xref: /linux/drivers/net/ethernet/intel/i40e/i40e_main.c (revision 31368ce83c59a5422ee621a38aeea98142d0ecf7)
1 /*******************************************************************************
2  *
3  * Intel Ethernet Controller XL710 Family Linux Driver
4  * Copyright(c) 2013 - 2017 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 #include <linux/bpf.h>
31 
32 /* Local includes */
33 #include "i40e.h"
34 #include "i40e_diag.h"
35 #include <net/udp_tunnel.h>
36 /* All i40e tracepoints are defined by the include below, which
37  * must be included exactly once across the whole kernel with
38  * CREATE_TRACE_POINTS defined
39  */
40 #define CREATE_TRACE_POINTS
41 #include "i40e_trace.h"
42 
43 const char i40e_driver_name[] = "i40e";
44 static const char i40e_driver_string[] =
45 			"Intel(R) Ethernet Connection XL710 Network Driver";
46 
47 #define DRV_KERN "-k"
48 
49 #define DRV_VERSION_MAJOR 2
50 #define DRV_VERSION_MINOR 1
51 #define DRV_VERSION_BUILD 14
52 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
53 	     __stringify(DRV_VERSION_MINOR) "." \
54 	     __stringify(DRV_VERSION_BUILD)    DRV_KERN
55 const char i40e_driver_version_str[] = DRV_VERSION;
56 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
57 
58 /* a bit of forward declarations */
59 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
60 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired);
61 static int i40e_add_vsi(struct i40e_vsi *vsi);
62 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
63 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
64 static int i40e_setup_misc_vector(struct i40e_pf *pf);
65 static void i40e_determine_queue_usage(struct i40e_pf *pf);
66 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
67 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired);
68 static int i40e_reset(struct i40e_pf *pf);
69 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired);
70 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
71 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
72 
73 /* i40e_pci_tbl - PCI Device ID Table
74  *
75  * Last entry must be all 0s
76  *
77  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
78  *   Class, Class Mask, private data (not used) }
79  */
80 static const struct pci_device_id i40e_pci_tbl[] = {
81 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
82 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
83 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
84 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
85 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
86 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
87 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
88 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
89 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
90 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
91 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
92 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
93 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
94 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
95 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
96 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
97 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
98 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_B), 0},
99 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_SFP28), 0},
100 	/* required last entry */
101 	{0, }
102 };
103 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
104 
105 #define I40E_MAX_VF_COUNT 128
106 static int debug = -1;
107 module_param(debug, uint, 0);
108 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
109 
110 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
111 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
112 MODULE_LICENSE("GPL");
113 MODULE_VERSION(DRV_VERSION);
114 
115 static struct workqueue_struct *i40e_wq;
116 
117 /**
118  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
119  * @hw:   pointer to the HW structure
120  * @mem:  ptr to mem struct to fill out
121  * @size: size of memory requested
122  * @alignment: what to align the allocation to
123  **/
124 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
125 			    u64 size, u32 alignment)
126 {
127 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
128 
129 	mem->size = ALIGN(size, alignment);
130 	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
131 				      &mem->pa, GFP_KERNEL);
132 	if (!mem->va)
133 		return -ENOMEM;
134 
135 	return 0;
136 }
137 
138 /**
139  * i40e_free_dma_mem_d - OS specific memory free for shared code
140  * @hw:   pointer to the HW structure
141  * @mem:  ptr to mem struct to free
142  **/
143 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
144 {
145 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
146 
147 	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
148 	mem->va = NULL;
149 	mem->pa = 0;
150 	mem->size = 0;
151 
152 	return 0;
153 }
154 
155 /**
156  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
157  * @hw:   pointer to the HW structure
158  * @mem:  ptr to mem struct to fill out
159  * @size: size of memory requested
160  **/
161 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
162 			     u32 size)
163 {
164 	mem->size = size;
165 	mem->va = kzalloc(size, GFP_KERNEL);
166 
167 	if (!mem->va)
168 		return -ENOMEM;
169 
170 	return 0;
171 }
172 
173 /**
174  * i40e_free_virt_mem_d - OS specific memory free for shared code
175  * @hw:   pointer to the HW structure
176  * @mem:  ptr to mem struct to free
177  **/
178 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
179 {
180 	/* it's ok to kfree a NULL pointer */
181 	kfree(mem->va);
182 	mem->va = NULL;
183 	mem->size = 0;
184 
185 	return 0;
186 }
187 
188 /**
189  * i40e_get_lump - find a lump of free generic resource
190  * @pf: board private structure
191  * @pile: the pile of resource to search
192  * @needed: the number of items needed
193  * @id: an owner id to stick on the items assigned
194  *
195  * Returns the base item index of the lump, or negative for error
196  *
197  * The search_hint trick and lack of advanced fit-finding only work
198  * because we're highly likely to have all the same size lump requests.
199  * Linear search time and any fragmentation should be minimal.
200  **/
201 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
202 			 u16 needed, u16 id)
203 {
204 	int ret = -ENOMEM;
205 	int i, j;
206 
207 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
208 		dev_info(&pf->pdev->dev,
209 			 "param err: pile=%p needed=%d id=0x%04x\n",
210 			 pile, needed, id);
211 		return -EINVAL;
212 	}
213 
214 	/* start the linear search with an imperfect hint */
215 	i = pile->search_hint;
216 	while (i < pile->num_entries) {
217 		/* skip already allocated entries */
218 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
219 			i++;
220 			continue;
221 		}
222 
223 		/* do we have enough in this lump? */
224 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
225 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
226 				break;
227 		}
228 
229 		if (j == needed) {
230 			/* there was enough, so assign it to the requestor */
231 			for (j = 0; j < needed; j++)
232 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
233 			ret = i;
234 			pile->search_hint = i + j;
235 			break;
236 		}
237 
238 		/* not enough, so skip over it and continue looking */
239 		i += j;
240 	}
241 
242 	return ret;
243 }
244 
245 /**
246  * i40e_put_lump - return a lump of generic resource
247  * @pile: the pile of resource to search
248  * @index: the base item index
249  * @id: the owner id of the items assigned
250  *
251  * Returns the count of items in the lump
252  **/
253 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
254 {
255 	int valid_id = (id | I40E_PILE_VALID_BIT);
256 	int count = 0;
257 	int i;
258 
259 	if (!pile || index >= pile->num_entries)
260 		return -EINVAL;
261 
262 	for (i = index;
263 	     i < pile->num_entries && pile->list[i] == valid_id;
264 	     i++) {
265 		pile->list[i] = 0;
266 		count++;
267 	}
268 
269 	if (count && index < pile->search_hint)
270 		pile->search_hint = index;
271 
272 	return count;
273 }
274 
275 /**
276  * i40e_find_vsi_from_id - searches for the vsi with the given id
277  * @pf - the pf structure to search for the vsi
278  * @id - id of the vsi it is searching for
279  **/
280 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
281 {
282 	int i;
283 
284 	for (i = 0; i < pf->num_alloc_vsi; i++)
285 		if (pf->vsi[i] && (pf->vsi[i]->id == id))
286 			return pf->vsi[i];
287 
288 	return NULL;
289 }
290 
291 /**
292  * i40e_service_event_schedule - Schedule the service task to wake up
293  * @pf: board private structure
294  *
295  * If not already scheduled, this puts the task into the work queue
296  **/
297 void i40e_service_event_schedule(struct i40e_pf *pf)
298 {
299 	if (!test_bit(__I40E_DOWN, pf->state) &&
300 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
301 		queue_work(i40e_wq, &pf->service_task);
302 }
303 
304 /**
305  * i40e_tx_timeout - Respond to a Tx Hang
306  * @netdev: network interface device structure
307  *
308  * If any port has noticed a Tx timeout, it is likely that the whole
309  * device is munged, not just the one netdev port, so go for the full
310  * reset.
311  **/
312 static void i40e_tx_timeout(struct net_device *netdev)
313 {
314 	struct i40e_netdev_priv *np = netdev_priv(netdev);
315 	struct i40e_vsi *vsi = np->vsi;
316 	struct i40e_pf *pf = vsi->back;
317 	struct i40e_ring *tx_ring = NULL;
318 	unsigned int i, hung_queue = 0;
319 	u32 head, val;
320 
321 	pf->tx_timeout_count++;
322 
323 	/* find the stopped queue the same way the stack does */
324 	for (i = 0; i < netdev->num_tx_queues; i++) {
325 		struct netdev_queue *q;
326 		unsigned long trans_start;
327 
328 		q = netdev_get_tx_queue(netdev, i);
329 		trans_start = q->trans_start;
330 		if (netif_xmit_stopped(q) &&
331 		    time_after(jiffies,
332 			       (trans_start + netdev->watchdog_timeo))) {
333 			hung_queue = i;
334 			break;
335 		}
336 	}
337 
338 	if (i == netdev->num_tx_queues) {
339 		netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
340 	} else {
341 		/* now that we have an index, find the tx_ring struct */
342 		for (i = 0; i < vsi->num_queue_pairs; i++) {
343 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
344 				if (hung_queue ==
345 				    vsi->tx_rings[i]->queue_index) {
346 					tx_ring = vsi->tx_rings[i];
347 					break;
348 				}
349 			}
350 		}
351 	}
352 
353 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
354 		pf->tx_timeout_recovery_level = 1;  /* reset after some time */
355 	else if (time_before(jiffies,
356 		      (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
357 		return;   /* don't do any new action before the next timeout */
358 
359 	if (tx_ring) {
360 		head = i40e_get_head(tx_ring);
361 		/* Read interrupt register */
362 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
363 			val = rd32(&pf->hw,
364 			     I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
365 						tx_ring->vsi->base_vector - 1));
366 		else
367 			val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
368 
369 		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",
370 			    vsi->seid, hung_queue, tx_ring->next_to_clean,
371 			    head, tx_ring->next_to_use,
372 			    readl(tx_ring->tail), val);
373 	}
374 
375 	pf->tx_timeout_last_recovery = jiffies;
376 	netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
377 		    pf->tx_timeout_recovery_level, hung_queue);
378 
379 	switch (pf->tx_timeout_recovery_level) {
380 	case 1:
381 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
382 		break;
383 	case 2:
384 		set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
385 		break;
386 	case 3:
387 		set_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
388 		break;
389 	default:
390 		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
391 		break;
392 	}
393 
394 	i40e_service_event_schedule(pf);
395 	pf->tx_timeout_recovery_level++;
396 }
397 
398 /**
399  * i40e_get_vsi_stats_struct - Get System Network Statistics
400  * @vsi: the VSI we care about
401  *
402  * Returns the address of the device statistics structure.
403  * The statistics are actually updated from the service task.
404  **/
405 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
406 {
407 	return &vsi->net_stats;
408 }
409 
410 /**
411  * i40e_get_netdev_stats_struct_tx - populate stats from a Tx ring
412  * @ring: Tx ring to get statistics from
413  * @stats: statistics entry to be updated
414  **/
415 static void i40e_get_netdev_stats_struct_tx(struct i40e_ring *ring,
416 					    struct rtnl_link_stats64 *stats)
417 {
418 	u64 bytes, packets;
419 	unsigned int start;
420 
421 	do {
422 		start = u64_stats_fetch_begin_irq(&ring->syncp);
423 		packets = ring->stats.packets;
424 		bytes   = ring->stats.bytes;
425 	} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
426 
427 	stats->tx_packets += packets;
428 	stats->tx_bytes   += bytes;
429 }
430 
431 /**
432  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
433  * @netdev: network interface device structure
434  *
435  * Returns the address of the device statistics structure.
436  * The statistics are actually updated from the service task.
437  **/
438 static void i40e_get_netdev_stats_struct(struct net_device *netdev,
439 				  struct rtnl_link_stats64 *stats)
440 {
441 	struct i40e_netdev_priv *np = netdev_priv(netdev);
442 	struct i40e_ring *tx_ring, *rx_ring;
443 	struct i40e_vsi *vsi = np->vsi;
444 	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
445 	int i;
446 
447 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
448 		return;
449 
450 	if (!vsi->tx_rings)
451 		return;
452 
453 	rcu_read_lock();
454 	for (i = 0; i < vsi->num_queue_pairs; i++) {
455 		u64 bytes, packets;
456 		unsigned int start;
457 
458 		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
459 		if (!tx_ring)
460 			continue;
461 		i40e_get_netdev_stats_struct_tx(tx_ring, stats);
462 
463 		rx_ring = &tx_ring[1];
464 
465 		do {
466 			start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
467 			packets = rx_ring->stats.packets;
468 			bytes   = rx_ring->stats.bytes;
469 		} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
470 
471 		stats->rx_packets += packets;
472 		stats->rx_bytes   += bytes;
473 
474 		if (i40e_enabled_xdp_vsi(vsi))
475 			i40e_get_netdev_stats_struct_tx(&rx_ring[1], stats);
476 	}
477 	rcu_read_unlock();
478 
479 	/* following stats updated by i40e_watchdog_subtask() */
480 	stats->multicast	= vsi_stats->multicast;
481 	stats->tx_errors	= vsi_stats->tx_errors;
482 	stats->tx_dropped	= vsi_stats->tx_dropped;
483 	stats->rx_errors	= vsi_stats->rx_errors;
484 	stats->rx_dropped	= vsi_stats->rx_dropped;
485 	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
486 	stats->rx_length_errors	= vsi_stats->rx_length_errors;
487 }
488 
489 /**
490  * i40e_vsi_reset_stats - Resets all stats of the given vsi
491  * @vsi: the VSI to have its stats reset
492  **/
493 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
494 {
495 	struct rtnl_link_stats64 *ns;
496 	int i;
497 
498 	if (!vsi)
499 		return;
500 
501 	ns = i40e_get_vsi_stats_struct(vsi);
502 	memset(ns, 0, sizeof(*ns));
503 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
504 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
505 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
506 	if (vsi->rx_rings && vsi->rx_rings[0]) {
507 		for (i = 0; i < vsi->num_queue_pairs; i++) {
508 			memset(&vsi->rx_rings[i]->stats, 0,
509 			       sizeof(vsi->rx_rings[i]->stats));
510 			memset(&vsi->rx_rings[i]->rx_stats, 0,
511 			       sizeof(vsi->rx_rings[i]->rx_stats));
512 			memset(&vsi->tx_rings[i]->stats, 0,
513 			       sizeof(vsi->tx_rings[i]->stats));
514 			memset(&vsi->tx_rings[i]->tx_stats, 0,
515 			       sizeof(vsi->tx_rings[i]->tx_stats));
516 		}
517 	}
518 	vsi->stat_offsets_loaded = false;
519 }
520 
521 /**
522  * i40e_pf_reset_stats - Reset all of the stats for the given PF
523  * @pf: the PF to be reset
524  **/
525 void i40e_pf_reset_stats(struct i40e_pf *pf)
526 {
527 	int i;
528 
529 	memset(&pf->stats, 0, sizeof(pf->stats));
530 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
531 	pf->stat_offsets_loaded = false;
532 
533 	for (i = 0; i < I40E_MAX_VEB; i++) {
534 		if (pf->veb[i]) {
535 			memset(&pf->veb[i]->stats, 0,
536 			       sizeof(pf->veb[i]->stats));
537 			memset(&pf->veb[i]->stats_offsets, 0,
538 			       sizeof(pf->veb[i]->stats_offsets));
539 			pf->veb[i]->stat_offsets_loaded = false;
540 		}
541 	}
542 	pf->hw_csum_rx_error = 0;
543 }
544 
545 /**
546  * i40e_stat_update48 - read and update a 48 bit stat from the chip
547  * @hw: ptr to the hardware info
548  * @hireg: the high 32 bit reg to read
549  * @loreg: the low 32 bit reg to read
550  * @offset_loaded: has the initial offset been loaded yet
551  * @offset: ptr to current offset value
552  * @stat: ptr to the stat
553  *
554  * Since the device stats are not reset at PFReset, they likely will not
555  * be zeroed when the driver starts.  We'll save the first values read
556  * and use them as offsets to be subtracted from the raw values in order
557  * to report stats that count from zero.  In the process, we also manage
558  * the potential roll-over.
559  **/
560 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
561 			       bool offset_loaded, u64 *offset, u64 *stat)
562 {
563 	u64 new_data;
564 
565 	if (hw->device_id == I40E_DEV_ID_QEMU) {
566 		new_data = rd32(hw, loreg);
567 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
568 	} else {
569 		new_data = rd64(hw, loreg);
570 	}
571 	if (!offset_loaded)
572 		*offset = new_data;
573 	if (likely(new_data >= *offset))
574 		*stat = new_data - *offset;
575 	else
576 		*stat = (new_data + BIT_ULL(48)) - *offset;
577 	*stat &= 0xFFFFFFFFFFFFULL;
578 }
579 
580 /**
581  * i40e_stat_update32 - read and update a 32 bit stat from the chip
582  * @hw: ptr to the hardware info
583  * @reg: the hw reg to read
584  * @offset_loaded: has the initial offset been loaded yet
585  * @offset: ptr to current offset value
586  * @stat: ptr to the stat
587  **/
588 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
589 			       bool offset_loaded, u64 *offset, u64 *stat)
590 {
591 	u32 new_data;
592 
593 	new_data = rd32(hw, reg);
594 	if (!offset_loaded)
595 		*offset = new_data;
596 	if (likely(new_data >= *offset))
597 		*stat = (u32)(new_data - *offset);
598 	else
599 		*stat = (u32)((new_data + BIT_ULL(32)) - *offset);
600 }
601 
602 /**
603  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
604  * @vsi: the VSI to be updated
605  **/
606 void i40e_update_eth_stats(struct i40e_vsi *vsi)
607 {
608 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
609 	struct i40e_pf *pf = vsi->back;
610 	struct i40e_hw *hw = &pf->hw;
611 	struct i40e_eth_stats *oes;
612 	struct i40e_eth_stats *es;     /* device's eth stats */
613 
614 	es = &vsi->eth_stats;
615 	oes = &vsi->eth_stats_offsets;
616 
617 	/* Gather up the stats that the hw collects */
618 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
619 			   vsi->stat_offsets_loaded,
620 			   &oes->tx_errors, &es->tx_errors);
621 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
622 			   vsi->stat_offsets_loaded,
623 			   &oes->rx_discards, &es->rx_discards);
624 	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
625 			   vsi->stat_offsets_loaded,
626 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
627 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
628 			   vsi->stat_offsets_loaded,
629 			   &oes->tx_errors, &es->tx_errors);
630 
631 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
632 			   I40E_GLV_GORCL(stat_idx),
633 			   vsi->stat_offsets_loaded,
634 			   &oes->rx_bytes, &es->rx_bytes);
635 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
636 			   I40E_GLV_UPRCL(stat_idx),
637 			   vsi->stat_offsets_loaded,
638 			   &oes->rx_unicast, &es->rx_unicast);
639 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
640 			   I40E_GLV_MPRCL(stat_idx),
641 			   vsi->stat_offsets_loaded,
642 			   &oes->rx_multicast, &es->rx_multicast);
643 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
644 			   I40E_GLV_BPRCL(stat_idx),
645 			   vsi->stat_offsets_loaded,
646 			   &oes->rx_broadcast, &es->rx_broadcast);
647 
648 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
649 			   I40E_GLV_GOTCL(stat_idx),
650 			   vsi->stat_offsets_loaded,
651 			   &oes->tx_bytes, &es->tx_bytes);
652 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
653 			   I40E_GLV_UPTCL(stat_idx),
654 			   vsi->stat_offsets_loaded,
655 			   &oes->tx_unicast, &es->tx_unicast);
656 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
657 			   I40E_GLV_MPTCL(stat_idx),
658 			   vsi->stat_offsets_loaded,
659 			   &oes->tx_multicast, &es->tx_multicast);
660 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
661 			   I40E_GLV_BPTCL(stat_idx),
662 			   vsi->stat_offsets_loaded,
663 			   &oes->tx_broadcast, &es->tx_broadcast);
664 	vsi->stat_offsets_loaded = true;
665 }
666 
667 /**
668  * i40e_update_veb_stats - Update Switch component statistics
669  * @veb: the VEB being updated
670  **/
671 static void i40e_update_veb_stats(struct i40e_veb *veb)
672 {
673 	struct i40e_pf *pf = veb->pf;
674 	struct i40e_hw *hw = &pf->hw;
675 	struct i40e_eth_stats *oes;
676 	struct i40e_eth_stats *es;     /* device's eth stats */
677 	struct i40e_veb_tc_stats *veb_oes;
678 	struct i40e_veb_tc_stats *veb_es;
679 	int i, idx = 0;
680 
681 	idx = veb->stats_idx;
682 	es = &veb->stats;
683 	oes = &veb->stats_offsets;
684 	veb_es = &veb->tc_stats;
685 	veb_oes = &veb->tc_stats_offsets;
686 
687 	/* Gather up the stats that the hw collects */
688 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
689 			   veb->stat_offsets_loaded,
690 			   &oes->tx_discards, &es->tx_discards);
691 	if (hw->revision_id > 0)
692 		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
693 				   veb->stat_offsets_loaded,
694 				   &oes->rx_unknown_protocol,
695 				   &es->rx_unknown_protocol);
696 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
697 			   veb->stat_offsets_loaded,
698 			   &oes->rx_bytes, &es->rx_bytes);
699 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
700 			   veb->stat_offsets_loaded,
701 			   &oes->rx_unicast, &es->rx_unicast);
702 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
703 			   veb->stat_offsets_loaded,
704 			   &oes->rx_multicast, &es->rx_multicast);
705 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
706 			   veb->stat_offsets_loaded,
707 			   &oes->rx_broadcast, &es->rx_broadcast);
708 
709 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
710 			   veb->stat_offsets_loaded,
711 			   &oes->tx_bytes, &es->tx_bytes);
712 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
713 			   veb->stat_offsets_loaded,
714 			   &oes->tx_unicast, &es->tx_unicast);
715 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
716 			   veb->stat_offsets_loaded,
717 			   &oes->tx_multicast, &es->tx_multicast);
718 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
719 			   veb->stat_offsets_loaded,
720 			   &oes->tx_broadcast, &es->tx_broadcast);
721 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
722 		i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
723 				   I40E_GLVEBTC_RPCL(i, idx),
724 				   veb->stat_offsets_loaded,
725 				   &veb_oes->tc_rx_packets[i],
726 				   &veb_es->tc_rx_packets[i]);
727 		i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
728 				   I40E_GLVEBTC_RBCL(i, idx),
729 				   veb->stat_offsets_loaded,
730 				   &veb_oes->tc_rx_bytes[i],
731 				   &veb_es->tc_rx_bytes[i]);
732 		i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
733 				   I40E_GLVEBTC_TPCL(i, idx),
734 				   veb->stat_offsets_loaded,
735 				   &veb_oes->tc_tx_packets[i],
736 				   &veb_es->tc_tx_packets[i]);
737 		i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
738 				   I40E_GLVEBTC_TBCL(i, idx),
739 				   veb->stat_offsets_loaded,
740 				   &veb_oes->tc_tx_bytes[i],
741 				   &veb_es->tc_tx_bytes[i]);
742 	}
743 	veb->stat_offsets_loaded = true;
744 }
745 
746 /**
747  * i40e_update_vsi_stats - Update the vsi statistics counters.
748  * @vsi: the VSI to be updated
749  *
750  * There are a few instances where we store the same stat in a
751  * couple of different structs.  This is partly because we have
752  * the netdev stats that need to be filled out, which is slightly
753  * different from the "eth_stats" defined by the chip and used in
754  * VF communications.  We sort it out here.
755  **/
756 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
757 {
758 	struct i40e_pf *pf = vsi->back;
759 	struct rtnl_link_stats64 *ons;
760 	struct rtnl_link_stats64 *ns;   /* netdev stats */
761 	struct i40e_eth_stats *oes;
762 	struct i40e_eth_stats *es;     /* device's eth stats */
763 	u32 tx_restart, tx_busy;
764 	struct i40e_ring *p;
765 	u32 rx_page, rx_buf;
766 	u64 bytes, packets;
767 	unsigned int start;
768 	u64 tx_linearize;
769 	u64 tx_force_wb;
770 	u64 rx_p, rx_b;
771 	u64 tx_p, tx_b;
772 	u16 q;
773 
774 	if (test_bit(__I40E_VSI_DOWN, vsi->state) ||
775 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
776 		return;
777 
778 	ns = i40e_get_vsi_stats_struct(vsi);
779 	ons = &vsi->net_stats_offsets;
780 	es = &vsi->eth_stats;
781 	oes = &vsi->eth_stats_offsets;
782 
783 	/* Gather up the netdev and vsi stats that the driver collects
784 	 * on the fly during packet processing
785 	 */
786 	rx_b = rx_p = 0;
787 	tx_b = tx_p = 0;
788 	tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
789 	rx_page = 0;
790 	rx_buf = 0;
791 	rcu_read_lock();
792 	for (q = 0; q < vsi->num_queue_pairs; q++) {
793 		/* locate Tx ring */
794 		p = ACCESS_ONCE(vsi->tx_rings[q]);
795 
796 		do {
797 			start = u64_stats_fetch_begin_irq(&p->syncp);
798 			packets = p->stats.packets;
799 			bytes = p->stats.bytes;
800 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
801 		tx_b += bytes;
802 		tx_p += packets;
803 		tx_restart += p->tx_stats.restart_queue;
804 		tx_busy += p->tx_stats.tx_busy;
805 		tx_linearize += p->tx_stats.tx_linearize;
806 		tx_force_wb += p->tx_stats.tx_force_wb;
807 
808 		/* Rx queue is part of the same block as Tx queue */
809 		p = &p[1];
810 		do {
811 			start = u64_stats_fetch_begin_irq(&p->syncp);
812 			packets = p->stats.packets;
813 			bytes = p->stats.bytes;
814 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
815 		rx_b += bytes;
816 		rx_p += packets;
817 		rx_buf += p->rx_stats.alloc_buff_failed;
818 		rx_page += p->rx_stats.alloc_page_failed;
819 	}
820 	rcu_read_unlock();
821 	vsi->tx_restart = tx_restart;
822 	vsi->tx_busy = tx_busy;
823 	vsi->tx_linearize = tx_linearize;
824 	vsi->tx_force_wb = tx_force_wb;
825 	vsi->rx_page_failed = rx_page;
826 	vsi->rx_buf_failed = rx_buf;
827 
828 	ns->rx_packets = rx_p;
829 	ns->rx_bytes = rx_b;
830 	ns->tx_packets = tx_p;
831 	ns->tx_bytes = tx_b;
832 
833 	/* update netdev stats from eth stats */
834 	i40e_update_eth_stats(vsi);
835 	ons->tx_errors = oes->tx_errors;
836 	ns->tx_errors = es->tx_errors;
837 	ons->multicast = oes->rx_multicast;
838 	ns->multicast = es->rx_multicast;
839 	ons->rx_dropped = oes->rx_discards;
840 	ns->rx_dropped = es->rx_discards;
841 	ons->tx_dropped = oes->tx_discards;
842 	ns->tx_dropped = es->tx_discards;
843 
844 	/* pull in a couple PF stats if this is the main vsi */
845 	if (vsi == pf->vsi[pf->lan_vsi]) {
846 		ns->rx_crc_errors = pf->stats.crc_errors;
847 		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
848 		ns->rx_length_errors = pf->stats.rx_length_errors;
849 	}
850 }
851 
852 /**
853  * i40e_update_pf_stats - Update the PF statistics counters.
854  * @pf: the PF to be updated
855  **/
856 static void i40e_update_pf_stats(struct i40e_pf *pf)
857 {
858 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
859 	struct i40e_hw_port_stats *nsd = &pf->stats;
860 	struct i40e_hw *hw = &pf->hw;
861 	u32 val;
862 	int i;
863 
864 	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
865 			   I40E_GLPRT_GORCL(hw->port),
866 			   pf->stat_offsets_loaded,
867 			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
868 	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
869 			   I40E_GLPRT_GOTCL(hw->port),
870 			   pf->stat_offsets_loaded,
871 			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
872 	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
873 			   pf->stat_offsets_loaded,
874 			   &osd->eth.rx_discards,
875 			   &nsd->eth.rx_discards);
876 	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
877 			   I40E_GLPRT_UPRCL(hw->port),
878 			   pf->stat_offsets_loaded,
879 			   &osd->eth.rx_unicast,
880 			   &nsd->eth.rx_unicast);
881 	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
882 			   I40E_GLPRT_MPRCL(hw->port),
883 			   pf->stat_offsets_loaded,
884 			   &osd->eth.rx_multicast,
885 			   &nsd->eth.rx_multicast);
886 	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
887 			   I40E_GLPRT_BPRCL(hw->port),
888 			   pf->stat_offsets_loaded,
889 			   &osd->eth.rx_broadcast,
890 			   &nsd->eth.rx_broadcast);
891 	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
892 			   I40E_GLPRT_UPTCL(hw->port),
893 			   pf->stat_offsets_loaded,
894 			   &osd->eth.tx_unicast,
895 			   &nsd->eth.tx_unicast);
896 	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
897 			   I40E_GLPRT_MPTCL(hw->port),
898 			   pf->stat_offsets_loaded,
899 			   &osd->eth.tx_multicast,
900 			   &nsd->eth.tx_multicast);
901 	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
902 			   I40E_GLPRT_BPTCL(hw->port),
903 			   pf->stat_offsets_loaded,
904 			   &osd->eth.tx_broadcast,
905 			   &nsd->eth.tx_broadcast);
906 
907 	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
908 			   pf->stat_offsets_loaded,
909 			   &osd->tx_dropped_link_down,
910 			   &nsd->tx_dropped_link_down);
911 
912 	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
913 			   pf->stat_offsets_loaded,
914 			   &osd->crc_errors, &nsd->crc_errors);
915 
916 	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
917 			   pf->stat_offsets_loaded,
918 			   &osd->illegal_bytes, &nsd->illegal_bytes);
919 
920 	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
921 			   pf->stat_offsets_loaded,
922 			   &osd->mac_local_faults,
923 			   &nsd->mac_local_faults);
924 	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
925 			   pf->stat_offsets_loaded,
926 			   &osd->mac_remote_faults,
927 			   &nsd->mac_remote_faults);
928 
929 	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
930 			   pf->stat_offsets_loaded,
931 			   &osd->rx_length_errors,
932 			   &nsd->rx_length_errors);
933 
934 	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
935 			   pf->stat_offsets_loaded,
936 			   &osd->link_xon_rx, &nsd->link_xon_rx);
937 	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
938 			   pf->stat_offsets_loaded,
939 			   &osd->link_xon_tx, &nsd->link_xon_tx);
940 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
941 			   pf->stat_offsets_loaded,
942 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
943 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
944 			   pf->stat_offsets_loaded,
945 			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
946 
947 	for (i = 0; i < 8; i++) {
948 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
949 				   pf->stat_offsets_loaded,
950 				   &osd->priority_xoff_rx[i],
951 				   &nsd->priority_xoff_rx[i]);
952 		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
953 				   pf->stat_offsets_loaded,
954 				   &osd->priority_xon_rx[i],
955 				   &nsd->priority_xon_rx[i]);
956 		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
957 				   pf->stat_offsets_loaded,
958 				   &osd->priority_xon_tx[i],
959 				   &nsd->priority_xon_tx[i]);
960 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
961 				   pf->stat_offsets_loaded,
962 				   &osd->priority_xoff_tx[i],
963 				   &nsd->priority_xoff_tx[i]);
964 		i40e_stat_update32(hw,
965 				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
966 				   pf->stat_offsets_loaded,
967 				   &osd->priority_xon_2_xoff[i],
968 				   &nsd->priority_xon_2_xoff[i]);
969 	}
970 
971 	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
972 			   I40E_GLPRT_PRC64L(hw->port),
973 			   pf->stat_offsets_loaded,
974 			   &osd->rx_size_64, &nsd->rx_size_64);
975 	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
976 			   I40E_GLPRT_PRC127L(hw->port),
977 			   pf->stat_offsets_loaded,
978 			   &osd->rx_size_127, &nsd->rx_size_127);
979 	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
980 			   I40E_GLPRT_PRC255L(hw->port),
981 			   pf->stat_offsets_loaded,
982 			   &osd->rx_size_255, &nsd->rx_size_255);
983 	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
984 			   I40E_GLPRT_PRC511L(hw->port),
985 			   pf->stat_offsets_loaded,
986 			   &osd->rx_size_511, &nsd->rx_size_511);
987 	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
988 			   I40E_GLPRT_PRC1023L(hw->port),
989 			   pf->stat_offsets_loaded,
990 			   &osd->rx_size_1023, &nsd->rx_size_1023);
991 	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
992 			   I40E_GLPRT_PRC1522L(hw->port),
993 			   pf->stat_offsets_loaded,
994 			   &osd->rx_size_1522, &nsd->rx_size_1522);
995 	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
996 			   I40E_GLPRT_PRC9522L(hw->port),
997 			   pf->stat_offsets_loaded,
998 			   &osd->rx_size_big, &nsd->rx_size_big);
999 
1000 	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1001 			   I40E_GLPRT_PTC64L(hw->port),
1002 			   pf->stat_offsets_loaded,
1003 			   &osd->tx_size_64, &nsd->tx_size_64);
1004 	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1005 			   I40E_GLPRT_PTC127L(hw->port),
1006 			   pf->stat_offsets_loaded,
1007 			   &osd->tx_size_127, &nsd->tx_size_127);
1008 	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1009 			   I40E_GLPRT_PTC255L(hw->port),
1010 			   pf->stat_offsets_loaded,
1011 			   &osd->tx_size_255, &nsd->tx_size_255);
1012 	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1013 			   I40E_GLPRT_PTC511L(hw->port),
1014 			   pf->stat_offsets_loaded,
1015 			   &osd->tx_size_511, &nsd->tx_size_511);
1016 	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1017 			   I40E_GLPRT_PTC1023L(hw->port),
1018 			   pf->stat_offsets_loaded,
1019 			   &osd->tx_size_1023, &nsd->tx_size_1023);
1020 	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1021 			   I40E_GLPRT_PTC1522L(hw->port),
1022 			   pf->stat_offsets_loaded,
1023 			   &osd->tx_size_1522, &nsd->tx_size_1522);
1024 	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1025 			   I40E_GLPRT_PTC9522L(hw->port),
1026 			   pf->stat_offsets_loaded,
1027 			   &osd->tx_size_big, &nsd->tx_size_big);
1028 
1029 	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1030 			   pf->stat_offsets_loaded,
1031 			   &osd->rx_undersize, &nsd->rx_undersize);
1032 	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1033 			   pf->stat_offsets_loaded,
1034 			   &osd->rx_fragments, &nsd->rx_fragments);
1035 	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1036 			   pf->stat_offsets_loaded,
1037 			   &osd->rx_oversize, &nsd->rx_oversize);
1038 	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1039 			   pf->stat_offsets_loaded,
1040 			   &osd->rx_jabber, &nsd->rx_jabber);
1041 
1042 	/* FDIR stats */
1043 	i40e_stat_update32(hw,
1044 			   I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1045 			   pf->stat_offsets_loaded,
1046 			   &osd->fd_atr_match, &nsd->fd_atr_match);
1047 	i40e_stat_update32(hw,
1048 			   I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1049 			   pf->stat_offsets_loaded,
1050 			   &osd->fd_sb_match, &nsd->fd_sb_match);
1051 	i40e_stat_update32(hw,
1052 		      I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1053 		      pf->stat_offsets_loaded,
1054 		      &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1055 
1056 	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1057 	nsd->tx_lpi_status =
1058 		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1059 			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1060 	nsd->rx_lpi_status =
1061 		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1062 			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1063 	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1064 			   pf->stat_offsets_loaded,
1065 			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1066 	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1067 			   pf->stat_offsets_loaded,
1068 			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1069 
1070 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1071 	    !(pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED))
1072 		nsd->fd_sb_status = true;
1073 	else
1074 		nsd->fd_sb_status = false;
1075 
1076 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1077 	    !(pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED))
1078 		nsd->fd_atr_status = true;
1079 	else
1080 		nsd->fd_atr_status = false;
1081 
1082 	pf->stat_offsets_loaded = true;
1083 }
1084 
1085 /**
1086  * i40e_update_stats - Update the various statistics counters.
1087  * @vsi: the VSI to be updated
1088  *
1089  * Update the various stats for this VSI and its related entities.
1090  **/
1091 void i40e_update_stats(struct i40e_vsi *vsi)
1092 {
1093 	struct i40e_pf *pf = vsi->back;
1094 
1095 	if (vsi == pf->vsi[pf->lan_vsi])
1096 		i40e_update_pf_stats(pf);
1097 
1098 	i40e_update_vsi_stats(vsi);
1099 }
1100 
1101 /**
1102  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1103  * @vsi: the VSI to be searched
1104  * @macaddr: the MAC address
1105  * @vlan: the vlan
1106  *
1107  * Returns ptr to the filter object or NULL
1108  **/
1109 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1110 						const u8 *macaddr, s16 vlan)
1111 {
1112 	struct i40e_mac_filter *f;
1113 	u64 key;
1114 
1115 	if (!vsi || !macaddr)
1116 		return NULL;
1117 
1118 	key = i40e_addr_to_hkey(macaddr);
1119 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1120 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1121 		    (vlan == f->vlan))
1122 			return f;
1123 	}
1124 	return NULL;
1125 }
1126 
1127 /**
1128  * i40e_find_mac - Find a mac addr in the macvlan filters list
1129  * @vsi: the VSI to be searched
1130  * @macaddr: the MAC address we are searching for
1131  *
1132  * Returns the first filter with the provided MAC address or NULL if
1133  * MAC address was not found
1134  **/
1135 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1136 {
1137 	struct i40e_mac_filter *f;
1138 	u64 key;
1139 
1140 	if (!vsi || !macaddr)
1141 		return NULL;
1142 
1143 	key = i40e_addr_to_hkey(macaddr);
1144 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1145 		if ((ether_addr_equal(macaddr, f->macaddr)))
1146 			return f;
1147 	}
1148 	return NULL;
1149 }
1150 
1151 /**
1152  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1153  * @vsi: the VSI to be searched
1154  *
1155  * Returns true if VSI is in vlan mode or false otherwise
1156  **/
1157 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1158 {
1159 	/* If we have a PVID, always operate in VLAN mode */
1160 	if (vsi->info.pvid)
1161 		return true;
1162 
1163 	/* We need to operate in VLAN mode whenever we have any filters with
1164 	 * a VLAN other than I40E_VLAN_ALL. We could check the table each
1165 	 * time, incurring search cost repeatedly. However, we can notice two
1166 	 * things:
1167 	 *
1168 	 * 1) the only place where we can gain a VLAN filter is in
1169 	 *    i40e_add_filter.
1170 	 *
1171 	 * 2) the only place where filters are actually removed is in
1172 	 *    i40e_sync_filters_subtask.
1173 	 *
1174 	 * Thus, we can simply use a boolean value, has_vlan_filters which we
1175 	 * will set to true when we add a VLAN filter in i40e_add_filter. Then
1176 	 * we have to perform the full search after deleting filters in
1177 	 * i40e_sync_filters_subtask, but we already have to search
1178 	 * filters here and can perform the check at the same time. This
1179 	 * results in avoiding embedding a loop for VLAN mode inside another
1180 	 * loop over all the filters, and should maintain correctness as noted
1181 	 * above.
