1 /*******************************************************************************
2  *
3  * Intel Ethernet Controller XL710 Family Linux Driver
4  * Copyright(c) 2013 - 2015 Intel Corporation.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along
16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * The full GNU General Public License is included in this distribution in
19  * the file called "COPYING".
20  *
21  * Contact Information:
22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
24  *
25  ******************************************************************************/
26 
27 #include <linux/etherdevice.h>
28 #include <linux/of_net.h>
29 #include <linux/pci.h>
30 
31 #ifdef CONFIG_SPARC
32 #include <asm/idprom.h>
33 #include <asm/prom.h>
34 #endif
35 
36 /* Local includes */
37 #include "i40e.h"
38 #include "i40e_diag.h"
39 #if IS_ENABLED(CONFIG_VXLAN)
40 #include <net/vxlan.h>
41 #endif
42 #if IS_ENABLED(CONFIG_GENEVE)
43 #include <net/geneve.h>
44 #endif
45 
46 const char i40e_driver_name[] = "i40e";
47 static const char i40e_driver_string[] =
48 			"Intel(R) Ethernet Connection XL710 Network Driver";
49 
50 #define DRV_KERN "-k"
51 
52 #define DRV_VERSION_MAJOR 1
53 #define DRV_VERSION_MINOR 4
54 #define DRV_VERSION_BUILD 8
55 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
56 	     __stringify(DRV_VERSION_MINOR) "." \
57 	     __stringify(DRV_VERSION_BUILD)    DRV_KERN
58 const char i40e_driver_version_str[] = DRV_VERSION;
59 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
60 
61 /* a bit of forward declarations */
62 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
63 static void i40e_handle_reset_warning(struct i40e_pf *pf);
64 static int i40e_add_vsi(struct i40e_vsi *vsi);
65 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
66 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
67 static int i40e_setup_misc_vector(struct i40e_pf *pf);
68 static void i40e_determine_queue_usage(struct i40e_pf *pf);
69 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
70 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
71 			      u16 rss_table_size, u16 rss_size);
72 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
73 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
74 
75 /* i40e_pci_tbl - PCI Device ID Table
76  *
77  * Last entry must be all 0s
78  *
79  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
80  *   Class, Class Mask, private data (not used) }
81  */
82 static const struct pci_device_id i40e_pci_tbl[] = {
83 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
84 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
85 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
86 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
87 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
88 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
89 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
90 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
91 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
92 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
93 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
94 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
95 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
96 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
97 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
98 	/* required last entry */
99 	{0, }
100 };
101 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
102 
103 #define I40E_MAX_VF_COUNT 128
104 static int debug = -1;
105 module_param(debug, int, 0);
106 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
107 
108 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
109 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
110 MODULE_LICENSE("GPL");
111 MODULE_VERSION(DRV_VERSION);
112 
113 /**
114  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
115  * @hw:   pointer to the HW structure
116  * @mem:  ptr to mem struct to fill out
117  * @size: size of memory requested
118  * @alignment: what to align the allocation to
119  **/
120 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
121 			    u64 size, u32 alignment)
122 {
123 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
124 
125 	mem->size = ALIGN(size, alignment);
126 	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
127 				      &mem->pa, GFP_KERNEL);
128 	if (!mem->va)
129 		return -ENOMEM;
130 
131 	return 0;
132 }
133 
134 /**
135  * i40e_free_dma_mem_d - OS specific memory free for shared code
136  * @hw:   pointer to the HW structure
137  * @mem:  ptr to mem struct to free
138  **/
139 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
140 {
141 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
142 
143 	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
144 	mem->va = NULL;
145 	mem->pa = 0;
146 	mem->size = 0;
147 
148 	return 0;
149 }
150 
151 /**
152  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
153  * @hw:   pointer to the HW structure
154  * @mem:  ptr to mem struct to fill out
155  * @size: size of memory requested
156  **/
157 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
158 			     u32 size)
159 {
160 	mem->size = size;
161 	mem->va = kzalloc(size, GFP_KERNEL);
162 
163 	if (!mem->va)
164 		return -ENOMEM;
165 
166 	return 0;
167 }
168 
169 /**
170  * i40e_free_virt_mem_d - OS specific memory free for shared code
171  * @hw:   pointer to the HW structure
172  * @mem:  ptr to mem struct to free
173  **/
174 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
175 {
176 	/* it's ok to kfree a NULL pointer */
177 	kfree(mem->va);
178 	mem->va = NULL;
179 	mem->size = 0;
180 
181 	return 0;
182 }
183 
184 /**
185  * i40e_get_lump - find a lump of free generic resource
186  * @pf: board private structure
187  * @pile: the pile of resource to search
188  * @needed: the number of items needed
189  * @id: an owner id to stick on the items assigned
190  *
191  * Returns the base item index of the lump, or negative for error
192  *
193  * The search_hint trick and lack of advanced fit-finding only work
194  * because we're highly likely to have all the same size lump requests.
195  * Linear search time and any fragmentation should be minimal.
196  **/
197 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
198 			 u16 needed, u16 id)
199 {
200 	int ret = -ENOMEM;
201 	int i, j;
202 
203 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
204 		dev_info(&pf->pdev->dev,
205 			 "param err: pile=%p needed=%d id=0x%04x\n",
206 			 pile, needed, id);
207 		return -EINVAL;
208 	}
209 
210 	/* start the linear search with an imperfect hint */
211 	i = pile->search_hint;
212 	while (i < pile->num_entries) {
213 		/* skip already allocated entries */
214 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
215 			i++;
216 			continue;
217 		}
218 
219 		/* do we have enough in this lump? */
220 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
221 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
222 				break;
223 		}
224 
225 		if (j == needed) {
226 			/* there was enough, so assign it to the requestor */
227 			for (j = 0; j < needed; j++)
228 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
229 			ret = i;
230 			pile->search_hint = i + j;
231 			break;
232 		}
233 
234 		/* not enough, so skip over it and continue looking */
235 		i += j;
236 	}
237 
238 	return ret;
239 }
240 
241 /**
242  * i40e_put_lump - return a lump of generic resource
243  * @pile: the pile of resource to search
244  * @index: the base item index
245  * @id: the owner id of the items assigned
246  *
247  * Returns the count of items in the lump
248  **/
249 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
250 {
251 	int valid_id = (id | I40E_PILE_VALID_BIT);
252 	int count = 0;
253 	int i;
254 
255 	if (!pile || index >= pile->num_entries)
256 		return -EINVAL;
257 
258 	for (i = index;
259 	     i < pile->num_entries && pile->list[i] == valid_id;
260 	     i++) {
261 		pile->list[i] = 0;
262 		count++;
263 	}
264 
265 	if (count && index < pile->search_hint)
266 		pile->search_hint = index;
267 
268 	return count;
269 }
270 
271 /**
272  * i40e_find_vsi_from_id - searches for the vsi with the given id
273  * @pf - the pf structure to search for the vsi
274  * @id - id of the vsi it is searching for
275  **/
276 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
277 {
278 	int i;
279 
280 	for (i = 0; i < pf->num_alloc_vsi; i++)
281 		if (pf->vsi[i] && (pf->vsi[i]->id == id))
282 			return pf->vsi[i];
283 
284 	return NULL;
285 }
286 
287 /**
288  * i40e_service_event_schedule - Schedule the service task to wake up
289  * @pf: board private structure
290  *
291  * If not already scheduled, this puts the task into the work queue
292  **/
293 static void i40e_service_event_schedule(struct i40e_pf *pf)
294 {
295 	if (!test_bit(__I40E_DOWN, &pf->state) &&
296 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
297 	    !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
298 		schedule_work(&pf->service_task);
299 }
300 
301 /**
302  * i40e_tx_timeout - Respond to a Tx Hang
303  * @netdev: network interface device structure
304  *
305  * If any port has noticed a Tx timeout, it is likely that the whole
306  * device is munged, not just the one netdev port, so go for the full
307  * reset.
308  **/
309 #ifdef I40E_FCOE
310 void i40e_tx_timeout(struct net_device *netdev)
311 #else
312 static void i40e_tx_timeout(struct net_device *netdev)
313 #endif
314 {
315 	struct i40e_netdev_priv *np = netdev_priv(netdev);
316 	struct i40e_vsi *vsi = np->vsi;
317 	struct i40e_pf *pf = vsi->back;
318 	struct i40e_ring *tx_ring = NULL;
319 	unsigned int i, hung_queue = 0;
320 	u32 head, val;
321 
322 	pf->tx_timeout_count++;
323 
324 	/* find the stopped queue the same way the stack does */
325 	for (i = 0; i < netdev->num_tx_queues; i++) {
326 		struct netdev_queue *q;
327 		unsigned long trans_start;
328 
329 		q = netdev_get_tx_queue(netdev, i);
330 		trans_start = q->trans_start ? : netdev->trans_start;
331 		if (netif_xmit_stopped(q) &&
332 		    time_after(jiffies,
333 			       (trans_start + netdev->watchdog_timeo))) {
334 			hung_queue = i;
335 			break;
336 		}
337 	}
338 
339 	if (i == netdev->num_tx_queues) {
340 		netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
341 	} else {
342 		/* now that we have an index, find the tx_ring struct */
343 		for (i = 0; i < vsi->num_queue_pairs; i++) {
344 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
345 				if (hung_queue ==
346 				    vsi->tx_rings[i]->queue_index) {
347 					tx_ring = vsi->tx_rings[i];
348 					break;
349 				}
350 			}
351 		}
352 	}
353 
354 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
355 		pf->tx_timeout_recovery_level = 1;  /* reset after some time */
356 	else if (time_before(jiffies,
357 		      (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
358 		return;   /* don't do any new action before the next timeout */
359 
360 	if (tx_ring) {
361 		head = i40e_get_head(tx_ring);
362 		/* Read interrupt register */
363 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
364 			val = rd32(&pf->hw,
365 			     I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
366 						tx_ring->vsi->base_vector - 1));
367 		else
368 			val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
369 
370 		netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
371 			    vsi->seid, hung_queue, tx_ring->next_to_clean,
372 			    head, tx_ring->next_to_use,
373 			    readl(tx_ring->tail), val);
374 	}
375 
376 	pf->tx_timeout_last_recovery = jiffies;
377 	netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
378 		    pf->tx_timeout_recovery_level, hung_queue);
379 
380 	switch (pf->tx_timeout_recovery_level) {
381 	case 1:
382 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
383 		break;
384 	case 2:
385 		set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
386 		break;
387 	case 3:
388 		set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
389 		break;
390 	default:
391 		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
392 		break;
393 	}
394 
395 	i40e_service_event_schedule(pf);
396 	pf->tx_timeout_recovery_level++;
397 }
398 
399 /**
400  * i40e_release_rx_desc - Store the new tail and head values
401  * @rx_ring: ring to bump
402  * @val: new head index
403  **/
404 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
405 {
406 	rx_ring->next_to_use = val;
407 
408 	/* Force memory writes to complete before letting h/w
409 	 * know there are new descriptors to fetch.  (Only
410 	 * applicable for weak-ordered memory model archs,
411 	 * such as IA-64).
412 	 */
413 	wmb();
414 	writel(val, rx_ring->tail);
415 }
416 
417 /**
418  * i40e_get_vsi_stats_struct - Get System Network Statistics
419  * @vsi: the VSI we care about
420  *
421  * Returns the address of the device statistics structure.
422  * The statistics are actually updated from the service task.
423  **/
424 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
425 {
426 	return &vsi->net_stats;
427 }
428 
429 /**
430  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
431  * @netdev: network interface device structure
432  *
433  * Returns the address of the device statistics structure.
434  * The statistics are actually updated from the service task.
435  **/
436 #ifdef I40E_FCOE
437 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
438 					     struct net_device *netdev,
439 					     struct rtnl_link_stats64 *stats)
440 #else
441 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
442 					     struct net_device *netdev,
443 					     struct rtnl_link_stats64 *stats)
444 #endif
445 {
446 	struct i40e_netdev_priv *np = netdev_priv(netdev);
447 	struct i40e_ring *tx_ring, *rx_ring;
448 	struct i40e_vsi *vsi = np->vsi;
449 	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
450 	int i;
451 
452 	if (test_bit(__I40E_DOWN, &vsi->state))
453 		return stats;
454 
455 	if (!vsi->tx_rings)
456 		return stats;
457 
458 	rcu_read_lock();
459 	for (i = 0; i < vsi->num_queue_pairs; i++) {
460 		u64 bytes, packets;
461 		unsigned int start;
462 
463 		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
464 		if (!tx_ring)
465 			continue;
466 
467 		do {
468 			start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
469 			packets = tx_ring->stats.packets;
470 			bytes   = tx_ring->stats.bytes;
471 		} while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
472 
473 		stats->tx_packets += packets;
474 		stats->tx_bytes   += bytes;
475 		rx_ring = &tx_ring[1];
476 
477 		do {
478 			start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
479 			packets = rx_ring->stats.packets;
480 			bytes   = rx_ring->stats.bytes;
481 		} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
482 
483 		stats->rx_packets += packets;
484 		stats->rx_bytes   += bytes;
485 	}
486 	rcu_read_unlock();
487 
488 	/* following stats updated by i40e_watchdog_subtask() */
489 	stats->multicast	= vsi_stats->multicast;
490 	stats->tx_errors	= vsi_stats->tx_errors;
491 	stats->tx_dropped	= vsi_stats->tx_dropped;
492 	stats->rx_errors	= vsi_stats->rx_errors;
493 	stats->rx_dropped	= vsi_stats->rx_dropped;
494 	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
495 	stats->rx_length_errors	= vsi_stats->rx_length_errors;
496 
497 	return stats;
498 }
499 
500 /**
501  * i40e_vsi_reset_stats - Resets all stats of the given vsi
502  * @vsi: the VSI to have its stats reset
503  **/
504 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
505 {
506 	struct rtnl_link_stats64 *ns;
507 	int i;
508 
509 	if (!vsi)
510 		return;
511 
512 	ns = i40e_get_vsi_stats_struct(vsi);
513 	memset(ns, 0, sizeof(*ns));
514 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
515 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
516 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
517 	if (vsi->rx_rings && vsi->rx_rings[0]) {
518 		for (i = 0; i < vsi->num_queue_pairs; i++) {
519 			memset(&vsi->rx_rings[i]->stats, 0,
520 			       sizeof(vsi->rx_rings[i]->stats));
521 			memset(&vsi->rx_rings[i]->rx_stats, 0,
522 			       sizeof(vsi->rx_rings[i]->rx_stats));
523 			memset(&vsi->tx_rings[i]->stats, 0,
524 			       sizeof(vsi->tx_rings[i]->stats));
525 			memset(&vsi->tx_rings[i]->tx_stats, 0,
526 			       sizeof(vsi->tx_rings[i]->tx_stats));
527 		}
528 	}
529 	vsi->stat_offsets_loaded = false;
530 }
531 
532 /**
533  * i40e_pf_reset_stats - Reset all of the stats for the given PF
534  * @pf: the PF to be reset
535  **/
536 void i40e_pf_reset_stats(struct i40e_pf *pf)
537 {
538 	int i;
539 
540 	memset(&pf->stats, 0, sizeof(pf->stats));
541 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
542 	pf->stat_offsets_loaded = false;
543 
544 	for (i = 0; i < I40E_MAX_VEB; i++) {
545 		if (pf->veb[i]) {
546 			memset(&pf->veb[i]->stats, 0,
547 			       sizeof(pf->veb[i]->stats));
548 			memset(&pf->veb[i]->stats_offsets, 0,
549 			       sizeof(pf->veb[i]->stats_offsets));
550 			pf->veb[i]->stat_offsets_loaded = false;
551 		}
552 	}
553 }
554 
555 /**
556  * i40e_stat_update48 - read and update a 48 bit stat from the chip
557  * @hw: ptr to the hardware info
558  * @hireg: the high 32 bit reg to read
559  * @loreg: the low 32 bit reg to read
560  * @offset_loaded: has the initial offset been loaded yet
561  * @offset: ptr to current offset value
562  * @stat: ptr to the stat
563  *
564  * Since the device stats are not reset at PFReset, they likely will not
565  * be zeroed when the driver starts.  We'll save the first values read
566  * and use them as offsets to be subtracted from the raw values in order
567  * to report stats that count from zero.  In the process, we also manage
568  * the potential roll-over.
569  **/
570 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
571 			       bool offset_loaded, u64 *offset, u64 *stat)
572 {
573 	u64 new_data;
574 
575 	if (hw->device_id == I40E_DEV_ID_QEMU) {
576 		new_data = rd32(hw, loreg);
577 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
578 	} else {
579 		new_data = rd64(hw, loreg);
580 	}
581 	if (!offset_loaded)
582 		*offset = new_data;
583 	if (likely(new_data >= *offset))
584 		*stat = new_data - *offset;
585 	else
586 		*stat = (new_data + BIT_ULL(48)) - *offset;
587 	*stat &= 0xFFFFFFFFFFFFULL;
588 }
589 
590 /**
591  * i40e_stat_update32 - read and update a 32 bit stat from the chip
592  * @hw: ptr to the hardware info
593  * @reg: the hw reg to read
594  * @offset_loaded: has the initial offset been loaded yet
595  * @offset: ptr to current offset value
596  * @stat: ptr to the stat
597  **/
598 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
599 			       bool offset_loaded, u64 *offset, u64 *stat)
600 {
601 	u32 new_data;
602 
603 	new_data = rd32(hw, reg);
604 	if (!offset_loaded)
605 		*offset = new_data;
606 	if (likely(new_data >= *offset))
607 		*stat = (u32)(new_data - *offset);
608 	else
609 		*stat = (u32)((new_data + BIT_ULL(32)) - *offset);
610 }
611 
612 /**
613  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
614  * @vsi: the VSI to be updated
615  **/
616 void i40e_update_eth_stats(struct i40e_vsi *vsi)
617 {
618 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
619 	struct i40e_pf *pf = vsi->back;
620 	struct i40e_hw *hw = &pf->hw;
621 	struct i40e_eth_stats *oes;
622 	struct i40e_eth_stats *es;     /* device's eth stats */
623 
624 	es = &vsi->eth_stats;
625 	oes = &vsi->eth_stats_offsets;
626 
627 	/* Gather up the stats that the hw collects */
628 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
629 			   vsi->stat_offsets_loaded,
630 			   &oes->tx_errors, &es->tx_errors);
631 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
632 			   vsi->stat_offsets_loaded,
633 			   &oes->rx_discards, &es->rx_discards);
634 	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
635 			   vsi->stat_offsets_loaded,
636 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
637 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
638 			   vsi->stat_offsets_loaded,
639 			   &oes->tx_errors, &es->tx_errors);
640 
641 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
642 			   I40E_GLV_GORCL(stat_idx),
643 			   vsi->stat_offsets_loaded,
644 			   &oes->rx_bytes, &es->rx_bytes);
645 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
646 			   I40E_GLV_UPRCL(stat_idx),
647 			   vsi->stat_offsets_loaded,
648 			   &oes->rx_unicast, &es->rx_unicast);
649 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
650 			   I40E_GLV_MPRCL(stat_idx),
651 			   vsi->stat_offsets_loaded,
652 			   &oes->rx_multicast, &es->rx_multicast);
653 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
654 			   I40E_GLV_BPRCL(stat_idx),
655 			   vsi->stat_offsets_loaded,
656 			   &oes->rx_broadcast, &es->rx_broadcast);
657 
658 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
659 			   I40E_GLV_GOTCL(stat_idx),
660 			   vsi->stat_offsets_loaded,
661 			   &oes->tx_bytes, &es->tx_bytes);
662 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
663 			   I40E_GLV_UPTCL(stat_idx),
664 			   vsi->stat_offsets_loaded,
665 			   &oes->tx_unicast, &es->tx_unicast);
666 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
667 			   I40E_GLV_MPTCL(stat_idx),
668 			   vsi->stat_offsets_loaded,
669 			   &oes->tx_multicast, &es->tx_multicast);
670 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
671 			   I40E_GLV_BPTCL(stat_idx),
672 			   vsi->stat_offsets_loaded,
673 			   &oes->tx_broadcast, &es->tx_broadcast);
674 	vsi->stat_offsets_loaded = true;
675 }
676 
677 /**
678  * i40e_update_veb_stats - Update Switch component statistics
679  * @veb: the VEB being updated
680  **/
681 static void i40e_update_veb_stats(struct i40e_veb *veb)
682 {
683 	struct i40e_pf *pf = veb->pf;
684 	struct i40e_hw *hw = &pf->hw;
685 	struct i40e_eth_stats *oes;
686 	struct i40e_eth_stats *es;     /* device's eth stats */
687 	struct i40e_veb_tc_stats *veb_oes;
688 	struct i40e_veb_tc_stats *veb_es;
689 	int i, idx = 0;
690 
691 	idx = veb->stats_idx;
692 	es = &veb->stats;
693 	oes = &veb->stats_offsets;
694 	veb_es = &veb->tc_stats;
695 	veb_oes = &veb->tc_stats_offsets;
696 
697 	/* Gather up the stats that the hw collects */
698 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
699 			   veb->stat_offsets_loaded,
700 			   &oes->tx_discards, &es->tx_discards);
701 	if (hw->revision_id > 0)
702 		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
703 				   veb->stat_offsets_loaded,
704 				   &oes->rx_unknown_protocol,
705 				   &es->rx_unknown_protocol);
706 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
707 			   veb->stat_offsets_loaded,
708 			   &oes->rx_bytes, &es->rx_bytes);
709 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
710 			   veb->stat_offsets_loaded,
711 			   &oes->rx_unicast, &es->rx_unicast);
712 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
713 			   veb->stat_offsets_loaded,
714 			   &oes->rx_multicast, &es->rx_multicast);
715 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
716 			   veb->stat_offsets_loaded,
717 			   &oes->rx_broadcast, &es->rx_broadcast);
718 
719 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
720 			   veb->stat_offsets_loaded,
721 			   &oes->tx_bytes, &es->tx_bytes);
722 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
723 			   veb->stat_offsets_loaded,
724 			   &oes->tx_unicast, &es->tx_unicast);
725 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
726 			   veb->stat_offsets_loaded,
727 			   &oes->tx_multicast, &es->tx_multicast);
728 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
729 			   veb->stat_offsets_loaded,
730 			   &oes->tx_broadcast, &es->tx_broadcast);
731 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
732 		i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
733 				   I40E_GLVEBTC_RPCL(i, idx),
734 				   veb->stat_offsets_loaded,
735 				   &veb_oes->tc_rx_packets[i],
736 				   &veb_es->tc_rx_packets[i]);
737 		i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
738 				   I40E_GLVEBTC_RBCL(i, idx),
739 				   veb->stat_offsets_loaded,
740 				   &veb_oes->tc_rx_bytes[i],
741 				   &veb_es->tc_rx_bytes[i]);
742 		i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
743 				   I40E_GLVEBTC_TPCL(i, idx),
744 				   veb->stat_offsets_loaded,
745 				   &veb_oes->tc_tx_packets[i],
746 				   &veb_es->tc_tx_packets[i]);
747 		i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
748 				   I40E_GLVEBTC_TBCL(i, idx),
749 				   veb->stat_offsets_loaded,
750 				   &veb_oes->tc_tx_bytes[i],
751 				   &veb_es->tc_tx_bytes[i]);
752 	}
753 	veb->stat_offsets_loaded = true;
754 }
755 
756 #ifdef I40E_FCOE
757 /**
758  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
759  * @vsi: the VSI that is capable of doing FCoE
760  **/
761 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
762 {
763 	struct i40e_pf *pf = vsi->back;
764 	struct i40e_hw *hw = &pf->hw;
765 	struct i40e_fcoe_stats *ofs;
766 	struct i40e_fcoe_stats *fs;     /* device's eth stats */
767 	int idx;
768 
769 	if (vsi->type != I40E_VSI_FCOE)
770 		return;
771 
772 	idx = (pf->pf_seid - I40E_BASE_PF_SEID) + I40E_FCOE_PF_STAT_OFFSET;
773 	fs = &vsi->fcoe_stats;
774 	ofs = &vsi->fcoe_stats_offsets;
775 
776 	i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
777 			   vsi->fcoe_stat_offsets_loaded,
778 			   &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
779 	i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
780 			   vsi->fcoe_stat_offsets_loaded,
781 			   &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
782 	i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
783 			   vsi->fcoe_stat_offsets_loaded,
784 			   &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
785 	i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
786 			   vsi->fcoe_stat_offsets_loaded,
787 			   &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
788 	i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
789 			   vsi->fcoe_stat_offsets_loaded,
790 			   &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
791 	i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
792 			   vsi->fcoe_stat_offsets_loaded,
793 			   &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
794 	i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
795 			   vsi->fcoe_stat_offsets_loaded,
796 			   &ofs->fcoe_last_error, &fs->fcoe_last_error);
797 	i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
798 			   vsi->fcoe_stat_offsets_loaded,
799 			   &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
800 
801 	vsi->fcoe_stat_offsets_loaded = true;
802 }
803 
804 #endif
805 /**
806  * i40e_update_vsi_stats - Update the vsi statistics counters.
807  * @vsi: the VSI to be updated
808  *
809  * There are a few instances where we store the same stat in a
810  * couple of different structs.  This is partly because we have
811  * the netdev stats that need to be filled out, which is slightly
812  * different from the "eth_stats" defined by the chip and used in
813  * VF communications.  We sort it out here.
814  **/
815 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
816 {
817 	struct i40e_pf *pf = vsi->back;
818 	struct rtnl_link_stats64 *ons;
819 	struct rtnl_link_stats64 *ns;   /* netdev stats */
820 	struct i40e_eth_stats *oes;
821 	struct i40e_eth_stats *es;     /* device's eth stats */
822 	u32 tx_restart, tx_busy;
823 	struct i40e_ring *p;
824 	u32 rx_page, rx_buf;
825 	u64 bytes, packets;
826 	unsigned int start;
827 	u64 tx_linearize;
828 	u64 tx_force_wb;
829 	u64 rx_p, rx_b;
830 	u64 tx_p, tx_b;
831 	u16 q;
832 
833 	if (test_bit(__I40E_DOWN, &vsi->state) ||
834 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
835 		return;
836 
837 	ns = i40e_get_vsi_stats_struct(vsi);
838 	ons = &vsi->net_stats_offsets;
839 	es = &vsi->eth_stats;
840 	oes = &vsi->eth_stats_offsets;
841 
842 	/* Gather up the netdev and vsi stats that the driver collects
843 	 * on the fly during packet processing
844 	 */
845 	rx_b = rx_p = 0;
846 	tx_b = tx_p = 0;
847 	tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
848 	rx_page = 0;
849 	rx_buf = 0;
850 	rcu_read_lock();
851 	for (q = 0; q < vsi->num_queue_pairs; q++) {
852 		/* locate Tx ring */
853 		p = ACCESS_ONCE(vsi->tx_rings[q]);
854 
855 		do {
856 			start = u64_stats_fetch_begin_irq(&p->syncp);
857 			packets = p->stats.packets;
858 			bytes = p->stats.bytes;
859 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
860 		tx_b += bytes;
861 		tx_p += packets;
862 		tx_restart += p->tx_stats.restart_queue;
863 		tx_busy += p->tx_stats.tx_busy;
864 		tx_linearize += p->tx_stats.tx_linearize;
865 		tx_force_wb += p->tx_stats.tx_force_wb;
866 
867 		/* Rx queue is part of the same block as Tx queue */
868 		p = &p[1];
869 		do {
870 			start = u64_stats_fetch_begin_irq(&p->syncp);
871 			packets = p->stats.packets;
872 			bytes = p->stats.bytes;
873 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
874 		rx_b += bytes;
875 		rx_p += packets;
876 		rx_buf += p->rx_stats.alloc_buff_failed;
877 		rx_page += p->rx_stats.alloc_page_failed;
878 	}
879 	rcu_read_unlock();
880 	vsi->tx_restart = tx_restart;
881 	vsi->tx_busy = tx_busy;
882 	vsi->tx_linearize = tx_linearize;
883 	vsi->tx_force_wb = tx_force_wb;
884 	vsi->rx_page_failed = rx_page;
885 	vsi->rx_buf_failed = rx_buf;
886 
887 	ns->rx_packets = rx_p;
888 	ns->rx_bytes = rx_b;
889 	ns->tx_packets = tx_p;
890 	ns->tx_bytes = tx_b;
891 
892 	/* update netdev stats from eth stats */
893 	i40e_update_eth_stats(vsi);
894 	ons->tx_errors = oes->tx_errors;
895 	ns->tx_errors = es->tx_errors;
896 	ons->multicast = oes->rx_multicast;
897 	ns->multicast = es->rx_multicast;
898 	ons->rx_dropped = oes->rx_discards;
899 	ns->rx_dropped = es->rx_discards;
900 	ons->tx_dropped = oes->tx_discards;
901 	ns->tx_dropped = es->tx_discards;
902 
903 	/* pull in a couple PF stats if this is the main vsi */
904 	if (vsi == pf->vsi[pf->lan_vsi]) {
905 		ns->rx_crc_errors = pf->stats.crc_errors;
906 		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
907 		ns->rx_length_errors = pf->stats.rx_length_errors;
908 	}
909 }
910 
911 /**
912  * i40e_update_pf_stats - Update the PF statistics counters.
913  * @pf: the PF to be updated
914  **/
915 static void i40e_update_pf_stats(struct i40e_pf *pf)
916 {
917 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
918 	struct i40e_hw_port_stats *nsd = &pf->stats;
919 	struct i40e_hw *hw = &pf->hw;
920 	u32 val;
921 	int i;
922 
923 	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
924 			   I40E_GLPRT_GORCL(hw->port),
925 			   pf->stat_offsets_loaded,
926 			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
927 	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
928 			   I40E_GLPRT_GOTCL(hw->port),
929 			   pf->stat_offsets_loaded,
930 			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
931 	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
932 			   pf->stat_offsets_loaded,
933 			   &osd->eth.rx_discards,
934 			   &nsd->eth.rx_discards);
935 	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
936 			   I40E_GLPRT_UPRCL(hw->port),
937 			   pf->stat_offsets_loaded,
938 			   &osd->eth.rx_unicast,
939 			   &nsd->eth.rx_unicast);
940 	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
941 			   I40E_GLPRT_MPRCL(hw->port),
942 			   pf->stat_offsets_loaded,
943 			   &osd->eth.rx_multicast,
944 			   &nsd->eth.rx_multicast);
945 	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
946 			   I40E_GLPRT_BPRCL(hw->port),
947 			   pf->stat_offsets_loaded,
948 			   &osd->eth.rx_broadcast,
949 			   &nsd->eth.rx_broadcast);
950 	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
951 			   I40E_GLPRT_UPTCL(hw->port),
952 			   pf->stat_offsets_loaded,
953 			   &osd->eth.tx_unicast,
954 			   &nsd->eth.tx_unicast);
955 	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
956 			   I40E_GLPRT_MPTCL(hw->port),
957 			   pf->stat_offsets_loaded,
958 			   &osd->eth.tx_multicast,
959 			   &nsd->eth.tx_multicast);
960 	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
961 			   I40E_GLPRT_BPTCL(hw->port),
962 			   pf->stat_offsets_loaded,
963 			   &osd->eth.tx_broadcast,
964 			   &nsd->eth.tx_broadcast);
965 
966 	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
967 			   pf->stat_offsets_loaded,
968 			   &osd->tx_dropped_link_down,
969 			   &nsd->tx_dropped_link_down);
970 
971 	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
972 			   pf->stat_offsets_loaded,
973 			   &osd->crc_errors, &nsd->crc_errors);
974 
975 	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
976 			   pf->stat_offsets_loaded,
977 			   &osd->illegal_bytes, &nsd->illegal_bytes);
978 
979 	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
980 			   pf->stat_offsets_loaded,
981 			   &osd->mac_local_faults,
982 			   &nsd->mac_local_faults);
983 	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
984 			   pf->stat_offsets_loaded,
985 			   &osd->mac_remote_faults,
986 			   &nsd->mac_remote_faults);
987 
988 	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
989 			   pf->stat_offsets_loaded,
990 			   &osd->rx_length_errors,
991 			   &nsd->rx_length_errors);
992 
993 	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
994 			   pf->stat_offsets_loaded,
995 			   &osd->link_xon_rx, &nsd->link_xon_rx);
996 	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
997 			   pf->stat_offsets_loaded,
998 			   &osd->link_xon_tx, &nsd->link_xon_tx);
999 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
1000 			   pf->stat_offsets_loaded,
1001 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
1002 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
1003 			   pf->stat_offsets_loaded,
1004 			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
1005 
1006 	for (i = 0; i < 8; i++) {
1007 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
1008 				   pf->stat_offsets_loaded,
1009 				   &osd->priority_xoff_rx[i],
1010 				   &nsd->priority_xoff_rx[i]);
1011 		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
1012 				   pf->stat_offsets_loaded,
1013 				   &osd->priority_xon_rx[i],
1014 				   &nsd->priority_xon_rx[i]);
1015 		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1016 				   pf->stat_offsets_loaded,
1017 				   &osd->priority_xon_tx[i],
1018 				   &nsd->priority_xon_tx[i]);
1019 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1020 				   pf->stat_offsets_loaded,
1021 				   &osd->priority_xoff_tx[i],
1022 				   &nsd->priority_xoff_tx[i]);
1023 		i40e_stat_update32(hw,
1024 				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1025 				   pf->stat_offsets_loaded,
1026 				   &osd->priority_xon_2_xoff[i],
1027 				   &nsd->priority_xon_2_xoff[i]);
1028 	}
1029 
1030 	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1031 			   I40E_GLPRT_PRC64L(hw->port),
1032 			   pf->stat_offsets_loaded,
1033 			   &osd->rx_size_64, &nsd->rx_size_64);
1034 	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1035 			   I40E_GLPRT_PRC127L(hw->port),
1036 			   pf->stat_offsets_loaded,
1037 			   &osd->rx_size_127, &nsd->rx_size_127);
1038 	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1039 			   I40E_GLPRT_PRC255L(hw->port),
1040 			   pf->stat_offsets_loaded,
1041 			   &osd->rx_size_255, &nsd->rx_size_255);
1042 	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1043 			   I40E_GLPRT_PRC511L(hw->port),
1044 			   pf->stat_offsets_loaded,
1045 			   &osd->rx_size_511, &nsd->rx_size_511);
1046 	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1047 			   I40E_GLPRT_PRC1023L(hw->port),
1048 			   pf->stat_offsets_loaded,
1049 			   &osd->rx_size_1023, &nsd->rx_size_1023);
1050 	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1051 			   I40E_GLPRT_PRC1522L(hw->port),
1052 			   pf->stat_offsets_loaded,
1053 			   &osd->rx_size_1522, &nsd->rx_size_1522);
1054 	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1055 			   I40E_GLPRT_PRC9522L(hw->port),
1056 			   pf->stat_offsets_loaded,
1057 			   &osd->rx_size_big, &nsd->rx_size_big);
1058 
1059 	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1060 			   I40E_GLPRT_PTC64L(hw->port),
1061 			   pf->stat_offsets_loaded,
1062 			   &osd->tx_size_64, &nsd->tx_size_64);
1063 	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1064 			   I40E_GLPRT_PTC127L(hw->port),
1065 			   pf->stat_offsets_loaded,
1066 			   &osd->tx_size_127, &nsd->tx_size_127);
1067 	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1068 			   I40E_GLPRT_PTC255L(hw->port),
1069 			   pf->stat_offsets_loaded,
1070 			   &osd->tx_size_255, &nsd->tx_size_255);
1071 	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1072 			   I40E_GLPRT_PTC511L(hw->port),
1073 			   pf->stat_offsets_loaded,
1074 			   &osd->tx_size_511, &nsd->tx_size_511);
1075 	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1076 			   I40E_GLPRT_PTC1023L(hw->port),
1077 			   pf->stat_offsets_loaded,
1078 			   &osd->tx_size_1023, &nsd->tx_size_1023);
1079 	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1080 			   I40E_GLPRT_PTC1522L(hw->port),
1081 			   pf->stat_offsets_loaded,
1082 			   &osd->tx_size_1522, &nsd->tx_size_1522);
1083 	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1084 			   I40E_GLPRT_PTC9522L(hw->port),
1085 			   pf->stat_offsets_loaded,
1086 			   &osd->tx_size_big, &nsd->tx_size_big);
1087 
1088 	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1089 			   pf->stat_offsets_loaded,
1090 			   &osd->rx_undersize, &nsd->rx_undersize);
1091 	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1092 			   pf->stat_offsets_loaded,
1093 			   &osd->rx_fragments, &nsd->rx_fragments);
1094 	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1095 			   pf->stat_offsets_loaded,
1096 			   &osd->rx_oversize, &nsd->rx_oversize);
1097 	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1098 			   pf->stat_offsets_loaded,
1099 			   &osd->rx_jabber, &nsd->rx_jabber);
1100 
1101 	/* FDIR stats */
1102 	i40e_stat_update32(hw,
1103 			   I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1104 			   pf->stat_offsets_loaded,
1105 			   &osd->fd_atr_match, &nsd->fd_atr_match);
1106 	i40e_stat_update32(hw,
1107 			   I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1108 			   pf->stat_offsets_loaded,
1109 			   &osd->fd_sb_match, &nsd->fd_sb_match);
1110 	i40e_stat_update32(hw,
1111 		      I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1112 		      pf->stat_offsets_loaded,
1113 		      &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1114 
1115 	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1116 	nsd->tx_lpi_status =
1117 		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1118 			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1119 	nsd->rx_lpi_status =
1120 		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1121 			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1122 	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1123 			   pf->stat_offsets_loaded,
1124 			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1125 	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1126 			   pf->stat_offsets_loaded,
1127 			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1128 
1129 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1130 	    !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1131 		nsd->fd_sb_status = true;
1132 	else
1133 		nsd->fd_sb_status = false;
1134 
1135 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1136 	    !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1137 		nsd->fd_atr_status = true;
1138 	else
1139 		nsd->fd_atr_status = false;
1140 
1141 	pf->stat_offsets_loaded = true;
1142 }
1143 
1144 /**
1145  * i40e_update_stats - Update the various statistics counters.
1146  * @vsi: the VSI to be updated
1147  *
1148  * Update the various stats for this VSI and its related entities.
1149  **/
1150 void i40e_update_stats(struct i40e_vsi *vsi)
1151 {
1152 	struct i40e_pf *pf = vsi->back;
1153 
1154 	if (vsi == pf->vsi[pf->lan_vsi])
1155 		i40e_update_pf_stats(pf);
1156 
1157 	i40e_update_vsi_stats(vsi);
1158 #ifdef I40E_FCOE
1159 	i40e_update_fcoe_stats(vsi);
1160 #endif
1161 }
1162 
1163 /**
1164  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1165  * @vsi: the VSI to be searched
1166  * @macaddr: the MAC address
1167  * @vlan: the vlan
1168  * @is_vf: make sure its a VF filter, else doesn't matter
1169  * @is_netdev: make sure its a netdev filter, else doesn't matter
1170  *
1171  * Returns ptr to the filter object or NULL
1172  **/
1173 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1174 						u8 *macaddr, s16 vlan,
1175 						bool is_vf, bool is_netdev)
1176 {
1177 	struct i40e_mac_filter *f;
1178 
1179 	if (!vsi || !macaddr)
1180 		return NULL;
1181 
1182 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1183 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1184 		    (vlan == f->vlan)    &&
1185 		    (!is_vf || f->is_vf) &&
1186 		    (!is_netdev || f->is_netdev))
1187 			return f;
1188 	}
1189 	return NULL;
1190 }
1191 
1192 /**
1193  * i40e_find_mac - Find a mac addr in the macvlan filters list
1194  * @vsi: the VSI to be searched
1195  * @macaddr: the MAC address we are searching for
1196  * @is_vf: make sure its a VF filter, else doesn't matter
1197  * @is_netdev: make sure its a netdev filter, else doesn't matter
1198  *
1199  * Returns the first filter with the provided MAC address or NULL if
1200  * MAC address was not found
1201  **/
1202 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1203 				      bool is_vf, bool is_netdev)
1204 {
1205 	struct i40e_mac_filter *f;
1206 
1207 	if (!vsi || !macaddr)
1208 		return NULL;
1209 
1210 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1211 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1212 		    (!is_vf || f->is_vf) &&
1213 		    (!is_netdev || f->is_netdev))
1214 			return f;
1215 	}
1216 	return NULL;
1217 }
1218 
1219 /**
1220  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1221  * @vsi: the VSI to be searched
1222  *
1223  * Returns true if VSI is in vlan mode or false otherwise
1224  **/
1225 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1226 {
1227 	struct i40e_mac_filter *f;
1228 
1229 	/* Only -1 for all the filters denotes not in vlan mode
1230 	 * so we have to go through all the list in order to make sure
1231 	 */
1232 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1233 		if (f->vlan >= 0 || vsi->info.pvid)
1234 			return true;
1235 	}
1236 
1237 	return false;
1238 }
1239 
1240 /**
1241  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1242  * @vsi: the VSI to be searched
1243  * @macaddr: the mac address to be filtered
1244  * @is_vf: true if it is a VF
1245  * @is_netdev: true if it is a netdev
1246  *
1247  * Goes through all the macvlan filters and adds a
1248  * macvlan filter for each unique vlan that already exists
1249  *
1250  * Returns first filter found on success, else NULL
1251  **/
1252 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1253 					     bool is_vf, bool is_netdev)
1254 {
1255 	struct i40e_mac_filter *f;
1256 
1257 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1258 		if (vsi->info.pvid)
1259 			f->vlan = le16_to_cpu(vsi->info.pvid);
1260 		if (!i40e_find_filter(vsi, macaddr, f->vlan,
1261 				      is_vf, is_netdev)) {
1262 			if (!i40e_add_filter(vsi, macaddr, f->vlan,
1263 					     is_vf, is_netdev))
1264 				return NULL;
1265 		}
1266 	}
1267 
1268 	return list_first_entry_or_null(&vsi->mac_filter_list,
1269 					struct i40e_mac_filter, list);
1270 }
1271 
1272 /**
1273  * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1274  * @vsi: the VSI to be searched
1275  * @macaddr: the mac address to be removed
1276  * @is_vf: true if it is a VF
1277  * @is_netdev: true if it is a netdev
1278  *
1279  * Removes a given MAC address from a VSI, regardless of VLAN
1280  *
1281  * Returns 0 for success, or error
1282  **/
1283 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1284 			  bool is_vf, bool is_netdev)
1285 {
1286 	struct i40e_mac_filter *f = NULL;
1287 	int changed = 0;
1288 
1289 	WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1290 	     "Missing mac_filter_list_lock\n");
1291 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1292 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1293 		    (is_vf == f->is_vf) &&
1294 		    (is_netdev == f->is_netdev)) {
1295 			f->counter--;
1296 			f->changed = true;
1297 			changed = 1;
1298 		}
1299 	}
1300 	if (changed) {
1301 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1302 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1303 		return 0;
1304 	}
1305 	return -ENOENT;
1306 }
1307 
1308 /**
1309  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1310  * @vsi: the PF Main VSI - inappropriate for any other VSI
1311  * @macaddr: the MAC address
1312  *
1313  * Some older firmware configurations set up a default promiscuous VLAN
1314  * filter that needs to be removed.
1315  **/
1316 static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1317 {
1318 	struct i40e_aqc_remove_macvlan_element_data element;
1319 	struct i40e_pf *pf = vsi->back;
1320 	i40e_status ret;
1321 
1322 	/* Only appropriate for the PF main VSI */
1323 	if (vsi->type != I40E_VSI_MAIN)
1324 		return -EINVAL;
1325 
1326 	memset(&element, 0, sizeof(element));
1327 	ether_addr_copy(element.mac_addr, macaddr);
1328 	element.vlan_tag = 0;
1329 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1330 			I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1331 	ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1332 	if (ret)
1333 		return -ENOENT;
1334 
1335 	return 0;
1336 }
1337 
1338 /**
1339  * i40e_add_filter - Add a mac/vlan filter to the VSI
1340  * @vsi: the VSI to be searched
1341  * @macaddr: the MAC address
1342  * @vlan: the vlan
1343  * @is_vf: make sure its a VF filter, else doesn't matter
1344  * @is_netdev: make sure its a netdev filter, else doesn't matter
1345  *
1346  * Returns ptr to the filter object or NULL when no memory available.
1347  *
1348  * NOTE: This function is expected to be called with mac_filter_list_lock
1349  * being held.
