dma_addr + meta_off,
					data_len);

1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705
		if (!dp->chained_metadata_format) {
			nfp_net_set_hash_desc(dp->netdev, &meta,
					      rxbuf->frag + meta_off, rxd);
		} else if (meta_len) {
			void *end;

			end = nfp_net_parse_meta(dp->netdev, &meta,
						 rxbuf->frag + meta_off,
						 meta_len);
			if (unlikely(end != rxbuf->frag + pkt_off)) {
				nn_dp_warn(dp, "invalid RX packet metadata\n");
				nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf,
						NULL);
				continue;
			}
		}

1706
		if (xdp_prog && !(rxd->rxd.flags & PCIE_DESC_RX_BPF &&
1707
				  dp->bpf_offload_xdp) && !meta.portid) {
1708
			void *orig_data = rxbuf->frag + pkt_off;
1709
			unsigned int dma_off;
1710 1711
			int act;

1712 1713 1714 1715 1716 1717 1718 1719 1720
			xdp.data_hard_start = rxbuf->frag + NFP_NET_RX_BUF_HEADROOM;
			xdp.data = orig_data;
			xdp.data_meta = orig_data;
			xdp.data_end = orig_data + pkt_len;

			act = bpf_prog_run_xdp(xdp_prog, &xdp);

			pkt_len -= xdp.data - orig_data;
			pkt_off += xdp.data - orig_data;
1721

1722 1723
			switch (act) {
			case XDP_PASS:
1724
				meta_len_xdp = xdp.data - xdp.data_meta;
1725 1726
				break;
			case XDP_TX:
1727
				dma_off = pkt_off - NFP_NET_RX_BUF_HEADROOM;
1728
				if (unlikely(!nfp_net_tx_xdp_buf(dp, rx_ring,
1729
								 tx_ring, rxbuf,
1730
								 dma_off,
1731 1732
								 pkt_len,
								 &xdp_tx_cmpl)))
1733 1734
					trace_xdp_exception(dp->netdev,
							    xdp_prog, act);
1735 1736 1737
				continue;
			default:
				bpf_warn_invalid_xdp_action(act);
1738
				/* fall through */
1739
			case XDP_ABORTED:
1740
				trace_xdp_exception(dp->netdev, xdp_prog, act);
1741
				/* fall through */
1742
			case XDP_DROP:
1743
				nfp_net_rx_give_one(dp, rx_ring, rxbuf->frag,
1744 1745 1746 1747 1748 1749
						    rxbuf->dma_addr);
				continue;
			}
		}

		skb = build_skb(rxbuf->frag, true_bufsz);
1750
		if (unlikely(!skb)) {
1751
			nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf, NULL);
1752 1753
			continue;
		}
1754
		new_frag = nfp_net_napi_alloc_one(dp, &new_dma_addr);
1755
		if (unlikely(!new_frag)) {
1756
			nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf, skb);
1757 1758 1759
			continue;
		}

1760 1761 1762 1763
		nfp_net_dma_unmap_rx(dp, rxbuf->dma_addr);

		nfp_net_rx_give_one(dp, rx_ring, new_frag, new_dma_addr);

1764 1765 1766 1767 1768 1769 1770 1771
		if (likely(!meta.portid)) {
			netdev = dp->netdev;
		} else {
			struct nfp_net *nn;

			nn = netdev_priv(dp->netdev);
			netdev = nfp_app_repr_get(nn->app, meta.portid);
			if (unlikely(!netdev)) {
1772
				nfp_net_rx_drop(dp, r_vec, rx_ring, NULL, skb);
1773 1774 1775 1776 1777
				continue;
			}
			nfp_repr_inc_rx_stats(netdev, pkt_len);
		}

1778
		skb_reserve(skb, pkt_off);
1779 1780
		skb_put(skb, pkt_len);

1781 1782
		skb->mark = meta.mark;
		skb_set_hash(skb, meta.hash, meta.hash_type);
1783

1784
		skb_record_rx_queue(skb, rx_ring->idx);
1785
		skb->protocol = eth_type_trans(skb, netdev);
1786

1787
		nfp_net_rx_csum(dp, r_vec, rxd, &meta, skb);
1788 1789 1790 1791

		if (rxd->rxd.flags & PCIE_DESC_RX_VLAN)
			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
					       le16_to_cpu(rxd->rxd.vlan));
1792 1793
		if (meta_len_xdp)
			skb_metadata_set(skb, meta_len_xdp);
1794 1795 1796 1797

		napi_gro_receive(&rx_ring->r_vec->napi, skb);
	}

1798 1799 1800 1801 1802 1803 1804 1805
	if (xdp_prog) {
		if (tx_ring->wr_ptr_add)
			nfp_net_tx_xmit_more_flush(tx_ring);
		else if (unlikely(tx_ring->wr_p != tx_ring->rd_p) &&
			 !xdp_tx_cmpl)
			if (!nfp_net_xdp_complete(tx_ring))
				pkts_polled = budget;
	}
1806 1807
	rcu_read_unlock();

1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
	return pkts_polled;
}

/**
 * nfp_net_poll() - napi poll function
 * @napi:    NAPI structure
 * @budget:  NAPI budget
 *
 * Return: number of packets polled.
 */
static int nfp_net_poll(struct napi_struct *napi, int budget)
{
	struct nfp_net_r_vector *r_vec =
		container_of(napi, struct nfp_net_r_vector, napi);
1822
	unsigned int pkts_polled = 0;
1823

1824 1825
	if (r_vec->tx_ring)
		nfp_net_tx_complete(r_vec->tx_ring);
1826
	if (r_vec->rx_ring)
1827
		pkts_polled = nfp_net_rx(r_vec->rx_ring, budget);
1828

1829 1830 1831
	if (pkts_polled < budget)
		if (napi_complete_done(napi, pkts_polled))
			nfp_net_irq_unmask(r_vec->nfp_net, r_vec->irq_entry);
1832 1833 1834 1835

	return pkts_polled;
}

J
Jakub Kicinski 已提交
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 1861 1862 1863 1864 1865 1866 1867 1868 1869 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
/* Control device data path
 */

static bool
nfp_ctrl_tx_one(struct nfp_net *nn, struct nfp_net_r_vector *r_vec,
		struct sk_buff *skb, bool old)
{
	unsigned int real_len = skb->len, meta_len = 0;
	struct nfp_net_tx_ring *tx_ring;
	struct nfp_net_tx_buf *txbuf;
	struct nfp_net_tx_desc *txd;
	struct nfp_net_dp *dp;
	dma_addr_t dma_addr;
	int wr_idx;

	dp = &r_vec->nfp_net->dp;
	tx_ring = r_vec->tx_ring;

	if (WARN_ON_ONCE(skb_shinfo(skb)->nr_frags)) {
		nn_dp_warn(dp, "Driver's CTRL TX does not implement gather\n");
		goto err_free;
	}

	if (unlikely(nfp_net_tx_full(tx_ring, 1))) {
		u64_stats_update_begin(&r_vec->tx_sync);
		r_vec->tx_busy++;
		u64_stats_update_end(&r_vec->tx_sync);
		if (!old)
			__skb_queue_tail(&r_vec->queue, skb);
		else
			__skb_queue_head(&r_vec->queue, skb);
		return true;
	}

	if (nfp_app_ctrl_has_meta(nn->app)) {
		if (unlikely(skb_headroom(skb) < 8)) {
			nn_dp_warn(dp, "CTRL TX on skb without headroom\n");
			goto err_free;
		}
		meta_len = 8;
		put_unaligned_be32(NFP_META_PORT_ID_CTRL, skb_push(skb, 4));
		put_unaligned_be32(NFP_NET_META_PORTID, skb_push(skb, 4));
	}

	/* Start with the head skbuf */
	dma_addr = dma_map_single(dp->dev, skb->data, skb_headlen(skb),
				  DMA_TO_DEVICE);
	if (dma_mapping_error(dp->dev, dma_addr))
		goto err_dma_warn;

	wr_idx = D_IDX(tx_ring, tx_ring->wr_p);

	/* Stash the soft descriptor of the head then initialize it */
	txbuf = &tx_ring->txbufs[wr_idx];
	txbuf->skb = skb;
	txbuf->dma_addr = dma_addr;
	txbuf->fidx = -1;
	txbuf->pkt_cnt = 1;
	txbuf->real_len = real_len;

	/* Build TX descriptor */
	txd = &tx_ring->txds[wr_idx];
	txd->offset_eop = meta_len | PCIE_DESC_TX_EOP;
	txd->dma_len = cpu_to_le16(skb_headlen(skb));
	nfp_desc_set_dma_addr(txd, dma_addr);
	txd->data_len = cpu_to_le16(skb->len);

	txd->flags = 0;
	txd->mss = 0;
	txd->lso_hdrlen = 0;

	tx_ring->wr_p++;
	tx_ring->wr_ptr_add++;
	nfp_net_tx_xmit_more_flush(tx_ring);

	return false;

err_dma_warn:
	nn_dp_warn(dp, "Failed to DMA map TX CTRL buffer\n");
err_free:
	u64_stats_update_begin(&r_vec->tx_sync);
	r_vec->tx_errors++;
	u64_stats_update_end(&r_vec->tx_sync);
	dev_kfree_skb_any(skb);
	return false;
}

1923 1924 1925 1926 1927 1928 1929
bool __nfp_ctrl_tx(struct nfp_net *nn, struct sk_buff *skb)
{
	struct nfp_net_r_vector *r_vec = &nn->r_vecs[0];

	return nfp_ctrl_tx_one(nn, r_vec, skb, false);
}

J
Jakub Kicinski 已提交
1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038
bool nfp_ctrl_tx(struct nfp_net *nn, struct sk_buff *skb)
{
	struct nfp_net_r_vector *r_vec = &nn->r_vecs[0];
	bool ret;

	spin_lock_bh(&r_vec->lock);
	ret = nfp_ctrl_tx_one(nn, r_vec, skb, false);
	spin_unlock_bh(&r_vec->lock);

	return ret;
}

static void __nfp_ctrl_tx_queued(struct nfp_net_r_vector *r_vec)
{
	struct sk_buff *skb;

	while ((skb = __skb_dequeue(&r_vec->queue)))
		if (nfp_ctrl_tx_one(r_vec->nfp_net, r_vec, skb, true))
			return;
}

static bool
nfp_ctrl_meta_ok(struct nfp_net *nn, void *data, unsigned int meta_len)
{
	u32 meta_type, meta_tag;

	if (!nfp_app_ctrl_has_meta(nn->app))
		return !meta_len;

	if (meta_len != 8)
		return false;

	meta_type = get_unaligned_be32(data);
	meta_tag = get_unaligned_be32(data + 4);

	return (meta_type == NFP_NET_META_PORTID &&
		meta_tag == NFP_META_PORT_ID_CTRL);
}

static bool
nfp_ctrl_rx_one(struct nfp_net *nn, struct nfp_net_dp *dp,
		struct nfp_net_r_vector *r_vec, struct nfp_net_rx_ring *rx_ring)
{
	unsigned int meta_len, data_len, meta_off, pkt_len, pkt_off;
	struct nfp_net_rx_buf *rxbuf;
	struct nfp_net_rx_desc *rxd;
	dma_addr_t new_dma_addr;
	struct sk_buff *skb;
	void *new_frag;
	int idx;

	idx = D_IDX(rx_ring, rx_ring->rd_p);

	rxd = &rx_ring->rxds[idx];
	if (!(rxd->rxd.meta_len_dd & PCIE_DESC_RX_DD))
		return false;