1182 	 */
1183 	return vsi->has_vlan_filter;
1184 }
1185 
1186 /**
1187  * i40e_correct_mac_vlan_filters - Correct non-VLAN filters if necessary
1188  * @vsi: the VSI to configure
1189  * @tmp_add_list: list of filters ready to be added
1190  * @tmp_del_list: list of filters ready to be deleted
1191  * @vlan_filters: the number of active VLAN filters
1192  *
1193  * Update VLAN=0 and VLAN=-1 (I40E_VLAN_ANY) filters properly so that they
1194  * behave as expected. If we have any active VLAN filters remaining or about
1195  * to be added then we need to update non-VLAN filters to be marked as VLAN=0
1196  * so that they only match against untagged traffic. If we no longer have any
1197  * active VLAN filters, we need to make all non-VLAN filters marked as VLAN=-1
1198  * so that they match against both tagged and untagged traffic. In this way,
1199  * we ensure that we correctly receive the desired traffic. This ensures that
1200  * when we have an active VLAN we will receive only untagged traffic and
1201  * traffic matching active VLANs. If we have no active VLANs then we will
1202  * operate in non-VLAN mode and receive all traffic, tagged or untagged.
1203  *
1204  * Finally, in a similar fashion, this function also corrects filters when
1205  * there is an active PVID assigned to this VSI.
1206  *
1207  * In case of memory allocation failure return -ENOMEM. Otherwise, return 0.
1208  *
1209  * This function is only expected to be called from within
1210  * i40e_sync_vsi_filters.
1211  *
1212  * NOTE: This function expects to be called while under the
1213  * mac_filter_hash_lock
1214  */
1215 static int i40e_correct_mac_vlan_filters(struct i40e_vsi *vsi,
1216 					 struct hlist_head *tmp_add_list,
1217 					 struct hlist_head *tmp_del_list,
1218 					 int vlan_filters)
1219 {
1220 	s16 pvid = le16_to_cpu(vsi->info.pvid);
1221 	struct i40e_mac_filter *f, *add_head;
1222 	struct i40e_new_mac_filter *new;
1223 	struct hlist_node *h;
1224 	int bkt, new_vlan;
1225 
1226 	/* To determine if a particular filter needs to be replaced we
1227 	 * have the three following conditions:
1228 	 *
1229 	 * a) if we have a PVID assigned, then all filters which are
1230 	 *    not marked as VLAN=PVID must be replaced with filters that
1231 	 *    are.
1232 	 * b) otherwise, if we have any active VLANS, all filters
1233 	 *    which are marked as VLAN=-1 must be replaced with
1234 	 *    filters marked as VLAN=0
1235 	 * c) finally, if we do not have any active VLANS, all filters
1236 	 *    which are marked as VLAN=0 must be replaced with filters
1237 	 *    marked as VLAN=-1
1238 	 */
1239 
1240 	/* Update the filters about to be added in place */
1241 	hlist_for_each_entry(new, tmp_add_list, hlist) {
1242 		if (pvid && new->f->vlan != pvid)
1243 			new->f->vlan = pvid;
1244 		else if (vlan_filters && new->f->vlan == I40E_VLAN_ANY)
1245 			new->f->vlan = 0;
1246 		else if (!vlan_filters && new->f->vlan == 0)
1247 			new->f->vlan = I40E_VLAN_ANY;
1248 	}
1249 
1250 	/* Update the remaining active filters */
1251 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1252 		/* Combine the checks for whether a filter needs to be changed
1253 		 * and then determine the new VLAN inside the if block, in
1254 		 * order to avoid duplicating code for adding the new filter
1255 		 * then deleting the old filter.
1256 		 */
1257 		if ((pvid && f->vlan != pvid) ||
1258 		    (vlan_filters && f->vlan == I40E_VLAN_ANY) ||
1259 		    (!vlan_filters && f->vlan == 0)) {
1260 			/* Determine the new vlan we will be adding */
1261 			if (pvid)
1262 				new_vlan = pvid;
1263 			else if (vlan_filters)
1264 				new_vlan = 0;
1265 			else
1266 				new_vlan = I40E_VLAN_ANY;
1267 
1268 			/* Create the new filter */
1269 			add_head = i40e_add_filter(vsi, f->macaddr, new_vlan);
1270 			if (!add_head)
1271 				return -ENOMEM;
1272 
1273 			/* Create a temporary i40e_new_mac_filter */
1274 			new = kzalloc(sizeof(*new), GFP_ATOMIC);
1275 			if (!new)
1276 				return -ENOMEM;
1277 
1278 			new->f = add_head;
1279 			new->state = add_head->state;
1280 
1281 			/* Add the new filter to the tmp list */
1282 			hlist_add_head(&new->hlist, tmp_add_list);
1283 
1284 			/* Put the original filter into the delete list */
1285 			f->state = I40E_FILTER_REMOVE;
1286 			hash_del(&f->hlist);
1287 			hlist_add_head(&f->hlist, tmp_del_list);
1288 		}
1289 	}
1290 
1291 	vsi->has_vlan_filter = !!vlan_filters;
1292 
1293 	return 0;
1294 }
1295 
1296 /**
1297  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1298  * @vsi: the PF Main VSI - inappropriate for any other VSI
1299  * @macaddr: the MAC address
1300  *
1301  * Remove whatever filter the firmware set up so the driver can manage
1302  * its own filtering intelligently.
1303  **/
1304 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1305 {
1306 	struct i40e_aqc_remove_macvlan_element_data element;
1307 	struct i40e_pf *pf = vsi->back;
1308 
1309 	/* Only appropriate for the PF main VSI */
1310 	if (vsi->type != I40E_VSI_MAIN)
1311 		return;
1312 
1313 	memset(&element, 0, sizeof(element));
1314 	ether_addr_copy(element.mac_addr, macaddr);
1315 	element.vlan_tag = 0;
1316 	/* Ignore error returns, some firmware does it this way... */
1317 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1318 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1319 
1320 	memset(&element, 0, sizeof(element));
1321 	ether_addr_copy(element.mac_addr, macaddr);
1322 	element.vlan_tag = 0;
1323 	/* ...and some firmware does it this way. */
1324 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1325 			I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1326 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1327 }
1328 
1329 /**
1330  * i40e_add_filter - Add a mac/vlan filter to the VSI
1331  * @vsi: the VSI to be searched
1332  * @macaddr: the MAC address
1333  * @vlan: the vlan
1334  *
1335  * Returns ptr to the filter object or NULL when no memory available.
1336  *
1337  * NOTE: This function is expected to be called with mac_filter_hash_lock
1338  * being held.
1339  **/
1340 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1341 					const u8 *macaddr, s16 vlan)
1342 {
1343 	struct i40e_mac_filter *f;
1344 	u64 key;
1345 
1346 	if (!vsi || !macaddr)
1347 		return NULL;
1348 
1349 	f = i40e_find_filter(vsi, macaddr, vlan);
1350 	if (!f) {
1351 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1352 		if (!f)
1353 			return NULL;
1354 
1355 		/* Update the boolean indicating if we need to function in
1356 		 * VLAN mode.
1357 		 */
1358 		if (vlan >= 0)
1359 			vsi->has_vlan_filter = true;
1360 
1361 		ether_addr_copy(f->macaddr, macaddr);
1362 		f->vlan = vlan;
1363 		/* If we're in overflow promisc mode, set the state directly
1364 		 * to failed, so we don't bother to try sending the filter
1365 		 * to the hardware.
1366 		 */
1367 		if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state))
1368 			f->state = I40E_FILTER_FAILED;
1369 		else
1370 			f->state = I40E_FILTER_NEW;
1371 		INIT_HLIST_NODE(&f->hlist);
1372 
1373 		key = i40e_addr_to_hkey(macaddr);
1374 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1375 
1376 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1377 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1378 	}
1379 
1380 	/* If we're asked to add a filter that has been marked for removal, it
1381 	 * is safe to simply restore it to active state. __i40e_del_filter
1382 	 * will have simply deleted any filters which were previously marked
1383 	 * NEW or FAILED, so if it is currently marked REMOVE it must have
1384 	 * previously been ACTIVE. Since we haven't yet run the sync filters
1385 	 * task, just restore this filter to the ACTIVE state so that the
1386 	 * sync task leaves it in place
1387 	 */
1388 	if (f->state == I40E_FILTER_REMOVE)
1389 		f->state = I40E_FILTER_ACTIVE;
1390 
1391 	return f;
1392 }
1393 
1394 /**
1395  * __i40e_del_filter - Remove a specific filter from the VSI
1396  * @vsi: VSI to remove from
1397  * @f: the filter to remove from the list
1398  *
1399  * This function should be called instead of i40e_del_filter only if you know
1400  * the exact filter you will remove already, such as via i40e_find_filter or
1401  * i40e_find_mac.
1402  *
1403  * NOTE: This function is expected to be called with mac_filter_hash_lock
1404  * being held.
1405  * ANOTHER NOTE: This function MUST be called from within the context of
1406  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1407  * instead of list_for_each_entry().
1408  **/
1409 void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1410 {
1411 	if (!f)
1412 		return;
1413 
1414 	/* If the filter was never added to firmware then we can just delete it
1415 	 * directly and we don't want to set the status to remove or else an
1416 	 * admin queue command will unnecessarily fire.
1417 	 */
1418 	if ((f->state == I40E_FILTER_FAILED) ||
1419 	    (f->state == I40E_FILTER_NEW)) {
1420 		hash_del(&f->hlist);
1421 		kfree(f);
1422 	} else {
1423 		f->state = I40E_FILTER_REMOVE;
1424 	}
1425 
1426 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1427 	vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1428 }
1429 
1430 /**
1431  * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1432  * @vsi: the VSI to be searched
1433  * @macaddr: the MAC address
1434  * @vlan: the VLAN
1435  *
1436  * NOTE: This function is expected to be called with mac_filter_hash_lock
1437  * being held.
1438  * ANOTHER NOTE: This function MUST be called from within the context of
1439  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1440  * instead of list_for_each_entry().
1441  **/
1442 void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1443 {
1444 	struct i40e_mac_filter *f;
1445 
1446 	if (!vsi || !macaddr)
1447 		return;
1448 
1449 	f = i40e_find_filter(vsi, macaddr, vlan);
1450 	__i40e_del_filter(vsi, f);
1451 }
1452 
1453 /**
1454  * i40e_add_mac_filter - Add a MAC filter for all active VLANs
1455  * @vsi: the VSI to be searched
1456  * @macaddr: the mac address to be filtered
1457  *
1458  * If we're not in VLAN mode, just add the filter to I40E_VLAN_ANY. Otherwise,
1459  * go through all the macvlan filters and add a macvlan filter for each
1460  * unique vlan that already exists. If a PVID has been assigned, instead only
1461  * add the macaddr to that VLAN.
1462  *
1463  * Returns last filter added on success, else NULL
1464  **/
1465 struct i40e_mac_filter *i40e_add_mac_filter(struct i40e_vsi *vsi,
1466 					    const u8 *macaddr)
1467 {
1468 	struct i40e_mac_filter *f, *add = NULL;
1469 	struct hlist_node *h;
1470 	int bkt;
1471 
1472 	if (vsi->info.pvid)
1473 		return i40e_add_filter(vsi, macaddr,
1474 				       le16_to_cpu(vsi->info.pvid));
1475 
1476 	if (!i40e_is_vsi_in_vlan(vsi))
1477 		return i40e_add_filter(vsi, macaddr, I40E_VLAN_ANY);
1478 
1479 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1480 		if (f->state == I40E_FILTER_REMOVE)
1481 			continue;
1482 		add = i40e_add_filter(vsi, macaddr, f->vlan);
1483 		if (!add)
1484 			return NULL;
1485 	}
1486 
1487 	return add;
1488 }
1489 
1490 /**
1491  * i40e_del_mac_filter - Remove a MAC filter from all VLANs
1492  * @vsi: the VSI to be searched
1493  * @macaddr: the mac address to be removed
1494  *
1495  * Removes a given MAC address from a VSI regardless of what VLAN it has been
1496  * associated with.
1497  *
1498  * Returns 0 for success, or error
1499  **/
1500 int i40e_del_mac_filter(struct i40e_vsi *vsi, const u8 *macaddr)
1501 {
1502 	struct i40e_mac_filter *f;
1503 	struct hlist_node *h;
1504 	bool found = false;
1505 	int bkt;
1506 
1507 	WARN(!spin_is_locked(&vsi->mac_filter_hash_lock),
1508 	     "Missing mac_filter_hash_lock\n");
1509 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1510 		if (ether_addr_equal(macaddr, f->macaddr)) {
1511 			__i40e_del_filter(vsi, f);
1512 			found = true;
1513 		}
1514 	}
1515 
1516 	if (found)
1517 		return 0;
1518 	else
1519 		return -ENOENT;
1520 }
1521 
1522 /**
1523  * i40e_set_mac - NDO callback to set mac address
1524  * @netdev: network interface device structure
1525  * @p: pointer to an address structure
1526  *
1527  * Returns 0 on success, negative on failure
1528  **/
1529 static int i40e_set_mac(struct net_device *netdev, void *p)
1530 {
1531 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1532 	struct i40e_vsi *vsi = np->vsi;
1533 	struct i40e_pf *pf = vsi->back;
1534 	struct i40e_hw *hw = &pf->hw;
1535 	struct sockaddr *addr = p;
1536 
1537 	if (!is_valid_ether_addr(addr->sa_data))
1538 		return -EADDRNOTAVAIL;
1539 
1540 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1541 		netdev_info(netdev, "already using mac address %pM\n",
1542 			    addr->sa_data);
1543 		return 0;
1544 	}
1545 
1546 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
1547 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
1548 		return -EADDRNOTAVAIL;
1549 
1550 	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1551 		netdev_info(netdev, "returning to hw mac address %pM\n",
1552 			    hw->mac.addr);
1553 	else
1554 		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1555 
1556 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1557 	i40e_del_mac_filter(vsi, netdev->dev_addr);
1558 	i40e_add_mac_filter(vsi, addr->sa_data);
1559 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1560 	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1561 	if (vsi->type == I40E_VSI_MAIN) {
1562 		i40e_status ret;
1563 
1564 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1565 						I40E_AQC_WRITE_TYPE_LAA_WOL,
1566 						addr->sa_data, NULL);
1567 		if (ret)
1568 			netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1569 				    i40e_stat_str(hw, ret),
1570 				    i40e_aq_str(hw, hw->aq.asq_last_status));
1571 	}
1572 
1573 	/* schedule our worker thread which will take care of
1574 	 * applying the new filter changes
1575 	 */
1576 	i40e_service_event_schedule(vsi->back);
1577 	return 0;
1578 }
1579 
1580 /**
1581  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1582  * @vsi: the VSI being setup
1583  * @ctxt: VSI context structure
1584  * @enabled_tc: Enabled TCs bitmap
1585  * @is_add: True if called before Add VSI
1586  *
1587  * Setup VSI queue mapping for enabled traffic classes.
1588  **/
1589 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1590 				     struct i40e_vsi_context *ctxt,
1591 				     u8 enabled_tc,
1592 				     bool is_add)
1593 {
1594 	struct i40e_pf *pf = vsi->back;
1595 	u16 sections = 0;
1596 	u8 netdev_tc = 0;
1597 	u16 numtc = 0;
1598 	u16 qcount;
1599 	u8 offset;
1600 	u16 qmap;
1601 	int i;
1602 	u16 num_tc_qps = 0;
1603 
1604 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1605 	offset = 0;
1606 
1607 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1608 		/* Find numtc from enabled TC bitmap */
1609 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1610 			if (enabled_tc & BIT(i)) /* TC is enabled */
1611 				numtc++;
1612 		}
1613 		if (!numtc) {
1614 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1615 			numtc = 1;
1616 		}
1617 	} else {
1618 		/* At least TC0 is enabled in case of non-DCB case */
1619 		numtc = 1;
1620 	}
1621 
1622 	vsi->tc_config.numtc = numtc;
1623 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1624 	/* Number of queues per enabled TC */
1625 	qcount = vsi->alloc_queue_pairs;
1626 
1627 	num_tc_qps = qcount / numtc;
1628 	num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1629 
1630 	/* Setup queue offset/count for all TCs for given VSI */
1631 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1632 		/* See if the given TC is enabled for the given VSI */
1633 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1634 			/* TC is enabled */
1635 			int pow, num_qps;
1636 
1637 			switch (vsi->type) {
1638 			case I40E_VSI_MAIN:
1639 				qcount = min_t(int, pf->alloc_rss_size,
1640 					       num_tc_qps);
1641 				break;
1642 			case I40E_VSI_FDIR:
1643 			case I40E_VSI_SRIOV:
1644 			case I40E_VSI_VMDQ2:
1645 			default:
1646 				qcount = num_tc_qps;
1647 				WARN_ON(i != 0);
1648 				break;
1649 			}
1650 			vsi->tc_config.tc_info[i].qoffset = offset;
1651 			vsi->tc_config.tc_info[i].qcount = qcount;
1652 
1653 			/* find the next higher power-of-2 of num queue pairs */
1654 			num_qps = qcount;
1655 			pow = 0;
1656 			while (num_qps && (BIT_ULL(pow) < qcount)) {
1657 				pow++;
1658 				num_qps >>= 1;
1659 			}
1660 
1661 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1662 			qmap =
1663 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1664 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1665 
1666 			offset += qcount;
1667 		} else {
1668 			/* TC is not enabled so set the offset to
1669 			 * default queue and allocate one queue
1670 			 * for the given TC.
1671 			 */
1672 			vsi->tc_config.tc_info[i].qoffset = 0;
1673 			vsi->tc_config.tc_info[i].qcount = 1;
1674 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1675 
1676 			qmap = 0;
1677 		}
1678 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1679 	}
1680 
1681 	/* Set actual Tx/Rx queue pairs */
1682 	vsi->num_queue_pairs = offset;
1683 	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1684 		if (vsi->req_queue_pairs > 0)
1685 			vsi->num_queue_pairs = vsi->req_queue_pairs;
1686 		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1687 			vsi->num_queue_pairs = pf->num_lan_msix;
1688 	}
1689 
1690 	/* Scheduler section valid can only be set for ADD VSI */
1691 	if (is_add) {
1692 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1693 
1694 		ctxt->info.up_enable_bits = enabled_tc;
1695 	}
1696 	if (vsi->type == I40E_VSI_SRIOV) {
1697 		ctxt->info.mapping_flags |=
1698 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1699 		for (i = 0; i < vsi->num_queue_pairs; i++)
1700 			ctxt->info.queue_mapping[i] =
1701 					       cpu_to_le16(vsi->base_queue + i);
1702 	} else {
1703 		ctxt->info.mapping_flags |=
1704 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1705 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1706 	}
1707 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1708 }
1709 
1710 /**
1711  * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1712  * @netdev: the netdevice
1713  * @addr: address to add
1714  *
1715  * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1716  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1717  */
1718 static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1719 {
1720 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1721 	struct i40e_vsi *vsi = np->vsi;
1722 
1723 	if (i40e_add_mac_filter(vsi, addr))
1724 		return 0;
1725 	else
1726 		return -ENOMEM;
1727 }
1728 
1729 /**
1730  * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1731  * @netdev: the netdevice
1732  * @addr: address to add
1733  *
1734  * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1735  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1736  */
1737 static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1738 {
1739 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1740 	struct i40e_vsi *vsi = np->vsi;
1741 
1742 	i40e_del_mac_filter(vsi, addr);
1743 
1744 	return 0;
1745 }
1746 
1747 /**
1748  * i40e_set_rx_mode - NDO callback to set the netdev filters
1749  * @netdev: network interface device structure
1750  **/
1751 static void i40e_set_rx_mode(struct net_device *netdev)
1752 {
1753 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1754 	struct i40e_vsi *vsi = np->vsi;
1755 
1756 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1757 
1758 	__dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1759 	__dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1760 
1761 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1762 
1763 	/* check for other flag changes */
1764 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1765 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1766 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1767 	}
1768 
1769 	/* schedule our worker thread which will take care of
1770 	 * applying the new filter changes
1771 	 */
1772 	i40e_service_event_schedule(vsi->back);
1773 }
1774 
1775 /**
1776  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1777  * @vsi: Pointer to VSI struct
1778  * @from: Pointer to list which contains MAC filter entries - changes to
1779  *        those entries needs to be undone.
1780  *
1781  * MAC filter entries from this list were slated for deletion.
1782  **/
1783 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1784 					 struct hlist_head *from)
1785 {
1786 	struct i40e_mac_filter *f;
1787 	struct hlist_node *h;
1788 
1789 	hlist_for_each_entry_safe(f, h, from, hlist) {
1790 		u64 key = i40e_addr_to_hkey(f->macaddr);
1791 
1792 		/* Move the element back into MAC filter list*/
1793 		hlist_del(&f->hlist);
1794 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1795 	}
1796 }
1797 
1798 /**
1799  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1800  * @vsi: Pointer to vsi struct
1801  * @from: Pointer to list which contains MAC filter entries - changes to
1802  *        those entries needs to be undone.
1803  *
1804  * MAC filter entries from this list were slated for addition.
1805  **/
1806 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi,
1807 					 struct hlist_head *from)
1808 {
1809 	struct i40e_new_mac_filter *new;
1810 	struct hlist_node *h;
1811 
1812 	hlist_for_each_entry_safe(new, h, from, hlist) {
1813 		/* We can simply free the wrapper structure */
1814 		hlist_del(&new->hlist);
1815 		kfree(new);
1816 	}
1817 }
1818 
1819 /**
1820  * i40e_next_entry - Get the next non-broadcast filter from a list
1821  * @next: pointer to filter in list
1822  *
1823  * Returns the next non-broadcast filter in the list. Required so that we
1824  * ignore broadcast filters within the list, since these are not handled via
1825  * the normal firmware update path.
1826  */
1827 static
1828 struct i40e_new_mac_filter *i40e_next_filter(struct i40e_new_mac_filter *next)
1829 {
1830 	hlist_for_each_entry_continue(next, hlist) {
1831 		if (!is_broadcast_ether_addr(next->f->macaddr))
1832 			return next;
1833 	}
1834 
1835 	return NULL;
1836 }
1837 
1838 /**
1839  * i40e_update_filter_state - Update filter state based on return data
1840  * from firmware
1841  * @count: Number of filters added
1842  * @add_list: return data from fw
1843  * @head: pointer to first filter in current batch
1844  *
1845  * MAC filter entries from list were slated to be added to device. Returns
1846  * number of successful filters. Note that 0 does NOT mean success!
1847  **/
1848 static int
1849 i40e_update_filter_state(int count,
1850 			 struct i40e_aqc_add_macvlan_element_data *add_list,
1851 			 struct i40e_new_mac_filter *add_head)
1852 {
1853 	int retval = 0;
1854 	int i;
1855 
1856 	for (i = 0; i < count; i++) {
1857 		/* Always check status of each filter. We don't need to check
1858 		 * the firmware return status because we pre-set the filter
1859 		 * status to I40E_AQC_MM_ERR_NO_RES when sending the filter
1860 		 * request to the adminq. Thus, if it no longer matches then
1861 		 * we know the filter is active.
1862 		 */
1863 		if (add_list[i].match_method == I40E_AQC_MM_ERR_NO_RES) {
1864 			add_head->state = I40E_FILTER_FAILED;
1865 		} else {
1866 			add_head->state = I40E_FILTER_ACTIVE;
1867 			retval++;
1868 		}
1869 
1870 		add_head = i40e_next_filter(add_head);
1871 		if (!add_head)
1872 			break;
1873 	}
1874 
1875 	return retval;
1876 }
1877 
1878 /**
1879  * i40e_aqc_del_filters - Request firmware to delete a set of filters
1880  * @vsi: ptr to the VSI
1881  * @vsi_name: name to display in messages
1882  * @list: the list of filters to send to firmware
1883  * @num_del: the number of filters to delete
1884  * @retval: Set to -EIO on failure to delete
1885  *
1886  * Send a request to firmware via AdminQ to delete a set of filters. Uses
1887  * *retval instead of a return value so that success does not force ret_val to
1888  * be set to 0. This ensures that a sequence of calls to this function
1889  * preserve the previous value of *retval on successful delete.
1890  */
1891 static
1892 void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
1893 			  struct i40e_aqc_remove_macvlan_element_data *list,
1894 			  int num_del, int *retval)
1895 {
1896 	struct i40e_hw *hw = &vsi->back->hw;
1897 	i40e_status aq_ret;
1898 	int aq_err;
1899 
1900 	aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
1901 	aq_err = hw->aq.asq_last_status;
1902 
1903 	/* Explicitly ignore and do not report when firmware returns ENOENT */
1904 	if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1905 		*retval = -EIO;
1906 		dev_info(&vsi->back->pdev->dev,
1907 			 "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1908 			 vsi_name, i40e_stat_str(hw, aq_ret),
1909 			 i40e_aq_str(hw, aq_err));
1910 	}
1911 }
1912 
1913 /**
1914  * i40e_aqc_add_filters - Request firmware to add a set of filters
1915  * @vsi: ptr to the VSI
1916  * @vsi_name: name to display in messages
1917  * @list: the list of filters to send to firmware
1918  * @add_head: Position in the add hlist
1919  * @num_add: the number of filters to add
1920  * @promisc_change: set to true on exit if promiscuous mode was forced on
1921  *
1922  * Send a request to firmware via AdminQ to add a chunk of filters. Will set
1923  * promisc_changed to true if the firmware has run out of space for more
1924  * filters.
1925  */
1926 static
1927 void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
1928 			  struct i40e_aqc_add_macvlan_element_data *list,
1929 			  struct i40e_new_mac_filter *add_head,
1930 			  int num_add, bool *promisc_changed)
1931 {
1932 	struct i40e_hw *hw = &vsi->back->hw;
1933 	int aq_err, fcnt;
1934 
1935 	i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
1936 	aq_err = hw->aq.asq_last_status;
1937 	fcnt = i40e_update_filter_state(num_add, list, add_head);
1938 
1939 	if (fcnt != num_add) {
1940 		*promisc_changed = true;
1941 		set_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
1942 		dev_warn(&vsi->back->pdev->dev,
1943 			 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
1944 			 i40e_aq_str(hw, aq_err),
1945 			 vsi_name);
1946 	}
1947 }
1948 
1949 /**
1950  * i40e_aqc_broadcast_filter - Set promiscuous broadcast flags
1951  * @vsi: pointer to the VSI
1952  * @f: filter data
1953  *
1954  * This function sets or clears the promiscuous broadcast flags for VLAN
1955  * filters in order to properly receive broadcast frames. Assumes that only
1956  * broadcast filters are passed.
1957  *
1958  * Returns status indicating success or failure;
1959  **/
1960 static i40e_status
1961 i40e_aqc_broadcast_filter(struct i40e_vsi *vsi, const char *vsi_name,
1962 			  struct i40e_mac_filter *f)
1963 {
1964 	bool enable = f->state == I40E_FILTER_NEW;
1965 	struct i40e_hw *hw = &vsi->back->hw;
1966 	i40e_status aq_ret;
1967 
1968 	if (f->vlan == I40E_VLAN_ANY) {
1969 		aq_ret = i40e_aq_set_vsi_broadcast(hw,
1970 						   vsi->seid,
1971 						   enable,
1972 						   NULL);
1973 	} else {
1974 		aq_ret = i40e_aq_set_vsi_bc_promisc_on_vlan(hw,
1975 							    vsi->seid,
1976 							    enable,
1977 							    f->vlan,
1978 							    NULL);
1979 	}
1980 
1981 	if (aq_ret)
1982 		dev_warn(&vsi->back->pdev->dev,
1983 			 "Error %s setting broadcast promiscuous mode on %s\n",
1984 			 i40e_aq_str(hw, hw->aq.asq_last_status),
1985 			 vsi_name);
1986 
1987 	return aq_ret;
1988 }
1989 
1990 /**
1991  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1992  * @vsi: ptr to the VSI
1993  *
1994  * Push any outstanding VSI filter changes through the AdminQ.
1995  *
1996  * Returns 0 or error value
1997  **/
1998 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1999 {
2000 	struct hlist_head tmp_add_list, tmp_del_list;
2001 	struct i40e_mac_filter *f;
2002 	struct i40e_new_mac_filter *new, *add_head = NULL;
2003 	struct i40e_hw *hw = &vsi->back->hw;
2004 	unsigned int failed_filters = 0;
2005 	unsigned int vlan_filters = 0;
2006 	bool promisc_changed = false;
2007 	char vsi_name[16] = "PF";
2008 	int filter_list_len = 0;
2009 	i40e_status aq_ret = 0;
2010 	u32 changed_flags = 0;
2011 	struct hlist_node *h;
2012 	struct i40e_pf *pf;
2013 	int num_add = 0;
2014 	int num_del = 0;
2015 	int retval = 0;
2016 	u16 cmd_flags;
2017 	int list_size;
2018 	int bkt;
2019 
2020 	/* empty array typed pointers, kcalloc later */
2021 	struct i40e_aqc_add_macvlan_element_data *add_list;
2022 	struct i40e_aqc_remove_macvlan_element_data *del_list;
2023 
2024 	while (test_and_set_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state))
2025 		usleep_range(1000, 2000);
2026 	pf = vsi->back;
2027 
2028 	if (vsi->netdev) {
2029 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
2030 		vsi->current_netdev_flags = vsi->netdev->flags;
2031 	}
2032 
2033 	INIT_HLIST_HEAD(&tmp_add_list);
2034 	INIT_HLIST_HEAD(&tmp_del_list);
2035 
2036 	if (vsi->type == I40E_VSI_SRIOV)
2037 		snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
2038 	else if (vsi->type != I40E_VSI_MAIN)
2039 		snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
2040 
2041 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
2042 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
2043 
2044 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2045 		/* Create a list of filters to delete. */
2046 		hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2047 			if (f->state == I40E_FILTER_REMOVE) {
2048 				/* Move the element into temporary del_list */
2049 				hash_del(&f->hlist);
2050 				hlist_add_head(&f->hlist, &tmp_del_list);
2051 
2052 				/* Avoid counting removed filters */
2053 				continue;
2054 			}
2055 			if (f->state == I40E_FILTER_NEW) {
2056 				/* Create a temporary i40e_new_mac_filter */
2057 				new = kzalloc(sizeof(*new), GFP_ATOMIC);
2058 				if (!new)
2059 					goto err_no_memory_locked;
2060 
2061 				/* Store pointer to the real filter */
2062 				new->f = f;
2063 				new->state = f->state;
2064 
2065 				/* Add it to the hash list */
2066 				hlist_add_head(&new->hlist, &tmp_add_list);
2067 			}
2068 
2069 			/* Count the number of active (current and new) VLAN
2070 			 * filters we have now. Does not count filters which
2071 			 * are marked for deletion.
2072 			 */
2073 			if (f->vlan > 0)
2074 				vlan_filters++;
2075 		}
2076 
2077 		retval = i40e_correct_mac_vlan_filters(vsi,
2078 						       &tmp_add_list,
2079 						       &tmp_del_list,
2080 						       vlan_filters);
2081 		if (retval)
2082 			goto err_no_memory_locked;
2083 
2084 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2085 	}
2086 
2087 	/* Now process 'del_list' outside the lock */
2088 	if (!hlist_empty(&tmp_del_list)) {
2089 		filter_list_len = hw->aq.asq_buf_size /
2090 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2091 		list_size = filter_list_len *
2092 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2093 		del_list = kzalloc(list_size, GFP_ATOMIC);
2094 		if (!del_list)
2095 			goto err_no_memory;
2096 
2097 		hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2098 			cmd_flags = 0;
2099 
2100 			/* handle broadcast filters by updating the broadcast
2101 			 * promiscuous flag and release filter list.
2102 			 */
2103 			if (is_broadcast_ether_addr(f->macaddr)) {
2104 				i40e_aqc_broadcast_filter(vsi, vsi_name, f);
2105 
2106 				hlist_del(&f->hlist);
2107 				kfree(f);
2108 				continue;
2109 			}
2110 
2111 			/* add to delete list */
2112 			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2113 			if (f->vlan == I40E_VLAN_ANY) {
2114 				del_list[num_del].vlan_tag = 0;
2115 				cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2116 			} else {
2117 				del_list[num_del].vlan_tag =
2118 					cpu_to_le16((u16)(f->vlan));
2119 			}
2120 
2121 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2122 			del_list[num_del].flags = cmd_flags;
2123 			num_del++;
2124 
2125 			/* flush a full buffer */
2126 			if (num_del == filter_list_len) {
2127 				i40e_aqc_del_filters(vsi, vsi_name, del_list,
2128 						     num_del, &retval);
2129 				memset(del_list, 0, list_size);
2130 				num_del = 0;
2131 			}
2132 			/* Release memory for MAC filter entries which were
2133 			 * synced up with HW.
2134 			 */
2135 			hlist_del(&f->hlist);
2136 			kfree(f);
2137 		}
2138 
2139 		if (num_del) {
2140 			i40e_aqc_del_filters(vsi, vsi_name, del_list,
2141 					     num_del, &retval);
2142 		}
2143 
2144 		kfree(del_list);
2145 		del_list = NULL;
2146 	}
2147 
2148 	if (!hlist_empty(&tmp_add_list)) {
2149 		/* Do all the adds now. */
2150 		filter_list_len = hw->aq.asq_buf_size /
2151 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2152 		list_size = filter_list_len *
2153 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2154 		add_list = kzalloc(list_size, GFP_ATOMIC);
2155 		if (!add_list)
2156 			goto err_no_memory;
2157 
2158 		num_add = 0;
2159 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2160 			if (test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2161 				     vsi->state)) {
2162 				new->state = I40E_FILTER_FAILED;
2163 				continue;
2164 			}
2165 
2166 			/* handle broadcast filters by updating the broadcast
2167 			 * promiscuous flag instead of adding a MAC filter.
2168 			 */
2169 			if (is_broadcast_ether_addr(new->f->macaddr)) {
2170 				if (i40e_aqc_broadcast_filter(vsi, vsi_name,
2171 							      new->f))
2172 					new->state = I40E_FILTER_FAILED;
2173 				else
2174 					new->state = I40E_FILTER_ACTIVE;
2175 				continue;
2176 			}
2177 
2178 			/* add to add array */
2179 			if (num_add == 0)
2180 				add_head = new;
2181 			cmd_flags = 0;
2182 			ether_addr_copy(add_list[num_add].mac_addr,
2183 					new->f->macaddr);
2184 			if (new->f->vlan == I40E_VLAN_ANY) {
2185 				add_list[num_add].vlan_tag = 0;
2186 				cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2187 			} else {
2188 				add_list[num_add].vlan_tag =
2189 					cpu_to_le16((u16)(new->f->vlan));
2190 			}
2191 			add_list[num_add].queue_number = 0;
2192 			/* set invalid match method for later detection */
2193 			add_list[num_add].match_method = I40E_AQC_MM_ERR_NO_RES;
2194 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2195 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
2196 			num_add++;
2197 
2198 			/* flush a full buffer */
2199 			if (num_add == filter_list_len) {
2200 				i40e_aqc_add_filters(vsi, vsi_name, add_list,
2201 						     add_head, num_add,
2202 						     &promisc_changed);
2203 				memset(add_list, 0, list_size);
2204 				num_add = 0;
2205 			}
2206 		}
2207 		if (num_add) {
2208 			i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2209 					     num_add, &promisc_changed);
2210 		}
2211 		/* Now move all of the filters from the temp add list back to
2212 		 * the VSI's list.
2213 		 */
2214 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2215 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2216 			/* Only update the state if we're still NEW */
2217 			if (new->f->state == I40E_FILTER_NEW)
2218 				new->f->state = new->state;
2219 			hlist_del(&new->hlist);
2220 			kfree(new);
2221 		}
2222 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2223 		kfree(add_list);
2224 		add_list = NULL;
2225 	}
2226 
2227 	/* Determine the number of active and failed filters. */
2228 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2229 	vsi->active_filters = 0;
2230 	hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2231 		if (f->state == I40E_FILTER_ACTIVE)
2232 			vsi->active_filters++;
2233 		else if (f->state == I40E_FILTER_FAILED)
2234 			failed_filters++;
2235 	}
2236 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2237 
2238 	/* If promiscuous mode has changed, we need to calculate a new
2239 	 * threshold for when we are safe to exit
2240 	 */
2241 	if (promisc_changed)
2242 		vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
2243 
2244 	/* Check if we are able to exit overflow promiscuous mode. We can
2245 	 * safely exit if we didn't just enter, we no longer have any failed
2246 	 * filters, and we have reduced filters below the threshold value.
2247 	 */
2248 	if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state) &&
2249 	    !promisc_changed && !failed_filters &&
2250 	    (vsi->active_filters < vsi->promisc_threshold)) {
2251 		dev_info(&pf->pdev->dev,
2252 			 "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2253 			 vsi_name);
2254 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2255 		promisc_changed = true;
2256 		vsi->promisc_threshold = 0;
2257 	}
2258 
2259 	/* if the VF is not trusted do not do promisc */
2260 	if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2261 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2262 		goto out;
2263 	}
2264 
2265 	/* check for changes in promiscuous modes */
2266 	if (changed_flags & IFF_ALLMULTI) {
2267 		bool cur_multipromisc;
2268 
2269 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2270 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2271 							       vsi->seid,
2272 							       cur_multipromisc,
2273 							       NULL);
2274 		if (aq_ret) {
2275 			retval = i40e_aq_rc_to_posix(aq_ret,
2276 						     hw->aq.asq_last_status);
2277 			dev_info(&pf->pdev->dev,
2278 				 "set multi promisc failed on %s, err %s aq_err %s\n",
2279 				 vsi_name,
2280 				 i40e_stat_str(hw, aq_ret),
2281 				 i40e_aq_str(hw, hw->aq.asq_last_status));
2282 		}
2283 	}
2284 
2285 	if ((changed_flags & IFF_PROMISC) || promisc_changed) {
2286 		bool cur_promisc;
2287 
2288 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2289 			       test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2290 					vsi->state));
2291 		if ((vsi->type == I40E_VSI_MAIN) &&
2292 		    (pf->lan_veb != I40E_NO_VEB) &&
2293 		    !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2294 			/* set defport ON for Main VSI instead of true promisc
2295 			 * this way we will get all unicast/multicast and VLAN
2296 			 * promisc behavior but will not get VF or VMDq traffic
2297 			 * replicated on the Main VSI.