1350  **/
1351 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1352 					u8 *macaddr, s16 vlan,
1353 					bool is_vf, bool is_netdev)
1354 {
1355 	struct i40e_mac_filter *f;
1356 
1357 	if (!vsi || !macaddr)
1358 		return NULL;
1359 
1360 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1361 	if (!f) {
1362 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1363 		if (!f)
1364 			goto add_filter_out;
1365 
1366 		ether_addr_copy(f->macaddr, macaddr);
1367 		f->vlan = vlan;
1368 		f->changed = true;
1369 
1370 		INIT_LIST_HEAD(&f->list);
1371 		list_add(&f->list, &vsi->mac_filter_list);
1372 	}
1373 
1374 	/* increment counter and add a new flag if needed */
1375 	if (is_vf) {
1376 		if (!f->is_vf) {
1377 			f->is_vf = true;
1378 			f->counter++;
1379 		}
1380 	} else if (is_netdev) {
1381 		if (!f->is_netdev) {
1382 			f->is_netdev = true;
1383 			f->counter++;
1384 		}
1385 	} else {
1386 		f->counter++;
1387 	}
1388 
1389 	/* changed tells sync_filters_subtask to
1390 	 * push the filter down to the firmware
1391 	 */
1392 	if (f->changed) {
1393 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1394 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1395 	}
1396 
1397 add_filter_out:
1398 	return f;
1399 }
1400 
1401 /**
1402  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1403  * @vsi: the VSI to be searched
1404  * @macaddr: the MAC address
1405  * @vlan: the vlan
1406  * @is_vf: make sure it's a VF filter, else doesn't matter
1407  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1408  *
1409  * NOTE: This function is expected to be called with mac_filter_list_lock
1410  * being held.
1411  **/
1412 void i40e_del_filter(struct i40e_vsi *vsi,
1413 		     u8 *macaddr, s16 vlan,
1414 		     bool is_vf, bool is_netdev)
1415 {
1416 	struct i40e_mac_filter *f;
1417 
1418 	if (!vsi || !macaddr)
1419 		return;
1420 
1421 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1422 	if (!f || f->counter == 0)
1423 		return;
1424 
1425 	if (is_vf) {
1426 		if (f->is_vf) {
1427 			f->is_vf = false;
1428 			f->counter--;
1429 		}
1430 	} else if (is_netdev) {
1431 		if (f->is_netdev) {
1432 			f->is_netdev = false;
1433 			f->counter--;
1434 		}
1435 	} else {
1436 		/* make sure we don't remove a filter in use by VF or netdev */
1437 		int min_f = 0;
1438 
1439 		min_f += (f->is_vf ? 1 : 0);
1440 		min_f += (f->is_netdev ? 1 : 0);
1441 
1442 		if (f->counter > min_f)
1443 			f->counter--;
1444 	}
1445 
1446 	/* counter == 0 tells sync_filters_subtask to
1447 	 * remove the filter from the firmware's list
1448 	 */
1449 	if (f->counter == 0) {
1450 		f->changed = true;
1451 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1452 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1453 	}
1454 }
1455 
1456 /**
1457  * i40e_set_mac - NDO callback to set mac address
1458  * @netdev: network interface device structure
1459  * @p: pointer to an address structure
1460  *
1461  * Returns 0 on success, negative on failure
1462  **/
1463 #ifdef I40E_FCOE
1464 int i40e_set_mac(struct net_device *netdev, void *p)
1465 #else
1466 static int i40e_set_mac(struct net_device *netdev, void *p)
1467 #endif
1468 {
1469 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1470 	struct i40e_vsi *vsi = np->vsi;
1471 	struct i40e_pf *pf = vsi->back;
1472 	struct i40e_hw *hw = &pf->hw;
1473 	struct sockaddr *addr = p;
1474 	struct i40e_mac_filter *f;
1475 
1476 	if (!is_valid_ether_addr(addr->sa_data))
1477 		return -EADDRNOTAVAIL;
1478 
1479 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1480 		netdev_info(netdev, "already using mac address %pM\n",
1481 			    addr->sa_data);
1482 		return 0;
1483 	}
1484 
1485 	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1486 	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1487 		return -EADDRNOTAVAIL;
1488 
1489 	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1490 		netdev_info(netdev, "returning to hw mac address %pM\n",
1491 			    hw->mac.addr);
1492 	else
1493 		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1494 
1495 	if (vsi->type == I40E_VSI_MAIN) {
1496 		i40e_status ret;
1497 
1498 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1499 						I40E_AQC_WRITE_TYPE_LAA_WOL,
1500 						addr->sa_data, NULL);
1501 		if (ret) {
1502 			netdev_info(netdev,
1503 				    "Addr change for Main VSI failed: %d\n",
1504 				    ret);
1505 			return -EADDRNOTAVAIL;
1506 		}
1507 	}
1508 
1509 	if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1510 		struct i40e_aqc_remove_macvlan_element_data element;
1511 
1512 		memset(&element, 0, sizeof(element));
1513 		ether_addr_copy(element.mac_addr, netdev->dev_addr);
1514 		element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1515 		i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1516 	} else {
1517 		spin_lock_bh(&vsi->mac_filter_list_lock);
1518 		i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1519 				false, false);
1520 		spin_unlock_bh(&vsi->mac_filter_list_lock);
1521 	}
1522 
1523 	if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1524 		struct i40e_aqc_add_macvlan_element_data element;
1525 
1526 		memset(&element, 0, sizeof(element));
1527 		ether_addr_copy(element.mac_addr, hw->mac.addr);
1528 		element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1529 		i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1530 	} else {
1531 		spin_lock_bh(&vsi->mac_filter_list_lock);
1532 		f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1533 				    false, false);
1534 		if (f)
1535 			f->is_laa = true;
1536 		spin_unlock_bh(&vsi->mac_filter_list_lock);
1537 	}
1538 
1539 	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1540 
1541 	return i40e_sync_vsi_filters(vsi);
1542 }
1543 
1544 /**
1545  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1546  * @vsi: the VSI being setup
1547  * @ctxt: VSI context structure
1548  * @enabled_tc: Enabled TCs bitmap
1549  * @is_add: True if called before Add VSI
1550  *
1551  * Setup VSI queue mapping for enabled traffic classes.
1552  **/
1553 #ifdef I40E_FCOE
1554 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1555 			      struct i40e_vsi_context *ctxt,
1556 			      u8 enabled_tc,
1557 			      bool is_add)
1558 #else
1559 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1560 				     struct i40e_vsi_context *ctxt,
1561 				     u8 enabled_tc,
1562 				     bool is_add)
1563 #endif
1564 {
1565 	struct i40e_pf *pf = vsi->back;
1566 	u16 sections = 0;
1567 	u8 netdev_tc = 0;
1568 	u16 numtc = 0;
1569 	u16 qcount;
1570 	u8 offset;
1571 	u16 qmap;
1572 	int i;
1573 	u16 num_tc_qps = 0;
1574 
1575 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1576 	offset = 0;
1577 
1578 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1579 		/* Find numtc from enabled TC bitmap */
1580 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1581 			if (enabled_tc & BIT(i)) /* TC is enabled */
1582 				numtc++;
1583 		}
1584 		if (!numtc) {
1585 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1586 			numtc = 1;
1587 		}
1588 	} else {
1589 		/* At least TC0 is enabled in case of non-DCB case */
1590 		numtc = 1;
1591 	}
1592 
1593 	vsi->tc_config.numtc = numtc;
1594 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1595 	/* Number of queues per enabled TC */
1596 	/* In MFP case we can have a much lower count of MSIx
1597 	 * vectors available and so we need to lower the used
1598 	 * q count.
1599 	 */
1600 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1601 		qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1602 	else
1603 		qcount = vsi->alloc_queue_pairs;
1604 	num_tc_qps = qcount / numtc;
1605 	num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1606 
1607 	/* Setup queue offset/count for all TCs for given VSI */
1608 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1609 		/* See if the given TC is enabled for the given VSI */
1610 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1611 			/* TC is enabled */
1612 			int pow, num_qps;
1613 
1614 			switch (vsi->type) {
1615 			case I40E_VSI_MAIN:
1616 				qcount = min_t(int, pf->alloc_rss_size,
1617 					       num_tc_qps);
1618 				break;
1619 #ifdef I40E_FCOE
1620 			case I40E_VSI_FCOE:
1621 				qcount = num_tc_qps;
1622 				break;
1623 #endif
1624 			case I40E_VSI_FDIR:
1625 			case I40E_VSI_SRIOV:
1626 			case I40E_VSI_VMDQ2:
1627 			default:
1628 				qcount = num_tc_qps;
1629 				WARN_ON(i != 0);
1630 				break;
1631 			}
1632 			vsi->tc_config.tc_info[i].qoffset = offset;
1633 			vsi->tc_config.tc_info[i].qcount = qcount;
1634 
1635 			/* find the next higher power-of-2 of num queue pairs */
1636 			num_qps = qcount;
1637 			pow = 0;
1638 			while (num_qps && (BIT_ULL(pow) < qcount)) {
1639 				pow++;
1640 				num_qps >>= 1;
1641 			}
1642 
1643 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1644 			qmap =
1645 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1646 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1647 
1648 			offset += qcount;
1649 		} else {
1650 			/* TC is not enabled so set the offset to
1651 			 * default queue and allocate one queue
1652 			 * for the given TC.
1653 			 */
1654 			vsi->tc_config.tc_info[i].qoffset = 0;
1655 			vsi->tc_config.tc_info[i].qcount = 1;
1656 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1657 
1658 			qmap = 0;
1659 		}
1660 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1661 	}
1662 
1663 	/* Set actual Tx/Rx queue pairs */
1664 	vsi->num_queue_pairs = offset;
1665 	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1666 		if (vsi->req_queue_pairs > 0)
1667 			vsi->num_queue_pairs = vsi->req_queue_pairs;
1668 		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1669 			vsi->num_queue_pairs = pf->num_lan_msix;
1670 	}
1671 
1672 	/* Scheduler section valid can only be set for ADD VSI */
1673 	if (is_add) {
1674 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1675 
1676 		ctxt->info.up_enable_bits = enabled_tc;
1677 	}
1678 	if (vsi->type == I40E_VSI_SRIOV) {
1679 		ctxt->info.mapping_flags |=
1680 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1681 		for (i = 0; i < vsi->num_queue_pairs; i++)
1682 			ctxt->info.queue_mapping[i] =
1683 					       cpu_to_le16(vsi->base_queue + i);
1684 	} else {
1685 		ctxt->info.mapping_flags |=
1686 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1687 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1688 	}
1689 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1690 }
1691 
1692 /**
1693  * i40e_set_rx_mode - NDO callback to set the netdev filters
1694  * @netdev: network interface device structure
1695  **/
1696 #ifdef I40E_FCOE
1697 void i40e_set_rx_mode(struct net_device *netdev)
1698 #else
1699 static void i40e_set_rx_mode(struct net_device *netdev)
1700 #endif
1701 {
1702 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1703 	struct i40e_mac_filter *f, *ftmp;
1704 	struct i40e_vsi *vsi = np->vsi;
1705 	struct netdev_hw_addr *uca;
1706 	struct netdev_hw_addr *mca;
1707 	struct netdev_hw_addr *ha;
1708 
1709 	spin_lock_bh(&vsi->mac_filter_list_lock);
1710 
1711 	/* add addr if not already in the filter list */
1712 	netdev_for_each_uc_addr(uca, netdev) {
1713 		if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1714 			if (i40e_is_vsi_in_vlan(vsi))
1715 				i40e_put_mac_in_vlan(vsi, uca->addr,
1716 						     false, true);
1717 			else
1718 				i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1719 						false, true);
1720 		}
1721 	}
1722 
1723 	netdev_for_each_mc_addr(mca, netdev) {
1724 		if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1725 			if (i40e_is_vsi_in_vlan(vsi))
1726 				i40e_put_mac_in_vlan(vsi, mca->addr,
1727 						     false, true);
1728 			else
1729 				i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1730 						false, true);
1731 		}
1732 	}
1733 
1734 	/* remove filter if not in netdev list */
1735 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1736 
1737 		if (!f->is_netdev)
1738 			continue;
1739 
1740 		netdev_for_each_mc_addr(mca, netdev)
1741 			if (ether_addr_equal(mca->addr, f->macaddr))
1742 				goto bottom_of_search_loop;
1743 
1744 		netdev_for_each_uc_addr(uca, netdev)
1745 			if (ether_addr_equal(uca->addr, f->macaddr))
1746 				goto bottom_of_search_loop;
1747 
1748 		for_each_dev_addr(netdev, ha)
1749 			if (ether_addr_equal(ha->addr, f->macaddr))
1750 				goto bottom_of_search_loop;
1751 
1752 		/* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1753 		i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1754 
1755 bottom_of_search_loop:
1756 		continue;
1757 	}
1758 	spin_unlock_bh(&vsi->mac_filter_list_lock);
1759 
1760 	/* check for other flag changes */
1761 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1762 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1763 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1764 	}
1765 }
1766 
1767 /**
1768  * i40e_mac_filter_entry_clone - Clones a MAC filter entry
1769  * @src: source MAC filter entry to be clones
1770  *
1771  * Returns the pointer to newly cloned MAC filter entry or NULL
1772  * in case of error
1773  **/
1774 static struct i40e_mac_filter *i40e_mac_filter_entry_clone(
1775 					struct i40e_mac_filter *src)
1776 {
1777 	struct i40e_mac_filter *f;
1778 
1779 	f = kzalloc(sizeof(*f), GFP_ATOMIC);
1780 	if (!f)
1781 		return NULL;
1782 	*f = *src;
1783 
1784 	INIT_LIST_HEAD(&f->list);
1785 
1786 	return f;
1787 }
1788 
1789 /**
1790  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1791  * @vsi: pointer to vsi struct
1792  * @from: Pointer to list which contains MAC filter entries - changes to
1793  *        those entries needs to be undone.
1794  *
1795  * MAC filter entries from list were slated to be removed from device.
1796  **/
1797 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1798 					 struct list_head *from)
1799 {
1800 	struct i40e_mac_filter *f, *ftmp;
1801 
1802 	list_for_each_entry_safe(f, ftmp, from, list) {
1803 		f->changed = true;
1804 		/* Move the element back into MAC filter list*/
1805 		list_move_tail(&f->list, &vsi->mac_filter_list);
1806 	}
1807 }
1808 
1809 /**
1810  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1811  * @vsi: pointer to vsi struct
1812  *
1813  * MAC filter entries from list were slated to be added from device.
1814  **/
1815 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi)
1816 {
1817 	struct i40e_mac_filter *f, *ftmp;
1818 
1819 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1820 		if (!f->changed && f->counter)
1821 			f->changed = true;
1822 	}
1823 }
1824 
1825 /**
1826  * i40e_cleanup_add_list - Deletes the element from add list and release
1827  *			memory
1828  * @add_list: Pointer to list which contains MAC filter entries
1829  **/
1830 static void i40e_cleanup_add_list(struct list_head *add_list)
1831 {
1832 	struct i40e_mac_filter *f, *ftmp;
1833 
1834 	list_for_each_entry_safe(f, ftmp, add_list, list) {
1835 		list_del(&f->list);
1836 		kfree(f);
1837 	}
1838 }
1839 
1840 /**
1841  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1842  * @vsi: ptr to the VSI
1843  *
1844  * Push any outstanding VSI filter changes through the AdminQ.
1845  *
1846  * Returns 0 or error value
1847  **/
1848 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1849 {
1850 	struct list_head tmp_del_list, tmp_add_list;
1851 	struct i40e_mac_filter *f, *ftmp, *fclone;
1852 	bool promisc_forced_on = false;
1853 	bool add_happened = false;
1854 	int filter_list_len = 0;
1855 	u32 changed_flags = 0;
1856 	i40e_status aq_ret = 0;
1857 	bool err_cond = false;
1858 	int retval = 0;
1859 	struct i40e_pf *pf;
1860 	int num_add = 0;
1861 	int num_del = 0;
1862 	int aq_err = 0;
1863 	u16 cmd_flags;
1864 
1865 	/* empty array typed pointers, kcalloc later */
1866 	struct i40e_aqc_add_macvlan_element_data *add_list;
1867 	struct i40e_aqc_remove_macvlan_element_data *del_list;
1868 
1869 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1870 		usleep_range(1000, 2000);
1871 	pf = vsi->back;
1872 
1873 	if (vsi->netdev) {
1874 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1875 		vsi->current_netdev_flags = vsi->netdev->flags;
1876 	}
1877 
1878 	INIT_LIST_HEAD(&tmp_del_list);
1879 	INIT_LIST_HEAD(&tmp_add_list);
1880 
1881 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1882 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1883 
1884 		spin_lock_bh(&vsi->mac_filter_list_lock);
1885 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1886 			if (!f->changed)
1887 				continue;
1888 
1889 			if (f->counter != 0)
1890 				continue;
1891 			f->changed = false;
1892 
1893 			/* Move the element into temporary del_list */
1894 			list_move_tail(&f->list, &tmp_del_list);
1895 		}
1896 
1897 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1898 			if (!f->changed)
1899 				continue;
1900 
1901 			if (f->counter == 0)
1902 				continue;
1903 			f->changed = false;
1904 
1905 			/* Clone MAC filter entry and add into temporary list */
1906 			fclone = i40e_mac_filter_entry_clone(f);
1907 			if (!fclone) {
1908 				err_cond = true;
1909 				break;
1910 			}
1911 			list_add_tail(&fclone->list, &tmp_add_list);
1912 		}
1913 
1914 		/* if failed to clone MAC filter entry - undo */
1915 		if (err_cond) {
1916 			i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1917 			i40e_undo_add_filter_entries(vsi);
1918 		}
1919 		spin_unlock_bh(&vsi->mac_filter_list_lock);
1920 
1921 		if (err_cond) {
1922 			i40e_cleanup_add_list(&tmp_add_list);
1923 			retval = -ENOMEM;
1924 			goto out;
1925 		}
1926 	}
1927 
1928 	/* Now process 'del_list' outside the lock */
1929 	if (!list_empty(&tmp_del_list)) {
1930 		int del_list_size;
1931 
1932 		filter_list_len = pf->hw.aq.asq_buf_size /
1933 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
1934 		del_list_size = filter_list_len *
1935 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
1936 		del_list = kzalloc(del_list_size, GFP_KERNEL);
1937 		if (!del_list) {
1938 			i40e_cleanup_add_list(&tmp_add_list);
1939 
1940 			/* Undo VSI's MAC filter entry element updates */
1941 			spin_lock_bh(&vsi->mac_filter_list_lock);
1942 			i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1943 			i40e_undo_add_filter_entries(vsi);
1944 			spin_unlock_bh(&vsi->mac_filter_list_lock);
1945 			retval = -ENOMEM;
1946 			goto out;
1947 		}
1948 
1949 		list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1950 			cmd_flags = 0;
1951 
1952 			/* add to delete list */
1953 			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1954 			del_list[num_del].vlan_tag =
1955 				cpu_to_le16((u16)(f->vlan ==
1956 					    I40E_VLAN_ANY ? 0 : f->vlan));
1957 
1958 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1959 			del_list[num_del].flags = cmd_flags;
1960 			num_del++;
1961 
1962 			/* flush a full buffer */
1963 			if (num_del == filter_list_len) {
1964 				aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1965 								vsi->seid,
1966 								del_list,
1967 								num_del,
1968 								NULL);
1969 				aq_err = pf->hw.aq.asq_last_status;
1970 				num_del = 0;
1971 				memset(del_list, 0, del_list_size);
1972 
1973 				if (aq_ret && aq_err != I40E_AQ_RC_ENOENT) {
1974 					retval = -EIO;
1975 					dev_err(&pf->pdev->dev,
1976 						"ignoring delete macvlan error, err %s, aq_err %s while flushing a full buffer\n",
1977 						i40e_stat_str(&pf->hw, aq_ret),
1978 						i40e_aq_str(&pf->hw, aq_err));
1979 				}
1980 			}
1981 			/* Release memory for MAC filter entries which were
1982 			 * synced up with HW.
1983 			 */
1984 			list_del(&f->list);
1985 			kfree(f);
1986 		}
1987 
1988 		if (num_del) {
1989 			aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1990 							del_list, num_del,
1991 							NULL);
1992 			aq_err = pf->hw.aq.asq_last_status;
1993 			num_del = 0;
1994 
1995 			if (aq_ret && aq_err != I40E_AQ_RC_ENOENT)
1996 				dev_info(&pf->pdev->dev,
1997 					 "ignoring delete macvlan error, err %s aq_err %s\n",
1998 					 i40e_stat_str(&pf->hw, aq_ret),
1999 					 i40e_aq_str(&pf->hw, aq_err));
2000 		}
2001 
2002 		kfree(del_list);
2003 		del_list = NULL;
2004 	}
2005 
2006 	if (!list_empty(&tmp_add_list)) {
2007 		int add_list_size;
2008 
2009 		/* do all the adds now */
2010 		filter_list_len = pf->hw.aq.asq_buf_size /
2011 			       sizeof(struct i40e_aqc_add_macvlan_element_data),
2012 		add_list_size = filter_list_len *
2013 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2014 		add_list = kzalloc(add_list_size, GFP_KERNEL);
2015 		if (!add_list) {
2016 			/* Purge element from temporary lists */
2017 			i40e_cleanup_add_list(&tmp_add_list);
2018 
2019 			/* Undo add filter entries from VSI MAC filter list */
2020 			spin_lock_bh(&vsi->mac_filter_list_lock);
2021 			i40e_undo_add_filter_entries(vsi);
2022 			spin_unlock_bh(&vsi->mac_filter_list_lock);
2023 			retval = -ENOMEM;
2024 			goto out;
2025 		}
2026 
2027 		list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2028 
2029 			add_happened = true;
2030 			cmd_flags = 0;
2031 
2032 			/* add to add array */
2033 			ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
2034 			add_list[num_add].vlan_tag =
2035 				cpu_to_le16(
2036 				 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
2037 			add_list[num_add].queue_number = 0;
2038 
2039 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2040 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
2041 			num_add++;
2042 
2043 			/* flush a full buffer */
2044 			if (num_add == filter_list_len) {
2045 				aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2046 							     add_list, num_add,
2047 							     NULL);
2048 				aq_err = pf->hw.aq.asq_last_status;
2049 				num_add = 0;
2050 
2051 				if (aq_ret)
2052 					break;
2053 				memset(add_list, 0, add_list_size);
2054 			}
2055 			/* Entries from tmp_add_list were cloned from MAC
2056 			 * filter list, hence clean those cloned entries
2057 			 */
2058 			list_del(&f->list);
2059 			kfree(f);
2060 		}
2061 
2062 		if (num_add) {
2063 			aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2064 						     add_list, num_add, NULL);
2065 			aq_err = pf->hw.aq.asq_last_status;
2066 			num_add = 0;
2067 		}
2068 		kfree(add_list);
2069 		add_list = NULL;
2070 
2071 		if (add_happened && aq_ret && aq_err != I40E_AQ_RC_EINVAL) {
2072 			retval = i40e_aq_rc_to_posix(aq_ret, aq_err);
2073 			dev_info(&pf->pdev->dev,
2074 				 "add filter failed, err %s aq_err %s\n",
2075 				 i40e_stat_str(&pf->hw, aq_ret),
2076 				 i40e_aq_str(&pf->hw, aq_err));
2077 			if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
2078 			    !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2079 				      &vsi->state)) {
2080 				promisc_forced_on = true;
2081 				set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2082 					&vsi->state);
2083 				dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
2084 			}
2085 		}
2086 	}
2087 
2088 	/* check for changes in promiscuous modes */
2089 	if (changed_flags & IFF_ALLMULTI) {
2090 		bool cur_multipromisc;
2091 
2092 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2093 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2094 							       vsi->seid,
2095 							       cur_multipromisc,
2096 							       NULL);
2097 		if (aq_ret) {
2098 			retval = i40e_aq_rc_to_posix(aq_ret,
2099 						     pf->hw.aq.asq_last_status);
2100 			dev_info(&pf->pdev->dev,
2101 				 "set multi promisc failed, err %s aq_err %s\n",
2102 				 i40e_stat_str(&pf->hw, aq_ret),
2103 				 i40e_aq_str(&pf->hw,
2104 					     pf->hw.aq.asq_last_status));
2105 		}
2106 	}
2107 	if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
2108 		bool cur_promisc;
2109 
2110 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2111 			       test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2112 					&vsi->state));
2113 		if (vsi->type == I40E_VSI_MAIN && pf->lan_veb != I40E_NO_VEB) {
2114 			/* set defport ON for Main VSI instead of true promisc
2115 			 * this way we will get all unicast/multicast and VLAN
2116 			 * promisc behavior but will not get VF or VMDq traffic
2117 			 * replicated on the Main VSI.
2118 			 */
2119 			if (pf->cur_promisc != cur_promisc) {
2120 				pf->cur_promisc = cur_promisc;
2121 				set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2122 			}
2123 		} else {
2124 			aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2125 							  &vsi->back->hw,
2126 							  vsi->seid,
2127 							  cur_promisc, NULL);
2128 			if (aq_ret) {
2129 				retval =
2130 				i40e_aq_rc_to_posix(aq_ret,
2131 						    pf->hw.aq.asq_last_status);
2132 				dev_info(&pf->pdev->dev,
2133 					 "set unicast promisc failed, err %d, aq_err %d\n",
2134 					 aq_ret, pf->hw.aq.asq_last_status);
2135 			}
2136 			aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2137 							  &vsi->back->hw,
2138 							  vsi->seid,
2139 							  cur_promisc, NULL);
2140 			if (aq_ret) {
2141 				retval =
2142 				i40e_aq_rc_to_posix(aq_ret,
2143 						    pf->hw.aq.asq_last_status);
2144 				dev_info(&pf->pdev->dev,
2145 					 "set multicast promisc failed, err %d, aq_err %d\n",
2146 					 aq_ret, pf->hw.aq.asq_last_status);
2147 			}
2148 		}
2149 		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2150 						   vsi->seid,
2151 						   cur_promisc, NULL);
2152 		if (aq_ret) {
2153 			retval = i40e_aq_rc_to_posix(aq_ret,
2154 						     pf->hw.aq.asq_last_status);
2155 			dev_info(&pf->pdev->dev,
2156 				 "set brdcast promisc failed, err %s, aq_err %s\n",
2157 				 i40e_stat_str(&pf->hw, aq_ret),
2158 				 i40e_aq_str(&pf->hw,
2159 					     pf->hw.aq.asq_last_status));
2160 		}
2161 	}
2162 out:
2163 	clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2164 	return retval;
2165 }
2166 
2167 /**
2168  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2169  * @pf: board private structure
2170  **/
2171 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2172 {
2173 	int v;
2174 
2175 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2176 		return;
2177 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2178 
2179 	for (v = 0; v < pf->num_alloc_vsi; v++) {
2180 		if (pf->vsi[v] &&
2181 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2182 			int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2183 
2184 			if (ret) {
2185 				/* come back and try again later */
2186 				pf->flags |= I40E_FLAG_FILTER_SYNC;
2187 				break;
2188 			}
2189 		}
2190 	}
2191 }
2192 
2193 /**
2194  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2195  * @netdev: network interface device structure
2196  * @new_mtu: new value for maximum frame size
2197  *
2198  * Returns 0 on success, negative on failure
2199  **/
2200 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2201 {
2202 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2203 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2204 	struct i40e_vsi *vsi = np->vsi;
2205 
2206 	/* MTU < 68 is an error and causes problems on some kernels */
2207 	if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2208 		return -EINVAL;
2209 
2210 	netdev_info(netdev, "changing MTU from %d to %d\n",
2211 		    netdev->mtu, new_mtu);
2212 	netdev->mtu = new_mtu;
2213 	if (netif_running(netdev))
2214 		i40e_vsi_reinit_locked(vsi);
2215 
2216 	return 0;
2217 }
2218 
2219 /**
2220  * i40e_ioctl - Access the hwtstamp interface
2221  * @netdev: network interface device structure
2222  * @ifr: interface request data
2223  * @cmd: ioctl command
2224  **/
2225 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2226 {
2227 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2228 	struct i40e_pf *pf = np->vsi->back;
2229 
2230 	switch (cmd) {
2231 	case SIOCGHWTSTAMP:
2232 		return i40e_ptp_get_ts_config(pf, ifr);
2233 	case SIOCSHWTSTAMP:
2234 		return i40e_ptp_set_ts_config(pf, ifr);
2235 	default:
2236 		return -EOPNOTSUPP;
2237 	}
2238 }
2239 
2240 /**
2241  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2242  * @vsi: the vsi being adjusted
2243  **/
2244 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2245 {
2246 	struct i40e_vsi_context ctxt;
2247 	i40e_status ret;
2248 
2249 	if ((vsi->info.valid_sections &
2250 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2251 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2252 		return;  /* already enabled */
2253 
2254 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2255 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2256 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2257 
2258 	ctxt.seid = vsi->seid;
2259 	ctxt.info = vsi->info;
2260 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2261 	if (ret) {
2262 		dev_info(&vsi->back->pdev->dev,
2263 			 "update vlan stripping failed, err %s aq_err %s\n",
2264 			 i40e_stat_str(&vsi->back->hw, ret),
2265 			 i40e_aq_str(&vsi->back->hw,
2266 				     vsi->back->hw.aq.asq_last_status));
2267 	}
2268 }
2269 
2270 /**
2271  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2272  * @vsi: the vsi being adjusted
2273  **/
2274 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2275 {
2276 	struct i40e_vsi_context ctxt;
2277 	i40e_status ret;
2278 
2279 	if ((vsi->info.valid_sections &
2280 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2281 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2282 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2283 		return;  /* already disabled */
2284 
2285 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2286 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2287 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2288 
2289 	ctxt.seid = vsi->seid;
2290 	ctxt.info = vsi->info;
2291 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2292 	if (ret) {
2293 		dev_info(&vsi->back->pdev->dev,
2294 			 "update vlan stripping failed, err %s aq_err %s\n",
2295 			 i40e_stat_str(&vsi->back->hw, ret),
2296 			 i40e_aq_str(&vsi->back->hw,
2297 				     vsi->back->hw.aq.asq_last_status));
2298 	}
2299 }
2300 
2301 /**
2302  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2303  * @netdev: network interface to be adjusted
2304  * @features: netdev features to test if VLAN offload is enabled or not
2305  **/
2306 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2307 {
2308 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2309 	struct i40e_vsi *vsi = np->vsi;
2310 
2311 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2312 		i40e_vlan_stripping_enable(vsi);
2313 	else
2314 		i40e_vlan_stripping_disable(vsi);
2315 }
2316 
2317 /**
2318  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2319  * @vsi: the vsi being configured
2320  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2321  **/
2322 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2323 {
2324 	struct i40e_mac_filter *f, *add_f;
2325 	bool is_netdev, is_vf;
2326 
2327 	is_vf = (vsi->type == I40E_VSI_SRIOV);
2328 	is_netdev = !!(vsi->netdev);
2329 
2330 	/* Locked once because all functions invoked below iterates list*/
2331 	spin_lock_bh(&vsi->mac_filter_list_lock);
2332 
2333 	if (is_netdev) {
2334 		add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2335 					is_vf, is_netdev);
2336 		if (!add_f) {
2337 			dev_info(&vsi->back->pdev->dev,
2338 				 "Could not add vlan filter %d for %pM\n",
2339 				 vid, vsi->netdev->dev_addr);
2340 			spin_unlock_bh(&vsi->mac_filter_list_lock);
2341 			return -ENOMEM;
2342 		}
2343 	}
2344 
2345 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
2346 		add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2347 		if (!add_f) {
2348 			dev_info(&vsi->back->pdev->dev,
2349 				 "Could not add vlan filter %d for %pM\n",
2350 				 vid, f->macaddr);
2351 			spin_unlock_bh(&vsi->mac_filter_list_lock);
2352 			return -ENOMEM;
2353 		}
2354 	}
2355 
2356 	/* Now if we add a vlan tag, make sure to check if it is the first
2357 	 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2358 	 * with 0, so we now accept untagged and specified tagged traffic
2359 	 * (and not any taged and untagged)
2360 	 */
2361 	if (vid > 0) {
2362 		if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2363 						  I40E_VLAN_ANY,
2364 						  is_vf, is_netdev)) {
2365 			i40e_del_filter(vsi, vsi->netdev->dev_addr,
2366 					I40E_VLAN_ANY, is_vf, is_netdev);
2367 			add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2368 						is_vf, is_netdev);
2369 			if (!add_f) {
2370 				dev_info(&vsi->back->pdev->dev,
2371 					 "Could not add filter 0 for %pM\n",
2372 					 vsi->netdev->dev_addr);
2373 				spin_unlock_bh(&vsi->mac_filter_list_lock);
2374 				return -ENOMEM;
2375 			}
2376 		}
2377 	}
2378 
2379 	/* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2380 	if (vid > 0 && !vsi->info.pvid) {
2381 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
2382 			if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2383 					      is_vf, is_netdev))
2384 				continue;
2385 			i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2386 					is_vf, is_netdev);
2387 			add_f = i40e_add_filter(vsi, f->macaddr,
2388 						0, is_vf, is_netdev);
2389 			if (!add_f) {
2390 				dev_info(&vsi->back->pdev->dev,
2391 					 "Could not add filter 0 for %pM\n",
2392 					f->macaddr);
2393 				spin_unlock_bh(&vsi->mac_filter_list_lock);
2394 				return -ENOMEM;
2395 			}
2396 		}
2397 	}
2398 
2399 	spin_unlock_bh(&vsi->mac_filter_list_lock);
2400 
2401 	/* schedule our worker thread which will take care of
2402 	 * applying the new filter changes
2403 	 */
2404 	i40e_service_event_schedule(vsi->back);
2405 	return 0;
2406 }
2407 
2408 /**
2409  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2410  * @vsi: the vsi being configured
2411  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2412  *
2413  * Return: 0 on success or negative otherwise
2414  **/
2415 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2416 {
2417 	struct net_device *netdev = vsi->netdev;
2418 	struct i40e_mac_filter *f, *add_f;
2419 	bool is_vf, is_netdev;
2420 	int filter_count = 0;
2421 
2422 	is_vf = (vsi->type == I40E_VSI_SRIOV);
2423 	is_netdev = !!(netdev);
2424 
2425 	/* Locked once because all functions invoked below iterates list */
2426 	spin_lock_bh(&vsi->mac_filter_list_lock);
2427 
2428 	if (is_netdev)
2429 		i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2430 
2431 	list_for_each_entry(f, &vsi->mac_filter_list, list)
2432 		i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2433 
2434 	/* go through all the filters for this VSI and if there is only
2435 	 * vid == 0 it means there are no other filters, so vid 0 must
2436 	 * be replaced with -1. This signifies that we should from now
2437 	 * on accept any traffic (with any tag present, or untagged)
2438 	 */
2439 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
2440 		if (is_netdev) {
2441 			if (f->vlan &&
2442 			    ether_addr_equal(netdev->dev_addr, f->macaddr))
2443 				filter_count++;
2444 		}
2445 
2446 		if (f->vlan)
2447 			filter_count++;
2448 	}
2449 
2450 	if (!filter_count && is_netdev) {
2451 		i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2452 		f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2453 				    is_vf, is_netdev);
2454 		if (!f) {
2455 			dev_info(&vsi->back->pdev->dev,
2456 				 "Could not add filter %d for %pM\n",
2457 				 I40E_VLAN_ANY, netdev->dev_addr);
2458 			spin_unlock_bh(&vsi->mac_filter_list_lock);
2459 			return -ENOMEM;
2460 		}
2461 	}
2462 
2463 	if (!filter_count) {
2464 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
2465 			i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2466 			add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2467 						is_vf, is_netdev);
2468 			if (!add_f) {
2469 				dev_info(&vsi->back->pdev->dev,
2470 					 "Could not add filter %d for %pM\n",
2471 					 I40E_VLAN_ANY, f->macaddr);
2472 				spin_unlock_bh(&vsi->mac_filter_list_lock);
2473 				return -ENOMEM;
2474 			}
2475 		}
2476 	}
2477 
2478 	spin_unlock_bh(&vsi->mac_filter_list_lock);
2479 
2480 	/* schedule our worker thread which will take care of
2481 	 * applying the new filter changes
2482 	 */
2483 	i40e_service_event_schedule(vsi->back);
2484 	return 0;
2485 }
2486 
2487 /**
2488  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2489  * @netdev: network interface to be adjusted
2490  * @vid: vlan id to be added
2491  *
2492  * net_device_ops implementation for adding vlan ids
2493  **/
2494 #ifdef I40E_FCOE
2495 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2496 			 __always_unused __be16 proto, u16 vid)
2497 #else
2498 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2499 				__always_unused __be16 proto, u16 vid)
2500 #endif
2501 {
2502 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2503 	struct i40e_vsi *vsi = np->vsi;
2504 	int ret = 0;
2505 
2506 	if (vid > 4095)
2507 		return -EINVAL;
2508 
2509 	netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2510 
2511 	/* If the network stack called us with vid = 0 then
2512 	 * it is asking to receive priority tagged packets with
2513 	 * vlan id 0.  Our HW receives them by default when configured
2514 	 * to receive untagged packets so there is no need to add an
2515 	 * extra filter for vlan 0 tagged packets.
2516 	 */
2517 	if (vid)
2518 		ret = i40e_vsi_add_vlan(vsi, vid);
2519 
2520 	if (!ret && (vid < VLAN_N_VID))
2521 		set_bit(vid, vsi->active_vlans);
2522 
2523 	return ret;
2524 }
2525 
2526 /**
2527  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2528  * @netdev: network interface to be adjusted
2529  * @vid: vlan id to be removed
2530  *
2531  * net_device_ops implementation for removing vlan ids
2532  **/
2533 #ifdef I40E_FCOE
2534 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2535 			  __always_unused __be16 proto, u16 vid)
2536 #else
2537 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2538 				 __always_unused __be16 proto, u16 vid)
2539 #endif
2540 {
2541 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2542 	struct i40e_vsi *vsi = np->vsi;
2543 
2544 	netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2545 
2546 	/* return code is ignored as there is nothing a user
2547 	 * can do about failure to remove and a log message was
2548 	 * already printed from the other function
2549 	 */
2550 	i40e_vsi_kill_vlan(vsi, vid);
2551 
2552 	clear_bit(vid, vsi->active_vlans);
2553 
2554 	return 0;
2555 }
2556 
2557 /**
2558  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2559  * @vsi: the vsi being brought back up
2560  **/
2561 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2562 {
2563 	u16 vid;
2564 
2565 	if (!vsi->netdev)
2566 		return;
2567 
2568 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2569 
2570 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2571 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2572 				     vid);
2573 }
2574 
2575 /**
2576  * i40e_vsi_add_pvid - Add pvid for the VSI
2577  * @vsi: the vsi being adjusted
2578  * @vid: the vlan id to set as a PVID
2579  **/
2580 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2581 {
2582 	struct i40e_vsi_context ctxt;
2583 	i40e_status ret;
2584 
2585 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2586 	vsi->info.pvid = cpu_to_le16(vid);
2587 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2588 				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2589 				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2590 
2591 	ctxt.seid = vsi->seid;
2592 	ctxt.info = vsi->info;
2593 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2594 	if (ret) {
2595 		dev_info(&vsi->back->pdev->dev,
2596 			 "add pvid failed, err %s aq_err %s\n",
2597 			 i40e_stat_str(&vsi->back->hw, ret),
2598 			 i40e_aq_str(&vsi->back->hw,
2599 				     vsi->back->hw.aq.asq_last_status));
2600 		return -ENOENT;
2601 	}
2602 
2603 	return 0;
2604 }
2605 
2606 /**
2607  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2608  * @vsi: the vsi being adjusted
2609  *
2610  * Just use the vlan_rx_register() service to put it back to normal
2611  **/
2612 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2613 {
2614 	i40e_vlan_stripping_disable(vsi);
2615 
2616 	vsi->info.pvid = 0;
2617 }
2618 
2619 /**
2620  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2621  * @vsi: ptr to the VSI
2622  *
2623  * If this function returns with an error, then it's possible one or
2624  * more of the rings is populated (while the rest are not).  It is the
2625  * callers duty to clean those orphaned rings.
2626  *
2627  * Return 0 on success, negative on failure
2628  **/
2629 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2630 {
2631 	int i, err = 0;
2632 
2633 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2634 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2635 
2636 	return err;
2637 }
2638 
2639 /**
2640  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2641  * @vsi: ptr to the VSI
2642  *
2643  * Free VSI's transmit software resources
2644  **/
2645 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2646 {
2647 	int i;
2648 
2649 	if (!vsi->tx_rings)
2650 		return;
2651 
2652 	for (i = 0; i < vsi->num_queue_pairs; i++)
2653 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2654 			i40e_free_tx_resources(vsi->tx_rings[i]);
2655 }
2656 
2657 /**
2658  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2659  * @vsi: ptr to the VSI
2660  *
2661  * If this function returns with an error, then it's possible one or
2662  * more of the rings is populated (while the rest are not).  It is the
2663  * callers duty to clean those orphaned rings.
2664  *
2665  * Return 0 on success, negative on failure
2666  **/
2667 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2668 {
2669 	int i, err = 0;
2670 
2671 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2672 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2673 #ifdef I40E_FCOE
2674 	i40e_fcoe_setup_ddp_resources(vsi);
2675 #endif
2676 	return err;
2677 }
2678 
2679 /**
2680  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2681  * @vsi: ptr to the VSI
2682  *
2683  * Free all receive software resources
2684  **/
2685 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2686 {
2687 	int i;
2688 
2689 	if (!vsi->rx_rings)
2690 		return;
2691 
2692 	for (i = 0; i < vsi->num_queue_pairs; i++)
2693 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2694 			i40e_free_rx_resources(vsi->rx_rings[i]);
2695 #ifdef I40E_FCOE
2696 	i40e_fcoe_free_ddp_resources(vsi);
2697 #endif
2698 }
2699 
2700 /**
2701  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2702  * @ring: The Tx ring to configure
2703  *
2704  * This enables/disables XPS for a given Tx descriptor ring
2705  * based on the TCs enabled for the VSI that ring belongs to.
2706  **/
2707 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2708 {
2709 	struct i40e_vsi *vsi = ring->vsi;
2710 	cpumask_var_t mask;
2711 
2712 	if (!ring->q_vector || !ring->netdev)
2713 		return;
2714 
2715 	/* Single TC mode enable XPS */
2716 	if (vsi->tc_config.numtc <= 1) {
2717 		if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2718 			netif_set_xps_queue(ring->netdev,
2719 					    &ring->q_vector->affinity_mask,
2720 					    ring->queue_index);
2721 	} else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2722 		/* Disable XPS to allow selection based on TC */
2723 		bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2724 		netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2725 		free_cpumask_var(mask);
2726 	}
2727 
2728 	/* schedule our worker thread which will take care of
2729 	 * applying the new filter changes
2730 	 */
2731 	i40e_service_event_schedule(vsi->back);
2732 }
2733 
2734 /**
2735  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2736  * @ring: The Tx ring to configure
2737  *
2738  * Configure the Tx descriptor ring in the HMC context.
2739  **/
2740 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2741 {
2742 	struct i40e_vsi *vsi = ring->vsi;
2743 	u16 pf_q = vsi->base_queue + ring->queue_index;
2744 	struct i40e_hw *hw = &vsi->back->hw;
2745 	struct i40e_hmc_obj_txq tx_ctx;
2746 	i40e_status err = 0;
2747 	u32 qtx_ctl = 0;
2748 
2749 	/* some ATR related tx ring init */
2750 	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2751 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2752 		ring->atr_count = 0;
2753 	} else {
2754 		ring->atr_sample_rate = 0;
2755 	}
2756 
2757 	/* configure XPS */
2758 	i40e_config_xps_tx_ring(ring);
2759 
2760 	/* clear the context structure first */
2761 	memset(&tx_ctx, 0, sizeof(tx_ctx));
2762 
2763 	tx_ctx.new_context = 1;
2764 	tx_ctx.base = (ring->dma / 128);
2765 	tx_ctx.qlen = ring->count;
2766 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2767 					       I40E_FLAG_FD_ATR_ENABLED));
2768 #ifdef I40E_FCOE
2769 	tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2770 #endif
2771 	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2772 	/* FDIR VSI tx ring can still use RS bit and writebacks */
2773 	if (vsi->type != I40E_VSI_FDIR)
2774 		tx_ctx.head_wb_ena = 1;
2775 	tx_ctx.head_wb_addr = ring->dma +
2776 			      (ring->count * sizeof(struct i40e_tx_desc));
2777 
2778 	/* As part of VSI creation/update, FW allocates certain
2779 	 * Tx arbitration queue sets for each TC enabled for
2780 	 * the VSI. The FW returns the handles to these queue
2781 	 * sets as part of the response buffer to Add VSI,
2782 	 * Update VSI, etc. AQ commands. It is expected that
2783 	 * these queue set handles be associated with the Tx
2784 	 * queues by the driver as part of the TX queue context
2785 	 * initialization. This has to be done regardless of
2786 	 * DCB as by default everything is mapped to TC0.