	/* Memory barrier to ensure that we won't do other reads
	 * before the DD bit.
	 */
	dma_rmb();

	rx_ring->rd_p++;

	rxbuf =	&rx_ring->rxbufs[idx];
	meta_len = rxd->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK;
	data_len = le16_to_cpu(rxd->rxd.data_len);
	pkt_len = data_len - meta_len;

	pkt_off = NFP_NET_RX_BUF_HEADROOM + dp->rx_dma_off;
	if (dp->rx_offset == NFP_NET_CFG_RX_OFFSET_DYNAMIC)
		pkt_off += meta_len;
	else
		pkt_off += dp->rx_offset;
	meta_off = pkt_off - meta_len;

	/* Stats update */
	u64_stats_update_begin(&r_vec->rx_sync);
	r_vec->rx_pkts++;
	r_vec->rx_bytes += pkt_len;
	u64_stats_update_end(&r_vec->rx_sync);

	nfp_net_dma_sync_cpu_rx(dp, rxbuf->dma_addr + meta_off,	data_len);

	if (unlikely(!nfp_ctrl_meta_ok(nn, rxbuf->frag + meta_off, meta_len))) {
		nn_dp_warn(dp, "incorrect metadata for ctrl packet (%d)\n",
			   meta_len);
		nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf, NULL);
		return true;
	}

	skb = build_skb(rxbuf->frag, dp->fl_bufsz);
	if (unlikely(!skb)) {
		nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf, NULL);
		return true;
	}
	new_frag = nfp_net_napi_alloc_one(dp, &new_dma_addr);
	if (unlikely(!new_frag)) {
		nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf, skb);
		return true;
	}

	nfp_net_dma_unmap_rx(dp, rxbuf->dma_addr);

	nfp_net_rx_give_one(dp, rx_ring, new_frag, new_dma_addr);

	skb_reserve(skb, pkt_off);
	skb_put(skb, pkt_len);

2039
	nfp_app_ctrl_rx(nn->app, skb);
J
Jakub Kicinski 已提交
2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067

	return true;
}

static void nfp_ctrl_rx(struct nfp_net_r_vector *r_vec)
{
	struct nfp_net_rx_ring *rx_ring = r_vec->rx_ring;
	struct nfp_net *nn = r_vec->nfp_net;
	struct nfp_net_dp *dp = &nn->dp;

	while (nfp_ctrl_rx_one(nn, dp, r_vec, rx_ring))
		continue;
}

static void nfp_ctrl_poll(unsigned long arg)
{
	struct nfp_net_r_vector *r_vec = (void *)arg;

	spin_lock_bh(&r_vec->lock);
	nfp_net_tx_complete(r_vec->tx_ring);
	__nfp_ctrl_tx_queued(r_vec);
	spin_unlock_bh(&r_vec->lock);

	nfp_ctrl_rx(r_vec);

	nfp_net_irq_unmask(r_vec->nfp_net, r_vec->irq_entry);
}

2068 2069 2070
/* Setup and Configuration
 */

J
Jakub Kicinski 已提交
2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092
/**
 * nfp_net_vecs_init() - Assign IRQs and setup rvecs.
 * @nn:		NFP Network structure
 */
static void nfp_net_vecs_init(struct nfp_net *nn)
{
	struct nfp_net_r_vector *r_vec;
	int r;

	nn->lsc_handler = nfp_net_irq_lsc;
	nn->exn_handler = nfp_net_irq_exn;

	for (r = 0; r < nn->max_r_vecs; r++) {
		struct msix_entry *entry;

		entry = &nn->irq_entries[NFP_NET_NON_Q_VECTORS + r];

		r_vec = &nn->r_vecs[r];
		r_vec->nfp_net = nn;
		r_vec->irq_entry = entry->entry;
		r_vec->irq_vector = entry->vector;

J
Jakub Kicinski 已提交
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104
		if (nn->dp.netdev) {
			r_vec->handler = nfp_net_irq_rxtx;
		} else {
			r_vec->handler = nfp_ctrl_irq_rxtx;

			__skb_queue_head_init(&r_vec->queue);
			spin_lock_init(&r_vec->lock);
			tasklet_init(&r_vec->tasklet, nfp_ctrl_poll,
				     (unsigned long)r_vec);
			tasklet_disable(&r_vec->tasklet);
		}

J
Jakub Kicinski 已提交
2105 2106 2107 2108
		cpumask_set_cpu(r, &r_vec->affinity_mask);
	}
}

2109 2110 2111 2112 2113 2114 2115
/**
 * nfp_net_tx_ring_free() - Free resources allocated to a TX ring
 * @tx_ring:   TX ring to free
 */
static void nfp_net_tx_ring_free(struct nfp_net_tx_ring *tx_ring)
{
	struct nfp_net_r_vector *r_vec = tx_ring->r_vec;
2116
	struct nfp_net_dp *dp = &r_vec->nfp_net->dp;
2117 2118 2119 2120

	kfree(tx_ring->txbufs);

	if (tx_ring->txds)
2121
		dma_free_coherent(dp->dev, tx_ring->size,
2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132
				  tx_ring->txds, tx_ring->dma);

	tx_ring->cnt = 0;
	tx_ring->txbufs = NULL;
	tx_ring->txds = NULL;
	tx_ring->dma = 0;
	tx_ring->size = 0;
}

/**
 * nfp_net_tx_ring_alloc() - Allocate resource for a TX ring
2133
 * @dp:        NFP Net data path struct
2134 2135 2136 2137
 * @tx_ring:   TX Ring structure to allocate
 *
 * Return: 0 on success, negative errno otherwise.
 */
2138
static int
2139
nfp_net_tx_ring_alloc(struct nfp_net_dp *dp, struct nfp_net_tx_ring *tx_ring)
2140 2141 2142 2143
{
	struct nfp_net_r_vector *r_vec = tx_ring->r_vec;
	int sz;

2144
	tx_ring->cnt = dp->txd_cnt;
2145 2146

	tx_ring->size = sizeof(*tx_ring->txds) * tx_ring->cnt;
2147
	tx_ring->txds = dma_zalloc_coherent(dp->dev, tx_ring->size,
2148 2149 2150 2151 2152 2153 2154 2155 2156
					    &tx_ring->dma, GFP_KERNEL);
	if (!tx_ring->txds)
		goto err_alloc;

	sz = sizeof(*tx_ring->txbufs) * tx_ring->cnt;
	tx_ring->txbufs = kzalloc(sz, GFP_KERNEL);
	if (!tx_ring->txbufs)
		goto err_alloc;

2157
	if (!tx_ring->is_xdp && dp->netdev)
2158
		netif_set_xps_queue(dp->netdev, &r_vec->affinity_mask,
2159
				    tx_ring->idx);
2160 2161 2162 2163 2164 2165 2166 2167

	return 0;

err_alloc:
	nfp_net_tx_ring_free(tx_ring);
	return -ENOMEM;
}

2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
static void
nfp_net_tx_ring_bufs_free(struct nfp_net_dp *dp,
			  struct nfp_net_tx_ring *tx_ring)
{
	unsigned int i;

	if (!tx_ring->is_xdp)
		return;

	for (i = 0; i < tx_ring->cnt; i++) {
		if (!tx_ring->txbufs[i].frag)
			return;

		nfp_net_dma_unmap_rx(dp, tx_ring->txbufs[i].dma_addr);
		__free_page(virt_to_page(tx_ring->txbufs[i].frag));
	}
}

static int
nfp_net_tx_ring_bufs_alloc(struct nfp_net_dp *dp,
			   struct nfp_net_tx_ring *tx_ring)
{
	struct nfp_net_tx_buf *txbufs = tx_ring->txbufs;
	unsigned int i;

	if (!tx_ring->is_xdp)
		return 0;

	for (i = 0; i < tx_ring->cnt; i++) {
		txbufs[i].frag = nfp_net_rx_alloc_one(dp, &txbufs[i].dma_addr);
		if (!txbufs[i].frag) {
			nfp_net_tx_ring_bufs_free(dp, tx_ring);
			return -ENOMEM;
		}
	}

	return 0;
}

2207
static int nfp_net_tx_rings_prepare(struct nfp_net *nn, struct nfp_net_dp *dp)
2208 2209 2210
{
	unsigned int r;

2211 2212 2213 2214
	dp->tx_rings = kcalloc(dp->num_tx_rings, sizeof(*dp->tx_rings),
			       GFP_KERNEL);
	if (!dp->tx_rings)
		return -ENOMEM;
2215

2216
	for (r = 0; r < dp->num_tx_rings; r++) {
2217 2218
		int bias = 0;

2219 2220
		if (r >= dp->num_stack_tx_rings)
			bias = dp->num_stack_tx_rings;
2221

2222
		nfp_net_tx_ring_init(&dp->tx_rings[r], &nn->r_vecs[r - bias],
2223
				     r, bias);
2224

2225
		if (nfp_net_tx_ring_alloc(dp, &dp->tx_rings[r]))
2226
			goto err_free_prev;
2227 2228 2229

		if (nfp_net_tx_ring_bufs_alloc(dp, &dp->tx_rings[r]))
			goto err_free_ring;
2230 2231
	}

2232
	return 0;
2233 2234

err_free_prev:
2235 2236 2237
	while (r--) {
		nfp_net_tx_ring_bufs_free(dp, &dp->tx_rings[r]);
err_free_ring:
2238
		nfp_net_tx_ring_free(&dp->tx_rings[r]);
2239
	}
2240 2241
	kfree(dp->tx_rings);
	return -ENOMEM;
2242 2243
}

2244
static void nfp_net_tx_rings_free(struct nfp_net_dp *dp)
2245 2246 2247
{
	unsigned int r;

2248 2249
	for (r = 0; r < dp->num_tx_rings; r++) {
		nfp_net_tx_ring_bufs_free(dp, &dp->tx_rings[r]);
2250
		nfp_net_tx_ring_free(&dp->tx_rings[r]);
2251
	}
2252

2253
	kfree(dp->tx_rings);
2254 2255
}

2256 2257 2258 2259 2260 2261 2262
/**
 * nfp_net_rx_ring_free() - Free resources allocated to a RX ring
 * @rx_ring:  RX ring to free
 */
static void nfp_net_rx_ring_free(struct nfp_net_rx_ring *rx_ring)
{
	struct nfp_net_r_vector *r_vec = rx_ring->r_vec;
2263
	struct nfp_net_dp *dp = &r_vec->nfp_net->dp;
2264

2265 2266
	if (dp->netdev)
		xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
2267 2268 2269
	kfree(rx_ring->rxbufs);

	if (rx_ring->rxds)
2270
		dma_free_coherent(dp->dev, rx_ring->size,
2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281
				  rx_ring->rxds, rx_ring->dma);

	rx_ring->cnt = 0;
	rx_ring->rxbufs = NULL;
	rx_ring->rxds = NULL;
	rx_ring->dma = 0;
	rx_ring->size = 0;
}

/**
 * nfp_net_rx_ring_alloc() - Allocate resource for a RX ring
2282
 * @dp:	      NFP Net data path struct
2283 2284 2285 2286
 * @rx_ring:  RX ring to allocate
 *
 * Return: 0 on success, negative errno otherwise.
 */
2287
static int
2288
nfp_net_rx_ring_alloc(struct nfp_net_dp *dp, struct nfp_net_rx_ring *rx_ring)
2289
{
J
Jesper Dangaard Brouer 已提交
2290 2291
	int sz, err;

