1 /* Intel Ethernet Switch Host Interface Driver
2  * Copyright(c) 2013 - 2015 Intel Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * The full GNU General Public License is included in this distribution in
14  * the file called "COPYING".
15  *
16  * Contact Information:
17  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
18  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
19  */
20 
21 #include <linux/module.h>
22 #include <linux/aer.h>
23 
24 #include "fm10k.h"
25 
26 static const struct fm10k_info *fm10k_info_tbl[] = {
27 	[fm10k_device_pf] = &fm10k_pf_info,
28 	[fm10k_device_vf] = &fm10k_vf_info,
29 };
30 
31 /**
32  * fm10k_pci_tbl - PCI Device ID Table
33  *
34  * Wildcard entries (PCI_ANY_ID) should come last
35  * Last entry must be all 0s
36  *
37  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
38  *   Class, Class Mask, private data (not used) }
39  */
40 static const struct pci_device_id fm10k_pci_tbl[] = {
41 	{ PCI_VDEVICE(INTEL, FM10K_DEV_ID_PF), fm10k_device_pf },
42 	{ PCI_VDEVICE(INTEL, FM10K_DEV_ID_VF), fm10k_device_vf },
43 	/* required last entry */
44 	{ 0, }
45 };
46 MODULE_DEVICE_TABLE(pci, fm10k_pci_tbl);
47 
48 u16 fm10k_read_pci_cfg_word(struct fm10k_hw *hw, u32 reg)
49 {
50 	struct fm10k_intfc *interface = hw->back;
51 	u16 value = 0;
52 
53 	if (FM10K_REMOVED(hw->hw_addr))
54 		return ~value;
55 
56 	pci_read_config_word(interface->pdev, reg, &value);
57 	if (value == 0xFFFF)
58 		fm10k_write_flush(hw);
59 
60 	return value;
61 }
62 
63 u32 fm10k_read_reg(struct fm10k_hw *hw, int reg)
64 {
65 	u32 __iomem *hw_addr = ACCESS_ONCE(hw->hw_addr);
66 	u32 value = 0;
67 
68 	if (FM10K_REMOVED(hw_addr))
69 		return ~value;
70 
71 	value = readl(&hw_addr[reg]);
72 	if (!(~value) && (!reg || !(~readl(hw_addr)))) {
73 		struct fm10k_intfc *interface = hw->back;
74 		struct net_device *netdev = interface->netdev;
75 
76 		hw->hw_addr = NULL;
77 		netif_device_detach(netdev);
78 		netdev_err(netdev, "PCIe link lost, device now detached\n");
79 	}
80 
81 	return value;
82 }
83 
84 static int fm10k_hw_ready(struct fm10k_intfc *interface)
85 {
86 	struct fm10k_hw *hw = &interface->hw;
87 
88 	fm10k_write_flush(hw);
89 
90 	return FM10K_REMOVED(hw->hw_addr) ? -ENODEV : 0;
91 }
92 
93 void fm10k_service_event_schedule(struct fm10k_intfc *interface)
94 {
95 	if (!test_bit(__FM10K_SERVICE_DISABLE, &interface->state) &&
96 	    !test_and_set_bit(__FM10K_SERVICE_SCHED, &interface->state))
97 		queue_work(fm10k_workqueue, &interface->service_task);
98 }
99 
100 static void fm10k_service_event_complete(struct fm10k_intfc *interface)
101 {
102 	BUG_ON(!test_bit(__FM10K_SERVICE_SCHED, &interface->state));
103 
104 	/* flush memory to make sure state is correct before next watchog */
105 	smp_mb__before_atomic();
106 	clear_bit(__FM10K_SERVICE_SCHED, &interface->state);
107 }
108 
109 /**
110  * fm10k_service_timer - Timer Call-back
111  * @data: pointer to interface cast into an unsigned long
112  **/
113 static void fm10k_service_timer(unsigned long data)
114 {
115 	struct fm10k_intfc *interface = (struct fm10k_intfc *)data;
116 
117 	/* Reset the timer */
118 	mod_timer(&interface->service_timer, (HZ * 2) + jiffies);
119 
120 	fm10k_service_event_schedule(interface);
121 }
122 
123 static void fm10k_detach_subtask(struct fm10k_intfc *interface)
124 {
125 	struct net_device *netdev = interface->netdev;
126 
127 	/* do nothing if device is still present or hw_addr is set */
128 	if (netif_device_present(netdev) || interface->hw.hw_addr)
129 		return;
130 
131 	rtnl_lock();
132 
133 	if (netif_running(netdev))
134 		dev_close(netdev);
135 
136 	rtnl_unlock();
137 }
138 
139 static void fm10k_reinit(struct fm10k_intfc *interface)
140 {
141 	struct net_device *netdev = interface->netdev;
142 	struct fm10k_hw *hw = &interface->hw;
143 	int err;
144 
145 	WARN_ON(in_interrupt());
146 
147 	/* put off any impending NetWatchDogTimeout */
148 	netdev->trans_start = jiffies;
149 
150 	while (test_and_set_bit(__FM10K_RESETTING, &interface->state))
151 		usleep_range(1000, 2000);
152 
153 	rtnl_lock();
154 
155 	fm10k_iov_suspend(interface->pdev);
156 
157 	if (netif_running(netdev))
158 		fm10k_close(netdev);
159 
160 	fm10k_mbx_free_irq(interface);
161 
162 	/* free interrupts */
163 	fm10k_clear_queueing_scheme(interface);
164 
165 	/* delay any future reset requests */
166 	interface->last_reset = jiffies + (10 * HZ);
167 
168 	/* reset and initialize the hardware so it is in a known state */
169 	err = hw->mac.ops.reset_hw(hw);
170 	if (err) {
171 		dev_err(&interface->pdev->dev, "reset_hw failed: %d\n", err);
172 		goto reinit_err;
173 	}
174 
175 	err = hw->mac.ops.init_hw(hw);
176 	if (err) {
177 		dev_err(&interface->pdev->dev, "init_hw failed: %d\n", err);
178 		goto reinit_err;
179 	}
180 
181 	err = fm10k_init_queueing_scheme(interface);
182 	if (err) {
183 		dev_err(&interface->pdev->dev,
184 			"init_queueing_scheme failed: %d\n", err);
185 		goto reinit_err;
186 	}
187 
188 	/* reassociate interrupts */
189 	err = fm10k_mbx_request_irq(interface);
190 	if (err)
191 		goto err_mbx_irq;
192 
193 	err = fm10k_hw_ready(interface);
194 	if (err)
195 		goto err_open;
196 
197 	/* update hardware address for VFs if perm_addr has changed */
198 	if (hw->mac.type == fm10k_mac_vf) {
199 		if (is_valid_ether_addr(hw->mac.perm_addr)) {
200 			ether_addr_copy(hw->mac.addr, hw->mac.perm_addr);
201 			ether_addr_copy(netdev->perm_addr, hw->mac.perm_addr);
202 			ether_addr_copy(netdev->dev_addr, hw->mac.perm_addr);
203 			netdev->addr_assign_type &= ~NET_ADDR_RANDOM;
204 		}
205 
206 		if (hw->mac.vlan_override)
207 			netdev->features &= ~NETIF_F_HW_VLAN_CTAG_RX;
208 		else
209 			netdev->features |= NETIF_F_HW_VLAN_CTAG_RX;
210 	}
211 
212 	/* reset clock */
213 	fm10k_ts_reset(interface);
214 
215 	err = netif_running(netdev) ? fm10k_open(netdev) : 0;
216 	if (err)
217 		goto err_open;
218 
219 	fm10k_iov_resume(interface->pdev);
220 
221 	rtnl_unlock();
222 
223 	clear_bit(__FM10K_RESETTING, &interface->state);
224 
225 	return;
226 err_open:
227 	fm10k_mbx_free_irq(interface);
228 err_mbx_irq:
229 	fm10k_clear_queueing_scheme(interface);
230 reinit_err:
231 	netif_device_detach(netdev);
232 
233 	rtnl_unlock();
234 
235 	clear_bit(__FM10K_RESETTING, &interface->state);
236 }
237 
238 static void fm10k_reset_subtask(struct fm10k_intfc *interface)
239 {
240 	if (!(interface->flags & FM10K_FLAG_RESET_REQUESTED))
241 		return;
242 
243 	interface->flags &= ~FM10K_FLAG_RESET_REQUESTED;
244 
245 	netdev_err(interface->netdev, "Reset interface\n");
246 
247 	fm10k_reinit(interface);
248 }
249 
250 /**
251  * fm10k_configure_swpri_map - Configure Receive SWPRI to PC mapping
252  * @interface: board private structure
253  *
254  * Configure the SWPRI to PC mapping for the port.
255  **/
256 static void fm10k_configure_swpri_map(struct fm10k_intfc *interface)
257 {
258 	struct net_device *netdev = interface->netdev;
259 	struct fm10k_hw *hw = &interface->hw;
260 	int i;
261 
262 	/* clear flag indicating update is needed */
263 	interface->flags &= ~FM10K_FLAG_SWPRI_CONFIG;
264 
265 	/* these registers are only available on the PF */
266 	if (hw->mac.type != fm10k_mac_pf)
267 		return;
268 
269 	/* configure SWPRI to PC map */
270 	for (i = 0; i < FM10K_SWPRI_MAX; i++)
271 		fm10k_write_reg(hw, FM10K_SWPRI_MAP(i),
272 				netdev_get_prio_tc_map(netdev, i));
273 }
274 
275 /**
276  * fm10k_watchdog_update_host_state - Update the link status based on host.
277  * @interface: board private structure
278  **/
279 static void fm10k_watchdog_update_host_state(struct fm10k_intfc *interface)
280 {
281 	struct fm10k_hw *hw = &interface->hw;
282 	s32 err;
283 
284 	if (test_bit(__FM10K_LINK_DOWN, &interface->state)) {
285 		interface->host_ready = false;
286 		if (time_is_after_jiffies(interface->link_down_event))
287 			return;
288 		clear_bit(__FM10K_LINK_DOWN, &interface->state);
289 	}
290 
291 	if (interface->flags & FM10K_FLAG_SWPRI_CONFIG) {
292 		if (rtnl_trylock()) {
293 			fm10k_configure_swpri_map(interface);
294 			rtnl_unlock();
295 		}
296 	}
297 
298 	/* lock the mailbox for transmit and receive */
299 	fm10k_mbx_lock(interface);
300 
301 	err = hw->mac.ops.get_host_state(hw, &interface->host_ready);
302 	if (err && time_is_before_jiffies(interface->last_reset))
303 		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
304 
305 	/* free the lock */
306 	fm10k_mbx_unlock(interface);
307 }
308 
309 /**
310  * fm10k_mbx_subtask - Process upstream and downstream mailboxes
311  * @interface: board private structure
312  *
313  * This function will process both the upstream and downstream mailboxes.
314  **/
315 static void fm10k_mbx_subtask(struct fm10k_intfc *interface)
316 {
317 	/* process upstream mailbox and update device state */
318 	fm10k_watchdog_update_host_state(interface);
319 
320 	/* process downstream mailboxes */
321 	fm10k_iov_mbx(interface);
322 }
323 
324 /**
325  * fm10k_watchdog_host_is_ready - Update netdev status based on host ready
326  * @interface: board private structure
327  **/
328 static void fm10k_watchdog_host_is_ready(struct fm10k_intfc *interface)
329 {
330 	struct net_device *netdev = interface->netdev;
331 
332 	/* only continue if link state is currently down */
333 	if (netif_carrier_ok(netdev))
334 		return;
335 
336 	netif_info(interface, drv, netdev, "NIC Link is up\n");
337 
338 	netif_carrier_on(netdev);
339 	netif_tx_wake_all_queues(netdev);
340 }
341 
342 /**
343  * fm10k_watchdog_host_not_ready - Update netdev status based on host not ready
344  * @interface: board private structure
345  **/
346 static void fm10k_watchdog_host_not_ready(struct fm10k_intfc *interface)
347 {
348 	struct net_device *netdev = interface->netdev;
349 
350 	/* only continue if link state is currently up */
351 	if (!netif_carrier_ok(netdev))
352 		return;
353 
354 	netif_info(interface, drv, netdev, "NIC Link is down\n");
355 
356 	netif_carrier_off(netdev);
357 	netif_tx_stop_all_queues(netdev);
358 }
359 
360 /**
361  * fm10k_update_stats - Update the board statistics counters.