2787 	 */
2788 	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2789 	tx_ctx.rdylist_act = 0;
2790 
2791 	/* clear the context in the HMC */
2792 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2793 	if (err) {
2794 		dev_info(&vsi->back->pdev->dev,
2795 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2796 			 ring->queue_index, pf_q, err);
2797 		return -ENOMEM;
2798 	}
2799 
2800 	/* set the context in the HMC */
2801 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2802 	if (err) {
2803 		dev_info(&vsi->back->pdev->dev,
2804 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2805 			 ring->queue_index, pf_q, err);
2806 		return -ENOMEM;
2807 	}
2808 
2809 	/* Now associate this queue with this PCI function */
2810 	if (vsi->type == I40E_VSI_VMDQ2) {
2811 		qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2812 		qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2813 			   I40E_QTX_CTL_VFVM_INDX_MASK;
2814 	} else {
2815 		qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2816 	}
2817 
2818 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2819 		    I40E_QTX_CTL_PF_INDX_MASK);
2820 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2821 	i40e_flush(hw);
2822 
2823 	/* cache tail off for easier writes later */
2824 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2825 
2826 	return 0;
2827 }
2828 
2829 /**
2830  * i40e_configure_rx_ring - Configure a receive ring context
2831  * @ring: The Rx ring to configure
2832  *
2833  * Configure the Rx descriptor ring in the HMC context.
2834  **/
2835 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2836 {
2837 	struct i40e_vsi *vsi = ring->vsi;
2838 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2839 	u16 pf_q = vsi->base_queue + ring->queue_index;
2840 	struct i40e_hw *hw = &vsi->back->hw;
2841 	struct i40e_hmc_obj_rxq rx_ctx;
2842 	i40e_status err = 0;
2843 
2844 	ring->state = 0;
2845 
2846 	/* clear the context structure first */
2847 	memset(&rx_ctx, 0, sizeof(rx_ctx));
2848 
2849 	ring->rx_buf_len = vsi->rx_buf_len;
2850 	ring->rx_hdr_len = vsi->rx_hdr_len;
2851 
2852 	rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2853 	rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2854 
2855 	rx_ctx.base = (ring->dma / 128);
2856 	rx_ctx.qlen = ring->count;
2857 
2858 	if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2859 		set_ring_16byte_desc_enabled(ring);
2860 		rx_ctx.dsize = 0;
2861 	} else {
2862 		rx_ctx.dsize = 1;
2863 	}
2864 
2865 	rx_ctx.dtype = vsi->dtype;
2866 	if (vsi->dtype) {
2867 		set_ring_ps_enabled(ring);
2868 		rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2869 				  I40E_RX_SPLIT_IP      |
2870 				  I40E_RX_SPLIT_TCP_UDP |
2871 				  I40E_RX_SPLIT_SCTP;
2872 	} else {
2873 		rx_ctx.hsplit_0 = 0;
2874 	}
2875 
2876 	rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2877 				  (chain_len * ring->rx_buf_len));
2878 	if (hw->revision_id == 0)
2879 		rx_ctx.lrxqthresh = 0;
2880 	else
2881 		rx_ctx.lrxqthresh = 2;
2882 	rx_ctx.crcstrip = 1;
2883 	rx_ctx.l2tsel = 1;
2884 	/* this controls whether VLAN is stripped from inner headers */
2885 	rx_ctx.showiv = 0;
2886 #ifdef I40E_FCOE
2887 	rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2888 #endif
2889 	/* set the prefena field to 1 because the manual says to */
2890 	rx_ctx.prefena = 1;
2891 
2892 	/* clear the context in the HMC */
2893 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2894 	if (err) {
2895 		dev_info(&vsi->back->pdev->dev,
2896 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2897 			 ring->queue_index, pf_q, err);
2898 		return -ENOMEM;
2899 	}
2900 
2901 	/* set the context in the HMC */
2902 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2903 	if (err) {
2904 		dev_info(&vsi->back->pdev->dev,
2905 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2906 			 ring->queue_index, pf_q, err);
2907 		return -ENOMEM;
2908 	}
2909 
2910 	/* cache tail for quicker writes, and clear the reg before use */
2911 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2912 	writel(0, ring->tail);
2913 
2914 	if (ring_is_ps_enabled(ring)) {
2915 		i40e_alloc_rx_headers(ring);
2916 		i40e_alloc_rx_buffers_ps(ring, I40E_DESC_UNUSED(ring));
2917 	} else {
2918 		i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring));
2919 	}
2920 
2921 	return 0;
2922 }
2923 
2924 /**
2925  * i40e_vsi_configure_tx - Configure the VSI for Tx
2926  * @vsi: VSI structure describing this set of rings and resources
2927  *
2928  * Configure the Tx VSI for operation.
2929  **/
2930 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2931 {
2932 	int err = 0;
2933 	u16 i;
2934 
2935 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2936 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2937 
2938 	return err;
2939 }
2940 
2941 /**
2942  * i40e_vsi_configure_rx - Configure the VSI for Rx
2943  * @vsi: the VSI being configured
2944  *
2945  * Configure the Rx VSI for operation.
2946  **/
2947 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2948 {
2949 	int err = 0;
2950 	u16 i;
2951 
2952 	if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2953 		vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2954 			       + ETH_FCS_LEN + VLAN_HLEN;
2955 	else
2956 		vsi->max_frame = I40E_RXBUFFER_2048;
2957 
2958 	/* figure out correct receive buffer length */
2959 	switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2960 				    I40E_FLAG_RX_PS_ENABLED)) {
2961 	case I40E_FLAG_RX_1BUF_ENABLED:
2962 		vsi->rx_hdr_len = 0;
2963 		vsi->rx_buf_len = vsi->max_frame;
2964 		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2965 		break;
2966 	case I40E_FLAG_RX_PS_ENABLED:
2967 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2968 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2969 		vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2970 		break;
2971 	default:
2972 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2973 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2974 		vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2975 		break;
2976 	}
2977 
2978 #ifdef I40E_FCOE
2979 	/* setup rx buffer for FCoE */
2980 	if ((vsi->type == I40E_VSI_FCOE) &&
2981 	    (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2982 		vsi->rx_hdr_len = 0;
2983 		vsi->rx_buf_len = I40E_RXBUFFER_3072;
2984 		vsi->max_frame = I40E_RXBUFFER_3072;
2985 		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2986 	}
2987 
2988 #endif /* I40E_FCOE */
2989 	/* round up for the chip's needs */
2990 	vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2991 				BIT_ULL(I40E_RXQ_CTX_HBUFF_SHIFT));
2992 	vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2993 				BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
2994 
2995 	/* set up individual rings */
2996 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2997 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2998 
2999 	return err;
3000 }
3001 
3002 /**
3003  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3004  * @vsi: ptr to the VSI
3005  **/
3006 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3007 {
3008 	struct i40e_ring *tx_ring, *rx_ring;
3009 	u16 qoffset, qcount;
3010 	int i, n;
3011 
3012 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3013 		/* Reset the TC information */
3014 		for (i = 0; i < vsi->num_queue_pairs; i++) {
3015 			rx_ring = vsi->rx_rings[i];
3016 			tx_ring = vsi->tx_rings[i];
3017 			rx_ring->dcb_tc = 0;
3018 			tx_ring->dcb_tc = 0;
3019 		}
3020 	}
3021 
3022 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3023 		if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3024 			continue;
3025 
3026 		qoffset = vsi->tc_config.tc_info[n].qoffset;
3027 		qcount = vsi->tc_config.tc_info[n].qcount;
3028 		for (i = qoffset; i < (qoffset + qcount); i++) {
3029 			rx_ring = vsi->rx_rings[i];
3030 			tx_ring = vsi->tx_rings[i];
3031 			rx_ring->dcb_tc = n;
3032 			tx_ring->dcb_tc = n;
3033 		}
3034 	}
3035 }
3036 
3037 /**
3038  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3039  * @vsi: ptr to the VSI
3040  **/
3041 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3042 {
3043 	if (vsi->netdev)
3044 		i40e_set_rx_mode(vsi->netdev);
3045 }
3046 
3047 /**
3048  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3049  * @vsi: Pointer to the targeted VSI
3050  *
3051  * This function replays the hlist on the hw where all the SB Flow Director
3052  * filters were saved.
3053  **/
3054 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3055 {
3056 	struct i40e_fdir_filter *filter;
3057 	struct i40e_pf *pf = vsi->back;
3058 	struct hlist_node *node;
3059 
3060 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3061 		return;
3062 
3063 	hlist_for_each_entry_safe(filter, node,
3064 				  &pf->fdir_filter_list, fdir_node) {
3065 		i40e_add_del_fdir(vsi, filter, true);
3066 	}
3067 }
3068 
3069 /**
3070  * i40e_vsi_configure - Set up the VSI for action
3071  * @vsi: the VSI being configured
3072  **/
3073 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3074 {
3075 	int err;
3076 
3077 	i40e_set_vsi_rx_mode(vsi);
3078 	i40e_restore_vlan(vsi);
3079 	i40e_vsi_config_dcb_rings(vsi);
3080 	err = i40e_vsi_configure_tx(vsi);
3081 	if (!err)
3082 		err = i40e_vsi_configure_rx(vsi);
3083 
3084 	return err;
3085 }
3086 
3087 /**
3088  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3089  * @vsi: the VSI being configured
3090  **/
3091 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3092 {
3093 	struct i40e_pf *pf = vsi->back;
3094 	struct i40e_hw *hw = &pf->hw;
3095 	u16 vector;
3096 	int i, q;
3097 	u32 qp;
3098 
3099 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
3100 	 * and PFINT_LNKLSTn registers, e.g.:
3101 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3102 	 */
3103 	qp = vsi->base_queue;
3104 	vector = vsi->base_vector;
3105 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3106 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3107 
3108 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
3109 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3110 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
3111 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3112 		     q_vector->rx.itr);
3113 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3114 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
3115 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3116 		     q_vector->tx.itr);
3117 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
3118 		     INTRL_USEC_TO_REG(vsi->int_rate_limit));
3119 
3120 		/* Linked list for the queuepairs assigned to this vector */
3121 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3122 		for (q = 0; q < q_vector->num_ringpairs; q++) {
3123 			u32 val;
3124 
3125 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3126 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3127 			      (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3128 			      (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3129 			      (I40E_QUEUE_TYPE_TX
3130 				      << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3131 
3132 			wr32(hw, I40E_QINT_RQCTL(qp), val);
3133 
3134 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3135 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3136 			      (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3137 			      ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3138 			      (I40E_QUEUE_TYPE_RX
3139 				      << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3140 
3141 			/* Terminate the linked list */
3142 			if (q == (q_vector->num_ringpairs - 1))
3143 				val |= (I40E_QUEUE_END_OF_LIST
3144 					   << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3145 
3146 			wr32(hw, I40E_QINT_TQCTL(qp), val);
3147 			qp++;
3148 		}
3149 	}
3150 
3151 	i40e_flush(hw);
3152 }
3153 
3154 /**
3155  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3156  * @hw: ptr to the hardware info
3157  **/
3158 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3159 {
3160 	struct i40e_hw *hw = &pf->hw;
3161 	u32 val;
3162 
3163 	/* clear things first */
3164 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3165 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3166 
3167 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3168 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3169 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
3170 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3171 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3172 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3173 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3174 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3175 
3176 	if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3177 		val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3178 
3179 	if (pf->flags & I40E_FLAG_PTP)
3180 		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3181 
3182 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
3183 
3184 	/* SW_ITR_IDX = 0, but don't change INTENA */
3185 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3186 					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3187 
3188 	/* OTHER_ITR_IDX = 0 */
3189 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3190 }
3191 
3192 /**
3193  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3194  * @vsi: the VSI being configured
3195  **/
3196 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3197 {
3198 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3199 	struct i40e_pf *pf = vsi->back;
3200 	struct i40e_hw *hw = &pf->hw;
3201 	u32 val;
3202 
3203 	/* set the ITR configuration */
3204 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
3205 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3206 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
3207 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3208 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3209 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
3210 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3211 
3212 	i40e_enable_misc_int_causes(pf);
3213 
3214 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3215 	wr32(hw, I40E_PFINT_LNKLST0, 0);
3216 
3217 	/* Associate the queue pair to the vector and enable the queue int */
3218 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		      |
3219 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3220 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3221 
3222 	wr32(hw, I40E_QINT_RQCTL(0), val);
3223 
3224 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
3225 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3226 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3227 
3228 	wr32(hw, I40E_QINT_TQCTL(0), val);
3229 	i40e_flush(hw);
3230 }
3231 
3232 /**
3233  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3234  * @pf: board private structure
3235  **/
3236 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3237 {
3238 	struct i40e_hw *hw = &pf->hw;
3239 
3240 	wr32(hw, I40E_PFINT_DYN_CTL0,
3241 	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3242 	i40e_flush(hw);
3243 }
3244 
3245 /**
3246  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3247  * @pf: board private structure
3248  **/
3249 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3250 {
3251 	struct i40e_hw *hw = &pf->hw;
3252 	u32 val;
3253 
3254 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3255 	      I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3256 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3257 
3258 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
3259 	i40e_flush(hw);
3260 }
3261 
3262 /**
3263  * i40e_irq_dynamic_disable - Disable default interrupt generation settings
3264  * @vsi: pointer to a vsi
3265  * @vector: disable a particular Hw Interrupt vector
3266  **/
3267 void i40e_irq_dynamic_disable(struct i40e_vsi *vsi, int vector)
3268 {
3269 	struct i40e_pf *pf = vsi->back;
3270 	struct i40e_hw *hw = &pf->hw;
3271 	u32 val;
3272 
3273 	val = I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT;
3274 	wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
3275 	i40e_flush(hw);
3276 }
3277 
3278 /**
3279  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3280  * @irq: interrupt number
3281  * @data: pointer to a q_vector
3282  **/
3283 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3284 {
3285 	struct i40e_q_vector *q_vector = data;
3286 
3287 	if (!q_vector->tx.ring && !q_vector->rx.ring)
3288 		return IRQ_HANDLED;
3289 
3290 	napi_schedule_irqoff(&q_vector->napi);
3291 
3292 	return IRQ_HANDLED;
3293 }
3294 
3295 /**
3296  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3297  * @vsi: the VSI being configured
3298  * @basename: name for the vector
3299  *
3300  * Allocates MSI-X vectors and requests interrupts from the kernel.
3301  **/
3302 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3303 {
3304 	int q_vectors = vsi->num_q_vectors;
3305 	struct i40e_pf *pf = vsi->back;
3306 	int base = vsi->base_vector;
3307 	int rx_int_idx = 0;
3308 	int tx_int_idx = 0;
3309 	int vector, err;
3310 
3311 	for (vector = 0; vector < q_vectors; vector++) {
3312 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3313 
3314 		if (q_vector->tx.ring && q_vector->rx.ring) {
3315 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3316 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3317 			tx_int_idx++;
3318 		} else if (q_vector->rx.ring) {
3319 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3320 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3321 		} else if (q_vector->tx.ring) {
3322 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3323 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3324 		} else {
3325 			/* skip this unused q_vector */
3326 			continue;
3327 		}
3328 		err = request_irq(pf->msix_entries[base + vector].vector,
3329 				  vsi->irq_handler,
3330 				  0,
3331 				  q_vector->name,
3332 				  q_vector);
3333 		if (err) {
3334 			dev_info(&pf->pdev->dev,
3335 				 "MSIX request_irq failed, error: %d\n", err);
3336 			goto free_queue_irqs;
3337 		}
3338 		/* assign the mask for this irq */
3339 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3340 				      &q_vector->affinity_mask);
3341 	}
3342 
3343 	vsi->irqs_ready = true;
3344 	return 0;
3345 
3346 free_queue_irqs:
3347 	while (vector) {
3348 		vector--;
3349 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3350 				      NULL);
3351 		free_irq(pf->msix_entries[base + vector].vector,
3352 			 &(vsi->q_vectors[vector]));
3353 	}
3354 	return err;
3355 }
3356 
3357 /**
3358  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3359  * @vsi: the VSI being un-configured
3360  **/
3361 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3362 {
3363 	struct i40e_pf *pf = vsi->back;
3364 	struct i40e_hw *hw = &pf->hw;
3365 	int base = vsi->base_vector;
3366 	int i;
3367 
3368 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3369 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3370 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3371 	}
3372 
3373 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3374 		for (i = vsi->base_vector;
3375 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3376 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3377 
3378 		i40e_flush(hw);
3379 		for (i = 0; i < vsi->num_q_vectors; i++)
3380 			synchronize_irq(pf->msix_entries[i + base].vector);
3381 	} else {
3382 		/* Legacy and MSI mode - this stops all interrupt handling */
3383 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3384 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3385 		i40e_flush(hw);
3386 		synchronize_irq(pf->pdev->irq);
3387 	}
3388 }
3389 
3390 /**
3391  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3392  * @vsi: the VSI being configured
3393  **/
3394 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3395 {
3396 	struct i40e_pf *pf = vsi->back;
3397 	int i;
3398 
3399 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3400 		for (i = 0; i < vsi->num_q_vectors; i++)
3401 			i40e_irq_dynamic_enable(vsi, i);
3402 	} else {
3403 		i40e_irq_dynamic_enable_icr0(pf);
3404 	}
3405 
3406 	i40e_flush(&pf->hw);
3407 	return 0;
3408 }
3409 
3410 /**
3411  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3412  * @pf: board private structure
3413  **/
3414 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3415 {
3416 	/* Disable ICR 0 */
3417 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3418 	i40e_flush(&pf->hw);
3419 }
3420 
3421 /**
3422  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3423  * @irq: interrupt number
3424  * @data: pointer to a q_vector
3425  *
3426  * This is the handler used for all MSI/Legacy interrupts, and deals
3427  * with both queue and non-queue interrupts.  This is also used in
3428  * MSIX mode to handle the non-queue interrupts.
3429  **/
3430 static irqreturn_t i40e_intr(int irq, void *data)
3431 {
3432 	struct i40e_pf *pf = (struct i40e_pf *)data;
3433 	struct i40e_hw *hw = &pf->hw;
3434 	irqreturn_t ret = IRQ_NONE;
3435 	u32 icr0, icr0_remaining;
3436 	u32 val, ena_mask;
3437 
3438 	icr0 = rd32(hw, I40E_PFINT_ICR0);
3439 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3440 
3441 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3442 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3443 		goto enable_intr;
3444 
3445 	/* if interrupt but no bits showing, must be SWINT */
3446 	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3447 	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3448 		pf->sw_int_count++;
3449 
3450 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3451 	    (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3452 		ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3453 		icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3454 		dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3455 	}
3456 
3457 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3458 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3459 		struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3460 		struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3461 
3462 		/* temporarily disable queue cause for NAPI processing */
3463 		u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
3464 
3465 		qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3466 		wr32(hw, I40E_QINT_RQCTL(0), qval);
3467 
3468 		qval = rd32(hw, I40E_QINT_TQCTL(0));
3469 		qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3470 		wr32(hw, I40E_QINT_TQCTL(0), qval);
3471 
3472 		if (!test_bit(__I40E_DOWN, &pf->state))
3473 			napi_schedule_irqoff(&q_vector->napi);
3474 	}
3475 
3476 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3477 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3478 		set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3479 	}
3480 
3481 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3482 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3483 		set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3484 	}
3485 
3486 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3487 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3488 		set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3489 	}
3490 
3491 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3492 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3493 			set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3494 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3495 		val = rd32(hw, I40E_GLGEN_RSTAT);
3496 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3497 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3498 		if (val == I40E_RESET_CORER) {
3499 			pf->corer_count++;
3500 		} else if (val == I40E_RESET_GLOBR) {
3501 			pf->globr_count++;
3502 		} else if (val == I40E_RESET_EMPR) {
3503 			pf->empr_count++;
3504 			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3505 		}
3506 	}
3507 
3508 	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3509 		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3510 		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3511 		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3512 			 rd32(hw, I40E_PFHMC_ERRORINFO),
3513 			 rd32(hw, I40E_PFHMC_ERRORDATA));
3514 	}
3515 
3516 	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3517 		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3518 
3519 		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3520 			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3521 			i40e_ptp_tx_hwtstamp(pf);
3522 		}
3523 	}
3524 
3525 	/* If a critical error is pending we have no choice but to reset the
3526 	 * device.
3527 	 * Report and mask out any remaining unexpected interrupts.
3528 	 */
3529 	icr0_remaining = icr0 & ena_mask;
3530 	if (icr0_remaining) {
3531 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3532 			 icr0_remaining);
3533 		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3534 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3535 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3536 			dev_info(&pf->pdev->dev, "device will be reset\n");
3537 			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3538 			i40e_service_event_schedule(pf);
3539 		}
3540 		ena_mask &= ~icr0_remaining;
3541 	}
3542 	ret = IRQ_HANDLED;
3543 
3544 enable_intr:
3545 	/* re-enable interrupt causes */
3546 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3547 	if (!test_bit(__I40E_DOWN, &pf->state)) {
3548 		i40e_service_event_schedule(pf);
3549 		i40e_irq_dynamic_enable_icr0(pf);
3550 	}
3551 
3552 	return ret;
3553 }
3554 
3555 /**
3556  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3557  * @tx_ring:  tx ring to clean
3558  * @budget:   how many cleans we're allowed
3559  *
3560  * Returns true if there's any budget left (e.g. the clean is finished)
3561  **/
3562 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3563 {
3564 	struct i40e_vsi *vsi = tx_ring->vsi;
3565 	u16 i = tx_ring->next_to_clean;
3566 	struct i40e_tx_buffer *tx_buf;
3567 	struct i40e_tx_desc *tx_desc;
3568 
3569 	tx_buf = &tx_ring->tx_bi[i];
3570 	tx_desc = I40E_TX_DESC(tx_ring, i);
3571 	i -= tx_ring->count;
3572 
3573 	do {
3574 		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3575 
3576 		/* if next_to_watch is not set then there is no work pending */
3577 		if (!eop_desc)
3578 			break;
3579 
3580 		/* prevent any other reads prior to eop_desc */
3581 		read_barrier_depends();
3582 
3583 		/* if the descriptor isn't done, no work yet to do */
3584 		if (!(eop_desc->cmd_type_offset_bsz &
3585 		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3586 			break;
3587 
3588 		/* clear next_to_watch to prevent false hangs */
3589 		tx_buf->next_to_watch = NULL;
3590 
3591 		tx_desc->buffer_addr = 0;
3592 		tx_desc->cmd_type_offset_bsz = 0;
3593 		/* move past filter desc */
3594 		tx_buf++;
3595 		tx_desc++;
3596 		i++;
3597 		if (unlikely(!i)) {
3598 			i -= tx_ring->count;
3599 			tx_buf = tx_ring->tx_bi;
3600 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3601 		}
3602 		/* unmap skb header data */
3603 		dma_unmap_single(tx_ring->dev,
3604 				 dma_unmap_addr(tx_buf, dma),
3605 				 dma_unmap_len(tx_buf, len),
3606 				 DMA_TO_DEVICE);
3607 		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3608 			kfree(tx_buf->raw_buf);
3609 
3610 		tx_buf->raw_buf = NULL;
3611 		tx_buf->tx_flags = 0;
3612 		tx_buf->next_to_watch = NULL;
3613 		dma_unmap_len_set(tx_buf, len, 0);
3614 		tx_desc->buffer_addr = 0;
3615 		tx_desc->cmd_type_offset_bsz = 0;
3616 
3617 		/* move us past the eop_desc for start of next FD desc */
3618 		tx_buf++;
3619 		tx_desc++;
3620 		i++;
3621 		if (unlikely(!i)) {
3622 			i -= tx_ring->count;
3623 			tx_buf = tx_ring->tx_bi;
3624 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3625 		}
3626 
3627 		/* update budget accounting */
3628 		budget--;
3629 	} while (likely(budget));
3630 
3631 	i += tx_ring->count;
3632 	tx_ring->next_to_clean = i;
3633 
3634 	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3635 		i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3636 
3637 	return budget > 0;
3638 }
3639 
3640 /**
3641  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3642  * @irq: interrupt number
3643  * @data: pointer to a q_vector
3644  **/
3645 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3646 {
3647 	struct i40e_q_vector *q_vector = data;
3648 	struct i40e_vsi *vsi;
3649 
3650 	if (!q_vector->tx.ring)
3651 		return IRQ_HANDLED;
3652 
3653 	vsi = q_vector->tx.ring->vsi;
3654 	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3655 
3656 	return IRQ_HANDLED;
3657 }
3658 
3659 /**
3660  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3661  * @vsi: the VSI being configured
3662  * @v_idx: vector index
3663  * @qp_idx: queue pair index
3664  **/
3665 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3666 {
3667 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3668 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3669 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3670 
3671 	tx_ring->q_vector = q_vector;
3672 	tx_ring->next = q_vector->tx.ring;
3673 	q_vector->tx.ring = tx_ring;
3674 	q_vector->tx.count++;
3675 
3676 	rx_ring->q_vector = q_vector;
3677 	rx_ring->next = q_vector->rx.ring;
3678 	q_vector->rx.ring = rx_ring;
3679 	q_vector->rx.count++;
3680 }
3681 
3682 /**
3683  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3684  * @vsi: the VSI being configured
3685  *
3686  * This function maps descriptor rings to the queue-specific vectors
3687  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3688  * one vector per queue pair, but on a constrained vector budget, we
3689  * group the queue pairs as "efficiently" as possible.
3690  **/
3691 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3692 {
3693 	int qp_remaining = vsi->num_queue_pairs;
3694 	int q_vectors = vsi->num_q_vectors;
3695 	int num_ringpairs;
3696 	int v_start = 0;
3697 	int qp_idx = 0;
3698 
3699 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3700 	 * group them so there are multiple queues per vector.
3701 	 * It is also important to go through all the vectors available to be
3702 	 * sure that if we don't use all the vectors, that the remaining vectors
3703 	 * are cleared. This is especially important when decreasing the
3704 	 * number of queues in use.
3705 	 */
3706 	for (; v_start < q_vectors; v_start++) {
3707 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3708 
3709 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3710 
3711 		q_vector->num_ringpairs = num_ringpairs;
3712 
3713 		q_vector->rx.count = 0;
3714 		q_vector->tx.count = 0;
3715 		q_vector->rx.ring = NULL;
3716 		q_vector->tx.ring = NULL;
3717 
3718 		while (num_ringpairs--) {
3719 			i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3720 			qp_idx++;
3721 			qp_remaining--;
3722 		}
3723 	}
3724 }
3725 
3726 /**
3727  * i40e_vsi_request_irq - Request IRQ from the OS
3728  * @vsi: the VSI being configured
3729  * @basename: name for the vector
3730  **/
3731 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3732 {
3733 	struct i40e_pf *pf = vsi->back;
3734 	int err;
3735 
3736 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3737 		err = i40e_vsi_request_irq_msix(vsi, basename);
3738 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3739 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
3740 				  pf->int_name, pf);
3741 	else
3742 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3743 				  pf->int_name, pf);
3744 
3745 	if (err)
3746 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3747 
3748 	return err;
3749 }
3750 
3751 #ifdef CONFIG_NET_POLL_CONTROLLER
3752 /**
3753  * i40e_netpoll - A Polling 'interrupt'handler
3754  * @netdev: network interface device structure
3755  *
3756  * This is used by netconsole to send skbs without having to re-enable
3757  * interrupts.  It's not called while the normal interrupt routine is executing.
3758  **/
3759 #ifdef I40E_FCOE
3760 void i40e_netpoll(struct net_device *netdev)
3761 #else
3762 static void i40e_netpoll(struct net_device *netdev)
3763 #endif
3764 {
3765 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3766 	struct i40e_vsi *vsi = np->vsi;
3767 	struct i40e_pf *pf = vsi->back;
3768 	int i;
3769 
3770 	/* if interface is down do nothing */
3771 	if (test_bit(__I40E_DOWN, &vsi->state))
3772 		return;
3773 
3774 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3775 		for (i = 0; i < vsi->num_q_vectors; i++)
3776 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3777 	} else {
3778 		i40e_intr(pf->pdev->irq, netdev);
3779 	}
3780 }
3781 #endif
3782 
3783 /**
3784  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3785  * @pf: the PF being configured
3786  * @pf_q: the PF queue
3787  * @enable: enable or disable state of the queue
3788  *
3789  * This routine will wait for the given Tx queue of the PF to reach the
3790  * enabled or disabled state.
3791  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3792  * multiple retries; else will return 0 in case of success.
3793  **/
3794 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3795 {
3796 	int i;
3797 	u32 tx_reg;
3798 
3799 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3800 		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3801 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3802 			break;
3803 
3804 		usleep_range(10, 20);
3805 	}
3806 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3807 		return -ETIMEDOUT;
3808 
3809 	return 0;
3810 }
3811 
3812 /**
3813  * i40e_vsi_control_tx - Start or stop a VSI's rings
3814  * @vsi: the VSI being configured
3815  * @enable: start or stop the rings
3816  **/
3817 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3818 {
3819 	struct i40e_pf *pf = vsi->back;
3820 	struct i40e_hw *hw = &pf->hw;
3821 	int i, j, pf_q, ret = 0;
3822 	u32 tx_reg;
3823 
3824 	pf_q = vsi->base_queue;
3825 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3826 
3827 		/* warn the TX unit of coming changes */
3828 		i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3829 		if (!enable)
3830 			usleep_range(10, 20);
3831 
3832 		for (j = 0; j < 50; j++) {
3833 			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3834 			if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3835 			    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3836 				break;
3837 			usleep_range(1000, 2000);
3838 		}
3839 		/* Skip if the queue is already in the requested state */
3840 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3841 			continue;
3842 
3843 		/* turn on/off the queue */
3844 		if (enable) {
3845 			wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3846 			tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3847 		} else {
3848 			tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3849 		}
3850 
3851 		wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3852 		/* No waiting for the Tx queue to disable */
3853 		if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3854 			continue;
3855 
3856 		/* wait for the change to finish */
3857 		ret = i40e_pf_txq_wait(pf, pf_q, enable);
3858 		if (ret) {
3859 			dev_info(&pf->pdev->dev,
3860 				 "VSI seid %d Tx ring %d %sable timeout\n",
3861 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
3862 			break;
3863 		}
3864 	}
3865 
3866 	if (hw->revision_id == 0)
3867 		mdelay(50);
3868 	return ret;
3869 }
3870 
3871 /**
3872  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3873  * @pf: the PF being configured
3874  * @pf_q: the PF queue
3875  * @enable: enable or disable state of the queue
3876  *
3877  * This routine will wait for the given Rx queue of the PF to reach the
3878  * enabled or disabled state.
3879  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3880  * multiple retries; else will return 0 in case of success.
3881  **/
3882 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3883 {
3884 	int i;
3885 	u32 rx_reg;
3886 
3887 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3888 		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3889 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3890 			break;
3891 
3892 		usleep_range(10, 20);
3893 	}
3894 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3895 		return -ETIMEDOUT;
3896 
3897 	return 0;
3898 }
3899 
3900 /**
3901  * i40e_vsi_control_rx - Start or stop a VSI's rings
3902  * @vsi: the VSI being configured
3903  * @enable: start or stop the rings
3904  **/
3905 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3906 {
3907 	struct i40e_pf *pf = vsi->back;
3908 	struct i40e_hw *hw = &pf->hw;
3909 	int i, j, pf_q, ret = 0;
3910 	u32 rx_reg;
3911 
3912 	pf_q = vsi->base_queue;
3913 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3914 		for (j = 0; j < 50; j++) {
3915 			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3916 			if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3917 			    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3918 				break;
3919 			usleep_range(1000, 2000);
3920 		}
3921 
3922 		/* Skip if the queue is already in the requested state */
3923 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3924 			continue;
3925 
3926 		/* turn on/off the queue */
3927 		if (enable)
3928 			rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3929 		else
3930 			rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3931 		wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3932 
3933 		/* wait for the change to finish */
3934 		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3935 		if (ret) {
3936 			dev_info(&pf->pdev->dev,
3937 				 "VSI seid %d Rx ring %d %sable timeout\n",
3938 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
3939 			break;
3940 		}
3941 	}
3942 
3943 	return ret;
3944 }
3945 
3946 /**
3947  * i40e_vsi_control_rings - Start or stop a VSI's rings
3948  * @vsi: the VSI being configured
3949  * @enable: start or stop the rings
3950  **/
3951 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3952 {
3953 	int ret = 0;
3954 
3955 	/* do rx first for enable and last for disable */
3956 	if (request) {
3957 		ret = i40e_vsi_control_rx(vsi, request);
3958 		if (ret)
3959 			return ret;
3960 		ret = i40e_vsi_control_tx(vsi, request);
3961 	} else {
3962 		/* Ignore return value, we need to shutdown whatever we can */
3963 		i40e_vsi_control_tx(vsi, request);
3964 		i40e_vsi_control_rx(vsi, request);
3965 	}
3966 
3967 	return ret;
3968 }
3969 
3970 /**
3971  * i40e_vsi_free_irq - Free the irq association with the OS
3972  * @vsi: the VSI being configured
3973  **/
3974 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3975 {
3976 	struct i40e_pf *pf = vsi->back;
3977 	struct i40e_hw *hw = &pf->hw;
3978 	int base = vsi->base_vector;
3979 	u32 val, qp;
3980 	int i;
3981 
3982 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3983 		if (!vsi->q_vectors)
3984 			return;
3985 
3986 		if (!vsi->irqs_ready)
3987 			return;
3988 
3989 		vsi->irqs_ready = false;
3990 		for (i = 0; i < vsi->num_q_vectors; i++) {
3991 			u16 vector = i + base;
3992 
3993 			/* free only the irqs that were actually requested */
3994 			if (!vsi->q_vectors[i] ||
3995 			    !vsi->q_vectors[i]->num_ringpairs)
3996 				continue;
3997 
3998 			/* clear the affinity_mask in the IRQ descriptor */
3999 			irq_set_affinity_hint(pf->msix_entries[vector].vector,
4000 					      NULL);
4001 			free_irq(pf->msix_entries[vector].vector,
4002 				 vsi->q_vectors[i]);
4003 
4004 			/* Tear down the interrupt queue link list
4005 			 *
4006 			 * We know that they come in pairs and always
4007 			 * the Rx first, then the Tx.  To clear the
4008 			 * link list, stick the EOL value into the
4009 			 * next_q field of the registers.
4010 			 */
4011 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4012 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4013 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4014 			val |= I40E_QUEUE_END_OF_LIST
4015 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4016 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4017 
4018 			while (qp != I40E_QUEUE_END_OF_LIST) {
4019 				u32 next;
4020 
4021 				val = rd32(hw, I40E_QINT_RQCTL(qp));
4022 
4023 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4024 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4025 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4026 					 I40E_QINT_RQCTL_INTEVENT_MASK);
4027 
4028 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4029 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4030 
4031 				wr32(hw, I40E_QINT_RQCTL(qp), val);
4032 
4033 				val = rd32(hw, I40E_QINT_TQCTL(qp));
4034 
4035 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4036 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4037 
4038 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4039 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4040 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4041 					 I40E_QINT_TQCTL_INTEVENT_MASK);
4042 
4043 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4044 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4045 
4046 				wr32(hw, I40E_QINT_TQCTL(qp), val);
4047 				qp = next;
4048 			}
4049 		}
4050 	} else {
4051 		free_irq(pf->pdev->irq, pf);
4052 
4053 		val = rd32(hw, I40E_PFINT_LNKLST0);
4054 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4055 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4056 		val |= I40E_QUEUE_END_OF_LIST
4057 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4058 		wr32(hw, I40E_PFINT_LNKLST0, val);
4059 
4060 		val = rd32(hw, I40E_QINT_RQCTL(qp));
4061 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4062 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4063 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4064 			 I40E_QINT_RQCTL_INTEVENT_MASK);
4065 
4066 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4067 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4068 
4069 		wr32(hw, I40E_QINT_RQCTL(qp), val);
4070 
4071 		val = rd32(hw, I40E_QINT_TQCTL(qp));
4072 
4073 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4074 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4075 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4076 			 I40E_QINT_TQCTL_INTEVENT_MASK);
4077 
4078 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4079 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4080 
4081 		wr32(hw, I40E_QINT_TQCTL(qp), val);
4082 	}
4083 }
4084 
4085 /**
4086  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4087  * @vsi: the VSI being configured
4088  * @v_idx: Index of vector to be freed
4089  *
4090  * This function frees the memory allocated to the q_vector.  In addition if
4091  * NAPI is enabled it will delete any references to the NAPI struct prior
4092  * to freeing the q_vector.
4093  **/
4094 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4095 {
4096 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4097 	struct i40e_ring *ring;
4098 
4099 	if (!q_vector)
4100 		return;
4101 
4102 	/* disassociate q_vector from rings */
4103 	i40e_for_each_ring(ring, q_vector->tx)
4104 		ring->q_vector = NULL;
4105 
4106 	i40e_for_each_ring(ring, q_vector->rx)
4107 		ring->q_vector = NULL;
4108 
4109 	/* only VSI w/ an associated netdev is set up w/ NAPI */
4110 	if (vsi->netdev)
4111 		netif_napi_del(&q_vector->napi);
4112 
4113 	vsi->q_vectors[v_idx] = NULL;
4114 
4115 	kfree_rcu(q_vector, rcu);
4116 }
4117 
4118 /**
4119  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4120  * @vsi: the VSI being un-configured
4121  *
4122  * This frees the memory allocated to the q_vectors and
4123  * deletes references to the NAPI struct.
4124  **/
4125 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4126 {
4127 	int v_idx;
4128 
4129 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4130 		i40e_free_q_vector(vsi, v_idx);
4131 }
4132 
4133 /**
4134  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4135  * @pf: board private structure
4136  **/
4137 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4138 {
4139 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4140 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4141 		pci_disable_msix(pf->pdev);
4142 		kfree(pf->msix_entries);
4143 		pf->msix_entries = NULL;
4144 		kfree(pf->irq_pile);
4145 		pf->irq_pile = NULL;
4146 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4147 		pci_disable_msi(pf->pdev);
4148 	}
4149 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4150 }
4151 
4152 /**
4153  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4154  * @pf: board private structure
4155  *
4156  * We go through and clear interrupt specific resources and reset the structure
4157  * to pre-load conditions
4158  **/
4159 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4160 {
4161 	int i;
4162 
4163 	i40e_stop_misc_vector(pf);
4164 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4165 		synchronize_irq(pf->msix_entries[0].vector);
4166 		free_irq(pf->msix_entries[0].vector, pf);
4167 	}
4168 
4169 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4170 	for (i = 0; i < pf->num_alloc_vsi; i++)
4171 		if (pf->vsi[i])
4172 			i40e_vsi_free_q_vectors(pf->vsi[i]);
4173 	i40e_reset_interrupt_capability(pf);
4174 }
4175 
4176 /**
4177  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4178  * @vsi: the VSI being configured
4179  **/
4180 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4181 {
4182 	int q_idx;
4183 
4184 	if (!vsi->netdev)
4185 		return;
4186 
4187 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4188 		napi_enable(&vsi->q_vectors[q_idx]->napi);
4189 }
4190 
4191 /**
4192  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4193  * @vsi: the VSI being configured
4194  **/
4195 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4196 {
4197 	int q_idx;
4198 
4199 	if (!vsi->netdev)
4200 		return;
4201 
4202 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4203 		napi_disable(&vsi->q_vectors[q_idx]->napi);
4204 }
4205 
4206 /**
4207  * i40e_vsi_close - Shut down a VSI
4208  * @vsi: the vsi to be quelled
4209  **/
4210 static void i40e_vsi_close(struct i40e_vsi *vsi)
4211 {
4212 	if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4213 		i40e_down(vsi);
4214 	i40e_vsi_free_irq(vsi);
4215 	i40e_vsi_free_tx_resources(vsi);
4216 	i40e_vsi_free_rx_resources(vsi);
4217 	vsi->current_netdev_flags = 0;
4218 }
4219 
4220 /**
4221  * i40e_quiesce_vsi - Pause a given VSI
4222  * @vsi: the VSI being paused
4223  **/
4224 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4225 {
4226 	if (test_bit(__I40E_DOWN, &vsi->state))
4227 		return;
4228 
4229 	/* No need to disable FCoE VSI when Tx suspended */
4230 	if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4231 	    vsi->type == I40E_VSI_FCOE) {
4232 		dev_dbg(&vsi->back->pdev->dev,
4233 			 "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4234 		return;
4235 	}
4236 
4237 	set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4238 	if (vsi->netdev && netif_running(vsi->netdev))
4239 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4240 	else
4241 		i40e_vsi_close(vsi);
4242 }
4243 
4244 /**
4245  * i40e_unquiesce_vsi - Resume a given VSI
4246  * @vsi: the VSI being resumed
4247  **/
4248 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4249 {
4250 	if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4251 		return;
4252 
4253 	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4254 	if (vsi->netdev && netif_running(vsi->netdev))
4255 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4256 	else
4257 		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4258 }
4259 
4260 /**
4261  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4262  * @pf: the PF
4263  **/
4264 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4265 {
4266 	int v;
4267 
4268 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4269 		if (pf->vsi[v])
4270 			i40e_quiesce_vsi(pf->vsi[v]);
4271 	}
4272 }
4273 
4274 /**
4275  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4276  * @pf: the PF
4277  **/
4278 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4279 {
4280 	int v;
4281 
4282 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4283 		if (pf->vsi[v])
4284 			i40e_unquiesce_vsi(pf->vsi[v]);
4285 	}
4286 }
4287 
4288 #ifdef CONFIG_I40E_DCB
4289 /**
4290  * i40e_vsi_wait_txq_disabled - Wait for VSI's queues to be disabled
4291  * @vsi: the VSI being configured
4292  *
4293  * This function waits for the given VSI's Tx queues to be disabled.
4294  **/
4295 static int i40e_vsi_wait_txq_disabled(struct i40e_vsi *vsi)
4296 {
4297 	struct i40e_pf *pf = vsi->back;
4298 	int i, pf_q, ret;
4299 
4300 	pf_q = vsi->base_queue;
4301 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4302 		/* Check and wait for the disable status of the queue */
4303 		ret = i40e_pf_txq_wait(pf, pf_q, false);
4304 		if (ret) {
4305 			dev_info(&pf->pdev->dev,
4306 				 "VSI seid %d Tx ring %d disable timeout\n",
4307 				 vsi->seid, pf_q);
4308 			return ret;
4309 		}
4310 	}
4311 
4312 	return 0;
4313 }
4314 
4315 /**
4316  * i40e_pf_wait_txq_disabled - Wait for all queues of PF VSIs to be disabled
4317  * @pf: the PF
4318  *
4319  * This function waits for the Tx queues to be in disabled state for all the
4320  * VSIs that are managed by this PF.
4321  **/
4322 static int i40e_pf_wait_txq_disabled(struct i40e_pf *pf)
4323 {
4324 	int v, ret = 0;
4325 
4326 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4327 		/* No need to wait for FCoE VSI queues */
4328 		if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4329 			ret = i40e_vsi_wait_txq_disabled(pf->vsi[v]);
4330 			if (ret)
4331 				break;
4332 		}
4333 	}
4334 
4335 	return ret;
4336 }
4337 
4338 #endif
4339 
4340 /**
4341  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4342  * @q_idx: TX queue number
4343  * @vsi: Pointer to VSI struct
4344  *
4345  * This function checks specified queue for given VSI. Detects hung condition.
4346  * Sets hung bit since it is two step process. Before next run of service task
4347  * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4348  * hung condition remain unchanged and during subsequent run, this function
4349  * issues SW interrupt to recover from hung condition.
4350  **/
4351 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4352 {
4353 	struct i40e_ring *tx_ring = NULL;
4354 	struct i40e_pf	*pf;
4355 	u32 head, val, tx_pending;
4356 	int i;
4357 
4358 	pf = vsi->back;
4359 
4360 	/* now that we have an index, find the tx_ring struct */
4361 	for (i = 0; i < vsi->num_queue_pairs; i++) {
4362 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4363 			if (q_idx == vsi->tx_rings[i]->queue_index) {
4364 				tx_ring = vsi->tx_rings[i];
4365 				break;
4366 			}
4367 		}
4368 	}
4369 
4370 	if (!tx_ring)
4371 		return;
4372 
4373 	/* Read interrupt register */
4374 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4375 		val = rd32(&pf->hw,
4376 			   I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4377 					       tx_ring->vsi->base_vector - 1));
4378 	else
4379 		val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4380 
4381 	/* Bail out if interrupts are disabled because napi_poll
4382 	 * execution in-progress or will get scheduled soon.
4383 	 * napi_poll cleans TX and RX queues and updates 'next_to_clean'.
4384 	 */
4385 	if (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))
4386 		return;
4387 
4388 	head = i40e_get_head(tx_ring);
4389 
4390 	tx_pending = i40e_get_tx_pending(tx_ring);
4391 
4392 	/* HW is done executing descriptors, updated HEAD write back,
4393 	 * but SW hasn't processed those descriptors. If interrupt is
4394 	 * not generated from this point ON, it could result into
4395 	 * dev_watchdog detecting timeout on those netdev_queue,
4396 	 * hence proactively trigger SW interrupt.
