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 /* Local includes */
28 #include "i40e.h"
29 #include "i40e_diag.h"
30 #ifdef CONFIG_I40E_VXLAN
31 #include <net/vxlan.h>
32 #endif
33 
34 const char i40e_driver_name[] = "i40e";
35 static const char i40e_driver_string[] =
36 			"Intel(R) Ethernet Connection XL710 Network Driver";
37 
38 #define DRV_KERN "-k"
39 
40 #define DRV_VERSION_MAJOR 1
41 #define DRV_VERSION_MINOR 3
42 #define DRV_VERSION_BUILD 2
43 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
44 	     __stringify(DRV_VERSION_MINOR) "." \
45 	     __stringify(DRV_VERSION_BUILD)    DRV_KERN
46 const char i40e_driver_version_str[] = DRV_VERSION;
47 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
48 
49 /* a bit of forward declarations */
50 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
51 static void i40e_handle_reset_warning(struct i40e_pf *pf);
52 static int i40e_add_vsi(struct i40e_vsi *vsi);
53 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
54 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
55 static int i40e_setup_misc_vector(struct i40e_pf *pf);
56 static void i40e_determine_queue_usage(struct i40e_pf *pf);
57 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
58 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
59 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
60 
61 /* i40e_pci_tbl - PCI Device ID Table
62  *
63  * Last entry must be all 0s
64  *
65  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
66  *   Class, Class Mask, private data (not used) }
67  */
68 static const struct pci_device_id i40e_pci_tbl[] = {
69 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
70 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
71 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_A), 0},
72 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
73 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
74 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
75 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
76 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
77 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
78 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
79 	/* required last entry */
80 	{0, }
81 };
82 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
83 
84 #define I40E_MAX_VF_COUNT 128
85 static int debug = -1;
86 module_param(debug, int, 0);
87 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
88 
89 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
90 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
91 MODULE_LICENSE("GPL");
92 MODULE_VERSION(DRV_VERSION);
93 
94 /**
95  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
96  * @hw:   pointer to the HW structure
97  * @mem:  ptr to mem struct to fill out
98  * @size: size of memory requested
99  * @alignment: what to align the allocation to
100  **/
101 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
102 			    u64 size, u32 alignment)
103 {
104 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
105 
106 	mem->size = ALIGN(size, alignment);
107 	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
108 				      &mem->pa, GFP_KERNEL);
109 	if (!mem->va)
110 		return -ENOMEM;
111 
112 	return 0;
113 }
114 
115 /**
116  * i40e_free_dma_mem_d - OS specific memory free for shared code
117  * @hw:   pointer to the HW structure
118  * @mem:  ptr to mem struct to free
119  **/
120 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
121 {
122 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
123 
124 	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
125 	mem->va = NULL;
126 	mem->pa = 0;
127 	mem->size = 0;
128 
129 	return 0;
130 }
131 
132 /**
133  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
134  * @hw:   pointer to the HW structure
135  * @mem:  ptr to mem struct to fill out
136  * @size: size of memory requested
137  **/
138 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
139 			     u32 size)
140 {
141 	mem->size = size;
142 	mem->va = kzalloc(size, GFP_KERNEL);
143 
144 	if (!mem->va)
145 		return -ENOMEM;
146 
147 	return 0;
148 }
149 
150 /**
151  * i40e_free_virt_mem_d - OS specific memory free for shared code
152  * @hw:   pointer to the HW structure
153  * @mem:  ptr to mem struct to free
154  **/
155 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
156 {
157 	/* it's ok to kfree a NULL pointer */
158 	kfree(mem->va);
159 	mem->va = NULL;
160 	mem->size = 0;
161 
162 	return 0;
163 }
164 
165 /**
166  * i40e_get_lump - find a lump of free generic resource
167  * @pf: board private structure
168  * @pile: the pile of resource to search
169  * @needed: the number of items needed
170  * @id: an owner id to stick on the items assigned
171  *
172  * Returns the base item index of the lump, or negative for error
173  *
174  * The search_hint trick and lack of advanced fit-finding only work
175  * because we're highly likely to have all the same size lump requests.
176  * Linear search time and any fragmentation should be minimal.
177  **/
178 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
179 			 u16 needed, u16 id)
180 {
181 	int ret = -ENOMEM;
182 	int i, j;
183 
184 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
185 		dev_info(&pf->pdev->dev,
186 			 "param err: pile=%p needed=%d id=0x%04x\n",
187 			 pile, needed, id);
188 		return -EINVAL;
189 	}
190 
191 	/* start the linear search with an imperfect hint */
192 	i = pile->search_hint;
193 	while (i < pile->num_entries) {
194 		/* skip already allocated entries */
195 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
196 			i++;
197 			continue;
198 		}
199 
200 		/* do we have enough in this lump? */
201 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
202 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
203 				break;
204 		}
205 
206 		if (j == needed) {
207 			/* there was enough, so assign it to the requestor */
208 			for (j = 0; j < needed; j++)
209 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
210 			ret = i;
211 			pile->search_hint = i + j;
212 			break;
213 		} else {
214 			/* not enough, so skip over it and continue looking */
215 			i += j;
216 		}
217 	}
218 
219 	return ret;
220 }
221 
222 /**
223  * i40e_put_lump - return a lump of generic resource
224  * @pile: the pile of resource to search
225  * @index: the base item index
226  * @id: the owner id of the items assigned
227  *
228  * Returns the count of items in the lump
229  **/
230 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
231 {
232 	int valid_id = (id | I40E_PILE_VALID_BIT);
233 	int count = 0;
234 	int i;
235 
236 	if (!pile || index >= pile->num_entries)
237 		return -EINVAL;
238 
239 	for (i = index;
240 	     i < pile->num_entries && pile->list[i] == valid_id;
241 	     i++) {
242 		pile->list[i] = 0;
243 		count++;
244 	}
245 
246 	if (count && index < pile->search_hint)
247 		pile->search_hint = index;
248 
249 	return count;
250 }
251 
252 /**
253  * i40e_find_vsi_from_id - searches for the vsi with the given id
254  * @pf - the pf structure to search for the vsi
255  * @id - id of the vsi it is searching for
256  **/
257 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
258 {
259 	int i;
260 
261 	for (i = 0; i < pf->num_alloc_vsi; i++)
262 		if (pf->vsi[i] && (pf->vsi[i]->id == id))
263 			return pf->vsi[i];
264 
265 	return NULL;
266 }
267 
268 /**
269  * i40e_service_event_schedule - Schedule the service task to wake up
270  * @pf: board private structure
271  *
272  * If not already scheduled, this puts the task into the work queue
273  **/
274 static void i40e_service_event_schedule(struct i40e_pf *pf)
275 {
276 	if (!test_bit(__I40E_DOWN, &pf->state) &&
277 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
278 	    !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
279 		schedule_work(&pf->service_task);
280 }
281 
282 /**
283  * i40e_tx_timeout - Respond to a Tx Hang
284  * @netdev: network interface device structure
285  *
286  * If any port has noticed a Tx timeout, it is likely that the whole
287  * device is munged, not just the one netdev port, so go for the full
288  * reset.
289  **/
290 #ifdef I40E_FCOE
291 void i40e_tx_timeout(struct net_device *netdev)
292 #else
293 static void i40e_tx_timeout(struct net_device *netdev)
294 #endif
295 {
296 	struct i40e_netdev_priv *np = netdev_priv(netdev);
297 	struct i40e_vsi *vsi = np->vsi;
298 	struct i40e_pf *pf = vsi->back;
299 
300 	pf->tx_timeout_count++;
301 
302 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
303 		pf->tx_timeout_recovery_level = 1;
304 	pf->tx_timeout_last_recovery = jiffies;
305 	netdev_info(netdev, "tx_timeout recovery level %d\n",
306 		    pf->tx_timeout_recovery_level);
307 
308 	switch (pf->tx_timeout_recovery_level) {
309 	case 0:
310 		/* disable and re-enable queues for the VSI */
311 		if (in_interrupt()) {
312 			set_bit(__I40E_REINIT_REQUESTED, &pf->state);
313 			set_bit(__I40E_REINIT_REQUESTED, &vsi->state);
314 		} else {
315 			i40e_vsi_reinit_locked(vsi);
316 		}
317 		break;
318 	case 1:
319 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
320 		break;
321 	case 2:
322 		set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
323 		break;
324 	case 3:
325 		set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
326 		break;
327 	default:
328 		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
329 		set_bit(__I40E_DOWN_REQUESTED, &pf->state);
330 		set_bit(__I40E_DOWN_REQUESTED, &vsi->state);
331 		break;
332 	}
333 	i40e_service_event_schedule(pf);
334 	pf->tx_timeout_recovery_level++;
335 }
336 
337 /**
338  * i40e_release_rx_desc - Store the new tail and head values
339  * @rx_ring: ring to bump
340  * @val: new head index
341  **/
342 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
343 {
344 	rx_ring->next_to_use = val;
345 
346 	/* Force memory writes to complete before letting h/w
347 	 * know there are new descriptors to fetch.  (Only
348 	 * applicable for weak-ordered memory model archs,
349 	 * such as IA-64).
350 	 */
351 	wmb();
352 	writel(val, rx_ring->tail);
353 }
354 
355 /**
356  * i40e_get_vsi_stats_struct - Get System Network Statistics
357  * @vsi: the VSI we care about
358  *
359  * Returns the address of the device statistics structure.
360  * The statistics are actually updated from the service task.
361  **/
362 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
363 {
364 	return &vsi->net_stats;
365 }
366 
367 /**
368  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
369  * @netdev: network interface device structure
370  *
371  * Returns the address of the device statistics structure.
372  * The statistics are actually updated from the service task.
373  **/
374 #ifdef I40E_FCOE
375 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
376 					     struct net_device *netdev,
377 					     struct rtnl_link_stats64 *stats)
378 #else
379 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
380 					     struct net_device *netdev,
381 					     struct rtnl_link_stats64 *stats)
382 #endif
383 {
384 	struct i40e_netdev_priv *np = netdev_priv(netdev);
385 	struct i40e_ring *tx_ring, *rx_ring;
386 	struct i40e_vsi *vsi = np->vsi;
387 	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
388 	int i;
389 
390 	if (test_bit(__I40E_DOWN, &vsi->state))
391 		return stats;
392 
393 	if (!vsi->tx_rings)
394 		return stats;
395 
396 	rcu_read_lock();
397 	for (i = 0; i < vsi->num_queue_pairs; i++) {
398 		u64 bytes, packets;
399 		unsigned int start;
400 
401 		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
402 		if (!tx_ring)
403 			continue;
404 
405 		do {
406 			start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
407 			packets = tx_ring->stats.packets;
408 			bytes   = tx_ring->stats.bytes;
409 		} while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
410 
411 		stats->tx_packets += packets;
412 		stats->tx_bytes   += bytes;
413 		rx_ring = &tx_ring[1];
414 
415 		do {
416 			start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
417 			packets = rx_ring->stats.packets;
418 			bytes   = rx_ring->stats.bytes;
419 		} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
420 
421 		stats->rx_packets += packets;
422 		stats->rx_bytes   += bytes;
423 	}
424 	rcu_read_unlock();
425 
426 	/* following stats updated by i40e_watchdog_subtask() */
427 	stats->multicast	= vsi_stats->multicast;
428 	stats->tx_errors	= vsi_stats->tx_errors;
429 	stats->tx_dropped	= vsi_stats->tx_dropped;
430 	stats->rx_errors	= vsi_stats->rx_errors;
431 	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
432 	stats->rx_length_errors	= vsi_stats->rx_length_errors;
433 
434 	return stats;
435 }
436 
437 /**
438  * i40e_vsi_reset_stats - Resets all stats of the given vsi
439  * @vsi: the VSI to have its stats reset
440  **/
441 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
442 {
443 	struct rtnl_link_stats64 *ns;
444 	int i;
445 
446 	if (!vsi)
447 		return;
448 
449 	ns = i40e_get_vsi_stats_struct(vsi);
450 	memset(ns, 0, sizeof(*ns));
451 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
452 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
453 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
454 	if (vsi->rx_rings && vsi->rx_rings[0]) {
455 		for (i = 0; i < vsi->num_queue_pairs; i++) {
456 			memset(&vsi->rx_rings[i]->stats, 0 ,
457 			       sizeof(vsi->rx_rings[i]->stats));
458 			memset(&vsi->rx_rings[i]->rx_stats, 0 ,
459 			       sizeof(vsi->rx_rings[i]->rx_stats));
460 			memset(&vsi->tx_rings[i]->stats, 0 ,
461 			       sizeof(vsi->tx_rings[i]->stats));
462 			memset(&vsi->tx_rings[i]->tx_stats, 0,
463 			       sizeof(vsi->tx_rings[i]->tx_stats));
464 		}
465 	}
466 	vsi->stat_offsets_loaded = false;
467 }
468 
469 /**
470  * i40e_pf_reset_stats - Reset all of the stats for the given PF
471  * @pf: the PF to be reset
472  **/
473 void i40e_pf_reset_stats(struct i40e_pf *pf)
474 {
475 	int i;
476 
477 	memset(&pf->stats, 0, sizeof(pf->stats));
478 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
479 	pf->stat_offsets_loaded = false;
480 
481 	for (i = 0; i < I40E_MAX_VEB; i++) {
482 		if (pf->veb[i]) {
483 			memset(&pf->veb[i]->stats, 0,
484 			       sizeof(pf->veb[i]->stats));
485 			memset(&pf->veb[i]->stats_offsets, 0,
486 			       sizeof(pf->veb[i]->stats_offsets));
487 			pf->veb[i]->stat_offsets_loaded = false;
488 		}
489 	}
490 }
491 
492 /**
493  * i40e_stat_update48 - read and update a 48 bit stat from the chip
494  * @hw: ptr to the hardware info
495  * @hireg: the high 32 bit reg to read
496  * @loreg: the low 32 bit reg to read
497  * @offset_loaded: has the initial offset been loaded yet
498  * @offset: ptr to current offset value
499  * @stat: ptr to the stat
500  *
501  * Since the device stats are not reset at PFReset, they likely will not
502  * be zeroed when the driver starts.  We'll save the first values read
503  * and use them as offsets to be subtracted from the raw values in order
504  * to report stats that count from zero.  In the process, we also manage
505  * the potential roll-over.
506  **/
507 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
508 			       bool offset_loaded, u64 *offset, u64 *stat)
509 {
510 	u64 new_data;
511 
512 	if (hw->device_id == I40E_DEV_ID_QEMU) {
513 		new_data = rd32(hw, loreg);
514 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
515 	} else {
516 		new_data = rd64(hw, loreg);
517 	}
518 	if (!offset_loaded)
519 		*offset = new_data;
520 	if (likely(new_data >= *offset))
521 		*stat = new_data - *offset;
522 	else
523 		*stat = (new_data + ((u64)1 << 48)) - *offset;
524 	*stat &= 0xFFFFFFFFFFFFULL;
525 }
526 
527 /**
528  * i40e_stat_update32 - read and update a 32 bit stat from the chip
529  * @hw: ptr to the hardware info
530  * @reg: the hw reg to read
531  * @offset_loaded: has the initial offset been loaded yet
532  * @offset: ptr to current offset value
533  * @stat: ptr to the stat
534  **/
535 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
536 			       bool offset_loaded, u64 *offset, u64 *stat)
537 {
538 	u32 new_data;
539 
540 	new_data = rd32(hw, reg);
541 	if (!offset_loaded)
542 		*offset = new_data;
543 	if (likely(new_data >= *offset))
544 		*stat = (u32)(new_data - *offset);
545 	else
546 		*stat = (u32)((new_data + ((u64)1 << 32)) - *offset);
547 }
548 
549 /**
550  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
551  * @vsi: the VSI to be updated
552  **/
553 void i40e_update_eth_stats(struct i40e_vsi *vsi)
554 {
555 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
556 	struct i40e_pf *pf = vsi->back;
557 	struct i40e_hw *hw = &pf->hw;
558 	struct i40e_eth_stats *oes;
559 	struct i40e_eth_stats *es;     /* device's eth stats */
560 
561 	es = &vsi->eth_stats;
562 	oes = &vsi->eth_stats_offsets;
563 
564 	/* Gather up the stats that the hw collects */
565 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
566 			   vsi->stat_offsets_loaded,
567 			   &oes->tx_errors, &es->tx_errors);
568 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
569 			   vsi->stat_offsets_loaded,
570 			   &oes->rx_discards, &es->rx_discards);
571 	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
572 			   vsi->stat_offsets_loaded,
573 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
574 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
575 			   vsi->stat_offsets_loaded,
576 			   &oes->tx_errors, &es->tx_errors);
577 
578 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
579 			   I40E_GLV_GORCL(stat_idx),
580 			   vsi->stat_offsets_loaded,
581 			   &oes->rx_bytes, &es->rx_bytes);
582 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
583 			   I40E_GLV_UPRCL(stat_idx),
584 			   vsi->stat_offsets_loaded,
585 			   &oes->rx_unicast, &es->rx_unicast);
586 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
587 			   I40E_GLV_MPRCL(stat_idx),
588 			   vsi->stat_offsets_loaded,
589 			   &oes->rx_multicast, &es->rx_multicast);
590 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
591 			   I40E_GLV_BPRCL(stat_idx),
592 			   vsi->stat_offsets_loaded,
593 			   &oes->rx_broadcast, &es->rx_broadcast);
594 
595 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
596 			   I40E_GLV_GOTCL(stat_idx),
597 			   vsi->stat_offsets_loaded,
598 			   &oes->tx_bytes, &es->tx_bytes);
599 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
600 			   I40E_GLV_UPTCL(stat_idx),
601 			   vsi->stat_offsets_loaded,
602 			   &oes->tx_unicast, &es->tx_unicast);
603 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
604 			   I40E_GLV_MPTCL(stat_idx),
605 			   vsi->stat_offsets_loaded,
606 			   &oes->tx_multicast, &es->tx_multicast);
607 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
608 			   I40E_GLV_BPTCL(stat_idx),
609 			   vsi->stat_offsets_loaded,
610 			   &oes->tx_broadcast, &es->tx_broadcast);
611 	vsi->stat_offsets_loaded = true;
612 }
613 
614 /**
615  * i40e_update_veb_stats - Update Switch component statistics
616  * @veb: the VEB being updated
617  **/
618 static void i40e_update_veb_stats(struct i40e_veb *veb)
619 {
620 	struct i40e_pf *pf = veb->pf;
621 	struct i40e_hw *hw = &pf->hw;
622 	struct i40e_eth_stats *oes;
623 	struct i40e_eth_stats *es;     /* device's eth stats */
624 	int idx = 0;
625 
626 	idx = veb->stats_idx;
627 	es = &veb->stats;
628 	oes = &veb->stats_offsets;
629 
630 	/* Gather up the stats that the hw collects */
631 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
632 			   veb->stat_offsets_loaded,
633 			   &oes->tx_discards, &es->tx_discards);
634 	if (hw->revision_id > 0)
635 		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
636 				   veb->stat_offsets_loaded,
637 				   &oes->rx_unknown_protocol,
638 				   &es->rx_unknown_protocol);
639 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
640 			   veb->stat_offsets_loaded,
641 			   &oes->rx_bytes, &es->rx_bytes);
642 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
643 			   veb->stat_offsets_loaded,
644 			   &oes->rx_unicast, &es->rx_unicast);
645 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
646 			   veb->stat_offsets_loaded,
647 			   &oes->rx_multicast, &es->rx_multicast);
648 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
649 			   veb->stat_offsets_loaded,
650 			   &oes->rx_broadcast, &es->rx_broadcast);
651 
652 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
653 			   veb->stat_offsets_loaded,
654 			   &oes->tx_bytes, &es->tx_bytes);
655 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
656 			   veb->stat_offsets_loaded,
657 			   &oes->tx_unicast, &es->tx_unicast);
658 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
659 			   veb->stat_offsets_loaded,
660 			   &oes->tx_multicast, &es->tx_multicast);
661 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
662 			   veb->stat_offsets_loaded,
663 			   &oes->tx_broadcast, &es->tx_broadcast);
664 	veb->stat_offsets_loaded = true;
665 }
666 
667 #ifdef I40E_FCOE
668 /**
669  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
670  * @vsi: the VSI that is capable of doing FCoE
671  **/
672 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
673 {
674 	struct i40e_pf *pf = vsi->back;
675 	struct i40e_hw *hw = &pf->hw;
676 	struct i40e_fcoe_stats *ofs;
677 	struct i40e_fcoe_stats *fs;     /* device's eth stats */
678 	int idx;
679 
680 	if (vsi->type != I40E_VSI_FCOE)
681 		return;
682 
683 	idx = (pf->pf_seid - I40E_BASE_PF_SEID) + I40E_FCOE_PF_STAT_OFFSET;
684 	fs = &vsi->fcoe_stats;
685 	ofs = &vsi->fcoe_stats_offsets;
686 
687 	i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
688 			   vsi->fcoe_stat_offsets_loaded,
689 			   &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
690 	i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
691 			   vsi->fcoe_stat_offsets_loaded,
692 			   &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
693 	i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
694 			   vsi->fcoe_stat_offsets_loaded,
695 			   &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
696 	i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
697 			   vsi->fcoe_stat_offsets_loaded,
698 			   &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
699 	i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
700 			   vsi->fcoe_stat_offsets_loaded,
701 			   &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
702 	i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
703 			   vsi->fcoe_stat_offsets_loaded,
704 			   &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
705 	i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
706 			   vsi->fcoe_stat_offsets_loaded,
707 			   &ofs->fcoe_last_error, &fs->fcoe_last_error);
708 	i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
709 			   vsi->fcoe_stat_offsets_loaded,
710 			   &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
711 
712 	vsi->fcoe_stat_offsets_loaded = true;
713 }
714 
715 #endif
716 /**
717  * i40e_update_link_xoff_rx - Update XOFF received in link flow control mode
718  * @pf: the corresponding PF
719  *
720  * Update the Rx XOFF counter (PAUSE frames) in link flow control mode
721  **/
722 static void i40e_update_link_xoff_rx(struct i40e_pf *pf)
723 {
724 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
725 	struct i40e_hw_port_stats *nsd = &pf->stats;
726 	struct i40e_hw *hw = &pf->hw;
727 	u64 xoff = 0;
728 	u16 i, v;
729 
730 	if ((hw->fc.current_mode != I40E_FC_FULL) &&
731 	    (hw->fc.current_mode != I40E_FC_RX_PAUSE))
732 		return;
733 
734 	xoff = nsd->link_xoff_rx;
735 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
736 			   pf->stat_offsets_loaded,
737 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
738 
739 	/* No new LFC xoff rx */
740 	if (!(nsd->link_xoff_rx - xoff))
741 		return;
742 
743 	/* Clear the __I40E_HANG_CHECK_ARMED bit for all Tx rings */
744 	for (v = 0; v < pf->num_alloc_vsi; v++) {
745 		struct i40e_vsi *vsi = pf->vsi[v];
746 
747 		if (!vsi || !vsi->tx_rings[0])
748 			continue;
749 
750 		for (i = 0; i < vsi->num_queue_pairs; i++) {
751 			struct i40e_ring *ring = vsi->tx_rings[i];
752 			clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
753 		}
754 	}
755 }
756 
757 /**
758  * i40e_update_prio_xoff_rx - Update XOFF received in PFC mode
759  * @pf: the corresponding PF
760  *
761  * Update the Rx XOFF counter (PAUSE frames) in PFC mode
762  **/
763 static void i40e_update_prio_xoff_rx(struct i40e_pf *pf)
764 {
765 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
766 	struct i40e_hw_port_stats *nsd = &pf->stats;
767 	bool xoff[I40E_MAX_TRAFFIC_CLASS] = {false};
768 	struct i40e_dcbx_config *dcb_cfg;
769 	struct i40e_hw *hw = &pf->hw;
770 	u16 i, v;
771 	u8 tc;
772 
773 	dcb_cfg = &hw->local_dcbx_config;
774 
775 	/* See if DCB enabled with PFC TC */
776 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED) ||
777 	    !(dcb_cfg->pfc.pfcenable)) {
778 		i40e_update_link_xoff_rx(pf);
779 		return;
780 	}
781 
782 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
783 		u64 prio_xoff = nsd->priority_xoff_rx[i];
784 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
785 				   pf->stat_offsets_loaded,
786 				   &osd->priority_xoff_rx[i],
787 				   &nsd->priority_xoff_rx[i]);
788 
789 		/* No new PFC xoff rx */
790 		if (!(nsd->priority_xoff_rx[i] - prio_xoff))
791 			continue;
792 		/* Get the TC for given priority */
793 		tc = dcb_cfg->etscfg.prioritytable[i];
794 		xoff[tc] = true;
795 	}
796 
797 	/* Clear the __I40E_HANG_CHECK_ARMED bit for Tx rings */
798 	for (v = 0; v < pf->num_alloc_vsi; v++) {
799 		struct i40e_vsi *vsi = pf->vsi[v];
800 
801 		if (!vsi || !vsi->tx_rings[0])
802 			continue;
803 
804 		for (i = 0; i < vsi->num_queue_pairs; i++) {
805 			struct i40e_ring *ring = vsi->tx_rings[i];
806 
807 			tc = ring->dcb_tc;
808 			if (xoff[tc])
809 				clear_bit(__I40E_HANG_CHECK_ARMED,
810 					  &ring->state);
811 		}
812 	}
813 }
814 
815 /**
816  * i40e_update_vsi_stats - Update the vsi statistics counters.
817  * @vsi: the VSI to be updated
818  *
819  * There are a few instances where we store the same stat in a
820  * couple of different structs.  This is partly because we have
821  * the netdev stats that need to be filled out, which is slightly
822  * different from the "eth_stats" defined by the chip and used in
823  * VF communications.  We sort it out here.
824  **/
825 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
826 {
827 	struct i40e_pf *pf = vsi->back;
828 	struct rtnl_link_stats64 *ons;
829 	struct rtnl_link_stats64 *ns;   /* netdev stats */
830 	struct i40e_eth_stats *oes;
831 	struct i40e_eth_stats *es;     /* device's eth stats */
832 	u32 tx_restart, tx_busy;
833 	struct i40e_ring *p;
834 	u32 rx_page, rx_buf;
835 	u64 bytes, packets;
836 	unsigned int start;
837 	u64 rx_p, rx_b;
838 	u64 tx_p, tx_b;
839 	u16 q;
840 
841 	if (test_bit(__I40E_DOWN, &vsi->state) ||
842 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
843 		return;
844 
845 	ns = i40e_get_vsi_stats_struct(vsi);
846 	ons = &vsi->net_stats_offsets;
847 	es = &vsi->eth_stats;
848 	oes = &vsi->eth_stats_offsets;
849 
850 	/* Gather up the netdev and vsi stats that the driver collects
851 	 * on the fly during packet processing
852 	 */
853 	rx_b = rx_p = 0;
854 	tx_b = tx_p = 0;
855 	tx_restart = tx_busy = 0;
856 	rx_page = 0;
857 	rx_buf = 0;
858 	rcu_read_lock();
859 	for (q = 0; q < vsi->num_queue_pairs; q++) {
860 		/* locate Tx ring */
861 		p = ACCESS_ONCE(vsi->tx_rings[q]);
862 
863 		do {
864 			start = u64_stats_fetch_begin_irq(&p->syncp);
865 			packets = p->stats.packets;
866 			bytes = p->stats.bytes;
867 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
868 		tx_b += bytes;
869 		tx_p += packets;
870 		tx_restart += p->tx_stats.restart_queue;
871 		tx_busy += p->tx_stats.tx_busy;
872 
873 		/* Rx queue is part of the same block as Tx queue */
874 		p = &p[1];
875 		do {
876 			start = u64_stats_fetch_begin_irq(&p->syncp);
877 			packets = p->stats.packets;
878 			bytes = p->stats.bytes;
879 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
880 		rx_b += bytes;
881 		rx_p += packets;
882 		rx_buf += p->rx_stats.alloc_buff_failed;
883 		rx_page += p->rx_stats.alloc_page_failed;
884 	}
885 	rcu_read_unlock();
886 	vsi->tx_restart = tx_restart;
887 	vsi->tx_busy = tx_busy;
888 	vsi->rx_page_failed = rx_page;
889 	vsi->rx_buf_failed = rx_buf;
890 
891 	ns->rx_packets = rx_p;
892 	ns->rx_bytes = rx_b;
893 	ns->tx_packets = tx_p;
894 	ns->tx_bytes = tx_b;
895 
896 	/* update netdev stats from eth stats */
897 	i40e_update_eth_stats(vsi);
898 	ons->tx_errors = oes->tx_errors;
899 	ns->tx_errors = es->tx_errors;
900 	ons->multicast = oes->rx_multicast;
901 	ns->multicast = es->rx_multicast;
902 	ons->rx_dropped = oes->rx_discards;
903 	ns->rx_dropped = es->rx_discards;
904 	ons->tx_dropped = oes->tx_discards;
905 	ns->tx_dropped = es->tx_discards;
906 
907 	/* pull in a couple PF stats if this is the main vsi */
908 	if (vsi == pf->vsi[pf->lan_vsi]) {
909 		ns->rx_crc_errors = pf->stats.crc_errors;
910 		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
911 		ns->rx_length_errors = pf->stats.rx_length_errors;
912 	}
913 }
914 
915 /**
916  * i40e_update_pf_stats - Update the PF statistics counters.
917  * @pf: the PF to be updated
918  **/
919 static void i40e_update_pf_stats(struct i40e_pf *pf)
920 {
921 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
922 	struct i40e_hw_port_stats *nsd = &pf->stats;
923 	struct i40e_hw *hw = &pf->hw;
924 	u32 val;
925 	int i;
926 
927 	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
928 			   I40E_GLPRT_GORCL(hw->port),
929 			   pf->stat_offsets_loaded,
930 			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
931 	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
932 			   I40E_GLPRT_GOTCL(hw->port),
933 			   pf->stat_offsets_loaded,
934 			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
935 	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
936 			   pf->stat_offsets_loaded,
937 			   &osd->eth.rx_discards,
938 			   &nsd->eth.rx_discards);
939 	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
940 			   I40E_GLPRT_UPRCL(hw->port),
941 			   pf->stat_offsets_loaded,
942 			   &osd->eth.rx_unicast,
943 			   &nsd->eth.rx_unicast);
944 	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
945 			   I40E_GLPRT_MPRCL(hw->port),
946 			   pf->stat_offsets_loaded,
947 			   &osd->eth.rx_multicast,
948 			   &nsd->eth.rx_multicast);
949 	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
950 			   I40E_GLPRT_BPRCL(hw->port),
951 			   pf->stat_offsets_loaded,
952 			   &osd->eth.rx_broadcast,
953 			   &nsd->eth.rx_broadcast);
954 	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
955 			   I40E_GLPRT_UPTCL(hw->port),
956 			   pf->stat_offsets_loaded,
957 			   &osd->eth.tx_unicast,
958 			   &nsd->eth.tx_unicast);
959 	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
960 			   I40E_GLPRT_MPTCL(hw->port),
961 			   pf->stat_offsets_loaded,
962 			   &osd->eth.tx_multicast,
963 			   &nsd->eth.tx_multicast);
964 	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
965 			   I40E_GLPRT_BPTCL(hw->port),
966 			   pf->stat_offsets_loaded,
967 			   &osd->eth.tx_broadcast,
968 			   &nsd->eth.tx_broadcast);
969 
970 	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
971 			   pf->stat_offsets_loaded,
972 			   &osd->tx_dropped_link_down,
973 			   &nsd->tx_dropped_link_down);
974 
975 	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
976 			   pf->stat_offsets_loaded,
977 			   &osd->crc_errors, &nsd->crc_errors);
978 
979 	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
980 			   pf->stat_offsets_loaded,
981 			   &osd->illegal_bytes, &nsd->illegal_bytes);
982 
983 	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
984 			   pf->stat_offsets_loaded,
985 			   &osd->mac_local_faults,
986 			   &nsd->mac_local_faults);
987 	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
988 			   pf->stat_offsets_loaded,
989 			   &osd->mac_remote_faults,
990 			   &nsd->mac_remote_faults);
991 
992 	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
993 			   pf->stat_offsets_loaded,
994 			   &osd->rx_length_errors,
995 			   &nsd->rx_length_errors);
996 
997 	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
998 			   pf->stat_offsets_loaded,
999 			   &osd->link_xon_rx, &nsd->link_xon_rx);
1000 	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
1001 			   pf->stat_offsets_loaded,
1002 			   &osd->link_xon_tx, &nsd->link_xon_tx);
1003 	i40e_update_prio_xoff_rx(pf);  /* handles I40E_GLPRT_LXOFFRXC */
1004 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
1005 			   pf->stat_offsets_loaded,
1006 			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
1007 
1008 	for (i = 0; i < 8; i++) {
1009 		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
1010 				   pf->stat_offsets_loaded,
1011 				   &osd->priority_xon_rx[i],
1012 				   &nsd->priority_xon_rx[i]);
1013 		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1014 				   pf->stat_offsets_loaded,
1015 				   &osd->priority_xon_tx[i],
1016 				   &nsd->priority_xon_tx[i]);
1017 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1018 				   pf->stat_offsets_loaded,
1019 				   &osd->priority_xoff_tx[i],
1020 				   &nsd->priority_xoff_tx[i]);
1021 		i40e_stat_update32(hw,
1022 				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1023 				   pf->stat_offsets_loaded,
1024 				   &osd->priority_xon_2_xoff[i],
1025 				   &nsd->priority_xon_2_xoff[i]);
1026 	}
1027 
1028 	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1029 			   I40E_GLPRT_PRC64L(hw->port),
1030 			   pf->stat_offsets_loaded,
1031 			   &osd->rx_size_64, &nsd->rx_size_64);
1032 	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1033 			   I40E_GLPRT_PRC127L(hw->port),
1034 			   pf->stat_offsets_loaded,
1035 			   &osd->rx_size_127, &nsd->rx_size_127);
1036 	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1037 			   I40E_GLPRT_PRC255L(hw->port),
1038 			   pf->stat_offsets_loaded,
1039 			   &osd->rx_size_255, &nsd->rx_size_255);
1040 	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1041 			   I40E_GLPRT_PRC511L(hw->port),
1042 			   pf->stat_offsets_loaded,
1043 			   &osd->rx_size_511, &nsd->rx_size_511);
1044 	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1045 			   I40E_GLPRT_PRC1023L(hw->port),
1046 			   pf->stat_offsets_loaded,
1047 			   &osd->rx_size_1023, &nsd->rx_size_1023);
1048 	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1049 			   I40E_GLPRT_PRC1522L(hw->port),
1050 			   pf->stat_offsets_loaded,
1051 			   &osd->rx_size_1522, &nsd->rx_size_1522);
1052 	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1053 			   I40E_GLPRT_PRC9522L(hw->port),
1054 			   pf->stat_offsets_loaded,
1055 			   &osd->rx_size_big, &nsd->rx_size_big);
1056 
1057 	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1058 			   I40E_GLPRT_PTC64L(hw->port),
1059 			   pf->stat_offsets_loaded,
1060 			   &osd->tx_size_64, &nsd->tx_size_64);
1061 	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1062 			   I40E_GLPRT_PTC127L(hw->port),
1063 			   pf->stat_offsets_loaded,
1064 			   &osd->tx_size_127, &nsd->tx_size_127);
1065 	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1066 			   I40E_GLPRT_PTC255L(hw->port),
1067 			   pf->stat_offsets_loaded,
1068 			   &osd->tx_size_255, &nsd->tx_size_255);
1069 	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1070 			   I40E_GLPRT_PTC511L(hw->port),
1071 			   pf->stat_offsets_loaded,
1072 			   &osd->tx_size_511, &nsd->tx_size_511);
1073 	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1074 			   I40E_GLPRT_PTC1023L(hw->port),
1075 			   pf->stat_offsets_loaded,
1076 			   &osd->tx_size_1023, &nsd->tx_size_1023);
1077 	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1078 			   I40E_GLPRT_PTC1522L(hw->port),
1079 			   pf->stat_offsets_loaded,
1080 			   &osd->tx_size_1522, &nsd->tx_size_1522);
1081 	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1082 			   I40E_GLPRT_PTC9522L(hw->port),
1083 			   pf->stat_offsets_loaded,
1084 			   &osd->tx_size_big, &nsd->tx_size_big);
1085 
1086 	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1087 			   pf->stat_offsets_loaded,
1088 			   &osd->rx_undersize, &nsd->rx_undersize);
1089 	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1090 			   pf->stat_offsets_loaded,
1091 			   &osd->rx_fragments, &nsd->rx_fragments);
1092 	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1093 			   pf->stat_offsets_loaded,
1094 			   &osd->rx_oversize, &nsd->rx_oversize);
1095 	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1096 			   pf->stat_offsets_loaded,
1097 			   &osd->rx_jabber, &nsd->rx_jabber);
1098 
1099 	/* FDIR stats */
1100 	i40e_stat_update32(hw, I40E_GLQF_PCNT(pf->fd_atr_cnt_idx),
1101 			   pf->stat_offsets_loaded,
1102 			   &osd->fd_atr_match, &nsd->fd_atr_match);
1103 	i40e_stat_update32(hw, I40E_GLQF_PCNT(pf->fd_sb_cnt_idx),
1104 			   pf->stat_offsets_loaded,
1105 			   &osd->fd_sb_match, &nsd->fd_sb_match);
1106 
1107 	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1108 	nsd->tx_lpi_status =
1109 		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1110 			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1111 	nsd->rx_lpi_status =
1112 		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1113 			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1114 	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1115 			   pf->stat_offsets_loaded,
1116 			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1117 	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1118 			   pf->stat_offsets_loaded,
1119 			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1120 
1121 	pf->stat_offsets_loaded = true;
1122 }
1123 
1124 /**
1125  * i40e_update_stats - Update the various statistics counters.
1126  * @vsi: the VSI to be updated
1127  *
1128  * Update the various stats for this VSI and its related entities.
1129  **/
1130 void i40e_update_stats(struct i40e_vsi *vsi)
1131 {
1132 	struct i40e_pf *pf = vsi->back;
1133 
1134 	if (vsi == pf->vsi[pf->lan_vsi])
1135 		i40e_update_pf_stats(pf);
1136 
1137 	i40e_update_vsi_stats(vsi);
1138 #ifdef I40E_FCOE
1139 	i40e_update_fcoe_stats(vsi);
1140 #endif
1141 }
1142 
1143 /**
1144  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1145  * @vsi: the VSI to be searched
1146  * @macaddr: the MAC address
1147  * @vlan: the vlan
1148  * @is_vf: make sure its a VF filter, else doesn't matter
1149  * @is_netdev: make sure its a netdev filter, else doesn't matter
1150  *
1151  * Returns ptr to the filter object or NULL
1152  **/
1153 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1154 						u8 *macaddr, s16 vlan,
1155 						bool is_vf, bool is_netdev)
1156 {
1157 	struct i40e_mac_filter *f;
1158 
1159 	if (!vsi || !macaddr)
1160 		return NULL;
1161 
1162 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1163 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1164 		    (vlan == f->vlan)    &&
1165 		    (!is_vf || f->is_vf) &&
1166 		    (!is_netdev || f->is_netdev))
1167 			return f;
1168 	}
1169 	return NULL;
1170 }
1171 
1172 /**
1173  * i40e_find_mac - Find a mac addr in the macvlan filters list
1174  * @vsi: the VSI to be searched
1175  * @macaddr: the MAC address we are searching for
1176  * @is_vf: make sure its a VF filter, else doesn't matter
1177  * @is_netdev: make sure its a netdev filter, else doesn't matter
1178  *
1179  * Returns the first filter with the provided MAC address or NULL if
1180  * MAC address was not found
1181  **/
1182 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1183 				      bool is_vf, bool is_netdev)
1184 {
1185 	struct i40e_mac_filter *f;
1186 
1187 	if (!vsi || !macaddr)
1188 		return NULL;
1189 
1190 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1191 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1192 		    (!is_vf || f->is_vf) &&
1193 		    (!is_netdev || f->is_netdev))
1194 			return f;
1195 	}
1196 	return NULL;
1197 }
1198 
1199 /**
1200  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1201  * @vsi: the VSI to be searched
1202  *
1203  * Returns true if VSI is in vlan mode or false otherwise
1204  **/
1205 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1206 {
1207 	struct i40e_mac_filter *f;
1208 
1209 	/* Only -1 for all the filters denotes not in vlan mode
1210 	 * so we have to go through all the list in order to make sure
1211 	 */
1212 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1213 		if (f->vlan >= 0)
1214 			return true;
1215 	}
1216 
1217 	return false;
1218 }
1219 
1220 /**
1221  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1222  * @vsi: the VSI to be searched
1223  * @macaddr: the mac address to be filtered
1224  * @is_vf: true if it is a VF
1225  * @is_netdev: true if it is a netdev
1226  *
1227  * Goes through all the macvlan filters and adds a
1228  * macvlan filter for each unique vlan that already exists
1229  *
1230  * Returns first filter found on success, else NULL
1231  **/
1232 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1233 					     bool is_vf, bool is_netdev)
1234 {
1235 	struct i40e_mac_filter *f;
1236 
1237 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1238 		if (!i40e_find_filter(vsi, macaddr, f->vlan,
1239 				      is_vf, is_netdev)) {
1240 			if (!i40e_add_filter(vsi, macaddr, f->vlan,
1241 					     is_vf, is_netdev))
1242 				return NULL;
1243 		}
1244 	}
1245 
1246 	return list_first_entry_or_null(&vsi->mac_filter_list,
1247 					struct i40e_mac_filter, list);
1248 }
1249 
1250 /**
1251  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1252  * @vsi: the PF Main VSI - inappropriate for any other VSI
1253  * @macaddr: the MAC address
1254  *
1255  * Some older firmware configurations set up a default promiscuous VLAN
1256  * filter that needs to be removed.