2292 2293 2294 2295 2296 2297
	if (dp->netdev) {
		err = xdp_rxq_info_reg(&rx_ring->xdp_rxq, dp->netdev,
				       rx_ring->idx);
		if (err < 0)
			return err;
	}
2298

2299
	rx_ring->cnt = dp->rxd_cnt;
2300
	rx_ring->size = sizeof(*rx_ring->rxds) * rx_ring->cnt;
2301
	rx_ring->rxds = dma_zalloc_coherent(dp->dev, rx_ring->size,
2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317
					    &rx_ring->dma, GFP_KERNEL);
	if (!rx_ring->rxds)
		goto err_alloc;

	sz = sizeof(*rx_ring->rxbufs) * rx_ring->cnt;
	rx_ring->rxbufs = kzalloc(sz, GFP_KERNEL);
	if (!rx_ring->rxbufs)
		goto err_alloc;

	return 0;

err_alloc:
	nfp_net_rx_ring_free(rx_ring);
	return -ENOMEM;
}

2318
static int nfp_net_rx_rings_prepare(struct nfp_net *nn, struct nfp_net_dp *dp)
2319 2320 2321
{
	unsigned int r;

2322 2323 2324 2325
	dp->rx_rings = kcalloc(dp->num_rx_rings, sizeof(*dp->rx_rings),
			       GFP_KERNEL);
	if (!dp->rx_rings)
		return -ENOMEM;
2326

2327 2328
	for (r = 0; r < dp->num_rx_rings; r++) {
		nfp_net_rx_ring_init(&dp->rx_rings[r], &nn->r_vecs[r], r);
2329

2330
		if (nfp_net_rx_ring_alloc(dp, &dp->rx_rings[r]))
2331 2332
			goto err_free_prev;

2333
		if (nfp_net_rx_ring_bufs_alloc(dp, &dp->rx_rings[r]))
2334 2335 2336
			goto err_free_ring;
	}

2337
	return 0;
2338 2339 2340

err_free_prev:
	while (r--) {
2341
		nfp_net_rx_ring_bufs_free(dp, &dp->rx_rings[r]);
2342
err_free_ring:
2343
		nfp_net_rx_ring_free(&dp->rx_rings[r]);
2344
	}
2345 2346
	kfree(dp->rx_rings);
	return -ENOMEM;
2347 2348
}

2349
static void nfp_net_rx_rings_free(struct nfp_net_dp *dp)
2350 2351 2352
{
	unsigned int r;

2353 2354 2355
	for (r = 0; r < dp->num_rx_rings; r++) {
		nfp_net_rx_ring_bufs_free(dp, &dp->rx_rings[r]);
		nfp_net_rx_ring_free(&dp->rx_rings[r]);
2356 2357
	}

2358
	kfree(dp->rx_rings);
2359 2360
}

2361
static void
2362 2363
nfp_net_vector_assign_rings(struct nfp_net_dp *dp,
			    struct nfp_net_r_vector *r_vec, int idx)
2364
{
2365
	r_vec->rx_ring = idx < dp->num_rx_rings ? &dp->rx_rings[idx] : NULL;
2366
	r_vec->tx_ring =
2367
		idx < dp->num_stack_tx_rings ? &dp->tx_rings[idx] : NULL;
2368

2369 2370
	r_vec->xdp_ring = idx < dp->num_tx_rings - dp->num_stack_tx_rings ?
		&dp->tx_rings[dp->num_stack_tx_rings + idx] : NULL;
2371 2372
}

2373 2374 2375
static int
nfp_net_prepare_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec,
		       int idx)
2376
{
2377
	int err;
2378

2379
	/* Setup NAPI */
J
Jakub Kicinski 已提交
2380 2381 2382 2383 2384
	if (nn->dp.netdev)
		netif_napi_add(nn->dp.netdev, &r_vec->napi,
			       nfp_net_poll, NAPI_POLL_WEIGHT);
	else
		tasklet_enable(&r_vec->tasklet);
2385

2386
	snprintf(r_vec->name, sizeof(r_vec->name),
J
Jakub Kicinski 已提交
2387
		 "%s-rxtx-%d", nfp_net_name(nn), idx);
2388 2389
	err = request_irq(r_vec->irq_vector, r_vec->handler, 0, r_vec->name,
			  r_vec);
2390
	if (err) {
J
Jakub Kicinski 已提交
2391 2392 2393 2394 2395
		if (nn->dp.netdev)
			netif_napi_del(&r_vec->napi);
		else
			tasklet_disable(&r_vec->tasklet);

2396
		nn_err(nn, "Error requesting IRQ %d\n", r_vec->irq_vector);
2397 2398
		return err;
	}
2399
	disable_irq(r_vec->irq_vector);
2400

2401
	irq_set_affinity_hint(r_vec->irq_vector, &r_vec->affinity_mask);
2402

2403 2404
	nn_dbg(nn, "RV%02d: irq=%03d/%03d\n", idx, r_vec->irq_vector,
	       r_vec->irq_entry);
2405

2406
	return 0;
2407 2408
}

2409 2410
static void
nfp_net_cleanup_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec)
2411
{
2412
	irq_set_affinity_hint(r_vec->irq_vector, NULL);
J
Jakub Kicinski 已提交
2413 2414 2415 2416 2417
	if (nn->dp.netdev)
		netif_napi_del(&r_vec->napi);
	else
		tasklet_disable(&r_vec->tasklet);

2418
	free_irq(r_vec->irq_vector, r_vec);
2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441
}

/**
 * nfp_net_rss_write_itbl() - Write RSS indirection table to device
 * @nn:      NFP Net device to reconfigure
 */
void nfp_net_rss_write_itbl(struct nfp_net *nn)
{
	int i;

	for (i = 0; i < NFP_NET_CFG_RSS_ITBL_SZ; i += 4)
		nn_writel(nn, NFP_NET_CFG_RSS_ITBL + i,
			  get_unaligned_le32(nn->rss_itbl + i));
}

/**
 * nfp_net_rss_write_key() - Write RSS hash key to device
 * @nn:      NFP Net device to reconfigure
 */
void nfp_net_rss_write_key(struct nfp_net *nn)
{
	int i;

2442
	for (i = 0; i < nfp_net_rss_key_sz(nn); i += 4)
2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465
		nn_writel(nn, NFP_NET_CFG_RSS_KEY + i,
			  get_unaligned_le32(nn->rss_key + i));
}

/**
 * nfp_net_coalesce_write_cfg() - Write irq coalescence configuration to HW
 * @nn:      NFP Net device to reconfigure
 */
void nfp_net_coalesce_write_cfg(struct nfp_net *nn)
{
	u8 i;
	u32 factor;
	u32 value;

	/* Compute factor used to convert coalesce '_usecs' parameters to
	 * ME timestamp ticks.  There are 16 ME clock cycles for each timestamp
	 * count.
	 */
	factor = nn->me_freq_mhz / 16;

	/* copy RX interrupt coalesce parameters */
	value = (nn->rx_coalesce_max_frames << 16) |
		(factor * nn->rx_coalesce_usecs);
2466
	for (i = 0; i < nn->dp.num_rx_rings; i++)
2467 2468 2469 2470 2471
		nn_writel(nn, NFP_NET_CFG_RXR_IRQ_MOD(i), value);

	/* copy TX interrupt coalesce parameters */
	value = (nn->tx_coalesce_max_frames << 16) |
		(factor * nn->tx_coalesce_usecs);
2472
	for (i = 0; i < nn->dp.num_tx_rings; i++)
2473 2474 2475 2476
		nn_writel(nn, NFP_NET_CFG_TXR_IRQ_MOD(i), value);
}

/**
2477
 * nfp_net_write_mac_addr() - Write mac address to the device control BAR
2478
 * @nn:      NFP Net device to reconfigure
2479
 * @addr:    MAC address to write
2480
 *
2481 2482 2483
 * Writes the MAC address from the netdev to the device control BAR.  Does not
 * perform the required reconfig.  We do a bit of byte swapping dance because
 * firmware is LE.
2484
 */
2485
static void nfp_net_write_mac_addr(struct nfp_net *nn, const u8 *addr)
2486
{
2487 2488
	nn_writel(nn, NFP_NET_CFG_MACADDR + 0, get_unaligned_be32(addr));
	nn_writew(nn, NFP_NET_CFG_MACADDR + 6, get_unaligned_be16(addr + 4));
2489 2490
}

2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501
static void nfp_net_vec_clear_ring_data(struct nfp_net *nn, unsigned int idx)
{
	nn_writeq(nn, NFP_NET_CFG_RXR_ADDR(idx), 0);
	nn_writeb(nn, NFP_NET_CFG_RXR_SZ(idx), 0);
	nn_writeb(nn, NFP_NET_CFG_RXR_VEC(idx), 0);

	nn_writeq(nn, NFP_NET_CFG_TXR_ADDR(idx), 0);
	nn_writeb(nn, NFP_NET_CFG_TXR_SZ(idx), 0);
	nn_writeb(nn, NFP_NET_CFG_TXR_VEC(idx), 0);
}

2502 2503 2504 2505 2506 2507 2508
/**
 * nfp_net_clear_config_and_disable() - Clear control BAR and disable NFP
 * @nn:      NFP Net device to reconfigure
 */
static void nfp_net_clear_config_and_disable(struct nfp_net *nn)
{
	u32 new_ctrl, update;
2509
	unsigned int r;
2510 2511
	int err;

2512
	new_ctrl = nn->dp.ctrl;
2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525
	new_ctrl &= ~NFP_NET_CFG_CTRL_ENABLE;
	update = NFP_NET_CFG_UPDATE_GEN;
	update |= NFP_NET_CFG_UPDATE_MSIX;
	update |= NFP_NET_CFG_UPDATE_RING;

	if (nn->cap & NFP_NET_CFG_CTRL_RINGCFG)
		new_ctrl &= ~NFP_NET_CFG_CTRL_RINGCFG;

	nn_writeq(nn, NFP_NET_CFG_TXRS_ENABLE, 0);
	nn_writeq(nn, NFP_NET_CFG_RXRS_ENABLE, 0);

	nn_writel(nn, NFP_NET_CFG_CTRL, new_ctrl);
	err = nfp_net_reconfig(nn, update);
2526
	if (err)
2527 2528
		nn_err(nn, "Could not disable device: %d\n", err);

2529 2530 2531 2532 2533
	for (r = 0; r < nn->dp.num_rx_rings; r++)
		nfp_net_rx_ring_reset(&nn->dp.rx_rings[r]);
	for (r = 0; r < nn->dp.num_tx_rings; r++)
		nfp_net_tx_ring_reset(&nn->dp, &nn->dp.tx_rings[r]);
	for (r = 0; r < nn->dp.num_r_vecs; r++)
2534 2535
		nfp_net_vec_clear_ring_data(nn, r);

2536
	nn->dp.ctrl = new_ctrl;
2537 2538
}

2539
static void
2540 2541
nfp_net_rx_ring_hw_cfg_write(struct nfp_net *nn,
			     struct nfp_net_rx_ring *rx_ring, unsigned int idx)
2542 2543
{
	/* Write the DMA address, size and MSI-X info to the device */
2544 2545
	nn_writeq(nn, NFP_NET_CFG_RXR_ADDR(idx), rx_ring->dma);
	nn_writeb(nn, NFP_NET_CFG_RXR_SZ(idx), ilog2(rx_ring->cnt));
2546
	nn_writeb(nn, NFP_NET_CFG_RXR_VEC(idx), rx_ring->r_vec->irq_entry);
2547
}
2548

