xref: /linux/drivers/net/ethernet/intel/i40e/i40e_main.c (revision 1fc31357ad194fb98691f3d122bcd47e59239e83)
1 /*******************************************************************************
2  *
3  * Intel Ethernet Controller XL710 Family Linux Driver
4  * Copyright(c) 2013 - 2016 Intel Corporation.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along
16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * The full GNU General Public License is included in this distribution in
19  * the file called "COPYING".
20  *
21  * Contact Information:
22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
24  *
25  ******************************************************************************/
26 
27 #include <linux/etherdevice.h>
28 #include <linux/of_net.h>
29 #include <linux/pci.h>
30 
31 /* Local includes */
32 #include "i40e.h"
33 #include "i40e_diag.h"
34 #include <net/udp_tunnel.h>
35 
36 const char i40e_driver_name[] = "i40e";
37 static const char i40e_driver_string[] =
38 			"Intel(R) Ethernet Connection XL710 Network Driver";
39 
40 #define DRV_KERN "-k"
41 
42 #define DRV_VERSION_MAJOR 1
43 #define DRV_VERSION_MINOR 6
44 #define DRV_VERSION_BUILD 21
45 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
46 	     __stringify(DRV_VERSION_MINOR) "." \
47 	     __stringify(DRV_VERSION_BUILD)    DRV_KERN
48 const char i40e_driver_version_str[] = DRV_VERSION;
49 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
50 
51 /* a bit of forward declarations */
52 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
53 static void i40e_handle_reset_warning(struct i40e_pf *pf);
54 static int i40e_add_vsi(struct i40e_vsi *vsi);
55 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
56 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
57 static int i40e_setup_misc_vector(struct i40e_pf *pf);
58 static void i40e_determine_queue_usage(struct i40e_pf *pf);
59 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
60 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
61 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
62 
63 /* i40e_pci_tbl - PCI Device ID Table
64  *
65  * Last entry must be all 0s
66  *
67  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
68  *   Class, Class Mask, private data (not used) }
69  */
70 static const struct pci_device_id i40e_pci_tbl[] = {
71 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
72 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
73 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
74 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
75 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
76 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
77 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
78 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
79 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
80 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
81 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
82 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
83 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
84 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
85 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
86 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
87 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
88 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
89 	/* required last entry */
90 	{0, }
91 };
92 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
93 
94 #define I40E_MAX_VF_COUNT 128
95 static int debug = -1;
96 module_param(debug, uint, 0);
97 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
98 
99 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
100 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
101 MODULE_LICENSE("GPL");
102 MODULE_VERSION(DRV_VERSION);
103 
104 static struct workqueue_struct *i40e_wq;
105 
106 /**
107  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
108  * @hw:   pointer to the HW structure
109  * @mem:  ptr to mem struct to fill out
110  * @size: size of memory requested
111  * @alignment: what to align the allocation to
112  **/
113 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
114 			    u64 size, u32 alignment)
115 {
116 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
117 
118 	mem->size = ALIGN(size, alignment);
119 	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
120 				      &mem->pa, GFP_KERNEL);
121 	if (!mem->va)
122 		return -ENOMEM;
123 
124 	return 0;
125 }
126 
127 /**
128  * i40e_free_dma_mem_d - OS specific memory free for shared code
129  * @hw:   pointer to the HW structure
130  * @mem:  ptr to mem struct to free
131  **/
132 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
133 {
134 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
135 
136 	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
137 	mem->va = NULL;
138 	mem->pa = 0;
139 	mem->size = 0;
140 
141 	return 0;
142 }
143 
144 /**
145  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
146  * @hw:   pointer to the HW structure
147  * @mem:  ptr to mem struct to fill out
148  * @size: size of memory requested
149  **/
150 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
151 			     u32 size)
152 {
153 	mem->size = size;
154 	mem->va = kzalloc(size, GFP_KERNEL);
155 
156 	if (!mem->va)
157 		return -ENOMEM;
158 
159 	return 0;
160 }
161 
162 /**
163  * i40e_free_virt_mem_d - OS specific memory free for shared code
164  * @hw:   pointer to the HW structure
165  * @mem:  ptr to mem struct to free
166  **/
167 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
168 {
169 	/* it's ok to kfree a NULL pointer */
170 	kfree(mem->va);
171 	mem->va = NULL;
172 	mem->size = 0;
173 
174 	return 0;
175 }
176 
177 /**
178  * i40e_get_lump - find a lump of free generic resource
179  * @pf: board private structure
180  * @pile: the pile of resource to search
181  * @needed: the number of items needed
182  * @id: an owner id to stick on the items assigned
183  *
184  * Returns the base item index of the lump, or negative for error
185  *
186  * The search_hint trick and lack of advanced fit-finding only work
187  * because we're highly likely to have all the same size lump requests.
188  * Linear search time and any fragmentation should be minimal.
189  **/
190 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
191 			 u16 needed, u16 id)
192 {
193 	int ret = -ENOMEM;
194 	int i, j;
195 
196 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
197 		dev_info(&pf->pdev->dev,
198 			 "param err: pile=%p needed=%d id=0x%04x\n",
199 			 pile, needed, id);
200 		return -EINVAL;
201 	}
202 
203 	/* start the linear search with an imperfect hint */
204 	i = pile->search_hint;
205 	while (i < pile->num_entries) {
206 		/* skip already allocated entries */
207 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
208 			i++;
209 			continue;
210 		}
211 
212 		/* do we have enough in this lump? */
213 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
214 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
215 				break;
216 		}
217 
218 		if (j == needed) {
219 			/* there was enough, so assign it to the requestor */
220 			for (j = 0; j < needed; j++)
221 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
222 			ret = i;
223 			pile->search_hint = i + j;
224 			break;
225 		}
226 
227 		/* not enough, so skip over it and continue looking */
228 		i += j;
229 	}
230 
231 	return ret;
232 }
233 
234 /**
235  * i40e_put_lump - return a lump of generic resource
236  * @pile: the pile of resource to search
237  * @index: the base item index
238  * @id: the owner id of the items assigned
239  *
240  * Returns the count of items in the lump
241  **/
242 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
243 {
244 	int valid_id = (id | I40E_PILE_VALID_BIT);
245 	int count = 0;
246 	int i;
247 
248 	if (!pile || index >= pile->num_entries)
249 		return -EINVAL;
250 
251 	for (i = index;
252 	     i < pile->num_entries && pile->list[i] == valid_id;
253 	     i++) {
254 		pile->list[i] = 0;
255 		count++;
256 	}
257 
258 	if (count && index < pile->search_hint)
259 		pile->search_hint = index;
260 
261 	return count;
262 }
263 
264 /**
265  * i40e_find_vsi_from_id - searches for the vsi with the given id
266  * @pf - the pf structure to search for the vsi
267  * @id - id of the vsi it is searching for
268  **/
269 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
270 {
271 	int i;
272 
273 	for (i = 0; i < pf->num_alloc_vsi; i++)
274 		if (pf->vsi[i] && (pf->vsi[i]->id == id))
275 			return pf->vsi[i];
276 
277 	return NULL;
278 }
279 
280 /**
281  * i40e_service_event_schedule - Schedule the service task to wake up
282  * @pf: board private structure
283  *
284  * If not already scheduled, this puts the task into the work queue
285  **/
286 void i40e_service_event_schedule(struct i40e_pf *pf)
287 {
288 	if (!test_bit(__I40E_DOWN, &pf->state) &&
289 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
290 	    !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
291 		queue_work(i40e_wq, &pf->service_task);
292 }
293 
294 /**
295  * i40e_tx_timeout - Respond to a Tx Hang
296  * @netdev: network interface device structure
297  *
298  * If any port has noticed a Tx timeout, it is likely that the whole
299  * device is munged, not just the one netdev port, so go for the full
300  * reset.
301  **/
302 #ifdef I40E_FCOE
303 void i40e_tx_timeout(struct net_device *netdev)
304 #else
305 static void i40e_tx_timeout(struct net_device *netdev)
306 #endif
307 {
308 	struct i40e_netdev_priv *np = netdev_priv(netdev);
309 	struct i40e_vsi *vsi = np->vsi;
310 	struct i40e_pf *pf = vsi->back;
311 	struct i40e_ring *tx_ring = NULL;
312 	unsigned int i, hung_queue = 0;
313 	u32 head, val;
314 
315 	pf->tx_timeout_count++;
316 
317 	/* find the stopped queue the same way the stack does */
318 	for (i = 0; i < netdev->num_tx_queues; i++) {
319 		struct netdev_queue *q;
320 		unsigned long trans_start;
321 
322 		q = netdev_get_tx_queue(netdev, i);
323 		trans_start = q->trans_start;
324 		if (netif_xmit_stopped(q) &&
325 		    time_after(jiffies,
326 			       (trans_start + netdev->watchdog_timeo))) {
327 			hung_queue = i;
328 			break;
329 		}
330 	}
331 
332 	if (i == netdev->num_tx_queues) {
333 		netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
334 	} else {
335 		/* now that we have an index, find the tx_ring struct */
336 		for (i = 0; i < vsi->num_queue_pairs; i++) {
337 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
338 				if (hung_queue ==
339 				    vsi->tx_rings[i]->queue_index) {
340 					tx_ring = vsi->tx_rings[i];
341 					break;
342 				}
343 			}
344 		}
345 	}
346 
347 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
348 		pf->tx_timeout_recovery_level = 1;  /* reset after some time */
349 	else if (time_before(jiffies,
350 		      (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
351 		return;   /* don't do any new action before the next timeout */
352 
353 	if (tx_ring) {
354 		head = i40e_get_head(tx_ring);
355 		/* Read interrupt register */
356 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
357 			val = rd32(&pf->hw,
358 			     I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
359 						tx_ring->vsi->base_vector - 1));
360 		else
361 			val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
362 
363 		netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
364 			    vsi->seid, hung_queue, tx_ring->next_to_clean,
365 			    head, tx_ring->next_to_use,
366 			    readl(tx_ring->tail), val);
367 	}
368 
369 	pf->tx_timeout_last_recovery = jiffies;
370 	netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
371 		    pf->tx_timeout_recovery_level, hung_queue);
372 
373 	switch (pf->tx_timeout_recovery_level) {
374 	case 1:
375 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
376 		break;
377 	case 2:
378 		set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
379 		break;
380 	case 3:
381 		set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
382 		break;
383 	default:
384 		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
385 		break;
386 	}
387 
388 	i40e_service_event_schedule(pf);
389 	pf->tx_timeout_recovery_level++;
390 }
391 
392 /**
393  * i40e_get_vsi_stats_struct - Get System Network Statistics
394  * @vsi: the VSI we care about
395  *
396  * Returns the address of the device statistics structure.
397  * The statistics are actually updated from the service task.
398  **/
399 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
400 {
401 	return &vsi->net_stats;
402 }
403 
404 /**
405  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
406  * @netdev: network interface device structure
407  *
408  * Returns the address of the device statistics structure.
409  * The statistics are actually updated from the service task.
410  **/
411 #ifdef I40E_FCOE
412 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
413 					     struct net_device *netdev,
414 					     struct rtnl_link_stats64 *stats)
415 #else
416 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
417 					     struct net_device *netdev,
418 					     struct rtnl_link_stats64 *stats)
419 #endif
420 {
421 	struct i40e_netdev_priv *np = netdev_priv(netdev);
422 	struct i40e_ring *tx_ring, *rx_ring;
423 	struct i40e_vsi *vsi = np->vsi;
424 	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
425 	int i;
426 
427 	if (test_bit(__I40E_DOWN, &vsi->state))
428 		return stats;
429 
430 	if (!vsi->tx_rings)
431 		return stats;
432 
433 	rcu_read_lock();
434 	for (i = 0; i < vsi->num_queue_pairs; i++) {
435 		u64 bytes, packets;
436 		unsigned int start;
437 
438 		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
439 		if (!tx_ring)
440 			continue;
441 
442 		do {
443 			start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
444 			packets = tx_ring->stats.packets;
445 			bytes   = tx_ring->stats.bytes;
446 		} while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
447 
448 		stats->tx_packets += packets;
449 		stats->tx_bytes   += bytes;
450 		rx_ring = &tx_ring[1];
451 
452 		do {
453 			start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
454 			packets = rx_ring->stats.packets;
455 			bytes   = rx_ring->stats.bytes;
456 		} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
457 
458 		stats->rx_packets += packets;
459 		stats->rx_bytes   += bytes;
460 	}
461 	rcu_read_unlock();
462 
463 	/* following stats updated by i40e_watchdog_subtask() */
464 	stats->multicast	= vsi_stats->multicast;
465 	stats->tx_errors	= vsi_stats->tx_errors;
466 	stats->tx_dropped	= vsi_stats->tx_dropped;
467 	stats->rx_errors	= vsi_stats->rx_errors;
468 	stats->rx_dropped	= vsi_stats->rx_dropped;
469 	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
470 	stats->rx_length_errors	= vsi_stats->rx_length_errors;
471 
472 	return stats;
473 }
474 
475 /**
476  * i40e_vsi_reset_stats - Resets all stats of the given vsi
477  * @vsi: the VSI to have its stats reset
478  **/
479 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
480 {
481 	struct rtnl_link_stats64 *ns;
482 	int i;
483 
484 	if (!vsi)
485 		return;
486 
487 	ns = i40e_get_vsi_stats_struct(vsi);
488 	memset(ns, 0, sizeof(*ns));
489 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
490 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
491 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
492 	if (vsi->rx_rings && vsi->rx_rings[0]) {
493 		for (i = 0; i < vsi->num_queue_pairs; i++) {
494 			memset(&vsi->rx_rings[i]->stats, 0,
495 			       sizeof(vsi->rx_rings[i]->stats));
496 			memset(&vsi->rx_rings[i]->rx_stats, 0,
497 			       sizeof(vsi->rx_rings[i]->rx_stats));
498 			memset(&vsi->tx_rings[i]->stats, 0,
499 			       sizeof(vsi->tx_rings[i]->stats));
500 			memset(&vsi->tx_rings[i]->tx_stats, 0,
501 			       sizeof(vsi->tx_rings[i]->tx_stats));
502 		}
503 	}
504 	vsi->stat_offsets_loaded = false;
505 }
506 
507 /**
508  * i40e_pf_reset_stats - Reset all of the stats for the given PF
509  * @pf: the PF to be reset
510  **/
511 void i40e_pf_reset_stats(struct i40e_pf *pf)
512 {
513 	int i;
514 
515 	memset(&pf->stats, 0, sizeof(pf->stats));
516 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
517 	pf->stat_offsets_loaded = false;
518 
519 	for (i = 0; i < I40E_MAX_VEB; i++) {
520 		if (pf->veb[i]) {
521 			memset(&pf->veb[i]->stats, 0,
522 			       sizeof(pf->veb[i]->stats));
523 			memset(&pf->veb[i]->stats_offsets, 0,
524 			       sizeof(pf->veb[i]->stats_offsets));
525 			pf->veb[i]->stat_offsets_loaded = false;
526 		}
527 	}
528 	pf->hw_csum_rx_error = 0;
529 }
530 
531 /**
532  * i40e_stat_update48 - read and update a 48 bit stat from the chip
533  * @hw: ptr to the hardware info
534  * @hireg: the high 32 bit reg to read
535  * @loreg: the low 32 bit reg to read
536  * @offset_loaded: has the initial offset been loaded yet
537  * @offset: ptr to current offset value
538  * @stat: ptr to the stat
539  *
540  * Since the device stats are not reset at PFReset, they likely will not
541  * be zeroed when the driver starts.  We'll save the first values read
542  * and use them as offsets to be subtracted from the raw values in order
543  * to report stats that count from zero.  In the process, we also manage
544  * the potential roll-over.
545  **/
546 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
547 			       bool offset_loaded, u64 *offset, u64 *stat)
548 {
549 	u64 new_data;
550 
551 	if (hw->device_id == I40E_DEV_ID_QEMU) {
552 		new_data = rd32(hw, loreg);
553 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
554 	} else {
555 		new_data = rd64(hw, loreg);
556 	}
557 	if (!offset_loaded)
558 		*offset = new_data;
559 	if (likely(new_data >= *offset))
560 		*stat = new_data - *offset;
561 	else
562 		*stat = (new_data + BIT_ULL(48)) - *offset;
563 	*stat &= 0xFFFFFFFFFFFFULL;
564 }
565 
566 /**
567  * i40e_stat_update32 - read and update a 32 bit stat from the chip
568  * @hw: ptr to the hardware info
569  * @reg: the hw reg to read
570  * @offset_loaded: has the initial offset been loaded yet
571  * @offset: ptr to current offset value
572  * @stat: ptr to the stat
573  **/
574 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
575 			       bool offset_loaded, u64 *offset, u64 *stat)
576 {
577 	u32 new_data;
578 
579 	new_data = rd32(hw, reg);
580 	if (!offset_loaded)
581 		*offset = new_data;
582 	if (likely(new_data >= *offset))
583 		*stat = (u32)(new_data - *offset);
584 	else
585 		*stat = (u32)((new_data + BIT_ULL(32)) - *offset);
586 }
587 
588 /**
589  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
590  * @vsi: the VSI to be updated
591  **/
592 void i40e_update_eth_stats(struct i40e_vsi *vsi)
593 {
594 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
595 	struct i40e_pf *pf = vsi->back;
596 	struct i40e_hw *hw = &pf->hw;
597 	struct i40e_eth_stats *oes;
598 	struct i40e_eth_stats *es;     /* device's eth stats */
599 
600 	es = &vsi->eth_stats;
601 	oes = &vsi->eth_stats_offsets;
602 
603 	/* Gather up the stats that the hw collects */
604 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
605 			   vsi->stat_offsets_loaded,
606 			   &oes->tx_errors, &es->tx_errors);
607 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
608 			   vsi->stat_offsets_loaded,
609 			   &oes->rx_discards, &es->rx_discards);
610 	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
611 			   vsi->stat_offsets_loaded,
612 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
613 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
614 			   vsi->stat_offsets_loaded,
615 			   &oes->tx_errors, &es->tx_errors);
616 
617 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
618 			   I40E_GLV_GORCL(stat_idx),
619 			   vsi->stat_offsets_loaded,
620 			   &oes->rx_bytes, &es->rx_bytes);
621 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
622 			   I40E_GLV_UPRCL(stat_idx),
623 			   vsi->stat_offsets_loaded,
624 			   &oes->rx_unicast, &es->rx_unicast);
625 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
626 			   I40E_GLV_MPRCL(stat_idx),
627 			   vsi->stat_offsets_loaded,
628 			   &oes->rx_multicast, &es->rx_multicast);
629 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
630 			   I40E_GLV_BPRCL(stat_idx),
631 			   vsi->stat_offsets_loaded,
632 			   &oes->rx_broadcast, &es->rx_broadcast);
633 
634 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
635 			   I40E_GLV_GOTCL(stat_idx),
636 			   vsi->stat_offsets_loaded,
637 			   &oes->tx_bytes, &es->tx_bytes);
638 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
639 			   I40E_GLV_UPTCL(stat_idx),
640 			   vsi->stat_offsets_loaded,
641 			   &oes->tx_unicast, &es->tx_unicast);
642 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
643 			   I40E_GLV_MPTCL(stat_idx),
644 			   vsi->stat_offsets_loaded,
645 			   &oes->tx_multicast, &es->tx_multicast);
646 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
647 			   I40E_GLV_BPTCL(stat_idx),
648 			   vsi->stat_offsets_loaded,
649 			   &oes->tx_broadcast, &es->tx_broadcast);
650 	vsi->stat_offsets_loaded = true;
651 }
652 
653 /**
654  * i40e_update_veb_stats - Update Switch component statistics
655  * @veb: the VEB being updated
656  **/
657 static void i40e_update_veb_stats(struct i40e_veb *veb)
658 {
659 	struct i40e_pf *pf = veb->pf;
660 	struct i40e_hw *hw = &pf->hw;
661 	struct i40e_eth_stats *oes;
662 	struct i40e_eth_stats *es;     /* device's eth stats */
663 	struct i40e_veb_tc_stats *veb_oes;
664 	struct i40e_veb_tc_stats *veb_es;
665 	int i, idx = 0;
666 
667 	idx = veb->stats_idx;
668 	es = &veb->stats;
669 	oes = &veb->stats_offsets;
670 	veb_es = &veb->tc_stats;
671 	veb_oes = &veb->tc_stats_offsets;
672 
673 	/* Gather up the stats that the hw collects */
674 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
675 			   veb->stat_offsets_loaded,
676 			   &oes->tx_discards, &es->tx_discards);
677 	if (hw->revision_id > 0)
678 		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
679 				   veb->stat_offsets_loaded,
680 				   &oes->rx_unknown_protocol,
681 				   &es->rx_unknown_protocol);
682 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
683 			   veb->stat_offsets_loaded,
684 			   &oes->rx_bytes, &es->rx_bytes);
685 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
686 			   veb->stat_offsets_loaded,
687 			   &oes->rx_unicast, &es->rx_unicast);
688 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
689 			   veb->stat_offsets_loaded,
690 			   &oes->rx_multicast, &es->rx_multicast);
691 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
692 			   veb->stat_offsets_loaded,
693 			   &oes->rx_broadcast, &es->rx_broadcast);
694 
695 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
696 			   veb->stat_offsets_loaded,
697 			   &oes->tx_bytes, &es->tx_bytes);
698 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
699 			   veb->stat_offsets_loaded,
700 			   &oes->tx_unicast, &es->tx_unicast);
701 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
702 			   veb->stat_offsets_loaded,
703 			   &oes->tx_multicast, &es->tx_multicast);
704 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
705 			   veb->stat_offsets_loaded,
706 			   &oes->tx_broadcast, &es->tx_broadcast);
707 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
708 		i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
709 				   I40E_GLVEBTC_RPCL(i, idx),
710 				   veb->stat_offsets_loaded,
711 				   &veb_oes->tc_rx_packets[i],
712 				   &veb_es->tc_rx_packets[i]);
713 		i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
714 				   I40E_GLVEBTC_RBCL(i, idx),
715 				   veb->stat_offsets_loaded,
716 				   &veb_oes->tc_rx_bytes[i],
717 				   &veb_es->tc_rx_bytes[i]);
718 		i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
719 				   I40E_GLVEBTC_TPCL(i, idx),
720 				   veb->stat_offsets_loaded,
721 				   &veb_oes->tc_tx_packets[i],
722 				   &veb_es->tc_tx_packets[i]);
723 		i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
724 				   I40E_GLVEBTC_TBCL(i, idx),
725 				   veb->stat_offsets_loaded,
726 				   &veb_oes->tc_tx_bytes[i],
727 				   &veb_es->tc_tx_bytes[i]);
728 	}
729 	veb->stat_offsets_loaded = true;
730 }
731 
732 #ifdef I40E_FCOE
733 /**
734  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
735  * @vsi: the VSI that is capable of doing FCoE
736  **/
737 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
738 {
739 	struct i40e_pf *pf = vsi->back;
740 	struct i40e_hw *hw = &pf->hw;
741 	struct i40e_fcoe_stats *ofs;
742 	struct i40e_fcoe_stats *fs;     /* device's eth stats */
743 	int idx;
744 
745 	if (vsi->type != I40E_VSI_FCOE)
746 		return;
747 
748 	idx = hw->pf_id + I40E_FCOE_PF_STAT_OFFSET;
749 	fs = &vsi->fcoe_stats;
750 	ofs = &vsi->fcoe_stats_offsets;
751 
752 	i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
753 			   vsi->fcoe_stat_offsets_loaded,
754 			   &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
755 	i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
756 			   vsi->fcoe_stat_offsets_loaded,
757 			   &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
758 	i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
759 			   vsi->fcoe_stat_offsets_loaded,
760 			   &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
761 	i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
762 			   vsi->fcoe_stat_offsets_loaded,
763 			   &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
764 	i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
765 			   vsi->fcoe_stat_offsets_loaded,
766 			   &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
767 	i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
768 			   vsi->fcoe_stat_offsets_loaded,
769 			   &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
770 	i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
771 			   vsi->fcoe_stat_offsets_loaded,
772 			   &ofs->fcoe_last_error, &fs->fcoe_last_error);
773 	i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
774 			   vsi->fcoe_stat_offsets_loaded,
775 			   &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
776 
777 	vsi->fcoe_stat_offsets_loaded = true;
778 }
779 
780 #endif
781 /**
782  * i40e_update_vsi_stats - Update the vsi statistics counters.
783  * @vsi: the VSI to be updated
784  *
785  * There are a few instances where we store the same stat in a
786  * couple of different structs.  This is partly because we have
787  * the netdev stats that need to be filled out, which is slightly
788  * different from the "eth_stats" defined by the chip and used in
789  * VF communications.  We sort it out here.
790  **/
791 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
792 {
793 	struct i40e_pf *pf = vsi->back;
794 	struct rtnl_link_stats64 *ons;
795 	struct rtnl_link_stats64 *ns;   /* netdev stats */
796 	struct i40e_eth_stats *oes;
797 	struct i40e_eth_stats *es;     /* device's eth stats */
798 	u32 tx_restart, tx_busy;
799 	u64 tx_lost_interrupt;
800 	struct i40e_ring *p;
801 	u32 rx_page, rx_buf;
802 	u64 bytes, packets;
803 	unsigned int start;
804 	u64 tx_linearize;
805 	u64 tx_force_wb;
806 	u64 rx_p, rx_b;
807 	u64 tx_p, tx_b;
808 	u16 q;
809 
810 	if (test_bit(__I40E_DOWN, &vsi->state) ||
811 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
812 		return;
813 
814 	ns = i40e_get_vsi_stats_struct(vsi);
815 	ons = &vsi->net_stats_offsets;
816 	es = &vsi->eth_stats;
817 	oes = &vsi->eth_stats_offsets;
818 
819 	/* Gather up the netdev and vsi stats that the driver collects
820 	 * on the fly during packet processing
821 	 */
822 	rx_b = rx_p = 0;
823 	tx_b = tx_p = 0;
824 	tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
825 	tx_lost_interrupt = 0;
826 	rx_page = 0;
827 	rx_buf = 0;
828 	rcu_read_lock();
829 	for (q = 0; q < vsi->num_queue_pairs; q++) {
830 		/* locate Tx ring */
831 		p = ACCESS_ONCE(vsi->tx_rings[q]);
832 
833 		do {
834 			start = u64_stats_fetch_begin_irq(&p->syncp);
835 			packets = p->stats.packets;
836 			bytes = p->stats.bytes;
837 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
838 		tx_b += bytes;
839 		tx_p += packets;
840 		tx_restart += p->tx_stats.restart_queue;
841 		tx_busy += p->tx_stats.tx_busy;
842 		tx_linearize += p->tx_stats.tx_linearize;
843 		tx_force_wb += p->tx_stats.tx_force_wb;
844 		tx_lost_interrupt += p->tx_stats.tx_lost_interrupt;
845 
846 		/* Rx queue is part of the same block as Tx queue */
847 		p = &p[1];
848 		do {
849 			start = u64_stats_fetch_begin_irq(&p->syncp);
850 			packets = p->stats.packets;
851 			bytes = p->stats.bytes;
852 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
853 		rx_b += bytes;
854 		rx_p += packets;
855 		rx_buf += p->rx_stats.alloc_buff_failed;
856 		rx_page += p->rx_stats.alloc_page_failed;
857 	}
858 	rcu_read_unlock();
859 	vsi->tx_restart = tx_restart;
860 	vsi->tx_busy = tx_busy;
861 	vsi->tx_linearize = tx_linearize;
862 	vsi->tx_force_wb = tx_force_wb;
863 	vsi->tx_lost_interrupt = tx_lost_interrupt;
864 	vsi->rx_page_failed = rx_page;
865 	vsi->rx_buf_failed = rx_buf;
866 
867 	ns->rx_packets = rx_p;
868 	ns->rx_bytes = rx_b;
869 	ns->tx_packets = tx_p;
870 	ns->tx_bytes = tx_b;
871 
872 	/* update netdev stats from eth stats */
873 	i40e_update_eth_stats(vsi);
874 	ons->tx_errors = oes->tx_errors;
875 	ns->tx_errors = es->tx_errors;
876 	ons->multicast = oes->rx_multicast;
877 	ns->multicast = es->rx_multicast;
878 	ons->rx_dropped = oes->rx_discards;
879 	ns->rx_dropped = es->rx_discards;
880 	ons->tx_dropped = oes->tx_discards;
881 	ns->tx_dropped = es->tx_discards;
882 
883 	/* pull in a couple PF stats if this is the main vsi */
884 	if (vsi == pf->vsi[pf->lan_vsi]) {
885 		ns->rx_crc_errors = pf->stats.crc_errors;
886 		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
887 		ns->rx_length_errors = pf->stats.rx_length_errors;
888 	}
889 }
890 
891 /**
892  * i40e_update_pf_stats - Update the PF statistics counters.
893  * @pf: the PF to be updated
894  **/
895 static void i40e_update_pf_stats(struct i40e_pf *pf)
896 {
897 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
898 	struct i40e_hw_port_stats *nsd = &pf->stats;
899 	struct i40e_hw *hw = &pf->hw;
900 	u32 val;
901 	int i;
902 
903 	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
904 			   I40E_GLPRT_GORCL(hw->port),
905 			   pf->stat_offsets_loaded,
906 			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
907 	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
908 			   I40E_GLPRT_GOTCL(hw->port),
909 			   pf->stat_offsets_loaded,
910 			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
911 	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
912 			   pf->stat_offsets_loaded,
913 			   &osd->eth.rx_discards,
914 			   &nsd->eth.rx_discards);
915 	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
916 			   I40E_GLPRT_UPRCL(hw->port),
917 			   pf->stat_offsets_loaded,
918 			   &osd->eth.rx_unicast,
919 			   &nsd->eth.rx_unicast);
920 	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
921 			   I40E_GLPRT_MPRCL(hw->port),
922 			   pf->stat_offsets_loaded,
923 			   &osd->eth.rx_multicast,
924 			   &nsd->eth.rx_multicast);
925 	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
926 			   I40E_GLPRT_BPRCL(hw->port),
927 			   pf->stat_offsets_loaded,
928 			   &osd->eth.rx_broadcast,
929 			   &nsd->eth.rx_broadcast);
930 	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
931 			   I40E_GLPRT_UPTCL(hw->port),
932 			   pf->stat_offsets_loaded,
933 			   &osd->eth.tx_unicast,
934 			   &nsd->eth.tx_unicast);
935 	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
936 			   I40E_GLPRT_MPTCL(hw->port),
937 			   pf->stat_offsets_loaded,
938 			   &osd->eth.tx_multicast,
939 			   &nsd->eth.tx_multicast);
940 	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
941 			   I40E_GLPRT_BPTCL(hw->port),
942 			   pf->stat_offsets_loaded,
943 			   &osd->eth.tx_broadcast,
944 			   &nsd->eth.tx_broadcast);
945 
946 	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
947 			   pf->stat_offsets_loaded,
948 			   &osd->tx_dropped_link_down,
949 			   &nsd->tx_dropped_link_down);
950 
951 	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
952 			   pf->stat_offsets_loaded,
953 			   &osd->crc_errors, &nsd->crc_errors);
954 
955 	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
956 			   pf->stat_offsets_loaded,
957 			   &osd->illegal_bytes, &nsd->illegal_bytes);
958 
959 	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
960 			   pf->stat_offsets_loaded,
961 			   &osd->mac_local_faults,
962 			   &nsd->mac_local_faults);
963 	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
964 			   pf->stat_offsets_loaded,
965 			   &osd->mac_remote_faults,
966 			   &nsd->mac_remote_faults);
967 
968 	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
969 			   pf->stat_offsets_loaded,
970 			   &osd->rx_length_errors,
971 			   &nsd->rx_length_errors);
972 
973 	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
974 			   pf->stat_offsets_loaded,
975 			   &osd->link_xon_rx, &nsd->link_xon_rx);
976 	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
977 			   pf->stat_offsets_loaded,
978 			   &osd->link_xon_tx, &nsd->link_xon_tx);
979 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
980 			   pf->stat_offsets_loaded,
981 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
982 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
983 			   pf->stat_offsets_loaded,
984 			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
985 
986 	for (i = 0; i < 8; i++) {
987 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
988 				   pf->stat_offsets_loaded,
989 				   &osd->priority_xoff_rx[i],
990 				   &nsd->priority_xoff_rx[i]);
991 		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
992 				   pf->stat_offsets_loaded,
993 				   &osd->priority_xon_rx[i],
994 				   &nsd->priority_xon_rx[i]);
995 		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
996 				   pf->stat_offsets_loaded,
997 				   &osd->priority_xon_tx[i],
998 				   &nsd->priority_xon_tx[i]);
999 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1000 				   pf->stat_offsets_loaded,
1001 				   &osd->priority_xoff_tx[i],
1002 				   &nsd->priority_xoff_tx[i]);
1003 		i40e_stat_update32(hw,
1004 				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1005 				   pf->stat_offsets_loaded,
1006 				   &osd->priority_xon_2_xoff[i],
1007 				   &nsd->priority_xon_2_xoff[i]);
1008 	}
1009 
1010 	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1011 			   I40E_GLPRT_PRC64L(hw->port),
1012 			   pf->stat_offsets_loaded,
1013 			   &osd->rx_size_64, &nsd->rx_size_64);
1014 	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1015 			   I40E_GLPRT_PRC127L(hw->port),
1016 			   pf->stat_offsets_loaded,
1017 			   &osd->rx_size_127, &nsd->rx_size_127);
1018 	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1019 			   I40E_GLPRT_PRC255L(hw->port),
1020 			   pf->stat_offsets_loaded,
1021 			   &osd->rx_size_255, &nsd->rx_size_255);
1022 	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1023 			   I40E_GLPRT_PRC511L(hw->port),
1024 			   pf->stat_offsets_loaded,
1025 			   &osd->rx_size_511, &nsd->rx_size_511);
1026 	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1027 			   I40E_GLPRT_PRC1023L(hw->port),
1028 			   pf->stat_offsets_loaded,
1029 			   &osd->rx_size_1023, &nsd->rx_size_1023);
1030 	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1031 			   I40E_GLPRT_PRC1522L(hw->port),
1032 			   pf->stat_offsets_loaded,
1033 			   &osd->rx_size_1522, &nsd->rx_size_1522);
1034 	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1035 			   I40E_GLPRT_PRC9522L(hw->port),
1036 			   pf->stat_offsets_loaded,
1037 			   &osd->rx_size_big, &nsd->rx_size_big);
1038 
1039 	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1040 			   I40E_GLPRT_PTC64L(hw->port),
1041 			   pf->stat_offsets_loaded,
1042 			   &osd->tx_size_64, &nsd->tx_size_64);
1043 	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1044 			   I40E_GLPRT_PTC127L(hw->port),
1045 			   pf->stat_offsets_loaded,
1046 			   &osd->tx_size_127, &nsd->tx_size_127);
1047 	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1048 			   I40E_GLPRT_PTC255L(hw->port),
1049 			   pf->stat_offsets_loaded,
1050 			   &osd->tx_size_255, &nsd->tx_size_255);
1051 	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1052 			   I40E_GLPRT_PTC511L(hw->port),
1053 			   pf->stat_offsets_loaded,
1054 			   &osd->tx_size_511, &nsd->tx_size_511);
1055 	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1056 			   I40E_GLPRT_PTC1023L(hw->port),
1057 			   pf->stat_offsets_loaded,
1058 			   &osd->tx_size_1023, &nsd->tx_size_1023);
1059 	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1060 			   I40E_GLPRT_PTC1522L(hw->port),
1061 			   pf->stat_offsets_loaded,
1062 			   &osd->tx_size_1522, &nsd->tx_size_1522);
1063 	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1064 			   I40E_GLPRT_PTC9522L(hw->port),
1065 			   pf->stat_offsets_loaded,
1066 			   &osd->tx_size_big, &nsd->tx_size_big);
1067 
1068 	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1069 			   pf->stat_offsets_loaded,
1070 			   &osd->rx_undersize, &nsd->rx_undersize);
1071 	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1072 			   pf->stat_offsets_loaded,
1073 			   &osd->rx_fragments, &nsd->rx_fragments);
1074 	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1075 			   pf->stat_offsets_loaded,
1076 			   &osd->rx_oversize, &nsd->rx_oversize);
1077 	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1078 			   pf->stat_offsets_loaded,
1079 			   &osd->rx_jabber, &nsd->rx_jabber);
1080 
1081 	/* FDIR stats */
1082 	i40e_stat_update32(hw,
1083 			   I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1084 			   pf->stat_offsets_loaded,
1085 			   &osd->fd_atr_match, &nsd->fd_atr_match);
1086 	i40e_stat_update32(hw,
1087 			   I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1088 			   pf->stat_offsets_loaded,
1089 			   &osd->fd_sb_match, &nsd->fd_sb_match);
1090 	i40e_stat_update32(hw,
1091 		      I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1092 		      pf->stat_offsets_loaded,
1093 		      &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1094 
1095 	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1096 	nsd->tx_lpi_status =
1097 		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1098 			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1099 	nsd->rx_lpi_status =
1100 		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1101 			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1102 	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1103 			   pf->stat_offsets_loaded,
1104 			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1105 	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1106 			   pf->stat_offsets_loaded,
1107 			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1108 
1109 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1110 	    !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1111 		nsd->fd_sb_status = true;
1112 	else
1113 		nsd->fd_sb_status = false;
1114 
1115 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1116 	    !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1117 		nsd->fd_atr_status = true;
1118 	else
1119 		nsd->fd_atr_status = false;
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  *
1149  * Returns ptr to the filter object or NULL
1150  **/
1151 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1152 						const u8 *macaddr, s16 vlan)
1153 {
1154 	struct i40e_mac_filter *f;
1155 	u64 key;
1156 
1157 	if (!vsi || !macaddr)
1158 		return NULL;
1159 
1160 	key = i40e_addr_to_hkey(macaddr);
1161 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1162 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1163 		    (vlan == f->vlan))
1164 			return f;
1165 	}
1166 	return NULL;
1167 }
1168 
1169 /**
1170  * i40e_find_mac - Find a mac addr in the macvlan filters list
1171  * @vsi: the VSI to be searched
1172  * @macaddr: the MAC address we are searching for
1173  *
1174  * Returns the first filter with the provided MAC address or NULL if
1175  * MAC address was not found
1176  **/
1177 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1178 {
1179 	struct i40e_mac_filter *f;
1180 	u64 key;
1181 
1182 	if (!vsi || !macaddr)
1183 		return NULL;
1184 
1185 	key = i40e_addr_to_hkey(macaddr);
1186 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1187 		if ((ether_addr_equal(macaddr, f->macaddr)))
1188 			return f;
1189 	}
1190 	return NULL;
1191 }
1192 
1193 /**
1194  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1195  * @vsi: the VSI to be searched
1196  *
1197  * Returns true if VSI is in vlan mode or false otherwise
1198  **/
1199 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1200 {
1201 	/* If we have a PVID, always operate in VLAN mode */
1202 	if (vsi->info.pvid)
1203 		return true;
1204 
1205 	/* We need to operate in VLAN mode whenever we have any filters with
1206 	 * a VLAN other than I40E_VLAN_ALL. We could check the table each
1207 	 * time, incurring search cost repeatedly. However, we can notice two
1208 	 * things:
1209 	 *
1210 	 * 1) the only place where we can gain a VLAN filter is in
1211 	 *    i40e_add_filter.
1212 	 *
1213 	 * 2) the only place where filters are actually removed is in
1214 	 *    i40e_vsi_sync_filters_subtask.
1215 	 *
1216 	 * Thus, we can simply use a boolean value, has_vlan_filters which we
1217 	 * will set to true when we add a VLAN filter in i40e_add_filter. Then
1218 	 * we have to perform the full search after deleting filters in
1219 	 * i40e_vsi_sync_filters_subtask, but we already have to search
1220 	 * filters here and can perform the check at the same time. This
1221 	 * results in avoiding embedding a loop for VLAN mode inside another
1222 	 * loop over all the filters, and should maintain correctness as noted
1223 	 * above.
1224 	 */
1225 	return vsi->has_vlan_filter;
1226 }
1227 
1228 /**
1229  * i40e_add_filter - Add a mac/vlan filter to the VSI
1230  * @vsi: the VSI to be searched
1231  * @macaddr: the MAC address
1232  * @vlan: the vlan
1233  *
1234  * Returns ptr to the filter object or NULL when no memory available.
1235  *
1236  * NOTE: This function is expected to be called with mac_filter_hash_lock
1237  * being held.
1238  **/
1239 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1240 					const u8 *macaddr, s16 vlan)
1241 {
1242 	struct i40e_mac_filter *f;
1243 	u64 key;
1244 
1245 	if (!vsi || !macaddr)
1246 		return NULL;
1247 
1248 	/* Do not allow broadcast filter to be added since broadcast filter
1249 	 * is added as part of add VSI for any newly created VSI except
1250 	 * FDIR VSI
1251 	 */
1252 	if (is_broadcast_ether_addr(macaddr))
1253 		return NULL;
1254 
1255 	f = i40e_find_filter(vsi, macaddr, vlan);
1256 	if (!f) {
1257 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1258 		if (!f)
1259 			return NULL;
1260 
1261 		/* Update the boolean indicating if we need to function in
1262 		 * VLAN mode.
1263 		 */
1264 		if (vlan >= 0)
1265 			vsi->has_vlan_filter = true;
1266 
1267 		ether_addr_copy(f->macaddr, macaddr);
1268 		f->vlan = vlan;
1269 		/* If we're in overflow promisc mode, set the state directly
1270 		 * to failed, so we don't bother to try sending the filter
1271 		 * to the hardware.
1272 		 */
1273 		if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))
1274 			f->state = I40E_FILTER_FAILED;
1275 		else
1276 			f->state = I40E_FILTER_NEW;
1277 		INIT_HLIST_NODE(&f->hlist);
1278 
1279 		key = i40e_addr_to_hkey(macaddr);
1280 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1281 
1282 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1283 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1284 	}
1285 
1286 	/* If we're asked to add a filter that has been marked for removal, it
1287 	 * is safe to simply restore it to active state. __i40e_del_filter
1288 	 * will have simply deleted any filters which were previously marked
1289 	 * NEW or FAILED, so if it is currently marked REMOVE it must have
1290 	 * previously been ACTIVE. Since we haven't yet run the sync filters
1291 	 * task, just restore this filter to the ACTIVE state so that the
1292 	 * sync task leaves it in place
1293 	 */
1294 	if (f->state == I40E_FILTER_REMOVE)
1295 		f->state = I40E_FILTER_ACTIVE;
1296 
1297 	return f;
1298 }
1299 
1300 /**
1301  * __i40e_del_filter - Remove a specific filter from the VSI
1302  * @vsi: VSI to remove from
1303  * @f: the filter to remove from the list
1304  *
1305  * This function should be called instead of i40e_del_filter only if you know
1306  * the exact filter you will remove already, such as via i40e_find_filter or
1307  * i40e_find_mac.
1308  *
1309  * NOTE: This function is expected to be called with mac_filter_hash_lock
1310  * being held.
1311  * ANOTHER NOTE: This function MUST be called from within the context of
1312  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1313  * instead of list_for_each_entry().
1314  **/
1315 static void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1316 {
1317 	if (!f)
1318 		return;
1319 
1320 	if ((f->state == I40E_FILTER_FAILED) ||
1321 	    (f->state == I40E_FILTER_NEW)) {
1322 		/* this one never got added by the FW. Just remove it,
1323 		 * no need to sync anything.
1324 		 */
1325 		hash_del(&f->hlist);
1326 		kfree(f);
1327 	} else {
1328 		f->state = I40E_FILTER_REMOVE;
1329 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1330 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1331 	}
1332 }
1333 
1334 /**
1335  * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1336  * @vsi: the VSI to be searched
1337  * @macaddr: the MAC address
1338  * @vlan: the VLAN
1339  *
1340  * NOTE: This function is expected to be called with mac_filter_hash_lock
1341  * being held.
1342  * ANOTHER NOTE: This function MUST be called from within the context of
1343  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1344  * instead of list_for_each_entry().
1345  **/
1346 void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1347 {
1348 	struct i40e_mac_filter *f;
1349 
1350 	if (!vsi || !macaddr)
1351 		return;
1352 
1353 	f = i40e_find_filter(vsi, macaddr, vlan);
1354 	__i40e_del_filter(vsi, f);
1355 }
1356 
1357 /**
1358  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1359  * @vsi: the VSI to be searched
1360  * @macaddr: the mac address to be filtered
1361  *
1362  * Goes through all the macvlan filters and adds a macvlan filter for each
1363  * unique vlan that already exists. If a PVID has been assigned, instead only
1364  * add the macaddr to that VLAN.