4397 	 */
4398 	if (tx_pending) {
4399 		/* NAPI Poll didn't run and clear since it was set */
4400 		if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4401 				       &tx_ring->q_vector->hung_detected)) {
4402 			netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4403 				    vsi->seid, q_idx, tx_pending,
4404 				    tx_ring->next_to_clean, head,
4405 				    tx_ring->next_to_use,
4406 				    readl(tx_ring->tail));
4407 			netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4408 				    vsi->seid, q_idx, val);
4409 			i40e_force_wb(vsi, tx_ring->q_vector);
4410 		} else {
4411 			/* First Chance - detected possible hung */
4412 			set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4413 				&tx_ring->q_vector->hung_detected);
4414 		}
4415 	}
4416 }
4417 
4418 /**
4419  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4420  * @pf:  pointer to PF struct
4421  *
4422  * LAN VSI has netdev and netdev has TX queues. This function is to check
4423  * each of those TX queues if they are hung, trigger recovery by issuing
4424  * SW interrupt.
4425  **/
4426 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4427 {
4428 	struct net_device *netdev;
4429 	struct i40e_vsi *vsi;
4430 	int i;
4431 
4432 	/* Only for LAN VSI */
4433 	vsi = pf->vsi[pf->lan_vsi];
4434 
4435 	if (!vsi)
4436 		return;
4437 
4438 	/* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4439 	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4440 	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4441 		return;
4442 
4443 	/* Make sure type is MAIN VSI */
4444 	if (vsi->type != I40E_VSI_MAIN)
4445 		return;
4446 
4447 	netdev = vsi->netdev;
4448 	if (!netdev)
4449 		return;
4450 
4451 	/* Bail out if netif_carrier is not OK */
4452 	if (!netif_carrier_ok(netdev))
4453 		return;
4454 
4455 	/* Go thru' TX queues for netdev */
4456 	for (i = 0; i < netdev->num_tx_queues; i++) {
4457 		struct netdev_queue *q;
4458 
4459 		q = netdev_get_tx_queue(netdev, i);
4460 		if (q)
4461 			i40e_detect_recover_hung_queue(i, vsi);
4462 	}
4463 }
4464 
4465 /**
4466  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4467  * @pf: pointer to PF
4468  *
4469  * Get TC map for ISCSI PF type that will include iSCSI TC
4470  * and LAN TC.
4471  **/
4472 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4473 {
4474 	struct i40e_dcb_app_priority_table app;
4475 	struct i40e_hw *hw = &pf->hw;
4476 	u8 enabled_tc = 1; /* TC0 is always enabled */
4477 	u8 tc, i;
4478 	/* Get the iSCSI APP TLV */
4479 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4480 
4481 	for (i = 0; i < dcbcfg->numapps; i++) {
4482 		app = dcbcfg->app[i];
4483 		if (app.selector == I40E_APP_SEL_TCPIP &&
4484 		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4485 			tc = dcbcfg->etscfg.prioritytable[app.priority];
4486 			enabled_tc |= BIT(tc);
4487 			break;
4488 		}
4489 	}
4490 
4491 	return enabled_tc;
4492 }
4493 
4494 /**
4495  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4496  * @dcbcfg: the corresponding DCBx configuration structure
4497  *
4498  * Return the number of TCs from given DCBx configuration
4499  **/
4500 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4501 {
4502 	u8 num_tc = 0;
4503 	int i;
4504 
4505 	/* Scan the ETS Config Priority Table to find
4506 	 * traffic class enabled for a given priority
4507 	 * and use the traffic class index to get the
4508 	 * number of traffic classes enabled
4509 	 */
4510 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4511 		if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4512 			num_tc = dcbcfg->etscfg.prioritytable[i];
4513 	}
4514 
4515 	/* Traffic class index starts from zero so
4516 	 * increment to return the actual count
4517 	 */
4518 	return num_tc + 1;
4519 }
4520 
4521 /**
4522  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4523  * @dcbcfg: the corresponding DCBx configuration structure
4524  *
4525  * Query the current DCB configuration and return the number of
4526  * traffic classes enabled from the given DCBX config
4527  **/
4528 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4529 {
4530 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4531 	u8 enabled_tc = 1;
4532 	u8 i;
4533 
4534 	for (i = 0; i < num_tc; i++)
4535 		enabled_tc |= BIT(i);
4536 
4537 	return enabled_tc;
4538 }
4539 
4540 /**
4541  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4542  * @pf: PF being queried
4543  *
4544  * Return number of traffic classes enabled for the given PF
4545  **/
4546 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4547 {
4548 	struct i40e_hw *hw = &pf->hw;
4549 	u8 i, enabled_tc;
4550 	u8 num_tc = 0;
4551 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4552 
4553 	/* If DCB is not enabled then always in single TC */
4554 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4555 		return 1;
4556 
4557 	/* SFP mode will be enabled for all TCs on port */
4558 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4559 		return i40e_dcb_get_num_tc(dcbcfg);
4560 
4561 	/* MFP mode return count of enabled TCs for this PF */
4562 	if (pf->hw.func_caps.iscsi)
4563 		enabled_tc =  i40e_get_iscsi_tc_map(pf);
4564 	else
4565 		return 1; /* Only TC0 */
4566 
4567 	/* At least have TC0 */
4568 	enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4569 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4570 		if (enabled_tc & BIT(i))
4571 			num_tc++;
4572 	}
4573 	return num_tc;
4574 }
4575 
4576 /**
4577  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4578  * @pf: PF being queried
4579  *
4580  * Return a bitmap for first enabled traffic class for this PF.
4581  **/
4582 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4583 {
4584 	u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4585 	u8 i = 0;
4586 
4587 	if (!enabled_tc)
4588 		return 0x1; /* TC0 */
4589 
4590 	/* Find the first enabled TC */
4591 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4592 		if (enabled_tc & BIT(i))
4593 			break;
4594 	}
4595 
4596 	return BIT(i);
4597 }
4598 
4599 /**
4600  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4601  * @pf: PF being queried
4602  *
4603  * Return a bitmap for enabled traffic classes for this PF.
4604  **/
4605 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4606 {
4607 	/* If DCB is not enabled for this PF then just return default TC */
4608 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4609 		return i40e_pf_get_default_tc(pf);
4610 
4611 	/* SFP mode we want PF to be enabled for all TCs */
4612 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4613 		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4614 
4615 	/* MFP enabled and iSCSI PF type */
4616 	if (pf->hw.func_caps.iscsi)
4617 		return i40e_get_iscsi_tc_map(pf);
4618 	else
4619 		return i40e_pf_get_default_tc(pf);
4620 }
4621 
4622 /**
4623  * i40e_vsi_get_bw_info - Query VSI BW Information
4624  * @vsi: the VSI being queried
4625  *
4626  * Returns 0 on success, negative value on failure
4627  **/
4628 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4629 {
4630 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4631 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4632 	struct i40e_pf *pf = vsi->back;
4633 	struct i40e_hw *hw = &pf->hw;
4634 	i40e_status ret;
4635 	u32 tc_bw_max;
4636 	int i;
4637 
4638 	/* Get the VSI level BW configuration */
4639 	ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4640 	if (ret) {
4641 		dev_info(&pf->pdev->dev,
4642 			 "couldn't get PF vsi bw config, err %s aq_err %s\n",
4643 			 i40e_stat_str(&pf->hw, ret),
4644 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4645 		return -EINVAL;
4646 	}
4647 
4648 	/* Get the VSI level BW configuration per TC */
4649 	ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4650 					       NULL);
4651 	if (ret) {
4652 		dev_info(&pf->pdev->dev,
4653 			 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4654 			 i40e_stat_str(&pf->hw, ret),
4655 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4656 		return -EINVAL;
4657 	}
4658 
4659 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4660 		dev_info(&pf->pdev->dev,
4661 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4662 			 bw_config.tc_valid_bits,
4663 			 bw_ets_config.tc_valid_bits);
4664 		/* Still continuing */
4665 	}
4666 
4667 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4668 	vsi->bw_max_quanta = bw_config.max_bw;
4669 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4670 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4671 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4672 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4673 		vsi->bw_ets_limit_credits[i] =
4674 					le16_to_cpu(bw_ets_config.credits[i]);
4675 		/* 3 bits out of 4 for each TC */
4676 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4677 	}
4678 
4679 	return 0;
4680 }
4681 
4682 /**
4683  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4684  * @vsi: the VSI being configured
4685  * @enabled_tc: TC bitmap
4686  * @bw_credits: BW shared credits per TC
4687  *
4688  * Returns 0 on success, negative value on failure
4689  **/
4690 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4691 				       u8 *bw_share)
4692 {
4693 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4694 	i40e_status ret;
4695 	int i;
4696 
4697 	bw_data.tc_valid_bits = enabled_tc;
4698 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4699 		bw_data.tc_bw_credits[i] = bw_share[i];
4700 
4701 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4702 				       NULL);
4703 	if (ret) {
4704 		dev_info(&vsi->back->pdev->dev,
4705 			 "AQ command Config VSI BW allocation per TC failed = %d\n",
4706 			 vsi->back->hw.aq.asq_last_status);
4707 		return -EINVAL;
4708 	}
4709 
4710 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4711 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4712 
4713 	return 0;
4714 }
4715 
4716 /**
4717  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4718  * @vsi: the VSI being configured
4719  * @enabled_tc: TC map to be enabled
4720  *
4721  **/
4722 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4723 {
4724 	struct net_device *netdev = vsi->netdev;
4725 	struct i40e_pf *pf = vsi->back;
4726 	struct i40e_hw *hw = &pf->hw;
4727 	u8 netdev_tc = 0;
4728 	int i;
4729 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4730 
4731 	if (!netdev)
4732 		return;
4733 
4734 	if (!enabled_tc) {
4735 		netdev_reset_tc(netdev);
4736 		return;
4737 	}
4738 
4739 	/* Set up actual enabled TCs on the VSI */
4740 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4741 		return;
4742 
4743 	/* set per TC queues for the VSI */
4744 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4745 		/* Only set TC queues for enabled tcs
4746 		 *
4747 		 * e.g. For a VSI that has TC0 and TC3 enabled the
4748 		 * enabled_tc bitmap would be 0x00001001; the driver
4749 		 * will set the numtc for netdev as 2 that will be
4750 		 * referenced by the netdev layer as TC 0 and 1.
4751 		 */
4752 		if (vsi->tc_config.enabled_tc & BIT(i))
4753 			netdev_set_tc_queue(netdev,
4754 					vsi->tc_config.tc_info[i].netdev_tc,
4755 					vsi->tc_config.tc_info[i].qcount,
4756 					vsi->tc_config.tc_info[i].qoffset);
4757 	}
4758 
4759 	/* Assign UP2TC map for the VSI */
4760 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4761 		/* Get the actual TC# for the UP */
4762 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4763 		/* Get the mapped netdev TC# for the UP */
4764 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4765 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
4766 	}
4767 }
4768 
4769 /**
4770  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4771  * @vsi: the VSI being configured
4772  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4773  **/
4774 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4775 				      struct i40e_vsi_context *ctxt)
4776 {
4777 	/* copy just the sections touched not the entire info
4778 	 * since not all sections are valid as returned by
4779 	 * update vsi params
4780 	 */
4781 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
4782 	memcpy(&vsi->info.queue_mapping,
4783 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4784 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4785 	       sizeof(vsi->info.tc_mapping));
4786 }
4787 
4788 /**
4789  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4790  * @vsi: VSI to be configured
4791  * @enabled_tc: TC bitmap
4792  *
4793  * This configures a particular VSI for TCs that are mapped to the
4794  * given TC bitmap. It uses default bandwidth share for TCs across
4795  * VSIs to configure TC for a particular VSI.
4796  *
4797  * NOTE:
4798  * It is expected that the VSI queues have been quisced before calling
4799  * this function.
4800  **/
4801 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4802 {
4803 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4804 	struct i40e_vsi_context ctxt;
4805 	int ret = 0;
4806 	int i;
4807 
4808 	/* Check if enabled_tc is same as existing or new TCs */
4809 	if (vsi->tc_config.enabled_tc == enabled_tc)
4810 		return ret;
4811 
4812 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
4813 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4814 		if (enabled_tc & BIT(i))
4815 			bw_share[i] = 1;
4816 	}
4817 
4818 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4819 	if (ret) {
4820 		dev_info(&vsi->back->pdev->dev,
4821 			 "Failed configuring TC map %d for VSI %d\n",
4822 			 enabled_tc, vsi->seid);
4823 		goto out;
4824 	}
4825 
4826 	/* Update Queue Pairs Mapping for currently enabled UPs */
4827 	ctxt.seid = vsi->seid;
4828 	ctxt.pf_num = vsi->back->hw.pf_id;
4829 	ctxt.vf_num = 0;
4830 	ctxt.uplink_seid = vsi->uplink_seid;
4831 	ctxt.info = vsi->info;
4832 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4833 
4834 	/* Update the VSI after updating the VSI queue-mapping information */
4835 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4836 	if (ret) {
4837 		dev_info(&vsi->back->pdev->dev,
4838 			 "Update vsi tc config failed, err %s aq_err %s\n",
4839 			 i40e_stat_str(&vsi->back->hw, ret),
4840 			 i40e_aq_str(&vsi->back->hw,
4841 				     vsi->back->hw.aq.asq_last_status));
4842 		goto out;
4843 	}
4844 	/* update the local VSI info with updated queue map */
4845 	i40e_vsi_update_queue_map(vsi, &ctxt);
4846 	vsi->info.valid_sections = 0;
4847 
4848 	/* Update current VSI BW information */
4849 	ret = i40e_vsi_get_bw_info(vsi);
4850 	if (ret) {
4851 		dev_info(&vsi->back->pdev->dev,
4852 			 "Failed updating vsi bw info, err %s aq_err %s\n",
4853 			 i40e_stat_str(&vsi->back->hw, ret),
4854 			 i40e_aq_str(&vsi->back->hw,
4855 				     vsi->back->hw.aq.asq_last_status));
4856 		goto out;
4857 	}
4858 
4859 	/* Update the netdev TC setup */
4860 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4861 out:
4862 	return ret;
4863 }
4864 
4865 /**
4866  * i40e_veb_config_tc - Configure TCs for given VEB
4867  * @veb: given VEB
4868  * @enabled_tc: TC bitmap
4869  *
4870  * Configures given TC bitmap for VEB (switching) element
4871  **/
4872 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4873 {
4874 	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4875 	struct i40e_pf *pf = veb->pf;
4876 	int ret = 0;
4877 	int i;
4878 
4879 	/* No TCs or already enabled TCs just return */
4880 	if (!enabled_tc || veb->enabled_tc == enabled_tc)
4881 		return ret;
4882 
4883 	bw_data.tc_valid_bits = enabled_tc;
4884 	/* bw_data.absolute_credits is not set (relative) */
4885 
4886 	/* Enable ETS TCs with equal BW Share for now */
4887 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4888 		if (enabled_tc & BIT(i))
4889 			bw_data.tc_bw_share_credits[i] = 1;
4890 	}
4891 
4892 	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4893 						   &bw_data, NULL);
4894 	if (ret) {
4895 		dev_info(&pf->pdev->dev,
4896 			 "VEB bw config failed, err %s aq_err %s\n",
4897 			 i40e_stat_str(&pf->hw, ret),
4898 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4899 		goto out;
4900 	}
4901 
4902 	/* Update the BW information */
4903 	ret = i40e_veb_get_bw_info(veb);
4904 	if (ret) {
4905 		dev_info(&pf->pdev->dev,
4906 			 "Failed getting veb bw config, err %s aq_err %s\n",
4907 			 i40e_stat_str(&pf->hw, ret),
4908 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4909 	}
4910 
4911 out:
4912 	return ret;
4913 }
4914 
4915 #ifdef CONFIG_I40E_DCB
4916 /**
4917  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4918  * @pf: PF struct
4919  *
4920  * Reconfigure VEB/VSIs on a given PF; it is assumed that
4921  * the caller would've quiesce all the VSIs before calling
4922  * this function
4923  **/
4924 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4925 {
4926 	u8 tc_map = 0;
4927 	int ret;
4928 	u8 v;
4929 
4930 	/* Enable the TCs available on PF to all VEBs */
4931 	tc_map = i40e_pf_get_tc_map(pf);
4932 	for (v = 0; v < I40E_MAX_VEB; v++) {
4933 		if (!pf->veb[v])
4934 			continue;
4935 		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4936 		if (ret) {
4937 			dev_info(&pf->pdev->dev,
4938 				 "Failed configuring TC for VEB seid=%d\n",
4939 				 pf->veb[v]->seid);
4940 			/* Will try to configure as many components */
4941 		}
4942 	}
4943 
4944 	/* Update each VSI */
4945 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4946 		if (!pf->vsi[v])
4947 			continue;
4948 
4949 		/* - Enable all TCs for the LAN VSI
4950 #ifdef I40E_FCOE
4951 		 * - For FCoE VSI only enable the TC configured
4952 		 *   as per the APP TLV
4953 #endif
4954 		 * - For all others keep them at TC0 for now
4955 		 */
4956 		if (v == pf->lan_vsi)
4957 			tc_map = i40e_pf_get_tc_map(pf);
4958 		else
4959 			tc_map = i40e_pf_get_default_tc(pf);
4960 #ifdef I40E_FCOE
4961 		if (pf->vsi[v]->type == I40E_VSI_FCOE)
4962 			tc_map = i40e_get_fcoe_tc_map(pf);
4963 #endif /* #ifdef I40E_FCOE */
4964 
4965 		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4966 		if (ret) {
4967 			dev_info(&pf->pdev->dev,
4968 				 "Failed configuring TC for VSI seid=%d\n",
4969 				 pf->vsi[v]->seid);
4970 			/* Will try to configure as many components */
4971 		} else {
4972 			/* Re-configure VSI vectors based on updated TC map */
4973 			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4974 			if (pf->vsi[v]->netdev)
4975 				i40e_dcbnl_set_all(pf->vsi[v]);
4976 		}
4977 	}
4978 }
4979 
4980 /**
4981  * i40e_resume_port_tx - Resume port Tx
4982  * @pf: PF struct
4983  *
4984  * Resume a port's Tx and issue a PF reset in case of failure to
4985  * resume.
4986  **/
4987 static int i40e_resume_port_tx(struct i40e_pf *pf)
4988 {
4989 	struct i40e_hw *hw = &pf->hw;
4990 	int ret;
4991 
4992 	ret = i40e_aq_resume_port_tx(hw, NULL);
4993 	if (ret) {
4994 		dev_info(&pf->pdev->dev,
4995 			 "Resume Port Tx failed, err %s aq_err %s\n",
4996 			  i40e_stat_str(&pf->hw, ret),
4997 			  i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4998 		/* Schedule PF reset to recover */
4999 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5000 		i40e_service_event_schedule(pf);
5001 	}
5002 
5003 	return ret;
5004 }
5005 
5006 /**
5007  * i40e_init_pf_dcb - Initialize DCB configuration
5008  * @pf: PF being configured
5009  *
5010  * Query the current DCB configuration and cache it
5011  * in the hardware structure
5012  **/
5013 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5014 {
5015 	struct i40e_hw *hw = &pf->hw;
5016 	int err = 0;
5017 
5018 	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5019 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
5020 	    (pf->hw.aq.fw_maj_ver < 4))
5021 		goto out;
5022 
5023 	/* Get the initial DCB configuration */
5024 	err = i40e_init_dcb(hw);
5025 	if (!err) {
5026 		/* Device/Function is not DCBX capable */
5027 		if ((!hw->func_caps.dcb) ||
5028 		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5029 			dev_info(&pf->pdev->dev,
5030 				 "DCBX offload is not supported or is disabled for this PF.\n");
5031 
5032 			if (pf->flags & I40E_FLAG_MFP_ENABLED)
5033 				goto out;
5034 
5035 		} else {
5036 			/* When status is not DISABLED then DCBX in FW */
5037 			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5038 				       DCB_CAP_DCBX_VER_IEEE;
5039 
5040 			pf->flags |= I40E_FLAG_DCB_CAPABLE;
5041 			/* Enable DCB tagging only when more than one TC */
5042 			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5043 				pf->flags |= I40E_FLAG_DCB_ENABLED;
5044 			dev_dbg(&pf->pdev->dev,
5045 				"DCBX offload is supported for this PF.\n");
5046 		}
5047 	} else {
5048 		dev_info(&pf->pdev->dev,
5049 			 "Query for DCB configuration failed, err %s aq_err %s\n",
5050 			 i40e_stat_str(&pf->hw, err),
5051 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5052 	}
5053 
5054 out:
5055 	return err;
5056 }
5057 #endif /* CONFIG_I40E_DCB */
5058 #define SPEED_SIZE 14
5059 #define FC_SIZE 8
5060 /**
5061  * i40e_print_link_message - print link up or down
5062  * @vsi: the VSI for which link needs a message
5063  */
5064 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5065 {
5066 	char *speed = "Unknown";
5067 	char *fc = "Unknown";
5068 
5069 	if (vsi->current_isup == isup)
5070 		return;
5071 	vsi->current_isup = isup;
5072 	if (!isup) {
5073 		netdev_info(vsi->netdev, "NIC Link is Down\n");
5074 		return;
5075 	}
5076 
5077 	/* Warn user if link speed on NPAR enabled partition is not at
5078 	 * least 10GB
5079 	 */
5080 	if (vsi->back->hw.func_caps.npar_enable &&
5081 	    (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5082 	     vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5083 		netdev_warn(vsi->netdev,
5084 			    "The partition detected link speed that is less than 10Gbps\n");
5085 
5086 	switch (vsi->back->hw.phy.link_info.link_speed) {
5087 	case I40E_LINK_SPEED_40GB:
5088 		speed = "40 G";
5089 		break;
5090 	case I40E_LINK_SPEED_20GB:
5091 		speed = "20 G";
5092 		break;
5093 	case I40E_LINK_SPEED_10GB:
5094 		speed = "10 G";
5095 		break;
5096 	case I40E_LINK_SPEED_1GB:
5097 		speed = "1000 M";
5098 		break;
5099 	case I40E_LINK_SPEED_100MB:
5100 		speed = "100 M";
5101 		break;
5102 	default:
5103 		break;
5104 	}
5105 
5106 	switch (vsi->back->hw.fc.current_mode) {
5107 	case I40E_FC_FULL:
5108 		fc = "RX/TX";
5109 		break;
5110 	case I40E_FC_TX_PAUSE:
5111 		fc = "TX";
5112 		break;
5113 	case I40E_FC_RX_PAUSE:
5114 		fc = "RX";
5115 		break;
5116 	default:
5117 		fc = "None";
5118 		break;
5119 	}
5120 
5121 	netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5122 		    speed, fc);
5123 }
5124 
5125 /**
5126  * i40e_up_complete - Finish the last steps of bringing up a connection
5127  * @vsi: the VSI being configured
5128  **/
5129 static int i40e_up_complete(struct i40e_vsi *vsi)
5130 {
5131 	struct i40e_pf *pf = vsi->back;
5132 	int err;
5133 
5134 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5135 		i40e_vsi_configure_msix(vsi);
5136 	else
5137 		i40e_configure_msi_and_legacy(vsi);
5138 
5139 	/* start rings */
5140 	err = i40e_vsi_control_rings(vsi, true);
5141 	if (err)
5142 		return err;
5143 
5144 	clear_bit(__I40E_DOWN, &vsi->state);
5145 	i40e_napi_enable_all(vsi);
5146 	i40e_vsi_enable_irq(vsi);
5147 
5148 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5149 	    (vsi->netdev)) {
5150 		i40e_print_link_message(vsi, true);
5151 		netif_tx_start_all_queues(vsi->netdev);
5152 		netif_carrier_on(vsi->netdev);
5153 	} else if (vsi->netdev) {
5154 		i40e_print_link_message(vsi, false);
5155 		/* need to check for qualified module here*/
5156 		if ((pf->hw.phy.link_info.link_info &
5157 			I40E_AQ_MEDIA_AVAILABLE) &&
5158 		    (!(pf->hw.phy.link_info.an_info &
5159 			I40E_AQ_QUALIFIED_MODULE)))
5160 			netdev_err(vsi->netdev,
5161 				   "the driver failed to link because an unqualified module was detected.");
5162 	}
5163 
5164 	/* replay FDIR SB filters */
5165 	if (vsi->type == I40E_VSI_FDIR) {
5166 		/* reset fd counters */
5167 		pf->fd_add_err = pf->fd_atr_cnt = 0;
5168 		if (pf->fd_tcp_rule > 0) {
5169 			pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5170 			if (I40E_DEBUG_FD & pf->hw.debug_mask)
5171 				dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5172 			pf->fd_tcp_rule = 0;
5173 		}
5174 		i40e_fdir_filter_restore(vsi);
5175 	}
5176 	i40e_service_event_schedule(pf);
5177 
5178 	return 0;
5179 }
5180 
5181 /**
5182  * i40e_vsi_reinit_locked - Reset the VSI
5183  * @vsi: the VSI being configured
5184  *
5185  * Rebuild the ring structs after some configuration
5186  * has changed, e.g. MTU size.
5187  **/
5188 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5189 {
5190 	struct i40e_pf *pf = vsi->back;
5191 
5192 	WARN_ON(in_interrupt());
5193 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5194 		usleep_range(1000, 2000);
5195 	i40e_down(vsi);
5196 
5197 	/* Give a VF some time to respond to the reset.  The
5198 	 * two second wait is based upon the watchdog cycle in
5199 	 * the VF driver.
5200 	 */
5201 	if (vsi->type == I40E_VSI_SRIOV)
5202 		msleep(2000);
5203 	i40e_up(vsi);
5204 	clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5205 }
5206 
5207 /**
5208  * i40e_up - Bring the connection back up after being down
5209  * @vsi: the VSI being configured
5210  **/
5211 int i40e_up(struct i40e_vsi *vsi)
5212 {
5213 	int err;
5214 
5215 	err = i40e_vsi_configure(vsi);
5216 	if (!err)
5217 		err = i40e_up_complete(vsi);
5218 
5219 	return err;
5220 }
5221 
5222 /**
5223  * i40e_down - Shutdown the connection processing
5224  * @vsi: the VSI being stopped
5225  **/
5226 void i40e_down(struct i40e_vsi *vsi)
5227 {
5228 	int i;
5229 
5230 	/* It is assumed that the caller of this function
5231 	 * sets the vsi->state __I40E_DOWN bit.
5232 	 */
5233 	if (vsi->netdev) {
5234 		netif_carrier_off(vsi->netdev);
5235 		netif_tx_disable(vsi->netdev);
5236 	}
5237 	i40e_vsi_disable_irq(vsi);
5238 	i40e_vsi_control_rings(vsi, false);
5239 	i40e_napi_disable_all(vsi);
5240 
5241 	for (i = 0; i < vsi->num_queue_pairs; i++) {
5242 		i40e_clean_tx_ring(vsi->tx_rings[i]);
5243 		i40e_clean_rx_ring(vsi->rx_rings[i]);
5244 	}
5245 }
5246 
5247 /**
5248  * i40e_setup_tc - configure multiple traffic classes
5249  * @netdev: net device to configure
5250  * @tc: number of traffic classes to enable
5251  **/
5252 #ifdef I40E_FCOE
5253 int i40e_setup_tc(struct net_device *netdev, u8 tc)
5254 #else
5255 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5256 #endif
5257 {
5258 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5259 	struct i40e_vsi *vsi = np->vsi;
5260 	struct i40e_pf *pf = vsi->back;
5261 	u8 enabled_tc = 0;
5262 	int ret = -EINVAL;
5263 	int i;
5264 
5265 	/* Check if DCB enabled to continue */
5266 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5267 		netdev_info(netdev, "DCB is not enabled for adapter\n");
5268 		goto exit;
5269 	}
5270 
5271 	/* Check if MFP enabled */
5272 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5273 		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5274 		goto exit;
5275 	}
5276 
5277 	/* Check whether tc count is within enabled limit */
5278 	if (tc > i40e_pf_get_num_tc(pf)) {
5279 		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5280 		goto exit;
5281 	}
5282 
5283 	/* Generate TC map for number of tc requested */
5284 	for (i = 0; i < tc; i++)
5285 		enabled_tc |= BIT(i);
5286 
5287 	/* Requesting same TC configuration as already enabled */
5288 	if (enabled_tc == vsi->tc_config.enabled_tc)
5289 		return 0;
5290 
5291 	/* Quiesce VSI queues */
5292 	i40e_quiesce_vsi(vsi);
5293 
5294 	/* Configure VSI for enabled TCs */
5295 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
5296 	if (ret) {
5297 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5298 			    vsi->seid);
5299 		goto exit;
5300 	}
5301 
5302 	/* Unquiesce VSI */
5303 	i40e_unquiesce_vsi(vsi);
5304 
5305 exit:
5306 	return ret;
5307 }
5308 
5309 /**
5310  * i40e_open - Called when a network interface is made active
5311  * @netdev: network interface device structure
5312  *
5313  * The open entry point is called when a network interface is made
5314  * active by the system (IFF_UP).  At this point all resources needed
5315  * for transmit and receive operations are allocated, the interrupt
5316  * handler is registered with the OS, the netdev watchdog subtask is
5317  * enabled, and the stack is notified that the interface is ready.
5318  *
5319  * Returns 0 on success, negative value on failure
5320  **/
5321 int i40e_open(struct net_device *netdev)
5322 {
5323 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5324 	struct i40e_vsi *vsi = np->vsi;
5325 	struct i40e_pf *pf = vsi->back;
5326 	int err;
5327 
5328 	/* disallow open during test or if eeprom is broken */
5329 	if (test_bit(__I40E_TESTING, &pf->state) ||
5330 	    test_bit(__I40E_BAD_EEPROM, &pf->state))
5331 		return -EBUSY;
5332 
5333 	netif_carrier_off(netdev);
5334 
5335 	err = i40e_vsi_open(vsi);
5336 	if (err)
5337 		return err;
5338 
5339 	/* configure global TSO hardware offload settings */
5340 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5341 						       TCP_FLAG_FIN) >> 16);
5342 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5343 						       TCP_FLAG_FIN |
5344 						       TCP_FLAG_CWR) >> 16);
5345 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5346 
5347 #ifdef CONFIG_I40E_VXLAN
5348 	vxlan_get_rx_port(netdev);
5349 #endif
5350 #ifdef CONFIG_I40E_GENEVE
5351 	geneve_get_rx_port(netdev);
5352 #endif
5353 
5354 	return 0;
5355 }
5356 
5357 /**
5358  * i40e_vsi_open -
5359  * @vsi: the VSI to open
5360  *
5361  * Finish initialization of the VSI.
5362  *
5363  * Returns 0 on success, negative value on failure
5364  **/
5365 int i40e_vsi_open(struct i40e_vsi *vsi)
5366 {
5367 	struct i40e_pf *pf = vsi->back;
5368 	char int_name[I40E_INT_NAME_STR_LEN];
5369 	int err;
5370 
5371 	/* allocate descriptors */
5372 	err = i40e_vsi_setup_tx_resources(vsi);
5373 	if (err)
5374 		goto err_setup_tx;
5375 	err = i40e_vsi_setup_rx_resources(vsi);
5376 	if (err)
5377 		goto err_setup_rx;
5378 
5379 	err = i40e_vsi_configure(vsi);
5380 	if (err)
5381 		goto err_setup_rx;
5382 
5383 	if (vsi->netdev) {
5384 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5385 			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5386 		err = i40e_vsi_request_irq(vsi, int_name);
5387 		if (err)
5388 			goto err_setup_rx;
5389 
5390 		/* Notify the stack of the actual queue counts. */
5391 		err = netif_set_real_num_tx_queues(vsi->netdev,
5392 						   vsi->num_queue_pairs);
5393 		if (err)
5394 			goto err_set_queues;
5395 
5396 		err = netif_set_real_num_rx_queues(vsi->netdev,
5397 						   vsi->num_queue_pairs);
5398 		if (err)
5399 			goto err_set_queues;
5400 
5401 	} else if (vsi->type == I40E_VSI_FDIR) {
5402 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5403 			 dev_driver_string(&pf->pdev->dev),
5404 			 dev_name(&pf->pdev->dev));
5405 		err = i40e_vsi_request_irq(vsi, int_name);
5406 
5407 	} else {
5408 		err = -EINVAL;
5409 		goto err_setup_rx;
5410 	}
5411 
5412 	err = i40e_up_complete(vsi);
5413 	if (err)
5414 		goto err_up_complete;
5415 
5416 	return 0;
5417 
5418 err_up_complete:
5419 	i40e_down(vsi);
5420 err_set_queues:
5421 	i40e_vsi_free_irq(vsi);
5422 err_setup_rx:
5423 	i40e_vsi_free_rx_resources(vsi);
5424 err_setup_tx:
5425 	i40e_vsi_free_tx_resources(vsi);
5426 	if (vsi == pf->vsi[pf->lan_vsi])
5427 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5428 
5429 	return err;
5430 }
5431 
5432 /**
5433  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5434  * @pf: Pointer to PF
5435  *
5436  * This function destroys the hlist where all the Flow Director
5437  * filters were saved.
5438  **/
5439 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5440 {
5441 	struct i40e_fdir_filter *filter;
5442 	struct hlist_node *node2;
5443 
5444 	hlist_for_each_entry_safe(filter, node2,
5445 				  &pf->fdir_filter_list, fdir_node) {
5446 		hlist_del(&filter->fdir_node);
5447 		kfree(filter);
5448 	}
5449 	pf->fdir_pf_active_filters = 0;
5450 }
5451 
5452 /**
5453  * i40e_close - Disables a network interface
5454  * @netdev: network interface device structure
5455  *
5456  * The close entry point is called when an interface is de-activated
5457  * by the OS.  The hardware is still under the driver's control, but
5458  * this netdev interface is disabled.
5459  *
5460  * Returns 0, this is not allowed to fail
5461  **/
5462 #ifdef I40E_FCOE
5463 int i40e_close(struct net_device *netdev)
5464 #else
5465 static int i40e_close(struct net_device *netdev)
5466 #endif
5467 {
5468 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5469 	struct i40e_vsi *vsi = np->vsi;
5470 
5471 	i40e_vsi_close(vsi);
5472 
5473 	return 0;
5474 }
5475 
5476 /**
5477  * i40e_do_reset - Start a PF or Core Reset sequence
5478  * @pf: board private structure
5479  * @reset_flags: which reset is requested
5480  *
5481  * The essential difference in resets is that the PF Reset
5482  * doesn't clear the packet buffers, doesn't reset the PE
5483  * firmware, and doesn't bother the other PFs on the chip.
5484  **/
5485 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5486 {
5487 	u32 val;
5488 
5489 	WARN_ON(in_interrupt());
5490 
5491 	if (i40e_check_asq_alive(&pf->hw))
5492 		i40e_vc_notify_reset(pf);
5493 
5494 	/* do the biggest reset indicated */
5495 	if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5496 
5497 		/* Request a Global Reset
5498 		 *
5499 		 * This will start the chip's countdown to the actual full
5500 		 * chip reset event, and a warning interrupt to be sent
5501 		 * to all PFs, including the requestor.  Our handler
5502 		 * for the warning interrupt will deal with the shutdown
5503 		 * and recovery of the switch setup.
5504 		 */
5505 		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5506 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5507 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5508 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5509 
5510 	} else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5511 
5512 		/* Request a Core Reset
5513 		 *
5514 		 * Same as Global Reset, except does *not* include the MAC/PHY
5515 		 */
5516 		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5517 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5518 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
5519 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5520 		i40e_flush(&pf->hw);
5521 
5522 	} else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5523 
5524 		/* Request a PF Reset
5525 		 *
5526 		 * Resets only the PF-specific registers
5527 		 *
5528 		 * This goes directly to the tear-down and rebuild of
5529 		 * the switch, since we need to do all the recovery as
5530 		 * for the Core Reset.
5531 		 */
5532 		dev_dbg(&pf->pdev->dev, "PFR requested\n");
5533 		i40e_handle_reset_warning(pf);
5534 
5535 	} else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5536 		int v;
5537 
5538 		/* Find the VSI(s) that requested a re-init */
5539 		dev_info(&pf->pdev->dev,
5540 			 "VSI reinit requested\n");
5541 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5542 			struct i40e_vsi *vsi = pf->vsi[v];
5543 
5544 			if (vsi != NULL &&
5545 			    test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5546 				i40e_vsi_reinit_locked(pf->vsi[v]);
5547 				clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5548 			}
5549 		}
5550 	} else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5551 		int v;
5552 
5553 		/* Find the VSI(s) that needs to be brought down */
5554 		dev_info(&pf->pdev->dev, "VSI down requested\n");
5555 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5556 			struct i40e_vsi *vsi = pf->vsi[v];
5557 
5558 			if (vsi != NULL &&
5559 			    test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5560 				set_bit(__I40E_DOWN, &vsi->state);
5561 				i40e_down(vsi);
5562 				clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5563 			}
5564 		}
5565 	} else {
5566 		dev_info(&pf->pdev->dev,
5567 			 "bad reset request 0x%08x\n", reset_flags);
5568 	}
5569 }
5570 
5571 #ifdef CONFIG_I40E_DCB
5572 /**
5573  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5574  * @pf: board private structure
5575  * @old_cfg: current DCB config
5576  * @new_cfg: new DCB config
5577  **/
5578 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5579 			    struct i40e_dcbx_config *old_cfg,
5580 			    struct i40e_dcbx_config *new_cfg)
5581 {
5582 	bool need_reconfig = false;
5583 
5584 	/* Check if ETS configuration has changed */
5585 	if (memcmp(&new_cfg->etscfg,
5586 		   &old_cfg->etscfg,
5587 		   sizeof(new_cfg->etscfg))) {
5588 		/* If Priority Table has changed reconfig is needed */
5589 		if (memcmp(&new_cfg->etscfg.prioritytable,
5590 			   &old_cfg->etscfg.prioritytable,
5591 			   sizeof(new_cfg->etscfg.prioritytable))) {
5592 			need_reconfig = true;
5593 			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5594 		}
5595 
5596 		if (memcmp(&new_cfg->etscfg.tcbwtable,
5597 			   &old_cfg->etscfg.tcbwtable,
5598 			   sizeof(new_cfg->etscfg.tcbwtable)))
5599 			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5600 
5601 		if (memcmp(&new_cfg->etscfg.tsatable,
5602 			   &old_cfg->etscfg.tsatable,
5603 			   sizeof(new_cfg->etscfg.tsatable)))
5604 			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5605 	}
5606 
5607 	/* Check if PFC configuration has changed */
5608 	if (memcmp(&new_cfg->pfc,
5609 		   &old_cfg->pfc,
5610 		   sizeof(new_cfg->pfc))) {
5611 		need_reconfig = true;
5612 		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5613 	}
5614 
5615 	/* Check if APP Table has changed */
5616 	if (memcmp(&new_cfg->app,
5617 		   &old_cfg->app,
5618 		   sizeof(new_cfg->app))) {
5619 		need_reconfig = true;
5620 		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5621 	}
5622 
5623 	dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5624 	return need_reconfig;
5625 }
5626 
5627 /**
5628  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5629  * @pf: board private structure
5630  * @e: event info posted on ARQ
5631  **/
5632 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5633 				  struct i40e_arq_event_info *e)
5634 {
5635 	struct i40e_aqc_lldp_get_mib *mib =
5636 		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5637 	struct i40e_hw *hw = &pf->hw;
5638 	struct i40e_dcbx_config tmp_dcbx_cfg;
5639 	bool need_reconfig = false;
5640 	int ret = 0;
5641 	u8 type;
5642 
5643 	/* Not DCB capable or capability disabled */
5644 	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5645 		return ret;
5646 
5647 	/* Ignore if event is not for Nearest Bridge */
5648 	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5649 		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5650 	dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5651 	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5652 		return ret;
5653 
5654 	/* Check MIB Type and return if event for Remote MIB update */
5655 	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5656 	dev_dbg(&pf->pdev->dev,
5657 		"LLDP event mib type %s\n", type ? "remote" : "local");
5658 	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5659 		/* Update the remote cached instance and return */
5660 		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5661 				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5662 				&hw->remote_dcbx_config);
5663 		goto exit;
5664 	}
5665 
5666 	/* Store the old configuration */
5667 	tmp_dcbx_cfg = hw->local_dcbx_config;
5668 
5669 	/* Reset the old DCBx configuration data */
5670 	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5671 	/* Get updated DCBX data from firmware */
5672 	ret = i40e_get_dcb_config(&pf->hw);
5673 	if (ret) {
5674 		dev_info(&pf->pdev->dev,
5675 			 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5676 			 i40e_stat_str(&pf->hw, ret),
5677 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5678 		goto exit;
5679 	}
5680 
5681 	/* No change detected in DCBX configs */
5682 	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5683 		    sizeof(tmp_dcbx_cfg))) {
5684 		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5685 		goto exit;
5686 	}
5687 
5688 	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5689 					       &hw->local_dcbx_config);
5690 
5691 	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5692 
5693 	if (!need_reconfig)
5694 		goto exit;
5695 
5696 	/* Enable DCB tagging only when more than one TC */
5697 	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5698 		pf->flags |= I40E_FLAG_DCB_ENABLED;
5699 	else
5700 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5701 
5702 	set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5703 	/* Reconfiguration needed quiesce all VSIs */
5704 	i40e_pf_quiesce_all_vsi(pf);
5705 
5706 	/* Changes in configuration update VEB/VSI */
5707 	i40e_dcb_reconfigure(pf);
5708 
5709 	ret = i40e_resume_port_tx(pf);
5710 
5711 	clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5712 	/* In case of error no point in resuming VSIs */
5713 	if (ret)
5714 		goto exit;
5715 
5716 	/* Wait for the PF's Tx queues to be disabled */
5717 	ret = i40e_pf_wait_txq_disabled(pf);
5718 	if (ret) {
5719 		/* Schedule PF reset to recover */
5720 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5721 		i40e_service_event_schedule(pf);
5722 	} else {
5723 		i40e_pf_unquiesce_all_vsi(pf);
5724 	}
5725 
5726 exit:
5727 	return ret;
5728 }
5729 #endif /* CONFIG_I40E_DCB */
5730 
5731 /**
5732  * i40e_do_reset_safe - Protected reset path for userland calls.
5733  * @pf: board private structure
5734  * @reset_flags: which reset is requested
5735  *
5736  **/
5737 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5738 {
5739 	rtnl_lock();
5740 	i40e_do_reset(pf, reset_flags);
5741 	rtnl_unlock();
5742 }
5743 
5744 /**
5745  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5746  * @pf: board private structure
5747  * @e: event info posted on ARQ
5748  *
5749  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5750  * and VF queues
5751  **/
5752 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5753 					   struct i40e_arq_event_info *e)
5754 {
5755 	struct i40e_aqc_lan_overflow *data =
5756 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5757 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
5758 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5759 	struct i40e_hw *hw = &pf->hw;
5760 	struct i40e_vf *vf;
5761 	u16 vf_id;
5762 
5763 	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5764 		queue, qtx_ctl);
5765 
5766 	/* Queue belongs to VF, find the VF and issue VF reset */
5767 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5768 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5769 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5770 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5771 		vf_id -= hw->func_caps.vf_base_id;
5772 		vf = &pf->vf[vf_id];
5773 		i40e_vc_notify_vf_reset(vf);
5774 		/* Allow VF to process pending reset notification */
5775 		msleep(20);
5776 		i40e_reset_vf(vf, false);
5777 	}
5778 }
5779 
5780 /**
5781  * i40e_service_event_complete - Finish up the service event
5782  * @pf: board private structure
5783  **/
5784 static void i40e_service_event_complete(struct i40e_pf *pf)
5785 {
5786 	WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5787 
5788 	/* flush memory to make sure state is correct before next watchog */
5789 	smp_mb__before_atomic();
5790 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5791 }
5792 
5793 /**
5794  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5795  * @pf: board private structure
5796  **/
5797 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5798 {
5799 	u32 val, fcnt_prog;
5800 
5801 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5802 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5803 	return fcnt_prog;
5804 }
5805 
5806 /**
5807  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5808  * @pf: board private structure
5809  **/
5810 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5811 {
5812 	u32 val, fcnt_prog;
5813 
5814 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5815 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5816 		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5817 		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5818 	return fcnt_prog;
5819 }
5820 
5821 /**
5822  * i40e_get_global_fd_count - Get total FD filters programmed on device
5823  * @pf: board private structure
5824  **/
5825 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5826 {
5827 	u32 val, fcnt_prog;
5828 
5829 	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5830 	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5831 		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5832 		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5833 	return fcnt_prog;
5834 }
5835 
5836 /**
5837  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5838  * @pf: board private structure
5839  **/
5840 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5841 {
5842 	struct i40e_fdir_filter *filter;
5843 	u32 fcnt_prog, fcnt_avail;
5844 	struct hlist_node *node;
5845 
5846 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5847 		return;
5848 
5849 	/* Check if, FD SB or ATR was auto disabled and if there is enough room
5850 	 * to re-enable
5851 	 */
5852 	fcnt_prog = i40e_get_global_fd_count(pf);
5853 	fcnt_avail = pf->fdir_pf_filter_count;
5854 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5855 	    (pf->fd_add_err == 0) ||
5856 	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5857 		if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5858 		    (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5859 			pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5860 			if (I40E_DEBUG_FD & pf->hw.debug_mask)
5861 				dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5862 		}
5863 	}
5864 	/* Wait for some more space to be available to turn on ATR */
5865 	if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5866 		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5867 		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5868 			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5869 			if (I40E_DEBUG_FD & pf->hw.debug_mask)
5870 				dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5871 		}
5872 	}
5873 
5874 	/* if hw had a problem adding a filter, delete it */
5875 	if (pf->fd_inv > 0) {
5876 		hlist_for_each_entry_safe(filter, node,
5877 					  &pf->fdir_filter_list, fdir_node) {
5878 			if (filter->fd_id == pf->fd_inv) {
5879 				hlist_del(&filter->fdir_node);
5880 				kfree(filter);
5881 				pf->fdir_pf_active_filters--;
5882 			}
5883 		}
5884 	}
5885 }
5886 
5887 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5888 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5889 /**
5890  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5891  * @pf: board private structure
5892  **/
5893 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5894 {
5895 	unsigned long min_flush_time;
5896 	int flush_wait_retry = 50;
5897 	bool disable_atr = false;
5898 	int fd_room;
5899 	int reg;
5900 
5901 	if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5902 		return;
5903 
5904 	if (!time_after(jiffies, pf->fd_flush_timestamp +
5905 				 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5906 		return;
5907 
5908 	/* If the flush is happening too quick and we have mostly SB rules we
5909 	 * should not re-enable ATR for some time.