2549 2550 2551 2552 2553 2554
static void
nfp_net_tx_ring_hw_cfg_write(struct nfp_net *nn,
			     struct nfp_net_tx_ring *tx_ring, unsigned int idx)
{
	nn_writeq(nn, NFP_NET_CFG_TXR_ADDR(idx), tx_ring->dma);
	nn_writeb(nn, NFP_NET_CFG_TXR_SZ(idx), ilog2(tx_ring->cnt));
2555
	nn_writeb(nn, NFP_NET_CFG_TXR_VEC(idx), tx_ring->r_vec->irq_entry);
2556 2557
}

2558 2559 2560 2561 2562
/**
 * nfp_net_set_config_and_enable() - Write control BAR and enable NFP
 * @nn:      NFP Net device to reconfigure
 */
static int nfp_net_set_config_and_enable(struct nfp_net *nn)
2563
{
2564
	u32 bufsz, new_ctrl, update = 0;
2565 2566 2567
	unsigned int r;
	int err;

2568
	new_ctrl = nn->dp.ctrl;
2569

2570
	if (nn->dp.ctrl & NFP_NET_CFG_CTRL_RSS_ANY) {
2571 2572 2573 2574 2575 2576
		nfp_net_rss_write_key(nn);
		nfp_net_rss_write_itbl(nn);
		nn_writel(nn, NFP_NET_CFG_RSS_CTRL, nn->rss_cfg);
		update |= NFP_NET_CFG_UPDATE_RSS;
	}

2577
	if (nn->dp.ctrl & NFP_NET_CFG_CTRL_IRQMOD) {
2578 2579 2580 2581
		nfp_net_coalesce_write_cfg(nn);
		update |= NFP_NET_CFG_UPDATE_IRQMOD;
	}

2582 2583 2584 2585
	for (r = 0; r < nn->dp.num_tx_rings; r++)
		nfp_net_tx_ring_hw_cfg_write(nn, &nn->dp.tx_rings[r], r);
	for (r = 0; r < nn->dp.num_rx_rings; r++)
		nfp_net_rx_ring_hw_cfg_write(nn, &nn->dp.rx_rings[r], r);
2586

2587 2588
	nn_writeq(nn, NFP_NET_CFG_TXRS_ENABLE, nn->dp.num_tx_rings == 64 ?
		  0xffffffffffffffffULL : ((u64)1 << nn->dp.num_tx_rings) - 1);
2589

2590 2591
	nn_writeq(nn, NFP_NET_CFG_RXRS_ENABLE, nn->dp.num_rx_rings == 64 ?
		  0xffffffffffffffffULL : ((u64)1 << nn->dp.num_rx_rings) - 1);
2592

2593 2594
	if (nn->dp.netdev)
		nfp_net_write_mac_addr(nn, nn->dp.netdev->dev_addr);
2595

2596
	nn_writel(nn, NFP_NET_CFG_MTU, nn->dp.mtu);
2597 2598 2599

	bufsz = nn->dp.fl_bufsz - nn->dp.rx_dma_off - NFP_NET_RX_BUF_NON_DATA;
	nn_writel(nn, NFP_NET_CFG_FLBUFSZ, bufsz);
2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610

	/* Enable device */
	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
	update |= NFP_NET_CFG_UPDATE_GEN;
	update |= NFP_NET_CFG_UPDATE_MSIX;
	update |= NFP_NET_CFG_UPDATE_RING;
	if (nn->cap & NFP_NET_CFG_CTRL_RINGCFG)
		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;

	nn_writel(nn, NFP_NET_CFG_CTRL, new_ctrl);
	err = nfp_net_reconfig(nn, update);
2611 2612 2613 2614
	if (err) {
		nfp_net_clear_config_and_disable(nn);
		return err;
	}
2615

2616
	nn->dp.ctrl = new_ctrl;
2617

2618
	for (r = 0; r < nn->dp.num_rx_rings; r++)
2619
		nfp_net_rx_ring_fill_freelist(&nn->dp, &nn->dp.rx_rings[r]);
2620

2621 2622 2623
	/* Since reconfiguration requests while NFP is down are ignored we
	 * have to wipe the entire VXLAN configuration and reinitialize it.
	 */
2624
	if (nn->dp.ctrl & NFP_NET_CFG_CTRL_VXLAN) {
2625 2626
		memset(&nn->vxlan_ports, 0, sizeof(nn->vxlan_ports));
		memset(&nn->vxlan_usecnt, 0, sizeof(nn->vxlan_usecnt));
2627
		udp_tunnel_get_rx_info(nn->dp.netdev);
2628 2629
	}

2630
	return 0;
2631 2632
}

2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660
/**
 * nfp_net_close_stack() - Quiesce the stack (part of close)
 * @nn:	     NFP Net device to reconfigure
 */
static void nfp_net_close_stack(struct nfp_net *nn)
{
	unsigned int r;

	disable_irq(nn->irq_entries[NFP_NET_IRQ_LSC_IDX].vector);
	netif_carrier_off(nn->dp.netdev);
	nn->link_up = false;

	for (r = 0; r < nn->dp.num_r_vecs; r++) {
		disable_irq(nn->r_vecs[r].irq_vector);
		napi_disable(&nn->r_vecs[r].napi);
	}

	netif_tx_disable(nn->dp.netdev);
}

/**
 * nfp_net_close_free_all() - Free all runtime resources
 * @nn:      NFP Net device to reconfigure
 */
static void nfp_net_close_free_all(struct nfp_net *nn)
{
	unsigned int r;

2661 2662 2663
	nfp_net_tx_rings_free(&nn->dp);
	nfp_net_rx_rings_free(&nn->dp);

2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685
	for (r = 0; r < nn->dp.num_r_vecs; r++)
		nfp_net_cleanup_vector(nn, &nn->r_vecs[r]);

	nfp_net_aux_irq_free(nn, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
	nfp_net_aux_irq_free(nn, NFP_NET_CFG_EXN, NFP_NET_IRQ_EXN_IDX);
}

/**
 * nfp_net_netdev_close() - Called when the device is downed
 * @netdev:      netdev structure
 */
static int nfp_net_netdev_close(struct net_device *netdev)
{
	struct nfp_net *nn = netdev_priv(netdev);

	/* Step 1: Disable RX and TX rings from the Linux kernel perspective
	 */
	nfp_net_close_stack(nn);

	/* Step 2: Tell NFP
	 */
	nfp_net_clear_config_and_disable(nn);
2686
	nfp_port_configure(netdev, false);
2687 2688 2689 2690 2691 2692 2693 2694 2695

	/* Step 3: Free resources
	 */
	nfp_net_close_free_all(nn);

	nn_dbg(nn, "%s down", netdev->name);
	return 0;
}

J
Jakub Kicinski 已提交
2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713
void nfp_ctrl_close(struct nfp_net *nn)
{
	int r;

	rtnl_lock();

	for (r = 0; r < nn->dp.num_r_vecs; r++) {
		disable_irq(nn->r_vecs[r].irq_vector);
		tasklet_disable(&nn->r_vecs[r].tasklet);
	}

	nfp_net_clear_config_and_disable(nn);

	nfp_net_close_free_all(nn);

	rtnl_unlock();
}

2714 2715 2716 2717 2718 2719 2720 2721
/**
 * nfp_net_open_stack() - Start the device from stack's perspective
 * @nn:      NFP Net device to reconfigure
 */
static void nfp_net_open_stack(struct nfp_net *nn)
{
	unsigned int r;

2722
	for (r = 0; r < nn->dp.num_r_vecs; r++) {
2723
		napi_enable(&nn->r_vecs[r].napi);
2724
		enable_irq(nn->r_vecs[r].irq_vector);
2725
	}
2726

2727
	netif_tx_wake_all_queues(nn->dp.netdev);
2728

2729
	enable_irq(nn->irq_entries[NFP_NET_IRQ_LSC_IDX].vector);
2730 2731 2732
	nfp_net_read_link_status(nn);
}

2733
static int nfp_net_open_alloc_all(struct nfp_net *nn)
2734 2735 2736 2737 2738 2739 2740 2741
{
	int err, r;

	err = nfp_net_aux_irq_request(nn, NFP_NET_CFG_EXN, "%s-exn",
				      nn->exn_name, sizeof(nn->exn_name),
				      NFP_NET_IRQ_EXN_IDX, nn->exn_handler);
	if (err)
		return err;
2742 2743 2744 2745 2746
	err = nfp_net_aux_irq_request(nn, NFP_NET_CFG_LSC, "%s-lsc",
				      nn->lsc_name, sizeof(nn->lsc_name),
				      NFP_NET_IRQ_LSC_IDX, nn->lsc_handler);
	if (err)
		goto err_free_exn;
2747
	disable_irq(nn->irq_entries[NFP_NET_IRQ_LSC_IDX].vector);
2748

2749
	for (r = 0; r < nn->dp.num_r_vecs; r++) {
2750 2751
		err = nfp_net_prepare_vector(nn, &nn->r_vecs[r], r);
		if (err)
2752 2753
			goto err_cleanup_vec_p;
	}
2754

2755 2756
	err = nfp_net_rx_rings_prepare(nn, &nn->dp);
	if (err)
2757
		goto err_cleanup_vec;
2758

2759 2760
	err = nfp_net_tx_rings_prepare(nn, &nn->dp);
	if (err)
2761
		goto err_free_rx_rings;
2762

2763
	for (r = 0; r < nn->max_r_vecs; r++)
2764
		nfp_net_vector_assign_rings(&nn->dp, &nn->r_vecs[r], r);
2765

2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794
	return 0;

err_free_rx_rings:
	nfp_net_rx_rings_free(&nn->dp);
err_cleanup_vec:
	r = nn->dp.num_r_vecs;
err_cleanup_vec_p:
	while (r--)
		nfp_net_cleanup_vector(nn, &nn->r_vecs[r]);
	nfp_net_aux_irq_free(nn, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
err_free_exn:
	nfp_net_aux_irq_free(nn, NFP_NET_CFG_EXN, NFP_NET_IRQ_EXN_IDX);
	return err;
}

static int nfp_net_netdev_open(struct net_device *netdev)
{
	struct nfp_net *nn = netdev_priv(netdev);
	int err;

	/* Step 1: Allocate resources for rings and the like
	 * - Request interrupts
	 * - Allocate RX and TX ring resources
	 * - Setup initial RSS table
	 */
	err = nfp_net_open_alloc_all(nn);
	if (err)
		return err;

2795
	err = netif_set_real_num_tx_queues(netdev, nn->dp.num_stack_tx_rings);
2796
	if (err)
2797
		goto err_free_all;
2798

2799
	err = netif_set_real_num_rx_queues(netdev, nn->dp.num_rx_rings);
2800
	if (err)
2801
		goto err_free_all;
2802 2803

	/* Step 2: Configure the NFP
2804
	 * - Ifup the physical interface if it exists
2805 2806 2807 2808 2809 2810
	 * - Enable rings from 0 to tx_rings/rx_rings - 1.
	 * - Write MAC address (in case it changed)
	 * - Set the MTU
	 * - Set the Freelist buffer size
	 * - Enable the FW
	 */
2811
	err = nfp_port_configure(netdev, true);
2812
	if (err)
2813
		goto err_free_all;
2814