1365  *
1366  * Returns last filter added on success, else NULL
1367  **/
1368 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi,
1369 					     const u8 *macaddr)
1370 {
1371 	struct i40e_mac_filter *f, *add = NULL;
1372 	struct hlist_node *h;
1373 	int bkt;
1374 
1375 	if (vsi->info.pvid)
1376 		return i40e_add_filter(vsi, macaddr,
1377 				       le16_to_cpu(vsi->info.pvid));
1378 
1379 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1380 		if (f->state == I40E_FILTER_REMOVE)
1381 			continue;
1382 		add = i40e_add_filter(vsi, macaddr, f->vlan);
1383 		if (!add)
1384 			return NULL;
1385 	}
1386 
1387 	return add;
1388 }
1389 
1390 /**
1391  * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1392  * @vsi: the VSI to be searched
1393  * @macaddr: the mac address to be removed
1394  *
1395  * Removes a given MAC address from a VSI, regardless of VLAN
1396  *
1397  * Returns 0 for success, or error
1398  **/
1399 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, const u8 *macaddr)
1400 {
1401 	struct i40e_mac_filter *f;
1402 	struct hlist_node *h;
1403 	bool found = false;
1404 	int bkt;
1405 
1406 	WARN(!spin_is_locked(&vsi->mac_filter_hash_lock),
1407 	     "Missing mac_filter_hash_lock\n");
1408 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1409 		if (ether_addr_equal(macaddr, f->macaddr)) {
1410 			__i40e_del_filter(vsi, f);
1411 			found = true;
1412 		}
1413 	}
1414 
1415 	if (found)
1416 		return 0;
1417 	else
1418 		return -ENOENT;
1419 }
1420 
1421 /**
1422  * i40e_set_mac - NDO callback to set mac address
1423  * @netdev: network interface device structure
1424  * @p: pointer to an address structure
1425  *
1426  * Returns 0 on success, negative on failure
1427  **/
1428 #ifdef I40E_FCOE
1429 int i40e_set_mac(struct net_device *netdev, void *p)
1430 #else
1431 static int i40e_set_mac(struct net_device *netdev, void *p)
1432 #endif
1433 {
1434 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1435 	struct i40e_vsi *vsi = np->vsi;
1436 	struct i40e_pf *pf = vsi->back;
1437 	struct i40e_hw *hw = &pf->hw;
1438 	struct sockaddr *addr = p;
1439 
1440 	if (!is_valid_ether_addr(addr->sa_data))
1441 		return -EADDRNOTAVAIL;
1442 
1443 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1444 		netdev_info(netdev, "already using mac address %pM\n",
1445 			    addr->sa_data);
1446 		return 0;
1447 	}
1448 
1449 	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1450 	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1451 		return -EADDRNOTAVAIL;
1452 
1453 	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1454 		netdev_info(netdev, "returning to hw mac address %pM\n",
1455 			    hw->mac.addr);
1456 	else
1457 		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1458 
1459 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1460 	i40e_del_mac_all_vlan(vsi, netdev->dev_addr);
1461 	i40e_put_mac_in_vlan(vsi, addr->sa_data);
1462 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1463 	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1464 	if (vsi->type == I40E_VSI_MAIN) {
1465 		i40e_status ret;
1466 
1467 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1468 						I40E_AQC_WRITE_TYPE_LAA_WOL,
1469 						addr->sa_data, NULL);
1470 		if (ret)
1471 			netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1472 				    i40e_stat_str(hw, ret),
1473 				    i40e_aq_str(hw, hw->aq.asq_last_status));
1474 	}
1475 
1476 	/* schedule our worker thread which will take care of
1477 	 * applying the new filter changes
1478 	 */
1479 	i40e_service_event_schedule(vsi->back);
1480 	return 0;
1481 }
1482 
1483 /**
1484  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1485  * @vsi: the VSI being setup
1486  * @ctxt: VSI context structure
1487  * @enabled_tc: Enabled TCs bitmap
1488  * @is_add: True if called before Add VSI
1489  *
1490  * Setup VSI queue mapping for enabled traffic classes.
1491  **/
1492 #ifdef I40E_FCOE
1493 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1494 			      struct i40e_vsi_context *ctxt,
1495 			      u8 enabled_tc,
1496 			      bool is_add)
1497 #else
1498 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1499 				     struct i40e_vsi_context *ctxt,
1500 				     u8 enabled_tc,
1501 				     bool is_add)
1502 #endif
1503 {
1504 	struct i40e_pf *pf = vsi->back;
1505 	u16 sections = 0;
1506 	u8 netdev_tc = 0;
1507 	u16 numtc = 0;
1508 	u16 qcount;
1509 	u8 offset;
1510 	u16 qmap;
1511 	int i;
1512 	u16 num_tc_qps = 0;
1513 
1514 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1515 	offset = 0;
1516 
1517 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1518 		/* Find numtc from enabled TC bitmap */
1519 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1520 			if (enabled_tc & BIT(i)) /* TC is enabled */
1521 				numtc++;
1522 		}
1523 		if (!numtc) {
1524 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1525 			numtc = 1;
1526 		}
1527 	} else {
1528 		/* At least TC0 is enabled in case of non-DCB case */
1529 		numtc = 1;
1530 	}
1531 
1532 	vsi->tc_config.numtc = numtc;
1533 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1534 	/* Number of queues per enabled TC */
1535 	qcount = vsi->alloc_queue_pairs;
1536 
1537 	num_tc_qps = qcount / numtc;
1538 	num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1539 
1540 	/* Setup queue offset/count for all TCs for given VSI */
1541 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1542 		/* See if the given TC is enabled for the given VSI */
1543 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1544 			/* TC is enabled */
1545 			int pow, num_qps;
1546 
1547 			switch (vsi->type) {
1548 			case I40E_VSI_MAIN:
1549 				qcount = min_t(int, pf->alloc_rss_size,
1550 					       num_tc_qps);
1551 				break;
1552 #ifdef I40E_FCOE
1553 			case I40E_VSI_FCOE:
1554 				qcount = num_tc_qps;
1555 				break;
1556 #endif
1557 			case I40E_VSI_FDIR:
1558 			case I40E_VSI_SRIOV:
1559 			case I40E_VSI_VMDQ2:
1560 			default:
1561 				qcount = num_tc_qps;
1562 				WARN_ON(i != 0);
1563 				break;
1564 			}
1565 			vsi->tc_config.tc_info[i].qoffset = offset;
1566 			vsi->tc_config.tc_info[i].qcount = qcount;
1567 
1568 			/* find the next higher power-of-2 of num queue pairs */
1569 			num_qps = qcount;
1570 			pow = 0;
1571 			while (num_qps && (BIT_ULL(pow) < qcount)) {
1572 				pow++;
1573 				num_qps >>= 1;
1574 			}
1575 
1576 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1577 			qmap =
1578 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1579 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1580 
1581 			offset += qcount;
1582 		} else {
1583 			/* TC is not enabled so set the offset to
1584 			 * default queue and allocate one queue
1585 			 * for the given TC.
1586 			 */
1587 			vsi->tc_config.tc_info[i].qoffset = 0;
1588 			vsi->tc_config.tc_info[i].qcount = 1;
1589 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1590 
1591 			qmap = 0;
1592 		}
1593 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1594 	}
1595 
1596 	/* Set actual Tx/Rx queue pairs */
1597 	vsi->num_queue_pairs = offset;
1598 	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1599 		if (vsi->req_queue_pairs > 0)
1600 			vsi->num_queue_pairs = vsi->req_queue_pairs;
1601 		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1602 			vsi->num_queue_pairs = pf->num_lan_msix;
1603 	}
1604 
1605 	/* Scheduler section valid can only be set for ADD VSI */
1606 	if (is_add) {
1607 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1608 
1609 		ctxt->info.up_enable_bits = enabled_tc;
1610 	}
1611 	if (vsi->type == I40E_VSI_SRIOV) {
1612 		ctxt->info.mapping_flags |=
1613 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1614 		for (i = 0; i < vsi->num_queue_pairs; i++)
1615 			ctxt->info.queue_mapping[i] =
1616 					       cpu_to_le16(vsi->base_queue + i);
1617 	} else {
1618 		ctxt->info.mapping_flags |=
1619 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1620 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1621 	}
1622 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1623 }
1624 
1625 /**
1626  * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1627  * @netdev: the netdevice
1628  * @addr: address to add
1629  *
1630  * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1631  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1632  */
1633 static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1634 {
1635 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1636 	struct i40e_vsi *vsi = np->vsi;
1637 	struct i40e_mac_filter *f;
1638 
1639 	if (i40e_is_vsi_in_vlan(vsi))
1640 		f = i40e_put_mac_in_vlan(vsi, addr);
1641 	else
1642 		f = i40e_add_filter(vsi, addr, I40E_VLAN_ANY);
1643 
1644 	if (f)
1645 		return 0;
1646 	else
1647 		return -ENOMEM;
1648 }
1649 
1650 /**
1651  * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1652  * @netdev: the netdevice
1653  * @addr: address to add
1654  *
1655  * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1656  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1657  */
1658 static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1659 {
1660 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1661 	struct i40e_vsi *vsi = np->vsi;
1662 
1663 	if (i40e_is_vsi_in_vlan(vsi))
1664 		i40e_del_mac_all_vlan(vsi, addr);
1665 	else
1666 		i40e_del_filter(vsi, addr, I40E_VLAN_ANY);
1667 
1668 	return 0;
1669 }
1670 
1671 /**
1672  * i40e_set_rx_mode - NDO callback to set the netdev filters
1673  * @netdev: network interface device structure
1674  **/
1675 #ifdef I40E_FCOE
1676 void i40e_set_rx_mode(struct net_device *netdev)
1677 #else
1678 static void i40e_set_rx_mode(struct net_device *netdev)
1679 #endif
1680 {
1681 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1682 	struct i40e_vsi *vsi = np->vsi;
1683 
1684 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1685 
1686 	__dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1687 	__dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1688 
1689 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1690 
1691 	/* check for other flag changes */
1692 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1693 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1694 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1695 	}
1696 
1697 	/* schedule our worker thread which will take care of
1698 	 * applying the new filter changes
1699 	 */
1700 	i40e_service_event_schedule(vsi->back);
1701 }
1702 
1703 /**
1704  * i40e_undo_filter_entries - Undo the changes made to MAC filter entries
1705  * @vsi: Pointer to VSI struct
1706  * @from: Pointer to list which contains MAC filter entries - changes to
1707  *        those entries needs to be undone.
1708  *
1709  * MAC filter entries from list were slated to be sent to firmware, either for
1710  * addition or deletion.
1711  **/
1712 static void i40e_undo_filter_entries(struct i40e_vsi *vsi,
1713 				     struct hlist_head *from)
1714 {
1715 	struct i40e_mac_filter *f;
1716 	struct hlist_node *h;
1717 
1718 	hlist_for_each_entry_safe(f, h, from, hlist) {
1719 		u64 key = i40e_addr_to_hkey(f->macaddr);
1720 
1721 		/* Move the element back into MAC filter list*/
1722 		hlist_del(&f->hlist);
1723 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1724 	}
1725 }
1726 
1727 /**
1728  * i40e_update_filter_state - Update filter state based on return data
1729  * from firmware
1730  * @count: Number of filters added
1731  * @add_list: return data from fw
1732  * @head: pointer to first filter in current batch
1733  * @aq_err: status from fw
1734  *
1735  * MAC filter entries from list were slated to be added to device. Returns
1736  * number of successful filters. Note that 0 does NOT mean success!
1737  **/
1738 static int
1739 i40e_update_filter_state(int count,
1740 			 struct i40e_aqc_add_macvlan_element_data *add_list,
1741 			 struct i40e_mac_filter *add_head, int aq_err)
1742 {
1743 	int retval = 0;
1744 	int i;
1745 
1746 
1747 	if (!aq_err) {
1748 		retval = count;
1749 		/* Everything's good, mark all filters active. */
1750 		for (i = 0; i < count ; i++) {
1751 			add_head->state = I40E_FILTER_ACTIVE;
1752 			add_head = hlist_entry(add_head->hlist.next,
1753 					       typeof(struct i40e_mac_filter),
1754 					       hlist);
1755 		}
1756 	} else if (aq_err == I40E_AQ_RC_ENOSPC) {
1757 		/* Device ran out of filter space. Check the return value
1758 		 * for each filter to see which ones are active.
1759 		 */
1760 		for (i = 0; i < count ; i++) {
1761 			if (add_list[i].match_method ==
1762 			    I40E_AQC_MM_ERR_NO_RES) {
1763 				add_head->state = I40E_FILTER_FAILED;
1764 			} else {
1765 				add_head->state = I40E_FILTER_ACTIVE;
1766 				retval++;
1767 			}
1768 			add_head = hlist_entry(add_head->hlist.next,
1769 					       typeof(struct i40e_mac_filter),
1770 					       hlist);
1771 		}
1772 	} else {
1773 		/* Some other horrible thing happened, fail all filters */
1774 		retval = 0;
1775 		for (i = 0; i < count ; i++) {
1776 			add_head->state = I40E_FILTER_FAILED;
1777 			add_head = hlist_entry(add_head->hlist.next,
1778 					       typeof(struct i40e_mac_filter),
1779 					       hlist);
1780 		}
1781 	}
1782 	return retval;
1783 }
1784 
1785 /**
1786  * i40e_aqc_del_filters - Request firmware to delete a set of filters
1787  * @vsi: ptr to the VSI
1788  * @vsi_name: name to display in messages
1789  * @list: the list of filters to send to firmware
1790  * @num_del: the number of filters to delete
1791  * @retval: Set to -EIO on failure to delete
1792  *
1793  * Send a request to firmware via AdminQ to delete a set of filters. Uses
1794  * *retval instead of a return value so that success does not force ret_val to
1795  * be set to 0. This ensures that a sequence of calls to this function
1796  * preserve the previous value of *retval on successful delete.
1797  */
1798 static
1799 void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
1800 			  struct i40e_aqc_remove_macvlan_element_data *list,
1801 			  int num_del, int *retval)
1802 {
1803 	struct i40e_hw *hw = &vsi->back->hw;
1804 	i40e_status aq_ret;
1805 	int aq_err;
1806 
1807 	aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
1808 	aq_err = hw->aq.asq_last_status;
1809 
1810 	/* Explicitly ignore and do not report when firmware returns ENOENT */
1811 	if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1812 		*retval = -EIO;
1813 		dev_info(&vsi->back->pdev->dev,
1814 			 "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1815 			 vsi_name, i40e_stat_str(hw, aq_ret),
1816 			 i40e_aq_str(hw, aq_err));
1817 	}
1818 }
1819 
1820 /**
1821  * i40e_aqc_add_filters - Request firmware to add a set of filters
1822  * @vsi: ptr to the VSI
1823  * @vsi_name: name to display in messages
1824  * @list: the list of filters to send to firmware
1825  * @add_head: Position in the add hlist
1826  * @num_add: the number of filters to add
1827  * @promisc_change: set to true on exit if promiscuous mode was forced on
1828  *
1829  * Send a request to firmware via AdminQ to add a chunk of filters. Will set
1830  * promisc_changed to true if the firmware has run out of space for more
1831  * filters.
1832  */
1833 static
1834 void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
1835 			  struct i40e_aqc_add_macvlan_element_data *list,
1836 			  struct i40e_mac_filter *add_head,
1837 			  int num_add, bool *promisc_changed)
1838 {
1839 	struct i40e_hw *hw = &vsi->back->hw;
1840 	i40e_status aq_ret;
1841 	int aq_err, fcnt;
1842 
1843 	aq_ret = i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
1844 	aq_err = hw->aq.asq_last_status;
1845 	fcnt = i40e_update_filter_state(num_add, list, add_head, aq_ret);
1846 	vsi->active_filters += fcnt;
1847 
1848 	if (fcnt != num_add) {
1849 		*promisc_changed = true;
1850 		set_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
1851 		vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
1852 		dev_warn(&vsi->back->pdev->dev,
1853 			 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
1854 			 i40e_aq_str(hw, aq_err),
1855 			 vsi_name);
1856 	}
1857 }
1858 
1859 /**
1860  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1861  * @vsi: ptr to the VSI
1862  *
1863  * Push any outstanding VSI filter changes through the AdminQ.
1864  *
1865  * Returns 0 or error value
1866  **/
1867 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1868 {
1869 	struct hlist_head tmp_add_list, tmp_del_list;
1870 	struct i40e_mac_filter *f, *add_head = NULL;
1871 	struct i40e_hw *hw = &vsi->back->hw;
1872 	unsigned int vlan_any_filters = 0;
1873 	unsigned int non_vlan_filters = 0;
1874 	unsigned int vlan_filters = 0;
1875 	bool promisc_changed = false;
1876 	char vsi_name[16] = "PF";
1877 	int filter_list_len = 0;
1878 	i40e_status aq_ret = 0;
1879 	u32 changed_flags = 0;
1880 	struct hlist_node *h;
1881 	struct i40e_pf *pf;
1882 	int num_add = 0;
1883 	int num_del = 0;
1884 	int retval = 0;
1885 	u16 cmd_flags;
1886 	int list_size;
1887 	int bkt;
1888 
1889 	/* empty array typed pointers, kcalloc later */
1890 	struct i40e_aqc_add_macvlan_element_data *add_list;
1891 	struct i40e_aqc_remove_macvlan_element_data *del_list;
1892 
1893 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1894 		usleep_range(1000, 2000);
1895 	pf = vsi->back;
1896 
1897 	if (vsi->netdev) {
1898 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1899 		vsi->current_netdev_flags = vsi->netdev->flags;
1900 	}
1901 
1902 	INIT_HLIST_HEAD(&tmp_add_list);
1903 	INIT_HLIST_HEAD(&tmp_del_list);
1904 
1905 	if (vsi->type == I40E_VSI_SRIOV)
1906 		snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
1907 	else if (vsi->type != I40E_VSI_MAIN)
1908 		snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
1909 
1910 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1911 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1912 
1913 		spin_lock_bh(&vsi->mac_filter_hash_lock);
1914 		/* Create a list of filters to delete. */
1915 		hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1916 			if (f->state == I40E_FILTER_REMOVE) {
1917 				/* Move the element into temporary del_list */
1918 				hash_del(&f->hlist);
1919 				hlist_add_head(&f->hlist, &tmp_del_list);
1920 				vsi->active_filters--;
1921 
1922 				/* Avoid counting removed filters */
1923 				continue;
1924 			}
1925 			if (f->state == I40E_FILTER_NEW) {
1926 				hash_del(&f->hlist);
1927 				hlist_add_head(&f->hlist, &tmp_add_list);
1928 			}
1929 
1930 			/* Count the number of each type of filter we have
1931 			 * remaining, ignoring any filters we're about to
1932 			 * delete.
1933 			 */
1934 			if (f->vlan > 0)
1935 				vlan_filters++;
1936 			else if (!f->vlan)
1937 				non_vlan_filters++;
1938 			else
1939 				vlan_any_filters++;
1940 		}
1941 
1942 		/* We should never have VLAN=-1 filters at the same time as we
1943 		 * have either VLAN=0 or VLAN>0 filters, so warn about this
1944 		 * case here to help catch any issues.
1945 		 */
1946 		WARN_ON(vlan_any_filters && (vlan_filters + non_vlan_filters));
1947 
1948 		/* If we only have VLAN=0 filters remaining, and don't have
1949 		 * any other VLAN filters, we need to convert these VLAN=0
1950 		 * filters into VLAN=-1 (I40E_VLAN_ANY) so that we operate
1951 		 * correctly in non-VLAN mode and receive all traffic tagged
1952 		 * or untagged.
1953 		 */
1954 		if (non_vlan_filters && !vlan_filters) {
1955 			hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f,
1956 					   hlist) {
1957 				/* Only replace VLAN=0 filters */
1958 				if (f->vlan)
1959 					continue;
1960 
1961 				/* Allocate a replacement element */
1962 				add_head = kzalloc(sizeof(*add_head),
1963 						   GFP_KERNEL);
1964 				if (!add_head)
1965 					goto err_no_memory_locked;
1966 
1967 				/* Copy the filter, with new state and VLAN */
1968 				*add_head = *f;
1969 				add_head->state = I40E_FILTER_NEW;
1970 				add_head->vlan = I40E_VLAN_ANY;
1971 
1972 				/* Move the replacement to the add list */
1973 				INIT_HLIST_NODE(&add_head->hlist);
1974 				hlist_add_head(&add_head->hlist,
1975 					       &tmp_add_list);
1976 
1977 				/* Move the original to the delete list */
1978 				f->state = I40E_FILTER_REMOVE;
1979 				hash_del(&f->hlist);
1980 				hlist_add_head(&f->hlist, &tmp_del_list);
1981 				vsi->active_filters--;
1982 			}
1983 
1984 			/* Also update any filters on the tmp_add list */
1985 			hlist_for_each_entry(f, &tmp_add_list, hlist) {
1986 				if (!f->vlan)
1987 					f->vlan = I40E_VLAN_ANY;
1988 			}
1989 			add_head = NULL;
1990 		}
1991 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
1992 	}
1993 
1994 	/* Now process 'del_list' outside the lock */
1995 	if (!hlist_empty(&tmp_del_list)) {
1996 		filter_list_len = hw->aq.asq_buf_size /
1997 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
1998 		list_size = filter_list_len *
1999 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2000 		del_list = kzalloc(list_size, GFP_ATOMIC);
2001 		if (!del_list)
2002 			goto err_no_memory;
2003 
2004 		hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2005 			cmd_flags = 0;
2006 
2007 			/* add to delete list */
2008 			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2009 			if (f->vlan == I40E_VLAN_ANY) {
2010 				del_list[num_del].vlan_tag = 0;
2011 				cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2012 			} else {
2013 				del_list[num_del].vlan_tag =
2014 					cpu_to_le16((u16)(f->vlan));
2015 			}
2016 
2017 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2018 			del_list[num_del].flags = cmd_flags;
2019 			num_del++;
2020 
2021 			/* flush a full buffer */
2022 			if (num_del == filter_list_len) {
2023 				i40e_aqc_del_filters(vsi, vsi_name, del_list,
2024 						     num_del, &retval);
2025 				memset(del_list, 0, list_size);
2026 				num_del = 0;
2027 			}
2028 			/* Release memory for MAC filter entries which were
2029 			 * synced up with HW.
2030 			 */
2031 			hlist_del(&f->hlist);
2032 			kfree(f);
2033 		}
2034 
2035 		if (num_del) {
2036 			i40e_aqc_del_filters(vsi, vsi_name, del_list,
2037 					     num_del, &retval);
2038 		}
2039 
2040 		kfree(del_list);
2041 		del_list = NULL;
2042 	}
2043 
2044 	/* After finishing notifying firmware of the deleted filters, update
2045 	 * the cached value of vsi->has_vlan_filter. Note that we are safe to
2046 	 * use just !!vlan_filters here because if we only have VLAN=0 (that
2047 	 * is, non_vlan_filters) these will all be converted to VLAN=-1 in the
2048 	 * logic above already so this value would still be correct.
2049 	 */
2050 	vsi->has_vlan_filter = !!vlan_filters;
2051 
2052 	if (!hlist_empty(&tmp_add_list)) {
2053 		/* Do all the adds now. */
2054 		filter_list_len = hw->aq.asq_buf_size /
2055 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2056 		list_size = filter_list_len *
2057 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2058 		add_list = kzalloc(list_size, GFP_ATOMIC);
2059 		if (!add_list)
2060 			goto err_no_memory;
2061 
2062 		num_add = 0;
2063 		hlist_for_each_entry(f, &tmp_add_list, hlist) {
2064 			if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2065 				     &vsi->state)) {
2066 				f->state = I40E_FILTER_FAILED;
2067 				continue;
2068 			}
2069 			/* add to add array */
2070 			if (num_add == 0)
2071 				add_head = f;
2072 			cmd_flags = 0;
2073 			ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
2074 			if (f->vlan == I40E_VLAN_ANY) {
2075 				add_list[num_add].vlan_tag = 0;
2076 				cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2077 			} else {
2078 				add_list[num_add].vlan_tag =
2079 					cpu_to_le16((u16)(f->vlan));
2080 			}
2081 			add_list[num_add].queue_number = 0;
2082 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2083 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
2084 			num_add++;
2085 
2086 			/* flush a full buffer */
2087 			if (num_add == filter_list_len) {
2088 				i40e_aqc_add_filters(vsi, vsi_name, add_list,
2089 						     add_head, num_add,
2090 						     &promisc_changed);
2091 				memset(add_list, 0, list_size);
2092 				num_add = 0;
2093 			}
2094 		}
2095 		if (num_add) {
2096 			i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2097 					     num_add, &promisc_changed);
2098 		}
2099 		/* Now move all of the filters from the temp add list back to
2100 		 * the VSI's list.
2101 		 */
2102 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2103 		hlist_for_each_entry_safe(f, h, &tmp_add_list, hlist) {
2104 			u64 key = i40e_addr_to_hkey(f->macaddr);
2105 
2106 			hlist_del(&f->hlist);
2107 			hash_add(vsi->mac_filter_hash, &f->hlist, key);
2108 		}
2109 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2110 		kfree(add_list);
2111 		add_list = NULL;
2112 	}
2113 
2114 	/* Check to see if we can drop out of overflow promiscuous mode. */
2115 	if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state) &&
2116 	    (vsi->active_filters < vsi->promisc_threshold)) {
2117 		int failed_count = 0;
2118 		/* See if we have any failed filters. We can't drop out of
2119 		 * promiscuous until these have all been deleted.
2120 		 */
2121 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2122 		hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2123 			if (f->state == I40E_FILTER_FAILED)
2124 				failed_count++;
2125 		}
2126 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2127 		if (!failed_count) {
2128 			dev_info(&pf->pdev->dev,
2129 				 "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2130 				 vsi_name);
2131 			clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2132 			promisc_changed = true;
2133 			vsi->promisc_threshold = 0;
2134 		}
2135 	}
2136 
2137 	/* if the VF is not trusted do not do promisc */
2138 	if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2139 		clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2140 		goto out;
2141 	}
2142 
2143 	/* check for changes in promiscuous modes */
2144 	if (changed_flags & IFF_ALLMULTI) {
2145 		bool cur_multipromisc;
2146 
2147 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2148 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2149 							       vsi->seid,
2150 							       cur_multipromisc,
2151 							       NULL);
2152 		if (aq_ret) {
2153 			retval = i40e_aq_rc_to_posix(aq_ret,
2154 						     hw->aq.asq_last_status);
2155 			dev_info(&pf->pdev->dev,
2156 				 "set multi promisc failed on %s, err %s aq_err %s\n",
2157 				 vsi_name,
2158 				 i40e_stat_str(hw, aq_ret),
2159 				 i40e_aq_str(hw, hw->aq.asq_last_status));
2160 		}
2161 	}
2162 	if ((changed_flags & IFF_PROMISC) ||
2163 	    (promisc_changed &&
2164 	     test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))) {
2165 		bool cur_promisc;
2166 
2167 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2168 			       test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2169 					&vsi->state));
2170 		if ((vsi->type == I40E_VSI_MAIN) &&
2171 		    (pf->lan_veb != I40E_NO_VEB) &&
2172 		    !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2173 			/* set defport ON for Main VSI instead of true promisc
2174 			 * this way we will get all unicast/multicast and VLAN
2175 			 * promisc behavior but will not get VF or VMDq traffic
2176 			 * replicated on the Main VSI.
2177 			 */
2178 			if (pf->cur_promisc != cur_promisc) {
2179 				pf->cur_promisc = cur_promisc;
2180 				if (cur_promisc)
2181 					aq_ret =
2182 					      i40e_aq_set_default_vsi(hw,
2183 								      vsi->seid,
2184 								      NULL);
2185 				else
2186 					aq_ret =
2187 					    i40e_aq_clear_default_vsi(hw,
2188 								      vsi->seid,
2189 								      NULL);
2190 				if (aq_ret) {
2191 					retval = i40e_aq_rc_to_posix(aq_ret,
2192 							hw->aq.asq_last_status);
2193 					dev_info(&pf->pdev->dev,
2194 						 "Set default VSI failed on %s, err %s, aq_err %s\n",
2195 						 vsi_name,
2196 						 i40e_stat_str(hw, aq_ret),
2197 						 i40e_aq_str(hw,
2198 						     hw->aq.asq_last_status));
2199 				}
2200 			}
2201 		} else {
2202 			aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2203 							  hw,
2204 							  vsi->seid,
2205 							  cur_promisc, NULL,
2206 							  true);
2207 			if (aq_ret) {
2208 				retval =
2209 				i40e_aq_rc_to_posix(aq_ret,
2210 						    hw->aq.asq_last_status);
2211 				dev_info(&pf->pdev->dev,
2212 					 "set unicast promisc failed on %s, err %s, aq_err %s\n",
2213 					 vsi_name,
2214 					 i40e_stat_str(hw, aq_ret),
2215 					 i40e_aq_str(hw,
2216 						     hw->aq.asq_last_status));
2217 			}
2218 			aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2219 							  hw,
2220 							  vsi->seid,
2221 							  cur_promisc, NULL);
2222 			if (aq_ret) {
2223 				retval =
2224 				i40e_aq_rc_to_posix(aq_ret,
2225 						    hw->aq.asq_last_status);
2226 				dev_info(&pf->pdev->dev,
2227 					 "set multicast promisc failed on %s, err %s, aq_err %s\n",
2228 					 vsi_name,
2229 					 i40e_stat_str(hw, aq_ret),
2230 					 i40e_aq_str(hw,
2231 						     hw->aq.asq_last_status));
2232 			}
2233 		}
2234 		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2235 						   vsi->seid,
2236 						   cur_promisc, NULL);
2237 		if (aq_ret) {
2238 			retval = i40e_aq_rc_to_posix(aq_ret,
2239 						     pf->hw.aq.asq_last_status);
2240 			dev_info(&pf->pdev->dev,
2241 				 "set brdcast promisc failed, err %s, aq_err %s\n",
2242 					 i40e_stat_str(hw, aq_ret),
2243 					 i40e_aq_str(hw,
2244 						     hw->aq.asq_last_status));
2245 		}
2246 	}
2247 out:
2248 	/* if something went wrong then set the changed flag so we try again */
2249 	if (retval)
2250 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2251 
2252 	clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2253 	return retval;
2254 
2255 err_no_memory:
2256 	/* Restore elements on the temporary add and delete lists */
2257 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2258 err_no_memory_locked:
2259 	i40e_undo_filter_entries(vsi, &tmp_del_list);
2260 	i40e_undo_filter_entries(vsi, &tmp_add_list);
2261 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2262 
2263 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2264 	clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2265 	return -ENOMEM;
2266 }
2267 
2268 /**
2269  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2270  * @pf: board private structure
2271  **/
2272 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2273 {
2274 	int v;
2275 
2276 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2277 		return;
2278 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2279 
2280 	for (v = 0; v < pf->num_alloc_vsi; v++) {
2281 		if (pf->vsi[v] &&
2282 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2283 			int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2284 
2285 			if (ret) {
2286 				/* come back and try again later */
2287 				pf->flags |= I40E_FLAG_FILTER_SYNC;
2288 				break;
2289 			}
2290 		}
2291 	}
2292 }
2293 
2294 /**
2295  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2296  * @netdev: network interface device structure
2297  * @new_mtu: new value for maximum frame size
2298  *
2299  * Returns 0 on success, negative on failure
2300  **/
2301 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2302 {
2303 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2304 	struct i40e_vsi *vsi = np->vsi;
2305 
2306 	netdev_info(netdev, "changing MTU from %d to %d\n",
2307 		    netdev->mtu, new_mtu);
2308 	netdev->mtu = new_mtu;
2309 	if (netif_running(netdev))
2310 		i40e_vsi_reinit_locked(vsi);
2311 	i40e_notify_client_of_l2_param_changes(vsi);
2312 	return 0;
2313 }
2314 
2315 /**
2316  * i40e_ioctl - Access the hwtstamp interface
2317  * @netdev: network interface device structure
2318  * @ifr: interface request data
2319  * @cmd: ioctl command
2320  **/
2321 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2322 {
2323 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2324 	struct i40e_pf *pf = np->vsi->back;
2325 
2326 	switch (cmd) {
2327 	case SIOCGHWTSTAMP:
2328 		return i40e_ptp_get_ts_config(pf, ifr);
2329 	case SIOCSHWTSTAMP:
2330 		return i40e_ptp_set_ts_config(pf, ifr);
2331 	default:
2332 		return -EOPNOTSUPP;
2333 	}
2334 }
2335 
2336 /**
2337  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2338  * @vsi: the vsi being adjusted
2339  **/
2340 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2341 {
2342 	struct i40e_vsi_context ctxt;
2343 	i40e_status ret;
2344 
2345 	if ((vsi->info.valid_sections &
2346 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2347 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2348 		return;  /* already enabled */
2349 
2350 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2351 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2352 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2353 
2354 	ctxt.seid = vsi->seid;
2355 	ctxt.info = vsi->info;
2356 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2357 	if (ret) {
2358 		dev_info(&vsi->back->pdev->dev,
2359 			 "update vlan stripping failed, err %s aq_err %s\n",
2360 			 i40e_stat_str(&vsi->back->hw, ret),
2361 			 i40e_aq_str(&vsi->back->hw,
2362 				     vsi->back->hw.aq.asq_last_status));
2363 	}
2364 }
2365 
2366 /**
2367  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2368  * @vsi: the vsi being adjusted
2369  **/
2370 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2371 {
2372 	struct i40e_vsi_context ctxt;
2373 	i40e_status ret;
2374 
2375 	if ((vsi->info.valid_sections &
2376 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2377 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2378 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2379 		return;  /* already disabled */
2380 
2381 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2382 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2383 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2384 
2385 	ctxt.seid = vsi->seid;
2386 	ctxt.info = vsi->info;
2387 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2388 	if (ret) {
2389 		dev_info(&vsi->back->pdev->dev,
2390 			 "update vlan stripping failed, err %s aq_err %s\n",
2391 			 i40e_stat_str(&vsi->back->hw, ret),
2392 			 i40e_aq_str(&vsi->back->hw,
2393 				     vsi->back->hw.aq.asq_last_status));
2394 	}
2395 }
2396 
2397 /**
2398  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2399  * @netdev: network interface to be adjusted
2400  * @features: netdev features to test if VLAN offload is enabled or not
2401  **/
2402 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2403 {
2404 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2405 	struct i40e_vsi *vsi = np->vsi;
2406 
2407 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2408 		i40e_vlan_stripping_enable(vsi);
2409 	else
2410 		i40e_vlan_stripping_disable(vsi);
2411 }
2412 
2413 /**
2414  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2415  * @vsi: the vsi being configured
2416  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2417  **/
2418 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2419 {
2420 	struct i40e_mac_filter *f, *add_f, *del_f;
2421 	struct hlist_node *h;
2422 	int bkt;
2423 
2424 	/* Locked once because all functions invoked below iterates list*/
2425 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2426 
2427 	if (vsi->netdev) {
2428 		add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid);
2429 		if (!add_f) {
2430 			dev_info(&vsi->back->pdev->dev,
2431 				 "Could not add vlan filter %d for %pM\n",
2432 				 vid, vsi->netdev->dev_addr);
2433 			spin_unlock_bh(&vsi->mac_filter_hash_lock);
2434 			return -ENOMEM;
2435 		}
2436 	}
2437 
2438 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2439 		if (f->state == I40E_FILTER_REMOVE)
2440 			continue;
2441 		add_f = i40e_add_filter(vsi, f->macaddr, vid);
2442 		if (!add_f) {
2443 			dev_info(&vsi->back->pdev->dev,
2444 				 "Could not add vlan filter %d for %pM\n",
2445 				 vid, f->macaddr);
2446 			spin_unlock_bh(&vsi->mac_filter_hash_lock);
2447 			return -ENOMEM;
2448 		}
2449 	}
2450 
2451 	/* Now if we add a vlan tag, make sure to check if it is the first
2452 	 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2453 	 * with 0, so we now accept untagged and specified tagged traffic
2454 	 * (and not all tags along with untagged)
2455 	 */
2456 	if (vid > 0 && vsi->netdev) {
2457 		del_f = i40e_find_filter(vsi, vsi->netdev->dev_addr,
2458 					 I40E_VLAN_ANY);
2459 		if (del_f) {
2460 			__i40e_del_filter(vsi, del_f);
2461 			add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0);
2462 			if (!add_f) {
2463 				dev_info(&vsi->back->pdev->dev,
2464 					 "Could not add filter 0 for %pM\n",
2465 					 vsi->netdev->dev_addr);
2466 				spin_unlock_bh(&vsi->mac_filter_hash_lock);
2467 				return -ENOMEM;
2468 			}
2469 		}
2470 	}
2471 
2472 	/* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2473 	if (vid > 0 && !vsi->info.pvid) {
2474 		hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2475 			if (f->state == I40E_FILTER_REMOVE)
2476 				continue;
2477 			del_f = i40e_find_filter(vsi, f->macaddr,
2478 						 I40E_VLAN_ANY);
2479 			if (!del_f)
2480 				continue;
2481 			__i40e_del_filter(vsi, del_f);
2482 			add_f = i40e_add_filter(vsi, f->macaddr, 0);
2483 			if (!add_f) {
2484 				dev_info(&vsi->back->pdev->dev,
2485 					 "Could not add filter 0 for %pM\n",
2486 					f->macaddr);
2487 				spin_unlock_bh(&vsi->mac_filter_hash_lock);
2488 				return -ENOMEM;
2489 			}
2490 		}
2491 	}
2492 
2493 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2494 
2495 	/* schedule our worker thread which will take care of
2496 	 * applying the new filter changes
2497 	 */
2498 	i40e_service_event_schedule(vsi->back);
2499 	return 0;
2500 }
2501 
2502 /**
2503  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2504  * @vsi: the vsi being configured
2505  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2506  **/
2507 void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2508 {
2509 	struct net_device *netdev = vsi->netdev;
2510 	struct i40e_mac_filter *f;
2511 	struct hlist_node *h;
2512 	int bkt;
2513 
2514 	/* Locked once because all functions invoked below iterates list */
2515 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2516 
2517 	if (vsi->netdev)
2518 		i40e_del_filter(vsi, netdev->dev_addr, vid);
2519 
2520 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2521 		if (f->vlan == vid)
2522 			__i40e_del_filter(vsi, f);
2523 	}
2524 
2525 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2526 
2527 	/* schedule our worker thread which will take care of
2528 	 * applying the new filter changes
2529 	 */
2530 	i40e_service_event_schedule(vsi->back);
2531 }
2532 
2533 /**
2534  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2535  * @netdev: network interface to be adjusted
2536  * @vid: vlan id to be added
2537  *
2538  * net_device_ops implementation for adding vlan ids
2539  **/
2540 #ifdef I40E_FCOE
2541 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2542 			 __always_unused __be16 proto, u16 vid)
2543 #else
2544 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2545 				__always_unused __be16 proto, u16 vid)
2546 #endif
2547 {
2548 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2549 	struct i40e_vsi *vsi = np->vsi;
2550 	int ret = 0;
2551 
2552 	if (vid > 4095)
2553 		return -EINVAL;
2554 
2555 	/* If the network stack called us with vid = 0 then
2556 	 * it is asking to receive priority tagged packets with
2557 	 * vlan id 0.  Our HW receives them by default when configured
2558 	 * to receive untagged packets so there is no need to add an
2559 	 * extra filter for vlan 0 tagged packets.
2560 	 */
2561 	if (vid)
2562 		ret = i40e_vsi_add_vlan(vsi, vid);
2563 
2564 	if (!ret && (vid < VLAN_N_VID))
2565 		set_bit(vid, vsi->active_vlans);
2566 
2567 	return ret;
2568 }
2569 
2570 /**
2571  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2572  * @netdev: network interface to be adjusted
2573  * @vid: vlan id to be removed
2574  *
2575  * net_device_ops implementation for removing vlan ids
2576  **/
2577 #ifdef I40E_FCOE
2578 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2579 			  __always_unused __be16 proto, u16 vid)
2580 #else
2581 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2582 				 __always_unused __be16 proto, u16 vid)
2583 #endif
2584 {
2585 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2586 	struct i40e_vsi *vsi = np->vsi;
2587 
2588 	/* return code is ignored as there is nothing a user
2589 	 * can do about failure to remove and a log message was
2590 	 * already printed from the other function
2591 	 */
2592 	i40e_vsi_kill_vlan(vsi, vid);
2593 
2594 	clear_bit(vid, vsi->active_vlans);
2595 
2596 	return 0;
2597 }
2598 
2599 /**
2600  * i40e_macaddr_init - explicitly write the mac address filters
2601  *
2602  * @vsi: pointer to the vsi
2603  * @macaddr: the MAC address
2604  *
2605  * This is needed when the macaddr has been obtained by other
2606  * means than the default, e.g., from Open Firmware or IDPROM.
2607  * Returns 0 on success, negative on failure
2608  **/
2609 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
2610 {
2611 	int ret;
2612 	struct i40e_aqc_add_macvlan_element_data element;
2613 
2614 	ret = i40e_aq_mac_address_write(&vsi->back->hw,
2615 					I40E_AQC_WRITE_TYPE_LAA_WOL,
2616 					macaddr, NULL);
2617 	if (ret) {
2618 		dev_info(&vsi->back->pdev->dev,
2619 			 "Addr change for VSI failed: %d\n", ret);
2620 		return -EADDRNOTAVAIL;
2621 	}
2622 
2623 	memset(&element, 0, sizeof(element));
2624 	ether_addr_copy(element.mac_addr, macaddr);
2625 	element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
2626 	ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
2627 	if (ret) {
2628 		dev_info(&vsi->back->pdev->dev,
2629 			 "add filter failed err %s aq_err %s\n",
2630 			 i40e_stat_str(&vsi->back->hw, ret),
2631 			 i40e_aq_str(&vsi->back->hw,
2632 				     vsi->back->hw.aq.asq_last_status));
2633 	}
2634 	return ret;
2635 }
2636 
2637 /**
2638  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2639  * @vsi: the vsi being brought back up
2640  **/
2641 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2642 {
2643 	u16 vid;
2644 
2645 	if (!vsi->netdev)
2646 		return;
2647 
2648 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2649 
2650 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2651 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2652 				     vid);
2653 }
2654 
2655 /**
2656  * i40e_vsi_add_pvid - Add pvid for the VSI
2657  * @vsi: the vsi being adjusted
2658  * @vid: the vlan id to set as a PVID
2659  **/
2660 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2661 {
2662 	struct i40e_vsi_context ctxt;
2663 	i40e_status ret;
2664 
2665 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2666 	vsi->info.pvid = cpu_to_le16(vid);
2667 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2668 				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2669 				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2670 
2671 	ctxt.seid = vsi->seid;
2672 	ctxt.info = vsi->info;
2673 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2674 	if (ret) {
2675 		dev_info(&vsi->back->pdev->dev,
2676 			 "add pvid failed, err %s aq_err %s\n",
2677 			 i40e_stat_str(&vsi->back->hw, ret),
2678 			 i40e_aq_str(&vsi->back->hw,
2679 				     vsi->back->hw.aq.asq_last_status));
2680 		return -ENOENT;
2681 	}
2682 
2683 	return 0;
2684 }
2685 
2686 /**
2687  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2688  * @vsi: the vsi being adjusted
2689  *
2690  * Just use the vlan_rx_register() service to put it back to normal
2691  **/
2692 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2693 {
2694 	i40e_vlan_stripping_disable(vsi);
2695 
2696 	vsi->info.pvid = 0;
2697 }
2698 
2699 /**
2700  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2701  * @vsi: ptr to the VSI
2702  *
2703  * If this function returns with an error, then it's possible one or
2704  * more of the rings is populated (while the rest are not).  It is the
2705  * callers duty to clean those orphaned rings.
2706  *
2707  * Return 0 on success, negative on failure
2708  **/
2709 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2710 {
2711 	int i, err = 0;
2712 
2713 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2714 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2715 
2716 	return err;
2717 }
2718 
2719 /**
2720  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2721  * @vsi: ptr to the VSI
2722  *
2723  * Free VSI's transmit software resources
2724  **/
2725 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2726 {
2727 	int i;
2728 
2729 	if (!vsi->tx_rings)
2730 		return;
2731 
2732 	for (i = 0; i < vsi->num_queue_pairs; i++)
2733 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2734 			i40e_free_tx_resources(vsi->tx_rings[i]);
2735 }
2736 
2737 /**
2738  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2739  * @vsi: ptr to the VSI
2740  *
2741  * If this function returns with an error, then it's possible one or
2742  * more of the rings is populated (while the rest are not).  It is the
2743  * callers duty to clean those orphaned rings.
2744  *
2745  * Return 0 on success, negative on failure
2746  **/
2747 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2748 {
2749 	int i, err = 0;
2750 
2751 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2752 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2753 #ifdef I40E_FCOE
2754 	i40e_fcoe_setup_ddp_resources(vsi);
2755 #endif
2756 	return err;
2757 }
2758 
2759 /**
2760  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2761  * @vsi: ptr to the VSI
2762  *
2763  * Free all receive software resources
2764  **/
2765 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2766 {
2767 	int i;
2768 
2769 	if (!vsi->rx_rings)
2770 		return;
2771 
2772 	for (i = 0; i < vsi->num_queue_pairs; i++)
2773 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2774 			i40e_free_rx_resources(vsi->rx_rings[i]);
2775 #ifdef I40E_FCOE
2776 	i40e_fcoe_free_ddp_resources(vsi);
2777 #endif
2778 }
2779 
2780 /**
2781  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2782  * @ring: The Tx ring to configure
2783  *
2784  * This enables/disables XPS for a given Tx descriptor ring
2785  * based on the TCs enabled for the VSI that ring belongs to.
2786  **/
2787 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2788 {
2789 	struct i40e_vsi *vsi = ring->vsi;
2790 	cpumask_var_t mask;
2791 
2792 	if (!ring->q_vector || !ring->netdev)
2793 		return;
2794 
2795 	/* Single TC mode enable XPS */
2796 	if (vsi->tc_config.numtc <= 1) {
2797 		if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2798 			netif_set_xps_queue(ring->netdev,
2799 					    &ring->q_vector->affinity_mask,
2800 					    ring->queue_index);
2801 	} else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2802 		/* Disable XPS to allow selection based on TC */
2803 		bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2804 		netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2805 		free_cpumask_var(mask);
2806 	}
2807 
2808 	/* schedule our worker thread which will take care of
2809 	 * applying the new filter changes
2810 	 */
2811 	i40e_service_event_schedule(vsi->back);
2812 }
2813 
2814 /**
2815  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2816  * @ring: The Tx ring to configure
2817  *
2818  * Configure the Tx descriptor ring in the HMC context.