5910 	 */
5911 	min_flush_time = pf->fd_flush_timestamp +
5912 			 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5913 	fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5914 
5915 	if (!(time_after(jiffies, min_flush_time)) &&
5916 	    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5917 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
5918 			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5919 		disable_atr = true;
5920 	}
5921 
5922 	pf->fd_flush_timestamp = jiffies;
5923 	pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5924 	/* flush all filters */
5925 	wr32(&pf->hw, I40E_PFQF_CTL_1,
5926 	     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5927 	i40e_flush(&pf->hw);
5928 	pf->fd_flush_cnt++;
5929 	pf->fd_add_err = 0;
5930 	do {
5931 		/* Check FD flush status every 5-6msec */
5932 		usleep_range(5000, 6000);
5933 		reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5934 		if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5935 			break;
5936 	} while (flush_wait_retry--);
5937 	if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5938 		dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5939 	} else {
5940 		/* replay sideband filters */
5941 		i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5942 		if (!disable_atr)
5943 			pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5944 		clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5945 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
5946 			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5947 	}
5948 
5949 }
5950 
5951 /**
5952  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5953  * @pf: board private structure
5954  **/
5955 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5956 {
5957 	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5958 }
5959 
5960 /* We can see up to 256 filter programming desc in transit if the filters are
5961  * being applied really fast; before we see the first
5962  * filter miss error on Rx queue 0. Accumulating enough error messages before
5963  * reacting will make sure we don't cause flush too often.
5964  */
5965 #define I40E_MAX_FD_PROGRAM_ERROR 256
5966 
5967 /**
5968  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
5969  * @pf: board private structure
5970  **/
5971 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
5972 {
5973 
5974 	/* if interface is down do nothing */
5975 	if (test_bit(__I40E_DOWN, &pf->state))
5976 		return;
5977 
5978 	if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5979 		return;
5980 
5981 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5982 		i40e_fdir_flush_and_replay(pf);
5983 
5984 	i40e_fdir_check_and_reenable(pf);
5985 
5986 }
5987 
5988 /**
5989  * i40e_vsi_link_event - notify VSI of a link event
5990  * @vsi: vsi to be notified
5991  * @link_up: link up or down
5992  **/
5993 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
5994 {
5995 	if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
5996 		return;
5997 
5998 	switch (vsi->type) {
5999 	case I40E_VSI_MAIN:
6000 #ifdef I40E_FCOE
6001 	case I40E_VSI_FCOE:
6002 #endif
6003 		if (!vsi->netdev || !vsi->netdev_registered)
6004 			break;
6005 
6006 		if (link_up) {
6007 			netif_carrier_on(vsi->netdev);
6008 			netif_tx_wake_all_queues(vsi->netdev);
6009 		} else {
6010 			netif_carrier_off(vsi->netdev);
6011 			netif_tx_stop_all_queues(vsi->netdev);
6012 		}
6013 		break;
6014 
6015 	case I40E_VSI_SRIOV:
6016 	case I40E_VSI_VMDQ2:
6017 	case I40E_VSI_CTRL:
6018 	case I40E_VSI_MIRROR:
6019 	default:
6020 		/* there is no notification for other VSIs */
6021 		break;
6022 	}
6023 }
6024 
6025 /**
6026  * i40e_veb_link_event - notify elements on the veb of a link event
6027  * @veb: veb to be notified
6028  * @link_up: link up or down
6029  **/
6030 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6031 {
6032 	struct i40e_pf *pf;
6033 	int i;
6034 
6035 	if (!veb || !veb->pf)
6036 		return;
6037 	pf = veb->pf;
6038 
6039 	/* depth first... */
6040 	for (i = 0; i < I40E_MAX_VEB; i++)
6041 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6042 			i40e_veb_link_event(pf->veb[i], link_up);
6043 
6044 	/* ... now the local VSIs */
6045 	for (i = 0; i < pf->num_alloc_vsi; i++)
6046 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6047 			i40e_vsi_link_event(pf->vsi[i], link_up);
6048 }
6049 
6050 /**
6051  * i40e_link_event - Update netif_carrier status
6052  * @pf: board private structure
6053  **/
6054 static void i40e_link_event(struct i40e_pf *pf)
6055 {
6056 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6057 	u8 new_link_speed, old_link_speed;
6058 	i40e_status status;
6059 	bool new_link, old_link;
6060 
6061 	/* save off old link status information */
6062 	pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6063 
6064 	/* set this to force the get_link_status call to refresh state */
6065 	pf->hw.phy.get_link_info = true;
6066 
6067 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6068 
6069 	status = i40e_get_link_status(&pf->hw, &new_link);
6070 	if (status) {
6071 		dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6072 			status);
6073 		return;
6074 	}
6075 
6076 	old_link_speed = pf->hw.phy.link_info_old.link_speed;
6077 	new_link_speed = pf->hw.phy.link_info.link_speed;
6078 
6079 	if (new_link == old_link &&
6080 	    new_link_speed == old_link_speed &&
6081 	    (test_bit(__I40E_DOWN, &vsi->state) ||
6082 	     new_link == netif_carrier_ok(vsi->netdev)))
6083 		return;
6084 
6085 	if (!test_bit(__I40E_DOWN, &vsi->state))
6086 		i40e_print_link_message(vsi, new_link);
6087 
6088 	/* Notify the base of the switch tree connected to
6089 	 * the link.  Floating VEBs are not notified.
6090 	 */
6091 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6092 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6093 	else
6094 		i40e_vsi_link_event(vsi, new_link);
6095 
6096 	if (pf->vf)
6097 		i40e_vc_notify_link_state(pf);
6098 
6099 	if (pf->flags & I40E_FLAG_PTP)
6100 		i40e_ptp_set_increment(pf);
6101 }
6102 
6103 /**
6104  * i40e_watchdog_subtask - periodic checks not using event driven response
6105  * @pf: board private structure
6106  **/
6107 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6108 {
6109 	int i;
6110 
6111 	/* if interface is down do nothing */
6112 	if (test_bit(__I40E_DOWN, &pf->state) ||
6113 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
6114 		return;
6115 
6116 	/* make sure we don't do these things too often */
6117 	if (time_before(jiffies, (pf->service_timer_previous +
6118 				  pf->service_timer_period)))
6119 		return;
6120 	pf->service_timer_previous = jiffies;
6121 
6122 	if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6123 		i40e_link_event(pf);
6124 
6125 	/* Update the stats for active netdevs so the network stack
6126 	 * can look at updated numbers whenever it cares to
6127 	 */
6128 	for (i = 0; i < pf->num_alloc_vsi; i++)
6129 		if (pf->vsi[i] && pf->vsi[i]->netdev)
6130 			i40e_update_stats(pf->vsi[i]);
6131 
6132 	if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6133 		/* Update the stats for the active switching components */
6134 		for (i = 0; i < I40E_MAX_VEB; i++)
6135 			if (pf->veb[i])
6136 				i40e_update_veb_stats(pf->veb[i]);
6137 	}
6138 
6139 	i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6140 }
6141 
6142 /**
6143  * i40e_reset_subtask - Set up for resetting the device and driver
6144  * @pf: board private structure
6145  **/
6146 static void i40e_reset_subtask(struct i40e_pf *pf)
6147 {
6148 	u32 reset_flags = 0;
6149 
6150 	rtnl_lock();
6151 	if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6152 		reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6153 		clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6154 	}
6155 	if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6156 		reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6157 		clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6158 	}
6159 	if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6160 		reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6161 		clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6162 	}
6163 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6164 		reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6165 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6166 	}
6167 	if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6168 		reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6169 		clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6170 	}
6171 
6172 	/* If there's a recovery already waiting, it takes
6173 	 * precedence before starting a new reset sequence.
6174 	 */
6175 	if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6176 		i40e_handle_reset_warning(pf);
6177 		goto unlock;
6178 	}
6179 
6180 	/* If we're already down or resetting, just bail */
6181 	if (reset_flags &&
6182 	    !test_bit(__I40E_DOWN, &pf->state) &&
6183 	    !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6184 		i40e_do_reset(pf, reset_flags);
6185 
6186 unlock:
6187 	rtnl_unlock();
6188 }
6189 
6190 /**
6191  * i40e_handle_link_event - Handle link event
6192  * @pf: board private structure
6193  * @e: event info posted on ARQ
6194  **/
6195 static void i40e_handle_link_event(struct i40e_pf *pf,
6196 				   struct i40e_arq_event_info *e)
6197 {
6198 	struct i40e_aqc_get_link_status *status =
6199 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6200 
6201 	/* Do a new status request to re-enable LSE reporting
6202 	 * and load new status information into the hw struct
6203 	 * This completely ignores any state information
6204 	 * in the ARQ event info, instead choosing to always
6205 	 * issue the AQ update link status command.
6206 	 */
6207 	i40e_link_event(pf);
6208 
6209 	/* check for unqualified module, if link is down */
6210 	if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6211 	    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6212 	    (!(status->link_info & I40E_AQ_LINK_UP)))
6213 		dev_err(&pf->pdev->dev,
6214 			"The driver failed to link because an unqualified module was detected.\n");
6215 }
6216 
6217 /**
6218  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6219  * @pf: board private structure
6220  **/
6221 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6222 {
6223 	struct i40e_arq_event_info event;
6224 	struct i40e_hw *hw = &pf->hw;
6225 	u16 pending, i = 0;
6226 	i40e_status ret;
6227 	u16 opcode;
6228 	u32 oldval;
6229 	u32 val;
6230 
6231 	/* Do not run clean AQ when PF reset fails */
6232 	if (test_bit(__I40E_RESET_FAILED, &pf->state))
6233 		return;
6234 
6235 	/* check for error indications */
6236 	val = rd32(&pf->hw, pf->hw.aq.arq.len);
6237 	oldval = val;
6238 	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6239 		if (hw->debug_mask & I40E_DEBUG_AQ)
6240 			dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6241 		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6242 	}
6243 	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6244 		if (hw->debug_mask & I40E_DEBUG_AQ)
6245 			dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6246 		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6247 	}
6248 	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6249 		if (hw->debug_mask & I40E_DEBUG_AQ)
6250 			dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6251 		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6252 	}
6253 	if (oldval != val)
6254 		wr32(&pf->hw, pf->hw.aq.arq.len, val);
6255 
6256 	val = rd32(&pf->hw, pf->hw.aq.asq.len);
6257 	oldval = val;
6258 	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6259 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6260 			dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6261 		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6262 	}
6263 	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6264 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6265 			dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6266 		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6267 	}
6268 	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6269 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6270 			dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6271 		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6272 	}
6273 	if (oldval != val)
6274 		wr32(&pf->hw, pf->hw.aq.asq.len, val);
6275 
6276 	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6277 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6278 	if (!event.msg_buf)
6279 		return;
6280 
6281 	do {
6282 		ret = i40e_clean_arq_element(hw, &event, &pending);
6283 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6284 			break;
6285 		else if (ret) {
6286 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6287 			break;
6288 		}
6289 
6290 		opcode = le16_to_cpu(event.desc.opcode);
6291 		switch (opcode) {
6292 
6293 		case i40e_aqc_opc_get_link_status:
6294 			i40e_handle_link_event(pf, &event);
6295 			break;
6296 		case i40e_aqc_opc_send_msg_to_pf:
6297 			ret = i40e_vc_process_vf_msg(pf,
6298 					le16_to_cpu(event.desc.retval),
6299 					le32_to_cpu(event.desc.cookie_high),
6300 					le32_to_cpu(event.desc.cookie_low),
6301 					event.msg_buf,
6302 					event.msg_len);
6303 			break;
6304 		case i40e_aqc_opc_lldp_update_mib:
6305 			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6306 #ifdef CONFIG_I40E_DCB
6307 			rtnl_lock();
6308 			ret = i40e_handle_lldp_event(pf, &event);
6309 			rtnl_unlock();
6310 #endif /* CONFIG_I40E_DCB */
6311 			break;
6312 		case i40e_aqc_opc_event_lan_overflow:
6313 			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6314 			i40e_handle_lan_overflow_event(pf, &event);
6315 			break;
6316 		case i40e_aqc_opc_send_msg_to_peer:
6317 			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6318 			break;
6319 		case i40e_aqc_opc_nvm_erase:
6320 		case i40e_aqc_opc_nvm_update:
6321 		case i40e_aqc_opc_oem_post_update:
6322 			i40e_debug(&pf->hw, I40E_DEBUG_NVM, "ARQ NVM operation completed\n");
6323 			break;
6324 		default:
6325 			dev_info(&pf->pdev->dev,
6326 				 "ARQ Error: Unknown event 0x%04x received\n",
6327 				 opcode);
6328 			break;
6329 		}
6330 	} while (pending && (i++ < pf->adminq_work_limit));
6331 
6332 	clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6333 	/* re-enable Admin queue interrupt cause */
6334 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
6335 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6336 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
6337 	i40e_flush(hw);
6338 
6339 	kfree(event.msg_buf);
6340 }
6341 
6342 /**
6343  * i40e_verify_eeprom - make sure eeprom is good to use
6344  * @pf: board private structure
6345  **/
6346 static void i40e_verify_eeprom(struct i40e_pf *pf)
6347 {
6348 	int err;
6349 
6350 	err = i40e_diag_eeprom_test(&pf->hw);
6351 	if (err) {
6352 		/* retry in case of garbage read */
6353 		err = i40e_diag_eeprom_test(&pf->hw);
6354 		if (err) {
6355 			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6356 				 err);
6357 			set_bit(__I40E_BAD_EEPROM, &pf->state);
6358 		}
6359 	}
6360 
6361 	if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6362 		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6363 		clear_bit(__I40E_BAD_EEPROM, &pf->state);
6364 	}
6365 }
6366 
6367 /**
6368  * i40e_enable_pf_switch_lb
6369  * @pf: pointer to the PF structure
6370  *
6371  * enable switch loop back or die - no point in a return value
6372  **/
6373 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6374 {
6375 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6376 	struct i40e_vsi_context ctxt;
6377 	int ret;
6378 
6379 	ctxt.seid = pf->main_vsi_seid;
6380 	ctxt.pf_num = pf->hw.pf_id;
6381 	ctxt.vf_num = 0;
6382 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6383 	if (ret) {
6384 		dev_info(&pf->pdev->dev,
6385 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6386 			 i40e_stat_str(&pf->hw, ret),
6387 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6388 		return;
6389 	}
6390 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6391 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6392 	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6393 
6394 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6395 	if (ret) {
6396 		dev_info(&pf->pdev->dev,
6397 			 "update vsi switch failed, err %s aq_err %s\n",
6398 			 i40e_stat_str(&pf->hw, ret),
6399 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6400 	}
6401 }
6402 
6403 /**
6404  * i40e_disable_pf_switch_lb
6405  * @pf: pointer to the PF structure
6406  *
6407  * disable switch loop back or die - no point in a return value
6408  **/
6409 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6410 {
6411 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6412 	struct i40e_vsi_context ctxt;
6413 	int ret;
6414 
6415 	ctxt.seid = pf->main_vsi_seid;
6416 	ctxt.pf_num = pf->hw.pf_id;
6417 	ctxt.vf_num = 0;
6418 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6419 	if (ret) {
6420 		dev_info(&pf->pdev->dev,
6421 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6422 			 i40e_stat_str(&pf->hw, ret),
6423 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6424 		return;
6425 	}
6426 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6427 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6428 	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6429 
6430 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6431 	if (ret) {
6432 		dev_info(&pf->pdev->dev,
6433 			 "update vsi switch failed, err %s aq_err %s\n",
6434 			 i40e_stat_str(&pf->hw, ret),
6435 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6436 	}
6437 }
6438 
6439 /**
6440  * i40e_config_bridge_mode - Configure the HW bridge mode
6441  * @veb: pointer to the bridge instance
6442  *
6443  * Configure the loop back mode for the LAN VSI that is downlink to the
6444  * specified HW bridge instance. It is expected this function is called
6445  * when a new HW bridge is instantiated.
6446  **/
6447 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6448 {
6449 	struct i40e_pf *pf = veb->pf;
6450 
6451 	if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6452 		dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6453 			 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6454 	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6455 		i40e_disable_pf_switch_lb(pf);
6456 	else
6457 		i40e_enable_pf_switch_lb(pf);
6458 }
6459 
6460 /**
6461  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6462  * @veb: pointer to the VEB instance
6463  *
6464  * This is a recursive function that first builds the attached VSIs then
6465  * recurses in to build the next layer of VEB.  We track the connections
6466  * through our own index numbers because the seid's from the HW could
6467  * change across the reset.
6468  **/
6469 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6470 {
6471 	struct i40e_vsi *ctl_vsi = NULL;
6472 	struct i40e_pf *pf = veb->pf;
6473 	int v, veb_idx;
6474 	int ret;
6475 
6476 	/* build VSI that owns this VEB, temporarily attached to base VEB */
6477 	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6478 		if (pf->vsi[v] &&
6479 		    pf->vsi[v]->veb_idx == veb->idx &&
6480 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6481 			ctl_vsi = pf->vsi[v];
6482 			break;
6483 		}
6484 	}
6485 	if (!ctl_vsi) {
6486 		dev_info(&pf->pdev->dev,
6487 			 "missing owner VSI for veb_idx %d\n", veb->idx);
6488 		ret = -ENOENT;
6489 		goto end_reconstitute;
6490 	}
6491 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
6492 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6493 	ret = i40e_add_vsi(ctl_vsi);
6494 	if (ret) {
6495 		dev_info(&pf->pdev->dev,
6496 			 "rebuild of veb_idx %d owner VSI failed: %d\n",
6497 			 veb->idx, ret);
6498 		goto end_reconstitute;
6499 	}
6500 	i40e_vsi_reset_stats(ctl_vsi);
6501 
6502 	/* create the VEB in the switch and move the VSI onto the VEB */
6503 	ret = i40e_add_veb(veb, ctl_vsi);
6504 	if (ret)
6505 		goto end_reconstitute;
6506 
6507 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6508 		veb->bridge_mode = BRIDGE_MODE_VEB;
6509 	else
6510 		veb->bridge_mode = BRIDGE_MODE_VEPA;
6511 	i40e_config_bridge_mode(veb);
6512 
6513 	/* create the remaining VSIs attached to this VEB */
6514 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6515 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6516 			continue;
6517 
6518 		if (pf->vsi[v]->veb_idx == veb->idx) {
6519 			struct i40e_vsi *vsi = pf->vsi[v];
6520 
6521 			vsi->uplink_seid = veb->seid;
6522 			ret = i40e_add_vsi(vsi);
6523 			if (ret) {
6524 				dev_info(&pf->pdev->dev,
6525 					 "rebuild of vsi_idx %d failed: %d\n",
6526 					 v, ret);
6527 				goto end_reconstitute;
6528 			}
6529 			i40e_vsi_reset_stats(vsi);
6530 		}
6531 	}
6532 
6533 	/* create any VEBs attached to this VEB - RECURSION */
6534 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6535 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6536 			pf->veb[veb_idx]->uplink_seid = veb->seid;
6537 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6538 			if (ret)
6539 				break;
6540 		}
6541 	}
6542 
6543 end_reconstitute:
6544 	return ret;
6545 }
6546 
6547 /**
6548  * i40e_get_capabilities - get info about the HW
6549  * @pf: the PF struct
6550  **/
6551 static int i40e_get_capabilities(struct i40e_pf *pf)
6552 {
6553 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6554 	u16 data_size;
6555 	int buf_len;
6556 	int err;
6557 
6558 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6559 	do {
6560 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
6561 		if (!cap_buf)
6562 			return -ENOMEM;
6563 
6564 		/* this loads the data into the hw struct for us */
6565 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6566 					    &data_size,
6567 					    i40e_aqc_opc_list_func_capabilities,
6568 					    NULL);
6569 		/* data loaded, buffer no longer needed */
6570 		kfree(cap_buf);
6571 
6572 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6573 			/* retry with a larger buffer */
6574 			buf_len = data_size;
6575 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6576 			dev_info(&pf->pdev->dev,
6577 				 "capability discovery failed, err %s aq_err %s\n",
6578 				 i40e_stat_str(&pf->hw, err),
6579 				 i40e_aq_str(&pf->hw,
6580 					     pf->hw.aq.asq_last_status));
6581 			return -ENODEV;
6582 		}
6583 	} while (err);
6584 
6585 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
6586 		dev_info(&pf->pdev->dev,
6587 			 "pf=%d, num_vfs=%d, msix_pf=%d, msix_vf=%d, fd_g=%d, fd_b=%d, pf_max_q=%d num_vsi=%d\n",
6588 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6589 			 pf->hw.func_caps.num_msix_vectors,
6590 			 pf->hw.func_caps.num_msix_vectors_vf,
6591 			 pf->hw.func_caps.fd_filters_guaranteed,
6592 			 pf->hw.func_caps.fd_filters_best_effort,
6593 			 pf->hw.func_caps.num_tx_qp,
6594 			 pf->hw.func_caps.num_vsis);
6595 
6596 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6597 		       + pf->hw.func_caps.num_vfs)
6598 	if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6599 		dev_info(&pf->pdev->dev,
6600 			 "got num_vsis %d, setting num_vsis to %d\n",
6601 			 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6602 		pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6603 	}
6604 
6605 	return 0;
6606 }
6607 
6608 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6609 
6610 /**
6611  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6612  * @pf: board private structure
6613  **/
6614 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6615 {
6616 	struct i40e_vsi *vsi;
6617 	int i;
6618 
6619 	/* quick workaround for an NVM issue that leaves a critical register
6620 	 * uninitialized
6621 	 */
6622 	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6623 		static const u32 hkey[] = {
6624 			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6625 			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6626 			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6627 			0x95b3a76d};
6628 
6629 		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6630 			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6631 	}
6632 
6633 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6634 		return;
6635 
6636 	/* find existing VSI and see if it needs configuring */
6637 	vsi = NULL;
6638 	for (i = 0; i < pf->num_alloc_vsi; i++) {
6639 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6640 			vsi = pf->vsi[i];
6641 			break;
6642 		}
6643 	}
6644 
6645 	/* create a new VSI if none exists */
6646 	if (!vsi) {
6647 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6648 				     pf->vsi[pf->lan_vsi]->seid, 0);
6649 		if (!vsi) {
6650 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6651 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6652 			return;
6653 		}
6654 	}
6655 
6656 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6657 }
6658 
6659 /**
6660  * i40e_fdir_teardown - release the Flow Director resources
6661  * @pf: board private structure
6662  **/
6663 static void i40e_fdir_teardown(struct i40e_pf *pf)
6664 {
6665 	int i;
6666 
6667 	i40e_fdir_filter_exit(pf);
6668 	for (i = 0; i < pf->num_alloc_vsi; i++) {
6669 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6670 			i40e_vsi_release(pf->vsi[i]);
6671 			break;
6672 		}
6673 	}
6674 }
6675 
6676 /**
6677  * i40e_prep_for_reset - prep for the core to reset
6678  * @pf: board private structure
6679  *
6680  * Close up the VFs and other things in prep for PF Reset.
6681   **/
6682 static void i40e_prep_for_reset(struct i40e_pf *pf)
6683 {
6684 	struct i40e_hw *hw = &pf->hw;
6685 	i40e_status ret = 0;
6686 	u32 v;
6687 
6688 	clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6689 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6690 		return;
6691 
6692 	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6693 
6694 	/* quiesce the VSIs and their queues that are not already DOWN */
6695 	i40e_pf_quiesce_all_vsi(pf);
6696 
6697 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6698 		if (pf->vsi[v])
6699 			pf->vsi[v]->seid = 0;
6700 	}
6701 
6702 	i40e_shutdown_adminq(&pf->hw);
6703 
6704 	/* call shutdown HMC */
6705 	if (hw->hmc.hmc_obj) {
6706 		ret = i40e_shutdown_lan_hmc(hw);
6707 		if (ret)
6708 			dev_warn(&pf->pdev->dev,
6709 				 "shutdown_lan_hmc failed: %d\n", ret);
6710 	}
6711 }
6712 
6713 /**
6714  * i40e_send_version - update firmware with driver version
6715  * @pf: PF struct
6716  */
6717 static void i40e_send_version(struct i40e_pf *pf)
6718 {
6719 	struct i40e_driver_version dv;
6720 
6721 	dv.major_version = DRV_VERSION_MAJOR;
6722 	dv.minor_version = DRV_VERSION_MINOR;
6723 	dv.build_version = DRV_VERSION_BUILD;
6724 	dv.subbuild_version = 0;
6725 	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6726 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6727 }
6728 
6729 /**
6730  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6731  * @pf: board private structure
6732  * @reinit: if the Main VSI needs to re-initialized.
6733  **/
6734 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6735 {
6736 	struct i40e_hw *hw = &pf->hw;
6737 	u8 set_fc_aq_fail = 0;
6738 	i40e_status ret;
6739 	u32 val;
6740 	u32 v;
6741 
6742 	/* Now we wait for GRST to settle out.
6743 	 * We don't have to delete the VEBs or VSIs from the hw switch
6744 	 * because the reset will make them disappear.
6745 	 */
6746 	ret = i40e_pf_reset(hw);
6747 	if (ret) {
6748 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6749 		set_bit(__I40E_RESET_FAILED, &pf->state);
6750 		goto clear_recovery;
6751 	}
6752 	pf->pfr_count++;
6753 
6754 	if (test_bit(__I40E_DOWN, &pf->state))
6755 		goto clear_recovery;
6756 	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6757 
6758 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6759 	ret = i40e_init_adminq(&pf->hw);
6760 	if (ret) {
6761 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6762 			 i40e_stat_str(&pf->hw, ret),
6763 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6764 		goto clear_recovery;
6765 	}
6766 
6767 	/* re-verify the eeprom if we just had an EMP reset */
6768 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6769 		i40e_verify_eeprom(pf);
6770 
6771 	i40e_clear_pxe_mode(hw);
6772 	ret = i40e_get_capabilities(pf);
6773 	if (ret)
6774 		goto end_core_reset;
6775 
6776 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6777 				hw->func_caps.num_rx_qp,
6778 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6779 	if (ret) {
6780 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6781 		goto end_core_reset;
6782 	}
6783 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6784 	if (ret) {
6785 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6786 		goto end_core_reset;
6787 	}
6788 
6789 #ifdef CONFIG_I40E_DCB
6790 	ret = i40e_init_pf_dcb(pf);
6791 	if (ret) {
6792 		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6793 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6794 		/* Continue without DCB enabled */
6795 	}
6796 #endif /* CONFIG_I40E_DCB */
6797 #ifdef I40E_FCOE
6798 	i40e_init_pf_fcoe(pf);
6799 
6800 #endif
6801 	/* do basic switch setup */
6802 	ret = i40e_setup_pf_switch(pf, reinit);
6803 	if (ret)
6804 		goto end_core_reset;
6805 
6806 	/* driver is only interested in link up/down and module qualification
6807 	 * reports from firmware
6808 	 */
6809 	ret = i40e_aq_set_phy_int_mask(&pf->hw,
6810 				       I40E_AQ_EVENT_LINK_UPDOWN |
6811 				       I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
6812 	if (ret)
6813 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6814 			 i40e_stat_str(&pf->hw, ret),
6815 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6816 
6817 	/* make sure our flow control settings are restored */
6818 	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6819 	if (ret)
6820 		dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6821 			i40e_stat_str(&pf->hw, ret),
6822 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6823 
6824 	/* Rebuild the VSIs and VEBs that existed before reset.
6825 	 * They are still in our local switch element arrays, so only
6826 	 * need to rebuild the switch model in the HW.
6827 	 *
6828 	 * If there were VEBs but the reconstitution failed, we'll try
6829 	 * try to recover minimal use by getting the basic PF VSI working.
6830 	 */
6831 	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6832 		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6833 		/* find the one VEB connected to the MAC, and find orphans */
6834 		for (v = 0; v < I40E_MAX_VEB; v++) {
6835 			if (!pf->veb[v])
6836 				continue;
6837 
6838 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6839 			    pf->veb[v]->uplink_seid == 0) {
6840 				ret = i40e_reconstitute_veb(pf->veb[v]);
6841 
6842 				if (!ret)
6843 					continue;
6844 
6845 				/* If Main VEB failed, we're in deep doodoo,
6846 				 * so give up rebuilding the switch and set up
6847 				 * for minimal rebuild of PF VSI.
6848 				 * If orphan failed, we'll report the error
6849 				 * but try to keep going.
6850 				 */
6851 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6852 					dev_info(&pf->pdev->dev,
6853 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6854 						 ret);
6855 					pf->vsi[pf->lan_vsi]->uplink_seid
6856 								= pf->mac_seid;
6857 					break;
6858 				} else if (pf->veb[v]->uplink_seid == 0) {
6859 					dev_info(&pf->pdev->dev,
6860 						 "rebuild of orphan VEB failed: %d\n",
6861 						 ret);
6862 				}
6863 			}
6864 		}
6865 	}
6866 
6867 	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6868 		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6869 		/* no VEB, so rebuild only the Main VSI */
6870 		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6871 		if (ret) {
6872 			dev_info(&pf->pdev->dev,
6873 				 "rebuild of Main VSI failed: %d\n", ret);
6874 			goto end_core_reset;
6875 		}
6876 	}
6877 
6878 	/* Reconfigure hardware for allowing smaller MSS in the case
6879 	 * of TSO, so that we avoid the MDD being fired and causing
6880 	 * a reset in the case of small MSS+TSO.
6881 	 */
6882 #define I40E_REG_MSS          0x000E64DC
6883 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6884 #define I40E_64BYTE_MSS       0x400000
6885 	val = rd32(hw, I40E_REG_MSS);
6886 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6887 		val &= ~I40E_REG_MSS_MIN_MASK;
6888 		val |= I40E_64BYTE_MSS;
6889 		wr32(hw, I40E_REG_MSS, val);
6890 	}
6891 
6892 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
6893 	    (pf->hw.aq.fw_maj_ver < 4)) {
6894 		msleep(75);
6895 		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6896 		if (ret)
6897 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6898 				 i40e_stat_str(&pf->hw, ret),
6899 				 i40e_aq_str(&pf->hw,
6900 					     pf->hw.aq.asq_last_status));
6901 	}
6902 	/* reinit the misc interrupt */
6903 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6904 		ret = i40e_setup_misc_vector(pf);
6905 
6906 	/* Add a filter to drop all Flow control frames from any VSI from being
6907 	 * transmitted. By doing so we stop a malicious VF from sending out
6908 	 * PAUSE or PFC frames and potentially controlling traffic for other
6909 	 * PF/VF VSIs.
6910 	 * The FW can still send Flow control frames if enabled.
6911 	 */
6912 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6913 						       pf->main_vsi_seid);
6914 
6915 	/* restart the VSIs that were rebuilt and running before the reset */
6916 	i40e_pf_unquiesce_all_vsi(pf);
6917 
6918 	if (pf->num_alloc_vfs) {
6919 		for (v = 0; v < pf->num_alloc_vfs; v++)
6920 			i40e_reset_vf(&pf->vf[v], true);
6921 	}
6922 
6923 	/* tell the firmware that we're starting */
6924 	i40e_send_version(pf);
6925 
6926 end_core_reset:
6927 	clear_bit(__I40E_RESET_FAILED, &pf->state);
6928 clear_recovery:
6929 	clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6930 }
6931 
6932 /**
6933  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6934  * @pf: board private structure
6935  *
6936  * Close up the VFs and other things in prep for a Core Reset,
6937  * then get ready to rebuild the world.
6938  **/
6939 static void i40e_handle_reset_warning(struct i40e_pf *pf)
6940 {
6941 	i40e_prep_for_reset(pf);
6942 	i40e_reset_and_rebuild(pf, false);
6943 }
6944 
6945 /**
6946  * i40e_handle_mdd_event
6947  * @pf: pointer to the PF structure
6948  *
6949  * Called from the MDD irq handler to identify possibly malicious vfs
6950  **/
6951 static void i40e_handle_mdd_event(struct i40e_pf *pf)
6952 {
6953 	struct i40e_hw *hw = &pf->hw;
6954 	bool mdd_detected = false;
6955 	bool pf_mdd_detected = false;
6956 	struct i40e_vf *vf;
6957 	u32 reg;
6958 	int i;
6959 
6960 	if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
6961 		return;
6962 
6963 	/* find what triggered the MDD event */
6964 	reg = rd32(hw, I40E_GL_MDET_TX);
6965 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
6966 		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
6967 				I40E_GL_MDET_TX_PF_NUM_SHIFT;
6968 		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
6969 				I40E_GL_MDET_TX_VF_NUM_SHIFT;
6970 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
6971 				I40E_GL_MDET_TX_EVENT_SHIFT;
6972 		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
6973 				I40E_GL_MDET_TX_QUEUE_SHIFT) -
6974 				pf->hw.func_caps.base_queue;
6975 		if (netif_msg_tx_err(pf))
6976 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
6977 				 event, queue, pf_num, vf_num);
6978 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
6979 		mdd_detected = true;
6980 	}
6981 	reg = rd32(hw, I40E_GL_MDET_RX);
6982 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
6983 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
6984 				I40E_GL_MDET_RX_FUNCTION_SHIFT;
6985 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
6986 				I40E_GL_MDET_RX_EVENT_SHIFT;
6987 		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
6988 				I40E_GL_MDET_RX_QUEUE_SHIFT) -
6989 				pf->hw.func_caps.base_queue;
6990 		if (netif_msg_rx_err(pf))
6991 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
6992 				 event, queue, func);
6993 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
6994 		mdd_detected = true;
6995 	}
6996 
6997 	if (mdd_detected) {
6998 		reg = rd32(hw, I40E_PF_MDET_TX);
6999 		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7000 			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7001 			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7002 			pf_mdd_detected = true;
7003 		}
7004 		reg = rd32(hw, I40E_PF_MDET_RX);
7005 		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7006 			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7007 			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7008 			pf_mdd_detected = true;
7009 		}
7010 		/* Queue belongs to the PF, initiate a reset */
7011 		if (pf_mdd_detected) {
7012 			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7013 			i40e_service_event_schedule(pf);
7014 		}
7015 	}
7016 
7017 	/* see if one of the VFs needs its hand slapped */
7018 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7019 		vf = &(pf->vf[i]);
7020 		reg = rd32(hw, I40E_VP_MDET_TX(i));
7021 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7022 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7023 			vf->num_mdd_events++;
7024 			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7025 				 i);
7026 		}
7027 
7028 		reg = rd32(hw, I40E_VP_MDET_RX(i));
7029 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7030 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7031 			vf->num_mdd_events++;
7032 			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7033 				 i);
7034 		}
7035 
7036 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7037 			dev_info(&pf->pdev->dev,
7038 				 "Too many MDD events on VF %d, disabled\n", i);
7039 			dev_info(&pf->pdev->dev,
7040 				 "Use PF Control I/F to re-enable the VF\n");
7041 			set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7042 		}
7043 	}
7044 
7045 	/* re-enable mdd interrupt cause */
7046 	clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7047 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7048 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7049 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7050 	i40e_flush(hw);
7051 }
7052 
7053 /**
7054  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7055  * @pf: board private structure
7056  **/
7057 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7058 {
7059 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
7060 	struct i40e_hw *hw = &pf->hw;
7061 	i40e_status ret;
7062 	__be16 port;
7063 	int i;
7064 
7065 	if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7066 		return;
7067 
7068 	pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7069 
7070 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7071 		if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7072 			pf->pending_udp_bitmap &= ~BIT_ULL(i);
7073 			port = pf->udp_ports[i].index;
7074 			if (port)
7075 				ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7076 						     pf->udp_ports[i].type,
7077 						     NULL, NULL);
7078 			else
7079 				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7080 
7081 			if (ret) {
7082 				dev_info(&pf->pdev->dev,
7083 					 "%s vxlan port %d, index %d failed, err %s aq_err %s\n",
7084 					 port ? "add" : "delete",
7085 					 ntohs(port), i,
7086 					 i40e_stat_str(&pf->hw, ret),
7087 					 i40e_aq_str(&pf->hw,
7088 						    pf->hw.aq.asq_last_status));
7089 				pf->udp_ports[i].index = 0;
7090 			}
7091 		}
7092 	}
7093 #endif
7094 }
7095 
7096 /**
7097  * i40e_service_task - Run the driver's async subtasks
7098  * @work: pointer to work_struct containing our data
7099  **/
7100 static void i40e_service_task(struct work_struct *work)
7101 {
7102 	struct i40e_pf *pf = container_of(work,
7103 					  struct i40e_pf,
7104 					  service_task);
7105 	unsigned long start_time = jiffies;
7106 
7107 	/* don't bother with service tasks if a reset is in progress */
7108 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7109 		i40e_service_event_complete(pf);
7110 		return;
7111 	}
7112 
7113 	i40e_detect_recover_hung(pf);
7114 	i40e_reset_subtask(pf);
7115 	i40e_handle_mdd_event(pf);
7116 	i40e_vc_process_vflr_event(pf);
7117 	i40e_watchdog_subtask(pf);
7118 	i40e_fdir_reinit_subtask(pf);
7119 	i40e_sync_filters_subtask(pf);
7120 	i40e_sync_udp_filters_subtask(pf);
7121 	i40e_clean_adminq_subtask(pf);
7122 
7123 	i40e_service_event_complete(pf);
7124 
7125 	/* If the tasks have taken longer than one timer cycle or there
7126 	 * is more work to be done, reschedule the service task now
7127 	 * rather than wait for the timer to tick again.
7128 	 */
7129 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7130 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)		 ||
7131 	    test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)		 ||
7132 	    test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7133 		i40e_service_event_schedule(pf);
7134 }
7135 
7136 /**
7137  * i40e_service_timer - timer callback
7138  * @data: pointer to PF struct
7139  **/
7140 static void i40e_service_timer(unsigned long data)
7141 {
7142 	struct i40e_pf *pf = (struct i40e_pf *)data;
7143 
7144 	mod_timer(&pf->service_timer,
7145 		  round_jiffies(jiffies + pf->service_timer_period));
7146 	i40e_service_event_schedule(pf);
7147 }
7148 
7149 /**
7150  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7151  * @vsi: the VSI being configured
7152  **/
7153 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7154 {
7155 	struct i40e_pf *pf = vsi->back;
7156 
7157 	switch (vsi->type) {
7158 	case I40E_VSI_MAIN:
7159 		vsi->alloc_queue_pairs = pf->num_lan_qps;
7160 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7161 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7162 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7163 			vsi->num_q_vectors = pf->num_lan_msix;
7164 		else
7165 			vsi->num_q_vectors = 1;
7166 
7167 		break;
7168 
7169 	case I40E_VSI_FDIR:
7170 		vsi->alloc_queue_pairs = 1;
7171 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7172 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7173 		vsi->num_q_vectors = 1;
7174 		break;
7175 
7176 	case I40E_VSI_VMDQ2:
7177 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7178 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7179 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7180 		vsi->num_q_vectors = pf->num_vmdq_msix;
7181 		break;
7182 
7183 	case I40E_VSI_SRIOV:
7184 		vsi->alloc_queue_pairs = pf->num_vf_qps;
7185 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7186 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7187 		break;
7188 
7189 #ifdef I40E_FCOE
7190 	case I40E_VSI_FCOE:
7191 		vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7192 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7193 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7194 		vsi->num_q_vectors = pf->num_fcoe_msix;
7195 		break;
7196 
7197 #endif /* I40E_FCOE */
7198 	default:
7199 		WARN_ON(1);
7200 		return -ENODATA;
7201 	}
7202 
7203 	return 0;
7204 }
7205 
7206 /**
7207  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7208  * @type: VSI pointer
7209  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7210  *
7211  * On error: returns error code (negative)
7212  * On success: returns 0
7213  **/
7214 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7215 {
7216 	int size;
7217 	int ret = 0;
7218 
7219 	/* allocate memory for both Tx and Rx ring pointers */
7220 	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7221 	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7222 	if (!vsi->tx_rings)
7223 		return -ENOMEM;
7224 	vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7225 
7226 	if (alloc_qvectors) {
7227 		/* allocate memory for q_vector pointers */
7228 		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7229 		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7230 		if (!vsi->q_vectors) {
7231 			ret = -ENOMEM;
7232 			goto err_vectors;
7233 		}
7234 	}
7235 	return ret;
7236 
7237 err_vectors:
7238 	kfree(vsi->tx_rings);
7239 	return ret;
7240 }
7241 
7242 /**
7243  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7244  * @pf: board private structure
7245  * @type: type of VSI
7246  *
7247  * On error: returns error code (negative)
7248  * On success: returns vsi index in PF (positive)
7249  **/
7250 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7251 {
7252 	int ret = -ENODEV;
7253 	struct i40e_vsi *vsi;
7254 	int vsi_idx;
7255 	int i;
7256 
7257 	/* Need to protect the allocation of the VSIs at the PF level */
7258 	mutex_lock(&pf->switch_mutex);
7259 
7260 	/* VSI list may be fragmented if VSI creation/destruction has
7261 	 * been happening.  We can afford to do a quick scan to look
7262 	 * for any free VSIs in the list.