2815 2816 2817 2818
	err = nfp_net_set_config_and_enable(nn);
	if (err)
		goto err_port_disable;

2819 2820 2821 2822 2823 2824
	/* Step 3: Enable for kernel
	 * - put some freelist descriptors on each RX ring
	 * - enable NAPI on each ring
	 * - enable all TX queues
	 * - set link state
	 */
2825
	nfp_net_open_stack(nn);
2826 2827 2828

	return 0;

2829 2830
err_port_disable:
	nfp_port_configure(netdev, false);
2831 2832
err_free_all:
	nfp_net_close_free_all(nn);
2833 2834 2835
	return err;
}

J
Jakub Kicinski 已提交
2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
int nfp_ctrl_open(struct nfp_net *nn)
{
	int err, r;

	/* ring dumping depends on vNICs being opened/closed under rtnl */
	rtnl_lock();

	err = nfp_net_open_alloc_all(nn);
	if (err)
		goto err_unlock;

	err = nfp_net_set_config_and_enable(nn);
	if (err)
		goto err_free_all;

	for (r = 0; r < nn->dp.num_r_vecs; r++)
		enable_irq(nn->r_vecs[r].irq_vector);

	rtnl_unlock();

	return 0;

err_free_all:
	nfp_net_close_free_all(nn);
err_unlock:
	rtnl_unlock();
	return err;
}

2865 2866 2867 2868 2869
static void nfp_net_set_rx_mode(struct net_device *netdev)
{
	struct nfp_net *nn = netdev_priv(netdev);
	u32 new_ctrl;

2870
	new_ctrl = nn->dp.ctrl;
2871

2872 2873 2874 2875 2876
	if (!netdev_mc_empty(netdev) || netdev->flags & IFF_ALLMULTI)
		new_ctrl |= nn->cap & NFP_NET_CFG_CTRL_L2MC;
	else
		new_ctrl &= ~NFP_NET_CFG_CTRL_L2MC;

2877 2878 2879 2880 2881 2882 2883 2884 2885
	if (netdev->flags & IFF_PROMISC) {
		if (nn->cap & NFP_NET_CFG_CTRL_PROMISC)
			new_ctrl |= NFP_NET_CFG_CTRL_PROMISC;
		else
			nn_warn(nn, "FW does not support promiscuous mode\n");
	} else {
		new_ctrl &= ~NFP_NET_CFG_CTRL_PROMISC;
	}

2886
	if (new_ctrl == nn->dp.ctrl)
2887 2888 2889
		return;

	nn_writel(nn, NFP_NET_CFG_CTRL, new_ctrl);
2890
	nfp_net_reconfig_post(nn, NFP_NET_CFG_UPDATE_GEN);
2891

2892
	nn->dp.ctrl = new_ctrl;
2893 2894
}

2895 2896 2897 2898 2899 2900
static void nfp_net_rss_init_itbl(struct nfp_net *nn)
{
	int i;

	for (i = 0; i < sizeof(nn->rss_itbl); i++)
		nn->rss_itbl[i] =
2901
			ethtool_rxfh_indir_default(i, nn->dp.num_rx_rings);
2902 2903
}

2904 2905 2906 2907 2908 2909
static void nfp_net_dp_swap(struct nfp_net *nn, struct nfp_net_dp *dp)
{
	struct nfp_net_dp new_dp = *dp;

	*dp = nn->dp;
	nn->dp = new_dp;
2910 2911

	nn->dp.netdev->mtu = new_dp.mtu;
2912 2913 2914

	if (!netif_is_rxfh_configured(nn->dp.netdev))
		nfp_net_rss_init_itbl(nn);
2915 2916
}

2917
static int nfp_net_dp_swap_enable(struct nfp_net *nn, struct nfp_net_dp *dp)
2918
{
2919
	unsigned int r;
2920
	int err;
2921

2922
	nfp_net_dp_swap(nn, dp);
2923

2924
	for (r = 0; r <	nn->max_r_vecs; r++)
2925
		nfp_net_vector_assign_rings(&nn->dp, &nn->r_vecs[r], r);
2926

2927
	err = netif_set_real_num_rx_queues(nn->dp.netdev, nn->dp.num_rx_rings);
2928 2929
	if (err)
		return err;
2930

2931 2932 2933
	if (nn->dp.netdev->real_num_tx_queues != nn->dp.num_stack_tx_rings) {
		err = netif_set_real_num_tx_queues(nn->dp.netdev,
						   nn->dp.num_stack_tx_rings);
2934 2935 2936 2937
		if (err)
			return err;
	}

2938
	return nfp_net_set_config_and_enable(nn);
2939
}
2940

2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960
struct nfp_net_dp *nfp_net_clone_dp(struct nfp_net *nn)
{
	struct nfp_net_dp *new;

	new = kmalloc(sizeof(*new), GFP_KERNEL);
	if (!new)
		return NULL;

	*new = nn->dp;

	/* Clear things which need to be recomputed */
	new->fl_bufsz = 0;
	new->tx_rings = NULL;
	new->rx_rings = NULL;
	new->num_r_vecs = 0;
	new->num_stack_tx_rings = 0;

	return new;
}

2961 2962 2963
static int
nfp_net_check_config(struct nfp_net *nn, struct nfp_net_dp *dp,
		     struct netlink_ext_ack *extack)
2964 2965
{
	/* XDP-enabled tests */
2966
	if (!dp->xdp_prog)
2967
		return 0;
2968
	if (dp->fl_bufsz > PAGE_SIZE) {
2969
		NL_SET_ERR_MSG_MOD(extack, "MTU too large w/ XDP enabled");
2970 2971
		return -EINVAL;
	}
2972
	if (dp->num_tx_rings > nn->max_tx_rings) {
2973
		NL_SET_ERR_MSG_MOD(extack, "Insufficient number of TX rings w/ XDP enabled");
2974 2975 2976 2977 2978 2979
		return -EINVAL;
	}

	return 0;
}

2980 2981
int nfp_net_ring_reconfig(struct nfp_net *nn, struct nfp_net_dp *dp,
			  struct netlink_ext_ack *extack)
2982
{
2983
	int r, err;
2984

2985
	dp->fl_bufsz = nfp_net_calc_fl_bufsz(dp);
2986

2987
	dp->num_stack_tx_rings = dp->num_tx_rings;
2988
	if (dp->xdp_prog)
2989
		dp->num_stack_tx_rings -= dp->num_rx_rings;
2990

2991
	dp->num_r_vecs = max(dp->num_rx_rings, dp->num_stack_tx_rings);
2992

2993
	err = nfp_net_check_config(nn, dp, extack);
2994
	if (err)
2995
		goto exit_free_dp;
2996

2997
	if (!netif_running(dp->netdev)) {
2998
		nfp_net_dp_swap(nn, dp);
2999 3000
		err = 0;
		goto exit_free_dp;
3001 3002 3003
	}

	/* Prepare new rings */
3004
	for (r = nn->dp.num_r_vecs; r < dp->num_r_vecs; r++) {
3005 3006
		err = nfp_net_prepare_vector(nn, &nn->r_vecs[r], r);
		if (err) {
3007
			dp->num_r_vecs = r;
3008 3009 3010
			goto err_cleanup_vecs;
		}
	}
3011 3012 3013 3014 3015 3016 3017 3018

	err = nfp_net_rx_rings_prepare(nn, dp);
	if (err)
		goto err_cleanup_vecs;

	err = nfp_net_tx_rings_prepare(nn, dp);
	if (err)
		goto err_free_rx;
3019 3020 3021 3022 3023

	/* Stop device, swap in new rings, try to start the firmware */
	nfp_net_close_stack(nn);
	nfp_net_clear_config_and_disable(nn);

3024
	err = nfp_net_dp_swap_enable(nn, dp);
3025
	if (err) {
3026
		int err2;
3027

3028
		nfp_net_clear_config_and_disable(nn);
3029

3030
		/* Try with old configuration and old rings */
3031
		err2 = nfp_net_dp_swap_enable(nn, dp);
3032
		if (err2)
3033
			nn_err(nn, "Can't restore ring config - FW communication failed (%d,%d)\n",
3034
			       err, err2);
3035
	}
3036
	for (r = dp->num_r_vecs - 1; r >= nn->dp.num_r_vecs; r--)
3037
		nfp_net_cleanup_vector(nn, &nn->r_vecs[r]);
3038

3039 3040
	nfp_net_rx_rings_free(dp);
	nfp_net_tx_rings_free(dp);
3041 3042

	nfp_net_open_stack(nn);
3043 3044
exit_free_dp:
	kfree(dp);
3045 3046

	return err;
3047 3048

err_free_rx:
3049
	nfp_net_rx_rings_free(dp);
3050
err_cleanup_vecs:
3051
	for (r = dp->num_r_vecs - 1; r >= nn->dp.num_r_vecs; r--)
3052
		nfp_net_cleanup_vector(nn, &nn->r_vecs[r]);
3053
	kfree(dp);
3054 3055 3056 3057 3058 3059
	return err;
}

static int nfp_net_change_mtu(struct net_device *netdev, int new_mtu)
{
	struct nfp_net *nn = netdev_priv(netdev);
3060
	struct nfp_net_dp *dp;
3061 3062 3063 3064 3065
	int err;

	err = nfp_app_change_mtu(nn->app, netdev, new_mtu);
	if (err)
		return err;
3066 3067 3068 3069

	dp = nfp_net_clone_dp(nn);
	if (!dp)
		return -ENOMEM;
3070

3071 3072
	dp->mtu = new_mtu;

3073
	return nfp_net_ring_reconfig(nn, dp, NULL);
3074 3075
}

P
Pablo Cascón 已提交
3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
static int
nfp_net_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
{
	struct nfp_net *nn = netdev_priv(netdev);

	/* Priority tagged packets with vlan id 0 are processed by the
	 * NFP as untagged packets
	 */
	if (!vid)
		return 0;

	nn_writew(nn, NFP_NET_CFG_VLAN_FILTER_VID, vid);
	nn_writew(nn, NFP_NET_CFG_VLAN_FILTER_PROTO, ETH_P_8021Q);

	return nfp_net_reconfig_mbox(nn, NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_ADD);
}

static int
nfp_net_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
{
	struct nfp_net *nn = netdev_priv(netdev);

	/* Priority tagged packets with vlan id 0 are processed by the
	 * NFP as untagged packets
	 */
	if (!vid)
		return 0;

	nn_writew(nn, NFP_NET_CFG_VLAN_FILTER_VID, vid);
	nn_writew(nn, NFP_NET_CFG_VLAN_FILTER_PROTO, ETH_P_8021Q);

	return nfp_net_reconfig_mbox(nn, NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_KILL);
}

3110 3111
static void nfp_net_stat64(struct net_device *netdev,
			   struct rtnl_link_stats64 *stats)
3112 3113 3114 3115
{
	struct nfp_net *nn = netdev_priv(netdev);
	int r;