2298 			 */
2299 			if (pf->cur_promisc != cur_promisc) {
2300 				pf->cur_promisc = cur_promisc;
2301 				if (cur_promisc)
2302 					aq_ret =
2303 					      i40e_aq_set_default_vsi(hw,
2304 								      vsi->seid,
2305 								      NULL);
2306 				else
2307 					aq_ret =
2308 					    i40e_aq_clear_default_vsi(hw,
2309 								      vsi->seid,
2310 								      NULL);
2311 				if (aq_ret) {
2312 					retval = i40e_aq_rc_to_posix(aq_ret,
2313 							hw->aq.asq_last_status);
2314 					dev_info(&pf->pdev->dev,
2315 						 "Set default VSI failed on %s, err %s, aq_err %s\n",
2316 						 vsi_name,
2317 						 i40e_stat_str(hw, aq_ret),
2318 						 i40e_aq_str(hw,
2319 						     hw->aq.asq_last_status));
2320 				}
2321 			}
2322 		} else {
2323 			aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2324 							  hw,
2325 							  vsi->seid,
2326 							  cur_promisc, NULL,
2327 							  true);
2328 			if (aq_ret) {
2329 				retval =
2330 				i40e_aq_rc_to_posix(aq_ret,
2331 						    hw->aq.asq_last_status);
2332 				dev_info(&pf->pdev->dev,
2333 					 "set unicast promisc failed on %s, err %s, aq_err %s\n",
2334 					 vsi_name,
2335 					 i40e_stat_str(hw, aq_ret),
2336 					 i40e_aq_str(hw,
2337 						     hw->aq.asq_last_status));
2338 			}
2339 			aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2340 							  hw,
2341 							  vsi->seid,
2342 							  cur_promisc, NULL);
2343 			if (aq_ret) {
2344 				retval =
2345 				i40e_aq_rc_to_posix(aq_ret,
2346 						    hw->aq.asq_last_status);
2347 				dev_info(&pf->pdev->dev,
2348 					 "set multicast promisc failed on %s, err %s, aq_err %s\n",
2349 					 vsi_name,
2350 					 i40e_stat_str(hw, aq_ret),
2351 					 i40e_aq_str(hw,
2352 						     hw->aq.asq_last_status));
2353 			}
2354 		}
2355 		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2356 						   vsi->seid,
2357 						   cur_promisc, NULL);
2358 		if (aq_ret) {
2359 			retval = i40e_aq_rc_to_posix(aq_ret,
2360 						     pf->hw.aq.asq_last_status);
2361 			dev_info(&pf->pdev->dev,
2362 				 "set brdcast promisc failed, err %s, aq_err %s\n",
2363 					 i40e_stat_str(hw, aq_ret),
2364 					 i40e_aq_str(hw,
2365 						     hw->aq.asq_last_status));
2366 		}
2367 	}
2368 out:
2369 	/* if something went wrong then set the changed flag so we try again */
2370 	if (retval)
2371 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2372 
2373 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2374 	return retval;
2375 
2376 err_no_memory:
2377 	/* Restore elements on the temporary add and delete lists */
2378 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2379 err_no_memory_locked:
2380 	i40e_undo_del_filter_entries(vsi, &tmp_del_list);
2381 	i40e_undo_add_filter_entries(vsi, &tmp_add_list);
2382 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2383 
2384 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2385 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2386 	return -ENOMEM;
2387 }
2388 
2389 /**
2390  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2391  * @pf: board private structure
2392  **/
2393 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2394 {
2395 	int v;
2396 
2397 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2398 		return;
2399 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2400 
2401 	for (v = 0; v < pf->num_alloc_vsi; v++) {
2402 		if (pf->vsi[v] &&
2403 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2404 			int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2405 
2406 			if (ret) {
2407 				/* come back and try again later */
2408 				pf->flags |= I40E_FLAG_FILTER_SYNC;
2409 				break;
2410 			}
2411 		}
2412 	}
2413 }
2414 
2415 /**
2416  * i40e_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2417  * @vsi: the vsi
2418  **/
2419 static int i40e_max_xdp_frame_size(struct i40e_vsi *vsi)
2420 {
2421 	if (PAGE_SIZE >= 8192 || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
2422 		return I40E_RXBUFFER_2048;
2423 	else
2424 		return I40E_RXBUFFER_3072;
2425 }
2426 
2427 /**
2428  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2429  * @netdev: network interface device structure
2430  * @new_mtu: new value for maximum frame size
2431  *
2432  * Returns 0 on success, negative on failure
2433  **/
2434 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2435 {
2436 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2437 	struct i40e_vsi *vsi = np->vsi;
2438 	struct i40e_pf *pf = vsi->back;
2439 
2440 	if (i40e_enabled_xdp_vsi(vsi)) {
2441 		int frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2442 
2443 		if (frame_size > i40e_max_xdp_frame_size(vsi))
2444 			return -EINVAL;
2445 	}
2446 
2447 	netdev_info(netdev, "changing MTU from %d to %d\n",
2448 		    netdev->mtu, new_mtu);
2449 	netdev->mtu = new_mtu;
2450 	if (netif_running(netdev))
2451 		i40e_vsi_reinit_locked(vsi);
2452 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
2453 		      I40E_FLAG_CLIENT_L2_CHANGE);
2454 	return 0;
2455 }
2456 
2457 /**
2458  * i40e_ioctl - Access the hwtstamp interface
2459  * @netdev: network interface device structure
2460  * @ifr: interface request data
2461  * @cmd: ioctl command
2462  **/
2463 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2464 {
2465 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2466 	struct i40e_pf *pf = np->vsi->back;
2467 
2468 	switch (cmd) {
2469 	case SIOCGHWTSTAMP:
2470 		return i40e_ptp_get_ts_config(pf, ifr);
2471 	case SIOCSHWTSTAMP:
2472 		return i40e_ptp_set_ts_config(pf, ifr);
2473 	default:
2474 		return -EOPNOTSUPP;
2475 	}
2476 }
2477 
2478 /**
2479  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2480  * @vsi: the vsi being adjusted
2481  **/
2482 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2483 {
2484 	struct i40e_vsi_context ctxt;
2485 	i40e_status ret;
2486 
2487 	if ((vsi->info.valid_sections &
2488 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2489 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2490 		return;  /* already enabled */
2491 
2492 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2493 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2494 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2495 
2496 	ctxt.seid = vsi->seid;
2497 	ctxt.info = vsi->info;
2498 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2499 	if (ret) {
2500 		dev_info(&vsi->back->pdev->dev,
2501 			 "update vlan stripping failed, err %s aq_err %s\n",
2502 			 i40e_stat_str(&vsi->back->hw, ret),
2503 			 i40e_aq_str(&vsi->back->hw,
2504 				     vsi->back->hw.aq.asq_last_status));
2505 	}
2506 }
2507 
2508 /**
2509  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2510  * @vsi: the vsi being adjusted
2511  **/
2512 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2513 {
2514 	struct i40e_vsi_context ctxt;
2515 	i40e_status ret;
2516 
2517 	if ((vsi->info.valid_sections &
2518 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2519 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2520 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2521 		return;  /* already disabled */
2522 
2523 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2524 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2525 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2526 
2527 	ctxt.seid = vsi->seid;
2528 	ctxt.info = vsi->info;
2529 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2530 	if (ret) {
2531 		dev_info(&vsi->back->pdev->dev,
2532 			 "update vlan stripping failed, err %s aq_err %s\n",
2533 			 i40e_stat_str(&vsi->back->hw, ret),
2534 			 i40e_aq_str(&vsi->back->hw,
2535 				     vsi->back->hw.aq.asq_last_status));
2536 	}
2537 }
2538 
2539 /**
2540  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2541  * @netdev: network interface to be adjusted
2542  * @features: netdev features to test if VLAN offload is enabled or not
2543  **/
2544 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2545 {
2546 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2547 	struct i40e_vsi *vsi = np->vsi;
2548 
2549 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2550 		i40e_vlan_stripping_enable(vsi);
2551 	else
2552 		i40e_vlan_stripping_disable(vsi);
2553 }
2554 
2555 /**
2556  * i40e_add_vlan_all_mac - Add a MAC/VLAN filter for each existing MAC address
2557  * @vsi: the vsi being configured
2558  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2559  *
2560  * This is a helper function for adding a new MAC/VLAN filter with the
2561  * specified VLAN for each existing MAC address already in the hash table.
2562  * This function does *not* perform any accounting to update filters based on
2563  * VLAN mode.
2564  *
2565  * NOTE: this function expects to be called while under the
2566  * mac_filter_hash_lock
2567  **/
2568 int i40e_add_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2569 {
2570 	struct i40e_mac_filter *f, *add_f;
2571 	struct hlist_node *h;
2572 	int bkt;
2573 
2574 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2575 		if (f->state == I40E_FILTER_REMOVE)
2576 			continue;
2577 		add_f = i40e_add_filter(vsi, f->macaddr, vid);
2578 		if (!add_f) {
2579 			dev_info(&vsi->back->pdev->dev,
2580 				 "Could not add vlan filter %d for %pM\n",
2581 				 vid, f->macaddr);
2582 			return -ENOMEM;
2583 		}
2584 	}
2585 
2586 	return 0;
2587 }
2588 
2589 /**
2590  * i40e_vsi_add_vlan - Add VSI membership for given VLAN
2591  * @vsi: the VSI being configured
2592  * @vid: VLAN id to be added
2593  **/
2594 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, u16 vid)
2595 {
2596 	int err;
2597 
2598 	if (!vid || vsi->info.pvid)
2599 		return -EINVAL;
2600 
2601 	/* Locked once because all functions invoked below iterates list*/
2602 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2603 	err = i40e_add_vlan_all_mac(vsi, vid);
2604 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2605 	if (err)
2606 		return err;
2607 
2608 	/* schedule our worker thread which will take care of
2609 	 * applying the new filter changes
2610 	 */
2611 	i40e_service_event_schedule(vsi->back);
2612 	return 0;
2613 }
2614 
2615 /**
2616  * i40e_rm_vlan_all_mac - Remove MAC/VLAN pair for all MAC with the given VLAN
2617  * @vsi: the vsi being configured
2618  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2619  *
2620  * This function should be used to remove all VLAN filters which match the
2621  * given VID. It does not schedule the service event and does not take the
2622  * mac_filter_hash_lock so it may be combined with other operations under
2623  * a single invocation of the mac_filter_hash_lock.
2624  *
2625  * NOTE: this function expects to be called while under the
2626  * mac_filter_hash_lock
2627  */
2628 void i40e_rm_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2629 {
2630 	struct i40e_mac_filter *f;
2631 	struct hlist_node *h;
2632 	int bkt;
2633 
2634 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2635 		if (f->vlan == vid)
2636 			__i40e_del_filter(vsi, f);
2637 	}
2638 }
2639 
2640 /**
2641  * i40e_vsi_kill_vlan - Remove VSI membership for given VLAN
2642  * @vsi: the VSI being configured
2643  * @vid: VLAN id to be removed
2644  **/
2645 void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, u16 vid)
2646 {
2647 	if (!vid || vsi->info.pvid)
2648 		return;
2649 
2650 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2651 	i40e_rm_vlan_all_mac(vsi, vid);
2652 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2653 
2654 	/* schedule our worker thread which will take care of
2655 	 * applying the new filter changes
2656 	 */
2657 	i40e_service_event_schedule(vsi->back);
2658 }
2659 
2660 /**
2661  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2662  * @netdev: network interface to be adjusted
2663  * @vid: vlan id to be added
2664  *
2665  * net_device_ops implementation for adding vlan ids
2666  **/
2667 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2668 				__always_unused __be16 proto, u16 vid)
2669 {
2670 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2671 	struct i40e_vsi *vsi = np->vsi;
2672 	int ret = 0;
2673 
2674 	if (vid >= VLAN_N_VID)
2675 		return -EINVAL;
2676 
2677 	/* If the network stack called us with vid = 0 then
2678 	 * it is asking to receive priority tagged packets with
2679 	 * vlan id 0.  Our HW receives them by default when configured
2680 	 * to receive untagged packets so there is no need to add an
2681 	 * extra filter for vlan 0 tagged packets.
2682 	 */
2683 	if (vid)
2684 		ret = i40e_vsi_add_vlan(vsi, vid);
2685 
2686 	if (!ret)
2687 		set_bit(vid, vsi->active_vlans);
2688 
2689 	return ret;
2690 }
2691 
2692 /**
2693  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2694  * @netdev: network interface to be adjusted
2695  * @vid: vlan id to be removed
2696  *
2697  * net_device_ops implementation for removing vlan ids
2698  **/
2699 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2700 				 __always_unused __be16 proto, u16 vid)
2701 {
2702 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2703 	struct i40e_vsi *vsi = np->vsi;
2704 
2705 	/* return code is ignored as there is nothing a user
2706 	 * can do about failure to remove and a log message was
2707 	 * already printed from the other function
2708 	 */
2709 	i40e_vsi_kill_vlan(vsi, vid);
2710 
2711 	clear_bit(vid, vsi->active_vlans);
2712 
2713 	return 0;
2714 }
2715 
2716 /**
2717  * i40e_macaddr_init - explicitly write the mac address filters
2718  *
2719  * @vsi: pointer to the vsi
2720  * @macaddr: the MAC address
2721  *
2722  * This is needed when the macaddr has been obtained by other
2723  * means than the default, e.g., from Open Firmware or IDPROM.
2724  * Returns 0 on success, negative on failure
2725  **/
2726 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
2727 {
2728 	int ret;
2729 	struct i40e_aqc_add_macvlan_element_data element;
2730 
2731 	ret = i40e_aq_mac_address_write(&vsi->back->hw,
2732 					I40E_AQC_WRITE_TYPE_LAA_WOL,
2733 					macaddr, NULL);
2734 	if (ret) {
2735 		dev_info(&vsi->back->pdev->dev,
2736 			 "Addr change for VSI failed: %d\n", ret);
2737 		return -EADDRNOTAVAIL;
2738 	}
2739 
2740 	memset(&element, 0, sizeof(element));
2741 	ether_addr_copy(element.mac_addr, macaddr);
2742 	element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
2743 	ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
2744 	if (ret) {
2745 		dev_info(&vsi->back->pdev->dev,
2746 			 "add filter failed err %s aq_err %s\n",
2747 			 i40e_stat_str(&vsi->back->hw, ret),
2748 			 i40e_aq_str(&vsi->back->hw,
2749 				     vsi->back->hw.aq.asq_last_status));
2750 	}
2751 	return ret;
2752 }
2753 
2754 /**
2755  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2756  * @vsi: the vsi being brought back up
2757  **/
2758 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2759 {
2760 	u16 vid;
2761 
2762 	if (!vsi->netdev)
2763 		return;
2764 
2765 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2766 
2767 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2768 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2769 				     vid);
2770 }
2771 
2772 /**
2773  * i40e_vsi_add_pvid - Add pvid for the VSI
2774  * @vsi: the vsi being adjusted
2775  * @vid: the vlan id to set as a PVID
2776  **/
2777 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2778 {
2779 	struct i40e_vsi_context ctxt;
2780 	i40e_status ret;
2781 
2782 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2783 	vsi->info.pvid = cpu_to_le16(vid);
2784 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2785 				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2786 				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2787 
2788 	ctxt.seid = vsi->seid;
2789 	ctxt.info = vsi->info;
2790 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2791 	if (ret) {
2792 		dev_info(&vsi->back->pdev->dev,
2793 			 "add pvid failed, err %s aq_err %s\n",
2794 			 i40e_stat_str(&vsi->back->hw, ret),
2795 			 i40e_aq_str(&vsi->back->hw,
2796 				     vsi->back->hw.aq.asq_last_status));
2797 		return -ENOENT;
2798 	}
2799 
2800 	return 0;
2801 }
2802 
2803 /**
2804  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2805  * @vsi: the vsi being adjusted
2806  *
2807  * Just use the vlan_rx_register() service to put it back to normal
2808  **/
2809 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2810 {
2811 	i40e_vlan_stripping_disable(vsi);
2812 
2813 	vsi->info.pvid = 0;
2814 }
2815 
2816 /**
2817  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2818  * @vsi: ptr to the VSI
2819  *
2820  * If this function returns with an error, then it's possible one or
2821  * more of the rings is populated (while the rest are not).  It is the
2822  * callers duty to clean those orphaned rings.
2823  *
2824  * Return 0 on success, negative on failure
2825  **/
2826 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2827 {
2828 	int i, err = 0;
2829 
2830 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2831 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2832 
2833 	if (!i40e_enabled_xdp_vsi(vsi))
2834 		return err;
2835 
2836 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2837 		err = i40e_setup_tx_descriptors(vsi->xdp_rings[i]);
2838 
2839 	return err;
2840 }
2841 
2842 /**
2843  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2844  * @vsi: ptr to the VSI
2845  *
2846  * Free VSI's transmit software resources
2847  **/
2848 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2849 {
2850 	int i;
2851 
2852 	if (vsi->tx_rings) {
2853 		for (i = 0; i < vsi->num_queue_pairs; i++)
2854 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2855 				i40e_free_tx_resources(vsi->tx_rings[i]);
2856 	}
2857 
2858 	if (vsi->xdp_rings) {
2859 		for (i = 0; i < vsi->num_queue_pairs; i++)
2860 			if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2861 				i40e_free_tx_resources(vsi->xdp_rings[i]);
2862 	}
2863 }
2864 
2865 /**
2866  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2867  * @vsi: ptr to the VSI
2868  *
2869  * If this function returns with an error, then it's possible one or
2870  * more of the rings is populated (while the rest are not).  It is the
2871  * callers duty to clean those orphaned rings.
2872  *
2873  * Return 0 on success, negative on failure
2874  **/
2875 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2876 {
2877 	int i, err = 0;
2878 
2879 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2880 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2881 	return err;
2882 }
2883 
2884 /**
2885  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2886  * @vsi: ptr to the VSI
2887  *
2888  * Free all receive software resources
2889  **/
2890 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2891 {
2892 	int i;
2893 
2894 	if (!vsi->rx_rings)
2895 		return;
2896 
2897 	for (i = 0; i < vsi->num_queue_pairs; i++)
2898 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2899 			i40e_free_rx_resources(vsi->rx_rings[i]);
2900 }
2901 
2902 /**
2903  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2904  * @ring: The Tx ring to configure
2905  *
2906  * This enables/disables XPS for a given Tx descriptor ring
2907  * based on the TCs enabled for the VSI that ring belongs to.
2908  **/
2909 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2910 {
2911 	struct i40e_vsi *vsi = ring->vsi;
2912 	cpumask_var_t mask;
2913 
2914 	if (!ring->q_vector || !ring->netdev)
2915 		return;
2916 
2917 	/* Single TC mode enable XPS */
2918 	if (vsi->tc_config.numtc <= 1) {
2919 		if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2920 			netif_set_xps_queue(ring->netdev,
2921 					    &ring->q_vector->affinity_mask,
2922 					    ring->queue_index);
2923 	} else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2924 		/* Disable XPS to allow selection based on TC */
2925 		bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2926 		netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2927 		free_cpumask_var(mask);
2928 	}
2929 
2930 	/* schedule our worker thread which will take care of
2931 	 * applying the new filter changes
2932 	 */
2933 	i40e_service_event_schedule(vsi->back);
2934 }
2935 
2936 /**
2937  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2938  * @ring: The Tx ring to configure
2939  *
2940  * Configure the Tx descriptor ring in the HMC context.
2941  **/
2942 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2943 {
2944 	struct i40e_vsi *vsi = ring->vsi;
2945 	u16 pf_q = vsi->base_queue + ring->queue_index;
2946 	struct i40e_hw *hw = &vsi->back->hw;
2947 	struct i40e_hmc_obj_txq tx_ctx;
2948 	i40e_status err = 0;
2949 	u32 qtx_ctl = 0;
2950 
2951 	/* some ATR related tx ring init */
2952 	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2953 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2954 		ring->atr_count = 0;
2955 	} else {
2956 		ring->atr_sample_rate = 0;
2957 	}
2958 
2959 	/* configure XPS */
2960 	i40e_config_xps_tx_ring(ring);
2961 
2962 	/* clear the context structure first */
2963 	memset(&tx_ctx, 0, sizeof(tx_ctx));
2964 
2965 	tx_ctx.new_context = 1;
2966 	tx_ctx.base = (ring->dma / 128);
2967 	tx_ctx.qlen = ring->count;
2968 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2969 					       I40E_FLAG_FD_ATR_ENABLED));
2970 	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2971 	/* FDIR VSI tx ring can still use RS bit and writebacks */
2972 	if (vsi->type != I40E_VSI_FDIR)
2973 		tx_ctx.head_wb_ena = 1;
2974 	tx_ctx.head_wb_addr = ring->dma +
2975 			      (ring->count * sizeof(struct i40e_tx_desc));
2976 
2977 	/* As part of VSI creation/update, FW allocates certain
2978 	 * Tx arbitration queue sets for each TC enabled for
2979 	 * the VSI. The FW returns the handles to these queue
2980 	 * sets as part of the response buffer to Add VSI,
2981 	 * Update VSI, etc. AQ commands. It is expected that
2982 	 * these queue set handles be associated with the Tx
2983 	 * queues by the driver as part of the TX queue context
2984 	 * initialization. This has to be done regardless of
2985 	 * DCB as by default everything is mapped to TC0.
2986 	 */
2987 	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2988 	tx_ctx.rdylist_act = 0;
2989 
2990 	/* clear the context in the HMC */
2991 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2992 	if (err) {
2993 		dev_info(&vsi->back->pdev->dev,
2994 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2995 			 ring->queue_index, pf_q, err);
2996 		return -ENOMEM;
2997 	}
2998 
2999 	/* set the context in the HMC */
3000 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
3001 	if (err) {
3002 		dev_info(&vsi->back->pdev->dev,
3003 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
3004 			 ring->queue_index, pf_q, err);
3005 		return -ENOMEM;
3006 	}
3007 
3008 	/* Now associate this queue with this PCI function */
3009 	if (vsi->type == I40E_VSI_VMDQ2) {
3010 		qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
3011 		qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
3012 			   I40E_QTX_CTL_VFVM_INDX_MASK;
3013 	} else {
3014 		qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
3015 	}
3016 
3017 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
3018 		    I40E_QTX_CTL_PF_INDX_MASK);
3019 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
3020 	i40e_flush(hw);
3021 
3022 	/* cache tail off for easier writes later */
3023 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
3024 
3025 	return 0;
3026 }
3027 
3028 /**
3029  * i40e_configure_rx_ring - Configure a receive ring context
3030  * @ring: The Rx ring to configure
3031  *
3032  * Configure the Rx descriptor ring in the HMC context.
3033  **/
3034 static int i40e_configure_rx_ring(struct i40e_ring *ring)
3035 {
3036 	struct i40e_vsi *vsi = ring->vsi;
3037 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
3038 	u16 pf_q = vsi->base_queue + ring->queue_index;
3039 	struct i40e_hw *hw = &vsi->back->hw;
3040 	struct i40e_hmc_obj_rxq rx_ctx;
3041 	i40e_status err = 0;
3042 
3043 	ring->state = 0;
3044 
3045 	/* clear the context structure first */
3046 	memset(&rx_ctx, 0, sizeof(rx_ctx));
3047 
3048 	ring->rx_buf_len = vsi->rx_buf_len;
3049 
3050 	rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3051 				    BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3052 
3053 	rx_ctx.base = (ring->dma / 128);
3054 	rx_ctx.qlen = ring->count;
3055 
3056 	/* use 32 byte descriptors */
3057 	rx_ctx.dsize = 1;
3058 
3059 	/* descriptor type is always zero
3060 	 * rx_ctx.dtype = 0;
3061 	 */
3062 	rx_ctx.hsplit_0 = 0;
3063 
3064 	rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3065 	if (hw->revision_id == 0)
3066 		rx_ctx.lrxqthresh = 0;
3067 	else
3068 		rx_ctx.lrxqthresh = 2;
3069 	rx_ctx.crcstrip = 1;
3070 	rx_ctx.l2tsel = 1;
3071 	/* this controls whether VLAN is stripped from inner headers */
3072 	rx_ctx.showiv = 0;
3073 	/* set the prefena field to 1 because the manual says to */
3074 	rx_ctx.prefena = 1;
3075 
3076 	/* clear the context in the HMC */
3077 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3078 	if (err) {
3079 		dev_info(&vsi->back->pdev->dev,
3080 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3081 			 ring->queue_index, pf_q, err);
3082 		return -ENOMEM;
3083 	}
3084 
3085 	/* set the context in the HMC */
3086 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3087 	if (err) {
3088 		dev_info(&vsi->back->pdev->dev,
3089 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3090 			 ring->queue_index, pf_q, err);
3091 		return -ENOMEM;
3092 	}
3093 
3094 	/* configure Rx buffer alignment */
3095 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3096 		clear_ring_build_skb_enabled(ring);
3097 	else
3098 		set_ring_build_skb_enabled(ring);
3099 
3100 	/* cache tail for quicker writes, and clear the reg before use */
3101 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3102 	writel(0, ring->tail);
3103 
3104 	i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3105 
3106 	return 0;
3107 }
3108 
3109 /**
3110  * i40e_vsi_configure_tx - Configure the VSI for Tx
3111  * @vsi: VSI structure describing this set of rings and resources
3112  *
3113  * Configure the Tx VSI for operation.
3114  **/
3115 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3116 {
3117 	int err = 0;
3118 	u16 i;
3119 
3120 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3121 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3122 
3123 	if (!i40e_enabled_xdp_vsi(vsi))
3124 		return err;
3125 
3126 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3127 		err = i40e_configure_tx_ring(vsi->xdp_rings[i]);
3128 
3129 	return err;
3130 }
3131 
3132 /**
3133  * i40e_vsi_configure_rx - Configure the VSI for Rx
3134  * @vsi: the VSI being configured
3135  *
3136  * Configure the Rx VSI for operation.
3137  **/
3138 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3139 {
3140 	int err = 0;
3141 	u16 i;
3142 
3143 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3144 		vsi->max_frame = I40E_MAX_RXBUFFER;
3145 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
3146 #if (PAGE_SIZE < 8192)
3147 	} else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3148 		   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3149 		vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3150 		vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3151 #endif
3152 	} else {
3153 		vsi->max_frame = I40E_MAX_RXBUFFER;
3154 		vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3155 						       I40E_RXBUFFER_2048;
3156 	}
3157 
3158 	/* set up individual rings */
3159 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3160 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3161 
3162 	return err;
3163 }
3164 
3165 /**
3166  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3167  * @vsi: ptr to the VSI
3168  **/
3169 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3170 {
3171 	struct i40e_ring *tx_ring, *rx_ring;
3172 	u16 qoffset, qcount;
3173 	int i, n;
3174 
3175 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3176 		/* Reset the TC information */
3177 		for (i = 0; i < vsi->num_queue_pairs; i++) {
3178 			rx_ring = vsi->rx_rings[i];
3179 			tx_ring = vsi->tx_rings[i];
3180 			rx_ring->dcb_tc = 0;
3181 			tx_ring->dcb_tc = 0;
3182 		}
3183 	}
3184 
3185 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3186 		if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3187 			continue;
3188 
3189 		qoffset = vsi->tc_config.tc_info[n].qoffset;
3190 		qcount = vsi->tc_config.tc_info[n].qcount;
3191 		for (i = qoffset; i < (qoffset + qcount); i++) {
3192 			rx_ring = vsi->rx_rings[i];
3193 			tx_ring = vsi->tx_rings[i];
3194 			rx_ring->dcb_tc = n;
3195 			tx_ring->dcb_tc = n;
3196 		}
3197 	}
3198 }
3199 
3200 /**
3201  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3202  * @vsi: ptr to the VSI
3203  **/
3204 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3205 {
3206 	struct i40e_pf *pf = vsi->back;
3207 	int err;
3208 
3209 	if (vsi->netdev)
3210 		i40e_set_rx_mode(vsi->netdev);
3211 
3212 	if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
3213 		err = i40e_macaddr_init(vsi, pf->hw.mac.addr);
3214 		if (err) {
3215 			dev_warn(&pf->pdev->dev,
3216 				 "could not set up macaddr; err %d\n", err);
3217 		}
3218 	}
3219 }
3220 
3221 /**
3222  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3223  * @vsi: Pointer to the targeted VSI
3224  *
3225  * This function replays the hlist on the hw where all the SB Flow Director
3226  * filters were saved.
3227  **/
3228 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3229 {
3230 	struct i40e_fdir_filter *filter;
3231 	struct i40e_pf *pf = vsi->back;
3232 	struct hlist_node *node;
3233 
3234 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3235 		return;
3236 
3237 	/* Reset FDir counters as we're replaying all existing filters */
3238 	pf->fd_tcp4_filter_cnt = 0;
3239 	pf->fd_udp4_filter_cnt = 0;
3240 	pf->fd_sctp4_filter_cnt = 0;
3241 	pf->fd_ip4_filter_cnt = 0;
3242 
3243 	hlist_for_each_entry_safe(filter, node,
3244 				  &pf->fdir_filter_list, fdir_node) {
3245 		i40e_add_del_fdir(vsi, filter, true);
3246 	}
3247 }
3248 
3249 /**
3250  * i40e_vsi_configure - Set up the VSI for action
3251  * @vsi: the VSI being configured
3252  **/
3253 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3254 {
3255 	int err;
3256 
3257 	i40e_set_vsi_rx_mode(vsi);
3258 	i40e_restore_vlan(vsi);
3259 	i40e_vsi_config_dcb_rings(vsi);
3260 	err = i40e_vsi_configure_tx(vsi);
3261 	if (!err)
3262 		err = i40e_vsi_configure_rx(vsi);
3263 
3264 	return err;
3265 }
3266 
3267 /**
3268  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3269  * @vsi: the VSI being configured
3270  **/
3271 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3272 {
3273 	bool has_xdp = i40e_enabled_xdp_vsi(vsi);
3274 	struct i40e_pf *pf = vsi->back;
3275 	struct i40e_hw *hw = &pf->hw;
3276 	u16 vector;
3277 	int i, q;
3278 	u32 qp;
3279 
3280 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
3281 	 * and PFINT_LNKLSTn registers, e.g.:
3282 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3283 	 */
3284 	qp = vsi->base_queue;
3285 	vector = vsi->base_vector;
3286 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3287 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3288 
3289 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
3290 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3291 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
3292 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3293 		     q_vector->rx.itr);
3294 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3295 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
3296 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3297 		     q_vector->tx.itr);
3298 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
3299 		     i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3300 
3301 		/* Linked list for the queuepairs assigned to this vector */
3302 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3303 		for (q = 0; q < q_vector->num_ringpairs; q++) {
3304 			u32 nextqp = has_xdp ? qp + vsi->alloc_queue_pairs : qp;
3305 			u32 val;
3306 
3307 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3308 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3309 			      (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3310 			      (nextqp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT) |
3311 			      (I40E_QUEUE_TYPE_TX <<
3312 			       I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3313 
3314 			wr32(hw, I40E_QINT_RQCTL(qp), val);
3315 
3316 			if (has_xdp) {
3317 				val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3318 				      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3319 				      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3320 				      (qp << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3321 				      (I40E_QUEUE_TYPE_TX <<
3322 				       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3323 
3324 				wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3325 			}
3326 
3327 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3328 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3329 			      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3330 			      ((qp + 1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3331 			      (I40E_QUEUE_TYPE_RX <<
3332 			       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3333 
3334 			/* Terminate the linked list */
3335 			if (q == (q_vector->num_ringpairs - 1))
3336 				val |= (I40E_QUEUE_END_OF_LIST <<
3337 					I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3338 
3339 			wr32(hw, I40E_QINT_TQCTL(qp), val);
3340 			qp++;
3341 		}
3342 	}
3343 
3344 	i40e_flush(hw);
3345 }
3346 
3347 /**
3348  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3349  * @hw: ptr to the hardware info
3350  **/
3351 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3352 {
3353 	struct i40e_hw *hw = &pf->hw;
3354 	u32 val;
3355 
3356 	/* clear things first */
3357 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3358 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3359 
3360 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3361 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3362 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
3363 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3364 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3365 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3366 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3367 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3368 
3369 	if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3370 		val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3371 
3372 	if (pf->flags & I40E_FLAG_PTP)
3373 		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3374 
3375 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
3376 
3377 	/* SW_ITR_IDX = 0, but don't change INTENA */
3378 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3379 					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3380 
3381 	/* OTHER_ITR_IDX = 0 */
3382 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3383 }
3384 
3385 /**
3386  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3387  * @vsi: the VSI being configured
3388  **/
3389 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3390 {
3391 	u32 nextqp = i40e_enabled_xdp_vsi(vsi) ? vsi->alloc_queue_pairs : 0;
3392 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3393 	struct i40e_pf *pf = vsi->back;
3394 	struct i40e_hw *hw = &pf->hw;
3395 	u32 val;
3396 
3397 	/* set the ITR configuration */
3398 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
3399 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3400 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
3401 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3402 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3403 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
3404 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3405 
3406 	i40e_enable_misc_int_causes(pf);
3407 
3408 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3409 	wr32(hw, I40E_PFINT_LNKLST0, 0);
3410 
3411 	/* Associate the queue pair to the vector and enable the queue int */
3412 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		       |
3413 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3414 	      (nextqp	   << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3415 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3416 
3417 	wr32(hw, I40E_QINT_RQCTL(0), val);
3418 
3419 	if (i40e_enabled_xdp_vsi(vsi)) {
3420 		val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		     |
3421 		      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)|
3422 		      (I40E_QUEUE_TYPE_TX
3423 		       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3424 
3425 	       wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3426 	}
3427 
3428 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
3429 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3430 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3431 
3432 	wr32(hw, I40E_QINT_TQCTL(0), val);
3433 	i40e_flush(hw);
3434 }
3435 
3436 /**
3437  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3438  * @pf: board private structure
3439  **/
3440 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3441 {
3442 	struct i40e_hw *hw = &pf->hw;
3443 
3444 	wr32(hw, I40E_PFINT_DYN_CTL0,
3445 	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3446 	i40e_flush(hw);
3447 }
3448 
3449 /**
3450  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3451  * @pf: board private structure
3452  * @clearpba: true when all pending interrupt events should be cleared
3453  **/
3454 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3455 {
3456 	struct i40e_hw *hw = &pf->hw;
3457 	u32 val;
3458 
3459 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3460 	      (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3461 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3462 
3463 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
3464 	i40e_flush(hw);
3465 }
3466 
3467 /**
3468  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3469  * @irq: interrupt number
3470  * @data: pointer to a q_vector
3471  **/
3472 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3473 {
3474 	struct i40e_q_vector *q_vector = data;
3475 
3476 	if (!q_vector->tx.ring && !q_vector->rx.ring)
3477 		return IRQ_HANDLED;
3478 
3479 	napi_schedule_irqoff(&q_vector->napi);
3480 
3481 	return IRQ_HANDLED;
3482 }
3483 
3484 /**
3485  * i40e_irq_affinity_notify - Callback for affinity changes
3486  * @notify: context as to what irq was changed
3487  * @mask: the new affinity mask
3488  *
3489  * This is a callback function used by the irq_set_affinity_notifier function
3490  * so that we may register to receive changes to the irq affinity masks.
3491  **/
3492 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3493 				     const cpumask_t *mask)
3494 {
3495 	struct i40e_q_vector *q_vector =
3496 		container_of(notify, struct i40e_q_vector, affinity_notify);
3497 
3498 	q_vector->affinity_mask = *mask;
3499 }
3500 
3501 /**
3502  * i40e_irq_affinity_release - Callback for affinity notifier release
3503  * @ref: internal core kernel usage
3504  *
3505  * This is a callback function used by the irq_set_affinity_notifier function
3506  * to inform the current notification subscriber that they will no longer
3507  * receive notifications.
3508  **/
3509 static void i40e_irq_affinity_release(struct kref *ref) {}
3510 
3511 /**
3512  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3513  * @vsi: the VSI being configured
3514  * @basename: name for the vector
3515  *
3516  * Allocates MSI-X vectors and requests interrupts from the kernel.
3517  **/
3518 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3519 {
3520 	int q_vectors = vsi->num_q_vectors;
3521 	struct i40e_pf *pf = vsi->back;
3522 	int base = vsi->base_vector;
3523 	int rx_int_idx = 0;
3524 	int tx_int_idx = 0;
3525 	int vector, err;
3526 	int irq_num;
3527 
3528 	for (vector = 0; vector < q_vectors; vector++) {
3529 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3530 
3531 		irq_num = pf->msix_entries[base + vector].vector;
3532 
3533 		if (q_vector->tx.ring && q_vector->rx.ring) {
3534 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3535 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3536 			tx_int_idx++;
3537 		} else if (q_vector->rx.ring) {
3538 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3539 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3540 		} else if (q_vector->tx.ring) {
3541 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3542 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3543 		} else {
3544 			/* skip this unused q_vector */
3545 			continue;
3546 		}
3547 		err = request_irq(irq_num,
3548 				  vsi->irq_handler,
3549 				  0,
3550 				  q_vector->name,
3551 				  q_vector);
3552 		if (err) {
3553 			dev_info(&pf->pdev->dev,
3554 				 "MSIX request_irq failed, error: %d\n", err);
3555 			goto free_queue_irqs;
3556 		}
3557 
3558 		/* register for affinity change notifications */
3559 		q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3560 		q_vector->affinity_notify.release = i40e_irq_affinity_release;
3561 		irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3562 		/* assign the mask for this irq */
3563 		irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
3564 	}
3565 
3566 	vsi->irqs_ready = true;
3567 	return 0;
3568 
3569 free_queue_irqs:
3570 	while (vector) {
3571 		vector--;
3572 		irq_num = pf->msix_entries[base + vector].vector;
3573 		irq_set_affinity_notifier(irq_num, NULL);
3574 		irq_set_affinity_hint(irq_num, NULL);
3575 		free_irq(irq_num, &vsi->q_vectors[vector]);
3576 	}
3577 	return err;
3578 }
3579 
3580 /**
3581  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3582  * @vsi: the VSI being un-configured
3583  **/
3584 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3585 {
3586 	struct i40e_pf *pf = vsi->back;
3587 	struct i40e_hw *hw = &pf->hw;
3588 	int base = vsi->base_vector;
3589 	int i;
3590 
3591 	/* disable interrupt causation from each queue */
3592 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3593 		u32 val;
3594 
3595 		val = rd32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx));
3596 		val &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3597 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), val);
3598 
3599 		val = rd32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx));
3600 		val &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3601 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), val);
3602 
3603 		if (!i40e_enabled_xdp_vsi(vsi))
3604 			continue;
3605 		wr32(hw, I40E_QINT_TQCTL(vsi->xdp_rings[i]->reg_idx), 0);
3606 	}
3607 
3608 	/* disable each interrupt */
3609 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3610 		for (i = vsi->base_vector;
3611 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3612 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3613 
3614 		i40e_flush(hw);
3615 		for (i = 0; i < vsi->num_q_vectors; i++)
3616 			synchronize_irq(pf->msix_entries[i + base].vector);
3617 	} else {
3618 		/* Legacy and MSI mode - this stops all interrupt handling */
3619 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3620 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3621 		i40e_flush(hw);
3622 		synchronize_irq(pf->pdev->irq);
3623 	}
3624 }
3625 
3626 /**
3627  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3628  * @vsi: the VSI being configured
3629  **/
3630 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3631 {
3632 	struct i40e_pf *pf = vsi->back;
3633 	int i;
3634 
3635 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3636 		for (i = 0; i < vsi->num_q_vectors; i++)
3637 			i40e_irq_dynamic_enable(vsi, i);
3638 	} else {
3639 		i40e_irq_dynamic_enable_icr0(pf, true);
3640 	}
3641 
3642 	i40e_flush(&pf->hw);
3643 	return 0;
3644 }
3645 
3646 /**
3647  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3648  * @pf: board private structure
3649  **/
3650 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3651 {
3652 	/* Disable ICR 0 */
3653 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3654 	i40e_flush(&pf->hw);
3655 }
3656 
3657 /**
3658  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3659  * @irq: interrupt number
3660  * @data: pointer to a q_vector
3661  *
3662  * This is the handler used for all MSI/Legacy interrupts, and deals
3663  * with both queue and non-queue interrupts.  This is also used in
3664  * MSIX mode to handle the non-queue interrupts.