7263 	 *
7264 	 * find next empty vsi slot, looping back around if necessary
7265 	 */
7266 	i = pf->next_vsi;
7267 	while (i < pf->num_alloc_vsi && pf->vsi[i])
7268 		i++;
7269 	if (i >= pf->num_alloc_vsi) {
7270 		i = 0;
7271 		while (i < pf->next_vsi && pf->vsi[i])
7272 			i++;
7273 	}
7274 
7275 	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7276 		vsi_idx = i;             /* Found one! */
7277 	} else {
7278 		ret = -ENODEV;
7279 		goto unlock_pf;  /* out of VSI slots! */
7280 	}
7281 	pf->next_vsi = ++i;
7282 
7283 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7284 	if (!vsi) {
7285 		ret = -ENOMEM;
7286 		goto unlock_pf;
7287 	}
7288 	vsi->type = type;
7289 	vsi->back = pf;
7290 	set_bit(__I40E_DOWN, &vsi->state);
7291 	vsi->flags = 0;
7292 	vsi->idx = vsi_idx;
7293 	vsi->rx_itr_setting = pf->rx_itr_default;
7294 	vsi->tx_itr_setting = pf->tx_itr_default;
7295 	vsi->int_rate_limit = 0;
7296 	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7297 				pf->rss_table_size : 64;
7298 	vsi->netdev_registered = false;
7299 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7300 	INIT_LIST_HEAD(&vsi->mac_filter_list);
7301 	vsi->irqs_ready = false;
7302 
7303 	ret = i40e_set_num_rings_in_vsi(vsi);
7304 	if (ret)
7305 		goto err_rings;
7306 
7307 	ret = i40e_vsi_alloc_arrays(vsi, true);
7308 	if (ret)
7309 		goto err_rings;
7310 
7311 	/* Setup default MSIX irq handler for VSI */
7312 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7313 
7314 	/* Initialize VSI lock */
7315 	spin_lock_init(&vsi->mac_filter_list_lock);
7316 	pf->vsi[vsi_idx] = vsi;
7317 	ret = vsi_idx;
7318 	goto unlock_pf;
7319 
7320 err_rings:
7321 	pf->next_vsi = i - 1;
7322 	kfree(vsi);
7323 unlock_pf:
7324 	mutex_unlock(&pf->switch_mutex);
7325 	return ret;
7326 }
7327 
7328 /**
7329  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7330  * @type: VSI pointer
7331  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7332  *
7333  * On error: returns error code (negative)
7334  * On success: returns 0
7335  **/
7336 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7337 {
7338 	/* free the ring and vector containers */
7339 	if (free_qvectors) {
7340 		kfree(vsi->q_vectors);
7341 		vsi->q_vectors = NULL;
7342 	}
7343 	kfree(vsi->tx_rings);
7344 	vsi->tx_rings = NULL;
7345 	vsi->rx_rings = NULL;
7346 }
7347 
7348 /**
7349  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7350  * and lookup table
7351  * @vsi: Pointer to VSI structure
7352  */
7353 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7354 {
7355 	if (!vsi)
7356 		return;
7357 
7358 	kfree(vsi->rss_hkey_user);
7359 	vsi->rss_hkey_user = NULL;
7360 
7361 	kfree(vsi->rss_lut_user);
7362 	vsi->rss_lut_user = NULL;
7363 }
7364 
7365 /**
7366  * i40e_vsi_clear - Deallocate the VSI provided
7367  * @vsi: the VSI being un-configured
7368  **/
7369 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7370 {
7371 	struct i40e_pf *pf;
7372 
7373 	if (!vsi)
7374 		return 0;
7375 
7376 	if (!vsi->back)
7377 		goto free_vsi;
7378 	pf = vsi->back;
7379 
7380 	mutex_lock(&pf->switch_mutex);
7381 	if (!pf->vsi[vsi->idx]) {
7382 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7383 			vsi->idx, vsi->idx, vsi, vsi->type);
7384 		goto unlock_vsi;
7385 	}
7386 
7387 	if (pf->vsi[vsi->idx] != vsi) {
7388 		dev_err(&pf->pdev->dev,
7389 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7390 			pf->vsi[vsi->idx]->idx,
7391 			pf->vsi[vsi->idx],
7392 			pf->vsi[vsi->idx]->type,
7393 			vsi->idx, vsi, vsi->type);
7394 		goto unlock_vsi;
7395 	}
7396 
7397 	/* updates the PF for this cleared vsi */
7398 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7399 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7400 
7401 	i40e_vsi_free_arrays(vsi, true);
7402 	i40e_clear_rss_config_user(vsi);
7403 
7404 	pf->vsi[vsi->idx] = NULL;
7405 	if (vsi->idx < pf->next_vsi)
7406 		pf->next_vsi = vsi->idx;
7407 
7408 unlock_vsi:
7409 	mutex_unlock(&pf->switch_mutex);
7410 free_vsi:
7411 	kfree(vsi);
7412 
7413 	return 0;
7414 }
7415 
7416 /**
7417  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7418  * @vsi: the VSI being cleaned
7419  **/
7420 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7421 {
7422 	int i;
7423 
7424 	if (vsi->tx_rings && vsi->tx_rings[0]) {
7425 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7426 			kfree_rcu(vsi->tx_rings[i], rcu);
7427 			vsi->tx_rings[i] = NULL;
7428 			vsi->rx_rings[i] = NULL;
7429 		}
7430 	}
7431 }
7432 
7433 /**
7434  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7435  * @vsi: the VSI being configured
7436  **/
7437 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7438 {
7439 	struct i40e_ring *tx_ring, *rx_ring;
7440 	struct i40e_pf *pf = vsi->back;
7441 	int i;
7442 
7443 	/* Set basic values in the rings to be used later during open() */
7444 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7445 		/* allocate space for both Tx and Rx in one shot */
7446 		tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7447 		if (!tx_ring)
7448 			goto err_out;
7449 
7450 		tx_ring->queue_index = i;
7451 		tx_ring->reg_idx = vsi->base_queue + i;
7452 		tx_ring->ring_active = false;
7453 		tx_ring->vsi = vsi;
7454 		tx_ring->netdev = vsi->netdev;
7455 		tx_ring->dev = &pf->pdev->dev;
7456 		tx_ring->count = vsi->num_desc;
7457 		tx_ring->size = 0;
7458 		tx_ring->dcb_tc = 0;
7459 		if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7460 			tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7461 		if (vsi->back->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE)
7462 			tx_ring->flags |= I40E_TXR_FLAGS_OUTER_UDP_CSUM;
7463 		vsi->tx_rings[i] = tx_ring;
7464 
7465 		rx_ring = &tx_ring[1];
7466 		rx_ring->queue_index = i;
7467 		rx_ring->reg_idx = vsi->base_queue + i;
7468 		rx_ring->ring_active = false;
7469 		rx_ring->vsi = vsi;
7470 		rx_ring->netdev = vsi->netdev;
7471 		rx_ring->dev = &pf->pdev->dev;
7472 		rx_ring->count = vsi->num_desc;
7473 		rx_ring->size = 0;
7474 		rx_ring->dcb_tc = 0;
7475 		if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
7476 			set_ring_16byte_desc_enabled(rx_ring);
7477 		else
7478 			clear_ring_16byte_desc_enabled(rx_ring);
7479 		vsi->rx_rings[i] = rx_ring;
7480 	}
7481 
7482 	return 0;
7483 
7484 err_out:
7485 	i40e_vsi_clear_rings(vsi);
7486 	return -ENOMEM;
7487 }
7488 
7489 /**
7490  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7491  * @pf: board private structure
7492  * @vectors: the number of MSI-X vectors to request
7493  *
7494  * Returns the number of vectors reserved, or error
7495  **/
7496 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7497 {
7498 	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7499 					I40E_MIN_MSIX, vectors);
7500 	if (vectors < 0) {
7501 		dev_info(&pf->pdev->dev,
7502 			 "MSI-X vector reservation failed: %d\n", vectors);
7503 		vectors = 0;
7504 	}
7505 
7506 	return vectors;
7507 }
7508 
7509 /**
7510  * i40e_init_msix - Setup the MSIX capability
7511  * @pf: board private structure
7512  *
7513  * Work with the OS to set up the MSIX vectors needed.
7514  *
7515  * Returns the number of vectors reserved or negative on failure
7516  **/
7517 static int i40e_init_msix(struct i40e_pf *pf)
7518 {
7519 	struct i40e_hw *hw = &pf->hw;
7520 	int vectors_left;
7521 	int v_budget, i;
7522 	int v_actual;
7523 
7524 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7525 		return -ENODEV;
7526 
7527 	/* The number of vectors we'll request will be comprised of:
7528 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
7529 	 *   - The number of LAN queue pairs
7530 	 *	- Queues being used for RSS.
7531 	 *		We don't need as many as max_rss_size vectors.
7532 	 *		use rss_size instead in the calculation since that
7533 	 *		is governed by number of cpus in the system.
7534 	 *	- assumes symmetric Tx/Rx pairing
7535 	 *   - The number of VMDq pairs
7536 #ifdef I40E_FCOE
7537 	 *   - The number of FCOE qps.
7538 #endif
7539 	 * Once we count this up, try the request.
7540 	 *
7541 	 * If we can't get what we want, we'll simplify to nearly nothing
7542 	 * and try again.  If that still fails, we punt.
7543 	 */
7544 	vectors_left = hw->func_caps.num_msix_vectors;
7545 	v_budget = 0;
7546 
7547 	/* reserve one vector for miscellaneous handler */
7548 	if (vectors_left) {
7549 		v_budget++;
7550 		vectors_left--;
7551 	}
7552 
7553 	/* reserve vectors for the main PF traffic queues */
7554 	pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7555 	vectors_left -= pf->num_lan_msix;
7556 	v_budget += pf->num_lan_msix;
7557 
7558 	/* reserve one vector for sideband flow director */
7559 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7560 		if (vectors_left) {
7561 			v_budget++;
7562 			vectors_left--;
7563 		} else {
7564 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7565 		}
7566 	}
7567 
7568 #ifdef I40E_FCOE
7569 	/* can we reserve enough for FCoE? */
7570 	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7571 		if (!vectors_left)
7572 			pf->num_fcoe_msix = 0;
7573 		else if (vectors_left >= pf->num_fcoe_qps)
7574 			pf->num_fcoe_msix = pf->num_fcoe_qps;
7575 		else
7576 			pf->num_fcoe_msix = 1;
7577 		v_budget += pf->num_fcoe_msix;
7578 		vectors_left -= pf->num_fcoe_msix;
7579 	}
7580 
7581 #endif
7582 	/* any vectors left over go for VMDq support */
7583 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7584 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7585 		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7586 
7587 		/* if we're short on vectors for what's desired, we limit
7588 		 * the queues per vmdq.  If this is still more than are
7589 		 * available, the user will need to change the number of
7590 		 * queues/vectors used by the PF later with the ethtool
7591 		 * channels command
7592 		 */
7593 		if (vmdq_vecs < vmdq_vecs_wanted)
7594 			pf->num_vmdq_qps = 1;
7595 		pf->num_vmdq_msix = pf->num_vmdq_qps;
7596 
7597 		v_budget += vmdq_vecs;
7598 		vectors_left -= vmdq_vecs;
7599 	}
7600 
7601 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7602 				   GFP_KERNEL);
7603 	if (!pf->msix_entries)
7604 		return -ENOMEM;
7605 
7606 	for (i = 0; i < v_budget; i++)
7607 		pf->msix_entries[i].entry = i;
7608 	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7609 
7610 	if (v_actual != v_budget) {
7611 		/* If we have limited resources, we will start with no vectors
7612 		 * for the special features and then allocate vectors to some
7613 		 * of these features based on the policy and at the end disable
7614 		 * the features that did not get any vectors.
7615 		 */
7616 #ifdef I40E_FCOE
7617 		pf->num_fcoe_qps = 0;
7618 		pf->num_fcoe_msix = 0;
7619 #endif
7620 		pf->num_vmdq_msix = 0;
7621 	}
7622 
7623 	if (v_actual < I40E_MIN_MSIX) {
7624 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7625 		kfree(pf->msix_entries);
7626 		pf->msix_entries = NULL;
7627 		return -ENODEV;
7628 
7629 	} else if (v_actual == I40E_MIN_MSIX) {
7630 		/* Adjust for minimal MSIX use */
7631 		pf->num_vmdq_vsis = 0;
7632 		pf->num_vmdq_qps = 0;
7633 		pf->num_lan_qps = 1;
7634 		pf->num_lan_msix = 1;
7635 
7636 	} else if (v_actual != v_budget) {
7637 		int vec;
7638 
7639 		/* reserve the misc vector */
7640 		vec = v_actual - 1;
7641 
7642 		/* Scale vector usage down */
7643 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7644 		pf->num_vmdq_vsis = 1;
7645 		pf->num_vmdq_qps = 1;
7646 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7647 
7648 		/* partition out the remaining vectors */
7649 		switch (vec) {
7650 		case 2:
7651 			pf->num_lan_msix = 1;
7652 			break;
7653 		case 3:
7654 #ifdef I40E_FCOE
7655 			/* give one vector to FCoE */
7656 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7657 				pf->num_lan_msix = 1;
7658 				pf->num_fcoe_msix = 1;
7659 			}
7660 #else
7661 			pf->num_lan_msix = 2;
7662 #endif
7663 			break;
7664 		default:
7665 #ifdef I40E_FCOE
7666 			/* give one vector to FCoE */
7667 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7668 				pf->num_fcoe_msix = 1;
7669 				vec--;
7670 			}
7671 #endif
7672 			/* give the rest to the PF */
7673 			pf->num_lan_msix = min_t(int, vec, pf->num_lan_qps);
7674 			break;
7675 		}
7676 	}
7677 
7678 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7679 	    (pf->num_vmdq_msix == 0)) {
7680 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7681 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7682 	}
7683 #ifdef I40E_FCOE
7684 
7685 	if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7686 		dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7687 		pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7688 	}
7689 #endif
7690 	return v_actual;
7691 }
7692 
7693 /**
7694  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7695  * @vsi: the VSI being configured
7696  * @v_idx: index of the vector in the vsi struct
7697  *
7698  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7699  **/
7700 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
7701 {
7702 	struct i40e_q_vector *q_vector;
7703 
7704 	/* allocate q_vector */
7705 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7706 	if (!q_vector)
7707 		return -ENOMEM;
7708 
7709 	q_vector->vsi = vsi;
7710 	q_vector->v_idx = v_idx;
7711 	cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
7712 	if (vsi->netdev)
7713 		netif_napi_add(vsi->netdev, &q_vector->napi,
7714 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
7715 
7716 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
7717 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
7718 
7719 	/* tie q_vector and vsi together */
7720 	vsi->q_vectors[v_idx] = q_vector;
7721 
7722 	return 0;
7723 }
7724 
7725 /**
7726  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7727  * @vsi: the VSI being configured
7728  *
7729  * We allocate one q_vector per queue interrupt.  If allocation fails we
7730  * return -ENOMEM.
7731  **/
7732 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7733 {
7734 	struct i40e_pf *pf = vsi->back;
7735 	int v_idx, num_q_vectors;
7736 	int err;
7737 
7738 	/* if not MSIX, give the one vector only to the LAN VSI */
7739 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7740 		num_q_vectors = vsi->num_q_vectors;
7741 	else if (vsi == pf->vsi[pf->lan_vsi])
7742 		num_q_vectors = 1;
7743 	else
7744 		return -EINVAL;
7745 
7746 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7747 		err = i40e_vsi_alloc_q_vector(vsi, v_idx);
7748 		if (err)
7749 			goto err_out;
7750 	}
7751 
7752 	return 0;
7753 
7754 err_out:
7755 	while (v_idx--)
7756 		i40e_free_q_vector(vsi, v_idx);
7757 
7758 	return err;
7759 }
7760 
7761 /**
7762  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7763  * @pf: board private structure to initialize
7764  **/
7765 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7766 {
7767 	int vectors = 0;
7768 	ssize_t size;
7769 
7770 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7771 		vectors = i40e_init_msix(pf);
7772 		if (vectors < 0) {
7773 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
7774 #ifdef I40E_FCOE
7775 				       I40E_FLAG_FCOE_ENABLED	|
7776 #endif
7777 				       I40E_FLAG_RSS_ENABLED	|
7778 				       I40E_FLAG_DCB_CAPABLE	|
7779 				       I40E_FLAG_SRIOV_ENABLED	|
7780 				       I40E_FLAG_FD_SB_ENABLED	|
7781 				       I40E_FLAG_FD_ATR_ENABLED	|
7782 				       I40E_FLAG_VMDQ_ENABLED);
7783 
7784 			/* rework the queue expectations without MSIX */
7785 			i40e_determine_queue_usage(pf);
7786 		}
7787 	}
7788 
7789 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7790 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7791 		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7792 		vectors = pci_enable_msi(pf->pdev);
7793 		if (vectors < 0) {
7794 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7795 				 vectors);
7796 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7797 		}
7798 		vectors = 1;  /* one MSI or Legacy vector */
7799 	}
7800 
7801 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7802 		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7803 
7804 	/* set up vector assignment tracking */
7805 	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7806 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
7807 	if (!pf->irq_pile) {
7808 		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7809 		return -ENOMEM;
7810 	}
7811 	pf->irq_pile->num_entries = vectors;
7812 	pf->irq_pile->search_hint = 0;
7813 
7814 	/* track first vector for misc interrupts, ignore return */
7815 	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7816 
7817 	return 0;
7818 }
7819 
7820 /**
7821  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7822  * @pf: board private structure
7823  *
7824  * This sets up the handler for MSIX 0, which is used to manage the
7825  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7826  * when in MSI or Legacy interrupt mode.
7827  **/
7828 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7829 {
7830 	struct i40e_hw *hw = &pf->hw;
7831 	int err = 0;
7832 
7833 	/* Only request the irq if this is the first time through, and
7834 	 * not when we're rebuilding after a Reset
7835 	 */
7836 	if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7837 		err = request_irq(pf->msix_entries[0].vector,
7838 				  i40e_intr, 0, pf->int_name, pf);
7839 		if (err) {
7840 			dev_info(&pf->pdev->dev,
7841 				 "request_irq for %s failed: %d\n",
7842 				 pf->int_name, err);
7843 			return -EFAULT;
7844 		}
7845 	}
7846 
7847 	i40e_enable_misc_int_causes(pf);
7848 
7849 	/* associate no queues to the misc vector */
7850 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7851 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7852 
7853 	i40e_flush(hw);
7854 
7855 	i40e_irq_dynamic_enable_icr0(pf);
7856 
7857 	return err;
7858 }
7859 
7860 /**
7861  * i40e_config_rss_aq - Prepare for RSS using AQ commands
7862  * @vsi: vsi structure
7863  * @seed: RSS hash seed
7864  **/
7865 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7866 			      u8 *lut, u16 lut_size)
7867 {
7868 	struct i40e_aqc_get_set_rss_key_data rss_key;
7869 	struct i40e_pf *pf = vsi->back;
7870 	struct i40e_hw *hw = &pf->hw;
7871 	bool pf_lut = false;
7872 	u8 *rss_lut;
7873 	int ret, i;
7874 
7875 	memset(&rss_key, 0, sizeof(rss_key));
7876 	memcpy(&rss_key, seed, sizeof(rss_key));
7877 
7878 	rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
7879 	if (!rss_lut)
7880 		return -ENOMEM;
7881 
7882 	/* Populate the LUT with max no. of queues in round robin fashion */
7883 	for (i = 0; i < vsi->rss_table_size; i++)
7884 		rss_lut[i] = i % vsi->rss_size;
7885 
7886 	ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
7887 	if (ret) {
7888 		dev_info(&pf->pdev->dev,
7889 			 "Cannot set RSS key, err %s aq_err %s\n",
7890 			 i40e_stat_str(&pf->hw, ret),
7891 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7892 		goto config_rss_aq_out;
7893 	}
7894 
7895 	if (vsi->type == I40E_VSI_MAIN)
7896 		pf_lut = true;
7897 
7898 	ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
7899 				  vsi->rss_table_size);
7900 	if (ret)
7901 		dev_info(&pf->pdev->dev,
7902 			 "Cannot set RSS lut, err %s aq_err %s\n",
7903 			 i40e_stat_str(&pf->hw, ret),
7904 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7905 
7906 config_rss_aq_out:
7907 	kfree(rss_lut);
7908 	return ret;
7909 }
7910 
7911 /**
7912  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
7913  * @vsi: VSI structure
7914  **/
7915 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
7916 {
7917 	u8 seed[I40E_HKEY_ARRAY_SIZE];
7918 	struct i40e_pf *pf = vsi->back;
7919 	u8 *lut;
7920 	int ret;
7921 
7922 	if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
7923 		return 0;
7924 
7925 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
7926 	if (!lut)
7927 		return -ENOMEM;
7928 
7929 	i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
7930 	netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
7931 	vsi->rss_size = min_t(int, pf->alloc_rss_size, vsi->num_queue_pairs);
7932 	ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
7933 	kfree(lut);
7934 
7935 	return ret;
7936 }
7937 
7938 /**
7939  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
7940  * @vsi: Pointer to vsi structure
7941  * @seed: RSS hash seed
7942  * @lut: Lookup table
7943  * @lut_size: Lookup table size
7944  *
7945  * Returns 0 on success, negative on failure
7946  **/
7947 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
7948 			       const u8 *lut, u16 lut_size)
7949 {
7950 	struct i40e_pf *pf = vsi->back;
7951 	struct i40e_hw *hw = &pf->hw;
7952 	u8 i;
7953 
7954 	/* Fill out hash function seed */
7955 	if (seed) {
7956 		u32 *seed_dw = (u32 *)seed;
7957 
7958 		for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
7959 			wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
7960 	}
7961 
7962 	if (lut) {
7963 		u32 *lut_dw = (u32 *)lut;
7964 
7965 		if (lut_size != I40E_HLUT_ARRAY_SIZE)
7966 			return -EINVAL;
7967 
7968 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
7969 			wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
7970 	}
7971 	i40e_flush(hw);
7972 
7973 	return 0;
7974 }
7975 
7976 /**
7977  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
7978  * @vsi: Pointer to VSI structure
7979  * @seed: Buffer to store the keys
7980  * @lut: Buffer to store the lookup table entries
7981  * @lut_size: Size of buffer to store the lookup table entries
7982  *
7983  * Returns 0 on success, negative on failure
7984  */
7985 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
7986 			    u8 *lut, u16 lut_size)
7987 {
7988 	struct i40e_pf *pf = vsi->back;
7989 	struct i40e_hw *hw = &pf->hw;
7990 	u16 i;
7991 
7992 	if (seed) {
7993 		u32 *seed_dw = (u32 *)seed;
7994 
7995 		for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
7996 			seed_dw[i] = rd32(hw, I40E_PFQF_HKEY(i));
7997 	}
7998 	if (lut) {
7999 		u32 *lut_dw = (u32 *)lut;
8000 
8001 		if (lut_size != I40E_HLUT_ARRAY_SIZE)
8002 			return -EINVAL;
8003 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8004 			lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8005 	}
8006 
8007 	return 0;
8008 }
8009 
8010 /**
8011  * i40e_config_rss - Configure RSS keys and lut
8012  * @vsi: Pointer to VSI structure
8013  * @seed: RSS hash seed
8014  * @lut: Lookup table
8015  * @lut_size: Lookup table size
8016  *
8017  * Returns 0 on success, negative on failure
8018  */
8019 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8020 {
8021 	struct i40e_pf *pf = vsi->back;
8022 
8023 	if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8024 		return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8025 	else
8026 		return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8027 }
8028 
8029 /**
8030  * i40e_get_rss - Get RSS keys and lut
8031  * @vsi: Pointer to VSI structure
8032  * @seed: Buffer to store the keys
8033  * @lut: Buffer to store the lookup table entries
8034  * lut_size: Size of buffer to store the lookup table entries
8035  *
8036  * Returns 0 on success, negative on failure
8037  */
8038 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8039 {
8040 	return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8041 }
8042 
8043 /**
8044  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8045  * @pf: Pointer to board private structure
8046  * @lut: Lookup table
8047  * @rss_table_size: Lookup table size
8048  * @rss_size: Range of queue number for hashing
8049  */
8050 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8051 			      u16 rss_table_size, u16 rss_size)
8052 {
8053 	u16 i;
8054 
8055 	for (i = 0; i < rss_table_size; i++)
8056 		lut[i] = i % rss_size;
8057 }
8058 
8059 /**
8060  * i40e_pf_config_rss - Prepare for RSS if used
8061  * @pf: board private structure
8062  **/
8063 static int i40e_pf_config_rss(struct i40e_pf *pf)
8064 {
8065 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8066 	u8 seed[I40E_HKEY_ARRAY_SIZE];
8067 	u8 *lut;
8068 	struct i40e_hw *hw = &pf->hw;
8069 	u32 reg_val;
8070 	u64 hena;
8071 	int ret;
8072 
8073 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8074 	hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
8075 		((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
8076 	hena |= i40e_pf_get_default_rss_hena(pf);
8077 
8078 	wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
8079 	wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8080 
8081 	/* Determine the RSS table size based on the hardware capabilities */
8082 	reg_val = rd32(hw, I40E_PFQF_CTL_0);
8083 	reg_val = (pf->rss_table_size == 512) ?
8084 			(reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8085 			(reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8086 	wr32(hw, I40E_PFQF_CTL_0, reg_val);
8087 
8088 	/* Determine the RSS size of the VSI */
8089 	if (!vsi->rss_size)
8090 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
8091 				      vsi->num_queue_pairs);
8092 
8093 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8094 	if (!lut)
8095 		return -ENOMEM;
8096 
8097 	/* Use user configured lut if there is one, otherwise use default */
8098 	if (vsi->rss_lut_user)
8099 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8100 	else
8101 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8102 
8103 	/* Use user configured hash key if there is one, otherwise
8104 	 * use default.
8105 	 */
8106 	if (vsi->rss_hkey_user)
8107 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8108 	else
8109 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8110 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8111 	kfree(lut);
8112 
8113 	return ret;
8114 }
8115 
8116 /**
8117  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8118  * @pf: board private structure
8119  * @queue_count: the requested queue count for rss.
8120  *
8121  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8122  * count which may be different from the requested queue count.
8123  **/
8124 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8125 {
8126 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8127 	int new_rss_size;
8128 
8129 	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8130 		return 0;
8131 
8132 	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8133 
8134 	if (queue_count != vsi->num_queue_pairs) {
8135 		vsi->req_queue_pairs = queue_count;
8136 		i40e_prep_for_reset(pf);
8137 
8138 		pf->alloc_rss_size = new_rss_size;
8139 
8140 		i40e_reset_and_rebuild(pf, true);
8141 
8142 		/* Discard the user configured hash keys and lut, if less
8143 		 * queues are enabled.
8144 		 */
8145 		if (queue_count < vsi->rss_size) {
8146 			i40e_clear_rss_config_user(vsi);
8147 			dev_dbg(&pf->pdev->dev,
8148 				"discard user configured hash keys and lut\n");
8149 		}
8150 
8151 		/* Reset vsi->rss_size, as number of enabled queues changed */
8152 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
8153 				      vsi->num_queue_pairs);
8154 
8155 		i40e_pf_config_rss(pf);
8156 	}
8157 	dev_info(&pf->pdev->dev, "RSS count/HW max RSS count:  %d/%d\n",
8158 		 pf->alloc_rss_size, pf->rss_size_max);
8159 	return pf->alloc_rss_size;
8160 }
8161 
8162 /**
8163  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8164  * @pf: board private structure
8165  **/
8166 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8167 {
8168 	i40e_status status;
8169 	bool min_valid, max_valid;
8170 	u32 max_bw, min_bw;
8171 
8172 	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8173 					   &min_valid, &max_valid);
8174 
8175 	if (!status) {
8176 		if (min_valid)
8177 			pf->npar_min_bw = min_bw;
8178 		if (max_valid)
8179 			pf->npar_max_bw = max_bw;
8180 	}
8181 
8182 	return status;
8183 }
8184 
8185 /**
8186  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8187  * @pf: board private structure
8188  **/
8189 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8190 {
8191 	struct i40e_aqc_configure_partition_bw_data bw_data;
8192 	i40e_status status;
8193 
8194 	/* Set the valid bit for this PF */
8195 	bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8196 	bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8197 	bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8198 
8199 	/* Set the new bandwidths */
8200 	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8201 
8202 	return status;
8203 }
8204 
8205 /**
8206  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8207  * @pf: board private structure
8208  **/
8209 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8210 {
8211 	/* Commit temporary BW setting to permanent NVM image */
8212 	enum i40e_admin_queue_err last_aq_status;
8213 	i40e_status ret;
8214 	u16 nvm_word;
8215 
8216 	if (pf->hw.partition_id != 1) {
8217 		dev_info(&pf->pdev->dev,
8218 			 "Commit BW only works on partition 1! This is partition %d",
8219 			 pf->hw.partition_id);
8220 		ret = I40E_NOT_SUPPORTED;
8221 		goto bw_commit_out;
8222 	}
8223 
8224 	/* Acquire NVM for read access */
8225 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8226 	last_aq_status = pf->hw.aq.asq_last_status;
8227 	if (ret) {
8228 		dev_info(&pf->pdev->dev,
8229 			 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8230 			 i40e_stat_str(&pf->hw, ret),
8231 			 i40e_aq_str(&pf->hw, last_aq_status));
8232 		goto bw_commit_out;
8233 	}
8234 
8235 	/* Read word 0x10 of NVM - SW compatibility word 1 */
8236 	ret = i40e_aq_read_nvm(&pf->hw,
8237 			       I40E_SR_NVM_CONTROL_WORD,
8238 			       0x10, sizeof(nvm_word), &nvm_word,
8239 			       false, NULL);
8240 	/* Save off last admin queue command status before releasing
8241 	 * the NVM
8242 	 */
8243 	last_aq_status = pf->hw.aq.asq_last_status;
8244 	i40e_release_nvm(&pf->hw);
8245 	if (ret) {
8246 		dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8247 			 i40e_stat_str(&pf->hw, ret),
8248 			 i40e_aq_str(&pf->hw, last_aq_status));
8249 		goto bw_commit_out;
8250 	}
8251 
8252 	/* Wait a bit for NVM release to complete */
8253 	msleep(50);
8254 
8255 	/* Acquire NVM for write access */
8256 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8257 	last_aq_status = pf->hw.aq.asq_last_status;
8258 	if (ret) {
8259 		dev_info(&pf->pdev->dev,
8260 			 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8261 			 i40e_stat_str(&pf->hw, ret),
8262 			 i40e_aq_str(&pf->hw, last_aq_status));
8263 		goto bw_commit_out;
8264 	}
8265 	/* Write it back out unchanged to initiate update NVM,
8266 	 * which will force a write of the shadow (alt) RAM to
8267 	 * the NVM - thus storing the bandwidth values permanently.
8268 	 */
8269 	ret = i40e_aq_update_nvm(&pf->hw,
8270 				 I40E_SR_NVM_CONTROL_WORD,
8271 				 0x10, sizeof(nvm_word),
8272 				 &nvm_word, true, NULL);
8273 	/* Save off last admin queue command status before releasing
8274 	 * the NVM
8275 	 */
8276 	last_aq_status = pf->hw.aq.asq_last_status;
8277 	i40e_release_nvm(&pf->hw);
8278 	if (ret)
8279 		dev_info(&pf->pdev->dev,
8280 			 "BW settings NOT SAVED, err %s aq_err %s\n",
8281 			 i40e_stat_str(&pf->hw, ret),
8282 			 i40e_aq_str(&pf->hw, last_aq_status));
8283 bw_commit_out:
8284 
8285 	return ret;
8286 }
8287 
8288 /**
8289  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8290  * @pf: board private structure to initialize
8291  *
8292  * i40e_sw_init initializes the Adapter private data structure.
8293  * Fields are initialized based on PCI device information and
8294  * OS network device settings (MTU size).
8295  **/
8296 static int i40e_sw_init(struct i40e_pf *pf)
8297 {
8298 	int err = 0;
8299 	int size;
8300 
8301 	pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8302 				(NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8303 	pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
8304 	if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8305 		if (I40E_DEBUG_USER & debug)
8306 			pf->hw.debug_mask = debug;
8307 		pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8308 						I40E_DEFAULT_MSG_ENABLE);
8309 	}
8310 
8311 	/* Set default capability flags */
8312 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8313 		    I40E_FLAG_MSI_ENABLED     |
8314 		    I40E_FLAG_LINK_POLLING_ENABLED |
8315 		    I40E_FLAG_MSIX_ENABLED;
8316 
8317 	if (iommu_present(&pci_bus_type))
8318 		pf->flags |= I40E_FLAG_RX_PS_ENABLED;
8319 	else
8320 		pf->flags |= I40E_FLAG_RX_1BUF_ENABLED;
8321 
8322 	/* Set default ITR */
8323 	pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8324 	pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8325 
8326 	/* Depending on PF configurations, it is possible that the RSS
8327 	 * maximum might end up larger than the available queues
8328 	 */
8329 	pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8330 	pf->alloc_rss_size = 1;
8331 	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8332 	pf->rss_size_max = min_t(int, pf->rss_size_max,
8333 				 pf->hw.func_caps.num_tx_qp);
8334 	if (pf->hw.func_caps.rss) {
8335 		pf->flags |= I40E_FLAG_RSS_ENABLED;
8336 		pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8337 					   num_online_cpus());
8338 	}
8339 
8340 	/* MFP mode enabled */
8341 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8342 		pf->flags |= I40E_FLAG_MFP_ENABLED;
8343 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8344 		if (i40e_get_npar_bw_setting(pf))
8345 			dev_warn(&pf->pdev->dev,
8346 				 "Could not get NPAR bw settings\n");
8347 		else
8348 			dev_info(&pf->pdev->dev,
8349 				 "Min BW = %8.8x, Max BW = %8.8x\n",
8350 				 pf->npar_min_bw, pf->npar_max_bw);
8351 	}
8352 
8353 	/* FW/NVM is not yet fixed in this regard */
8354 	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8355 	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8356 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8357 		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8358 		if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8359 		    pf->hw.num_partitions > 1)
8360 			dev_info(&pf->pdev->dev,
8361 				 "Flow Director Sideband mode Disabled in MFP mode\n");
8362 		else
8363 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8364 		pf->fdir_pf_filter_count =
8365 				 pf->hw.func_caps.fd_filters_guaranteed;
8366 		pf->hw.fdir_shared_filter_count =
8367 				 pf->hw.func_caps.fd_filters_best_effort;
8368 	}
8369 
8370 	if (pf->hw.func_caps.vmdq) {
8371 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8372 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8373 		pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8374 	}
8375 
8376 #ifdef I40E_FCOE
8377 	i40e_init_pf_fcoe(pf);
8378 
8379 #endif /* I40E_FCOE */
8380 #ifdef CONFIG_PCI_IOV
8381 	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8382 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8383 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8384 		pf->num_req_vfs = min_t(int,
8385 					pf->hw.func_caps.num_vfs,
8386 					I40E_MAX_VF_COUNT);
8387 	}
8388 #endif /* CONFIG_PCI_IOV */
8389 	if (pf->hw.mac.type == I40E_MAC_X722) {
8390 		pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8391 			     I40E_FLAG_128_QP_RSS_CAPABLE |
8392 			     I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8393 			     I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8394 			     I40E_FLAG_WB_ON_ITR_CAPABLE |
8395 			     I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8396 			     I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8397 	}
8398 	pf->eeprom_version = 0xDEAD;
8399 	pf->lan_veb = I40E_NO_VEB;
8400 	pf->lan_vsi = I40E_NO_VSI;
8401 
8402 	/* By default FW has this off for performance reasons */
8403 	pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8404 
8405 	/* set up queue assignment tracking */
8406 	size = sizeof(struct i40e_lump_tracking)
8407 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8408 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
8409 	if (!pf->qp_pile) {
8410 		err = -ENOMEM;
8411 		goto sw_init_done;
8412 	}
8413 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8414 	pf->qp_pile->search_hint = 0;
8415 
8416 	pf->tx_timeout_recovery_level = 1;
8417 
8418 	mutex_init(&pf->switch_mutex);
8419 
8420 	/* If NPAR is enabled nudge the Tx scheduler */
8421 	if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8422 		i40e_set_npar_bw_setting(pf);
8423 
8424 sw_init_done:
8425 	return err;
8426 }
8427 
8428 /**
8429  * i40e_set_ntuple - set the ntuple feature flag and take action
8430  * @pf: board private structure to initialize
8431  * @features: the feature set that the stack is suggesting
8432  *
8433  * returns a bool to indicate if reset needs to happen
8434  **/
8435 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8436 {
8437 	bool need_reset = false;
8438 
8439 	/* Check if Flow Director n-tuple support was enabled or disabled.  If
8440 	 * the state changed, we need to reset.
8441 	 */
8442 	if (features & NETIF_F_NTUPLE) {
8443 		/* Enable filters and mark for reset */
8444 		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8445 			need_reset = true;
8446 		pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8447 	} else {
8448 		/* turn off filters, mark for reset and clear SW filter list */
8449 		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8450 			need_reset = true;
8451 			i40e_fdir_filter_exit(pf);
8452 		}
8453 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8454 		pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8455 		/* reset fd counters */
8456 		pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8457 		pf->fdir_pf_active_filters = 0;
8458 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8459 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
8460 			dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8461 		/* if ATR was auto disabled it can be re-enabled. */
8462 		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8463 		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8464 			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8465 	}
8466 	return need_reset;
8467 }
8468 
8469 /**
8470  * i40e_set_features - set the netdev feature flags
8471  * @netdev: ptr to the netdev being adjusted
8472  * @features: the feature set that the stack is suggesting
8473  **/
8474 static int i40e_set_features(struct net_device *netdev,
8475 			     netdev_features_t features)
8476 {
8477 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8478 	struct i40e_vsi *vsi = np->vsi;
8479 	struct i40e_pf *pf = vsi->back;
8480 	bool need_reset;
8481 
8482 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
8483 		i40e_vlan_stripping_enable(vsi);
8484 	else
8485 		i40e_vlan_stripping_disable(vsi);
8486 
8487 	need_reset = i40e_set_ntuple(pf, features);
8488 
8489 	if (need_reset)
8490 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8491 
8492 	return 0;
8493 }
8494 
8495 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
8496 /**
8497  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8498  * @pf: board private structure
8499  * @port: The UDP port to look up
8500  *
8501  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8502  **/
8503 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8504 {
8505 	u8 i;
8506 
8507 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8508 		if (pf->udp_ports[i].index == port)
8509 			return i;
8510 	}
8511 
8512 	return i;
8513 }
8514 
8515 #endif
8516 
8517 #if IS_ENABLED(CONFIG_VXLAN)
8518 /**
8519  * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
8520  * @netdev: This physical port's netdev
8521  * @sa_family: Socket Family that VXLAN is notifying us about
8522  * @port: New UDP port number that VXLAN started listening to
8523  **/
8524 static void i40e_add_vxlan_port(struct net_device *netdev,
8525 				sa_family_t sa_family, __be16 port)
8526 {
8527 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8528 	struct i40e_vsi *vsi = np->vsi;
8529 	struct i40e_pf *pf = vsi->back;
8530 	u8 next_idx;
8531 	u8 idx;
8532 
8533 	if (sa_family == AF_INET6)
8534 		return;
8535 
8536 	idx = i40e_get_udp_port_idx(pf, port);
8537 
8538 	/* Check if port already exists */
8539 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8540 		netdev_info(netdev, "vxlan port %d already offloaded\n",
8541 			    ntohs(port));
8542 		return;
8543 	}
8544 
8545 	/* Now check if there is space to add the new port */
8546 	next_idx = i40e_get_udp_port_idx(pf, 0);
8547 
8548 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8549 		netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
8550 			    ntohs(port));
8551 		return;
8552 	}
8553 
8554 	/* New port: add it and mark its index in the bitmap */
8555 	pf->udp_ports[next_idx].index = port;
8556 	pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8557 	pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8558 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8559 }
8560 
8561 /**
8562  * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
8563  * @netdev: This physical port's netdev
8564  * @sa_family: Socket Family that VXLAN is notifying us about
8565  * @port: UDP port number that VXLAN stopped listening to
8566  **/
8567 static void i40e_del_vxlan_port(struct net_device *netdev,
8568 				sa_family_t sa_family, __be16 port)
8569 {
8570 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8571 	struct i40e_vsi *vsi = np->vsi;
8572 	struct i40e_pf *pf = vsi->back;
8573 	u8 idx;
8574 
8575 	if (sa_family == AF_INET6)
8576 		return;
8577 
8578 	idx = i40e_get_udp_port_idx(pf, port);
8579 
8580 	/* Check if port already exists */
8581 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8582 		/* if port exists, set it to 0 (mark for deletion)
8583 		 * and make it pending
8584 		 */
8585 		pf->udp_ports[idx].index = 0;
8586 		pf->pending_udp_bitmap |= BIT_ULL(idx);
8587 		pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8588 	} else {
8589 		netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
8590 			    ntohs(port));
8591 	}
8592 }
8593 #endif
8594 
8595 #if IS_ENABLED(CONFIG_GENEVE)
8596 /**
8597  * i40e_add_geneve_port - Get notifications about GENEVE ports that come up
8598  * @netdev: This physical port's netdev
8599  * @sa_family: Socket Family that GENEVE is notifying us about
8600  * @port: New UDP port number that GENEVE started listening to
8601  **/
8602 static void i40e_add_geneve_port(struct net_device *netdev,
8603 				 sa_family_t sa_family, __be16 port)
8604 {
8605 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8606 	struct i40e_vsi *vsi = np->vsi;
8607 	struct i40e_pf *pf = vsi->back;
8608 	u8 next_idx;
8609 	u8 idx;
8610 
8611 	if (sa_family == AF_INET6)
8612 		return;
8613 
8614 	idx = i40e_get_udp_port_idx(pf, port);
8615 
8616 	/* Check if port already exists */
8617 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8618 		netdev_info(netdev, "udp port %d already offloaded\n",
8619 			    ntohs(port));
8620 		return;
8621 	}
8622 
8623 	/* Now check if there is space to add the new port */
8624 	next_idx = i40e_get_udp_port_idx(pf, 0);
8625 
8626 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8627 		netdev_info(netdev, "maximum number of UDP ports reached, not adding port %d\n",
8628 			    ntohs(port));
8629 		return;
8630 	}
8631 
8632 	/* New port: add it and mark its index in the bitmap */
8633 	pf->udp_ports[next_idx].index = port;
8634 	pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8635 	pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8636 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8637 
8638 	dev_info(&pf->pdev->dev, "adding geneve port %d\n", ntohs(port));
8639 }
8640 
8641 /**
8642  * i40e_del_geneve_port - Get notifications about GENEVE ports that go away
8643  * @netdev: This physical port's netdev
8644  * @sa_family: Socket Family that GENEVE is notifying us about
8645  * @port: UDP port number that GENEVE stopped listening to
8646  **/
8647 static void i40e_del_geneve_port(struct net_device *netdev,
8648 				 sa_family_t sa_family, __be16 port)
8649 {
8650 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8651 	struct i40e_vsi *vsi = np->vsi;
8652 	struct i40e_pf *pf = vsi->back;
8653 	u8 idx;
8654 
8655 	if (sa_family == AF_INET6)
8656 		return;
8657 
8658 	idx = i40e_get_udp_port_idx(pf, port);
8659 
8660 	/* Check if port already exists */
8661 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8662 		/* if port exists, set it to 0 (mark for deletion)
8663 		 * and make it pending
8664 		 */
8665 		pf->udp_ports[idx].index = 0;
8666 		pf->pending_udp_bitmap |= BIT_ULL(idx);
8667 		pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8668 
8669 		dev_info(&pf->pdev->dev, "deleting geneve port %d\n",
8670 			 ntohs(port));
8671 	} else {
8672 		netdev_warn(netdev, "geneve port %d was not found, not deleting\n",
8673 			    ntohs(port));
8674 	}
8675 }
8676 #endif
8677 
8678 static int i40e_get_phys_port_id(struct net_device *netdev,
8679 				 struct netdev_phys_item_id *ppid)
8680 {
8681 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8682 	struct i40e_pf *pf = np->vsi->back;
8683 	struct i40e_hw *hw = &pf->hw;
8684 
8685 	if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8686 		return -EOPNOTSUPP;
8687 
8688 	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8689 	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8690 
8691 	return 0;
8692 }
8693 
8694 /**
8695  * i40e_ndo_fdb_add - add an entry to the hardware database
8696  * @ndm: the input from the stack
8697  * @tb: pointer to array of nladdr (unused)
8698  * @dev: the net device pointer
8699  * @addr: the MAC address entry being added
8700  * @flags: instructions from stack about fdb operation
8701  */
8702 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8703 			    struct net_device *dev,
8704 			    const unsigned char *addr, u16 vid,
8705 			    u16 flags)
8706 {
8707 	struct i40e_netdev_priv *np = netdev_priv(dev);
8708 	struct i40e_pf *pf = np->vsi->back;
8709 	int err = 0;
8710 
8711 	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8712 		return -EOPNOTSUPP;
8713 
8714 	if (vid) {
8715 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8716 		return -EINVAL;
8717 	}
8718 
8719 	/* Hardware does not support aging addresses so if a
8720 	 * ndm_state is given only allow permanent addresses
8721 	 */
8722 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8723 		netdev_info(dev, "FDB only supports static addresses\n");
8724 		return -EINVAL;
8725 	}
8726 
8727 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8728 		err = dev_uc_add_excl(dev, addr);
8729 	else if (is_multicast_ether_addr(addr))
8730 		err = dev_mc_add_excl(dev, addr);
8731 	else
8732 		err = -EINVAL;
8733 
8734 	/* Only return duplicate errors if NLM_F_EXCL is set */
8735 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
8736 		err = 0;
8737 
8738 	return err;
8739 }
8740 
8741 /**
8742  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8743  * @dev: the netdev being configured
8744  * @nlh: RTNL message
8745  *
8746  * Inserts a new hardware bridge if not already created and
8747  * enables the bridging mode requested (VEB or VEPA). If the
8748  * hardware bridge has already been inserted and the request
8749  * is to change the mode then that requires a PF reset to
8750  * allow rebuild of the components with required hardware
8751  * bridge mode enabled.