3116
	for (r = 0; r < nn->dp.num_r_vecs; r++) {
3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152
		struct nfp_net_r_vector *r_vec = &nn->r_vecs[r];
		u64 data[3];
		unsigned int start;

		do {
			start = u64_stats_fetch_begin(&r_vec->rx_sync);
			data[0] = r_vec->rx_pkts;
			data[1] = r_vec->rx_bytes;
			data[2] = r_vec->rx_drops;
		} while (u64_stats_fetch_retry(&r_vec->rx_sync, start));
		stats->rx_packets += data[0];
		stats->rx_bytes += data[1];
		stats->rx_dropped += data[2];

		do {
			start = u64_stats_fetch_begin(&r_vec->tx_sync);
			data[0] = r_vec->tx_pkts;
			data[1] = r_vec->tx_bytes;
			data[2] = r_vec->tx_errors;
		} while (u64_stats_fetch_retry(&r_vec->tx_sync, start));
		stats->tx_packets += data[0];
		stats->tx_bytes += data[1];
		stats->tx_errors += data[2];
	}
}

static int nfp_net_set_features(struct net_device *netdev,
				netdev_features_t features)
{
	netdev_features_t changed = netdev->features ^ features;
	struct nfp_net *nn = netdev_priv(netdev);
	u32 new_ctrl;
	int err;

	/* Assume this is not called with features we have not advertised */

3153
	new_ctrl = nn->dp.ctrl;
3154 3155 3156

	if (changed & NETIF_F_RXCSUM) {
		if (features & NETIF_F_RXCSUM)
3157
			new_ctrl |= nn->cap & NFP_NET_CFG_CTRL_RXCSUM_ANY;
3158
		else
3159
			new_ctrl &= ~NFP_NET_CFG_CTRL_RXCSUM_ANY;
3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170
	}

	if (changed & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)) {
		if (features & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM))
			new_ctrl |= NFP_NET_CFG_CTRL_TXCSUM;
		else
			new_ctrl &= ~NFP_NET_CFG_CTRL_TXCSUM;
	}

	if (changed & (NETIF_F_TSO | NETIF_F_TSO6)) {
		if (features & (NETIF_F_TSO | NETIF_F_TSO6))
E
Edwin Peer 已提交
3171 3172
			new_ctrl |= nn->cap & NFP_NET_CFG_CTRL_LSO2 ?:
					      NFP_NET_CFG_CTRL_LSO;
3173
		else
E
Edwin Peer 已提交
3174
			new_ctrl &= ~NFP_NET_CFG_CTRL_LSO_ANY;
3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190
	}

	if (changed & NETIF_F_HW_VLAN_CTAG_RX) {
		if (features & NETIF_F_HW_VLAN_CTAG_RX)
			new_ctrl |= NFP_NET_CFG_CTRL_RXVLAN;
		else
			new_ctrl &= ~NFP_NET_CFG_CTRL_RXVLAN;
	}

	if (changed & NETIF_F_HW_VLAN_CTAG_TX) {
		if (features & NETIF_F_HW_VLAN_CTAG_TX)
			new_ctrl |= NFP_NET_CFG_CTRL_TXVLAN;
		else
			new_ctrl &= ~NFP_NET_CFG_CTRL_TXVLAN;
	}

P
Pablo Cascón 已提交
3191 3192 3193 3194 3195 3196 3197
	if (changed & NETIF_F_HW_VLAN_CTAG_FILTER) {
		if (features & NETIF_F_HW_VLAN_CTAG_FILTER)
			new_ctrl |= NFP_NET_CFG_CTRL_CTAG_FILTER;
		else
			new_ctrl &= ~NFP_NET_CFG_CTRL_CTAG_FILTER;
	}

3198 3199 3200 3201 3202 3203 3204
	if (changed & NETIF_F_SG) {
		if (features & NETIF_F_SG)
			new_ctrl |= NFP_NET_CFG_CTRL_GATHER;
		else
			new_ctrl &= ~NFP_NET_CFG_CTRL_GATHER;
	}

3205
	if (changed & NETIF_F_HW_TC && nfp_app_tc_busy(nn->app, nn)) {
3206 3207 3208 3209
		nn_err(nn, "Cannot disable HW TC offload while in use\n");
		return -EBUSY;
	}

3210 3211 3212
	nn_dbg(nn, "Feature change 0x%llx -> 0x%llx (changed=0x%llx)\n",
	       netdev->features, features, changed);

3213
	if (new_ctrl == nn->dp.ctrl)
3214 3215
		return 0;

3216
	nn_dbg(nn, "NIC ctrl: 0x%x -> 0x%x\n", nn->dp.ctrl, new_ctrl);
3217 3218 3219 3220 3221
	nn_writel(nn, NFP_NET_CFG_CTRL, new_ctrl);
	err = nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_GEN);
	if (err)
		return err;

3222
	nn->dp.ctrl = new_ctrl;
3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258

	return 0;
}

static netdev_features_t
nfp_net_features_check(struct sk_buff *skb, struct net_device *dev,
		       netdev_features_t features)
{
	u8 l4_hdr;

	/* We can't do TSO over double tagged packets (802.1AD) */
	features &= vlan_features_check(skb, features);

	if (!skb->encapsulation)
		return features;

	/* Ensure that inner L4 header offset fits into TX descriptor field */
	if (skb_is_gso(skb)) {
		u32 hdrlen;

		hdrlen = skb_inner_transport_header(skb) - skb->data +
			inner_tcp_hdrlen(skb);

		if (unlikely(hdrlen > NFP_NET_LSO_MAX_HDR_SZ))
			features &= ~NETIF_F_GSO_MASK;
	}

	/* VXLAN/GRE check */
	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:
3259
		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
3260 3261 3262 3263 3264 3265 3266 3267
	}

	if (skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
	    skb->inner_protocol != htons(ETH_P_TEB) ||
	    (l4_hdr != IPPROTO_UDP && l4_hdr != IPPROTO_GRE) ||
	    (l4_hdr == IPPROTO_UDP &&
	     (skb_inner_mac_header(skb) - skb_transport_header(skb) !=
	      sizeof(struct udphdr) + sizeof(struct vxlanhdr))))
3268
		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284

	return features;
}

/**
 * nfp_net_set_vxlan_port() - set vxlan port in SW and reconfigure HW
 * @nn:   NFP Net device to reconfigure
 * @idx:  Index into the port table where new port should be written
 * @port: UDP port to configure (pass zero to remove VXLAN port)
 */
static void nfp_net_set_vxlan_port(struct nfp_net *nn, int idx, __be16 port)
{
	int i;

	nn->vxlan_ports[idx] = port;

3285
	if (!(nn->dp.ctrl & NFP_NET_CFG_CTRL_VXLAN))
3286 3287 3288 3289 3290 3291 3292 3293
		return;

	BUILD_BUG_ON(NFP_NET_N_VXLAN_PORTS & 1);
	for (i = 0; i < NFP_NET_N_VXLAN_PORTS; i += 2)
		nn_writel(nn, NFP_NET_CFG_VXLAN_PORT + i * sizeof(port),
			  be16_to_cpu(nn->vxlan_ports[i + 1]) << 16 |
			  be16_to_cpu(nn->vxlan_ports[i]));

3294
	nfp_net_reconfig_post(nn, NFP_NET_CFG_UPDATE_VXLAN);
3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320
}

/**
 * nfp_net_find_vxlan_idx() - find table entry of the port or a free one
 * @nn:   NFP Network structure
 * @port: UDP port to look for
 *
 * Return: if the port is already in the table -- it's position;
 *	   if the port is not in the table -- free position to use;
 *	   if the table is full -- -ENOSPC.
 */
static int nfp_net_find_vxlan_idx(struct nfp_net *nn, __be16 port)
{
	int i, free_idx = -ENOSPC;

	for (i = 0; i < NFP_NET_N_VXLAN_PORTS; i++) {
		if (nn->vxlan_ports[i] == port)
			return i;
		if (!nn->vxlan_usecnt[i])
			free_idx = i;
	}

	return free_idx;
}

static void nfp_net_add_vxlan_port(struct net_device *netdev,
3321
				   struct udp_tunnel_info *ti)
3322 3323 3324 3325
{
	struct nfp_net *nn = netdev_priv(netdev);
	int idx;

3326 3327 3328 3329
	if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
		return;

	idx = nfp_net_find_vxlan_idx(nn, ti->port);
3330 3331 3332 3333
	if (idx == -ENOSPC)
		return;

	if (!nn->vxlan_usecnt[idx]++)
3334
		nfp_net_set_vxlan_port(nn, idx, ti->port);
3335 3336 3337
}

static void nfp_net_del_vxlan_port(struct net_device *netdev,
3338
				   struct udp_tunnel_info *ti)
3339 3340 3341 3342
{
	struct nfp_net *nn = netdev_priv(netdev);
	int idx;

3343 3344 3345 3346
	if (ti->type != UDP_TUNNEL_TYPE_VXLAN)
		return;

	idx = nfp_net_find_vxlan_idx(nn, ti->port);
3347
	if (idx == -ENOSPC || !nn->vxlan_usecnt[idx])
3348 3349 3350 3351 3352 3353
		return;

	if (!--nn->vxlan_usecnt[idx])
		nfp_net_set_vxlan_port(nn, idx, 0);
}

3354 3355 3356
static int
nfp_net_xdp_setup_drv(struct nfp_net *nn, struct bpf_prog *prog,
		      struct netlink_ext_ack *extack)
3357
{
3358
	struct nfp_net_dp *dp;
3359

3360 3361
	if (!prog == !nn->dp.xdp_prog) {
		WRITE_ONCE(nn->dp.xdp_prog, prog);
3362 3363 3364
		return 0;
	}

3365 3366 3367 3368
	dp = nfp_net_clone_dp(nn);
	if (!dp)
		return -ENOMEM;

3369
	dp->xdp_prog = prog;
3370
	dp->num_tx_rings += prog ? nn->dp.num_rx_rings : -nn->dp.num_rx_rings;
3371
	dp->rx_dma_dir = prog ? DMA_BIDIRECTIONAL : DMA_FROM_DEVICE;
3372
	dp->rx_dma_off = prog ? XDP_PACKET_HEADROOM - nn->dp.rx_offset : 0;
3373 3374

	/* We need RX reconfig to remap the buffers (BIDIR vs FROM_DEV) */
3375
	return nfp_net_ring_reconfig(nn, dp, extack);
3376 3377 3378
}

static int
3379
nfp_net_xdp_setup(struct nfp_net *nn, struct bpf_prog *prog, u32 flags,
3380 3381
		  struct netlink_ext_ack *extack)
{
3382
	struct bpf_prog *drv_prog, *offload_prog;
3383 3384
	int err;

3385
	if (nn->xdp_prog && (flags ^ nn->xdp_flags) & XDP_FLAGS_MODES)
3386 3387
		return -EBUSY;

3388 3389 3390 3391
	/* Load both when no flags set to allow easy activation of driver path
	 * when program is replaced by one which can't be offloaded.
	 */
	drv_prog     = flags & XDP_FLAGS_HW_MODE  ? NULL : prog;
3392 3393
	offload_prog = flags & XDP_FLAGS_DRV_MODE ? NULL : prog;

3394
	err = nfp_net_xdp_setup_drv(nn, drv_prog, extack);
3395 3396 3397
	if (err)
		return err;

3398 3399 3400
	err = nfp_app_xdp_offload(nn->app, nn, offload_prog);
	if (err && flags & XDP_FLAGS_HW_MODE)
		return err;
3401 3402 3403 3404

	if (nn->xdp_prog)
		bpf_prog_put(nn->xdp_prog);
	nn->xdp_prog = prog;
3405
	nn->xdp_flags = flags;
3406