2819  **/
2820 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2821 {
2822 	struct i40e_vsi *vsi = ring->vsi;
2823 	u16 pf_q = vsi->base_queue + ring->queue_index;
2824 	struct i40e_hw *hw = &vsi->back->hw;
2825 	struct i40e_hmc_obj_txq tx_ctx;
2826 	i40e_status err = 0;
2827 	u32 qtx_ctl = 0;
2828 
2829 	/* some ATR related tx ring init */
2830 	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2831 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2832 		ring->atr_count = 0;
2833 	} else {
2834 		ring->atr_sample_rate = 0;
2835 	}
2836 
2837 	/* configure XPS */
2838 	i40e_config_xps_tx_ring(ring);
2839 
2840 	/* clear the context structure first */
2841 	memset(&tx_ctx, 0, sizeof(tx_ctx));
2842 
2843 	tx_ctx.new_context = 1;
2844 	tx_ctx.base = (ring->dma / 128);
2845 	tx_ctx.qlen = ring->count;
2846 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2847 					       I40E_FLAG_FD_ATR_ENABLED));
2848 #ifdef I40E_FCOE
2849 	tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2850 #endif
2851 	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2852 	/* FDIR VSI tx ring can still use RS bit and writebacks */
2853 	if (vsi->type != I40E_VSI_FDIR)
2854 		tx_ctx.head_wb_ena = 1;
2855 	tx_ctx.head_wb_addr = ring->dma +
2856 			      (ring->count * sizeof(struct i40e_tx_desc));
2857 
2858 	/* As part of VSI creation/update, FW allocates certain
2859 	 * Tx arbitration queue sets for each TC enabled for
2860 	 * the VSI. The FW returns the handles to these queue
2861 	 * sets as part of the response buffer to Add VSI,
2862 	 * Update VSI, etc. AQ commands. It is expected that
2863 	 * these queue set handles be associated with the Tx
2864 	 * queues by the driver as part of the TX queue context
2865 	 * initialization. This has to be done regardless of
2866 	 * DCB as by default everything is mapped to TC0.
2867 	 */
2868 	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2869 	tx_ctx.rdylist_act = 0;
2870 
2871 	/* clear the context in the HMC */
2872 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2873 	if (err) {
2874 		dev_info(&vsi->back->pdev->dev,
2875 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2876 			 ring->queue_index, pf_q, err);
2877 		return -ENOMEM;
2878 	}
2879 
2880 	/* set the context in the HMC */
2881 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2882 	if (err) {
2883 		dev_info(&vsi->back->pdev->dev,
2884 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2885 			 ring->queue_index, pf_q, err);
2886 		return -ENOMEM;
2887 	}
2888 
2889 	/* Now associate this queue with this PCI function */
2890 	if (vsi->type == I40E_VSI_VMDQ2) {
2891 		qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2892 		qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2893 			   I40E_QTX_CTL_VFVM_INDX_MASK;
2894 	} else {
2895 		qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2896 	}
2897 
2898 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2899 		    I40E_QTX_CTL_PF_INDX_MASK);
2900 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2901 	i40e_flush(hw);
2902 
2903 	/* cache tail off for easier writes later */
2904 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2905 
2906 	return 0;
2907 }
2908 
2909 /**
2910  * i40e_configure_rx_ring - Configure a receive ring context
2911  * @ring: The Rx ring to configure
2912  *
2913  * Configure the Rx descriptor ring in the HMC context.
2914  **/
2915 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2916 {
2917 	struct i40e_vsi *vsi = ring->vsi;
2918 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2919 	u16 pf_q = vsi->base_queue + ring->queue_index;
2920 	struct i40e_hw *hw = &vsi->back->hw;
2921 	struct i40e_hmc_obj_rxq rx_ctx;
2922 	i40e_status err = 0;
2923 
2924 	ring->state = 0;
2925 
2926 	/* clear the context structure first */
2927 	memset(&rx_ctx, 0, sizeof(rx_ctx));
2928 
2929 	ring->rx_buf_len = vsi->rx_buf_len;
2930 
2931 	rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2932 
2933 	rx_ctx.base = (ring->dma / 128);
2934 	rx_ctx.qlen = ring->count;
2935 
2936 	/* use 32 byte descriptors */
2937 	rx_ctx.dsize = 1;
2938 
2939 	/* descriptor type is always zero
2940 	 * rx_ctx.dtype = 0;
2941 	 */
2942 	rx_ctx.hsplit_0 = 0;
2943 
2944 	rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
2945 	if (hw->revision_id == 0)
2946 		rx_ctx.lrxqthresh = 0;
2947 	else
2948 		rx_ctx.lrxqthresh = 2;
2949 	rx_ctx.crcstrip = 1;
2950 	rx_ctx.l2tsel = 1;
2951 	/* this controls whether VLAN is stripped from inner headers */
2952 	rx_ctx.showiv = 0;
2953 #ifdef I40E_FCOE
2954 	rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2955 #endif
2956 	/* set the prefena field to 1 because the manual says to */
2957 	rx_ctx.prefena = 1;
2958 
2959 	/* clear the context in the HMC */
2960 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2961 	if (err) {
2962 		dev_info(&vsi->back->pdev->dev,
2963 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2964 			 ring->queue_index, pf_q, err);
2965 		return -ENOMEM;
2966 	}
2967 
2968 	/* set the context in the HMC */
2969 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2970 	if (err) {
2971 		dev_info(&vsi->back->pdev->dev,
2972 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2973 			 ring->queue_index, pf_q, err);
2974 		return -ENOMEM;
2975 	}
2976 
2977 	/* cache tail for quicker writes, and clear the reg before use */
2978 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2979 	writel(0, ring->tail);
2980 
2981 	i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2982 
2983 	return 0;
2984 }
2985 
2986 /**
2987  * i40e_vsi_configure_tx - Configure the VSI for Tx
2988  * @vsi: VSI structure describing this set of rings and resources
2989  *
2990  * Configure the Tx VSI for operation.
2991  **/
2992 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2993 {
2994 	int err = 0;
2995 	u16 i;
2996 
2997 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2998 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2999 
3000 	return err;
3001 }
3002 
3003 /**
3004  * i40e_vsi_configure_rx - Configure the VSI for Rx
3005  * @vsi: the VSI being configured
3006  *
3007  * Configure the Rx VSI for operation.
3008  **/
3009 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3010 {
3011 	int err = 0;
3012 	u16 i;
3013 
3014 	if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
3015 		vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
3016 			       + ETH_FCS_LEN + VLAN_HLEN;
3017 	else
3018 		vsi->max_frame = I40E_RXBUFFER_2048;
3019 
3020 	vsi->rx_buf_len = I40E_RXBUFFER_2048;
3021 
3022 #ifdef I40E_FCOE
3023 	/* setup rx buffer for FCoE */
3024 	if ((vsi->type == I40E_VSI_FCOE) &&
3025 	    (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
3026 		vsi->rx_buf_len = I40E_RXBUFFER_3072;
3027 		vsi->max_frame = I40E_RXBUFFER_3072;
3028 	}
3029 
3030 #endif /* I40E_FCOE */
3031 	/* round up for the chip's needs */
3032 	vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
3033 				BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3034 
3035 	/* set up individual rings */
3036 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3037 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3038 
3039 	return err;
3040 }
3041 
3042 /**
3043  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3044  * @vsi: ptr to the VSI
3045  **/
3046 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3047 {
3048 	struct i40e_ring *tx_ring, *rx_ring;
3049 	u16 qoffset, qcount;
3050 	int i, n;
3051 
3052 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3053 		/* Reset the TC information */
3054 		for (i = 0; i < vsi->num_queue_pairs; i++) {
3055 			rx_ring = vsi->rx_rings[i];
3056 			tx_ring = vsi->tx_rings[i];
3057 			rx_ring->dcb_tc = 0;
3058 			tx_ring->dcb_tc = 0;
3059 		}
3060 	}
3061 
3062 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3063 		if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3064 			continue;
3065 
3066 		qoffset = vsi->tc_config.tc_info[n].qoffset;
3067 		qcount = vsi->tc_config.tc_info[n].qcount;
3068 		for (i = qoffset; i < (qoffset + qcount); i++) {
3069 			rx_ring = vsi->rx_rings[i];
3070 			tx_ring = vsi->tx_rings[i];
3071 			rx_ring->dcb_tc = n;
3072 			tx_ring->dcb_tc = n;
3073 		}
3074 	}
3075 }
3076 
3077 /**
3078  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3079  * @vsi: ptr to the VSI
3080  **/
3081 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3082 {
3083 	struct i40e_pf *pf = vsi->back;
3084 	int err;
3085 
3086 	if (vsi->netdev)
3087 		i40e_set_rx_mode(vsi->netdev);
3088 
3089 	if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
3090 		err = i40e_macaddr_init(vsi, pf->hw.mac.addr);
3091 		if (err) {
3092 			dev_warn(&pf->pdev->dev,
3093 				 "could not set up macaddr; err %d\n", err);
3094 		}
3095 	}
3096 }
3097 
3098 /**
3099  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3100  * @vsi: Pointer to the targeted VSI
3101  *
3102  * This function replays the hlist on the hw where all the SB Flow Director
3103  * filters were saved.
3104  **/
3105 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3106 {
3107 	struct i40e_fdir_filter *filter;
3108 	struct i40e_pf *pf = vsi->back;
3109 	struct hlist_node *node;
3110 
3111 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3112 		return;
3113 
3114 	hlist_for_each_entry_safe(filter, node,
3115 				  &pf->fdir_filter_list, fdir_node) {
3116 		i40e_add_del_fdir(vsi, filter, true);
3117 	}
3118 }
3119 
3120 /**
3121  * i40e_vsi_configure - Set up the VSI for action
3122  * @vsi: the VSI being configured
3123  **/
3124 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3125 {
3126 	int err;
3127 
3128 	i40e_set_vsi_rx_mode(vsi);
3129 	i40e_restore_vlan(vsi);
3130 	i40e_vsi_config_dcb_rings(vsi);
3131 	err = i40e_vsi_configure_tx(vsi);
3132 	if (!err)
3133 		err = i40e_vsi_configure_rx(vsi);
3134 
3135 	return err;
3136 }
3137 
3138 /**
3139  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3140  * @vsi: the VSI being configured
3141  **/
3142 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3143 {
3144 	struct i40e_pf *pf = vsi->back;
3145 	struct i40e_hw *hw = &pf->hw;
3146 	u16 vector;
3147 	int i, q;
3148 	u32 qp;
3149 
3150 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
3151 	 * and PFINT_LNKLSTn registers, e.g.:
3152 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3153 	 */
3154 	qp = vsi->base_queue;
3155 	vector = vsi->base_vector;
3156 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3157 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3158 
3159 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
3160 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3161 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
3162 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3163 		     q_vector->rx.itr);
3164 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3165 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
3166 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3167 		     q_vector->tx.itr);
3168 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
3169 		     INTRL_USEC_TO_REG(vsi->int_rate_limit));
3170 
3171 		/* Linked list for the queuepairs assigned to this vector */
3172 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3173 		for (q = 0; q < q_vector->num_ringpairs; q++) {
3174 			u32 val;
3175 
3176 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3177 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3178 			      (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3179 			      (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3180 			      (I40E_QUEUE_TYPE_TX
3181 				      << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3182 
3183 			wr32(hw, I40E_QINT_RQCTL(qp), val);
3184 
3185 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3186 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3187 			      (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3188 			      ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3189 			      (I40E_QUEUE_TYPE_RX
3190 				      << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3191 
3192 			/* Terminate the linked list */
3193 			if (q == (q_vector->num_ringpairs - 1))
3194 				val |= (I40E_QUEUE_END_OF_LIST
3195 					   << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3196 
3197 			wr32(hw, I40E_QINT_TQCTL(qp), val);
3198 			qp++;
3199 		}
3200 	}
3201 
3202 	i40e_flush(hw);
3203 }
3204 
3205 /**
3206  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3207  * @hw: ptr to the hardware info
3208  **/
3209 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3210 {
3211 	struct i40e_hw *hw = &pf->hw;
3212 	u32 val;
3213 
3214 	/* clear things first */
3215 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3216 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3217 
3218 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3219 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3220 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
3221 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3222 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3223 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3224 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3225 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3226 
3227 	if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3228 		val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3229 
3230 	if (pf->flags & I40E_FLAG_PTP)
3231 		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3232 
3233 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
3234 
3235 	/* SW_ITR_IDX = 0, but don't change INTENA */
3236 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3237 					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3238 
3239 	/* OTHER_ITR_IDX = 0 */
3240 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3241 }
3242 
3243 /**
3244  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3245  * @vsi: the VSI being configured
3246  **/
3247 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3248 {
3249 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3250 	struct i40e_pf *pf = vsi->back;
3251 	struct i40e_hw *hw = &pf->hw;
3252 	u32 val;
3253 
3254 	/* set the ITR configuration */
3255 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
3256 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3257 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
3258 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3259 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3260 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
3261 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3262 
3263 	i40e_enable_misc_int_causes(pf);
3264 
3265 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3266 	wr32(hw, I40E_PFINT_LNKLST0, 0);
3267 
3268 	/* Associate the queue pair to the vector and enable the queue int */
3269 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		      |
3270 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3271 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3272 
3273 	wr32(hw, I40E_QINT_RQCTL(0), val);
3274 
3275 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
3276 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3277 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3278 
3279 	wr32(hw, I40E_QINT_TQCTL(0), val);
3280 	i40e_flush(hw);
3281 }
3282 
3283 /**
3284  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3285  * @pf: board private structure
3286  **/
3287 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3288 {
3289 	struct i40e_hw *hw = &pf->hw;
3290 
3291 	wr32(hw, I40E_PFINT_DYN_CTL0,
3292 	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3293 	i40e_flush(hw);
3294 }
3295 
3296 /**
3297  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3298  * @pf: board private structure
3299  * @clearpba: true when all pending interrupt events should be cleared
3300  **/
3301 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3302 {
3303 	struct i40e_hw *hw = &pf->hw;
3304 	u32 val;
3305 
3306 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3307 	      (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3308 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3309 
3310 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
3311 	i40e_flush(hw);
3312 }
3313 
3314 /**
3315  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3316  * @irq: interrupt number
3317  * @data: pointer to a q_vector
3318  **/
3319 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3320 {
3321 	struct i40e_q_vector *q_vector = data;
3322 
3323 	if (!q_vector->tx.ring && !q_vector->rx.ring)
3324 		return IRQ_HANDLED;
3325 
3326 	napi_schedule_irqoff(&q_vector->napi);
3327 
3328 	return IRQ_HANDLED;
3329 }
3330 
3331 /**
3332  * i40e_irq_affinity_notify - Callback for affinity changes
3333  * @notify: context as to what irq was changed
3334  * @mask: the new affinity mask
3335  *
3336  * This is a callback function used by the irq_set_affinity_notifier function
3337  * so that we may register to receive changes to the irq affinity masks.
3338  **/
3339 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3340 				     const cpumask_t *mask)
3341 {
3342 	struct i40e_q_vector *q_vector =
3343 		container_of(notify, struct i40e_q_vector, affinity_notify);
3344 
3345 	q_vector->affinity_mask = *mask;
3346 }
3347 
3348 /**
3349  * i40e_irq_affinity_release - Callback for affinity notifier release
3350  * @ref: internal core kernel usage
3351  *
3352  * This is a callback function used by the irq_set_affinity_notifier function
3353  * to inform the current notification subscriber that they will no longer
3354  * receive notifications.
3355  **/
3356 static void i40e_irq_affinity_release(struct kref *ref) {}
3357 
3358 /**
3359  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3360  * @vsi: the VSI being configured
3361  * @basename: name for the vector
3362  *
3363  * Allocates MSI-X vectors and requests interrupts from the kernel.
3364  **/
3365 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3366 {
3367 	int q_vectors = vsi->num_q_vectors;
3368 	struct i40e_pf *pf = vsi->back;
3369 	int base = vsi->base_vector;
3370 	int rx_int_idx = 0;
3371 	int tx_int_idx = 0;
3372 	int vector, err;
3373 	int irq_num;
3374 
3375 	for (vector = 0; vector < q_vectors; vector++) {
3376 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3377 
3378 		irq_num = pf->msix_entries[base + vector].vector;
3379 
3380 		if (q_vector->tx.ring && q_vector->rx.ring) {
3381 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3382 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3383 			tx_int_idx++;
3384 		} else if (q_vector->rx.ring) {
3385 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3386 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3387 		} else if (q_vector->tx.ring) {
3388 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3389 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3390 		} else {
3391 			/* skip this unused q_vector */
3392 			continue;
3393 		}
3394 		err = request_irq(irq_num,
3395 				  vsi->irq_handler,
3396 				  0,
3397 				  q_vector->name,
3398 				  q_vector);
3399 		if (err) {
3400 			dev_info(&pf->pdev->dev,
3401 				 "MSIX request_irq failed, error: %d\n", err);
3402 			goto free_queue_irqs;
3403 		}
3404 
3405 		/* register for affinity change notifications */
3406 		q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3407 		q_vector->affinity_notify.release = i40e_irq_affinity_release;
3408 		irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3409 		/* assign the mask for this irq */
3410 		irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
3411 	}
3412 
3413 	vsi->irqs_ready = true;
3414 	return 0;
3415 
3416 free_queue_irqs:
3417 	while (vector) {
3418 		vector--;
3419 		irq_num = pf->msix_entries[base + vector].vector;
3420 		irq_set_affinity_notifier(irq_num, NULL);
3421 		irq_set_affinity_hint(irq_num, NULL);
3422 		free_irq(irq_num, &vsi->q_vectors[vector]);
3423 	}
3424 	return err;
3425 }
3426 
3427 /**
3428  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3429  * @vsi: the VSI being un-configured
3430  **/
3431 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3432 {
3433 	struct i40e_pf *pf = vsi->back;
3434 	struct i40e_hw *hw = &pf->hw;
3435 	int base = vsi->base_vector;
3436 	int i;
3437 
3438 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3439 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3440 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3441 	}
3442 
3443 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3444 		for (i = vsi->base_vector;
3445 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3446 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3447 
3448 		i40e_flush(hw);
3449 		for (i = 0; i < vsi->num_q_vectors; i++)
3450 			synchronize_irq(pf->msix_entries[i + base].vector);
3451 	} else {
3452 		/* Legacy and MSI mode - this stops all interrupt handling */
3453 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3454 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3455 		i40e_flush(hw);
3456 		synchronize_irq(pf->pdev->irq);
3457 	}
3458 }
3459 
3460 /**
3461  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3462  * @vsi: the VSI being configured
3463  **/
3464 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3465 {
3466 	struct i40e_pf *pf = vsi->back;
3467 	int i;
3468 
3469 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3470 		for (i = 0; i < vsi->num_q_vectors; i++)
3471 			i40e_irq_dynamic_enable(vsi, i);
3472 	} else {
3473 		i40e_irq_dynamic_enable_icr0(pf, true);
3474 	}
3475 
3476 	i40e_flush(&pf->hw);
3477 	return 0;
3478 }
3479 
3480 /**
3481  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3482  * @pf: board private structure
3483  **/
3484 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3485 {
3486 	/* Disable ICR 0 */
3487 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3488 	i40e_flush(&pf->hw);
3489 }
3490 
3491 /**
3492  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3493  * @irq: interrupt number
3494  * @data: pointer to a q_vector
3495  *
3496  * This is the handler used for all MSI/Legacy interrupts, and deals
3497  * with both queue and non-queue interrupts.  This is also used in
3498  * MSIX mode to handle the non-queue interrupts.
3499  **/
3500 static irqreturn_t i40e_intr(int irq, void *data)
3501 {
3502 	struct i40e_pf *pf = (struct i40e_pf *)data;
3503 	struct i40e_hw *hw = &pf->hw;
3504 	irqreturn_t ret = IRQ_NONE;
3505 	u32 icr0, icr0_remaining;
3506 	u32 val, ena_mask;
3507 
3508 	icr0 = rd32(hw, I40E_PFINT_ICR0);
3509 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3510 
3511 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3512 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3513 		goto enable_intr;
3514 
3515 	/* if interrupt but no bits showing, must be SWINT */
3516 	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3517 	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3518 		pf->sw_int_count++;
3519 
3520 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3521 	    (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3522 		ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3523 		icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3524 		dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3525 	}
3526 
3527 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3528 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3529 		struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3530 		struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3531 
3532 		/* We do not have a way to disarm Queue causes while leaving
3533 		 * interrupt enabled for all other causes, ideally
3534 		 * interrupt should be disabled while we are in NAPI but
3535 		 * this is not a performance path and napi_schedule()
3536 		 * can deal with rescheduling.
3537 		 */
3538 		if (!test_bit(__I40E_DOWN, &pf->state))
3539 			napi_schedule_irqoff(&q_vector->napi);
3540 	}
3541 
3542 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3543 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3544 		set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3545 		i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3546 	}
3547 
3548 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3549 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3550 		set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3551 	}
3552 
3553 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3554 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3555 		set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3556 	}
3557 
3558 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3559 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3560 			set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3561 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3562 		val = rd32(hw, I40E_GLGEN_RSTAT);
3563 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3564 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3565 		if (val == I40E_RESET_CORER) {
3566 			pf->corer_count++;
3567 		} else if (val == I40E_RESET_GLOBR) {
3568 			pf->globr_count++;
3569 		} else if (val == I40E_RESET_EMPR) {
3570 			pf->empr_count++;
3571 			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3572 		}
3573 	}
3574 
3575 	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3576 		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3577 		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3578 		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3579 			 rd32(hw, I40E_PFHMC_ERRORINFO),
3580 			 rd32(hw, I40E_PFHMC_ERRORDATA));
3581 	}
3582 
3583 	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3584 		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3585 
3586 		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3587 			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3588 			i40e_ptp_tx_hwtstamp(pf);
3589 		}
3590 	}
3591 
3592 	/* If a critical error is pending we have no choice but to reset the
3593 	 * device.
3594 	 * Report and mask out any remaining unexpected interrupts.
3595 	 */
3596 	icr0_remaining = icr0 & ena_mask;
3597 	if (icr0_remaining) {
3598 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3599 			 icr0_remaining);
3600 		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3601 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3602 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3603 			dev_info(&pf->pdev->dev, "device will be reset\n");
3604 			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3605 			i40e_service_event_schedule(pf);
3606 		}
3607 		ena_mask &= ~icr0_remaining;
3608 	}
3609 	ret = IRQ_HANDLED;
3610 
3611 enable_intr:
3612 	/* re-enable interrupt causes */
3613 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3614 	if (!test_bit(__I40E_DOWN, &pf->state)) {
3615 		i40e_service_event_schedule(pf);
3616 		i40e_irq_dynamic_enable_icr0(pf, false);
3617 	}
3618 
3619 	return ret;
3620 }
3621 
3622 /**
3623  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3624  * @tx_ring:  tx ring to clean
3625  * @budget:   how many cleans we're allowed
3626  *
3627  * Returns true if there's any budget left (e.g. the clean is finished)
3628  **/
3629 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3630 {
3631 	struct i40e_vsi *vsi = tx_ring->vsi;
3632 	u16 i = tx_ring->next_to_clean;
3633 	struct i40e_tx_buffer *tx_buf;
3634 	struct i40e_tx_desc *tx_desc;
3635 
3636 	tx_buf = &tx_ring->tx_bi[i];
3637 	tx_desc = I40E_TX_DESC(tx_ring, i);
3638 	i -= tx_ring->count;
3639 
3640 	do {
3641 		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3642 
3643 		/* if next_to_watch is not set then there is no work pending */
3644 		if (!eop_desc)
3645 			break;
3646 
3647 		/* prevent any other reads prior to eop_desc */
3648 		read_barrier_depends();
3649 
3650 		/* if the descriptor isn't done, no work yet to do */
3651 		if (!(eop_desc->cmd_type_offset_bsz &
3652 		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3653 			break;
3654 
3655 		/* clear next_to_watch to prevent false hangs */
3656 		tx_buf->next_to_watch = NULL;
3657 
3658 		tx_desc->buffer_addr = 0;
3659 		tx_desc->cmd_type_offset_bsz = 0;
3660 		/* move past filter desc */
3661 		tx_buf++;
3662 		tx_desc++;
3663 		i++;
3664 		if (unlikely(!i)) {
3665 			i -= tx_ring->count;
3666 			tx_buf = tx_ring->tx_bi;
3667 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3668 		}
3669 		/* unmap skb header data */
3670 		dma_unmap_single(tx_ring->dev,
3671 				 dma_unmap_addr(tx_buf, dma),
3672 				 dma_unmap_len(tx_buf, len),
3673 				 DMA_TO_DEVICE);
3674 		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3675 			kfree(tx_buf->raw_buf);
3676 
3677 		tx_buf->raw_buf = NULL;
3678 		tx_buf->tx_flags = 0;
3679 		tx_buf->next_to_watch = NULL;
3680 		dma_unmap_len_set(tx_buf, len, 0);
3681 		tx_desc->buffer_addr = 0;
3682 		tx_desc->cmd_type_offset_bsz = 0;
3683 
3684 		/* move us past the eop_desc for start of next FD desc */
3685 		tx_buf++;
3686 		tx_desc++;
3687 		i++;
3688 		if (unlikely(!i)) {
3689 			i -= tx_ring->count;
3690 			tx_buf = tx_ring->tx_bi;
3691 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3692 		}
3693 
3694 		/* update budget accounting */
3695 		budget--;
3696 	} while (likely(budget));
3697 
3698 	i += tx_ring->count;
3699 	tx_ring->next_to_clean = i;
3700 
3701 	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3702 		i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3703 
3704 	return budget > 0;
3705 }
3706 
3707 /**
3708  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3709  * @irq: interrupt number
3710  * @data: pointer to a q_vector
3711  **/
3712 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3713 {
3714 	struct i40e_q_vector *q_vector = data;
3715 	struct i40e_vsi *vsi;
3716 
3717 	if (!q_vector->tx.ring)
3718 		return IRQ_HANDLED;
3719 
3720 	vsi = q_vector->tx.ring->vsi;
3721 	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3722 
3723 	return IRQ_HANDLED;
3724 }
3725 
3726 /**
3727  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3728  * @vsi: the VSI being configured
3729  * @v_idx: vector index
3730  * @qp_idx: queue pair index
3731  **/
3732 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3733 {
3734 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3735 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3736 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3737 
3738 	tx_ring->q_vector = q_vector;
3739 	tx_ring->next = q_vector->tx.ring;
3740 	q_vector->tx.ring = tx_ring;
3741 	q_vector->tx.count++;
3742 
3743 	rx_ring->q_vector = q_vector;
3744 	rx_ring->next = q_vector->rx.ring;
3745 	q_vector->rx.ring = rx_ring;
3746 	q_vector->rx.count++;
3747 }
3748 
3749 /**
3750  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3751  * @vsi: the VSI being configured
3752  *
3753  * This function maps descriptor rings to the queue-specific vectors
3754  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3755  * one vector per queue pair, but on a constrained vector budget, we
3756  * group the queue pairs as "efficiently" as possible.
3757  **/
3758 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3759 {
3760 	int qp_remaining = vsi->num_queue_pairs;
3761 	int q_vectors = vsi->num_q_vectors;
3762 	int num_ringpairs;
3763 	int v_start = 0;
3764 	int qp_idx = 0;
3765 
3766 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3767 	 * group them so there are multiple queues per vector.
3768 	 * It is also important to go through all the vectors available to be
3769 	 * sure that if we don't use all the vectors, that the remaining vectors
3770 	 * are cleared. This is especially important when decreasing the
3771 	 * number of queues in use.
3772 	 */
3773 	for (; v_start < q_vectors; v_start++) {
3774 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3775 
3776 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3777 
3778 		q_vector->num_ringpairs = num_ringpairs;
3779 
3780 		q_vector->rx.count = 0;
3781 		q_vector->tx.count = 0;
3782 		q_vector->rx.ring = NULL;
3783 		q_vector->tx.ring = NULL;
3784 
3785 		while (num_ringpairs--) {
3786 			i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3787 			qp_idx++;
3788 			qp_remaining--;
3789 		}
3790 	}
3791 }
3792 
3793 /**
3794  * i40e_vsi_request_irq - Request IRQ from the OS
3795  * @vsi: the VSI being configured
3796  * @basename: name for the vector
3797  **/
3798 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3799 {
3800 	struct i40e_pf *pf = vsi->back;
3801 	int err;
3802 
3803 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3804 		err = i40e_vsi_request_irq_msix(vsi, basename);
3805 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3806 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
3807 				  pf->int_name, pf);
3808 	else
3809 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3810 				  pf->int_name, pf);
3811 
3812 	if (err)
3813 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3814 
3815 	return err;
3816 }
3817 
3818 #ifdef CONFIG_NET_POLL_CONTROLLER
3819 /**
3820  * i40e_netpoll - A Polling 'interrupt' handler
3821  * @netdev: network interface device structure
3822  *
3823  * This is used by netconsole to send skbs without having to re-enable
3824  * interrupts.  It's not called while the normal interrupt routine is executing.
3825  **/
3826 #ifdef I40E_FCOE
3827 void i40e_netpoll(struct net_device *netdev)
3828 #else
3829 static void i40e_netpoll(struct net_device *netdev)
3830 #endif
3831 {
3832 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3833 	struct i40e_vsi *vsi = np->vsi;
3834 	struct i40e_pf *pf = vsi->back;
3835 	int i;
3836 
3837 	/* if interface is down do nothing */
3838 	if (test_bit(__I40E_DOWN, &vsi->state))
3839 		return;
3840 
3841 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3842 		for (i = 0; i < vsi->num_q_vectors; i++)
3843 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3844 	} else {
3845 		i40e_intr(pf->pdev->irq, netdev);
3846 	}
3847 }
3848 #endif
3849 
3850 /**
3851  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3852  * @pf: the PF being configured
3853  * @pf_q: the PF queue
3854  * @enable: enable or disable state of the queue
3855  *
3856  * This routine will wait for the given Tx queue of the PF to reach the
3857  * enabled or disabled state.
3858  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3859  * multiple retries; else will return 0 in case of success.
3860  **/
3861 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3862 {
3863 	int i;
3864 	u32 tx_reg;
3865 
3866 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3867 		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3868 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3869 			break;
3870 
3871 		usleep_range(10, 20);
3872 	}
3873 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3874 		return -ETIMEDOUT;
3875 
3876 	return 0;
3877 }
3878 
3879 /**
3880  * i40e_vsi_control_tx - Start or stop a VSI's rings
3881  * @vsi: the VSI being configured
3882  * @enable: start or stop the rings
3883  **/
3884 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3885 {
3886 	struct i40e_pf *pf = vsi->back;
3887 	struct i40e_hw *hw = &pf->hw;
3888 	int i, j, pf_q, ret = 0;
3889 	u32 tx_reg;
3890 
3891 	pf_q = vsi->base_queue;
3892 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3893 
3894 		/* warn the TX unit of coming changes */
3895 		i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3896 		if (!enable)
3897 			usleep_range(10, 20);
3898 
3899 		for (j = 0; j < 50; j++) {
3900 			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3901 			if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3902 			    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3903 				break;
3904 			usleep_range(1000, 2000);
3905 		}
3906 		/* Skip if the queue is already in the requested state */
3907 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3908 			continue;
3909 
3910 		/* turn on/off the queue */
3911 		if (enable) {
3912 			wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3913 			tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3914 		} else {
3915 			tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3916 		}
3917 
3918 		wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3919 		/* No waiting for the Tx queue to disable */
3920 		if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3921 			continue;
3922 
3923 		/* wait for the change to finish */
3924 		ret = i40e_pf_txq_wait(pf, pf_q, enable);
3925 		if (ret) {
3926 			dev_info(&pf->pdev->dev,
3927 				 "VSI seid %d Tx ring %d %sable timeout\n",
3928 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
3929 			break;
3930 		}
3931 	}
3932 
3933 	if (hw->revision_id == 0)
3934 		mdelay(50);
3935 	return ret;
3936 }
3937 
3938 /**
3939  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3940  * @pf: the PF being configured
3941  * @pf_q: the PF queue
3942  * @enable: enable or disable state of the queue
3943  *
3944  * This routine will wait for the given Rx queue of the PF to reach the
3945  * enabled or disabled state.
3946  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3947  * multiple retries; else will return 0 in case of success.
3948  **/
3949 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3950 {
3951 	int i;
3952 	u32 rx_reg;
3953 
3954 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3955 		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3956 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3957 			break;
3958 
3959 		usleep_range(10, 20);
3960 	}
3961 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3962 		return -ETIMEDOUT;
3963 
3964 	return 0;
3965 }
3966 
3967 /**
3968  * i40e_vsi_control_rx - Start or stop a VSI's rings
3969  * @vsi: the VSI being configured
3970  * @enable: start or stop the rings
3971  **/
3972 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3973 {
3974 	struct i40e_pf *pf = vsi->back;
3975 	struct i40e_hw *hw = &pf->hw;
3976 	int i, j, pf_q, ret = 0;
3977 	u32 rx_reg;
3978 
3979 	pf_q = vsi->base_queue;
3980 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3981 		for (j = 0; j < 50; j++) {
3982 			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3983 			if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3984 			    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3985 				break;
3986 			usleep_range(1000, 2000);
3987 		}
3988 
3989 		/* Skip if the queue is already in the requested state */
3990 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3991 			continue;
3992 
3993 		/* turn on/off the queue */
3994 		if (enable)
3995 			rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3996 		else
3997 			rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3998 		wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3999 		/* No waiting for the Tx queue to disable */
4000 		if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
4001 			continue;
4002 
4003 		/* wait for the change to finish */
4004 		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4005 		if (ret) {
4006 			dev_info(&pf->pdev->dev,
4007 				 "VSI seid %d Rx ring %d %sable timeout\n",
4008 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
4009 			break;
4010 		}
4011 	}
4012 
4013 	return ret;
4014 }
4015 
4016 /**
4017  * i40e_vsi_start_rings - Start a VSI's rings
4018  * @vsi: the VSI being configured
4019  **/
4020 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4021 {
4022 	int ret = 0;
4023 
4024 	/* do rx first for enable and last for disable */
4025 	ret = i40e_vsi_control_rx(vsi, true);
4026 	if (ret)
4027 		return ret;
4028 	ret = i40e_vsi_control_tx(vsi, true);
4029 
4030 	return ret;
4031 }
4032 
4033 /**
4034  * i40e_vsi_stop_rings - Stop a VSI's rings
4035  * @vsi: the VSI being configured
4036  **/
4037 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4038 {
4039 	/* do rx first for enable and last for disable
4040 	 * Ignore return value, we need to shutdown whatever we can
4041 	 */
4042 	i40e_vsi_control_tx(vsi, false);
4043 	i40e_vsi_control_rx(vsi, false);
4044 }
4045 
4046 /**
4047  * i40e_vsi_free_irq - Free the irq association with the OS
4048  * @vsi: the VSI being configured
4049  **/
4050 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4051 {
4052 	struct i40e_pf *pf = vsi->back;
4053 	struct i40e_hw *hw = &pf->hw;
4054 	int base = vsi->base_vector;
4055 	u32 val, qp;
4056 	int i;
4057 
4058 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4059 		if (!vsi->q_vectors)
4060 			return;
4061 
4062 		if (!vsi->irqs_ready)
4063 			return;
4064 
4065 		vsi->irqs_ready = false;
4066 		for (i = 0; i < vsi->num_q_vectors; i++) {
4067 			int irq_num;
4068 			u16 vector;
4069 
4070 			vector = i + base;
4071 			irq_num = pf->msix_entries[vector].vector;
4072 
4073 			/* free only the irqs that were actually requested */
4074 			if (!vsi->q_vectors[i] ||
4075 			    !vsi->q_vectors[i]->num_ringpairs)
4076 				continue;
4077 
4078 			/* clear the affinity notifier in the IRQ descriptor */
4079 			irq_set_affinity_notifier(irq_num, NULL);
4080 			/* clear the affinity_mask in the IRQ descriptor */
4081 			irq_set_affinity_hint(irq_num, NULL);
4082 			synchronize_irq(irq_num);
4083 			free_irq(irq_num, vsi->q_vectors[i]);
4084 
4085 			/* Tear down the interrupt queue link list
4086 			 *
4087 			 * We know that they come in pairs and always
4088 			 * the Rx first, then the Tx.  To clear the
4089 			 * link list, stick the EOL value into the
4090 			 * next_q field of the registers.
4091 			 */
4092 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4093 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4094 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4095 			val |= I40E_QUEUE_END_OF_LIST
4096 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4097 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4098 
4099 			while (qp != I40E_QUEUE_END_OF_LIST) {
4100 				u32 next;
4101 
4102 				val = rd32(hw, I40E_QINT_RQCTL(qp));
4103 
4104 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4105 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4106 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4107 					 I40E_QINT_RQCTL_INTEVENT_MASK);
4108 
4109 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4110 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4111 
4112 				wr32(hw, I40E_QINT_RQCTL(qp), val);
4113 
4114 				val = rd32(hw, I40E_QINT_TQCTL(qp));
4115 
4116 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4117 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4118 
4119 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4120 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4121 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4122 					 I40E_QINT_TQCTL_INTEVENT_MASK);
4123 
4124 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4125 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4126 
4127 				wr32(hw, I40E_QINT_TQCTL(qp), val);
4128 				qp = next;
4129 			}
4130 		}
4131 	} else {
4132 		free_irq(pf->pdev->irq, pf);
4133 
4134 		val = rd32(hw, I40E_PFINT_LNKLST0);
4135 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4136 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4137 		val |= I40E_QUEUE_END_OF_LIST
4138 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4139 		wr32(hw, I40E_PFINT_LNKLST0, val);
4140 
4141 		val = rd32(hw, I40E_QINT_RQCTL(qp));
4142 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4143 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4144 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4145 			 I40E_QINT_RQCTL_INTEVENT_MASK);
4146 
4147 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4148 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4149 
4150 		wr32(hw, I40E_QINT_RQCTL(qp), val);
4151 
4152 		val = rd32(hw, I40E_QINT_TQCTL(qp));
4153 
4154 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4155 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4156 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4157 			 I40E_QINT_TQCTL_INTEVENT_MASK);
4158 
4159 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4160 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4161 
4162 		wr32(hw, I40E_QINT_TQCTL(qp), val);
4163 	}
4164 }
4165 
4166 /**
4167  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4168  * @vsi: the VSI being configured
4169  * @v_idx: Index of vector to be freed
4170  *
4171  * This function frees the memory allocated to the q_vector.  In addition if
4172  * NAPI is enabled it will delete any references to the NAPI struct prior
4173  * to freeing the q_vector.
4174  **/
4175 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4176 {
4177 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4178 	struct i40e_ring *ring;
4179 
4180 	if (!q_vector)
4181 		return;
4182 
4183 	/* disassociate q_vector from rings */
4184 	i40e_for_each_ring(ring, q_vector->tx)
4185 		ring->q_vector = NULL;
4186 
4187 	i40e_for_each_ring(ring, q_vector->rx)
4188 		ring->q_vector = NULL;
4189 
4190 	/* only VSI w/ an associated netdev is set up w/ NAPI */
4191 	if (vsi->netdev)
4192 		netif_napi_del(&q_vector->napi);
4193 
4194 	vsi->q_vectors[v_idx] = NULL;
4195 
4196 	kfree_rcu(q_vector, rcu);
4197 }
4198 
4199 /**
4200  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4201  * @vsi: the VSI being un-configured
4202  *
4203  * This frees the memory allocated to the q_vectors and
4204  * deletes references to the NAPI struct.
4205  **/
4206 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4207 {
4208 	int v_idx;
4209 
4210 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4211 		i40e_free_q_vector(vsi, v_idx);
4212 }
4213 
4214 /**
4215  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4216  * @pf: board private structure
4217  **/
4218 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4219 {
4220 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4221 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4222 		pci_disable_msix(pf->pdev);
4223 		kfree(pf->msix_entries);
4224 		pf->msix_entries = NULL;
4225 		kfree(pf->irq_pile);
4226 		pf->irq_pile = NULL;
4227 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4228 		pci_disable_msi(pf->pdev);
4229 	}
4230 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4231 }
4232 
4233 /**
4234  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4235  * @pf: board private structure
4236  *
4237  * We go through and clear interrupt specific resources and reset the structure
4238  * to pre-load conditions
4239  **/
4240 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4241 {
4242 	int i;
4243 
4244 	i40e_stop_misc_vector(pf);
4245 	if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4246 		synchronize_irq(pf->msix_entries[0].vector);
4247 		free_irq(pf->msix_entries[0].vector, pf);
4248 	}
4249 
4250 	i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4251 		      I40E_IWARP_IRQ_PILE_ID);
4252 
4253 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4254 	for (i = 0; i < pf->num_alloc_vsi; i++)
4255 		if (pf->vsi[i])
4256 			i40e_vsi_free_q_vectors(pf->vsi[i]);
4257 	i40e_reset_interrupt_capability(pf);
4258 }
4259 
4260 /**
4261  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4262  * @vsi: the VSI being configured
4263  **/
4264 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4265 {
4266 	int q_idx;
4267 
4268 	if (!vsi->netdev)
4269 		return;
4270 
4271 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4272 		napi_enable(&vsi->q_vectors[q_idx]->napi);
4273 }
4274 
4275 /**
4276  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4277  * @vsi: the VSI being configured
4278  **/
4279 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4280 {
4281 	int q_idx;
4282 
4283 	if (!vsi->netdev)
4284 		return;
4285 
4286 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4287 		napi_disable(&vsi->q_vectors[q_idx]->napi);
4288 }
4289 
4290 /**
4291  * i40e_vsi_close - Shut down a VSI
4292  * @vsi: the vsi to be quelled
4293  **/
4294 static void i40e_vsi_close(struct i40e_vsi *vsi)
4295 {
4296 	bool reset = false;
4297 
4298 	if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4299 		i40e_down(vsi);
4300 	i40e_vsi_free_irq(vsi);
4301 	i40e_vsi_free_tx_resources(vsi);
4302 	i40e_vsi_free_rx_resources(vsi);
4303 	vsi->current_netdev_flags = 0;
4304 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4305 		reset = true;
4306 	i40e_notify_client_of_netdev_close(vsi, reset);
4307 }
4308 
4309 /**
4310  * i40e_quiesce_vsi - Pause a given VSI
4311  * @vsi: the VSI being paused
4312  **/
4313 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4314 {
4315 	if (test_bit(__I40E_DOWN, &vsi->state))
4316 		return;
4317 
4318 	/* No need to disable FCoE VSI when Tx suspended */
4319 	if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4320 	    vsi->type == I40E_VSI_FCOE) {
4321 		dev_dbg(&vsi->back->pdev->dev,
4322 			 "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4323 		return;
4324 	}
4325 
4326 	set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4327 	if (vsi->netdev && netif_running(vsi->netdev))
4328 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4329 	else
4330 		i40e_vsi_close(vsi);
4331 }
4332 
4333 /**
4334  * i40e_unquiesce_vsi - Resume a given VSI
4335  * @vsi: the VSI being resumed
4336  **/
4337 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4338 {
4339 	if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4340 		return;
4341 
4342 	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4343 	if (vsi->netdev && netif_running(vsi->netdev))
4344 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4345 	else
4346 		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4347 }
4348 
4349 /**
4350  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4351  * @pf: the PF
4352  **/
4353 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4354 {
4355 	int v;
4356 
4357 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4358 		if (pf->vsi[v])
4359 			i40e_quiesce_vsi(pf->vsi[v]);
4360 	}
4361 }
4362 
4363 /**
4364  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4365  * @pf: the PF
4366  **/
4367 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4368 {
4369 	int v;
4370 
4371 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4372 		if (pf->vsi[v])
4373 			i40e_unquiesce_vsi(pf->vsi[v]);
4374 	}
4375 }
4376 
4377 #ifdef CONFIG_I40E_DCB
4378 /**
4379  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4380  * @vsi: the VSI being configured
4381  *
4382  * This function waits for the given VSI's queues to be disabled.
4383  **/
4384 static int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4385 {
4386 	struct i40e_pf *pf = vsi->back;
4387 	int i, pf_q, ret;
4388 
4389 	pf_q = vsi->base_queue;
4390 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4391 		/* Check and wait for the disable status of the queue */
4392 		ret = i40e_pf_txq_wait(pf, pf_q, false);
4393 		if (ret) {
4394 			dev_info(&pf->pdev->dev,
4395 				 "VSI seid %d Tx ring %d disable timeout\n",
4396 				 vsi->seid, pf_q);
4397 			return ret;
4398 		}
4399 	}
4400 
4401 	pf_q = vsi->base_queue;
4402 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4403 		/* Check and wait for the disable status of the queue */
4404 		ret = i40e_pf_rxq_wait(pf, pf_q, false);
4405 		if (ret) {
4406 			dev_info(&pf->pdev->dev,
4407 				 "VSI seid %d Rx ring %d disable timeout\n",
4408 				 vsi->seid, pf_q);
4409 			return ret;
4410 		}
4411 	}
4412 
4413 	return 0;
4414 }
4415 
4416 /**
4417  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4418  * @pf: the PF
4419  *
4420  * This function waits for the queues to be in disabled state for all the
4421  * VSIs that are managed by this PF.
4422  **/
4423 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4424 {
4425 	int v, ret = 0;
4426 
4427 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4428 		/* No need to wait for FCoE VSI queues */
4429 		if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4430 			ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4431 			if (ret)
4432 				break;
4433 		}
4434 	}
4435 
4436 	return ret;
4437 }
4438 
4439 #endif
4440 
4441 /**
4442  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4443  * @q_idx: TX queue number
4444  * @vsi: Pointer to VSI struct
4445  *
4446  * This function checks specified queue for given VSI. Detects hung condition.
4447  * Sets hung bit since it is two step process. Before next run of service task
4448  * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4449  * hung condition remain unchanged and during subsequent run, this function
4450  * issues SW interrupt to recover from hung condition.