8752  **/
8753 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8754 				   struct nlmsghdr *nlh,
8755 				   u16 flags)
8756 {
8757 	struct i40e_netdev_priv *np = netdev_priv(dev);
8758 	struct i40e_vsi *vsi = np->vsi;
8759 	struct i40e_pf *pf = vsi->back;
8760 	struct i40e_veb *veb = NULL;
8761 	struct nlattr *attr, *br_spec;
8762 	int i, rem;
8763 
8764 	/* Only for PF VSI for now */
8765 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8766 		return -EOPNOTSUPP;
8767 
8768 	/* Find the HW bridge for PF VSI */
8769 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8770 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8771 			veb = pf->veb[i];
8772 	}
8773 
8774 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8775 
8776 	nla_for_each_nested(attr, br_spec, rem) {
8777 		__u16 mode;
8778 
8779 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8780 			continue;
8781 
8782 		mode = nla_get_u16(attr);
8783 		if ((mode != BRIDGE_MODE_VEPA) &&
8784 		    (mode != BRIDGE_MODE_VEB))
8785 			return -EINVAL;
8786 
8787 		/* Insert a new HW bridge */
8788 		if (!veb) {
8789 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8790 					     vsi->tc_config.enabled_tc);
8791 			if (veb) {
8792 				veb->bridge_mode = mode;
8793 				i40e_config_bridge_mode(veb);
8794 			} else {
8795 				/* No Bridge HW offload available */
8796 				return -ENOENT;
8797 			}
8798 			break;
8799 		} else if (mode != veb->bridge_mode) {
8800 			/* Existing HW bridge but different mode needs reset */
8801 			veb->bridge_mode = mode;
8802 			/* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8803 			if (mode == BRIDGE_MODE_VEB)
8804 				pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8805 			else
8806 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8807 			i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8808 			break;
8809 		}
8810 	}
8811 
8812 	return 0;
8813 }
8814 
8815 /**
8816  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8817  * @skb: skb buff
8818  * @pid: process id
8819  * @seq: RTNL message seq #
8820  * @dev: the netdev being configured
8821  * @filter_mask: unused
8822  * @nlflags: netlink flags passed in
8823  *
8824  * Return the mode in which the hardware bridge is operating in
8825  * i.e VEB or VEPA.
8826  **/
8827 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8828 				   struct net_device *dev,
8829 				   u32 __always_unused filter_mask,
8830 				   int nlflags)
8831 {
8832 	struct i40e_netdev_priv *np = netdev_priv(dev);
8833 	struct i40e_vsi *vsi = np->vsi;
8834 	struct i40e_pf *pf = vsi->back;
8835 	struct i40e_veb *veb = NULL;
8836 	int i;
8837 
8838 	/* Only for PF VSI for now */
8839 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8840 		return -EOPNOTSUPP;
8841 
8842 	/* Find the HW bridge for the PF VSI */
8843 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8844 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8845 			veb = pf->veb[i];
8846 	}
8847 
8848 	if (!veb)
8849 		return 0;
8850 
8851 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
8852 				       nlflags, 0, 0, filter_mask, NULL);
8853 }
8854 
8855 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
8856  * inner mac plus all inner ethertypes.
8857  */
8858 #define I40E_MAX_TUNNEL_HDR_LEN 128
8859 /**
8860  * i40e_features_check - Validate encapsulated packet conforms to limits
8861  * @skb: skb buff
8862  * @dev: This physical port's netdev
8863  * @features: Offload features that the stack believes apply
8864  **/
8865 static netdev_features_t i40e_features_check(struct sk_buff *skb,
8866 					     struct net_device *dev,
8867 					     netdev_features_t features)
8868 {
8869 	if (skb->encapsulation &&
8870 	    ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
8871 	     I40E_MAX_TUNNEL_HDR_LEN))
8872 		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
8873 
8874 	return features;
8875 }
8876 
8877 static const struct net_device_ops i40e_netdev_ops = {
8878 	.ndo_open		= i40e_open,
8879 	.ndo_stop		= i40e_close,
8880 	.ndo_start_xmit		= i40e_lan_xmit_frame,
8881 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
8882 	.ndo_set_rx_mode	= i40e_set_rx_mode,
8883 	.ndo_validate_addr	= eth_validate_addr,
8884 	.ndo_set_mac_address	= i40e_set_mac,
8885 	.ndo_change_mtu		= i40e_change_mtu,
8886 	.ndo_do_ioctl		= i40e_ioctl,
8887 	.ndo_tx_timeout		= i40e_tx_timeout,
8888 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
8889 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
8890 #ifdef CONFIG_NET_POLL_CONTROLLER
8891 	.ndo_poll_controller	= i40e_netpoll,
8892 #endif
8893 	.ndo_setup_tc		= i40e_setup_tc,
8894 #ifdef I40E_FCOE
8895 	.ndo_fcoe_enable	= i40e_fcoe_enable,
8896 	.ndo_fcoe_disable	= i40e_fcoe_disable,
8897 #endif
8898 	.ndo_set_features	= i40e_set_features,
8899 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
8900 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
8901 	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
8902 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
8903 	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
8904 	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
8905 #if IS_ENABLED(CONFIG_VXLAN)
8906 	.ndo_add_vxlan_port	= i40e_add_vxlan_port,
8907 	.ndo_del_vxlan_port	= i40e_del_vxlan_port,
8908 #endif
8909 #if IS_ENABLED(CONFIG_GENEVE)
8910 	.ndo_add_geneve_port	= i40e_add_geneve_port,
8911 	.ndo_del_geneve_port	= i40e_del_geneve_port,
8912 #endif
8913 	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
8914 	.ndo_fdb_add		= i40e_ndo_fdb_add,
8915 	.ndo_features_check	= i40e_features_check,
8916 	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
8917 	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
8918 };
8919 
8920 /**
8921  * i40e_config_netdev - Setup the netdev flags
8922  * @vsi: the VSI being configured
8923  *
8924  * Returns 0 on success, negative value on failure
8925  **/
8926 static int i40e_config_netdev(struct i40e_vsi *vsi)
8927 {
8928 	u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
8929 	struct i40e_pf *pf = vsi->back;
8930 	struct i40e_hw *hw = &pf->hw;
8931 	struct i40e_netdev_priv *np;
8932 	struct net_device *netdev;
8933 	u8 mac_addr[ETH_ALEN];
8934 	int etherdev_size;
8935 
8936 	etherdev_size = sizeof(struct i40e_netdev_priv);
8937 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
8938 	if (!netdev)
8939 		return -ENOMEM;
8940 
8941 	vsi->netdev = netdev;
8942 	np = netdev_priv(netdev);
8943 	np->vsi = vsi;
8944 
8945 	netdev->hw_enc_features |= NETIF_F_IP_CSUM	 |
8946 				  NETIF_F_RXCSUM	 |
8947 				  NETIF_F_GSO_UDP_TUNNEL |
8948 				  NETIF_F_GSO_GRE	 |
8949 				  NETIF_F_TSO;
8950 
8951 	netdev->features = NETIF_F_SG		       |
8952 			   NETIF_F_IP_CSUM	       |
8953 			   NETIF_F_SCTP_CRC	       |
8954 			   NETIF_F_HIGHDMA	       |
8955 			   NETIF_F_GSO_UDP_TUNNEL      |
8956 			   NETIF_F_GSO_GRE	       |
8957 			   NETIF_F_HW_VLAN_CTAG_TX     |
8958 			   NETIF_F_HW_VLAN_CTAG_RX     |
8959 			   NETIF_F_HW_VLAN_CTAG_FILTER |
8960 			   NETIF_F_IPV6_CSUM	       |
8961 			   NETIF_F_TSO		       |
8962 			   NETIF_F_TSO_ECN	       |
8963 			   NETIF_F_TSO6		       |
8964 			   NETIF_F_RXCSUM	       |
8965 			   NETIF_F_RXHASH	       |
8966 			   0;
8967 
8968 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
8969 		netdev->features |= NETIF_F_NTUPLE;
8970 
8971 	/* copy netdev features into list of user selectable features */
8972 	netdev->hw_features |= netdev->features;
8973 
8974 	if (vsi->type == I40E_VSI_MAIN) {
8975 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
8976 		ether_addr_copy(mac_addr, hw->mac.perm_addr);
8977 		/* The following steps are necessary to prevent reception
8978 		 * of tagged packets - some older NVM configurations load a
8979 		 * default a MAC-VLAN filter that accepts any tagged packet
8980 		 * which must be replaced by a normal filter.
8981 		 */
8982 		if (!i40e_rm_default_mac_filter(vsi, mac_addr)) {
8983 			spin_lock_bh(&vsi->mac_filter_list_lock);
8984 			i40e_add_filter(vsi, mac_addr,
8985 					I40E_VLAN_ANY, false, true);
8986 			spin_unlock_bh(&vsi->mac_filter_list_lock);
8987 		}
8988 	} else {
8989 		/* relate the VSI_VMDQ name to the VSI_MAIN name */
8990 		snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
8991 			 pf->vsi[pf->lan_vsi]->netdev->name);
8992 		random_ether_addr(mac_addr);
8993 
8994 		spin_lock_bh(&vsi->mac_filter_list_lock);
8995 		i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
8996 		spin_unlock_bh(&vsi->mac_filter_list_lock);
8997 	}
8998 
8999 	spin_lock_bh(&vsi->mac_filter_list_lock);
9000 	i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
9001 	spin_unlock_bh(&vsi->mac_filter_list_lock);
9002 
9003 	ether_addr_copy(netdev->dev_addr, mac_addr);
9004 	ether_addr_copy(netdev->perm_addr, mac_addr);
9005 	/* vlan gets same features (except vlan offload)
9006 	 * after any tweaks for specific VSI types
9007 	 */
9008 	netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
9009 						     NETIF_F_HW_VLAN_CTAG_RX |
9010 						   NETIF_F_HW_VLAN_CTAG_FILTER);
9011 	netdev->priv_flags |= IFF_UNICAST_FLT;
9012 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9013 	/* Setup netdev TC information */
9014 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9015 
9016 	netdev->netdev_ops = &i40e_netdev_ops;
9017 	netdev->watchdog_timeo = 5 * HZ;
9018 	i40e_set_ethtool_ops(netdev);
9019 #ifdef I40E_FCOE
9020 	i40e_fcoe_config_netdev(netdev, vsi);
9021 #endif
9022 
9023 	return 0;
9024 }
9025 
9026 /**
9027  * i40e_vsi_delete - Delete a VSI from the switch
9028  * @vsi: the VSI being removed
9029  *
9030  * Returns 0 on success, negative value on failure
9031  **/
9032 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9033 {
9034 	/* remove default VSI is not allowed */
9035 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9036 		return;
9037 
9038 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9039 }
9040 
9041 /**
9042  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9043  * @vsi: the VSI being queried
9044  *
9045  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9046  **/
9047 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9048 {
9049 	struct i40e_veb *veb;
9050 	struct i40e_pf *pf = vsi->back;
9051 
9052 	/* Uplink is not a bridge so default to VEB */
9053 	if (vsi->veb_idx == I40E_NO_VEB)
9054 		return 1;
9055 
9056 	veb = pf->veb[vsi->veb_idx];
9057 	if (!veb) {
9058 		dev_info(&pf->pdev->dev,
9059 			 "There is no veb associated with the bridge\n");
9060 		return -ENOENT;
9061 	}
9062 
9063 	/* Uplink is a bridge in VEPA mode */
9064 	if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9065 		return 0;
9066 	} else {
9067 		/* Uplink is a bridge in VEB mode */
9068 		return 1;
9069 	}
9070 
9071 	/* VEPA is now default bridge, so return 0 */
9072 	return 0;
9073 }
9074 
9075 /**
9076  * i40e_add_vsi - Add a VSI to the switch
9077  * @vsi: the VSI being configured
9078  *
9079  * This initializes a VSI context depending on the VSI type to be added and
9080  * passes it down to the add_vsi aq command.
9081  **/
9082 static int i40e_add_vsi(struct i40e_vsi *vsi)
9083 {
9084 	int ret = -ENODEV;
9085 	u8 laa_macaddr[ETH_ALEN];
9086 	bool found_laa_mac_filter = false;
9087 	struct i40e_pf *pf = vsi->back;
9088 	struct i40e_hw *hw = &pf->hw;
9089 	struct i40e_vsi_context ctxt;
9090 	struct i40e_mac_filter *f, *ftmp;
9091 
9092 	u8 enabled_tc = 0x1; /* TC0 enabled */
9093 	int f_count = 0;
9094 
9095 	memset(&ctxt, 0, sizeof(ctxt));
9096 	switch (vsi->type) {
9097 	case I40E_VSI_MAIN:
9098 		/* The PF's main VSI is already setup as part of the
9099 		 * device initialization, so we'll not bother with
9100 		 * the add_vsi call, but we will retrieve the current
9101 		 * VSI context.
9102 		 */
9103 		ctxt.seid = pf->main_vsi_seid;
9104 		ctxt.pf_num = pf->hw.pf_id;
9105 		ctxt.vf_num = 0;
9106 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9107 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9108 		if (ret) {
9109 			dev_info(&pf->pdev->dev,
9110 				 "couldn't get PF vsi config, err %s aq_err %s\n",
9111 				 i40e_stat_str(&pf->hw, ret),
9112 				 i40e_aq_str(&pf->hw,
9113 					     pf->hw.aq.asq_last_status));
9114 			return -ENOENT;
9115 		}
9116 		vsi->info = ctxt.info;
9117 		vsi->info.valid_sections = 0;
9118 
9119 		vsi->seid = ctxt.seid;
9120 		vsi->id = ctxt.vsi_number;
9121 
9122 		enabled_tc = i40e_pf_get_tc_map(pf);
9123 
9124 		/* MFP mode setup queue map and update VSI */
9125 		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9126 		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9127 			memset(&ctxt, 0, sizeof(ctxt));
9128 			ctxt.seid = pf->main_vsi_seid;
9129 			ctxt.pf_num = pf->hw.pf_id;
9130 			ctxt.vf_num = 0;
9131 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9132 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9133 			if (ret) {
9134 				dev_info(&pf->pdev->dev,
9135 					 "update vsi failed, err %s aq_err %s\n",
9136 					 i40e_stat_str(&pf->hw, ret),
9137 					 i40e_aq_str(&pf->hw,
9138 						    pf->hw.aq.asq_last_status));
9139 				ret = -ENOENT;
9140 				goto err;
9141 			}
9142 			/* update the local VSI info queue map */
9143 			i40e_vsi_update_queue_map(vsi, &ctxt);
9144 			vsi->info.valid_sections = 0;
9145 		} else {
9146 			/* Default/Main VSI is only enabled for TC0
9147 			 * reconfigure it to enable all TCs that are
9148 			 * available on the port in SFP mode.
9149 			 * For MFP case the iSCSI PF would use this
9150 			 * flow to enable LAN+iSCSI TC.
9151 			 */
9152 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
9153 			if (ret) {
9154 				dev_info(&pf->pdev->dev,
9155 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9156 					 enabled_tc,
9157 					 i40e_stat_str(&pf->hw, ret),
9158 					 i40e_aq_str(&pf->hw,
9159 						    pf->hw.aq.asq_last_status));
9160 				ret = -ENOENT;
9161 			}
9162 		}
9163 		break;
9164 
9165 	case I40E_VSI_FDIR:
9166 		ctxt.pf_num = hw->pf_id;
9167 		ctxt.vf_num = 0;
9168 		ctxt.uplink_seid = vsi->uplink_seid;
9169 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9170 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9171 		if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9172 		    (i40e_is_vsi_uplink_mode_veb(vsi))) {
9173 			ctxt.info.valid_sections |=
9174 			     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9175 			ctxt.info.switch_id =
9176 			   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9177 		}
9178 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9179 		break;
9180 
9181 	case I40E_VSI_VMDQ2:
9182 		ctxt.pf_num = hw->pf_id;
9183 		ctxt.vf_num = 0;
9184 		ctxt.uplink_seid = vsi->uplink_seid;
9185 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9186 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9187 
9188 		/* This VSI is connected to VEB so the switch_id
9189 		 * should be set to zero by default.
9190 		 */
9191 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9192 			ctxt.info.valid_sections |=
9193 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9194 			ctxt.info.switch_id =
9195 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9196 		}
9197 
9198 		/* Setup the VSI tx/rx queue map for TC0 only for now */
9199 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9200 		break;
9201 
9202 	case I40E_VSI_SRIOV:
9203 		ctxt.pf_num = hw->pf_id;
9204 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9205 		ctxt.uplink_seid = vsi->uplink_seid;
9206 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9207 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9208 
9209 		/* This VSI is connected to VEB so the switch_id
9210 		 * should be set to zero by default.
9211 		 */
9212 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9213 			ctxt.info.valid_sections |=
9214 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9215 			ctxt.info.switch_id =
9216 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9217 		}
9218 
9219 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9220 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9221 		if (pf->vf[vsi->vf_id].spoofchk) {
9222 			ctxt.info.valid_sections |=
9223 				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9224 			ctxt.info.sec_flags |=
9225 				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9226 				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9227 		}
9228 		/* Setup the VSI tx/rx queue map for TC0 only for now */
9229 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9230 		break;
9231 
9232 #ifdef I40E_FCOE
9233 	case I40E_VSI_FCOE:
9234 		ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9235 		if (ret) {
9236 			dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9237 			return ret;
9238 		}
9239 		break;
9240 
9241 #endif /* I40E_FCOE */
9242 	default:
9243 		return -ENODEV;
9244 	}
9245 
9246 	if (vsi->type != I40E_VSI_MAIN) {
9247 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9248 		if (ret) {
9249 			dev_info(&vsi->back->pdev->dev,
9250 				 "add vsi failed, err %s aq_err %s\n",
9251 				 i40e_stat_str(&pf->hw, ret),
9252 				 i40e_aq_str(&pf->hw,
9253 					     pf->hw.aq.asq_last_status));
9254 			ret = -ENOENT;
9255 			goto err;
9256 		}
9257 		vsi->info = ctxt.info;
9258 		vsi->info.valid_sections = 0;
9259 		vsi->seid = ctxt.seid;
9260 		vsi->id = ctxt.vsi_number;
9261 	}
9262 
9263 	spin_lock_bh(&vsi->mac_filter_list_lock);
9264 	/* If macvlan filters already exist, force them to get loaded */
9265 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9266 		f->changed = true;
9267 		f_count++;
9268 
9269 		/* Expected to have only one MAC filter entry for LAA in list */
9270 		if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
9271 			ether_addr_copy(laa_macaddr, f->macaddr);
9272 			found_laa_mac_filter = true;
9273 		}
9274 	}
9275 	spin_unlock_bh(&vsi->mac_filter_list_lock);
9276 
9277 	if (found_laa_mac_filter) {
9278 		struct i40e_aqc_remove_macvlan_element_data element;
9279 
9280 		memset(&element, 0, sizeof(element));
9281 		ether_addr_copy(element.mac_addr, laa_macaddr);
9282 		element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
9283 		ret = i40e_aq_remove_macvlan(hw, vsi->seid,
9284 					     &element, 1, NULL);
9285 		if (ret) {
9286 			/* some older FW has a different default */
9287 			element.flags |=
9288 				       I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
9289 			i40e_aq_remove_macvlan(hw, vsi->seid,
9290 					       &element, 1, NULL);
9291 		}
9292 
9293 		i40e_aq_mac_address_write(hw,
9294 					  I40E_AQC_WRITE_TYPE_LAA_WOL,
9295 					  laa_macaddr, NULL);
9296 	}
9297 
9298 	if (f_count) {
9299 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9300 		pf->flags |= I40E_FLAG_FILTER_SYNC;
9301 	}
9302 
9303 	/* Update VSI BW information */
9304 	ret = i40e_vsi_get_bw_info(vsi);
9305 	if (ret) {
9306 		dev_info(&pf->pdev->dev,
9307 			 "couldn't get vsi bw info, err %s aq_err %s\n",
9308 			 i40e_stat_str(&pf->hw, ret),
9309 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9310 		/* VSI is already added so not tearing that up */
9311 		ret = 0;
9312 	}
9313 
9314 err:
9315 	return ret;
9316 }
9317 
9318 /**
9319  * i40e_vsi_release - Delete a VSI and free its resources
9320  * @vsi: the VSI being removed
9321  *
9322  * Returns 0 on success or < 0 on error
9323  **/
9324 int i40e_vsi_release(struct i40e_vsi *vsi)
9325 {
9326 	struct i40e_mac_filter *f, *ftmp;
9327 	struct i40e_veb *veb = NULL;
9328 	struct i40e_pf *pf;
9329 	u16 uplink_seid;
9330 	int i, n;
9331 
9332 	pf = vsi->back;
9333 
9334 	/* release of a VEB-owner or last VSI is not allowed */
9335 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9336 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9337 			 vsi->seid, vsi->uplink_seid);
9338 		return -ENODEV;
9339 	}
9340 	if (vsi == pf->vsi[pf->lan_vsi] &&
9341 	    !test_bit(__I40E_DOWN, &pf->state)) {
9342 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9343 		return -ENODEV;
9344 	}
9345 
9346 	uplink_seid = vsi->uplink_seid;
9347 	if (vsi->type != I40E_VSI_SRIOV) {
9348 		if (vsi->netdev_registered) {
9349 			vsi->netdev_registered = false;
9350 			if (vsi->netdev) {
9351 				/* results in a call to i40e_close() */
9352 				unregister_netdev(vsi->netdev);
9353 			}
9354 		} else {
9355 			i40e_vsi_close(vsi);
9356 		}
9357 		i40e_vsi_disable_irq(vsi);
9358 	}
9359 
9360 	spin_lock_bh(&vsi->mac_filter_list_lock);
9361 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9362 		i40e_del_filter(vsi, f->macaddr, f->vlan,
9363 				f->is_vf, f->is_netdev);
9364 	spin_unlock_bh(&vsi->mac_filter_list_lock);
9365 
9366 	i40e_sync_vsi_filters(vsi);
9367 
9368 	i40e_vsi_delete(vsi);
9369 	i40e_vsi_free_q_vectors(vsi);
9370 	if (vsi->netdev) {
9371 		free_netdev(vsi->netdev);
9372 		vsi->netdev = NULL;
9373 	}
9374 	i40e_vsi_clear_rings(vsi);
9375 	i40e_vsi_clear(vsi);
9376 
9377 	/* If this was the last thing on the VEB, except for the
9378 	 * controlling VSI, remove the VEB, which puts the controlling
9379 	 * VSI onto the next level down in the switch.
9380 	 *
9381 	 * Well, okay, there's one more exception here: don't remove
9382 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
9383 	 * from up the network stack.
9384 	 */
9385 	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9386 		if (pf->vsi[i] &&
9387 		    pf->vsi[i]->uplink_seid == uplink_seid &&
9388 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9389 			n++;      /* count the VSIs */
9390 		}
9391 	}
9392 	for (i = 0; i < I40E_MAX_VEB; i++) {
9393 		if (!pf->veb[i])
9394 			continue;
9395 		if (pf->veb[i]->uplink_seid == uplink_seid)
9396 			n++;     /* count the VEBs */
9397 		if (pf->veb[i]->seid == uplink_seid)
9398 			veb = pf->veb[i];
9399 	}
9400 	if (n == 0 && veb && veb->uplink_seid != 0)
9401 		i40e_veb_release(veb);
9402 
9403 	return 0;
9404 }
9405 
9406 /**
9407  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9408  * @vsi: ptr to the VSI
9409  *
9410  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9411  * corresponding SW VSI structure and initializes num_queue_pairs for the
9412  * newly allocated VSI.
9413  *
9414  * Returns 0 on success or negative on failure
9415  **/
9416 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9417 {
9418 	int ret = -ENOENT;
9419 	struct i40e_pf *pf = vsi->back;
9420 
9421 	if (vsi->q_vectors[0]) {
9422 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9423 			 vsi->seid);
9424 		return -EEXIST;
9425 	}
9426 
9427 	if (vsi->base_vector) {
9428 		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9429 			 vsi->seid, vsi->base_vector);
9430 		return -EEXIST;
9431 	}
9432 
9433 	ret = i40e_vsi_alloc_q_vectors(vsi);
9434 	if (ret) {
9435 		dev_info(&pf->pdev->dev,
9436 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9437 			 vsi->num_q_vectors, vsi->seid, ret);
9438 		vsi->num_q_vectors = 0;
9439 		goto vector_setup_out;
9440 	}
9441 
9442 	/* In Legacy mode, we do not have to get any other vector since we
9443 	 * piggyback on the misc/ICR0 for queue interrupts.
9444 	*/
9445 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9446 		return ret;
9447 	if (vsi->num_q_vectors)
9448 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9449 						 vsi->num_q_vectors, vsi->idx);
9450 	if (vsi->base_vector < 0) {
9451 		dev_info(&pf->pdev->dev,
9452 			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9453 			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9454 		i40e_vsi_free_q_vectors(vsi);
9455 		ret = -ENOENT;
9456 		goto vector_setup_out;
9457 	}
9458 
9459 vector_setup_out:
9460 	return ret;
9461 }
9462 
9463 /**
9464  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9465  * @vsi: pointer to the vsi.
9466  *
9467  * This re-allocates a vsi's queue resources.
9468  *
9469  * Returns pointer to the successfully allocated and configured VSI sw struct
9470  * on success, otherwise returns NULL on failure.
9471  **/
9472 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9473 {
9474 	struct i40e_pf *pf = vsi->back;
9475 	u8 enabled_tc;
9476 	int ret;
9477 
9478 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9479 	i40e_vsi_clear_rings(vsi);
9480 
9481 	i40e_vsi_free_arrays(vsi, false);
9482 	i40e_set_num_rings_in_vsi(vsi);
9483 	ret = i40e_vsi_alloc_arrays(vsi, false);
9484 	if (ret)
9485 		goto err_vsi;
9486 
9487 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9488 	if (ret < 0) {
9489 		dev_info(&pf->pdev->dev,
9490 			 "failed to get tracking for %d queues for VSI %d err %d\n",
9491 			 vsi->alloc_queue_pairs, vsi->seid, ret);
9492 		goto err_vsi;
9493 	}
9494 	vsi->base_queue = ret;
9495 
9496 	/* Update the FW view of the VSI. Force a reset of TC and queue
9497 	 * layout configurations.
9498 	 */
9499 	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9500 	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9501 	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9502 	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9503 
9504 	/* assign it some queues */
9505 	ret = i40e_alloc_rings(vsi);
9506 	if (ret)
9507 		goto err_rings;
9508 
9509 	/* map all of the rings to the q_vectors */
9510 	i40e_vsi_map_rings_to_vectors(vsi);
9511 	return vsi;
9512 
9513 err_rings:
9514 	i40e_vsi_free_q_vectors(vsi);
9515 	if (vsi->netdev_registered) {
9516 		vsi->netdev_registered = false;
9517 		unregister_netdev(vsi->netdev);
9518 		free_netdev(vsi->netdev);
9519 		vsi->netdev = NULL;
9520 	}
9521 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9522 err_vsi:
9523 	i40e_vsi_clear(vsi);
9524 	return NULL;
9525 }
9526 
9527 /**
9528  * i40e_macaddr_init - explicitly write the mac address filters.
9529  *
9530  * @vsi: pointer to the vsi.
9531  * @macaddr: the MAC address
9532  *
9533  * This is needed when the macaddr has been obtained by other
9534  * means than the default, e.g., from Open Firmware or IDPROM.
9535  * Returns 0 on success, negative on failure
9536  **/
9537 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
9538 {
9539 	int ret;
9540 	struct i40e_aqc_add_macvlan_element_data element;
9541 
9542 	ret = i40e_aq_mac_address_write(&vsi->back->hw,
9543 					I40E_AQC_WRITE_TYPE_LAA_WOL,
9544 					macaddr, NULL);
9545 	if (ret) {
9546 		dev_info(&vsi->back->pdev->dev,
9547 			 "Addr change for VSI failed: %d\n", ret);
9548 		return -EADDRNOTAVAIL;
9549 	}
9550 
9551 	memset(&element, 0, sizeof(element));
9552 	ether_addr_copy(element.mac_addr, macaddr);
9553 	element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
9554 	ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
9555 	if (ret) {
9556 		dev_info(&vsi->back->pdev->dev,
9557 			 "add filter failed err %s aq_err %s\n",
9558 			 i40e_stat_str(&vsi->back->hw, ret),
9559 			 i40e_aq_str(&vsi->back->hw,
9560 				     vsi->back->hw.aq.asq_last_status));
9561 	}
9562 	return ret;
9563 }
9564 
9565 /**
9566  * i40e_vsi_setup - Set up a VSI by a given type
9567  * @pf: board private structure
9568  * @type: VSI type
9569  * @uplink_seid: the switch element to link to
9570  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9571  *
9572  * This allocates the sw VSI structure and its queue resources, then add a VSI
9573  * to the identified VEB.
9574  *
9575  * Returns pointer to the successfully allocated and configure VSI sw struct on
9576  * success, otherwise returns NULL on failure.
9577  **/
9578 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9579 				u16 uplink_seid, u32 param1)
9580 {
9581 	struct i40e_vsi *vsi = NULL;
9582 	struct i40e_veb *veb = NULL;
9583 	int ret, i;
9584 	int v_idx;
9585 
9586 	/* The requested uplink_seid must be either
9587 	 *     - the PF's port seid
9588 	 *              no VEB is needed because this is the PF
9589 	 *              or this is a Flow Director special case VSI
9590 	 *     - seid of an existing VEB
9591 	 *     - seid of a VSI that owns an existing VEB
9592 	 *     - seid of a VSI that doesn't own a VEB
9593 	 *              a new VEB is created and the VSI becomes the owner
9594 	 *     - seid of the PF VSI, which is what creates the first VEB
9595 	 *              this is a special case of the previous
9596 	 *
9597 	 * Find which uplink_seid we were given and create a new VEB if needed
9598 	 */
9599 	for (i = 0; i < I40E_MAX_VEB; i++) {
9600 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9601 			veb = pf->veb[i];
9602 			break;
9603 		}
9604 	}
9605 
9606 	if (!veb && uplink_seid != pf->mac_seid) {
9607 
9608 		for (i = 0; i < pf->num_alloc_vsi; i++) {
9609 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9610 				vsi = pf->vsi[i];
9611 				break;
9612 			}
9613 		}
9614 		if (!vsi) {
9615 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9616 				 uplink_seid);
9617 			return NULL;
9618 		}
9619 
9620 		if (vsi->uplink_seid == pf->mac_seid)
9621 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9622 					     vsi->tc_config.enabled_tc);
9623 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9624 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9625 					     vsi->tc_config.enabled_tc);
9626 		if (veb) {
9627 			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9628 				dev_info(&vsi->back->pdev->dev,
9629 					 "New VSI creation error, uplink seid of LAN VSI expected.\n");
9630 				return NULL;
9631 			}
9632 			/* We come up by default in VEPA mode if SRIOV is not
9633 			 * already enabled, in which case we can't force VEPA
9634 			 * mode.
9635 			 */
9636 			if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9637 				veb->bridge_mode = BRIDGE_MODE_VEPA;
9638 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9639 			}
9640 			i40e_config_bridge_mode(veb);
9641 		}
9642 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9643 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9644 				veb = pf->veb[i];
9645 		}
9646 		if (!veb) {
9647 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9648 			return NULL;
9649 		}
9650 
9651 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9652 		uplink_seid = veb->seid;
9653 	}
9654 
9655 	/* get vsi sw struct */
9656 	v_idx = i40e_vsi_mem_alloc(pf, type);
9657 	if (v_idx < 0)
9658 		goto err_alloc;
9659 	vsi = pf->vsi[v_idx];
9660 	if (!vsi)
9661 		goto err_alloc;
9662 	vsi->type = type;
9663 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9664 
9665 	if (type == I40E_VSI_MAIN)
9666 		pf->lan_vsi = v_idx;
9667 	else if (type == I40E_VSI_SRIOV)
9668 		vsi->vf_id = param1;
9669 	/* assign it some queues */
9670 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9671 				vsi->idx);
9672 	if (ret < 0) {
9673 		dev_info(&pf->pdev->dev,
9674 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
9675 			 vsi->alloc_queue_pairs, vsi->seid, ret);
9676 		goto err_vsi;
9677 	}
9678 	vsi->base_queue = ret;
9679 
9680 	/* get a VSI from the hardware */
9681 	vsi->uplink_seid = uplink_seid;
9682 	ret = i40e_add_vsi(vsi);
9683 	if (ret)
9684 		goto err_vsi;
9685 
9686 	switch (vsi->type) {
9687 	/* setup the netdev if needed */
9688 	case I40E_VSI_MAIN:
9689 		/* Apply relevant filters if a platform-specific mac
9690 		 * address was selected.
9691 		 */
9692 		if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9693 			ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9694 			if (ret) {
9695 				dev_warn(&pf->pdev->dev,
9696 					 "could not set up macaddr; err %d\n",
9697 					 ret);
9698 			}
9699 		}
9700 	case I40E_VSI_VMDQ2:
9701 	case I40E_VSI_FCOE:
9702 		ret = i40e_config_netdev(vsi);
9703 		if (ret)
9704 			goto err_netdev;
9705 		ret = register_netdev(vsi->netdev);
9706 		if (ret)
9707 			goto err_netdev;
9708 		vsi->netdev_registered = true;
9709 		netif_carrier_off(vsi->netdev);
9710 #ifdef CONFIG_I40E_DCB
9711 		/* Setup DCB netlink interface */
9712 		i40e_dcbnl_setup(vsi);
9713 #endif /* CONFIG_I40E_DCB */
9714 		/* fall through */
9715 
9716 	case I40E_VSI_FDIR:
9717 		/* set up vectors and rings if needed */
9718 		ret = i40e_vsi_setup_vectors(vsi);
9719 		if (ret)
9720 			goto err_msix;
9721 
9722 		ret = i40e_alloc_rings(vsi);
9723 		if (ret)
9724 			goto err_rings;
9725 
9726 		/* map all of the rings to the q_vectors */
9727 		i40e_vsi_map_rings_to_vectors(vsi);
9728 
9729 		i40e_vsi_reset_stats(vsi);
9730 		break;
9731 
9732 	default:
9733 		/* no netdev or rings for the other VSI types */
9734 		break;
9735 	}
9736 
9737 	if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9738 	    (vsi->type == I40E_VSI_VMDQ2)) {
9739 		ret = i40e_vsi_config_rss(vsi);
9740 	}
9741 	return vsi;
9742 
9743 err_rings:
9744 	i40e_vsi_free_q_vectors(vsi);
9745 err_msix:
9746 	if (vsi->netdev_registered) {
9747 		vsi->netdev_registered = false;
9748 		unregister_netdev(vsi->netdev);
9749 		free_netdev(vsi->netdev);
9750 		vsi->netdev = NULL;
9751 	}
9752 err_netdev:
9753 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9754 err_vsi:
9755 	i40e_vsi_clear(vsi);
9756 err_alloc:
9757 	return NULL;
9758 }
9759 
9760 /**
9761  * i40e_veb_get_bw_info - Query VEB BW information
9762  * @veb: the veb to query
9763  *
9764  * Query the Tx scheduler BW configuration data for given VEB
9765  **/
9766 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9767 {
9768 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9769 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9770 	struct i40e_pf *pf = veb->pf;
9771 	struct i40e_hw *hw = &pf->hw;
9772 	u32 tc_bw_max;
9773 	int ret = 0;
9774 	int i;
9775 
9776 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9777 						  &bw_data, NULL);
9778 	if (ret) {
9779 		dev_info(&pf->pdev->dev,
9780 			 "query veb bw config failed, err %s aq_err %s\n",
9781 			 i40e_stat_str(&pf->hw, ret),
9782 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9783 		goto out;
9784 	}
9785 
9786 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9787 						   &ets_data, NULL);
9788 	if (ret) {
9789 		dev_info(&pf->pdev->dev,
9790 			 "query veb bw ets config failed, err %s aq_err %s\n",
9791 			 i40e_stat_str(&pf->hw, ret),
9792 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9793 		goto out;
9794 	}
9795 
9796 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9797 	veb->bw_max_quanta = ets_data.tc_bw_max;
9798 	veb->is_abs_credits = bw_data.absolute_credits_enable;
9799 	veb->enabled_tc = ets_data.tc_valid_bits;
9800 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9801 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9802 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9803 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9804 		veb->bw_tc_limit_credits[i] =
9805 					le16_to_cpu(bw_data.tc_bw_limits[i]);
9806 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9807 	}
9808 
9809 out:
9810 	return ret;
9811 }
9812 
9813 /**
9814  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9815  * @pf: board private structure
9816  *
9817  * On error: returns error code (negative)
9818  * On success: returns vsi index in PF (positive)
9819  **/
9820 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9821 {
9822 	int ret = -ENOENT;
9823 	struct i40e_veb *veb;
9824 	int i;
9825 
9826 	/* Need to protect the allocation of switch elements at the PF level */
9827 	mutex_lock(&pf->switch_mutex);
9828 
9829 	/* VEB list may be fragmented if VEB creation/destruction has
9830 	 * been happening.  We can afford to do a quick scan to look
9831 	 * for any free slots in the list.
9832 	 *
9833 	 * find next empty veb slot, looping back around if necessary
9834 	 */
9835 	i = 0;
9836 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9837 		i++;
9838 	if (i >= I40E_MAX_VEB) {
9839 		ret = -ENOMEM;
9840 		goto err_alloc_veb;  /* out of VEB slots! */
9841 	}
9842 
9843 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9844 	if (!veb) {
9845 		ret = -ENOMEM;
9846 		goto err_alloc_veb;
9847 	}
9848 	veb->pf = pf;
9849 	veb->idx = i;
9850 	veb->enabled_tc = 1;
9851 
9852 	pf->veb[i] = veb;
9853 	ret = i;
9854 err_alloc_veb:
9855 	mutex_unlock(&pf->switch_mutex);
9856 	return ret;
9857 }
9858 
9859 /**
9860  * i40e_switch_branch_release - Delete a branch of the switch tree
9861  * @branch: where to start deleting
9862  *
9863  * This uses recursion to find the tips of the branch to be
9864  * removed, deleting until we get back to and can delete this VEB.
9865  **/
9866 static void i40e_switch_branch_release(struct i40e_veb *branch)
9867 {
9868 	struct i40e_pf *pf = branch->pf;
9869 	u16 branch_seid = branch->seid;
9870 	u16 veb_idx = branch->idx;
9871 	int i;
9872 
9873 	/* release any VEBs on this VEB - RECURSION */
9874 	for (i = 0; i < I40E_MAX_VEB; i++) {
9875 		if (!pf->veb[i])
9876 			continue;
9877 		if (pf->veb[i]->uplink_seid == branch->seid)
9878 			i40e_switch_branch_release(pf->veb[i]);
9879 	}
9880 
9881 	/* Release the VSIs on this VEB, but not the owner VSI.
9882 	 *
9883 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
9884 	 *       the VEB itself, so don't use (*branch) after this loop.
9885 	 */
9886 	for (i = 0; i < pf->num_alloc_vsi; i++) {
9887 		if (!pf->vsi[i])
9888 			continue;
9889 		if (pf->vsi[i]->uplink_seid == branch_seid &&
9890 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9891 			i40e_vsi_release(pf->vsi[i]);
9892 		}
9893 	}
9894 
9895 	/* There's one corner case where the VEB might not have been
9896 	 * removed, so double check it here and remove it if needed.
9897 	 * This case happens if the veb was created from the debugfs
9898 	 * commands and no VSIs were added to it.
9899 	 */
9900 	if (pf->veb[veb_idx])
9901 		i40e_veb_release(pf->veb[veb_idx]);
9902 }
9903 
9904 /**
9905  * i40e_veb_clear - remove veb struct
9906  * @veb: the veb to remove
9907  **/
9908 static void i40e_veb_clear(struct i40e_veb *veb)
9909 {
9910 	if (!veb)
9911 		return;
9912 
9913 	if (veb->pf) {
9914 		struct i40e_pf *pf = veb->pf;
9915 
9916 		mutex_lock(&pf->switch_mutex);
9917 		if (pf->veb[veb->idx] == veb)
9918 			pf->veb[veb->idx] = NULL;
9919 		mutex_unlock(&pf->switch_mutex);
9920 	}
9921 
9922 	kfree(veb);
9923 }
9924 
9925 /**
9926  * i40e_veb_release - Delete a VEB and free its resources
9927  * @veb: the VEB being removed
9928  **/
9929 void i40e_veb_release(struct i40e_veb *veb)
9930 {
9931 	struct i40e_vsi *vsi = NULL;
9932 	struct i40e_pf *pf;
9933 	int i, n = 0;
9934 
9935 	pf = veb->pf;
9936 
9937 	/* find the remaining VSI and check for extras */
9938 	for (i = 0; i < pf->num_alloc_vsi; i++) {
9939 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
9940 			n++;
9941 			vsi = pf->vsi[i];
9942 		}
9943 	}
9944 	if (n != 1) {
9945 		dev_info(&pf->pdev->dev,
9946 			 "can't remove VEB %d with %d VSIs left\n",
9947 			 veb->seid, n);
9948 		return;
9949 	}
9950 
9951 	/* move the remaining VSI to uplink veb */
9952 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
9953 	if (veb->uplink_seid) {
9954 		vsi->uplink_seid = veb->uplink_seid;
9955 		if (veb->uplink_seid == pf->mac_seid)
9956 			vsi->veb_idx = I40E_NO_VEB;
9957 		else
9958 			vsi->veb_idx = veb->veb_idx;
9959 	} else {
9960 		/* floating VEB */
9961 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9962 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
9963 	}
9964 
9965 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
9966 	i40e_veb_clear(veb);
9967 }
9968 
9969 /**
9970  * i40e_add_veb - create the VEB in the switch
9971  * @veb: the VEB to be instantiated
9972  * @vsi: the controlling VSI
9973  **/
9974 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
9975 {
9976 	struct i40e_pf *pf = veb->pf;
9977 	bool is_default = veb->pf->cur_promisc;
9978 	bool is_cloud = false;
9979 	int ret;
9980 
9981 	/* get a VEB from the hardware */
9982 	ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
9983 			      veb->enabled_tc, is_default,
9984 			      is_cloud, &veb->seid, NULL);
9985 	if (ret) {
9986 		dev_info(&pf->pdev->dev,
9987 			 "couldn't add VEB, err %s aq_err %s\n",
9988 			 i40e_stat_str(&pf->hw, ret),
9989 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9990 		return -EPERM;
9991 	}
9992 
9993 	/* get statistics counter */
9994 	ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
9995 					 &veb->stats_idx, NULL, NULL, NULL);
9996 	if (ret) {
9997 		dev_info(&pf->pdev->dev,
9998 			 "couldn't get VEB statistics idx, err %s aq_err %s\n",
9999 			 i40e_stat_str(&pf->hw, ret),
10000 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10001 		return -EPERM;
10002 	}
10003 	ret = i40e_veb_get_bw_info(veb);
10004 	if (ret) {
10005 		dev_info(&pf->pdev->dev,
10006 			 "couldn't get VEB bw info, err %s aq_err %s\n",
10007 			 i40e_stat_str(&pf->hw, ret),
10008 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10009 		i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10010 		return -ENOENT;
10011 	}
10012 
10013 	vsi->uplink_seid = veb->seid;
10014 	vsi->veb_idx = veb->idx;
10015 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10016 
10017 	return 0;
10018 }
10019 
10020 /**
10021  * i40e_veb_setup - Set up a VEB
10022  * @pf: board private structure
10023  * @flags: VEB setup flags
10024  * @uplink_seid: the switch element to link to
10025  * @vsi_seid: the initial VSI seid
10026  * @enabled_tc: Enabled TC bit-map
10027  *
10028  * This allocates the sw VEB structure and links it into the switch
10029  * It is possible and legal for this to be a duplicate of an already
10030  * existing VEB.  It is also possible for both uplink and vsi seids
10031  * to be zero, in order to create a floating VEB.
10032  *
10033  * Returns pointer to the successfully allocated VEB sw struct on
10034  * success, otherwise returns NULL on failure.
10035  **/
10036 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10037 				u16 uplink_seid, u16 vsi_seid,
10038 				u8 enabled_tc)
10039 {
10040 	struct i40e_veb *veb, *uplink_veb = NULL;
10041 	int vsi_idx, veb_idx;
10042 	int ret;
10043 
10044 	/* if one seid is 0, the other must be 0 to create a floating relay */
10045 	if ((uplink_seid == 0 || vsi_seid == 0) &&
10046 	    (uplink_seid + vsi_seid != 0)) {
10047 		dev_info(&pf->pdev->dev,
10048 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
10049 			 uplink_seid, vsi_seid);
10050 		return NULL;
10051 	}
10052 
10053 	/* make sure there is such a vsi and uplink */
10054 	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10055 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10056 			break;
10057 	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10058 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10059 			 vsi_seid);
10060 		return NULL;
10061 	}
10062 
10063 	if (uplink_seid && uplink_seid != pf->mac_seid) {
10064 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10065 			if (pf->veb[veb_idx] &&
10066 			    pf->veb[veb_idx]->seid == uplink_seid) {
10067 				uplink_veb = pf->veb[veb_idx];
10068 				break;
10069 			}
10070 		}
10071 		if (!uplink_veb) {
10072 			dev_info(&pf->pdev->dev,
10073 				 "uplink seid %d not found\n", uplink_seid);
10074 			return NULL;
10075 		}
10076 	}
10077 
10078 	/* get veb sw struct */
10079 	veb_idx = i40e_veb_mem_alloc(pf);
10080 	if (veb_idx < 0)
10081 		goto err_alloc;
10082 	veb = pf->veb[veb_idx];
10083 	veb->flags = flags;
10084 	veb->uplink_seid = uplink_seid;
10085 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10086 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10087 
10088 	/* create the VEB in the switch */
10089 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10090 	if (ret)
10091 		goto err_veb;
10092 	if (vsi_idx == pf->lan_vsi)
10093 		pf->lan_veb = veb->idx;
10094 
10095 	return veb;
10096 
10097 err_veb:
10098 	i40e_veb_clear(veb);
10099 err_alloc:
10100 	return NULL;
10101 }
10102 
10103 /**
10104  * i40e_setup_pf_switch_element - set PF vars based on switch type
10105  * @pf: board private structure
10106  * @ele: element we are building info from
10107  * @num_reported: total number of elements
10108  * @printconfig: should we print the contents
10109  *
10110  * helper function to assist in extracting a few useful SEID values.