1257  **/
1258 static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1259 {
1260 	struct i40e_aqc_remove_macvlan_element_data element;
1261 	struct i40e_pf *pf = vsi->back;
1262 	i40e_status aq_ret;
1263 
1264 	/* Only appropriate for the PF main VSI */
1265 	if (vsi->type != I40E_VSI_MAIN)
1266 		return -EINVAL;
1267 
1268 	memset(&element, 0, sizeof(element));
1269 	ether_addr_copy(element.mac_addr, macaddr);
1270 	element.vlan_tag = 0;
1271 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1272 			I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1273 	aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1274 	if (aq_ret)
1275 		return -ENOENT;
1276 
1277 	return 0;
1278 }
1279 
1280 /**
1281  * i40e_add_filter - Add a mac/vlan filter to the VSI
1282  * @vsi: the VSI to be searched
1283  * @macaddr: the MAC address
1284  * @vlan: the vlan
1285  * @is_vf: make sure its a VF filter, else doesn't matter
1286  * @is_netdev: make sure its a netdev filter, else doesn't matter
1287  *
1288  * Returns ptr to the filter object or NULL when no memory available.
1289  **/
1290 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1291 					u8 *macaddr, s16 vlan,
1292 					bool is_vf, bool is_netdev)
1293 {
1294 	struct i40e_mac_filter *f;
1295 
1296 	if (!vsi || !macaddr)
1297 		return NULL;
1298 
1299 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1300 	if (!f) {
1301 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1302 		if (!f)
1303 			goto add_filter_out;
1304 
1305 		ether_addr_copy(f->macaddr, macaddr);
1306 		f->vlan = vlan;
1307 		f->changed = true;
1308 
1309 		INIT_LIST_HEAD(&f->list);
1310 		list_add(&f->list, &vsi->mac_filter_list);
1311 	}
1312 
1313 	/* increment counter and add a new flag if needed */
1314 	if (is_vf) {
1315 		if (!f->is_vf) {
1316 			f->is_vf = true;
1317 			f->counter++;
1318 		}
1319 	} else if (is_netdev) {
1320 		if (!f->is_netdev) {
1321 			f->is_netdev = true;
1322 			f->counter++;
1323 		}
1324 	} else {
1325 		f->counter++;
1326 	}
1327 
1328 	/* changed tells sync_filters_subtask to
1329 	 * push the filter down to the firmware
1330 	 */
1331 	if (f->changed) {
1332 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1333 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1334 	}
1335 
1336 add_filter_out:
1337 	return f;
1338 }
1339 
1340 /**
1341  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1342  * @vsi: the VSI to be searched
1343  * @macaddr: the MAC address
1344  * @vlan: the vlan
1345  * @is_vf: make sure it's a VF filter, else doesn't matter
1346  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1347  **/
1348 void i40e_del_filter(struct i40e_vsi *vsi,
1349 		     u8 *macaddr, s16 vlan,
1350 		     bool is_vf, bool is_netdev)
1351 {
1352 	struct i40e_mac_filter *f;
1353 
1354 	if (!vsi || !macaddr)
1355 		return;
1356 
1357 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1358 	if (!f || f->counter == 0)
1359 		return;
1360 
1361 	if (is_vf) {
1362 		if (f->is_vf) {
1363 			f->is_vf = false;
1364 			f->counter--;
1365 		}
1366 	} else if (is_netdev) {
1367 		if (f->is_netdev) {
1368 			f->is_netdev = false;
1369 			f->counter--;
1370 		}
1371 	} else {
1372 		/* make sure we don't remove a filter in use by VF or netdev */
1373 		int min_f = 0;
1374 		min_f += (f->is_vf ? 1 : 0);
1375 		min_f += (f->is_netdev ? 1 : 0);
1376 
1377 		if (f->counter > min_f)
1378 			f->counter--;
1379 	}
1380 
1381 	/* counter == 0 tells sync_filters_subtask to
1382 	 * remove the filter from the firmware's list
1383 	 */
1384 	if (f->counter == 0) {
1385 		f->changed = true;
1386 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1387 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1388 	}
1389 }
1390 
1391 /**
1392  * i40e_set_mac - NDO callback to set mac address
1393  * @netdev: network interface device structure
1394  * @p: pointer to an address structure
1395  *
1396  * Returns 0 on success, negative on failure
1397  **/
1398 #ifdef I40E_FCOE
1399 int i40e_set_mac(struct net_device *netdev, void *p)
1400 #else
1401 static int i40e_set_mac(struct net_device *netdev, void *p)
1402 #endif
1403 {
1404 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1405 	struct i40e_vsi *vsi = np->vsi;
1406 	struct i40e_pf *pf = vsi->back;
1407 	struct i40e_hw *hw = &pf->hw;
1408 	struct sockaddr *addr = p;
1409 	struct i40e_mac_filter *f;
1410 
1411 	if (!is_valid_ether_addr(addr->sa_data))
1412 		return -EADDRNOTAVAIL;
1413 
1414 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1415 		netdev_info(netdev, "already using mac address %pM\n",
1416 			    addr->sa_data);
1417 		return 0;
1418 	}
1419 
1420 	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1421 	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1422 		return -EADDRNOTAVAIL;
1423 
1424 	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1425 		netdev_info(netdev, "returning to hw mac address %pM\n",
1426 			    hw->mac.addr);
1427 	else
1428 		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1429 
1430 	if (vsi->type == I40E_VSI_MAIN) {
1431 		i40e_status ret;
1432 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1433 						I40E_AQC_WRITE_TYPE_LAA_WOL,
1434 						addr->sa_data, NULL);
1435 		if (ret) {
1436 			netdev_info(netdev,
1437 				    "Addr change for Main VSI failed: %d\n",
1438 				    ret);
1439 			return -EADDRNOTAVAIL;
1440 		}
1441 	}
1442 
1443 	if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1444 		struct i40e_aqc_remove_macvlan_element_data element;
1445 
1446 		memset(&element, 0, sizeof(element));
1447 		ether_addr_copy(element.mac_addr, netdev->dev_addr);
1448 		element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1449 		i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1450 	} else {
1451 		i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1452 				false, false);
1453 	}
1454 
1455 	if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1456 		struct i40e_aqc_add_macvlan_element_data element;
1457 
1458 		memset(&element, 0, sizeof(element));
1459 		ether_addr_copy(element.mac_addr, hw->mac.addr);
1460 		element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1461 		i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1462 	} else {
1463 		f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1464 				    false, false);
1465 		if (f)
1466 			f->is_laa = true;
1467 	}
1468 
1469 	i40e_sync_vsi_filters(vsi);
1470 	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1471 
1472 	return 0;
1473 }
1474 
1475 /**
1476  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1477  * @vsi: the VSI being setup
1478  * @ctxt: VSI context structure
1479  * @enabled_tc: Enabled TCs bitmap
1480  * @is_add: True if called before Add VSI
1481  *
1482  * Setup VSI queue mapping for enabled traffic classes.
1483  **/
1484 #ifdef I40E_FCOE
1485 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1486 			      struct i40e_vsi_context *ctxt,
1487 			      u8 enabled_tc,
1488 			      bool is_add)
1489 #else
1490 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1491 				     struct i40e_vsi_context *ctxt,
1492 				     u8 enabled_tc,
1493 				     bool is_add)
1494 #endif
1495 {
1496 	struct i40e_pf *pf = vsi->back;
1497 	u16 sections = 0;
1498 	u8 netdev_tc = 0;
1499 	u16 numtc = 0;
1500 	u16 qcount;
1501 	u8 offset;
1502 	u16 qmap;
1503 	int i;
1504 	u16 num_tc_qps = 0;
1505 
1506 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1507 	offset = 0;
1508 
1509 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1510 		/* Find numtc from enabled TC bitmap */
1511 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1512 			if (enabled_tc & (1 << i)) /* TC is enabled */
1513 				numtc++;
1514 		}
1515 		if (!numtc) {
1516 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1517 			numtc = 1;
1518 		}
1519 	} else {
1520 		/* At least TC0 is enabled in case of non-DCB case */
1521 		numtc = 1;
1522 	}
1523 
1524 	vsi->tc_config.numtc = numtc;
1525 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1526 	/* Number of queues per enabled TC */
1527 	/* In MFP case we can have a much lower count of MSIx
1528 	 * vectors available and so we need to lower the used
1529 	 * q count.
1530 	 */
1531 	qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1532 	num_tc_qps = qcount / numtc;
1533 	num_tc_qps = min_t(int, num_tc_qps, I40E_MAX_QUEUES_PER_TC);
1534 
1535 	/* Setup queue offset/count for all TCs for given VSI */
1536 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1537 		/* See if the given TC is enabled for the given VSI */
1538 		if (vsi->tc_config.enabled_tc & (1 << i)) { /* TC is enabled */
1539 			int pow, num_qps;
1540 
1541 			switch (vsi->type) {
1542 			case I40E_VSI_MAIN:
1543 				qcount = min_t(int, pf->rss_size, num_tc_qps);
1544 				break;
1545 #ifdef I40E_FCOE
1546 			case I40E_VSI_FCOE:
1547 				qcount = num_tc_qps;
1548 				break;
1549 #endif
1550 			case I40E_VSI_FDIR:
1551 			case I40E_VSI_SRIOV:
1552 			case I40E_VSI_VMDQ2:
1553 			default:
1554 				qcount = num_tc_qps;
1555 				WARN_ON(i != 0);
1556 				break;
1557 			}
1558 			vsi->tc_config.tc_info[i].qoffset = offset;
1559 			vsi->tc_config.tc_info[i].qcount = qcount;
1560 
1561 			/* find the next higher power-of-2 of num queue pairs */
1562 			num_qps = qcount;
1563 			pow = 0;
1564 			while (num_qps && ((1 << pow) < qcount)) {
1565 				pow++;
1566 				num_qps >>= 1;
1567 			}
1568 
1569 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1570 			qmap =
1571 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1572 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1573 
1574 			offset += qcount;
1575 		} else {
1576 			/* TC is not enabled so set the offset to
1577 			 * default queue and allocate one queue
1578 			 * for the given TC.
1579 			 */
1580 			vsi->tc_config.tc_info[i].qoffset = 0;
1581 			vsi->tc_config.tc_info[i].qcount = 1;
1582 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1583 
1584 			qmap = 0;
1585 		}
1586 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1587 	}
1588 
1589 	/* Set actual Tx/Rx queue pairs */
1590 	vsi->num_queue_pairs = offset;
1591 	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1592 		if (vsi->req_queue_pairs > 0)
1593 			vsi->num_queue_pairs = vsi->req_queue_pairs;
1594 		else
1595 			vsi->num_queue_pairs = pf->num_lan_msix;
1596 	}
1597 
1598 	/* Scheduler section valid can only be set for ADD VSI */
1599 	if (is_add) {
1600 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1601 
1602 		ctxt->info.up_enable_bits = enabled_tc;
1603 	}
1604 	if (vsi->type == I40E_VSI_SRIOV) {
1605 		ctxt->info.mapping_flags |=
1606 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1607 		for (i = 0; i < vsi->num_queue_pairs; i++)
1608 			ctxt->info.queue_mapping[i] =
1609 					       cpu_to_le16(vsi->base_queue + i);
1610 	} else {
1611 		ctxt->info.mapping_flags |=
1612 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1613 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1614 	}
1615 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1616 }
1617 
1618 /**
1619  * i40e_set_rx_mode - NDO callback to set the netdev filters
1620  * @netdev: network interface device structure
1621  **/
1622 #ifdef I40E_FCOE
1623 void i40e_set_rx_mode(struct net_device *netdev)
1624 #else
1625 static void i40e_set_rx_mode(struct net_device *netdev)
1626 #endif
1627 {
1628 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1629 	struct i40e_mac_filter *f, *ftmp;
1630 	struct i40e_vsi *vsi = np->vsi;
1631 	struct netdev_hw_addr *uca;
1632 	struct netdev_hw_addr *mca;
1633 	struct netdev_hw_addr *ha;
1634 
1635 	/* add addr if not already in the filter list */
1636 	netdev_for_each_uc_addr(uca, netdev) {
1637 		if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1638 			if (i40e_is_vsi_in_vlan(vsi))
1639 				i40e_put_mac_in_vlan(vsi, uca->addr,
1640 						     false, true);
1641 			else
1642 				i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1643 						false, true);
1644 		}
1645 	}
1646 
1647 	netdev_for_each_mc_addr(mca, netdev) {
1648 		if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1649 			if (i40e_is_vsi_in_vlan(vsi))
1650 				i40e_put_mac_in_vlan(vsi, mca->addr,
1651 						     false, true);
1652 			else
1653 				i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1654 						false, true);
1655 		}
1656 	}
1657 
1658 	/* remove filter if not in netdev list */
1659 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1660 		bool found = false;
1661 
1662 		if (!f->is_netdev)
1663 			continue;
1664 
1665 		if (is_multicast_ether_addr(f->macaddr)) {
1666 			netdev_for_each_mc_addr(mca, netdev) {
1667 				if (ether_addr_equal(mca->addr, f->macaddr)) {
1668 					found = true;
1669 					break;
1670 				}
1671 			}
1672 		} else {
1673 			netdev_for_each_uc_addr(uca, netdev) {
1674 				if (ether_addr_equal(uca->addr, f->macaddr)) {
1675 					found = true;
1676 					break;
1677 				}
1678 			}
1679 
1680 			for_each_dev_addr(netdev, ha) {
1681 				if (ether_addr_equal(ha->addr, f->macaddr)) {
1682 					found = true;
1683 					break;
1684 				}
1685 			}
1686 		}
1687 		if (!found)
1688 			i40e_del_filter(
1689 			   vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1690 	}
1691 
1692 	/* check for other flag changes */
1693 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1694 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1695 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1696 	}
1697 }
1698 
1699 /**
1700  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1701  * @vsi: ptr to the VSI
1702  *
1703  * Push any outstanding VSI filter changes through the AdminQ.
1704  *
1705  * Returns 0 or error value
1706  **/
1707 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1708 {
1709 	struct i40e_mac_filter *f, *ftmp;
1710 	bool promisc_forced_on = false;
1711 	bool add_happened = false;
1712 	int filter_list_len = 0;
1713 	u32 changed_flags = 0;
1714 	i40e_status aq_ret = 0;
1715 	struct i40e_pf *pf;
1716 	int num_add = 0;
1717 	int num_del = 0;
1718 	u16 cmd_flags;
1719 
1720 	/* empty array typed pointers, kcalloc later */
1721 	struct i40e_aqc_add_macvlan_element_data *add_list;
1722 	struct i40e_aqc_remove_macvlan_element_data *del_list;
1723 
1724 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1725 		usleep_range(1000, 2000);
1726 	pf = vsi->back;
1727 
1728 	if (vsi->netdev) {
1729 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1730 		vsi->current_netdev_flags = vsi->netdev->flags;
1731 	}
1732 
1733 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1734 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1735 
1736 		filter_list_len = pf->hw.aq.asq_buf_size /
1737 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
1738 		del_list = kcalloc(filter_list_len,
1739 			    sizeof(struct i40e_aqc_remove_macvlan_element_data),
1740 			    GFP_KERNEL);
1741 		if (!del_list)
1742 			return -ENOMEM;
1743 
1744 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1745 			if (!f->changed)
1746 				continue;
1747 
1748 			if (f->counter != 0)
1749 				continue;
1750 			f->changed = false;
1751 			cmd_flags = 0;
1752 
1753 			/* add to delete list */
1754 			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1755 			del_list[num_del].vlan_tag =
1756 				cpu_to_le16((u16)(f->vlan ==
1757 					    I40E_VLAN_ANY ? 0 : f->vlan));
1758 
1759 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1760 			del_list[num_del].flags = cmd_flags;
1761 			num_del++;
1762 
1763 			/* unlink from filter list */
1764 			list_del(&f->list);
1765 			kfree(f);
1766 
1767 			/* flush a full buffer */
1768 			if (num_del == filter_list_len) {
1769 				aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1770 					    vsi->seid, del_list, num_del,
1771 					    NULL);
1772 				num_del = 0;
1773 				memset(del_list, 0, sizeof(*del_list));
1774 
1775 				if (aq_ret &&
1776 				    pf->hw.aq.asq_last_status !=
1777 							      I40E_AQ_RC_ENOENT)
1778 					dev_info(&pf->pdev->dev,
1779 						 "ignoring delete macvlan error, err %d, aq_err %d while flushing a full buffer\n",
1780 						 aq_ret,
1781 						 pf->hw.aq.asq_last_status);
1782 			}
1783 		}
1784 		if (num_del) {
1785 			aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1786 						     del_list, num_del, NULL);
1787 			num_del = 0;
1788 
1789 			if (aq_ret &&
1790 			    pf->hw.aq.asq_last_status != I40E_AQ_RC_ENOENT)
1791 				dev_info(&pf->pdev->dev,
1792 					 "ignoring delete macvlan error, err %d, aq_err %d\n",
1793 					 aq_ret, pf->hw.aq.asq_last_status);
1794 		}
1795 
1796 		kfree(del_list);
1797 		del_list = NULL;
1798 
1799 		/* do all the adds now */
1800 		filter_list_len = pf->hw.aq.asq_buf_size /
1801 			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1802 		add_list = kcalloc(filter_list_len,
1803 			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1804 			       GFP_KERNEL);
1805 		if (!add_list)
1806 			return -ENOMEM;
1807 
1808 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1809 			if (!f->changed)
1810 				continue;
1811 
1812 			if (f->counter == 0)
1813 				continue;
1814 			f->changed = false;
1815 			add_happened = true;
1816 			cmd_flags = 0;
1817 
1818 			/* add to add array */
1819 			ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
1820 			add_list[num_add].vlan_tag =
1821 				cpu_to_le16(
1822 				 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
1823 			add_list[num_add].queue_number = 0;
1824 
1825 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
1826 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
1827 			num_add++;
1828 
1829 			/* flush a full buffer */
1830 			if (num_add == filter_list_len) {
1831 				aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1832 							     add_list, num_add,
1833 							     NULL);
1834 				num_add = 0;
1835 
1836 				if (aq_ret)
1837 					break;
1838 				memset(add_list, 0, sizeof(*add_list));
1839 			}
1840 		}
1841 		if (num_add) {
1842 			aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1843 						     add_list, num_add, NULL);
1844 			num_add = 0;
1845 		}
1846 		kfree(add_list);
1847 		add_list = NULL;
1848 
1849 		if (add_happened && aq_ret &&
1850 		    pf->hw.aq.asq_last_status != I40E_AQ_RC_EINVAL) {
1851 			dev_info(&pf->pdev->dev,
1852 				 "add filter failed, err %d, aq_err %d\n",
1853 				 aq_ret, pf->hw.aq.asq_last_status);
1854 			if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
1855 			    !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1856 				      &vsi->state)) {
1857 				promisc_forced_on = true;
1858 				set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1859 					&vsi->state);
1860 				dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
1861 			}
1862 		}
1863 	}
1864 
1865 	/* check for changes in promiscuous modes */
1866 	if (changed_flags & IFF_ALLMULTI) {
1867 		bool cur_multipromisc;
1868 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
1869 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
1870 							       vsi->seid,
1871 							       cur_multipromisc,
1872 							       NULL);
1873 		if (aq_ret)
1874 			dev_info(&pf->pdev->dev,
1875 				 "set multi promisc failed, err %d, aq_err %d\n",
1876 				 aq_ret, pf->hw.aq.asq_last_status);
1877 	}
1878 	if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
1879 		bool cur_promisc;
1880 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
1881 			       test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1882 					&vsi->state));
1883 		aq_ret = i40e_aq_set_vsi_unicast_promiscuous(&vsi->back->hw,
1884 							     vsi->seid,
1885 							     cur_promisc, NULL);
1886 		if (aq_ret)
1887 			dev_info(&pf->pdev->dev,
1888 				 "set uni promisc failed, err %d, aq_err %d\n",
1889 				 aq_ret, pf->hw.aq.asq_last_status);
1890 		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
1891 						   vsi->seid,
1892 						   cur_promisc, NULL);
1893 		if (aq_ret)
1894 			dev_info(&pf->pdev->dev,
1895 				 "set brdcast promisc failed, err %d, aq_err %d\n",
1896 				 aq_ret, pf->hw.aq.asq_last_status);
1897 	}
1898 
1899 	clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
1900 	return 0;
1901 }
1902 
1903 /**
1904  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
1905  * @pf: board private structure
1906  **/
1907 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
1908 {
1909 	int v;
1910 
1911 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
1912 		return;
1913 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
1914 
1915 	for (v = 0; v < pf->num_alloc_vsi; v++) {
1916 		if (pf->vsi[v] &&
1917 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED))
1918 			i40e_sync_vsi_filters(pf->vsi[v]);
1919 	}
1920 }
1921 
1922 /**
1923  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
1924  * @netdev: network interface device structure
1925  * @new_mtu: new value for maximum frame size
1926  *
1927  * Returns 0 on success, negative on failure
1928  **/
1929 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
1930 {
1931 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1932 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
1933 	struct i40e_vsi *vsi = np->vsi;
1934 
1935 	/* MTU < 68 is an error and causes problems on some kernels */
1936 	if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
1937 		return -EINVAL;
1938 
1939 	netdev_info(netdev, "changing MTU from %d to %d\n",
1940 		    netdev->mtu, new_mtu);
1941 	netdev->mtu = new_mtu;
1942 	if (netif_running(netdev))
1943 		i40e_vsi_reinit_locked(vsi);
1944 
1945 	return 0;
1946 }
1947 
1948 /**
1949  * i40e_ioctl - Access the hwtstamp interface
1950  * @netdev: network interface device structure
1951  * @ifr: interface request data
1952  * @cmd: ioctl command
1953  **/
1954 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
1955 {
1956 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1957 	struct i40e_pf *pf = np->vsi->back;
1958 
1959 	switch (cmd) {
1960 	case SIOCGHWTSTAMP:
1961 		return i40e_ptp_get_ts_config(pf, ifr);
1962 	case SIOCSHWTSTAMP:
1963 		return i40e_ptp_set_ts_config(pf, ifr);
1964 	default:
1965 		return -EOPNOTSUPP;
1966 	}
1967 }
1968 
1969 /**
1970  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
1971  * @vsi: the vsi being adjusted
1972  **/
1973 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
1974 {
1975 	struct i40e_vsi_context ctxt;
1976 	i40e_status ret;
1977 
1978 	if ((vsi->info.valid_sections &
1979 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1980 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
1981 		return;  /* already enabled */
1982 
1983 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1984 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1985 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
1986 
1987 	ctxt.seid = vsi->seid;
1988 	ctxt.info = vsi->info;
1989 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1990 	if (ret) {
1991 		dev_info(&vsi->back->pdev->dev,
1992 			 "%s: update vsi failed, aq_err=%d\n",
1993 			 __func__, vsi->back->hw.aq.asq_last_status);
1994 	}
1995 }
1996 
1997 /**
1998  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
1999  * @vsi: the vsi being adjusted
2000  **/
2001 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2002 {
2003 	struct i40e_vsi_context ctxt;
2004 	i40e_status ret;
2005 
2006 	if ((vsi->info.valid_sections &
2007 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2008 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2009 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2010 		return;  /* already disabled */
2011 
2012 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2013 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2014 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2015 
2016 	ctxt.seid = vsi->seid;
2017 	ctxt.info = vsi->info;
2018 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2019 	if (ret) {
2020 		dev_info(&vsi->back->pdev->dev,
2021 			 "%s: update vsi failed, aq_err=%d\n",
2022 			 __func__, vsi->back->hw.aq.asq_last_status);
2023 	}
2024 }
2025 
2026 /**
2027  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2028  * @netdev: network interface to be adjusted
2029  * @features: netdev features to test if VLAN offload is enabled or not
2030  **/
2031 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2032 {
2033 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2034 	struct i40e_vsi *vsi = np->vsi;
2035 
2036 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2037 		i40e_vlan_stripping_enable(vsi);
2038 	else
2039 		i40e_vlan_stripping_disable(vsi);
2040 }
2041 
2042 /**
2043  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2044  * @vsi: the vsi being configured
2045  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2046  **/
2047 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2048 {
2049 	struct i40e_mac_filter *f, *add_f;
2050 	bool is_netdev, is_vf;
2051 
2052 	is_vf = (vsi->type == I40E_VSI_SRIOV);
2053 	is_netdev = !!(vsi->netdev);
2054 
2055 	if (is_netdev) {
2056 		add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2057 					is_vf, is_netdev);
2058 		if (!add_f) {
2059 			dev_info(&vsi->back->pdev->dev,
2060 				 "Could not add vlan filter %d for %pM\n",
2061 				 vid, vsi->netdev->dev_addr);
2062 			return -ENOMEM;
2063 		}
2064 	}
2065 
2066 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
2067 		add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2068 		if (!add_f) {
2069 			dev_info(&vsi->back->pdev->dev,
2070 				 "Could not add vlan filter %d for %pM\n",
2071 				 vid, f->macaddr);
2072 			return -ENOMEM;
2073 		}
2074 	}
2075 
2076 	/* Now if we add a vlan tag, make sure to check if it is the first
2077 	 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2078 	 * with 0, so we now accept untagged and specified tagged traffic
2079 	 * (and not any taged and untagged)
2080 	 */
2081 	if (vid > 0) {
2082 		if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2083 						  I40E_VLAN_ANY,
2084 						  is_vf, is_netdev)) {
2085 			i40e_del_filter(vsi, vsi->netdev->dev_addr,
2086 					I40E_VLAN_ANY, is_vf, is_netdev);
2087 			add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2088 						is_vf, is_netdev);
2089 			if (!add_f) {
2090 				dev_info(&vsi->back->pdev->dev,
2091 					 "Could not add filter 0 for %pM\n",
2092 					 vsi->netdev->dev_addr);
2093 				return -ENOMEM;
2094 			}
2095 		}
2096 	}
2097 
2098 	/* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2099 	if (vid > 0 && !vsi->info.pvid) {
2100 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
2101 			if (i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2102 					     is_vf, is_netdev)) {
2103 				i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2104 						is_vf, is_netdev);
2105 				add_f = i40e_add_filter(vsi, f->macaddr,
2106 							0, is_vf, is_netdev);
2107 				if (!add_f) {
2108 					dev_info(&vsi->back->pdev->dev,
2109 						 "Could not add filter 0 for %pM\n",
2110 						 f->macaddr);
2111 					return -ENOMEM;
2112 				}
2113 			}
2114 		}
2115 	}
2116 
2117 	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
2118 	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
2119 		return 0;
2120 
2121 	return i40e_sync_vsi_filters(vsi);
2122 }
2123 
2124 /**
2125  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2126  * @vsi: the vsi being configured
2127  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2128  *
2129  * Return: 0 on success or negative otherwise
2130  **/
2131 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2132 {
2133 	struct net_device *netdev = vsi->netdev;
2134 	struct i40e_mac_filter *f, *add_f;
2135 	bool is_vf, is_netdev;
2136 	int filter_count = 0;
2137 
2138 	is_vf = (vsi->type == I40E_VSI_SRIOV);
2139 	is_netdev = !!(netdev);
2140 
2141 	if (is_netdev)
2142 		i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2143 
2144 	list_for_each_entry(f, &vsi->mac_filter_list, list)
2145 		i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2146 
2147 	/* go through all the filters for this VSI and if there is only
2148 	 * vid == 0 it means there are no other filters, so vid 0 must
2149 	 * be replaced with -1. This signifies that we should from now
2150 	 * on accept any traffic (with any tag present, or untagged)
2151 	 */
2152 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
2153 		if (is_netdev) {
2154 			if (f->vlan &&
2155 			    ether_addr_equal(netdev->dev_addr, f->macaddr))
2156 				filter_count++;
2157 		}
2158 
2159 		if (f->vlan)
2160 			filter_count++;
2161 	}
2162 
2163 	if (!filter_count && is_netdev) {
2164 		i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2165 		f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2166 				    is_vf, is_netdev);
2167 		if (!f) {
2168 			dev_info(&vsi->back->pdev->dev,
2169 				 "Could not add filter %d for %pM\n",
2170 				 I40E_VLAN_ANY, netdev->dev_addr);
2171 			return -ENOMEM;
2172 		}
2173 	}
2174 
2175 	if (!filter_count) {
2176 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
2177 			i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2178 			add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2179 					    is_vf, is_netdev);
2180 			if (!add_f) {
2181 				dev_info(&vsi->back->pdev->dev,
2182 					 "Could not add filter %d for %pM\n",
2183 					 I40E_VLAN_ANY, f->macaddr);
2184 				return -ENOMEM;
2185 			}
2186 		}
2187 	}
2188 
2189 	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
2190 	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
2191 		return 0;
2192 
2193 	return i40e_sync_vsi_filters(vsi);
2194 }
2195 
2196 /**
2197  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2198  * @netdev: network interface to be adjusted
2199  * @vid: vlan id to be added
2200  *
2201  * net_device_ops implementation for adding vlan ids
2202  **/
2203 #ifdef I40E_FCOE
2204 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2205 			 __always_unused __be16 proto, u16 vid)
2206 #else
2207 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2208 				__always_unused __be16 proto, u16 vid)
2209 #endif
2210 {
2211 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2212 	struct i40e_vsi *vsi = np->vsi;
2213 	int ret = 0;
2214 
2215 	if (vid > 4095)
2216 		return -EINVAL;
2217 
2218 	netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2219 
2220 	/* If the network stack called us with vid = 0 then
2221 	 * it is asking to receive priority tagged packets with
2222 	 * vlan id 0.  Our HW receives them by default when configured
2223 	 * to receive untagged packets so there is no need to add an
2224 	 * extra filter for vlan 0 tagged packets.
2225 	 */
2226 	if (vid)
2227 		ret = i40e_vsi_add_vlan(vsi, vid);
2228 
2229 	if (!ret && (vid < VLAN_N_VID))
2230 		set_bit(vid, vsi->active_vlans);
2231 
2232 	return ret;
2233 }
2234 
2235 /**
2236  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2237  * @netdev: network interface to be adjusted
2238  * @vid: vlan id to be removed
2239  *
2240  * net_device_ops implementation for removing vlan ids
2241  **/
2242 #ifdef I40E_FCOE
2243 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2244 			  __always_unused __be16 proto, u16 vid)
2245 #else
2246 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2247 				 __always_unused __be16 proto, u16 vid)
2248 #endif
2249 {
2250 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2251 	struct i40e_vsi *vsi = np->vsi;
2252 
2253 	netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2254 
2255 	/* return code is ignored as there is nothing a user
2256 	 * can do about failure to remove and a log message was
2257 	 * already printed from the other function
2258 	 */
2259 	i40e_vsi_kill_vlan(vsi, vid);
2260 
2261 	clear_bit(vid, vsi->active_vlans);
2262 
2263 	return 0;
2264 }
2265 
2266 /**
2267  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2268  * @vsi: the vsi being brought back up
2269  **/
2270 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2271 {
2272 	u16 vid;
2273 
2274 	if (!vsi->netdev)
2275 		return;
2276 
2277 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2278 
2279 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2280 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2281 				     vid);
2282 }
2283 
2284 /**
2285  * i40e_vsi_add_pvid - Add pvid for the VSI
2286  * @vsi: the vsi being adjusted
2287  * @vid: the vlan id to set as a PVID
2288  **/
2289 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2290 {
2291 	struct i40e_vsi_context ctxt;
2292 	i40e_status aq_ret;
2293 
2294 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2295 	vsi->info.pvid = cpu_to_le16(vid);
2296 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2297 				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2298 				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2299 
2300 	ctxt.seid = vsi->seid;
2301 	ctxt.info = vsi->info;
2302 	aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2303 	if (aq_ret) {
2304 		dev_info(&vsi->back->pdev->dev,
2305 			 "%s: update vsi failed, aq_err=%d\n",
2306 			 __func__, vsi->back->hw.aq.asq_last_status);
2307 		return -ENOENT;
2308 	}
2309 
2310 	return 0;
2311 }
2312 
2313 /**
2314  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2315  * @vsi: the vsi being adjusted
2316  *
2317  * Just use the vlan_rx_register() service to put it back to normal
2318  **/
2319 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2320 {
2321 	i40e_vlan_stripping_disable(vsi);
2322 
2323 	vsi->info.pvid = 0;
2324 }
2325 
2326 /**
2327  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2328  * @vsi: ptr to the VSI
2329  *
2330  * If this function returns with an error, then it's possible one or
2331  * more of the rings is populated (while the rest are not).  It is the
2332  * callers duty to clean those orphaned rings.
2333  *
2334  * Return 0 on success, negative on failure
2335  **/
2336 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2337 {
2338 	int i, err = 0;
2339 
2340 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2341 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2342 
2343 	return err;
2344 }
2345 
2346 /**
2347  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2348  * @vsi: ptr to the VSI
2349  *
2350  * Free VSI's transmit software resources
2351  **/
2352 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2353 {
2354 	int i;
2355 
2356 	if (!vsi->tx_rings)
2357 		return;
2358 
2359 	for (i = 0; i < vsi->num_queue_pairs; i++)
2360 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2361 			i40e_free_tx_resources(vsi->tx_rings[i]);
2362 }
2363 
2364 /**
2365  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2366  * @vsi: ptr to the VSI
2367  *
2368  * If this function returns with an error, then it's possible one or
2369  * more of the rings is populated (while the rest are not).  It is the
2370  * callers duty to clean those orphaned rings.
2371  *
2372  * Return 0 on success, negative on failure
2373  **/
2374 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2375 {
2376 	int i, err = 0;
2377 
2378 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2379 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2380 #ifdef I40E_FCOE
2381 	i40e_fcoe_setup_ddp_resources(vsi);
2382 #endif
2383 	return err;
2384 }
2385 
2386 /**
2387  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2388  * @vsi: ptr to the VSI
2389  *
2390  * Free all receive software resources
2391  **/
2392 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2393 {
2394 	int i;
2395 
2396 	if (!vsi->rx_rings)
2397 		return;
2398 
2399 	for (i = 0; i < vsi->num_queue_pairs; i++)
2400 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2401 			i40e_free_rx_resources(vsi->rx_rings[i]);
2402 #ifdef I40E_FCOE
2403 	i40e_fcoe_free_ddp_resources(vsi);
2404 #endif
2405 }
2406 
2407 /**
2408  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2409  * @ring: The Tx ring to configure
2410  *
2411  * This enables/disables XPS for a given Tx descriptor ring
2412  * based on the TCs enabled for the VSI that ring belongs to.
2413  **/
2414 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2415 {
2416 	struct i40e_vsi *vsi = ring->vsi;
2417 	cpumask_var_t mask;
2418 
2419 	if (!ring->q_vector || !ring->netdev)
2420 		return;
2421 
2422 	/* Single TC mode enable XPS */
2423 	if (vsi->tc_config.numtc <= 1) {
2424 		if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2425 			netif_set_xps_queue(ring->netdev,
2426 					    &ring->q_vector->affinity_mask,
2427 					    ring->queue_index);
2428 	} else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2429 		/* Disable XPS to allow selection based on TC */
2430 		bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2431 		netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2432 		free_cpumask_var(mask);
2433 	}
2434 }
2435 
2436 /**
2437  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2438  * @ring: The Tx ring to configure
2439  *
2440  * Configure the Tx descriptor ring in the HMC context.
2441  **/
2442 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2443 {
2444 	struct i40e_vsi *vsi = ring->vsi;
2445 	u16 pf_q = vsi->base_queue + ring->queue_index;
2446 	struct i40e_hw *hw = &vsi->back->hw;
2447 	struct i40e_hmc_obj_txq tx_ctx;
2448 	i40e_status err = 0;
2449 	u32 qtx_ctl = 0;
2450 
2451 	/* some ATR related tx ring init */
2452 	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2453 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2454 		ring->atr_count = 0;
2455 	} else {
2456 		ring->atr_sample_rate = 0;
2457 	}
2458 
2459 	/* configure XPS */
2460 	i40e_config_xps_tx_ring(ring);
2461 
2462 	/* clear the context structure first */
2463 	memset(&tx_ctx, 0, sizeof(tx_ctx));
2464 
2465 	tx_ctx.new_context = 1;
2466 	tx_ctx.base = (ring->dma / 128);
2467 	tx_ctx.qlen = ring->count;
2468 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2469 					       I40E_FLAG_FD_ATR_ENABLED));
2470 #ifdef I40E_FCOE
2471 	tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2472 #endif
2473 	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2474 	/* FDIR VSI tx ring can still use RS bit and writebacks */
2475 	if (vsi->type != I40E_VSI_FDIR)
2476 		tx_ctx.head_wb_ena = 1;
2477 	tx_ctx.head_wb_addr = ring->dma +
2478 			      (ring->count * sizeof(struct i40e_tx_desc));
2479 
2480 	/* As part of VSI creation/update, FW allocates certain
2481 	 * Tx arbitration queue sets for each TC enabled for
2482 	 * the VSI. The FW returns the handles to these queue
2483 	 * sets as part of the response buffer to Add VSI,
2484 	 * Update VSI, etc. AQ commands. It is expected that
2485 	 * these queue set handles be associated with the Tx
2486 	 * queues by the driver as part of the TX queue context
2487 	 * initialization. This has to be done regardless of
2488 	 * DCB as by default everything is mapped to TC0.
2489 	 */
2490 	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2491 	tx_ctx.rdylist_act = 0;
2492 
2493 	/* clear the context in the HMC */
2494 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2495 	if (err) {
2496 		dev_info(&vsi->back->pdev->dev,
2497 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2498 			 ring->queue_index, pf_q, err);
2499 		return -ENOMEM;
2500 	}
2501 
2502 	/* set the context in the HMC */
2503 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2504 	if (err) {
2505 		dev_info(&vsi->back->pdev->dev,
2506 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2507 			 ring->queue_index, pf_q, err);
2508 		return -ENOMEM;
2509 	}
2510 
2511 	/* Now associate this queue with this PCI function */
2512 	if (vsi->type == I40E_VSI_VMDQ2) {
2513 		qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2514 		qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2515 			   I40E_QTX_CTL_VFVM_INDX_MASK;
2516 	} else {
2517 		qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2518 	}
2519 
2520 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2521 		    I40E_QTX_CTL_PF_INDX_MASK);
2522 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2523 	i40e_flush(hw);
2524 
2525 	clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
2526 
2527 	/* cache tail off for easier writes later */
2528 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2529 
2530 	return 0;
2531 }
2532 
2533 /**
2534  * i40e_configure_rx_ring - Configure a receive ring context
2535  * @ring: The Rx ring to configure
2536  *
2537  * Configure the Rx descriptor ring in the HMC context.
2538  **/
2539 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2540 {
2541 	struct i40e_vsi *vsi = ring->vsi;
2542 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2543 	u16 pf_q = vsi->base_queue + ring->queue_index;
2544 	struct i40e_hw *hw = &vsi->back->hw;
2545 	struct i40e_hmc_obj_rxq rx_ctx;
2546 	i40e_status err = 0;
2547 
2548 	ring->state = 0;
2549 
2550 	/* clear the context structure first */
2551 	memset(&rx_ctx, 0, sizeof(rx_ctx));
2552 
2553 	ring->rx_buf_len = vsi->rx_buf_len;
2554 	ring->rx_hdr_len = vsi->rx_hdr_len;
2555 
2556 	rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2557 	rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2558 
2559 	rx_ctx.base = (ring->dma / 128);
2560 	rx_ctx.qlen = ring->count;
2561 
2562 	if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2563 		set_ring_16byte_desc_enabled(ring);
2564 		rx_ctx.dsize = 0;
2565 	} else {
2566 		rx_ctx.dsize = 1;
2567 	}
2568 
2569 	rx_ctx.dtype = vsi->dtype;
2570 	if (vsi->dtype) {
2571 		set_ring_ps_enabled(ring);
2572 		rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2573 				  I40E_RX_SPLIT_IP      |
2574 				  I40E_RX_SPLIT_TCP_UDP |
2575 				  I40E_RX_SPLIT_SCTP;
2576 	} else {
2577 		rx_ctx.hsplit_0 = 0;
2578 	}
2579 
2580 	rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2581 				  (chain_len * ring->rx_buf_len));
2582 	if (hw->revision_id == 0)
2583 		rx_ctx.lrxqthresh = 0;
2584 	else
2585 		rx_ctx.lrxqthresh = 2;
2586 	rx_ctx.crcstrip = 1;
2587 	rx_ctx.l2tsel = 1;
2588 	rx_ctx.showiv = 1;
2589 #ifdef I40E_FCOE
2590 	rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2591 #endif
2592 	/* set the prefena field to 1 because the manual says to */
2593 	rx_ctx.prefena = 1;
2594 
2595 	/* clear the context in the HMC */
2596 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2597 	if (err) {
2598 		dev_info(&vsi->back->pdev->dev,
2599 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2600 			 ring->queue_index, pf_q, err);
2601 		return -ENOMEM;
2602 	}
2603 
2604 	/* set the context in the HMC */
2605 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2606 	if (err) {
2607 		dev_info(&vsi->back->pdev->dev,
2608 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2609 			 ring->queue_index, pf_q, err);
2610 		return -ENOMEM;
2611 	}
2612 
2613 	/* cache tail for quicker writes, and clear the reg before use */
2614 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2615 	writel(0, ring->tail);
2616 
2617 	if (ring_is_ps_enabled(ring)) {
2618 		i40e_alloc_rx_headers(ring);
2619 		i40e_alloc_rx_buffers_ps(ring, I40E_DESC_UNUSED(ring));
2620 	} else {
2621 		i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring));
2622 	}
2623 
2624 	return 0;
2625 }
2626 
2627 /**
2628  * i40e_vsi_configure_tx - Configure the VSI for Tx
2629  * @vsi: VSI structure describing this set of rings and resources
2630  *
2631  * Configure the Tx VSI for operation.