3665  **/
3666 static irqreturn_t i40e_intr(int irq, void *data)
3667 {
3668 	struct i40e_pf *pf = (struct i40e_pf *)data;
3669 	struct i40e_hw *hw = &pf->hw;
3670 	irqreturn_t ret = IRQ_NONE;
3671 	u32 icr0, icr0_remaining;
3672 	u32 val, ena_mask;
3673 
3674 	icr0 = rd32(hw, I40E_PFINT_ICR0);
3675 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3676 
3677 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3678 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3679 		goto enable_intr;
3680 
3681 	/* if interrupt but no bits showing, must be SWINT */
3682 	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3683 	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3684 		pf->sw_int_count++;
3685 
3686 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3687 	    (icr0 & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3688 		ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3689 		dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3690 		set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
3691 	}
3692 
3693 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3694 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3695 		struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3696 		struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3697 
3698 		/* We do not have a way to disarm Queue causes while leaving
3699 		 * interrupt enabled for all other causes, ideally
3700 		 * interrupt should be disabled while we are in NAPI but
3701 		 * this is not a performance path and napi_schedule()
3702 		 * can deal with rescheduling.
3703 		 */
3704 		if (!test_bit(__I40E_DOWN, pf->state))
3705 			napi_schedule_irqoff(&q_vector->napi);
3706 	}
3707 
3708 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3709 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3710 		set_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
3711 		i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3712 	}
3713 
3714 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3715 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3716 		set_bit(__I40E_MDD_EVENT_PENDING, pf->state);
3717 	}
3718 
3719 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3720 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3721 		set_bit(__I40E_VFLR_EVENT_PENDING, pf->state);
3722 	}
3723 
3724 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3725 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
3726 			set_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
3727 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3728 		val = rd32(hw, I40E_GLGEN_RSTAT);
3729 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3730 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3731 		if (val == I40E_RESET_CORER) {
3732 			pf->corer_count++;
3733 		} else if (val == I40E_RESET_GLOBR) {
3734 			pf->globr_count++;
3735 		} else if (val == I40E_RESET_EMPR) {
3736 			pf->empr_count++;
3737 			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state);
3738 		}
3739 	}
3740 
3741 	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3742 		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3743 		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3744 		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3745 			 rd32(hw, I40E_PFHMC_ERRORINFO),
3746 			 rd32(hw, I40E_PFHMC_ERRORDATA));
3747 	}
3748 
3749 	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3750 		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3751 
3752 		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3753 			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3754 			i40e_ptp_tx_hwtstamp(pf);
3755 		}
3756 	}
3757 
3758 	/* If a critical error is pending we have no choice but to reset the
3759 	 * device.
3760 	 * Report and mask out any remaining unexpected interrupts.
3761 	 */
3762 	icr0_remaining = icr0 & ena_mask;
3763 	if (icr0_remaining) {
3764 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3765 			 icr0_remaining);
3766 		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3767 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3768 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3769 			dev_info(&pf->pdev->dev, "device will be reset\n");
3770 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
3771 			i40e_service_event_schedule(pf);
3772 		}
3773 		ena_mask &= ~icr0_remaining;
3774 	}
3775 	ret = IRQ_HANDLED;
3776 
3777 enable_intr:
3778 	/* re-enable interrupt causes */
3779 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3780 	if (!test_bit(__I40E_DOWN, pf->state)) {
3781 		i40e_service_event_schedule(pf);
3782 		i40e_irq_dynamic_enable_icr0(pf, false);
3783 	}
3784 
3785 	return ret;
3786 }
3787 
3788 /**
3789  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3790  * @tx_ring:  tx ring to clean
3791  * @budget:   how many cleans we're allowed
3792  *
3793  * Returns true if there's any budget left (e.g. the clean is finished)
3794  **/
3795 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3796 {
3797 	struct i40e_vsi *vsi = tx_ring->vsi;
3798 	u16 i = tx_ring->next_to_clean;
3799 	struct i40e_tx_buffer *tx_buf;
3800 	struct i40e_tx_desc *tx_desc;
3801 
3802 	tx_buf = &tx_ring->tx_bi[i];
3803 	tx_desc = I40E_TX_DESC(tx_ring, i);
3804 	i -= tx_ring->count;
3805 
3806 	do {
3807 		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3808 
3809 		/* if next_to_watch is not set then there is no work pending */
3810 		if (!eop_desc)
3811 			break;
3812 
3813 		/* prevent any other reads prior to eop_desc */
3814 		read_barrier_depends();
3815 
3816 		/* if the descriptor isn't done, no work yet to do */
3817 		if (!(eop_desc->cmd_type_offset_bsz &
3818 		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3819 			break;
3820 
3821 		/* clear next_to_watch to prevent false hangs */
3822 		tx_buf->next_to_watch = NULL;
3823 
3824 		tx_desc->buffer_addr = 0;
3825 		tx_desc->cmd_type_offset_bsz = 0;
3826 		/* move past filter desc */
3827 		tx_buf++;
3828 		tx_desc++;
3829 		i++;
3830 		if (unlikely(!i)) {
3831 			i -= tx_ring->count;
3832 			tx_buf = tx_ring->tx_bi;
3833 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3834 		}
3835 		/* unmap skb header data */
3836 		dma_unmap_single(tx_ring->dev,
3837 				 dma_unmap_addr(tx_buf, dma),
3838 				 dma_unmap_len(tx_buf, len),
3839 				 DMA_TO_DEVICE);
3840 		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3841 			kfree(tx_buf->raw_buf);
3842 
3843 		tx_buf->raw_buf = NULL;
3844 		tx_buf->tx_flags = 0;
3845 		tx_buf->next_to_watch = NULL;
3846 		dma_unmap_len_set(tx_buf, len, 0);
3847 		tx_desc->buffer_addr = 0;
3848 		tx_desc->cmd_type_offset_bsz = 0;
3849 
3850 		/* move us past the eop_desc for start of next FD desc */
3851 		tx_buf++;
3852 		tx_desc++;
3853 		i++;
3854 		if (unlikely(!i)) {
3855 			i -= tx_ring->count;
3856 			tx_buf = tx_ring->tx_bi;
3857 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3858 		}
3859 
3860 		/* update budget accounting */
3861 		budget--;
3862 	} while (likely(budget));
3863 
3864 	i += tx_ring->count;
3865 	tx_ring->next_to_clean = i;
3866 
3867 	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3868 		i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3869 
3870 	return budget > 0;
3871 }
3872 
3873 /**
3874  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3875  * @irq: interrupt number
3876  * @data: pointer to a q_vector
3877  **/
3878 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3879 {
3880 	struct i40e_q_vector *q_vector = data;
3881 	struct i40e_vsi *vsi;
3882 
3883 	if (!q_vector->tx.ring)
3884 		return IRQ_HANDLED;
3885 
3886 	vsi = q_vector->tx.ring->vsi;
3887 	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3888 
3889 	return IRQ_HANDLED;
3890 }
3891 
3892 /**
3893  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3894  * @vsi: the VSI being configured
3895  * @v_idx: vector index
3896  * @qp_idx: queue pair index
3897  **/
3898 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3899 {
3900 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3901 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3902 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3903 
3904 	tx_ring->q_vector = q_vector;
3905 	tx_ring->next = q_vector->tx.ring;
3906 	q_vector->tx.ring = tx_ring;
3907 	q_vector->tx.count++;
3908 
3909 	/* Place XDP Tx ring in the same q_vector ring list as regular Tx */
3910 	if (i40e_enabled_xdp_vsi(vsi)) {
3911 		struct i40e_ring *xdp_ring = vsi->xdp_rings[qp_idx];
3912 
3913 		xdp_ring->q_vector = q_vector;
3914 		xdp_ring->next = q_vector->tx.ring;
3915 		q_vector->tx.ring = xdp_ring;
3916 		q_vector->tx.count++;
3917 	}
3918 
3919 	rx_ring->q_vector = q_vector;
3920 	rx_ring->next = q_vector->rx.ring;
3921 	q_vector->rx.ring = rx_ring;
3922 	q_vector->rx.count++;
3923 }
3924 
3925 /**
3926  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3927  * @vsi: the VSI being configured
3928  *
3929  * This function maps descriptor rings to the queue-specific vectors
3930  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3931  * one vector per queue pair, but on a constrained vector budget, we
3932  * group the queue pairs as "efficiently" as possible.
3933  **/
3934 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3935 {
3936 	int qp_remaining = vsi->num_queue_pairs;
3937 	int q_vectors = vsi->num_q_vectors;
3938 	int num_ringpairs;
3939 	int v_start = 0;
3940 	int qp_idx = 0;
3941 
3942 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3943 	 * group them so there are multiple queues per vector.
3944 	 * It is also important to go through all the vectors available to be
3945 	 * sure that if we don't use all the vectors, that the remaining vectors
3946 	 * are cleared. This is especially important when decreasing the
3947 	 * number of queues in use.
3948 	 */
3949 	for (; v_start < q_vectors; v_start++) {
3950 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3951 
3952 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3953 
3954 		q_vector->num_ringpairs = num_ringpairs;
3955 
3956 		q_vector->rx.count = 0;
3957 		q_vector->tx.count = 0;
3958 		q_vector->rx.ring = NULL;
3959 		q_vector->tx.ring = NULL;
3960 
3961 		while (num_ringpairs--) {
3962 			i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3963 			qp_idx++;
3964 			qp_remaining--;
3965 		}
3966 	}
3967 }
3968 
3969 /**
3970  * i40e_vsi_request_irq - Request IRQ from the OS
3971  * @vsi: the VSI being configured
3972  * @basename: name for the vector
3973  **/
3974 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3975 {
3976 	struct i40e_pf *pf = vsi->back;
3977 	int err;
3978 
3979 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3980 		err = i40e_vsi_request_irq_msix(vsi, basename);
3981 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3982 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
3983 				  pf->int_name, pf);
3984 	else
3985 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3986 				  pf->int_name, pf);
3987 
3988 	if (err)
3989 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3990 
3991 	return err;
3992 }
3993 
3994 #ifdef CONFIG_NET_POLL_CONTROLLER
3995 /**
3996  * i40e_netpoll - A Polling 'interrupt' handler
3997  * @netdev: network interface device structure
3998  *
3999  * This is used by netconsole to send skbs without having to re-enable
4000  * interrupts.  It's not called while the normal interrupt routine is executing.
4001  **/
4002 static void i40e_netpoll(struct net_device *netdev)
4003 {
4004 	struct i40e_netdev_priv *np = netdev_priv(netdev);
4005 	struct i40e_vsi *vsi = np->vsi;
4006 	struct i40e_pf *pf = vsi->back;
4007 	int i;
4008 
4009 	/* if interface is down do nothing */
4010 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
4011 		return;
4012 
4013 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4014 		for (i = 0; i < vsi->num_q_vectors; i++)
4015 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
4016 	} else {
4017 		i40e_intr(pf->pdev->irq, netdev);
4018 	}
4019 }
4020 #endif
4021 
4022 #define I40E_QTX_ENA_WAIT_COUNT 50
4023 
4024 /**
4025  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
4026  * @pf: the PF being configured
4027  * @pf_q: the PF queue
4028  * @enable: enable or disable state of the queue
4029  *
4030  * This routine will wait for the given Tx queue of the PF to reach the
4031  * enabled or disabled state.
4032  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4033  * multiple retries; else will return 0 in case of success.
4034  **/
4035 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4036 {
4037 	int i;
4038 	u32 tx_reg;
4039 
4040 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4041 		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
4042 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4043 			break;
4044 
4045 		usleep_range(10, 20);
4046 	}
4047 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4048 		return -ETIMEDOUT;
4049 
4050 	return 0;
4051 }
4052 
4053 /**
4054  * i40e_control_tx_q - Start or stop a particular Tx queue
4055  * @pf: the PF structure
4056  * @pf_q: the PF queue to configure
4057  * @enable: start or stop the queue
4058  *
4059  * This function enables or disables a single queue. Note that any delay
4060  * required after the operation is expected to be handled by the caller of
4061  * this function.
4062  **/
4063 static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
4064 {
4065 	struct i40e_hw *hw = &pf->hw;
4066 	u32 tx_reg;
4067 	int i;
4068 
4069 	/* warn the TX unit of coming changes */
4070 	i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
4071 	if (!enable)
4072 		usleep_range(10, 20);
4073 
4074 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4075 		tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
4076 		if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
4077 		    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
4078 			break;
4079 		usleep_range(1000, 2000);
4080 	}
4081 
4082 	/* Skip if the queue is already in the requested state */
4083 	if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4084 		return;
4085 
4086 	/* turn on/off the queue */
4087 	if (enable) {
4088 		wr32(hw, I40E_QTX_HEAD(pf_q), 0);
4089 		tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
4090 	} else {
4091 		tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
4092 	}
4093 
4094 	wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
4095 }
4096 
4097 /**
4098  * i40e_control_wait_tx_q - Start/stop Tx queue and wait for completion
4099  * @seid: VSI SEID
4100  * @pf: the PF structure
4101  * @pf_q: the PF queue to configure
4102  * @is_xdp: true if the queue is used for XDP
4103  * @enable: start or stop the queue
4104  **/
4105 static int i40e_control_wait_tx_q(int seid, struct i40e_pf *pf, int pf_q,
4106 				  bool is_xdp, bool enable)
4107 {
4108 	int ret;
4109 
4110 	i40e_control_tx_q(pf, pf_q, enable);
4111 
4112 	/* wait for the change to finish */
4113 	ret = i40e_pf_txq_wait(pf, pf_q, enable);
4114 	if (ret) {
4115 		dev_info(&pf->pdev->dev,
4116 			 "VSI seid %d %sTx ring %d %sable timeout\n",
4117 			 seid, (is_xdp ? "XDP " : ""), pf_q,
4118 			 (enable ? "en" : "dis"));
4119 	}
4120 
4121 	return ret;
4122 }
4123 
4124 /**
4125  * i40e_vsi_control_tx - Start or stop a VSI's rings
4126  * @vsi: the VSI being configured
4127  * @enable: start or stop the rings
4128  **/
4129 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4130 {
4131 	struct i40e_pf *pf = vsi->back;
4132 	int i, pf_q, ret = 0;
4133 
4134 	pf_q = vsi->base_queue;
4135 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4136 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4137 					     pf_q,
4138 					     false /*is xdp*/, enable);
4139 		if (ret)
4140 			break;
4141 
4142 		if (!i40e_enabled_xdp_vsi(vsi))
4143 			continue;
4144 
4145 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4146 					     pf_q + vsi->alloc_queue_pairs,
4147 					     true /*is xdp*/, enable);
4148 		if (ret)
4149 			break;
4150 	}
4151 
4152 	return ret;
4153 }
4154 
4155 /**
4156  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4157  * @pf: the PF being configured
4158  * @pf_q: the PF queue
4159  * @enable: enable or disable state of the queue
4160  *
4161  * This routine will wait for the given Rx queue of the PF to reach the
4162  * enabled or disabled state.
4163  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4164  * multiple retries; else will return 0 in case of success.
4165  **/
4166 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4167 {
4168 	int i;
4169 	u32 rx_reg;
4170 
4171 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4172 		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4173 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4174 			break;
4175 
4176 		usleep_range(10, 20);
4177 	}
4178 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4179 		return -ETIMEDOUT;
4180 
4181 	return 0;
4182 }
4183 
4184 /**
4185  * i40e_control_rx_q - Start or stop a particular Rx queue
4186  * @pf: the PF structure
4187  * @pf_q: the PF queue to configure
4188  * @enable: start or stop the queue
4189  *
4190  * This function enables or disables a single queue. Note that any delay
4191  * required after the operation is expected to be handled by the caller of
4192  * this function.
4193  **/
4194 static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4195 {
4196 	struct i40e_hw *hw = &pf->hw;
4197 	u32 rx_reg;
4198 	int i;
4199 
4200 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4201 		rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4202 		if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4203 		    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4204 			break;
4205 		usleep_range(1000, 2000);
4206 	}
4207 
4208 	/* Skip if the queue is already in the requested state */
4209 	if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4210 		return;
4211 
4212 	/* turn on/off the queue */
4213 	if (enable)
4214 		rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4215 	else
4216 		rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4217 
4218 	wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4219 }
4220 
4221 /**
4222  * i40e_vsi_control_rx - Start or stop a VSI's rings
4223  * @vsi: the VSI being configured
4224  * @enable: start or stop the rings
4225  **/
4226 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4227 {
4228 	struct i40e_pf *pf = vsi->back;
4229 	int i, pf_q, ret = 0;
4230 
4231 	pf_q = vsi->base_queue;
4232 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4233 		i40e_control_rx_q(pf, pf_q, enable);
4234 
4235 		/* wait for the change to finish */
4236 		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4237 		if (ret) {
4238 			dev_info(&pf->pdev->dev,
4239 				 "VSI seid %d Rx ring %d %sable timeout\n",
4240 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
4241 			break;
4242 		}
4243 	}
4244 
4245 	/* Due to HW errata, on Rx disable only, the register can indicate done
4246 	 * before it really is. Needs 50ms to be sure
4247 	 */
4248 	if (!enable)
4249 		mdelay(50);
4250 
4251 	return ret;
4252 }
4253 
4254 /**
4255  * i40e_vsi_start_rings - Start a VSI's rings
4256  * @vsi: the VSI being configured
4257  **/
4258 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4259 {
4260 	int ret = 0;
4261 
4262 	/* do rx first for enable and last for disable */
4263 	ret = i40e_vsi_control_rx(vsi, true);
4264 	if (ret)
4265 		return ret;
4266 	ret = i40e_vsi_control_tx(vsi, true);
4267 
4268 	return ret;
4269 }
4270 
4271 /**
4272  * i40e_vsi_stop_rings - Stop a VSI's rings
4273  * @vsi: the VSI being configured
4274  **/
4275 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4276 {
4277 	/* When port TX is suspended, don't wait */
4278 	if (test_bit(__I40E_PORT_SUSPENDED, vsi->back->state))
4279 		return i40e_vsi_stop_rings_no_wait(vsi);
4280 
4281 	/* do rx first for enable and last for disable
4282 	 * Ignore return value, we need to shutdown whatever we can
4283 	 */
4284 	i40e_vsi_control_tx(vsi, false);
4285 	i40e_vsi_control_rx(vsi, false);
4286 }
4287 
4288 /**
4289  * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4290  * @vsi: the VSI being shutdown
4291  *
4292  * This function stops all the rings for a VSI but does not delay to verify
4293  * that rings have been disabled. It is expected that the caller is shutting
4294  * down multiple VSIs at once and will delay together for all the VSIs after
4295  * initiating the shutdown. This is particularly useful for shutting down lots
4296  * of VFs together. Otherwise, a large delay can be incurred while configuring
4297  * each VSI in serial.
4298  **/
4299 void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4300 {
4301 	struct i40e_pf *pf = vsi->back;
4302 	int i, pf_q;
4303 
4304 	pf_q = vsi->base_queue;
4305 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4306 		i40e_control_tx_q(pf, pf_q, false);
4307 		i40e_control_rx_q(pf, pf_q, false);
4308 	}
4309 }
4310 
4311 /**
4312  * i40e_vsi_free_irq - Free the irq association with the OS
4313  * @vsi: the VSI being configured
4314  **/
4315 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4316 {
4317 	struct i40e_pf *pf = vsi->back;
4318 	struct i40e_hw *hw = &pf->hw;
4319 	int base = vsi->base_vector;
4320 	u32 val, qp;
4321 	int i;
4322 
4323 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4324 		if (!vsi->q_vectors)
4325 			return;
4326 
4327 		if (!vsi->irqs_ready)
4328 			return;
4329 
4330 		vsi->irqs_ready = false;
4331 		for (i = 0; i < vsi->num_q_vectors; i++) {
4332 			int irq_num;
4333 			u16 vector;
4334 
4335 			vector = i + base;
4336 			irq_num = pf->msix_entries[vector].vector;
4337 
4338 			/* free only the irqs that were actually requested */
4339 			if (!vsi->q_vectors[i] ||
4340 			    !vsi->q_vectors[i]->num_ringpairs)
4341 				continue;
4342 
4343 			/* clear the affinity notifier in the IRQ descriptor */
4344 			irq_set_affinity_notifier(irq_num, NULL);
4345 			/* clear the affinity_mask in the IRQ descriptor */
4346 			irq_set_affinity_hint(irq_num, NULL);
4347 			synchronize_irq(irq_num);
4348 			free_irq(irq_num, vsi->q_vectors[i]);
4349 
4350 			/* Tear down the interrupt queue link list
4351 			 *
4352 			 * We know that they come in pairs and always
4353 			 * the Rx first, then the Tx.  To clear the
4354 			 * link list, stick the EOL value into the
4355 			 * next_q field of the registers.
4356 			 */
4357 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4358 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4359 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4360 			val |= I40E_QUEUE_END_OF_LIST
4361 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4362 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4363 
4364 			while (qp != I40E_QUEUE_END_OF_LIST) {
4365 				u32 next;
4366 
4367 				val = rd32(hw, I40E_QINT_RQCTL(qp));
4368 
4369 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4370 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4371 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4372 					 I40E_QINT_RQCTL_INTEVENT_MASK);
4373 
4374 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4375 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4376 
4377 				wr32(hw, I40E_QINT_RQCTL(qp), val);
4378 
4379 				val = rd32(hw, I40E_QINT_TQCTL(qp));
4380 
4381 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4382 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4383 
4384 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4385 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4386 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4387 					 I40E_QINT_TQCTL_INTEVENT_MASK);
4388 
4389 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4390 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4391 
4392 				wr32(hw, I40E_QINT_TQCTL(qp), val);
4393 				qp = next;
4394 			}
4395 		}
4396 	} else {
4397 		free_irq(pf->pdev->irq, pf);
4398 
4399 		val = rd32(hw, I40E_PFINT_LNKLST0);
4400 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4401 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4402 		val |= I40E_QUEUE_END_OF_LIST
4403 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4404 		wr32(hw, I40E_PFINT_LNKLST0, val);
4405 
4406 		val = rd32(hw, I40E_QINT_RQCTL(qp));
4407 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4408 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4409 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4410 			 I40E_QINT_RQCTL_INTEVENT_MASK);
4411 
4412 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4413 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4414 
4415 		wr32(hw, I40E_QINT_RQCTL(qp), val);
4416 
4417 		val = rd32(hw, I40E_QINT_TQCTL(qp));
4418 
4419 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4420 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4421 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4422 			 I40E_QINT_TQCTL_INTEVENT_MASK);
4423 
4424 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4425 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4426 
4427 		wr32(hw, I40E_QINT_TQCTL(qp), val);
4428 	}
4429 }
4430 
4431 /**
4432  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4433  * @vsi: the VSI being configured
4434  * @v_idx: Index of vector to be freed
4435  *
4436  * This function frees the memory allocated to the q_vector.  In addition if
4437  * NAPI is enabled it will delete any references to the NAPI struct prior
4438  * to freeing the q_vector.
4439  **/
4440 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4441 {
4442 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4443 	struct i40e_ring *ring;
4444 
4445 	if (!q_vector)
4446 		return;
4447 
4448 	/* disassociate q_vector from rings */
4449 	i40e_for_each_ring(ring, q_vector->tx)
4450 		ring->q_vector = NULL;
4451 
4452 	i40e_for_each_ring(ring, q_vector->rx)
4453 		ring->q_vector = NULL;
4454 
4455 	/* only VSI w/ an associated netdev is set up w/ NAPI */
4456 	if (vsi->netdev)
4457 		netif_napi_del(&q_vector->napi);
4458 
4459 	vsi->q_vectors[v_idx] = NULL;
4460 
4461 	kfree_rcu(q_vector, rcu);
4462 }
4463 
4464 /**
4465  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4466  * @vsi: the VSI being un-configured
4467  *
4468  * This frees the memory allocated to the q_vectors and
4469  * deletes references to the NAPI struct.
4470  **/
4471 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4472 {
4473 	int v_idx;
4474 
4475 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4476 		i40e_free_q_vector(vsi, v_idx);
4477 }
4478 
4479 /**
4480  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4481  * @pf: board private structure
4482  **/
4483 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4484 {
4485 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4486 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4487 		pci_disable_msix(pf->pdev);
4488 		kfree(pf->msix_entries);
4489 		pf->msix_entries = NULL;
4490 		kfree(pf->irq_pile);
4491 		pf->irq_pile = NULL;
4492 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4493 		pci_disable_msi(pf->pdev);
4494 	}
4495 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4496 }
4497 
4498 /**
4499  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4500  * @pf: board private structure
4501  *
4502  * We go through and clear interrupt specific resources and reset the structure
4503  * to pre-load conditions
4504  **/
4505 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4506 {
4507 	int i;
4508 
4509 	i40e_stop_misc_vector(pf);
4510 	if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4511 		synchronize_irq(pf->msix_entries[0].vector);
4512 		free_irq(pf->msix_entries[0].vector, pf);
4513 	}
4514 
4515 	i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4516 		      I40E_IWARP_IRQ_PILE_ID);
4517 
4518 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4519 	for (i = 0; i < pf->num_alloc_vsi; i++)
4520 		if (pf->vsi[i])
4521 			i40e_vsi_free_q_vectors(pf->vsi[i]);
4522 	i40e_reset_interrupt_capability(pf);
4523 }
4524 
4525 /**
4526  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4527  * @vsi: the VSI being configured
4528  **/
4529 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4530 {
4531 	int q_idx;
4532 
4533 	if (!vsi->netdev)
4534 		return;
4535 
4536 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4537 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4538 
4539 		if (q_vector->rx.ring || q_vector->tx.ring)
4540 			napi_enable(&q_vector->napi);
4541 	}
4542 }
4543 
4544 /**
4545  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4546  * @vsi: the VSI being configured
4547  **/
4548 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4549 {
4550 	int q_idx;
4551 
4552 	if (!vsi->netdev)
4553 		return;
4554 
4555 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4556 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4557 
4558 		if (q_vector->rx.ring || q_vector->tx.ring)
4559 			napi_disable(&q_vector->napi);
4560 	}
4561 }
4562 
4563 /**
4564  * i40e_vsi_close - Shut down a VSI
4565  * @vsi: the vsi to be quelled
4566  **/
4567 static void i40e_vsi_close(struct i40e_vsi *vsi)
4568 {
4569 	struct i40e_pf *pf = vsi->back;
4570 	if (!test_and_set_bit(__I40E_VSI_DOWN, vsi->state))
4571 		i40e_down(vsi);
4572 	i40e_vsi_free_irq(vsi);
4573 	i40e_vsi_free_tx_resources(vsi);
4574 	i40e_vsi_free_rx_resources(vsi);
4575 	vsi->current_netdev_flags = 0;
4576 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
4577 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
4578 		pf->flags |=  I40E_FLAG_CLIENT_RESET;
4579 }
4580 
4581 /**
4582  * i40e_quiesce_vsi - Pause a given VSI
4583  * @vsi: the VSI being paused
4584  **/
4585 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4586 {
4587 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
4588 		return;
4589 
4590 	set_bit(__I40E_VSI_NEEDS_RESTART, vsi->state);
4591 	if (vsi->netdev && netif_running(vsi->netdev))
4592 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4593 	else
4594 		i40e_vsi_close(vsi);
4595 }
4596 
4597 /**
4598  * i40e_unquiesce_vsi - Resume a given VSI
4599  * @vsi: the VSI being resumed
4600  **/
4601 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4602 {
4603 	if (!test_and_clear_bit(__I40E_VSI_NEEDS_RESTART, vsi->state))
4604 		return;
4605 
4606 	if (vsi->netdev && netif_running(vsi->netdev))
4607 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4608 	else
4609 		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4610 }
4611 
4612 /**
4613  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4614  * @pf: the PF
4615  **/
4616 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4617 {
4618 	int v;
4619 
4620 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4621 		if (pf->vsi[v])
4622 			i40e_quiesce_vsi(pf->vsi[v]);
4623 	}
4624 }
4625 
4626 /**
4627  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4628  * @pf: the PF
4629  **/
4630 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4631 {
4632 	int v;
4633 
4634 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4635 		if (pf->vsi[v])
4636 			i40e_unquiesce_vsi(pf->vsi[v]);
4637 	}
4638 }
4639 
4640 /**
4641  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4642  * @vsi: the VSI being configured
4643  *
4644  * Wait until all queues on a given VSI have been disabled.
4645  **/
4646 int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4647 {
4648 	struct i40e_pf *pf = vsi->back;
4649 	int i, pf_q, ret;
4650 
4651 	pf_q = vsi->base_queue;
4652 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4653 		/* Check and wait for the Tx queue */
4654 		ret = i40e_pf_txq_wait(pf, pf_q, false);
4655 		if (ret) {
4656 			dev_info(&pf->pdev->dev,
4657 				 "VSI seid %d Tx ring %d disable timeout\n",
4658 				 vsi->seid, pf_q);
4659 			return ret;
4660 		}
4661 
4662 		if (!i40e_enabled_xdp_vsi(vsi))
4663 			goto wait_rx;
4664 
4665 		/* Check and wait for the XDP Tx queue */
4666 		ret = i40e_pf_txq_wait(pf, pf_q + vsi->alloc_queue_pairs,
4667 				       false);
4668 		if (ret) {
4669 			dev_info(&pf->pdev->dev,
4670 				 "VSI seid %d XDP Tx ring %d disable timeout\n",
4671 				 vsi->seid, pf_q);
4672 			return ret;
4673 		}
4674 wait_rx:
4675 		/* Check and wait for the Rx queue */
4676 		ret = i40e_pf_rxq_wait(pf, pf_q, false);
4677 		if (ret) {
4678 			dev_info(&pf->pdev->dev,
4679 				 "VSI seid %d Rx ring %d disable timeout\n",
4680 				 vsi->seid, pf_q);
4681 			return ret;
4682 		}
4683 	}
4684 
4685 	return 0;
4686 }
4687 
4688 #ifdef CONFIG_I40E_DCB
4689 /**
4690  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4691  * @pf: the PF
4692  *
4693  * This function waits for the queues to be in disabled state for all the
4694  * VSIs that are managed by this PF.
4695  **/
4696 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4697 {
4698 	int v, ret = 0;
4699 
4700 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4701 		if (pf->vsi[v]) {
4702 			ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4703 			if (ret)
4704 				break;
4705 		}
4706 	}
4707 
4708 	return ret;
4709 }
4710 
4711 #endif
4712 
4713 /**
4714  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4715  * @q_idx: TX queue number
4716  * @vsi: Pointer to VSI struct
4717  *
4718  * This function checks specified queue for given VSI. Detects hung condition.
4719  * We proactively detect hung TX queues by checking if interrupts are disabled
4720  * but there are pending descriptors.  If it appears hung, attempt to recover
4721  * by triggering a SW interrupt.
4722  **/
4723 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4724 {
4725 	struct i40e_ring *tx_ring = NULL;
4726 	struct i40e_pf	*pf;
4727 	u32 val, tx_pending;
4728 	int i;
4729 
4730 	pf = vsi->back;
4731 
4732 	/* now that we have an index, find the tx_ring struct */
4733 	for (i = 0; i < vsi->num_queue_pairs; i++) {
4734 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4735 			if (q_idx == vsi->tx_rings[i]->queue_index) {
4736 				tx_ring = vsi->tx_rings[i];
4737 				break;
4738 			}
4739 		}
4740 	}
4741 
4742 	if (!tx_ring)
4743 		return;
4744 
4745 	/* Read interrupt register */
4746 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4747 		val = rd32(&pf->hw,
4748 			   I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4749 					       tx_ring->vsi->base_vector - 1));
4750 	else
4751 		val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4752 
4753 	tx_pending = i40e_get_tx_pending(tx_ring);
4754 
4755 	/* Interrupts are disabled and TX pending is non-zero,
4756 	 * trigger the SW interrupt (don't wait). Worst case
4757 	 * there will be one extra interrupt which may result
4758 	 * into not cleaning any queues because queues are cleaned.
4759 	 */
4760 	if (tx_pending && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)))
4761 		i40e_force_wb(vsi, tx_ring->q_vector);
4762 }
4763 
4764 /**
4765  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4766  * @pf:  pointer to PF struct
4767  *
4768  * LAN VSI has netdev and netdev has TX queues. This function is to check
4769  * each of those TX queues if they are hung, trigger recovery by issuing
4770  * SW interrupt.
4771  **/
4772 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4773 {
4774 	struct net_device *netdev;
4775 	struct i40e_vsi *vsi;
4776 	int i;
4777 
4778 	/* Only for LAN VSI */
4779 	vsi = pf->vsi[pf->lan_vsi];
4780 
4781 	if (!vsi)
4782 		return;
4783 
4784 	/* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4785 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
4786 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
4787 		return;
4788 
4789 	/* Make sure type is MAIN VSI */
4790 	if (vsi->type != I40E_VSI_MAIN)
4791 		return;
4792 
4793 	netdev = vsi->netdev;
4794 	if (!netdev)
4795 		return;
4796 
4797 	/* Bail out if netif_carrier is not OK */
4798 	if (!netif_carrier_ok(netdev))
4799 		return;
4800 
4801 	/* Go thru' TX queues for netdev */
4802 	for (i = 0; i < netdev->num_tx_queues; i++) {
4803 		struct netdev_queue *q;
4804 
4805 		q = netdev_get_tx_queue(netdev, i);
4806 		if (q)
4807 			i40e_detect_recover_hung_queue(i, vsi);
4808 	}
4809 }
4810 
4811 /**
4812  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4813  * @pf: pointer to PF
4814  *
4815  * Get TC map for ISCSI PF type that will include iSCSI TC
4816  * and LAN TC.
4817  **/
4818 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4819 {
4820 	struct i40e_dcb_app_priority_table app;
4821 	struct i40e_hw *hw = &pf->hw;
4822 	u8 enabled_tc = 1; /* TC0 is always enabled */
4823 	u8 tc, i;
4824 	/* Get the iSCSI APP TLV */
4825 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4826 
4827 	for (i = 0; i < dcbcfg->numapps; i++) {
4828 		app = dcbcfg->app[i];
4829 		if (app.selector == I40E_APP_SEL_TCPIP &&
4830 		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4831 			tc = dcbcfg->etscfg.prioritytable[app.priority];
4832 			enabled_tc |= BIT(tc);
4833 			break;
4834 		}
4835 	}
4836 
4837 	return enabled_tc;
4838 }
4839 
4840 /**
4841  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4842  * @dcbcfg: the corresponding DCBx configuration structure
4843  *
4844  * Return the number of TCs from given DCBx configuration
4845  **/
4846 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4847 {
4848 	int i, tc_unused = 0;
4849 	u8 num_tc = 0;
4850 	u8 ret = 0;
4851 
4852 	/* Scan the ETS Config Priority Table to find
4853 	 * traffic class enabled for a given priority
4854 	 * and create a bitmask of enabled TCs
4855 	 */
4856 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4857 		num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4858 
4859 	/* Now scan the bitmask to check for
4860 	 * contiguous TCs starting with TC0
4861 	 */
4862 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4863 		if (num_tc & BIT(i)) {
4864 			if (!tc_unused) {
4865 				ret++;
4866 			} else {
4867 				pr_err("Non-contiguous TC - Disabling DCB\n");
4868 				return 1;
4869 			}
4870 		} else {
4871 			tc_unused = 1;
4872 		}
4873 	}
4874 
4875 	/* There is always at least TC0 */
4876 	if (!ret)
4877 		ret = 1;
4878 
4879 	return ret;
4880 }
4881 
4882 /**
4883  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4884  * @dcbcfg: the corresponding DCBx configuration structure
4885  *
4886  * Query the current DCB configuration and return the number of
4887  * traffic classes enabled from the given DCBX config
4888  **/
4889 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4890 {
4891 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4892 	u8 enabled_tc = 1;
4893 	u8 i;
4894 
4895 	for (i = 0; i < num_tc; i++)
4896 		enabled_tc |= BIT(i);
4897 
4898 	return enabled_tc;
4899 }
4900 
4901 /**
4902  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4903  * @pf: PF being queried
4904  *
4905  * Return number of traffic classes enabled for the given PF
4906  **/
4907 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4908 {
4909 	struct i40e_hw *hw = &pf->hw;
4910 	u8 i, enabled_tc = 1;
4911 	u8 num_tc = 0;
4912 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4913 
4914 	/* If DCB is not enabled then always in single TC */
4915 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4916 		return 1;
4917 
4918 	/* SFP mode will be enabled for all TCs on port */
4919 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4920 		return i40e_dcb_get_num_tc(dcbcfg);
4921 
4922 	/* MFP mode return count of enabled TCs for this PF */
4923 	if (pf->hw.func_caps.iscsi)
4924 		enabled_tc =  i40e_get_iscsi_tc_map(pf);
4925 	else
4926 		return 1; /* Only TC0 */
4927 
4928 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4929 		if (enabled_tc & BIT(i))
4930 			num_tc++;
4931 	}
4932 	return num_tc;
4933 }
4934 
4935 /**
4936  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4937  * @pf: PF being queried
4938  *
4939  * Return a bitmap for enabled traffic classes for this PF.
4940  **/
4941 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4942 {
4943 	/* If DCB is not enabled for this PF then just return default TC */
4944 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4945 		return I40E_DEFAULT_TRAFFIC_CLASS;
4946 
4947 	/* SFP mode we want PF to be enabled for all TCs */
4948 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4949 		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4950 
4951 	/* MFP enabled and iSCSI PF type */
4952 	if (pf->hw.func_caps.iscsi)
4953 		return i40e_get_iscsi_tc_map(pf);
4954 	else
4955 		return I40E_DEFAULT_TRAFFIC_CLASS;
4956 }
4957 
4958 /**
4959  * i40e_vsi_get_bw_info - Query VSI BW Information
4960  * @vsi: the VSI being queried
4961  *
4962  * Returns 0 on success, negative value on failure
4963  **/
4964 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4965 {
4966 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4967 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4968 	struct i40e_pf *pf = vsi->back;
4969 	struct i40e_hw *hw = &pf->hw;
4970 	i40e_status ret;
4971 	u32 tc_bw_max;
4972 	int i;
4973 
4974 	/* Get the VSI level BW configuration */
4975 	ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4976 	if (ret) {
4977 		dev_info(&pf->pdev->dev,
4978 			 "couldn't get PF vsi bw config, err %s aq_err %s\n",
4979 			 i40e_stat_str(&pf->hw, ret),
4980 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4981 		return -EINVAL;
4982 	}
4983 
4984 	/* Get the VSI level BW configuration per TC */
4985 	ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4986 					       NULL);
4987 	if (ret) {
4988 		dev_info(&pf->pdev->dev,
4989 			 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4990 			 i40e_stat_str(&pf->hw, ret),
4991 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4992 		return -EINVAL;
4993 	}
4994 
4995 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4996 		dev_info(&pf->pdev->dev,
4997 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4998 			 bw_config.tc_valid_bits,
4999 			 bw_ets_config.tc_valid_bits);
5000 		/* Still continuing */
5001 	}
5002 
5003 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
5004 	vsi->bw_max_quanta = bw_config.max_bw;
5005 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
5006 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
5007 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5008 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
5009 		vsi->bw_ets_limit_credits[i] =
5010 					le16_to_cpu(bw_ets_config.credits[i]);
5011 		/* 3 bits out of 4 for each TC */
5012 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
5013 	}
5014 
5015 	return 0;
5016 }
5017 
5018 /**
5019  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
5020  * @vsi: the VSI being configured
5021  * @enabled_tc: TC bitmap
5022  * @bw_credits: BW shared credits per TC
5023  *
5024  * Returns 0 on success, negative value on failure
5025  **/
5026 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
5027 				       u8 *bw_share)
5028 {
5029 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5030 	i40e_status ret;
5031 	int i;
5032 
5033 	bw_data.tc_valid_bits = enabled_tc;
5034 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5035 		bw_data.tc_bw_credits[i] = bw_share[i];
5036 
5037 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
5038 				       NULL);
5039 	if (ret) {
5040 		dev_info(&vsi->back->pdev->dev,
5041 			 "AQ command Config VSI BW allocation per TC failed = %d\n",
5042 			 vsi->back->hw.aq.asq_last_status);
5043 		return -EINVAL;
5044 	}
5045 
5046 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5047 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
5048 
5049 	return 0;
5050 }
5051 
5052 /**
5053  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
5054  * @vsi: the VSI being configured
5055  * @enabled_tc: TC map to be enabled
5056  *
5057  **/
5058 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5059 {
5060 	struct net_device *netdev = vsi->netdev;
5061 	struct i40e_pf *pf = vsi->back;
5062 	struct i40e_hw *hw = &pf->hw;
5063 	u8 netdev_tc = 0;
5064 	int i;
5065 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5066 
5067 	if (!netdev)
5068 		return;
5069 
5070 	if (!enabled_tc) {
5071 		netdev_reset_tc(netdev);
5072 		return;
5073 	}
5074 
5075 	/* Set up actual enabled TCs on the VSI */
5076 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
5077 		return;
5078 
5079 	/* set per TC queues for the VSI */
5080 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5081 		/* Only set TC queues for enabled tcs
5082 		 *
5083 		 * e.g. For a VSI that has TC0 and TC3 enabled the
5084 		 * enabled_tc bitmap would be 0x00001001; the driver
5085 		 * will set the numtc for netdev as 2 that will be
5086 		 * referenced by the netdev layer as TC 0 and 1.