362  * @interface: board private structure
363  **/
364 void fm10k_update_stats(struct fm10k_intfc *interface)
365 {
366 	struct net_device_stats *net_stats = &interface->netdev->stats;
367 	struct fm10k_hw *hw = &interface->hw;
368 	u64 hw_csum_tx_good = 0, hw_csum_rx_good = 0, rx_length_errors = 0;
369 	u64 rx_switch_errors = 0, rx_drops = 0, rx_pp_errors = 0;
370 	u64 rx_link_errors = 0;
371 	u64 rx_errors = 0, rx_csum_errors = 0, tx_csum_errors = 0;
372 	u64 restart_queue = 0, tx_busy = 0, alloc_failed = 0;
373 	u64 rx_bytes_nic = 0, rx_pkts_nic = 0, rx_drops_nic = 0;
374 	u64 tx_bytes_nic = 0, tx_pkts_nic = 0;
375 	u64 bytes, pkts;
376 	int i;
377 
378 	/* do not allow stats update via service task for next second */
379 	interface->next_stats_update = jiffies + HZ;
380 
381 	/* gather some stats to the interface struct that are per queue */
382 	for (bytes = 0, pkts = 0, i = 0; i < interface->num_tx_queues; i++) {
383 		struct fm10k_ring *tx_ring = interface->tx_ring[i];
384 
385 		restart_queue += tx_ring->tx_stats.restart_queue;
386 		tx_busy += tx_ring->tx_stats.tx_busy;
387 		tx_csum_errors += tx_ring->tx_stats.csum_err;
388 		bytes += tx_ring->stats.bytes;
389 		pkts += tx_ring->stats.packets;
390 		hw_csum_tx_good += tx_ring->tx_stats.csum_good;
391 	}
392 
393 	interface->restart_queue = restart_queue;
394 	interface->tx_busy = tx_busy;
395 	net_stats->tx_bytes = bytes;
396 	net_stats->tx_packets = pkts;
397 	interface->tx_csum_errors = tx_csum_errors;
398 	interface->hw_csum_tx_good = hw_csum_tx_good;
399 
400 	/* gather some stats to the interface struct that are per queue */
401 	for (bytes = 0, pkts = 0, i = 0; i < interface->num_rx_queues; i++) {
402 		struct fm10k_ring *rx_ring = interface->rx_ring[i];
403 
404 		bytes += rx_ring->stats.bytes;
405 		pkts += rx_ring->stats.packets;
406 		alloc_failed += rx_ring->rx_stats.alloc_failed;
407 		rx_csum_errors += rx_ring->rx_stats.csum_err;
408 		rx_errors += rx_ring->rx_stats.errors;
409 		hw_csum_rx_good += rx_ring->rx_stats.csum_good;
410 		rx_switch_errors += rx_ring->rx_stats.switch_errors;
411 		rx_drops += rx_ring->rx_stats.drops;
412 		rx_pp_errors += rx_ring->rx_stats.pp_errors;
413 		rx_link_errors += rx_ring->rx_stats.link_errors;
414 		rx_length_errors += rx_ring->rx_stats.length_errors;
415 	}
416 
417 	net_stats->rx_bytes = bytes;
418 	net_stats->rx_packets = pkts;
419 	interface->alloc_failed = alloc_failed;
420 	interface->rx_csum_errors = rx_csum_errors;
421 	interface->hw_csum_rx_good = hw_csum_rx_good;
422 	interface->rx_switch_errors = rx_switch_errors;
423 	interface->rx_drops = rx_drops;
424 	interface->rx_pp_errors = rx_pp_errors;
425 	interface->rx_link_errors = rx_link_errors;
426 	interface->rx_length_errors = rx_length_errors;
427 
428 	hw->mac.ops.update_hw_stats(hw, &interface->stats);
429 
430 	for (i = 0; i < hw->mac.max_queues; i++) {
431 		struct fm10k_hw_stats_q *q = &interface->stats.q[i];
432 
433 		tx_bytes_nic += q->tx_bytes.count;
434 		tx_pkts_nic += q->tx_packets.count;
435 		rx_bytes_nic += q->rx_bytes.count;
436 		rx_pkts_nic += q->rx_packets.count;
437 		rx_drops_nic += q->rx_drops.count;
438 	}
439 
440 	interface->tx_bytes_nic = tx_bytes_nic;
441 	interface->tx_packets_nic = tx_pkts_nic;
442 	interface->rx_bytes_nic = rx_bytes_nic;
443 	interface->rx_packets_nic = rx_pkts_nic;
444 	interface->rx_drops_nic = rx_drops_nic;
445 
446 	/* Fill out the OS statistics structure */
447 	net_stats->rx_errors = rx_errors;
448 	net_stats->rx_dropped = interface->stats.nodesc_drop.count;
449 }
450 
451 /**
452  * fm10k_watchdog_flush_tx - flush queues on host not ready
453  * @interface - pointer to the device interface structure
454  **/
455 static void fm10k_watchdog_flush_tx(struct fm10k_intfc *interface)
456 {
457 	int some_tx_pending = 0;
458 	int i;
459 
460 	/* nothing to do if carrier is up */
461 	if (netif_carrier_ok(interface->netdev))
462 		return;
463 
464 	for (i = 0; i < interface->num_tx_queues; i++) {
465 		struct fm10k_ring *tx_ring = interface->tx_ring[i];
466 
467 		if (tx_ring->next_to_use != tx_ring->next_to_clean) {
468 			some_tx_pending = 1;
469 			break;
470 		}
471 	}
472 
473 	/* We've lost link, so the controller stops DMA, but we've got
474 	 * queued Tx work that's never going to get done, so reset
475 	 * controller to flush Tx.
476 	 */
477 	if (some_tx_pending)
478 		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
479 }
480 
481 /**
482  * fm10k_watchdog_subtask - check and bring link up
483  * @interface - pointer to the device interface structure
484  **/
485 static void fm10k_watchdog_subtask(struct fm10k_intfc *interface)
486 {
487 	/* if interface is down do nothing */
488 	if (test_bit(__FM10K_DOWN, &interface->state) ||
489 	    test_bit(__FM10K_RESETTING, &interface->state))
490 		return;
491 
492 	if (interface->host_ready)
493 		fm10k_watchdog_host_is_ready(interface);
494 	else
495 		fm10k_watchdog_host_not_ready(interface);
496 
497 	/* update stats only once every second */
498 	if (time_is_before_jiffies(interface->next_stats_update))
499 		fm10k_update_stats(interface);
500 
501 	/* flush any uncompleted work */
502 	fm10k_watchdog_flush_tx(interface);
503 }
504 
505 /**
506  * fm10k_check_hang_subtask - check for hung queues and dropped interrupts
507  * @interface - pointer to the device interface structure
508  *
509  * This function serves two purposes.  First it strobes the interrupt lines
510  * in order to make certain interrupts are occurring.  Secondly it sets the
511  * bits needed to check for TX hangs.  As a result we should immediately
512  * determine if a hang has occurred.
513  */
514 static void fm10k_check_hang_subtask(struct fm10k_intfc *interface)
515 {
516 	int i;
517 
518 	/* If we're down or resetting, just bail */
519 	if (test_bit(__FM10K_DOWN, &interface->state) ||
520 	    test_bit(__FM10K_RESETTING, &interface->state))
521 		return;
522 
523 	/* rate limit tx hang checks to only once every 2 seconds */
524 	if (time_is_after_eq_jiffies(interface->next_tx_hang_check))
525 		return;
526 	interface->next_tx_hang_check = jiffies + (2 * HZ);
527 
528 	if (netif_carrier_ok(interface->netdev)) {
529 		/* Force detection of hung controller */
530 		for (i = 0; i < interface->num_tx_queues; i++)
531 			set_check_for_tx_hang(interface->tx_ring[i]);
532 
533 		/* Rearm all in-use q_vectors for immediate firing */
534 		for (i = 0; i < interface->num_q_vectors; i++) {
535 			struct fm10k_q_vector *qv = interface->q_vector[i];
536 
537 			if (!qv->tx.count && !qv->rx.count)
538 				continue;
539 			writel(FM10K_ITR_ENABLE | FM10K_ITR_PENDING2, qv->itr);
540 		}
541 	}
542 }
543 
544 /**
545  * fm10k_service_task - manages and runs subtasks
546  * @work: pointer to work_struct containing our data
547  **/
548 static void fm10k_service_task(struct work_struct *work)
549 {
550 	struct fm10k_intfc *interface;
551 
552 	interface = container_of(work, struct fm10k_intfc, service_task);
553 
554 	/* tasks run even when interface is down */
555 	fm10k_mbx_subtask(interface);
556 	fm10k_detach_subtask(interface);
557 	fm10k_reset_subtask(interface);
558 
559 	/* tasks only run when interface is up */
560 	fm10k_watchdog_subtask(interface);
561 	fm10k_check_hang_subtask(interface);
562 	fm10k_ts_tx_subtask(interface);
563 
564 	/* release lock on service events to allow scheduling next event */
565 	fm10k_service_event_complete(interface);
566 }
567 
568 /**
569  * fm10k_configure_tx_ring - Configure Tx ring after Reset
570  * @interface: board private structure
571  * @ring: structure containing ring specific data
572  *
573  * Configure the Tx descriptor ring after a reset.
574  **/
575 static void fm10k_configure_tx_ring(struct fm10k_intfc *interface,
576 				    struct fm10k_ring *ring)
577 {
578 	struct fm10k_hw *hw = &interface->hw;
579 	u64 tdba = ring->dma;
580 	u32 size = ring->count * sizeof(struct fm10k_tx_desc);
581 	u32 txint = FM10K_INT_MAP_DISABLE;
582 	u32 txdctl = FM10K_TXDCTL_ENABLE | (1 << FM10K_TXDCTL_MAX_TIME_SHIFT);
583 	u8 reg_idx = ring->reg_idx;
584 
585 	/* disable queue to avoid issues while updating state */
586 	fm10k_write_reg(hw, FM10K_TXDCTL(reg_idx), 0);
587 	fm10k_write_flush(hw);
588 
589 	/* possible poll here to verify ring resources have been cleaned */
590 
591 	/* set location and size for descriptor ring */
592 	fm10k_write_reg(hw, FM10K_TDBAL(reg_idx), tdba & DMA_BIT_MASK(32));
593 	fm10k_write_reg(hw, FM10K_TDBAH(reg_idx), tdba >> 32);
594 	fm10k_write_reg(hw, FM10K_TDLEN(reg_idx), size);
595 
596 	/* reset head and tail pointers */
597 	fm10k_write_reg(hw, FM10K_TDH(reg_idx), 0);
598 	fm10k_write_reg(hw, FM10K_TDT(reg_idx), 0);
599 
600 	/* store tail pointer */
601 	ring->tail = &interface->uc_addr[FM10K_TDT(reg_idx)];
602 
603 	/* reset ntu and ntc to place SW in sync with hardware */
604 	ring->next_to_clean = 0;
605 	ring->next_to_use = 0;
606 
607 	/* Map interrupt */
608 	if (ring->q_vector) {
609 		txint = ring->q_vector->v_idx + NON_Q_VECTORS(hw);
610 		txint |= FM10K_INT_MAP_TIMER0;
611 	}
612 
613 	fm10k_write_reg(hw, FM10K_TXINT(reg_idx), txint);
614 
615 	/* enable use of FTAG bit in Tx descriptor, register is RO for VF */
616 	fm10k_write_reg(hw, FM10K_PFVTCTL(reg_idx),
617 			FM10K_PFVTCTL_FTAG_DESC_ENABLE);
618 
619 	/* Initialize XPS */
620 	if (!test_and_set_bit(__FM10K_TX_XPS_INIT_DONE, &ring->state) &&
621 	    ring->q_vector)
622 		netif_set_xps_queue(ring->netdev,
623 				    &ring->q_vector->affinity_mask,
624 				    ring->queue_index);
625 
626 	/* enable queue */
627 	fm10k_write_reg(hw, FM10K_TXDCTL(reg_idx), txdctl);
628 }
629 
630 /**
631  * fm10k_enable_tx_ring - Verify Tx ring is enabled after configuration
632  * @interface: board private structure
633  * @ring: structure containing ring specific data
634  *
635  * Verify the Tx descriptor ring is ready for transmit.