4451  **/
4452 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4453 {
4454 	struct i40e_ring *tx_ring = NULL;
4455 	struct i40e_pf	*pf;
4456 	u32 head, val, tx_pending_hw;
4457 	int i;
4458 
4459 	pf = vsi->back;
4460 
4461 	/* now that we have an index, find the tx_ring struct */
4462 	for (i = 0; i < vsi->num_queue_pairs; i++) {
4463 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4464 			if (q_idx == vsi->tx_rings[i]->queue_index) {
4465 				tx_ring = vsi->tx_rings[i];
4466 				break;
4467 			}
4468 		}
4469 	}
4470 
4471 	if (!tx_ring)
4472 		return;
4473 
4474 	/* Read interrupt register */
4475 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4476 		val = rd32(&pf->hw,
4477 			   I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4478 					       tx_ring->vsi->base_vector - 1));
4479 	else
4480 		val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4481 
4482 	head = i40e_get_head(tx_ring);
4483 
4484 	tx_pending_hw = i40e_get_tx_pending(tx_ring, false);
4485 
4486 	/* HW is done executing descriptors, updated HEAD write back,
4487 	 * but SW hasn't processed those descriptors. If interrupt is
4488 	 * not generated from this point ON, it could result into
4489 	 * dev_watchdog detecting timeout on those netdev_queue,
4490 	 * hence proactively trigger SW interrupt.
4491 	 */
4492 	if (tx_pending_hw && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4493 		/* NAPI Poll didn't run and clear since it was set */
4494 		if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4495 				       &tx_ring->q_vector->hung_detected)) {
4496 			netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending_hw: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4497 				    vsi->seid, q_idx, tx_pending_hw,
4498 				    tx_ring->next_to_clean, head,
4499 				    tx_ring->next_to_use,
4500 				    readl(tx_ring->tail));
4501 			netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4502 				    vsi->seid, q_idx, val);
4503 			i40e_force_wb(vsi, tx_ring->q_vector);
4504 		} else {
4505 			/* First Chance - detected possible hung */
4506 			set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4507 				&tx_ring->q_vector->hung_detected);
4508 		}
4509 	}
4510 
4511 	/* This is the case where we have interrupts missing,
4512 	 * so the tx_pending in HW will most likely be 0, but we
4513 	 * will have tx_pending in SW since the WB happened but the
4514 	 * interrupt got lost.
4515 	 */
4516 	if ((!tx_pending_hw) && i40e_get_tx_pending(tx_ring, true) &&
4517 	    (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4518 		if (napi_reschedule(&tx_ring->q_vector->napi))
4519 			tx_ring->tx_stats.tx_lost_interrupt++;
4520 	}
4521 }
4522 
4523 /**
4524  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4525  * @pf:  pointer to PF struct
4526  *
4527  * LAN VSI has netdev and netdev has TX queues. This function is to check
4528  * each of those TX queues if they are hung, trigger recovery by issuing
4529  * SW interrupt.
4530  **/
4531 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4532 {
4533 	struct net_device *netdev;
4534 	struct i40e_vsi *vsi;
4535 	int i;
4536 
4537 	/* Only for LAN VSI */
4538 	vsi = pf->vsi[pf->lan_vsi];
4539 
4540 	if (!vsi)
4541 		return;
4542 
4543 	/* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4544 	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4545 	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4546 		return;
4547 
4548 	/* Make sure type is MAIN VSI */
4549 	if (vsi->type != I40E_VSI_MAIN)
4550 		return;
4551 
4552 	netdev = vsi->netdev;
4553 	if (!netdev)
4554 		return;
4555 
4556 	/* Bail out if netif_carrier is not OK */
4557 	if (!netif_carrier_ok(netdev))
4558 		return;
4559 
4560 	/* Go thru' TX queues for netdev */
4561 	for (i = 0; i < netdev->num_tx_queues; i++) {
4562 		struct netdev_queue *q;
4563 
4564 		q = netdev_get_tx_queue(netdev, i);
4565 		if (q)
4566 			i40e_detect_recover_hung_queue(i, vsi);
4567 	}
4568 }
4569 
4570 /**
4571  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4572  * @pf: pointer to PF
4573  *
4574  * Get TC map for ISCSI PF type that will include iSCSI TC
4575  * and LAN TC.
4576  **/
4577 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4578 {
4579 	struct i40e_dcb_app_priority_table app;
4580 	struct i40e_hw *hw = &pf->hw;
4581 	u8 enabled_tc = 1; /* TC0 is always enabled */
4582 	u8 tc, i;
4583 	/* Get the iSCSI APP TLV */
4584 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4585 
4586 	for (i = 0; i < dcbcfg->numapps; i++) {
4587 		app = dcbcfg->app[i];
4588 		if (app.selector == I40E_APP_SEL_TCPIP &&
4589 		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4590 			tc = dcbcfg->etscfg.prioritytable[app.priority];
4591 			enabled_tc |= BIT(tc);
4592 			break;
4593 		}
4594 	}
4595 
4596 	return enabled_tc;
4597 }
4598 
4599 /**
4600  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4601  * @dcbcfg: the corresponding DCBx configuration structure
4602  *
4603  * Return the number of TCs from given DCBx configuration
4604  **/
4605 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4606 {
4607 	int i, tc_unused = 0;
4608 	u8 num_tc = 0;
4609 	u8 ret = 0;
4610 
4611 	/* Scan the ETS Config Priority Table to find
4612 	 * traffic class enabled for a given priority
4613 	 * and create a bitmask of enabled TCs
4614 	 */
4615 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4616 		num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4617 
4618 	/* Now scan the bitmask to check for
4619 	 * contiguous TCs starting with TC0
4620 	 */
4621 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4622 		if (num_tc & BIT(i)) {
4623 			if (!tc_unused) {
4624 				ret++;
4625 			} else {
4626 				pr_err("Non-contiguous TC - Disabling DCB\n");
4627 				return 1;
4628 			}
4629 		} else {
4630 			tc_unused = 1;
4631 		}
4632 	}
4633 
4634 	/* There is always at least TC0 */
4635 	if (!ret)
4636 		ret = 1;
4637 
4638 	return ret;
4639 }
4640 
4641 /**
4642  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4643  * @dcbcfg: the corresponding DCBx configuration structure
4644  *
4645  * Query the current DCB configuration and return the number of
4646  * traffic classes enabled from the given DCBX config
4647  **/
4648 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4649 {
4650 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4651 	u8 enabled_tc = 1;
4652 	u8 i;
4653 
4654 	for (i = 0; i < num_tc; i++)
4655 		enabled_tc |= BIT(i);
4656 
4657 	return enabled_tc;
4658 }
4659 
4660 /**
4661  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4662  * @pf: PF being queried
4663  *
4664  * Return number of traffic classes enabled for the given PF
4665  **/
4666 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4667 {
4668 	struct i40e_hw *hw = &pf->hw;
4669 	u8 i, enabled_tc = 1;
4670 	u8 num_tc = 0;
4671 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4672 
4673 	/* If DCB is not enabled then always in single TC */
4674 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4675 		return 1;
4676 
4677 	/* SFP mode will be enabled for all TCs on port */
4678 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4679 		return i40e_dcb_get_num_tc(dcbcfg);
4680 
4681 	/* MFP mode return count of enabled TCs for this PF */
4682 	if (pf->hw.func_caps.iscsi)
4683 		enabled_tc =  i40e_get_iscsi_tc_map(pf);
4684 	else
4685 		return 1; /* Only TC0 */
4686 
4687 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4688 		if (enabled_tc & BIT(i))
4689 			num_tc++;
4690 	}
4691 	return num_tc;
4692 }
4693 
4694 /**
4695  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4696  * @pf: PF being queried
4697  *
4698  * Return a bitmap for enabled traffic classes for this PF.
4699  **/
4700 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4701 {
4702 	/* If DCB is not enabled for this PF then just return default TC */
4703 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4704 		return I40E_DEFAULT_TRAFFIC_CLASS;
4705 
4706 	/* SFP mode we want PF to be enabled for all TCs */
4707 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4708 		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4709 
4710 	/* MFP enabled and iSCSI PF type */
4711 	if (pf->hw.func_caps.iscsi)
4712 		return i40e_get_iscsi_tc_map(pf);
4713 	else
4714 		return I40E_DEFAULT_TRAFFIC_CLASS;
4715 }
4716 
4717 /**
4718  * i40e_vsi_get_bw_info - Query VSI BW Information
4719  * @vsi: the VSI being queried
4720  *
4721  * Returns 0 on success, negative value on failure
4722  **/
4723 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4724 {
4725 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4726 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4727 	struct i40e_pf *pf = vsi->back;
4728 	struct i40e_hw *hw = &pf->hw;
4729 	i40e_status ret;
4730 	u32 tc_bw_max;
4731 	int i;
4732 
4733 	/* Get the VSI level BW configuration */
4734 	ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4735 	if (ret) {
4736 		dev_info(&pf->pdev->dev,
4737 			 "couldn't get PF vsi bw config, err %s aq_err %s\n",
4738 			 i40e_stat_str(&pf->hw, ret),
4739 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4740 		return -EINVAL;
4741 	}
4742 
4743 	/* Get the VSI level BW configuration per TC */
4744 	ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4745 					       NULL);
4746 	if (ret) {
4747 		dev_info(&pf->pdev->dev,
4748 			 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4749 			 i40e_stat_str(&pf->hw, ret),
4750 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4751 		return -EINVAL;
4752 	}
4753 
4754 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4755 		dev_info(&pf->pdev->dev,
4756 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4757 			 bw_config.tc_valid_bits,
4758 			 bw_ets_config.tc_valid_bits);
4759 		/* Still continuing */
4760 	}
4761 
4762 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4763 	vsi->bw_max_quanta = bw_config.max_bw;
4764 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4765 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4766 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4767 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4768 		vsi->bw_ets_limit_credits[i] =
4769 					le16_to_cpu(bw_ets_config.credits[i]);
4770 		/* 3 bits out of 4 for each TC */
4771 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4772 	}
4773 
4774 	return 0;
4775 }
4776 
4777 /**
4778  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4779  * @vsi: the VSI being configured
4780  * @enabled_tc: TC bitmap
4781  * @bw_credits: BW shared credits per TC
4782  *
4783  * Returns 0 on success, negative value on failure
4784  **/
4785 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4786 				       u8 *bw_share)
4787 {
4788 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4789 	i40e_status ret;
4790 	int i;
4791 
4792 	bw_data.tc_valid_bits = enabled_tc;
4793 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4794 		bw_data.tc_bw_credits[i] = bw_share[i];
4795 
4796 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4797 				       NULL);
4798 	if (ret) {
4799 		dev_info(&vsi->back->pdev->dev,
4800 			 "AQ command Config VSI BW allocation per TC failed = %d\n",
4801 			 vsi->back->hw.aq.asq_last_status);
4802 		return -EINVAL;
4803 	}
4804 
4805 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4806 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4807 
4808 	return 0;
4809 }
4810 
4811 /**
4812  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4813  * @vsi: the VSI being configured
4814  * @enabled_tc: TC map to be enabled
4815  *
4816  **/
4817 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4818 {
4819 	struct net_device *netdev = vsi->netdev;
4820 	struct i40e_pf *pf = vsi->back;
4821 	struct i40e_hw *hw = &pf->hw;
4822 	u8 netdev_tc = 0;
4823 	int i;
4824 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4825 
4826 	if (!netdev)
4827 		return;
4828 
4829 	if (!enabled_tc) {
4830 		netdev_reset_tc(netdev);
4831 		return;
4832 	}
4833 
4834 	/* Set up actual enabled TCs on the VSI */
4835 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4836 		return;
4837 
4838 	/* set per TC queues for the VSI */
4839 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4840 		/* Only set TC queues for enabled tcs
4841 		 *
4842 		 * e.g. For a VSI that has TC0 and TC3 enabled the
4843 		 * enabled_tc bitmap would be 0x00001001; the driver
4844 		 * will set the numtc for netdev as 2 that will be
4845 		 * referenced by the netdev layer as TC 0 and 1.
4846 		 */
4847 		if (vsi->tc_config.enabled_tc & BIT(i))
4848 			netdev_set_tc_queue(netdev,
4849 					vsi->tc_config.tc_info[i].netdev_tc,
4850 					vsi->tc_config.tc_info[i].qcount,
4851 					vsi->tc_config.tc_info[i].qoffset);
4852 	}
4853 
4854 	/* Assign UP2TC map for the VSI */
4855 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4856 		/* Get the actual TC# for the UP */
4857 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4858 		/* Get the mapped netdev TC# for the UP */
4859 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4860 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
4861 	}
4862 }
4863 
4864 /**
4865  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4866  * @vsi: the VSI being configured
4867  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4868  **/
4869 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4870 				      struct i40e_vsi_context *ctxt)
4871 {
4872 	/* copy just the sections touched not the entire info
4873 	 * since not all sections are valid as returned by
4874 	 * update vsi params
4875 	 */
4876 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
4877 	memcpy(&vsi->info.queue_mapping,
4878 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4879 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4880 	       sizeof(vsi->info.tc_mapping));
4881 }
4882 
4883 /**
4884  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4885  * @vsi: VSI to be configured
4886  * @enabled_tc: TC bitmap
4887  *
4888  * This configures a particular VSI for TCs that are mapped to the
4889  * given TC bitmap. It uses default bandwidth share for TCs across
4890  * VSIs to configure TC for a particular VSI.
4891  *
4892  * NOTE:
4893  * It is expected that the VSI queues have been quisced before calling
4894  * this function.
4895  **/
4896 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4897 {
4898 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4899 	struct i40e_vsi_context ctxt;
4900 	int ret = 0;
4901 	int i;
4902 
4903 	/* Check if enabled_tc is same as existing or new TCs */
4904 	if (vsi->tc_config.enabled_tc == enabled_tc)
4905 		return ret;
4906 
4907 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
4908 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4909 		if (enabled_tc & BIT(i))
4910 			bw_share[i] = 1;
4911 	}
4912 
4913 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4914 	if (ret) {
4915 		dev_info(&vsi->back->pdev->dev,
4916 			 "Failed configuring TC map %d for VSI %d\n",
4917 			 enabled_tc, vsi->seid);
4918 		goto out;
4919 	}
4920 
4921 	/* Update Queue Pairs Mapping for currently enabled UPs */
4922 	ctxt.seid = vsi->seid;
4923 	ctxt.pf_num = vsi->back->hw.pf_id;
4924 	ctxt.vf_num = 0;
4925 	ctxt.uplink_seid = vsi->uplink_seid;
4926 	ctxt.info = vsi->info;
4927 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4928 
4929 	if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
4930 		ctxt.info.valid_sections |=
4931 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
4932 		ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
4933 	}
4934 
4935 	/* Update the VSI after updating the VSI queue-mapping information */
4936 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4937 	if (ret) {
4938 		dev_info(&vsi->back->pdev->dev,
4939 			 "Update vsi tc config failed, err %s aq_err %s\n",
4940 			 i40e_stat_str(&vsi->back->hw, ret),
4941 			 i40e_aq_str(&vsi->back->hw,
4942 				     vsi->back->hw.aq.asq_last_status));
4943 		goto out;
4944 	}
4945 	/* update the local VSI info with updated queue map */
4946 	i40e_vsi_update_queue_map(vsi, &ctxt);
4947 	vsi->info.valid_sections = 0;
4948 
4949 	/* Update current VSI BW information */
4950 	ret = i40e_vsi_get_bw_info(vsi);
4951 	if (ret) {
4952 		dev_info(&vsi->back->pdev->dev,
4953 			 "Failed updating vsi bw info, err %s aq_err %s\n",
4954 			 i40e_stat_str(&vsi->back->hw, ret),
4955 			 i40e_aq_str(&vsi->back->hw,
4956 				     vsi->back->hw.aq.asq_last_status));
4957 		goto out;
4958 	}
4959 
4960 	/* Update the netdev TC setup */
4961 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4962 out:
4963 	return ret;
4964 }
4965 
4966 /**
4967  * i40e_veb_config_tc - Configure TCs for given VEB
4968  * @veb: given VEB
4969  * @enabled_tc: TC bitmap
4970  *
4971  * Configures given TC bitmap for VEB (switching) element
4972  **/
4973 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4974 {
4975 	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4976 	struct i40e_pf *pf = veb->pf;
4977 	int ret = 0;
4978 	int i;
4979 
4980 	/* No TCs or already enabled TCs just return */
4981 	if (!enabled_tc || veb->enabled_tc == enabled_tc)
4982 		return ret;
4983 
4984 	bw_data.tc_valid_bits = enabled_tc;
4985 	/* bw_data.absolute_credits is not set (relative) */
4986 
4987 	/* Enable ETS TCs with equal BW Share for now */
4988 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4989 		if (enabled_tc & BIT(i))
4990 			bw_data.tc_bw_share_credits[i] = 1;
4991 	}
4992 
4993 	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4994 						   &bw_data, NULL);
4995 	if (ret) {
4996 		dev_info(&pf->pdev->dev,
4997 			 "VEB bw config failed, err %s aq_err %s\n",
4998 			 i40e_stat_str(&pf->hw, ret),
4999 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5000 		goto out;
5001 	}
5002 
5003 	/* Update the BW information */
5004 	ret = i40e_veb_get_bw_info(veb);
5005 	if (ret) {
5006 		dev_info(&pf->pdev->dev,
5007 			 "Failed getting veb bw config, err %s aq_err %s\n",
5008 			 i40e_stat_str(&pf->hw, ret),
5009 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5010 	}
5011 
5012 out:
5013 	return ret;
5014 }
5015 
5016 #ifdef CONFIG_I40E_DCB
5017 /**
5018  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
5019  * @pf: PF struct
5020  *
5021  * Reconfigure VEB/VSIs on a given PF; it is assumed that
5022  * the caller would've quiesce all the VSIs before calling
5023  * this function
5024  **/
5025 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
5026 {
5027 	u8 tc_map = 0;
5028 	int ret;
5029 	u8 v;
5030 
5031 	/* Enable the TCs available on PF to all VEBs */
5032 	tc_map = i40e_pf_get_tc_map(pf);
5033 	for (v = 0; v < I40E_MAX_VEB; v++) {
5034 		if (!pf->veb[v])
5035 			continue;
5036 		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
5037 		if (ret) {
5038 			dev_info(&pf->pdev->dev,
5039 				 "Failed configuring TC for VEB seid=%d\n",
5040 				 pf->veb[v]->seid);
5041 			/* Will try to configure as many components */
5042 		}
5043 	}
5044 
5045 	/* Update each VSI */
5046 	for (v = 0; v < pf->num_alloc_vsi; v++) {
5047 		if (!pf->vsi[v])
5048 			continue;
5049 
5050 		/* - Enable all TCs for the LAN VSI
5051 #ifdef I40E_FCOE
5052 		 * - For FCoE VSI only enable the TC configured
5053 		 *   as per the APP TLV
5054 #endif
5055 		 * - For all others keep them at TC0 for now
5056 		 */
5057 		if (v == pf->lan_vsi)
5058 			tc_map = i40e_pf_get_tc_map(pf);
5059 		else
5060 			tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
5061 #ifdef I40E_FCOE
5062 		if (pf->vsi[v]->type == I40E_VSI_FCOE)
5063 			tc_map = i40e_get_fcoe_tc_map(pf);
5064 #endif /* #ifdef I40E_FCOE */
5065 
5066 		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5067 		if (ret) {
5068 			dev_info(&pf->pdev->dev,
5069 				 "Failed configuring TC for VSI seid=%d\n",
5070 				 pf->vsi[v]->seid);
5071 			/* Will try to configure as many components */
5072 		} else {
5073 			/* Re-configure VSI vectors based on updated TC map */
5074 			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5075 			if (pf->vsi[v]->netdev)
5076 				i40e_dcbnl_set_all(pf->vsi[v]);
5077 		}
5078 	}
5079 }
5080 
5081 /**
5082  * i40e_resume_port_tx - Resume port Tx
5083  * @pf: PF struct
5084  *
5085  * Resume a port's Tx and issue a PF reset in case of failure to
5086  * resume.
5087  **/
5088 static int i40e_resume_port_tx(struct i40e_pf *pf)
5089 {
5090 	struct i40e_hw *hw = &pf->hw;
5091 	int ret;
5092 
5093 	ret = i40e_aq_resume_port_tx(hw, NULL);
5094 	if (ret) {
5095 		dev_info(&pf->pdev->dev,
5096 			 "Resume Port Tx failed, err %s aq_err %s\n",
5097 			  i40e_stat_str(&pf->hw, ret),
5098 			  i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5099 		/* Schedule PF reset to recover */
5100 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5101 		i40e_service_event_schedule(pf);
5102 	}
5103 
5104 	return ret;
5105 }
5106 
5107 /**
5108  * i40e_init_pf_dcb - Initialize DCB configuration
5109  * @pf: PF being configured
5110  *
5111  * Query the current DCB configuration and cache it
5112  * in the hardware structure
5113  **/
5114 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5115 {
5116 	struct i40e_hw *hw = &pf->hw;
5117 	int err = 0;
5118 
5119 	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5120 	if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5121 		goto out;
5122 
5123 	/* Get the initial DCB configuration */
5124 	err = i40e_init_dcb(hw);
5125 	if (!err) {
5126 		/* Device/Function is not DCBX capable */
5127 		if ((!hw->func_caps.dcb) ||
5128 		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5129 			dev_info(&pf->pdev->dev,
5130 				 "DCBX offload is not supported or is disabled for this PF.\n");
5131 
5132 			if (pf->flags & I40E_FLAG_MFP_ENABLED)
5133 				goto out;
5134 
5135 		} else {
5136 			/* When status is not DISABLED then DCBX in FW */
5137 			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5138 				       DCB_CAP_DCBX_VER_IEEE;
5139 
5140 			pf->flags |= I40E_FLAG_DCB_CAPABLE;
5141 			/* Enable DCB tagging only when more than one TC
5142 			 * or explicitly disable if only one TC
5143 			 */
5144 			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5145 				pf->flags |= I40E_FLAG_DCB_ENABLED;
5146 			else
5147 				pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5148 			dev_dbg(&pf->pdev->dev,
5149 				"DCBX offload is supported for this PF.\n");
5150 		}
5151 	} else {
5152 		dev_info(&pf->pdev->dev,
5153 			 "Query for DCB configuration failed, err %s aq_err %s\n",
5154 			 i40e_stat_str(&pf->hw, err),
5155 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5156 	}
5157 
5158 out:
5159 	return err;
5160 }
5161 #endif /* CONFIG_I40E_DCB */
5162 #define SPEED_SIZE 14
5163 #define FC_SIZE 8
5164 /**
5165  * i40e_print_link_message - print link up or down
5166  * @vsi: the VSI for which link needs a message
5167  */
5168 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5169 {
5170 	char *speed = "Unknown";
5171 	char *fc = "Unknown";
5172 
5173 	if (vsi->current_isup == isup)
5174 		return;
5175 	vsi->current_isup = isup;
5176 	if (!isup) {
5177 		netdev_info(vsi->netdev, "NIC Link is Down\n");
5178 		return;
5179 	}
5180 
5181 	/* Warn user if link speed on NPAR enabled partition is not at
5182 	 * least 10GB
5183 	 */
5184 	if (vsi->back->hw.func_caps.npar_enable &&
5185 	    (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5186 	     vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5187 		netdev_warn(vsi->netdev,
5188 			    "The partition detected link speed that is less than 10Gbps\n");
5189 
5190 	switch (vsi->back->hw.phy.link_info.link_speed) {
5191 	case I40E_LINK_SPEED_40GB:
5192 		speed = "40 G";
5193 		break;
5194 	case I40E_LINK_SPEED_20GB:
5195 		speed = "20 G";
5196 		break;
5197 	case I40E_LINK_SPEED_10GB:
5198 		speed = "10 G";
5199 		break;
5200 	case I40E_LINK_SPEED_1GB:
5201 		speed = "1000 M";
5202 		break;
5203 	case I40E_LINK_SPEED_100MB:
5204 		speed = "100 M";
5205 		break;
5206 	default:
5207 		break;
5208 	}
5209 
5210 	switch (vsi->back->hw.fc.current_mode) {
5211 	case I40E_FC_FULL:
5212 		fc = "RX/TX";
5213 		break;
5214 	case I40E_FC_TX_PAUSE:
5215 		fc = "TX";
5216 		break;
5217 	case I40E_FC_RX_PAUSE:
5218 		fc = "RX";
5219 		break;
5220 	default:
5221 		fc = "None";
5222 		break;
5223 	}
5224 
5225 	netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5226 		    speed, fc);
5227 }
5228 
5229 /**
5230  * i40e_up_complete - Finish the last steps of bringing up a connection
5231  * @vsi: the VSI being configured
5232  **/
5233 static int i40e_up_complete(struct i40e_vsi *vsi)
5234 {
5235 	struct i40e_pf *pf = vsi->back;
5236 	int err;
5237 
5238 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5239 		i40e_vsi_configure_msix(vsi);
5240 	else
5241 		i40e_configure_msi_and_legacy(vsi);
5242 
5243 	/* start rings */
5244 	err = i40e_vsi_start_rings(vsi);
5245 	if (err)
5246 		return err;
5247 
5248 	clear_bit(__I40E_DOWN, &vsi->state);
5249 	i40e_napi_enable_all(vsi);
5250 	i40e_vsi_enable_irq(vsi);
5251 
5252 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5253 	    (vsi->netdev)) {
5254 		i40e_print_link_message(vsi, true);
5255 		netif_tx_start_all_queues(vsi->netdev);
5256 		netif_carrier_on(vsi->netdev);
5257 	} else if (vsi->netdev) {
5258 		i40e_print_link_message(vsi, false);
5259 		/* need to check for qualified module here*/
5260 		if ((pf->hw.phy.link_info.link_info &
5261 			I40E_AQ_MEDIA_AVAILABLE) &&
5262 		    (!(pf->hw.phy.link_info.an_info &
5263 			I40E_AQ_QUALIFIED_MODULE)))
5264 			netdev_err(vsi->netdev,
5265 				   "the driver failed to link because an unqualified module was detected.");
5266 	}
5267 
5268 	/* replay FDIR SB filters */
5269 	if (vsi->type == I40E_VSI_FDIR) {
5270 		/* reset fd counters */
5271 		pf->fd_add_err = pf->fd_atr_cnt = 0;
5272 		if (pf->fd_tcp_rule > 0) {
5273 			pf->auto_disable_flags |= I40E_FLAG_FD_ATR_ENABLED;
5274 			if (I40E_DEBUG_FD & pf->hw.debug_mask)
5275 				dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5276 			pf->fd_tcp_rule = 0;
5277 		}
5278 		i40e_fdir_filter_restore(vsi);
5279 	}
5280 
5281 	/* On the next run of the service_task, notify any clients of the new
5282 	 * opened netdev
5283 	 */
5284 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5285 	i40e_service_event_schedule(pf);
5286 
5287 	return 0;
5288 }
5289 
5290 /**
5291  * i40e_vsi_reinit_locked - Reset the VSI
5292  * @vsi: the VSI being configured
5293  *
5294  * Rebuild the ring structs after some configuration
5295  * has changed, e.g. MTU size.
5296  **/
5297 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5298 {
5299 	struct i40e_pf *pf = vsi->back;
5300 
5301 	WARN_ON(in_interrupt());
5302 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5303 		usleep_range(1000, 2000);
5304 	i40e_down(vsi);
5305 
5306 	i40e_up(vsi);
5307 	clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5308 }
5309 
5310 /**
5311  * i40e_up - Bring the connection back up after being down
5312  * @vsi: the VSI being configured
5313  **/
5314 int i40e_up(struct i40e_vsi *vsi)
5315 {
5316 	int err;
5317 
5318 	err = i40e_vsi_configure(vsi);
5319 	if (!err)
5320 		err = i40e_up_complete(vsi);
5321 
5322 	return err;
5323 }
5324 
5325 /**
5326  * i40e_down - Shutdown the connection processing
5327  * @vsi: the VSI being stopped
5328  **/
5329 void i40e_down(struct i40e_vsi *vsi)
5330 {
5331 	int i;
5332 
5333 	/* It is assumed that the caller of this function
5334 	 * sets the vsi->state __I40E_DOWN bit.
5335 	 */
5336 	if (vsi->netdev) {
5337 		netif_carrier_off(vsi->netdev);
5338 		netif_tx_disable(vsi->netdev);
5339 	}
5340 	i40e_vsi_disable_irq(vsi);
5341 	i40e_vsi_stop_rings(vsi);
5342 	i40e_napi_disable_all(vsi);
5343 
5344 	for (i = 0; i < vsi->num_queue_pairs; i++) {
5345 		i40e_clean_tx_ring(vsi->tx_rings[i]);
5346 		i40e_clean_rx_ring(vsi->rx_rings[i]);
5347 	}
5348 
5349 	i40e_notify_client_of_netdev_close(vsi, false);
5350 
5351 }
5352 
5353 /**
5354  * i40e_setup_tc - configure multiple traffic classes
5355  * @netdev: net device to configure
5356  * @tc: number of traffic classes to enable
5357  **/
5358 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5359 {
5360 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5361 	struct i40e_vsi *vsi = np->vsi;
5362 	struct i40e_pf *pf = vsi->back;
5363 	u8 enabled_tc = 0;
5364 	int ret = -EINVAL;
5365 	int i;
5366 
5367 	/* Check if DCB enabled to continue */
5368 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5369 		netdev_info(netdev, "DCB is not enabled for adapter\n");
5370 		goto exit;
5371 	}
5372 
5373 	/* Check if MFP enabled */
5374 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5375 		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5376 		goto exit;
5377 	}
5378 
5379 	/* Check whether tc count is within enabled limit */
5380 	if (tc > i40e_pf_get_num_tc(pf)) {
5381 		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5382 		goto exit;
5383 	}
5384 
5385 	/* Generate TC map for number of tc requested */
5386 	for (i = 0; i < tc; i++)
5387 		enabled_tc |= BIT(i);
5388 
5389 	/* Requesting same TC configuration as already enabled */
5390 	if (enabled_tc == vsi->tc_config.enabled_tc)
5391 		return 0;
5392 
5393 	/* Quiesce VSI queues */
5394 	i40e_quiesce_vsi(vsi);
5395 
5396 	/* Configure VSI for enabled TCs */
5397 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
5398 	if (ret) {
5399 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5400 			    vsi->seid);
5401 		goto exit;
5402 	}
5403 
5404 	/* Unquiesce VSI */
5405 	i40e_unquiesce_vsi(vsi);
5406 
5407 exit:
5408 	return ret;
5409 }
5410 
5411 #ifdef I40E_FCOE
5412 int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5413 		    struct tc_to_netdev *tc)
5414 #else
5415 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5416 			   struct tc_to_netdev *tc)
5417 #endif
5418 {
5419 	if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)
5420 		return -EINVAL;
5421 	return i40e_setup_tc(netdev, tc->tc);
5422 }
5423 
5424 /**
5425  * i40e_open - Called when a network interface is made active
5426  * @netdev: network interface device structure
5427  *
5428  * The open entry point is called when a network interface is made
5429  * active by the system (IFF_UP).  At this point all resources needed
5430  * for transmit and receive operations are allocated, the interrupt
5431  * handler is registered with the OS, the netdev watchdog subtask is
5432  * enabled, and the stack is notified that the interface is ready.
5433  *
5434  * Returns 0 on success, negative value on failure
5435  **/
5436 int i40e_open(struct net_device *netdev)
5437 {
5438 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5439 	struct i40e_vsi *vsi = np->vsi;
5440 	struct i40e_pf *pf = vsi->back;
5441 	int err;
5442 
5443 	/* disallow open during test or if eeprom is broken */
5444 	if (test_bit(__I40E_TESTING, &pf->state) ||
5445 	    test_bit(__I40E_BAD_EEPROM, &pf->state))
5446 		return -EBUSY;
5447 
5448 	netif_carrier_off(netdev);
5449 
5450 	err = i40e_vsi_open(vsi);
5451 	if (err)
5452 		return err;
5453 
5454 	/* configure global TSO hardware offload settings */
5455 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5456 						       TCP_FLAG_FIN) >> 16);
5457 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5458 						       TCP_FLAG_FIN |
5459 						       TCP_FLAG_CWR) >> 16);
5460 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5461 
5462 	udp_tunnel_get_rx_info(netdev);
5463 
5464 	return 0;
5465 }
5466 
5467 /**
5468  * i40e_vsi_open -
5469  * @vsi: the VSI to open
5470  *
5471  * Finish initialization of the VSI.
5472  *
5473  * Returns 0 on success, negative value on failure
5474  **/
5475 int i40e_vsi_open(struct i40e_vsi *vsi)
5476 {
5477 	struct i40e_pf *pf = vsi->back;
5478 	char int_name[I40E_INT_NAME_STR_LEN];
5479 	int err;
5480 
5481 	/* allocate descriptors */
5482 	err = i40e_vsi_setup_tx_resources(vsi);
5483 	if (err)
5484 		goto err_setup_tx;
5485 	err = i40e_vsi_setup_rx_resources(vsi);
5486 	if (err)
5487 		goto err_setup_rx;
5488 
5489 	err = i40e_vsi_configure(vsi);
5490 	if (err)
5491 		goto err_setup_rx;
5492 
5493 	if (vsi->netdev) {
5494 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5495 			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5496 		err = i40e_vsi_request_irq(vsi, int_name);
5497 		if (err)
5498 			goto err_setup_rx;
5499 
5500 		/* Notify the stack of the actual queue counts. */
5501 		err = netif_set_real_num_tx_queues(vsi->netdev,
5502 						   vsi->num_queue_pairs);
5503 		if (err)
5504 			goto err_set_queues;
5505 
5506 		err = netif_set_real_num_rx_queues(vsi->netdev,
5507 						   vsi->num_queue_pairs);
5508 		if (err)
5509 			goto err_set_queues;
5510 
5511 	} else if (vsi->type == I40E_VSI_FDIR) {
5512 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5513 			 dev_driver_string(&pf->pdev->dev),
5514 			 dev_name(&pf->pdev->dev));
5515 		err = i40e_vsi_request_irq(vsi, int_name);
5516 
5517 	} else {
5518 		err = -EINVAL;
5519 		goto err_setup_rx;
5520 	}
5521 
5522 	err = i40e_up_complete(vsi);
5523 	if (err)
5524 		goto err_up_complete;
5525 
5526 	return 0;
5527 
5528 err_up_complete:
5529 	i40e_down(vsi);
5530 err_set_queues:
5531 	i40e_vsi_free_irq(vsi);
5532 err_setup_rx:
5533 	i40e_vsi_free_rx_resources(vsi);
5534 err_setup_tx:
5535 	i40e_vsi_free_tx_resources(vsi);
5536 	if (vsi == pf->vsi[pf->lan_vsi])
5537 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5538 
5539 	return err;
5540 }
5541 
5542 /**
5543  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5544  * @pf: Pointer to PF
5545  *
5546  * This function destroys the hlist where all the Flow Director
5547  * filters were saved.
5548  **/
5549 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5550 {
5551 	struct i40e_fdir_filter *filter;
5552 	struct hlist_node *node2;
5553 
5554 	hlist_for_each_entry_safe(filter, node2,
5555 				  &pf->fdir_filter_list, fdir_node) {
5556 		hlist_del(&filter->fdir_node);
5557 		kfree(filter);
5558 	}
5559 	pf->fdir_pf_active_filters = 0;
5560 }
5561 
5562 /**
5563  * i40e_close - Disables a network interface
5564  * @netdev: network interface device structure
5565  *
5566  * The close entry point is called when an interface is de-activated
5567  * by the OS.  The hardware is still under the driver's control, but
5568  * this netdev interface is disabled.
5569  *
5570  * Returns 0, this is not allowed to fail
5571  **/
5572 int i40e_close(struct net_device *netdev)
5573 {
5574 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5575 	struct i40e_vsi *vsi = np->vsi;
5576 
5577 	i40e_vsi_close(vsi);
5578 
5579 	return 0;
5580 }
5581 
5582 /**
5583  * i40e_do_reset - Start a PF or Core Reset sequence
5584  * @pf: board private structure
5585  * @reset_flags: which reset is requested
5586  *
5587  * The essential difference in resets is that the PF Reset
5588  * doesn't clear the packet buffers, doesn't reset the PE
5589  * firmware, and doesn't bother the other PFs on the chip.
5590  **/
5591 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5592 {
5593 	u32 val;
5594 
5595 	WARN_ON(in_interrupt());
5596 
5597 
5598 	/* do the biggest reset indicated */
5599 	if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5600 
5601 		/* Request a Global Reset
5602 		 *
5603 		 * This will start the chip's countdown to the actual full
5604 		 * chip reset event, and a warning interrupt to be sent
5605 		 * to all PFs, including the requestor.  Our handler
5606 		 * for the warning interrupt will deal with the shutdown
5607 		 * and recovery of the switch setup.
5608 		 */
5609 		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5610 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5611 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5612 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5613 
5614 	} else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5615 
5616 		/* Request a Core Reset
5617 		 *
5618 		 * Same as Global Reset, except does *not* include the MAC/PHY
5619 		 */
5620 		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5621 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5622 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
5623 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5624 		i40e_flush(&pf->hw);
5625 
5626 	} else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5627 
5628 		/* Request a PF Reset
5629 		 *
5630 		 * Resets only the PF-specific registers
5631 		 *
5632 		 * This goes directly to the tear-down and rebuild of
5633 		 * the switch, since we need to do all the recovery as
5634 		 * for the Core Reset.
5635 		 */
5636 		dev_dbg(&pf->pdev->dev, "PFR requested\n");
5637 		i40e_handle_reset_warning(pf);
5638 
5639 	} else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5640 		int v;
5641 
5642 		/* Find the VSI(s) that requested a re-init */
5643 		dev_info(&pf->pdev->dev,
5644 			 "VSI reinit requested\n");
5645 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5646 			struct i40e_vsi *vsi = pf->vsi[v];
5647 
5648 			if (vsi != NULL &&
5649 			    test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5650 				i40e_vsi_reinit_locked(pf->vsi[v]);
5651 				clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5652 			}
5653 		}
5654 	} else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5655 		int v;
5656 
5657 		/* Find the VSI(s) that needs to be brought down */
5658 		dev_info(&pf->pdev->dev, "VSI down requested\n");
5659 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5660 			struct i40e_vsi *vsi = pf->vsi[v];
5661 
5662 			if (vsi != NULL &&
5663 			    test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5664 				set_bit(__I40E_DOWN, &vsi->state);
5665 				i40e_down(vsi);
5666 				clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5667 			}
5668 		}
5669 	} else {
5670 		dev_info(&pf->pdev->dev,
5671 			 "bad reset request 0x%08x\n", reset_flags);
5672 	}
5673 }
5674 
5675 #ifdef CONFIG_I40E_DCB
5676 /**
5677  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5678  * @pf: board private structure
5679  * @old_cfg: current DCB config
5680  * @new_cfg: new DCB config
5681  **/
5682 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5683 			    struct i40e_dcbx_config *old_cfg,
5684 			    struct i40e_dcbx_config *new_cfg)
5685 {
5686 	bool need_reconfig = false;
5687 
5688 	/* Check if ETS configuration has changed */
5689 	if (memcmp(&new_cfg->etscfg,
5690 		   &old_cfg->etscfg,
5691 		   sizeof(new_cfg->etscfg))) {
5692 		/* If Priority Table has changed reconfig is needed */
5693 		if (memcmp(&new_cfg->etscfg.prioritytable,
5694 			   &old_cfg->etscfg.prioritytable,
5695 			   sizeof(new_cfg->etscfg.prioritytable))) {
5696 			need_reconfig = true;
5697 			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5698 		}
5699 
5700 		if (memcmp(&new_cfg->etscfg.tcbwtable,
5701 			   &old_cfg->etscfg.tcbwtable,
5702 			   sizeof(new_cfg->etscfg.tcbwtable)))
5703 			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5704 
5705 		if (memcmp(&new_cfg->etscfg.tsatable,
5706 			   &old_cfg->etscfg.tsatable,
5707 			   sizeof(new_cfg->etscfg.tsatable)))
5708 			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5709 	}
5710 
5711 	/* Check if PFC configuration has changed */
5712 	if (memcmp(&new_cfg->pfc,
5713 		   &old_cfg->pfc,
5714 		   sizeof(new_cfg->pfc))) {
5715 		need_reconfig = true;
5716 		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5717 	}
5718 
5719 	/* Check if APP Table has changed */
5720 	if (memcmp(&new_cfg->app,
5721 		   &old_cfg->app,
5722 		   sizeof(new_cfg->app))) {
5723 		need_reconfig = true;
5724 		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5725 	}
5726 
5727 	dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5728 	return need_reconfig;
5729 }
5730 
5731 /**
5732  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5733  * @pf: board private structure
5734  * @e: event info posted on ARQ
5735  **/
5736 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5737 				  struct i40e_arq_event_info *e)
5738 {
5739 	struct i40e_aqc_lldp_get_mib *mib =
5740 		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5741 	struct i40e_hw *hw = &pf->hw;
5742 	struct i40e_dcbx_config tmp_dcbx_cfg;
5743 	bool need_reconfig = false;
5744 	int ret = 0;
5745 	u8 type;
5746 
5747 	/* Not DCB capable or capability disabled */
5748 	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5749 		return ret;
5750 
5751 	/* Ignore if event is not for Nearest Bridge */
5752 	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5753 		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5754 	dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5755 	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5756 		return ret;
5757 
5758 	/* Check MIB Type and return if event for Remote MIB update */
5759 	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5760 	dev_dbg(&pf->pdev->dev,
5761 		"LLDP event mib type %s\n", type ? "remote" : "local");
5762 	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5763 		/* Update the remote cached instance and return */
5764 		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5765 				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5766 				&hw->remote_dcbx_config);
5767 		goto exit;
5768 	}
5769 
5770 	/* Store the old configuration */
5771 	tmp_dcbx_cfg = hw->local_dcbx_config;
5772 
5773 	/* Reset the old DCBx configuration data */
5774 	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5775 	/* Get updated DCBX data from firmware */
5776 	ret = i40e_get_dcb_config(&pf->hw);
5777 	if (ret) {
5778 		dev_info(&pf->pdev->dev,
5779 			 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5780 			 i40e_stat_str(&pf->hw, ret),
5781 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5782 		goto exit;
5783 	}
5784 
5785 	/* No change detected in DCBX configs */
5786 	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5787 		    sizeof(tmp_dcbx_cfg))) {
5788 		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5789 		goto exit;
5790 	}
5791 
5792 	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5793 					       &hw->local_dcbx_config);
5794 
5795 	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5796 
5797 	if (!need_reconfig)
5798 		goto exit;
5799 
5800 	/* Enable DCB tagging only when more than one TC */
5801 	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5802 		pf->flags |= I40E_FLAG_DCB_ENABLED;
5803 	else
5804 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5805 
5806 	set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5807 	/* Reconfiguration needed quiesce all VSIs */
5808 	i40e_pf_quiesce_all_vsi(pf);
5809 
5810 	/* Changes in configuration update VEB/VSI */
5811 	i40e_dcb_reconfigure(pf);
5812 
5813 	ret = i40e_resume_port_tx(pf);
5814 
5815 	clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5816 	/* In case of error no point in resuming VSIs */
5817 	if (ret)
5818 		goto exit;
5819 
5820 	/* Wait for the PF's queues to be disabled */
5821 	ret = i40e_pf_wait_queues_disabled(pf);
5822 	if (ret) {
5823 		/* Schedule PF reset to recover */
5824 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5825 		i40e_service_event_schedule(pf);
5826 	} else {
5827 		i40e_pf_unquiesce_all_vsi(pf);
5828 		/* Notify the client for the DCB changes */
5829 		i40e_notify_client_of_l2_param_changes(pf->vsi[pf->lan_vsi]);
5830 	}
5831 
5832 exit:
5833 	return ret;
5834 }
5835 #endif /* CONFIG_I40E_DCB */
5836 
5837 /**
5838  * i40e_do_reset_safe - Protected reset path for userland calls.