2632  **/
2633 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2634 {
2635 	int err = 0;
2636 	u16 i;
2637 
2638 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2639 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2640 
2641 	return err;
2642 }
2643 
2644 /**
2645  * i40e_vsi_configure_rx - Configure the VSI for Rx
2646  * @vsi: the VSI being configured
2647  *
2648  * Configure the Rx VSI for operation.
2649  **/
2650 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2651 {
2652 	int err = 0;
2653 	u16 i;
2654 
2655 	if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2656 		vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2657 			       + ETH_FCS_LEN + VLAN_HLEN;
2658 	else
2659 		vsi->max_frame = I40E_RXBUFFER_2048;
2660 
2661 	/* figure out correct receive buffer length */
2662 	switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2663 				    I40E_FLAG_RX_PS_ENABLED)) {
2664 	case I40E_FLAG_RX_1BUF_ENABLED:
2665 		vsi->rx_hdr_len = 0;
2666 		vsi->rx_buf_len = vsi->max_frame;
2667 		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2668 		break;
2669 	case I40E_FLAG_RX_PS_ENABLED:
2670 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2671 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2672 		vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2673 		break;
2674 	default:
2675 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2676 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2677 		vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2678 		break;
2679 	}
2680 
2681 #ifdef I40E_FCOE
2682 	/* setup rx buffer for FCoE */
2683 	if ((vsi->type == I40E_VSI_FCOE) &&
2684 	    (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2685 		vsi->rx_hdr_len = 0;
2686 		vsi->rx_buf_len = I40E_RXBUFFER_3072;
2687 		vsi->max_frame = I40E_RXBUFFER_3072;
2688 		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2689 	}
2690 
2691 #endif /* I40E_FCOE */
2692 	/* round up for the chip's needs */
2693 	vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2694 				(1 << I40E_RXQ_CTX_HBUFF_SHIFT));
2695 	vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2696 				(1 << I40E_RXQ_CTX_DBUFF_SHIFT));
2697 
2698 	/* set up individual rings */
2699 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2700 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2701 
2702 	return err;
2703 }
2704 
2705 /**
2706  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2707  * @vsi: ptr to the VSI
2708  **/
2709 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2710 {
2711 	struct i40e_ring *tx_ring, *rx_ring;
2712 	u16 qoffset, qcount;
2713 	int i, n;
2714 
2715 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
2716 		/* Reset the TC information */
2717 		for (i = 0; i < vsi->num_queue_pairs; i++) {
2718 			rx_ring = vsi->rx_rings[i];
2719 			tx_ring = vsi->tx_rings[i];
2720 			rx_ring->dcb_tc = 0;
2721 			tx_ring->dcb_tc = 0;
2722 		}
2723 	}
2724 
2725 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
2726 		if (!(vsi->tc_config.enabled_tc & (1 << n)))
2727 			continue;
2728 
2729 		qoffset = vsi->tc_config.tc_info[n].qoffset;
2730 		qcount = vsi->tc_config.tc_info[n].qcount;
2731 		for (i = qoffset; i < (qoffset + qcount); i++) {
2732 			rx_ring = vsi->rx_rings[i];
2733 			tx_ring = vsi->tx_rings[i];
2734 			rx_ring->dcb_tc = n;
2735 			tx_ring->dcb_tc = n;
2736 		}
2737 	}
2738 }
2739 
2740 /**
2741  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
2742  * @vsi: ptr to the VSI
2743  **/
2744 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
2745 {
2746 	if (vsi->netdev)
2747 		i40e_set_rx_mode(vsi->netdev);
2748 }
2749 
2750 /**
2751  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
2752  * @vsi: Pointer to the targeted VSI
2753  *
2754  * This function replays the hlist on the hw where all the SB Flow Director
2755  * filters were saved.
2756  **/
2757 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
2758 {
2759 	struct i40e_fdir_filter *filter;
2760 	struct i40e_pf *pf = vsi->back;
2761 	struct hlist_node *node;
2762 
2763 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
2764 		return;
2765 
2766 	hlist_for_each_entry_safe(filter, node,
2767 				  &pf->fdir_filter_list, fdir_node) {
2768 		i40e_add_del_fdir(vsi, filter, true);
2769 	}
2770 }
2771 
2772 /**
2773  * i40e_vsi_configure - Set up the VSI for action
2774  * @vsi: the VSI being configured
2775  **/
2776 static int i40e_vsi_configure(struct i40e_vsi *vsi)
2777 {
2778 	int err;
2779 
2780 	i40e_set_vsi_rx_mode(vsi);
2781 	i40e_restore_vlan(vsi);
2782 	i40e_vsi_config_dcb_rings(vsi);
2783 	err = i40e_vsi_configure_tx(vsi);
2784 	if (!err)
2785 		err = i40e_vsi_configure_rx(vsi);
2786 
2787 	return err;
2788 }
2789 
2790 /**
2791  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
2792  * @vsi: the VSI being configured
2793  **/
2794 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
2795 {
2796 	struct i40e_pf *pf = vsi->back;
2797 	struct i40e_q_vector *q_vector;
2798 	struct i40e_hw *hw = &pf->hw;
2799 	u16 vector;
2800 	int i, q;
2801 	u32 val;
2802 	u32 qp;
2803 
2804 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
2805 	 * and PFINT_LNKLSTn registers, e.g.:
2806 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
2807 	 */
2808 	qp = vsi->base_queue;
2809 	vector = vsi->base_vector;
2810 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
2811 		q_vector = vsi->q_vectors[i];
2812 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2813 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
2814 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
2815 		     q_vector->rx.itr);
2816 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2817 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
2818 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
2819 		     q_vector->tx.itr);
2820 
2821 		/* Linked list for the queuepairs assigned to this vector */
2822 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
2823 		for (q = 0; q < q_vector->num_ringpairs; q++) {
2824 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2825 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
2826 			      (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
2827 			      (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
2828 			      (I40E_QUEUE_TYPE_TX
2829 				      << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
2830 
2831 			wr32(hw, I40E_QINT_RQCTL(qp), val);
2832 
2833 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2834 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
2835 			      (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
2836 			      ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
2837 			      (I40E_QUEUE_TYPE_RX
2838 				      << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2839 
2840 			/* Terminate the linked list */
2841 			if (q == (q_vector->num_ringpairs - 1))
2842 				val |= (I40E_QUEUE_END_OF_LIST
2843 					   << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2844 
2845 			wr32(hw, I40E_QINT_TQCTL(qp), val);
2846 			qp++;
2847 		}
2848 	}
2849 
2850 	i40e_flush(hw);
2851 }
2852 
2853 /**
2854  * i40e_enable_misc_int_causes - enable the non-queue interrupts
2855  * @hw: ptr to the hardware info
2856  **/
2857 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
2858 {
2859 	struct i40e_hw *hw = &pf->hw;
2860 	u32 val;
2861 
2862 	/* clear things first */
2863 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
2864 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
2865 
2866 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
2867 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
2868 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
2869 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
2870 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
2871 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
2872 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
2873 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2874 
2875 	if (pf->flags & I40E_FLAG_PTP)
2876 		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
2877 
2878 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
2879 
2880 	/* SW_ITR_IDX = 0, but don't change INTENA */
2881 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
2882 					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
2883 
2884 	/* OTHER_ITR_IDX = 0 */
2885 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
2886 }
2887 
2888 /**
2889  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
2890  * @vsi: the VSI being configured
2891  **/
2892 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
2893 {
2894 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
2895 	struct i40e_pf *pf = vsi->back;
2896 	struct i40e_hw *hw = &pf->hw;
2897 	u32 val;
2898 
2899 	/* set the ITR configuration */
2900 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2901 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
2902 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
2903 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2904 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
2905 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
2906 
2907 	i40e_enable_misc_int_causes(pf);
2908 
2909 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
2910 	wr32(hw, I40E_PFINT_LNKLST0, 0);
2911 
2912 	/* Associate the queue pair to the vector and enable the queue int */
2913 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		      |
2914 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2915 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2916 
2917 	wr32(hw, I40E_QINT_RQCTL(0), val);
2918 
2919 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
2920 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2921 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2922 
2923 	wr32(hw, I40E_QINT_TQCTL(0), val);
2924 	i40e_flush(hw);
2925 }
2926 
2927 /**
2928  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
2929  * @pf: board private structure
2930  **/
2931 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
2932 {
2933 	struct i40e_hw *hw = &pf->hw;
2934 
2935 	wr32(hw, I40E_PFINT_DYN_CTL0,
2936 	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2937 	i40e_flush(hw);
2938 }
2939 
2940 /**
2941  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
2942  * @pf: board private structure
2943  **/
2944 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
2945 {
2946 	struct i40e_hw *hw = &pf->hw;
2947 	u32 val;
2948 
2949 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
2950 	      I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
2951 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
2952 
2953 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
2954 	i40e_flush(hw);
2955 }
2956 
2957 /**
2958  * i40e_irq_dynamic_enable - Enable default interrupt generation settings
2959  * @vsi: pointer to a vsi
2960  * @vector: enable a particular Hw Interrupt vector
2961  **/
2962 void i40e_irq_dynamic_enable(struct i40e_vsi *vsi, int vector)
2963 {
2964 	struct i40e_pf *pf = vsi->back;
2965 	struct i40e_hw *hw = &pf->hw;
2966 	u32 val;
2967 
2968 	val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
2969 	      I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
2970 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2971 	wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2972 	/* skip the flush */
2973 }
2974 
2975 /**
2976  * i40e_irq_dynamic_disable - Disable default interrupt generation settings
2977  * @vsi: pointer to a vsi
2978  * @vector: disable a particular Hw Interrupt vector
2979  **/
2980 void i40e_irq_dynamic_disable(struct i40e_vsi *vsi, int vector)
2981 {
2982 	struct i40e_pf *pf = vsi->back;
2983 	struct i40e_hw *hw = &pf->hw;
2984 	u32 val;
2985 
2986 	val = I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT;
2987 	wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2988 	i40e_flush(hw);
2989 }
2990 
2991 /**
2992  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
2993  * @irq: interrupt number
2994  * @data: pointer to a q_vector
2995  **/
2996 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
2997 {
2998 	struct i40e_q_vector *q_vector = data;
2999 
3000 	if (!q_vector->tx.ring && !q_vector->rx.ring)
3001 		return IRQ_HANDLED;
3002 
3003 	napi_schedule(&q_vector->napi);
3004 
3005 	return IRQ_HANDLED;
3006 }
3007 
3008 /**
3009  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3010  * @vsi: the VSI being configured
3011  * @basename: name for the vector
3012  *
3013  * Allocates MSI-X vectors and requests interrupts from the kernel.
3014  **/
3015 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3016 {
3017 	int q_vectors = vsi->num_q_vectors;
3018 	struct i40e_pf *pf = vsi->back;
3019 	int base = vsi->base_vector;
3020 	int rx_int_idx = 0;
3021 	int tx_int_idx = 0;
3022 	int vector, err;
3023 
3024 	for (vector = 0; vector < q_vectors; vector++) {
3025 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3026 
3027 		if (q_vector->tx.ring && q_vector->rx.ring) {
3028 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3029 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3030 			tx_int_idx++;
3031 		} else if (q_vector->rx.ring) {
3032 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3033 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3034 		} else if (q_vector->tx.ring) {
3035 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3036 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3037 		} else {
3038 			/* skip this unused q_vector */
3039 			continue;
3040 		}
3041 		err = request_irq(pf->msix_entries[base + vector].vector,
3042 				  vsi->irq_handler,
3043 				  0,
3044 				  q_vector->name,
3045 				  q_vector);
3046 		if (err) {
3047 			dev_info(&pf->pdev->dev,
3048 				 "%s: request_irq failed, error: %d\n",
3049 				 __func__, err);
3050 			goto free_queue_irqs;
3051 		}
3052 		/* assign the mask for this irq */
3053 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3054 				      &q_vector->affinity_mask);
3055 	}
3056 
3057 	vsi->irqs_ready = true;
3058 	return 0;
3059 
3060 free_queue_irqs:
3061 	while (vector) {
3062 		vector--;
3063 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3064 				      NULL);
3065 		free_irq(pf->msix_entries[base + vector].vector,
3066 			 &(vsi->q_vectors[vector]));
3067 	}
3068 	return err;
3069 }
3070 
3071 /**
3072  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3073  * @vsi: the VSI being un-configured
3074  **/
3075 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3076 {
3077 	struct i40e_pf *pf = vsi->back;
3078 	struct i40e_hw *hw = &pf->hw;
3079 	int base = vsi->base_vector;
3080 	int i;
3081 
3082 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3083 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3084 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3085 	}
3086 
3087 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3088 		for (i = vsi->base_vector;
3089 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3090 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3091 
3092 		i40e_flush(hw);
3093 		for (i = 0; i < vsi->num_q_vectors; i++)
3094 			synchronize_irq(pf->msix_entries[i + base].vector);
3095 	} else {
3096 		/* Legacy and MSI mode - this stops all interrupt handling */
3097 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3098 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3099 		i40e_flush(hw);
3100 		synchronize_irq(pf->pdev->irq);
3101 	}
3102 }
3103 
3104 /**
3105  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3106  * @vsi: the VSI being configured
3107  **/
3108 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3109 {
3110 	struct i40e_pf *pf = vsi->back;
3111 	int i;
3112 
3113 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3114 		for (i = vsi->base_vector;
3115 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3116 			i40e_irq_dynamic_enable(vsi, i);
3117 	} else {
3118 		i40e_irq_dynamic_enable_icr0(pf);
3119 	}
3120 
3121 	i40e_flush(&pf->hw);
3122 	return 0;
3123 }
3124 
3125 /**
3126  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3127  * @pf: board private structure
3128  **/
3129 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3130 {
3131 	/* Disable ICR 0 */
3132 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3133 	i40e_flush(&pf->hw);
3134 }
3135 
3136 /**
3137  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3138  * @irq: interrupt number
3139  * @data: pointer to a q_vector
3140  *
3141  * This is the handler used for all MSI/Legacy interrupts, and deals
3142  * with both queue and non-queue interrupts.  This is also used in
3143  * MSIX mode to handle the non-queue interrupts.
3144  **/
3145 static irqreturn_t i40e_intr(int irq, void *data)
3146 {
3147 	struct i40e_pf *pf = (struct i40e_pf *)data;
3148 	struct i40e_hw *hw = &pf->hw;
3149 	irqreturn_t ret = IRQ_NONE;
3150 	u32 icr0, icr0_remaining;
3151 	u32 val, ena_mask;
3152 
3153 	icr0 = rd32(hw, I40E_PFINT_ICR0);
3154 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3155 
3156 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3157 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3158 		goto enable_intr;
3159 
3160 	/* if interrupt but no bits showing, must be SWINT */
3161 	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3162 	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3163 		pf->sw_int_count++;
3164 
3165 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3166 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3167 
3168 		/* temporarily disable queue cause for NAPI processing */
3169 		u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
3170 		qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3171 		wr32(hw, I40E_QINT_RQCTL(0), qval);
3172 
3173 		qval = rd32(hw, I40E_QINT_TQCTL(0));
3174 		qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3175 		wr32(hw, I40E_QINT_TQCTL(0), qval);
3176 
3177 		if (!test_bit(__I40E_DOWN, &pf->state))
3178 			napi_schedule(&pf->vsi[pf->lan_vsi]->q_vectors[0]->napi);
3179 	}
3180 
3181 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3182 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3183 		set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3184 	}
3185 
3186 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3187 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3188 		set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3189 	}
3190 
3191 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3192 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3193 		set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3194 	}
3195 
3196 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3197 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3198 			set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3199 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3200 		val = rd32(hw, I40E_GLGEN_RSTAT);
3201 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3202 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3203 		if (val == I40E_RESET_CORER) {
3204 			pf->corer_count++;
3205 		} else if (val == I40E_RESET_GLOBR) {
3206 			pf->globr_count++;
3207 		} else if (val == I40E_RESET_EMPR) {
3208 			pf->empr_count++;
3209 			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3210 		}
3211 	}
3212 
3213 	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3214 		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3215 		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3216 		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3217 			 rd32(hw, I40E_PFHMC_ERRORINFO),
3218 			 rd32(hw, I40E_PFHMC_ERRORDATA));
3219 	}
3220 
3221 	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3222 		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3223 
3224 		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3225 			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3226 			i40e_ptp_tx_hwtstamp(pf);
3227 		}
3228 	}
3229 
3230 	/* If a critical error is pending we have no choice but to reset the
3231 	 * device.
3232 	 * Report and mask out any remaining unexpected interrupts.
3233 	 */
3234 	icr0_remaining = icr0 & ena_mask;
3235 	if (icr0_remaining) {
3236 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3237 			 icr0_remaining);
3238 		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3239 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3240 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3241 			dev_info(&pf->pdev->dev, "device will be reset\n");
3242 			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3243 			i40e_service_event_schedule(pf);
3244 		}
3245 		ena_mask &= ~icr0_remaining;
3246 	}
3247 	ret = IRQ_HANDLED;
3248 
3249 enable_intr:
3250 	/* re-enable interrupt causes */
3251 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3252 	if (!test_bit(__I40E_DOWN, &pf->state)) {
3253 		i40e_service_event_schedule(pf);
3254 		i40e_irq_dynamic_enable_icr0(pf);
3255 	}
3256 
3257 	return ret;
3258 }
3259 
3260 /**
3261  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3262  * @tx_ring:  tx ring to clean
3263  * @budget:   how many cleans we're allowed
3264  *
3265  * Returns true if there's any budget left (e.g. the clean is finished)
3266  **/
3267 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3268 {
3269 	struct i40e_vsi *vsi = tx_ring->vsi;
3270 	u16 i = tx_ring->next_to_clean;
3271 	struct i40e_tx_buffer *tx_buf;
3272 	struct i40e_tx_desc *tx_desc;
3273 
3274 	tx_buf = &tx_ring->tx_bi[i];
3275 	tx_desc = I40E_TX_DESC(tx_ring, i);
3276 	i -= tx_ring->count;
3277 
3278 	do {
3279 		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3280 
3281 		/* if next_to_watch is not set then there is no work pending */
3282 		if (!eop_desc)
3283 			break;
3284 
3285 		/* prevent any other reads prior to eop_desc */
3286 		read_barrier_depends();
3287 
3288 		/* if the descriptor isn't done, no work yet to do */
3289 		if (!(eop_desc->cmd_type_offset_bsz &
3290 		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3291 			break;
3292 
3293 		/* clear next_to_watch to prevent false hangs */
3294 		tx_buf->next_to_watch = NULL;
3295 
3296 		tx_desc->buffer_addr = 0;
3297 		tx_desc->cmd_type_offset_bsz = 0;
3298 		/* move past filter desc */
3299 		tx_buf++;
3300 		tx_desc++;
3301 		i++;
3302 		if (unlikely(!i)) {
3303 			i -= tx_ring->count;
3304 			tx_buf = tx_ring->tx_bi;
3305 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3306 		}
3307 		/* unmap skb header data */
3308 		dma_unmap_single(tx_ring->dev,
3309 				 dma_unmap_addr(tx_buf, dma),
3310 				 dma_unmap_len(tx_buf, len),
3311 				 DMA_TO_DEVICE);
3312 		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3313 			kfree(tx_buf->raw_buf);
3314 
3315 		tx_buf->raw_buf = NULL;
3316 		tx_buf->tx_flags = 0;
3317 		tx_buf->next_to_watch = NULL;
3318 		dma_unmap_len_set(tx_buf, len, 0);
3319 		tx_desc->buffer_addr = 0;
3320 		tx_desc->cmd_type_offset_bsz = 0;
3321 
3322 		/* move us past the eop_desc for start of next FD desc */
3323 		tx_buf++;
3324 		tx_desc++;
3325 		i++;
3326 		if (unlikely(!i)) {
3327 			i -= tx_ring->count;
3328 			tx_buf = tx_ring->tx_bi;
3329 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3330 		}
3331 
3332 		/* update budget accounting */
3333 		budget--;
3334 	} while (likely(budget));
3335 
3336 	i += tx_ring->count;
3337 	tx_ring->next_to_clean = i;
3338 
3339 	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED) {
3340 		i40e_irq_dynamic_enable(vsi,
3341 				tx_ring->q_vector->v_idx + vsi->base_vector);
3342 	}
3343 	return budget > 0;
3344 }
3345 
3346 /**
3347  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3348  * @irq: interrupt number
3349  * @data: pointer to a q_vector
3350  **/
3351 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3352 {
3353 	struct i40e_q_vector *q_vector = data;
3354 	struct i40e_vsi *vsi;
3355 
3356 	if (!q_vector->tx.ring)
3357 		return IRQ_HANDLED;
3358 
3359 	vsi = q_vector->tx.ring->vsi;
3360 	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3361 
3362 	return IRQ_HANDLED;
3363 }
3364 
3365 /**
3366  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3367  * @vsi: the VSI being configured
3368  * @v_idx: vector index
3369  * @qp_idx: queue pair index
3370  **/
3371 static void map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3372 {
3373 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3374 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3375 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3376 
3377 	tx_ring->q_vector = q_vector;
3378 	tx_ring->next = q_vector->tx.ring;
3379 	q_vector->tx.ring = tx_ring;
3380 	q_vector->tx.count++;
3381 
3382 	rx_ring->q_vector = q_vector;
3383 	rx_ring->next = q_vector->rx.ring;
3384 	q_vector->rx.ring = rx_ring;
3385 	q_vector->rx.count++;
3386 }
3387 
3388 /**
3389  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3390  * @vsi: the VSI being configured
3391  *
3392  * This function maps descriptor rings to the queue-specific vectors
3393  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3394  * one vector per queue pair, but on a constrained vector budget, we
3395  * group the queue pairs as "efficiently" as possible.
3396  **/
3397 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3398 {
3399 	int qp_remaining = vsi->num_queue_pairs;
3400 	int q_vectors = vsi->num_q_vectors;
3401 	int num_ringpairs;
3402 	int v_start = 0;
3403 	int qp_idx = 0;
3404 
3405 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3406 	 * group them so there are multiple queues per vector.
3407 	 * It is also important to go through all the vectors available to be
3408 	 * sure that if we don't use all the vectors, that the remaining vectors
3409 	 * are cleared. This is especially important when decreasing the
3410 	 * number of queues in use.
3411 	 */
3412 	for (; v_start < q_vectors; v_start++) {
3413 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3414 
3415 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3416 
3417 		q_vector->num_ringpairs = num_ringpairs;
3418 
3419 		q_vector->rx.count = 0;
3420 		q_vector->tx.count = 0;
3421 		q_vector->rx.ring = NULL;
3422 		q_vector->tx.ring = NULL;
3423 
3424 		while (num_ringpairs--) {
3425 			map_vector_to_qp(vsi, v_start, qp_idx);
3426 			qp_idx++;
3427 			qp_remaining--;
3428 		}
3429 	}
3430 }
3431 
3432 /**
3433  * i40e_vsi_request_irq - Request IRQ from the OS
3434  * @vsi: the VSI being configured
3435  * @basename: name for the vector
3436  **/
3437 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3438 {
3439 	struct i40e_pf *pf = vsi->back;
3440 	int err;
3441 
3442 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3443 		err = i40e_vsi_request_irq_msix(vsi, basename);
3444 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3445 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
3446 				  pf->int_name, pf);
3447 	else
3448 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3449 				  pf->int_name, pf);
3450 
3451 	if (err)
3452 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3453 
3454 	return err;
3455 }
3456 
3457 #ifdef CONFIG_NET_POLL_CONTROLLER
3458 /**
3459  * i40e_netpoll - A Polling 'interrupt'handler
3460  * @netdev: network interface device structure
3461  *
3462  * This is used by netconsole to send skbs without having to re-enable
3463  * interrupts.  It's not called while the normal interrupt routine is executing.
3464  **/
3465 #ifdef I40E_FCOE
3466 void i40e_netpoll(struct net_device *netdev)
3467 #else
3468 static void i40e_netpoll(struct net_device *netdev)
3469 #endif
3470 {
3471 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3472 	struct i40e_vsi *vsi = np->vsi;
3473 	struct i40e_pf *pf = vsi->back;
3474 	int i;
3475 
3476 	/* if interface is down do nothing */
3477 	if (test_bit(__I40E_DOWN, &vsi->state))
3478 		return;
3479 
3480 	pf->flags |= I40E_FLAG_IN_NETPOLL;
3481 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3482 		for (i = 0; i < vsi->num_q_vectors; i++)
3483 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3484 	} else {
3485 		i40e_intr(pf->pdev->irq, netdev);
3486 	}
3487 	pf->flags &= ~I40E_FLAG_IN_NETPOLL;
3488 }
3489 #endif
3490 
3491 /**
3492  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3493  * @pf: the PF being configured
3494  * @pf_q: the PF queue
3495  * @enable: enable or disable state of the queue
3496  *
3497  * This routine will wait for the given Tx queue of the PF to reach the
3498  * enabled or disabled state.
3499  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3500  * multiple retries; else will return 0 in case of success.
3501  **/
3502 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3503 {
3504 	int i;
3505 	u32 tx_reg;
3506 
3507 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3508 		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3509 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3510 			break;
3511 
3512 		usleep_range(10, 20);
3513 	}
3514 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3515 		return -ETIMEDOUT;
3516 
3517 	return 0;
3518 }
3519 
3520 /**
3521  * i40e_vsi_control_tx - Start or stop a VSI's rings
3522  * @vsi: the VSI being configured
3523  * @enable: start or stop the rings
3524  **/
3525 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3526 {
3527 	struct i40e_pf *pf = vsi->back;
3528 	struct i40e_hw *hw = &pf->hw;
3529 	int i, j, pf_q, ret = 0;
3530 	u32 tx_reg;
3531 
3532 	pf_q = vsi->base_queue;
3533 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3534 
3535 		/* warn the TX unit of coming changes */
3536 		i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3537 		if (!enable)
3538 			usleep_range(10, 20);
3539 
3540 		for (j = 0; j < 50; j++) {
3541 			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3542 			if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3543 			    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3544 				break;
3545 			usleep_range(1000, 2000);
3546 		}
3547 		/* Skip if the queue is already in the requested state */
3548 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3549 			continue;
3550 
3551 		/* turn on/off the queue */
3552 		if (enable) {
3553 			wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3554 			tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3555 		} else {
3556 			tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3557 		}
3558 
3559 		wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3560 		/* No waiting for the Tx queue to disable */
3561 		if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3562 			continue;
3563 
3564 		/* wait for the change to finish */
3565 		ret = i40e_pf_txq_wait(pf, pf_q, enable);
3566 		if (ret) {
3567 			dev_info(&pf->pdev->dev,
3568 				 "%s: VSI seid %d Tx ring %d %sable timeout\n",
3569 				 __func__, vsi->seid, pf_q,
3570 				 (enable ? "en" : "dis"));
3571 			break;
3572 		}
3573 	}
3574 
3575 	if (hw->revision_id == 0)
3576 		mdelay(50);
3577 	return ret;
3578 }
3579 
3580 /**
3581  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3582  * @pf: the PF being configured
3583  * @pf_q: the PF queue
3584  * @enable: enable or disable state of the queue
3585  *
3586  * This routine will wait for the given Rx queue of the PF to reach the
3587  * enabled or disabled state.
3588  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3589  * multiple retries; else will return 0 in case of success.
3590  **/
3591 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3592 {
3593 	int i;
3594 	u32 rx_reg;
3595 
3596 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3597 		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3598 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3599 			break;
3600 
3601 		usleep_range(10, 20);
3602 	}
3603 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3604 		return -ETIMEDOUT;
3605 
3606 	return 0;
3607 }
3608 
3609 /**
3610  * i40e_vsi_control_rx - Start or stop a VSI's rings
3611  * @vsi: the VSI being configured
3612  * @enable: start or stop the rings
3613  **/
3614 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3615 {
3616 	struct i40e_pf *pf = vsi->back;
3617 	struct i40e_hw *hw = &pf->hw;
3618 	int i, j, pf_q, ret = 0;
3619 	u32 rx_reg;
3620 
3621 	pf_q = vsi->base_queue;
3622 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3623 		for (j = 0; j < 50; j++) {
3624 			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3625 			if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3626 			    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3627 				break;
3628 			usleep_range(1000, 2000);
3629 		}
3630 
3631 		/* Skip if the queue is already in the requested state */
3632 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3633 			continue;
3634 
3635 		/* turn on/off the queue */
3636 		if (enable)
3637 			rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3638 		else
3639 			rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3640 		wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3641 
3642 		/* wait for the change to finish */
3643 		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3644 		if (ret) {
3645 			dev_info(&pf->pdev->dev,
3646 				 "%s: VSI seid %d Rx ring %d %sable timeout\n",
3647 				 __func__, vsi->seid, pf_q,
3648 				 (enable ? "en" : "dis"));
3649 			break;
3650 		}
3651 	}
3652 
3653 	return ret;
3654 }
3655 
3656 /**
3657  * i40e_vsi_control_rings - Start or stop a VSI's rings
3658  * @vsi: the VSI being configured
3659  * @enable: start or stop the rings
3660  **/
3661 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3662 {
3663 	int ret = 0;
3664 
3665 	/* do rx first for enable and last for disable */
3666 	if (request) {
3667 		ret = i40e_vsi_control_rx(vsi, request);
3668 		if (ret)
3669 			return ret;
3670 		ret = i40e_vsi_control_tx(vsi, request);
3671 	} else {
3672 		/* Ignore return value, we need to shutdown whatever we can */
3673 		i40e_vsi_control_tx(vsi, request);
3674 		i40e_vsi_control_rx(vsi, request);
3675 	}
3676 
3677 	return ret;
3678 }
3679 
3680 /**
3681  * i40e_vsi_free_irq - Free the irq association with the OS
3682  * @vsi: the VSI being configured
3683  **/
3684 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3685 {
3686 	struct i40e_pf *pf = vsi->back;
3687 	struct i40e_hw *hw = &pf->hw;
3688 	int base = vsi->base_vector;
3689 	u32 val, qp;
3690 	int i;
3691 
3692 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3693 		if (!vsi->q_vectors)
3694 			return;
3695 
3696 		if (!vsi->irqs_ready)
3697 			return;
3698 
3699 		vsi->irqs_ready = false;
3700 		for (i = 0; i < vsi->num_q_vectors; i++) {
3701 			u16 vector = i + base;
3702 
3703 			/* free only the irqs that were actually requested */
3704 			if (!vsi->q_vectors[i] ||
3705 			    !vsi->q_vectors[i]->num_ringpairs)
3706 				continue;
3707 
3708 			/* clear the affinity_mask in the IRQ descriptor */
3709 			irq_set_affinity_hint(pf->msix_entries[vector].vector,
3710 					      NULL);
3711 			free_irq(pf->msix_entries[vector].vector,
3712 				 vsi->q_vectors[i]);
3713 
3714 			/* Tear down the interrupt queue link list
3715 			 *
3716 			 * We know that they come in pairs and always
3717 			 * the Rx first, then the Tx.  To clear the
3718 			 * link list, stick the EOL value into the
3719 			 * next_q field of the registers.
3720 			 */
3721 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3722 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3723 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3724 			val |= I40E_QUEUE_END_OF_LIST
3725 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3726 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
3727 
3728 			while (qp != I40E_QUEUE_END_OF_LIST) {
3729 				u32 next;
3730 
3731 				val = rd32(hw, I40E_QINT_RQCTL(qp));
3732 
3733 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3734 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3735 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3736 					 I40E_QINT_RQCTL_INTEVENT_MASK);
3737 
3738 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3739 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3740 
3741 				wr32(hw, I40E_QINT_RQCTL(qp), val);
3742 
3743 				val = rd32(hw, I40E_QINT_TQCTL(qp));
3744 
3745 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
3746 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
3747 
3748 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3749 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3750 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3751 					 I40E_QINT_TQCTL_INTEVENT_MASK);
3752 
3753 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3754 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3755 
3756 				wr32(hw, I40E_QINT_TQCTL(qp), val);
3757 				qp = next;
3758 			}
3759 		}
3760 	} else {
3761 		free_irq(pf->pdev->irq, pf);
3762 
3763 		val = rd32(hw, I40E_PFINT_LNKLST0);
3764 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3765 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3766 		val |= I40E_QUEUE_END_OF_LIST
3767 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
3768 		wr32(hw, I40E_PFINT_LNKLST0, val);
3769 
3770 		val = rd32(hw, I40E_QINT_RQCTL(qp));
3771 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3772 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3773 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3774 			 I40E_QINT_RQCTL_INTEVENT_MASK);
3775 
3776 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3777 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3778 
3779 		wr32(hw, I40E_QINT_RQCTL(qp), val);
3780 
3781 		val = rd32(hw, I40E_QINT_TQCTL(qp));
3782 
3783 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3784 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3785 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3786 			 I40E_QINT_TQCTL_INTEVENT_MASK);
3787 
3788 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3789 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3790 
3791 		wr32(hw, I40E_QINT_TQCTL(qp), val);
3792 	}
3793 }
3794 
3795 /**
3796  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
3797  * @vsi: the VSI being configured
3798  * @v_idx: Index of vector to be freed
3799  *
3800  * This function frees the memory allocated to the q_vector.  In addition if
3801  * NAPI is enabled it will delete any references to the NAPI struct prior
3802  * to freeing the q_vector.
3803  **/
3804 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
3805 {
3806 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3807 	struct i40e_ring *ring;
3808 
3809 	if (!q_vector)
3810 		return;
3811 
3812 	/* disassociate q_vector from rings */
3813 	i40e_for_each_ring(ring, q_vector->tx)
3814 		ring->q_vector = NULL;
3815 
3816 	i40e_for_each_ring(ring, q_vector->rx)
3817 		ring->q_vector = NULL;
3818 
3819 	/* only VSI w/ an associated netdev is set up w/ NAPI */
3820 	if (vsi->netdev)
3821 		netif_napi_del(&q_vector->napi);
3822 
3823 	vsi->q_vectors[v_idx] = NULL;
3824 
3825 	kfree_rcu(q_vector, rcu);
3826 }
3827 
3828 /**
3829  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
3830  * @vsi: the VSI being un-configured
3831  *
3832  * This frees the memory allocated to the q_vectors and
3833  * deletes references to the NAPI struct.
3834  **/
3835 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
3836 {
3837 	int v_idx;
3838 
3839 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
3840 		i40e_free_q_vector(vsi, v_idx);
3841 }
3842 
3843 /**
3844  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
3845  * @pf: board private structure
3846  **/
3847 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
3848 {
3849 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
3850 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3851 		pci_disable_msix(pf->pdev);
3852 		kfree(pf->msix_entries);
3853 		pf->msix_entries = NULL;
3854 		kfree(pf->irq_pile);
3855 		pf->irq_pile = NULL;
3856 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
3857 		pci_disable_msi(pf->pdev);
3858 	}
3859 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
3860 }
3861 
3862 /**
3863  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
3864  * @pf: board private structure
3865  *
3866  * We go through and clear interrupt specific resources and reset the structure
3867  * to pre-load conditions
3868  **/
3869 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
3870 {
3871 	int i;
3872 
3873 	i40e_stop_misc_vector(pf);
3874 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3875 		synchronize_irq(pf->msix_entries[0].vector);
3876 		free_irq(pf->msix_entries[0].vector, pf);
3877 	}
3878 
3879 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
3880 	for (i = 0; i < pf->num_alloc_vsi; i++)
3881 		if (pf->vsi[i])
3882 			i40e_vsi_free_q_vectors(pf->vsi[i]);
3883 	i40e_reset_interrupt_capability(pf);
3884 }
3885 
3886 /**
3887  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
3888  * @vsi: the VSI being configured
3889  **/
3890 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
3891 {
3892 	int q_idx;
3893 
3894 	if (!vsi->netdev)
3895 		return;
3896 
3897 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3898 		napi_enable(&vsi->q_vectors[q_idx]->napi);
3899 }
3900 
3901 /**
3902  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
3903  * @vsi: the VSI being configured
3904  **/
3905 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
3906 {
3907 	int q_idx;
3908 
3909 	if (!vsi->netdev)
3910 		return;
3911 
3912 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3913 		napi_disable(&vsi->q_vectors[q_idx]->napi);
3914 }
3915 
3916 /**
3917  * i40e_vsi_close - Shut down a VSI
3918  * @vsi: the vsi to be quelled
3919  **/
3920 static void i40e_vsi_close(struct i40e_vsi *vsi)
3921 {
3922 	if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
3923 		i40e_down(vsi);
3924 	i40e_vsi_free_irq(vsi);
3925 	i40e_vsi_free_tx_resources(vsi);
3926 	i40e_vsi_free_rx_resources(vsi);
3927 }
3928 
3929 /**
3930  * i40e_quiesce_vsi - Pause a given VSI
3931  * @vsi: the VSI being paused
3932  **/
3933 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
3934 {
3935 	if (test_bit(__I40E_DOWN, &vsi->state))
3936 		return;
3937 
3938 	/* No need to disable FCoE VSI when Tx suspended */
3939 	if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
3940 	    vsi->type == I40E_VSI_FCOE) {
3941 		dev_dbg(&vsi->back->pdev->dev,
3942 			"%s: VSI seid %d skipping FCoE VSI disable\n",
3943 			 __func__, vsi->seid);
3944 		return;
3945 	}
3946 
3947 	set_bit(__I40E_NEEDS_RESTART, &vsi->state);
3948 	if (vsi->netdev && netif_running(vsi->netdev)) {
3949 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3950 	} else {
3951 		i40e_vsi_close(vsi);
3952 	}
3953 }
3954 
3955 /**
3956  * i40e_unquiesce_vsi - Resume a given VSI
3957  * @vsi: the VSI being resumed
3958  **/
3959 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
3960 {
3961 	if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
3962 		return;
3963 
3964 	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
3965 	if (vsi->netdev && netif_running(vsi->netdev))
3966 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3967 	else
3968 		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
3969 }
3970 
3971 /**
3972  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
3973  * @pf: the PF
3974  **/
3975 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
3976 {
3977 	int v;
3978 
3979 	for (v = 0; v < pf->num_alloc_vsi; v++) {
3980 		if (pf->vsi[v])
3981 			i40e_quiesce_vsi(pf->vsi[v]);
3982 	}
3983 }
3984 
3985 /**
3986  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
3987  * @pf: the PF
3988  **/
3989 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
3990 {
3991 	int v;
3992 
3993 	for (v = 0; v < pf->num_alloc_vsi; v++) {
3994 		if (pf->vsi[v])
3995 			i40e_unquiesce_vsi(pf->vsi[v]);
3996 	}
3997 }
3998 
3999 #ifdef CONFIG_I40E_DCB
4000 /**
4001  * i40e_vsi_wait_txq_disabled - Wait for VSI's queues to be disabled
4002  * @vsi: the VSI being configured
4003  *
4004  * This function waits for the given VSI's Tx queues to be disabled.
4005  **/
4006 static int i40e_vsi_wait_txq_disabled(struct i40e_vsi *vsi)
4007 {
4008 	struct i40e_pf *pf = vsi->back;
4009 	int i, pf_q, ret;
4010 
4011 	pf_q = vsi->base_queue;
4012 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4013 		/* Check and wait for the disable status of the queue */
4014 		ret = i40e_pf_txq_wait(pf, pf_q, false);
4015 		if (ret) {
4016 			dev_info(&pf->pdev->dev,
4017 				 "%s: VSI seid %d Tx ring %d disable timeout\n",
4018 				 __func__, vsi->seid, pf_q);
4019 			return ret;
4020 		}
4021 	}
4022 
4023 	return 0;
4024 }
4025 
4026 /**
4027  * i40e_pf_wait_txq_disabled - Wait for all queues of PF VSIs to be disabled
4028  * @pf: the PF
4029  *
4030  * This function waits for the Tx queues to be in disabled state for all the
4031  * VSIs that are managed by this PF.
4032  **/
4033 static int i40e_pf_wait_txq_disabled(struct i40e_pf *pf)
4034 {
4035 	int v, ret = 0;
4036 
4037 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4038 		/* No need to wait for FCoE VSI queues */
4039 		if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4040 			ret = i40e_vsi_wait_txq_disabled(pf->vsi[v]);
4041 			if (ret)
4042 				break;
4043 		}
4044 	}
4045 
4046 	return ret;
4047 }
4048 
4049 #endif
4050 /**
4051  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4052  * @pf: pointer to PF
4053  *
4054  * Get TC map for ISCSI PF type that will include iSCSI TC
4055  * and LAN TC.