636  **/
637 static void fm10k_enable_tx_ring(struct fm10k_intfc *interface,
638 				 struct fm10k_ring *ring)
639 {
640 	struct fm10k_hw *hw = &interface->hw;
641 	int wait_loop = 10;
642 	u32 txdctl;
643 	u8 reg_idx = ring->reg_idx;
644 
645 	/* if we are already enabled just exit */
646 	if (fm10k_read_reg(hw, FM10K_TXDCTL(reg_idx)) & FM10K_TXDCTL_ENABLE)
647 		return;
648 
649 	/* poll to verify queue is enabled */
650 	do {
651 		usleep_range(1000, 2000);
652 		txdctl = fm10k_read_reg(hw, FM10K_TXDCTL(reg_idx));
653 	} while (!(txdctl & FM10K_TXDCTL_ENABLE) && --wait_loop);
654 	if (!wait_loop)
655 		netif_err(interface, drv, interface->netdev,
656 			  "Could not enable Tx Queue %d\n", reg_idx);
657 }
658 
659 /**
660  * fm10k_configure_tx - Configure Transmit Unit after Reset
661  * @interface: board private structure
662  *
663  * Configure the Tx unit of the MAC after a reset.
664  **/
665 static void fm10k_configure_tx(struct fm10k_intfc *interface)
666 {
667 	int i;
668 
669 	/* Setup the HW Tx Head and Tail descriptor pointers */
670 	for (i = 0; i < interface->num_tx_queues; i++)
671 		fm10k_configure_tx_ring(interface, interface->tx_ring[i]);
672 
673 	/* poll here to verify that Tx rings are now enabled */
674 	for (i = 0; i < interface->num_tx_queues; i++)
675 		fm10k_enable_tx_ring(interface, interface->tx_ring[i]);
676 }
677 
678 /**
679  * fm10k_configure_rx_ring - Configure Rx ring after Reset
680  * @interface: board private structure
681  * @ring: structure containing ring specific data
682  *
683  * Configure the Rx descriptor ring after a reset.
684  **/
685 static void fm10k_configure_rx_ring(struct fm10k_intfc *interface,
686 				    struct fm10k_ring *ring)
687 {
688 	u64 rdba = ring->dma;
689 	struct fm10k_hw *hw = &interface->hw;
690 	u32 size = ring->count * sizeof(union fm10k_rx_desc);
691 	u32 rxqctl = FM10K_RXQCTL_ENABLE | FM10K_RXQCTL_PF;
692 	u32 rxdctl = FM10K_RXDCTL_WRITE_BACK_MIN_DELAY;
693 	u32 srrctl = FM10K_SRRCTL_BUFFER_CHAINING_EN;
694 	u32 rxint = FM10K_INT_MAP_DISABLE;
695 	u8 rx_pause = interface->rx_pause;
696 	u8 reg_idx = ring->reg_idx;
697 
698 	/* disable queue to avoid issues while updating state */
699 	fm10k_write_reg(hw, FM10K_RXQCTL(reg_idx), 0);
700 	fm10k_write_flush(hw);
701 
702 	/* possible poll here to verify ring resources have been cleaned */
703 
704 	/* set location and size for descriptor ring */
705 	fm10k_write_reg(hw, FM10K_RDBAL(reg_idx), rdba & DMA_BIT_MASK(32));
706 	fm10k_write_reg(hw, FM10K_RDBAH(reg_idx), rdba >> 32);
707 	fm10k_write_reg(hw, FM10K_RDLEN(reg_idx), size);
708 
709 	/* reset head and tail pointers */
710 	fm10k_write_reg(hw, FM10K_RDH(reg_idx), 0);
711 	fm10k_write_reg(hw, FM10K_RDT(reg_idx), 0);
712 
713 	/* store tail pointer */
714 	ring->tail = &interface->uc_addr[FM10K_RDT(reg_idx)];
715 
716 	/* reset ntu and ntc to place SW in sync with hardware */
717 	ring->next_to_clean = 0;
718 	ring->next_to_use = 0;
719 	ring->next_to_alloc = 0;
720 
721 	/* Configure the Rx buffer size for one buff without split */
722 	srrctl |= FM10K_RX_BUFSZ >> FM10K_SRRCTL_BSIZEPKT_SHIFT;
723 
724 	/* Configure the Rx ring to suppress loopback packets */
725 	srrctl |= FM10K_SRRCTL_LOOPBACK_SUPPRESS;
726 	fm10k_write_reg(hw, FM10K_SRRCTL(reg_idx), srrctl);
727 
728 	/* Enable drop on empty */
729 #ifdef CONFIG_DCB
730 	if (interface->pfc_en)
731 		rx_pause = interface->pfc_en;
732 #endif
733 	if (!(rx_pause & (1 << ring->qos_pc)))
734 		rxdctl |= FM10K_RXDCTL_DROP_ON_EMPTY;
735 
736 	fm10k_write_reg(hw, FM10K_RXDCTL(reg_idx), rxdctl);
737 
738 	/* assign default VLAN to queue */
739 	ring->vid = hw->mac.default_vid;
740 
741 	/* if we have an active VLAN, disable default VLAN ID */
742 	if (test_bit(hw->mac.default_vid, interface->active_vlans))
743 		ring->vid |= FM10K_VLAN_CLEAR;
744 
745 	/* Map interrupt */
746 	if (ring->q_vector) {
747 		rxint = ring->q_vector->v_idx + NON_Q_VECTORS(hw);
748 		rxint |= FM10K_INT_MAP_TIMER1;
749 	}
750 
751 	fm10k_write_reg(hw, FM10K_RXINT(reg_idx), rxint);
752 
753 	/* enable queue */
754 	fm10k_write_reg(hw, FM10K_RXQCTL(reg_idx), rxqctl);
755 
756 	/* place buffers on ring for receive data */
757 	fm10k_alloc_rx_buffers(ring, fm10k_desc_unused(ring));
758 }
759 
760 /**
761  * fm10k_update_rx_drop_en - Configures the drop enable bits for Rx rings
762  * @interface: board private structure
763  *
764  * Configure the drop enable bits for the Rx rings.
765  **/
766 void fm10k_update_rx_drop_en(struct fm10k_intfc *interface)
767 {
768 	struct fm10k_hw *hw = &interface->hw;
769 	u8 rx_pause = interface->rx_pause;
770 	int i;
771 
772 #ifdef CONFIG_DCB
773 	if (interface->pfc_en)
774 		rx_pause = interface->pfc_en;
775 
776 #endif
777 	for (i = 0; i < interface->num_rx_queues; i++) {
778 		struct fm10k_ring *ring = interface->rx_ring[i];
779 		u32 rxdctl = FM10K_RXDCTL_WRITE_BACK_MIN_DELAY;
780 		u8 reg_idx = ring->reg_idx;
781 
782 		if (!(rx_pause & (1 << ring->qos_pc)))
783 			rxdctl |= FM10K_RXDCTL_DROP_ON_EMPTY;
784 
785 		fm10k_write_reg(hw, FM10K_RXDCTL(reg_idx), rxdctl);
786 	}
787 }
788 
789 /**
790  * fm10k_configure_dglort - Configure Receive DGLORT after reset
791  * @interface: board private structure
792  *
793  * Configure the DGLORT description and RSS tables.
794  **/
795 static void fm10k_configure_dglort(struct fm10k_intfc *interface)
796 {
797 	struct fm10k_dglort_cfg dglort = { 0 };
798 	struct fm10k_hw *hw = &interface->hw;
799 	int i;
800 	u32 mrqc;
801 
802 	/* Fill out hash function seeds */
803 	for (i = 0; i < FM10K_RSSRK_SIZE; i++)
804 		fm10k_write_reg(hw, FM10K_RSSRK(0, i), interface->rssrk[i]);
805 
806 	/* Write RETA table to hardware */
807 	for (i = 0; i < FM10K_RETA_SIZE; i++)
808 		fm10k_write_reg(hw, FM10K_RETA(0, i), interface->reta[i]);
809 
810 	/* Generate RSS hash based on packet types, TCP/UDP
811 	 * port numbers and/or IPv4/v6 src and dst addresses
812 	 */
813 	mrqc = FM10K_MRQC_IPV4 |
814 	       FM10K_MRQC_TCP_IPV4 |
815 	       FM10K_MRQC_IPV6 |
816 	       FM10K_MRQC_TCP_IPV6;
817 
818 	if (interface->flags & FM10K_FLAG_RSS_FIELD_IPV4_UDP)
819 		mrqc |= FM10K_MRQC_UDP_IPV4;
820 	if (interface->flags & FM10K_FLAG_RSS_FIELD_IPV6_UDP)
821 		mrqc |= FM10K_MRQC_UDP_IPV6;
822 
823 	fm10k_write_reg(hw, FM10K_MRQC(0), mrqc);
824 
825 	/* configure default DGLORT mapping for RSS/DCB */
826 	dglort.inner_rss = 1;
827 	dglort.rss_l = fls(interface->ring_feature[RING_F_RSS].mask);
828 	dglort.pc_l = fls(interface->ring_feature[RING_F_QOS].mask);
829 	hw->mac.ops.configure_dglort_map(hw, &dglort);
830 
831 	/* assign GLORT per queue for queue mapped testing */
832 	if (interface->glort_count > 64) {
833 		memset(&dglort, 0, sizeof(dglort));
834 		dglort.inner_rss = 1;
835 		dglort.glort = interface->glort + 64;
836 		dglort.idx = fm10k_dglort_pf_queue;
837 		dglort.queue_l = fls(interface->num_rx_queues - 1);
838 		hw->mac.ops.configure_dglort_map(hw, &dglort);
839 	}
840 
841 	/* assign glort value for RSS/DCB specific to this interface */
842 	memset(&dglort, 0, sizeof(dglort));
843 	dglort.inner_rss = 1;
844 	dglort.glort = interface->glort;
845 	dglort.rss_l = fls(interface->ring_feature[RING_F_RSS].mask);
846 	dglort.pc_l = fls(interface->ring_feature[RING_F_QOS].mask);
847 	/* configure DGLORT mapping for RSS/DCB */
848 	dglort.idx = fm10k_dglort_pf_rss;
849 	if (interface->l2_accel)
850 		dglort.shared_l = fls(interface->l2_accel->size);
851 	hw->mac.ops.configure_dglort_map(hw, &dglort);
852 }
853 
854 /**
855  * fm10k_configure_rx - Configure Receive Unit after Reset
856  * @interface: board private structure
857  *
858  * Configure the Rx unit of the MAC after a reset.
859  **/
860 static void fm10k_configure_rx(struct fm10k_intfc *interface)
861 {
862 	int i;
863 
864 	/* Configure SWPRI to PC map */
865 	fm10k_configure_swpri_map(interface);
866 
867 	/* Configure RSS and DGLORT map */
868 	fm10k_configure_dglort(interface);
869 
870 	/* Setup the HW Rx Head and Tail descriptor pointers */
871 	for (i = 0; i < interface->num_rx_queues; i++)
872 		fm10k_configure_rx_ring(interface, interface->rx_ring[i]);
873 
874 	/* possible poll here to verify that Rx rings are now enabled */
875 }
876 
877 static void fm10k_napi_enable_all(struct fm10k_intfc *interface)
878 {
879 	struct fm10k_q_vector *q_vector;
880 	int q_idx;
881 
882 	for (q_idx = 0; q_idx < interface->num_q_vectors; q_idx++) {
883 		q_vector = interface->q_vector[q_idx];
884 		napi_enable(&q_vector->napi);
885 	}
886 }
887 
888 static irqreturn_t fm10k_msix_clean_rings(int __always_unused irq, void *data)
889 {
890 	struct fm10k_q_vector *q_vector = data;
891 
892 	if (q_vector->rx.count || q_vector->tx.count)
893 		napi_schedule_irqoff(&q_vector->napi);
894 
895 	return IRQ_HANDLED;
896 }
897 
898 static irqreturn_t fm10k_msix_mbx_vf(int __always_unused irq, void *data)
899 {
900 	struct fm10k_intfc *interface = data;
901 	struct fm10k_hw *hw = &interface->hw;
902 	struct fm10k_mbx_info *mbx = &hw->mbx;
903 
904 	/* re-enable mailbox interrupt and indicate 20us delay */
905 	fm10k_write_reg(hw, FM10K_VFITR(FM10K_MBX_VECTOR),
906 			FM10K_ITR_ENABLE | (FM10K_MBX_INT_DELAY >>
907 					    hw->mac.itr_scale));
908 
909 	/* service upstream mailbox */
910 	if (fm10k_mbx_trylock(interface)) {
911 		mbx->ops.process(hw, mbx);
912 		fm10k_mbx_unlock(interface);
913 	}
914 
915 	hw->mac.get_host_state = true;
916 	fm10k_service_event_schedule(interface);
917 
918 	return IRQ_HANDLED;
919 }
920 
921 #ifdef CONFIG_NET_POLL_CONTROLLER
922 /**
923  *  fm10k_netpoll - A Polling 'interrupt' handler
924  *  @netdev: network interface device structure
925  *
926  *  This is used by netconsole to send skbs without having to re-enable
927  *  interrupts. It's not called while the normal interrupt routine is executing.