5839  * @pf: board private structure
5840  * @reset_flags: which reset is requested
5841  *
5842  **/
5843 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5844 {
5845 	rtnl_lock();
5846 	i40e_do_reset(pf, reset_flags);
5847 	rtnl_unlock();
5848 }
5849 
5850 /**
5851  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5852  * @pf: board private structure
5853  * @e: event info posted on ARQ
5854  *
5855  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5856  * and VF queues
5857  **/
5858 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5859 					   struct i40e_arq_event_info *e)
5860 {
5861 	struct i40e_aqc_lan_overflow *data =
5862 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5863 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
5864 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5865 	struct i40e_hw *hw = &pf->hw;
5866 	struct i40e_vf *vf;
5867 	u16 vf_id;
5868 
5869 	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5870 		queue, qtx_ctl);
5871 
5872 	/* Queue belongs to VF, find the VF and issue VF reset */
5873 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5874 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5875 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5876 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5877 		vf_id -= hw->func_caps.vf_base_id;
5878 		vf = &pf->vf[vf_id];
5879 		i40e_vc_notify_vf_reset(vf);
5880 		/* Allow VF to process pending reset notification */
5881 		msleep(20);
5882 		i40e_reset_vf(vf, false);
5883 	}
5884 }
5885 
5886 /**
5887  * i40e_service_event_complete - Finish up the service event
5888  * @pf: board private structure
5889  **/
5890 static void i40e_service_event_complete(struct i40e_pf *pf)
5891 {
5892 	WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5893 
5894 	/* flush memory to make sure state is correct before next watchog */
5895 	smp_mb__before_atomic();
5896 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5897 }
5898 
5899 /**
5900  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5901  * @pf: board private structure
5902  **/
5903 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5904 {
5905 	u32 val, fcnt_prog;
5906 
5907 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5908 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5909 	return fcnt_prog;
5910 }
5911 
5912 /**
5913  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5914  * @pf: board private structure
5915  **/
5916 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5917 {
5918 	u32 val, fcnt_prog;
5919 
5920 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5921 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5922 		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5923 		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5924 	return fcnt_prog;
5925 }
5926 
5927 /**
5928  * i40e_get_global_fd_count - Get total FD filters programmed on device
5929  * @pf: board private structure
5930  **/
5931 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5932 {
5933 	u32 val, fcnt_prog;
5934 
5935 	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5936 	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5937 		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5938 		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5939 	return fcnt_prog;
5940 }
5941 
5942 /**
5943  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5944  * @pf: board private structure
5945  **/
5946 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5947 {
5948 	struct i40e_fdir_filter *filter;
5949 	u32 fcnt_prog, fcnt_avail;
5950 	struct hlist_node *node;
5951 
5952 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5953 		return;
5954 
5955 	/* Check if, FD SB or ATR was auto disabled and if there is enough room
5956 	 * to re-enable
5957 	 */
5958 	fcnt_prog = i40e_get_global_fd_count(pf);
5959 	fcnt_avail = pf->fdir_pf_filter_count;
5960 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5961 	    (pf->fd_add_err == 0) ||
5962 	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5963 		if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5964 		    (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5965 			pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5966 			if (I40E_DEBUG_FD & pf->hw.debug_mask)
5967 				dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5968 		}
5969 	}
5970 
5971 	/* Wait for some more space to be available to turn on ATR. We also
5972 	 * must check that no existing ntuple rules for TCP are in effect
5973 	 */
5974 	if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5975 		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5976 		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED) &&
5977 		    (pf->fd_tcp_rule == 0)) {
5978 			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5979 			if (I40E_DEBUG_FD & pf->hw.debug_mask)
5980 				dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
5981 		}
5982 	}
5983 
5984 	/* if hw had a problem adding a filter, delete it */
5985 	if (pf->fd_inv > 0) {
5986 		hlist_for_each_entry_safe(filter, node,
5987 					  &pf->fdir_filter_list, fdir_node) {
5988 			if (filter->fd_id == pf->fd_inv) {
5989 				hlist_del(&filter->fdir_node);
5990 				kfree(filter);
5991 				pf->fdir_pf_active_filters--;
5992 			}
5993 		}
5994 	}
5995 }
5996 
5997 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5998 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5999 /**
6000  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
6001  * @pf: board private structure
6002  **/
6003 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
6004 {
6005 	unsigned long min_flush_time;
6006 	int flush_wait_retry = 50;
6007 	bool disable_atr = false;
6008 	int fd_room;
6009 	int reg;
6010 
6011 	if (!time_after(jiffies, pf->fd_flush_timestamp +
6012 				 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
6013 		return;
6014 
6015 	/* If the flush is happening too quick and we have mostly SB rules we
6016 	 * should not re-enable ATR for some time.
6017 	 */
6018 	min_flush_time = pf->fd_flush_timestamp +
6019 			 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
6020 	fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
6021 
6022 	if (!(time_after(jiffies, min_flush_time)) &&
6023 	    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
6024 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
6025 			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
6026 		disable_atr = true;
6027 	}
6028 
6029 	pf->fd_flush_timestamp = jiffies;
6030 	pf->auto_disable_flags |= I40E_FLAG_FD_ATR_ENABLED;
6031 	/* flush all filters */
6032 	wr32(&pf->hw, I40E_PFQF_CTL_1,
6033 	     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6034 	i40e_flush(&pf->hw);
6035 	pf->fd_flush_cnt++;
6036 	pf->fd_add_err = 0;
6037 	do {
6038 		/* Check FD flush status every 5-6msec */
6039 		usleep_range(5000, 6000);
6040 		reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6041 		if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6042 			break;
6043 	} while (flush_wait_retry--);
6044 	if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6045 		dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6046 	} else {
6047 		/* replay sideband filters */
6048 		i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6049 		if (!disable_atr)
6050 			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
6051 		clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
6052 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
6053 			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6054 	}
6055 }
6056 
6057 /**
6058  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6059  * @pf: board private structure
6060  **/
6061 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6062 {
6063 	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6064 }
6065 
6066 /* We can see up to 256 filter programming desc in transit if the filters are
6067  * being applied really fast; before we see the first
6068  * filter miss error on Rx queue 0. Accumulating enough error messages before
6069  * reacting will make sure we don't cause flush too often.
6070  */
6071 #define I40E_MAX_FD_PROGRAM_ERROR 256
6072 
6073 /**
6074  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6075  * @pf: board private structure
6076  **/
6077 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6078 {
6079 
6080 	/* if interface is down do nothing */
6081 	if (test_bit(__I40E_DOWN, &pf->state))
6082 		return;
6083 
6084 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6085 		i40e_fdir_flush_and_replay(pf);
6086 
6087 	i40e_fdir_check_and_reenable(pf);
6088 
6089 }
6090 
6091 /**
6092  * i40e_vsi_link_event - notify VSI of a link event
6093  * @vsi: vsi to be notified
6094  * @link_up: link up or down
6095  **/
6096 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6097 {
6098 	if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6099 		return;
6100 
6101 	switch (vsi->type) {
6102 	case I40E_VSI_MAIN:
6103 #ifdef I40E_FCOE
6104 	case I40E_VSI_FCOE:
6105 #endif
6106 		if (!vsi->netdev || !vsi->netdev_registered)
6107 			break;
6108 
6109 		if (link_up) {
6110 			netif_carrier_on(vsi->netdev);
6111 			netif_tx_wake_all_queues(vsi->netdev);
6112 		} else {
6113 			netif_carrier_off(vsi->netdev);
6114 			netif_tx_stop_all_queues(vsi->netdev);
6115 		}
6116 		break;
6117 
6118 	case I40E_VSI_SRIOV:
6119 	case I40E_VSI_VMDQ2:
6120 	case I40E_VSI_CTRL:
6121 	case I40E_VSI_IWARP:
6122 	case I40E_VSI_MIRROR:
6123 	default:
6124 		/* there is no notification for other VSIs */
6125 		break;
6126 	}
6127 }
6128 
6129 /**
6130  * i40e_veb_link_event - notify elements on the veb of a link event
6131  * @veb: veb to be notified
6132  * @link_up: link up or down
6133  **/
6134 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6135 {
6136 	struct i40e_pf *pf;
6137 	int i;
6138 
6139 	if (!veb || !veb->pf)
6140 		return;
6141 	pf = veb->pf;
6142 
6143 	/* depth first... */
6144 	for (i = 0; i < I40E_MAX_VEB; i++)
6145 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6146 			i40e_veb_link_event(pf->veb[i], link_up);
6147 
6148 	/* ... now the local VSIs */
6149 	for (i = 0; i < pf->num_alloc_vsi; i++)
6150 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6151 			i40e_vsi_link_event(pf->vsi[i], link_up);
6152 }
6153 
6154 /**
6155  * i40e_link_event - Update netif_carrier status
6156  * @pf: board private structure
6157  **/
6158 static void i40e_link_event(struct i40e_pf *pf)
6159 {
6160 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6161 	u8 new_link_speed, old_link_speed;
6162 	i40e_status status;
6163 	bool new_link, old_link;
6164 
6165 	/* save off old link status information */
6166 	pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6167 
6168 	/* set this to force the get_link_status call to refresh state */
6169 	pf->hw.phy.get_link_info = true;
6170 
6171 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6172 
6173 	status = i40e_get_link_status(&pf->hw, &new_link);
6174 	if (status) {
6175 		dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6176 			status);
6177 		return;
6178 	}
6179 
6180 	old_link_speed = pf->hw.phy.link_info_old.link_speed;
6181 	new_link_speed = pf->hw.phy.link_info.link_speed;
6182 
6183 	if (new_link == old_link &&
6184 	    new_link_speed == old_link_speed &&
6185 	    (test_bit(__I40E_DOWN, &vsi->state) ||
6186 	     new_link == netif_carrier_ok(vsi->netdev)))
6187 		return;
6188 
6189 	if (!test_bit(__I40E_DOWN, &vsi->state))
6190 		i40e_print_link_message(vsi, new_link);
6191 
6192 	/* Notify the base of the switch tree connected to
6193 	 * the link.  Floating VEBs are not notified.
6194 	 */
6195 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6196 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6197 	else
6198 		i40e_vsi_link_event(vsi, new_link);
6199 
6200 	if (pf->vf)
6201 		i40e_vc_notify_link_state(pf);
6202 
6203 	if (pf->flags & I40E_FLAG_PTP)
6204 		i40e_ptp_set_increment(pf);
6205 }
6206 
6207 /**
6208  * i40e_watchdog_subtask - periodic checks not using event driven response
6209  * @pf: board private structure
6210  **/
6211 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6212 {
6213 	int i;
6214 
6215 	/* if interface is down do nothing */
6216 	if (test_bit(__I40E_DOWN, &pf->state) ||
6217 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
6218 		return;
6219 
6220 	/* make sure we don't do these things too often */
6221 	if (time_before(jiffies, (pf->service_timer_previous +
6222 				  pf->service_timer_period)))
6223 		return;
6224 	pf->service_timer_previous = jiffies;
6225 
6226 	if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6227 		i40e_link_event(pf);
6228 
6229 	/* Update the stats for active netdevs so the network stack
6230 	 * can look at updated numbers whenever it cares to
6231 	 */
6232 	for (i = 0; i < pf->num_alloc_vsi; i++)
6233 		if (pf->vsi[i] && pf->vsi[i]->netdev)
6234 			i40e_update_stats(pf->vsi[i]);
6235 
6236 	if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6237 		/* Update the stats for the active switching components */
6238 		for (i = 0; i < I40E_MAX_VEB; i++)
6239 			if (pf->veb[i])
6240 				i40e_update_veb_stats(pf->veb[i]);
6241 	}
6242 
6243 	i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6244 }
6245 
6246 /**
6247  * i40e_reset_subtask - Set up for resetting the device and driver
6248  * @pf: board private structure
6249  **/
6250 static void i40e_reset_subtask(struct i40e_pf *pf)
6251 {
6252 	u32 reset_flags = 0;
6253 
6254 	rtnl_lock();
6255 	if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6256 		reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6257 		clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6258 	}
6259 	if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6260 		reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6261 		clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6262 	}
6263 	if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6264 		reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6265 		clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6266 	}
6267 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6268 		reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6269 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6270 	}
6271 	if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6272 		reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6273 		clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6274 	}
6275 
6276 	/* If there's a recovery already waiting, it takes
6277 	 * precedence before starting a new reset sequence.
6278 	 */
6279 	if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6280 		i40e_handle_reset_warning(pf);
6281 		goto unlock;
6282 	}
6283 
6284 	/* If we're already down or resetting, just bail */
6285 	if (reset_flags &&
6286 	    !test_bit(__I40E_DOWN, &pf->state) &&
6287 	    !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6288 		i40e_do_reset(pf, reset_flags);
6289 
6290 unlock:
6291 	rtnl_unlock();
6292 }
6293 
6294 /**
6295  * i40e_handle_link_event - Handle link event
6296  * @pf: board private structure
6297  * @e: event info posted on ARQ
6298  **/
6299 static void i40e_handle_link_event(struct i40e_pf *pf,
6300 				   struct i40e_arq_event_info *e)
6301 {
6302 	struct i40e_aqc_get_link_status *status =
6303 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6304 
6305 	/* Do a new status request to re-enable LSE reporting
6306 	 * and load new status information into the hw struct
6307 	 * This completely ignores any state information
6308 	 * in the ARQ event info, instead choosing to always
6309 	 * issue the AQ update link status command.
6310 	 */
6311 	i40e_link_event(pf);
6312 
6313 	/* check for unqualified module, if link is down */
6314 	if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6315 	    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6316 	    (!(status->link_info & I40E_AQ_LINK_UP)))
6317 		dev_err(&pf->pdev->dev,
6318 			"The driver failed to link because an unqualified module was detected.\n");
6319 }
6320 
6321 /**
6322  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6323  * @pf: board private structure
6324  **/
6325 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6326 {
6327 	struct i40e_arq_event_info event;
6328 	struct i40e_hw *hw = &pf->hw;
6329 	u16 pending, i = 0;
6330 	i40e_status ret;
6331 	u16 opcode;
6332 	u32 oldval;
6333 	u32 val;
6334 
6335 	/* Do not run clean AQ when PF reset fails */
6336 	if (test_bit(__I40E_RESET_FAILED, &pf->state))
6337 		return;
6338 
6339 	/* check for error indications */
6340 	val = rd32(&pf->hw, pf->hw.aq.arq.len);
6341 	oldval = val;
6342 	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6343 		if (hw->debug_mask & I40E_DEBUG_AQ)
6344 			dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6345 		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6346 	}
6347 	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6348 		if (hw->debug_mask & I40E_DEBUG_AQ)
6349 			dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6350 		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6351 		pf->arq_overflows++;
6352 	}
6353 	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6354 		if (hw->debug_mask & I40E_DEBUG_AQ)
6355 			dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6356 		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6357 	}
6358 	if (oldval != val)
6359 		wr32(&pf->hw, pf->hw.aq.arq.len, val);
6360 
6361 	val = rd32(&pf->hw, pf->hw.aq.asq.len);
6362 	oldval = val;
6363 	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6364 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6365 			dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6366 		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6367 	}
6368 	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6369 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6370 			dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6371 		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6372 	}
6373 	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6374 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6375 			dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6376 		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6377 	}
6378 	if (oldval != val)
6379 		wr32(&pf->hw, pf->hw.aq.asq.len, val);
6380 
6381 	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6382 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6383 	if (!event.msg_buf)
6384 		return;
6385 
6386 	do {
6387 		ret = i40e_clean_arq_element(hw, &event, &pending);
6388 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6389 			break;
6390 		else if (ret) {
6391 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6392 			break;
6393 		}
6394 
6395 		opcode = le16_to_cpu(event.desc.opcode);
6396 		switch (opcode) {
6397 
6398 		case i40e_aqc_opc_get_link_status:
6399 			i40e_handle_link_event(pf, &event);
6400 			break;
6401 		case i40e_aqc_opc_send_msg_to_pf:
6402 			ret = i40e_vc_process_vf_msg(pf,
6403 					le16_to_cpu(event.desc.retval),
6404 					le32_to_cpu(event.desc.cookie_high),
6405 					le32_to_cpu(event.desc.cookie_low),
6406 					event.msg_buf,
6407 					event.msg_len);
6408 			break;
6409 		case i40e_aqc_opc_lldp_update_mib:
6410 			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6411 #ifdef CONFIG_I40E_DCB
6412 			rtnl_lock();
6413 			ret = i40e_handle_lldp_event(pf, &event);
6414 			rtnl_unlock();
6415 #endif /* CONFIG_I40E_DCB */
6416 			break;
6417 		case i40e_aqc_opc_event_lan_overflow:
6418 			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6419 			i40e_handle_lan_overflow_event(pf, &event);
6420 			break;
6421 		case i40e_aqc_opc_send_msg_to_peer:
6422 			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6423 			break;
6424 		case i40e_aqc_opc_nvm_erase:
6425 		case i40e_aqc_opc_nvm_update:
6426 		case i40e_aqc_opc_oem_post_update:
6427 			i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6428 				   "ARQ NVM operation 0x%04x completed\n",
6429 				   opcode);
6430 			break;
6431 		default:
6432 			dev_info(&pf->pdev->dev,
6433 				 "ARQ: Unknown event 0x%04x ignored\n",
6434 				 opcode);
6435 			break;
6436 		}
6437 	} while (pending && (i++ < pf->adminq_work_limit));
6438 
6439 	clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6440 	/* re-enable Admin queue interrupt cause */
6441 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
6442 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6443 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
6444 	i40e_flush(hw);
6445 
6446 	kfree(event.msg_buf);
6447 }
6448 
6449 /**
6450  * i40e_verify_eeprom - make sure eeprom is good to use
6451  * @pf: board private structure
6452  **/
6453 static void i40e_verify_eeprom(struct i40e_pf *pf)
6454 {
6455 	int err;
6456 
6457 	err = i40e_diag_eeprom_test(&pf->hw);
6458 	if (err) {
6459 		/* retry in case of garbage read */
6460 		err = i40e_diag_eeprom_test(&pf->hw);
6461 		if (err) {
6462 			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6463 				 err);
6464 			set_bit(__I40E_BAD_EEPROM, &pf->state);
6465 		}
6466 	}
6467 
6468 	if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6469 		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6470 		clear_bit(__I40E_BAD_EEPROM, &pf->state);
6471 	}
6472 }
6473 
6474 /**
6475  * i40e_enable_pf_switch_lb
6476  * @pf: pointer to the PF structure
6477  *
6478  * enable switch loop back or die - no point in a return value
6479  **/
6480 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6481 {
6482 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6483 	struct i40e_vsi_context ctxt;
6484 	int ret;
6485 
6486 	ctxt.seid = pf->main_vsi_seid;
6487 	ctxt.pf_num = pf->hw.pf_id;
6488 	ctxt.vf_num = 0;
6489 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6490 	if (ret) {
6491 		dev_info(&pf->pdev->dev,
6492 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6493 			 i40e_stat_str(&pf->hw, ret),
6494 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6495 		return;
6496 	}
6497 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6498 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6499 	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6500 
6501 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6502 	if (ret) {
6503 		dev_info(&pf->pdev->dev,
6504 			 "update vsi switch failed, err %s aq_err %s\n",
6505 			 i40e_stat_str(&pf->hw, ret),
6506 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6507 	}
6508 }
6509 
6510 /**
6511  * i40e_disable_pf_switch_lb
6512  * @pf: pointer to the PF structure
6513  *
6514  * disable switch loop back or die - no point in a return value
6515  **/
6516 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6517 {
6518 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6519 	struct i40e_vsi_context ctxt;
6520 	int ret;
6521 
6522 	ctxt.seid = pf->main_vsi_seid;
6523 	ctxt.pf_num = pf->hw.pf_id;
6524 	ctxt.vf_num = 0;
6525 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6526 	if (ret) {
6527 		dev_info(&pf->pdev->dev,
6528 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6529 			 i40e_stat_str(&pf->hw, ret),
6530 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6531 		return;
6532 	}
6533 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6534 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6535 	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6536 
6537 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6538 	if (ret) {
6539 		dev_info(&pf->pdev->dev,
6540 			 "update vsi switch failed, err %s aq_err %s\n",
6541 			 i40e_stat_str(&pf->hw, ret),
6542 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6543 	}
6544 }
6545 
6546 /**
6547  * i40e_config_bridge_mode - Configure the HW bridge mode
6548  * @veb: pointer to the bridge instance
6549  *
6550  * Configure the loop back mode for the LAN VSI that is downlink to the
6551  * specified HW bridge instance. It is expected this function is called
6552  * when a new HW bridge is instantiated.
6553  **/
6554 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6555 {
6556 	struct i40e_pf *pf = veb->pf;
6557 
6558 	if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6559 		dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6560 			 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6561 	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6562 		i40e_disable_pf_switch_lb(pf);
6563 	else
6564 		i40e_enable_pf_switch_lb(pf);
6565 }
6566 
6567 /**
6568  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6569  * @veb: pointer to the VEB instance
6570  *
6571  * This is a recursive function that first builds the attached VSIs then
6572  * recurses in to build the next layer of VEB.  We track the connections
6573  * through our own index numbers because the seid's from the HW could
6574  * change across the reset.
6575  **/
6576 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6577 {
6578 	struct i40e_vsi *ctl_vsi = NULL;
6579 	struct i40e_pf *pf = veb->pf;
6580 	int v, veb_idx;
6581 	int ret;
6582 
6583 	/* build VSI that owns this VEB, temporarily attached to base VEB */
6584 	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6585 		if (pf->vsi[v] &&
6586 		    pf->vsi[v]->veb_idx == veb->idx &&
6587 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6588 			ctl_vsi = pf->vsi[v];
6589 			break;
6590 		}
6591 	}
6592 	if (!ctl_vsi) {
6593 		dev_info(&pf->pdev->dev,
6594 			 "missing owner VSI for veb_idx %d\n", veb->idx);
6595 		ret = -ENOENT;
6596 		goto end_reconstitute;
6597 	}
6598 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
6599 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6600 	ret = i40e_add_vsi(ctl_vsi);
6601 	if (ret) {
6602 		dev_info(&pf->pdev->dev,
6603 			 "rebuild of veb_idx %d owner VSI failed: %d\n",
6604 			 veb->idx, ret);
6605 		goto end_reconstitute;
6606 	}
6607 	i40e_vsi_reset_stats(ctl_vsi);
6608 
6609 	/* create the VEB in the switch and move the VSI onto the VEB */
6610 	ret = i40e_add_veb(veb, ctl_vsi);
6611 	if (ret)
6612 		goto end_reconstitute;
6613 
6614 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6615 		veb->bridge_mode = BRIDGE_MODE_VEB;
6616 	else
6617 		veb->bridge_mode = BRIDGE_MODE_VEPA;
6618 	i40e_config_bridge_mode(veb);
6619 
6620 	/* create the remaining VSIs attached to this VEB */
6621 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6622 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6623 			continue;
6624 
6625 		if (pf->vsi[v]->veb_idx == veb->idx) {
6626 			struct i40e_vsi *vsi = pf->vsi[v];
6627 
6628 			vsi->uplink_seid = veb->seid;
6629 			ret = i40e_add_vsi(vsi);
6630 			if (ret) {
6631 				dev_info(&pf->pdev->dev,
6632 					 "rebuild of vsi_idx %d failed: %d\n",
6633 					 v, ret);
6634 				goto end_reconstitute;
6635 			}
6636 			i40e_vsi_reset_stats(vsi);
6637 		}
6638 	}
6639 
6640 	/* create any VEBs attached to this VEB - RECURSION */
6641 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6642 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6643 			pf->veb[veb_idx]->uplink_seid = veb->seid;
6644 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6645 			if (ret)
6646 				break;
6647 		}
6648 	}
6649 
6650 end_reconstitute:
6651 	return ret;
6652 }
6653 
6654 /**
6655  * i40e_get_capabilities - get info about the HW
6656  * @pf: the PF struct
6657  **/
6658 static int i40e_get_capabilities(struct i40e_pf *pf)
6659 {
6660 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6661 	u16 data_size;
6662 	int buf_len;
6663 	int err;
6664 
6665 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6666 	do {
6667 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
6668 		if (!cap_buf)
6669 			return -ENOMEM;
6670 
6671 		/* this loads the data into the hw struct for us */
6672 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6673 					    &data_size,
6674 					    i40e_aqc_opc_list_func_capabilities,
6675 					    NULL);
6676 		/* data loaded, buffer no longer needed */
6677 		kfree(cap_buf);
6678 
6679 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6680 			/* retry with a larger buffer */
6681 			buf_len = data_size;
6682 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6683 			dev_info(&pf->pdev->dev,
6684 				 "capability discovery failed, err %s aq_err %s\n",
6685 				 i40e_stat_str(&pf->hw, err),
6686 				 i40e_aq_str(&pf->hw,
6687 					     pf->hw.aq.asq_last_status));
6688 			return -ENODEV;
6689 		}
6690 	} while (err);
6691 
6692 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
6693 		dev_info(&pf->pdev->dev,
6694 			 "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",
6695 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6696 			 pf->hw.func_caps.num_msix_vectors,
6697 			 pf->hw.func_caps.num_msix_vectors_vf,
6698 			 pf->hw.func_caps.fd_filters_guaranteed,
6699 			 pf->hw.func_caps.fd_filters_best_effort,
6700 			 pf->hw.func_caps.num_tx_qp,
6701 			 pf->hw.func_caps.num_vsis);
6702 
6703 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6704 		       + pf->hw.func_caps.num_vfs)
6705 	if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6706 		dev_info(&pf->pdev->dev,
6707 			 "got num_vsis %d, setting num_vsis to %d\n",
6708 			 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6709 		pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6710 	}
6711 
6712 	return 0;
6713 }
6714 
6715 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6716 
6717 /**
6718  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6719  * @pf: board private structure
6720  **/
6721 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6722 {
6723 	struct i40e_vsi *vsi;
6724 
6725 	/* quick workaround for an NVM issue that leaves a critical register
6726 	 * uninitialized
6727 	 */
6728 	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6729 		static const u32 hkey[] = {
6730 			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6731 			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6732 			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6733 			0x95b3a76d};
6734 		int i;
6735 
6736 		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6737 			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6738 	}
6739 
6740 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6741 		return;
6742 
6743 	/* find existing VSI and see if it needs configuring */
6744 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
6745 
6746 	/* create a new VSI if none exists */
6747 	if (!vsi) {
6748 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6749 				     pf->vsi[pf->lan_vsi]->seid, 0);
6750 		if (!vsi) {
6751 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6752 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6753 			return;
6754 		}
6755 	}
6756 
6757 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6758 }
6759 
6760 /**
6761  * i40e_fdir_teardown - release the Flow Director resources
6762  * @pf: board private structure
6763  **/
6764 static void i40e_fdir_teardown(struct i40e_pf *pf)
6765 {
6766 	struct i40e_vsi *vsi;
6767 
6768 	i40e_fdir_filter_exit(pf);
6769 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
6770 	if (vsi)
6771 		i40e_vsi_release(vsi);
6772 }
6773 
6774 /**
6775  * i40e_prep_for_reset - prep for the core to reset
6776  * @pf: board private structure
6777  *
6778  * Close up the VFs and other things in prep for PF Reset.
6779   **/
6780 static void i40e_prep_for_reset(struct i40e_pf *pf)
6781 {
6782 	struct i40e_hw *hw = &pf->hw;
6783 	i40e_status ret = 0;
6784 	u32 v;
6785 
6786 	clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6787 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6788 		return;
6789 	if (i40e_check_asq_alive(&pf->hw))
6790 		i40e_vc_notify_reset(pf);
6791 
6792 	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6793 
6794 	/* quiesce the VSIs and their queues that are not already DOWN */
6795 	i40e_pf_quiesce_all_vsi(pf);
6796 
6797 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6798 		if (pf->vsi[v])
6799 			pf->vsi[v]->seid = 0;
6800 	}
6801 
6802 	i40e_shutdown_adminq(&pf->hw);
6803 
6804 	/* call shutdown HMC */
6805 	if (hw->hmc.hmc_obj) {
6806 		ret = i40e_shutdown_lan_hmc(hw);
6807 		if (ret)
6808 			dev_warn(&pf->pdev->dev,
6809 				 "shutdown_lan_hmc failed: %d\n", ret);
6810 	}
6811 }
6812 
6813 /**
6814  * i40e_send_version - update firmware with driver version
6815  * @pf: PF struct
6816  */
6817 static void i40e_send_version(struct i40e_pf *pf)
6818 {
6819 	struct i40e_driver_version dv;
6820 
6821 	dv.major_version = DRV_VERSION_MAJOR;
6822 	dv.minor_version = DRV_VERSION_MINOR;
6823 	dv.build_version = DRV_VERSION_BUILD;
6824 	dv.subbuild_version = 0;
6825 	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6826 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6827 }
6828 
6829 /**
6830  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6831  * @pf: board private structure
6832  * @reinit: if the Main VSI needs to re-initialized.
6833  **/
6834 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6835 {
6836 	struct i40e_hw *hw = &pf->hw;
6837 	u8 set_fc_aq_fail = 0;
6838 	i40e_status ret;
6839 	u32 val;
6840 	u32 v;
6841 
6842 	/* Now we wait for GRST to settle out.
6843 	 * We don't have to delete the VEBs or VSIs from the hw switch
6844 	 * because the reset will make them disappear.
6845 	 */
6846 	ret = i40e_pf_reset(hw);
6847 	if (ret) {
6848 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6849 		set_bit(__I40E_RESET_FAILED, &pf->state);
6850 		goto clear_recovery;
6851 	}
6852 	pf->pfr_count++;
6853 
6854 	if (test_bit(__I40E_DOWN, &pf->state))
6855 		goto clear_recovery;
6856 	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6857 
6858 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6859 	ret = i40e_init_adminq(&pf->hw);
6860 	if (ret) {
6861 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6862 			 i40e_stat_str(&pf->hw, ret),
6863 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6864 		goto clear_recovery;
6865 	}
6866 
6867 	/* re-verify the eeprom if we just had an EMP reset */
6868 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6869 		i40e_verify_eeprom(pf);
6870 
6871 	i40e_clear_pxe_mode(hw);
6872 	ret = i40e_get_capabilities(pf);
6873 	if (ret)
6874 		goto end_core_reset;
6875 
6876 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6877 				hw->func_caps.num_rx_qp,
6878 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6879 	if (ret) {
6880 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6881 		goto end_core_reset;
6882 	}
6883 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6884 	if (ret) {
6885 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6886 		goto end_core_reset;
6887 	}
6888 
6889 #ifdef CONFIG_I40E_DCB
6890 	ret = i40e_init_pf_dcb(pf);
6891 	if (ret) {
6892 		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6893 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6894 		/* Continue without DCB enabled */
6895 	}
6896 #endif /* CONFIG_I40E_DCB */
6897 #ifdef I40E_FCOE
6898 	i40e_init_pf_fcoe(pf);
6899 
6900 #endif
6901 	/* do basic switch setup */
6902 	ret = i40e_setup_pf_switch(pf, reinit);
6903 	if (ret)
6904 		goto end_core_reset;
6905 
6906 	/* The driver only wants link up/down and module qualification
6907 	 * reports from firmware.  Note the negative logic.
6908 	 */
6909 	ret = i40e_aq_set_phy_int_mask(&pf->hw,
6910 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
6911 					 I40E_AQ_EVENT_MEDIA_NA |
6912 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
6913 	if (ret)
6914 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6915 			 i40e_stat_str(&pf->hw, ret),
6916 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6917 
6918 	/* make sure our flow control settings are restored */
6919 	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6920 	if (ret)
6921 		dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6922 			i40e_stat_str(&pf->hw, ret),
6923 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6924 
6925 	/* Rebuild the VSIs and VEBs that existed before reset.
6926 	 * They are still in our local switch element arrays, so only
6927 	 * need to rebuild the switch model in the HW.
6928 	 *
6929 	 * If there were VEBs but the reconstitution failed, we'll try
6930 	 * try to recover minimal use by getting the basic PF VSI working.
6931 	 */
6932 	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6933 		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6934 		/* find the one VEB connected to the MAC, and find orphans */
6935 		for (v = 0; v < I40E_MAX_VEB; v++) {
6936 			if (!pf->veb[v])
6937 				continue;
6938 
6939 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6940 			    pf->veb[v]->uplink_seid == 0) {
6941 				ret = i40e_reconstitute_veb(pf->veb[v]);
6942 
6943 				if (!ret)
6944 					continue;
6945 
6946 				/* If Main VEB failed, we're in deep doodoo,
6947 				 * so give up rebuilding the switch and set up
6948 				 * for minimal rebuild of PF VSI.
6949 				 * If orphan failed, we'll report the error
6950 				 * but try to keep going.
6951 				 */
6952 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6953 					dev_info(&pf->pdev->dev,
6954 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6955 						 ret);
6956 					pf->vsi[pf->lan_vsi]->uplink_seid
6957 								= pf->mac_seid;
6958 					break;
6959 				} else if (pf->veb[v]->uplink_seid == 0) {
6960 					dev_info(&pf->pdev->dev,
6961 						 "rebuild of orphan VEB failed: %d\n",
6962 						 ret);
6963 				}
6964 			}
6965 		}
6966 	}
6967 
6968 	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6969 		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6970 		/* no VEB, so rebuild only the Main VSI */
6971 		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6972 		if (ret) {
6973 			dev_info(&pf->pdev->dev,
6974 				 "rebuild of Main VSI failed: %d\n", ret);
6975 			goto end_core_reset;
6976 		}
6977 	}
6978 
6979 	/* Reconfigure hardware for allowing smaller MSS in the case
6980 	 * of TSO, so that we avoid the MDD being fired and causing
6981 	 * a reset in the case of small MSS+TSO.
6982 	 */
6983 #define I40E_REG_MSS          0x000E64DC
6984 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6985 #define I40E_64BYTE_MSS       0x400000
6986 	val = rd32(hw, I40E_REG_MSS);
6987 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6988 		val &= ~I40E_REG_MSS_MIN_MASK;
6989 		val |= I40E_64BYTE_MSS;
6990 		wr32(hw, I40E_REG_MSS, val);
6991 	}
6992 
6993 	if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
6994 		msleep(75);
6995 		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6996 		if (ret)
6997 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6998 				 i40e_stat_str(&pf->hw, ret),
6999 				 i40e_aq_str(&pf->hw,
7000 					     pf->hw.aq.asq_last_status));
7001 	}
7002 	/* reinit the misc interrupt */
7003 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7004 		ret = i40e_setup_misc_vector(pf);
7005 
7006 	/* Add a filter to drop all Flow control frames from any VSI from being
7007 	 * transmitted. By doing so we stop a malicious VF from sending out
7008 	 * PAUSE or PFC frames and potentially controlling traffic for other
7009 	 * PF/VF VSIs.
7010 	 * The FW can still send Flow control frames if enabled.
7011 	 */
7012 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
7013 						       pf->main_vsi_seid);
7014 
7015 	/* restart the VSIs that were rebuilt and running before the reset */
7016 	i40e_pf_unquiesce_all_vsi(pf);
7017 
7018 	if (pf->num_alloc_vfs) {
7019 		for (v = 0; v < pf->num_alloc_vfs; v++)
7020 			i40e_reset_vf(&pf->vf[v], true);
7021 	}
7022 
7023 	/* tell the firmware that we're starting */
7024 	i40e_send_version(pf);
7025 
7026 end_core_reset:
7027 	clear_bit(__I40E_RESET_FAILED, &pf->state);
7028 clear_recovery:
7029 	clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
7030 }
7031 
7032 /**
7033  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7034  * @pf: board private structure
7035  *
7036  * Close up the VFs and other things in prep for a Core Reset,
7037  * then get ready to rebuild the world.
7038  **/
7039 static void i40e_handle_reset_warning(struct i40e_pf *pf)
7040 {
7041 	i40e_prep_for_reset(pf);
7042 	i40e_reset_and_rebuild(pf, false);
7043 }
7044 
7045 /**
7046  * i40e_handle_mdd_event
7047  * @pf: pointer to the PF structure
7048  *
7049  * Called from the MDD irq handler to identify possibly malicious vfs
7050  **/
7051 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7052 {
7053 	struct i40e_hw *hw = &pf->hw;
7054 	bool mdd_detected = false;
7055 	bool pf_mdd_detected = false;
7056 	struct i40e_vf *vf;
7057 	u32 reg;
7058 	int i;
7059 
7060 	if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
7061 		return;
7062 
7063 	/* find what triggered the MDD event */
7064 	reg = rd32(hw, I40E_GL_MDET_TX);
7065 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7066 		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7067 				I40E_GL_MDET_TX_PF_NUM_SHIFT;
7068 		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7069 				I40E_GL_MDET_TX_VF_NUM_SHIFT;
7070 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7071 				I40E_GL_MDET_TX_EVENT_SHIFT;
7072 		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7073 				I40E_GL_MDET_TX_QUEUE_SHIFT) -
7074 				pf->hw.func_caps.base_queue;
7075 		if (netif_msg_tx_err(pf))
7076 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7077 				 event, queue, pf_num, vf_num);
7078 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7079 		mdd_detected = true;
7080 	}
7081 	reg = rd32(hw, I40E_GL_MDET_RX);
7082 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7083 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7084 				I40E_GL_MDET_RX_FUNCTION_SHIFT;
7085 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7086 				I40E_GL_MDET_RX_EVENT_SHIFT;
7087 		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7088 				I40E_GL_MDET_RX_QUEUE_SHIFT) -
7089 				pf->hw.func_caps.base_queue;
7090 		if (netif_msg_rx_err(pf))
7091 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7092 				 event, queue, func);
7093 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7094 		mdd_detected = true;
7095 	}
7096 
7097 	if (mdd_detected) {
7098 		reg = rd32(hw, I40E_PF_MDET_TX);
7099 		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7100 			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7101 			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7102 			pf_mdd_detected = true;
7103 		}
7104 		reg = rd32(hw, I40E_PF_MDET_RX);
7105 		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7106 			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7107 			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7108 			pf_mdd_detected = true;
7109 		}
7110 		/* Queue belongs to the PF, initiate a reset */
7111 		if (pf_mdd_detected) {
7112 			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7113 			i40e_service_event_schedule(pf);
7114 		}
7115 	}
7116 
7117 	/* see if one of the VFs needs its hand slapped */
7118 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7119 		vf = &(pf->vf[i]);
7120 		reg = rd32(hw, I40E_VP_MDET_TX(i));
7121 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7122 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7123 			vf->num_mdd_events++;
7124 			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7125 				 i);
7126 		}
7127 
7128 		reg = rd32(hw, I40E_VP_MDET_RX(i));
7129 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7130 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7131 			vf->num_mdd_events++;
7132 			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7133 				 i);
7134 		}
7135 
7136 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7137 			dev_info(&pf->pdev->dev,
7138 				 "Too many MDD events on VF %d, disabled\n", i);
7139 			dev_info(&pf->pdev->dev,
7140 				 "Use PF Control I/F to re-enable the VF\n");
7141 			set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7142 		}
7143 	}
7144 
7145 	/* re-enable mdd interrupt cause */
7146 	clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7147 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7148 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7149 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7150 	i40e_flush(hw);
7151 }
7152 
7153 /**
7154  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7155  * @pf: board private structure
7156  **/
7157 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7158 {
7159 	struct i40e_hw *hw = &pf->hw;
7160 	i40e_status ret;
7161 	__be16 port;
7162 	int i;
7163 
7164 	if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7165 		return;
7166 
7167 	pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7168 
7169 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7170 		if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7171 			pf->pending_udp_bitmap &= ~BIT_ULL(i);
7172 			port = pf->udp_ports[i].index;
7173 			if (port)
7174 				ret = i40e_aq_add_udp_tunnel(hw, port,
7175 							pf->udp_ports[i].type,
7176 							NULL, NULL);
7177 			else
7178 				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7179 
7180 			if (ret) {
7181 				dev_dbg(&pf->pdev->dev,
7182 					"%s %s port %d, index %d failed, err %s aq_err %s\n",
7183 					pf->udp_ports[i].type ? "vxlan" : "geneve",
7184 					port ? "add" : "delete",
7185 					ntohs(port), i,
7186 					i40e_stat_str(&pf->hw, ret),
7187 					i40e_aq_str(&pf->hw,
7188 						    pf->hw.aq.asq_last_status));
7189 				pf->udp_ports[i].index = 0;
7190 			}
7191 		}
7192 	}
7193 }
7194 
7195 /**
7196  * i40e_service_task - Run the driver's async subtasks
7197  * @work: pointer to work_struct containing our data
7198  **/
7199 static void i40e_service_task(struct work_struct *work)
7200 {
7201 	struct i40e_pf *pf = container_of(work,
7202 					  struct i40e_pf,
7203 					  service_task);
7204 	unsigned long start_time = jiffies;
7205 
7206 	/* don't bother with service tasks if a reset is in progress */
7207 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7208 		i40e_service_event_complete(pf);
7209 		return;
7210 	}
7211 
7212 	i40e_detect_recover_hung(pf);
7213 	i40e_sync_filters_subtask(pf);
7214 	i40e_reset_subtask(pf);
7215 	i40e_handle_mdd_event(pf);
7216 	i40e_vc_process_vflr_event(pf);
7217 	i40e_watchdog_subtask(pf);
7218 	i40e_fdir_reinit_subtask(pf);
7219 	i40e_client_subtask(pf);
7220 	i40e_sync_filters_subtask(pf);
7221 	i40e_sync_udp_filters_subtask(pf);
7222 	i40e_clean_adminq_subtask(pf);
7223 
7224 	i40e_service_event_complete(pf);
7225 
7226 	/* If the tasks have taken longer than one timer cycle or there
7227 	 * is more work to be done, reschedule the service task now
7228 	 * rather than wait for the timer to tick again.
7229 	 */
7230 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7231 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)		 ||
7232 	    test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)		 ||
7233 	    test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7234 		i40e_service_event_schedule(pf);
7235 }
7236 
7237 /**
7238  * i40e_service_timer - timer callback
7239  * @data: pointer to PF struct
7240  **/
7241 static void i40e_service_timer(unsigned long data)
7242 {
7243 	struct i40e_pf *pf = (struct i40e_pf *)data;
7244 
7245 	mod_timer(&pf->service_timer,
7246 		  round_jiffies(jiffies + pf->service_timer_period));
7247 	i40e_service_event_schedule(pf);
7248 }
7249 
7250 /**
7251  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7252  * @vsi: the VSI being configured
7253  **/
7254 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7255 {
7256 	struct i40e_pf *pf = vsi->back;
7257 
7258 	switch (vsi->type) {
7259 	case I40E_VSI_MAIN:
7260 		vsi->alloc_queue_pairs = pf->num_lan_qps;
7261 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7262 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7263 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7264 			vsi->num_q_vectors = pf->num_lan_msix;
7265 		else
7266 			vsi->num_q_vectors = 1;
7267 
7268 		break;
7269 
7270 	case I40E_VSI_FDIR:
7271 		vsi->alloc_queue_pairs = 1;
7272 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7273 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7274 		vsi->num_q_vectors = pf->num_fdsb_msix;
7275 		break;
7276 
7277 	case I40E_VSI_VMDQ2:
7278 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7279 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7280 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7281 		vsi->num_q_vectors = pf->num_vmdq_msix;
7282 		break;
7283 
7284 	case I40E_VSI_SRIOV:
7285 		vsi->alloc_queue_pairs = pf->num_vf_qps;
7286 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7287 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7288 		break;
7289 
7290 #ifdef I40E_FCOE
7291 	case I40E_VSI_FCOE:
7292 		vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7293 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7294 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7295 		vsi->num_q_vectors = pf->num_fcoe_msix;
7296 		break;
7297 
7298 #endif /* I40E_FCOE */
7299 	default:
7300 		WARN_ON(1);
7301 		return -ENODATA;
7302 	}
7303 
7304 	return 0;
7305 }
7306 
7307 /**
7308  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7309  * @type: VSI pointer
7310  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7311  *
7312  * On error: returns error code (negative)
7313  * On success: returns 0
7314  **/
7315 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7316 {
7317 	int size;
7318 	int ret = 0;
7319 
7320 	/* allocate memory for both Tx and Rx ring pointers */
7321 	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7322 	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7323 	if (!vsi->tx_rings)
7324 		return -ENOMEM;
7325 	vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7326 
7327 	if (alloc_qvectors) {
7328 		/* allocate memory for q_vector pointers */
7329 		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7330 		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7331 		if (!vsi->q_vectors) {
7332 			ret = -ENOMEM;
7333 			goto err_vectors;
7334 		}
7335 	}
7336 	return ret;
7337 
7338 err_vectors:
7339 	kfree(vsi->tx_rings);
7340 	return ret;
7341 }
7342 
7343 /**
7344  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7345  * @pf: board private structure
7346  * @type: type of VSI
7347  *
7348  * On error: returns error code (negative)
7349  * On success: returns vsi index in PF (positive)
7350  **/
7351 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7352 {
7353 	int ret = -ENODEV;
7354 	struct i40e_vsi *vsi;
7355 	int vsi_idx;
7356 	int i;
7357 
7358 	/* Need to protect the allocation of the VSIs at the PF level */
7359 	mutex_lock(&pf->switch_mutex);
7360 
7361 	/* VSI list may be fragmented if VSI creation/destruction has
7362 	 * been happening.  We can afford to do a quick scan to look
7363 	 * for any free VSIs in the list.