10111  **/
10112 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10113 				struct i40e_aqc_switch_config_element_resp *ele,
10114 				u16 num_reported, bool printconfig)
10115 {
10116 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10117 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10118 	u8 element_type = ele->element_type;
10119 	u16 seid = le16_to_cpu(ele->seid);
10120 
10121 	if (printconfig)
10122 		dev_info(&pf->pdev->dev,
10123 			 "type=%d seid=%d uplink=%d downlink=%d\n",
10124 			 element_type, seid, uplink_seid, downlink_seid);
10125 
10126 	switch (element_type) {
10127 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
10128 		pf->mac_seid = seid;
10129 		break;
10130 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
10131 		/* Main VEB? */
10132 		if (uplink_seid != pf->mac_seid)
10133 			break;
10134 		if (pf->lan_veb == I40E_NO_VEB) {
10135 			int v;
10136 
10137 			/* find existing or else empty VEB */
10138 			for (v = 0; v < I40E_MAX_VEB; v++) {
10139 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10140 					pf->lan_veb = v;
10141 					break;
10142 				}
10143 			}
10144 			if (pf->lan_veb == I40E_NO_VEB) {
10145 				v = i40e_veb_mem_alloc(pf);
10146 				if (v < 0)
10147 					break;
10148 				pf->lan_veb = v;
10149 			}
10150 		}
10151 
10152 		pf->veb[pf->lan_veb]->seid = seid;
10153 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10154 		pf->veb[pf->lan_veb]->pf = pf;
10155 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10156 		break;
10157 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
10158 		if (num_reported != 1)
10159 			break;
10160 		/* This is immediately after a reset so we can assume this is
10161 		 * the PF's VSI
10162 		 */
10163 		pf->mac_seid = uplink_seid;
10164 		pf->pf_seid = downlink_seid;
10165 		pf->main_vsi_seid = seid;
10166 		if (printconfig)
10167 			dev_info(&pf->pdev->dev,
10168 				 "pf_seid=%d main_vsi_seid=%d\n",
10169 				 pf->pf_seid, pf->main_vsi_seid);
10170 		break;
10171 	case I40E_SWITCH_ELEMENT_TYPE_PF:
10172 	case I40E_SWITCH_ELEMENT_TYPE_VF:
10173 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
10174 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
10175 	case I40E_SWITCH_ELEMENT_TYPE_PE:
10176 	case I40E_SWITCH_ELEMENT_TYPE_PA:
10177 		/* ignore these for now */
10178 		break;
10179 	default:
10180 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10181 			 element_type, seid);
10182 		break;
10183 	}
10184 }
10185 
10186 /**
10187  * i40e_fetch_switch_configuration - Get switch config from firmware
10188  * @pf: board private structure
10189  * @printconfig: should we print the contents
10190  *
10191  * Get the current switch configuration from the device and
10192  * extract a few useful SEID values.
10193  **/
10194 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10195 {
10196 	struct i40e_aqc_get_switch_config_resp *sw_config;
10197 	u16 next_seid = 0;
10198 	int ret = 0;
10199 	u8 *aq_buf;
10200 	int i;
10201 
10202 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10203 	if (!aq_buf)
10204 		return -ENOMEM;
10205 
10206 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10207 	do {
10208 		u16 num_reported, num_total;
10209 
10210 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10211 						I40E_AQ_LARGE_BUF,
10212 						&next_seid, NULL);
10213 		if (ret) {
10214 			dev_info(&pf->pdev->dev,
10215 				 "get switch config failed err %s aq_err %s\n",
10216 				 i40e_stat_str(&pf->hw, ret),
10217 				 i40e_aq_str(&pf->hw,
10218 					     pf->hw.aq.asq_last_status));
10219 			kfree(aq_buf);
10220 			return -ENOENT;
10221 		}
10222 
10223 		num_reported = le16_to_cpu(sw_config->header.num_reported);
10224 		num_total = le16_to_cpu(sw_config->header.num_total);
10225 
10226 		if (printconfig)
10227 			dev_info(&pf->pdev->dev,
10228 				 "header: %d reported %d total\n",
10229 				 num_reported, num_total);
10230 
10231 		for (i = 0; i < num_reported; i++) {
10232 			struct i40e_aqc_switch_config_element_resp *ele =
10233 				&sw_config->element[i];
10234 
10235 			i40e_setup_pf_switch_element(pf, ele, num_reported,
10236 						     printconfig);
10237 		}
10238 	} while (next_seid != 0);
10239 
10240 	kfree(aq_buf);
10241 	return ret;
10242 }
10243 
10244 /**
10245  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10246  * @pf: board private structure
10247  * @reinit: if the Main VSI needs to re-initialized.
10248  *
10249  * Returns 0 on success, negative value on failure
10250  **/
10251 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10252 {
10253 	int ret;
10254 
10255 	/* find out what's out there already */
10256 	ret = i40e_fetch_switch_configuration(pf, false);
10257 	if (ret) {
10258 		dev_info(&pf->pdev->dev,
10259 			 "couldn't fetch switch config, err %s aq_err %s\n",
10260 			 i40e_stat_str(&pf->hw, ret),
10261 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10262 		return ret;
10263 	}
10264 	i40e_pf_reset_stats(pf);
10265 
10266 	/* first time setup */
10267 	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10268 		struct i40e_vsi *vsi = NULL;
10269 		u16 uplink_seid;
10270 
10271 		/* Set up the PF VSI associated with the PF's main VSI
10272 		 * that is already in the HW switch
10273 		 */
10274 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10275 			uplink_seid = pf->veb[pf->lan_veb]->seid;
10276 		else
10277 			uplink_seid = pf->mac_seid;
10278 		if (pf->lan_vsi == I40E_NO_VSI)
10279 			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10280 		else if (reinit)
10281 			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10282 		if (!vsi) {
10283 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10284 			i40e_fdir_teardown(pf);
10285 			return -EAGAIN;
10286 		}
10287 	} else {
10288 		/* force a reset of TC and queue layout configurations */
10289 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10290 
10291 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10292 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10293 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10294 	}
10295 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10296 
10297 	i40e_fdir_sb_setup(pf);
10298 
10299 	/* Setup static PF queue filter control settings */
10300 	ret = i40e_setup_pf_filter_control(pf);
10301 	if (ret) {
10302 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10303 			 ret);
10304 		/* Failure here should not stop continuing other steps */
10305 	}
10306 
10307 	/* enable RSS in the HW, even for only one queue, as the stack can use
10308 	 * the hash
10309 	 */
10310 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10311 		i40e_pf_config_rss(pf);
10312 
10313 	/* fill in link information and enable LSE reporting */
10314 	i40e_update_link_info(&pf->hw);
10315 	i40e_link_event(pf);
10316 
10317 	/* Initialize user-specific link properties */
10318 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10319 				  I40E_AQ_AN_COMPLETED) ? true : false);
10320 
10321 	i40e_ptp_init(pf);
10322 
10323 	return ret;
10324 }
10325 
10326 /**
10327  * i40e_determine_queue_usage - Work out queue distribution
10328  * @pf: board private structure
10329  **/
10330 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10331 {
10332 	int queues_left;
10333 
10334 	pf->num_lan_qps = 0;
10335 #ifdef I40E_FCOE
10336 	pf->num_fcoe_qps = 0;
10337 #endif
10338 
10339 	/* Find the max queues to be put into basic use.  We'll always be
10340 	 * using TC0, whether or not DCB is running, and TC0 will get the
10341 	 * big RSS set.
10342 	 */
10343 	queues_left = pf->hw.func_caps.num_tx_qp;
10344 
10345 	if ((queues_left == 1) ||
10346 	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10347 		/* one qp for PF, no queues for anything else */
10348 		queues_left = 0;
10349 		pf->alloc_rss_size = pf->num_lan_qps = 1;
10350 
10351 		/* make sure all the fancies are disabled */
10352 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
10353 #ifdef I40E_FCOE
10354 			       I40E_FLAG_FCOE_ENABLED	|
10355 #endif
10356 			       I40E_FLAG_FD_SB_ENABLED	|
10357 			       I40E_FLAG_FD_ATR_ENABLED	|
10358 			       I40E_FLAG_DCB_CAPABLE	|
10359 			       I40E_FLAG_SRIOV_ENABLED	|
10360 			       I40E_FLAG_VMDQ_ENABLED);
10361 	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10362 				  I40E_FLAG_FD_SB_ENABLED |
10363 				  I40E_FLAG_FD_ATR_ENABLED |
10364 				  I40E_FLAG_DCB_CAPABLE))) {
10365 		/* one qp for PF */
10366 		pf->alloc_rss_size = pf->num_lan_qps = 1;
10367 		queues_left -= pf->num_lan_qps;
10368 
10369 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
10370 #ifdef I40E_FCOE
10371 			       I40E_FLAG_FCOE_ENABLED	|
10372 #endif
10373 			       I40E_FLAG_FD_SB_ENABLED	|
10374 			       I40E_FLAG_FD_ATR_ENABLED	|
10375 			       I40E_FLAG_DCB_ENABLED	|
10376 			       I40E_FLAG_VMDQ_ENABLED);
10377 	} else {
10378 		/* Not enough queues for all TCs */
10379 		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10380 		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10381 			pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10382 			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10383 		}
10384 		pf->num_lan_qps = max_t(int, pf->rss_size_max,
10385 					num_online_cpus());
10386 		pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10387 					pf->hw.func_caps.num_tx_qp);
10388 
10389 		queues_left -= pf->num_lan_qps;
10390 	}
10391 
10392 #ifdef I40E_FCOE
10393 	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10394 		if (I40E_DEFAULT_FCOE <= queues_left) {
10395 			pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10396 		} else if (I40E_MINIMUM_FCOE <= queues_left) {
10397 			pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10398 		} else {
10399 			pf->num_fcoe_qps = 0;
10400 			pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10401 			dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10402 		}
10403 
10404 		queues_left -= pf->num_fcoe_qps;
10405 	}
10406 
10407 #endif
10408 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10409 		if (queues_left > 1) {
10410 			queues_left -= 1; /* save 1 queue for FD */
10411 		} else {
10412 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10413 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10414 		}
10415 	}
10416 
10417 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10418 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10419 		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10420 					(queues_left / pf->num_vf_qps));
10421 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10422 	}
10423 
10424 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10425 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10426 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10427 					  (queues_left / pf->num_vmdq_qps));
10428 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10429 	}
10430 
10431 	pf->queues_left = queues_left;
10432 	dev_dbg(&pf->pdev->dev,
10433 		"qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10434 		pf->hw.func_caps.num_tx_qp,
10435 		!!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10436 		pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10437 		pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10438 		queues_left);
10439 #ifdef I40E_FCOE
10440 	dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10441 #endif
10442 }
10443 
10444 /**
10445  * i40e_setup_pf_filter_control - Setup PF static filter control
10446  * @pf: PF to be setup
10447  *
10448  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10449  * settings. If PE/FCoE are enabled then it will also set the per PF
10450  * based filter sizes required for them. It also enables Flow director,
10451  * ethertype and macvlan type filter settings for the pf.
10452  *
10453  * Returns 0 on success, negative on failure
10454  **/
10455 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10456 {
10457 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
10458 
10459 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10460 
10461 	/* Flow Director is enabled */
10462 	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10463 		settings->enable_fdir = true;
10464 
10465 	/* Ethtype and MACVLAN filters enabled for PF */
10466 	settings->enable_ethtype = true;
10467 	settings->enable_macvlan = true;
10468 
10469 	if (i40e_set_filter_control(&pf->hw, settings))
10470 		return -ENOENT;
10471 
10472 	return 0;
10473 }
10474 
10475 #define INFO_STRING_LEN 255
10476 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10477 static void i40e_print_features(struct i40e_pf *pf)
10478 {
10479 	struct i40e_hw *hw = &pf->hw;
10480 	char *buf;
10481 	int i;
10482 
10483 	buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10484 	if (!buf)
10485 		return;
10486 
10487 	i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10488 #ifdef CONFIG_PCI_IOV
10489 	i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10490 #endif
10491 	i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d RX: %s",
10492 		      pf->hw.func_caps.num_vsis,
10493 		      pf->vsi[pf->lan_vsi]->num_queue_pairs,
10494 		      pf->flags & I40E_FLAG_RX_PS_ENABLED ? "PS" : "1BUF");
10495 
10496 	if (pf->flags & I40E_FLAG_RSS_ENABLED)
10497 		i += snprintf(&buf[i], REMAIN(i), " RSS");
10498 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10499 		i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10500 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10501 		i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10502 		i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10503 	}
10504 	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10505 		i += snprintf(&buf[i], REMAIN(i), " DCB");
10506 #if IS_ENABLED(CONFIG_VXLAN)
10507 	i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10508 #endif
10509 #if IS_ENABLED(CONFIG_GENEVE)
10510 	i += snprintf(&buf[i], REMAIN(i), " Geneve");
10511 #endif
10512 	if (pf->flags & I40E_FLAG_PTP)
10513 		i += snprintf(&buf[i], REMAIN(i), " PTP");
10514 #ifdef I40E_FCOE
10515 	if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10516 		i += snprintf(&buf[i], REMAIN(i), " FCOE");
10517 #endif
10518 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10519 		i += snprintf(&buf[i], REMAIN(i), " VEB");
10520 	else
10521 		i += snprintf(&buf[i], REMAIN(i), " VEPA");
10522 
10523 	dev_info(&pf->pdev->dev, "%s\n", buf);
10524 	kfree(buf);
10525 	WARN_ON(i > INFO_STRING_LEN);
10526 }
10527 
10528 /**
10529  * i40e_get_platform_mac_addr - get platform-specific MAC address
10530  *
10531  * @pdev: PCI device information struct
10532  * @pf: board private structure
10533  *
10534  * Look up the MAC address in Open Firmware  on systems that support it,
10535  * and use IDPROM on SPARC if no OF address is found. On return, the
10536  * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10537  * has been selected.
10538  **/
10539 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10540 {
10541 	struct device_node *dp = pci_device_to_OF_node(pdev);
10542 	const unsigned char *addr;
10543 	u8 *mac_addr = pf->hw.mac.addr;
10544 
10545 	pf->flags &= ~I40E_FLAG_PF_MAC;
10546 	addr = of_get_mac_address(dp);
10547 	if (addr) {
10548 		ether_addr_copy(mac_addr, addr);
10549 		pf->flags |= I40E_FLAG_PF_MAC;
10550 #ifdef CONFIG_SPARC
10551 	} else {
10552 		ether_addr_copy(mac_addr, idprom->id_ethaddr);
10553 		pf->flags |= I40E_FLAG_PF_MAC;
10554 #endif /* CONFIG_SPARC */
10555 	}
10556 }
10557 
10558 /**
10559  * i40e_probe - Device initialization routine
10560  * @pdev: PCI device information struct
10561  * @ent: entry in i40e_pci_tbl
10562  *
10563  * i40e_probe initializes a PF identified by a pci_dev structure.
10564  * The OS initialization, configuring of the PF private structure,
10565  * and a hardware reset occur.
10566  *
10567  * Returns 0 on success, negative on failure
10568  **/
10569 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10570 {
10571 	struct i40e_aq_get_phy_abilities_resp abilities;
10572 	struct i40e_pf *pf;
10573 	struct i40e_hw *hw;
10574 	static u16 pfs_found;
10575 	u16 wol_nvm_bits;
10576 	u16 link_status;
10577 	int err;
10578 	u32 len;
10579 	u32 val;
10580 	u32 i;
10581 	u8 set_fc_aq_fail;
10582 
10583 	err = pci_enable_device_mem(pdev);
10584 	if (err)
10585 		return err;
10586 
10587 	/* set up for high or low dma */
10588 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10589 	if (err) {
10590 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10591 		if (err) {
10592 			dev_err(&pdev->dev,
10593 				"DMA configuration failed: 0x%x\n", err);
10594 			goto err_dma;
10595 		}
10596 	}
10597 
10598 	/* set up pci connections */
10599 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
10600 					   IORESOURCE_MEM), i40e_driver_name);
10601 	if (err) {
10602 		dev_info(&pdev->dev,
10603 			 "pci_request_selected_regions failed %d\n", err);
10604 		goto err_pci_reg;
10605 	}
10606 
10607 	pci_enable_pcie_error_reporting(pdev);
10608 	pci_set_master(pdev);
10609 
10610 	/* Now that we have a PCI connection, we need to do the
10611 	 * low level device setup.  This is primarily setting up
10612 	 * the Admin Queue structures and then querying for the
10613 	 * device's current profile information.
10614 	 */
10615 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10616 	if (!pf) {
10617 		err = -ENOMEM;
10618 		goto err_pf_alloc;
10619 	}
10620 	pf->next_vsi = 0;
10621 	pf->pdev = pdev;
10622 	set_bit(__I40E_DOWN, &pf->state);
10623 
10624 	hw = &pf->hw;
10625 	hw->back = pf;
10626 
10627 	pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10628 				I40E_MAX_CSR_SPACE);
10629 
10630 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10631 	if (!hw->hw_addr) {
10632 		err = -EIO;
10633 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10634 			 (unsigned int)pci_resource_start(pdev, 0),
10635 			 pf->ioremap_len, err);
10636 		goto err_ioremap;
10637 	}
10638 	hw->vendor_id = pdev->vendor;
10639 	hw->device_id = pdev->device;
10640 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10641 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
10642 	hw->subsystem_device_id = pdev->subsystem_device;
10643 	hw->bus.device = PCI_SLOT(pdev->devfn);
10644 	hw->bus.func = PCI_FUNC(pdev->devfn);
10645 	pf->instance = pfs_found;
10646 
10647 	if (debug != -1) {
10648 		pf->msg_enable = pf->hw.debug_mask;
10649 		pf->msg_enable = debug;
10650 	}
10651 
10652 	/* do a special CORER for clearing PXE mode once at init */
10653 	if (hw->revision_id == 0 &&
10654 	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10655 		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10656 		i40e_flush(hw);
10657 		msleep(200);
10658 		pf->corer_count++;
10659 
10660 		i40e_clear_pxe_mode(hw);
10661 	}
10662 
10663 	/* Reset here to make sure all is clean and to define PF 'n' */
10664 	i40e_clear_hw(hw);
10665 	err = i40e_pf_reset(hw);
10666 	if (err) {
10667 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10668 		goto err_pf_reset;
10669 	}
10670 	pf->pfr_count++;
10671 
10672 	hw->aq.num_arq_entries = I40E_AQ_LEN;
10673 	hw->aq.num_asq_entries = I40E_AQ_LEN;
10674 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10675 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10676 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10677 
10678 	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10679 		 "%s-%s:misc",
10680 		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10681 
10682 	err = i40e_init_shared_code(hw);
10683 	if (err) {
10684 		dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10685 			 err);
10686 		goto err_pf_reset;
10687 	}
10688 
10689 	/* set up a default setting for link flow control */
10690 	pf->hw.fc.requested_mode = I40E_FC_NONE;
10691 
10692 	/* set up the locks for the AQ, do this only once in probe
10693 	 * and destroy them only once in remove
10694 	 */
10695 	mutex_init(&hw->aq.asq_mutex);
10696 	mutex_init(&hw->aq.arq_mutex);
10697 
10698 	err = i40e_init_adminq(hw);
10699 	if (err) {
10700 		if (err == I40E_ERR_FIRMWARE_API_VERSION)
10701 			dev_info(&pdev->dev,
10702 				 "The driver for the device stopped because the NVM image is newer than expected. You must install the most recent version of the network driver.\n");
10703 		else
10704 			dev_info(&pdev->dev,
10705 				 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10706 
10707 		goto err_pf_reset;
10708 	}
10709 
10710 	/* provide nvm, fw, api versions */
10711 	dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10712 		 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10713 		 hw->aq.api_maj_ver, hw->aq.api_min_ver,
10714 		 i40e_nvm_version_str(hw));
10715 
10716 	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10717 	    hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10718 		dev_info(&pdev->dev,
10719 			 "The driver for the device detected a newer version of the NVM image than expected. Please install the most recent version of the network driver.\n");
10720 	else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10721 		 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10722 		dev_info(&pdev->dev,
10723 			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10724 
10725 	i40e_verify_eeprom(pf);
10726 
10727 	/* Rev 0 hardware was never productized */
10728 	if (hw->revision_id < 1)
10729 		dev_warn(&pdev->dev, "This device is a pre-production adapter/LOM. Please be aware there may be issues with your hardware. If you are experiencing problems please contact your Intel or hardware representative who provided you with this hardware.\n");
10730 
10731 	i40e_clear_pxe_mode(hw);
10732 	err = i40e_get_capabilities(pf);
10733 	if (err)
10734 		goto err_adminq_setup;
10735 
10736 	err = i40e_sw_init(pf);
10737 	if (err) {
10738 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10739 		goto err_sw_init;
10740 	}
10741 
10742 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10743 				hw->func_caps.num_rx_qp,
10744 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10745 	if (err) {
10746 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10747 		goto err_init_lan_hmc;
10748 	}
10749 
10750 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10751 	if (err) {
10752 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10753 		err = -ENOENT;
10754 		goto err_configure_lan_hmc;
10755 	}
10756 
10757 	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
10758 	 * Ignore error return codes because if it was already disabled via
10759 	 * hardware settings this will fail
10760 	 */
10761 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
10762 	    (pf->hw.aq.fw_maj_ver < 4)) {
10763 		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10764 		i40e_aq_stop_lldp(hw, true, NULL);
10765 	}
10766 
10767 	i40e_get_mac_addr(hw, hw->mac.addr);
10768 	/* allow a platform config to override the HW addr */
10769 	i40e_get_platform_mac_addr(pdev, pf);
10770 	if (!is_valid_ether_addr(hw->mac.addr)) {
10771 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10772 		err = -EIO;
10773 		goto err_mac_addr;
10774 	}
10775 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10776 	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10777 	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10778 	if (is_valid_ether_addr(hw->mac.port_addr))
10779 		pf->flags |= I40E_FLAG_PORT_ID_VALID;
10780 #ifdef I40E_FCOE
10781 	err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10782 	if (err)
10783 		dev_info(&pdev->dev,
10784 			 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10785 	if (!is_valid_ether_addr(hw->mac.san_addr)) {
10786 		dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10787 			 hw->mac.san_addr);
10788 		ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10789 	}
10790 	dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10791 #endif /* I40E_FCOE */
10792 
10793 	pci_set_drvdata(pdev, pf);
10794 	pci_save_state(pdev);
10795 #ifdef CONFIG_I40E_DCB
10796 	err = i40e_init_pf_dcb(pf);
10797 	if (err) {
10798 		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10799 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10800 		/* Continue without DCB enabled */
10801 	}
10802 #endif /* CONFIG_I40E_DCB */
10803 
10804 	/* set up periodic task facility */
10805 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10806 	pf->service_timer_period = HZ;
10807 
10808 	INIT_WORK(&pf->service_task, i40e_service_task);
10809 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10810 	pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10811 
10812 	/* NVM bit on means WoL disabled for the port */
10813 	i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10814 	if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
10815 		pf->wol_en = false;
10816 	else
10817 		pf->wol_en = true;
10818 	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10819 
10820 	/* set up the main switch operations */
10821 	i40e_determine_queue_usage(pf);
10822 	err = i40e_init_interrupt_scheme(pf);
10823 	if (err)
10824 		goto err_switch_setup;
10825 
10826 	/* The number of VSIs reported by the FW is the minimum guaranteed
10827 	 * to us; HW supports far more and we share the remaining pool with
10828 	 * the other PFs. We allocate space for more than the guarantee with
10829 	 * the understanding that we might not get them all later.
10830 	 */
10831 	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10832 		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10833 	else
10834 		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10835 
10836 	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
10837 	len = sizeof(struct i40e_vsi *) * pf->num_alloc_vsi;
10838 	pf->vsi = kzalloc(len, GFP_KERNEL);
10839 	if (!pf->vsi) {
10840 		err = -ENOMEM;
10841 		goto err_switch_setup;
10842 	}
10843 
10844 #ifdef CONFIG_PCI_IOV
10845 	/* prep for VF support */
10846 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10847 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10848 	    !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10849 		if (pci_num_vf(pdev))
10850 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
10851 	}
10852 #endif
10853 	err = i40e_setup_pf_switch(pf, false);
10854 	if (err) {
10855 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
10856 		goto err_vsis;
10857 	}
10858 
10859 	/* Make sure flow control is set according to current settings */
10860 	err = i40e_set_fc(hw, &set_fc_aq_fail, true);
10861 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
10862 		dev_dbg(&pf->pdev->dev,
10863 			"Set fc with err %s aq_err %s on get_phy_cap\n",
10864 			i40e_stat_str(hw, err),
10865 			i40e_aq_str(hw, hw->aq.asq_last_status));
10866 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
10867 		dev_dbg(&pf->pdev->dev,
10868 			"Set fc with err %s aq_err %s on set_phy_config\n",
10869 			i40e_stat_str(hw, err),
10870 			i40e_aq_str(hw, hw->aq.asq_last_status));
10871 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
10872 		dev_dbg(&pf->pdev->dev,
10873 			"Set fc with err %s aq_err %s on get_link_info\n",
10874 			i40e_stat_str(hw, err),
10875 			i40e_aq_str(hw, hw->aq.asq_last_status));
10876 
10877 	/* if FDIR VSI was set up, start it now */
10878 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10879 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
10880 			i40e_vsi_open(pf->vsi[i]);
10881 			break;
10882 		}
10883 	}
10884 
10885 	/* driver is only interested in link up/down and module qualification
10886 	 * reports from firmware
10887 	 */
10888 	err = i40e_aq_set_phy_int_mask(&pf->hw,
10889 				       I40E_AQ_EVENT_LINK_UPDOWN |
10890 				       I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
10891 	if (err)
10892 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
10893 			 i40e_stat_str(&pf->hw, err),
10894 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10895 
10896 	/* Reconfigure hardware for allowing smaller MSS in the case
10897 	 * of TSO, so that we avoid the MDD being fired and causing
10898 	 * a reset in the case of small MSS+TSO.
10899 	 */
10900 	val = rd32(hw, I40E_REG_MSS);
10901 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
10902 		val &= ~I40E_REG_MSS_MIN_MASK;
10903 		val |= I40E_64BYTE_MSS;
10904 		wr32(hw, I40E_REG_MSS, val);
10905 	}
10906 
10907 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
10908 	    (pf->hw.aq.fw_maj_ver < 4)) {
10909 		msleep(75);
10910 		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
10911 		if (err)
10912 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
10913 				 i40e_stat_str(&pf->hw, err),
10914 				 i40e_aq_str(&pf->hw,
10915 					     pf->hw.aq.asq_last_status));
10916 	}
10917 	/* The main driver is (mostly) up and happy. We need to set this state
10918 	 * before setting up the misc vector or we get a race and the vector
10919 	 * ends up disabled forever.
10920 	 */
10921 	clear_bit(__I40E_DOWN, &pf->state);
10922 
10923 	/* In case of MSIX we are going to setup the misc vector right here
10924 	 * to handle admin queue events etc. In case of legacy and MSI
10925 	 * the misc functionality and queue processing is combined in
10926 	 * the same vector and that gets setup at open.
10927 	 */
10928 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
10929 		err = i40e_setup_misc_vector(pf);
10930 		if (err) {
10931 			dev_info(&pdev->dev,
10932 				 "setup of misc vector failed: %d\n", err);
10933 			goto err_vsis;
10934 		}
10935 	}
10936 
10937 #ifdef CONFIG_PCI_IOV
10938 	/* prep for VF support */
10939 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10940 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10941 	    !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10942 		u32 val;
10943 
10944 		/* disable link interrupts for VFs */
10945 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
10946 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
10947 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
10948 		i40e_flush(hw);
10949 
10950 		if (pci_num_vf(pdev)) {
10951 			dev_info(&pdev->dev,
10952 				 "Active VFs found, allocating resources.\n");
10953 			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
10954 			if (err)
10955 				dev_info(&pdev->dev,
10956 					 "Error %d allocating resources for existing VFs\n",
10957 					 err);
10958 		}
10959 	}
10960 #endif /* CONFIG_PCI_IOV */
10961 
10962 	pfs_found++;
10963 
10964 	i40e_dbg_pf_init(pf);
10965 
10966 	/* tell the firmware that we're starting */
10967 	i40e_send_version(pf);
10968 
10969 	/* since everything's happy, start the service_task timer */
10970 	mod_timer(&pf->service_timer,
10971 		  round_jiffies(jiffies + pf->service_timer_period));
10972 
10973 #ifdef I40E_FCOE
10974 	/* create FCoE interface */
10975 	i40e_fcoe_vsi_setup(pf);
10976 
10977 #endif
10978 #define PCI_SPEED_SIZE 8
10979 #define PCI_WIDTH_SIZE 8
10980 	/* Devices on the IOSF bus do not have this information
10981 	 * and will report PCI Gen 1 x 1 by default so don't bother
10982 	 * checking them.
10983 	 */
10984 	if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
10985 		char speed[PCI_SPEED_SIZE] = "Unknown";
10986 		char width[PCI_WIDTH_SIZE] = "Unknown";
10987 
10988 		/* Get the negotiated link width and speed from PCI config
10989 		 * space
10990 		 */
10991 		pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
10992 					  &link_status);
10993 
10994 		i40e_set_pci_config_data(hw, link_status);
10995 
10996 		switch (hw->bus.speed) {
10997 		case i40e_bus_speed_8000:
10998 			strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
10999 		case i40e_bus_speed_5000:
11000 			strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11001 		case i40e_bus_speed_2500:
11002 			strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11003 		default:
11004 			break;
11005 		}
11006 		switch (hw->bus.width) {
11007 		case i40e_bus_width_pcie_x8:
11008 			strncpy(width, "8", PCI_WIDTH_SIZE); break;
11009 		case i40e_bus_width_pcie_x4:
11010 			strncpy(width, "4", PCI_WIDTH_SIZE); break;
11011 		case i40e_bus_width_pcie_x2:
11012 			strncpy(width, "2", PCI_WIDTH_SIZE); break;
11013 		case i40e_bus_width_pcie_x1:
11014 			strncpy(width, "1", PCI_WIDTH_SIZE); break;
11015 		default:
11016 			break;
11017 		}
11018 
11019 		dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11020 			 speed, width);
11021 
11022 		if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11023 		    hw->bus.speed < i40e_bus_speed_8000) {
11024 			dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11025 			dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11026 		}
11027 	}
11028 
11029 	/* get the requested speeds from the fw */
11030 	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11031 	if (err)
11032 		dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11033 			i40e_stat_str(&pf->hw, err),
11034 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11035 	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11036 
11037 	/* get the supported phy types from the fw */
11038 	err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11039 	if (err)
11040 		dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11041 			i40e_stat_str(&pf->hw, err),
11042 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11043 	pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11044 
11045 	/* Add a filter to drop all Flow control frames from any VSI from being
11046 	 * transmitted. By doing so we stop a malicious VF from sending out
11047 	 * PAUSE or PFC frames and potentially controlling traffic for other
11048 	 * PF/VF VSIs.
11049 	 * The FW can still send Flow control frames if enabled.
11050 	 */
11051 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11052 						       pf->main_vsi_seid);
11053 
11054 	/* print a string summarizing features */
11055 	i40e_print_features(pf);
11056 
11057 	return 0;
11058 
11059 	/* Unwind what we've done if something failed in the setup */
11060 err_vsis:
11061 	set_bit(__I40E_DOWN, &pf->state);
11062 	i40e_clear_interrupt_scheme(pf);
11063 	kfree(pf->vsi);
11064 err_switch_setup:
11065 	i40e_reset_interrupt_capability(pf);
11066 	del_timer_sync(&pf->service_timer);
11067 err_mac_addr:
11068 err_configure_lan_hmc:
11069 	(void)i40e_shutdown_lan_hmc(hw);
11070 err_init_lan_hmc:
11071 	kfree(pf->qp_pile);
11072 err_sw_init:
11073 err_adminq_setup:
11074 	(void)i40e_shutdown_adminq(hw);
11075 err_pf_reset:
11076 	iounmap(hw->hw_addr);
11077 err_ioremap:
11078 	kfree(pf);
11079 err_pf_alloc:
11080 	pci_disable_pcie_error_reporting(pdev);
11081 	pci_release_selected_regions(pdev,
11082 				     pci_select_bars(pdev, IORESOURCE_MEM));
11083 err_pci_reg:
11084 err_dma:
11085 	pci_disable_device(pdev);
11086 	return err;
11087 }
11088 
11089 /**
11090  * i40e_remove - Device removal routine
11091  * @pdev: PCI device information struct
11092  *
11093  * i40e_remove is called by the PCI subsystem to alert the driver
11094  * that is should release a PCI device.  This could be caused by a
11095  * Hot-Plug event, or because the driver is going to be removed from
11096  * memory.
11097  **/
11098 static void i40e_remove(struct pci_dev *pdev)
11099 {
11100 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11101 	struct i40e_hw *hw = &pf->hw;
11102 	i40e_status ret_code;
11103 	int i;
11104 
11105 	i40e_dbg_pf_exit(pf);
11106 
11107 	i40e_ptp_stop(pf);
11108 
11109 	/* Disable RSS in hw */
11110 	wr32(hw, I40E_PFQF_HENA(0), 0);
11111 	wr32(hw, I40E_PFQF_HENA(1), 0);
11112 
11113 	/* no more scheduling of any task */
11114 	set_bit(__I40E_DOWN, &pf->state);
11115 	del_timer_sync(&pf->service_timer);
11116 	cancel_work_sync(&pf->service_task);
11117 
11118 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11119 		i40e_free_vfs(pf);
11120 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11121 	}
11122 
11123 	i40e_fdir_teardown(pf);
11124 
11125 	/* If there is a switch structure or any orphans, remove them.
11126 	 * This will leave only the PF's VSI remaining.
11127 	 */
11128 	for (i = 0; i < I40E_MAX_VEB; i++) {
11129 		if (!pf->veb[i])
11130 			continue;
11131 
11132 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11133 		    pf->veb[i]->uplink_seid == 0)
11134 			i40e_switch_branch_release(pf->veb[i]);
11135 	}
11136 
11137 	/* Now we can shutdown the PF's VSI, just before we kill
11138 	 * adminq and hmc.
11139 	 */
11140 	if (pf->vsi[pf->lan_vsi])
11141 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11142 
11143 	/* shutdown and destroy the HMC */
11144 	if (pf->hw.hmc.hmc_obj) {
11145 		ret_code = i40e_shutdown_lan_hmc(&pf->hw);
11146 		if (ret_code)
11147 			dev_warn(&pdev->dev,
11148 				 "Failed to destroy the HMC resources: %d\n",
11149 				 ret_code);
11150 	}
11151 
11152 	/* shutdown the adminq */
11153 	ret_code = i40e_shutdown_adminq(&pf->hw);
11154 	if (ret_code)
11155 		dev_warn(&pdev->dev,
11156 			 "Failed to destroy the Admin Queue resources: %d\n",
11157 			 ret_code);
11158 
11159 	/* destroy the locks only once, here */
11160 	mutex_destroy(&hw->aq.arq_mutex);
11161 	mutex_destroy(&hw->aq.asq_mutex);
11162 
11163 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11164 	i40e_clear_interrupt_scheme(pf);
11165 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11166 		if (pf->vsi[i]) {
11167 			i40e_vsi_clear_rings(pf->vsi[i]);
11168 			i40e_vsi_clear(pf->vsi[i]);
11169 			pf->vsi[i] = NULL;
11170 		}
11171 	}
11172 
11173 	for (i = 0; i < I40E_MAX_VEB; i++) {
11174 		kfree(pf->veb[i]);
11175 		pf->veb[i] = NULL;
11176 	}
11177 
11178 	kfree(pf->qp_pile);
11179 	kfree(pf->vsi);
11180 
11181 	iounmap(pf->hw.hw_addr);
11182 	kfree(pf);
11183 	pci_release_selected_regions(pdev,
11184 				     pci_select_bars(pdev, IORESOURCE_MEM));
11185 
11186 	pci_disable_pcie_error_reporting(pdev);
11187 	pci_disable_device(pdev);
11188 }
11189 
11190 /**
11191  * i40e_pci_error_detected - warning that something funky happened in PCI land
11192  * @pdev: PCI device information struct
11193  *
11194  * Called to warn that something happened and the error handling steps
11195  * are in progress.  Allows the driver to quiesce things, be ready for
11196  * remediation.
11197  **/
11198 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11199 						enum pci_channel_state error)
11200 {
11201 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11202 
11203 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11204 
11205 	/* shutdown all operations */
11206 	if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11207 		rtnl_lock();
11208 		i40e_prep_for_reset(pf);
11209 		rtnl_unlock();
11210 	}
11211 
11212 	/* Request a slot reset */
11213 	return PCI_ERS_RESULT_NEED_RESET;
11214 }
11215 
11216 /**
11217  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11218  * @pdev: PCI device information struct
11219  *
11220  * Called to find if the driver can work with the device now that
11221  * the pci slot has been reset.  If a basic connection seems good
11222  * (registers are readable and have sane content) then return a
11223  * happy little PCI_ERS_RESULT_xxx.
11224  **/
11225 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11226 {
11227 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11228 	pci_ers_result_t result;
11229 	int err;
11230 	u32 reg;
11231 
11232 	dev_dbg(&pdev->dev, "%s\n", __func__);
11233 	if (pci_enable_device_mem(pdev)) {
11234 		dev_info(&pdev->dev,
11235 			 "Cannot re-enable PCI device after reset.\n");
11236 		result = PCI_ERS_RESULT_DISCONNECT;
11237 	} else {
11238 		pci_set_master(pdev);
11239 		pci_restore_state(pdev);
11240 		pci_save_state(pdev);
11241 		pci_wake_from_d3(pdev, false);
11242 
11243 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11244 		if (reg == 0)
11245 			result = PCI_ERS_RESULT_RECOVERED;
11246 		else
11247 			result = PCI_ERS_RESULT_DISCONNECT;
11248 	}
11249 
11250 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
11251 	if (err) {
11252 		dev_info(&pdev->dev,
11253 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11254 			 err);
11255 		/* non-fatal, continue */
11256 	}
11257 
11258 	return result;
11259 }
11260 
11261 /**
11262  * i40e_pci_error_resume - restart operations after PCI error recovery
11263  * @pdev: PCI device information struct
11264  *
11265  * Called to allow the driver to bring things back up after PCI error
11266  * and/or reset recovery has finished.
11267  **/
11268 static void i40e_pci_error_resume(struct pci_dev *pdev)
11269 {
11270 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11271 
11272 	dev_dbg(&pdev->dev, "%s\n", __func__);
11273 	if (test_bit(__I40E_SUSPENDED, &pf->state))
11274 		return;
11275 
11276 	rtnl_lock();
11277 	i40e_handle_reset_warning(pf);
11278 	rtnl_unlock();
11279 }
11280 
11281 /**
11282  * i40e_shutdown - PCI callback for shutting down
11283  * @pdev: PCI device information struct
11284  **/
11285 static void i40e_shutdown(struct pci_dev *pdev)
11286 {
11287 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11288 	struct i40e_hw *hw = &pf->hw;
11289 
11290 	set_bit(__I40E_SUSPENDED, &pf->state);
11291 	set_bit(__I40E_DOWN, &pf->state);
11292 	rtnl_lock();
11293 	i40e_prep_for_reset(pf);
11294 	rtnl_unlock();
11295 
11296 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11297 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11298 
11299 	del_timer_sync(&pf->service_timer);
11300 	cancel_work_sync(&pf->service_task);
11301 	i40e_fdir_teardown(pf);
11302 
11303 	rtnl_lock();
11304 	i40e_prep_for_reset(pf);
11305 	rtnl_unlock();
11306 
11307 	wr32(hw, I40E_PFPM_APM,
11308 	     (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11309 	wr32(hw, I40E_PFPM_WUFC,
11310 	     (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11311 
11312 	i40e_clear_interrupt_scheme(pf);
11313 
11314 	if (system_state == SYSTEM_POWER_OFF) {
11315 		pci_wake_from_d3(pdev, pf->wol_en);
11316 		pci_set_power_state(pdev, PCI_D3hot);
11317 	}
11318 }
11319 
11320 #ifdef CONFIG_PM
11321 /**
11322  * i40e_suspend - PCI callback for moving to D3
11323  * @pdev: PCI device information struct
11324  **/
11325 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11326 {
11327 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11328 	struct i40e_hw *hw = &pf->hw;
11329 
11330 	set_bit(__I40E_SUSPENDED, &pf->state);
11331 	set_bit(__I40E_DOWN, &pf->state);
11332 
11333 	rtnl_lock();
11334 	i40e_prep_for_reset(pf);
11335 	rtnl_unlock();
11336 
11337 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11338 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11339 
11340 	pci_wake_from_d3(pdev, pf->wol_en);
11341 	pci_set_power_state(pdev, PCI_D3hot);
11342 
11343 	return 0;
11344 }
11345 
11346 /**
11347  * i40e_resume - PCI callback for waking up from D3
11348  * @pdev: PCI device information struct
11349  **/
11350 static int i40e_resume(struct pci_dev *pdev)
11351 {
11352 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11353 	u32 err;
11354 
11355 	pci_set_power_state(pdev, PCI_D0);
11356 	pci_restore_state(pdev);
11357 	/* pci_restore_state() clears dev->state_saves, so
11358 	 * call pci_save_state() again to restore it.
11359 	 */
11360 	pci_save_state(pdev);
11361 
11362 	err = pci_enable_device_mem(pdev);
11363 	if (err) {
11364 		dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11365 		return err;
11366 	}
11367 	pci_set_master(pdev);
11368 
11369 	/* no wakeup events while running */
11370 	pci_wake_from_d3(pdev, false);
11371 
11372 	/* handling the reset will rebuild the device state */
11373 	if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11374 		clear_bit(__I40E_DOWN, &pf->state);
11375 		rtnl_lock();
11376 		i40e_reset_and_rebuild(pf, false);
11377 		rtnl_unlock();
11378 	}
11379 
11380 	return 0;
11381 }
11382 
11383 #endif
11384 static const struct pci_error_handlers i40e_err_handler = {
11385 	.error_detected = i40e_pci_error_detected,
11386 	.slot_reset = i40e_pci_error_slot_reset,
11387 	.resume = i40e_pci_error_resume,
11388 };
11389 
11390 static struct pci_driver i40e_driver = {
11391 	.name     = i40e_driver_name,
11392 	.id_table = i40e_pci_tbl,
11393 	.probe    = i40e_probe,
11394 	.remove   = i40e_remove,
11395 #ifdef CONFIG_PM
11396 	.suspend  = i40e_suspend,
11397 	.resume   = i40e_resume,
11398 #endif
11399 	.shutdown = i40e_shutdown,
11400 	.err_handler = &i40e_err_handler,
11401 	.sriov_configure = i40e_pci_sriov_configure,
11402 };
11403 
11404 /**
11405  * i40e_init_module - Driver registration routine
11406  *
11407  * i40e_init_module is the first routine called when the driver is
11408  * loaded. All it does is register with the PCI subsystem.
11409  **/
11410 static int __init i40e_init_module(void)
11411 {
11412 	pr_info("%s: %s - version %s\n", i40e_driver_name,
11413 		i40e_driver_string, i40e_driver_version_str);
11414 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11415 
11416 	i40e_dbg_init();
11417 	return pci_register_driver(&i40e_driver);
11418 }
11419 module_init(i40e_init_module);
11420 
11421 /**
11422  * i40e_exit_module - Driver exit cleanup routine
11423  *
11424  * i40e_exit_module is called just before the driver is removed
11425  * from memory.
11426  **/
11427 static void __exit i40e_exit_module(void)
11428 {
11429 	pci_unregister_driver(&i40e_driver);
11430 	i40e_dbg_exit();
11431 }
11432 module_exit(i40e_exit_module);
11433