928  **/
929 void fm10k_netpoll(struct net_device *netdev)
930 {
931 	struct fm10k_intfc *interface = netdev_priv(netdev);
932 	int i;
933 
934 	/* if interface is down do nothing */
935 	if (test_bit(__FM10K_DOWN, &interface->state))
936 		return;
937 
938 	for (i = 0; i < interface->num_q_vectors; i++)
939 		fm10k_msix_clean_rings(0, interface->q_vector[i]);
940 }
941 
942 #endif
943 #define FM10K_ERR_MSG(type) case (type): error = #type; break
944 static void fm10k_handle_fault(struct fm10k_intfc *interface, int type,
945 			       struct fm10k_fault *fault)
946 {
947 	struct pci_dev *pdev = interface->pdev;
948 	struct fm10k_hw *hw = &interface->hw;
949 	struct fm10k_iov_data *iov_data = interface->iov_data;
950 	char *error;
951 
952 	switch (type) {
953 	case FM10K_PCA_FAULT:
954 		switch (fault->type) {
955 		default:
956 			error = "Unknown PCA error";
957 			break;
958 		FM10K_ERR_MSG(PCA_NO_FAULT);
959 		FM10K_ERR_MSG(PCA_UNMAPPED_ADDR);
960 		FM10K_ERR_MSG(PCA_BAD_QACCESS_PF);
961 		FM10K_ERR_MSG(PCA_BAD_QACCESS_VF);
962 		FM10K_ERR_MSG(PCA_MALICIOUS_REQ);
963 		FM10K_ERR_MSG(PCA_POISONED_TLP);
964 		FM10K_ERR_MSG(PCA_TLP_ABORT);
965 		}
966 		break;
967 	case FM10K_THI_FAULT:
968 		switch (fault->type) {
969 		default:
970 			error = "Unknown THI error";
971 			break;
972 		FM10K_ERR_MSG(THI_NO_FAULT);
973 		FM10K_ERR_MSG(THI_MAL_DIS_Q_FAULT);
974 		}
975 		break;
976 	case FM10K_FUM_FAULT:
977 		switch (fault->type) {
978 		default:
979 			error = "Unknown FUM error";
980 			break;
981 		FM10K_ERR_MSG(FUM_NO_FAULT);
982 		FM10K_ERR_MSG(FUM_UNMAPPED_ADDR);
983 		FM10K_ERR_MSG(FUM_BAD_VF_QACCESS);
984 		FM10K_ERR_MSG(FUM_ADD_DECODE_ERR);
985 		FM10K_ERR_MSG(FUM_RO_ERROR);
986 		FM10K_ERR_MSG(FUM_QPRC_CRC_ERROR);
987 		FM10K_ERR_MSG(FUM_CSR_TIMEOUT);
988 		FM10K_ERR_MSG(FUM_INVALID_TYPE);
989 		FM10K_ERR_MSG(FUM_INVALID_LENGTH);
990 		FM10K_ERR_MSG(FUM_INVALID_BE);
991 		FM10K_ERR_MSG(FUM_INVALID_ALIGN);
992 		}
993 		break;
994 	default:
995 		error = "Undocumented fault";
996 		break;
997 	}
998 
999 	dev_warn(&pdev->dev,
1000 		 "%s Address: 0x%llx SpecInfo: 0x%x Func: %02x.%0x\n",
1001 		 error, fault->address, fault->specinfo,
1002 		 PCI_SLOT(fault->func), PCI_FUNC(fault->func));
1003 
1004 	/* For VF faults, clear out the respective LPORT, reset the queue
1005 	 * resources, and then reconnect to the mailbox. This allows the
1006 	 * VF in question to resume behavior. For transient faults that are
1007 	 * the result of non-malicious behavior this will log the fault and
1008 	 * allow the VF to resume functionality. Obviously for malicious VFs
1009 	 * they will be able to attempt malicious behavior again. In this
1010 	 * case, the system administrator will need to step in and manually
1011 	 * remove or disable the VF in question.
1012 	 */
1013 	if (fault->func && iov_data) {
1014 		int vf = fault->func - 1;
1015 		struct fm10k_vf_info *vf_info = &iov_data->vf_info[vf];
1016 
1017 		hw->iov.ops.reset_lport(hw, vf_info);
1018 		hw->iov.ops.reset_resources(hw, vf_info);
1019 
1020 		/* reset_lport disables the VF, so re-enable it */
1021 		hw->iov.ops.set_lport(hw, vf_info, vf,
1022 				      FM10K_VF_FLAG_MULTI_CAPABLE);
1023 
1024 		/* reset_resources will disconnect from the mbx  */
1025 		vf_info->mbx.ops.connect(hw, &vf_info->mbx);
1026 	}
1027 }
1028 
1029 static void fm10k_report_fault(struct fm10k_intfc *interface, u32 eicr)
1030 {
1031 	struct fm10k_hw *hw = &interface->hw;
1032 	struct fm10k_fault fault = { 0 };
1033 	int type, err;
1034 
1035 	for (eicr &= FM10K_EICR_FAULT_MASK, type = FM10K_PCA_FAULT;
1036 	     eicr;
1037 	     eicr >>= 1, type += FM10K_FAULT_SIZE) {
1038 		/* only check if there is an error reported */
1039 		if (!(eicr & 0x1))
1040 			continue;
1041 
1042 		/* retrieve fault info */
1043 		err = hw->mac.ops.get_fault(hw, type, &fault);
1044 		if (err) {
1045 			dev_err(&interface->pdev->dev,
1046 				"error reading fault\n");
1047 			continue;
1048 		}
1049 
1050 		fm10k_handle_fault(interface, type, &fault);
1051 	}
1052 }
1053 
1054 static void fm10k_reset_drop_on_empty(struct fm10k_intfc *interface, u32 eicr)
1055 {
1056 	struct fm10k_hw *hw = &interface->hw;
1057 	const u32 rxdctl = FM10K_RXDCTL_WRITE_BACK_MIN_DELAY;
1058 	u32 maxholdq;
1059 	int q;
1060 
1061 	if (!(eicr & FM10K_EICR_MAXHOLDTIME))
1062 		return;
1063 
1064 	maxholdq = fm10k_read_reg(hw, FM10K_MAXHOLDQ(7));
1065 	if (maxholdq)
1066 		fm10k_write_reg(hw, FM10K_MAXHOLDQ(7), maxholdq);
1067 	for (q = 255;;) {
1068 		if (maxholdq & (1 << 31)) {
1069 			if (q < FM10K_MAX_QUEUES_PF) {
1070 				interface->rx_overrun_pf++;
1071 				fm10k_write_reg(hw, FM10K_RXDCTL(q), rxdctl);
1072 			} else {
1073 				interface->rx_overrun_vf++;
1074 			}
1075 		}
1076 
1077 		maxholdq *= 2;
1078 		if (!maxholdq)
1079 			q &= ~(32 - 1);
1080 
1081 		if (!q)
1082 			break;
1083 
1084 		if (q-- % 32)
1085 			continue;
1086 
1087 		maxholdq = fm10k_read_reg(hw, FM10K_MAXHOLDQ(q / 32));
1088 		if (maxholdq)
1089 			fm10k_write_reg(hw, FM10K_MAXHOLDQ(q / 32), maxholdq);
1090 	}
1091 }
1092 
1093 static irqreturn_t fm10k_msix_mbx_pf(int __always_unused irq, void *data)
1094 {
1095 	struct fm10k_intfc *interface = data;
1096 	struct fm10k_hw *hw = &interface->hw;
1097 	struct fm10k_mbx_info *mbx = &hw->mbx;
1098 	u32 eicr;
1099 
1100 	/* unmask any set bits related to this interrupt */
1101 	eicr = fm10k_read_reg(hw, FM10K_EICR);
1102 	fm10k_write_reg(hw, FM10K_EICR, eicr & (FM10K_EICR_MAILBOX |
1103 						FM10K_EICR_SWITCHREADY |
1104 						FM10K_EICR_SWITCHNOTREADY));
1105 
1106 	/* report any faults found to the message log */
1107 	fm10k_report_fault(interface, eicr);
1108 
1109 	/* reset any queues disabled due to receiver overrun */
1110 	fm10k_reset_drop_on_empty(interface, eicr);
1111 
1112 	/* service mailboxes */
1113 	if (fm10k_mbx_trylock(interface)) {
1114 		mbx->ops.process(hw, mbx);
1115 		/* handle VFLRE events */
1116 		fm10k_iov_event(interface);
1117 		fm10k_mbx_unlock(interface);
1118 	}
1119 
1120 	/* if switch toggled state we should reset GLORTs */
1121 	if (eicr & FM10K_EICR_SWITCHNOTREADY) {
1122 		/* force link down for at least 4 seconds */
1123 		interface->link_down_event = jiffies + (4 * HZ);
1124 		set_bit(__FM10K_LINK_DOWN, &interface->state);
1125 
1126 		/* reset dglort_map back to no config */
1127 		hw->mac.dglort_map = FM10K_DGLORTMAP_NONE;
1128 	}
1129 
1130 	/* we should validate host state after interrupt event */
1131 	hw->mac.get_host_state = true;
1132 
1133 	/* validate host state, and handle VF mailboxes in the service task */
1134 	fm10k_service_event_schedule(interface);
1135 
1136 	/* re-enable mailbox interrupt and indicate 20us delay */
1137 	fm10k_write_reg(hw, FM10K_ITR(FM10K_MBX_VECTOR),
1138 			FM10K_ITR_ENABLE | (FM10K_MBX_INT_DELAY >>
1139 					    hw->mac.itr_scale));
1140 
1141 	return IRQ_HANDLED;
1142 }
1143 
1144 void fm10k_mbx_free_irq(struct fm10k_intfc *interface)
1145 {
1146 	struct msix_entry *entry = &interface->msix_entries[FM10K_MBX_VECTOR];
1147 	struct fm10k_hw *hw = &interface->hw;
1148 	int itr_reg;
1149 
1150 	/* no mailbox IRQ to free if MSI-X is not enabled */
1151 	if (!interface->msix_entries)
1152 		return;
1153 
1154 	/* disconnect the mailbox */
1155 	hw->mbx.ops.disconnect(hw, &hw->mbx);
1156 
1157 	/* disable Mailbox cause */
1158 	if (hw->mac.type == fm10k_mac_pf) {
1159 		fm10k_write_reg(hw, FM10K_EIMR,
1160 				FM10K_EIMR_DISABLE(PCA_FAULT) |
1161 				FM10K_EIMR_DISABLE(FUM_FAULT) |
1162 				FM10K_EIMR_DISABLE(MAILBOX) |
1163 				FM10K_EIMR_DISABLE(SWITCHREADY) |
1164 				FM10K_EIMR_DISABLE(SWITCHNOTREADY) |
1165 				FM10K_EIMR_DISABLE(SRAMERROR) |
1166 				FM10K_EIMR_DISABLE(VFLR) |
1167 				FM10K_EIMR_DISABLE(MAXHOLDTIME));
1168 		itr_reg = FM10K_ITR(FM10K_MBX_VECTOR);
1169 	} else {
1170 		itr_reg = FM10K_VFITR(FM10K_MBX_VECTOR);
1171 	}
1172 
1173 	fm10k_write_reg(hw, itr_reg, FM10K_ITR_MASK_SET);
1174 
1175 	free_irq(entry->vector, interface);
1176 }
1177 
1178 static s32 fm10k_mbx_mac_addr(struct fm10k_hw *hw, u32 **results,
1179 			      struct fm10k_mbx_info *mbx)
1180 {
1181 	bool vlan_override = hw->mac.vlan_override;
1182 	u16 default_vid = hw->mac.default_vid;
1183 	struct fm10k_intfc *interface;
1184 	s32 err;
1185 
1186 	err = fm10k_msg_mac_vlan_vf(hw, results, mbx);
1187 	if (err)
1188 		return err;
1189 
1190 	interface = container_of(hw, struct fm10k_intfc, hw);
1191 
1192 	/* MAC was changed so we need reset */
1193 	if (is_valid_ether_addr(hw->mac.perm_addr) &&
1194 	    !ether_addr_equal(hw->mac.perm_addr, hw->mac.addr))
1195 		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
1196 
1197 	/* VLAN override was changed, or default VLAN changed */
1198 	if ((vlan_override != hw->mac.vlan_override) ||
1199 	    (default_vid != hw->mac.default_vid))
1200 		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
1201 
1202 	return 0;
1203 }
1204 
1205 static s32 fm10k_1588_msg_vf(struct fm10k_hw *hw, u32 **results,
1206 			     struct fm10k_mbx_info __always_unused *mbx)
1207 {
1208 	struct fm10k_intfc *interface;
1209 	u64 timestamp;
1210 	s32 err;
1211 
1212 	err = fm10k_tlv_attr_get_u64(results[FM10K_1588_MSG_TIMESTAMP],
1213 				     &timestamp);
1214 	if (err)
1215 		return err;
1216 
1217 	interface = container_of(hw, struct fm10k_intfc, hw);
1218 
1219 	fm10k_ts_tx_hwtstamp(interface, 0, timestamp);
1220 
1221 	return 0;
1222 }
1223 
1224 /* generic error handler for mailbox issues */
1225 static s32 fm10k_mbx_error(struct fm10k_hw *hw, u32 **results,
1226 			   struct fm10k_mbx_info __always_unused *mbx)
1227 {
1228 	struct fm10k_intfc *interface;
1229 	struct pci_dev *pdev;
1230 