4056  **/
4057 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4058 {
4059 	struct i40e_dcb_app_priority_table app;
4060 	struct i40e_hw *hw = &pf->hw;
4061 	u8 enabled_tc = 1; /* TC0 is always enabled */
4062 	u8 tc, i;
4063 	/* Get the iSCSI APP TLV */
4064 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4065 
4066 	for (i = 0; i < dcbcfg->numapps; i++) {
4067 		app = dcbcfg->app[i];
4068 		if (app.selector == I40E_APP_SEL_TCPIP &&
4069 		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4070 			tc = dcbcfg->etscfg.prioritytable[app.priority];
4071 			enabled_tc |= (1 << tc);
4072 			break;
4073 		}
4074 	}
4075 
4076 	return enabled_tc;
4077 }
4078 
4079 /**
4080  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4081  * @dcbcfg: the corresponding DCBx configuration structure
4082  *
4083  * Return the number of TCs from given DCBx configuration
4084  **/
4085 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4086 {
4087 	u8 num_tc = 0;
4088 	int i;
4089 
4090 	/* Scan the ETS Config Priority Table to find
4091 	 * traffic class enabled for a given priority
4092 	 * and use the traffic class index to get the
4093 	 * number of traffic classes enabled
4094 	 */
4095 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4096 		if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4097 			num_tc = dcbcfg->etscfg.prioritytable[i];
4098 	}
4099 
4100 	/* Traffic class index starts from zero so
4101 	 * increment to return the actual count
4102 	 */
4103 	return num_tc + 1;
4104 }
4105 
4106 /**
4107  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4108  * @dcbcfg: the corresponding DCBx configuration structure
4109  *
4110  * Query the current DCB configuration and return the number of
4111  * traffic classes enabled from the given DCBX config
4112  **/
4113 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4114 {
4115 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4116 	u8 enabled_tc = 1;
4117 	u8 i;
4118 
4119 	for (i = 0; i < num_tc; i++)
4120 		enabled_tc |= 1 << i;
4121 
4122 	return enabled_tc;
4123 }
4124 
4125 /**
4126  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4127  * @pf: PF being queried
4128  *
4129  * Return number of traffic classes enabled for the given PF
4130  **/
4131 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4132 {
4133 	struct i40e_hw *hw = &pf->hw;
4134 	u8 i, enabled_tc;
4135 	u8 num_tc = 0;
4136 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4137 
4138 	/* If DCB is not enabled then always in single TC */
4139 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4140 		return 1;
4141 
4142 	/* SFP mode will be enabled for all TCs on port */
4143 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4144 		return i40e_dcb_get_num_tc(dcbcfg);
4145 
4146 	/* MFP mode return count of enabled TCs for this PF */
4147 	if (pf->hw.func_caps.iscsi)
4148 		enabled_tc =  i40e_get_iscsi_tc_map(pf);
4149 	else
4150 		return 1; /* Only TC0 */
4151 
4152 	/* At least have TC0 */
4153 	enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4154 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4155 		if (enabled_tc & (1 << i))
4156 			num_tc++;
4157 	}
4158 	return num_tc;
4159 }
4160 
4161 /**
4162  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4163  * @pf: PF being queried
4164  *
4165  * Return a bitmap for first enabled traffic class for this PF.
4166  **/
4167 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4168 {
4169 	u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4170 	u8 i = 0;
4171 
4172 	if (!enabled_tc)
4173 		return 0x1; /* TC0 */
4174 
4175 	/* Find the first enabled TC */
4176 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4177 		if (enabled_tc & (1 << i))
4178 			break;
4179 	}
4180 
4181 	return 1 << i;
4182 }
4183 
4184 /**
4185  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4186  * @pf: PF being queried
4187  *
4188  * Return a bitmap for enabled traffic classes for this PF.
4189  **/
4190 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4191 {
4192 	/* If DCB is not enabled for this PF then just return default TC */
4193 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4194 		return i40e_pf_get_default_tc(pf);
4195 
4196 	/* SFP mode we want PF to be enabled for all TCs */
4197 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4198 		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4199 
4200 	/* MFP enabled and iSCSI PF type */
4201 	if (pf->hw.func_caps.iscsi)
4202 		return i40e_get_iscsi_tc_map(pf);
4203 	else
4204 		return i40e_pf_get_default_tc(pf);
4205 }
4206 
4207 /**
4208  * i40e_vsi_get_bw_info - Query VSI BW Information
4209  * @vsi: the VSI being queried
4210  *
4211  * Returns 0 on success, negative value on failure
4212  **/
4213 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4214 {
4215 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4216 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4217 	struct i40e_pf *pf = vsi->back;
4218 	struct i40e_hw *hw = &pf->hw;
4219 	i40e_status aq_ret;
4220 	u32 tc_bw_max;
4221 	int i;
4222 
4223 	/* Get the VSI level BW configuration */
4224 	aq_ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4225 	if (aq_ret) {
4226 		dev_info(&pf->pdev->dev,
4227 			 "couldn't get PF vsi bw config, err %d, aq_err %d\n",
4228 			 aq_ret, pf->hw.aq.asq_last_status);
4229 		return -EINVAL;
4230 	}
4231 
4232 	/* Get the VSI level BW configuration per TC */
4233 	aq_ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4234 						  NULL);
4235 	if (aq_ret) {
4236 		dev_info(&pf->pdev->dev,
4237 			 "couldn't get PF vsi ets bw config, err %d, aq_err %d\n",
4238 			 aq_ret, pf->hw.aq.asq_last_status);
4239 		return -EINVAL;
4240 	}
4241 
4242 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4243 		dev_info(&pf->pdev->dev,
4244 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4245 			 bw_config.tc_valid_bits,
4246 			 bw_ets_config.tc_valid_bits);
4247 		/* Still continuing */
4248 	}
4249 
4250 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4251 	vsi->bw_max_quanta = bw_config.max_bw;
4252 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4253 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4254 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4255 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4256 		vsi->bw_ets_limit_credits[i] =
4257 					le16_to_cpu(bw_ets_config.credits[i]);
4258 		/* 3 bits out of 4 for each TC */
4259 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4260 	}
4261 
4262 	return 0;
4263 }
4264 
4265 /**
4266  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4267  * @vsi: the VSI being configured
4268  * @enabled_tc: TC bitmap
4269  * @bw_credits: BW shared credits per TC
4270  *
4271  * Returns 0 on success, negative value on failure
4272  **/
4273 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4274 				       u8 *bw_share)
4275 {
4276 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4277 	i40e_status aq_ret;
4278 	int i;
4279 
4280 	bw_data.tc_valid_bits = enabled_tc;
4281 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4282 		bw_data.tc_bw_credits[i] = bw_share[i];
4283 
4284 	aq_ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4285 					  NULL);
4286 	if (aq_ret) {
4287 		dev_info(&vsi->back->pdev->dev,
4288 			 "AQ command Config VSI BW allocation per TC failed = %d\n",
4289 			 vsi->back->hw.aq.asq_last_status);
4290 		return -EINVAL;
4291 	}
4292 
4293 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4294 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4295 
4296 	return 0;
4297 }
4298 
4299 /**
4300  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4301  * @vsi: the VSI being configured
4302  * @enabled_tc: TC map to be enabled
4303  *
4304  **/
4305 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4306 {
4307 	struct net_device *netdev = vsi->netdev;
4308 	struct i40e_pf *pf = vsi->back;
4309 	struct i40e_hw *hw = &pf->hw;
4310 	u8 netdev_tc = 0;
4311 	int i;
4312 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4313 
4314 	if (!netdev)
4315 		return;
4316 
4317 	if (!enabled_tc) {
4318 		netdev_reset_tc(netdev);
4319 		return;
4320 	}
4321 
4322 	/* Set up actual enabled TCs on the VSI */
4323 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4324 		return;
4325 
4326 	/* set per TC queues for the VSI */
4327 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4328 		/* Only set TC queues for enabled tcs
4329 		 *
4330 		 * e.g. For a VSI that has TC0 and TC3 enabled the
4331 		 * enabled_tc bitmap would be 0x00001001; the driver
4332 		 * will set the numtc for netdev as 2 that will be
4333 		 * referenced by the netdev layer as TC 0 and 1.
4334 		 */
4335 		if (vsi->tc_config.enabled_tc & (1 << i))
4336 			netdev_set_tc_queue(netdev,
4337 					vsi->tc_config.tc_info[i].netdev_tc,
4338 					vsi->tc_config.tc_info[i].qcount,
4339 					vsi->tc_config.tc_info[i].qoffset);
4340 	}
4341 
4342 	/* Assign UP2TC map for the VSI */
4343 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4344 		/* Get the actual TC# for the UP */
4345 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4346 		/* Get the mapped netdev TC# for the UP */
4347 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4348 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
4349 	}
4350 }
4351 
4352 /**
4353  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4354  * @vsi: the VSI being configured
4355  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4356  **/
4357 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4358 				      struct i40e_vsi_context *ctxt)
4359 {
4360 	/* copy just the sections touched not the entire info
4361 	 * since not all sections are valid as returned by
4362 	 * update vsi params
4363 	 */
4364 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
4365 	memcpy(&vsi->info.queue_mapping,
4366 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4367 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4368 	       sizeof(vsi->info.tc_mapping));
4369 }
4370 
4371 /**
4372  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4373  * @vsi: VSI to be configured
4374  * @enabled_tc: TC bitmap
4375  *
4376  * This configures a particular VSI for TCs that are mapped to the
4377  * given TC bitmap. It uses default bandwidth share for TCs across
4378  * VSIs to configure TC for a particular VSI.
4379  *
4380  * NOTE:
4381  * It is expected that the VSI queues have been quisced before calling
4382  * this function.
4383  **/
4384 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4385 {
4386 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4387 	struct i40e_vsi_context ctxt;
4388 	int ret = 0;
4389 	int i;
4390 
4391 	/* Check if enabled_tc is same as existing or new TCs */
4392 	if (vsi->tc_config.enabled_tc == enabled_tc)
4393 		return ret;
4394 
4395 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
4396 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4397 		if (enabled_tc & (1 << i))
4398 			bw_share[i] = 1;
4399 	}
4400 
4401 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4402 	if (ret) {
4403 		dev_info(&vsi->back->pdev->dev,
4404 			 "Failed configuring TC map %d for VSI %d\n",
4405 			 enabled_tc, vsi->seid);
4406 		goto out;
4407 	}
4408 
4409 	/* Update Queue Pairs Mapping for currently enabled UPs */
4410 	ctxt.seid = vsi->seid;
4411 	ctxt.pf_num = vsi->back->hw.pf_id;
4412 	ctxt.vf_num = 0;
4413 	ctxt.uplink_seid = vsi->uplink_seid;
4414 	ctxt.info = vsi->info;
4415 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4416 
4417 	/* Update the VSI after updating the VSI queue-mapping information */
4418 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4419 	if (ret) {
4420 		dev_info(&vsi->back->pdev->dev,
4421 			 "update vsi failed, aq_err=%d\n",
4422 			 vsi->back->hw.aq.asq_last_status);
4423 		goto out;
4424 	}
4425 	/* update the local VSI info with updated queue map */
4426 	i40e_vsi_update_queue_map(vsi, &ctxt);
4427 	vsi->info.valid_sections = 0;
4428 
4429 	/* Update current VSI BW information */
4430 	ret = i40e_vsi_get_bw_info(vsi);
4431 	if (ret) {
4432 		dev_info(&vsi->back->pdev->dev,
4433 			 "Failed updating vsi bw info, aq_err=%d\n",
4434 			 vsi->back->hw.aq.asq_last_status);
4435 		goto out;
4436 	}
4437 
4438 	/* Update the netdev TC setup */
4439 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4440 out:
4441 	return ret;
4442 }
4443 
4444 /**
4445  * i40e_veb_config_tc - Configure TCs for given VEB
4446  * @veb: given VEB
4447  * @enabled_tc: TC bitmap
4448  *
4449  * Configures given TC bitmap for VEB (switching) element
4450  **/
4451 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4452 {
4453 	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4454 	struct i40e_pf *pf = veb->pf;
4455 	int ret = 0;
4456 	int i;
4457 
4458 	/* No TCs or already enabled TCs just return */
4459 	if (!enabled_tc || veb->enabled_tc == enabled_tc)
4460 		return ret;
4461 
4462 	bw_data.tc_valid_bits = enabled_tc;
4463 	/* bw_data.absolute_credits is not set (relative) */
4464 
4465 	/* Enable ETS TCs with equal BW Share for now */
4466 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4467 		if (enabled_tc & (1 << i))
4468 			bw_data.tc_bw_share_credits[i] = 1;
4469 	}
4470 
4471 	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4472 						   &bw_data, NULL);
4473 	if (ret) {
4474 		dev_info(&pf->pdev->dev,
4475 			 "veb bw config failed, aq_err=%d\n",
4476 			 pf->hw.aq.asq_last_status);
4477 		goto out;
4478 	}
4479 
4480 	/* Update the BW information */
4481 	ret = i40e_veb_get_bw_info(veb);
4482 	if (ret) {
4483 		dev_info(&pf->pdev->dev,
4484 			 "Failed getting veb bw config, aq_err=%d\n",
4485 			 pf->hw.aq.asq_last_status);
4486 	}
4487 
4488 out:
4489 	return ret;
4490 }
4491 
4492 #ifdef CONFIG_I40E_DCB
4493 /**
4494  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4495  * @pf: PF struct
4496  *
4497  * Reconfigure VEB/VSIs on a given PF; it is assumed that
4498  * the caller would've quiesce all the VSIs before calling
4499  * this function
4500  **/
4501 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4502 {
4503 	u8 tc_map = 0;
4504 	int ret;
4505 	u8 v;
4506 
4507 	/* Enable the TCs available on PF to all VEBs */
4508 	tc_map = i40e_pf_get_tc_map(pf);
4509 	for (v = 0; v < I40E_MAX_VEB; v++) {
4510 		if (!pf->veb[v])
4511 			continue;
4512 		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4513 		if (ret) {
4514 			dev_info(&pf->pdev->dev,
4515 				 "Failed configuring TC for VEB seid=%d\n",
4516 				 pf->veb[v]->seid);
4517 			/* Will try to configure as many components */
4518 		}
4519 	}
4520 
4521 	/* Update each VSI */
4522 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4523 		if (!pf->vsi[v])
4524 			continue;
4525 
4526 		/* - Enable all TCs for the LAN VSI
4527 #ifdef I40E_FCOE
4528 		 * - For FCoE VSI only enable the TC configured
4529 		 *   as per the APP TLV
4530 #endif
4531 		 * - For all others keep them at TC0 for now
4532 		 */
4533 		if (v == pf->lan_vsi)
4534 			tc_map = i40e_pf_get_tc_map(pf);
4535 		else
4536 			tc_map = i40e_pf_get_default_tc(pf);
4537 #ifdef I40E_FCOE
4538 		if (pf->vsi[v]->type == I40E_VSI_FCOE)
4539 			tc_map = i40e_get_fcoe_tc_map(pf);
4540 #endif /* #ifdef I40E_FCOE */
4541 
4542 		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4543 		if (ret) {
4544 			dev_info(&pf->pdev->dev,
4545 				 "Failed configuring TC for VSI seid=%d\n",
4546 				 pf->vsi[v]->seid);
4547 			/* Will try to configure as many components */
4548 		} else {
4549 			/* Re-configure VSI vectors based on updated TC map */
4550 			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4551 			if (pf->vsi[v]->netdev)
4552 				i40e_dcbnl_set_all(pf->vsi[v]);
4553 		}
4554 	}
4555 }
4556 
4557 /**
4558  * i40e_resume_port_tx - Resume port Tx
4559  * @pf: PF struct
4560  *
4561  * Resume a port's Tx and issue a PF reset in case of failure to
4562  * resume.
4563  **/
4564 static int i40e_resume_port_tx(struct i40e_pf *pf)
4565 {
4566 	struct i40e_hw *hw = &pf->hw;
4567 	int ret;
4568 
4569 	ret = i40e_aq_resume_port_tx(hw, NULL);
4570 	if (ret) {
4571 		dev_info(&pf->pdev->dev,
4572 			 "AQ command Resume Port Tx failed = %d\n",
4573 			  pf->hw.aq.asq_last_status);
4574 		/* Schedule PF reset to recover */
4575 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4576 		i40e_service_event_schedule(pf);
4577 	}
4578 
4579 	return ret;
4580 }
4581 
4582 /**
4583  * i40e_init_pf_dcb - Initialize DCB configuration
4584  * @pf: PF being configured
4585  *
4586  * Query the current DCB configuration and cache it
4587  * in the hardware structure
4588  **/
4589 static int i40e_init_pf_dcb(struct i40e_pf *pf)
4590 {
4591 	struct i40e_hw *hw = &pf->hw;
4592 	int err = 0;
4593 
4594 	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
4595 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
4596 	    (pf->hw.aq.fw_maj_ver < 4))
4597 		goto out;
4598 
4599 	/* Get the initial DCB configuration */
4600 	err = i40e_init_dcb(hw);
4601 	if (!err) {
4602 		/* Device/Function is not DCBX capable */
4603 		if ((!hw->func_caps.dcb) ||
4604 		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
4605 			dev_info(&pf->pdev->dev,
4606 				 "DCBX offload is not supported or is disabled for this PF.\n");
4607 
4608 			if (pf->flags & I40E_FLAG_MFP_ENABLED)
4609 				goto out;
4610 
4611 		} else {
4612 			/* When status is not DISABLED then DCBX in FW */
4613 			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
4614 				       DCB_CAP_DCBX_VER_IEEE;
4615 
4616 			pf->flags |= I40E_FLAG_DCB_CAPABLE;
4617 			/* Enable DCB tagging only when more than one TC */
4618 			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
4619 				pf->flags |= I40E_FLAG_DCB_ENABLED;
4620 			dev_dbg(&pf->pdev->dev,
4621 				"DCBX offload is supported for this PF.\n");
4622 		}
4623 	} else {
4624 		dev_info(&pf->pdev->dev,
4625 			 "AQ Querying DCB configuration failed: aq_err %d\n",
4626 			 pf->hw.aq.asq_last_status);
4627 	}
4628 
4629 out:
4630 	return err;
4631 }
4632 #endif /* CONFIG_I40E_DCB */
4633 #define SPEED_SIZE 14
4634 #define FC_SIZE 8
4635 /**
4636  * i40e_print_link_message - print link up or down
4637  * @vsi: the VSI for which link needs a message
4638  */
4639 static void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
4640 {
4641 	char speed[SPEED_SIZE] = "Unknown";
4642 	char fc[FC_SIZE] = "RX/TX";
4643 
4644 	if (!isup) {
4645 		netdev_info(vsi->netdev, "NIC Link is Down\n");
4646 		return;
4647 	}
4648 
4649 	/* Warn user if link speed on NPAR enabled partition is not at
4650 	 * least 10GB
4651 	 */
4652 	if (vsi->back->hw.func_caps.npar_enable &&
4653 	    (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
4654 	     vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
4655 		netdev_warn(vsi->netdev,
4656 			    "The partition detected link speed that is less than 10Gbps\n");
4657 
4658 	switch (vsi->back->hw.phy.link_info.link_speed) {
4659 	case I40E_LINK_SPEED_40GB:
4660 		strlcpy(speed, "40 Gbps", SPEED_SIZE);
4661 		break;
4662 	case I40E_LINK_SPEED_20GB:
4663 		strncpy(speed, "20 Gbps", SPEED_SIZE);
4664 		break;
4665 	case I40E_LINK_SPEED_10GB:
4666 		strlcpy(speed, "10 Gbps", SPEED_SIZE);
4667 		break;
4668 	case I40E_LINK_SPEED_1GB:
4669 		strlcpy(speed, "1000 Mbps", SPEED_SIZE);
4670 		break;
4671 	case I40E_LINK_SPEED_100MB:
4672 		strncpy(speed, "100 Mbps", SPEED_SIZE);
4673 		break;
4674 	default:
4675 		break;
4676 	}
4677 
4678 	switch (vsi->back->hw.fc.current_mode) {
4679 	case I40E_FC_FULL:
4680 		strlcpy(fc, "RX/TX", FC_SIZE);
4681 		break;
4682 	case I40E_FC_TX_PAUSE:
4683 		strlcpy(fc, "TX", FC_SIZE);
4684 		break;
4685 	case I40E_FC_RX_PAUSE:
4686 		strlcpy(fc, "RX", FC_SIZE);
4687 		break;
4688 	default:
4689 		strlcpy(fc, "None", FC_SIZE);
4690 		break;
4691 	}
4692 
4693 	netdev_info(vsi->netdev, "NIC Link is Up %s Full Duplex, Flow Control: %s\n",
4694 		    speed, fc);
4695 }
4696 
4697 /**
4698  * i40e_up_complete - Finish the last steps of bringing up a connection
4699  * @vsi: the VSI being configured
4700  **/
4701 static int i40e_up_complete(struct i40e_vsi *vsi)
4702 {
4703 	struct i40e_pf *pf = vsi->back;
4704 	int err;
4705 
4706 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4707 		i40e_vsi_configure_msix(vsi);
4708 	else
4709 		i40e_configure_msi_and_legacy(vsi);
4710 
4711 	/* start rings */
4712 	err = i40e_vsi_control_rings(vsi, true);
4713 	if (err)
4714 		return err;
4715 
4716 	clear_bit(__I40E_DOWN, &vsi->state);
4717 	i40e_napi_enable_all(vsi);
4718 	i40e_vsi_enable_irq(vsi);
4719 
4720 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
4721 	    (vsi->netdev)) {
4722 		i40e_print_link_message(vsi, true);
4723 		netif_tx_start_all_queues(vsi->netdev);
4724 		netif_carrier_on(vsi->netdev);
4725 	} else if (vsi->netdev) {
4726 		i40e_print_link_message(vsi, false);
4727 		/* need to check for qualified module here*/
4728 		if ((pf->hw.phy.link_info.link_info &
4729 			I40E_AQ_MEDIA_AVAILABLE) &&
4730 		    (!(pf->hw.phy.link_info.an_info &
4731 			I40E_AQ_QUALIFIED_MODULE)))
4732 			netdev_err(vsi->netdev,
4733 				   "the driver failed to link because an unqualified module was detected.");
4734 	}
4735 
4736 	/* replay FDIR SB filters */
4737 	if (vsi->type == I40E_VSI_FDIR) {
4738 		/* reset fd counters */
4739 		pf->fd_add_err = pf->fd_atr_cnt = 0;
4740 		if (pf->fd_tcp_rule > 0) {
4741 			pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
4742 			dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
4743 			pf->fd_tcp_rule = 0;
4744 		}
4745 		i40e_fdir_filter_restore(vsi);
4746 	}
4747 	i40e_service_event_schedule(pf);
4748 
4749 	return 0;
4750 }
4751 
4752 /**
4753  * i40e_vsi_reinit_locked - Reset the VSI
4754  * @vsi: the VSI being configured
4755  *
4756  * Rebuild the ring structs after some configuration
4757  * has changed, e.g. MTU size.
4758  **/
4759 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
4760 {
4761 	struct i40e_pf *pf = vsi->back;
4762 
4763 	WARN_ON(in_interrupt());
4764 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
4765 		usleep_range(1000, 2000);
4766 	i40e_down(vsi);
4767 
4768 	/* Give a VF some time to respond to the reset.  The
4769 	 * two second wait is based upon the watchdog cycle in
4770 	 * the VF driver.
4771 	 */
4772 	if (vsi->type == I40E_VSI_SRIOV)
4773 		msleep(2000);
4774 	i40e_up(vsi);
4775 	clear_bit(__I40E_CONFIG_BUSY, &pf->state);
4776 }
4777 
4778 /**
4779  * i40e_up - Bring the connection back up after being down
4780  * @vsi: the VSI being configured
4781  **/
4782 int i40e_up(struct i40e_vsi *vsi)
4783 {
4784 	int err;
4785 
4786 	err = i40e_vsi_configure(vsi);
4787 	if (!err)
4788 		err = i40e_up_complete(vsi);
4789 
4790 	return err;
4791 }
4792 
4793 /**
4794  * i40e_down - Shutdown the connection processing
4795  * @vsi: the VSI being stopped
4796  **/
4797 void i40e_down(struct i40e_vsi *vsi)
4798 {
4799 	int i;
4800 
4801 	/* It is assumed that the caller of this function
4802 	 * sets the vsi->state __I40E_DOWN bit.
4803 	 */
4804 	if (vsi->netdev) {
4805 		netif_carrier_off(vsi->netdev);
4806 		netif_tx_disable(vsi->netdev);
4807 	}
4808 	i40e_vsi_disable_irq(vsi);
4809 	i40e_vsi_control_rings(vsi, false);
4810 	i40e_napi_disable_all(vsi);
4811 
4812 	for (i = 0; i < vsi->num_queue_pairs; i++) {
4813 		i40e_clean_tx_ring(vsi->tx_rings[i]);
4814 		i40e_clean_rx_ring(vsi->rx_rings[i]);
4815 	}
4816 }
4817 
4818 /**
4819  * i40e_setup_tc - configure multiple traffic classes
4820  * @netdev: net device to configure
4821  * @tc: number of traffic classes to enable
4822  **/
4823 #ifdef I40E_FCOE
4824 int i40e_setup_tc(struct net_device *netdev, u8 tc)
4825 #else
4826 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
4827 #endif
4828 {
4829 	struct i40e_netdev_priv *np = netdev_priv(netdev);
4830 	struct i40e_vsi *vsi = np->vsi;
4831 	struct i40e_pf *pf = vsi->back;
4832 	u8 enabled_tc = 0;
4833 	int ret = -EINVAL;
4834 	int i;
4835 
4836 	/* Check if DCB enabled to continue */
4837 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
4838 		netdev_info(netdev, "DCB is not enabled for adapter\n");
4839 		goto exit;
4840 	}
4841 
4842 	/* Check if MFP enabled */
4843 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
4844 		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
4845 		goto exit;
4846 	}
4847 
4848 	/* Check whether tc count is within enabled limit */
4849 	if (tc > i40e_pf_get_num_tc(pf)) {
4850 		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
4851 		goto exit;
4852 	}
4853 
4854 	/* Generate TC map for number of tc requested */
4855 	for (i = 0; i < tc; i++)
4856 		enabled_tc |= (1 << i);
4857 
4858 	/* Requesting same TC configuration as already enabled */
4859 	if (enabled_tc == vsi->tc_config.enabled_tc)
4860 		return 0;
4861 
4862 	/* Quiesce VSI queues */
4863 	i40e_quiesce_vsi(vsi);
4864 
4865 	/* Configure VSI for enabled TCs */
4866 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
4867 	if (ret) {
4868 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
4869 			    vsi->seid);
4870 		goto exit;
4871 	}
4872 
4873 	/* Unquiesce VSI */
4874 	i40e_unquiesce_vsi(vsi);
4875 
4876 exit:
4877 	return ret;
4878 }
4879 
4880 /**
4881  * i40e_open - Called when a network interface is made active
4882  * @netdev: network interface device structure
4883  *
4884  * The open entry point is called when a network interface is made
4885  * active by the system (IFF_UP).  At this point all resources needed
4886  * for transmit and receive operations are allocated, the interrupt
4887  * handler is registered with the OS, the netdev watchdog subtask is
4888  * enabled, and the stack is notified that the interface is ready.
4889  *
4890  * Returns 0 on success, negative value on failure
4891  **/
4892 int i40e_open(struct net_device *netdev)
4893 {
4894 	struct i40e_netdev_priv *np = netdev_priv(netdev);
4895 	struct i40e_vsi *vsi = np->vsi;
4896 	struct i40e_pf *pf = vsi->back;
4897 	int err;
4898 
4899 	/* disallow open during test or if eeprom is broken */
4900 	if (test_bit(__I40E_TESTING, &pf->state) ||
4901 	    test_bit(__I40E_BAD_EEPROM, &pf->state))
4902 		return -EBUSY;
4903 
4904 	netif_carrier_off(netdev);
4905 
4906 	err = i40e_vsi_open(vsi);
4907 	if (err)
4908 		return err;
4909 
4910 	/* configure global TSO hardware offload settings */
4911 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
4912 						       TCP_FLAG_FIN) >> 16);
4913 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
4914 						       TCP_FLAG_FIN |
4915 						       TCP_FLAG_CWR) >> 16);
4916 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
4917 
4918 #ifdef CONFIG_I40E_VXLAN
4919 	vxlan_get_rx_port(netdev);
4920 #endif
4921 
4922 	return 0;
4923 }
4924 
4925 /**
4926  * i40e_vsi_open -
4927  * @vsi: the VSI to open
4928  *
4929  * Finish initialization of the VSI.
4930  *
4931  * Returns 0 on success, negative value on failure
4932  **/
4933 int i40e_vsi_open(struct i40e_vsi *vsi)
4934 {
4935 	struct i40e_pf *pf = vsi->back;
4936 	char int_name[I40E_INT_NAME_STR_LEN];
4937 	int err;
4938 
4939 	/* allocate descriptors */
4940 	err = i40e_vsi_setup_tx_resources(vsi);
4941 	if (err)
4942 		goto err_setup_tx;
4943 	err = i40e_vsi_setup_rx_resources(vsi);
4944 	if (err)
4945 		goto err_setup_rx;
4946 
4947 	err = i40e_vsi_configure(vsi);
4948 	if (err)
4949 		goto err_setup_rx;
4950 
4951 	if (vsi->netdev) {
4952 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
4953 			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
4954 		err = i40e_vsi_request_irq(vsi, int_name);
4955 		if (err)
4956 			goto err_setup_rx;
4957 
4958 		/* Notify the stack of the actual queue counts. */
4959 		err = netif_set_real_num_tx_queues(vsi->netdev,
4960 						   vsi->num_queue_pairs);
4961 		if (err)
4962 			goto err_set_queues;
4963 
4964 		err = netif_set_real_num_rx_queues(vsi->netdev,
4965 						   vsi->num_queue_pairs);
4966 		if (err)
4967 			goto err_set_queues;
4968 
4969 	} else if (vsi->type == I40E_VSI_FDIR) {
4970 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
4971 			 dev_driver_string(&pf->pdev->dev),
4972 			 dev_name(&pf->pdev->dev));
4973 		err = i40e_vsi_request_irq(vsi, int_name);
4974 
4975 	} else {
4976 		err = -EINVAL;
4977 		goto err_setup_rx;
4978 	}
4979 
4980 	err = i40e_up_complete(vsi);
4981 	if (err)
4982 		goto err_up_complete;
4983 
4984 	return 0;
4985 
4986 err_up_complete:
4987 	i40e_down(vsi);
4988 err_set_queues:
4989 	i40e_vsi_free_irq(vsi);
4990 err_setup_rx:
4991 	i40e_vsi_free_rx_resources(vsi);
4992 err_setup_tx:
4993 	i40e_vsi_free_tx_resources(vsi);
4994 	if (vsi == pf->vsi[pf->lan_vsi])
4995 		i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
4996 
4997 	return err;
4998 }
4999 
5000 /**
5001  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5002  * @pf: Pointer to PF
5003  *
5004  * This function destroys the hlist where all the Flow Director
5005  * filters were saved.
5006  **/
5007 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5008 {
5009 	struct i40e_fdir_filter *filter;
5010 	struct hlist_node *node2;
5011 
5012 	hlist_for_each_entry_safe(filter, node2,
5013 				  &pf->fdir_filter_list, fdir_node) {
5014 		hlist_del(&filter->fdir_node);
5015 		kfree(filter);
5016 	}
5017 	pf->fdir_pf_active_filters = 0;
5018 }
5019 
5020 /**
5021  * i40e_close - Disables a network interface
5022  * @netdev: network interface device structure
5023  *
5024  * The close entry point is called when an interface is de-activated
5025  * by the OS.  The hardware is still under the driver's control, but
5026  * this netdev interface is disabled.
5027  *
5028  * Returns 0, this is not allowed to fail
5029  **/
5030 #ifdef I40E_FCOE
5031 int i40e_close(struct net_device *netdev)
5032 #else
5033 static int i40e_close(struct net_device *netdev)
5034 #endif
5035 {
5036 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5037 	struct i40e_vsi *vsi = np->vsi;
5038 
5039 	i40e_vsi_close(vsi);
5040 
5041 	return 0;
5042 }
5043 
5044 /**
5045  * i40e_do_reset - Start a PF or Core Reset sequence
5046  * @pf: board private structure
5047  * @reset_flags: which reset is requested
5048  *
5049  * The essential difference in resets is that the PF Reset
5050  * doesn't clear the packet buffers, doesn't reset the PE
5051  * firmware, and doesn't bother the other PFs on the chip.
5052  **/
5053 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5054 {
5055 	u32 val;
5056 
5057 	WARN_ON(in_interrupt());
5058 
5059 	if (i40e_check_asq_alive(&pf->hw))
5060 		i40e_vc_notify_reset(pf);
5061 
5062 	/* do the biggest reset indicated */
5063 	if (reset_flags & (1 << __I40E_GLOBAL_RESET_REQUESTED)) {
5064 
5065 		/* Request a Global Reset
5066 		 *
5067 		 * This will start the chip's countdown to the actual full
5068 		 * chip reset event, and a warning interrupt to be sent
5069 		 * to all PFs, including the requestor.  Our handler
5070 		 * for the warning interrupt will deal with the shutdown
5071 		 * and recovery of the switch setup.
5072 		 */
5073 		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5074 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5075 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5076 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5077 
5078 	} else if (reset_flags & (1 << __I40E_CORE_RESET_REQUESTED)) {
5079 
5080 		/* Request a Core Reset
5081 		 *
5082 		 * Same as Global Reset, except does *not* include the MAC/PHY
5083 		 */
5084 		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5085 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5086 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
5087 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5088 		i40e_flush(&pf->hw);
5089 
5090 	} else if (reset_flags & (1 << __I40E_PF_RESET_REQUESTED)) {
5091 
5092 		/* Request a PF Reset
5093 		 *
5094 		 * Resets only the PF-specific registers
5095 		 *
5096 		 * This goes directly to the tear-down and rebuild of
5097 		 * the switch, since we need to do all the recovery as
5098 		 * for the Core Reset.
5099 		 */
5100 		dev_dbg(&pf->pdev->dev, "PFR requested\n");
5101 		i40e_handle_reset_warning(pf);
5102 
5103 	} else if (reset_flags & (1 << __I40E_REINIT_REQUESTED)) {
5104 		int v;
5105 
5106 		/* Find the VSI(s) that requested a re-init */
5107 		dev_info(&pf->pdev->dev,
5108 			 "VSI reinit requested\n");
5109 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5110 			struct i40e_vsi *vsi = pf->vsi[v];
5111 			if (vsi != NULL &&
5112 			    test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5113 				i40e_vsi_reinit_locked(pf->vsi[v]);
5114 				clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5115 			}
5116 		}
5117 
5118 		/* no further action needed, so return now */
5119 		return;
5120 	} else if (reset_flags & (1 << __I40E_DOWN_REQUESTED)) {
5121 		int v;
5122 
5123 		/* Find the VSI(s) that needs to be brought down */
5124 		dev_info(&pf->pdev->dev, "VSI down requested\n");
5125 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5126 			struct i40e_vsi *vsi = pf->vsi[v];
5127 			if (vsi != NULL &&
5128 			    test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5129 				set_bit(__I40E_DOWN, &vsi->state);
5130 				i40e_down(vsi);
5131 				clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5132 			}
5133 		}
5134 
5135 		/* no further action needed, so return now */
5136 		return;
5137 	} else {
5138 		dev_info(&pf->pdev->dev,
5139 			 "bad reset request 0x%08x\n", reset_flags);
5140 		return;
5141 	}
5142 }
5143 
5144 #ifdef CONFIG_I40E_DCB
5145 /**
5146  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5147  * @pf: board private structure
5148  * @old_cfg: current DCB config
5149  * @new_cfg: new DCB config
5150  **/
5151 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5152 			    struct i40e_dcbx_config *old_cfg,
5153 			    struct i40e_dcbx_config *new_cfg)
5154 {
5155 	bool need_reconfig = false;
5156 
5157 	/* Check if ETS configuration has changed */
5158 	if (memcmp(&new_cfg->etscfg,
5159 		   &old_cfg->etscfg,
5160 		   sizeof(new_cfg->etscfg))) {
5161 		/* If Priority Table has changed reconfig is needed */
5162 		if (memcmp(&new_cfg->etscfg.prioritytable,
5163 			   &old_cfg->etscfg.prioritytable,
5164 			   sizeof(new_cfg->etscfg.prioritytable))) {
5165 			need_reconfig = true;
5166 			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5167 		}
5168 
5169 		if (memcmp(&new_cfg->etscfg.tcbwtable,
5170 			   &old_cfg->etscfg.tcbwtable,
5171 			   sizeof(new_cfg->etscfg.tcbwtable)))
5172 			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5173 
5174 		if (memcmp(&new_cfg->etscfg.tsatable,
5175 			   &old_cfg->etscfg.tsatable,
5176 			   sizeof(new_cfg->etscfg.tsatable)))
5177 			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5178 	}
5179 
5180 	/* Check if PFC configuration has changed */
5181 	if (memcmp(&new_cfg->pfc,
5182 		   &old_cfg->pfc,
5183 		   sizeof(new_cfg->pfc))) {
5184 		need_reconfig = true;
5185 		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5186 	}
5187 
5188 	/* Check if APP Table has changed */
5189 	if (memcmp(&new_cfg->app,
5190 		   &old_cfg->app,
5191 		   sizeof(new_cfg->app))) {
5192 		need_reconfig = true;
5193 		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5194 	}
5195 
5196 	dev_dbg(&pf->pdev->dev, "%s: need_reconfig=%d\n", __func__,
5197 		need_reconfig);
5198 	return need_reconfig;
5199 }
5200 
5201 /**
5202  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5203  * @pf: board private structure
5204  * @e: event info posted on ARQ
5205  **/
5206 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5207 				  struct i40e_arq_event_info *e)
5208 {
5209 	struct i40e_aqc_lldp_get_mib *mib =
5210 		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5211 	struct i40e_hw *hw = &pf->hw;
5212 	struct i40e_dcbx_config tmp_dcbx_cfg;
5213 	bool need_reconfig = false;
5214 	int ret = 0;
5215 	u8 type;
5216 
5217 	/* Not DCB capable or capability disabled */
5218 	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5219 		return ret;
5220 
5221 	/* Ignore if event is not for Nearest Bridge */
5222 	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5223 		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5224 	dev_dbg(&pf->pdev->dev,
5225 		"%s: LLDP event mib bridge type 0x%x\n", __func__, type);
5226 	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5227 		return ret;
5228 
5229 	/* Check MIB Type and return if event for Remote MIB update */
5230 	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5231 	dev_dbg(&pf->pdev->dev,
5232 		"%s: LLDP event mib type %s\n", __func__,
5233 		type ? "remote" : "local");
5234 	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5235 		/* Update the remote cached instance and return */
5236 		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5237 				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5238 				&hw->remote_dcbx_config);
5239 		goto exit;
5240 	}
5241 
5242 	/* Store the old configuration */
5243 	tmp_dcbx_cfg = hw->local_dcbx_config;
5244 
5245 	/* Reset the old DCBx configuration data */
5246 	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5247 	/* Get updated DCBX data from firmware */
5248 	ret = i40e_get_dcb_config(&pf->hw);
5249 	if (ret) {
5250 		dev_info(&pf->pdev->dev, "Failed querying DCB configuration data from firmware.\n");
5251 		goto exit;
5252 	}
5253 
5254 	/* No change detected in DCBX configs */
5255 	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5256 		    sizeof(tmp_dcbx_cfg))) {
5257 		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5258 		goto exit;
5259 	}
5260 
5261 	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5262 					       &hw->local_dcbx_config);
5263 
5264 	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5265 
5266 	if (!need_reconfig)
5267 		goto exit;
5268 
5269 	/* Enable DCB tagging only when more than one TC */
5270 	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5271 		pf->flags |= I40E_FLAG_DCB_ENABLED;
5272 	else
5273 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5274 
5275 	set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5276 	/* Reconfiguration needed quiesce all VSIs */
5277 	i40e_pf_quiesce_all_vsi(pf);
5278 
5279 	/* Changes in configuration update VEB/VSI */
5280 	i40e_dcb_reconfigure(pf);
5281 
5282 	ret = i40e_resume_port_tx(pf);
5283 
5284 	clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5285 	/* In case of error no point in resuming VSIs */
5286 	if (ret)
5287 		goto exit;
5288 
5289 	/* Wait for the PF's Tx queues to be disabled */
5290 	ret = i40e_pf_wait_txq_disabled(pf);
5291 	if (ret) {
5292 		/* Schedule PF reset to recover */
5293 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5294 		i40e_service_event_schedule(pf);
5295 	} else {
5296 		i40e_pf_unquiesce_all_vsi(pf);
5297 	}
5298 
5299 exit:
5300 	return ret;
5301 }
5302 #endif /* CONFIG_I40E_DCB */
5303 
5304 /**
5305  * i40e_do_reset_safe - Protected reset path for userland calls.