7364 	 *
7365 	 * find next empty vsi slot, looping back around if necessary
7366 	 */
7367 	i = pf->next_vsi;
7368 	while (i < pf->num_alloc_vsi && pf->vsi[i])
7369 		i++;
7370 	if (i >= pf->num_alloc_vsi) {
7371 		i = 0;
7372 		while (i < pf->next_vsi && pf->vsi[i])
7373 			i++;
7374 	}
7375 
7376 	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7377 		vsi_idx = i;             /* Found one! */
7378 	} else {
7379 		ret = -ENODEV;
7380 		goto unlock_pf;  /* out of VSI slots! */
7381 	}
7382 	pf->next_vsi = ++i;
7383 
7384 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7385 	if (!vsi) {
7386 		ret = -ENOMEM;
7387 		goto unlock_pf;
7388 	}
7389 	vsi->type = type;
7390 	vsi->back = pf;
7391 	set_bit(__I40E_DOWN, &vsi->state);
7392 	vsi->flags = 0;
7393 	vsi->idx = vsi_idx;
7394 	vsi->int_rate_limit = 0;
7395 	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7396 				pf->rss_table_size : 64;
7397 	vsi->netdev_registered = false;
7398 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7399 	hash_init(vsi->mac_filter_hash);
7400 	vsi->irqs_ready = false;
7401 
7402 	ret = i40e_set_num_rings_in_vsi(vsi);
7403 	if (ret)
7404 		goto err_rings;
7405 
7406 	ret = i40e_vsi_alloc_arrays(vsi, true);
7407 	if (ret)
7408 		goto err_rings;
7409 
7410 	/* Setup default MSIX irq handler for VSI */
7411 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7412 
7413 	/* Initialize VSI lock */
7414 	spin_lock_init(&vsi->mac_filter_hash_lock);
7415 	pf->vsi[vsi_idx] = vsi;
7416 	ret = vsi_idx;
7417 	goto unlock_pf;
7418 
7419 err_rings:
7420 	pf->next_vsi = i - 1;
7421 	kfree(vsi);
7422 unlock_pf:
7423 	mutex_unlock(&pf->switch_mutex);
7424 	return ret;
7425 }
7426 
7427 /**
7428  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7429  * @type: VSI pointer
7430  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7431  *
7432  * On error: returns error code (negative)
7433  * On success: returns 0
7434  **/
7435 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7436 {
7437 	/* free the ring and vector containers */
7438 	if (free_qvectors) {
7439 		kfree(vsi->q_vectors);
7440 		vsi->q_vectors = NULL;
7441 	}
7442 	kfree(vsi->tx_rings);
7443 	vsi->tx_rings = NULL;
7444 	vsi->rx_rings = NULL;
7445 }
7446 
7447 /**
7448  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7449  * and lookup table
7450  * @vsi: Pointer to VSI structure
7451  */
7452 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7453 {
7454 	if (!vsi)
7455 		return;
7456 
7457 	kfree(vsi->rss_hkey_user);
7458 	vsi->rss_hkey_user = NULL;
7459 
7460 	kfree(vsi->rss_lut_user);
7461 	vsi->rss_lut_user = NULL;
7462 }
7463 
7464 /**
7465  * i40e_vsi_clear - Deallocate the VSI provided
7466  * @vsi: the VSI being un-configured
7467  **/
7468 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7469 {
7470 	struct i40e_pf *pf;
7471 
7472 	if (!vsi)
7473 		return 0;
7474 
7475 	if (!vsi->back)
7476 		goto free_vsi;
7477 	pf = vsi->back;
7478 
7479 	mutex_lock(&pf->switch_mutex);
7480 	if (!pf->vsi[vsi->idx]) {
7481 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7482 			vsi->idx, vsi->idx, vsi, vsi->type);
7483 		goto unlock_vsi;
7484 	}
7485 
7486 	if (pf->vsi[vsi->idx] != vsi) {
7487 		dev_err(&pf->pdev->dev,
7488 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7489 			pf->vsi[vsi->idx]->idx,
7490 			pf->vsi[vsi->idx],
7491 			pf->vsi[vsi->idx]->type,
7492 			vsi->idx, vsi, vsi->type);
7493 		goto unlock_vsi;
7494 	}
7495 
7496 	/* updates the PF for this cleared vsi */
7497 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7498 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7499 
7500 	i40e_vsi_free_arrays(vsi, true);
7501 	i40e_clear_rss_config_user(vsi);
7502 
7503 	pf->vsi[vsi->idx] = NULL;
7504 	if (vsi->idx < pf->next_vsi)
7505 		pf->next_vsi = vsi->idx;
7506 
7507 unlock_vsi:
7508 	mutex_unlock(&pf->switch_mutex);
7509 free_vsi:
7510 	kfree(vsi);
7511 
7512 	return 0;
7513 }
7514 
7515 /**
7516  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7517  * @vsi: the VSI being cleaned
7518  **/
7519 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7520 {
7521 	int i;
7522 
7523 	if (vsi->tx_rings && vsi->tx_rings[0]) {
7524 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7525 			kfree_rcu(vsi->tx_rings[i], rcu);
7526 			vsi->tx_rings[i] = NULL;
7527 			vsi->rx_rings[i] = NULL;
7528 		}
7529 	}
7530 }
7531 
7532 /**
7533  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7534  * @vsi: the VSI being configured
7535  **/
7536 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7537 {
7538 	struct i40e_ring *tx_ring, *rx_ring;
7539 	struct i40e_pf *pf = vsi->back;
7540 	int i;
7541 
7542 	/* Set basic values in the rings to be used later during open() */
7543 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7544 		/* allocate space for both Tx and Rx in one shot */
7545 		tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7546 		if (!tx_ring)
7547 			goto err_out;
7548 
7549 		tx_ring->queue_index = i;
7550 		tx_ring->reg_idx = vsi->base_queue + i;
7551 		tx_ring->ring_active = false;
7552 		tx_ring->vsi = vsi;
7553 		tx_ring->netdev = vsi->netdev;
7554 		tx_ring->dev = &pf->pdev->dev;
7555 		tx_ring->count = vsi->num_desc;
7556 		tx_ring->size = 0;
7557 		tx_ring->dcb_tc = 0;
7558 		if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7559 			tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7560 		tx_ring->tx_itr_setting = pf->tx_itr_default;
7561 		vsi->tx_rings[i] = tx_ring;
7562 
7563 		rx_ring = &tx_ring[1];
7564 		rx_ring->queue_index = i;
7565 		rx_ring->reg_idx = vsi->base_queue + i;
7566 		rx_ring->ring_active = false;
7567 		rx_ring->vsi = vsi;
7568 		rx_ring->netdev = vsi->netdev;
7569 		rx_ring->dev = &pf->pdev->dev;
7570 		rx_ring->count = vsi->num_desc;
7571 		rx_ring->size = 0;
7572 		rx_ring->dcb_tc = 0;
7573 		rx_ring->rx_itr_setting = pf->rx_itr_default;
7574 		vsi->rx_rings[i] = rx_ring;
7575 	}
7576 
7577 	return 0;
7578 
7579 err_out:
7580 	i40e_vsi_clear_rings(vsi);
7581 	return -ENOMEM;
7582 }
7583 
7584 /**
7585  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7586  * @pf: board private structure
7587  * @vectors: the number of MSI-X vectors to request
7588  *
7589  * Returns the number of vectors reserved, or error
7590  **/
7591 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7592 {
7593 	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7594 					I40E_MIN_MSIX, vectors);
7595 	if (vectors < 0) {
7596 		dev_info(&pf->pdev->dev,
7597 			 "MSI-X vector reservation failed: %d\n", vectors);
7598 		vectors = 0;
7599 	}
7600 
7601 	return vectors;
7602 }
7603 
7604 /**
7605  * i40e_init_msix - Setup the MSIX capability
7606  * @pf: board private structure
7607  *
7608  * Work with the OS to set up the MSIX vectors needed.
7609  *
7610  * Returns the number of vectors reserved or negative on failure
7611  **/
7612 static int i40e_init_msix(struct i40e_pf *pf)
7613 {
7614 	struct i40e_hw *hw = &pf->hw;
7615 	int vectors_left;
7616 	int v_budget, i;
7617 	int v_actual;
7618 	int iwarp_requested = 0;
7619 
7620 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7621 		return -ENODEV;
7622 
7623 	/* The number of vectors we'll request will be comprised of:
7624 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
7625 	 *   - The number of LAN queue pairs
7626 	 *	- Queues being used for RSS.
7627 	 *		We don't need as many as max_rss_size vectors.
7628 	 *		use rss_size instead in the calculation since that
7629 	 *		is governed by number of cpus in the system.
7630 	 *	- assumes symmetric Tx/Rx pairing
7631 	 *   - The number of VMDq pairs
7632 	 *   - The CPU count within the NUMA node if iWARP is enabled
7633 #ifdef I40E_FCOE
7634 	 *   - The number of FCOE qps.
7635 #endif
7636 	 * Once we count this up, try the request.
7637 	 *
7638 	 * If we can't get what we want, we'll simplify to nearly nothing
7639 	 * and try again.  If that still fails, we punt.
7640 	 */
7641 	vectors_left = hw->func_caps.num_msix_vectors;
7642 	v_budget = 0;
7643 
7644 	/* reserve one vector for miscellaneous handler */
7645 	if (vectors_left) {
7646 		v_budget++;
7647 		vectors_left--;
7648 	}
7649 
7650 	/* reserve vectors for the main PF traffic queues */
7651 	pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7652 	vectors_left -= pf->num_lan_msix;
7653 	v_budget += pf->num_lan_msix;
7654 
7655 	/* reserve one vector for sideband flow director */
7656 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7657 		if (vectors_left) {
7658 			pf->num_fdsb_msix = 1;
7659 			v_budget++;
7660 			vectors_left--;
7661 		} else {
7662 			pf->num_fdsb_msix = 0;
7663 		}
7664 	}
7665 
7666 #ifdef I40E_FCOE
7667 	/* can we reserve enough for FCoE? */
7668 	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7669 		if (!vectors_left)
7670 			pf->num_fcoe_msix = 0;
7671 		else if (vectors_left >= pf->num_fcoe_qps)
7672 			pf->num_fcoe_msix = pf->num_fcoe_qps;
7673 		else
7674 			pf->num_fcoe_msix = 1;
7675 		v_budget += pf->num_fcoe_msix;
7676 		vectors_left -= pf->num_fcoe_msix;
7677 	}
7678 
7679 #endif
7680 	/* can we reserve enough for iWARP? */
7681 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7682 		iwarp_requested = pf->num_iwarp_msix;
7683 
7684 		if (!vectors_left)
7685 			pf->num_iwarp_msix = 0;
7686 		else if (vectors_left < pf->num_iwarp_msix)
7687 			pf->num_iwarp_msix = 1;
7688 		v_budget += pf->num_iwarp_msix;
7689 		vectors_left -= pf->num_iwarp_msix;
7690 	}
7691 
7692 	/* any vectors left over go for VMDq support */
7693 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7694 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7695 		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7696 
7697 		if (!vectors_left) {
7698 			pf->num_vmdq_msix = 0;
7699 			pf->num_vmdq_qps = 0;
7700 		} else {
7701 			/* if we're short on vectors for what's desired, we limit
7702 			 * the queues per vmdq.  If this is still more than are
7703 			 * available, the user will need to change the number of
7704 			 * queues/vectors used by the PF later with the ethtool
7705 			 * channels command
7706 			 */
7707 			if (vmdq_vecs < vmdq_vecs_wanted)
7708 				pf->num_vmdq_qps = 1;
7709 			pf->num_vmdq_msix = pf->num_vmdq_qps;
7710 
7711 			v_budget += vmdq_vecs;
7712 			vectors_left -= vmdq_vecs;
7713 		}
7714 	}
7715 
7716 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7717 				   GFP_KERNEL);
7718 	if (!pf->msix_entries)
7719 		return -ENOMEM;
7720 
7721 	for (i = 0; i < v_budget; i++)
7722 		pf->msix_entries[i].entry = i;
7723 	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7724 
7725 	if (v_actual < I40E_MIN_MSIX) {
7726 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7727 		kfree(pf->msix_entries);
7728 		pf->msix_entries = NULL;
7729 		pci_disable_msix(pf->pdev);
7730 		return -ENODEV;
7731 
7732 	} else if (v_actual == I40E_MIN_MSIX) {
7733 		/* Adjust for minimal MSIX use */
7734 		pf->num_vmdq_vsis = 0;
7735 		pf->num_vmdq_qps = 0;
7736 		pf->num_lan_qps = 1;
7737 		pf->num_lan_msix = 1;
7738 
7739 	} else if (!vectors_left) {
7740 		/* If we have limited resources, we will start with no vectors
7741 		 * for the special features and then allocate vectors to some
7742 		 * of these features based on the policy and at the end disable
7743 		 * the features that did not get any vectors.
7744 		 */
7745 		int vec;
7746 
7747 		dev_info(&pf->pdev->dev,
7748 			 "MSI-X vector limit reached, attempting to redistribute vectors\n");
7749 		/* reserve the misc vector */
7750 		vec = v_actual - 1;
7751 
7752 		/* Scale vector usage down */
7753 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7754 		pf->num_vmdq_vsis = 1;
7755 		pf->num_vmdq_qps = 1;
7756 #ifdef I40E_FCOE
7757 		pf->num_fcoe_qps = 0;
7758 		pf->num_fcoe_msix = 0;
7759 #endif
7760 
7761 		/* partition out the remaining vectors */
7762 		switch (vec) {
7763 		case 2:
7764 			pf->num_lan_msix = 1;
7765 			break;
7766 		case 3:
7767 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7768 				pf->num_lan_msix = 1;
7769 				pf->num_iwarp_msix = 1;
7770 			} else {
7771 				pf->num_lan_msix = 2;
7772 			}
7773 #ifdef I40E_FCOE
7774 			/* give one vector to FCoE */
7775 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7776 				pf->num_lan_msix = 1;
7777 				pf->num_fcoe_msix = 1;
7778 			}
7779 #endif
7780 			break;
7781 		default:
7782 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7783 				pf->num_iwarp_msix = min_t(int, (vec / 3),
7784 						 iwarp_requested);
7785 				pf->num_vmdq_vsis = min_t(int, (vec / 3),
7786 						  I40E_DEFAULT_NUM_VMDQ_VSI);
7787 			} else {
7788 				pf->num_vmdq_vsis = min_t(int, (vec / 2),
7789 						  I40E_DEFAULT_NUM_VMDQ_VSI);
7790 			}
7791 			if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7792 				pf->num_fdsb_msix = 1;
7793 				vec--;
7794 			}
7795 			pf->num_lan_msix = min_t(int,
7796 			       (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7797 							      pf->num_lan_msix);
7798 			pf->num_lan_qps = pf->num_lan_msix;
7799 #ifdef I40E_FCOE
7800 			/* give one vector to FCoE */
7801 			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7802 				pf->num_fcoe_msix = 1;
7803 				vec--;
7804 			}
7805 #endif
7806 			break;
7807 		}
7808 	}
7809 
7810 	if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
7811 	    (pf->num_fdsb_msix == 0)) {
7812 		dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
7813 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7814 	}
7815 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7816 	    (pf->num_vmdq_msix == 0)) {
7817 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7818 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7819 	}
7820 
7821 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7822 	    (pf->num_iwarp_msix == 0)) {
7823 		dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7824 		pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7825 	}
7826 #ifdef I40E_FCOE
7827 
7828 	if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7829 		dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7830 		pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7831 	}
7832 #endif
7833 	i40e_debug(&pf->hw, I40E_DEBUG_INIT,
7834 		   "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
7835 		   pf->num_lan_msix,
7836 		   pf->num_vmdq_msix * pf->num_vmdq_vsis,
7837 		   pf->num_fdsb_msix,
7838 		   pf->num_iwarp_msix);
7839 
7840 	return v_actual;
7841 }
7842 
7843 /**
7844  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7845  * @vsi: the VSI being configured
7846  * @v_idx: index of the vector in the vsi struct
7847  * @cpu: cpu to be used on affinity_mask
7848  *
7849  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7850  **/
7851 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
7852 {
7853 	struct i40e_q_vector *q_vector;
7854 
7855 	/* allocate q_vector */
7856 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7857 	if (!q_vector)
7858 		return -ENOMEM;
7859 
7860 	q_vector->vsi = vsi;
7861 	q_vector->v_idx = v_idx;
7862 	cpumask_set_cpu(cpu, &q_vector->affinity_mask);
7863 
7864 	if (vsi->netdev)
7865 		netif_napi_add(vsi->netdev, &q_vector->napi,
7866 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
7867 
7868 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
7869 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
7870 
7871 	/* tie q_vector and vsi together */
7872 	vsi->q_vectors[v_idx] = q_vector;
7873 
7874 	return 0;
7875 }
7876 
7877 /**
7878  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7879  * @vsi: the VSI being configured
7880  *
7881  * We allocate one q_vector per queue interrupt.  If allocation fails we
7882  * return -ENOMEM.
7883  **/
7884 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7885 {
7886 	struct i40e_pf *pf = vsi->back;
7887 	int err, v_idx, num_q_vectors, current_cpu;
7888 
7889 	/* if not MSIX, give the one vector only to the LAN VSI */
7890 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7891 		num_q_vectors = vsi->num_q_vectors;
7892 	else if (vsi == pf->vsi[pf->lan_vsi])
7893 		num_q_vectors = 1;
7894 	else
7895 		return -EINVAL;
7896 
7897 	current_cpu = cpumask_first(cpu_online_mask);
7898 
7899 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7900 		err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
7901 		if (err)
7902 			goto err_out;
7903 		current_cpu = cpumask_next(current_cpu, cpu_online_mask);
7904 		if (unlikely(current_cpu >= nr_cpu_ids))
7905 			current_cpu = cpumask_first(cpu_online_mask);
7906 	}
7907 
7908 	return 0;
7909 
7910 err_out:
7911 	while (v_idx--)
7912 		i40e_free_q_vector(vsi, v_idx);
7913 
7914 	return err;
7915 }
7916 
7917 /**
7918  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7919  * @pf: board private structure to initialize
7920  **/
7921 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7922 {
7923 	int vectors = 0;
7924 	ssize_t size;
7925 
7926 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7927 		vectors = i40e_init_msix(pf);
7928 		if (vectors < 0) {
7929 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
7930 				       I40E_FLAG_IWARP_ENABLED	|
7931 #ifdef I40E_FCOE
7932 				       I40E_FLAG_FCOE_ENABLED	|
7933 #endif
7934 				       I40E_FLAG_RSS_ENABLED	|
7935 				       I40E_FLAG_DCB_CAPABLE	|
7936 				       I40E_FLAG_DCB_ENABLED	|
7937 				       I40E_FLAG_SRIOV_ENABLED	|
7938 				       I40E_FLAG_FD_SB_ENABLED	|
7939 				       I40E_FLAG_FD_ATR_ENABLED	|
7940 				       I40E_FLAG_VMDQ_ENABLED);
7941 
7942 			/* rework the queue expectations without MSIX */
7943 			i40e_determine_queue_usage(pf);
7944 		}
7945 	}
7946 
7947 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7948 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7949 		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7950 		vectors = pci_enable_msi(pf->pdev);
7951 		if (vectors < 0) {
7952 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7953 				 vectors);
7954 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7955 		}
7956 		vectors = 1;  /* one MSI or Legacy vector */
7957 	}
7958 
7959 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7960 		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7961 
7962 	/* set up vector assignment tracking */
7963 	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7964 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
7965 	if (!pf->irq_pile) {
7966 		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7967 		return -ENOMEM;
7968 	}
7969 	pf->irq_pile->num_entries = vectors;
7970 	pf->irq_pile->search_hint = 0;
7971 
7972 	/* track first vector for misc interrupts, ignore return */
7973 	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7974 
7975 	return 0;
7976 }
7977 
7978 /**
7979  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7980  * @pf: board private structure
7981  *
7982  * This sets up the handler for MSIX 0, which is used to manage the
7983  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7984  * when in MSI or Legacy interrupt mode.
7985  **/
7986 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7987 {
7988 	struct i40e_hw *hw = &pf->hw;
7989 	int err = 0;
7990 
7991 	/* Only request the irq if this is the first time through, and
7992 	 * not when we're rebuilding after a Reset
7993 	 */
7994 	if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7995 		err = request_irq(pf->msix_entries[0].vector,
7996 				  i40e_intr, 0, pf->int_name, pf);
7997 		if (err) {
7998 			dev_info(&pf->pdev->dev,
7999 				 "request_irq for %s failed: %d\n",
8000 				 pf->int_name, err);
8001 			return -EFAULT;
8002 		}
8003 	}
8004 
8005 	i40e_enable_misc_int_causes(pf);
8006 
8007 	/* associate no queues to the misc vector */
8008 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
8009 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
8010 
8011 	i40e_flush(hw);
8012 
8013 	i40e_irq_dynamic_enable_icr0(pf, true);
8014 
8015 	return err;
8016 }
8017 
8018 /**
8019  * i40e_config_rss_aq - Prepare for RSS using AQ commands
8020  * @vsi: vsi structure
8021  * @seed: RSS hash seed
8022  **/
8023 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8024 			      u8 *lut, u16 lut_size)
8025 {
8026 	struct i40e_pf *pf = vsi->back;
8027 	struct i40e_hw *hw = &pf->hw;
8028 	int ret = 0;
8029 
8030 	if (seed) {
8031 		struct i40e_aqc_get_set_rss_key_data *seed_dw =
8032 			(struct i40e_aqc_get_set_rss_key_data *)seed;
8033 		ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
8034 		if (ret) {
8035 			dev_info(&pf->pdev->dev,
8036 				 "Cannot set RSS key, err %s aq_err %s\n",
8037 				 i40e_stat_str(hw, ret),
8038 				 i40e_aq_str(hw, hw->aq.asq_last_status));
8039 			return ret;
8040 		}
8041 	}
8042 	if (lut) {
8043 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8044 
8045 		ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8046 		if (ret) {
8047 			dev_info(&pf->pdev->dev,
8048 				 "Cannot set RSS lut, err %s aq_err %s\n",
8049 				 i40e_stat_str(hw, ret),
8050 				 i40e_aq_str(hw, hw->aq.asq_last_status));
8051 			return ret;
8052 		}
8053 	}
8054 	return ret;
8055 }
8056 
8057 /**
8058  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8059  * @vsi: Pointer to vsi structure
8060  * @seed: Buffter to store the hash keys
8061  * @lut: Buffer to store the lookup table entries
8062  * @lut_size: Size of buffer to store the lookup table entries
8063  *
8064  * Return 0 on success, negative on failure
8065  */
8066 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8067 			   u8 *lut, u16 lut_size)
8068 {
8069 	struct i40e_pf *pf = vsi->back;
8070 	struct i40e_hw *hw = &pf->hw;
8071 	int ret = 0;
8072 
8073 	if (seed) {
8074 		ret = i40e_aq_get_rss_key(hw, vsi->id,
8075 			(struct i40e_aqc_get_set_rss_key_data *)seed);
8076 		if (ret) {
8077 			dev_info(&pf->pdev->dev,
8078 				 "Cannot get RSS key, err %s aq_err %s\n",
8079 				 i40e_stat_str(&pf->hw, ret),
8080 				 i40e_aq_str(&pf->hw,
8081 					     pf->hw.aq.asq_last_status));
8082 			return ret;
8083 		}
8084 	}
8085 
8086 	if (lut) {
8087 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8088 
8089 		ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8090 		if (ret) {
8091 			dev_info(&pf->pdev->dev,
8092 				 "Cannot get RSS lut, err %s aq_err %s\n",
8093 				 i40e_stat_str(&pf->hw, ret),
8094 				 i40e_aq_str(&pf->hw,
8095 					     pf->hw.aq.asq_last_status));
8096 			return ret;
8097 		}
8098 	}
8099 
8100 	return ret;
8101 }
8102 
8103 /**
8104  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8105  * @vsi: VSI structure
8106  **/
8107 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8108 {
8109 	u8 seed[I40E_HKEY_ARRAY_SIZE];
8110 	struct i40e_pf *pf = vsi->back;
8111 	u8 *lut;
8112 	int ret;
8113 
8114 	if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8115 		return 0;
8116 
8117 	if (!vsi->rss_size)
8118 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
8119 				      vsi->num_queue_pairs);
8120 	if (!vsi->rss_size)
8121 		return -EINVAL;
8122 
8123 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8124 	if (!lut)
8125 		return -ENOMEM;
8126 	/* Use the user configured hash keys and lookup table if there is one,
8127 	 * otherwise use default
8128 	 */
8129 	if (vsi->rss_lut_user)
8130 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8131 	else
8132 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8133 	if (vsi->rss_hkey_user)
8134 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8135 	else
8136 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8137 	ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8138 	kfree(lut);
8139 
8140 	return ret;
8141 }
8142 
8143 /**
8144  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8145  * @vsi: Pointer to vsi structure
8146  * @seed: RSS hash seed
8147  * @lut: Lookup table
8148  * @lut_size: Lookup table size
8149  *
8150  * Returns 0 on success, negative on failure
8151  **/
8152 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8153 			       const u8 *lut, u16 lut_size)
8154 {
8155 	struct i40e_pf *pf = vsi->back;
8156 	struct i40e_hw *hw = &pf->hw;
8157 	u16 vf_id = vsi->vf_id;
8158 	u8 i;
8159 
8160 	/* Fill out hash function seed */
8161 	if (seed) {
8162 		u32 *seed_dw = (u32 *)seed;
8163 
8164 		if (vsi->type == I40E_VSI_MAIN) {
8165 			for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8166 				i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i),
8167 						  seed_dw[i]);
8168 		} else if (vsi->type == I40E_VSI_SRIOV) {
8169 			for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8170 				i40e_write_rx_ctl(hw,
8171 						  I40E_VFQF_HKEY1(i, vf_id),
8172 						  seed_dw[i]);
8173 		} else {
8174 			dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8175 		}
8176 	}
8177 
8178 	if (lut) {
8179 		u32 *lut_dw = (u32 *)lut;
8180 
8181 		if (vsi->type == I40E_VSI_MAIN) {
8182 			if (lut_size != I40E_HLUT_ARRAY_SIZE)
8183 				return -EINVAL;
8184 			for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8185 				wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8186 		} else if (vsi->type == I40E_VSI_SRIOV) {
8187 			if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8188 				return -EINVAL;
8189 			for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8190 				i40e_write_rx_ctl(hw,
8191 						  I40E_VFQF_HLUT1(i, vf_id),
8192 						  lut_dw[i]);
8193 		} else {
8194 			dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8195 		}
8196 	}
8197 	i40e_flush(hw);
8198 
8199 	return 0;
8200 }
8201 
8202 /**
8203  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8204  * @vsi: Pointer to VSI structure
8205  * @seed: Buffer to store the keys
8206  * @lut: Buffer to store the lookup table entries
8207  * @lut_size: Size of buffer to store the lookup table entries
8208  *
8209  * Returns 0 on success, negative on failure
8210  */
8211 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8212 			    u8 *lut, u16 lut_size)
8213 {
8214 	struct i40e_pf *pf = vsi->back;
8215 	struct i40e_hw *hw = &pf->hw;
8216 	u16 i;
8217 
8218 	if (seed) {
8219 		u32 *seed_dw = (u32 *)seed;
8220 
8221 		for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8222 			seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8223 	}
8224 	if (lut) {
8225 		u32 *lut_dw = (u32 *)lut;
8226 
8227 		if (lut_size != I40E_HLUT_ARRAY_SIZE)
8228 			return -EINVAL;
8229 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8230 			lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8231 	}
8232 
8233 	return 0;
8234 }
8235 
8236 /**
8237  * i40e_config_rss - Configure RSS keys and lut
8238  * @vsi: Pointer to VSI structure
8239  * @seed: RSS hash seed
8240  * @lut: Lookup table
8241  * @lut_size: Lookup table size
8242  *
8243  * Returns 0 on success, negative on failure
8244  */
8245 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8246 {
8247 	struct i40e_pf *pf = vsi->back;
8248 
8249 	if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8250 		return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8251 	else
8252 		return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8253 }
8254 
8255 /**
8256  * i40e_get_rss - Get RSS keys and lut
8257  * @vsi: Pointer to VSI structure
8258  * @seed: Buffer to store the keys
8259  * @lut: Buffer to store the lookup table entries
8260  * lut_size: Size of buffer to store the lookup table entries
8261  *
8262  * Returns 0 on success, negative on failure
8263  */
8264 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8265 {
8266 	struct i40e_pf *pf = vsi->back;
8267 
8268 	if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8269 		return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8270 	else
8271 		return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8272 }
8273 
8274 /**
8275  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8276  * @pf: Pointer to board private structure
8277  * @lut: Lookup table
8278  * @rss_table_size: Lookup table size
8279  * @rss_size: Range of queue number for hashing
8280  */
8281 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8282 		       u16 rss_table_size, u16 rss_size)
8283 {
8284 	u16 i;
8285 
8286 	for (i = 0; i < rss_table_size; i++)
8287 		lut[i] = i % rss_size;
8288 }
8289 
8290 /**
8291  * i40e_pf_config_rss - Prepare for RSS if used
8292  * @pf: board private structure
8293  **/
8294 static int i40e_pf_config_rss(struct i40e_pf *pf)
8295 {
8296 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8297 	u8 seed[I40E_HKEY_ARRAY_SIZE];
8298 	u8 *lut;
8299 	struct i40e_hw *hw = &pf->hw;
8300 	u32 reg_val;
8301 	u64 hena;
8302 	int ret;
8303 
8304 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8305 	hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8306 		((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8307 	hena |= i40e_pf_get_default_rss_hena(pf);
8308 
8309 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8310 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8311 
8312 	/* Determine the RSS table size based on the hardware capabilities */
8313 	reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8314 	reg_val = (pf->rss_table_size == 512) ?
8315 			(reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8316 			(reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8317 	i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8318 
8319 	/* Determine the RSS size of the VSI */
8320 	if (!vsi->rss_size)
8321 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
8322 				      vsi->num_queue_pairs);
8323 	if (!vsi->rss_size)
8324 		return -EINVAL;
8325 
8326 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8327 	if (!lut)
8328 		return -ENOMEM;
8329 
8330 	/* Use user configured lut if there is one, otherwise use default */
8331 	if (vsi->rss_lut_user)
8332 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8333 	else
8334 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8335 
8336 	/* Use user configured hash key if there is one, otherwise
8337 	 * use default.
8338 	 */
8339 	if (vsi->rss_hkey_user)
8340 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8341 	else
8342 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8343 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8344 	kfree(lut);
8345 
8346 	return ret;
8347 }
8348 
8349 /**
8350  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8351  * @pf: board private structure
8352  * @queue_count: the requested queue count for rss.
8353  *
8354  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8355  * count which may be different from the requested queue count.
8356  **/
8357 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8358 {
8359 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8360 	int new_rss_size;
8361 
8362 	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8363 		return 0;
8364 
8365 	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8366 
8367 	if (queue_count != vsi->num_queue_pairs) {
8368 		vsi->req_queue_pairs = queue_count;
8369 		i40e_prep_for_reset(pf);
8370 
8371 		pf->alloc_rss_size = new_rss_size;
8372 
8373 		i40e_reset_and_rebuild(pf, true);
8374 
8375 		/* Discard the user configured hash keys and lut, if less
8376 		 * queues are enabled.
8377 		 */
8378 		if (queue_count < vsi->rss_size) {
8379 			i40e_clear_rss_config_user(vsi);
8380 			dev_dbg(&pf->pdev->dev,
8381 				"discard user configured hash keys and lut\n");
8382 		}
8383 
8384 		/* Reset vsi->rss_size, as number of enabled queues changed */
8385 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
8386 				      vsi->num_queue_pairs);
8387 
8388 		i40e_pf_config_rss(pf);
8389 	}
8390 	dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
8391 		 vsi->req_queue_pairs, pf->rss_size_max);
8392 	return pf->alloc_rss_size;
8393 }
8394 
8395 /**
8396  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8397  * @pf: board private structure
8398  **/
8399 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8400 {
8401 	i40e_status status;
8402 	bool min_valid, max_valid;
8403 	u32 max_bw, min_bw;
8404 
8405 	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8406 					   &min_valid, &max_valid);
8407 
8408 	if (!status) {
8409 		if (min_valid)
8410 			pf->npar_min_bw = min_bw;
8411 		if (max_valid)
8412 			pf->npar_max_bw = max_bw;
8413 	}
8414 
8415 	return status;
8416 }
8417 
8418 /**
8419  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8420  * @pf: board private structure
8421  **/
8422 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8423 {
8424 	struct i40e_aqc_configure_partition_bw_data bw_data;
8425 	i40e_status status;
8426 
8427 	/* Set the valid bit for this PF */
8428 	bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8429 	bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8430 	bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8431 
8432 	/* Set the new bandwidths */
8433 	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8434 
8435 	return status;
8436 }
8437 
8438 /**
8439  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8440  * @pf: board private structure
8441  **/
8442 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8443 {
8444 	/* Commit temporary BW setting to permanent NVM image */
8445 	enum i40e_admin_queue_err last_aq_status;
8446 	i40e_status ret;
8447 	u16 nvm_word;
8448 
8449 	if (pf->hw.partition_id != 1) {
8450 		dev_info(&pf->pdev->dev,
8451 			 "Commit BW only works on partition 1! This is partition %d",
8452 			 pf->hw.partition_id);
8453 		ret = I40E_NOT_SUPPORTED;
8454 		goto bw_commit_out;
8455 	}
8456 
8457 	/* Acquire NVM for read access */
8458 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8459 	last_aq_status = pf->hw.aq.asq_last_status;
8460 	if (ret) {
8461 		dev_info(&pf->pdev->dev,
8462 			 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8463 			 i40e_stat_str(&pf->hw, ret),
8464 			 i40e_aq_str(&pf->hw, last_aq_status));
8465 		goto bw_commit_out;
8466 	}
8467 
8468 	/* Read word 0x10 of NVM - SW compatibility word 1 */
8469 	ret = i40e_aq_read_nvm(&pf->hw,
8470 			       I40E_SR_NVM_CONTROL_WORD,
8471 			       0x10, sizeof(nvm_word), &nvm_word,
8472 			       false, NULL);
8473 	/* Save off last admin queue command status before releasing
8474 	 * the NVM
8475 	 */
8476 	last_aq_status = pf->hw.aq.asq_last_status;
8477 	i40e_release_nvm(&pf->hw);
8478 	if (ret) {
8479 		dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8480 			 i40e_stat_str(&pf->hw, ret),
8481 			 i40e_aq_str(&pf->hw, last_aq_status));
8482 		goto bw_commit_out;
8483 	}
8484 
8485 	/* Wait a bit for NVM release to complete */
8486 	msleep(50);
8487 
8488 	/* Acquire NVM for write access */
8489 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8490 	last_aq_status = pf->hw.aq.asq_last_status;
8491 	if (ret) {
8492 		dev_info(&pf->pdev->dev,
8493 			 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8494 			 i40e_stat_str(&pf->hw, ret),
8495 			 i40e_aq_str(&pf->hw, last_aq_status));
8496 		goto bw_commit_out;
8497 	}
8498 	/* Write it back out unchanged to initiate update NVM,
8499 	 * which will force a write of the shadow (alt) RAM to
8500 	 * the NVM - thus storing the bandwidth values permanently.
8501 	 */
8502 	ret = i40e_aq_update_nvm(&pf->hw,
8503 				 I40E_SR_NVM_CONTROL_WORD,
8504 				 0x10, sizeof(nvm_word),
8505 				 &nvm_word, true, NULL);
8506 	/* Save off last admin queue command status before releasing
8507 	 * the NVM
8508 	 */
8509 	last_aq_status = pf->hw.aq.asq_last_status;
8510 	i40e_release_nvm(&pf->hw);
8511 	if (ret)
8512 		dev_info(&pf->pdev->dev,
8513 			 "BW settings NOT SAVED, err %s aq_err %s\n",
8514 			 i40e_stat_str(&pf->hw, ret),
8515 			 i40e_aq_str(&pf->hw, last_aq_status));
8516 bw_commit_out:
8517 
8518 	return ret;
8519 }
8520 
8521 /**
8522  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8523  * @pf: board private structure to initialize
8524  *
8525  * i40e_sw_init initializes the Adapter private data structure.
8526  * Fields are initialized based on PCI device information and
8527  * OS network device settings (MTU size).
8528  **/
8529 static int i40e_sw_init(struct i40e_pf *pf)
8530 {
8531 	int err = 0;
8532 	int size;
8533 
8534 	/* Set default capability flags */
8535 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8536 		    I40E_FLAG_MSI_ENABLED     |
8537 		    I40E_FLAG_MSIX_ENABLED;
8538 
8539 	/* Set default ITR */
8540 	pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8541 	pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8542 
8543 	/* Depending on PF configurations, it is possible that the RSS
8544 	 * maximum might end up larger than the available queues
8545 	 */
8546 	pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8547 	pf->alloc_rss_size = 1;
8548 	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8549 	pf->rss_size_max = min_t(int, pf->rss_size_max,
8550 				 pf->hw.func_caps.num_tx_qp);
8551 	if (pf->hw.func_caps.rss) {
8552 		pf->flags |= I40E_FLAG_RSS_ENABLED;
8553 		pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8554 					   num_online_cpus());
8555 	}
8556 
8557 	/* MFP mode enabled */
8558 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8559 		pf->flags |= I40E_FLAG_MFP_ENABLED;
8560 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8561 		if (i40e_get_npar_bw_setting(pf))
8562 			dev_warn(&pf->pdev->dev,
8563 				 "Could not get NPAR bw settings\n");
8564 		else
8565 			dev_info(&pf->pdev->dev,
8566 				 "Min BW = %8.8x, Max BW = %8.8x\n",
8567 				 pf->npar_min_bw, pf->npar_max_bw);
8568 	}
8569 
8570 	/* FW/NVM is not yet fixed in this regard */
8571 	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8572 	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8573 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8574 		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8575 		if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8576 		    pf->hw.num_partitions > 1)
8577 			dev_info(&pf->pdev->dev,
8578 				 "Flow Director Sideband mode Disabled in MFP mode\n");
8579 		else
8580 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8581 		pf->fdir_pf_filter_count =
8582 				 pf->hw.func_caps.fd_filters_guaranteed;
8583 		pf->hw.fdir_shared_filter_count =
8584 				 pf->hw.func_caps.fd_filters_best_effort;
8585 	}
8586 
8587 	if (i40e_is_mac_710(&pf->hw) &&
8588 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8589 	    (pf->hw.aq.fw_maj_ver < 4))) {
8590 		pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8591 		/* No DCB support  for FW < v4.33 */
8592 		pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8593 	}
8594 
8595 	/* Disable FW LLDP if FW < v4.3 */
8596 	if (i40e_is_mac_710(&pf->hw) &&
8597 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8598 	    (pf->hw.aq.fw_maj_ver < 4)))
8599 		pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8600 
8601 	/* Use the FW Set LLDP MIB API if FW > v4.40 */
8602 	if (i40e_is_mac_710(&pf->hw) &&
8603 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8604 	    (pf->hw.aq.fw_maj_ver >= 5)))
8605 		pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8606 
8607 	if (pf->hw.func_caps.vmdq) {
8608 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8609 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8610 		pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8611 	}
8612 
8613 	if (pf->hw.func_caps.iwarp) {
8614 		pf->flags |= I40E_FLAG_IWARP_ENABLED;
8615 		/* IWARP needs one extra vector for CQP just like MISC.*/
8616 		pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8617 	}
8618 
8619 #ifdef I40E_FCOE
8620 	i40e_init_pf_fcoe(pf);
8621 
8622 #endif /* I40E_FCOE */
8623 #ifdef CONFIG_PCI_IOV
8624 	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8625 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8626 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8627 		pf->num_req_vfs = min_t(int,
8628 					pf->hw.func_caps.num_vfs,
8629 					I40E_MAX_VF_COUNT);
8630 	}
8631 #endif /* CONFIG_PCI_IOV */
8632 	if (pf->hw.mac.type == I40E_MAC_X722) {
8633 		pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8634 			     I40E_FLAG_128_QP_RSS_CAPABLE |
8635 			     I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8636 			     I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8637 			     I40E_FLAG_WB_ON_ITR_CAPABLE |
8638 			     I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8639 			     I40E_FLAG_NO_PCI_LINK_CHECK |
8640 			     I40E_FLAG_USE_SET_LLDP_MIB |
8641 			     I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8642 	} else if ((pf->hw.aq.api_maj_ver > 1) ||
8643 		   ((pf->hw.aq.api_maj_ver == 1) &&
8644 		    (pf->hw.aq.api_min_ver > 4))) {
8645 		/* Supported in FW API version higher than 1.4 */
8646 		pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8647 		pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8648 	} else {
8649 		pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8650 	}
8651 
8652 	pf->eeprom_version = 0xDEAD;
8653 	pf->lan_veb = I40E_NO_VEB;
8654 	pf->lan_vsi = I40E_NO_VSI;
8655 
8656 	/* By default FW has this off for performance reasons */
8657 	pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8658 
8659 	/* set up queue assignment tracking */
8660 	size = sizeof(struct i40e_lump_tracking)
8661 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8662 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
8663 	if (!pf->qp_pile) {
8664 		err = -ENOMEM;
8665 		goto sw_init_done;
8666 	}
8667 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8668 	pf->qp_pile->search_hint = 0;
8669 
8670 	pf->tx_timeout_recovery_level = 1;
8671 
8672 	mutex_init(&pf->switch_mutex);
8673 
8674 	/* If NPAR is enabled nudge the Tx scheduler */
8675 	if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8676 		i40e_set_npar_bw_setting(pf);
8677 
8678 sw_init_done:
8679 	return err;
8680 }
8681 
8682 /**
8683  * i40e_set_ntuple - set the ntuple feature flag and take action
8684  * @pf: board private structure to initialize
8685  * @features: the feature set that the stack is suggesting
8686  *
8687  * returns a bool to indicate if reset needs to happen
8688  **/
8689 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8690 {
8691 	bool need_reset = false;
8692 
8693 	/* Check if Flow Director n-tuple support was enabled or disabled.  If
8694 	 * the state changed, we need to reset.
8695 	 */
8696 	if (features & NETIF_F_NTUPLE) {
8697 		/* Enable filters and mark for reset */
8698 		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8699 			need_reset = true;
8700 		/* enable FD_SB only if there is MSI-X vector */
8701 		if (pf->num_fdsb_msix > 0)
8702 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8703 	} else {
8704 		/* turn off filters, mark for reset and clear SW filter list */
8705 		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8706 			need_reset = true;
8707 			i40e_fdir_filter_exit(pf);
8708 		}
8709 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8710 		pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8711 		/* reset fd counters */
8712 		pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8713 		pf->fdir_pf_active_filters = 0;
8714 		/* if ATR was auto disabled it can be re-enabled. */
8715 		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8716 		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
8717 			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8718 			if (I40E_DEBUG_FD & pf->hw.debug_mask)
8719 				dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8720 		}
8721 	}
8722 	return need_reset;
8723 }
8724 
8725 /**
8726  * i40e_clear_rss_lut - clear the rx hash lookup table
8727  * @vsi: the VSI being configured
8728  **/
8729 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
8730 {
8731 	struct i40e_pf *pf = vsi->back;
8732 	struct i40e_hw *hw = &pf->hw;
8733 	u16 vf_id = vsi->vf_id;
8734 	u8 i;
8735 
8736 	if (vsi->type == I40E_VSI_MAIN) {
8737 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8738 			wr32(hw, I40E_PFQF_HLUT(i), 0);
8739 	} else if (vsi->type == I40E_VSI_SRIOV) {
8740 		for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8741 			i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
8742 	} else {
8743 		dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8744 	}
8745 }
8746 
8747 /**
8748  * i40e_set_features - set the netdev feature flags
8749  * @netdev: ptr to the netdev being adjusted
8750  * @features: the feature set that the stack is suggesting
8751  **/
8752 static int i40e_set_features(struct net_device *netdev,
8753 			     netdev_features_t features)
8754 {
8755 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8756 	struct i40e_vsi *vsi = np->vsi;
8757 	struct i40e_pf *pf = vsi->back;
8758 	bool need_reset;
8759 
8760 	if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
8761 		i40e_pf_config_rss(pf);
8762 	else if (!(features & NETIF_F_RXHASH) &&
8763 		 netdev->features & NETIF_F_RXHASH)
8764 		i40e_clear_rss_lut(vsi);
8765 
8766 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
8767 		i40e_vlan_stripping_enable(vsi);
8768 	else
8769 		i40e_vlan_stripping_disable(vsi);
8770 
8771 	need_reset = i40e_set_ntuple(pf, features);
8772 
8773 	if (need_reset)
8774 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8775 
8776 	return 0;
8777 }
8778 
8779 /**
8780  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8781  * @pf: board private structure
8782  * @port: The UDP port to look up
8783  *
8784  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8785  **/
8786 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8787 {
8788 	u8 i;
8789 
8790 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8791 		if (pf->udp_ports[i].index == port)
8792 			return i;
8793 	}
8794 
8795 	return i;
8796 }
8797 
8798 /**
8799  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
8800  * @netdev: This physical port's netdev
8801  * @ti: Tunnel endpoint information
8802  **/
8803 static void i40e_udp_tunnel_add(struct net_device *netdev,
8804 				struct udp_tunnel_info *ti)
8805 {
8806 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8807 	struct i40e_vsi *vsi = np->vsi;
8808 	struct i40e_pf *pf = vsi->back;
8809 	__be16 port = ti->port;
8810 	u8 next_idx;
8811 	u8 idx;
8812 
8813 	idx = i40e_get_udp_port_idx(pf, port);
8814 
8815 	/* Check if port already exists */
8816 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8817 		netdev_info(netdev, "port %d already offloaded\n",
8818 			    ntohs(port));
8819 		return;
8820 	}
8821 
8822 	/* Now check if there is space to add the new port */
8823 	next_idx = i40e_get_udp_port_idx(pf, 0);
8824 
8825 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8826 		netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
8827 			    ntohs(port));
8828 		return;
8829 	}
8830 
8831 	switch (ti->type) {
8832 	case UDP_TUNNEL_TYPE_VXLAN:
8833 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8834 		break;
8835 	case UDP_TUNNEL_TYPE_GENEVE:
8836 		if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8837 			return;
8838 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8839 		break;
8840 	default:
8841 		return;
8842 	}
8843 
8844 	/* New port: add it and mark its index in the bitmap */
8845 	pf->udp_ports[next_idx].index = port;
8846 	pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8847 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8848 }
8849 
8850 /**
8851  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
8852  * @netdev: This physical port's netdev
8853  * @ti: Tunnel endpoint information
8854  **/
8855 static void i40e_udp_tunnel_del(struct net_device *netdev,
8856 				struct udp_tunnel_info *ti)
8857 {
8858 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8859 	struct i40e_vsi *vsi = np->vsi;
8860 	struct i40e_pf *pf = vsi->back;
8861 	__be16 port = ti->port;
8862 	u8 idx;
8863 
8864 	idx = i40e_get_udp_port_idx(pf, port);
8865 
8866 	/* Check if port already exists */
8867 	if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
8868 		goto not_found;
8869 
8870 	switch (ti->type) {
8871 	case UDP_TUNNEL_TYPE_VXLAN:
8872 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
8873 			goto not_found;
8874 		break;
8875 	case UDP_TUNNEL_TYPE_GENEVE:
8876 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
8877 			goto not_found;
8878 		break;
8879 	default:
8880 		goto not_found;
8881 	}
8882 
8883 	/* if port exists, set it to 0 (mark for deletion)
8884 	 * and make it pending
8885 	 */
8886 	pf->udp_ports[idx].index = 0;
8887 	pf->pending_udp_bitmap |= BIT_ULL(idx);
8888 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8889 
8890 	return;
8891 not_found:
8892 	netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
8893 		    ntohs(port));
8894 }
8895 
8896 static int i40e_get_phys_port_id(struct net_device *netdev,
8897 				 struct netdev_phys_item_id *ppid)
8898 {
8899 	struct i40e_netdev_priv *np = netdev_priv(netdev);
8900 	struct i40e_pf *pf = np->vsi->back;
8901 	struct i40e_hw *hw = &pf->hw;
8902 
8903 	if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8904 		return -EOPNOTSUPP;
8905 
8906 	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8907 	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8908 
8909 	return 0;
8910 }
8911 
8912 /**
8913  * i40e_ndo_fdb_add - add an entry to the hardware database
8914  * @ndm: the input from the stack
8915  * @tb: pointer to array of nladdr (unused)
8916  * @dev: the net device pointer
8917  * @addr: the MAC address entry being added
8918  * @flags: instructions from stack about fdb operation
8919  */
8920 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8921 			    struct net_device *dev,
8922 			    const unsigned char *addr, u16 vid,
8923 			    u16 flags)
8924 {
8925 	struct i40e_netdev_priv *np = netdev_priv(dev);
8926 	struct i40e_pf *pf = np->vsi->back;
8927 	int err = 0;
8928 
8929 	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8930 		return -EOPNOTSUPP;
8931 
8932 	if (vid) {
8933 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8934 		return -EINVAL;
8935 	}
8936 
8937 	/* Hardware does not support aging addresses so if a
8938 	 * ndm_state is given only allow permanent addresses
8939 	 */
8940 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8941 		netdev_info(dev, "FDB only supports static addresses\n");
8942 		return -EINVAL;
8943 	}
8944 
8945 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8946 		err = dev_uc_add_excl(dev, addr);
8947 	else if (is_multicast_ether_addr(addr))
8948 		err = dev_mc_add_excl(dev, addr);
8949 	else
8950 		err = -EINVAL;
8951 
8952 	/* Only return duplicate errors if NLM_F_EXCL is set */
8953 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
8954 		err = 0;
8955 
8956 	return err;
8957 }
8958 
8959 /**
8960  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8961  * @dev: the netdev being configured
8962  * @nlh: RTNL message
8963  *
8964  * Inserts a new hardware bridge if not already created and
8965  * enables the bridging mode requested (VEB or VEPA). If the
8966  * hardware bridge has already been inserted and the request
8967  * is to change the mode then that requires a PF reset to
8968  * allow rebuild of the components with required hardware
8969  * bridge mode enabled.