1231 	interface = container_of(hw, struct fm10k_intfc, hw);
1232 	pdev = interface->pdev;
1233 
1234 	dev_err(&pdev->dev, "Unknown message ID %u\n",
1235 		**results & FM10K_TLV_ID_MASK);
1236 
1237 	return 0;
1238 }
1239 
1240 static const struct fm10k_msg_data vf_mbx_data[] = {
1241 	FM10K_TLV_MSG_TEST_HANDLER(fm10k_tlv_msg_test),
1242 	FM10K_VF_MSG_MAC_VLAN_HANDLER(fm10k_mbx_mac_addr),
1243 	FM10K_VF_MSG_LPORT_STATE_HANDLER(fm10k_msg_lport_state_vf),
1244 	FM10K_VF_MSG_1588_HANDLER(fm10k_1588_msg_vf),
1245 	FM10K_TLV_MSG_ERROR_HANDLER(fm10k_mbx_error),
1246 };
1247 
1248 static int fm10k_mbx_request_irq_vf(struct fm10k_intfc *interface)
1249 {
1250 	struct msix_entry *entry = &interface->msix_entries[FM10K_MBX_VECTOR];
1251 	struct net_device *dev = interface->netdev;
1252 	struct fm10k_hw *hw = &interface->hw;
1253 	int err;
1254 
1255 	/* Use timer0 for interrupt moderation on the mailbox */
1256 	u32 itr = FM10K_INT_MAP_TIMER0 | entry->entry;
1257 
1258 	/* register mailbox handlers */
1259 	err = hw->mbx.ops.register_handlers(&hw->mbx, vf_mbx_data);
1260 	if (err)
1261 		return err;
1262 
1263 	/* request the IRQ */
1264 	err = request_irq(entry->vector, fm10k_msix_mbx_vf, 0,
1265 			  dev->name, interface);
1266 	if (err) {
1267 		netif_err(interface, probe, dev,
1268 			  "request_irq for msix_mbx failed: %d\n", err);
1269 		return err;
1270 	}
1271 
1272 	/* map all of the interrupt sources */
1273 	fm10k_write_reg(hw, FM10K_VFINT_MAP, itr);
1274 
1275 	/* enable interrupt */
1276 	fm10k_write_reg(hw, FM10K_VFITR(entry->entry), FM10K_ITR_ENABLE);
1277 
1278 	return 0;
1279 }
1280 
1281 static s32 fm10k_lport_map(struct fm10k_hw *hw, u32 **results,
1282 			   struct fm10k_mbx_info *mbx)
1283 {
1284 	struct fm10k_intfc *interface;
1285 	u32 dglort_map = hw->mac.dglort_map;
1286 	s32 err;
1287 
1288 	err = fm10k_msg_lport_map_pf(hw, results, mbx);
1289 	if (err)
1290 		return err;
1291 
1292 	interface = container_of(hw, struct fm10k_intfc, hw);
1293 
1294 	/* we need to reset if port count was just updated */
1295 	if (dglort_map != hw->mac.dglort_map)
1296 		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
1297 
1298 	return 0;
1299 }
1300 
1301 static s32 fm10k_update_pvid(struct fm10k_hw *hw, u32 **results,
1302 			     struct fm10k_mbx_info __always_unused *mbx)
1303 {
1304 	struct fm10k_intfc *interface;
1305 	u16 glort, pvid;
1306 	u32 pvid_update;
1307 	s32 err;
1308 
1309 	err = fm10k_tlv_attr_get_u32(results[FM10K_PF_ATTR_ID_UPDATE_PVID],
1310 				     &pvid_update);
1311 	if (err)
1312 		return err;
1313 
1314 	/* extract values from the pvid update */
1315 	glort = FM10K_MSG_HDR_FIELD_GET(pvid_update, UPDATE_PVID_GLORT);
1316 	pvid = FM10K_MSG_HDR_FIELD_GET(pvid_update, UPDATE_PVID_PVID);
1317 
1318 	/* if glort is not valid return error */
1319 	if (!fm10k_glort_valid_pf(hw, glort))
1320 		return FM10K_ERR_PARAM;
1321 
1322 	/* verify VLAN ID is valid */
1323 	if (pvid >= FM10K_VLAN_TABLE_VID_MAX)
1324 		return FM10K_ERR_PARAM;
1325 
1326 	interface = container_of(hw, struct fm10k_intfc, hw);
1327 
1328 	/* check to see if this belongs to one of the VFs */
1329 	err = fm10k_iov_update_pvid(interface, glort, pvid);
1330 	if (!err)
1331 		return 0;
1332 
1333 	/* we need to reset if default VLAN was just updated */
1334 	if (pvid != hw->mac.default_vid)
1335 		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
1336 
1337 	hw->mac.default_vid = pvid;
1338 
1339 	return 0;
1340 }
1341 
1342 static s32 fm10k_1588_msg_pf(struct fm10k_hw *hw, u32 **results,
1343 			     struct fm10k_mbx_info __always_unused *mbx)
1344 {
1345 	struct fm10k_swapi_1588_timestamp timestamp;
1346 	struct fm10k_iov_data *iov_data;
1347 	struct fm10k_intfc *interface;
1348 	u16 sglort, vf_idx;
1349 	s32 err;
1350 
1351 	err = fm10k_tlv_attr_get_le_struct(
1352 				results[FM10K_PF_ATTR_ID_1588_TIMESTAMP],
1353 				&timestamp, sizeof(timestamp));
1354 	if (err)
1355 		return err;
1356 
1357 	interface = container_of(hw, struct fm10k_intfc, hw);
1358 
1359 	if (timestamp.dglort) {
1360 		fm10k_ts_tx_hwtstamp(interface, timestamp.dglort,
1361 				     le64_to_cpu(timestamp.egress));
1362 		return 0;
1363 	}
1364 
1365 	/* either dglort or sglort must be set */
1366 	if (!timestamp.sglort)
1367 		return FM10K_ERR_PARAM;
1368 
1369 	/* verify GLORT is at least one of the ones we own */
1370 	sglort = le16_to_cpu(timestamp.sglort);
1371 	if (!fm10k_glort_valid_pf(hw, sglort))
1372 		return FM10K_ERR_PARAM;
1373 
1374 	if (sglort == interface->glort) {
1375 		fm10k_ts_tx_hwtstamp(interface, 0,
1376 				     le64_to_cpu(timestamp.ingress));
1377 		return 0;
1378 	}
1379 
1380 	/* if there is no iov_data then there is no mailboxes to process */
1381 	if (!ACCESS_ONCE(interface->iov_data))
1382 		return FM10K_ERR_PARAM;
1383 
1384 	rcu_read_lock();
1385 
1386 	/* notify VF if this timestamp belongs to it */
1387 	iov_data = interface->iov_data;
1388 	vf_idx = (hw->mac.dglort_map & FM10K_DGLORTMAP_NONE) - sglort;
1389 
1390 	if (!iov_data || vf_idx >= iov_data->num_vfs) {
1391 		err = FM10K_ERR_PARAM;
1392 		goto err_unlock;
1393 	}
1394 
1395 	err = hw->iov.ops.report_timestamp(hw, &iov_data->vf_info[vf_idx],
1396 					   le64_to_cpu(timestamp.ingress));
1397 
1398 err_unlock:
1399 	rcu_read_unlock();
1400 
1401 	return err;
1402 }
1403 
1404 static const struct fm10k_msg_data pf_mbx_data[] = {
1405 	FM10K_PF_MSG_ERR_HANDLER(XCAST_MODES, fm10k_msg_err_pf),
1406 	FM10K_PF_MSG_ERR_HANDLER(UPDATE_MAC_FWD_RULE, fm10k_msg_err_pf),
1407 	FM10K_PF_MSG_LPORT_MAP_HANDLER(fm10k_lport_map),
1408 	FM10K_PF_MSG_ERR_HANDLER(LPORT_CREATE, fm10k_msg_err_pf),
1409 	FM10K_PF_MSG_ERR_HANDLER(LPORT_DELETE, fm10k_msg_err_pf),
1410 	FM10K_PF_MSG_UPDATE_PVID_HANDLER(fm10k_update_pvid),
1411 	FM10K_PF_MSG_1588_TIMESTAMP_HANDLER(fm10k_1588_msg_pf),
1412 	FM10K_TLV_MSG_ERROR_HANDLER(fm10k_mbx_error),
1413 };
1414 
1415 static int fm10k_mbx_request_irq_pf(struct fm10k_intfc *interface)
1416 {
1417 	struct msix_entry *entry = &interface->msix_entries[FM10K_MBX_VECTOR];
1418 	struct net_device *dev = interface->netdev;
1419 	struct fm10k_hw *hw = &interface->hw;
1420 	int err;
1421 
1422 	/* Use timer0 for interrupt moderation on the mailbox */
1423 	u32 mbx_itr = FM10K_INT_MAP_TIMER0 | entry->entry;
1424 	u32 other_itr = FM10K_INT_MAP_IMMEDIATE | entry->entry;
1425 
1426 	/* register mailbox handlers */
1427 	err = hw->mbx.ops.register_handlers(&hw->mbx, pf_mbx_data);
1428 	if (err)
1429 		return err;
1430 
1431 	/* request the IRQ */
1432 	err = request_irq(entry->vector, fm10k_msix_mbx_pf, 0,
1433 			  dev->name, interface);
1434 	if (err) {
1435 		netif_err(interface, probe, dev,
1436 			  "request_irq for msix_mbx failed: %d\n", err);
1437 		return err;
1438 	}
1439 
1440 	/* Enable interrupts w/ no moderation for "other" interrupts */
1441 	fm10k_write_reg(hw, FM10K_INT_MAP(fm10k_int_pcie_fault), other_itr);
1442 	fm10k_write_reg(hw, FM10K_INT_MAP(fm10k_int_switch_up_down), other_itr);
1443 	fm10k_write_reg(hw, FM10K_INT_MAP(fm10k_int_sram), other_itr);
1444 	fm10k_write_reg(hw, FM10K_INT_MAP(fm10k_int_max_hold_time), other_itr);
1445 	fm10k_write_reg(hw, FM10K_INT_MAP(fm10k_int_vflr), other_itr);
1446 
1447 	/* Enable interrupts w/ moderation for mailbox */
1448 	fm10k_write_reg(hw, FM10K_INT_MAP(fm10k_int_mailbox), mbx_itr);
1449 
1450 	/* Enable individual interrupt causes */
1451 	fm10k_write_reg(hw, FM10K_EIMR, FM10K_EIMR_ENABLE(PCA_FAULT) |
1452 					FM10K_EIMR_ENABLE(FUM_FAULT) |
1453 					FM10K_EIMR_ENABLE(MAILBOX) |
1454 					FM10K_EIMR_ENABLE(SWITCHREADY) |
1455 					FM10K_EIMR_ENABLE(SWITCHNOTREADY) |
1456 					FM10K_EIMR_ENABLE(SRAMERROR) |
1457 					FM10K_EIMR_ENABLE(VFLR) |
1458 					FM10K_EIMR_ENABLE(MAXHOLDTIME));
1459 
1460 	/* enable interrupt */
1461 	fm10k_write_reg(hw, FM10K_ITR(entry->entry), FM10K_ITR_ENABLE);
1462 
1463 	return 0;
1464 }
1465 
1466 int fm10k_mbx_request_irq(struct fm10k_intfc *interface)
1467 {
1468 	struct fm10k_hw *hw = &interface->hw;
1469 	int err;
1470 
1471 	/* enable Mailbox cause */
1472 	if (hw->mac.type == fm10k_mac_pf)
1473 		err = fm10k_mbx_request_irq_pf(interface);
1474 	else
1475 		err = fm10k_mbx_request_irq_vf(interface);
1476 	if (err)
1477 		return err;
1478 
1479 	/* connect mailbox */
1480 	err = hw->mbx.ops.connect(hw, &hw->mbx);
1481 
1482 	/* if the mailbox failed to connect, then free IRQ */
1483 	if (err)
1484 		fm10k_mbx_free_irq(interface);
1485 
1486 	return err;
1487 }
1488 
1489 /**
1490  * fm10k_qv_free_irq - release interrupts associated with queue vectors
1491  * @interface: board private structure
1492  *
1493  * Release all interrupts associated with this interface
1494  **/
1495 void fm10k_qv_free_irq(struct fm10k_intfc *interface)
1496 {
1497 	int vector = interface->num_q_vectors;
1498 	struct fm10k_hw *hw = &interface->hw;
1499 	struct msix_entry *entry;
1500 
1501 	entry = &interface->msix_entries[NON_Q_VECTORS(hw) + vector];
1502 
1503 	while (vector) {
1504 		struct fm10k_q_vector *q_vector;
1505 
1506 		vector--;
1507 		entry--;
1508 		q_vector = interface->q_vector[vector];
1509 
1510 		if (!q_vector->tx.count && !q_vector->rx.count)
1511 			continue;
1512 
1513 		/* clear the affinity_mask in the IRQ descriptor */
1514 		irq_set_affinity_hint(entry->vector, NULL);
1515 
1516 		/* disable interrupts */
1517 		writel(FM10K_ITR_MASK_SET, q_vector->itr);
1518 
1519 		free_irq(entry->vector, q_vector);
1520 	}
1521 }
1522 
1523 /**
1524  * fm10k_qv_request_irq - initialize interrupts for queue vectors
1525  * @interface: board private structure
1526  *
1527  * Attempts to configure interrupts using the best available
1528  * capabilities of the hardware and kernel.