3407 3408 3409
	return 0;
}

3410
static int nfp_net_xdp(struct net_device *netdev, struct netdev_bpf *xdp)
3411 3412 3413 3414 3415
{
	struct nfp_net *nn = netdev_priv(netdev);

	switch (xdp->command) {
	case XDP_SETUP_PROG:
3416
	case XDP_SETUP_PROG_HW:
3417 3418
		return nfp_net_xdp_setup(nn, xdp->prog, xdp->flags,
					 xdp->extack);
3419
	case XDP_QUERY_PROG:
3420
		xdp->prog_attached = !!nn->xdp_prog;
3421 3422
		if (nn->dp.bpf_offload_xdp)
			xdp->prog_attached = XDP_ATTACHED_HW;
3423
		xdp->prog_id = nn->xdp_prog ? nn->xdp_prog->aux->id : 0;
3424
		xdp->prog_flags = nn->xdp_prog ? nn->xdp_flags : 0;
3425 3426
		return 0;
	default:
3427
		return nfp_app_bpf(nn->app, nn, xdp);
3428 3429 3430
	}
}

3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451
static int nfp_net_set_mac_address(struct net_device *netdev, void *addr)
{
	struct nfp_net *nn = netdev_priv(netdev);
	struct sockaddr *saddr = addr;
	int err;

	err = eth_prepare_mac_addr_change(netdev, addr);
	if (err)
		return err;

	nfp_net_write_mac_addr(nn, saddr->sa_data);

	err = nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_MACADDR);
	if (err)
		return err;

	eth_commit_mac_addr_change(netdev, addr);

	return 0;
}

J
Jakub Kicinski 已提交
3452
const struct net_device_ops nfp_net_netdev_ops = {
3453 3454 3455 3456
	.ndo_open		= nfp_net_netdev_open,
	.ndo_stop		= nfp_net_netdev_close,
	.ndo_start_xmit		= nfp_net_tx,
	.ndo_get_stats64	= nfp_net_stat64,
P
Pablo Cascón 已提交
3457 3458
	.ndo_vlan_rx_add_vid	= nfp_net_vlan_rx_add_vid,
	.ndo_vlan_rx_kill_vid	= nfp_net_vlan_rx_kill_vid,
3459 3460 3461 3462 3463
	.ndo_set_vf_mac         = nfp_app_set_vf_mac,
	.ndo_set_vf_vlan        = nfp_app_set_vf_vlan,
	.ndo_set_vf_spoofchk    = nfp_app_set_vf_spoofchk,
	.ndo_get_vf_config	= nfp_app_get_vf_config,
	.ndo_set_vf_link_state  = nfp_app_set_vf_link_state,
3464
	.ndo_setup_tc		= nfp_port_setup_tc,
3465 3466 3467
	.ndo_tx_timeout		= nfp_net_tx_timeout,
	.ndo_set_rx_mode	= nfp_net_set_rx_mode,
	.ndo_change_mtu		= nfp_net_change_mtu,
3468
	.ndo_set_mac_address	= nfp_net_set_mac_address,
3469 3470
	.ndo_set_features	= nfp_net_set_features,
	.ndo_features_check	= nfp_net_features_check,
J
Jakub Kicinski 已提交
3471
	.ndo_get_phys_port_name	= nfp_port_get_phys_port_name,
3472 3473
	.ndo_udp_tunnel_add	= nfp_net_add_vxlan_port,
	.ndo_udp_tunnel_del	= nfp_net_del_vxlan_port,
3474
	.ndo_bpf		= nfp_net_xdp,
3475 3476 3477 3478 3479 3480 3481 3482
};

/**
 * nfp_net_info() - Print general info about the NIC
 * @nn:      NFP Net device to reconfigure
 */
void nfp_net_info(struct nfp_net *nn)
{
J
Jakub Kicinski 已提交
3483
	nn_info(nn, "Netronome NFP-6xxx %sNetdev: TxQs=%d/%d RxQs=%d/%d\n",
3484 3485 3486
		nn->dp.is_vf ? "VF " : "",
		nn->dp.num_tx_rings, nn->max_tx_rings,
		nn->dp.num_rx_rings, nn->max_rx_rings);
3487 3488 3489 3490
	nn_info(nn, "VER: %d.%d.%d.%d, Maximum supported MTU: %d\n",
		nn->fw_ver.resv, nn->fw_ver.class,
		nn->fw_ver.major, nn->fw_ver.minor,
		nn->max_mtu);
P
Pablo Cascón 已提交
3491
	nn_info(nn, "CAP: %#x %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
3492 3493 3494 3495 3496 3497 3498 3499 3500 3501
		nn->cap,
		nn->cap & NFP_NET_CFG_CTRL_PROMISC  ? "PROMISC "  : "",
		nn->cap & NFP_NET_CFG_CTRL_L2BC     ? "L2BCFILT " : "",
		nn->cap & NFP_NET_CFG_CTRL_L2MC     ? "L2MCFILT " : "",
		nn->cap & NFP_NET_CFG_CTRL_RXCSUM   ? "RXCSUM "   : "",
		nn->cap & NFP_NET_CFG_CTRL_TXCSUM   ? "TXCSUM "   : "",
		nn->cap & NFP_NET_CFG_CTRL_RXVLAN   ? "RXVLAN "   : "",
		nn->cap & NFP_NET_CFG_CTRL_TXVLAN   ? "TXVLAN "   : "",
		nn->cap & NFP_NET_CFG_CTRL_SCATTER  ? "SCATTER "  : "",
		nn->cap & NFP_NET_CFG_CTRL_GATHER   ? "GATHER "   : "",
E
Edwin Peer 已提交
3502 3503
		nn->cap & NFP_NET_CFG_CTRL_LSO      ? "TSO1 "     : "",
		nn->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSO2 "     : "",
3504 3505
		nn->cap & NFP_NET_CFG_CTRL_RSS      ? "RSS1 "     : "",
		nn->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSS2 "     : "",
P
Pablo Cascón 已提交
3506
		nn->cap & NFP_NET_CFG_CTRL_CTAG_FILTER ? "CTAG_FILTER " : "",
3507 3508 3509 3510
		nn->cap & NFP_NET_CFG_CTRL_L2SWITCH ? "L2SWITCH " : "",
		nn->cap & NFP_NET_CFG_CTRL_MSIXAUTO ? "AUTOMASK " : "",
		nn->cap & NFP_NET_CFG_CTRL_IRQMOD   ? "IRQMOD "   : "",
		nn->cap & NFP_NET_CFG_CTRL_VXLAN    ? "VXLAN "    : "",
3511
		nn->cap & NFP_NET_CFG_CTRL_NVGRE    ? "NVGRE "	  : "",
3512
		nn->cap & NFP_NET_CFG_CTRL_CSUM_COMPLETE ?
3513
						      "RXCSUM_COMPLETE " : "",
3514 3515
		nn->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR " : "",
		nfp_app_extra_cap(nn->app, nn));
3516 3517 3518
}

/**
3519
 * nfp_net_alloc() - Allocate netdev and related structure
3520
 * @pdev:         PCI device
3521
 * @needs_netdev: Whether to allocate a netdev for this vNIC
3522 3523 3524 3525
 * @max_tx_rings: Maximum number of TX rings supported by device
 * @max_rx_rings: Maximum number of RX rings supported by device
 *
 * This function allocates a netdev device and fills in the initial
3526 3527
 * part of the @struct nfp_net structure.  In case of control device
 * nfp_net structure is allocated without the netdev.
3528 3529 3530
 *
 * Return: NFP Net device structure, or ERR_PTR on error.
 */
3531
struct nfp_net *nfp_net_alloc(struct pci_dev *pdev, bool needs_netdev,
3532 3533
			      unsigned int max_tx_rings,
			      unsigned int max_rx_rings)
3534 3535 3536
{
	struct nfp_net *nn;

3537 3538
	if (needs_netdev) {
		struct net_device *netdev;
3539

3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552
		netdev = alloc_etherdev_mqs(sizeof(struct nfp_net),
					    max_tx_rings, max_rx_rings);
		if (!netdev)
			return ERR_PTR(-ENOMEM);

		SET_NETDEV_DEV(netdev, &pdev->dev);
		nn = netdev_priv(netdev);
		nn->dp.netdev = netdev;
	} else {
		nn = vzalloc(sizeof(*nn));
		if (!nn)
			return ERR_PTR(-ENOMEM);
	}
3553

3554
	nn->dp.dev = &pdev->dev;
3555 3556 3557 3558 3559
	nn->pdev = pdev;

	nn->max_tx_rings = max_tx_rings;
	nn->max_rx_rings = max_rx_rings;

3560 3561 3562
	nn->dp.num_tx_rings = min_t(unsigned int,
				    max_tx_rings, num_online_cpus());
	nn->dp.num_rx_rings = min_t(unsigned int, max_rx_rings,
3563
				 netif_get_num_default_rss_queues());
3564

3565 3566 3567
	nn->dp.num_r_vecs = max(nn->dp.num_tx_rings, nn->dp.num_rx_rings);
	nn->dp.num_r_vecs = min_t(unsigned int,
				  nn->dp.num_r_vecs, num_online_cpus());
J
Jakub Kicinski 已提交
3568

3569 3570
	nn->dp.txd_cnt = NFP_NET_TX_DESCS_DEFAULT;
	nn->dp.rxd_cnt = NFP_NET_RX_DESCS_DEFAULT;
3571 3572 3573 3574

	spin_lock_init(&nn->reconfig_lock);
	spin_lock_init(&nn->link_status_lock);

3575
	timer_setup(&nn->reconfig_timer, nfp_net_reconfig_timer, 0);
3576

3577 3578 3579 3580
	return nn;
}

/**
3581
 * nfp_net_free() - Undo what @nfp_net_alloc() did
3582 3583
 * @nn:      NFP Net device to reconfigure
 */
3584
void nfp_net_free(struct nfp_net *nn)
3585
{
3586 3587 3588 3589
	if (nn->dp.netdev)
		free_netdev(nn->dp.netdev);
	else
		vfree(nn);
3590 3591
}

3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612
/**
 * nfp_net_rss_key_sz() - Get current size of the RSS key
 * @nn:		NFP Net device instance
 *
 * Return: size of the RSS key for currently selected hash function.
 */
unsigned int nfp_net_rss_key_sz(struct nfp_net *nn)
{
	switch (nn->rss_hfunc) {
	case ETH_RSS_HASH_TOP:
		return NFP_NET_CFG_RSS_KEY_SZ;
	case ETH_RSS_HASH_XOR:
		return 0;
	case ETH_RSS_HASH_CRC32:
		return 4;
	}

	nn_warn(nn, "Unknown hash function: %u\n", nn->rss_hfunc);
	return 0;
}

3613 3614 3615 3616 3617 3618
/**
 * nfp_net_rss_init() - Set the initial RSS parameters
 * @nn:	     NFP Net device to reconfigure
 */
static void nfp_net_rss_init(struct nfp_net *nn)
{
3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630
	unsigned long func_bit, rss_cap_hfunc;
	u32 reg;