8970  **/
8971 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8972 				   struct nlmsghdr *nlh,
8973 				   u16 flags)
8974 {
8975 	struct i40e_netdev_priv *np = netdev_priv(dev);
8976 	struct i40e_vsi *vsi = np->vsi;
8977 	struct i40e_pf *pf = vsi->back;
8978 	struct i40e_veb *veb = NULL;
8979 	struct nlattr *attr, *br_spec;
8980 	int i, rem;
8981 
8982 	/* Only for PF VSI for now */
8983 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8984 		return -EOPNOTSUPP;
8985 
8986 	/* Find the HW bridge for PF VSI */
8987 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8988 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8989 			veb = pf->veb[i];
8990 	}
8991 
8992 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8993 
8994 	nla_for_each_nested(attr, br_spec, rem) {
8995 		__u16 mode;
8996 
8997 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8998 			continue;
8999 
9000 		mode = nla_get_u16(attr);
9001 		if ((mode != BRIDGE_MODE_VEPA) &&
9002 		    (mode != BRIDGE_MODE_VEB))
9003 			return -EINVAL;
9004 
9005 		/* Insert a new HW bridge */
9006 		if (!veb) {
9007 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9008 					     vsi->tc_config.enabled_tc);
9009 			if (veb) {
9010 				veb->bridge_mode = mode;
9011 				i40e_config_bridge_mode(veb);
9012 			} else {
9013 				/* No Bridge HW offload available */
9014 				return -ENOENT;
9015 			}
9016 			break;
9017 		} else if (mode != veb->bridge_mode) {
9018 			/* Existing HW bridge but different mode needs reset */
9019 			veb->bridge_mode = mode;
9020 			/* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
9021 			if (mode == BRIDGE_MODE_VEB)
9022 				pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
9023 			else
9024 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9025 			i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
9026 			break;
9027 		}
9028 	}
9029 
9030 	return 0;
9031 }
9032 
9033 /**
9034  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
9035  * @skb: skb buff
9036  * @pid: process id
9037  * @seq: RTNL message seq #
9038  * @dev: the netdev being configured
9039  * @filter_mask: unused
9040  * @nlflags: netlink flags passed in
9041  *
9042  * Return the mode in which the hardware bridge is operating in
9043  * i.e VEB or VEPA.
9044  **/
9045 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9046 				   struct net_device *dev,
9047 				   u32 __always_unused filter_mask,
9048 				   int nlflags)
9049 {
9050 	struct i40e_netdev_priv *np = netdev_priv(dev);
9051 	struct i40e_vsi *vsi = np->vsi;
9052 	struct i40e_pf *pf = vsi->back;
9053 	struct i40e_veb *veb = NULL;
9054 	int i;
9055 
9056 	/* Only for PF VSI for now */
9057 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9058 		return -EOPNOTSUPP;
9059 
9060 	/* Find the HW bridge for the PF VSI */
9061 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9062 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9063 			veb = pf->veb[i];
9064 	}
9065 
9066 	if (!veb)
9067 		return 0;
9068 
9069 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9070 				       0, 0, nlflags, filter_mask, NULL);
9071 }
9072 
9073 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
9074  * inner mac plus all inner ethertypes.
9075  */
9076 #define I40E_MAX_TUNNEL_HDR_LEN 128
9077 /**
9078  * i40e_features_check - Validate encapsulated packet conforms to limits
9079  * @skb: skb buff
9080  * @dev: This physical port's netdev
9081  * @features: Offload features that the stack believes apply
9082  **/
9083 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9084 					     struct net_device *dev,
9085 					     netdev_features_t features)
9086 {
9087 	if (skb->encapsulation &&
9088 	    ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
9089 	     I40E_MAX_TUNNEL_HDR_LEN))
9090 		return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9091 
9092 	return features;
9093 }
9094 
9095 static const struct net_device_ops i40e_netdev_ops = {
9096 	.ndo_open		= i40e_open,
9097 	.ndo_stop		= i40e_close,
9098 	.ndo_start_xmit		= i40e_lan_xmit_frame,
9099 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
9100 	.ndo_set_rx_mode	= i40e_set_rx_mode,
9101 	.ndo_validate_addr	= eth_validate_addr,
9102 	.ndo_set_mac_address	= i40e_set_mac,
9103 	.ndo_change_mtu		= i40e_change_mtu,
9104 	.ndo_do_ioctl		= i40e_ioctl,
9105 	.ndo_tx_timeout		= i40e_tx_timeout,
9106 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
9107 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
9108 #ifdef CONFIG_NET_POLL_CONTROLLER
9109 	.ndo_poll_controller	= i40e_netpoll,
9110 #endif
9111 	.ndo_setup_tc		= __i40e_setup_tc,
9112 #ifdef I40E_FCOE
9113 	.ndo_fcoe_enable	= i40e_fcoe_enable,
9114 	.ndo_fcoe_disable	= i40e_fcoe_disable,
9115 #endif
9116 	.ndo_set_features	= i40e_set_features,
9117 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
9118 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
9119 	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
9120 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
9121 	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
9122 	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
9123 	.ndo_set_vf_trust	= i40e_ndo_set_vf_trust,
9124 	.ndo_udp_tunnel_add	= i40e_udp_tunnel_add,
9125 	.ndo_udp_tunnel_del	= i40e_udp_tunnel_del,
9126 	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
9127 	.ndo_fdb_add		= i40e_ndo_fdb_add,
9128 	.ndo_features_check	= i40e_features_check,
9129 	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
9130 	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
9131 };
9132 
9133 /**
9134  * i40e_config_netdev - Setup the netdev flags
9135  * @vsi: the VSI being configured
9136  *
9137  * Returns 0 on success, negative value on failure
9138  **/
9139 static int i40e_config_netdev(struct i40e_vsi *vsi)
9140 {
9141 	struct i40e_pf *pf = vsi->back;
9142 	struct i40e_hw *hw = &pf->hw;
9143 	struct i40e_netdev_priv *np;
9144 	struct net_device *netdev;
9145 	u8 mac_addr[ETH_ALEN];
9146 	int etherdev_size;
9147 
9148 	etherdev_size = sizeof(struct i40e_netdev_priv);
9149 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9150 	if (!netdev)
9151 		return -ENOMEM;
9152 
9153 	vsi->netdev = netdev;
9154 	np = netdev_priv(netdev);
9155 	np->vsi = vsi;
9156 
9157 	netdev->hw_enc_features |= NETIF_F_SG			|
9158 				   NETIF_F_IP_CSUM		|
9159 				   NETIF_F_IPV6_CSUM		|
9160 				   NETIF_F_HIGHDMA		|
9161 				   NETIF_F_SOFT_FEATURES	|
9162 				   NETIF_F_TSO			|
9163 				   NETIF_F_TSO_ECN		|
9164 				   NETIF_F_TSO6			|
9165 				   NETIF_F_GSO_GRE		|
9166 				   NETIF_F_GSO_GRE_CSUM		|
9167 				   NETIF_F_GSO_IPXIP4		|
9168 				   NETIF_F_GSO_IPXIP6		|
9169 				   NETIF_F_GSO_UDP_TUNNEL	|
9170 				   NETIF_F_GSO_UDP_TUNNEL_CSUM	|
9171 				   NETIF_F_GSO_PARTIAL		|
9172 				   NETIF_F_SCTP_CRC		|
9173 				   NETIF_F_RXHASH		|
9174 				   NETIF_F_RXCSUM		|
9175 				   0;
9176 
9177 	if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9178 		netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9179 
9180 	netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9181 
9182 	/* record features VLANs can make use of */
9183 	netdev->vlan_features |= netdev->hw_enc_features |
9184 				 NETIF_F_TSO_MANGLEID;
9185 
9186 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9187 		netdev->hw_features |= NETIF_F_NTUPLE;
9188 
9189 	netdev->hw_features |= netdev->hw_enc_features	|
9190 			       NETIF_F_HW_VLAN_CTAG_TX	|
9191 			       NETIF_F_HW_VLAN_CTAG_RX;
9192 
9193 	netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9194 	netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9195 
9196 	if (vsi->type == I40E_VSI_MAIN) {
9197 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9198 		ether_addr_copy(mac_addr, hw->mac.perm_addr);
9199 		spin_lock_bh(&vsi->mac_filter_hash_lock);
9200 		i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY);
9201 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
9202 	} else {
9203 		/* relate the VSI_VMDQ name to the VSI_MAIN name */
9204 		snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9205 			 pf->vsi[pf->lan_vsi]->netdev->name);
9206 		random_ether_addr(mac_addr);
9207 
9208 		spin_lock_bh(&vsi->mac_filter_hash_lock);
9209 		i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY);
9210 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
9211 	}
9212 
9213 	ether_addr_copy(netdev->dev_addr, mac_addr);
9214 	ether_addr_copy(netdev->perm_addr, mac_addr);
9215 
9216 	netdev->priv_flags |= IFF_UNICAST_FLT;
9217 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9218 	/* Setup netdev TC information */
9219 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9220 
9221 	netdev->netdev_ops = &i40e_netdev_ops;
9222 	netdev->watchdog_timeo = 5 * HZ;
9223 	i40e_set_ethtool_ops(netdev);
9224 #ifdef I40E_FCOE
9225 	i40e_fcoe_config_netdev(netdev, vsi);
9226 #endif
9227 
9228 	/* MTU range: 68 - 9706 */
9229 	netdev->min_mtu = ETH_MIN_MTU;
9230 	netdev->max_mtu = I40E_MAX_RXBUFFER -
9231 			  (ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN);
9232 
9233 	return 0;
9234 }
9235 
9236 /**
9237  * i40e_vsi_delete - Delete a VSI from the switch
9238  * @vsi: the VSI being removed
9239  *
9240  * Returns 0 on success, negative value on failure
9241  **/
9242 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9243 {
9244 	/* remove default VSI is not allowed */
9245 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9246 		return;
9247 
9248 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9249 }
9250 
9251 /**
9252  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9253  * @vsi: the VSI being queried
9254  *
9255  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9256  **/
9257 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9258 {
9259 	struct i40e_veb *veb;
9260 	struct i40e_pf *pf = vsi->back;
9261 
9262 	/* Uplink is not a bridge so default to VEB */
9263 	if (vsi->veb_idx == I40E_NO_VEB)
9264 		return 1;
9265 
9266 	veb = pf->veb[vsi->veb_idx];
9267 	if (!veb) {
9268 		dev_info(&pf->pdev->dev,
9269 			 "There is no veb associated with the bridge\n");
9270 		return -ENOENT;
9271 	}
9272 
9273 	/* Uplink is a bridge in VEPA mode */
9274 	if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9275 		return 0;
9276 	} else {
9277 		/* Uplink is a bridge in VEB mode */
9278 		return 1;
9279 	}
9280 
9281 	/* VEPA is now default bridge, so return 0 */
9282 	return 0;
9283 }
9284 
9285 /**
9286  * i40e_add_vsi - Add a VSI to the switch
9287  * @vsi: the VSI being configured
9288  *
9289  * This initializes a VSI context depending on the VSI type to be added and
9290  * passes it down to the add_vsi aq command.
9291  **/
9292 static int i40e_add_vsi(struct i40e_vsi *vsi)
9293 {
9294 	int ret = -ENODEV;
9295 	i40e_status aq_ret = 0;
9296 	struct i40e_pf *pf = vsi->back;
9297 	struct i40e_hw *hw = &pf->hw;
9298 	struct i40e_vsi_context ctxt;
9299 	struct i40e_mac_filter *f;
9300 	struct hlist_node *h;
9301 	int bkt;
9302 
9303 	u8 enabled_tc = 0x1; /* TC0 enabled */
9304 	int f_count = 0;
9305 
9306 	memset(&ctxt, 0, sizeof(ctxt));
9307 	switch (vsi->type) {
9308 	case I40E_VSI_MAIN:
9309 		/* The PF's main VSI is already setup as part of the
9310 		 * device initialization, so we'll not bother with
9311 		 * the add_vsi call, but we will retrieve the current
9312 		 * VSI context.
9313 		 */
9314 		ctxt.seid = pf->main_vsi_seid;
9315 		ctxt.pf_num = pf->hw.pf_id;
9316 		ctxt.vf_num = 0;
9317 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9318 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9319 		if (ret) {
9320 			dev_info(&pf->pdev->dev,
9321 				 "couldn't get PF vsi config, err %s aq_err %s\n",
9322 				 i40e_stat_str(&pf->hw, ret),
9323 				 i40e_aq_str(&pf->hw,
9324 					     pf->hw.aq.asq_last_status));
9325 			return -ENOENT;
9326 		}
9327 		vsi->info = ctxt.info;
9328 		vsi->info.valid_sections = 0;
9329 
9330 		vsi->seid = ctxt.seid;
9331 		vsi->id = ctxt.vsi_number;
9332 
9333 		enabled_tc = i40e_pf_get_tc_map(pf);
9334 
9335 		/* MFP mode setup queue map and update VSI */
9336 		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9337 		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9338 			memset(&ctxt, 0, sizeof(ctxt));
9339 			ctxt.seid = pf->main_vsi_seid;
9340 			ctxt.pf_num = pf->hw.pf_id;
9341 			ctxt.vf_num = 0;
9342 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9343 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9344 			if (ret) {
9345 				dev_info(&pf->pdev->dev,
9346 					 "update vsi failed, err %s aq_err %s\n",
9347 					 i40e_stat_str(&pf->hw, ret),
9348 					 i40e_aq_str(&pf->hw,
9349 						    pf->hw.aq.asq_last_status));
9350 				ret = -ENOENT;
9351 				goto err;
9352 			}
9353 			/* update the local VSI info queue map */
9354 			i40e_vsi_update_queue_map(vsi, &ctxt);
9355 			vsi->info.valid_sections = 0;
9356 		} else {
9357 			/* Default/Main VSI is only enabled for TC0
9358 			 * reconfigure it to enable all TCs that are
9359 			 * available on the port in SFP mode.
9360 			 * For MFP case the iSCSI PF would use this
9361 			 * flow to enable LAN+iSCSI TC.
9362 			 */
9363 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
9364 			if (ret) {
9365 				dev_info(&pf->pdev->dev,
9366 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9367 					 enabled_tc,
9368 					 i40e_stat_str(&pf->hw, ret),
9369 					 i40e_aq_str(&pf->hw,
9370 						    pf->hw.aq.asq_last_status));
9371 				ret = -ENOENT;
9372 			}
9373 		}
9374 		break;
9375 
9376 	case I40E_VSI_FDIR:
9377 		ctxt.pf_num = hw->pf_id;
9378 		ctxt.vf_num = 0;
9379 		ctxt.uplink_seid = vsi->uplink_seid;
9380 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9381 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9382 		if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9383 		    (i40e_is_vsi_uplink_mode_veb(vsi))) {
9384 			ctxt.info.valid_sections |=
9385 			     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9386 			ctxt.info.switch_id =
9387 			   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9388 		}
9389 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9390 		break;
9391 
9392 	case I40E_VSI_VMDQ2:
9393 		ctxt.pf_num = hw->pf_id;
9394 		ctxt.vf_num = 0;
9395 		ctxt.uplink_seid = vsi->uplink_seid;
9396 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9397 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9398 
9399 		/* This VSI is connected to VEB so the switch_id
9400 		 * should be set to zero by default.
9401 		 */
9402 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9403 			ctxt.info.valid_sections |=
9404 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9405 			ctxt.info.switch_id =
9406 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9407 		}
9408 
9409 		/* Setup the VSI tx/rx queue map for TC0 only for now */
9410 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9411 		break;
9412 
9413 	case I40E_VSI_SRIOV:
9414 		ctxt.pf_num = hw->pf_id;
9415 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9416 		ctxt.uplink_seid = vsi->uplink_seid;
9417 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9418 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9419 
9420 		/* This VSI is connected to VEB so the switch_id
9421 		 * should be set to zero by default.
9422 		 */
9423 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9424 			ctxt.info.valid_sections |=
9425 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9426 			ctxt.info.switch_id =
9427 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9428 		}
9429 
9430 		if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9431 			ctxt.info.valid_sections |=
9432 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9433 			ctxt.info.queueing_opt_flags |=
9434 				(I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9435 				 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9436 		}
9437 
9438 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9439 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9440 		if (pf->vf[vsi->vf_id].spoofchk) {
9441 			ctxt.info.valid_sections |=
9442 				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9443 			ctxt.info.sec_flags |=
9444 				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9445 				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9446 		}
9447 		/* Setup the VSI tx/rx queue map for TC0 only for now */
9448 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9449 		break;
9450 
9451 #ifdef I40E_FCOE
9452 	case I40E_VSI_FCOE:
9453 		ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9454 		if (ret) {
9455 			dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9456 			return ret;
9457 		}
9458 		break;
9459 
9460 #endif /* I40E_FCOE */
9461 	case I40E_VSI_IWARP:
9462 		/* send down message to iWARP */
9463 		break;
9464 
9465 	default:
9466 		return -ENODEV;
9467 	}
9468 
9469 	if (vsi->type != I40E_VSI_MAIN) {
9470 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9471 		if (ret) {
9472 			dev_info(&vsi->back->pdev->dev,
9473 				 "add vsi failed, err %s aq_err %s\n",
9474 				 i40e_stat_str(&pf->hw, ret),
9475 				 i40e_aq_str(&pf->hw,
9476 					     pf->hw.aq.asq_last_status));
9477 			ret = -ENOENT;
9478 			goto err;
9479 		}
9480 		vsi->info = ctxt.info;
9481 		vsi->info.valid_sections = 0;
9482 		vsi->seid = ctxt.seid;
9483 		vsi->id = ctxt.vsi_number;
9484 	}
9485 	/* Except FDIR VSI, for all othet VSI set the broadcast filter */
9486 	if (vsi->type != I40E_VSI_FDIR) {
9487 		aq_ret = i40e_aq_set_vsi_broadcast(hw, vsi->seid, true, NULL);
9488 		if (aq_ret) {
9489 			ret = i40e_aq_rc_to_posix(aq_ret,
9490 						  hw->aq.asq_last_status);
9491 			dev_info(&pf->pdev->dev,
9492 				 "set brdcast promisc failed, err %s, aq_err %s\n",
9493 				 i40e_stat_str(hw, aq_ret),
9494 				 i40e_aq_str(hw, hw->aq.asq_last_status));
9495 		}
9496 	}
9497 
9498 	vsi->active_filters = 0;
9499 	clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
9500 	spin_lock_bh(&vsi->mac_filter_hash_lock);
9501 	/* If macvlan filters already exist, force them to get loaded */
9502 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
9503 		f->state = I40E_FILTER_NEW;
9504 		f_count++;
9505 	}
9506 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
9507 
9508 	if (f_count) {
9509 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9510 		pf->flags |= I40E_FLAG_FILTER_SYNC;
9511 	}
9512 
9513 	/* Update VSI BW information */
9514 	ret = i40e_vsi_get_bw_info(vsi);
9515 	if (ret) {
9516 		dev_info(&pf->pdev->dev,
9517 			 "couldn't get vsi bw info, err %s aq_err %s\n",
9518 			 i40e_stat_str(&pf->hw, ret),
9519 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9520 		/* VSI is already added so not tearing that up */
9521 		ret = 0;
9522 	}
9523 
9524 err:
9525 	return ret;
9526 }
9527 
9528 /**
9529  * i40e_vsi_release - Delete a VSI and free its resources
9530  * @vsi: the VSI being removed
9531  *
9532  * Returns 0 on success or < 0 on error
9533  **/
9534 int i40e_vsi_release(struct i40e_vsi *vsi)
9535 {
9536 	struct i40e_mac_filter *f;
9537 	struct hlist_node *h;
9538 	struct i40e_veb *veb = NULL;
9539 	struct i40e_pf *pf;
9540 	u16 uplink_seid;
9541 	int i, n, bkt;
9542 
9543 	pf = vsi->back;
9544 
9545 	/* release of a VEB-owner or last VSI is not allowed */
9546 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9547 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9548 			 vsi->seid, vsi->uplink_seid);
9549 		return -ENODEV;
9550 	}
9551 	if (vsi == pf->vsi[pf->lan_vsi] &&
9552 	    !test_bit(__I40E_DOWN, &pf->state)) {
9553 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9554 		return -ENODEV;
9555 	}
9556 
9557 	uplink_seid = vsi->uplink_seid;
9558 	if (vsi->type != I40E_VSI_SRIOV) {
9559 		if (vsi->netdev_registered) {
9560 			vsi->netdev_registered = false;
9561 			if (vsi->netdev) {
9562 				/* results in a call to i40e_close() */
9563 				unregister_netdev(vsi->netdev);
9564 			}
9565 		} else {
9566 			i40e_vsi_close(vsi);
9567 		}
9568 		i40e_vsi_disable_irq(vsi);
9569 	}
9570 
9571 	spin_lock_bh(&vsi->mac_filter_hash_lock);
9572 
9573 	/* clear the sync flag on all filters */
9574 	if (vsi->netdev) {
9575 		__dev_uc_unsync(vsi->netdev, NULL);
9576 		__dev_mc_unsync(vsi->netdev, NULL);
9577 	}
9578 
9579 	/* make sure any remaining filters are marked for deletion */
9580 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
9581 		__i40e_del_filter(vsi, f);
9582 
9583 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
9584 
9585 	i40e_sync_vsi_filters(vsi);
9586 
9587 	i40e_vsi_delete(vsi);
9588 	i40e_vsi_free_q_vectors(vsi);
9589 	if (vsi->netdev) {
9590 		free_netdev(vsi->netdev);
9591 		vsi->netdev = NULL;
9592 	}
9593 	i40e_vsi_clear_rings(vsi);
9594 	i40e_vsi_clear(vsi);
9595 
9596 	/* If this was the last thing on the VEB, except for the
9597 	 * controlling VSI, remove the VEB, which puts the controlling
9598 	 * VSI onto the next level down in the switch.
9599 	 *
9600 	 * Well, okay, there's one more exception here: don't remove
9601 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
9602 	 * from up the network stack.
9603 	 */
9604 	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9605 		if (pf->vsi[i] &&
9606 		    pf->vsi[i]->uplink_seid == uplink_seid &&
9607 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9608 			n++;      /* count the VSIs */
9609 		}
9610 	}
9611 	for (i = 0; i < I40E_MAX_VEB; i++) {
9612 		if (!pf->veb[i])
9613 			continue;
9614 		if (pf->veb[i]->uplink_seid == uplink_seid)
9615 			n++;     /* count the VEBs */
9616 		if (pf->veb[i]->seid == uplink_seid)
9617 			veb = pf->veb[i];
9618 	}
9619 	if (n == 0 && veb && veb->uplink_seid != 0)
9620 		i40e_veb_release(veb);
9621 
9622 	return 0;
9623 }
9624 
9625 /**
9626  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9627  * @vsi: ptr to the VSI
9628  *
9629  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9630  * corresponding SW VSI structure and initializes num_queue_pairs for the
9631  * newly allocated VSI.
9632  *
9633  * Returns 0 on success or negative on failure
9634  **/
9635 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9636 {
9637 	int ret = -ENOENT;
9638 	struct i40e_pf *pf = vsi->back;
9639 
9640 	if (vsi->q_vectors[0]) {
9641 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9642 			 vsi->seid);
9643 		return -EEXIST;
9644 	}
9645 
9646 	if (vsi->base_vector) {
9647 		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9648 			 vsi->seid, vsi->base_vector);
9649 		return -EEXIST;
9650 	}
9651 
9652 	ret = i40e_vsi_alloc_q_vectors(vsi);
9653 	if (ret) {
9654 		dev_info(&pf->pdev->dev,
9655 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9656 			 vsi->num_q_vectors, vsi->seid, ret);
9657 		vsi->num_q_vectors = 0;
9658 		goto vector_setup_out;
9659 	}
9660 
9661 	/* In Legacy mode, we do not have to get any other vector since we
9662 	 * piggyback on the misc/ICR0 for queue interrupts.
9663 	*/
9664 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9665 		return ret;
9666 	if (vsi->num_q_vectors)
9667 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9668 						 vsi->num_q_vectors, vsi->idx);
9669 	if (vsi->base_vector < 0) {
9670 		dev_info(&pf->pdev->dev,
9671 			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9672 			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9673 		i40e_vsi_free_q_vectors(vsi);
9674 		ret = -ENOENT;
9675 		goto vector_setup_out;
9676 	}
9677 
9678 vector_setup_out:
9679 	return ret;
9680 }
9681 
9682 /**
9683  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9684  * @vsi: pointer to the vsi.
9685  *
9686  * This re-allocates a vsi's queue resources.
9687  *
9688  * Returns pointer to the successfully allocated and configured VSI sw struct
9689  * on success, otherwise returns NULL on failure.
9690  **/
9691 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9692 {
9693 	struct i40e_pf *pf;
9694 	u8 enabled_tc;
9695 	int ret;
9696 
9697 	if (!vsi)
9698 		return NULL;
9699 
9700 	pf = vsi->back;
9701 
9702 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9703 	i40e_vsi_clear_rings(vsi);
9704 
9705 	i40e_vsi_free_arrays(vsi, false);
9706 	i40e_set_num_rings_in_vsi(vsi);
9707 	ret = i40e_vsi_alloc_arrays(vsi, false);
9708 	if (ret)
9709 		goto err_vsi;
9710 
9711 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9712 	if (ret < 0) {
9713 		dev_info(&pf->pdev->dev,
9714 			 "failed to get tracking for %d queues for VSI %d err %d\n",
9715 			 vsi->alloc_queue_pairs, vsi->seid, ret);
9716 		goto err_vsi;
9717 	}
9718 	vsi->base_queue = ret;
9719 
9720 	/* Update the FW view of the VSI. Force a reset of TC and queue
9721 	 * layout configurations.
9722 	 */
9723 	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9724 	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9725 	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9726 	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9727 
9728 	/* assign it some queues */
9729 	ret = i40e_alloc_rings(vsi);
9730 	if (ret)
9731 		goto err_rings;
9732 
9733 	/* map all of the rings to the q_vectors */
9734 	i40e_vsi_map_rings_to_vectors(vsi);
9735 	return vsi;
9736 
9737 err_rings:
9738 	i40e_vsi_free_q_vectors(vsi);
9739 	if (vsi->netdev_registered) {
9740 		vsi->netdev_registered = false;
9741 		unregister_netdev(vsi->netdev);
9742 		free_netdev(vsi->netdev);
9743 		vsi->netdev = NULL;
9744 	}
9745 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9746 err_vsi:
9747 	i40e_vsi_clear(vsi);
9748 	return NULL;
9749 }
9750 
9751 /**
9752  * i40e_vsi_setup - Set up a VSI by a given type
9753  * @pf: board private structure
9754  * @type: VSI type
9755  * @uplink_seid: the switch element to link to
9756  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9757  *
9758  * This allocates the sw VSI structure and its queue resources, then add a VSI
9759  * to the identified VEB.
9760  *
9761  * Returns pointer to the successfully allocated and configure VSI sw struct on
9762  * success, otherwise returns NULL on failure.
9763  **/
9764 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9765 				u16 uplink_seid, u32 param1)
9766 {
9767 	struct i40e_vsi *vsi = NULL;
9768 	struct i40e_veb *veb = NULL;
9769 	int ret, i;
9770 	int v_idx;
9771 
9772 	/* The requested uplink_seid must be either
9773 	 *     - the PF's port seid
9774 	 *              no VEB is needed because this is the PF
9775 	 *              or this is a Flow Director special case VSI
9776 	 *     - seid of an existing VEB
9777 	 *     - seid of a VSI that owns an existing VEB
9778 	 *     - seid of a VSI that doesn't own a VEB
9779 	 *              a new VEB is created and the VSI becomes the owner
9780 	 *     - seid of the PF VSI, which is what creates the first VEB
9781 	 *              this is a special case of the previous
9782 	 *
9783 	 * Find which uplink_seid we were given and create a new VEB if needed
9784 	 */
9785 	for (i = 0; i < I40E_MAX_VEB; i++) {
9786 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9787 			veb = pf->veb[i];
9788 			break;
9789 		}
9790 	}
9791 
9792 	if (!veb && uplink_seid != pf->mac_seid) {
9793 
9794 		for (i = 0; i < pf->num_alloc_vsi; i++) {
9795 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9796 				vsi = pf->vsi[i];
9797 				break;
9798 			}
9799 		}
9800 		if (!vsi) {
9801 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9802 				 uplink_seid);
9803 			return NULL;
9804 		}
9805 
9806 		if (vsi->uplink_seid == pf->mac_seid)
9807 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9808 					     vsi->tc_config.enabled_tc);
9809 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9810 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9811 					     vsi->tc_config.enabled_tc);
9812 		if (veb) {
9813 			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9814 				dev_info(&vsi->back->pdev->dev,
9815 					 "New VSI creation error, uplink seid of LAN VSI expected.\n");
9816 				return NULL;
9817 			}
9818 			/* We come up by default in VEPA mode if SRIOV is not
9819 			 * already enabled, in which case we can't force VEPA
9820 			 * mode.
9821 			 */
9822 			if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9823 				veb->bridge_mode = BRIDGE_MODE_VEPA;
9824 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9825 			}
9826 			i40e_config_bridge_mode(veb);
9827 		}
9828 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9829 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9830 				veb = pf->veb[i];
9831 		}
9832 		if (!veb) {
9833 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9834 			return NULL;
9835 		}
9836 
9837 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9838 		uplink_seid = veb->seid;
9839 	}
9840 
9841 	/* get vsi sw struct */
9842 	v_idx = i40e_vsi_mem_alloc(pf, type);
9843 	if (v_idx < 0)
9844 		goto err_alloc;
9845 	vsi = pf->vsi[v_idx];
9846 	if (!vsi)
9847 		goto err_alloc;
9848 	vsi->type = type;
9849 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9850 
9851 	if (type == I40E_VSI_MAIN)
9852 		pf->lan_vsi = v_idx;
9853 	else if (type == I40E_VSI_SRIOV)
9854 		vsi->vf_id = param1;
9855 	/* assign it some queues */
9856 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9857 				vsi->idx);
9858 	if (ret < 0) {
9859 		dev_info(&pf->pdev->dev,
9860 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
9861 			 vsi->alloc_queue_pairs, vsi->seid, ret);
9862 		goto err_vsi;
9863 	}
9864 	vsi->base_queue = ret;
9865 
9866 	/* get a VSI from the hardware */
9867 	vsi->uplink_seid = uplink_seid;
9868 	ret = i40e_add_vsi(vsi);
9869 	if (ret)
9870 		goto err_vsi;
9871 
9872 	switch (vsi->type) {
9873 	/* setup the netdev if needed */
9874 	case I40E_VSI_MAIN:
9875 		/* Apply relevant filters if a platform-specific mac
9876 		 * address was selected.
9877 		 */
9878 		if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9879 			ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9880 			if (ret) {
9881 				dev_warn(&pf->pdev->dev,
9882 					 "could not set up macaddr; err %d\n",
9883 					 ret);
9884 			}
9885 		}
9886 	case I40E_VSI_VMDQ2:
9887 	case I40E_VSI_FCOE:
9888 		ret = i40e_config_netdev(vsi);
9889 		if (ret)
9890 			goto err_netdev;
9891 		ret = register_netdev(vsi->netdev);
9892 		if (ret)
9893 			goto err_netdev;
9894 		vsi->netdev_registered = true;
9895 		netif_carrier_off(vsi->netdev);
9896 #ifdef CONFIG_I40E_DCB
9897 		/* Setup DCB netlink interface */
9898 		i40e_dcbnl_setup(vsi);
9899 #endif /* CONFIG_I40E_DCB */
9900 		/* fall through */
9901 
9902 	case I40E_VSI_FDIR:
9903 		/* set up vectors and rings if needed */
9904 		ret = i40e_vsi_setup_vectors(vsi);
9905 		if (ret)
9906 			goto err_msix;
9907 
9908 		ret = i40e_alloc_rings(vsi);
9909 		if (ret)
9910 			goto err_rings;
9911 
9912 		/* map all of the rings to the q_vectors */
9913 		i40e_vsi_map_rings_to_vectors(vsi);
9914 
9915 		i40e_vsi_reset_stats(vsi);
9916 		break;
9917 
9918 	default:
9919 		/* no netdev or rings for the other VSI types */
9920 		break;
9921 	}
9922 
9923 	if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9924 	    (vsi->type == I40E_VSI_VMDQ2)) {
9925 		ret = i40e_vsi_config_rss(vsi);
9926 	}
9927 	return vsi;
9928 
9929 err_rings:
9930 	i40e_vsi_free_q_vectors(vsi);
9931 err_msix:
9932 	if (vsi->netdev_registered) {
9933 		vsi->netdev_registered = false;
9934 		unregister_netdev(vsi->netdev);
9935 		free_netdev(vsi->netdev);
9936 		vsi->netdev = NULL;
9937 	}
9938 err_netdev:
9939 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9940 err_vsi:
9941 	i40e_vsi_clear(vsi);
9942 err_alloc:
9943 	return NULL;
9944 }
9945 
9946 /**
9947  * i40e_veb_get_bw_info - Query VEB BW information
9948  * @veb: the veb to query
9949  *
9950  * Query the Tx scheduler BW configuration data for given VEB
9951  **/
9952 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9953 {
9954 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9955 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9956 	struct i40e_pf *pf = veb->pf;
9957 	struct i40e_hw *hw = &pf->hw;
9958 	u32 tc_bw_max;
9959 	int ret = 0;
9960 	int i;
9961 
9962 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9963 						  &bw_data, NULL);
9964 	if (ret) {
9965 		dev_info(&pf->pdev->dev,
9966 			 "query veb bw config failed, err %s aq_err %s\n",
9967 			 i40e_stat_str(&pf->hw, ret),
9968 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9969 		goto out;
9970 	}
9971 
9972 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9973 						   &ets_data, NULL);
9974 	if (ret) {
9975 		dev_info(&pf->pdev->dev,
9976 			 "query veb bw ets config failed, err %s aq_err %s\n",
9977 			 i40e_stat_str(&pf->hw, ret),
9978 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9979 		goto out;
9980 	}
9981 
9982 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9983 	veb->bw_max_quanta = ets_data.tc_bw_max;
9984 	veb->is_abs_credits = bw_data.absolute_credits_enable;
9985 	veb->enabled_tc = ets_data.tc_valid_bits;
9986 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9987 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9988 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9989 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9990 		veb->bw_tc_limit_credits[i] =
9991 					le16_to_cpu(bw_data.tc_bw_limits[i]);
9992 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9993 	}
9994 
9995 out:
9996 	return ret;
9997 }
9998 
9999 /**
10000  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
10001  * @pf: board private structure
10002  *
10003  * On error: returns error code (negative)
10004  * On success: returns vsi index in PF (positive)
10005  **/
10006 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
10007 {
10008 	int ret = -ENOENT;
10009 	struct i40e_veb *veb;
10010 	int i;
10011 
10012 	/* Need to protect the allocation of switch elements at the PF level */
10013 	mutex_lock(&pf->switch_mutex);
10014 
10015 	/* VEB list may be fragmented if VEB creation/destruction has
10016 	 * been happening.  We can afford to do a quick scan to look
10017 	 * for any free slots in the list.
10018 	 *
10019 	 * find next empty veb slot, looping back around if necessary
10020 	 */
10021 	i = 0;
10022 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
10023 		i++;
10024 	if (i >= I40E_MAX_VEB) {
10025 		ret = -ENOMEM;
10026 		goto err_alloc_veb;  /* out of VEB slots! */
10027 	}
10028 
10029 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
10030 	if (!veb) {
10031 		ret = -ENOMEM;
10032 		goto err_alloc_veb;
10033 	}
10034 	veb->pf = pf;
10035 	veb->idx = i;
10036 	veb->enabled_tc = 1;
10037 
10038 	pf->veb[i] = veb;
10039 	ret = i;
10040 err_alloc_veb:
10041 	mutex_unlock(&pf->switch_mutex);
10042 	return ret;
10043 }
10044 
10045 /**
10046  * i40e_switch_branch_release - Delete a branch of the switch tree
10047  * @branch: where to start deleting
10048  *
10049  * This uses recursion to find the tips of the branch to be
10050  * removed, deleting until we get back to and can delete this VEB.
10051  **/
10052 static void i40e_switch_branch_release(struct i40e_veb *branch)
10053 {
10054 	struct i40e_pf *pf = branch->pf;
10055 	u16 branch_seid = branch->seid;
10056 	u16 veb_idx = branch->idx;
10057 	int i;
10058 
10059 	/* release any VEBs on this VEB - RECURSION */
10060 	for (i = 0; i < I40E_MAX_VEB; i++) {
10061 		if (!pf->veb[i])
10062 			continue;
10063 		if (pf->veb[i]->uplink_seid == branch->seid)
10064 			i40e_switch_branch_release(pf->veb[i]);
10065 	}
10066 
10067 	/* Release the VSIs on this VEB, but not the owner VSI.
10068 	 *
10069 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10070 	 *       the VEB itself, so don't use (*branch) after this loop.
10071 	 */
10072 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10073 		if (!pf->vsi[i])
10074 			continue;
10075 		if (pf->vsi[i]->uplink_seid == branch_seid &&
10076 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10077 			i40e_vsi_release(pf->vsi[i]);
10078 		}
10079 	}
10080 
10081 	/* There's one corner case where the VEB might not have been
10082 	 * removed, so double check it here and remove it if needed.
10083 	 * This case happens if the veb was created from the debugfs
10084 	 * commands and no VSIs were added to it.
10085 	 */
10086 	if (pf->veb[veb_idx])
10087 		i40e_veb_release(pf->veb[veb_idx]);
10088 }
10089 
10090 /**
10091  * i40e_veb_clear - remove veb struct
10092  * @veb: the veb to remove
10093  **/
10094 static void i40e_veb_clear(struct i40e_veb *veb)
10095 {
10096 	if (!veb)
10097 		return;
10098 
10099 	if (veb->pf) {
10100 		struct i40e_pf *pf = veb->pf;
10101 
10102 		mutex_lock(&pf->switch_mutex);
10103 		if (pf->veb[veb->idx] == veb)
10104 			pf->veb[veb->idx] = NULL;
10105 		mutex_unlock(&pf->switch_mutex);
10106 	}
10107 
10108 	kfree(veb);
10109 }
10110 
10111 /**
10112  * i40e_veb_release - Delete a VEB and free its resources
10113  * @veb: the VEB being removed
10114  **/
10115 void i40e_veb_release(struct i40e_veb *veb)
10116 {
10117 	struct i40e_vsi *vsi = NULL;
10118 	struct i40e_pf *pf;
10119 	int i, n = 0;
10120 
10121 	pf = veb->pf;
10122 
10123 	/* find the remaining VSI and check for extras */
10124 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10125 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10126 			n++;
10127 			vsi = pf->vsi[i];
10128 		}
10129 	}
10130 	if (n != 1) {
10131 		dev_info(&pf->pdev->dev,
10132 			 "can't remove VEB %d with %d VSIs left\n",
10133 			 veb->seid, n);
10134 		return;
10135 	}
10136 
10137 	/* move the remaining VSI to uplink veb */
10138 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10139 	if (veb->uplink_seid) {
10140 		vsi->uplink_seid = veb->uplink_seid;
10141 		if (veb->uplink_seid == pf->mac_seid)
10142 			vsi->veb_idx = I40E_NO_VEB;
10143 		else
10144 			vsi->veb_idx = veb->veb_idx;
10145 	} else {
10146 		/* floating VEB */
10147 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10148 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10149 	}
10150 
10151 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10152 	i40e_veb_clear(veb);
10153 }
10154 
10155 /**
10156  * i40e_add_veb - create the VEB in the switch
10157  * @veb: the VEB to be instantiated
10158  * @vsi: the controlling VSI
10159  **/
10160 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10161 {
10162 	struct i40e_pf *pf = veb->pf;
10163 	bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10164 	int ret;
10165 
10166 	ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10167 			      veb->enabled_tc, false,
10168 			      &veb->seid, enable_stats, NULL);
10169 
10170 	/* get a VEB from the hardware */
10171 	if (ret) {
10172 		dev_info(&pf->pdev->dev,
10173 			 "couldn't add VEB, err %s aq_err %s\n",
10174 			 i40e_stat_str(&pf->hw, ret),
10175 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10176 		return -EPERM;
10177 	}
10178 
10179 	/* get statistics counter */
10180 	ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10181 					 &veb->stats_idx, NULL, NULL, NULL);
10182 	if (ret) {
10183 		dev_info(&pf->pdev->dev,
10184 			 "couldn't get VEB statistics idx, err %s aq_err %s\n",
10185 			 i40e_stat_str(&pf->hw, ret),
10186 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10187 		return -EPERM;
10188 	}
10189 	ret = i40e_veb_get_bw_info(veb);
10190 	if (ret) {
10191 		dev_info(&pf->pdev->dev,
10192 			 "couldn't get VEB bw info, err %s aq_err %s\n",
10193 			 i40e_stat_str(&pf->hw, ret),
10194 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10195 		i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10196 		return -ENOENT;
10197 	}
10198 
10199 	vsi->uplink_seid = veb->seid;
10200 	vsi->veb_idx = veb->idx;
10201 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10202 
10203 	return 0;
10204 }
10205 
10206 /**
10207  * i40e_veb_setup - Set up a VEB
10208  * @pf: board private structure
10209  * @flags: VEB setup flags
10210  * @uplink_seid: the switch element to link to
10211  * @vsi_seid: the initial VSI seid
10212  * @enabled_tc: Enabled TC bit-map
10213  *
10214  * This allocates the sw VEB structure and links it into the switch
10215  * It is possible and legal for this to be a duplicate of an already
10216  * existing VEB.  It is also possible for both uplink and vsi seids
10217  * to be zero, in order to create a floating VEB.
10218  *
10219  * Returns pointer to the successfully allocated VEB sw struct on
10220  * success, otherwise returns NULL on failure.
10221  **/
10222 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10223 				u16 uplink_seid, u16 vsi_seid,
10224 				u8 enabled_tc)
10225 {
10226 	struct i40e_veb *veb, *uplink_veb = NULL;
10227 	int vsi_idx, veb_idx;
10228 	int ret;
10229 
10230 	/* if one seid is 0, the other must be 0 to create a floating relay */
10231 	if ((uplink_seid == 0 || vsi_seid == 0) &&
10232 	    (uplink_seid + vsi_seid != 0)) {
10233 		dev_info(&pf->pdev->dev,
10234 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
10235 			 uplink_seid, vsi_seid);
10236 		return NULL;
10237 	}
10238 
10239 	/* make sure there is such a vsi and uplink */
10240 	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10241 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10242 			break;
10243 	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10244 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10245 			 vsi_seid);
10246 		return NULL;
10247 	}
10248 
10249 	if (uplink_seid && uplink_seid != pf->mac_seid) {
10250 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10251 			if (pf->veb[veb_idx] &&
10252 			    pf->veb[veb_idx]->seid == uplink_seid) {
10253 				uplink_veb = pf->veb[veb_idx];
10254 				break;
10255 			}
10256 		}
10257 		if (!uplink_veb) {
10258 			dev_info(&pf->pdev->dev,
10259 				 "uplink seid %d not found\n", uplink_seid);
10260 			return NULL;
10261 		}
10262 	}
10263 
10264 	/* get veb sw struct */
10265 	veb_idx = i40e_veb_mem_alloc(pf);
10266 	if (veb_idx < 0)
10267 		goto err_alloc;
10268 	veb = pf->veb[veb_idx];
10269 	veb->flags = flags;
10270 	veb->uplink_seid = uplink_seid;
10271 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10272 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10273 
10274 	/* create the VEB in the switch */
10275 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10276 	if (ret)
10277 		goto err_veb;
10278 	if (vsi_idx == pf->lan_vsi)
10279 		pf->lan_veb = veb->idx;
10280 
10281 	return veb;
10282 
10283 err_veb:
10284 	i40e_veb_clear(veb);
10285 err_alloc:
10286 	return NULL;
10287 }
10288 
10289 /**
10290  * i40e_setup_pf_switch_element - set PF vars based on switch type
10291  * @pf: board private structure
10292  * @ele: element we are building info from
10293  * @num_reported: total number of elements
10294  * @printconfig: should we print the contents
10295  *
10296  * helper function to assist in extracting a few useful SEID values.