1529  **/
1530 int fm10k_qv_request_irq(struct fm10k_intfc *interface)
1531 {
1532 	struct net_device *dev = interface->netdev;
1533 	struct fm10k_hw *hw = &interface->hw;
1534 	struct msix_entry *entry;
1535 	int ri = 0, ti = 0;
1536 	int vector, err;
1537 
1538 	entry = &interface->msix_entries[NON_Q_VECTORS(hw)];
1539 
1540 	for (vector = 0; vector < interface->num_q_vectors; vector++) {
1541 		struct fm10k_q_vector *q_vector = interface->q_vector[vector];
1542 
1543 		/* name the vector */
1544 		if (q_vector->tx.count && q_vector->rx.count) {
1545 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1546 				 "%s-TxRx-%d", dev->name, ri++);
1547 			ti++;
1548 		} else if (q_vector->rx.count) {
1549 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1550 				 "%s-rx-%d", dev->name, ri++);
1551 		} else if (q_vector->tx.count) {
1552 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1553 				 "%s-tx-%d", dev->name, ti++);
1554 		} else {
1555 			/* skip this unused q_vector */
1556 			continue;
1557 		}
1558 
1559 		/* Assign ITR register to q_vector */
1560 		q_vector->itr = (hw->mac.type == fm10k_mac_pf) ?
1561 				&interface->uc_addr[FM10K_ITR(entry->entry)] :
1562 				&interface->uc_addr[FM10K_VFITR(entry->entry)];
1563 
1564 		/* request the IRQ */
1565 		err = request_irq(entry->vector, &fm10k_msix_clean_rings, 0,
1566 				  q_vector->name, q_vector);
1567 		if (err) {
1568 			netif_err(interface, probe, dev,
1569 				  "request_irq failed for MSIX interrupt Error: %d\n",
1570 				  err);
1571 			goto err_out;
1572 		}
1573 
1574 		/* assign the mask for this irq */
1575 		irq_set_affinity_hint(entry->vector, &q_vector->affinity_mask);
1576 
1577 		/* Enable q_vector */
1578 		writel(FM10K_ITR_ENABLE, q_vector->itr);
1579 
1580 		entry++;
1581 	}
1582 
1583 	return 0;
1584 
1585 err_out:
1586 	/* wind through the ring freeing all entries and vectors */
1587 	while (vector) {
1588 		struct fm10k_q_vector *q_vector;
1589 
1590 		entry--;
1591 		vector--;
1592 		q_vector = interface->q_vector[vector];
1593 
1594 		if (!q_vector->tx.count && !q_vector->rx.count)
1595 			continue;
1596 
1597 		/* clear the affinity_mask in the IRQ descriptor */
1598 		irq_set_affinity_hint(entry->vector, NULL);
1599 
1600 		/* disable interrupts */
1601 		writel(FM10K_ITR_MASK_SET, q_vector->itr);
1602 
1603 		free_irq(entry->vector, q_vector);
1604 	}
1605 
1606 	return err;
1607 }
1608 
1609 void fm10k_up(struct fm10k_intfc *interface)
1610 {
1611 	struct fm10k_hw *hw = &interface->hw;
1612 
1613 	/* Enable Tx/Rx DMA */
1614 	hw->mac.ops.start_hw(hw);
1615 
1616 	/* configure Tx descriptor rings */
1617 	fm10k_configure_tx(interface);
1618 
1619 	/* configure Rx descriptor rings */
1620 	fm10k_configure_rx(interface);
1621 
1622 	/* configure interrupts */
1623 	hw->mac.ops.update_int_moderator(hw);
1624 
1625 	/* clear down bit to indicate we are ready to go */
1626 	clear_bit(__FM10K_DOWN, &interface->state);
1627 
1628 	/* enable polling cleanups */
1629 	fm10k_napi_enable_all(interface);
1630 
1631 	/* re-establish Rx filters */
1632 	fm10k_restore_rx_state(interface);
1633 
1634 	/* enable transmits */
1635 	netif_tx_start_all_queues(interface->netdev);
1636 
1637 	/* kick off the service timer now */
1638 	hw->mac.get_host_state = true;
1639 	mod_timer(&interface->service_timer, jiffies);
1640 }
1641 
1642 static void fm10k_napi_disable_all(struct fm10k_intfc *interface)
1643 {
1644 	struct fm10k_q_vector *q_vector;
1645 	int q_idx;
1646 
1647 	for (q_idx = 0; q_idx < interface->num_q_vectors; q_idx++) {
1648 		q_vector = interface->q_vector[q_idx];
1649 		napi_disable(&q_vector->napi);
1650 	}
1651 }
1652 
1653 void fm10k_down(struct fm10k_intfc *interface)
1654 {
1655 	struct net_device *netdev = interface->netdev;
1656 	struct fm10k_hw *hw = &interface->hw;
1657 
1658 	/* signal that we are down to the interrupt handler and service task */
1659 	set_bit(__FM10K_DOWN, &interface->state);
1660 
1661 	/* call carrier off first to avoid false dev_watchdog timeouts */
1662 	netif_carrier_off(netdev);
1663 
1664 	/* disable transmits */
1665 	netif_tx_stop_all_queues(netdev);
1666 	netif_tx_disable(netdev);
1667 
1668 	/* reset Rx filters */
1669 	fm10k_reset_rx_state(interface);
1670 
1671 	/* allow 10ms for device to quiesce */
1672 	usleep_range(10000, 20000);
1673 
1674 	/* disable polling routines */
1675 	fm10k_napi_disable_all(interface);
1676 
1677 	/* capture stats one last time before stopping interface */
1678 	fm10k_update_stats(interface);
1679 
1680 	/* Disable DMA engine for Tx/Rx */
1681 	hw->mac.ops.stop_hw(hw);
1682 
1683 	/* free any buffers still on the rings */
1684 	fm10k_clean_all_tx_rings(interface);
1685 	fm10k_clean_all_rx_rings(interface);
1686 }
1687 
1688 /**
1689  * fm10k_sw_init - Initialize general software structures
1690  * @interface: host interface private structure to initialize
1691  *
1692  * fm10k_sw_init initializes the interface private data structure.
1693  * Fields are initialized based on PCI device information and
1694  * OS network device settings (MTU size).