	/* Read the RSS function capability and select first supported func */
	reg = nn_readl(nn, NFP_NET_CFG_RSS_CAP);
	rss_cap_hfunc =	FIELD_GET(NFP_NET_CFG_RSS_CAP_HFUNC, reg);
	if (!rss_cap_hfunc)
		rss_cap_hfunc =	FIELD_GET(NFP_NET_CFG_RSS_CAP_HFUNC,
					  NFP_NET_CFG_RSS_TOEPLITZ);

	func_bit = find_first_bit(&rss_cap_hfunc, NFP_NET_CFG_RSS_HFUNCS);
	if (func_bit == NFP_NET_CFG_RSS_HFUNCS) {
3631
		dev_warn(nn->dp.dev,
3632 3633 3634 3635 3636 3637
			 "Bad RSS config, defaulting to Toeplitz hash\n");
		func_bit = ETH_RSS_HASH_TOP_BIT;
	}
	nn->rss_hfunc = 1 << func_bit;

	netdev_rss_key_fill(nn->rss_key, nfp_net_rss_key_sz(nn));
3638

3639
	nfp_net_rss_init_itbl(nn);
3640 3641 3642 3643

	/* Enable IPv4/IPv6 TCP by default */
	nn->rss_cfg = NFP_NET_CFG_RSS_IPV4_TCP |
		      NFP_NET_CFG_RSS_IPV6_TCP |
3644
		      FIELD_PREP(NFP_NET_CFG_RSS_HFUNC, nn->rss_hfunc) |
3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659
		      NFP_NET_CFG_RSS_MASK;
}

/**
 * nfp_net_irqmod_init() - Set the initial IRQ moderation parameters
 * @nn:	     NFP Net device to reconfigure
 */
static void nfp_net_irqmod_init(struct nfp_net *nn)
{
	nn->rx_coalesce_usecs      = 50;
	nn->rx_coalesce_max_frames = 64;
	nn->tx_coalesce_usecs      = 50;
	nn->tx_coalesce_max_frames = 64;
}

3660
static void nfp_net_netdev_init(struct nfp_net *nn)
3661
{
3662
	struct net_device *netdev = nn->dp.netdev;
3663

3664
	nfp_net_write_mac_addr(nn, nn->dp.netdev->dev_addr);
3665

3666
	netdev->mtu = nn->dp.mtu;
3667 3668 3669 3670 3671 3672 3673

	/* Advertise/enable offloads based on capabilities
	 *
	 * Note: netdev->features show the currently enabled features
	 * and netdev->hw_features advertises which features are
	 * supported.  By default we enable most features.
	 */
3674 3675 3676
	if (nn->cap & NFP_NET_CFG_CTRL_LIVE_ADDR)
		netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE;

3677
	netdev->hw_features = NETIF_F_HIGHDMA;
3678
	if (nn->cap & NFP_NET_CFG_CTRL_RXCSUM_ANY) {
3679
		netdev->hw_features |= NETIF_F_RXCSUM;
3680
		nn->dp.ctrl |= nn->cap & NFP_NET_CFG_CTRL_RXCSUM_ANY;
3681 3682 3683
	}
	if (nn->cap & NFP_NET_CFG_CTRL_TXCSUM) {
		netdev->hw_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
3684
		nn->dp.ctrl |= NFP_NET_CFG_CTRL_TXCSUM;
3685 3686 3687
	}
	if (nn->cap & NFP_NET_CFG_CTRL_GATHER) {
		netdev->hw_features |= NETIF_F_SG;
3688
		nn->dp.ctrl |= NFP_NET_CFG_CTRL_GATHER;
3689
	}
E
Edwin Peer 已提交
3690 3691
	if ((nn->cap & NFP_NET_CFG_CTRL_LSO && nn->fw_ver.major > 2) ||
	    nn->cap & NFP_NET_CFG_CTRL_LSO2) {
3692
		netdev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
E
Edwin Peer 已提交
3693 3694
		nn->dp.ctrl |= nn->cap & NFP_NET_CFG_CTRL_LSO2 ?:
					 NFP_NET_CFG_CTRL_LSO;
3695
	}
3696
	if (nn->cap & NFP_NET_CFG_CTRL_RSS_ANY)
3697 3698 3699 3700 3701 3702
		netdev->hw_features |= NETIF_F_RXHASH;
	if (nn->cap & NFP_NET_CFG_CTRL_VXLAN &&
	    nn->cap & NFP_NET_CFG_CTRL_NVGRE) {
		if (nn->cap & NFP_NET_CFG_CTRL_LSO)
			netdev->hw_features |= NETIF_F_GSO_GRE |
					       NETIF_F_GSO_UDP_TUNNEL;
3703
		nn->dp.ctrl |= NFP_NET_CFG_CTRL_VXLAN | NFP_NET_CFG_CTRL_NVGRE;
3704 3705 3706 3707 3708 3709 3710 3711

		netdev->hw_enc_features = netdev->hw_features;
	}

	netdev->vlan_features = netdev->hw_features;

	if (nn->cap & NFP_NET_CFG_CTRL_RXVLAN) {
		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_RX;
3712
		nn->dp.ctrl |= NFP_NET_CFG_CTRL_RXVLAN;
3713 3714
	}
	if (nn->cap & NFP_NET_CFG_CTRL_TXVLAN) {
E
Edwin Peer 已提交
3715 3716 3717 3718 3719 3720
		if (nn->cap & NFP_NET_CFG_CTRL_LSO2) {
			nn_warn(nn, "Device advertises both TSO2 and TXVLAN. Refusing to enable TXVLAN.\n");
		} else {
			netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX;
			nn->dp.ctrl |= NFP_NET_CFG_CTRL_TXVLAN;
		}
3721
	}
P
Pablo Cascón 已提交
3722 3723 3724 3725
	if (nn->cap & NFP_NET_CFG_CTRL_CTAG_FILTER) {
		netdev->hw_features |= NETIF_F_HW_VLAN_CTAG_FILTER;
		nn->dp.ctrl |= NFP_NET_CFG_CTRL_CTAG_FILTER;
	}
3726 3727 3728

	netdev->features = netdev->hw_features;

3729
	if (nfp_app_has_tc(nn->app))
3730 3731
		netdev->hw_features |= NETIF_F_HW_TC;

3732 3733
	/* Advertise but disable TSO by default. */
	netdev->features &= ~(NETIF_F_TSO | NETIF_F_TSO6);
E
Edwin Peer 已提交
3734
	nn->dp.ctrl &= ~NFP_NET_CFG_CTRL_LSO_ANY;
3735

3736 3737 3738 3739
	/* Finalise the netdev setup */
	netdev->netdev_ops = &nfp_net_netdev_ops;
	netdev->watchdog_timeo = msecs_to_jiffies(5 * 1000);

S
Simon Horman 已提交
3740 3741
	SWITCHDEV_SET_OPS(netdev, &nfp_port_switchdev_ops);

3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766
	/* MTU range: 68 - hw-specific max */
	netdev->min_mtu = ETH_MIN_MTU;
	netdev->max_mtu = nn->max_mtu;

	netif_carrier_off(netdev);

	nfp_net_set_ethtool_ops(netdev);
}

/**
 * nfp_net_init() - Initialise/finalise the nfp_net structure
 * @nn:		NFP Net device structure
 *
 * Return: 0 on success or negative errno on error.
 */
int nfp_net_init(struct nfp_net *nn)
{
	int err;

	nn->dp.rx_dma_dir = DMA_FROM_DEVICE;

	/* Get some of the read-only fields from the BAR */
	nn->cap = nn_readl(nn, NFP_NET_CFG_CAP);
	nn->max_mtu = nn_readl(nn, NFP_NET_CFG_MAX_MTU);

3767 3768 3769 3770
	/* ABI 4.x and ctrl vNIC always use chained metadata, in other cases
	 * we allow use of non-chained metadata if RSS(v1) is the only
	 * advertised capability requiring metadata.
	 */
3771
	nn->dp.chained_metadata_format = nn->fw_ver.major == 4 ||
J
Jakub Kicinski 已提交
3772
					 !nn->dp.netdev ||
3773
					 !(nn->cap & NFP_NET_CFG_CTRL_RSS) ||
3774
					 nn->cap & NFP_NET_CFG_CTRL_CHAIN_META;
3775 3776 3777
	/* RSS(v1) uses non-chained metadata format, except in ABI 4.x where
	 * it has the same meaning as RSSv2.
	 */
3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807
	if (nn->dp.chained_metadata_format && nn->fw_ver.major != 4)
		nn->cap &= ~NFP_NET_CFG_CTRL_RSS;

	/* Determine RX packet/metadata boundary offset */
	if (nn->fw_ver.major >= 2) {
		u32 reg;

		reg = nn_readl(nn, NFP_NET_CFG_RX_OFFSET);
		if (reg > NFP_NET_MAX_PREPEND) {
			nn_err(nn, "Invalid rx offset: %d\n", reg);
			return -EINVAL;
		}
		nn->dp.rx_offset = reg;
	} else {
		nn->dp.rx_offset = NFP_NET_RX_OFFSET;
	}

	/* Set default MTU and Freelist buffer size */
	if (nn->max_mtu < NFP_NET_DEFAULT_MTU)
		nn->dp.mtu = nn->max_mtu;
	else
		nn->dp.mtu = NFP_NET_DEFAULT_MTU;
	nn->dp.fl_bufsz = nfp_net_calc_fl_bufsz(&nn->dp);

	if (nn->cap & NFP_NET_CFG_CTRL_RSS_ANY) {
		nfp_net_rss_init(nn);
		nn->dp.ctrl |= nn->cap & NFP_NET_CFG_CTRL_RSS2 ?:
					 NFP_NET_CFG_CTRL_RSS;
	}

3808 3809
	/* Allow L2 Broadcast and Multicast through by default, if supported */
	if (nn->cap & NFP_NET_CFG_CTRL_L2BC)
3810
		nn->dp.ctrl |= NFP_NET_CFG_CTRL_L2BC;
3811 3812 3813 3814

	/* Allow IRQ moderation, if supported */
	if (nn->cap & NFP_NET_CFG_CTRL_IRQMOD) {
		nfp_net_irqmod_init(nn);
3815
		nn->dp.ctrl |= NFP_NET_CFG_CTRL_IRQMOD;
3816 3817
	}

3818 3819 3820 3821 3822
	err = nfp_net_tlv_caps_parse(&nn->pdev->dev, nn->dp.ctrl_bar,
				     &nn->tlv_caps);
	if (err)
		return err;

3823 3824 3825
	if (nn->dp.netdev)
		nfp_net_netdev_init(nn);

3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837
	/* Stash the re-configuration queue away.  First odd queue in TX Bar */
	nn->qcp_cfg = nn->tx_bar + NFP_QCP_QUEUE_ADDR_SZ;

	/* Make sure the FW knows the netdev is supposed to be disabled here */
	nn_writel(nn, NFP_NET_CFG_CTRL, 0);
	nn_writeq(nn, NFP_NET_CFG_TXRS_ENABLE, 0);
	nn_writeq(nn, NFP_NET_CFG_RXRS_ENABLE, 0);
	err = nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_RING |
				   NFP_NET_CFG_UPDATE_GEN);
	if (err)
		return err;

3838
	nfp_net_vecs_init(nn);
3839

3840 3841 3842
	if (!nn->dp.netdev)
		return 0;
	return register_netdev(nn->dp.netdev);
3843 3844 3845
}

/**
3846 3847
 * nfp_net_clean() - Undo what nfp_net_init() did.
 * @nn:		NFP Net device structure
3848
 */
3849
void nfp_net_clean(struct nfp_net *nn)
3850
{
3851 3852 3853
	if (!nn->dp.netdev)
		return;

3854
	unregister_netdev(nn->dp.netdev);
3855
}