10297  **/
10298 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10299 				struct i40e_aqc_switch_config_element_resp *ele,
10300 				u16 num_reported, bool printconfig)
10301 {
10302 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10303 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10304 	u8 element_type = ele->element_type;
10305 	u16 seid = le16_to_cpu(ele->seid);
10306 
10307 	if (printconfig)
10308 		dev_info(&pf->pdev->dev,
10309 			 "type=%d seid=%d uplink=%d downlink=%d\n",
10310 			 element_type, seid, uplink_seid, downlink_seid);
10311 
10312 	switch (element_type) {
10313 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
10314 		pf->mac_seid = seid;
10315 		break;
10316 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
10317 		/* Main VEB? */
10318 		if (uplink_seid != pf->mac_seid)
10319 			break;
10320 		if (pf->lan_veb == I40E_NO_VEB) {
10321 			int v;
10322 
10323 			/* find existing or else empty VEB */
10324 			for (v = 0; v < I40E_MAX_VEB; v++) {
10325 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10326 					pf->lan_veb = v;
10327 					break;
10328 				}
10329 			}
10330 			if (pf->lan_veb == I40E_NO_VEB) {
10331 				v = i40e_veb_mem_alloc(pf);
10332 				if (v < 0)
10333 					break;
10334 				pf->lan_veb = v;
10335 			}
10336 		}
10337 
10338 		pf->veb[pf->lan_veb]->seid = seid;
10339 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10340 		pf->veb[pf->lan_veb]->pf = pf;
10341 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10342 		break;
10343 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
10344 		if (num_reported != 1)
10345 			break;
10346 		/* This is immediately after a reset so we can assume this is
10347 		 * the PF's VSI
10348 		 */
10349 		pf->mac_seid = uplink_seid;
10350 		pf->pf_seid = downlink_seid;
10351 		pf->main_vsi_seid = seid;
10352 		if (printconfig)
10353 			dev_info(&pf->pdev->dev,
10354 				 "pf_seid=%d main_vsi_seid=%d\n",
10355 				 pf->pf_seid, pf->main_vsi_seid);
10356 		break;
10357 	case I40E_SWITCH_ELEMENT_TYPE_PF:
10358 	case I40E_SWITCH_ELEMENT_TYPE_VF:
10359 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
10360 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
10361 	case I40E_SWITCH_ELEMENT_TYPE_PE:
10362 	case I40E_SWITCH_ELEMENT_TYPE_PA:
10363 		/* ignore these for now */
10364 		break;
10365 	default:
10366 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10367 			 element_type, seid);
10368 		break;
10369 	}
10370 }
10371 
10372 /**
10373  * i40e_fetch_switch_configuration - Get switch config from firmware
10374  * @pf: board private structure
10375  * @printconfig: should we print the contents
10376  *
10377  * Get the current switch configuration from the device and
10378  * extract a few useful SEID values.
10379  **/
10380 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10381 {
10382 	struct i40e_aqc_get_switch_config_resp *sw_config;
10383 	u16 next_seid = 0;
10384 	int ret = 0;
10385 	u8 *aq_buf;
10386 	int i;
10387 
10388 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10389 	if (!aq_buf)
10390 		return -ENOMEM;
10391 
10392 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10393 	do {
10394 		u16 num_reported, num_total;
10395 
10396 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10397 						I40E_AQ_LARGE_BUF,
10398 						&next_seid, NULL);
10399 		if (ret) {
10400 			dev_info(&pf->pdev->dev,
10401 				 "get switch config failed err %s aq_err %s\n",
10402 				 i40e_stat_str(&pf->hw, ret),
10403 				 i40e_aq_str(&pf->hw,
10404 					     pf->hw.aq.asq_last_status));
10405 			kfree(aq_buf);
10406 			return -ENOENT;
10407 		}
10408 
10409 		num_reported = le16_to_cpu(sw_config->header.num_reported);
10410 		num_total = le16_to_cpu(sw_config->header.num_total);
10411 
10412 		if (printconfig)
10413 			dev_info(&pf->pdev->dev,
10414 				 "header: %d reported %d total\n",
10415 				 num_reported, num_total);
10416 
10417 		for (i = 0; i < num_reported; i++) {
10418 			struct i40e_aqc_switch_config_element_resp *ele =
10419 				&sw_config->element[i];
10420 
10421 			i40e_setup_pf_switch_element(pf, ele, num_reported,
10422 						     printconfig);
10423 		}
10424 	} while (next_seid != 0);
10425 
10426 	kfree(aq_buf);
10427 	return ret;
10428 }
10429 
10430 /**
10431  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10432  * @pf: board private structure
10433  * @reinit: if the Main VSI needs to re-initialized.
10434  *
10435  * Returns 0 on success, negative value on failure
10436  **/
10437 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10438 {
10439 	u16 flags = 0;
10440 	int ret;
10441 
10442 	/* find out what's out there already */
10443 	ret = i40e_fetch_switch_configuration(pf, false);
10444 	if (ret) {
10445 		dev_info(&pf->pdev->dev,
10446 			 "couldn't fetch switch config, err %s aq_err %s\n",
10447 			 i40e_stat_str(&pf->hw, ret),
10448 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10449 		return ret;
10450 	}
10451 	i40e_pf_reset_stats(pf);
10452 
10453 	/* set the switch config bit for the whole device to
10454 	 * support limited promisc or true promisc
10455 	 * when user requests promisc. The default is limited
10456 	 * promisc.
10457 	*/
10458 
10459 	if ((pf->hw.pf_id == 0) &&
10460 	    !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10461 		flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10462 
10463 	if (pf->hw.pf_id == 0) {
10464 		u16 valid_flags;
10465 
10466 		valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10467 		ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10468 						NULL);
10469 		if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10470 			dev_info(&pf->pdev->dev,
10471 				 "couldn't set switch config bits, err %s aq_err %s\n",
10472 				 i40e_stat_str(&pf->hw, ret),
10473 				 i40e_aq_str(&pf->hw,
10474 					     pf->hw.aq.asq_last_status));
10475 			/* not a fatal problem, just keep going */
10476 		}
10477 	}
10478 
10479 	/* first time setup */
10480 	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10481 		struct i40e_vsi *vsi = NULL;
10482 		u16 uplink_seid;
10483 
10484 		/* Set up the PF VSI associated with the PF's main VSI
10485 		 * that is already in the HW switch
10486 		 */
10487 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10488 			uplink_seid = pf->veb[pf->lan_veb]->seid;
10489 		else
10490 			uplink_seid = pf->mac_seid;
10491 		if (pf->lan_vsi == I40E_NO_VSI)
10492 			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10493 		else if (reinit)
10494 			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10495 		if (!vsi) {
10496 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10497 			i40e_fdir_teardown(pf);
10498 			return -EAGAIN;
10499 		}
10500 	} else {
10501 		/* force a reset of TC and queue layout configurations */
10502 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10503 
10504 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10505 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10506 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10507 	}
10508 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10509 
10510 	i40e_fdir_sb_setup(pf);
10511 
10512 	/* Setup static PF queue filter control settings */
10513 	ret = i40e_setup_pf_filter_control(pf);
10514 	if (ret) {
10515 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10516 			 ret);
10517 		/* Failure here should not stop continuing other steps */
10518 	}
10519 
10520 	/* enable RSS in the HW, even for only one queue, as the stack can use
10521 	 * the hash
10522 	 */
10523 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10524 		i40e_pf_config_rss(pf);
10525 
10526 	/* fill in link information and enable LSE reporting */
10527 	i40e_update_link_info(&pf->hw);
10528 	i40e_link_event(pf);
10529 
10530 	/* Initialize user-specific link properties */
10531 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10532 				  I40E_AQ_AN_COMPLETED) ? true : false);
10533 
10534 	i40e_ptp_init(pf);
10535 
10536 	return ret;
10537 }
10538 
10539 /**
10540  * i40e_determine_queue_usage - Work out queue distribution
10541  * @pf: board private structure
10542  **/
10543 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10544 {
10545 	int queues_left;
10546 
10547 	pf->num_lan_qps = 0;
10548 #ifdef I40E_FCOE
10549 	pf->num_fcoe_qps = 0;
10550 #endif
10551 
10552 	/* Find the max queues to be put into basic use.  We'll always be
10553 	 * using TC0, whether or not DCB is running, and TC0 will get the
10554 	 * big RSS set.
10555 	 */
10556 	queues_left = pf->hw.func_caps.num_tx_qp;
10557 
10558 	if ((queues_left == 1) ||
10559 	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10560 		/* one qp for PF, no queues for anything else */
10561 		queues_left = 0;
10562 		pf->alloc_rss_size = pf->num_lan_qps = 1;
10563 
10564 		/* make sure all the fancies are disabled */
10565 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
10566 			       I40E_FLAG_IWARP_ENABLED	|
10567 #ifdef I40E_FCOE
10568 			       I40E_FLAG_FCOE_ENABLED	|
10569 #endif
10570 			       I40E_FLAG_FD_SB_ENABLED	|
10571 			       I40E_FLAG_FD_ATR_ENABLED	|
10572 			       I40E_FLAG_DCB_CAPABLE	|
10573 			       I40E_FLAG_DCB_ENABLED	|
10574 			       I40E_FLAG_SRIOV_ENABLED	|
10575 			       I40E_FLAG_VMDQ_ENABLED);
10576 	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10577 				  I40E_FLAG_FD_SB_ENABLED |
10578 				  I40E_FLAG_FD_ATR_ENABLED |
10579 				  I40E_FLAG_DCB_CAPABLE))) {
10580 		/* one qp for PF */
10581 		pf->alloc_rss_size = pf->num_lan_qps = 1;
10582 		queues_left -= pf->num_lan_qps;
10583 
10584 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
10585 			       I40E_FLAG_IWARP_ENABLED	|
10586 #ifdef I40E_FCOE
10587 			       I40E_FLAG_FCOE_ENABLED	|
10588 #endif
10589 			       I40E_FLAG_FD_SB_ENABLED	|
10590 			       I40E_FLAG_FD_ATR_ENABLED	|
10591 			       I40E_FLAG_DCB_ENABLED	|
10592 			       I40E_FLAG_VMDQ_ENABLED);
10593 	} else {
10594 		/* Not enough queues for all TCs */
10595 		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10596 		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10597 			pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
10598 					I40E_FLAG_DCB_ENABLED);
10599 			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10600 		}
10601 		pf->num_lan_qps = max_t(int, pf->rss_size_max,
10602 					num_online_cpus());
10603 		pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10604 					pf->hw.func_caps.num_tx_qp);
10605 
10606 		queues_left -= pf->num_lan_qps;
10607 	}
10608 
10609 #ifdef I40E_FCOE
10610 	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10611 		if (I40E_DEFAULT_FCOE <= queues_left) {
10612 			pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10613 		} else if (I40E_MINIMUM_FCOE <= queues_left) {
10614 			pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10615 		} else {
10616 			pf->num_fcoe_qps = 0;
10617 			pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10618 			dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10619 		}
10620 
10621 		queues_left -= pf->num_fcoe_qps;
10622 	}
10623 
10624 #endif
10625 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10626 		if (queues_left > 1) {
10627 			queues_left -= 1; /* save 1 queue for FD */
10628 		} else {
10629 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10630 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10631 		}
10632 	}
10633 
10634 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10635 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10636 		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10637 					(queues_left / pf->num_vf_qps));
10638 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10639 	}
10640 
10641 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10642 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10643 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10644 					  (queues_left / pf->num_vmdq_qps));
10645 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10646 	}
10647 
10648 	pf->queues_left = queues_left;
10649 	dev_dbg(&pf->pdev->dev,
10650 		"qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10651 		pf->hw.func_caps.num_tx_qp,
10652 		!!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10653 		pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10654 		pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10655 		queues_left);
10656 #ifdef I40E_FCOE
10657 	dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10658 #endif
10659 }
10660 
10661 /**
10662  * i40e_setup_pf_filter_control - Setup PF static filter control
10663  * @pf: PF to be setup
10664  *
10665  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10666  * settings. If PE/FCoE are enabled then it will also set the per PF
10667  * based filter sizes required for them. It also enables Flow director,
10668  * ethertype and macvlan type filter settings for the pf.
10669  *
10670  * Returns 0 on success, negative on failure
10671  **/
10672 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10673 {
10674 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
10675 
10676 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10677 
10678 	/* Flow Director is enabled */
10679 	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10680 		settings->enable_fdir = true;
10681 
10682 	/* Ethtype and MACVLAN filters enabled for PF */
10683 	settings->enable_ethtype = true;
10684 	settings->enable_macvlan = true;
10685 
10686 	if (i40e_set_filter_control(&pf->hw, settings))
10687 		return -ENOENT;
10688 
10689 	return 0;
10690 }
10691 
10692 #define INFO_STRING_LEN 255
10693 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10694 static void i40e_print_features(struct i40e_pf *pf)
10695 {
10696 	struct i40e_hw *hw = &pf->hw;
10697 	char *buf;
10698 	int i;
10699 
10700 	buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10701 	if (!buf)
10702 		return;
10703 
10704 	i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10705 #ifdef CONFIG_PCI_IOV
10706 	i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10707 #endif
10708 	i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10709 		      pf->hw.func_caps.num_vsis,
10710 		      pf->vsi[pf->lan_vsi]->num_queue_pairs);
10711 	if (pf->flags & I40E_FLAG_RSS_ENABLED)
10712 		i += snprintf(&buf[i], REMAIN(i), " RSS");
10713 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10714 		i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10715 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10716 		i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10717 		i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10718 	}
10719 	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10720 		i += snprintf(&buf[i], REMAIN(i), " DCB");
10721 	i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10722 	i += snprintf(&buf[i], REMAIN(i), " Geneve");
10723 	if (pf->flags & I40E_FLAG_PTP)
10724 		i += snprintf(&buf[i], REMAIN(i), " PTP");
10725 #ifdef I40E_FCOE
10726 	if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10727 		i += snprintf(&buf[i], REMAIN(i), " FCOE");
10728 #endif
10729 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10730 		i += snprintf(&buf[i], REMAIN(i), " VEB");
10731 	else
10732 		i += snprintf(&buf[i], REMAIN(i), " VEPA");
10733 
10734 	dev_info(&pf->pdev->dev, "%s\n", buf);
10735 	kfree(buf);
10736 	WARN_ON(i > INFO_STRING_LEN);
10737 }
10738 
10739 /**
10740  * i40e_get_platform_mac_addr - get platform-specific MAC address
10741  *
10742  * @pdev: PCI device information struct
10743  * @pf: board private structure
10744  *
10745  * Look up the MAC address in Open Firmware  on systems that support it,
10746  * and use IDPROM on SPARC if no OF address is found. On return, the
10747  * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10748  * has been selected.
10749  **/
10750 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10751 {
10752 	pf->flags &= ~I40E_FLAG_PF_MAC;
10753 	if (!eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10754 		pf->flags |= I40E_FLAG_PF_MAC;
10755 }
10756 
10757 /**
10758  * i40e_probe - Device initialization routine
10759  * @pdev: PCI device information struct
10760  * @ent: entry in i40e_pci_tbl
10761  *
10762  * i40e_probe initializes a PF identified by a pci_dev structure.
10763  * The OS initialization, configuring of the PF private structure,
10764  * and a hardware reset occur.
10765  *
10766  * Returns 0 on success, negative on failure
10767  **/
10768 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10769 {
10770 	struct i40e_aq_get_phy_abilities_resp abilities;
10771 	struct i40e_pf *pf;
10772 	struct i40e_hw *hw;
10773 	static u16 pfs_found;
10774 	u16 wol_nvm_bits;
10775 	u16 link_status;
10776 	int err;
10777 	u32 val;
10778 	u32 i;
10779 	u8 set_fc_aq_fail;
10780 
10781 	err = pci_enable_device_mem(pdev);
10782 	if (err)
10783 		return err;
10784 
10785 	/* set up for high or low dma */
10786 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10787 	if (err) {
10788 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10789 		if (err) {
10790 			dev_err(&pdev->dev,
10791 				"DMA configuration failed: 0x%x\n", err);
10792 			goto err_dma;
10793 		}
10794 	}
10795 
10796 	/* set up pci connections */
10797 	err = pci_request_mem_regions(pdev, i40e_driver_name);
10798 	if (err) {
10799 		dev_info(&pdev->dev,
10800 			 "pci_request_selected_regions failed %d\n", err);
10801 		goto err_pci_reg;
10802 	}
10803 
10804 	pci_enable_pcie_error_reporting(pdev);
10805 	pci_set_master(pdev);
10806 
10807 	/* Now that we have a PCI connection, we need to do the
10808 	 * low level device setup.  This is primarily setting up
10809 	 * the Admin Queue structures and then querying for the
10810 	 * device's current profile information.
10811 	 */
10812 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10813 	if (!pf) {
10814 		err = -ENOMEM;
10815 		goto err_pf_alloc;
10816 	}
10817 	pf->next_vsi = 0;
10818 	pf->pdev = pdev;
10819 	set_bit(__I40E_DOWN, &pf->state);
10820 
10821 	hw = &pf->hw;
10822 	hw->back = pf;
10823 
10824 	pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10825 				I40E_MAX_CSR_SPACE);
10826 
10827 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10828 	if (!hw->hw_addr) {
10829 		err = -EIO;
10830 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10831 			 (unsigned int)pci_resource_start(pdev, 0),
10832 			 pf->ioremap_len, err);
10833 		goto err_ioremap;
10834 	}
10835 	hw->vendor_id = pdev->vendor;
10836 	hw->device_id = pdev->device;
10837 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10838 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
10839 	hw->subsystem_device_id = pdev->subsystem_device;
10840 	hw->bus.device = PCI_SLOT(pdev->devfn);
10841 	hw->bus.func = PCI_FUNC(pdev->devfn);
10842 	pf->instance = pfs_found;
10843 
10844 	/* set up the locks for the AQ, do this only once in probe
10845 	 * and destroy them only once in remove
10846 	 */
10847 	mutex_init(&hw->aq.asq_mutex);
10848 	mutex_init(&hw->aq.arq_mutex);
10849 
10850 	pf->msg_enable = netif_msg_init(debug,
10851 					NETIF_MSG_DRV |
10852 					NETIF_MSG_PROBE |
10853 					NETIF_MSG_LINK);
10854 	if (debug < -1)
10855 		pf->hw.debug_mask = debug;
10856 
10857 	/* do a special CORER for clearing PXE mode once at init */
10858 	if (hw->revision_id == 0 &&
10859 	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10860 		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10861 		i40e_flush(hw);
10862 		msleep(200);
10863 		pf->corer_count++;
10864 
10865 		i40e_clear_pxe_mode(hw);
10866 	}
10867 
10868 	/* Reset here to make sure all is clean and to define PF 'n' */
10869 	i40e_clear_hw(hw);
10870 	err = i40e_pf_reset(hw);
10871 	if (err) {
10872 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10873 		goto err_pf_reset;
10874 	}
10875 	pf->pfr_count++;
10876 
10877 	hw->aq.num_arq_entries = I40E_AQ_LEN;
10878 	hw->aq.num_asq_entries = I40E_AQ_LEN;
10879 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10880 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10881 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10882 
10883 	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10884 		 "%s-%s:misc",
10885 		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10886 
10887 	err = i40e_init_shared_code(hw);
10888 	if (err) {
10889 		dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10890 			 err);
10891 		goto err_pf_reset;
10892 	}
10893 
10894 	/* set up a default setting for link flow control */
10895 	pf->hw.fc.requested_mode = I40E_FC_NONE;
10896 
10897 	err = i40e_init_adminq(hw);
10898 	if (err) {
10899 		if (err == I40E_ERR_FIRMWARE_API_VERSION)
10900 			dev_info(&pdev->dev,
10901 				 "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");
10902 		else
10903 			dev_info(&pdev->dev,
10904 				 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10905 
10906 		goto err_pf_reset;
10907 	}
10908 
10909 	/* provide nvm, fw, api versions */
10910 	dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10911 		 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10912 		 hw->aq.api_maj_ver, hw->aq.api_min_ver,
10913 		 i40e_nvm_version_str(hw));
10914 
10915 	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10916 	    hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10917 		dev_info(&pdev->dev,
10918 			 "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");
10919 	else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10920 		 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10921 		dev_info(&pdev->dev,
10922 			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10923 
10924 	i40e_verify_eeprom(pf);
10925 
10926 	/* Rev 0 hardware was never productized */
10927 	if (hw->revision_id < 1)
10928 		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");
10929 
10930 	i40e_clear_pxe_mode(hw);
10931 	err = i40e_get_capabilities(pf);
10932 	if (err)
10933 		goto err_adminq_setup;
10934 
10935 	err = i40e_sw_init(pf);
10936 	if (err) {
10937 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10938 		goto err_sw_init;
10939 	}
10940 
10941 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10942 				hw->func_caps.num_rx_qp,
10943 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10944 	if (err) {
10945 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10946 		goto err_init_lan_hmc;
10947 	}
10948 
10949 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10950 	if (err) {
10951 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10952 		err = -ENOENT;
10953 		goto err_configure_lan_hmc;
10954 	}
10955 
10956 	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
10957 	 * Ignore error return codes because if it was already disabled via
10958 	 * hardware settings this will fail
10959 	 */
10960 	if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
10961 		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10962 		i40e_aq_stop_lldp(hw, true, NULL);
10963 	}
10964 
10965 	i40e_get_mac_addr(hw, hw->mac.addr);
10966 	/* allow a platform config to override the HW addr */
10967 	i40e_get_platform_mac_addr(pdev, pf);
10968 	if (!is_valid_ether_addr(hw->mac.addr)) {
10969 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10970 		err = -EIO;
10971 		goto err_mac_addr;
10972 	}
10973 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10974 	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10975 	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10976 	if (is_valid_ether_addr(hw->mac.port_addr))
10977 		pf->flags |= I40E_FLAG_PORT_ID_VALID;
10978 #ifdef I40E_FCOE
10979 	err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10980 	if (err)
10981 		dev_info(&pdev->dev,
10982 			 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10983 	if (!is_valid_ether_addr(hw->mac.san_addr)) {
10984 		dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10985 			 hw->mac.san_addr);
10986 		ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10987 	}
10988 	dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10989 #endif /* I40E_FCOE */
10990 
10991 	pci_set_drvdata(pdev, pf);
10992 	pci_save_state(pdev);
10993 #ifdef CONFIG_I40E_DCB
10994 	err = i40e_init_pf_dcb(pf);
10995 	if (err) {
10996 		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10997 		pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
10998 		/* Continue without DCB enabled */
10999 	}
11000 #endif /* CONFIG_I40E_DCB */
11001 
11002 	/* set up periodic task facility */
11003 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
11004 	pf->service_timer_period = HZ;
11005 
11006 	INIT_WORK(&pf->service_task, i40e_service_task);
11007 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
11008 	pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
11009 
11010 	/* NVM bit on means WoL disabled for the port */
11011 	i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
11012 	if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
11013 		pf->wol_en = false;
11014 	else
11015 		pf->wol_en = true;
11016 	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
11017 
11018 	/* set up the main switch operations */
11019 	i40e_determine_queue_usage(pf);
11020 	err = i40e_init_interrupt_scheme(pf);
11021 	if (err)
11022 		goto err_switch_setup;
11023 
11024 	/* The number of VSIs reported by the FW is the minimum guaranteed
11025 	 * to us; HW supports far more and we share the remaining pool with
11026 	 * the other PFs. We allocate space for more than the guarantee with
11027 	 * the understanding that we might not get them all later.
11028 	 */
11029 	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
11030 		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
11031 	else
11032 		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
11033 
11034 	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11035 	pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11036 			  GFP_KERNEL);
11037 	if (!pf->vsi) {
11038 		err = -ENOMEM;
11039 		goto err_switch_setup;
11040 	}
11041 
11042 #ifdef CONFIG_PCI_IOV
11043 	/* prep for VF support */
11044 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11045 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11046 	    !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11047 		if (pci_num_vf(pdev))
11048 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11049 	}
11050 #endif
11051 	err = i40e_setup_pf_switch(pf, false);
11052 	if (err) {
11053 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11054 		goto err_vsis;
11055 	}
11056 
11057 	/* Make sure flow control is set according to current settings */
11058 	err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11059 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11060 		dev_dbg(&pf->pdev->dev,
11061 			"Set fc with err %s aq_err %s on get_phy_cap\n",
11062 			i40e_stat_str(hw, err),
11063 			i40e_aq_str(hw, hw->aq.asq_last_status));
11064 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11065 		dev_dbg(&pf->pdev->dev,
11066 			"Set fc with err %s aq_err %s on set_phy_config\n",
11067 			i40e_stat_str(hw, err),
11068 			i40e_aq_str(hw, hw->aq.asq_last_status));
11069 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11070 		dev_dbg(&pf->pdev->dev,
11071 			"Set fc with err %s aq_err %s on get_link_info\n",
11072 			i40e_stat_str(hw, err),
11073 			i40e_aq_str(hw, hw->aq.asq_last_status));
11074 
11075 	/* if FDIR VSI was set up, start it now */
11076 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11077 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11078 			i40e_vsi_open(pf->vsi[i]);
11079 			break;
11080 		}
11081 	}
11082 
11083 	/* The driver only wants link up/down and module qualification
11084 	 * reports from firmware.  Note the negative logic.
11085 	 */
11086 	err = i40e_aq_set_phy_int_mask(&pf->hw,
11087 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
11088 					 I40E_AQ_EVENT_MEDIA_NA |
11089 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11090 	if (err)
11091 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11092 			 i40e_stat_str(&pf->hw, err),
11093 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11094 
11095 	/* Reconfigure hardware for allowing smaller MSS in the case
11096 	 * of TSO, so that we avoid the MDD being fired and causing
11097 	 * a reset in the case of small MSS+TSO.
11098 	 */
11099 	val = rd32(hw, I40E_REG_MSS);
11100 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11101 		val &= ~I40E_REG_MSS_MIN_MASK;
11102 		val |= I40E_64BYTE_MSS;
11103 		wr32(hw, I40E_REG_MSS, val);
11104 	}
11105 
11106 	if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11107 		msleep(75);
11108 		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11109 		if (err)
11110 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11111 				 i40e_stat_str(&pf->hw, err),
11112 				 i40e_aq_str(&pf->hw,
11113 					     pf->hw.aq.asq_last_status));
11114 	}
11115 	/* The main driver is (mostly) up and happy. We need to set this state
11116 	 * before setting up the misc vector or we get a race and the vector
11117 	 * ends up disabled forever.
11118 	 */
11119 	clear_bit(__I40E_DOWN, &pf->state);
11120 
11121 	/* In case of MSIX we are going to setup the misc vector right here
11122 	 * to handle admin queue events etc. In case of legacy and MSI
11123 	 * the misc functionality and queue processing is combined in
11124 	 * the same vector and that gets setup at open.
11125 	 */
11126 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11127 		err = i40e_setup_misc_vector(pf);
11128 		if (err) {
11129 			dev_info(&pdev->dev,
11130 				 "setup of misc vector failed: %d\n", err);
11131 			goto err_vsis;
11132 		}
11133 	}
11134 
11135 #ifdef CONFIG_PCI_IOV
11136 	/* prep for VF support */
11137 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11138 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11139 	    !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11140 		/* disable link interrupts for VFs */
11141 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11142 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11143 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11144 		i40e_flush(hw);
11145 
11146 		if (pci_num_vf(pdev)) {
11147 			dev_info(&pdev->dev,
11148 				 "Active VFs found, allocating resources.\n");
11149 			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11150 			if (err)
11151 				dev_info(&pdev->dev,
11152 					 "Error %d allocating resources for existing VFs\n",
11153 					 err);
11154 		}
11155 	}
11156 #endif /* CONFIG_PCI_IOV */
11157 
11158 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11159 		pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11160 						      pf->num_iwarp_msix,
11161 						      I40E_IWARP_IRQ_PILE_ID);
11162 		if (pf->iwarp_base_vector < 0) {
11163 			dev_info(&pdev->dev,
11164 				 "failed to get tracking for %d vectors for IWARP err=%d\n",
11165 				 pf->num_iwarp_msix, pf->iwarp_base_vector);
11166 			pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11167 		}
11168 	}
11169 
11170 	i40e_dbg_pf_init(pf);
11171 
11172 	/* tell the firmware that we're starting */
11173 	i40e_send_version(pf);
11174 
11175 	/* since everything's happy, start the service_task timer */
11176 	mod_timer(&pf->service_timer,
11177 		  round_jiffies(jiffies + pf->service_timer_period));
11178 
11179 	/* add this PF to client device list and launch a client service task */
11180 	err = i40e_lan_add_device(pf);
11181 	if (err)
11182 		dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11183 			 err);
11184 
11185 #ifdef I40E_FCOE
11186 	/* create FCoE interface */
11187 	i40e_fcoe_vsi_setup(pf);
11188 
11189 #endif
11190 #define PCI_SPEED_SIZE 8
11191 #define PCI_WIDTH_SIZE 8
11192 	/* Devices on the IOSF bus do not have this information
11193 	 * and will report PCI Gen 1 x 1 by default so don't bother
11194 	 * checking them.
11195 	 */
11196 	if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11197 		char speed[PCI_SPEED_SIZE] = "Unknown";
11198 		char width[PCI_WIDTH_SIZE] = "Unknown";
11199 
11200 		/* Get the negotiated link width and speed from PCI config
11201 		 * space
11202 		 */
11203 		pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11204 					  &link_status);
11205 
11206 		i40e_set_pci_config_data(hw, link_status);
11207 
11208 		switch (hw->bus.speed) {
11209 		case i40e_bus_speed_8000:
11210 			strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11211 		case i40e_bus_speed_5000:
11212 			strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11213 		case i40e_bus_speed_2500:
11214 			strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11215 		default:
11216 			break;
11217 		}
11218 		switch (hw->bus.width) {
11219 		case i40e_bus_width_pcie_x8:
11220 			strncpy(width, "8", PCI_WIDTH_SIZE); break;
11221 		case i40e_bus_width_pcie_x4:
11222 			strncpy(width, "4", PCI_WIDTH_SIZE); break;
11223 		case i40e_bus_width_pcie_x2:
11224 			strncpy(width, "2", PCI_WIDTH_SIZE); break;
11225 		case i40e_bus_width_pcie_x1:
11226 			strncpy(width, "1", PCI_WIDTH_SIZE); break;
11227 		default:
11228 			break;
11229 		}
11230 
11231 		dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11232 			 speed, width);
11233 
11234 		if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11235 		    hw->bus.speed < i40e_bus_speed_8000) {
11236 			dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11237 			dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11238 		}
11239 	}
11240 
11241 	/* get the requested speeds from the fw */
11242 	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11243 	if (err)
11244 		dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11245 			i40e_stat_str(&pf->hw, err),
11246 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11247 	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11248 
11249 	/* get the supported phy types from the fw */
11250 	err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11251 	if (err)
11252 		dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11253 			i40e_stat_str(&pf->hw, err),
11254 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11255 	pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11256 
11257 	/* Add a filter to drop all Flow control frames from any VSI from being
11258 	 * transmitted. By doing so we stop a malicious VF from sending out
11259 	 * PAUSE or PFC frames and potentially controlling traffic for other
11260 	 * PF/VF VSIs.
11261 	 * The FW can still send Flow control frames if enabled.
11262 	 */
11263 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11264 						       pf->main_vsi_seid);
11265 
11266 	if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11267 	    (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11268 		pf->flags |= I40E_FLAG_HAVE_10GBASET_PHY;
11269 
11270 	/* print a string summarizing features */
11271 	i40e_print_features(pf);
11272 
11273 	return 0;
11274 
11275 	/* Unwind what we've done if something failed in the setup */
11276 err_vsis:
11277 	set_bit(__I40E_DOWN, &pf->state);
11278 	i40e_clear_interrupt_scheme(pf);
11279 	kfree(pf->vsi);
11280 err_switch_setup:
11281 	i40e_reset_interrupt_capability(pf);
11282 	del_timer_sync(&pf->service_timer);
11283 err_mac_addr:
11284 err_configure_lan_hmc:
11285 	(void)i40e_shutdown_lan_hmc(hw);
11286 err_init_lan_hmc:
11287 	kfree(pf->qp_pile);
11288 err_sw_init:
11289 err_adminq_setup:
11290 err_pf_reset:
11291 	iounmap(hw->hw_addr);
11292 err_ioremap:
11293 	kfree(pf);
11294 err_pf_alloc:
11295 	pci_disable_pcie_error_reporting(pdev);
11296 	pci_release_mem_regions(pdev);
11297 err_pci_reg:
11298 err_dma:
11299 	pci_disable_device(pdev);
11300 	return err;
11301 }
11302 
11303 /**
11304  * i40e_remove - Device removal routine
11305  * @pdev: PCI device information struct
11306  *
11307  * i40e_remove is called by the PCI subsystem to alert the driver
11308  * that is should release a PCI device.  This could be caused by a
11309  * Hot-Plug event, or because the driver is going to be removed from
11310  * memory.
11311  **/
11312 static void i40e_remove(struct pci_dev *pdev)
11313 {
11314 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11315 	struct i40e_hw *hw = &pf->hw;
11316 	i40e_status ret_code;
11317 	int i;
11318 
11319 	i40e_dbg_pf_exit(pf);
11320 
11321 	i40e_ptp_stop(pf);
11322 
11323 	/* Disable RSS in hw */
11324 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11325 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11326 
11327 	/* no more scheduling of any task */
11328 	set_bit(__I40E_SUSPENDED, &pf->state);
11329 	set_bit(__I40E_DOWN, &pf->state);
11330 	if (pf->service_timer.data)
11331 		del_timer_sync(&pf->service_timer);
11332 	if (pf->service_task.func)
11333 		cancel_work_sync(&pf->service_task);
11334 
11335 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11336 		i40e_free_vfs(pf);
11337 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11338 	}
11339 
11340 	i40e_fdir_teardown(pf);
11341 
11342 	/* If there is a switch structure or any orphans, remove them.
11343 	 * This will leave only the PF's VSI remaining.
11344 	 */
11345 	for (i = 0; i < I40E_MAX_VEB; i++) {
11346 		if (!pf->veb[i])
11347 			continue;
11348 
11349 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11350 		    pf->veb[i]->uplink_seid == 0)
11351 			i40e_switch_branch_release(pf->veb[i]);
11352 	}
11353 
11354 	/* Now we can shutdown the PF's VSI, just before we kill
11355 	 * adminq and hmc.
11356 	 */
11357 	if (pf->vsi[pf->lan_vsi])
11358 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11359 
11360 	/* remove attached clients */
11361 	ret_code = i40e_lan_del_device(pf);
11362 	if (ret_code) {
11363 		dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11364 			 ret_code);
11365 	}
11366 
11367 	/* shutdown and destroy the HMC */
11368 	if (hw->hmc.hmc_obj) {
11369 		ret_code = i40e_shutdown_lan_hmc(hw);
11370 		if (ret_code)
11371 			dev_warn(&pdev->dev,
11372 				 "Failed to destroy the HMC resources: %d\n",
11373 				 ret_code);
11374 	}
11375 
11376 	/* shutdown the adminq */
11377 	i40e_shutdown_adminq(hw);
11378 
11379 	/* destroy the locks only once, here */
11380 	mutex_destroy(&hw->aq.arq_mutex);
11381 	mutex_destroy(&hw->aq.asq_mutex);
11382 
11383 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11384 	i40e_clear_interrupt_scheme(pf);
11385 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11386 		if (pf->vsi[i]) {
11387 			i40e_vsi_clear_rings(pf->vsi[i]);
11388 			i40e_vsi_clear(pf->vsi[i]);
11389 			pf->vsi[i] = NULL;
11390 		}
11391 	}
11392 
11393 	for (i = 0; i < I40E_MAX_VEB; i++) {
11394 		kfree(pf->veb[i]);
11395 		pf->veb[i] = NULL;
11396 	}
11397 
11398 	kfree(pf->qp_pile);
11399 	kfree(pf->vsi);
11400 
11401 	iounmap(hw->hw_addr);
11402 	kfree(pf);
11403 	pci_release_mem_regions(pdev);
11404 
11405 	pci_disable_pcie_error_reporting(pdev);
11406 	pci_disable_device(pdev);
11407 }
11408 
11409 /**
11410  * i40e_pci_error_detected - warning that something funky happened in PCI land
11411  * @pdev: PCI device information struct
11412  *
11413  * Called to warn that something happened and the error handling steps
11414  * are in progress.  Allows the driver to quiesce things, be ready for
11415  * remediation.
11416  **/
11417 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11418 						enum pci_channel_state error)
11419 {
11420 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11421 
11422 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11423 
11424 	if (!pf) {
11425 		dev_info(&pdev->dev,
11426 			 "Cannot recover - error happened during device probe\n");
11427 		return PCI_ERS_RESULT_DISCONNECT;
11428 	}
11429 
11430 	/* shutdown all operations */
11431 	if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11432 		rtnl_lock();
11433 		i40e_prep_for_reset(pf);
11434 		rtnl_unlock();
11435 	}
11436 
11437 	/* Request a slot reset */
11438 	return PCI_ERS_RESULT_NEED_RESET;
11439 }
11440 
11441 /**
11442  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11443  * @pdev: PCI device information struct
11444  *
11445  * Called to find if the driver can work with the device now that
11446  * the pci slot has been reset.  If a basic connection seems good
11447  * (registers are readable and have sane content) then return a
11448  * happy little PCI_ERS_RESULT_xxx.
11449  **/
11450 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11451 {
11452 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11453 	pci_ers_result_t result;
11454 	int err;
11455 	u32 reg;
11456 
11457 	dev_dbg(&pdev->dev, "%s\n", __func__);
11458 	if (pci_enable_device_mem(pdev)) {
11459 		dev_info(&pdev->dev,
11460 			 "Cannot re-enable PCI device after reset.\n");
11461 		result = PCI_ERS_RESULT_DISCONNECT;
11462 	} else {
11463 		pci_set_master(pdev);
11464 		pci_restore_state(pdev);
11465 		pci_save_state(pdev);
11466 		pci_wake_from_d3(pdev, false);
11467 
11468 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11469 		if (reg == 0)
11470 			result = PCI_ERS_RESULT_RECOVERED;
11471 		else
11472 			result = PCI_ERS_RESULT_DISCONNECT;
11473 	}
11474 
11475 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
11476 	if (err) {
11477 		dev_info(&pdev->dev,
11478 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11479 			 err);
11480 		/* non-fatal, continue */
11481 	}
11482 
11483 	return result;
11484 }
11485 
11486 /**
11487  * i40e_pci_error_resume - restart operations after PCI error recovery
11488  * @pdev: PCI device information struct
11489  *
11490  * Called to allow the driver to bring things back up after PCI error
11491  * and/or reset recovery has finished.
11492  **/
11493 static void i40e_pci_error_resume(struct pci_dev *pdev)
11494 {
11495 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11496 
11497 	dev_dbg(&pdev->dev, "%s\n", __func__);
11498 	if (test_bit(__I40E_SUSPENDED, &pf->state))
11499 		return;
11500 
11501 	rtnl_lock();
11502 	i40e_handle_reset_warning(pf);
11503 	rtnl_unlock();
11504 }
11505 
11506 /**
11507  * i40e_shutdown - PCI callback for shutting down
11508  * @pdev: PCI device information struct
11509  **/
11510 static void i40e_shutdown(struct pci_dev *pdev)
11511 {
11512 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11513 	struct i40e_hw *hw = &pf->hw;
11514 
11515 	set_bit(__I40E_SUSPENDED, &pf->state);
11516 	set_bit(__I40E_DOWN, &pf->state);
11517 	rtnl_lock();
11518 	i40e_prep_for_reset(pf);
11519 	rtnl_unlock();
11520 
11521 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11522 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11523 
11524 	del_timer_sync(&pf->service_timer);
11525 	cancel_work_sync(&pf->service_task);
11526 	i40e_fdir_teardown(pf);
11527 
11528 	rtnl_lock();
11529 	i40e_prep_for_reset(pf);
11530 	rtnl_unlock();
11531 
11532 	wr32(hw, I40E_PFPM_APM,
11533 	     (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11534 	wr32(hw, I40E_PFPM_WUFC,
11535 	     (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11536 
11537 	i40e_clear_interrupt_scheme(pf);
11538 
11539 	if (system_state == SYSTEM_POWER_OFF) {
11540 		pci_wake_from_d3(pdev, pf->wol_en);
11541 		pci_set_power_state(pdev, PCI_D3hot);
11542 	}
11543 }
11544 
11545 #ifdef CONFIG_PM
11546 /**
11547  * i40e_suspend - PCI callback for moving to D3
11548  * @pdev: PCI device information struct
11549  **/
11550 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11551 {
11552 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11553 	struct i40e_hw *hw = &pf->hw;
11554 	int retval = 0;
11555 
11556 	set_bit(__I40E_SUSPENDED, &pf->state);
11557 	set_bit(__I40E_DOWN, &pf->state);
11558 
11559 	rtnl_lock();
11560 	i40e_prep_for_reset(pf);
11561 	rtnl_unlock();
11562 
11563 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11564 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11565 
11566 	i40e_stop_misc_vector(pf);
11567 
11568 	retval = pci_save_state(pdev);
11569 	if (retval)
11570 		return retval;
11571 
11572 	pci_wake_from_d3(pdev, pf->wol_en);
11573 	pci_set_power_state(pdev, PCI_D3hot);
11574 
11575 	return retval;
11576 }
11577 
11578 /**
11579  * i40e_resume - PCI callback for waking up from D3
11580  * @pdev: PCI device information struct
11581  **/
11582 static int i40e_resume(struct pci_dev *pdev)
11583 {
11584 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11585 	u32 err;
11586 
11587 	pci_set_power_state(pdev, PCI_D0);
11588 	pci_restore_state(pdev);
11589 	/* pci_restore_state() clears dev->state_saves, so
11590 	 * call pci_save_state() again to restore it.
11591 	 */
11592 	pci_save_state(pdev);
11593 
11594 	err = pci_enable_device_mem(pdev);
11595 	if (err) {
11596 		dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11597 		return err;
11598 	}
11599 	pci_set_master(pdev);
11600 
11601 	/* no wakeup events while running */
11602 	pci_wake_from_d3(pdev, false);
11603 
11604 	/* handling the reset will rebuild the device state */
11605 	if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11606 		clear_bit(__I40E_DOWN, &pf->state);
11607 		rtnl_lock();
11608 		i40e_reset_and_rebuild(pf, false);
11609 		rtnl_unlock();
11610 	}
11611 
11612 	return 0;
11613 }
11614 
11615 #endif
11616 static const struct pci_error_handlers i40e_err_handler = {
11617 	.error_detected = i40e_pci_error_detected,
11618 	.slot_reset = i40e_pci_error_slot_reset,
11619 	.resume = i40e_pci_error_resume,
11620 };
11621 
11622 static struct pci_driver i40e_driver = {
11623 	.name     = i40e_driver_name,
11624 	.id_table = i40e_pci_tbl,
11625 	.probe    = i40e_probe,
11626 	.remove   = i40e_remove,
11627 #ifdef CONFIG_PM
11628 	.suspend  = i40e_suspend,
11629 	.resume   = i40e_resume,
11630 #endif
11631 	.shutdown = i40e_shutdown,
11632 	.err_handler = &i40e_err_handler,
11633 	.sriov_configure = i40e_pci_sriov_configure,
11634 };
11635 
11636 /**
11637  * i40e_init_module - Driver registration routine
11638  *
11639  * i40e_init_module is the first routine called when the driver is
11640  * loaded. All it does is register with the PCI subsystem.
11641  **/
11642 static int __init i40e_init_module(void)
11643 {
11644 	pr_info("%s: %s - version %s\n", i40e_driver_name,
11645 		i40e_driver_string, i40e_driver_version_str);
11646 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11647 
11648 	/* we will see if single thread per module is enough for now,
11649 	 * it can't be any worse than using the system workqueue which
11650 	 * was already single threaded
11651 	 */
11652 	i40e_wq = alloc_workqueue("%s", WQ_UNBOUND | WQ_MEM_RECLAIM, 1,
11653 				  i40e_driver_name);
11654 	if (!i40e_wq) {
11655 		pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11656 		return -ENOMEM;
11657 	}
11658 
11659 	i40e_dbg_init();
11660 	return pci_register_driver(&i40e_driver);
11661 }
11662 module_init(i40e_init_module);
11663 
11664 /**
11665  * i40e_exit_module - Driver exit cleanup routine
11666  *
11667  * i40e_exit_module is called just before the driver is removed
11668  * from memory.
11669  **/
11670 static void __exit i40e_exit_module(void)
11671 {
11672 	pci_unregister_driver(&i40e_driver);
11673 	destroy_workqueue(i40e_wq);
11674 	i40e_dbg_exit();
11675 }
11676 module_exit(i40e_exit_module);
11677