1695  **/
1696 static int fm10k_sw_init(struct fm10k_intfc *interface,
1697 			 const struct pci_device_id *ent)
1698 {
1699 	const struct fm10k_info *fi = fm10k_info_tbl[ent->driver_data];
1700 	struct fm10k_hw *hw = &interface->hw;
1701 	struct pci_dev *pdev = interface->pdev;
1702 	struct net_device *netdev = interface->netdev;
1703 	u32 rss_key[FM10K_RSSRK_SIZE];
1704 	unsigned int rss;
1705 	int err;
1706 
1707 	/* initialize back pointer */
1708 	hw->back = interface;
1709 	hw->hw_addr = interface->uc_addr;
1710 
1711 	/* PCI config space info */
1712 	hw->vendor_id = pdev->vendor;
1713 	hw->device_id = pdev->device;
1714 	hw->revision_id = pdev->revision;
1715 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
1716 	hw->subsystem_device_id = pdev->subsystem_device;
1717 
1718 	/* Setup hw api */
1719 	memcpy(&hw->mac.ops, fi->mac_ops, sizeof(hw->mac.ops));
1720 	hw->mac.type = fi->mac;
1721 
1722 	/* Setup IOV handlers */
1723 	if (fi->iov_ops)
1724 		memcpy(&hw->iov.ops, fi->iov_ops, sizeof(hw->iov.ops));
1725 
1726 	/* Set common capability flags and settings */
1727 	rss = min_t(int, FM10K_MAX_RSS_INDICES, num_online_cpus());
1728 	interface->ring_feature[RING_F_RSS].limit = rss;
1729 	fi->get_invariants(hw);
1730 
1731 	/* pick up the PCIe bus settings for reporting later */
1732 	if (hw->mac.ops.get_bus_info)
1733 		hw->mac.ops.get_bus_info(hw);
1734 
1735 	/* limit the usable DMA range */
1736 	if (hw->mac.ops.set_dma_mask)
1737 		hw->mac.ops.set_dma_mask(hw, dma_get_mask(&pdev->dev));
1738 
1739 	/* update netdev with DMA restrictions */
1740 	if (dma_get_mask(&pdev->dev) > DMA_BIT_MASK(32)) {
1741 		netdev->features |= NETIF_F_HIGHDMA;
1742 		netdev->vlan_features |= NETIF_F_HIGHDMA;
1743 	}
1744 
1745 	/* delay any future reset requests */
1746 	interface->last_reset = jiffies + (10 * HZ);
1747 
1748 	/* reset and initialize the hardware so it is in a known state */
1749 	err = hw->mac.ops.reset_hw(hw);
1750 	if (err) {
1751 		dev_err(&pdev->dev, "reset_hw failed: %d\n", err);
1752 		return err;
1753 	}
1754 
1755 	err = hw->mac.ops.init_hw(hw);
1756 	if (err) {
1757 		dev_err(&pdev->dev, "init_hw failed: %d\n", err);
1758 		return err;
1759 	}
1760 
1761 	/* initialize hardware statistics */
1762 	hw->mac.ops.update_hw_stats(hw, &interface->stats);
1763 
1764 	/* Set upper limit on IOV VFs that can be allocated */
1765 	pci_sriov_set_totalvfs(pdev, hw->iov.total_vfs);
1766 
1767 	/* Start with random Ethernet address */
1768 	eth_random_addr(hw->mac.addr);
1769 
1770 	/* Initialize MAC address from hardware */
1771 	err = hw->mac.ops.read_mac_addr(hw);
1772 	if (err) {
1773 		dev_warn(&pdev->dev,
1774 			 "Failed to obtain MAC address defaulting to random\n");
1775 		/* tag address assignment as random */
1776 		netdev->addr_assign_type |= NET_ADDR_RANDOM;
1777 	}
1778 
1779 	memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
1780 	memcpy(netdev->perm_addr, hw->mac.addr, netdev->addr_len);
1781 
1782 	if (!is_valid_ether_addr(netdev->perm_addr)) {
1783 		dev_err(&pdev->dev, "Invalid MAC Address\n");
1784 		return -EIO;
1785 	}
1786 
1787 	/* assign BAR 4 resources for use with PTP */
1788 	if (fm10k_read_reg(hw, FM10K_CTRL) & FM10K_CTRL_BAR4_ALLOWED)
1789 		interface->sw_addr = ioremap(pci_resource_start(pdev, 4),
1790 					     pci_resource_len(pdev, 4));
1791 	hw->sw_addr = interface->sw_addr;
1792 
1793 	/* initialize DCBNL interface */
1794 	fm10k_dcbnl_set_ops(netdev);
1795 
1796 	/* Initialize service timer and service task */
1797 	set_bit(__FM10K_SERVICE_DISABLE, &interface->state);
1798 	setup_timer(&interface->service_timer, &fm10k_service_timer,
1799 		    (unsigned long)interface);
1800 	INIT_WORK(&interface->service_task, fm10k_service_task);
1801 
1802 	/* kick off service timer now, even when interface is down */
1803 	mod_timer(&interface->service_timer, (HZ * 2) + jiffies);
1804 
1805 	/* Intitialize timestamp data */
1806 	fm10k_ts_init(interface);
1807 
1808 	/* set default ring sizes */
1809 	interface->tx_ring_count = FM10K_DEFAULT_TXD;
1810 	interface->rx_ring_count = FM10K_DEFAULT_RXD;
1811 
1812 	/* set default interrupt moderation */
1813 	interface->tx_itr = FM10K_TX_ITR_DEFAULT;
1814 	interface->rx_itr = FM10K_ITR_ADAPTIVE | FM10K_RX_ITR_DEFAULT;
1815 
1816 	/* initialize vxlan_port list */
1817 	INIT_LIST_HEAD(&interface->vxlan_port);
1818 
1819 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
1820 	memcpy(interface->rssrk, rss_key, sizeof(rss_key));
1821 
1822 	/* Start off interface as being down */
1823 	set_bit(__FM10K_DOWN, &interface->state);
1824 
1825 	return 0;
1826 }
1827 
1828 static void fm10k_slot_warn(struct fm10k_intfc *interface)
1829 {
1830 	enum pcie_link_width width = PCIE_LNK_WIDTH_UNKNOWN;
1831 	enum pci_bus_speed speed = PCI_SPEED_UNKNOWN;
1832 	struct fm10k_hw *hw = &interface->hw;
1833 	int max_gts = 0, expected_gts = 0;
1834 
1835 	if (pcie_get_minimum_link(interface->pdev, &speed, &width) ||
1836 	    speed == PCI_SPEED_UNKNOWN || width == PCIE_LNK_WIDTH_UNKNOWN) {
1837 		dev_warn(&interface->pdev->dev,
1838 			 "Unable to determine PCI Express bandwidth.\n");
1839 		return;
1840 	}
1841 
1842 	switch (speed) {
1843 	case PCIE_SPEED_2_5GT:
1844 		/* 8b/10b encoding reduces max throughput by 20% */
1845 		max_gts = 2 * width;
1846 		break;
1847 	case PCIE_SPEED_5_0GT:
1848 		/* 8b/10b encoding reduces max throughput by 20% */
1849 		max_gts = 4 * width;
1850 		break;
1851 	case PCIE_SPEED_8_0GT:
1852 		/* 128b/130b encoding has less than 2% impact on throughput */
1853 		max_gts = 8 * width;
1854 		break;
1855 	default:
1856 		dev_warn(&interface->pdev->dev,
1857 			 "Unable to determine PCI Express bandwidth.\n");
1858 		return;
1859 	}
1860 
1861 	dev_info(&interface->pdev->dev,
1862 		 "PCI Express bandwidth of %dGT/s available\n",
1863 		 max_gts);
1864 	dev_info(&interface->pdev->dev,
1865 		 "(Speed:%s, Width: x%d, Encoding Loss:%s, Payload:%s)\n",
1866 		 (speed == PCIE_SPEED_8_0GT ? "8.0GT/s" :
1867 		  speed == PCIE_SPEED_5_0GT ? "5.0GT/s" :
1868 		  speed == PCIE_SPEED_2_5GT ? "2.5GT/s" :
1869 		  "Unknown"),
1870 		 hw->bus.width,
1871 		 (speed == PCIE_SPEED_2_5GT ? "20%" :
1872 		  speed == PCIE_SPEED_5_0GT ? "20%" :
1873 		  speed == PCIE_SPEED_8_0GT ? "<2%" :
1874 		  "Unknown"),
1875 		 (hw->bus.payload == fm10k_bus_payload_128 ? "128B" :
1876 		  hw->bus.payload == fm10k_bus_payload_256 ? "256B" :
1877 		  hw->bus.payload == fm10k_bus_payload_512 ? "512B" :
1878 		  "Unknown"));
1879 
1880 	switch (hw->bus_caps.speed) {
1881 	case fm10k_bus_speed_2500:
1882 		/* 8b/10b encoding reduces max throughput by 20% */
1883 		expected_gts = 2 * hw->bus_caps.width;
1884 		break;
1885 	case fm10k_bus_speed_5000:
1886 		/* 8b/10b encoding reduces max throughput by 20% */
1887 		expected_gts = 4 * hw->bus_caps.width;
1888 		break;
1889 	case fm10k_bus_speed_8000:
1890 		/* 128b/130b encoding has less than 2% impact on throughput */
1891 		expected_gts = 8 * hw->bus_caps.width;
1892 		break;
1893 	default:
1894 		dev_warn(&interface->pdev->dev,
1895 			 "Unable to determine expected PCI Express bandwidth.\n");
1896 		return;
1897 	}
1898 
1899 	if (max_gts >= expected_gts)
1900 		return;
1901 
1902 	dev_warn(&interface->pdev->dev,
1903 		 "This device requires %dGT/s of bandwidth for optimal performance.\n",
1904 		 expected_gts);
1905 	dev_warn(&interface->pdev->dev,
1906 		 "A %sslot with x%d lanes is suggested.\n",
1907 		 (hw->bus_caps.speed == fm10k_bus_speed_2500 ? "2.5GT/s " :
1908 		  hw->bus_caps.speed == fm10k_bus_speed_5000 ? "5.0GT/s " :
1909 		  hw->bus_caps.speed == fm10k_bus_speed_8000 ? "8.0GT/s " : ""),
1910 		 hw->bus_caps.width);
1911 }
1912 
1913 /**
1914  * fm10k_probe - Device Initialization Routine
1915  * @pdev: PCI device information struct
1916  * @ent: entry in fm10k_pci_tbl
1917  *
1918  * Returns 0 on success, negative on failure
1919  *
1920  * fm10k_probe initializes an interface identified by a pci_dev structure.
1921  * The OS initialization, configuring of the interface private structure,
1922  * and a hardware reset occur.
1923  **/
1924 static int fm10k_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
1925 {
1926 	struct net_device *netdev;
1927 	struct fm10k_intfc *interface;
1928 	int err;
1929 
1930 	err = pci_enable_device_mem(pdev);
1931 	if (err)
1932 		return err;
1933 
1934 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(48));
1935 	if (err)
1936 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
1937 	if (err) {
1938 		dev_err(&pdev->dev,
1939 			"DMA configuration failed: %d\n", err);
1940 		goto err_dma;
1941 	}
1942 
1943 	err = pci_request_selected_regions(pdev,
1944 					   pci_select_bars(pdev,
1945 							   IORESOURCE_MEM),
1946 					   fm10k_driver_name);
1947 	if (err) {
1948 		dev_err(&pdev->dev,
1949 			"pci_request_selected_regions failed: %d\n", err);
1950 		goto err_pci_reg;
1951 	}
1952 
1953 	pci_enable_pcie_error_reporting(pdev);
1954 
1955 	pci_set_master(pdev);
1956 	pci_save_state(pdev);
1957 
1958 	netdev = fm10k_alloc_netdev(fm10k_info_tbl[ent->driver_data]);
1959 	if (!netdev) {
1960 		err = -ENOMEM;
1961 		goto err_alloc_netdev;
1962 	}
1963 
1964 	SET_NETDEV_DEV(netdev, &pdev->dev);
1965 
1966 	interface = netdev_priv(netdev);
1967 	pci_set_drvdata(pdev, interface);
1968 
1969 	interface->netdev = netdev;
1970 	interface->pdev = pdev;
1971 
1972 	interface->uc_addr = ioremap(pci_resource_start(pdev, 0),
1973 				     FM10K_UC_ADDR_SIZE);
1974 	if (!interface->uc_addr) {
1975 		err = -EIO;
1976 		goto err_ioremap;
1977 	}
1978 
1979 	err = fm10k_sw_init(interface, ent);
1980 	if (err)
1981 		goto err_sw_init;
1982 
1983 	/* enable debugfs support */
1984 	fm10k_dbg_intfc_init(interface);
1985 
1986 	err = fm10k_init_queueing_scheme(interface);
1987 	if (err)
1988 		goto err_sw_init;
1989 
1990 	err = fm10k_mbx_request_irq(interface);
1991 	if (err)
1992 		goto err_mbx_interrupt;
1993 
1994 	/* final check of hardware state before registering the interface */
1995 	err = fm10k_hw_ready(interface);
1996 	if (err)
1997 		goto err_register;
1998 
1999 	err = register_netdev(netdev);
2000 	if (err)
2001 		goto err_register;
2002 
2003 	/* carrier off reporting is important to ethtool even BEFORE open */
2004 	netif_carrier_off(netdev);
2005 
2006 	/* stop all the transmit queues from transmitting until link is up */
2007 	netif_tx_stop_all_queues(netdev);
2008 
2009 	/* Register PTP interface */
2010 	fm10k_ptp_register(interface);
2011 
2012 	/* print warning for non-optimal configurations */
2013 	fm10k_slot_warn(interface);
2014 
2015 	/* report MAC address for logging */
2016 	dev_info(&pdev->dev, "%pM\n", netdev->dev_addr);
2017 
2018 	/* enable SR-IOV after registering netdev to enforce PF/VF ordering */
2019 	fm10k_iov_configure(pdev, 0);
2020 
2021 	/* clear the service task disable bit to allow service task to start */
2022 	clear_bit(__FM10K_SERVICE_DISABLE, &interface->state);
2023 
2024 	return 0;
2025 
2026 err_register:
2027 	fm10k_mbx_free_irq(interface);
2028 err_mbx_interrupt:
2029 	fm10k_clear_queueing_scheme(interface);
2030 err_sw_init:
2031 	if (interface->sw_addr)
2032 		iounmap(interface->sw_addr);
2033 	iounmap(interface->uc_addr);
2034 err_ioremap:
2035 	free_netdev(netdev);
2036 err_alloc_netdev:
2037 	pci_release_selected_regions(pdev,
2038 				     pci_select_bars(pdev, IORESOURCE_MEM));
2039 err_pci_reg:
2040 err_dma:
2041 	pci_disable_device(pdev);
2042 	return err;
2043 }
2044 
2045 /**
2046  * fm10k_remove - Device Removal Routine
2047  * @pdev: PCI device information struct
2048  *
2049  * fm10k_remove is called by the PCI subsystem to alert the driver
2050  * that it should release a PCI device.  The could be caused by a
2051  * Hot-Plug event, or because the driver is going to be removed from
2052  * memory.