5306  * @pf: board private structure
5307  * @reset_flags: which reset is requested
5308  *
5309  **/
5310 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5311 {
5312 	rtnl_lock();
5313 	i40e_do_reset(pf, reset_flags);
5314 	rtnl_unlock();
5315 }
5316 
5317 /**
5318  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5319  * @pf: board private structure
5320  * @e: event info posted on ARQ
5321  *
5322  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5323  * and VF queues
5324  **/
5325 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5326 					   struct i40e_arq_event_info *e)
5327 {
5328 	struct i40e_aqc_lan_overflow *data =
5329 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5330 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
5331 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5332 	struct i40e_hw *hw = &pf->hw;
5333 	struct i40e_vf *vf;
5334 	u16 vf_id;
5335 
5336 	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5337 		queue, qtx_ctl);
5338 
5339 	/* Queue belongs to VF, find the VF and issue VF reset */
5340 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5341 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5342 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5343 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5344 		vf_id -= hw->func_caps.vf_base_id;
5345 		vf = &pf->vf[vf_id];
5346 		i40e_vc_notify_vf_reset(vf);
5347 		/* Allow VF to process pending reset notification */
5348 		msleep(20);
5349 		i40e_reset_vf(vf, false);
5350 	}
5351 }
5352 
5353 /**
5354  * i40e_service_event_complete - Finish up the service event
5355  * @pf: board private structure
5356  **/
5357 static void i40e_service_event_complete(struct i40e_pf *pf)
5358 {
5359 	BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5360 
5361 	/* flush memory to make sure state is correct before next watchog */
5362 	smp_mb__before_atomic();
5363 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5364 }
5365 
5366 /**
5367  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5368  * @pf: board private structure
5369  **/
5370 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5371 {
5372 	u32 val, fcnt_prog;
5373 
5374 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5375 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5376 	return fcnt_prog;
5377 }
5378 
5379 /**
5380  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5381  * @pf: board private structure
5382  **/
5383 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5384 {
5385 	u32 val, fcnt_prog;
5386 
5387 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5388 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5389 		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5390 		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5391 	return fcnt_prog;
5392 }
5393 
5394 /**
5395  * i40e_get_global_fd_count - Get total FD filters programmed on device
5396  * @pf: board private structure
5397  **/
5398 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5399 {
5400 	u32 val, fcnt_prog;
5401 
5402 	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5403 	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5404 		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5405 		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5406 	return fcnt_prog;
5407 }
5408 
5409 /**
5410  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5411  * @pf: board private structure
5412  **/
5413 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5414 {
5415 	u32 fcnt_prog, fcnt_avail;
5416 
5417 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5418 		return;
5419 
5420 	/* Check if, FD SB or ATR was auto disabled and if there is enough room
5421 	 * to re-enable
5422 	 */
5423 	fcnt_prog = i40e_get_global_fd_count(pf);
5424 	fcnt_avail = pf->fdir_pf_filter_count;
5425 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5426 	    (pf->fd_add_err == 0) ||
5427 	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5428 		if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5429 		    (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5430 			pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5431 			dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5432 		}
5433 	}
5434 	/* Wait for some more space to be available to turn on ATR */
5435 	if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5436 		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5437 		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5438 			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5439 			dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5440 		}
5441 	}
5442 }
5443 
5444 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5445 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5446 /**
5447  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5448  * @pf: board private structure
5449  **/
5450 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5451 {
5452 	unsigned long min_flush_time;
5453 	int flush_wait_retry = 50;
5454 	bool disable_atr = false;
5455 	int fd_room;
5456 	int reg;
5457 
5458 	if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5459 		return;
5460 
5461 	if (time_after(jiffies, pf->fd_flush_timestamp +
5462 				(I40E_MIN_FD_FLUSH_INTERVAL * HZ))) {
5463 		/* If the flush is happening too quick and we have mostly
5464 		 * SB rules we should not re-enable ATR for some time.
5465 		 */
5466 		min_flush_time = pf->fd_flush_timestamp
5467 				+ (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5468 		fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5469 
5470 		if (!(time_after(jiffies, min_flush_time)) &&
5471 		    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5472 			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5473 			disable_atr = true;
5474 		}
5475 
5476 		pf->fd_flush_timestamp = jiffies;
5477 		pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5478 		/* flush all filters */
5479 		wr32(&pf->hw, I40E_PFQF_CTL_1,
5480 		     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5481 		i40e_flush(&pf->hw);
5482 		pf->fd_flush_cnt++;
5483 		pf->fd_add_err = 0;
5484 		do {
5485 			/* Check FD flush status every 5-6msec */
5486 			usleep_range(5000, 6000);
5487 			reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5488 			if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5489 				break;
5490 		} while (flush_wait_retry--);
5491 		if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5492 			dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5493 		} else {
5494 			/* replay sideband filters */
5495 			i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5496 			if (!disable_atr)
5497 				pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5498 			clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5499 			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5500 		}
5501 	}
5502 }
5503 
5504 /**
5505  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5506  * @pf: board private structure
5507  **/
5508 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5509 {
5510 	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5511 }
5512 
5513 /* We can see up to 256 filter programming desc in transit if the filters are
5514  * being applied really fast; before we see the first
5515  * filter miss error on Rx queue 0. Accumulating enough error messages before
5516  * reacting will make sure we don't cause flush too often.
5517  */
5518 #define I40E_MAX_FD_PROGRAM_ERROR 256
5519 
5520 /**
5521  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
5522  * @pf: board private structure
5523  **/
5524 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
5525 {
5526 
5527 	/* if interface is down do nothing */
5528 	if (test_bit(__I40E_DOWN, &pf->state))
5529 		return;
5530 
5531 	if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5532 		return;
5533 
5534 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5535 		i40e_fdir_flush_and_replay(pf);
5536 
5537 	i40e_fdir_check_and_reenable(pf);
5538 
5539 }
5540 
5541 /**
5542  * i40e_vsi_link_event - notify VSI of a link event
5543  * @vsi: vsi to be notified
5544  * @link_up: link up or down
5545  **/
5546 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
5547 {
5548 	if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
5549 		return;
5550 
5551 	switch (vsi->type) {
5552 	case I40E_VSI_MAIN:
5553 #ifdef I40E_FCOE
5554 	case I40E_VSI_FCOE:
5555 #endif
5556 		if (!vsi->netdev || !vsi->netdev_registered)
5557 			break;
5558 
5559 		if (link_up) {
5560 			netif_carrier_on(vsi->netdev);
5561 			netif_tx_wake_all_queues(vsi->netdev);
5562 		} else {
5563 			netif_carrier_off(vsi->netdev);
5564 			netif_tx_stop_all_queues(vsi->netdev);
5565 		}
5566 		break;
5567 
5568 	case I40E_VSI_SRIOV:
5569 	case I40E_VSI_VMDQ2:
5570 	case I40E_VSI_CTRL:
5571 	case I40E_VSI_MIRROR:
5572 	default:
5573 		/* there is no notification for other VSIs */
5574 		break;
5575 	}
5576 }
5577 
5578 /**
5579  * i40e_veb_link_event - notify elements on the veb of a link event
5580  * @veb: veb to be notified
5581  * @link_up: link up or down
5582  **/
5583 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
5584 {
5585 	struct i40e_pf *pf;
5586 	int i;
5587 
5588 	if (!veb || !veb->pf)
5589 		return;
5590 	pf = veb->pf;
5591 
5592 	/* depth first... */
5593 	for (i = 0; i < I40E_MAX_VEB; i++)
5594 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
5595 			i40e_veb_link_event(pf->veb[i], link_up);
5596 
5597 	/* ... now the local VSIs */
5598 	for (i = 0; i < pf->num_alloc_vsi; i++)
5599 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
5600 			i40e_vsi_link_event(pf->vsi[i], link_up);
5601 }
5602 
5603 /**
5604  * i40e_link_event - Update netif_carrier status
5605  * @pf: board private structure
5606  **/
5607 static void i40e_link_event(struct i40e_pf *pf)
5608 {
5609 	bool new_link, old_link;
5610 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5611 	u8 new_link_speed, old_link_speed;
5612 
5613 	/* set this to force the get_link_status call to refresh state */
5614 	pf->hw.phy.get_link_info = true;
5615 
5616 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
5617 	new_link = i40e_get_link_status(&pf->hw);
5618 	old_link_speed = pf->hw.phy.link_info_old.link_speed;
5619 	new_link_speed = pf->hw.phy.link_info.link_speed;
5620 
5621 	if (new_link == old_link &&
5622 	    new_link_speed == old_link_speed &&
5623 	    (test_bit(__I40E_DOWN, &vsi->state) ||
5624 	     new_link == netif_carrier_ok(vsi->netdev)))
5625 		return;
5626 
5627 	if (!test_bit(__I40E_DOWN, &vsi->state))
5628 		i40e_print_link_message(vsi, new_link);
5629 
5630 	/* Notify the base of the switch tree connected to
5631 	 * the link.  Floating VEBs are not notified.
5632 	 */
5633 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
5634 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
5635 	else
5636 		i40e_vsi_link_event(vsi, new_link);
5637 
5638 	if (pf->vf)
5639 		i40e_vc_notify_link_state(pf);
5640 
5641 	if (pf->flags & I40E_FLAG_PTP)
5642 		i40e_ptp_set_increment(pf);
5643 }
5644 
5645 /**
5646  * i40e_check_hang_subtask - Check for hung queues and dropped interrupts
5647  * @pf: board private structure
5648  *
5649  * Set the per-queue flags to request a check for stuck queues in the irq
5650  * clean functions, then force interrupts to be sure the irq clean is called.
5651  **/
5652 static void i40e_check_hang_subtask(struct i40e_pf *pf)
5653 {
5654 	int i, v;
5655 
5656 	/* If we're down or resetting, just bail */
5657 	if (test_bit(__I40E_DOWN, &pf->state) ||
5658 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
5659 		return;
5660 
5661 	/* for each VSI/netdev
5662 	 *     for each Tx queue
5663 	 *         set the check flag
5664 	 *     for each q_vector
5665 	 *         force an interrupt
5666 	 */
5667 	for (v = 0; v < pf->num_alloc_vsi; v++) {
5668 		struct i40e_vsi *vsi = pf->vsi[v];
5669 		int armed = 0;
5670 
5671 		if (!pf->vsi[v] ||
5672 		    test_bit(__I40E_DOWN, &vsi->state) ||
5673 		    (vsi->netdev && !netif_carrier_ok(vsi->netdev)))
5674 			continue;
5675 
5676 		for (i = 0; i < vsi->num_queue_pairs; i++) {
5677 			set_check_for_tx_hang(vsi->tx_rings[i]);
5678 			if (test_bit(__I40E_HANG_CHECK_ARMED,
5679 				     &vsi->tx_rings[i]->state))
5680 				armed++;
5681 		}
5682 
5683 		if (armed) {
5684 			if (!(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
5685 				wr32(&vsi->back->hw, I40E_PFINT_DYN_CTL0,
5686 				     (I40E_PFINT_DYN_CTL0_INTENA_MASK |
5687 				      I40E_PFINT_DYN_CTL0_SWINT_TRIG_MASK |
5688 				      I40E_PFINT_DYN_CTL0_ITR_INDX_MASK |
5689 				      I40E_PFINT_DYN_CTL0_SW_ITR_INDX_ENA_MASK |
5690 				      I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK));
5691 			} else {
5692 				u16 vec = vsi->base_vector - 1;
5693 				u32 val = (I40E_PFINT_DYN_CTLN_INTENA_MASK |
5694 				      I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK |
5695 				      I40E_PFINT_DYN_CTLN_ITR_INDX_MASK |
5696 				      I40E_PFINT_DYN_CTLN_SW_ITR_INDX_ENA_MASK |
5697 				      I40E_PFINT_DYN_CTLN_SW_ITR_INDX_MASK);
5698 				for (i = 0; i < vsi->num_q_vectors; i++, vec++)
5699 					wr32(&vsi->back->hw,
5700 					     I40E_PFINT_DYN_CTLN(vec), val);
5701 			}
5702 			i40e_flush(&vsi->back->hw);
5703 		}
5704 	}
5705 }
5706 
5707 /**
5708  * i40e_watchdog_subtask - periodic checks not using event driven response
5709  * @pf: board private structure
5710  **/
5711 static void i40e_watchdog_subtask(struct i40e_pf *pf)
5712 {
5713 	int i;
5714 
5715 	/* if interface is down do nothing */
5716 	if (test_bit(__I40E_DOWN, &pf->state) ||
5717 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
5718 		return;
5719 
5720 	/* make sure we don't do these things too often */
5721 	if (time_before(jiffies, (pf->service_timer_previous +
5722 				  pf->service_timer_period)))
5723 		return;
5724 	pf->service_timer_previous = jiffies;
5725 
5726 	i40e_check_hang_subtask(pf);
5727 	i40e_link_event(pf);
5728 
5729 	/* Update the stats for active netdevs so the network stack
5730 	 * can look at updated numbers whenever it cares to
5731 	 */
5732 	for (i = 0; i < pf->num_alloc_vsi; i++)
5733 		if (pf->vsi[i] && pf->vsi[i]->netdev)
5734 			i40e_update_stats(pf->vsi[i]);
5735 
5736 	/* Update the stats for the active switching components */
5737 	for (i = 0; i < I40E_MAX_VEB; i++)
5738 		if (pf->veb[i])
5739 			i40e_update_veb_stats(pf->veb[i]);
5740 
5741 	i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
5742 }
5743 
5744 /**
5745  * i40e_reset_subtask - Set up for resetting the device and driver
5746  * @pf: board private structure
5747  **/
5748 static void i40e_reset_subtask(struct i40e_pf *pf)
5749 {
5750 	u32 reset_flags = 0;
5751 
5752 	rtnl_lock();
5753 	if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
5754 		reset_flags |= (1 << __I40E_REINIT_REQUESTED);
5755 		clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
5756 	}
5757 	if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
5758 		reset_flags |= (1 << __I40E_PF_RESET_REQUESTED);
5759 		clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5760 	}
5761 	if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
5762 		reset_flags |= (1 << __I40E_CORE_RESET_REQUESTED);
5763 		clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
5764 	}
5765 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
5766 		reset_flags |= (1 << __I40E_GLOBAL_RESET_REQUESTED);
5767 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
5768 	}
5769 	if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
5770 		reset_flags |= (1 << __I40E_DOWN_REQUESTED);
5771 		clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
5772 	}
5773 
5774 	/* If there's a recovery already waiting, it takes
5775 	 * precedence before starting a new reset sequence.
5776 	 */
5777 	if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
5778 		i40e_handle_reset_warning(pf);
5779 		goto unlock;
5780 	}
5781 
5782 	/* If we're already down or resetting, just bail */
5783 	if (reset_flags &&
5784 	    !test_bit(__I40E_DOWN, &pf->state) &&
5785 	    !test_bit(__I40E_CONFIG_BUSY, &pf->state))
5786 		i40e_do_reset(pf, reset_flags);
5787 
5788 unlock:
5789 	rtnl_unlock();
5790 }
5791 
5792 /**
5793  * i40e_handle_link_event - Handle link event
5794  * @pf: board private structure
5795  * @e: event info posted on ARQ
5796  **/
5797 static void i40e_handle_link_event(struct i40e_pf *pf,
5798 				   struct i40e_arq_event_info *e)
5799 {
5800 	struct i40e_hw *hw = &pf->hw;
5801 	struct i40e_aqc_get_link_status *status =
5802 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
5803 
5804 	/* save off old link status information */
5805 	hw->phy.link_info_old = hw->phy.link_info;
5806 
5807 	/* Do a new status request to re-enable LSE reporting
5808 	 * and load new status information into the hw struct
5809 	 * This completely ignores any state information
5810 	 * in the ARQ event info, instead choosing to always
5811 	 * issue the AQ update link status command.
5812 	 */
5813 	i40e_link_event(pf);
5814 
5815 	/* check for unqualified module, if link is down */
5816 	if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
5817 	    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
5818 	    (!(status->link_info & I40E_AQ_LINK_UP)))
5819 		dev_err(&pf->pdev->dev,
5820 			"The driver failed to link because an unqualified module was detected.\n");
5821 }
5822 
5823 /**
5824  * i40e_clean_adminq_subtask - Clean the AdminQ rings
5825  * @pf: board private structure
5826  **/
5827 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
5828 {
5829 	struct i40e_arq_event_info event;
5830 	struct i40e_hw *hw = &pf->hw;
5831 	u16 pending, i = 0;
5832 	i40e_status ret;
5833 	u16 opcode;
5834 	u32 oldval;
5835 	u32 val;
5836 
5837 	/* Do not run clean AQ when PF reset fails */
5838 	if (test_bit(__I40E_RESET_FAILED, &pf->state))
5839 		return;
5840 
5841 	/* check for error indications */
5842 	val = rd32(&pf->hw, pf->hw.aq.arq.len);
5843 	oldval = val;
5844 	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
5845 		dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
5846 		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
5847 	}
5848 	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
5849 		dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
5850 		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
5851 	}
5852 	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
5853 		dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
5854 		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
5855 	}
5856 	if (oldval != val)
5857 		wr32(&pf->hw, pf->hw.aq.arq.len, val);
5858 
5859 	val = rd32(&pf->hw, pf->hw.aq.asq.len);
5860 	oldval = val;
5861 	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
5862 		dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
5863 		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
5864 	}
5865 	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
5866 		dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
5867 		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
5868 	}
5869 	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
5870 		dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
5871 		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
5872 	}
5873 	if (oldval != val)
5874 		wr32(&pf->hw, pf->hw.aq.asq.len, val);
5875 
5876 	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
5877 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
5878 	if (!event.msg_buf)
5879 		return;
5880 
5881 	do {
5882 		ret = i40e_clean_arq_element(hw, &event, &pending);
5883 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
5884 			break;
5885 		else if (ret) {
5886 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
5887 			break;
5888 		}
5889 
5890 		opcode = le16_to_cpu(event.desc.opcode);
5891 		switch (opcode) {
5892 
5893 		case i40e_aqc_opc_get_link_status:
5894 			i40e_handle_link_event(pf, &event);
5895 			break;
5896 		case i40e_aqc_opc_send_msg_to_pf:
5897 			ret = i40e_vc_process_vf_msg(pf,
5898 					le16_to_cpu(event.desc.retval),
5899 					le32_to_cpu(event.desc.cookie_high),
5900 					le32_to_cpu(event.desc.cookie_low),
5901 					event.msg_buf,
5902 					event.msg_len);
5903 			break;
5904 		case i40e_aqc_opc_lldp_update_mib:
5905 			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
5906 #ifdef CONFIG_I40E_DCB
5907 			rtnl_lock();
5908 			ret = i40e_handle_lldp_event(pf, &event);
5909 			rtnl_unlock();
5910 #endif /* CONFIG_I40E_DCB */
5911 			break;
5912 		case i40e_aqc_opc_event_lan_overflow:
5913 			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
5914 			i40e_handle_lan_overflow_event(pf, &event);
5915 			break;
5916 		case i40e_aqc_opc_send_msg_to_peer:
5917 			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
5918 			break;
5919 		case i40e_aqc_opc_nvm_erase:
5920 		case i40e_aqc_opc_nvm_update:
5921 			i40e_debug(&pf->hw, I40E_DEBUG_NVM, "ARQ NVM operation completed\n");
5922 			break;
5923 		default:
5924 			dev_info(&pf->pdev->dev,
5925 				 "ARQ Error: Unknown event 0x%04x received\n",
5926 				 opcode);
5927 			break;
5928 		}
5929 	} while (pending && (i++ < pf->adminq_work_limit));
5930 
5931 	clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
5932 	/* re-enable Admin queue interrupt cause */
5933 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
5934 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
5935 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
5936 	i40e_flush(hw);
5937 
5938 	kfree(event.msg_buf);
5939 }
5940 
5941 /**
5942  * i40e_verify_eeprom - make sure eeprom is good to use
5943  * @pf: board private structure
5944  **/
5945 static void i40e_verify_eeprom(struct i40e_pf *pf)
5946 {
5947 	int err;
5948 
5949 	err = i40e_diag_eeprom_test(&pf->hw);
5950 	if (err) {
5951 		/* retry in case of garbage read */
5952 		err = i40e_diag_eeprom_test(&pf->hw);
5953 		if (err) {
5954 			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
5955 				 err);
5956 			set_bit(__I40E_BAD_EEPROM, &pf->state);
5957 		}
5958 	}
5959 
5960 	if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
5961 		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
5962 		clear_bit(__I40E_BAD_EEPROM, &pf->state);
5963 	}
5964 }
5965 
5966 /**
5967  * i40e_enable_pf_switch_lb
5968  * @pf: pointer to the PF structure
5969  *
5970  * enable switch loop back or die - no point in a return value
5971  **/
5972 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
5973 {
5974 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5975 	struct i40e_vsi_context ctxt;
5976 	int aq_ret;
5977 
5978 	ctxt.seid = pf->main_vsi_seid;
5979 	ctxt.pf_num = pf->hw.pf_id;
5980 	ctxt.vf_num = 0;
5981 	aq_ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
5982 	if (aq_ret) {
5983 		dev_info(&pf->pdev->dev,
5984 			 "%s couldn't get PF vsi config, err %d, aq_err %d\n",
5985 			 __func__, aq_ret, pf->hw.aq.asq_last_status);
5986 		return;
5987 	}
5988 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
5989 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5990 	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5991 
5992 	aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5993 	if (aq_ret) {
5994 		dev_info(&pf->pdev->dev,
5995 			 "%s: update vsi switch failed, aq_err=%d\n",
5996 			 __func__, vsi->back->hw.aq.asq_last_status);
5997 	}
5998 }
5999 
6000 /**
6001  * i40e_disable_pf_switch_lb
6002  * @pf: pointer to the PF structure
6003  *
6004  * disable switch loop back or die - no point in a return value
6005  **/
6006 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6007 {
6008 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6009 	struct i40e_vsi_context ctxt;
6010 	int aq_ret;
6011 
6012 	ctxt.seid = pf->main_vsi_seid;
6013 	ctxt.pf_num = pf->hw.pf_id;
6014 	ctxt.vf_num = 0;
6015 	aq_ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6016 	if (aq_ret) {
6017 		dev_info(&pf->pdev->dev,
6018 			 "%s couldn't get PF vsi config, err %d, aq_err %d\n",
6019 			 __func__, aq_ret, pf->hw.aq.asq_last_status);
6020 		return;
6021 	}
6022 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6023 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6024 	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6025 
6026 	aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6027 	if (aq_ret) {
6028 		dev_info(&pf->pdev->dev,
6029 			 "%s: update vsi switch failed, aq_err=%d\n",
6030 			 __func__, vsi->back->hw.aq.asq_last_status);
6031 	}
6032 }
6033 
6034 /**
6035  * i40e_config_bridge_mode - Configure the HW bridge mode
6036  * @veb: pointer to the bridge instance
6037  *
6038  * Configure the loop back mode for the LAN VSI that is downlink to the
6039  * specified HW bridge instance. It is expected this function is called
6040  * when a new HW bridge is instantiated.
6041  **/
6042 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6043 {
6044 	struct i40e_pf *pf = veb->pf;
6045 
6046 	dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6047 		 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6048 	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6049 		i40e_disable_pf_switch_lb(pf);
6050 	else
6051 		i40e_enable_pf_switch_lb(pf);
6052 }
6053 
6054 /**
6055  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6056  * @veb: pointer to the VEB instance
6057  *
6058  * This is a recursive function that first builds the attached VSIs then
6059  * recurses in to build the next layer of VEB.  We track the connections
6060  * through our own index numbers because the seid's from the HW could
6061  * change across the reset.
6062  **/
6063 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6064 {
6065 	struct i40e_vsi *ctl_vsi = NULL;
6066 	struct i40e_pf *pf = veb->pf;
6067 	int v, veb_idx;
6068 	int ret;
6069 
6070 	/* build VSI that owns this VEB, temporarily attached to base VEB */
6071 	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6072 		if (pf->vsi[v] &&
6073 		    pf->vsi[v]->veb_idx == veb->idx &&
6074 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6075 			ctl_vsi = pf->vsi[v];
6076 			break;
6077 		}
6078 	}
6079 	if (!ctl_vsi) {
6080 		dev_info(&pf->pdev->dev,
6081 			 "missing owner VSI for veb_idx %d\n", veb->idx);
6082 		ret = -ENOENT;
6083 		goto end_reconstitute;
6084 	}
6085 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
6086 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6087 	ret = i40e_add_vsi(ctl_vsi);
6088 	if (ret) {
6089 		dev_info(&pf->pdev->dev,
6090 			 "rebuild of owner VSI failed: %d\n", ret);
6091 		goto end_reconstitute;
6092 	}
6093 	i40e_vsi_reset_stats(ctl_vsi);
6094 
6095 	/* create the VEB in the switch and move the VSI onto the VEB */
6096 	ret = i40e_add_veb(veb, ctl_vsi);
6097 	if (ret)
6098 		goto end_reconstitute;
6099 
6100 	i40e_config_bridge_mode(veb);
6101 
6102 	/* create the remaining VSIs attached to this VEB */
6103 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6104 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6105 			continue;
6106 
6107 		if (pf->vsi[v]->veb_idx == veb->idx) {
6108 			struct i40e_vsi *vsi = pf->vsi[v];
6109 			vsi->uplink_seid = veb->seid;
6110 			ret = i40e_add_vsi(vsi);
6111 			if (ret) {
6112 				dev_info(&pf->pdev->dev,
6113 					 "rebuild of vsi_idx %d failed: %d\n",
6114 					 v, ret);
6115 				goto end_reconstitute;
6116 			}
6117 			i40e_vsi_reset_stats(vsi);
6118 		}
6119 	}
6120 
6121 	/* create any VEBs attached to this VEB - RECURSION */
6122 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6123 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6124 			pf->veb[veb_idx]->uplink_seid = veb->seid;
6125 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6126 			if (ret)
6127 				break;
6128 		}
6129 	}
6130 
6131 end_reconstitute:
6132 	return ret;
6133 }
6134 
6135 /**
6136  * i40e_get_capabilities - get info about the HW
6137  * @pf: the PF struct
6138  **/
6139 static int i40e_get_capabilities(struct i40e_pf *pf)
6140 {
6141 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6142 	u16 data_size;
6143 	int buf_len;
6144 	int err;
6145 
6146 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6147 	do {
6148 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
6149 		if (!cap_buf)
6150 			return -ENOMEM;
6151 
6152 		/* this loads the data into the hw struct for us */
6153 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6154 					    &data_size,
6155 					    i40e_aqc_opc_list_func_capabilities,
6156 					    NULL);
6157 		/* data loaded, buffer no longer needed */
6158 		kfree(cap_buf);
6159 
6160 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6161 			/* retry with a larger buffer */
6162 			buf_len = data_size;
6163 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6164 			dev_info(&pf->pdev->dev,
6165 				 "capability discovery failed: aq=%d\n",
6166 				 pf->hw.aq.asq_last_status);
6167 			return -ENODEV;
6168 		}
6169 	} while (err);
6170 
6171 	if (((pf->hw.aq.fw_maj_ver == 2) && (pf->hw.aq.fw_min_ver < 22)) ||
6172 	    (pf->hw.aq.fw_maj_ver < 2)) {
6173 		pf->hw.func_caps.num_msix_vectors++;
6174 		pf->hw.func_caps.num_msix_vectors_vf++;
6175 	}
6176 
6177 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
6178 		dev_info(&pf->pdev->dev,
6179 			 "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",
6180 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6181 			 pf->hw.func_caps.num_msix_vectors,
6182 			 pf->hw.func_caps.num_msix_vectors_vf,
6183 			 pf->hw.func_caps.fd_filters_guaranteed,
6184 			 pf->hw.func_caps.fd_filters_best_effort,
6185 			 pf->hw.func_caps.num_tx_qp,
6186 			 pf->hw.func_caps.num_vsis);
6187 
6188 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6189 		       + pf->hw.func_caps.num_vfs)
6190 	if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6191 		dev_info(&pf->pdev->dev,
6192 			 "got num_vsis %d, setting num_vsis to %d\n",
6193 			 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6194 		pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6195 	}
6196 
6197 	return 0;
6198 }
6199 
6200 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6201 
6202 /**
6203  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6204  * @pf: board private structure
6205  **/
6206 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6207 {
6208 	struct i40e_vsi *vsi;
6209 	int i;
6210 
6211 	/* quick workaround for an NVM issue that leaves a critical register
6212 	 * uninitialized
6213 	 */
6214 	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6215 		static const u32 hkey[] = {
6216 			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6217 			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6218 			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6219 			0x95b3a76d};
6220 
6221 		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6222 			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6223 	}
6224 
6225 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6226 		return;
6227 
6228 	/* find existing VSI and see if it needs configuring */
6229 	vsi = NULL;
6230 	for (i = 0; i < pf->num_alloc_vsi; i++) {
6231 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6232 			vsi = pf->vsi[i];
6233 			break;
6234 		}
6235 	}
6236 
6237 	/* create a new VSI if none exists */
6238 	if (!vsi) {
6239 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6240 				     pf->vsi[pf->lan_vsi]->seid, 0);
6241 		if (!vsi) {
6242 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6243 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6244 			return;
6245 		}
6246 	}
6247 
6248 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6249 }
6250 
6251 /**
6252  * i40e_fdir_teardown - release the Flow Director resources
6253  * @pf: board private structure
6254  **/
6255 static void i40e_fdir_teardown(struct i40e_pf *pf)
6256 {
6257 	int i;
6258 
6259 	i40e_fdir_filter_exit(pf);
6260 	for (i = 0; i < pf->num_alloc_vsi; i++) {
6261 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6262 			i40e_vsi_release(pf->vsi[i]);
6263 			break;
6264 		}
6265 	}
6266 }
6267 
6268 /**
6269  * i40e_prep_for_reset - prep for the core to reset
6270  * @pf: board private structure
6271  *
6272  * Close up the VFs and other things in prep for PF Reset.
6273   **/
6274 static void i40e_prep_for_reset(struct i40e_pf *pf)
6275 {
6276 	struct i40e_hw *hw = &pf->hw;
6277 	i40e_status ret = 0;
6278 	u32 v;
6279 
6280 	clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6281 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6282 		return;
6283 
6284 	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6285 
6286 	/* quiesce the VSIs and their queues that are not already DOWN */
6287 	i40e_pf_quiesce_all_vsi(pf);
6288 
6289 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6290 		if (pf->vsi[v])
6291 			pf->vsi[v]->seid = 0;
6292 	}
6293 
6294 	i40e_shutdown_adminq(&pf->hw);
6295 
6296 	/* call shutdown HMC */
6297 	if (hw->hmc.hmc_obj) {
6298 		ret = i40e_shutdown_lan_hmc(hw);
6299 		if (ret)
6300 			dev_warn(&pf->pdev->dev,
6301 				 "shutdown_lan_hmc failed: %d\n", ret);
6302 	}
6303 }
6304 
6305 /**
6306  * i40e_send_version - update firmware with driver version
6307  * @pf: PF struct
6308  */
6309 static void i40e_send_version(struct i40e_pf *pf)
6310 {
6311 	struct i40e_driver_version dv;
6312 
6313 	dv.major_version = DRV_VERSION_MAJOR;
6314 	dv.minor_version = DRV_VERSION_MINOR;
6315 	dv.build_version = DRV_VERSION_BUILD;
6316 	dv.subbuild_version = 0;
6317 	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6318 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6319 }
6320 
6321 /**
6322  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6323  * @pf: board private structure
6324  * @reinit: if the Main VSI needs to re-initialized.
6325  **/
6326 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6327 {
6328 	struct i40e_hw *hw = &pf->hw;
6329 	u8 set_fc_aq_fail = 0;
6330 	i40e_status ret;
6331 	u32 v;
6332 
6333 	/* Now we wait for GRST to settle out.
6334 	 * We don't have to delete the VEBs or VSIs from the hw switch
6335 	 * because the reset will make them disappear.
6336 	 */
6337 	ret = i40e_pf_reset(hw);
6338 	if (ret) {
6339 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6340 		set_bit(__I40E_RESET_FAILED, &pf->state);
6341 		goto clear_recovery;
6342 	}
6343 	pf->pfr_count++;
6344 
6345 	if (test_bit(__I40E_DOWN, &pf->state))
6346 		goto clear_recovery;
6347 	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6348 
6349 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6350 	ret = i40e_init_adminq(&pf->hw);
6351 	if (ret) {
6352 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, %d\n", ret);
6353 		goto clear_recovery;
6354 	}
6355 
6356 	/* re-verify the eeprom if we just had an EMP reset */
6357 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6358 		i40e_verify_eeprom(pf);
6359 
6360 	i40e_clear_pxe_mode(hw);
6361 	ret = i40e_get_capabilities(pf);
6362 	if (ret) {
6363 		dev_info(&pf->pdev->dev, "i40e_get_capabilities failed, %d\n",
6364 			 ret);
6365 		goto end_core_reset;
6366 	}
6367 
6368 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6369 				hw->func_caps.num_rx_qp,
6370 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6371 	if (ret) {
6372 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6373 		goto end_core_reset;
6374 	}
6375 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6376 	if (ret) {
6377 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6378 		goto end_core_reset;
6379 	}
6380 
6381 #ifdef CONFIG_I40E_DCB
6382 	ret = i40e_init_pf_dcb(pf);
6383 	if (ret) {
6384 		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6385 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6386 		/* Continue without DCB enabled */
6387 	}
6388 #endif /* CONFIG_I40E_DCB */
6389 #ifdef I40E_FCOE
6390 	ret = i40e_init_pf_fcoe(pf);
6391 	if (ret)
6392 		dev_info(&pf->pdev->dev, "init_pf_fcoe failed: %d\n", ret);
6393 
6394 #endif
6395 	/* do basic switch setup */
6396 	ret = i40e_setup_pf_switch(pf, reinit);
6397 	if (ret)
6398 		goto end_core_reset;
6399 
6400 	/* driver is only interested in link up/down and module qualification
6401 	 * reports from firmware
6402 	 */
6403 	ret = i40e_aq_set_phy_int_mask(&pf->hw,
6404 				       I40E_AQ_EVENT_LINK_UPDOWN |
6405 				       I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
6406 	if (ret)
6407 		dev_info(&pf->pdev->dev, "set phy mask fail, aq_err %d\n", ret);
6408 
6409 	/* make sure our flow control settings are restored */
6410 	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6411 	if (ret)
6412 		dev_info(&pf->pdev->dev, "set fc fail, aq_err %d\n", ret);
6413 
6414 	/* Rebuild the VSIs and VEBs that existed before reset.
6415 	 * They are still in our local switch element arrays, so only
6416 	 * need to rebuild the switch model in the HW.
6417 	 *
6418 	 * If there were VEBs but the reconstitution failed, we'll try
6419 	 * try to recover minimal use by getting the basic PF VSI working.
6420 	 */
6421 	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6422 		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6423 		/* find the one VEB connected to the MAC, and find orphans */
6424 		for (v = 0; v < I40E_MAX_VEB; v++) {
6425 			if (!pf->veb[v])
6426 				continue;
6427 
6428 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6429 			    pf->veb[v]->uplink_seid == 0) {
6430 				ret = i40e_reconstitute_veb(pf->veb[v]);
6431 
6432 				if (!ret)
6433 					continue;
6434 
6435 				/* If Main VEB failed, we're in deep doodoo,
6436 				 * so give up rebuilding the switch and set up
6437 				 * for minimal rebuild of PF VSI.
6438 				 * If orphan failed, we'll report the error
6439 				 * but try to keep going.
6440 				 */
6441 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6442 					dev_info(&pf->pdev->dev,
6443 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6444 						 ret);
6445 					pf->vsi[pf->lan_vsi]->uplink_seid
6446 								= pf->mac_seid;
6447 					break;
6448 				} else if (pf->veb[v]->uplink_seid == 0) {
6449 					dev_info(&pf->pdev->dev,
6450 						 "rebuild of orphan VEB failed: %d\n",
6451 						 ret);
6452 				}
6453 			}
6454 		}
6455 	}
6456 
6457 	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6458 		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6459 		/* no VEB, so rebuild only the Main VSI */
6460 		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6461 		if (ret) {
6462 			dev_info(&pf->pdev->dev,
6463 				 "rebuild of Main VSI failed: %d\n", ret);
6464 			goto end_core_reset;
6465 		}
6466 	}
6467 
6468 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
6469 	    (pf->hw.aq.fw_maj_ver < 4)) {
6470 		msleep(75);
6471 		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6472 		if (ret)
6473 			dev_info(&pf->pdev->dev, "link restart failed, aq_err=%d\n",
6474 				 pf->hw.aq.asq_last_status);
6475 	}
6476 	/* reinit the misc interrupt */
6477 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6478 		ret = i40e_setup_misc_vector(pf);
6479 
6480 	/* restart the VSIs that were rebuilt and running before the reset */
6481 	i40e_pf_unquiesce_all_vsi(pf);
6482 
6483 	if (pf->num_alloc_vfs) {
6484 		for (v = 0; v < pf->num_alloc_vfs; v++)
6485 			i40e_reset_vf(&pf->vf[v], true);
6486 	}
6487 
6488 	/* tell the firmware that we're starting */
6489 	i40e_send_version(pf);
6490 
6491 end_core_reset:
6492 	clear_bit(__I40E_RESET_FAILED, &pf->state);
6493 clear_recovery:
6494 	clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6495 }
6496 
6497 /**
6498  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6499  * @pf: board private structure
6500  *
6501  * Close up the VFs and other things in prep for a Core Reset,
6502  * then get ready to rebuild the world.
6503  **/
6504 static void i40e_handle_reset_warning(struct i40e_pf *pf)
6505 {
6506 	i40e_prep_for_reset(pf);
6507 	i40e_reset_and_rebuild(pf, false);
6508 }
6509 
6510 /**
6511  * i40e_handle_mdd_event
6512  * @pf: pointer to the PF structure
6513  *
6514  * Called from the MDD irq handler to identify possibly malicious vfs
6515  **/
6516 static void i40e_handle_mdd_event(struct i40e_pf *pf)
6517 {
6518 	struct i40e_hw *hw = &pf->hw;
6519 	bool mdd_detected = false;
6520 	bool pf_mdd_detected = false;
6521 	struct i40e_vf *vf;
6522 	u32 reg;
6523 	int i;
6524 
6525 	if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
6526 		return;
6527 
6528 	/* find what triggered the MDD event */
6529 	reg = rd32(hw, I40E_GL_MDET_TX);
6530 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
6531 		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
6532 				I40E_GL_MDET_TX_PF_NUM_SHIFT;
6533 		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
6534 				I40E_GL_MDET_TX_VF_NUM_SHIFT;
6535 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
6536 				I40E_GL_MDET_TX_EVENT_SHIFT;
6537 		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
6538 				I40E_GL_MDET_TX_QUEUE_SHIFT) -
6539 				pf->hw.func_caps.base_queue;
6540 		if (netif_msg_tx_err(pf))
6541 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
6542 				 event, queue, pf_num, vf_num);
6543 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
6544 		mdd_detected = true;
6545 	}
6546 	reg = rd32(hw, I40E_GL_MDET_RX);
6547 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
6548 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
6549 				I40E_GL_MDET_RX_FUNCTION_SHIFT;
6550 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
6551 				I40E_GL_MDET_RX_EVENT_SHIFT;
6552 		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
6553 				I40E_GL_MDET_RX_QUEUE_SHIFT) -
6554 				pf->hw.func_caps.base_queue;
6555 		if (netif_msg_rx_err(pf))
6556 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
6557 				 event, queue, func);
6558 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
6559 		mdd_detected = true;
6560 	}
6561 
6562 	if (mdd_detected) {
6563 		reg = rd32(hw, I40E_PF_MDET_TX);
6564 		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
6565 			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
6566 			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
6567 			pf_mdd_detected = true;
6568 		}
6569 		reg = rd32(hw, I40E_PF_MDET_RX);
6570 		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
6571 			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
6572 			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
6573 			pf_mdd_detected = true;
6574 		}
6575 		/* Queue belongs to the PF, initiate a reset */
6576 		if (pf_mdd_detected) {
6577 			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6578 			i40e_service_event_schedule(pf);
6579 		}
6580 	}
6581 
6582 	/* see if one of the VFs needs its hand slapped */
6583 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
6584 		vf = &(pf->vf[i]);
6585 		reg = rd32(hw, I40E_VP_MDET_TX(i));
6586 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
6587 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
6588 			vf->num_mdd_events++;
6589 			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
6590 				 i);
6591 		}
6592 
6593 		reg = rd32(hw, I40E_VP_MDET_RX(i));
6594 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
6595 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
6596 			vf->num_mdd_events++;
6597 			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
6598 				 i);
6599 		}
6600 
6601 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
6602 			dev_info(&pf->pdev->dev,
6603 				 "Too many MDD events on VF %d, disabled\n", i);
6604 			dev_info(&pf->pdev->dev,
6605 				 "Use PF Control I/F to re-enable the VF\n");
6606 			set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
6607 		}
6608 	}
6609 
6610 	/* re-enable mdd interrupt cause */
6611 	clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
6612 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
6613 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
6614 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
6615 	i40e_flush(hw);
6616 }
6617 
6618 #ifdef CONFIG_I40E_VXLAN
6619 /**
6620  * i40e_sync_vxlan_filters_subtask - Sync the VSI filter list with HW
6621  * @pf: board private structure
6622  **/
6623 static void i40e_sync_vxlan_filters_subtask(struct i40e_pf *pf)
6624 {
6625 	struct i40e_hw *hw = &pf->hw;
6626 	i40e_status ret;
6627 	__be16 port;
6628 	int i;
6629 
6630 	if (!(pf->flags & I40E_FLAG_VXLAN_FILTER_SYNC))
6631 		return;
6632 
6633 	pf->flags &= ~I40E_FLAG_VXLAN_FILTER_SYNC;
6634 
6635 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
6636 		if (pf->pending_vxlan_bitmap & (1 << i)) {
6637 			pf->pending_vxlan_bitmap &= ~(1 << i);
6638 			port = pf->vxlan_ports[i];
6639 			if (port)
6640 				ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
6641 						     I40E_AQC_TUNNEL_TYPE_VXLAN,
6642 						     NULL, NULL);
6643 			else
6644 				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
6645 
6646 			if (ret) {
6647 				dev_info(&pf->pdev->dev,
6648 					 "%s vxlan port %d, index %d failed, err %d, aq_err %d\n",
6649 					 port ? "add" : "delete",
6650 					 ntohs(port), i, ret,
6651 					 pf->hw.aq.asq_last_status);
6652 				pf->vxlan_ports[i] = 0;
6653 			}
6654 		}
6655 	}
6656 }
6657 
6658 #endif
6659 /**
6660  * i40e_service_task - Run the driver's async subtasks
6661  * @work: pointer to work_struct containing our data
6662  **/
6663 static void i40e_service_task(struct work_struct *work)
6664 {
6665 	struct i40e_pf *pf = container_of(work,
6666 					  struct i40e_pf,
6667 					  service_task);
6668 	unsigned long start_time = jiffies;
6669 
6670 	/* don't bother with service tasks if a reset is in progress */
6671 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
6672 		i40e_service_event_complete(pf);
6673 		return;
6674 	}
6675 
6676 	i40e_reset_subtask(pf);
6677 	i40e_handle_mdd_event(pf);
6678 	i40e_vc_process_vflr_event(pf);
6679 	i40e_watchdog_subtask(pf);
6680 	i40e_fdir_reinit_subtask(pf);
6681 	i40e_sync_filters_subtask(pf);
6682 #ifdef CONFIG_I40E_VXLAN
6683 	i40e_sync_vxlan_filters_subtask(pf);
6684 #endif
6685 	i40e_clean_adminq_subtask(pf);
6686 
6687 	i40e_service_event_complete(pf);
6688 
6689 	/* If the tasks have taken longer than one timer cycle or there
6690 	 * is more work to be done, reschedule the service task now
6691 	 * rather than wait for the timer to tick again.