5087 		 */
5088 		if (vsi->tc_config.enabled_tc & BIT(i))
5089 			netdev_set_tc_queue(netdev,
5090 					vsi->tc_config.tc_info[i].netdev_tc,
5091 					vsi->tc_config.tc_info[i].qcount,
5092 					vsi->tc_config.tc_info[i].qoffset);
5093 	}
5094 
5095 	/* Assign UP2TC map for the VSI */
5096 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
5097 		/* Get the actual TC# for the UP */
5098 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
5099 		/* Get the mapped netdev TC# for the UP */
5100 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
5101 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
5102 	}
5103 }
5104 
5105 /**
5106  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
5107  * @vsi: the VSI being configured
5108  * @ctxt: the ctxt buffer returned from AQ VSI update param command
5109  **/
5110 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
5111 				      struct i40e_vsi_context *ctxt)
5112 {
5113 	/* copy just the sections touched not the entire info
5114 	 * since not all sections are valid as returned by
5115 	 * update vsi params
5116 	 */
5117 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
5118 	memcpy(&vsi->info.queue_mapping,
5119 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
5120 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
5121 	       sizeof(vsi->info.tc_mapping));
5122 }
5123 
5124 /**
5125  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
5126  * @vsi: VSI to be configured
5127  * @enabled_tc: TC bitmap
5128  *
5129  * This configures a particular VSI for TCs that are mapped to the
5130  * given TC bitmap. It uses default bandwidth share for TCs across
5131  * VSIs to configure TC for a particular VSI.
5132  *
5133  * NOTE:
5134  * It is expected that the VSI queues have been quisced before calling
5135  * this function.
5136  **/
5137 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5138 {
5139 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5140 	struct i40e_vsi_context ctxt;
5141 	int ret = 0;
5142 	int i;
5143 
5144 	/* Check if enabled_tc is same as existing or new TCs */
5145 	if (vsi->tc_config.enabled_tc == enabled_tc)
5146 		return ret;
5147 
5148 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
5149 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5150 		if (enabled_tc & BIT(i))
5151 			bw_share[i] = 1;
5152 	}
5153 
5154 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5155 	if (ret) {
5156 		dev_info(&vsi->back->pdev->dev,
5157 			 "Failed configuring TC map %d for VSI %d\n",
5158 			 enabled_tc, vsi->seid);
5159 		goto out;
5160 	}
5161 
5162 	/* Update Queue Pairs Mapping for currently enabled UPs */
5163 	ctxt.seid = vsi->seid;
5164 	ctxt.pf_num = vsi->back->hw.pf_id;
5165 	ctxt.vf_num = 0;
5166 	ctxt.uplink_seid = vsi->uplink_seid;
5167 	ctxt.info = vsi->info;
5168 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5169 
5170 	if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5171 		ctxt.info.valid_sections |=
5172 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5173 		ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5174 	}
5175 
5176 	/* Update the VSI after updating the VSI queue-mapping information */
5177 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5178 	if (ret) {
5179 		dev_info(&vsi->back->pdev->dev,
5180 			 "Update vsi tc config failed, err %s aq_err %s\n",
5181 			 i40e_stat_str(&vsi->back->hw, ret),
5182 			 i40e_aq_str(&vsi->back->hw,
5183 				     vsi->back->hw.aq.asq_last_status));
5184 		goto out;
5185 	}
5186 	/* update the local VSI info with updated queue map */
5187 	i40e_vsi_update_queue_map(vsi, &ctxt);
5188 	vsi->info.valid_sections = 0;
5189 
5190 	/* Update current VSI BW information */
5191 	ret = i40e_vsi_get_bw_info(vsi);
5192 	if (ret) {
5193 		dev_info(&vsi->back->pdev->dev,
5194 			 "Failed updating vsi bw info, err %s aq_err %s\n",
5195 			 i40e_stat_str(&vsi->back->hw, ret),
5196 			 i40e_aq_str(&vsi->back->hw,
5197 				     vsi->back->hw.aq.asq_last_status));
5198 		goto out;
5199 	}
5200 
5201 	/* Update the netdev TC setup */
5202 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5203 out:
5204 	return ret;
5205 }
5206 
5207 /**
5208  * i40e_veb_config_tc - Configure TCs for given VEB
5209  * @veb: given VEB
5210  * @enabled_tc: TC bitmap
5211  *
5212  * Configures given TC bitmap for VEB (switching) element
5213  **/
5214 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
5215 {
5216 	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
5217 	struct i40e_pf *pf = veb->pf;
5218 	int ret = 0;
5219 	int i;
5220 
5221 	/* No TCs or already enabled TCs just return */
5222 	if (!enabled_tc || veb->enabled_tc == enabled_tc)
5223 		return ret;
5224 
5225 	bw_data.tc_valid_bits = enabled_tc;
5226 	/* bw_data.absolute_credits is not set (relative) */
5227 
5228 	/* Enable ETS TCs with equal BW Share for now */
5229 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5230 		if (enabled_tc & BIT(i))
5231 			bw_data.tc_bw_share_credits[i] = 1;
5232 	}
5233 
5234 	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
5235 						   &bw_data, NULL);
5236 	if (ret) {
5237 		dev_info(&pf->pdev->dev,
5238 			 "VEB bw config failed, err %s aq_err %s\n",
5239 			 i40e_stat_str(&pf->hw, ret),
5240 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5241 		goto out;
5242 	}
5243 
5244 	/* Update the BW information */
5245 	ret = i40e_veb_get_bw_info(veb);
5246 	if (ret) {
5247 		dev_info(&pf->pdev->dev,
5248 			 "Failed getting veb bw config, err %s aq_err %s\n",
5249 			 i40e_stat_str(&pf->hw, ret),
5250 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5251 	}
5252 
5253 out:
5254 	return ret;
5255 }
5256 
5257 #ifdef CONFIG_I40E_DCB
5258 /**
5259  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
5260  * @pf: PF struct
5261  *
5262  * Reconfigure VEB/VSIs on a given PF; it is assumed that
5263  * the caller would've quiesce all the VSIs before calling
5264  * this function
5265  **/
5266 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
5267 {
5268 	u8 tc_map = 0;
5269 	int ret;
5270 	u8 v;
5271 
5272 	/* Enable the TCs available on PF to all VEBs */
5273 	tc_map = i40e_pf_get_tc_map(pf);
5274 	for (v = 0; v < I40E_MAX_VEB; v++) {
5275 		if (!pf->veb[v])
5276 			continue;
5277 		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
5278 		if (ret) {
5279 			dev_info(&pf->pdev->dev,
5280 				 "Failed configuring TC for VEB seid=%d\n",
5281 				 pf->veb[v]->seid);
5282 			/* Will try to configure as many components */
5283 		}
5284 	}
5285 
5286 	/* Update each VSI */
5287 	for (v = 0; v < pf->num_alloc_vsi; v++) {
5288 		if (!pf->vsi[v])
5289 			continue;
5290 
5291 		/* - Enable all TCs for the LAN VSI
5292 		 * - For all others keep them at TC0 for now
5293 		 */
5294 		if (v == pf->lan_vsi)
5295 			tc_map = i40e_pf_get_tc_map(pf);
5296 		else
5297 			tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
5298 
5299 		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5300 		if (ret) {
5301 			dev_info(&pf->pdev->dev,
5302 				 "Failed configuring TC for VSI seid=%d\n",
5303 				 pf->vsi[v]->seid);
5304 			/* Will try to configure as many components */
5305 		} else {
5306 			/* Re-configure VSI vectors based on updated TC map */
5307 			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5308 			if (pf->vsi[v]->netdev)
5309 				i40e_dcbnl_set_all(pf->vsi[v]);
5310 		}
5311 	}
5312 }
5313 
5314 /**
5315  * i40e_resume_port_tx - Resume port Tx
5316  * @pf: PF struct
5317  *
5318  * Resume a port's Tx and issue a PF reset in case of failure to
5319  * resume.
5320  **/
5321 static int i40e_resume_port_tx(struct i40e_pf *pf)
5322 {
5323 	struct i40e_hw *hw = &pf->hw;
5324 	int ret;
5325 
5326 	ret = i40e_aq_resume_port_tx(hw, NULL);
5327 	if (ret) {
5328 		dev_info(&pf->pdev->dev,
5329 			 "Resume Port Tx failed, err %s aq_err %s\n",
5330 			  i40e_stat_str(&pf->hw, ret),
5331 			  i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5332 		/* Schedule PF reset to recover */
5333 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
5334 		i40e_service_event_schedule(pf);
5335 	}
5336 
5337 	return ret;
5338 }
5339 
5340 /**
5341  * i40e_init_pf_dcb - Initialize DCB configuration
5342  * @pf: PF being configured
5343  *
5344  * Query the current DCB configuration and cache it
5345  * in the hardware structure
5346  **/
5347 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5348 {
5349 	struct i40e_hw *hw = &pf->hw;
5350 	int err = 0;
5351 
5352 	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5353 	if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5354 		goto out;
5355 
5356 	/* Get the initial DCB configuration */
5357 	err = i40e_init_dcb(hw);
5358 	if (!err) {
5359 		/* Device/Function is not DCBX capable */
5360 		if ((!hw->func_caps.dcb) ||
5361 		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5362 			dev_info(&pf->pdev->dev,
5363 				 "DCBX offload is not supported or is disabled for this PF.\n");
5364 		} else {
5365 			/* When status is not DISABLED then DCBX in FW */
5366 			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5367 				       DCB_CAP_DCBX_VER_IEEE;
5368 
5369 			pf->flags |= I40E_FLAG_DCB_CAPABLE;
5370 			/* Enable DCB tagging only when more than one TC
5371 			 * or explicitly disable if only one TC
5372 			 */
5373 			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5374 				pf->flags |= I40E_FLAG_DCB_ENABLED;
5375 			else
5376 				pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5377 			dev_dbg(&pf->pdev->dev,
5378 				"DCBX offload is supported for this PF.\n");
5379 		}
5380 	} else {
5381 		dev_info(&pf->pdev->dev,
5382 			 "Query for DCB configuration failed, err %s aq_err %s\n",
5383 			 i40e_stat_str(&pf->hw, err),
5384 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5385 	}
5386 
5387 out:
5388 	return err;
5389 }
5390 #endif /* CONFIG_I40E_DCB */
5391 #define SPEED_SIZE 14
5392 #define FC_SIZE 8
5393 /**
5394  * i40e_print_link_message - print link up or down
5395  * @vsi: the VSI for which link needs a message
5396  */
5397 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5398 {
5399 	enum i40e_aq_link_speed new_speed;
5400 	char *speed = "Unknown";
5401 	char *fc = "Unknown";
5402 	char *fec = "";
5403 	char *an = "";
5404 
5405 	new_speed = vsi->back->hw.phy.link_info.link_speed;
5406 
5407 	if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
5408 		return;
5409 	vsi->current_isup = isup;
5410 	vsi->current_speed = new_speed;
5411 	if (!isup) {
5412 		netdev_info(vsi->netdev, "NIC Link is Down\n");
5413 		return;
5414 	}
5415 
5416 	/* Warn user if link speed on NPAR enabled partition is not at
5417 	 * least 10GB
5418 	 */
5419 	if (vsi->back->hw.func_caps.npar_enable &&
5420 	    (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5421 	     vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5422 		netdev_warn(vsi->netdev,
5423 			    "The partition detected link speed that is less than 10Gbps\n");
5424 
5425 	switch (vsi->back->hw.phy.link_info.link_speed) {
5426 	case I40E_LINK_SPEED_40GB:
5427 		speed = "40 G";
5428 		break;
5429 	case I40E_LINK_SPEED_20GB:
5430 		speed = "20 G";
5431 		break;
5432 	case I40E_LINK_SPEED_25GB:
5433 		speed = "25 G";
5434 		break;
5435 	case I40E_LINK_SPEED_10GB:
5436 		speed = "10 G";
5437 		break;
5438 	case I40E_LINK_SPEED_1GB:
5439 		speed = "1000 M";
5440 		break;
5441 	case I40E_LINK_SPEED_100MB:
5442 		speed = "100 M";
5443 		break;
5444 	default:
5445 		break;
5446 	}
5447 
5448 	switch (vsi->back->hw.fc.current_mode) {
5449 	case I40E_FC_FULL:
5450 		fc = "RX/TX";
5451 		break;
5452 	case I40E_FC_TX_PAUSE:
5453 		fc = "TX";
5454 		break;
5455 	case I40E_FC_RX_PAUSE:
5456 		fc = "RX";
5457 		break;
5458 	default:
5459 		fc = "None";
5460 		break;
5461 	}
5462 
5463 	if (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
5464 		fec = ", FEC: None";
5465 		an = ", Autoneg: False";
5466 
5467 		if (vsi->back->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
5468 			an = ", Autoneg: True";
5469 
5470 		if (vsi->back->hw.phy.link_info.fec_info &
5471 		    I40E_AQ_CONFIG_FEC_KR_ENA)
5472 			fec = ", FEC: CL74 FC-FEC/BASE-R";
5473 		else if (vsi->back->hw.phy.link_info.fec_info &
5474 			 I40E_AQ_CONFIG_FEC_RS_ENA)
5475 			fec = ", FEC: CL108 RS-FEC";
5476 	}
5477 
5478 	netdev_info(vsi->netdev, "NIC Link is Up, %sbps Full Duplex%s%s, Flow Control: %s\n",
5479 		    speed, fec, an, fc);
5480 }
5481 
5482 /**
5483  * i40e_up_complete - Finish the last steps of bringing up a connection
5484  * @vsi: the VSI being configured
5485  **/
5486 static int i40e_up_complete(struct i40e_vsi *vsi)
5487 {
5488 	struct i40e_pf *pf = vsi->back;
5489 	int err;
5490 
5491 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5492 		i40e_vsi_configure_msix(vsi);
5493 	else
5494 		i40e_configure_msi_and_legacy(vsi);
5495 
5496 	/* start rings */
5497 	err = i40e_vsi_start_rings(vsi);
5498 	if (err)
5499 		return err;
5500 
5501 	clear_bit(__I40E_VSI_DOWN, vsi->state);
5502 	i40e_napi_enable_all(vsi);
5503 	i40e_vsi_enable_irq(vsi);
5504 
5505 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5506 	    (vsi->netdev)) {
5507 		i40e_print_link_message(vsi, true);
5508 		netif_tx_start_all_queues(vsi->netdev);
5509 		netif_carrier_on(vsi->netdev);
5510 	} else if (vsi->netdev) {
5511 		i40e_print_link_message(vsi, false);
5512 		/* need to check for qualified module here*/
5513 		if ((pf->hw.phy.link_info.link_info &
5514 			I40E_AQ_MEDIA_AVAILABLE) &&
5515 		    (!(pf->hw.phy.link_info.an_info &
5516 			I40E_AQ_QUALIFIED_MODULE)))
5517 			netdev_err(vsi->netdev,
5518 				   "the driver failed to link because an unqualified module was detected.");
5519 	}
5520 
5521 	/* replay FDIR SB filters */
5522 	if (vsi->type == I40E_VSI_FDIR) {
5523 		/* reset fd counters */
5524 		pf->fd_add_err = 0;
5525 		pf->fd_atr_cnt = 0;
5526 		i40e_fdir_filter_restore(vsi);
5527 	}
5528 
5529 	/* On the next run of the service_task, notify any clients of the new
5530 	 * opened netdev
5531 	 */
5532 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5533 	i40e_service_event_schedule(pf);
5534 
5535 	return 0;
5536 }
5537 
5538 /**
5539  * i40e_vsi_reinit_locked - Reset the VSI
5540  * @vsi: the VSI being configured
5541  *
5542  * Rebuild the ring structs after some configuration
5543  * has changed, e.g. MTU size.
5544  **/
5545 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5546 {
5547 	struct i40e_pf *pf = vsi->back;
5548 
5549 	WARN_ON(in_interrupt());
5550 	while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state))
5551 		usleep_range(1000, 2000);
5552 	i40e_down(vsi);
5553 
5554 	i40e_up(vsi);
5555 	clear_bit(__I40E_CONFIG_BUSY, pf->state);
5556 }
5557 
5558 /**
5559  * i40e_up - Bring the connection back up after being down
5560  * @vsi: the VSI being configured
5561  **/
5562 int i40e_up(struct i40e_vsi *vsi)
5563 {
5564 	int err;
5565 
5566 	err = i40e_vsi_configure(vsi);
5567 	if (!err)
5568 		err = i40e_up_complete(vsi);
5569 
5570 	return err;
5571 }
5572 
5573 /**
5574  * i40e_down - Shutdown the connection processing
5575  * @vsi: the VSI being stopped
5576  **/
5577 void i40e_down(struct i40e_vsi *vsi)
5578 {
5579 	int i;
5580 
5581 	/* It is assumed that the caller of this function
5582 	 * sets the vsi->state __I40E_VSI_DOWN bit.
5583 	 */
5584 	if (vsi->netdev) {
5585 		netif_carrier_off(vsi->netdev);
5586 		netif_tx_disable(vsi->netdev);
5587 	}
5588 	i40e_vsi_disable_irq(vsi);
5589 	i40e_vsi_stop_rings(vsi);
5590 	i40e_napi_disable_all(vsi);
5591 
5592 	for (i = 0; i < vsi->num_queue_pairs; i++) {
5593 		i40e_clean_tx_ring(vsi->tx_rings[i]);
5594 		if (i40e_enabled_xdp_vsi(vsi))
5595 			i40e_clean_tx_ring(vsi->xdp_rings[i]);
5596 		i40e_clean_rx_ring(vsi->rx_rings[i]);
5597 	}
5598 
5599 }
5600 
5601 /**
5602  * i40e_setup_tc - configure multiple traffic classes
5603  * @netdev: net device to configure
5604  * @tc: number of traffic classes to enable
5605  **/
5606 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5607 {
5608 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5609 	struct i40e_vsi *vsi = np->vsi;
5610 	struct i40e_pf *pf = vsi->back;
5611 	u8 enabled_tc = 0;
5612 	int ret = -EINVAL;
5613 	int i;
5614 
5615 	/* Check if DCB enabled to continue */
5616 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5617 		netdev_info(netdev, "DCB is not enabled for adapter\n");
5618 		goto exit;
5619 	}
5620 
5621 	/* Check if MFP enabled */
5622 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5623 		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5624 		goto exit;
5625 	}
5626 
5627 	/* Check whether tc count is within enabled limit */
5628 	if (tc > i40e_pf_get_num_tc(pf)) {
5629 		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5630 		goto exit;
5631 	}
5632 
5633 	/* Generate TC map for number of tc requested */
5634 	for (i = 0; i < tc; i++)
5635 		enabled_tc |= BIT(i);
5636 
5637 	/* Requesting same TC configuration as already enabled */
5638 	if (enabled_tc == vsi->tc_config.enabled_tc)
5639 		return 0;
5640 
5641 	/* Quiesce VSI queues */
5642 	i40e_quiesce_vsi(vsi);
5643 
5644 	/* Configure VSI for enabled TCs */
5645 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
5646 	if (ret) {
5647 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5648 			    vsi->seid);
5649 		goto exit;
5650 	}
5651 
5652 	/* Unquiesce VSI */
5653 	i40e_unquiesce_vsi(vsi);
5654 
5655 exit:
5656 	return ret;
5657 }
5658 
5659 static int __i40e_setup_tc(struct net_device *netdev, u32 handle,
5660 			   u32 chain_index, __be16 proto,
5661 			   struct tc_to_netdev *tc)
5662 {
5663 	if (tc->type != TC_SETUP_MQPRIO)
5664 		return -EINVAL;
5665 
5666 	tc->mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
5667 
5668 	return i40e_setup_tc(netdev, tc->mqprio->num_tc);
5669 }
5670 
5671 /**
5672  * i40e_open - Called when a network interface is made active
5673  * @netdev: network interface device structure
5674  *
5675  * The open entry point is called when a network interface is made
5676  * active by the system (IFF_UP).  At this point all resources needed
5677  * for transmit and receive operations are allocated, the interrupt
5678  * handler is registered with the OS, the netdev watchdog subtask is
5679  * enabled, and the stack is notified that the interface is ready.
5680  *
5681  * Returns 0 on success, negative value on failure
5682  **/
5683 int i40e_open(struct net_device *netdev)
5684 {
5685 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5686 	struct i40e_vsi *vsi = np->vsi;
5687 	struct i40e_pf *pf = vsi->back;
5688 	int err;
5689 
5690 	/* disallow open during test or if eeprom is broken */
5691 	if (test_bit(__I40E_TESTING, pf->state) ||
5692 	    test_bit(__I40E_BAD_EEPROM, pf->state))
5693 		return -EBUSY;
5694 
5695 	netif_carrier_off(netdev);
5696 
5697 	err = i40e_vsi_open(vsi);
5698 	if (err)
5699 		return err;
5700 
5701 	/* configure global TSO hardware offload settings */
5702 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5703 						       TCP_FLAG_FIN) >> 16);
5704 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5705 						       TCP_FLAG_FIN |
5706 						       TCP_FLAG_CWR) >> 16);
5707 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5708 
5709 	udp_tunnel_get_rx_info(netdev);
5710 
5711 	return 0;
5712 }
5713 
5714 /**
5715  * i40e_vsi_open -
5716  * @vsi: the VSI to open
5717  *
5718  * Finish initialization of the VSI.
5719  *
5720  * Returns 0 on success, negative value on failure
5721  *
5722  * Note: expects to be called while under rtnl_lock()
5723  **/
5724 int i40e_vsi_open(struct i40e_vsi *vsi)
5725 {
5726 	struct i40e_pf *pf = vsi->back;
5727 	char int_name[I40E_INT_NAME_STR_LEN];
5728 	int err;
5729 
5730 	/* allocate descriptors */
5731 	err = i40e_vsi_setup_tx_resources(vsi);
5732 	if (err)
5733 		goto err_setup_tx;
5734 	err = i40e_vsi_setup_rx_resources(vsi);
5735 	if (err)
5736 		goto err_setup_rx;
5737 
5738 	err = i40e_vsi_configure(vsi);
5739 	if (err)
5740 		goto err_setup_rx;
5741 
5742 	if (vsi->netdev) {
5743 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5744 			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5745 		err = i40e_vsi_request_irq(vsi, int_name);
5746 		if (err)
5747 			goto err_setup_rx;
5748 
5749 		/* Notify the stack of the actual queue counts. */
5750 		err = netif_set_real_num_tx_queues(vsi->netdev,
5751 						   vsi->num_queue_pairs);
5752 		if (err)
5753 			goto err_set_queues;
5754 
5755 		err = netif_set_real_num_rx_queues(vsi->netdev,
5756 						   vsi->num_queue_pairs);
5757 		if (err)
5758 			goto err_set_queues;
5759 
5760 	} else if (vsi->type == I40E_VSI_FDIR) {
5761 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5762 			 dev_driver_string(&pf->pdev->dev),
5763 			 dev_name(&pf->pdev->dev));
5764 		err = i40e_vsi_request_irq(vsi, int_name);
5765 
5766 	} else {
5767 		err = -EINVAL;
5768 		goto err_setup_rx;
5769 	}
5770 
5771 	err = i40e_up_complete(vsi);
5772 	if (err)
5773 		goto err_up_complete;
5774 
5775 	return 0;
5776 
5777 err_up_complete:
5778 	i40e_down(vsi);
5779 err_set_queues:
5780 	i40e_vsi_free_irq(vsi);
5781 err_setup_rx:
5782 	i40e_vsi_free_rx_resources(vsi);
5783 err_setup_tx:
5784 	i40e_vsi_free_tx_resources(vsi);
5785 	if (vsi == pf->vsi[pf->lan_vsi])
5786 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
5787 
5788 	return err;
5789 }
5790 
5791 /**
5792  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5793  * @pf: Pointer to PF
5794  *
5795  * This function destroys the hlist where all the Flow Director
5796  * filters were saved.
5797  **/
5798 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5799 {
5800 	struct i40e_fdir_filter *filter;
5801 	struct i40e_flex_pit *pit_entry, *tmp;
5802 	struct hlist_node *node2;
5803 
5804 	hlist_for_each_entry_safe(filter, node2,
5805 				  &pf->fdir_filter_list, fdir_node) {
5806 		hlist_del(&filter->fdir_node);
5807 		kfree(filter);
5808 	}
5809 
5810 	list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
5811 		list_del(&pit_entry->list);
5812 		kfree(pit_entry);
5813 	}
5814 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
5815 
5816 	list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
5817 		list_del(&pit_entry->list);
5818 		kfree(pit_entry);
5819 	}
5820 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
5821 
5822 	pf->fdir_pf_active_filters = 0;
5823 	pf->fd_tcp4_filter_cnt = 0;
5824 	pf->fd_udp4_filter_cnt = 0;
5825 	pf->fd_sctp4_filter_cnt = 0;
5826 	pf->fd_ip4_filter_cnt = 0;
5827 
5828 	/* Reprogram the default input set for TCP/IPv4 */
5829 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
5830 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5831 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5832 
5833 	/* Reprogram the default input set for UDP/IPv4 */
5834 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
5835 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5836 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5837 
5838 	/* Reprogram the default input set for SCTP/IPv4 */
5839 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
5840 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5841 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5842 
5843 	/* Reprogram the default input set for Other/IPv4 */
5844 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
5845 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
5846 }
5847 
5848 /**
5849  * i40e_close - Disables a network interface
5850  * @netdev: network interface device structure
5851  *
5852  * The close entry point is called when an interface is de-activated
5853  * by the OS.  The hardware is still under the driver's control, but
5854  * this netdev interface is disabled.
5855  *
5856  * Returns 0, this is not allowed to fail
5857  **/
5858 int i40e_close(struct net_device *netdev)
5859 {
5860 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5861 	struct i40e_vsi *vsi = np->vsi;
5862 
5863 	i40e_vsi_close(vsi);
5864 
5865 	return 0;
5866 }
5867 
5868 /**
5869  * i40e_do_reset - Start a PF or Core Reset sequence
5870  * @pf: board private structure
5871  * @reset_flags: which reset is requested
5872  * @lock_acquired: indicates whether or not the lock has been acquired
5873  * before this function was called.
5874  *
5875  * The essential difference in resets is that the PF Reset
5876  * doesn't clear the packet buffers, doesn't reset the PE
5877  * firmware, and doesn't bother the other PFs on the chip.
5878  **/
5879 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
5880 {
5881 	u32 val;
5882 
5883 	WARN_ON(in_interrupt());
5884 
5885 
5886 	/* do the biggest reset indicated */
5887 	if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5888 
5889 		/* Request a Global Reset
5890 		 *
5891 		 * This will start the chip's countdown to the actual full
5892 		 * chip reset event, and a warning interrupt to be sent
5893 		 * to all PFs, including the requestor.  Our handler
5894 		 * for the warning interrupt will deal with the shutdown
5895 		 * and recovery of the switch setup.
5896 		 */
5897 		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5898 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5899 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5900 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5901 
5902 	} else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5903 
5904 		/* Request a Core Reset
5905 		 *
5906 		 * Same as Global Reset, except does *not* include the MAC/PHY
5907 		 */
5908 		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5909 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5910 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
5911 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5912 		i40e_flush(&pf->hw);
5913 
5914 	} else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5915 
5916 		/* Request a PF Reset
5917 		 *
5918 		 * Resets only the PF-specific registers
5919 		 *
5920 		 * This goes directly to the tear-down and rebuild of
5921 		 * the switch, since we need to do all the recovery as
5922 		 * for the Core Reset.
5923 		 */
5924 		dev_dbg(&pf->pdev->dev, "PFR requested\n");
5925 		i40e_handle_reset_warning(pf, lock_acquired);
5926 
5927 	} else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5928 		int v;
5929 
5930 		/* Find the VSI(s) that requested a re-init */
5931 		dev_info(&pf->pdev->dev,
5932 			 "VSI reinit requested\n");
5933 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5934 			struct i40e_vsi *vsi = pf->vsi[v];
5935 
5936 			if (vsi != NULL &&
5937 			    test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
5938 					       vsi->state))
5939 				i40e_vsi_reinit_locked(pf->vsi[v]);
5940 		}
5941 	} else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5942 		int v;
5943 
5944 		/* Find the VSI(s) that needs to be brought down */
5945 		dev_info(&pf->pdev->dev, "VSI down requested\n");
5946 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5947 			struct i40e_vsi *vsi = pf->vsi[v];
5948 
5949 			if (vsi != NULL &&
5950 			    test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
5951 					       vsi->state)) {
5952 				set_bit(__I40E_VSI_DOWN, vsi->state);
5953 				i40e_down(vsi);
5954 			}
5955 		}
5956 	} else {
5957 		dev_info(&pf->pdev->dev,
5958 			 "bad reset request 0x%08x\n", reset_flags);
5959 	}
5960 }
5961 
5962 #ifdef CONFIG_I40E_DCB
5963 /**
5964  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5965  * @pf: board private structure
5966  * @old_cfg: current DCB config
5967  * @new_cfg: new DCB config
5968  **/
5969 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5970 			    struct i40e_dcbx_config *old_cfg,
5971 			    struct i40e_dcbx_config *new_cfg)
5972 {
5973 	bool need_reconfig = false;
5974 
5975 	/* Check if ETS configuration has changed */
5976 	if (memcmp(&new_cfg->etscfg,
5977 		   &old_cfg->etscfg,
5978 		   sizeof(new_cfg->etscfg))) {
5979 		/* If Priority Table has changed reconfig is needed */
5980 		if (memcmp(&new_cfg->etscfg.prioritytable,
5981 			   &old_cfg->etscfg.prioritytable,
5982 			   sizeof(new_cfg->etscfg.prioritytable))) {
5983 			need_reconfig = true;
5984 			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5985 		}
5986 
5987 		if (memcmp(&new_cfg->etscfg.tcbwtable,
5988 			   &old_cfg->etscfg.tcbwtable,
5989 			   sizeof(new_cfg->etscfg.tcbwtable)))
5990 			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5991 
5992 		if (memcmp(&new_cfg->etscfg.tsatable,
5993 			   &old_cfg->etscfg.tsatable,
5994 			   sizeof(new_cfg->etscfg.tsatable)))
5995 			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5996 	}
5997 
5998 	/* Check if PFC configuration has changed */
5999 	if (memcmp(&new_cfg->pfc,
6000 		   &old_cfg->pfc,
6001 		   sizeof(new_cfg->pfc))) {
6002 		need_reconfig = true;
6003 		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
6004 	}
6005 
6006 	/* Check if APP Table has changed */
6007 	if (memcmp(&new_cfg->app,
6008 		   &old_cfg->app,
6009 		   sizeof(new_cfg->app))) {
6010 		need_reconfig = true;
6011 		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
6012 	}
6013 
6014 	dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
6015 	return need_reconfig;
6016 }
6017 
6018 /**
6019  * i40e_handle_lldp_event - Handle LLDP Change MIB event
6020  * @pf: board private structure
6021  * @e: event info posted on ARQ
6022  **/
6023 static int i40e_handle_lldp_event(struct i40e_pf *pf,
6024 				  struct i40e_arq_event_info *e)
6025 {
6026 	struct i40e_aqc_lldp_get_mib *mib =
6027 		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
6028 	struct i40e_hw *hw = &pf->hw;
6029 	struct i40e_dcbx_config tmp_dcbx_cfg;
6030 	bool need_reconfig = false;
6031 	int ret = 0;
6032 	u8 type;
6033 
6034 	/* Not DCB capable or capability disabled */
6035 	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
6036 		return ret;
6037 
6038 	/* Ignore if event is not for Nearest Bridge */
6039 	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
6040 		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
6041 	dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
6042 	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
6043 		return ret;
6044 
6045 	/* Check MIB Type and return if event for Remote MIB update */
6046 	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
6047 	dev_dbg(&pf->pdev->dev,
6048 		"LLDP event mib type %s\n", type ? "remote" : "local");
6049 	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
6050 		/* Update the remote cached instance and return */
6051 		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
6052 				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
6053 				&hw->remote_dcbx_config);
6054 		goto exit;
6055 	}
6056 
6057 	/* Store the old configuration */
6058 	tmp_dcbx_cfg = hw->local_dcbx_config;
6059 
6060 	/* Reset the old DCBx configuration data */
6061 	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
6062 	/* Get updated DCBX data from firmware */
6063 	ret = i40e_get_dcb_config(&pf->hw);
6064 	if (ret) {
6065 		dev_info(&pf->pdev->dev,
6066 			 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
6067 			 i40e_stat_str(&pf->hw, ret),
6068 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6069 		goto exit;
6070 	}
6071 
6072 	/* No change detected in DCBX configs */
6073 	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
6074 		    sizeof(tmp_dcbx_cfg))) {
6075 		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
6076 		goto exit;
6077 	}
6078 
6079 	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
6080 					       &hw->local_dcbx_config);
6081 
6082 	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
6083 
6084 	if (!need_reconfig)
6085 		goto exit;
6086 
6087 	/* Enable DCB tagging only when more than one TC */
6088 	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
6089 		pf->flags |= I40E_FLAG_DCB_ENABLED;
6090 	else
6091 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6092 
6093 	set_bit(__I40E_PORT_SUSPENDED, pf->state);
6094 	/* Reconfiguration needed quiesce all VSIs */
6095 	i40e_pf_quiesce_all_vsi(pf);
6096 
6097 	/* Changes in configuration update VEB/VSI */
6098 	i40e_dcb_reconfigure(pf);
6099 
6100 	ret = i40e_resume_port_tx(pf);
6101 
6102 	clear_bit(__I40E_PORT_SUSPENDED, pf->state);
6103 	/* In case of error no point in resuming VSIs */
6104 	if (ret)
6105 		goto exit;
6106 
6107 	/* Wait for the PF's queues to be disabled */
6108 	ret = i40e_pf_wait_queues_disabled(pf);
6109 	if (ret) {
6110 		/* Schedule PF reset to recover */
6111 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6112 		i40e_service_event_schedule(pf);
6113 	} else {
6114 		i40e_pf_unquiesce_all_vsi(pf);
6115 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
6116 		      I40E_FLAG_CLIENT_L2_CHANGE);
6117 	}
6118 
6119 exit:
6120 	return ret;
6121 }
6122 #endif /* CONFIG_I40E_DCB */
6123 
6124 /**
6125  * i40e_do_reset_safe - Protected reset path for userland calls.
6126  * @pf: board private structure
6127  * @reset_flags: which reset is requested
6128  *
6129  **/
6130 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
6131 {
6132 	rtnl_lock();
6133 	i40e_do_reset(pf, reset_flags, true);
6134 	rtnl_unlock();
6135 }
6136 
6137 /**
6138  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
6139  * @pf: board private structure
6140  * @e: event info posted on ARQ
6141  *
6142  * Handler for LAN Queue Overflow Event generated by the firmware for PF
6143  * and VF queues
6144  **/
6145 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
6146 					   struct i40e_arq_event_info *e)
6147 {
6148 	struct i40e_aqc_lan_overflow *data =
6149 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
6150 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
6151 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
6152 	struct i40e_hw *hw = &pf->hw;
6153 	struct i40e_vf *vf;
6154 	u16 vf_id;
6155 
6156 	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
6157 		queue, qtx_ctl);
6158 
6159 	/* Queue belongs to VF, find the VF and issue VF reset */
6160 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
6161 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
6162 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
6163 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
6164 		vf_id -= hw->func_caps.vf_base_id;
6165 		vf = &pf->vf[vf_id];
6166 		i40e_vc_notify_vf_reset(vf);
6167 		/* Allow VF to process pending reset notification */
6168 		msleep(20);
6169 		i40e_reset_vf(vf, false);
6170 	}
6171 }
6172 
6173 /**
6174  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
6175  * @pf: board private structure
6176  **/
6177 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
6178 {
6179 	u32 val, fcnt_prog;
6180 
6181 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6182 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
6183 	return fcnt_prog;
6184 }
6185 
6186 /**
6187  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
6188  * @pf: board private structure
6189  **/
6190 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
6191 {
6192 	u32 val, fcnt_prog;
6193 
6194 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6195 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
6196 		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
6197 		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
6198 	return fcnt_prog;
6199 }
6200 
6201 /**
6202  * i40e_get_global_fd_count - Get total FD filters programmed on device
6203  * @pf: board private structure
6204  **/
6205 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
6206 {
6207 	u32 val, fcnt_prog;
6208 
6209 	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
6210 	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
6211 		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
6212 		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
6213 	return fcnt_prog;
6214 }
6215 
6216 /**
6217  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
6218  * @pf: board private structure
6219  **/
6220 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
6221 {
6222 	struct i40e_fdir_filter *filter;
6223 	u32 fcnt_prog, fcnt_avail;
6224 	struct hlist_node *node;
6225 
6226 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
6227 		return;
6228 
6229 	/* Check if we have enough room to re-enable FDir SB capability. */
6230 	fcnt_prog = i40e_get_global_fd_count(pf);
6231 	fcnt_avail = pf->fdir_pf_filter_count;
6232 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
6233 	    (pf->fd_add_err == 0) ||
6234 	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
6235 		if (pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED) {
6236 			pf->flags &= ~I40E_FLAG_FD_SB_AUTO_DISABLED;
6237 			if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
6238 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
6239 				dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
6240 		}
6241 	}
6242 
6243 	/* We should wait for even more space before re-enabling ATR.
6244 	 * Additionally, we cannot enable ATR as long as we still have TCP SB
6245 	 * rules active.