2053  **/
2054 static void fm10k_remove(struct pci_dev *pdev)
2055 {
2056 	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
2057 	struct net_device *netdev = interface->netdev;
2058 
2059 	del_timer_sync(&interface->service_timer);
2060 
2061 	set_bit(__FM10K_SERVICE_DISABLE, &interface->state);
2062 	cancel_work_sync(&interface->service_task);
2063 
2064 	/* free netdev, this may bounce the interrupts due to setup_tc */
2065 	if (netdev->reg_state == NETREG_REGISTERED)
2066 		unregister_netdev(netdev);
2067 
2068 	/* cleanup timestamp handling */
2069 	fm10k_ptp_unregister(interface);
2070 
2071 	/* release VFs */
2072 	fm10k_iov_disable(pdev);
2073 
2074 	/* disable mailbox interrupt */
2075 	fm10k_mbx_free_irq(interface);
2076 
2077 	/* free interrupts */
2078 	fm10k_clear_queueing_scheme(interface);
2079 
2080 	/* remove any debugfs interfaces */
2081 	fm10k_dbg_intfc_exit(interface);
2082 
2083 	if (interface->sw_addr)
2084 		iounmap(interface->sw_addr);
2085 	iounmap(interface->uc_addr);
2086 
2087 	free_netdev(netdev);
2088 
2089 	pci_release_selected_regions(pdev,
2090 				     pci_select_bars(pdev, IORESOURCE_MEM));
2091 
2092 	pci_disable_pcie_error_reporting(pdev);
2093 
2094 	pci_disable_device(pdev);
2095 }
2096 
2097 #ifdef CONFIG_PM
2098 /**
2099  * fm10k_resume - Restore device to pre-sleep state
2100  * @pdev: PCI device information struct
2101  *
2102  * fm10k_resume is called after the system has powered back up from a sleep
2103  * state and is ready to resume operation.  This function is meant to restore
2104  * the device back to its pre-sleep state.
2105  **/
2106 static int fm10k_resume(struct pci_dev *pdev)
2107 {
2108 	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
2109 	struct net_device *netdev = interface->netdev;
2110 	struct fm10k_hw *hw = &interface->hw;
2111 	u32 err;
2112 
2113 	pci_set_power_state(pdev, PCI_D0);
2114 	pci_restore_state(pdev);
2115 
2116 	/* pci_restore_state clears dev->state_saved so call
2117 	 * pci_save_state to restore it.
2118 	 */
2119 	pci_save_state(pdev);
2120 
2121 	err = pci_enable_device_mem(pdev);
2122 	if (err) {
2123 		dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
2124 		return err;
2125 	}
2126 	pci_set_master(pdev);
2127 
2128 	pci_wake_from_d3(pdev, false);
2129 
2130 	/* refresh hw_addr in case it was dropped */
2131 	hw->hw_addr = interface->uc_addr;
2132 
2133 	/* reset hardware to known state */
2134 	err = hw->mac.ops.init_hw(&interface->hw);
2135 	if (err) {
2136 		dev_err(&pdev->dev, "init_hw failed: %d\n", err);
2137 		return err;
2138 	}
2139 
2140 	/* reset statistics starting values */
2141 	hw->mac.ops.rebind_hw_stats(hw, &interface->stats);
2142 
2143 	/* reset clock */
2144 	fm10k_ts_reset(interface);
2145 
2146 	rtnl_lock();
2147 
2148 	err = fm10k_init_queueing_scheme(interface);
2149 	if (err)
2150 		goto err_queueing_scheme;
2151 
2152 	err = fm10k_mbx_request_irq(interface);
2153 	if (err)
2154 		goto err_mbx_irq;
2155 
2156 	err = fm10k_hw_ready(interface);
2157 	if (err)
2158 		goto err_open;
2159 
2160 	err = netif_running(netdev) ? fm10k_open(netdev) : 0;
2161 	if (err)
2162 		goto err_open;
2163 
2164 	rtnl_unlock();
2165 
2166 	/* assume host is not ready, to prevent race with watchdog in case we
2167 	 * actually don't have connection to the switch
2168 	 */
2169 	interface->host_ready = false;
2170 	fm10k_watchdog_host_not_ready(interface);
2171 
2172 	/* clear the service task disable bit to allow service task to start */
2173 	clear_bit(__FM10K_SERVICE_DISABLE, &interface->state);
2174 	fm10k_service_event_schedule(interface);
2175 
2176 	/* restore SR-IOV interface */
2177 	fm10k_iov_resume(pdev);
2178 
2179 	netif_device_attach(netdev);
2180 
2181 	return 0;
2182 err_open:
2183 	fm10k_mbx_free_irq(interface);
2184 err_mbx_irq:
2185 	fm10k_clear_queueing_scheme(interface);
2186 err_queueing_scheme:
2187 	rtnl_unlock();
2188 
2189 	return err;
2190 }
2191 
2192 /**
2193  * fm10k_suspend - Prepare the device for a system sleep state
2194  * @pdev: PCI device information struct
2195  *
2196  * fm10k_suspend is meant to shutdown the device prior to the system entering
2197  * a sleep state.  The fm10k hardware does not support wake on lan so the
2198  * driver simply needs to shut down the device so it is in a low power state.
2199  **/
2200 static int fm10k_suspend(struct pci_dev *pdev,
2201 			 pm_message_t __always_unused state)
2202 {
2203 	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
2204 	struct net_device *netdev = interface->netdev;
2205 	int err = 0;
2206 
2207 	netif_device_detach(netdev);
2208 
2209 	fm10k_iov_suspend(pdev);
2210 
2211 	/* the watchdog tasks may read registers, which will appear like a
2212 	 * surprise-remove event once the PCI device is disabled. This will
2213 	 * cause us to close the netdevice, so we don't retain the open/closed
2214 	 * state post-resume. Prevent this by disabling the service task while
2215 	 * suspended, until we actually resume.
2216 	 */
2217 	set_bit(__FM10K_SERVICE_DISABLE, &interface->state);
2218 	cancel_work_sync(&interface->service_task);
2219 
2220 	rtnl_lock();
2221 
2222 	if (netif_running(netdev))
2223 		fm10k_close(netdev);
2224 
2225 	fm10k_mbx_free_irq(interface);
2226 
2227 	fm10k_clear_queueing_scheme(interface);
2228 
2229 	rtnl_unlock();
2230 
2231 	err = pci_save_state(pdev);
2232 	if (err)
2233 		return err;
2234 
2235 	pci_disable_device(pdev);
2236 	pci_wake_from_d3(pdev, false);
2237 	pci_set_power_state(pdev, PCI_D3hot);
2238 
2239 	return 0;
2240 }
2241 
2242 #endif /* CONFIG_PM */
2243 /**
2244  * fm10k_io_error_detected - called when PCI error is detected
2245  * @pdev: Pointer to PCI device
2246  * @state: The current pci connection state
2247  *
2248  * This function is called after a PCI bus error affecting
2249  * this device has been detected.
2250  */
2251 static pci_ers_result_t fm10k_io_error_detected(struct pci_dev *pdev,
2252 						pci_channel_state_t state)
2253 {
2254 	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
2255 	struct net_device *netdev = interface->netdev;
2256 
2257 	netif_device_detach(netdev);
2258 
2259 	if (state == pci_channel_io_perm_failure)
2260 		return PCI_ERS_RESULT_DISCONNECT;
2261 
2262 	if (netif_running(netdev))
2263 		fm10k_close(netdev);
2264 
2265 	/* free interrupts */
2266 	fm10k_clear_queueing_scheme(interface);
2267 
2268 	fm10k_mbx_free_irq(interface);
2269 
2270 	pci_disable_device(pdev);
2271 
2272 	/* Request a slot reset. */
2273 	return PCI_ERS_RESULT_NEED_RESET;
2274 }
2275 
2276 /**
2277  * fm10k_io_slot_reset - called after the pci bus has been reset.
2278  * @pdev: Pointer to PCI device
2279  *
2280  * Restart the card from scratch, as if from a cold-boot.
2281  */
2282 static pci_ers_result_t fm10k_io_slot_reset(struct pci_dev *pdev)
2283 {
2284 	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
2285 	pci_ers_result_t result;
2286 
2287 	if (pci_enable_device_mem(pdev)) {
2288 		dev_err(&pdev->dev,
2289 			"Cannot re-enable PCI device after reset.\n");
2290 		result = PCI_ERS_RESULT_DISCONNECT;
2291 	} else {
2292 		pci_set_master(pdev);
2293 		pci_restore_state(pdev);
2294 
2295 		/* After second error pci->state_saved is false, this
2296 		 * resets it so EEH doesn't break.
2297 		 */
2298 		pci_save_state(pdev);
2299 
2300 		pci_wake_from_d3(pdev, false);
2301 
2302 		/* refresh hw_addr in case it was dropped */
2303 		interface->hw.hw_addr = interface->uc_addr;
2304 
2305 		interface->flags |= FM10K_FLAG_RESET_REQUESTED;
2306 		fm10k_service_event_schedule(interface);
2307 
2308 		result = PCI_ERS_RESULT_RECOVERED;
2309 	}
2310 
2311 	pci_cleanup_aer_uncorrect_error_status(pdev);
2312 
2313 	return result;
2314 }
2315 
2316 /**
2317  * fm10k_io_resume - called when traffic can start flowing again.
2318  * @pdev: Pointer to PCI device
2319  *
2320  * This callback is called when the error recovery driver tells us that
2321  * its OK to resume normal operation.
2322  */
2323 static void fm10k_io_resume(struct pci_dev *pdev)
2324 {
2325 	struct fm10k_intfc *interface = pci_get_drvdata(pdev);
2326 	struct net_device *netdev = interface->netdev;
2327 	struct fm10k_hw *hw = &interface->hw;
2328 	int err = 0;
2329 
2330 	/* reset hardware to known state */
2331 	err = hw->mac.ops.init_hw(&interface->hw);
2332 	if (err) {
2333 		dev_err(&pdev->dev, "init_hw failed: %d\n", err);
2334 		return;
2335 	}
2336 
2337 	/* reset statistics starting values */
2338 	hw->mac.ops.rebind_hw_stats(hw, &interface->stats);
2339 
2340 	err = fm10k_init_queueing_scheme(interface);
2341 	if (err) {
2342 		dev_err(&interface->pdev->dev,
2343 			"init_queueing_scheme failed: %d\n", err);
2344 		return;
2345 	}
2346 
2347 	/* reassociate interrupts */
2348 	fm10k_mbx_request_irq(interface);
2349 
2350 	/* reset clock */
2351 	fm10k_ts_reset(interface);
2352 
2353 	if (netif_running(netdev))
2354 		err = fm10k_open(netdev);
2355 
2356 	/* final check of hardware state before registering the interface */
2357 	err = err ? : fm10k_hw_ready(interface);
2358 
2359 	if (!err)
2360 		netif_device_attach(netdev);
2361 }
2362 
2363 static const struct pci_error_handlers fm10k_err_handler = {
2364 	.error_detected = fm10k_io_error_detected,
2365 	.slot_reset = fm10k_io_slot_reset,
2366 	.resume = fm10k_io_resume,
2367 };
2368 
2369 static struct pci_driver fm10k_driver = {
2370 	.name			= fm10k_driver_name,
2371 	.id_table		= fm10k_pci_tbl,
2372 	.probe			= fm10k_probe,
2373 	.remove			= fm10k_remove,
2374 #ifdef CONFIG_PM
2375 	.suspend		= fm10k_suspend,
2376 	.resume			= fm10k_resume,
2377 #endif
2378 	.sriov_configure	= fm10k_iov_configure,
2379 	.err_handler		= &fm10k_err_handler
2380 };
2381 
2382 /**
2383  * fm10k_register_pci_driver - register driver interface
2384  *
2385  * This funciton is called on module load in order to register the driver.
2386  **/
2387 int fm10k_register_pci_driver(void)
2388 {
2389 	return pci_register_driver(&fm10k_driver);
2390 }
2391 
2392 /**
2393  * fm10k_unregister_pci_driver - unregister driver interface
2394  *
2395  * This funciton is called on module unload in order to remove the driver.
2396  **/
2397 void fm10k_unregister_pci_driver(void)
2398 {
2399 	pci_unregister_driver(&fm10k_driver);
2400 }
2401