6692 	 */
6693 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
6694 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)		 ||
6695 	    test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)		 ||
6696 	    test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
6697 		i40e_service_event_schedule(pf);
6698 }
6699 
6700 /**
6701  * i40e_service_timer - timer callback
6702  * @data: pointer to PF struct
6703  **/
6704 static void i40e_service_timer(unsigned long data)
6705 {
6706 	struct i40e_pf *pf = (struct i40e_pf *)data;
6707 
6708 	mod_timer(&pf->service_timer,
6709 		  round_jiffies(jiffies + pf->service_timer_period));
6710 	i40e_service_event_schedule(pf);
6711 }
6712 
6713 /**
6714  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
6715  * @vsi: the VSI being configured
6716  **/
6717 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
6718 {
6719 	struct i40e_pf *pf = vsi->back;
6720 
6721 	switch (vsi->type) {
6722 	case I40E_VSI_MAIN:
6723 		vsi->alloc_queue_pairs = pf->num_lan_qps;
6724 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6725 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6726 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6727 			vsi->num_q_vectors = pf->num_lan_msix;
6728 		else
6729 			vsi->num_q_vectors = 1;
6730 
6731 		break;
6732 
6733 	case I40E_VSI_FDIR:
6734 		vsi->alloc_queue_pairs = 1;
6735 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
6736 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6737 		vsi->num_q_vectors = 1;
6738 		break;
6739 
6740 	case I40E_VSI_VMDQ2:
6741 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
6742 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6743 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6744 		vsi->num_q_vectors = pf->num_vmdq_msix;
6745 		break;
6746 
6747 	case I40E_VSI_SRIOV:
6748 		vsi->alloc_queue_pairs = pf->num_vf_qps;
6749 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6750 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6751 		break;
6752 
6753 #ifdef I40E_FCOE
6754 	case I40E_VSI_FCOE:
6755 		vsi->alloc_queue_pairs = pf->num_fcoe_qps;
6756 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6757 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6758 		vsi->num_q_vectors = pf->num_fcoe_msix;
6759 		break;
6760 
6761 #endif /* I40E_FCOE */
6762 	default:
6763 		WARN_ON(1);
6764 		return -ENODATA;
6765 	}
6766 
6767 	return 0;
6768 }
6769 
6770 /**
6771  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
6772  * @type: VSI pointer
6773  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
6774  *
6775  * On error: returns error code (negative)
6776  * On success: returns 0
6777  **/
6778 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
6779 {
6780 	int size;
6781 	int ret = 0;
6782 
6783 	/* allocate memory for both Tx and Rx ring pointers */
6784 	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
6785 	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
6786 	if (!vsi->tx_rings)
6787 		return -ENOMEM;
6788 	vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
6789 
6790 	if (alloc_qvectors) {
6791 		/* allocate memory for q_vector pointers */
6792 		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
6793 		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
6794 		if (!vsi->q_vectors) {
6795 			ret = -ENOMEM;
6796 			goto err_vectors;
6797 		}
6798 	}
6799 	return ret;
6800 
6801 err_vectors:
6802 	kfree(vsi->tx_rings);
6803 	return ret;
6804 }
6805 
6806 /**
6807  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
6808  * @pf: board private structure
6809  * @type: type of VSI
6810  *
6811  * On error: returns error code (negative)
6812  * On success: returns vsi index in PF (positive)
6813  **/
6814 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
6815 {
6816 	int ret = -ENODEV;
6817 	struct i40e_vsi *vsi;
6818 	int vsi_idx;
6819 	int i;
6820 
6821 	/* Need to protect the allocation of the VSIs at the PF level */
6822 	mutex_lock(&pf->switch_mutex);
6823 
6824 	/* VSI list may be fragmented if VSI creation/destruction has
6825 	 * been happening.  We can afford to do a quick scan to look
6826 	 * for any free VSIs in the list.
6827 	 *
6828 	 * find next empty vsi slot, looping back around if necessary
6829 	 */
6830 	i = pf->next_vsi;
6831 	while (i < pf->num_alloc_vsi && pf->vsi[i])
6832 		i++;
6833 	if (i >= pf->num_alloc_vsi) {
6834 		i = 0;
6835 		while (i < pf->next_vsi && pf->vsi[i])
6836 			i++;
6837 	}
6838 
6839 	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
6840 		vsi_idx = i;             /* Found one! */
6841 	} else {
6842 		ret = -ENODEV;
6843 		goto unlock_pf;  /* out of VSI slots! */
6844 	}
6845 	pf->next_vsi = ++i;
6846 
6847 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
6848 	if (!vsi) {
6849 		ret = -ENOMEM;
6850 		goto unlock_pf;
6851 	}
6852 	vsi->type = type;
6853 	vsi->back = pf;
6854 	set_bit(__I40E_DOWN, &vsi->state);
6855 	vsi->flags = 0;
6856 	vsi->idx = vsi_idx;
6857 	vsi->rx_itr_setting = pf->rx_itr_default;
6858 	vsi->tx_itr_setting = pf->tx_itr_default;
6859 	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
6860 				pf->rss_table_size : 64;
6861 	vsi->netdev_registered = false;
6862 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
6863 	INIT_LIST_HEAD(&vsi->mac_filter_list);
6864 	vsi->irqs_ready = false;
6865 
6866 	ret = i40e_set_num_rings_in_vsi(vsi);
6867 	if (ret)
6868 		goto err_rings;
6869 
6870 	ret = i40e_vsi_alloc_arrays(vsi, true);
6871 	if (ret)
6872 		goto err_rings;
6873 
6874 	/* Setup default MSIX irq handler for VSI */
6875 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
6876 
6877 	pf->vsi[vsi_idx] = vsi;
6878 	ret = vsi_idx;
6879 	goto unlock_pf;
6880 
6881 err_rings:
6882 	pf->next_vsi = i - 1;
6883 	kfree(vsi);
6884 unlock_pf:
6885 	mutex_unlock(&pf->switch_mutex);
6886 	return ret;
6887 }
6888 
6889 /**
6890  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
6891  * @type: VSI pointer
6892  * @free_qvectors: a bool to specify if q_vectors need to be freed.
6893  *
6894  * On error: returns error code (negative)
6895  * On success: returns 0
6896  **/
6897 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
6898 {
6899 	/* free the ring and vector containers */
6900 	if (free_qvectors) {
6901 		kfree(vsi->q_vectors);
6902 		vsi->q_vectors = NULL;
6903 	}
6904 	kfree(vsi->tx_rings);
6905 	vsi->tx_rings = NULL;
6906 	vsi->rx_rings = NULL;
6907 }
6908 
6909 /**
6910  * i40e_vsi_clear - Deallocate the VSI provided
6911  * @vsi: the VSI being un-configured
6912  **/
6913 static int i40e_vsi_clear(struct i40e_vsi *vsi)
6914 {
6915 	struct i40e_pf *pf;
6916 
6917 	if (!vsi)
6918 		return 0;
6919 
6920 	if (!vsi->back)
6921 		goto free_vsi;
6922 	pf = vsi->back;
6923 
6924 	mutex_lock(&pf->switch_mutex);
6925 	if (!pf->vsi[vsi->idx]) {
6926 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
6927 			vsi->idx, vsi->idx, vsi, vsi->type);
6928 		goto unlock_vsi;
6929 	}
6930 
6931 	if (pf->vsi[vsi->idx] != vsi) {
6932 		dev_err(&pf->pdev->dev,
6933 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
6934 			pf->vsi[vsi->idx]->idx,
6935 			pf->vsi[vsi->idx],
6936 			pf->vsi[vsi->idx]->type,
6937 			vsi->idx, vsi, vsi->type);
6938 		goto unlock_vsi;
6939 	}
6940 
6941 	/* updates the PF for this cleared vsi */
6942 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
6943 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
6944 
6945 	i40e_vsi_free_arrays(vsi, true);
6946 
6947 	pf->vsi[vsi->idx] = NULL;
6948 	if (vsi->idx < pf->next_vsi)
6949 		pf->next_vsi = vsi->idx;
6950 
6951 unlock_vsi:
6952 	mutex_unlock(&pf->switch_mutex);
6953 free_vsi:
6954 	kfree(vsi);
6955 
6956 	return 0;
6957 }
6958 
6959 /**
6960  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
6961  * @vsi: the VSI being cleaned
6962  **/
6963 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
6964 {
6965 	int i;
6966 
6967 	if (vsi->tx_rings && vsi->tx_rings[0]) {
6968 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
6969 			kfree_rcu(vsi->tx_rings[i], rcu);
6970 			vsi->tx_rings[i] = NULL;
6971 			vsi->rx_rings[i] = NULL;
6972 		}
6973 	}
6974 }
6975 
6976 /**
6977  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
6978  * @vsi: the VSI being configured
6979  **/
6980 static int i40e_alloc_rings(struct i40e_vsi *vsi)
6981 {
6982 	struct i40e_ring *tx_ring, *rx_ring;
6983 	struct i40e_pf *pf = vsi->back;
6984 	int i;
6985 
6986 	/* Set basic values in the rings to be used later during open() */
6987 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
6988 		/* allocate space for both Tx and Rx in one shot */
6989 		tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
6990 		if (!tx_ring)
6991 			goto err_out;
6992 
6993 		tx_ring->queue_index = i;
6994 		tx_ring->reg_idx = vsi->base_queue + i;
6995 		tx_ring->ring_active = false;
6996 		tx_ring->vsi = vsi;
6997 		tx_ring->netdev = vsi->netdev;
6998 		tx_ring->dev = &pf->pdev->dev;
6999 		tx_ring->count = vsi->num_desc;
7000 		tx_ring->size = 0;
7001 		tx_ring->dcb_tc = 0;
7002 		vsi->tx_rings[i] = tx_ring;
7003 
7004 		rx_ring = &tx_ring[1];
7005 		rx_ring->queue_index = i;
7006 		rx_ring->reg_idx = vsi->base_queue + i;
7007 		rx_ring->ring_active = false;
7008 		rx_ring->vsi = vsi;
7009 		rx_ring->netdev = vsi->netdev;
7010 		rx_ring->dev = &pf->pdev->dev;
7011 		rx_ring->count = vsi->num_desc;
7012 		rx_ring->size = 0;
7013 		rx_ring->dcb_tc = 0;
7014 		if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
7015 			set_ring_16byte_desc_enabled(rx_ring);
7016 		else
7017 			clear_ring_16byte_desc_enabled(rx_ring);
7018 		vsi->rx_rings[i] = rx_ring;
7019 	}
7020 
7021 	return 0;
7022 
7023 err_out:
7024 	i40e_vsi_clear_rings(vsi);
7025 	return -ENOMEM;
7026 }
7027 
7028 /**
7029  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7030  * @pf: board private structure
7031  * @vectors: the number of MSI-X vectors to request
7032  *
7033  * Returns the number of vectors reserved, or error
7034  **/
7035 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7036 {
7037 	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7038 					I40E_MIN_MSIX, vectors);
7039 	if (vectors < 0) {
7040 		dev_info(&pf->pdev->dev,
7041 			 "MSI-X vector reservation failed: %d\n", vectors);
7042 		vectors = 0;
7043 	}
7044 
7045 	return vectors;
7046 }
7047 
7048 /**
7049  * i40e_init_msix - Setup the MSIX capability
7050  * @pf: board private structure
7051  *
7052  * Work with the OS to set up the MSIX vectors needed.
7053  *
7054  * Returns the number of vectors reserved or negative on failure
7055  **/
7056 static int i40e_init_msix(struct i40e_pf *pf)
7057 {
7058 	struct i40e_hw *hw = &pf->hw;
7059 	int vectors_left;
7060 	int v_budget, i;
7061 	int v_actual;
7062 
7063 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7064 		return -ENODEV;
7065 
7066 	/* The number of vectors we'll request will be comprised of:
7067 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
7068 	 *   - The number of LAN queue pairs
7069 	 *	- Queues being used for RSS.
7070 	 *		We don't need as many as max_rss_size vectors.
7071 	 *		use rss_size instead in the calculation since that
7072 	 *		is governed by number of cpus in the system.
7073 	 *	- assumes symmetric Tx/Rx pairing
7074 	 *   - The number of VMDq pairs
7075 #ifdef I40E_FCOE
7076 	 *   - The number of FCOE qps.
7077 #endif
7078 	 * Once we count this up, try the request.
7079 	 *
7080 	 * If we can't get what we want, we'll simplify to nearly nothing
7081 	 * and try again.  If that still fails, we punt.
7082 	 */
7083 	vectors_left = hw->func_caps.num_msix_vectors;
7084 	v_budget = 0;
7085 
7086 	/* reserve one vector for miscellaneous handler */
7087 	if (vectors_left) {
7088 		v_budget++;
7089 		vectors_left--;
7090 	}
7091 
7092 	/* reserve vectors for the main PF traffic queues */
7093 	pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7094 	vectors_left -= pf->num_lan_msix;
7095 	v_budget += pf->num_lan_msix;
7096 
7097 	/* reserve one vector for sideband flow director */
7098 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7099 		if (vectors_left) {
7100 			v_budget++;
7101 			vectors_left--;
7102 		} else {
7103 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7104 		}
7105 	}
7106 
7107 #ifdef I40E_FCOE
7108 	/* can we reserve enough for FCoE? */
7109 	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7110 		if (!vectors_left)
7111 			pf->num_fcoe_msix = 0;
7112 		else if (vectors_left >= pf->num_fcoe_qps)
7113 			pf->num_fcoe_msix = pf->num_fcoe_qps;
7114 		else
7115 			pf->num_fcoe_msix = 1;
7116 		v_budget += pf->num_fcoe_msix;
7117 		vectors_left -= pf->num_fcoe_msix;
7118 	}
7119 
7120 #endif
7121 	/* any vectors left over go for VMDq support */
7122 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7123 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7124 		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7125 
7126 		/* if we're short on vectors for what's desired, we limit
7127 		 * the queues per vmdq.  If this is still more than are
7128 		 * available, the user will need to change the number of
7129 		 * queues/vectors used by the PF later with the ethtool
7130 		 * channels command
7131 		 */
7132 		if (vmdq_vecs < vmdq_vecs_wanted)
7133 			pf->num_vmdq_qps = 1;
7134 		pf->num_vmdq_msix = pf->num_vmdq_qps;
7135 
7136 		v_budget += vmdq_vecs;
7137 		vectors_left -= vmdq_vecs;
7138 	}
7139 
7140 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7141 				   GFP_KERNEL);
7142 	if (!pf->msix_entries)
7143 		return -ENOMEM;
7144 
7145 	for (i = 0; i < v_budget; i++)
7146 		pf->msix_entries[i].entry = i;
7147 	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7148 
7149 	if (v_actual != v_budget) {
7150 		/* If we have limited resources, we will start with no vectors
7151 		 * for the special features and then allocate vectors to some
7152 		 * of these features based on the policy and at the end disable
7153 		 * the features that did not get any vectors.
7154 		 */
7155 #ifdef I40E_FCOE
7156 		pf->num_fcoe_qps = 0;
7157 		pf->num_fcoe_msix = 0;
7158 #endif
7159 		pf->num_vmdq_msix = 0;
7160 	}
7161 
7162 	if (v_actual < I40E_MIN_MSIX) {
7163 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7164 		kfree(pf->msix_entries);
7165 		pf->msix_entries = NULL;
7166 		return -ENODEV;
7167 
7168 	} else if (v_actual == I40E_MIN_MSIX) {
7169 		/* Adjust for minimal MSIX use */
7170 		pf->num_vmdq_vsis = 0;
7171 		pf->num_vmdq_qps = 0;
7172 		pf->num_lan_qps = 1;
7173 		pf->num_lan_msix = 1;
7174 
7175 	} else if (v_actual != v_budget) {
7176 		int vec;
7177 
7178 		/* reserve the misc vector */
7179 		vec = v_actual - 1;
7180 
7181 		/* Scale vector usage down */
7182 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7183 		pf->num_vmdq_vsis = 1;
7184 		pf->num_vmdq_qps = 1;
7185 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7186 
7187 		/* partition out the remaining vectors */
7188 		switch (vec) {
7189 		case 2:
7190 			pf->num_lan_msix = 1;
7191 			break;
7192 		case 3:
7193 #ifdef I40E_FCOE
7194 			/* give one vector to FCoE */
7195 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7196 				pf->num_lan_msix = 1;
7197 				pf->num_fcoe_msix = 1;
7198 			}
7199 #else
7200 			pf->num_lan_msix = 2;
7201 #endif
7202 			break;
7203 		default:
7204 #ifdef I40E_FCOE
7205 			/* give one vector to FCoE */
7206 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7207 				pf->num_fcoe_msix = 1;
7208 				vec--;
7209 			}
7210 #endif
7211 			/* give the rest to the PF */
7212 			pf->num_lan_msix = min_t(int, vec, pf->num_lan_qps);
7213 			break;
7214 		}
7215 	}
7216 
7217 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7218 	    (pf->num_vmdq_msix == 0)) {
7219 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7220 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7221 	}
7222 #ifdef I40E_FCOE
7223 
7224 	if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7225 		dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7226 		pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7227 	}
7228 #endif
7229 	return v_actual;
7230 }
7231 
7232 /**
7233  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7234  * @vsi: the VSI being configured
7235  * @v_idx: index of the vector in the vsi struct
7236  *
7237  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7238  **/
7239 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
7240 {
7241 	struct i40e_q_vector *q_vector;
7242 
7243 	/* allocate q_vector */
7244 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7245 	if (!q_vector)
7246 		return -ENOMEM;
7247 
7248 	q_vector->vsi = vsi;
7249 	q_vector->v_idx = v_idx;
7250 	cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
7251 	if (vsi->netdev)
7252 		netif_napi_add(vsi->netdev, &q_vector->napi,
7253 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
7254 
7255 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
7256 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
7257 
7258 	/* tie q_vector and vsi together */
7259 	vsi->q_vectors[v_idx] = q_vector;
7260 
7261 	return 0;
7262 }
7263 
7264 /**
7265  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7266  * @vsi: the VSI being configured
7267  *
7268  * We allocate one q_vector per queue interrupt.  If allocation fails we
7269  * return -ENOMEM.
7270  **/
7271 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7272 {
7273 	struct i40e_pf *pf = vsi->back;
7274 	int v_idx, num_q_vectors;
7275 	int err;
7276 
7277 	/* if not MSIX, give the one vector only to the LAN VSI */
7278 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7279 		num_q_vectors = vsi->num_q_vectors;
7280 	else if (vsi == pf->vsi[pf->lan_vsi])
7281 		num_q_vectors = 1;
7282 	else
7283 		return -EINVAL;
7284 
7285 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7286 		err = i40e_vsi_alloc_q_vector(vsi, v_idx);
7287 		if (err)
7288 			goto err_out;
7289 	}
7290 
7291 	return 0;
7292 
7293 err_out:
7294 	while (v_idx--)
7295 		i40e_free_q_vector(vsi, v_idx);
7296 
7297 	return err;
7298 }
7299 
7300 /**
7301  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7302  * @pf: board private structure to initialize
7303  **/
7304 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7305 {
7306 	int vectors = 0;
7307 	ssize_t size;
7308 
7309 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7310 		vectors = i40e_init_msix(pf);
7311 		if (vectors < 0) {
7312 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
7313 #ifdef I40E_FCOE
7314 				       I40E_FLAG_FCOE_ENABLED	|
7315 #endif
7316 				       I40E_FLAG_RSS_ENABLED	|
7317 				       I40E_FLAG_DCB_CAPABLE	|
7318 				       I40E_FLAG_SRIOV_ENABLED	|
7319 				       I40E_FLAG_FD_SB_ENABLED	|
7320 				       I40E_FLAG_FD_ATR_ENABLED	|
7321 				       I40E_FLAG_VMDQ_ENABLED);
7322 
7323 			/* rework the queue expectations without MSIX */
7324 			i40e_determine_queue_usage(pf);
7325 		}
7326 	}
7327 
7328 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7329 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7330 		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7331 		vectors = pci_enable_msi(pf->pdev);
7332 		if (vectors < 0) {
7333 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7334 				 vectors);
7335 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7336 		}
7337 		vectors = 1;  /* one MSI or Legacy vector */
7338 	}
7339 
7340 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7341 		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7342 
7343 	/* set up vector assignment tracking */
7344 	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7345 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
7346 	if (!pf->irq_pile) {
7347 		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7348 		return -ENOMEM;
7349 	}
7350 	pf->irq_pile->num_entries = vectors;
7351 	pf->irq_pile->search_hint = 0;
7352 
7353 	/* track first vector for misc interrupts, ignore return */
7354 	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7355 
7356 	return 0;
7357 }
7358 
7359 /**
7360  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7361  * @pf: board private structure
7362  *
7363  * This sets up the handler for MSIX 0, which is used to manage the
7364  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7365  * when in MSI or Legacy interrupt mode.
7366  **/
7367 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7368 {
7369 	struct i40e_hw *hw = &pf->hw;
7370 	int err = 0;
7371 
7372 	/* Only request the irq if this is the first time through, and
7373 	 * not when we're rebuilding after a Reset
7374 	 */
7375 	if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7376 		err = request_irq(pf->msix_entries[0].vector,
7377 				  i40e_intr, 0, pf->int_name, pf);
7378 		if (err) {
7379 			dev_info(&pf->pdev->dev,
7380 				 "request_irq for %s failed: %d\n",
7381 				 pf->int_name, err);
7382 			return -EFAULT;
7383 		}
7384 	}
7385 
7386 	i40e_enable_misc_int_causes(pf);
7387 
7388 	/* associate no queues to the misc vector */
7389 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7390 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7391 
7392 	i40e_flush(hw);
7393 
7394 	i40e_irq_dynamic_enable_icr0(pf);
7395 
7396 	return err;
7397 }
7398 
7399 /**
7400  * i40e_config_rss - Prepare for RSS if used
7401  * @pf: board private structure
7402  **/
7403 static int i40e_config_rss(struct i40e_pf *pf)
7404 {
7405 	u32 rss_key[I40E_PFQF_HKEY_MAX_INDEX + 1];
7406 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7407 	struct i40e_hw *hw = &pf->hw;
7408 	u32 lut = 0;
7409 	int i, j;
7410 	u64 hena;
7411 	u32 reg_val;
7412 
7413 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
7414 	for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
7415 		wr32(hw, I40E_PFQF_HKEY(i), rss_key[i]);
7416 
7417 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
7418 	hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
7419 		((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
7420 	hena |= I40E_DEFAULT_RSS_HENA;
7421 	wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
7422 	wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
7423 
7424 	vsi->rss_size = min_t(int, pf->rss_size, vsi->num_queue_pairs);
7425 
7426 	/* Check capability and Set table size and register per hw expectation*/
7427 	reg_val = rd32(hw, I40E_PFQF_CTL_0);
7428 	if (pf->rss_table_size == 512)
7429 		reg_val |= I40E_PFQF_CTL_0_HASHLUTSIZE_512;
7430 	else
7431 		reg_val &= ~I40E_PFQF_CTL_0_HASHLUTSIZE_512;
7432 	wr32(hw, I40E_PFQF_CTL_0, reg_val);
7433 
7434 	/* Populate the LUT with max no. of queues in round robin fashion */
7435 	for (i = 0, j = 0; i < pf->rss_table_size; i++, j++) {
7436 
7437 		/* The assumption is that lan qp count will be the highest
7438 		 * qp count for any PF VSI that needs RSS.
7439 		 * If multiple VSIs need RSS support, all the qp counts
7440 		 * for those VSIs should be a power of 2 for RSS to work.
7441 		 * If LAN VSI is the only consumer for RSS then this requirement
7442 		 * is not necessary.
7443 		 */
7444 		if (j == vsi->rss_size)
7445 			j = 0;
7446 		/* lut = 4-byte sliding window of 4 lut entries */
7447 		lut = (lut << 8) | (j &
7448 			 ((0x1 << pf->hw.func_caps.rss_table_entry_width) - 1));
7449 		/* On i = 3, we have 4 entries in lut; write to the register */
7450 		if ((i & 3) == 3)
7451 			wr32(hw, I40E_PFQF_HLUT(i >> 2), lut);
7452 	}
7453 	i40e_flush(hw);
7454 
7455 	return 0;
7456 }
7457 
7458 /**
7459  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
7460  * @pf: board private structure
7461  * @queue_count: the requested queue count for rss.
7462  *
7463  * returns 0 if rss is not enabled, if enabled returns the final rss queue
7464  * count which may be different from the requested queue count.
7465  **/
7466 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
7467 {
7468 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7469 	int new_rss_size;
7470 
7471 	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
7472 		return 0;
7473 
7474 	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
7475 
7476 	if (queue_count != vsi->num_queue_pairs) {
7477 		vsi->req_queue_pairs = queue_count;
7478 		i40e_prep_for_reset(pf);
7479 
7480 		pf->rss_size = new_rss_size;
7481 
7482 		i40e_reset_and_rebuild(pf, true);
7483 		i40e_config_rss(pf);
7484 	}
7485 	dev_info(&pf->pdev->dev, "RSS count:  %d\n", pf->rss_size);
7486 	return pf->rss_size;
7487 }
7488 
7489 /**
7490  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
7491  * @pf: board private structure
7492  **/
7493 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
7494 {
7495 	i40e_status status;
7496 	bool min_valid, max_valid;
7497 	u32 max_bw, min_bw;
7498 
7499 	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
7500 					   &min_valid, &max_valid);
7501 
7502 	if (!status) {
7503 		if (min_valid)
7504 			pf->npar_min_bw = min_bw;
7505 		if (max_valid)
7506 			pf->npar_max_bw = max_bw;
7507 	}
7508 
7509 	return status;
7510 }
7511 
7512 /**
7513  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
7514  * @pf: board private structure
7515  **/
7516 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
7517 {
7518 	struct i40e_aqc_configure_partition_bw_data bw_data;
7519 	i40e_status status;
7520 
7521 	/* Set the valid bit for this PF */
7522 	bw_data.pf_valid_bits = cpu_to_le16(1 << pf->hw.pf_id);
7523 	bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
7524 	bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
7525 
7526 	/* Set the new bandwidths */
7527 	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
7528 
7529 	return status;
7530 }
7531 
7532 /**
7533  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
7534  * @pf: board private structure
7535  **/
7536 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
7537 {
7538 	/* Commit temporary BW setting to permanent NVM image */
7539 	enum i40e_admin_queue_err last_aq_status;
7540 	i40e_status ret;
7541 	u16 nvm_word;
7542 
7543 	if (pf->hw.partition_id != 1) {
7544 		dev_info(&pf->pdev->dev,
7545 			 "Commit BW only works on partition 1! This is partition %d",
7546 			 pf->hw.partition_id);
7547 		ret = I40E_NOT_SUPPORTED;
7548 		goto bw_commit_out;
7549 	}
7550 
7551 	/* Acquire NVM for read access */
7552 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
7553 	last_aq_status = pf->hw.aq.asq_last_status;
7554 	if (ret) {
7555 		dev_info(&pf->pdev->dev,
7556 			 "Cannot acquire NVM for read access, err %d: aq_err %d\n",
7557 			 ret, last_aq_status);
7558 		goto bw_commit_out;
7559 	}
7560 
7561 	/* Read word 0x10 of NVM - SW compatibility word 1 */
7562 	ret = i40e_aq_read_nvm(&pf->hw,
7563 			       I40E_SR_NVM_CONTROL_WORD,
7564 			       0x10, sizeof(nvm_word), &nvm_word,
7565 			       false, NULL);
7566 	/* Save off last admin queue command status before releasing
7567 	 * the NVM
7568 	 */
7569 	last_aq_status = pf->hw.aq.asq_last_status;
7570 	i40e_release_nvm(&pf->hw);
7571 	if (ret) {
7572 		dev_info(&pf->pdev->dev, "NVM read error, err %d aq_err %d\n",
7573 			 ret, last_aq_status);
7574 		goto bw_commit_out;
7575 	}
7576 
7577 	/* Wait a bit for NVM release to complete */
7578 	msleep(50);
7579 
7580 	/* Acquire NVM for write access */
7581 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
7582 	last_aq_status = pf->hw.aq.asq_last_status;
7583 	if (ret) {
7584 		dev_info(&pf->pdev->dev,
7585 			 "Cannot acquire NVM for write access, err %d: aq_err %d\n",
7586 			 ret, last_aq_status);
7587 		goto bw_commit_out;
7588 	}
7589 	/* Write it back out unchanged to initiate update NVM,
7590 	 * which will force a write of the shadow (alt) RAM to
7591 	 * the NVM - thus storing the bandwidth values permanently.
7592 	 */
7593 	ret = i40e_aq_update_nvm(&pf->hw,
7594 				 I40E_SR_NVM_CONTROL_WORD,
7595 				 0x10, sizeof(nvm_word),
7596 				 &nvm_word, true, NULL);
7597 	/* Save off last admin queue command status before releasing
7598 	 * the NVM
7599 	 */
7600 	last_aq_status = pf->hw.aq.asq_last_status;
7601 	i40e_release_nvm(&pf->hw);
7602 	if (ret)
7603 		dev_info(&pf->pdev->dev,
7604 			 "BW settings NOT SAVED, err %d aq_err %d\n",
7605 			 ret, last_aq_status);
7606 bw_commit_out:
7607 
7608 	return ret;
7609 }
7610 
7611 /**
7612  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
7613  * @pf: board private structure to initialize
7614  *
7615  * i40e_sw_init initializes the Adapter private data structure.
7616  * Fields are initialized based on PCI device information and
7617  * OS network device settings (MTU size).
7618  **/
7619 static int i40e_sw_init(struct i40e_pf *pf)
7620 {
7621 	int err = 0;
7622 	int size;
7623 
7624 	pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
7625 				(NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
7626 	pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
7627 	if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
7628 		if (I40E_DEBUG_USER & debug)
7629 			pf->hw.debug_mask = debug;
7630 		pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
7631 						I40E_DEFAULT_MSG_ENABLE);
7632 	}
7633 
7634 	/* Set default capability flags */
7635 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
7636 		    I40E_FLAG_MSI_ENABLED     |
7637 		    I40E_FLAG_MSIX_ENABLED;
7638 
7639 	if (iommu_present(&pci_bus_type))
7640 		pf->flags |= I40E_FLAG_RX_PS_ENABLED;
7641 	else
7642 		pf->flags |= I40E_FLAG_RX_1BUF_ENABLED;
7643 
7644 	/* Set default ITR */
7645 	pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
7646 	pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
7647 
7648 	/* Depending on PF configurations, it is possible that the RSS
7649 	 * maximum might end up larger than the available queues
7650 	 */
7651 	pf->rss_size_max = 0x1 << pf->hw.func_caps.rss_table_entry_width;
7652 	pf->rss_size = 1;
7653 	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
7654 	pf->rss_size_max = min_t(int, pf->rss_size_max,
7655 				 pf->hw.func_caps.num_tx_qp);
7656 	if (pf->hw.func_caps.rss) {
7657 		pf->flags |= I40E_FLAG_RSS_ENABLED;
7658 		pf->rss_size = min_t(int, pf->rss_size_max, num_online_cpus());
7659 	}
7660 
7661 	/* MFP mode enabled */
7662 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.mfp_mode_1) {
7663 		pf->flags |= I40E_FLAG_MFP_ENABLED;
7664 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
7665 		if (i40e_get_npar_bw_setting(pf))
7666 			dev_warn(&pf->pdev->dev,
7667 				 "Could not get NPAR bw settings\n");
7668 		else
7669 			dev_info(&pf->pdev->dev,
7670 				 "Min BW = %8.8x, Max BW = %8.8x\n",
7671 				 pf->npar_min_bw, pf->npar_max_bw);
7672 	}
7673 
7674 	/* FW/NVM is not yet fixed in this regard */
7675 	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
7676 	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
7677 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
7678 		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
7679 		/* Setup a counter for fd_atr per PF */
7680 		pf->fd_atr_cnt_idx = I40E_FD_ATR_STAT_IDX(pf->hw.pf_id);
7681 		if (!(pf->flags & I40E_FLAG_MFP_ENABLED)) {
7682 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
7683 			/* Setup a counter for fd_sb per PF */
7684 			pf->fd_sb_cnt_idx = I40E_FD_SB_STAT_IDX(pf->hw.pf_id);
7685 		} else {
7686 			dev_info(&pf->pdev->dev,
7687 				 "Flow Director Sideband mode Disabled in MFP mode\n");
7688 		}
7689 		pf->fdir_pf_filter_count =
7690 				 pf->hw.func_caps.fd_filters_guaranteed;
7691 		pf->hw.fdir_shared_filter_count =
7692 				 pf->hw.func_caps.fd_filters_best_effort;
7693 	}
7694 
7695 	if (pf->hw.func_caps.vmdq) {
7696 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
7697 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
7698 		pf->num_vmdq_qps = I40E_DEFAULT_QUEUES_PER_VMDQ;
7699 	}
7700 
7701 #ifdef I40E_FCOE
7702 	err = i40e_init_pf_fcoe(pf);
7703 	if (err)
7704 		dev_info(&pf->pdev->dev, "init_pf_fcoe failed: %d\n", err);
7705 
7706 #endif /* I40E_FCOE */
7707 #ifdef CONFIG_PCI_IOV
7708 	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
7709 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
7710 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
7711 		pf->num_req_vfs = min_t(int,
7712 					pf->hw.func_caps.num_vfs,
7713 					I40E_MAX_VF_COUNT);
7714 	}
7715 #endif /* CONFIG_PCI_IOV */
7716 	pf->eeprom_version = 0xDEAD;
7717 	pf->lan_veb = I40E_NO_VEB;
7718 	pf->lan_vsi = I40E_NO_VSI;
7719 
7720 	/* set up queue assignment tracking */
7721 	size = sizeof(struct i40e_lump_tracking)
7722 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
7723 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
7724 	if (!pf->qp_pile) {
7725 		err = -ENOMEM;
7726 		goto sw_init_done;
7727 	}
7728 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
7729 	pf->qp_pile->search_hint = 0;
7730 
7731 	pf->tx_timeout_recovery_level = 1;
7732 
7733 	mutex_init(&pf->switch_mutex);
7734 
7735 	/* If NPAR is enabled nudge the Tx scheduler */
7736 	if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
7737 		i40e_set_npar_bw_setting(pf);
7738 
7739 sw_init_done:
7740 	return err;
7741 }
7742 
7743 /**
7744  * i40e_set_ntuple - set the ntuple feature flag and take action
7745  * @pf: board private structure to initialize
7746  * @features: the feature set that the stack is suggesting
7747  *
7748  * returns a bool to indicate if reset needs to happen
7749  **/
7750 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
7751 {
7752 	bool need_reset = false;
7753 
7754 	/* Check if Flow Director n-tuple support was enabled or disabled.  If
7755 	 * the state changed, we need to reset.
7756 	 */
7757 	if (features & NETIF_F_NTUPLE) {
7758 		/* Enable filters and mark for reset */
7759 		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
7760 			need_reset = true;
7761 		pf->flags |= I40E_FLAG_FD_SB_ENABLED;
7762 	} else {
7763 		/* turn off filters, mark for reset and clear SW filter list */
7764 		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7765 			need_reset = true;
7766 			i40e_fdir_filter_exit(pf);
7767 		}
7768 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7769 		pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
7770 		/* reset fd counters */
7771 		pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
7772 		pf->fdir_pf_active_filters = 0;
7773 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
7774 		dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
7775 		/* if ATR was auto disabled it can be re-enabled. */
7776 		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
7777 		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
7778 			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
7779 	}
7780 	return need_reset;
7781 }
7782 
7783 /**
7784  * i40e_set_features - set the netdev feature flags
7785  * @netdev: ptr to the netdev being adjusted
7786  * @features: the feature set that the stack is suggesting
7787  **/
7788 static int i40e_set_features(struct net_device *netdev,
7789 			     netdev_features_t features)
7790 {
7791 	struct i40e_netdev_priv *np = netdev_priv(netdev);
7792 	struct i40e_vsi *vsi = np->vsi;
7793 	struct i40e_pf *pf = vsi->back;
7794 	bool need_reset;
7795 
7796 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7797 		i40e_vlan_stripping_enable(vsi);
7798 	else
7799 		i40e_vlan_stripping_disable(vsi);
7800 
7801 	need_reset = i40e_set_ntuple(pf, features);
7802 
7803 	if (need_reset)
7804 		i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
7805 
7806 	return 0;
7807 }
7808 
7809 #ifdef CONFIG_I40E_VXLAN
7810 /**
7811  * i40e_get_vxlan_port_idx - Lookup a possibly offloaded for Rx UDP port
7812  * @pf: board private structure
7813  * @port: The UDP port to look up
7814  *
7815  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
7816  **/
7817 static u8 i40e_get_vxlan_port_idx(struct i40e_pf *pf, __be16 port)
7818 {
7819 	u8 i;
7820 
7821 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7822 		if (pf->vxlan_ports[i] == port)
7823 			return i;
7824 	}
7825 
7826 	return i;
7827 }
7828 
7829 /**
7830  * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
7831  * @netdev: This physical port's netdev
7832  * @sa_family: Socket Family that VXLAN is notifying us about
7833  * @port: New UDP port number that VXLAN started listening to
7834  **/
7835 static void i40e_add_vxlan_port(struct net_device *netdev,
7836 				sa_family_t sa_family, __be16 port)
7837 {
7838 	struct i40e_netdev_priv *np = netdev_priv(netdev);
7839 	struct i40e_vsi *vsi = np->vsi;
7840 	struct i40e_pf *pf = vsi->back;
7841 	u8 next_idx;
7842 	u8 idx;
7843 
7844 	if (sa_family == AF_INET6)
7845 		return;
7846 
7847 	idx = i40e_get_vxlan_port_idx(pf, port);
7848 
7849 	/* Check if port already exists */
7850 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
7851 		netdev_info(netdev, "vxlan port %d already offloaded\n",
7852 			    ntohs(port));
7853 		return;
7854 	}
7855 
7856 	/* Now check if there is space to add the new port */
7857 	next_idx = i40e_get_vxlan_port_idx(pf, 0);
7858 
7859 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
7860 		netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
7861 			    ntohs(port));
7862 		return;
7863 	}
7864 
7865 	/* New port: add it and mark its index in the bitmap */
7866 	pf->vxlan_ports[next_idx] = port;
7867 	pf->pending_vxlan_bitmap |= (1 << next_idx);
7868 	pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
7869 
7870 	dev_info(&pf->pdev->dev, "adding vxlan port %d\n", ntohs(port));
7871 }
7872 
7873 /**
7874  * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
7875  * @netdev: This physical port's netdev
7876  * @sa_family: Socket Family that VXLAN is notifying us about
7877  * @port: UDP port number that VXLAN stopped listening to
7878  **/
7879 static void i40e_del_vxlan_port(struct net_device *netdev,
7880 				sa_family_t sa_family, __be16 port)
7881 {
7882 	struct i40e_netdev_priv *np = netdev_priv(netdev);
7883 	struct i40e_vsi *vsi = np->vsi;
7884 	struct i40e_pf *pf = vsi->back;
7885 	u8 idx;
7886 
7887 	if (sa_family == AF_INET6)
7888 		return;
7889 
7890 	idx = i40e_get_vxlan_port_idx(pf, port);
7891 
7892 	/* Check if port already exists */
7893 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
7894 		/* if port exists, set it to 0 (mark for deletion)
7895 		 * and make it pending
7896 		 */
7897 		pf->vxlan_ports[idx] = 0;
7898 		pf->pending_vxlan_bitmap |= (1 << idx);
7899 		pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
7900 
7901 		dev_info(&pf->pdev->dev, "deleting vxlan port %d\n",
7902 			 ntohs(port));
7903 	} else {
7904 		netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
7905 			    ntohs(port));
7906 	}
7907 }
7908 
7909 #endif
7910 static int i40e_get_phys_port_id(struct net_device *netdev,
7911 				 struct netdev_phys_item_id *ppid)
7912 {
7913 	struct i40e_netdev_priv *np = netdev_priv(netdev);
7914 	struct i40e_pf *pf = np->vsi->back;
7915 	struct i40e_hw *hw = &pf->hw;
7916 
7917 	if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
7918 		return -EOPNOTSUPP;
7919 
7920 	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
7921 	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
7922 
7923 	return 0;
7924 }
7925 
7926 /**
7927  * i40e_ndo_fdb_add - add an entry to the hardware database
7928  * @ndm: the input from the stack
7929  * @tb: pointer to array of nladdr (unused)
7930  * @dev: the net device pointer
7931  * @addr: the MAC address entry being added
7932  * @flags: instructions from stack about fdb operation
7933  */
7934 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
7935 			    struct net_device *dev,
7936 			    const unsigned char *addr, u16 vid,
7937 			    u16 flags)
7938 {
7939 	struct i40e_netdev_priv *np = netdev_priv(dev);
7940 	struct i40e_pf *pf = np->vsi->back;
7941 	int err = 0;
7942 
7943 	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
7944 		return -EOPNOTSUPP;
7945 
7946 	if (vid) {
7947 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
7948 		return -EINVAL;
7949 	}
7950 
7951 	/* Hardware does not support aging addresses so if a
7952 	 * ndm_state is given only allow permanent addresses
7953 	 */
7954 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
7955 		netdev_info(dev, "FDB only supports static addresses\n");
7956 		return -EINVAL;
7957 	}
7958 
7959 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
7960 		err = dev_uc_add_excl(dev, addr);
7961 	else if (is_multicast_ether_addr(addr))
7962 		err = dev_mc_add_excl(dev, addr);
7963 	else
7964 		err = -EINVAL;
7965 
7966 	/* Only return duplicate errors if NLM_F_EXCL is set */
7967 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
7968 		err = 0;
7969 
7970 	return err;
7971 }
7972 
7973 #ifdef HAVE_BRIDGE_ATTRIBS
7974 /**
7975  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
7976  * @dev: the netdev being configured
7977  * @nlh: RTNL message
7978  *
7979  * Inserts a new hardware bridge if not already created and
7980  * enables the bridging mode requested (VEB or VEPA). If the
7981  * hardware bridge has already been inserted and the request
7982  * is to change the mode then that requires a PF reset to
7983  * allow rebuild of the components with required hardware
7984  * bridge mode enabled.