6246 	 */
6247 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
6248 	    (pf->fd_tcp4_filter_cnt == 0)) {
6249 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
6250 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
6251 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
6252 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
6253 				dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
6254 		}
6255 	}
6256 
6257 	/* if hw had a problem adding a filter, delete it */
6258 	if (pf->fd_inv > 0) {
6259 		hlist_for_each_entry_safe(filter, node,
6260 					  &pf->fdir_filter_list, fdir_node) {
6261 			if (filter->fd_id == pf->fd_inv) {
6262 				hlist_del(&filter->fdir_node);
6263 				kfree(filter);
6264 				pf->fdir_pf_active_filters--;
6265 			}
6266 		}
6267 	}
6268 }
6269 
6270 #define I40E_MIN_FD_FLUSH_INTERVAL 10
6271 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
6272 /**
6273  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
6274  * @pf: board private structure
6275  **/
6276 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
6277 {
6278 	unsigned long min_flush_time;
6279 	int flush_wait_retry = 50;
6280 	bool disable_atr = false;
6281 	int fd_room;
6282 	int reg;
6283 
6284 	if (!time_after(jiffies, pf->fd_flush_timestamp +
6285 				 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
6286 		return;
6287 
6288 	/* If the flush is happening too quick and we have mostly SB rules we
6289 	 * should not re-enable ATR for some time.
6290 	 */
6291 	min_flush_time = pf->fd_flush_timestamp +
6292 			 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
6293 	fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
6294 
6295 	if (!(time_after(jiffies, min_flush_time)) &&
6296 	    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
6297 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
6298 			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
6299 		disable_atr = true;
6300 	}
6301 
6302 	pf->fd_flush_timestamp = jiffies;
6303 	pf->flags |= I40E_FLAG_FD_ATR_AUTO_DISABLED;
6304 	/* flush all filters */
6305 	wr32(&pf->hw, I40E_PFQF_CTL_1,
6306 	     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6307 	i40e_flush(&pf->hw);
6308 	pf->fd_flush_cnt++;
6309 	pf->fd_add_err = 0;
6310 	do {
6311 		/* Check FD flush status every 5-6msec */
6312 		usleep_range(5000, 6000);
6313 		reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6314 		if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6315 			break;
6316 	} while (flush_wait_retry--);
6317 	if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6318 		dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6319 	} else {
6320 		/* replay sideband filters */
6321 		i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6322 		if (!disable_atr && !pf->fd_tcp4_filter_cnt)
6323 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
6324 		clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
6325 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
6326 			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6327 	}
6328 }
6329 
6330 /**
6331  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6332  * @pf: board private structure
6333  **/
6334 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6335 {
6336 	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6337 }
6338 
6339 /* We can see up to 256 filter programming desc in transit if the filters are
6340  * being applied really fast; before we see the first
6341  * filter miss error on Rx queue 0. Accumulating enough error messages before
6342  * reacting will make sure we don't cause flush too often.
6343  */
6344 #define I40E_MAX_FD_PROGRAM_ERROR 256
6345 
6346 /**
6347  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6348  * @pf: board private structure
6349  **/
6350 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6351 {
6352 
6353 	/* if interface is down do nothing */
6354 	if (test_bit(__I40E_DOWN, pf->state))
6355 		return;
6356 
6357 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
6358 		i40e_fdir_flush_and_replay(pf);
6359 
6360 	i40e_fdir_check_and_reenable(pf);
6361 
6362 }
6363 
6364 /**
6365  * i40e_vsi_link_event - notify VSI of a link event
6366  * @vsi: vsi to be notified
6367  * @link_up: link up or down
6368  **/
6369 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6370 {
6371 	if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
6372 		return;
6373 
6374 	switch (vsi->type) {
6375 	case I40E_VSI_MAIN:
6376 		if (!vsi->netdev || !vsi->netdev_registered)
6377 			break;
6378 
6379 		if (link_up) {
6380 			netif_carrier_on(vsi->netdev);
6381 			netif_tx_wake_all_queues(vsi->netdev);
6382 		} else {
6383 			netif_carrier_off(vsi->netdev);
6384 			netif_tx_stop_all_queues(vsi->netdev);
6385 		}
6386 		break;
6387 
6388 	case I40E_VSI_SRIOV:
6389 	case I40E_VSI_VMDQ2:
6390 	case I40E_VSI_CTRL:
6391 	case I40E_VSI_IWARP:
6392 	case I40E_VSI_MIRROR:
6393 	default:
6394 		/* there is no notification for other VSIs */
6395 		break;
6396 	}
6397 }
6398 
6399 /**
6400  * i40e_veb_link_event - notify elements on the veb of a link event
6401  * @veb: veb to be notified
6402  * @link_up: link up or down
6403  **/
6404 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6405 {
6406 	struct i40e_pf *pf;
6407 	int i;
6408 
6409 	if (!veb || !veb->pf)
6410 		return;
6411 	pf = veb->pf;
6412 
6413 	/* depth first... */
6414 	for (i = 0; i < I40E_MAX_VEB; i++)
6415 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6416 			i40e_veb_link_event(pf->veb[i], link_up);
6417 
6418 	/* ... now the local VSIs */
6419 	for (i = 0; i < pf->num_alloc_vsi; i++)
6420 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6421 			i40e_vsi_link_event(pf->vsi[i], link_up);
6422 }
6423 
6424 /**
6425  * i40e_link_event - Update netif_carrier status
6426  * @pf: board private structure
6427  **/
6428 static void i40e_link_event(struct i40e_pf *pf)
6429 {
6430 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6431 	u8 new_link_speed, old_link_speed;
6432 	i40e_status status;
6433 	bool new_link, old_link;
6434 
6435 	/* save off old link status information */
6436 	pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6437 
6438 	/* set this to force the get_link_status call to refresh state */
6439 	pf->hw.phy.get_link_info = true;
6440 
6441 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6442 
6443 	status = i40e_get_link_status(&pf->hw, &new_link);
6444 
6445 	/* On success, disable temp link polling */
6446 	if (status == I40E_SUCCESS) {
6447 		if (pf->flags & I40E_FLAG_TEMP_LINK_POLLING)
6448 			pf->flags &= ~I40E_FLAG_TEMP_LINK_POLLING;
6449 	} else {
6450 		/* Enable link polling temporarily until i40e_get_link_status
6451 		 * returns I40E_SUCCESS
6452 		 */
6453 		pf->flags |= I40E_FLAG_TEMP_LINK_POLLING;
6454 		dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6455 			status);
6456 		return;
6457 	}
6458 
6459 	old_link_speed = pf->hw.phy.link_info_old.link_speed;
6460 	new_link_speed = pf->hw.phy.link_info.link_speed;
6461 
6462 	if (new_link == old_link &&
6463 	    new_link_speed == old_link_speed &&
6464 	    (test_bit(__I40E_VSI_DOWN, vsi->state) ||
6465 	     new_link == netif_carrier_ok(vsi->netdev)))
6466 		return;
6467 
6468 	if (!test_bit(__I40E_VSI_DOWN, vsi->state))
6469 		i40e_print_link_message(vsi, new_link);
6470 
6471 	/* Notify the base of the switch tree connected to
6472 	 * the link.  Floating VEBs are not notified.
6473 	 */
6474 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6475 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6476 	else
6477 		i40e_vsi_link_event(vsi, new_link);
6478 
6479 	if (pf->vf)
6480 		i40e_vc_notify_link_state(pf);
6481 
6482 	if (pf->flags & I40E_FLAG_PTP)
6483 		i40e_ptp_set_increment(pf);
6484 }
6485 
6486 /**
6487  * i40e_watchdog_subtask - periodic checks not using event driven response
6488  * @pf: board private structure
6489  **/
6490 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6491 {
6492 	int i;
6493 
6494 	/* if interface is down do nothing */
6495 	if (test_bit(__I40E_DOWN, pf->state) ||
6496 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
6497 		return;
6498 
6499 	/* make sure we don't do these things too often */
6500 	if (time_before(jiffies, (pf->service_timer_previous +
6501 				  pf->service_timer_period)))
6502 		return;
6503 	pf->service_timer_previous = jiffies;
6504 
6505 	if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
6506 	    (pf->flags & I40E_FLAG_TEMP_LINK_POLLING))
6507 		i40e_link_event(pf);
6508 
6509 	/* Update the stats for active netdevs so the network stack
6510 	 * can look at updated numbers whenever it cares to
6511 	 */
6512 	for (i = 0; i < pf->num_alloc_vsi; i++)
6513 		if (pf->vsi[i] && pf->vsi[i]->netdev)
6514 			i40e_update_stats(pf->vsi[i]);
6515 
6516 	if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6517 		/* Update the stats for the active switching components */
6518 		for (i = 0; i < I40E_MAX_VEB; i++)
6519 			if (pf->veb[i])
6520 				i40e_update_veb_stats(pf->veb[i]);
6521 	}
6522 
6523 	i40e_ptp_rx_hang(pf);
6524 	i40e_ptp_tx_hang(pf);
6525 }
6526 
6527 /**
6528  * i40e_reset_subtask - Set up for resetting the device and driver
6529  * @pf: board private structure
6530  **/
6531 static void i40e_reset_subtask(struct i40e_pf *pf)
6532 {
6533 	u32 reset_flags = 0;
6534 
6535 	if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
6536 		reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6537 		clear_bit(__I40E_REINIT_REQUESTED, pf->state);
6538 	}
6539 	if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
6540 		reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6541 		clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6542 	}
6543 	if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
6544 		reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6545 		clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
6546 	}
6547 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
6548 		reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6549 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
6550 	}
6551 	if (test_bit(__I40E_DOWN_REQUESTED, pf->state)) {
6552 		reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6553 		clear_bit(__I40E_DOWN_REQUESTED, pf->state);
6554 	}
6555 
6556 	/* If there's a recovery already waiting, it takes
6557 	 * precedence before starting a new reset sequence.
6558 	 */
6559 	if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
6560 		i40e_prep_for_reset(pf, false);
6561 		i40e_reset(pf);
6562 		i40e_rebuild(pf, false, false);
6563 	}
6564 
6565 	/* If we're already down or resetting, just bail */
6566 	if (reset_flags &&
6567 	    !test_bit(__I40E_DOWN, pf->state) &&
6568 	    !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
6569 		i40e_do_reset(pf, reset_flags, false);
6570 	}
6571 }
6572 
6573 /**
6574  * i40e_handle_link_event - Handle link event
6575  * @pf: board private structure
6576  * @e: event info posted on ARQ
6577  **/
6578 static void i40e_handle_link_event(struct i40e_pf *pf,
6579 				   struct i40e_arq_event_info *e)
6580 {
6581 	struct i40e_aqc_get_link_status *status =
6582 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6583 
6584 	/* Do a new status request to re-enable LSE reporting
6585 	 * and load new status information into the hw struct
6586 	 * This completely ignores any state information
6587 	 * in the ARQ event info, instead choosing to always
6588 	 * issue the AQ update link status command.
6589 	 */
6590 	i40e_link_event(pf);
6591 
6592 	/* check for unqualified module, if link is down */
6593 	if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6594 	    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6595 	    (!(status->link_info & I40E_AQ_LINK_UP)))
6596 		dev_err(&pf->pdev->dev,
6597 			"The driver failed to link because an unqualified module was detected.\n");
6598 }
6599 
6600 /**
6601  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6602  * @pf: board private structure
6603  **/
6604 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6605 {
6606 	struct i40e_arq_event_info event;
6607 	struct i40e_hw *hw = &pf->hw;
6608 	u16 pending, i = 0;
6609 	i40e_status ret;
6610 	u16 opcode;
6611 	u32 oldval;
6612 	u32 val;
6613 
6614 	/* Do not run clean AQ when PF reset fails */
6615 	if (test_bit(__I40E_RESET_FAILED, pf->state))
6616 		return;
6617 
6618 	/* check for error indications */
6619 	val = rd32(&pf->hw, pf->hw.aq.arq.len);
6620 	oldval = val;
6621 	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6622 		if (hw->debug_mask & I40E_DEBUG_AQ)
6623 			dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6624 		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6625 	}
6626 	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6627 		if (hw->debug_mask & I40E_DEBUG_AQ)
6628 			dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6629 		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6630 		pf->arq_overflows++;
6631 	}
6632 	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6633 		if (hw->debug_mask & I40E_DEBUG_AQ)
6634 			dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6635 		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6636 	}
6637 	if (oldval != val)
6638 		wr32(&pf->hw, pf->hw.aq.arq.len, val);
6639 
6640 	val = rd32(&pf->hw, pf->hw.aq.asq.len);
6641 	oldval = val;
6642 	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6643 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6644 			dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6645 		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6646 	}
6647 	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6648 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6649 			dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6650 		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6651 	}
6652 	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6653 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6654 			dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6655 		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6656 	}
6657 	if (oldval != val)
6658 		wr32(&pf->hw, pf->hw.aq.asq.len, val);
6659 
6660 	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6661 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6662 	if (!event.msg_buf)
6663 		return;
6664 
6665 	do {
6666 		ret = i40e_clean_arq_element(hw, &event, &pending);
6667 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6668 			break;
6669 		else if (ret) {
6670 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6671 			break;
6672 		}
6673 
6674 		opcode = le16_to_cpu(event.desc.opcode);
6675 		switch (opcode) {
6676 
6677 		case i40e_aqc_opc_get_link_status:
6678 			i40e_handle_link_event(pf, &event);
6679 			break;
6680 		case i40e_aqc_opc_send_msg_to_pf:
6681 			ret = i40e_vc_process_vf_msg(pf,
6682 					le16_to_cpu(event.desc.retval),
6683 					le32_to_cpu(event.desc.cookie_high),
6684 					le32_to_cpu(event.desc.cookie_low),
6685 					event.msg_buf,
6686 					event.msg_len);
6687 			break;
6688 		case i40e_aqc_opc_lldp_update_mib:
6689 			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6690 #ifdef CONFIG_I40E_DCB
6691 			rtnl_lock();
6692 			ret = i40e_handle_lldp_event(pf, &event);
6693 			rtnl_unlock();
6694 #endif /* CONFIG_I40E_DCB */
6695 			break;
6696 		case i40e_aqc_opc_event_lan_overflow:
6697 			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6698 			i40e_handle_lan_overflow_event(pf, &event);
6699 			break;
6700 		case i40e_aqc_opc_send_msg_to_peer:
6701 			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6702 			break;
6703 		case i40e_aqc_opc_nvm_erase:
6704 		case i40e_aqc_opc_nvm_update:
6705 		case i40e_aqc_opc_oem_post_update:
6706 			i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6707 				   "ARQ NVM operation 0x%04x completed\n",
6708 				   opcode);
6709 			break;
6710 		default:
6711 			dev_info(&pf->pdev->dev,
6712 				 "ARQ: Unknown event 0x%04x ignored\n",
6713 				 opcode);
6714 			break;
6715 		}
6716 	} while (i++ < pf->adminq_work_limit);
6717 
6718 	if (i < pf->adminq_work_limit)
6719 		clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
6720 
6721 	/* re-enable Admin queue interrupt cause */
6722 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
6723 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6724 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
6725 	i40e_flush(hw);
6726 
6727 	kfree(event.msg_buf);
6728 }
6729 
6730 /**
6731  * i40e_verify_eeprom - make sure eeprom is good to use
6732  * @pf: board private structure
6733  **/
6734 static void i40e_verify_eeprom(struct i40e_pf *pf)
6735 {
6736 	int err;
6737 
6738 	err = i40e_diag_eeprom_test(&pf->hw);
6739 	if (err) {
6740 		/* retry in case of garbage read */
6741 		err = i40e_diag_eeprom_test(&pf->hw);
6742 		if (err) {
6743 			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6744 				 err);
6745 			set_bit(__I40E_BAD_EEPROM, pf->state);
6746 		}
6747 	}
6748 
6749 	if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
6750 		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6751 		clear_bit(__I40E_BAD_EEPROM, pf->state);
6752 	}
6753 }
6754 
6755 /**
6756  * i40e_enable_pf_switch_lb
6757  * @pf: pointer to the PF structure
6758  *
6759  * enable switch loop back or die - no point in a return value
6760  **/
6761 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6762 {
6763 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6764 	struct i40e_vsi_context ctxt;
6765 	int ret;
6766 
6767 	ctxt.seid = pf->main_vsi_seid;
6768 	ctxt.pf_num = pf->hw.pf_id;
6769 	ctxt.vf_num = 0;
6770 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6771 	if (ret) {
6772 		dev_info(&pf->pdev->dev,
6773 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6774 			 i40e_stat_str(&pf->hw, ret),
6775 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6776 		return;
6777 	}
6778 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6779 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6780 	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6781 
6782 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6783 	if (ret) {
6784 		dev_info(&pf->pdev->dev,
6785 			 "update vsi switch failed, err %s aq_err %s\n",
6786 			 i40e_stat_str(&pf->hw, ret),
6787 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6788 	}
6789 }
6790 
6791 /**
6792  * i40e_disable_pf_switch_lb
6793  * @pf: pointer to the PF structure
6794  *
6795  * disable switch loop back or die - no point in a return value
6796  **/
6797 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6798 {
6799 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6800 	struct i40e_vsi_context ctxt;
6801 	int ret;
6802 
6803 	ctxt.seid = pf->main_vsi_seid;
6804 	ctxt.pf_num = pf->hw.pf_id;
6805 	ctxt.vf_num = 0;
6806 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6807 	if (ret) {
6808 		dev_info(&pf->pdev->dev,
6809 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6810 			 i40e_stat_str(&pf->hw, ret),
6811 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6812 		return;
6813 	}
6814 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6815 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6816 	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6817 
6818 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6819 	if (ret) {
6820 		dev_info(&pf->pdev->dev,
6821 			 "update vsi switch failed, err %s aq_err %s\n",
6822 			 i40e_stat_str(&pf->hw, ret),
6823 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6824 	}
6825 }
6826 
6827 /**
6828  * i40e_config_bridge_mode - Configure the HW bridge mode
6829  * @veb: pointer to the bridge instance
6830  *
6831  * Configure the loop back mode for the LAN VSI that is downlink to the
6832  * specified HW bridge instance. It is expected this function is called
6833  * when a new HW bridge is instantiated.
6834  **/
6835 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6836 {
6837 	struct i40e_pf *pf = veb->pf;
6838 
6839 	if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6840 		dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6841 			 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6842 	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6843 		i40e_disable_pf_switch_lb(pf);
6844 	else
6845 		i40e_enable_pf_switch_lb(pf);
6846 }
6847 
6848 /**
6849  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6850  * @veb: pointer to the VEB instance
6851  *
6852  * This is a recursive function that first builds the attached VSIs then
6853  * recurses in to build the next layer of VEB.  We track the connections
6854  * through our own index numbers because the seid's from the HW could
6855  * change across the reset.
6856  **/
6857 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6858 {
6859 	struct i40e_vsi *ctl_vsi = NULL;
6860 	struct i40e_pf *pf = veb->pf;
6861 	int v, veb_idx;
6862 	int ret;
6863 
6864 	/* build VSI that owns this VEB, temporarily attached to base VEB */
6865 	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6866 		if (pf->vsi[v] &&
6867 		    pf->vsi[v]->veb_idx == veb->idx &&
6868 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6869 			ctl_vsi = pf->vsi[v];
6870 			break;
6871 		}
6872 	}
6873 	if (!ctl_vsi) {
6874 		dev_info(&pf->pdev->dev,
6875 			 "missing owner VSI for veb_idx %d\n", veb->idx);
6876 		ret = -ENOENT;
6877 		goto end_reconstitute;
6878 	}
6879 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
6880 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6881 	ret = i40e_add_vsi(ctl_vsi);
6882 	if (ret) {
6883 		dev_info(&pf->pdev->dev,
6884 			 "rebuild of veb_idx %d owner VSI failed: %d\n",
6885 			 veb->idx, ret);
6886 		goto end_reconstitute;
6887 	}
6888 	i40e_vsi_reset_stats(ctl_vsi);
6889 
6890 	/* create the VEB in the switch and move the VSI onto the VEB */
6891 	ret = i40e_add_veb(veb, ctl_vsi);
6892 	if (ret)
6893 		goto end_reconstitute;
6894 
6895 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6896 		veb->bridge_mode = BRIDGE_MODE_VEB;
6897 	else
6898 		veb->bridge_mode = BRIDGE_MODE_VEPA;
6899 	i40e_config_bridge_mode(veb);
6900 
6901 	/* create the remaining VSIs attached to this VEB */
6902 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6903 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6904 			continue;
6905 
6906 		if (pf->vsi[v]->veb_idx == veb->idx) {
6907 			struct i40e_vsi *vsi = pf->vsi[v];
6908 
6909 			vsi->uplink_seid = veb->seid;
6910 			ret = i40e_add_vsi(vsi);
6911 			if (ret) {
6912 				dev_info(&pf->pdev->dev,
6913 					 "rebuild of vsi_idx %d failed: %d\n",
6914 					 v, ret);
6915 				goto end_reconstitute;
6916 			}
6917 			i40e_vsi_reset_stats(vsi);
6918 		}
6919 	}
6920 
6921 	/* create any VEBs attached to this VEB - RECURSION */
6922 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6923 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6924 			pf->veb[veb_idx]->uplink_seid = veb->seid;
6925 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6926 			if (ret)
6927 				break;
6928 		}
6929 	}
6930 
6931 end_reconstitute:
6932 	return ret;
6933 }
6934 
6935 /**
6936  * i40e_get_capabilities - get info about the HW
6937  * @pf: the PF struct
6938  **/
6939 static int i40e_get_capabilities(struct i40e_pf *pf)
6940 {
6941 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6942 	u16 data_size;
6943 	int buf_len;
6944 	int err;
6945 
6946 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6947 	do {
6948 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
6949 		if (!cap_buf)
6950 			return -ENOMEM;
6951 
6952 		/* this loads the data into the hw struct for us */
6953 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6954 					    &data_size,
6955 					    i40e_aqc_opc_list_func_capabilities,
6956 					    NULL);
6957 		/* data loaded, buffer no longer needed */
6958 		kfree(cap_buf);
6959 
6960 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6961 			/* retry with a larger buffer */
6962 			buf_len = data_size;
6963 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6964 			dev_info(&pf->pdev->dev,
6965 				 "capability discovery failed, err %s aq_err %s\n",
6966 				 i40e_stat_str(&pf->hw, err),
6967 				 i40e_aq_str(&pf->hw,
6968 					     pf->hw.aq.asq_last_status));
6969 			return -ENODEV;
6970 		}
6971 	} while (err);
6972 
6973 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
6974 		dev_info(&pf->pdev->dev,
6975 			 "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",
6976 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6977 			 pf->hw.func_caps.num_msix_vectors,
6978 			 pf->hw.func_caps.num_msix_vectors_vf,
6979 			 pf->hw.func_caps.fd_filters_guaranteed,
6980 			 pf->hw.func_caps.fd_filters_best_effort,
6981 			 pf->hw.func_caps.num_tx_qp,
6982 			 pf->hw.func_caps.num_vsis);
6983 
6984 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6985 		       + pf->hw.func_caps.num_vfs)
6986 	if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6987 		dev_info(&pf->pdev->dev,
6988 			 "got num_vsis %d, setting num_vsis to %d\n",
6989 			 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6990 		pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6991 	}
6992 
6993 	return 0;
6994 }
6995 
6996 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6997 
6998 /**
6999  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
7000  * @pf: board private structure
7001  **/
7002 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
7003 {
7004 	struct i40e_vsi *vsi;
7005 
7006 	/* quick workaround for an NVM issue that leaves a critical register
7007 	 * uninitialized
7008 	 */
7009 	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
7010 		static const u32 hkey[] = {
7011 			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
7012 			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
7013 			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
7014 			0x95b3a76d};
7015 		int i;
7016 
7017 		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
7018 			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
7019 	}
7020 
7021 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
7022 		return;
7023 
7024 	/* find existing VSI and see if it needs configuring */
7025 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
7026 
7027 	/* create a new VSI if none exists */
7028 	if (!vsi) {
7029 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
7030 				     pf->vsi[pf->lan_vsi]->seid, 0);
7031 		if (!vsi) {
7032 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
7033 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7034 			return;
7035 		}
7036 	}
7037 
7038 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
7039 }
7040 
7041 /**
7042  * i40e_fdir_teardown - release the Flow Director resources
7043  * @pf: board private structure
7044  **/
7045 static void i40e_fdir_teardown(struct i40e_pf *pf)
7046 {
7047 	struct i40e_vsi *vsi;
7048 
7049 	i40e_fdir_filter_exit(pf);
7050 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
7051 	if (vsi)
7052 		i40e_vsi_release(vsi);
7053 }
7054 
7055 /**
7056  * i40e_prep_for_reset - prep for the core to reset
7057  * @pf: board private structure
7058  * @lock_acquired: indicates whether or not the lock has been acquired
7059  * before this function was called.
7060  *
7061  * Close up the VFs and other things in prep for PF Reset.
7062   **/
7063 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
7064 {
7065 	struct i40e_hw *hw = &pf->hw;
7066 	i40e_status ret = 0;
7067 	u32 v;
7068 
7069 	clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
7070 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
7071 		return;
7072 	if (i40e_check_asq_alive(&pf->hw))
7073 		i40e_vc_notify_reset(pf);
7074 
7075 	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
7076 
7077 	/* quiesce the VSIs and their queues that are not already DOWN */
7078 	/* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
7079 	if (!lock_acquired)
7080 		rtnl_lock();
7081 	i40e_pf_quiesce_all_vsi(pf);
7082 	if (!lock_acquired)
7083 		rtnl_unlock();
7084 
7085 	for (v = 0; v < pf->num_alloc_vsi; v++) {
7086 		if (pf->vsi[v])
7087 			pf->vsi[v]->seid = 0;
7088 	}
7089 
7090 	i40e_shutdown_adminq(&pf->hw);
7091 
7092 	/* call shutdown HMC */
7093 	if (hw->hmc.hmc_obj) {
7094 		ret = i40e_shutdown_lan_hmc(hw);
7095 		if (ret)
7096 			dev_warn(&pf->pdev->dev,
7097 				 "shutdown_lan_hmc failed: %d\n", ret);
7098 	}
7099 }
7100 
7101 /**
7102  * i40e_send_version - update firmware with driver version
7103  * @pf: PF struct
7104  */
7105 static void i40e_send_version(struct i40e_pf *pf)
7106 {
7107 	struct i40e_driver_version dv;
7108 
7109 	dv.major_version = DRV_VERSION_MAJOR;
7110 	dv.minor_version = DRV_VERSION_MINOR;
7111 	dv.build_version = DRV_VERSION_BUILD;
7112 	dv.subbuild_version = 0;
7113 	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
7114 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
7115 }
7116 
7117 /**
7118  * i40e_get_oem_version - get OEM specific version information
7119  * @hw: pointer to the hardware structure
7120  **/
7121 static void i40e_get_oem_version(struct i40e_hw *hw)
7122 {
7123 	u16 block_offset = 0xffff;
7124 	u16 block_length = 0;
7125 	u16 capabilities = 0;
7126 	u16 gen_snap = 0;
7127 	u16 release = 0;
7128 
7129 #define I40E_SR_NVM_OEM_VERSION_PTR		0x1B
7130 #define I40E_NVM_OEM_LENGTH_OFFSET		0x00
7131 #define I40E_NVM_OEM_CAPABILITIES_OFFSET	0x01
7132 #define I40E_NVM_OEM_GEN_OFFSET			0x02
7133 #define I40E_NVM_OEM_RELEASE_OFFSET		0x03
7134 #define I40E_NVM_OEM_CAPABILITIES_MASK		0x000F
7135 #define I40E_NVM_OEM_LENGTH			3
7136 
7137 	/* Check if pointer to OEM version block is valid. */
7138 	i40e_read_nvm_word(hw, I40E_SR_NVM_OEM_VERSION_PTR, &block_offset);
7139 	if (block_offset == 0xffff)
7140 		return;
7141 
7142 	/* Check if OEM version block has correct length. */
7143 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_LENGTH_OFFSET,
7144 			   &block_length);
7145 	if (block_length < I40E_NVM_OEM_LENGTH)
7146 		return;
7147 
7148 	/* Check if OEM version format is as expected. */
7149 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_CAPABILITIES_OFFSET,
7150 			   &capabilities);
7151 	if ((capabilities & I40E_NVM_OEM_CAPABILITIES_MASK) != 0)
7152 		return;
7153 
7154 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_GEN_OFFSET,
7155 			   &gen_snap);
7156 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_RELEASE_OFFSET,
7157 			   &release);
7158 	hw->nvm.oem_ver = (gen_snap << I40E_OEM_SNAP_SHIFT) | release;
7159 	hw->nvm.eetrack = I40E_OEM_EETRACK_ID;
7160 }
7161 
7162 /**
7163  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
7164  * @pf: board private structure
7165  **/
7166 static int i40e_reset(struct i40e_pf *pf)
7167 {
7168 	struct i40e_hw *hw = &pf->hw;
7169 	i40e_status ret;
7170 
7171 	ret = i40e_pf_reset(hw);
7172 	if (ret) {
7173 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
7174 		set_bit(__I40E_RESET_FAILED, pf->state);
7175 		clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
7176 	} else {
7177 		pf->pfr_count++;
7178 	}
7179 	return ret;
7180 }
7181 
7182 /**
7183  * i40e_rebuild - rebuild using a saved config
7184  * @pf: board private structure
7185  * @reinit: if the Main VSI needs to re-initialized.
7186  * @lock_acquired: indicates whether or not the lock has been acquired
7187  * before this function was called.
7188  **/
7189 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
7190 {
7191 	struct i40e_hw *hw = &pf->hw;
7192 	u8 set_fc_aq_fail = 0;
7193 	i40e_status ret;
7194 	u32 val;
7195 	int v;
7196 
7197 	if (test_bit(__I40E_DOWN, pf->state))
7198 		goto clear_recovery;
7199 	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
7200 
7201 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
7202 	ret = i40e_init_adminq(&pf->hw);
7203 	if (ret) {
7204 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
7205 			 i40e_stat_str(&pf->hw, ret),
7206 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7207 		goto clear_recovery;
7208 	}
7209 	i40e_get_oem_version(&pf->hw);
7210 
7211 	/* re-verify the eeprom if we just had an EMP reset */
7212 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
7213 		i40e_verify_eeprom(pf);
7214 
7215 	i40e_clear_pxe_mode(hw);
7216 	ret = i40e_get_capabilities(pf);
7217 	if (ret)
7218 		goto end_core_reset;
7219 
7220 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
7221 				hw->func_caps.num_rx_qp, 0, 0);
7222 	if (ret) {
7223 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
7224 		goto end_core_reset;
7225 	}
7226 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
7227 	if (ret) {
7228 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
7229 		goto end_core_reset;
7230 	}
7231 
7232 #ifdef CONFIG_I40E_DCB
7233 	ret = i40e_init_pf_dcb(pf);
7234 	if (ret) {
7235 		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
7236 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
7237 		/* Continue without DCB enabled */
7238 	}
7239 #endif /* CONFIG_I40E_DCB */
7240 	/* do basic switch setup */
7241 	if (!lock_acquired)
7242 		rtnl_lock();
7243 	ret = i40e_setup_pf_switch(pf, reinit);
7244 	if (ret)
7245 		goto end_unlock;
7246 
7247 	/* The driver only wants link up/down and module qualification
7248 	 * reports from firmware.  Note the negative logic.
7249 	 */
7250 	ret = i40e_aq_set_phy_int_mask(&pf->hw,
7251 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
7252 					 I40E_AQ_EVENT_MEDIA_NA |
7253 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
7254 	if (ret)
7255 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
7256 			 i40e_stat_str(&pf->hw, ret),
7257 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7258 
7259 	/* make sure our flow control settings are restored */
7260 	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
7261 	if (ret)
7262 		dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
7263 			i40e_stat_str(&pf->hw, ret),
7264 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7265 
7266 	/* Rebuild the VSIs and VEBs that existed before reset.
7267 	 * They are still in our local switch element arrays, so only
7268 	 * need to rebuild the switch model in the HW.
7269 	 *
7270 	 * If there were VEBs but the reconstitution failed, we'll try
7271 	 * try to recover minimal use by getting the basic PF VSI working.
7272 	 */
7273 	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
7274 		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
7275 		/* find the one VEB connected to the MAC, and find orphans */
7276 		for (v = 0; v < I40E_MAX_VEB; v++) {
7277 			if (!pf->veb[v])
7278 				continue;
7279 
7280 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
7281 			    pf->veb[v]->uplink_seid == 0) {
7282 				ret = i40e_reconstitute_veb(pf->veb[v]);
7283 
7284 				if (!ret)
7285 					continue;
7286 
7287 				/* If Main VEB failed, we're in deep doodoo,
7288 				 * so give up rebuilding the switch and set up
7289 				 * for minimal rebuild of PF VSI.
7290 				 * If orphan failed, we'll report the error
7291 				 * but try to keep going.
7292 				 */
7293 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
7294 					dev_info(&pf->pdev->dev,
7295 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
7296 						 ret);
7297 					pf->vsi[pf->lan_vsi]->uplink_seid
7298 								= pf->mac_seid;
7299 					break;
7300 				} else if (pf->veb[v]->uplink_seid == 0) {
7301 					dev_info(&pf->pdev->dev,
7302 						 "rebuild of orphan VEB failed: %d\n",
7303 						 ret);
7304 				}
7305 			}
7306 		}
7307 	}
7308 
7309 	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
7310 		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
7311 		/* no VEB, so rebuild only the Main VSI */
7312 		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
7313 		if (ret) {
7314 			dev_info(&pf->pdev->dev,
7315 				 "rebuild of Main VSI failed: %d\n", ret);
7316 			goto end_unlock;
7317 		}
7318 	}
7319 
7320 	/* Reconfigure hardware for allowing smaller MSS in the case
7321 	 * of TSO, so that we avoid the MDD being fired and causing
7322 	 * a reset in the case of small MSS+TSO.
7323 	 */
7324 #define I40E_REG_MSS          0x000E64DC
7325 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
7326 #define I40E_64BYTE_MSS       0x400000
7327 	val = rd32(hw, I40E_REG_MSS);
7328 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
7329 		val &= ~I40E_REG_MSS_MIN_MASK;
7330 		val |= I40E_64BYTE_MSS;
7331 		wr32(hw, I40E_REG_MSS, val);
7332 	}
7333 
7334 	if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
7335 		msleep(75);
7336 		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
7337 		if (ret)
7338 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
7339 				 i40e_stat_str(&pf->hw, ret),
7340 				 i40e_aq_str(&pf->hw,
7341 					     pf->hw.aq.asq_last_status));
7342 	}
7343 	/* reinit the misc interrupt */
7344 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7345 		ret = i40e_setup_misc_vector(pf);
7346 
7347 	/* Add a filter to drop all Flow control frames from any VSI from being
7348 	 * transmitted. By doing so we stop a malicious VF from sending out
7349 	 * PAUSE or PFC frames and potentially controlling traffic for other
7350 	 * PF/VF VSIs.
7351 	 * The FW can still send Flow control frames if enabled.
7352 	 */
7353 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
7354 						       pf->main_vsi_seid);
7355 
7356 	/* restart the VSIs that were rebuilt and running before the reset */
7357 	i40e_pf_unquiesce_all_vsi(pf);
7358 
7359 	/* Release the RTNL lock before we start resetting VFs */
7360 	if (!lock_acquired)
7361 		rtnl_unlock();
7362 
7363 	i40e_reset_all_vfs(pf, true);
7364 
7365 	/* tell the firmware that we're starting */
7366 	i40e_send_version(pf);
7367 
7368 	/* We've already released the lock, so don't do it again */
7369 	goto end_core_reset;
7370 
7371 end_unlock:
7372 	if (!lock_acquired)
7373 		rtnl_unlock();
7374 end_core_reset:
7375 	clear_bit(__I40E_RESET_FAILED, pf->state);
7376 clear_recovery:
7377 	clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
7378 }
7379 
7380 /**
7381  * i40e_reset_and_rebuild - reset and rebuild using a saved config
7382  * @pf: board private structure
7383  * @reinit: if the Main VSI needs to re-initialized.
7384  * @lock_acquired: indicates whether or not the lock has been acquired
7385  * before this function was called.
7386  **/
7387 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
7388 				   bool lock_acquired)
7389 {
7390 	int ret;
7391 	/* Now we wait for GRST to settle out.
7392 	 * We don't have to delete the VEBs or VSIs from the hw switch
7393 	 * because the reset will make them disappear.
7394 	 */
7395 	ret = i40e_reset(pf);
7396 	if (!ret)
7397 		i40e_rebuild(pf, reinit, lock_acquired);
7398 }
7399 
7400 /**
7401  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7402  * @pf: board private structure
7403  *
7404  * Close up the VFs and other things in prep for a Core Reset,
7405  * then get ready to rebuild the world.
7406  * @lock_acquired: indicates whether or not the lock has been acquired
7407  * before this function was called.
7408  **/
7409 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
7410 {
7411 	i40e_prep_for_reset(pf, lock_acquired);
7412 	i40e_reset_and_rebuild(pf, false, lock_acquired);
7413 }
7414 
7415 /**
7416  * i40e_handle_mdd_event
7417  * @pf: pointer to the PF structure
7418  *
7419  * Called from the MDD irq handler to identify possibly malicious vfs
7420  **/
7421 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7422 {
7423 	struct i40e_hw *hw = &pf->hw;
7424 	bool mdd_detected = false;
7425 	bool pf_mdd_detected = false;
7426 	struct i40e_vf *vf;
7427 	u32 reg;
7428 	int i;
7429 
7430 	if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
7431 		return;
7432 
7433 	/* find what triggered the MDD event */
7434 	reg = rd32(hw, I40E_GL_MDET_TX);
7435 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7436 		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7437 				I40E_GL_MDET_TX_PF_NUM_SHIFT;
7438 		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7439 				I40E_GL_MDET_TX_VF_NUM_SHIFT;
7440 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7441 				I40E_GL_MDET_TX_EVENT_SHIFT;
7442 		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7443 				I40E_GL_MDET_TX_QUEUE_SHIFT) -
7444 				pf->hw.func_caps.base_queue;
7445 		if (netif_msg_tx_err(pf))
7446 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7447 				 event, queue, pf_num, vf_num);
7448 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7449 		mdd_detected = true;
7450 	}
7451 	reg = rd32(hw, I40E_GL_MDET_RX);
7452 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7453 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7454 				I40E_GL_MDET_RX_FUNCTION_SHIFT;
7455 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7456 				I40E_GL_MDET_RX_EVENT_SHIFT;
7457 		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7458 				I40E_GL_MDET_RX_QUEUE_SHIFT) -
7459 				pf->hw.func_caps.base_queue;
7460 		if (netif_msg_rx_err(pf))
7461 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7462 				 event, queue, func);
7463 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7464 		mdd_detected = true;
7465 	}
7466 
7467 	if (mdd_detected) {
7468 		reg = rd32(hw, I40E_PF_MDET_TX);
7469 		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7470 			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7471 			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7472 			pf_mdd_detected = true;
7473 		}
7474 		reg = rd32(hw, I40E_PF_MDET_RX);
7475 		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7476 			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7477 			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7478 			pf_mdd_detected = true;
7479 		}
7480 		/* Queue belongs to the PF, initiate a reset */
7481 		if (pf_mdd_detected) {
7482 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
7483 			i40e_service_event_schedule(pf);
7484 		}
7485 	}
7486 
7487 	/* see if one of the VFs needs its hand slapped */
7488 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7489 		vf = &(pf->vf[i]);
7490 		reg = rd32(hw, I40E_VP_MDET_TX(i));
7491 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7492 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7493 			vf->num_mdd_events++;
7494 			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7495 				 i);
7496 		}
7497 
7498 		reg = rd32(hw, I40E_VP_MDET_RX(i));
7499 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7500 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7501 			vf->num_mdd_events++;
7502 			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7503 				 i);
7504 		}
7505 
7506 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7507 			dev_info(&pf->pdev->dev,
7508 				 "Too many MDD events on VF %d, disabled\n", i);
7509 			dev_info(&pf->pdev->dev,
7510 				 "Use PF Control I/F to re-enable the VF\n");
7511 			set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
7512 		}
7513 	}
7514 
7515 	/* re-enable mdd interrupt cause */
7516 	clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
7517 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7518 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7519 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7520 	i40e_flush(hw);
7521 }
7522 
7523 /**
7524  * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
7525  * @pf: board private structure
7526  **/
7527 static void i40e_sync_udp_filters(struct i40e_pf *pf)
7528 {
7529 	int i;
7530 
7531 	/* loop through and set pending bit for all active UDP filters */
7532 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7533 		if (pf->udp_ports[i].port)
7534 			pf->pending_udp_bitmap |= BIT_ULL(i);
7535 	}
7536 
7537 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
7538 }
7539 
7540 /**
7541  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7542  * @pf: board private structure
7543  **/
7544 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7545 {
7546 	struct i40e_hw *hw = &pf->hw;
7547 	i40e_status ret;
7548 	u16 port;
7549 	int i;
7550 
7551 	if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7552 		return;
7553 
7554 	pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7555 
7556 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7557 		if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7558 			pf->pending_udp_bitmap &= ~BIT_ULL(i);
7559 			port = pf->udp_ports[i].port;
7560 			if (port)
7561 				ret = i40e_aq_add_udp_tunnel(hw, port,
7562 							pf->udp_ports[i].type,
7563 							NULL, NULL);
7564 			else
7565 				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7566 
7567 			if (ret) {
7568 				dev_dbg(&pf->pdev->dev,
7569 					"%s %s port %d, index %d failed, err %s aq_err %s\n",
7570 					pf->udp_ports[i].type ? "vxlan" : "geneve",
7571 					port ? "add" : "delete",
7572 					port, i,
7573 					i40e_stat_str(&pf->hw, ret),
7574 					i40e_aq_str(&pf->hw,
7575 						    pf->hw.aq.asq_last_status));
7576 				pf->udp_ports[i].port = 0;
7577 			}
7578 		}
7579 	}
7580 }
7581 
7582 /**
7583  * i40e_service_task - Run the driver's async subtasks
7584  * @work: pointer to work_struct containing our data
7585  **/
7586 static void i40e_service_task(struct work_struct *work)
7587 {
7588 	struct i40e_pf *pf = container_of(work,
7589 					  struct i40e_pf,
7590 					  service_task);
7591 	unsigned long start_time = jiffies;
7592 
7593 	/* don't bother with service tasks if a reset is in progress */
7594 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
7595 		return;
7596 
7597 	if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
7598 		return;
7599 
7600 	i40e_detect_recover_hung(pf);
7601 	i40e_sync_filters_subtask(pf);
7602 	i40e_reset_subtask(pf);
7603 	i40e_handle_mdd_event(pf);
7604 	i40e_vc_process_vflr_event(pf);
7605 	i40e_watchdog_subtask(pf);
7606 	i40e_fdir_reinit_subtask(pf);
7607 	if (pf->flags & I40E_FLAG_CLIENT_RESET) {
7608 		/* Client subtask will reopen next time through. */
7609 		i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], true);
7610 		pf->flags &= ~I40E_FLAG_CLIENT_RESET;
7611 	} else {
7612 		i40e_client_subtask(pf);
7613 		if (pf->flags & I40E_FLAG_CLIENT_L2_CHANGE) {
7614 			i40e_notify_client_of_l2_param_changes(
7615 							pf->vsi[pf->lan_vsi]);
7616 			pf->flags &= ~I40E_FLAG_CLIENT_L2_CHANGE;
7617 		}
7618 	}
7619 	i40e_sync_filters_subtask(pf);
7620 	i40e_sync_udp_filters_subtask(pf);
7621 	i40e_clean_adminq_subtask(pf);
7622 
7623 	/* flush memory to make sure state is correct before next watchdog */
7624 	smp_mb__before_atomic();
7625 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
7626 
7627 	/* If the tasks have taken longer than one timer cycle or there
7628 	 * is more work to be done, reschedule the service task now
7629 	 * rather than wait for the timer to tick again.
7630 	 */
7631 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7632 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)		 ||
7633 	    test_bit(__I40E_MDD_EVENT_PENDING, pf->state)		 ||
7634 	    test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
7635 		i40e_service_event_schedule(pf);
7636 }
7637 
7638 /**
7639  * i40e_service_timer - timer callback
7640  * @data: pointer to PF struct
7641  **/
7642 static void i40e_service_timer(unsigned long data)
7643 {
7644 	struct i40e_pf *pf = (struct i40e_pf *)data;
7645 
7646 	mod_timer(&pf->service_timer,
7647 		  round_jiffies(jiffies + pf->service_timer_period));
7648 	i40e_service_event_schedule(pf);
7649 }
7650 
7651 /**
7652  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7653  * @vsi: the VSI being configured
7654  **/
7655 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7656 {
7657 	struct i40e_pf *pf = vsi->back;
7658 
7659 	switch (vsi->type) {
7660 	case I40E_VSI_MAIN:
7661 		vsi->alloc_queue_pairs = pf->num_lan_qps;
7662 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7663 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7664 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7665 			vsi->num_q_vectors = pf->num_lan_msix;
7666 		else
7667 			vsi->num_q_vectors = 1;
7668 
7669 		break;
7670 
7671 	case I40E_VSI_FDIR:
7672 		vsi->alloc_queue_pairs = 1;
7673 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7674 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7675 		vsi->num_q_vectors = pf->num_fdsb_msix;
7676 		break;
7677 
7678 	case I40E_VSI_VMDQ2:
7679 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7680 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7681 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7682 		vsi->num_q_vectors = pf->num_vmdq_msix;
7683 		break;
7684 
7685 	case I40E_VSI_SRIOV:
7686 		vsi->alloc_queue_pairs = pf->num_vf_qps;
7687 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7688 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7689 		break;
7690 
7691 	default:
7692 		WARN_ON(1);
7693 		return -ENODATA;
7694 	}
7695 
7696 	return 0;
7697 }
7698 
7699 /**
7700  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7701  * @type: VSI pointer
7702  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7703  *
7704  * On error: returns error code (negative)
7705  * On success: returns 0
7706  **/
7707 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7708 {
7709 	struct i40e_ring **next_rings;
7710 	int size;
7711 	int ret = 0;
7712 
7713 	/* allocate memory for both Tx, XDP Tx and Rx ring pointers */
7714 	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs *
7715 	       (i40e_enabled_xdp_vsi(vsi) ? 3 : 2);
7716 	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7717 	if (!vsi->tx_rings)
7718 		return -ENOMEM;
7719 	next_rings = vsi->tx_rings + vsi->alloc_queue_pairs;
7720 	if (i40e_enabled_xdp_vsi(vsi)) {
7721 		vsi->xdp_rings = next_rings;
7722 		next_rings += vsi->alloc_queue_pairs;
7723 	}
7724 	vsi->rx_rings = next_rings;
7725 
7726 	if (alloc_qvectors) {
7727 		/* allocate memory for q_vector pointers */
7728 		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7729 		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7730 		if (!vsi->q_vectors) {
7731 			ret = -ENOMEM;
7732 			goto err_vectors;
7733 		}
7734 	}
7735 	return ret;
7736 
7737 err_vectors:
7738 	kfree(vsi->tx_rings);
7739 	return ret;
7740 }
7741 
7742 /**
7743  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7744  * @pf: board private structure
7745  * @type: type of VSI
7746  *
7747  * On error: returns error code (negative)
7748  * On success: returns vsi index in PF (positive)
7749  **/
7750 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7751 {
7752 	int ret = -ENODEV;
7753 	struct i40e_vsi *vsi;
7754 	int vsi_idx;
7755 	int i;
7756 
7757 	/* Need to protect the allocation of the VSIs at the PF level */
7758 	mutex_lock(&pf->switch_mutex);
7759 
7760 	/* VSI list may be fragmented if VSI creation/destruction has
7761 	 * been happening.  We can afford to do a quick scan to look
7762 	 * for any free VSIs in the list.
7763 	 *
7764 	 * find next empty vsi slot, looping back around if necessary
7765 	 */
7766 	i = pf->next_vsi;
7767 	while (i < pf->num_alloc_vsi && pf->vsi[i])
7768 		i++;
7769 	if (i >= pf->num_alloc_vsi) {
7770 		i = 0;
7771 		while (i < pf->next_vsi && pf->vsi[i])
7772 			i++;
7773 	}
7774 
7775 	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7776 		vsi_idx = i;             /* Found one! */
7777 	} else {
7778 		ret = -ENODEV;
7779 		goto unlock_pf;  /* out of VSI slots! */
7780 	}
7781 	pf->next_vsi = ++i;
7782 
7783 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7784 	if (!vsi) {
7785 		ret = -ENOMEM;
7786 		goto unlock_pf;
7787 	}
7788 	vsi->type = type;
7789 	vsi->back = pf;
7790 	set_bit(__I40E_VSI_DOWN, vsi->state);
7791 	vsi->flags = 0;
7792 	vsi->idx = vsi_idx;
7793 	vsi->int_rate_limit = 0;
7794 	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7795 				pf->rss_table_size : 64;
7796 	vsi->netdev_registered = false;
7797 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7798 	hash_init(vsi->mac_filter_hash);
7799 	vsi->irqs_ready = false;
7800 
7801 	ret = i40e_set_num_rings_in_vsi(vsi);
7802 	if (ret)
7803 		goto err_rings;
7804 
7805 	ret = i40e_vsi_alloc_arrays(vsi, true);
7806 	if (ret)
7807 		goto err_rings;
7808 
7809 	/* Setup default MSIX irq handler for VSI */
7810 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7811 
7812 	/* Initialize VSI lock */
7813 	spin_lock_init(&vsi->mac_filter_hash_lock);
7814 	pf->vsi[vsi_idx] = vsi;
7815 	ret = vsi_idx;
7816 	goto unlock_pf;
7817 
7818 err_rings:
7819 	pf->next_vsi = i - 1;
7820 	kfree(vsi);
7821 unlock_pf:
7822 	mutex_unlock(&pf->switch_mutex);
7823 	return ret;
7824 }
7825 
7826 /**
7827  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7828  * @type: VSI pointer
7829  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7830  *
7831  * On error: returns error code (negative)
7832  * On success: returns 0
7833  **/
7834 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7835 {
7836 	/* free the ring and vector containers */
7837 	if (free_qvectors) {
7838 		kfree(vsi->q_vectors);
7839 		vsi->q_vectors = NULL;
7840 	}
7841 	kfree(vsi->tx_rings);
7842 	vsi->tx_rings = NULL;
7843 	vsi->rx_rings = NULL;
7844 	vsi->xdp_rings = NULL;
7845 }
7846 
7847 /**
7848  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7849  * and lookup table
7850  * @vsi: Pointer to VSI structure
7851  */
7852 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7853 {
7854 	if (!vsi)
7855 		return;
7856 
7857 	kfree(vsi->rss_hkey_user);
7858 	vsi->rss_hkey_user = NULL;
7859 
7860 	kfree(vsi->rss_lut_user);
7861 	vsi->rss_lut_user = NULL;
7862 }
7863 
7864 /**
7865  * i40e_vsi_clear - Deallocate the VSI provided
7866  * @vsi: the VSI being un-configured
7867  **/
7868 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7869 {
7870 	struct i40e_pf *pf;
7871 
7872 	if (!vsi)
7873 		return 0;
7874 
7875 	if (!vsi->back)
7876 		goto free_vsi;
7877 	pf = vsi->back;
7878 
7879 	mutex_lock(&pf->switch_mutex);
7880 	if (!pf->vsi[vsi->idx]) {
7881 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7882 			vsi->idx, vsi->idx, vsi, vsi->type);
7883 		goto unlock_vsi;
7884 	}
7885 
7886 	if (pf->vsi[vsi->idx] != vsi) {
7887 		dev_err(&pf->pdev->dev,
7888 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7889 			pf->vsi[vsi->idx]->idx,
7890 			pf->vsi[vsi->idx],
7891 			pf->vsi[vsi->idx]->type,
7892 			vsi->idx, vsi, vsi->type);
7893 		goto unlock_vsi;
7894 	}
7895 
7896 	/* updates the PF for this cleared vsi */
7897 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7898 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7899 
7900 	i40e_vsi_free_arrays(vsi, true);
7901 	i40e_clear_rss_config_user(vsi);
7902 
7903 	pf->vsi[vsi->idx] = NULL;
7904 	if (vsi->idx < pf->next_vsi)
7905 		pf->next_vsi = vsi->idx;
7906 
7907 unlock_vsi:
7908 	mutex_unlock(&pf->switch_mutex);
7909 free_vsi:
7910 	kfree(vsi);
7911 
7912 	return 0;
7913 }
7914 
7915 /**
7916  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7917  * @vsi: the VSI being cleaned
7918  **/
7919 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7920 {
7921 	int i;
7922 
7923 	if (vsi->tx_rings && vsi->tx_rings[0]) {
7924 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7925 			kfree_rcu(vsi->tx_rings[i], rcu);
7926 			vsi->tx_rings[i] = NULL;
7927 			vsi->rx_rings[i] = NULL;
7928 			if (vsi->xdp_rings)
7929 				vsi->xdp_rings[i] = NULL;
7930 		}
7931 	}
7932 }
7933 
7934 /**
7935  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7936  * @vsi: the VSI being configured
7937  **/
7938 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7939 {
7940 	int i, qpv = i40e_enabled_xdp_vsi(vsi) ? 3 : 2;
7941 	struct i40e_pf *pf = vsi->back;
7942 	struct i40e_ring *ring;
7943 
7944 	/* Set basic values in the rings to be used later during open() */
7945 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7946 		/* allocate space for both Tx and Rx in one shot */
7947 		ring = kcalloc(qpv, sizeof(struct i40e_ring), GFP_KERNEL);
7948 		if (!ring)
7949 			goto err_out;
7950 
7951 		ring->queue_index = i;
7952 		ring->reg_idx = vsi->base_queue + i;
7953 		ring->ring_active = false;
7954 		ring->vsi = vsi;
7955 		ring->netdev = vsi->netdev;
7956 		ring->dev = &pf->pdev->dev;
7957 		ring->count = vsi->num_desc;
7958 		ring->size = 0;
7959 		ring->dcb_tc = 0;
7960 		if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7961 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7962 		ring->tx_itr_setting = pf->tx_itr_default;
7963 		vsi->tx_rings[i] = ring++;
7964 
7965 		if (!i40e_enabled_xdp_vsi(vsi))
7966 			goto setup_rx;
7967 
7968 		ring->queue_index = vsi->alloc_queue_pairs + i;
7969 		ring->reg_idx = vsi->base_queue + ring->queue_index;
7970 		ring->ring_active = false;
7971 		ring->vsi = vsi;
7972 		ring->netdev = NULL;
7973 		ring->dev = &pf->pdev->dev;
7974 		ring->count = vsi->num_desc;
7975 		ring->size = 0;
7976 		ring->dcb_tc = 0;
7977 		if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7978 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7979 		set_ring_xdp(ring);
7980 		ring->tx_itr_setting = pf->tx_itr_default;
7981 		vsi->xdp_rings[i] = ring++;
7982 
7983 setup_rx:
7984 		ring->queue_index = i;
7985 		ring->reg_idx = vsi->base_queue + i;
7986 		ring->ring_active = false;
7987 		ring->vsi = vsi;
7988 		ring->netdev = vsi->netdev;
7989 		ring->dev = &pf->pdev->dev;
7990 		ring->count = vsi->num_desc;
7991 		ring->size = 0;
7992 		ring->dcb_tc = 0;
7993 		ring->rx_itr_setting = pf->rx_itr_default;
7994 		vsi->rx_rings[i] = ring;
7995 	}
7996 
7997 	return 0;
7998 
7999 err_out:
8000 	i40e_vsi_clear_rings(vsi);
8001 	return -ENOMEM;
8002 }
8003 
8004 /**
8005  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
8006  * @pf: board private structure
8007  * @vectors: the number of MSI-X vectors to request
8008  *
8009  * Returns the number of vectors reserved, or error
8010  **/
8011 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
8012 {
8013 	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
8014 					I40E_MIN_MSIX, vectors);
8015 	if (vectors < 0) {
8016 		dev_info(&pf->pdev->dev,
8017 			 "MSI-X vector reservation failed: %d\n", vectors);
8018 		vectors = 0;
8019 	}
8020 
8021 	return vectors;
8022 }
8023 
8024 /**
8025  * i40e_init_msix - Setup the MSIX capability
8026  * @pf: board private structure
8027  *
8028  * Work with the OS to set up the MSIX vectors needed.
8029  *
8030  * Returns the number of vectors reserved or negative on failure
8031  **/
8032 static int i40e_init_msix(struct i40e_pf *pf)
8033 {
8034 	struct i40e_hw *hw = &pf->hw;
8035 	int cpus, extra_vectors;
8036 	int vectors_left;
8037 	int v_budget, i;
8038 	int v_actual;
8039 	int iwarp_requested = 0;
8040 
8041 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
8042 		return -ENODEV;
8043 
8044 	/* The number of vectors we'll request will be comprised of:
8045 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
8046 	 *   - The number of LAN queue pairs
8047 	 *	- Queues being used for RSS.
8048 	 *		We don't need as many as max_rss_size vectors.
8049 	 *		use rss_size instead in the calculation since that
8050 	 *		is governed by number of cpus in the system.
8051 	 *	- assumes symmetric Tx/Rx pairing
8052 	 *   - The number of VMDq pairs
8053 	 *   - The CPU count within the NUMA node if iWARP is enabled
8054 	 * Once we count this up, try the request.
8055 	 *
8056 	 * If we can't get what we want, we'll simplify to nearly nothing
8057 	 * and try again.  If that still fails, we punt.
8058 	 */
8059 	vectors_left = hw->func_caps.num_msix_vectors;
8060 	v_budget = 0;
8061 
8062 	/* reserve one vector for miscellaneous handler */
8063 	if (vectors_left) {
8064 		v_budget++;
8065 		vectors_left--;
8066 	}
8067 
8068 	/* reserve some vectors for the main PF traffic queues. Initially we
8069 	 * only reserve at most 50% of the available vectors, in the case that
8070 	 * the number of online CPUs is large. This ensures that we can enable
8071 	 * extra features as well. Once we've enabled the other features, we
8072 	 * will use any remaining vectors to reach as close as we can to the
8073 	 * number of online CPUs.
8074 	 */
8075 	cpus = num_online_cpus();
8076 	pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
8077 	vectors_left -= pf->num_lan_msix;
8078 
8079 	/* reserve one vector for sideband flow director */
8080 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8081 		if (vectors_left) {
8082 			pf->num_fdsb_msix = 1;
8083 			v_budget++;
8084 			vectors_left--;
8085 		} else {
8086 			pf->num_fdsb_msix = 0;
8087 		}
8088 	}
8089 
8090 	/* can we reserve enough for iWARP? */
8091 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
8092 		iwarp_requested = pf->num_iwarp_msix;
8093 
8094 		if (!vectors_left)
8095 			pf->num_iwarp_msix = 0;
8096 		else if (vectors_left < pf->num_iwarp_msix)
8097 			pf->num_iwarp_msix = 1;
8098 		v_budget += pf->num_iwarp_msix;
8099 		vectors_left -= pf->num_iwarp_msix;
8100 	}
8101 
8102 	/* any vectors left over go for VMDq support */
8103 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
8104 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
8105 		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
8106 
8107 		if (!vectors_left) {
8108 			pf->num_vmdq_msix = 0;
8109 			pf->num_vmdq_qps = 0;
8110 		} else {
8111 			/* if we're short on vectors for what's desired, we limit
8112 			 * the queues per vmdq.  If this is still more than are
8113 			 * available, the user will need to change the number of
8114 			 * queues/vectors used by the PF later with the ethtool
8115 			 * channels command
8116 			 */
8117 			if (vmdq_vecs < vmdq_vecs_wanted)
8118 				pf->num_vmdq_qps = 1;
8119 			pf->num_vmdq_msix = pf->num_vmdq_qps;
8120 
8121 			v_budget += vmdq_vecs;
8122 			vectors_left -= vmdq_vecs;
8123 		}
8124 	}
8125 
8126 	/* On systems with a large number of SMP cores, we previously limited
8127 	 * the number of vectors for num_lan_msix to be at most 50% of the
8128 	 * available vectors, to allow for other features. Now, we add back
8129 	 * the remaining vectors. However, we ensure that the total
8130 	 * num_lan_msix will not exceed num_online_cpus(). To do this, we
8131 	 * calculate the number of vectors we can add without going over the
8132 	 * cap of CPUs. For systems with a small number of CPUs this will be
8133 	 * zero.
8134 	 */
8135 	extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
8136 	pf->num_lan_msix += extra_vectors;
8137 	vectors_left -= extra_vectors;
8138 
8139 	WARN(vectors_left < 0,
8140 	     "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
8141 
8142 	v_budget += pf->num_lan_msix;
8143 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
8144 				   GFP_KERNEL);
8145 	if (!pf->msix_entries)
8146 		return -ENOMEM;
8147 
8148 	for (i = 0; i < v_budget; i++)
8149 		pf->msix_entries[i].entry = i;
8150 	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
8151 
8152 	if (v_actual < I40E_MIN_MSIX) {
8153 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
8154 		kfree(pf->msix_entries);
8155 		pf->msix_entries = NULL;
8156 		pci_disable_msix(pf->pdev);
8157 		return -ENODEV;
8158 
8159 	} else if (v_actual == I40E_MIN_MSIX) {
8160 		/* Adjust for minimal MSIX use */
8161 		pf->num_vmdq_vsis = 0;
8162 		pf->num_vmdq_qps = 0;
8163 		pf->num_lan_qps = 1;
8164 		pf->num_lan_msix = 1;
8165 
8166 	} else if (!vectors_left) {
8167 		/* If we have limited resources, we will start with no vectors
8168 		 * for the special features and then allocate vectors to some
8169 		 * of these features based on the policy and at the end disable
8170 		 * the features that did not get any vectors.
8171 		 */
8172 		int vec;
8173 
8174 		dev_info(&pf->pdev->dev,
8175 			 "MSI-X vector limit reached, attempting to redistribute vectors\n");
8176 		/* reserve the misc vector */
8177 		vec = v_actual - 1;
8178 
8179 		/* Scale vector usage down */
8180 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
8181 		pf->num_vmdq_vsis = 1;
8182 		pf->num_vmdq_qps = 1;
8183 
8184 		/* partition out the remaining vectors */
8185 		switch (vec) {
8186 		case 2:
8187 			pf->num_lan_msix = 1;
8188 			break;
8189 		case 3:
8190 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
8191 				pf->num_lan_msix = 1;
8192 				pf->num_iwarp_msix = 1;
8193 			} else {
8194 				pf->num_lan_msix = 2;
8195 			}
8196 			break;
8197 		default:
8198 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
8199 				pf->num_iwarp_msix = min_t(int, (vec / 3),
8200 						 iwarp_requested);
8201 				pf->num_vmdq_vsis = min_t(int, (vec / 3),
8202 						  I40E_DEFAULT_NUM_VMDQ_VSI);
8203 			} else {
8204 				pf->num_vmdq_vsis = min_t(int, (vec / 2),
8205 						  I40E_DEFAULT_NUM_VMDQ_VSI);
8206 			}
8207 			if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8208 				pf->num_fdsb_msix = 1;
8209 				vec--;
8210 			}
8211 			pf->num_lan_msix = min_t(int,
8212 			       (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
8213 							      pf->num_lan_msix);
8214 			pf->num_lan_qps = pf->num_lan_msix;
8215 			break;
8216 		}
8217 	}
8218 
8219 	if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
8220 	    (pf->num_fdsb_msix == 0)) {
8221 		dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
8222 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8223 	}
8224 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
8225 	    (pf->num_vmdq_msix == 0)) {
8226 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
8227 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
8228 	}
8229 
8230 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
8231 	    (pf->num_iwarp_msix == 0)) {
8232 		dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
8233 		pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
8234 	}
8235 	i40e_debug(&pf->hw, I40E_DEBUG_INIT,
8236 		   "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
8237 		   pf->num_lan_msix,
8238 		   pf->num_vmdq_msix * pf->num_vmdq_vsis,
8239 		   pf->num_fdsb_msix,
8240 		   pf->num_iwarp_msix);
8241 
8242 	return v_actual;
8243 }
8244 
8245 /**
8246  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
8247  * @vsi: the VSI being configured
8248  * @v_idx: index of the vector in the vsi struct
8249  * @cpu: cpu to be used on affinity_mask
8250  *
8251  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
8252  **/
8253 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
8254 {
8255 	struct i40e_q_vector *q_vector;
8256 
8257 	/* allocate q_vector */
8258 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
8259 	if (!q_vector)
8260 		return -ENOMEM;
8261 
8262 	q_vector->vsi = vsi;
8263 	q_vector->v_idx = v_idx;
8264 	cpumask_set_cpu(cpu, &q_vector->affinity_mask);
8265 
8266 	if (vsi->netdev)
8267 		netif_napi_add(vsi->netdev, &q_vector->napi,
8268 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
8269 
8270 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
8271 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
8272 
8273 	/* tie q_vector and vsi together */
8274 	vsi->q_vectors[v_idx] = q_vector;
8275 
8276 	return 0;
8277 }
8278 
8279 /**
8280  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
8281  * @vsi: the VSI being configured
8282  *
8283  * We allocate one q_vector per queue interrupt.  If allocation fails we
8284  * return -ENOMEM.
8285  **/
8286 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
8287 {
8288 	struct i40e_pf *pf = vsi->back;
8289 	int err, v_idx, num_q_vectors, current_cpu;
8290 
8291 	/* if not MSIX, give the one vector only to the LAN VSI */
8292 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
8293 		num_q_vectors = vsi->num_q_vectors;
8294 	else if (vsi == pf->vsi[pf->lan_vsi])
8295 		num_q_vectors = 1;
8296 	else
8297 		return -EINVAL;
8298 
8299 	current_cpu = cpumask_first(cpu_online_mask);
8300 
8301 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
8302 		err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
8303 		if (err)
8304 			goto err_out;
8305 		current_cpu = cpumask_next(current_cpu, cpu_online_mask);
8306 		if (unlikely(current_cpu >= nr_cpu_ids))
8307 			current_cpu = cpumask_first(cpu_online_mask);
8308 	}
8309 
8310 	return 0;
8311 
8312 err_out:
8313 	while (v_idx--)
8314 		i40e_free_q_vector(vsi, v_idx);
8315 
8316 	return err;
8317 }
8318 
8319 /**
8320  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
8321  * @pf: board private structure to initialize
8322  **/
8323 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
8324 {
8325 	int vectors = 0;
8326 	ssize_t size;
8327 
8328 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
8329 		vectors = i40e_init_msix(pf);
8330 		if (vectors < 0) {
8331 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
8332 				       I40E_FLAG_IWARP_ENABLED	|
8333 				       I40E_FLAG_RSS_ENABLED	|
8334 				       I40E_FLAG_DCB_CAPABLE	|
8335 				       I40E_FLAG_DCB_ENABLED	|
8336 				       I40E_FLAG_SRIOV_ENABLED	|
8337 				       I40E_FLAG_FD_SB_ENABLED	|
8338 				       I40E_FLAG_FD_ATR_ENABLED	|
8339 				       I40E_FLAG_VMDQ_ENABLED);
8340 
8341 			/* rework the queue expectations without MSIX */
8342 			i40e_determine_queue_usage(pf);
8343 		}
8344 	}
8345 
8346 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
8347 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
8348 		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
8349 		vectors = pci_enable_msi(pf->pdev);
8350 		if (vectors < 0) {
8351 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
8352 				 vectors);
8353 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
8354 		}
8355 		vectors = 1;  /* one MSI or Legacy vector */
8356 	}
8357 
8358 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
8359 		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
8360 
8361 	/* set up vector assignment tracking */
8362 	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
8363 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
8364 	if (!pf->irq_pile) {
8365 		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
8366 		return -ENOMEM;
8367 	}
8368 	pf->irq_pile->num_entries = vectors;
8369 	pf->irq_pile->search_hint = 0;
8370 
8371 	/* track first vector for misc interrupts, ignore return */
8372 	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
8373 
8374 	return 0;
8375 }
8376 
8377 /**
8378  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
8379  * @pf: board private structure
8380  *
8381  * This sets up the handler for MSIX 0, which is used to manage the
8382  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
8383  * when in MSI or Legacy interrupt mode.
8384  **/
8385 static int i40e_setup_misc_vector(struct i40e_pf *pf)
8386 {
8387 	struct i40e_hw *hw = &pf->hw;
8388 	int err = 0;
8389 
8390 	/* Only request the irq if this is the first time through, and
8391 	 * not when we're rebuilding after a Reset
8392 	 */
8393 	if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state)) {
8394 		err = request_irq(pf->msix_entries[0].vector,
8395 				  i40e_intr, 0, pf->int_name, pf);
8396 		if (err) {
8397 			dev_info(&pf->pdev->dev,
8398 				 "request_irq for %s failed: %d\n",
8399 				 pf->int_name, err);
8400 			return -EFAULT;
8401 		}
8402 	}
8403 
8404 	i40e_enable_misc_int_causes(pf);
8405 
8406 	/* associate no queues to the misc vector */
8407 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
8408 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
8409 
8410 	i40e_flush(hw);
8411 
8412 	i40e_irq_dynamic_enable_icr0(pf, true);
8413 
8414 	return err;
8415 }
8416 
8417 /**
8418  * i40e_config_rss_aq - Prepare for RSS using AQ commands
8419  * @vsi: vsi structure
8420  * @seed: RSS hash seed
8421  **/
8422 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8423 			      u8 *lut, u16 lut_size)
8424 {
8425 	struct i40e_pf *pf = vsi->back;
8426 	struct i40e_hw *hw = &pf->hw;
8427 	int ret = 0;
8428 
8429 	if (seed) {
8430 		struct i40e_aqc_get_set_rss_key_data *seed_dw =
8431 			(struct i40e_aqc_get_set_rss_key_data *)seed;
8432 		ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
8433 		if (ret) {
8434 			dev_info(&pf->pdev->dev,
8435 				 "Cannot set RSS key, err %s aq_err %s\n",
8436 				 i40e_stat_str(hw, ret),
8437 				 i40e_aq_str(hw, hw->aq.asq_last_status));
8438 			return ret;
8439 		}
8440 	}
8441 	if (lut) {
8442 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8443 
8444 		ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8445 		if (ret) {
8446 			dev_info(&pf->pdev->dev,
8447 				 "Cannot set RSS lut, err %s aq_err %s\n",
8448 				 i40e_stat_str(hw, ret),
8449 				 i40e_aq_str(hw, hw->aq.asq_last_status));
8450 			return ret;
8451 		}
8452 	}
8453 	return ret;
8454 }
8455 
8456 /**
8457  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8458  * @vsi: Pointer to vsi structure
8459  * @seed: Buffter to store the hash keys
8460  * @lut: Buffer to store the lookup table entries
8461  * @lut_size: Size of buffer to store the lookup table entries
8462  *
8463  * Return 0 on success, negative on failure
8464  */
8465 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8466 			   u8 *lut, u16 lut_size)
8467 {
8468 	struct i40e_pf *pf = vsi->back;
8469 	struct i40e_hw *hw = &pf->hw;
8470 	int ret = 0;
8471 
8472 	if (seed) {
8473 		ret = i40e_aq_get_rss_key(hw, vsi->id,
8474 			(struct i40e_aqc_get_set_rss_key_data *)seed);
8475 		if (ret) {
8476 			dev_info(&pf->pdev->dev,
8477 				 "Cannot get RSS key, err %s aq_err %s\n",
8478 				 i40e_stat_str(&pf->hw, ret),
8479 				 i40e_aq_str(&pf->hw,
8480 					     pf->hw.aq.asq_last_status));
8481 			return ret;
8482 		}
8483 	}
8484 
8485 	if (lut) {
8486 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8487 
8488 		ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8489 		if (ret) {
8490 			dev_info(&pf->pdev->dev,
8491 				 "Cannot get RSS lut, err %s aq_err %s\n",
8492 				 i40e_stat_str(&pf->hw, ret),
8493 				 i40e_aq_str(&pf->hw,
8494 					     pf->hw.aq.asq_last_status));
8495 			return ret;
8496 		}
8497 	}
8498 
8499 	return ret;
8500 }
8501 
8502 /**
8503  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8504  * @vsi: VSI structure
8505  **/
8506 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8507 {
8508 	u8 seed[I40E_HKEY_ARRAY_SIZE];
8509 	struct i40e_pf *pf = vsi->back;
8510 	u8 *lut;
8511 	int ret;
8512 
8513 	if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8514 		return 0;
8515 
8516 	if (!vsi->rss_size)
8517 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
8518 				      vsi->num_queue_pairs);
8519 	if (!vsi->rss_size)
8520 		return -EINVAL;
8521 
8522 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8523 	if (!lut)
8524 		return -ENOMEM;
8525 	/* Use the user configured hash keys and lookup table if there is one,
8526 	 * otherwise use default
8527 	 */
8528 	if (vsi->rss_lut_user)
8529 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8530 	else
8531 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8532 	if (vsi->rss_hkey_user)
8533 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8534 	else
8535 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8536 	ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8537 	kfree(lut);
8538 
8539 	return ret;
8540 }
8541 
8542 /**
8543  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8544  * @vsi: Pointer to vsi structure
8545  * @seed: RSS hash seed
8546  * @lut: Lookup table
8547  * @lut_size: Lookup table size
8548  *
8549  * Returns 0 on success, negative on failure
8550  **/
8551 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8552 			       const u8 *lut, u16 lut_size)
8553 {
8554 	struct i40e_pf *pf = vsi->back;
8555 	struct i40e_hw *hw = &pf->hw;
8556 	u16 vf_id = vsi->vf_id;
8557 	u8 i;
8558 
8559 	/* Fill out hash function seed */
8560 	if (seed) {
8561 		u32 *seed_dw = (u32 *)seed;
8562 
8563 		if (vsi->type == I40E_VSI_MAIN) {
8564 			for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8565 				wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
8566 		} else if (vsi->type == I40E_VSI_SRIOV) {
8567 			for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8568 				wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
8569 		} else {
8570 			dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8571 		}
8572 	}
8573 
8574 	if (lut) {
8575 		u32 *lut_dw = (u32 *)lut;
8576 
8577 		if (vsi->type == I40E_VSI_MAIN) {
8578 			if (lut_size != I40E_HLUT_ARRAY_SIZE)
8579 				return -EINVAL;
8580 			for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8581 				wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8582 		} else if (vsi->type == I40E_VSI_SRIOV) {
8583 			if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8584 				return -EINVAL;
8585 			for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8586 				wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
8587 		} else {
8588 			dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8589 		}
8590 	}
8591 	i40e_flush(hw);
8592 
8593 	return 0;
8594 }
8595 
8596 /**
8597  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8598  * @vsi: Pointer to VSI structure
8599  * @seed: Buffer to store the keys
8600  * @lut: Buffer to store the lookup table entries
8601  * @lut_size: Size of buffer to store the lookup table entries
8602  *
8603  * Returns 0 on success, negative on failure
8604  */
8605 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8606 			    u8 *lut, u16 lut_size)
8607 {
8608 	struct i40e_pf *pf = vsi->back;
8609 	struct i40e_hw *hw = &pf->hw;
8610 	u16 i;
8611 
8612 	if (seed) {
8613 		u32 *seed_dw = (u32 *)seed;
8614 
8615 		for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8616 			seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8617 	}
8618 	if (lut) {
8619 		u32 *lut_dw = (u32 *)lut;
8620 
8621 		if (lut_size != I40E_HLUT_ARRAY_SIZE)
8622 			return -EINVAL;
8623 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8624 			lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8625 	}
8626 
8627 	return 0;
8628 }
8629 
8630 /**
8631  * i40e_config_rss - Configure RSS keys and lut
8632  * @vsi: Pointer to VSI structure
8633  * @seed: RSS hash seed
8634  * @lut: Lookup table
8635  * @lut_size: Lookup table size
8636  *
8637  * Returns 0 on success, negative on failure
8638  */
8639 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8640 {
8641 	struct i40e_pf *pf = vsi->back;
8642 
8643 	if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8644 		return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8645 	else
8646 		return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8647 }
8648 
8649 /**
8650  * i40e_get_rss - Get RSS keys and lut
8651  * @vsi: Pointer to VSI structure
8652  * @seed: Buffer to store the keys
8653  * @lut: Buffer to store the lookup table entries
8654  * lut_size: Size of buffer to store the lookup table entries
8655  *
8656  * Returns 0 on success, negative on failure
8657  */
8658 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8659 {
8660 	struct i40e_pf *pf = vsi->back;
8661 
8662 	if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8663 		return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8664 	else
8665 		return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8666 }
8667 
8668 /**
8669  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8670  * @pf: Pointer to board private structure
8671  * @lut: Lookup table
8672  * @rss_table_size: Lookup table size
8673  * @rss_size: Range of queue number for hashing
8674  */
8675 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8676 		       u16 rss_table_size, u16 rss_size)
8677 {
8678 	u16 i;
8679 
8680 	for (i = 0; i < rss_table_size; i++)
8681 		lut[i] = i % rss_size;
8682 }
8683 
8684 /**
8685  * i40e_pf_config_rss - Prepare for RSS if used
8686  * @pf: board private structure
8687  **/
8688 static int i40e_pf_config_rss(struct i40e_pf *pf)
8689 {
8690 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8691 	u8 seed[I40E_HKEY_ARRAY_SIZE];
8692 	u8 *lut;
8693 	struct i40e_hw *hw = &pf->hw;
8694 	u32 reg_val;
8695 	u64 hena;
8696 	int ret;
8697 
8698 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8699 	hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8700 		((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8701 	hena |= i40e_pf_get_default_rss_hena(pf);
8702 
8703 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8704 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8705 
8706 	/* Determine the RSS table size based on the hardware capabilities */
8707 	reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8708 	reg_val = (pf->rss_table_size == 512) ?
8709 			(reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8710 			(reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8711 	i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8712 
8713 	/* Determine the RSS size of the VSI */
8714 	if (!vsi->rss_size) {
8715 		u16 qcount;
8716 
8717 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8718 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8719 	}
8720 	if (!vsi->rss_size)
8721 		return -EINVAL;
8722 
8723 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8724 	if (!lut)
8725 		return -ENOMEM;
8726 
8727 	/* Use user configured lut if there is one, otherwise use default */
8728 	if (vsi->rss_lut_user)
8729 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8730 	else
8731 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8732 
8733 	/* Use user configured hash key if there is one, otherwise
8734 	 * use default.
8735 	 */
8736 	if (vsi->rss_hkey_user)
8737 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8738 	else
8739 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8740 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8741 	kfree(lut);
8742 
8743 	return ret;
8744 }
8745 
8746 /**
8747  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8748  * @pf: board private structure
8749  * @queue_count: the requested queue count for rss.
8750  *
8751  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8752  * count which may be different from the requested queue count.