7985  **/
7986 static int i40e_ndo_bridge_setlink(struct net_device *dev,
7987 				   struct nlmsghdr *nlh)
7988 {
7989 	struct i40e_netdev_priv *np = netdev_priv(dev);
7990 	struct i40e_vsi *vsi = np->vsi;
7991 	struct i40e_pf *pf = vsi->back;
7992 	struct i40e_veb *veb = NULL;
7993 	struct nlattr *attr, *br_spec;
7994 	int i, rem;
7995 
7996 	/* Only for PF VSI for now */
7997 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
7998 		return -EOPNOTSUPP;
7999 
8000 	/* Find the HW bridge for PF VSI */
8001 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8002 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8003 			veb = pf->veb[i];
8004 	}
8005 
8006 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8007 
8008 	nla_for_each_nested(attr, br_spec, rem) {
8009 		__u16 mode;
8010 
8011 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8012 			continue;
8013 
8014 		mode = nla_get_u16(attr);
8015 		if ((mode != BRIDGE_MODE_VEPA) &&
8016 		    (mode != BRIDGE_MODE_VEB))
8017 			return -EINVAL;
8018 
8019 		/* Insert a new HW bridge */
8020 		if (!veb) {
8021 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8022 					     vsi->tc_config.enabled_tc);
8023 			if (veb) {
8024 				veb->bridge_mode = mode;
8025 				i40e_config_bridge_mode(veb);
8026 			} else {
8027 				/* No Bridge HW offload available */
8028 				return -ENOENT;
8029 			}
8030 			break;
8031 		} else if (mode != veb->bridge_mode) {
8032 			/* Existing HW bridge but different mode needs reset */
8033 			veb->bridge_mode = mode;
8034 			i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
8035 			break;
8036 		}
8037 	}
8038 
8039 	return 0;
8040 }
8041 
8042 /**
8043  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8044  * @skb: skb buff
8045  * @pid: process id
8046  * @seq: RTNL message seq #
8047  * @dev: the netdev being configured
8048  * @filter_mask: unused
8049  *
8050  * Return the mode in which the hardware bridge is operating in
8051  * i.e VEB or VEPA.
8052  **/
8053 #ifdef HAVE_BRIDGE_FILTER
8054 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8055 				   struct net_device *dev,
8056 				   u32 __always_unused filter_mask, int nlflags)
8057 #else
8058 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8059 				   struct net_device *dev, int nlflags)
8060 #endif /* HAVE_BRIDGE_FILTER */
8061 {
8062 	struct i40e_netdev_priv *np = netdev_priv(dev);
8063 	struct i40e_vsi *vsi = np->vsi;
8064 	struct i40e_pf *pf = vsi->back;
8065 	struct i40e_veb *veb = NULL;
8066 	int i;
8067 
8068 	/* Only for PF VSI for now */
8069 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8070 		return -EOPNOTSUPP;
8071 
8072 	/* Find the HW bridge for the PF VSI */
8073 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8074 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8075 			veb = pf->veb[i];
8076 	}
8077 
8078 	if (!veb)
8079 		return 0;
8080 
8081 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
8082 				       nlflags);
8083 }
8084 #endif /* HAVE_BRIDGE_ATTRIBS */
8085 
8086 static const struct net_device_ops i40e_netdev_ops = {
8087 	.ndo_open		= i40e_open,
8088 	.ndo_stop		= i40e_close,
8089 	.ndo_start_xmit		= i40e_lan_xmit_frame,
8090 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
8091 	.ndo_set_rx_mode	= i40e_set_rx_mode,
8092 	.ndo_validate_addr	= eth_validate_addr,
8093 	.ndo_set_mac_address	= i40e_set_mac,
8094 	.ndo_change_mtu		= i40e_change_mtu,
8095 	.ndo_do_ioctl		= i40e_ioctl,
8096 	.ndo_tx_timeout		= i40e_tx_timeout,
8097 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
8098 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
8099 #ifdef CONFIG_NET_POLL_CONTROLLER
8100 	.ndo_poll_controller	= i40e_netpoll,
8101 #endif
8102 	.ndo_setup_tc		= i40e_setup_tc,
8103 #ifdef I40E_FCOE
8104 	.ndo_fcoe_enable	= i40e_fcoe_enable,
8105 	.ndo_fcoe_disable	= i40e_fcoe_disable,
8106 #endif
8107 	.ndo_set_features	= i40e_set_features,
8108 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
8109 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
8110 	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
8111 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
8112 	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
8113 	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
8114 #ifdef CONFIG_I40E_VXLAN
8115 	.ndo_add_vxlan_port	= i40e_add_vxlan_port,
8116 	.ndo_del_vxlan_port	= i40e_del_vxlan_port,
8117 #endif
8118 	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
8119 	.ndo_fdb_add		= i40e_ndo_fdb_add,
8120 #ifdef HAVE_BRIDGE_ATTRIBS
8121 	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
8122 	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
8123 #endif /* HAVE_BRIDGE_ATTRIBS */
8124 };
8125 
8126 /**
8127  * i40e_config_netdev - Setup the netdev flags
8128  * @vsi: the VSI being configured
8129  *
8130  * Returns 0 on success, negative value on failure
8131  **/
8132 static int i40e_config_netdev(struct i40e_vsi *vsi)
8133 {
8134 	u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
8135 	struct i40e_pf *pf = vsi->back;
8136 	struct i40e_hw *hw = &pf->hw;
8137 	struct i40e_netdev_priv *np;
8138 	struct net_device *netdev;
8139 	u8 mac_addr[ETH_ALEN];
8140 	int etherdev_size;
8141 
8142 	etherdev_size = sizeof(struct i40e_netdev_priv);
8143 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
8144 	if (!netdev)
8145 		return -ENOMEM;
8146 
8147 	vsi->netdev = netdev;
8148 	np = netdev_priv(netdev);
8149 	np->vsi = vsi;
8150 
8151 	netdev->hw_enc_features |= NETIF_F_IP_CSUM	 |
8152 				  NETIF_F_GSO_UDP_TUNNEL |
8153 				  NETIF_F_TSO;
8154 
8155 	netdev->features = NETIF_F_SG		       |
8156 			   NETIF_F_IP_CSUM	       |
8157 			   NETIF_F_SCTP_CSUM	       |
8158 			   NETIF_F_HIGHDMA	       |
8159 			   NETIF_F_GSO_UDP_TUNNEL      |
8160 			   NETIF_F_HW_VLAN_CTAG_TX     |
8161 			   NETIF_F_HW_VLAN_CTAG_RX     |
8162 			   NETIF_F_HW_VLAN_CTAG_FILTER |
8163 			   NETIF_F_IPV6_CSUM	       |
8164 			   NETIF_F_TSO		       |
8165 			   NETIF_F_TSO_ECN	       |
8166 			   NETIF_F_TSO6		       |
8167 			   NETIF_F_RXCSUM	       |
8168 			   NETIF_F_RXHASH	       |
8169 			   0;
8170 
8171 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
8172 		netdev->features |= NETIF_F_NTUPLE;
8173 
8174 	/* copy netdev features into list of user selectable features */
8175 	netdev->hw_features |= netdev->features;
8176 
8177 	if (vsi->type == I40E_VSI_MAIN) {
8178 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
8179 		ether_addr_copy(mac_addr, hw->mac.perm_addr);
8180 		/* The following steps are necessary to prevent reception
8181 		 * of tagged packets - some older NVM configurations load a
8182 		 * default a MAC-VLAN filter that accepts any tagged packet
8183 		 * which must be replaced by a normal filter.
8184 		 */
8185 		if (!i40e_rm_default_mac_filter(vsi, mac_addr))
8186 			i40e_add_filter(vsi, mac_addr,
8187 					I40E_VLAN_ANY, false, true);
8188 	} else {
8189 		/* relate the VSI_VMDQ name to the VSI_MAIN name */
8190 		snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
8191 			 pf->vsi[pf->lan_vsi]->netdev->name);
8192 		random_ether_addr(mac_addr);
8193 		i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
8194 	}
8195 	i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
8196 
8197 	ether_addr_copy(netdev->dev_addr, mac_addr);
8198 	ether_addr_copy(netdev->perm_addr, mac_addr);
8199 	/* vlan gets same features (except vlan offload)
8200 	 * after any tweaks for specific VSI types
8201 	 */
8202 	netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
8203 						     NETIF_F_HW_VLAN_CTAG_RX |
8204 						   NETIF_F_HW_VLAN_CTAG_FILTER);
8205 	netdev->priv_flags |= IFF_UNICAST_FLT;
8206 	netdev->priv_flags |= IFF_SUPP_NOFCS;
8207 	/* Setup netdev TC information */
8208 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
8209 
8210 	netdev->netdev_ops = &i40e_netdev_ops;
8211 	netdev->watchdog_timeo = 5 * HZ;
8212 	i40e_set_ethtool_ops(netdev);
8213 #ifdef I40E_FCOE
8214 	i40e_fcoe_config_netdev(netdev, vsi);
8215 #endif
8216 
8217 	return 0;
8218 }
8219 
8220 /**
8221  * i40e_vsi_delete - Delete a VSI from the switch
8222  * @vsi: the VSI being removed
8223  *
8224  * Returns 0 on success, negative value on failure
8225  **/
8226 static void i40e_vsi_delete(struct i40e_vsi *vsi)
8227 {
8228 	/* remove default VSI is not allowed */
8229 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
8230 		return;
8231 
8232 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
8233 }
8234 
8235 /**
8236  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
8237  * @vsi: the VSI being queried
8238  *
8239  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
8240  **/
8241 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
8242 {
8243 	struct i40e_veb *veb;
8244 	struct i40e_pf *pf = vsi->back;
8245 
8246 	/* Uplink is not a bridge so default to VEB */
8247 	if (vsi->veb_idx == I40E_NO_VEB)
8248 		return 1;
8249 
8250 	veb = pf->veb[vsi->veb_idx];
8251 	/* Uplink is a bridge in VEPA mode */
8252 	if (veb && (veb->bridge_mode & BRIDGE_MODE_VEPA))
8253 		return 0;
8254 
8255 	/* Uplink is a bridge in VEB mode */
8256 	return 1;
8257 }
8258 
8259 /**
8260  * i40e_add_vsi - Add a VSI to the switch
8261  * @vsi: the VSI being configured
8262  *
8263  * This initializes a VSI context depending on the VSI type to be added and
8264  * passes it down to the add_vsi aq command.
8265  **/
8266 static int i40e_add_vsi(struct i40e_vsi *vsi)
8267 {
8268 	int ret = -ENODEV;
8269 	struct i40e_mac_filter *f, *ftmp;
8270 	struct i40e_pf *pf = vsi->back;
8271 	struct i40e_hw *hw = &pf->hw;
8272 	struct i40e_vsi_context ctxt;
8273 	u8 enabled_tc = 0x1; /* TC0 enabled */
8274 	int f_count = 0;
8275 
8276 	memset(&ctxt, 0, sizeof(ctxt));
8277 	switch (vsi->type) {
8278 	case I40E_VSI_MAIN:
8279 		/* The PF's main VSI is already setup as part of the
8280 		 * device initialization, so we'll not bother with
8281 		 * the add_vsi call, but we will retrieve the current
8282 		 * VSI context.
8283 		 */
8284 		ctxt.seid = pf->main_vsi_seid;
8285 		ctxt.pf_num = pf->hw.pf_id;
8286 		ctxt.vf_num = 0;
8287 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
8288 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8289 		if (ret) {
8290 			dev_info(&pf->pdev->dev,
8291 				 "couldn't get PF vsi config, err %d, aq_err %d\n",
8292 				 ret, pf->hw.aq.asq_last_status);
8293 			return -ENOENT;
8294 		}
8295 		vsi->info = ctxt.info;
8296 		vsi->info.valid_sections = 0;
8297 
8298 		vsi->seid = ctxt.seid;
8299 		vsi->id = ctxt.vsi_number;
8300 
8301 		enabled_tc = i40e_pf_get_tc_map(pf);
8302 
8303 		/* MFP mode setup queue map and update VSI */
8304 		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
8305 		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
8306 			memset(&ctxt, 0, sizeof(ctxt));
8307 			ctxt.seid = pf->main_vsi_seid;
8308 			ctxt.pf_num = pf->hw.pf_id;
8309 			ctxt.vf_num = 0;
8310 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
8311 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
8312 			if (ret) {
8313 				dev_info(&pf->pdev->dev,
8314 					 "update vsi failed, aq_err=%d\n",
8315 					 pf->hw.aq.asq_last_status);
8316 				ret = -ENOENT;
8317 				goto err;
8318 			}
8319 			/* update the local VSI info queue map */
8320 			i40e_vsi_update_queue_map(vsi, &ctxt);
8321 			vsi->info.valid_sections = 0;
8322 		} else {
8323 			/* Default/Main VSI is only enabled for TC0
8324 			 * reconfigure it to enable all TCs that are
8325 			 * available on the port in SFP mode.
8326 			 * For MFP case the iSCSI PF would use this
8327 			 * flow to enable LAN+iSCSI TC.
8328 			 */
8329 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
8330 			if (ret) {
8331 				dev_info(&pf->pdev->dev,
8332 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %d, aq_err %d\n",
8333 					 enabled_tc, ret,
8334 					 pf->hw.aq.asq_last_status);
8335 				ret = -ENOENT;
8336 			}
8337 		}
8338 		break;
8339 
8340 	case I40E_VSI_FDIR:
8341 		ctxt.pf_num = hw->pf_id;
8342 		ctxt.vf_num = 0;
8343 		ctxt.uplink_seid = vsi->uplink_seid;
8344 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8345 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8346 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
8347 			ctxt.info.valid_sections |=
8348 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8349 			ctxt.info.switch_id =
8350 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8351 		}
8352 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8353 		break;
8354 
8355 	case I40E_VSI_VMDQ2:
8356 		ctxt.pf_num = hw->pf_id;
8357 		ctxt.vf_num = 0;
8358 		ctxt.uplink_seid = vsi->uplink_seid;
8359 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8360 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
8361 
8362 		/* This VSI is connected to VEB so the switch_id
8363 		 * should be set to zero by default.
8364 		 */
8365 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
8366 			ctxt.info.valid_sections |=
8367 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8368 			ctxt.info.switch_id =
8369 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8370 		}
8371 
8372 		/* Setup the VSI tx/rx queue map for TC0 only for now */
8373 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8374 		break;
8375 
8376 	case I40E_VSI_SRIOV:
8377 		ctxt.pf_num = hw->pf_id;
8378 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
8379 		ctxt.uplink_seid = vsi->uplink_seid;
8380 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8381 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
8382 
8383 		/* This VSI is connected to VEB so the switch_id
8384 		 * should be set to zero by default.
8385 		 */
8386 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
8387 			ctxt.info.valid_sections |=
8388 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8389 			ctxt.info.switch_id =
8390 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8391 		}
8392 
8393 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
8394 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
8395 		if (pf->vf[vsi->vf_id].spoofchk) {
8396 			ctxt.info.valid_sections |=
8397 				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
8398 			ctxt.info.sec_flags |=
8399 				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
8400 				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
8401 		}
8402 		/* Setup the VSI tx/rx queue map for TC0 only for now */
8403 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8404 		break;
8405 
8406 #ifdef I40E_FCOE
8407 	case I40E_VSI_FCOE:
8408 		ret = i40e_fcoe_vsi_init(vsi, &ctxt);
8409 		if (ret) {
8410 			dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
8411 			return ret;
8412 		}
8413 		break;
8414 
8415 #endif /* I40E_FCOE */
8416 	default:
8417 		return -ENODEV;
8418 	}
8419 
8420 	if (vsi->type != I40E_VSI_MAIN) {
8421 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
8422 		if (ret) {
8423 			dev_info(&vsi->back->pdev->dev,
8424 				 "add vsi failed, aq_err=%d\n",
8425 				 vsi->back->hw.aq.asq_last_status);
8426 			ret = -ENOENT;
8427 			goto err;
8428 		}
8429 		vsi->info = ctxt.info;
8430 		vsi->info.valid_sections = 0;
8431 		vsi->seid = ctxt.seid;
8432 		vsi->id = ctxt.vsi_number;
8433 	}
8434 
8435 	/* If macvlan filters already exist, force them to get loaded */
8436 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
8437 		f->changed = true;
8438 		f_count++;
8439 
8440 		if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
8441 			struct i40e_aqc_remove_macvlan_element_data element;
8442 
8443 			memset(&element, 0, sizeof(element));
8444 			ether_addr_copy(element.mac_addr, f->macaddr);
8445 			element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
8446 			ret = i40e_aq_remove_macvlan(hw, vsi->seid,
8447 						     &element, 1, NULL);
8448 			if (ret) {
8449 				/* some older FW has a different default */
8450 				element.flags |=
8451 					       I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
8452 				i40e_aq_remove_macvlan(hw, vsi->seid,
8453 						       &element, 1, NULL);
8454 			}
8455 
8456 			i40e_aq_mac_address_write(hw,
8457 						  I40E_AQC_WRITE_TYPE_LAA_WOL,
8458 						  f->macaddr, NULL);
8459 		}
8460 	}
8461 	if (f_count) {
8462 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
8463 		pf->flags |= I40E_FLAG_FILTER_SYNC;
8464 	}
8465 
8466 	/* Update VSI BW information */
8467 	ret = i40e_vsi_get_bw_info(vsi);
8468 	if (ret) {
8469 		dev_info(&pf->pdev->dev,
8470 			 "couldn't get vsi bw info, err %d, aq_err %d\n",
8471 			 ret, pf->hw.aq.asq_last_status);
8472 		/* VSI is already added so not tearing that up */
8473 		ret = 0;
8474 	}
8475 
8476 err:
8477 	return ret;
8478 }
8479 
8480 /**
8481  * i40e_vsi_release - Delete a VSI and free its resources
8482  * @vsi: the VSI being removed
8483  *
8484  * Returns 0 on success or < 0 on error
8485  **/
8486 int i40e_vsi_release(struct i40e_vsi *vsi)
8487 {
8488 	struct i40e_mac_filter *f, *ftmp;
8489 	struct i40e_veb *veb = NULL;
8490 	struct i40e_pf *pf;
8491 	u16 uplink_seid;
8492 	int i, n;
8493 
8494 	pf = vsi->back;
8495 
8496 	/* release of a VEB-owner or last VSI is not allowed */
8497 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
8498 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
8499 			 vsi->seid, vsi->uplink_seid);
8500 		return -ENODEV;
8501 	}
8502 	if (vsi == pf->vsi[pf->lan_vsi] &&
8503 	    !test_bit(__I40E_DOWN, &pf->state)) {
8504 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
8505 		return -ENODEV;
8506 	}
8507 
8508 	uplink_seid = vsi->uplink_seid;
8509 	if (vsi->type != I40E_VSI_SRIOV) {
8510 		if (vsi->netdev_registered) {
8511 			vsi->netdev_registered = false;
8512 			if (vsi->netdev) {
8513 				/* results in a call to i40e_close() */
8514 				unregister_netdev(vsi->netdev);
8515 			}
8516 		} else {
8517 			i40e_vsi_close(vsi);
8518 		}
8519 		i40e_vsi_disable_irq(vsi);
8520 	}
8521 
8522 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
8523 		i40e_del_filter(vsi, f->macaddr, f->vlan,
8524 				f->is_vf, f->is_netdev);
8525 	i40e_sync_vsi_filters(vsi);
8526 
8527 	i40e_vsi_delete(vsi);
8528 	i40e_vsi_free_q_vectors(vsi);
8529 	if (vsi->netdev) {
8530 		free_netdev(vsi->netdev);
8531 		vsi->netdev = NULL;
8532 	}
8533 	i40e_vsi_clear_rings(vsi);
8534 	i40e_vsi_clear(vsi);
8535 
8536 	/* If this was the last thing on the VEB, except for the
8537 	 * controlling VSI, remove the VEB, which puts the controlling
8538 	 * VSI onto the next level down in the switch.
8539 	 *
8540 	 * Well, okay, there's one more exception here: don't remove
8541 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
8542 	 * from up the network stack.
8543 	 */
8544 	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
8545 		if (pf->vsi[i] &&
8546 		    pf->vsi[i]->uplink_seid == uplink_seid &&
8547 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
8548 			n++;      /* count the VSIs */
8549 		}
8550 	}
8551 	for (i = 0; i < I40E_MAX_VEB; i++) {
8552 		if (!pf->veb[i])
8553 			continue;
8554 		if (pf->veb[i]->uplink_seid == uplink_seid)
8555 			n++;     /* count the VEBs */
8556 		if (pf->veb[i]->seid == uplink_seid)
8557 			veb = pf->veb[i];
8558 	}
8559 	if (n == 0 && veb && veb->uplink_seid != 0)
8560 		i40e_veb_release(veb);
8561 
8562 	return 0;
8563 }
8564 
8565 /**
8566  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
8567  * @vsi: ptr to the VSI
8568  *
8569  * This should only be called after i40e_vsi_mem_alloc() which allocates the
8570  * corresponding SW VSI structure and initializes num_queue_pairs for the
8571  * newly allocated VSI.
8572  *
8573  * Returns 0 on success or negative on failure
8574  **/
8575 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
8576 {
8577 	int ret = -ENOENT;
8578 	struct i40e_pf *pf = vsi->back;
8579 
8580 	if (vsi->q_vectors[0]) {
8581 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
8582 			 vsi->seid);
8583 		return -EEXIST;
8584 	}
8585 
8586 	if (vsi->base_vector) {
8587 		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
8588 			 vsi->seid, vsi->base_vector);
8589 		return -EEXIST;
8590 	}
8591 
8592 	ret = i40e_vsi_alloc_q_vectors(vsi);
8593 	if (ret) {
8594 		dev_info(&pf->pdev->dev,
8595 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
8596 			 vsi->num_q_vectors, vsi->seid, ret);
8597 		vsi->num_q_vectors = 0;
8598 		goto vector_setup_out;
8599 	}
8600 
8601 	if (vsi->num_q_vectors)
8602 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
8603 						 vsi->num_q_vectors, vsi->idx);
8604 	if (vsi->base_vector < 0) {
8605 		dev_info(&pf->pdev->dev,
8606 			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
8607 			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
8608 		i40e_vsi_free_q_vectors(vsi);
8609 		ret = -ENOENT;
8610 		goto vector_setup_out;
8611 	}
8612 
8613 vector_setup_out:
8614 	return ret;
8615 }
8616 
8617 /**
8618  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
8619  * @vsi: pointer to the vsi.
8620  *
8621  * This re-allocates a vsi's queue resources.
8622  *
8623  * Returns pointer to the successfully allocated and configured VSI sw struct
8624  * on success, otherwise returns NULL on failure.
8625  **/
8626 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
8627 {
8628 	struct i40e_pf *pf = vsi->back;
8629 	u8 enabled_tc;
8630 	int ret;
8631 
8632 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
8633 	i40e_vsi_clear_rings(vsi);
8634 
8635 	i40e_vsi_free_arrays(vsi, false);
8636 	i40e_set_num_rings_in_vsi(vsi);
8637 	ret = i40e_vsi_alloc_arrays(vsi, false);
8638 	if (ret)
8639 		goto err_vsi;
8640 
8641 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
8642 	if (ret < 0) {
8643 		dev_info(&pf->pdev->dev,
8644 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
8645 			 vsi->alloc_queue_pairs, vsi->seid, ret);
8646 		goto err_vsi;
8647 	}
8648 	vsi->base_queue = ret;
8649 
8650 	/* Update the FW view of the VSI. Force a reset of TC and queue
8651 	 * layout configurations.
8652 	 */
8653 	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
8654 	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
8655 	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
8656 	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
8657 
8658 	/* assign it some queues */
8659 	ret = i40e_alloc_rings(vsi);
8660 	if (ret)
8661 		goto err_rings;
8662 
8663 	/* map all of the rings to the q_vectors */
8664 	i40e_vsi_map_rings_to_vectors(vsi);
8665 	return vsi;
8666 
8667 err_rings:
8668 	i40e_vsi_free_q_vectors(vsi);
8669 	if (vsi->netdev_registered) {
8670 		vsi->netdev_registered = false;
8671 		unregister_netdev(vsi->netdev);
8672 		free_netdev(vsi->netdev);
8673 		vsi->netdev = NULL;
8674 	}
8675 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
8676 err_vsi:
8677 	i40e_vsi_clear(vsi);
8678 	return NULL;
8679 }
8680 
8681 /**
8682  * i40e_vsi_setup - Set up a VSI by a given type
8683  * @pf: board private structure
8684  * @type: VSI type
8685  * @uplink_seid: the switch element to link to
8686  * @param1: usage depends upon VSI type. For VF types, indicates VF id
8687  *
8688  * This allocates the sw VSI structure and its queue resources, then add a VSI
8689  * to the identified VEB.
8690  *
8691  * Returns pointer to the successfully allocated and configure VSI sw struct on
8692  * success, otherwise returns NULL on failure.
8693  **/
8694 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
8695 				u16 uplink_seid, u32 param1)
8696 {
8697 	struct i40e_vsi *vsi = NULL;
8698 	struct i40e_veb *veb = NULL;
8699 	int ret, i;
8700 	int v_idx;
8701 
8702 	/* The requested uplink_seid must be either
8703 	 *     - the PF's port seid
8704 	 *              no VEB is needed because this is the PF
8705 	 *              or this is a Flow Director special case VSI
8706 	 *     - seid of an existing VEB
8707 	 *     - seid of a VSI that owns an existing VEB
8708 	 *     - seid of a VSI that doesn't own a VEB
8709 	 *              a new VEB is created and the VSI becomes the owner
8710 	 *     - seid of the PF VSI, which is what creates the first VEB
8711 	 *              this is a special case of the previous
8712 	 *
8713 	 * Find which uplink_seid we were given and create a new VEB if needed
8714 	 */
8715 	for (i = 0; i < I40E_MAX_VEB; i++) {
8716 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
8717 			veb = pf->veb[i];
8718 			break;
8719 		}
8720 	}
8721 
8722 	if (!veb && uplink_seid != pf->mac_seid) {
8723 
8724 		for (i = 0; i < pf->num_alloc_vsi; i++) {
8725 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
8726 				vsi = pf->vsi[i];
8727 				break;
8728 			}
8729 		}
8730 		if (!vsi) {
8731 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
8732 				 uplink_seid);
8733 			return NULL;
8734 		}
8735 
8736 		if (vsi->uplink_seid == pf->mac_seid)
8737 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
8738 					     vsi->tc_config.enabled_tc);
8739 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
8740 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8741 					     vsi->tc_config.enabled_tc);
8742 		if (veb) {
8743 			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
8744 				dev_info(&vsi->back->pdev->dev,
8745 					 "%s: New VSI creation error, uplink seid of LAN VSI expected.\n",
8746 					 __func__);
8747 				return NULL;
8748 			}
8749 			i40e_config_bridge_mode(veb);
8750 		}
8751 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8752 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8753 				veb = pf->veb[i];
8754 		}
8755 		if (!veb) {
8756 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
8757 			return NULL;
8758 		}
8759 
8760 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
8761 		uplink_seid = veb->seid;
8762 	}
8763 
8764 	/* get vsi sw struct */
8765 	v_idx = i40e_vsi_mem_alloc(pf, type);
8766 	if (v_idx < 0)
8767 		goto err_alloc;
8768 	vsi = pf->vsi[v_idx];
8769 	if (!vsi)
8770 		goto err_alloc;
8771 	vsi->type = type;
8772 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
8773 
8774 	if (type == I40E_VSI_MAIN)
8775 		pf->lan_vsi = v_idx;
8776 	else if (type == I40E_VSI_SRIOV)
8777 		vsi->vf_id = param1;
8778 	/* assign it some queues */
8779 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
8780 				vsi->idx);
8781 	if (ret < 0) {
8782 		dev_info(&pf->pdev->dev,
8783 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
8784 			 vsi->alloc_queue_pairs, vsi->seid, ret);
8785 		goto err_vsi;
8786 	}
8787 	vsi->base_queue = ret;
8788 
8789 	/* get a VSI from the hardware */
8790 	vsi->uplink_seid = uplink_seid;
8791 	ret = i40e_add_vsi(vsi);
8792 	if (ret)
8793 		goto err_vsi;
8794 
8795 	switch (vsi->type) {
8796 	/* setup the netdev if needed */
8797 	case I40E_VSI_MAIN:
8798 	case I40E_VSI_VMDQ2:
8799 	case I40E_VSI_FCOE:
8800 		ret = i40e_config_netdev(vsi);
8801 		if (ret)
8802 			goto err_netdev;
8803 		ret = register_netdev(vsi->netdev);
8804 		if (ret)
8805 			goto err_netdev;
8806 		vsi->netdev_registered = true;
8807 		netif_carrier_off(vsi->netdev);
8808 #ifdef CONFIG_I40E_DCB
8809 		/* Setup DCB netlink interface */
8810 		i40e_dcbnl_setup(vsi);
8811 #endif /* CONFIG_I40E_DCB */
8812 		/* fall through */
8813 
8814 	case I40E_VSI_FDIR:
8815 		/* set up vectors and rings if needed */
8816 		ret = i40e_vsi_setup_vectors(vsi);
8817 		if (ret)
8818 			goto err_msix;
8819 
8820 		ret = i40e_alloc_rings(vsi);
8821 		if (ret)
8822 			goto err_rings;
8823 
8824 		/* map all of the rings to the q_vectors */
8825 		i40e_vsi_map_rings_to_vectors(vsi);
8826 
8827 		i40e_vsi_reset_stats(vsi);
8828 		break;
8829 
8830 	default:
8831 		/* no netdev or rings for the other VSI types */
8832 		break;
8833 	}
8834 
8835 	return vsi;
8836 
8837 err_rings:
8838 	i40e_vsi_free_q_vectors(vsi);
8839 err_msix:
8840 	if (vsi->netdev_registered) {
8841 		vsi->netdev_registered = false;
8842 		unregister_netdev(vsi->netdev);
8843 		free_netdev(vsi->netdev);
8844 		vsi->netdev = NULL;
8845 	}
8846 err_netdev:
8847 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
8848 err_vsi:
8849 	i40e_vsi_clear(vsi);
8850 err_alloc:
8851 	return NULL;
8852 }
8853 
8854 /**
8855  * i40e_veb_get_bw_info - Query VEB BW information
8856  * @veb: the veb to query
8857  *
8858  * Query the Tx scheduler BW configuration data for given VEB
8859  **/
8860 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
8861 {
8862 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
8863 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
8864 	struct i40e_pf *pf = veb->pf;
8865 	struct i40e_hw *hw = &pf->hw;
8866 	u32 tc_bw_max;
8867 	int ret = 0;
8868 	int i;
8869 
8870 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
8871 						  &bw_data, NULL);
8872 	if (ret) {
8873 		dev_info(&pf->pdev->dev,
8874 			 "query veb bw config failed, aq_err=%d\n",
8875 			 hw->aq.asq_last_status);
8876 		goto out;
8877 	}
8878 
8879 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
8880 						   &ets_data, NULL);
8881 	if (ret) {
8882 		dev_info(&pf->pdev->dev,
8883 			 "query veb bw ets config failed, aq_err=%d\n",
8884 			 hw->aq.asq_last_status);
8885 		goto out;
8886 	}
8887 
8888 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
8889 	veb->bw_max_quanta = ets_data.tc_bw_max;
8890 	veb->is_abs_credits = bw_data.absolute_credits_enable;
8891 	veb->enabled_tc = ets_data.tc_valid_bits;
8892 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
8893 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
8894 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
8895 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
8896 		veb->bw_tc_limit_credits[i] =
8897 					le16_to_cpu(bw_data.tc_bw_limits[i]);
8898 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
8899 	}
8900 
8901 out:
8902 	return ret;
8903 }
8904 
8905 /**
8906  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
8907  * @pf: board private structure
8908  *
8909  * On error: returns error code (negative)
8910  * On success: returns vsi index in PF (positive)
8911  **/
8912 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
8913 {
8914 	int ret = -ENOENT;
8915 	struct i40e_veb *veb;
8916 	int i;
8917 
8918 	/* Need to protect the allocation of switch elements at the PF level */
8919 	mutex_lock(&pf->switch_mutex);
8920 
8921 	/* VEB list may be fragmented if VEB creation/destruction has
8922 	 * been happening.  We can afford to do a quick scan to look
8923 	 * for any free slots in the list.
8924 	 *
8925 	 * find next empty veb slot, looping back around if necessary
8926 	 */
8927 	i = 0;
8928 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
8929 		i++;
8930 	if (i >= I40E_MAX_VEB) {
8931 		ret = -ENOMEM;
8932 		goto err_alloc_veb;  /* out of VEB slots! */
8933 	}
8934 
8935 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
8936 	if (!veb) {
8937 		ret = -ENOMEM;
8938 		goto err_alloc_veb;
8939 	}
8940 	veb->pf = pf;
8941 	veb->idx = i;
8942 	veb->enabled_tc = 1;
8943 
8944 	pf->veb[i] = veb;
8945 	ret = i;
8946 err_alloc_veb:
8947 	mutex_unlock(&pf->switch_mutex);
8948 	return ret;
8949 }
8950 
8951 /**
8952  * i40e_switch_branch_release - Delete a branch of the switch tree
8953  * @branch: where to start deleting
8954  *
8955  * This uses recursion to find the tips of the branch to be
8956  * removed, deleting until we get back to and can delete this VEB.
8957  **/
8958 static void i40e_switch_branch_release(struct i40e_veb *branch)
8959 {
8960 	struct i40e_pf *pf = branch->pf;
8961 	u16 branch_seid = branch->seid;
8962 	u16 veb_idx = branch->idx;
8963 	int i;
8964 
8965 	/* release any VEBs on this VEB - RECURSION */
8966 	for (i = 0; i < I40E_MAX_VEB; i++) {
8967 		if (!pf->veb[i])
8968 			continue;
8969 		if (pf->veb[i]->uplink_seid == branch->seid)
8970 			i40e_switch_branch_release(pf->veb[i]);
8971 	}
8972 
8973 	/* Release the VSIs on this VEB, but not the owner VSI.
8974 	 *
8975 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
8976 	 *       the VEB itself, so don't use (*branch) after this loop.
8977 	 */
8978 	for (i = 0; i < pf->num_alloc_vsi; i++) {
8979 		if (!pf->vsi[i])
8980 			continue;
8981 		if (pf->vsi[i]->uplink_seid == branch_seid &&
8982 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
8983 			i40e_vsi_release(pf->vsi[i]);
8984 		}
8985 	}
8986 
8987 	/* There's one corner case where the VEB might not have been
8988 	 * removed, so double check it here and remove it if needed.
8989 	 * This case happens if the veb was created from the debugfs
8990 	 * commands and no VSIs were added to it.
8991 	 */
8992 	if (pf->veb[veb_idx])
8993 		i40e_veb_release(pf->veb[veb_idx]);
8994 }
8995 
8996 /**
8997  * i40e_veb_clear - remove veb struct
8998  * @veb: the veb to remove
8999  **/
9000 static void i40e_veb_clear(struct i40e_veb *veb)
9001 {
9002 	if (!veb)
9003 		return;
9004 
9005 	if (veb->pf) {
9006 		struct i40e_pf *pf = veb->pf;
9007 
9008 		mutex_lock(&pf->switch_mutex);
9009 		if (pf->veb[veb->idx] == veb)
9010 			pf->veb[veb->idx] = NULL;
9011 		mutex_unlock(&pf->switch_mutex);
9012 	}
9013 
9014 	kfree(veb);
9015 }
9016 
9017 /**
9018  * i40e_veb_release - Delete a VEB and free its resources
9019  * @veb: the VEB being removed
9020  **/
9021 void i40e_veb_release(struct i40e_veb *veb)
9022 {
9023 	struct i40e_vsi *vsi = NULL;
9024 	struct i40e_pf *pf;
9025 	int i, n = 0;
9026 
9027 	pf = veb->pf;
9028 
9029 	/* find the remaining VSI and check for extras */
9030 	for (i = 0; i < pf->num_alloc_vsi; i++) {
9031 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
9032 			n++;
9033 			vsi = pf->vsi[i];
9034 		}
9035 	}
9036 	if (n != 1) {
9037 		dev_info(&pf->pdev->dev,
9038 			 "can't remove VEB %d with %d VSIs left\n",
9039 			 veb->seid, n);
9040 		return;
9041 	}
9042 
9043 	/* move the remaining VSI to uplink veb */
9044 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
9045 	if (veb->uplink_seid) {
9046 		vsi->uplink_seid = veb->uplink_seid;
9047 		if (veb->uplink_seid == pf->mac_seid)
9048 			vsi->veb_idx = I40E_NO_VEB;
9049 		else
9050 			vsi->veb_idx = veb->veb_idx;
9051 	} else {
9052 		/* floating VEB */
9053 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9054 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
9055 	}
9056 
9057 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
9058 	i40e_veb_clear(veb);
9059 }
9060 
9061 /**
9062  * i40e_add_veb - create the VEB in the switch
9063  * @veb: the VEB to be instantiated
9064  * @vsi: the controlling VSI
9065  **/
9066 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
9067 {
9068 	bool is_default = false;
9069 	bool is_cloud = false;
9070 	int ret;
9071 
9072 	/* get a VEB from the hardware */
9073 	ret = i40e_aq_add_veb(&veb->pf->hw, veb->uplink_seid, vsi->seid,
9074 			      veb->enabled_tc, is_default,
9075 			      is_cloud, &veb->seid, NULL);
9076 	if (ret) {
9077 		dev_info(&veb->pf->pdev->dev,
9078 			 "couldn't add VEB, err %d, aq_err %d\n",
9079 			 ret, veb->pf->hw.aq.asq_last_status);
9080 		return -EPERM;
9081 	}
9082 
9083 	/* get statistics counter */
9084 	ret = i40e_aq_get_veb_parameters(&veb->pf->hw, veb->seid, NULL, NULL,
9085 					 &veb->stats_idx, NULL, NULL, NULL);
9086 	if (ret) {
9087 		dev_info(&veb->pf->pdev->dev,
9088 			 "couldn't get VEB statistics idx, err %d, aq_err %d\n",
9089 			 ret, veb->pf->hw.aq.asq_last_status);
9090 		return -EPERM;
9091 	}
9092 	ret = i40e_veb_get_bw_info(veb);
9093 	if (ret) {
9094 		dev_info(&veb->pf->pdev->dev,
9095 			 "couldn't get VEB bw info, err %d, aq_err %d\n",
9096 			 ret, veb->pf->hw.aq.asq_last_status);
9097 		i40e_aq_delete_element(&veb->pf->hw, veb->seid, NULL);
9098 		return -ENOENT;
9099 	}
9100 
9101 	vsi->uplink_seid = veb->seid;
9102 	vsi->veb_idx = veb->idx;
9103 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9104 
9105 	return 0;
9106 }
9107 
9108 /**
9109  * i40e_veb_setup - Set up a VEB
9110  * @pf: board private structure
9111  * @flags: VEB setup flags
9112  * @uplink_seid: the switch element to link to
9113  * @vsi_seid: the initial VSI seid
9114  * @enabled_tc: Enabled TC bit-map
9115  *
9116  * This allocates the sw VEB structure and links it into the switch
9117  * It is possible and legal for this to be a duplicate of an already
9118  * existing VEB.  It is also possible for both uplink and vsi seids
9119  * to be zero, in order to create a floating VEB.
9120  *
9121  * Returns pointer to the successfully allocated VEB sw struct on
9122  * success, otherwise returns NULL on failure.
9123  **/
9124 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
9125 				u16 uplink_seid, u16 vsi_seid,
9126 				u8 enabled_tc)
9127 {
9128 	struct i40e_veb *veb, *uplink_veb = NULL;
9129 	int vsi_idx, veb_idx;
9130 	int ret;
9131 
9132 	/* if one seid is 0, the other must be 0 to create a floating relay */
9133 	if ((uplink_seid == 0 || vsi_seid == 0) &&
9134 	    (uplink_seid + vsi_seid != 0)) {
9135 		dev_info(&pf->pdev->dev,
9136 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
9137 			 uplink_seid, vsi_seid);
9138 		return NULL;
9139 	}
9140 
9141 	/* make sure there is such a vsi and uplink */
9142 	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
9143 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
9144 			break;
9145 	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
9146 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
9147 			 vsi_seid);
9148 		return NULL;
9149 	}
9150 
9151 	if (uplink_seid && uplink_seid != pf->mac_seid) {
9152 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
9153 			if (pf->veb[veb_idx] &&
9154 			    pf->veb[veb_idx]->seid == uplink_seid) {
9155 				uplink_veb = pf->veb[veb_idx];
9156 				break;
9157 			}
9158 		}
9159 		if (!uplink_veb) {
9160 			dev_info(&pf->pdev->dev,
9161 				 "uplink seid %d not found\n", uplink_seid);
9162 			return NULL;
9163 		}
9164 	}
9165 
9166 	/* get veb sw struct */
9167 	veb_idx = i40e_veb_mem_alloc(pf);
9168 	if (veb_idx < 0)
9169 		goto err_alloc;
9170 	veb = pf->veb[veb_idx];
9171 	veb->flags = flags;
9172 	veb->uplink_seid = uplink_seid;
9173 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
9174 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
9175 
9176 	/* create the VEB in the switch */
9177 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
9178 	if (ret)
9179 		goto err_veb;
9180 	if (vsi_idx == pf->lan_vsi)
9181 		pf->lan_veb = veb->idx;
9182 
9183 	return veb;
9184 
9185 err_veb:
9186 	i40e_veb_clear(veb);
9187 err_alloc:
9188 	return NULL;
9189 }
9190 
9191 /**
9192  * i40e_setup_pf_switch_element - set PF vars based on switch type
9193  * @pf: board private structure
9194  * @ele: element we are building info from
9195  * @num_reported: total number of elements
9196  * @printconfig: should we print the contents
9197  *
9198  * helper function to assist in extracting a few useful SEID values.
9199  **/
9200 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
9201 				struct i40e_aqc_switch_config_element_resp *ele,
9202 				u16 num_reported, bool printconfig)
9203 {
9204 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
9205 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
9206 	u8 element_type = ele->element_type;
9207 	u16 seid = le16_to_cpu(ele->seid);
9208 
9209 	if (printconfig)
9210 		dev_info(&pf->pdev->dev,
9211 			 "type=%d seid=%d uplink=%d downlink=%d\n",
9212 			 element_type, seid, uplink_seid, downlink_seid);
9213 
9214 	switch (element_type) {
9215 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
9216 		pf->mac_seid = seid;
9217 		break;
9218 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
9219 		/* Main VEB? */
9220 		if (uplink_seid != pf->mac_seid)
9221 			break;
9222 		if (pf->lan_veb == I40E_NO_VEB) {
9223 			int v;
9224 
9225 			/* find existing or else empty VEB */
9226 			for (v = 0; v < I40E_MAX_VEB; v++) {
9227 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
9228 					pf->lan_veb = v;
9229 					break;
9230 				}
9231 			}
9232 			if (pf->lan_veb == I40E_NO_VEB) {
9233 				v = i40e_veb_mem_alloc(pf);
9234 				if (v < 0)
9235 					break;
9236 				pf->lan_veb = v;
9237 			}
9238 		}
9239 
9240 		pf->veb[pf->lan_veb]->seid = seid;
9241 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
9242 		pf->veb[pf->lan_veb]->pf = pf;
9243 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
9244 		break;
9245 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
9246 		if (num_reported != 1)
9247 			break;
9248 		/* This is immediately after a reset so we can assume this is
9249 		 * the PF's VSI
9250 		 */
9251 		pf->mac_seid = uplink_seid;
9252 		pf->pf_seid = downlink_seid;
9253 		pf->main_vsi_seid = seid;
9254 		if (printconfig)
9255 			dev_info(&pf->pdev->dev,
9256 				 "pf_seid=%d main_vsi_seid=%d\n",
9257 				 pf->pf_seid, pf->main_vsi_seid);
9258 		break;
9259 	case I40E_SWITCH_ELEMENT_TYPE_PF:
9260 	case I40E_SWITCH_ELEMENT_TYPE_VF:
9261 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
9262 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
9263 	case I40E_SWITCH_ELEMENT_TYPE_PE:
9264 	case I40E_SWITCH_ELEMENT_TYPE_PA:
9265 		/* ignore these for now */
9266 		break;
9267 	default:
9268 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
9269 			 element_type, seid);
9270 		break;
9271 	}
9272 }
9273 
9274 /**
9275  * i40e_fetch_switch_configuration - Get switch config from firmware
9276  * @pf: board private structure
9277  * @printconfig: should we print the contents
9278  *
9279  * Get the current switch configuration from the device and
9280  * extract a few useful SEID values.
9281  **/
9282 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
9283 {
9284 	struct i40e_aqc_get_switch_config_resp *sw_config;
9285 	u16 next_seid = 0;
9286 	int ret = 0;
9287 	u8 *aq_buf;
9288 	int i;
9289 
9290 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
9291 	if (!aq_buf)
9292 		return -ENOMEM;
9293 
9294 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
9295 	do {
9296 		u16 num_reported, num_total;
9297 
9298 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
9299 						I40E_AQ_LARGE_BUF,
9300 						&next_seid, NULL);
9301 		if (ret) {
9302 			dev_info(&pf->pdev->dev,
9303 				 "get switch config failed %d aq_err=%x\n",
9304 				 ret, pf->hw.aq.asq_last_status);
9305 			kfree(aq_buf);
9306 			return -ENOENT;
9307 		}
9308 
9309 		num_reported = le16_to_cpu(sw_config->header.num_reported);
9310 		num_total = le16_to_cpu(sw_config->header.num_total);
9311 
9312 		if (printconfig)
9313 			dev_info(&pf->pdev->dev,
9314 				 "header: %d reported %d total\n",
9315 				 num_reported, num_total);
9316 
9317 		for (i = 0; i < num_reported; i++) {
9318 			struct i40e_aqc_switch_config_element_resp *ele =
9319 				&sw_config->element[i];
9320 
9321 			i40e_setup_pf_switch_element(pf, ele, num_reported,
9322 						     printconfig);
9323 		}
9324 	} while (next_seid != 0);
9325 
9326 	kfree(aq_buf);
9327 	return ret;
9328 }
9329 
9330 /**
9331  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
9332  * @pf: board private structure
9333  * @reinit: if the Main VSI needs to re-initialized.
9334  *
9335  * Returns 0 on success, negative value on failure
9336  **/
9337 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
9338 {
9339 	int ret;
9340 
9341 	/* find out what's out there already */
9342 	ret = i40e_fetch_switch_configuration(pf, false);
9343 	if (ret) {
9344 		dev_info(&pf->pdev->dev,
9345 			 "couldn't fetch switch config, err %d, aq_err %d\n",
9346 			 ret, pf->hw.aq.asq_last_status);
9347 		return ret;
9348 	}
9349 	i40e_pf_reset_stats(pf);
9350 
9351 	/* first time setup */
9352 	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
9353 		struct i40e_vsi *vsi = NULL;
9354 		u16 uplink_seid;
9355 
9356 		/* Set up the PF VSI associated with the PF's main VSI
9357 		 * that is already in the HW switch
9358 		 */
9359 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
9360 			uplink_seid = pf->veb[pf->lan_veb]->seid;
9361 		else
9362 			uplink_seid = pf->mac_seid;
9363 		if (pf->lan_vsi == I40E_NO_VSI)
9364 			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
9365 		else if (reinit)
9366 			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
9367 		if (!vsi) {
9368 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
9369 			i40e_fdir_teardown(pf);
9370 			return -EAGAIN;
9371 		}
9372 	} else {
9373 		/* force a reset of TC and queue layout configurations */
9374 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9375 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9376 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9377 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9378 	}
9379 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
9380 
9381 	i40e_fdir_sb_setup(pf);
9382 
9383 	/* Setup static PF queue filter control settings */
9384 	ret = i40e_setup_pf_filter_control(pf);
9385 	if (ret) {
9386 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
9387 			 ret);
9388 		/* Failure here should not stop continuing other steps */
9389 	}
9390 
9391 	/* enable RSS in the HW, even for only one queue, as the stack can use
9392 	 * the hash
9393 	 */
9394 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
9395 		i40e_config_rss(pf);
9396 
9397 	/* fill in link information and enable LSE reporting */
9398 	i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
9399 	i40e_link_event(pf);
9400 
9401 	/* Initialize user-specific link properties */
9402 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
9403 				  I40E_AQ_AN_COMPLETED) ? true : false);
9404 
9405 	i40e_ptp_init(pf);
9406 
9407 	return ret;
9408 }
9409 
9410 /**
9411  * i40e_determine_queue_usage - Work out queue distribution
9412  * @pf: board private structure
9413  **/
9414 static void i40e_determine_queue_usage(struct i40e_pf *pf)
9415 {
9416 	int queues_left;
9417 
9418 	pf->num_lan_qps = 0;
9419 #ifdef I40E_FCOE
9420 	pf->num_fcoe_qps = 0;
9421 #endif
9422 
9423 	/* Find the max queues to be put into basic use.  We'll always be
9424 	 * using TC0, whether or not DCB is running, and TC0 will get the
9425 	 * big RSS set.
9426 	 */
9427 	queues_left = pf->hw.func_caps.num_tx_qp;
9428 
9429 	if ((queues_left == 1) ||
9430 	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
9431 		/* one qp for PF, no queues for anything else */
9432 		queues_left = 0;
9433 		pf->rss_size = pf->num_lan_qps = 1;
9434 
9435 		/* make sure all the fancies are disabled */
9436 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
9437 #ifdef I40E_FCOE
9438 			       I40E_FLAG_FCOE_ENABLED	|
9439 #endif
9440 			       I40E_FLAG_FD_SB_ENABLED	|
9441 			       I40E_FLAG_FD_ATR_ENABLED	|
9442 			       I40E_FLAG_DCB_CAPABLE	|
9443 			       I40E_FLAG_SRIOV_ENABLED	|
9444 			       I40E_FLAG_VMDQ_ENABLED);
9445 	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
9446 				  I40E_FLAG_FD_SB_ENABLED |
9447 				  I40E_FLAG_FD_ATR_ENABLED |
9448 				  I40E_FLAG_DCB_CAPABLE))) {
9449 		/* one qp for PF */
9450 		pf->rss_size = pf->num_lan_qps = 1;
9451 		queues_left -= pf->num_lan_qps;
9452 
9453 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
9454 #ifdef I40E_FCOE
9455 			       I40E_FLAG_FCOE_ENABLED	|
9456 #endif
9457 			       I40E_FLAG_FD_SB_ENABLED	|
9458 			       I40E_FLAG_FD_ATR_ENABLED	|
9459 			       I40E_FLAG_DCB_ENABLED	|
9460 			       I40E_FLAG_VMDQ_ENABLED);
9461 	} else {
9462 		/* Not enough queues for all TCs */
9463 		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
9464 		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
9465 			pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
9466 			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
9467 		}
9468 		pf->num_lan_qps = max_t(int, pf->rss_size_max,
9469 					num_online_cpus());
9470 		pf->num_lan_qps = min_t(int, pf->num_lan_qps,
9471 					pf->hw.func_caps.num_tx_qp);
9472 
9473 		queues_left -= pf->num_lan_qps;
9474 	}
9475 
9476 #ifdef I40E_FCOE
9477 	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
9478 		if (I40E_DEFAULT_FCOE <= queues_left) {
9479 			pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
9480 		} else if (I40E_MINIMUM_FCOE <= queues_left) {
9481 			pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
9482 		} else {
9483 			pf->num_fcoe_qps = 0;
9484 			pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
9485 			dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
9486 		}
9487 
9488 		queues_left -= pf->num_fcoe_qps;
9489 	}
9490 
9491 #endif
9492 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9493 		if (queues_left > 1) {
9494 			queues_left -= 1; /* save 1 queue for FD */
9495 		} else {
9496 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
9497 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
9498 		}
9499 	}
9500 
9501 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
9502 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
9503 		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
9504 					(queues_left / pf->num_vf_qps));
9505 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
9506 	}
9507 
9508 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
9509 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
9510 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
9511 					  (queues_left / pf->num_vmdq_qps));
9512 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
9513 	}
9514 
9515 	pf->queues_left = queues_left;
9516 #ifdef I40E_FCOE
9517 	dev_info(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
9518 #endif
9519 }
9520 
9521 /**
9522  * i40e_setup_pf_filter_control - Setup PF static filter control
9523  * @pf: PF to be setup
9524  *
9525  * i40e_setup_pf_filter_control sets up a PF's initial filter control
9526  * settings. If PE/FCoE are enabled then it will also set the per PF
9527  * based filter sizes required for them. It also enables Flow director,
9528  * ethertype and macvlan type filter settings for the pf.
9529  *
9530  * Returns 0 on success, negative on failure
9531  **/
9532 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
9533 {
9534 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
9535 
9536 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
9537 
9538 	/* Flow Director is enabled */
9539 	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
9540 		settings->enable_fdir = true;
9541 
9542 	/* Ethtype and MACVLAN filters enabled for PF */
9543 	settings->enable_ethtype = true;
9544 	settings->enable_macvlan = true;
9545 
9546 	if (i40e_set_filter_control(&pf->hw, settings))
9547 		return -ENOENT;
9548 
9549 	return 0;
9550 }
9551 
9552 #define INFO_STRING_LEN 255
9553 static void i40e_print_features(struct i40e_pf *pf)
9554 {
9555 	struct i40e_hw *hw = &pf->hw;
9556 	char *buf, *string;
9557 
9558 	string = kzalloc(INFO_STRING_LEN, GFP_KERNEL);
9559 	if (!string) {
9560 		dev_err(&pf->pdev->dev, "Features string allocation failed\n");
9561 		return;
9562 	}
9563 
9564 	buf = string;
9565 
9566 	buf += sprintf(string, "Features: PF-id[%d] ", hw->pf_id);
9567 #ifdef CONFIG_PCI_IOV
9568 	buf += sprintf(buf, "VFs: %d ", pf->num_req_vfs);
9569 #endif
9570 	buf += sprintf(buf, "VSIs: %d QP: %d RX: %s ",
9571 		       pf->hw.func_caps.num_vsis,
9572 		       pf->vsi[pf->lan_vsi]->num_queue_pairs,
9573 		       pf->flags & I40E_FLAG_RX_PS_ENABLED ? "PS" : "1BUF");
9574 
9575 	if (pf->flags & I40E_FLAG_RSS_ENABLED)
9576 		buf += sprintf(buf, "RSS ");
9577 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
9578 		buf += sprintf(buf, "FD_ATR ");
9579 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9580 		buf += sprintf(buf, "FD_SB ");
9581 		buf += sprintf(buf, "NTUPLE ");
9582 	}
9583 	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
9584 		buf += sprintf(buf, "DCB ");
9585 	if (pf->flags & I40E_FLAG_PTP)
9586 		buf += sprintf(buf, "PTP ");
9587 #ifdef I40E_FCOE
9588 	if (pf->flags & I40E_FLAG_FCOE_ENABLED)
9589 		buf += sprintf(buf, "FCOE ");
9590 #endif
9591 
9592 	BUG_ON(buf > (string + INFO_STRING_LEN));
9593 	dev_info(&pf->pdev->dev, "%s\n", string);
9594 	kfree(string);
9595 }
9596 
9597 /**
9598  * i40e_probe - Device initialization routine
9599  * @pdev: PCI device information struct
9600  * @ent: entry in i40e_pci_tbl
9601  *
9602  * i40e_probe initializes a PF identified by a pci_dev structure.
9603  * The OS initialization, configuring of the PF private structure,
9604  * and a hardware reset occur.
9605  *
9606  * Returns 0 on success, negative on failure
9607  **/
9608 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
9609 {
9610 	struct i40e_aq_get_phy_abilities_resp abilities;
9611 	unsigned long ioremap_len;
9612 	struct i40e_pf *pf;
9613 	struct i40e_hw *hw;
9614 	static u16 pfs_found;
9615 	u16 link_status;
9616 	int err = 0;
9617 	u32 len;
9618 	u32 i;
9619 
9620 	err = pci_enable_device_mem(pdev);
9621 	if (err)
9622 		return err;
9623 
9624 	/* set up for high or low dma */
9625 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
9626 	if (err) {
9627 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
9628 		if (err) {
9629 			dev_err(&pdev->dev,
9630 				"DMA configuration failed: 0x%x\n", err);
9631 			goto err_dma;
9632 		}
9633 	}
9634 
9635 	/* set up pci connections */
9636 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
9637 					   IORESOURCE_MEM), i40e_driver_name);
9638 	if (err) {
9639 		dev_info(&pdev->dev,
9640 			 "pci_request_selected_regions failed %d\n", err);
9641 		goto err_pci_reg;
9642 	}
9643 
9644 	pci_enable_pcie_error_reporting(pdev);
9645 	pci_set_master(pdev);
9646 
9647 	/* Now that we have a PCI connection, we need to do the
9648 	 * low level device setup.  This is primarily setting up
9649 	 * the Admin Queue structures and then querying for the
9650 	 * device's current profile information.
9651 	 */
9652 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
9653 	if (!pf) {
9654 		err = -ENOMEM;
9655 		goto err_pf_alloc;
9656 	}
9657 	pf->next_vsi = 0;
9658 	pf->pdev = pdev;
9659 	set_bit(__I40E_DOWN, &pf->state);
9660 
9661 	hw = &pf->hw;
9662 	hw->back = pf;
9663 
9664 	ioremap_len = min_t(unsigned long, pci_resource_len(pdev, 0),
9665 			    I40E_MAX_CSR_SPACE);
9666 
9667 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), ioremap_len);
9668 	if (!hw->hw_addr) {
9669 		err = -EIO;
9670 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
9671 			 (unsigned int)pci_resource_start(pdev, 0),
9672 			 (unsigned int)pci_resource_len(pdev, 0), err);
9673 		goto err_ioremap;
9674 	}
9675 	hw->vendor_id = pdev->vendor;
9676 	hw->device_id = pdev->device;
9677 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
9678 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
9679 	hw->subsystem_device_id = pdev->subsystem_device;
9680 	hw->bus.device = PCI_SLOT(pdev->devfn);
9681 	hw->bus.func = PCI_FUNC(pdev->devfn);
9682 	pf->instance = pfs_found;
9683 
9684 	if (debug != -1) {
9685 		pf->msg_enable = pf->hw.debug_mask;
9686 		pf->msg_enable = debug;
9687 	}
9688 
9689 	/* do a special CORER for clearing PXE mode once at init */
9690 	if (hw->revision_id == 0 &&
9691 	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
9692 		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
9693 		i40e_flush(hw);
9694 		msleep(200);
9695 		pf->corer_count++;
9696 
9697 		i40e_clear_pxe_mode(hw);
9698 	}
9699 
9700 	/* Reset here to make sure all is clean and to define PF 'n' */
9701 	i40e_clear_hw(hw);
9702 	err = i40e_pf_reset(hw);
9703 	if (err) {
9704 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
9705 		goto err_pf_reset;
9706 	}
9707 	pf->pfr_count++;
9708 
9709 	hw->aq.num_arq_entries = I40E_AQ_LEN;
9710 	hw->aq.num_asq_entries = I40E_AQ_LEN;
9711 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
9712 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
9713 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
9714 
9715 	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
9716 		 "%s-%s:misc",
9717 		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
9718 
9719 	err = i40e_init_shared_code(hw);
9720 	if (err) {
9721 		dev_info(&pdev->dev, "init_shared_code failed: %d\n", err);
9722 		goto err_pf_reset;
9723 	}
9724 
9725 	/* set up a default setting for link flow control */
9726 	pf->hw.fc.requested_mode = I40E_FC_NONE;
9727 
9728 	err = i40e_init_adminq(hw);
9729 	dev_info(&pdev->dev, "%s\n", i40e_fw_version_str(hw));
9730 	if (err) {
9731 		dev_info(&pdev->dev,
9732 			 "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");
9733 		goto err_pf_reset;
9734 	}
9735 
9736 	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
9737 	    hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
9738 		dev_info(&pdev->dev,
9739 			 "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");
9740 	else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
9741 		 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
9742 		dev_info(&pdev->dev,
9743 			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
9744 
9745 	i40e_verify_eeprom(pf);
9746 
9747 	/* Rev 0 hardware was never productized */
9748 	if (hw->revision_id < 1)
9749 		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");
9750 
9751 	i40e_clear_pxe_mode(hw);
9752 	err = i40e_get_capabilities(pf);
9753 	if (err)
9754 		goto err_adminq_setup;
9755 
9756 	err = i40e_sw_init(pf);
9757 	if (err) {
9758 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
9759 		goto err_sw_init;
9760 	}
9761 
9762 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
9763 				hw->func_caps.num_rx_qp,
9764 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
9765 	if (err) {
9766 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
9767 		goto err_init_lan_hmc;
9768 	}
9769 
9770 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
9771 	if (err) {
9772 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
9773 		err = -ENOENT;
9774 		goto err_configure_lan_hmc;
9775 	}
9776 
9777 	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
9778 	 * Ignore error return codes because if it was already disabled via
9779 	 * hardware settings this will fail
9780 	 */
9781 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
9782 	    (pf->hw.aq.fw_maj_ver < 4)) {
9783 		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
9784 		i40e_aq_stop_lldp(hw, true, NULL);
9785 	}
9786 
9787 	i40e_get_mac_addr(hw, hw->mac.addr);
9788 	if (!is_valid_ether_addr(hw->mac.addr)) {
9789 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
9790 		err = -EIO;
9791 		goto err_mac_addr;
9792 	}
9793 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
9794 	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
9795 	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
9796 	if (is_valid_ether_addr(hw->mac.port_addr))
9797 		pf->flags |= I40E_FLAG_PORT_ID_VALID;
9798 #ifdef I40E_FCOE
9799 	err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
9800 	if (err)
9801 		dev_info(&pdev->dev,
9802 			 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
9803 	if (!is_valid_ether_addr(hw->mac.san_addr)) {
9804 		dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
9805 			 hw->mac.san_addr);
9806 		ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
9807 	}
9808 	dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
9809 #endif /* I40E_FCOE */
9810 
9811 	pci_set_drvdata(pdev, pf);
9812 	pci_save_state(pdev);
9813 #ifdef CONFIG_I40E_DCB
9814 	err = i40e_init_pf_dcb(pf);
9815 	if (err) {
9816 		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
9817 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
9818 		/* Continue without DCB enabled */
9819 	}
9820 #endif /* CONFIG_I40E_DCB */
9821 
9822 	/* set up periodic task facility */
9823 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
9824 	pf->service_timer_period = HZ;
9825 
9826 	INIT_WORK(&pf->service_task, i40e_service_task);
9827 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
9828 	pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
9829 	pf->link_check_timeout = jiffies;
9830 
9831 	/* WoL defaults to disabled */
9832 	pf->wol_en = false;
9833 	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
9834 
9835 	/* set up the main switch operations */
9836 	i40e_determine_queue_usage(pf);
9837 	err = i40e_init_interrupt_scheme(pf);
9838 	if (err)
9839 		goto err_switch_setup;
9840 
9841 	/* The number of VSIs reported by the FW is the minimum guaranteed
9842 	 * to us; HW supports far more and we share the remaining pool with
9843 	 * the other PFs. We allocate space for more than the guarantee with
9844 	 * the understanding that we might not get them all later.
9845 	 */
9846 	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
9847 		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
9848 	else
9849 		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
9850 
9851 	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
9852 	len = sizeof(struct i40e_vsi *) * pf->num_alloc_vsi;
9853 	pf->vsi = kzalloc(len, GFP_KERNEL);
9854 	if (!pf->vsi) {
9855 		err = -ENOMEM;
9856 		goto err_switch_setup;
9857 	}
9858 
9859 	err = i40e_setup_pf_switch(pf, false);
9860 	if (err) {
9861 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
9862 		goto err_vsis;
9863 	}
9864 	/* if FDIR VSI was set up, start it now */
9865 	for (i = 0; i < pf->num_alloc_vsi; i++) {
9866 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
9867 			i40e_vsi_open(pf->vsi[i]);
9868 			break;
9869 		}
9870 	}
9871 
9872 	/* driver is only interested in link up/down and module qualification
9873 	 * reports from firmware
9874 	 */
9875 	err = i40e_aq_set_phy_int_mask(&pf->hw,
9876 				       I40E_AQ_EVENT_LINK_UPDOWN |
9877 				       I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
9878 	if (err)
9879 		dev_info(&pf->pdev->dev, "set phy mask fail, aq_err %d\n", err);
9880 
9881 	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
9882 	    (pf->hw.aq.fw_maj_ver < 4)) {
9883 		msleep(75);
9884 		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
9885 		if (err)
9886 			dev_info(&pf->pdev->dev, "link restart failed, aq_err=%d\n",
9887 				 pf->hw.aq.asq_last_status);
9888 	}
9889 	/* The main driver is (mostly) up and happy. We need to set this state
9890 	 * before setting up the misc vector or we get a race and the vector
9891 	 * ends up disabled forever.
9892 	 */
9893 	clear_bit(__I40E_DOWN, &pf->state);
9894 
9895 	/* In case of MSIX we are going to setup the misc vector right here
9896 	 * to handle admin queue events etc. In case of legacy and MSI
9897 	 * the misc functionality and queue processing is combined in
9898 	 * the same vector and that gets setup at open.
9899 	 */
9900 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
9901 		err = i40e_setup_misc_vector(pf);
9902 		if (err) {
9903 			dev_info(&pdev->dev,
9904 				 "setup of misc vector failed: %d\n", err);
9905 			goto err_vsis;
9906 		}
9907 	}
9908 
9909 #ifdef CONFIG_PCI_IOV
9910 	/* prep for VF support */
9911 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
9912 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
9913 	    !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
9914 		u32 val;
9915 
9916 		/* disable link interrupts for VFs */
9917 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
9918 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
9919 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
9920 		i40e_flush(hw);
9921 
9922 		if (pci_num_vf(pdev)) {
9923 			dev_info(&pdev->dev,
9924 				 "Active VFs found, allocating resources.\n");
9925 			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
9926 			if (err)
9927 				dev_info(&pdev->dev,
9928 					 "Error %d allocating resources for existing VFs\n",
9929 					 err);
9930 		}
9931 	}
9932 #endif /* CONFIG_PCI_IOV */
9933 
9934 	pfs_found++;
9935 
9936 	i40e_dbg_pf_init(pf);
9937 
9938 	/* tell the firmware that we're starting */
9939 	i40e_send_version(pf);
9940 
9941 	/* since everything's happy, start the service_task timer */
9942 	mod_timer(&pf->service_timer,
9943 		  round_jiffies(jiffies + pf->service_timer_period));
9944 
9945 #ifdef I40E_FCOE
9946 	/* create FCoE interface */
9947 	i40e_fcoe_vsi_setup(pf);
9948 
9949 #endif
9950 	/* Get the negotiated link width and speed from PCI config space */
9951 	pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA, &link_status);
9952 
9953 	i40e_set_pci_config_data(hw, link_status);
9954 
9955 	dev_info(&pdev->dev, "PCI-Express: %s %s\n",
9956 		(hw->bus.speed == i40e_bus_speed_8000 ? "Speed 8.0GT/s" :
9957 		 hw->bus.speed == i40e_bus_speed_5000 ? "Speed 5.0GT/s" :
9958 		 hw->bus.speed == i40e_bus_speed_2500 ? "Speed 2.5GT/s" :
9959 		 "Unknown"),
9960 		(hw->bus.width == i40e_bus_width_pcie_x8 ? "Width x8" :
9961 		 hw->bus.width == i40e_bus_width_pcie_x4 ? "Width x4" :
9962 		 hw->bus.width == i40e_bus_width_pcie_x2 ? "Width x2" :
9963 		 hw->bus.width == i40e_bus_width_pcie_x1 ? "Width x1" :
9964 		 "Unknown"));
9965 
9966 	if (hw->bus.width < i40e_bus_width_pcie_x8 ||
9967 	    hw->bus.speed < i40e_bus_speed_8000) {
9968 		dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
9969 		dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
9970 	}
9971 
9972 	/* get the requested speeds from the fw */
9973 	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
9974 	if (err)
9975 		dev_info(&pf->pdev->dev, "get phy abilities failed, aq_err %d, advertised speed settings may not be correct\n",
9976 			 err);
9977 	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
9978 
9979 	/* print a string summarizing features */
9980 	i40e_print_features(pf);
9981 
9982 	return 0;
9983 
9984 	/* Unwind what we've done if something failed in the setup */
9985 err_vsis:
9986 	set_bit(__I40E_DOWN, &pf->state);
9987 	i40e_clear_interrupt_scheme(pf);
9988 	kfree(pf->vsi);
9989 err_switch_setup:
9990 	i40e_reset_interrupt_capability(pf);
9991 	del_timer_sync(&pf->service_timer);
9992 err_mac_addr:
9993 err_configure_lan_hmc:
9994 	(void)i40e_shutdown_lan_hmc(hw);
9995 err_init_lan_hmc:
9996 	kfree(pf->qp_pile);
9997 err_sw_init:
9998 err_adminq_setup:
9999 	(void)i40e_shutdown_adminq(hw);
10000 err_pf_reset:
10001 	iounmap(hw->hw_addr);
10002 err_ioremap:
10003 	kfree(pf);
10004 err_pf_alloc:
10005 	pci_disable_pcie_error_reporting(pdev);
10006 	pci_release_selected_regions(pdev,
10007 				     pci_select_bars(pdev, IORESOURCE_MEM));
10008 err_pci_reg:
10009 err_dma:
10010 	pci_disable_device(pdev);
10011 	return err;
10012 }
10013 
10014 /**
10015  * i40e_remove - Device removal routine
10016  * @pdev: PCI device information struct
10017  *
10018  * i40e_remove is called by the PCI subsystem to alert the driver
10019  * that is should release a PCI device.  This could be caused by a
10020  * Hot-Plug event, or because the driver is going to be removed from
10021  * memory.
10022  **/
10023 static void i40e_remove(struct pci_dev *pdev)
10024 {
10025 	struct i40e_pf *pf = pci_get_drvdata(pdev);
10026 	i40e_status ret_code;
10027 	int i;
10028 
10029 	i40e_dbg_pf_exit(pf);
10030 
10031 	i40e_ptp_stop(pf);
10032 
10033 	/* no more scheduling of any task */
10034 	set_bit(__I40E_DOWN, &pf->state);
10035 	del_timer_sync(&pf->service_timer);
10036 	cancel_work_sync(&pf->service_task);
10037 	i40e_fdir_teardown(pf);
10038 
10039 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
10040 		i40e_free_vfs(pf);
10041 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
10042 	}
10043 
10044 	i40e_fdir_teardown(pf);
10045 
10046 	/* If there is a switch structure or any orphans, remove them.
10047 	 * This will leave only the PF's VSI remaining.
10048 	 */
10049 	for (i = 0; i < I40E_MAX_VEB; i++) {
10050 		if (!pf->veb[i])
10051 			continue;
10052 
10053 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
10054 		    pf->veb[i]->uplink_seid == 0)
10055 			i40e_switch_branch_release(pf->veb[i]);
10056 	}
10057 
10058 	/* Now we can shutdown the PF's VSI, just before we kill
10059 	 * adminq and hmc.
10060 	 */
10061 	if (pf->vsi[pf->lan_vsi])
10062 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
10063 
10064 	/* shutdown and destroy the HMC */
10065 	if (pf->hw.hmc.hmc_obj) {
10066 		ret_code = i40e_shutdown_lan_hmc(&pf->hw);
10067 		if (ret_code)
10068 			dev_warn(&pdev->dev,
10069 				 "Failed to destroy the HMC resources: %d\n",
10070 				 ret_code);
10071 	}
10072 
10073 	/* shutdown the adminq */
10074 	ret_code = i40e_shutdown_adminq(&pf->hw);
10075 	if (ret_code)
10076 		dev_warn(&pdev->dev,
10077 			 "Failed to destroy the Admin Queue resources: %d\n",
10078 			 ret_code);
10079 
10080 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
10081 	i40e_clear_interrupt_scheme(pf);
10082 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10083 		if (pf->vsi[i]) {
10084 			i40e_vsi_clear_rings(pf->vsi[i]);
10085 			i40e_vsi_clear(pf->vsi[i]);
10086 			pf->vsi[i] = NULL;
10087 		}
10088 	}
10089 
10090 	for (i = 0; i < I40E_MAX_VEB; i++) {
10091 		kfree(pf->veb[i]);
10092 		pf->veb[i] = NULL;
10093 	}
10094 
10095 	kfree(pf->qp_pile);
10096 	kfree(pf->vsi);
10097 
10098 	iounmap(pf->hw.hw_addr);
10099 	kfree(pf);
10100 	pci_release_selected_regions(pdev,
10101 				     pci_select_bars(pdev, IORESOURCE_MEM));
10102 
10103 	pci_disable_pcie_error_reporting(pdev);
10104 	pci_disable_device(pdev);
10105 }
10106 
10107 /**
10108  * i40e_pci_error_detected - warning that something funky happened in PCI land
10109  * @pdev: PCI device information struct
10110  *
10111  * Called to warn that something happened and the error handling steps
10112  * are in progress.  Allows the driver to quiesce things, be ready for
10113  * remediation.
10114  **/
10115 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
10116 						enum pci_channel_state error)
10117 {
10118 	struct i40e_pf *pf = pci_get_drvdata(pdev);
10119 
10120 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
10121 
10122 	/* shutdown all operations */
10123 	if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
10124 		rtnl_lock();
10125 		i40e_prep_for_reset(pf);
10126 		rtnl_unlock();
10127 	}
10128 
10129 	/* Request a slot reset */
10130 	return PCI_ERS_RESULT_NEED_RESET;
10131 }
10132 
10133 /**
10134  * i40e_pci_error_slot_reset - a PCI slot reset just happened
10135  * @pdev: PCI device information struct
10136  *
10137  * Called to find if the driver can work with the device now that
10138  * the pci slot has been reset.  If a basic connection seems good
10139  * (registers are readable and have sane content) then return a
10140  * happy little PCI_ERS_RESULT_xxx.
10141  **/
10142 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
10143 {
10144 	struct i40e_pf *pf = pci_get_drvdata(pdev);
10145 	pci_ers_result_t result;
10146 	int err;
10147 	u32 reg;
10148 
10149 	dev_info(&pdev->dev, "%s\n", __func__);
10150 	if (pci_enable_device_mem(pdev)) {
10151 		dev_info(&pdev->dev,
10152 			 "Cannot re-enable PCI device after reset.\n");
10153 		result = PCI_ERS_RESULT_DISCONNECT;
10154 	} else {
10155 		pci_set_master(pdev);
10156 		pci_restore_state(pdev);
10157 		pci_save_state(pdev);
10158 		pci_wake_from_d3(pdev, false);
10159 
10160 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
10161 		if (reg == 0)
10162 			result = PCI_ERS_RESULT_RECOVERED;
10163 		else
10164 			result = PCI_ERS_RESULT_DISCONNECT;
10165 	}
10166 
10167 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10168 	if (err) {
10169 		dev_info(&pdev->dev,
10170 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
10171 			 err);
10172 		/* non-fatal, continue */
10173 	}
10174 
10175 	return result;
10176 }
10177 
10178 /**
10179  * i40e_pci_error_resume - restart operations after PCI error recovery
10180  * @pdev: PCI device information struct
10181  *
10182  * Called to allow the driver to bring things back up after PCI error
10183  * and/or reset recovery has finished.
10184  **/
10185 static void i40e_pci_error_resume(struct pci_dev *pdev)
10186 {
10187 	struct i40e_pf *pf = pci_get_drvdata(pdev);
10188 
10189 	dev_info(&pdev->dev, "%s\n", __func__);
10190 	if (test_bit(__I40E_SUSPENDED, &pf->state))
10191 		return;
10192 
10193 	rtnl_lock();
10194 	i40e_handle_reset_warning(pf);
10195 	rtnl_lock();
10196 }
10197 
10198 /**
10199  * i40e_shutdown - PCI callback for shutting down
10200  * @pdev: PCI device information struct
10201  **/
10202 static void i40e_shutdown(struct pci_dev *pdev)
10203 {
10204 	struct i40e_pf *pf = pci_get_drvdata(pdev);
10205 	struct i40e_hw *hw = &pf->hw;
10206 
10207 	set_bit(__I40E_SUSPENDED, &pf->state);
10208 	set_bit(__I40E_DOWN, &pf->state);
10209 	rtnl_lock();
10210 	i40e_prep_for_reset(pf);
10211 	rtnl_unlock();
10212 
10213 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
10214 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
10215 
10216 	i40e_clear_interrupt_scheme(pf);
10217 
10218 	if (system_state == SYSTEM_POWER_OFF) {
10219 		pci_wake_from_d3(pdev, pf->wol_en);
10220 		pci_set_power_state(pdev, PCI_D3hot);
10221 	}
10222 }
10223 
10224 #ifdef CONFIG_PM
10225 /**
10226  * i40e_suspend - PCI callback for moving to D3
10227  * @pdev: PCI device information struct
10228  **/
10229 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
10230 {
10231 	struct i40e_pf *pf = pci_get_drvdata(pdev);
10232 	struct i40e_hw *hw = &pf->hw;
10233 
10234 	set_bit(__I40E_SUSPENDED, &pf->state);
10235 	set_bit(__I40E_DOWN, &pf->state);
10236 	del_timer_sync(&pf->service_timer);
10237 	cancel_work_sync(&pf->service_task);
10238 	i40e_fdir_teardown(pf);
10239 
10240 	rtnl_lock();
10241 	i40e_prep_for_reset(pf);
10242 	rtnl_unlock();
10243 
10244 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
10245 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
10246 
10247 	pci_wake_from_d3(pdev, pf->wol_en);
10248 	pci_set_power_state(pdev, PCI_D3hot);
10249 
10250 	return 0;
10251 }
10252 
10253 /**
10254  * i40e_resume - PCI callback for waking up from D3
10255  * @pdev: PCI device information struct
10256  **/
10257 static int i40e_resume(struct pci_dev *pdev)
10258 {
10259 	struct i40e_pf *pf = pci_get_drvdata(pdev);
10260 	u32 err;
10261 
10262 	pci_set_power_state(pdev, PCI_D0);
10263 	pci_restore_state(pdev);
10264 	/* pci_restore_state() clears dev->state_saves, so
10265 	 * call pci_save_state() again to restore it.
10266 	 */
10267 	pci_save_state(pdev);
10268 
10269 	err = pci_enable_device_mem(pdev);
10270 	if (err) {
10271 		dev_err(&pdev->dev,
10272 			"%s: Cannot enable PCI device from suspend\n",
10273 			__func__);
10274 		return err;
10275 	}
10276 	pci_set_master(pdev);
10277 
10278 	/* no wakeup events while running */
10279 	pci_wake_from_d3(pdev, false);
10280 
10281 	/* handling the reset will rebuild the device state */
10282 	if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
10283 		clear_bit(__I40E_DOWN, &pf->state);
10284 		rtnl_lock();
10285 		i40e_reset_and_rebuild(pf, false);
10286 		rtnl_unlock();
10287 	}
10288 
10289 	return 0;
10290 }
10291 
10292 #endif
10293 static const struct pci_error_handlers i40e_err_handler = {
10294 	.error_detected = i40e_pci_error_detected,
10295 	.slot_reset = i40e_pci_error_slot_reset,
10296 	.resume = i40e_pci_error_resume,
10297 };
10298 
10299 static struct pci_driver i40e_driver = {
10300 	.name     = i40e_driver_name,
10301 	.id_table = i40e_pci_tbl,
10302 	.probe    = i40e_probe,
10303 	.remove   = i40e_remove,
10304 #ifdef CONFIG_PM
10305 	.suspend  = i40e_suspend,
10306 	.resume   = i40e_resume,
10307 #endif
10308 	.shutdown = i40e_shutdown,
10309 	.err_handler = &i40e_err_handler,
10310 	.sriov_configure = i40e_pci_sriov_configure,
10311 };
10312 
10313 /**
10314  * i40e_init_module - Driver registration routine
10315  *
10316  * i40e_init_module is the first routine called when the driver is
10317  * loaded. All it does is register with the PCI subsystem.
10318  **/
10319 static int __init i40e_init_module(void)
10320 {
10321 	pr_info("%s: %s - version %s\n", i40e_driver_name,
10322 		i40e_driver_string, i40e_driver_version_str);
10323 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
10324 
10325 	i40e_dbg_init();
10326 	return pci_register_driver(&i40e_driver);
10327 }
10328 module_init(i40e_init_module);
10329 
10330 /**
10331  * i40e_exit_module - Driver exit cleanup routine
10332  *
10333  * i40e_exit_module is called just before the driver is removed
10334  * from memory.
10335  **/
10336 static void __exit i40e_exit_module(void)
10337 {
10338 	pci_unregister_driver(&i40e_driver);
10339 	i40e_dbg_exit();
10340 }
10341 module_exit(i40e_exit_module);
10342