8753  * Note: expects to be called while under rtnl_lock()
8754  **/
8755 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8756 {
8757 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8758 	int new_rss_size;
8759 
8760 	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8761 		return 0;
8762 
8763 	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8764 
8765 	if (queue_count != vsi->num_queue_pairs) {
8766 		u16 qcount;
8767 
8768 		vsi->req_queue_pairs = queue_count;
8769 		i40e_prep_for_reset(pf, true);
8770 
8771 		pf->alloc_rss_size = new_rss_size;
8772 
8773 		i40e_reset_and_rebuild(pf, true, true);
8774 
8775 		/* Discard the user configured hash keys and lut, if less
8776 		 * queues are enabled.
8777 		 */
8778 		if (queue_count < vsi->rss_size) {
8779 			i40e_clear_rss_config_user(vsi);
8780 			dev_dbg(&pf->pdev->dev,
8781 				"discard user configured hash keys and lut\n");
8782 		}
8783 
8784 		/* Reset vsi->rss_size, as number of enabled queues changed */
8785 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8786 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8787 
8788 		i40e_pf_config_rss(pf);
8789 	}
8790 	dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
8791 		 vsi->req_queue_pairs, pf->rss_size_max);
8792 	return pf->alloc_rss_size;
8793 }
8794 
8795 /**
8796  * i40e_get_partition_bw_setting - Retrieve BW settings for this PF partition
8797  * @pf: board private structure
8798  **/
8799 i40e_status i40e_get_partition_bw_setting(struct i40e_pf *pf)
8800 {
8801 	i40e_status status;
8802 	bool min_valid, max_valid;
8803 	u32 max_bw, min_bw;
8804 
8805 	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8806 					   &min_valid, &max_valid);
8807 
8808 	if (!status) {
8809 		if (min_valid)
8810 			pf->min_bw = min_bw;
8811 		if (max_valid)
8812 			pf->max_bw = max_bw;
8813 	}
8814 
8815 	return status;
8816 }
8817 
8818 /**
8819  * i40e_set_partition_bw_setting - Set BW settings for this PF partition
8820  * @pf: board private structure
8821  **/
8822 i40e_status i40e_set_partition_bw_setting(struct i40e_pf *pf)
8823 {
8824 	struct i40e_aqc_configure_partition_bw_data bw_data;
8825 	i40e_status status;
8826 
8827 	/* Set the valid bit for this PF */
8828 	bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8829 	bw_data.max_bw[pf->hw.pf_id] = pf->max_bw & I40E_ALT_BW_VALUE_MASK;
8830 	bw_data.min_bw[pf->hw.pf_id] = pf->min_bw & I40E_ALT_BW_VALUE_MASK;
8831 
8832 	/* Set the new bandwidths */
8833 	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8834 
8835 	return status;
8836 }
8837 
8838 /**
8839  * i40e_commit_partition_bw_setting - Commit BW settings for this PF partition
8840  * @pf: board private structure
8841  **/
8842 i40e_status i40e_commit_partition_bw_setting(struct i40e_pf *pf)
8843 {
8844 	/* Commit temporary BW setting to permanent NVM image */
8845 	enum i40e_admin_queue_err last_aq_status;
8846 	i40e_status ret;
8847 	u16 nvm_word;
8848 
8849 	if (pf->hw.partition_id != 1) {
8850 		dev_info(&pf->pdev->dev,
8851 			 "Commit BW only works on partition 1! This is partition %d",
8852 			 pf->hw.partition_id);
8853 		ret = I40E_NOT_SUPPORTED;
8854 		goto bw_commit_out;
8855 	}
8856 
8857 	/* Acquire NVM for read access */
8858 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8859 	last_aq_status = pf->hw.aq.asq_last_status;
8860 	if (ret) {
8861 		dev_info(&pf->pdev->dev,
8862 			 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8863 			 i40e_stat_str(&pf->hw, ret),
8864 			 i40e_aq_str(&pf->hw, last_aq_status));
8865 		goto bw_commit_out;
8866 	}
8867 
8868 	/* Read word 0x10 of NVM - SW compatibility word 1 */
8869 	ret = i40e_aq_read_nvm(&pf->hw,
8870 			       I40E_SR_NVM_CONTROL_WORD,
8871 			       0x10, sizeof(nvm_word), &nvm_word,
8872 			       false, NULL);
8873 	/* Save off last admin queue command status before releasing
8874 	 * the NVM
8875 	 */
8876 	last_aq_status = pf->hw.aq.asq_last_status;
8877 	i40e_release_nvm(&pf->hw);
8878 	if (ret) {
8879 		dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8880 			 i40e_stat_str(&pf->hw, ret),
8881 			 i40e_aq_str(&pf->hw, last_aq_status));
8882 		goto bw_commit_out;
8883 	}
8884 
8885 	/* Wait a bit for NVM release to complete */
8886 	msleep(50);
8887 
8888 	/* Acquire NVM for write access */
8889 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8890 	last_aq_status = pf->hw.aq.asq_last_status;
8891 	if (ret) {
8892 		dev_info(&pf->pdev->dev,
8893 			 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8894 			 i40e_stat_str(&pf->hw, ret),
8895 			 i40e_aq_str(&pf->hw, last_aq_status));
8896 		goto bw_commit_out;
8897 	}
8898 	/* Write it back out unchanged to initiate update NVM,
8899 	 * which will force a write of the shadow (alt) RAM to
8900 	 * the NVM - thus storing the bandwidth values permanently.
8901 	 */
8902 	ret = i40e_aq_update_nvm(&pf->hw,
8903 				 I40E_SR_NVM_CONTROL_WORD,
8904 				 0x10, sizeof(nvm_word),
8905 				 &nvm_word, true, NULL);
8906 	/* Save off last admin queue command status before releasing
8907 	 * the NVM
8908 	 */
8909 	last_aq_status = pf->hw.aq.asq_last_status;
8910 	i40e_release_nvm(&pf->hw);
8911 	if (ret)
8912 		dev_info(&pf->pdev->dev,
8913 			 "BW settings NOT SAVED, err %s aq_err %s\n",
8914 			 i40e_stat_str(&pf->hw, ret),
8915 			 i40e_aq_str(&pf->hw, last_aq_status));
8916 bw_commit_out:
8917 
8918 	return ret;
8919 }
8920 
8921 /**
8922  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8923  * @pf: board private structure to initialize
8924  *
8925  * i40e_sw_init initializes the Adapter private data structure.
8926  * Fields are initialized based on PCI device information and
8927  * OS network device settings (MTU size).
8928  **/
8929 static int i40e_sw_init(struct i40e_pf *pf)
8930 {
8931 	int err = 0;
8932 	int size;
8933 
8934 	/* Set default capability flags */
8935 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8936 		    I40E_FLAG_MSI_ENABLED     |
8937 		    I40E_FLAG_MSIX_ENABLED;
8938 
8939 	/* Set default ITR */
8940 	pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8941 	pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8942 
8943 	/* Depending on PF configurations, it is possible that the RSS
8944 	 * maximum might end up larger than the available queues
8945 	 */
8946 	pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8947 	pf->alloc_rss_size = 1;
8948 	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8949 	pf->rss_size_max = min_t(int, pf->rss_size_max,
8950 				 pf->hw.func_caps.num_tx_qp);
8951 	if (pf->hw.func_caps.rss) {
8952 		pf->flags |= I40E_FLAG_RSS_ENABLED;
8953 		pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8954 					   num_online_cpus());
8955 	}
8956 
8957 	/* MFP mode enabled */
8958 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8959 		pf->flags |= I40E_FLAG_MFP_ENABLED;
8960 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8961 		if (i40e_get_partition_bw_setting(pf)) {
8962 			dev_warn(&pf->pdev->dev,
8963 				 "Could not get partition bw settings\n");
8964 		} else {
8965 			dev_info(&pf->pdev->dev,
8966 				 "Partition BW Min = %8.8x, Max = %8.8x\n",
8967 				 pf->min_bw, pf->max_bw);
8968 
8969 			/* nudge the Tx scheduler */
8970 			i40e_set_partition_bw_setting(pf);
8971 		}
8972 	}
8973 
8974 	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8975 	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8976 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8977 		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8978 		if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8979 		    pf->hw.num_partitions > 1)
8980 			dev_info(&pf->pdev->dev,
8981 				 "Flow Director Sideband mode Disabled in MFP mode\n");
8982 		else
8983 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8984 		pf->fdir_pf_filter_count =
8985 				 pf->hw.func_caps.fd_filters_guaranteed;
8986 		pf->hw.fdir_shared_filter_count =
8987 				 pf->hw.func_caps.fd_filters_best_effort;
8988 	}
8989 
8990 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
8991 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8992 	    (pf->hw.aq.fw_maj_ver < 4))) {
8993 		pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8994 		/* No DCB support  for FW < v4.33 */
8995 		pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8996 	}
8997 
8998 	/* Disable FW LLDP if FW < v4.3 */
8999 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
9000 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
9001 	    (pf->hw.aq.fw_maj_ver < 4)))
9002 		pf->flags |= I40E_FLAG_STOP_FW_LLDP;
9003 
9004 	/* Use the FW Set LLDP MIB API if FW > v4.40 */
9005 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
9006 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
9007 	    (pf->hw.aq.fw_maj_ver >= 5)))
9008 		pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
9009 
9010 	if (pf->hw.func_caps.vmdq) {
9011 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
9012 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
9013 		pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
9014 	}
9015 
9016 	if (pf->hw.func_caps.iwarp) {
9017 		pf->flags |= I40E_FLAG_IWARP_ENABLED;
9018 		/* IWARP needs one extra vector for CQP just like MISC.*/
9019 		pf->num_iwarp_msix = (int)num_online_cpus() + 1;
9020 	}
9021 
9022 #ifdef CONFIG_PCI_IOV
9023 	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
9024 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
9025 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
9026 		pf->num_req_vfs = min_t(int,
9027 					pf->hw.func_caps.num_vfs,
9028 					I40E_MAX_VF_COUNT);
9029 	}
9030 #endif /* CONFIG_PCI_IOV */
9031 	if (pf->hw.mac.type == I40E_MAC_X722) {
9032 		pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE
9033 			     | I40E_FLAG_128_QP_RSS_CAPABLE
9034 			     | I40E_FLAG_HW_ATR_EVICT_CAPABLE
9035 			     | I40E_FLAG_OUTER_UDP_CSUM_CAPABLE
9036 			     | I40E_FLAG_WB_ON_ITR_CAPABLE
9037 			     | I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE
9038 			     | I40E_FLAG_NO_PCI_LINK_CHECK
9039 			     | I40E_FLAG_USE_SET_LLDP_MIB
9040 			     | I40E_FLAG_GENEVE_OFFLOAD_CAPABLE
9041 			     | I40E_FLAG_PTP_L4_CAPABLE
9042 			     | I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE;
9043 	} else if ((pf->hw.aq.api_maj_ver > 1) ||
9044 		   ((pf->hw.aq.api_maj_ver == 1) &&
9045 		    (pf->hw.aq.api_min_ver > 4))) {
9046 		/* Supported in FW API version higher than 1.4 */
9047 		pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
9048 	}
9049 
9050 	/* Enable HW ATR eviction if possible */
9051 	if (pf->flags & I40E_FLAG_HW_ATR_EVICT_CAPABLE)
9052 		pf->flags |= I40E_FLAG_HW_ATR_EVICT_ENABLED;
9053 
9054 	pf->eeprom_version = 0xDEAD;
9055 	pf->lan_veb = I40E_NO_VEB;
9056 	pf->lan_vsi = I40E_NO_VSI;
9057 
9058 	/* By default FW has this off for performance reasons */
9059 	pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
9060 
9061 	/* set up queue assignment tracking */
9062 	size = sizeof(struct i40e_lump_tracking)
9063 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
9064 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
9065 	if (!pf->qp_pile) {
9066 		err = -ENOMEM;
9067 		goto sw_init_done;
9068 	}
9069 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
9070 	pf->qp_pile->search_hint = 0;
9071 
9072 	pf->tx_timeout_recovery_level = 1;
9073 
9074 	mutex_init(&pf->switch_mutex);
9075 
9076 sw_init_done:
9077 	return err;
9078 }
9079 
9080 /**
9081  * i40e_set_ntuple - set the ntuple feature flag and take action
9082  * @pf: board private structure to initialize
9083  * @features: the feature set that the stack is suggesting
9084  *
9085  * returns a bool to indicate if reset needs to happen
9086  **/
9087 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
9088 {
9089 	bool need_reset = false;
9090 
9091 	/* Check if Flow Director n-tuple support was enabled or disabled.  If
9092 	 * the state changed, we need to reset.
9093 	 */
9094 	if (features & NETIF_F_NTUPLE) {
9095 		/* Enable filters and mark for reset */
9096 		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
9097 			need_reset = true;
9098 		/* enable FD_SB only if there is MSI-X vector */
9099 		if (pf->num_fdsb_msix > 0)
9100 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
9101 	} else {
9102 		/* turn off filters, mark for reset and clear SW filter list */
9103 		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9104 			need_reset = true;
9105 			i40e_fdir_filter_exit(pf);
9106 		}
9107 		pf->flags &= ~(I40E_FLAG_FD_SB_ENABLED |
9108 			       I40E_FLAG_FD_SB_AUTO_DISABLED);
9109 		/* reset fd counters */
9110 		pf->fd_add_err = 0;
9111 		pf->fd_atr_cnt = 0;
9112 		/* if ATR was auto disabled it can be re-enabled. */
9113 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
9114 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
9115 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
9116 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
9117 				dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
9118 		}
9119 	}
9120 	return need_reset;
9121 }
9122 
9123 /**
9124  * i40e_clear_rss_lut - clear the rx hash lookup table
9125  * @vsi: the VSI being configured
9126  **/
9127 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
9128 {
9129 	struct i40e_pf *pf = vsi->back;
9130 	struct i40e_hw *hw = &pf->hw;
9131 	u16 vf_id = vsi->vf_id;
9132 	u8 i;
9133 
9134 	if (vsi->type == I40E_VSI_MAIN) {
9135 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
9136 			wr32(hw, I40E_PFQF_HLUT(i), 0);
9137 	} else if (vsi->type == I40E_VSI_SRIOV) {
9138 		for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
9139 			i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
9140 	} else {
9141 		dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
9142 	}
9143 }
9144 
9145 /**
9146  * i40e_set_features - set the netdev feature flags
9147  * @netdev: ptr to the netdev being adjusted
9148  * @features: the feature set that the stack is suggesting
9149  * Note: expects to be called while under rtnl_lock()
9150  **/
9151 static int i40e_set_features(struct net_device *netdev,
9152 			     netdev_features_t features)
9153 {
9154 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9155 	struct i40e_vsi *vsi = np->vsi;
9156 	struct i40e_pf *pf = vsi->back;
9157 	bool need_reset;
9158 
9159 	if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
9160 		i40e_pf_config_rss(pf);
9161 	else if (!(features & NETIF_F_RXHASH) &&
9162 		 netdev->features & NETIF_F_RXHASH)
9163 		i40e_clear_rss_lut(vsi);
9164 
9165 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
9166 		i40e_vlan_stripping_enable(vsi);
9167 	else
9168 		i40e_vlan_stripping_disable(vsi);
9169 
9170 	need_reset = i40e_set_ntuple(pf, features);
9171 
9172 	if (need_reset)
9173 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
9174 
9175 	return 0;
9176 }
9177 
9178 /**
9179  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
9180  * @pf: board private structure
9181  * @port: The UDP port to look up
9182  *
9183  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
9184  **/
9185 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
9186 {
9187 	u8 i;
9188 
9189 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
9190 		if (pf->udp_ports[i].port == port)
9191 			return i;
9192 	}
9193 
9194 	return i;
9195 }
9196 
9197 /**
9198  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
9199  * @netdev: This physical port's netdev
9200  * @ti: Tunnel endpoint information
9201  **/
9202 static void i40e_udp_tunnel_add(struct net_device *netdev,
9203 				struct udp_tunnel_info *ti)
9204 {
9205 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9206 	struct i40e_vsi *vsi = np->vsi;
9207 	struct i40e_pf *pf = vsi->back;
9208 	u16 port = ntohs(ti->port);
9209 	u8 next_idx;
9210 	u8 idx;
9211 
9212 	idx = i40e_get_udp_port_idx(pf, port);
9213 
9214 	/* Check if port already exists */
9215 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
9216 		netdev_info(netdev, "port %d already offloaded\n", port);
9217 		return;
9218 	}
9219 
9220 	/* Now check if there is space to add the new port */
9221 	next_idx = i40e_get_udp_port_idx(pf, 0);
9222 
9223 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
9224 		netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
9225 			    port);
9226 		return;
9227 	}
9228 
9229 	switch (ti->type) {
9230 	case UDP_TUNNEL_TYPE_VXLAN:
9231 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
9232 		break;
9233 	case UDP_TUNNEL_TYPE_GENEVE:
9234 		if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
9235 			return;
9236 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
9237 		break;
9238 	default:
9239 		return;
9240 	}
9241 
9242 	/* New port: add it and mark its index in the bitmap */
9243 	pf->udp_ports[next_idx].port = port;
9244 	pf->pending_udp_bitmap |= BIT_ULL(next_idx);
9245 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9246 }
9247 
9248 /**
9249  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
9250  * @netdev: This physical port's netdev
9251  * @ti: Tunnel endpoint information
9252  **/
9253 static void i40e_udp_tunnel_del(struct net_device *netdev,
9254 				struct udp_tunnel_info *ti)
9255 {
9256 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9257 	struct i40e_vsi *vsi = np->vsi;
9258 	struct i40e_pf *pf = vsi->back;
9259 	u16 port = ntohs(ti->port);
9260 	u8 idx;
9261 
9262 	idx = i40e_get_udp_port_idx(pf, port);
9263 
9264 	/* Check if port already exists */
9265 	if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
9266 		goto not_found;
9267 
9268 	switch (ti->type) {
9269 	case UDP_TUNNEL_TYPE_VXLAN:
9270 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
9271 			goto not_found;
9272 		break;
9273 	case UDP_TUNNEL_TYPE_GENEVE:
9274 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
9275 			goto not_found;
9276 		break;
9277 	default:
9278 		goto not_found;
9279 	}
9280 
9281 	/* if port exists, set it to 0 (mark for deletion)
9282 	 * and make it pending
9283 	 */
9284 	pf->udp_ports[idx].port = 0;
9285 	pf->pending_udp_bitmap |= BIT_ULL(idx);
9286 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9287 
9288 	return;
9289 not_found:
9290 	netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
9291 		    port);
9292 }
9293 
9294 static int i40e_get_phys_port_id(struct net_device *netdev,
9295 				 struct netdev_phys_item_id *ppid)
9296 {
9297 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9298 	struct i40e_pf *pf = np->vsi->back;
9299 	struct i40e_hw *hw = &pf->hw;
9300 
9301 	if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
9302 		return -EOPNOTSUPP;
9303 
9304 	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
9305 	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
9306 
9307 	return 0;
9308 }
9309 
9310 /**
9311  * i40e_ndo_fdb_add - add an entry to the hardware database
9312  * @ndm: the input from the stack
9313  * @tb: pointer to array of nladdr (unused)
9314  * @dev: the net device pointer
9315  * @addr: the MAC address entry being added
9316  * @flags: instructions from stack about fdb operation
9317  */
9318 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9319 			    struct net_device *dev,
9320 			    const unsigned char *addr, u16 vid,
9321 			    u16 flags)
9322 {
9323 	struct i40e_netdev_priv *np = netdev_priv(dev);
9324 	struct i40e_pf *pf = np->vsi->back;
9325 	int err = 0;
9326 
9327 	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
9328 		return -EOPNOTSUPP;
9329 
9330 	if (vid) {
9331 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
9332 		return -EINVAL;
9333 	}
9334 
9335 	/* Hardware does not support aging addresses so if a
9336 	 * ndm_state is given only allow permanent addresses
9337 	 */
9338 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
9339 		netdev_info(dev, "FDB only supports static addresses\n");
9340 		return -EINVAL;
9341 	}
9342 
9343 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
9344 		err = dev_uc_add_excl(dev, addr);
9345 	else if (is_multicast_ether_addr(addr))
9346 		err = dev_mc_add_excl(dev, addr);
9347 	else
9348 		err = -EINVAL;
9349 
9350 	/* Only return duplicate errors if NLM_F_EXCL is set */
9351 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
9352 		err = 0;
9353 
9354 	return err;
9355 }
9356 
9357 /**
9358  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
9359  * @dev: the netdev being configured
9360  * @nlh: RTNL message
9361  *
9362  * Inserts a new hardware bridge if not already created and
9363  * enables the bridging mode requested (VEB or VEPA). If the
9364  * hardware bridge has already been inserted and the request
9365  * is to change the mode then that requires a PF reset to
9366  * allow rebuild of the components with required hardware
9367  * bridge mode enabled.
9368  *
9369  * Note: expects to be called while under rtnl_lock()
9370  **/
9371 static int i40e_ndo_bridge_setlink(struct net_device *dev,
9372 				   struct nlmsghdr *nlh,
9373 				   u16 flags)
9374 {
9375 	struct i40e_netdev_priv *np = netdev_priv(dev);
9376 	struct i40e_vsi *vsi = np->vsi;
9377 	struct i40e_pf *pf = vsi->back;
9378 	struct i40e_veb *veb = NULL;
9379 	struct nlattr *attr, *br_spec;
9380 	int i, rem;
9381 
9382 	/* Only for PF VSI for now */
9383 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9384 		return -EOPNOTSUPP;
9385 
9386 	/* Find the HW bridge for PF VSI */
9387 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9388 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9389 			veb = pf->veb[i];
9390 	}
9391 
9392 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9393 
9394 	nla_for_each_nested(attr, br_spec, rem) {
9395 		__u16 mode;
9396 
9397 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9398 			continue;
9399 
9400 		mode = nla_get_u16(attr);
9401 		if ((mode != BRIDGE_MODE_VEPA) &&
9402 		    (mode != BRIDGE_MODE_VEB))
9403 			return -EINVAL;
9404 
9405 		/* Insert a new HW bridge */
9406 		if (!veb) {
9407 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9408 					     vsi->tc_config.enabled_tc);
9409 			if (veb) {
9410 				veb->bridge_mode = mode;
9411 				i40e_config_bridge_mode(veb);
9412 			} else {
9413 				/* No Bridge HW offload available */
9414 				return -ENOENT;
9415 			}
9416 			break;
9417 		} else if (mode != veb->bridge_mode) {
9418 			/* Existing HW bridge but different mode needs reset */
9419 			veb->bridge_mode = mode;
9420 			/* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
9421 			if (mode == BRIDGE_MODE_VEB)
9422 				pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
9423 			else
9424 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9425 			i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED),
9426 				      true);
9427 			break;
9428 		}
9429 	}
9430 
9431 	return 0;
9432 }
9433 
9434 /**
9435  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
9436  * @skb: skb buff
9437  * @pid: process id
9438  * @seq: RTNL message seq #
9439  * @dev: the netdev being configured
9440  * @filter_mask: unused
9441  * @nlflags: netlink flags passed in
9442  *
9443  * Return the mode in which the hardware bridge is operating in
9444  * i.e VEB or VEPA.
9445  **/
9446 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9447 				   struct net_device *dev,
9448 				   u32 __always_unused filter_mask,
9449 				   int nlflags)
9450 {
9451 	struct i40e_netdev_priv *np = netdev_priv(dev);
9452 	struct i40e_vsi *vsi = np->vsi;
9453 	struct i40e_pf *pf = vsi->back;
9454 	struct i40e_veb *veb = NULL;
9455 	int i;
9456 
9457 	/* Only for PF VSI for now */
9458 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9459 		return -EOPNOTSUPP;
9460 
9461 	/* Find the HW bridge for the PF VSI */
9462 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9463 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9464 			veb = pf->veb[i];
9465 	}
9466 
9467 	if (!veb)
9468 		return 0;
9469 
9470 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9471 				       0, 0, nlflags, filter_mask, NULL);
9472 }
9473 
9474 /**
9475  * i40e_features_check - Validate encapsulated packet conforms to limits
9476  * @skb: skb buff
9477  * @dev: This physical port's netdev
9478  * @features: Offload features that the stack believes apply
9479  **/
9480 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9481 					     struct net_device *dev,
9482 					     netdev_features_t features)
9483 {
9484 	size_t len;
9485 
9486 	/* No point in doing any of this if neither checksum nor GSO are
9487 	 * being requested for this frame.  We can rule out both by just
9488 	 * checking for CHECKSUM_PARTIAL
9489 	 */
9490 	if (skb->ip_summed != CHECKSUM_PARTIAL)
9491 		return features;
9492 
9493 	/* We cannot support GSO if the MSS is going to be less than
9494 	 * 64 bytes.  If it is then we need to drop support for GSO.
9495 	 */
9496 	if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
9497 		features &= ~NETIF_F_GSO_MASK;
9498 
9499 	/* MACLEN can support at most 63 words */
9500 	len = skb_network_header(skb) - skb->data;
9501 	if (len & ~(63 * 2))
9502 		goto out_err;
9503 
9504 	/* IPLEN and EIPLEN can support at most 127 dwords */
9505 	len = skb_transport_header(skb) - skb_network_header(skb);
9506 	if (len & ~(127 * 4))
9507 		goto out_err;
9508 
9509 	if (skb->encapsulation) {
9510 		/* L4TUNLEN can support 127 words */
9511 		len = skb_inner_network_header(skb) - skb_transport_header(skb);
9512 		if (len & ~(127 * 2))
9513 			goto out_err;
9514 
9515 		/* IPLEN can support at most 127 dwords */
9516 		len = skb_inner_transport_header(skb) -
9517 		      skb_inner_network_header(skb);
9518 		if (len & ~(127 * 4))
9519 			goto out_err;
9520 	}
9521 
9522 	/* No need to validate L4LEN as TCP is the only protocol with a
9523 	 * a flexible value and we support all possible values supported
9524 	 * by TCP, which is at most 15 dwords
9525 	 */
9526 
9527 	return features;
9528 out_err:
9529 	return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9530 }
9531 
9532 /**
9533  * i40e_xdp_setup - add/remove an XDP program
9534  * @vsi: VSI to changed
9535  * @prog: XDP program
9536  **/
9537 static int i40e_xdp_setup(struct i40e_vsi *vsi,
9538 			  struct bpf_prog *prog)
9539 {
9540 	int frame_size = vsi->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
9541 	struct i40e_pf *pf = vsi->back;
9542 	struct bpf_prog *old_prog;
9543 	bool need_reset;
9544 	int i;
9545 
9546 	/* Don't allow frames that span over multiple buffers */
9547 	if (frame_size > vsi->rx_buf_len)
9548 		return -EINVAL;
9549 
9550 	if (!i40e_enabled_xdp_vsi(vsi) && !prog)
9551 		return 0;
9552 
9553 	/* When turning XDP on->off/off->on we reset and rebuild the rings. */
9554 	need_reset = (i40e_enabled_xdp_vsi(vsi) != !!prog);
9555 
9556 	if (need_reset)
9557 		i40e_prep_for_reset(pf, true);
9558 
9559 	old_prog = xchg(&vsi->xdp_prog, prog);
9560 
9561 	if (need_reset)
9562 		i40e_reset_and_rebuild(pf, true, true);
9563 
9564 	for (i = 0; i < vsi->num_queue_pairs; i++)
9565 		WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
9566 
9567 	if (old_prog)
9568 		bpf_prog_put(old_prog);
9569 
9570 	return 0;
9571 }
9572 
9573 /**
9574  * i40e_xdp - implements ndo_xdp for i40e
9575  * @dev: netdevice
9576  * @xdp: XDP command
9577  **/
9578 static int i40e_xdp(struct net_device *dev,
9579 		    struct netdev_xdp *xdp)
9580 {
9581 	struct i40e_netdev_priv *np = netdev_priv(dev);
9582 	struct i40e_vsi *vsi = np->vsi;
9583 
9584 	if (vsi->type != I40E_VSI_MAIN)
9585 		return -EINVAL;
9586 
9587 	switch (xdp->command) {
9588 	case XDP_SETUP_PROG:
9589 		return i40e_xdp_setup(vsi, xdp->prog);
9590 	case XDP_QUERY_PROG:
9591 		xdp->prog_attached = i40e_enabled_xdp_vsi(vsi);
9592 		return 0;
9593 	default:
9594 		return -EINVAL;
9595 	}
9596 }
9597 
9598 static const struct net_device_ops i40e_netdev_ops = {
9599 	.ndo_open		= i40e_open,
9600 	.ndo_stop		= i40e_close,
9601 	.ndo_start_xmit		= i40e_lan_xmit_frame,
9602 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
9603 	.ndo_set_rx_mode	= i40e_set_rx_mode,
9604 	.ndo_validate_addr	= eth_validate_addr,
9605 	.ndo_set_mac_address	= i40e_set_mac,
9606 	.ndo_change_mtu		= i40e_change_mtu,
9607 	.ndo_do_ioctl		= i40e_ioctl,
9608 	.ndo_tx_timeout		= i40e_tx_timeout,
9609 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
9610 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
9611 #ifdef CONFIG_NET_POLL_CONTROLLER
9612 	.ndo_poll_controller	= i40e_netpoll,
9613 #endif
9614 	.ndo_setup_tc		= __i40e_setup_tc,
9615 	.ndo_set_features	= i40e_set_features,
9616 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
9617 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
9618 	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
9619 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
9620 	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
9621 	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
9622 	.ndo_set_vf_trust	= i40e_ndo_set_vf_trust,
9623 	.ndo_udp_tunnel_add	= i40e_udp_tunnel_add,
9624 	.ndo_udp_tunnel_del	= i40e_udp_tunnel_del,
9625 	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
9626 	.ndo_fdb_add		= i40e_ndo_fdb_add,
9627 	.ndo_features_check	= i40e_features_check,
9628 	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
9629 	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
9630 	.ndo_xdp		= i40e_xdp,
9631 };
9632 
9633 /**
9634  * i40e_config_netdev - Setup the netdev flags
9635  * @vsi: the VSI being configured
9636  *
9637  * Returns 0 on success, negative value on failure
9638  **/
9639 static int i40e_config_netdev(struct i40e_vsi *vsi)
9640 {
9641 	struct i40e_pf *pf = vsi->back;
9642 	struct i40e_hw *hw = &pf->hw;
9643 	struct i40e_netdev_priv *np;
9644 	struct net_device *netdev;
9645 	u8 broadcast[ETH_ALEN];
9646 	u8 mac_addr[ETH_ALEN];
9647 	int etherdev_size;
9648 	netdev_features_t hw_enc_features;
9649 	netdev_features_t hw_features;
9650 
9651 	etherdev_size = sizeof(struct i40e_netdev_priv);
9652 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9653 	if (!netdev)
9654 		return -ENOMEM;
9655 
9656 	vsi->netdev = netdev;
9657 	np = netdev_priv(netdev);
9658 	np->vsi = vsi;
9659 
9660 	hw_enc_features = NETIF_F_SG			|
9661 			  NETIF_F_IP_CSUM		|
9662 			  NETIF_F_IPV6_CSUM		|
9663 			  NETIF_F_HIGHDMA		|
9664 			  NETIF_F_SOFT_FEATURES		|
9665 			  NETIF_F_TSO			|
9666 			  NETIF_F_TSO_ECN		|
9667 			  NETIF_F_TSO6			|
9668 			  NETIF_F_GSO_GRE		|
9669 			  NETIF_F_GSO_GRE_CSUM		|
9670 			  NETIF_F_GSO_PARTIAL		|
9671 			  NETIF_F_GSO_UDP_TUNNEL	|
9672 			  NETIF_F_GSO_UDP_TUNNEL_CSUM	|
9673 			  NETIF_F_SCTP_CRC		|
9674 			  NETIF_F_RXHASH		|
9675 			  NETIF_F_RXCSUM		|
9676 			  0;
9677 
9678 	if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9679 		netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9680 
9681 	netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9682 
9683 	netdev->hw_enc_features |= hw_enc_features;
9684 
9685 	/* record features VLANs can make use of */
9686 	netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
9687 
9688 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9689 		netdev->hw_features |= NETIF_F_NTUPLE;
9690 	hw_features = hw_enc_features		|
9691 		      NETIF_F_HW_VLAN_CTAG_TX	|
9692 		      NETIF_F_HW_VLAN_CTAG_RX;
9693 
9694 	netdev->hw_features |= hw_features;
9695 
9696 	netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9697 	netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9698 
9699 	if (vsi->type == I40E_VSI_MAIN) {
9700 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9701 		ether_addr_copy(mac_addr, hw->mac.perm_addr);
9702 		/* The following steps are necessary for two reasons. First,
9703 		 * some older NVM configurations load a default MAC-VLAN
9704 		 * filter that will accept any tagged packet, and we want to
9705 		 * replace this with a normal filter. Additionally, it is
9706 		 * possible our MAC address was provided by the platform using
9707 		 * Open Firmware or similar.
9708 		 *
9709 		 * Thus, we need to remove the default filter and install one
9710 		 * specific to the MAC address.
9711 		 */
9712 		i40e_rm_default_mac_filter(vsi, mac_addr);
9713 		spin_lock_bh(&vsi->mac_filter_hash_lock);
9714 		i40e_add_mac_filter(vsi, mac_addr);
9715 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
9716 	} else {
9717 		/* relate the VSI_VMDQ name to the VSI_MAIN name */
9718 		snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9719 			 pf->vsi[pf->lan_vsi]->netdev->name);
9720 		random_ether_addr(mac_addr);
9721 
9722 		spin_lock_bh(&vsi->mac_filter_hash_lock);
9723 		i40e_add_mac_filter(vsi, mac_addr);
9724 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
9725 	}
9726 
9727 	/* Add the broadcast filter so that we initially will receive
9728 	 * broadcast packets. Note that when a new VLAN is first added the
9729 	 * driver will convert all filters marked I40E_VLAN_ANY into VLAN
9730 	 * specific filters as part of transitioning into "vlan" operation.
9731 	 * When more VLANs are added, the driver will copy each existing MAC
9732 	 * filter and add it for the new VLAN.
9733 	 *
9734 	 * Broadcast filters are handled specially by
9735 	 * i40e_sync_filters_subtask, as the driver must to set the broadcast
9736 	 * promiscuous bit instead of adding this directly as a MAC/VLAN
9737 	 * filter. The subtask will update the correct broadcast promiscuous
9738 	 * bits as VLANs become active or inactive.
9739 	 */
9740 	eth_broadcast_addr(broadcast);
9741 	spin_lock_bh(&vsi->mac_filter_hash_lock);
9742 	i40e_add_mac_filter(vsi, broadcast);
9743 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
9744 
9745 	ether_addr_copy(netdev->dev_addr, mac_addr);
9746 	ether_addr_copy(netdev->perm_addr, mac_addr);
9747 
9748 	netdev->priv_flags |= IFF_UNICAST_FLT;
9749 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9750 	/* Setup netdev TC information */
9751 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9752 
9753 	netdev->netdev_ops = &i40e_netdev_ops;
9754 	netdev->watchdog_timeo = 5 * HZ;
9755 	i40e_set_ethtool_ops(netdev);
9756 
9757 	/* MTU range: 68 - 9706 */
9758 	netdev->min_mtu = ETH_MIN_MTU;
9759 	netdev->max_mtu = I40E_MAX_RXBUFFER -
9760 			  (ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN);
9761 
9762 	return 0;
9763 }
9764 
9765 /**
9766  * i40e_vsi_delete - Delete a VSI from the switch
9767  * @vsi: the VSI being removed
9768  *
9769  * Returns 0 on success, negative value on failure
9770  **/
9771 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9772 {
9773 	/* remove default VSI is not allowed */
9774 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9775 		return;
9776 
9777 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9778 }
9779 
9780 /**
9781  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9782  * @vsi: the VSI being queried
9783  *
9784  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9785  **/
9786 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9787 {
9788 	struct i40e_veb *veb;
9789 	struct i40e_pf *pf = vsi->back;
9790 
9791 	/* Uplink is not a bridge so default to VEB */
9792 	if (vsi->veb_idx == I40E_NO_VEB)
9793 		return 1;
9794 
9795 	veb = pf->veb[vsi->veb_idx];
9796 	if (!veb) {
9797 		dev_info(&pf->pdev->dev,
9798 			 "There is no veb associated with the bridge\n");
9799 		return -ENOENT;
9800 	}
9801 
9802 	/* Uplink is a bridge in VEPA mode */
9803 	if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9804 		return 0;
9805 	} else {
9806 		/* Uplink is a bridge in VEB mode */
9807 		return 1;
9808 	}
9809 
9810 	/* VEPA is now default bridge, so return 0 */
9811 	return 0;
9812 }
9813 
9814 /**
9815  * i40e_add_vsi - Add a VSI to the switch
9816  * @vsi: the VSI being configured
9817  *
9818  * This initializes a VSI context depending on the VSI type to be added and
9819  * passes it down to the add_vsi aq command.
9820  **/
9821 static int i40e_add_vsi(struct i40e_vsi *vsi)
9822 {
9823 	int ret = -ENODEV;
9824 	struct i40e_pf *pf = vsi->back;
9825 	struct i40e_hw *hw = &pf->hw;
9826 	struct i40e_vsi_context ctxt;
9827 	struct i40e_mac_filter *f;
9828 	struct hlist_node *h;
9829 	int bkt;
9830 
9831 	u8 enabled_tc = 0x1; /* TC0 enabled */
9832 	int f_count = 0;
9833 
9834 	memset(&ctxt, 0, sizeof(ctxt));
9835 	switch (vsi->type) {
9836 	case I40E_VSI_MAIN:
9837 		/* The PF's main VSI is already setup as part of the
9838 		 * device initialization, so we'll not bother with
9839 		 * the add_vsi call, but we will retrieve the current
9840 		 * VSI context.
9841 		 */
9842 		ctxt.seid = pf->main_vsi_seid;
9843 		ctxt.pf_num = pf->hw.pf_id;
9844 		ctxt.vf_num = 0;
9845 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9846 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9847 		if (ret) {
9848 			dev_info(&pf->pdev->dev,
9849 				 "couldn't get PF vsi config, err %s aq_err %s\n",
9850 				 i40e_stat_str(&pf->hw, ret),
9851 				 i40e_aq_str(&pf->hw,
9852 					     pf->hw.aq.asq_last_status));
9853 			return -ENOENT;
9854 		}
9855 		vsi->info = ctxt.info;
9856 		vsi->info.valid_sections = 0;
9857 
9858 		vsi->seid = ctxt.seid;
9859 		vsi->id = ctxt.vsi_number;
9860 
9861 		enabled_tc = i40e_pf_get_tc_map(pf);
9862 
9863 		/* MFP mode setup queue map and update VSI */
9864 		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9865 		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9866 			memset(&ctxt, 0, sizeof(ctxt));
9867 			ctxt.seid = pf->main_vsi_seid;
9868 			ctxt.pf_num = pf->hw.pf_id;
9869 			ctxt.vf_num = 0;
9870 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9871 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9872 			if (ret) {
9873 				dev_info(&pf->pdev->dev,
9874 					 "update vsi failed, err %s aq_err %s\n",
9875 					 i40e_stat_str(&pf->hw, ret),
9876 					 i40e_aq_str(&pf->hw,
9877 						    pf->hw.aq.asq_last_status));
9878 				ret = -ENOENT;
9879 				goto err;
9880 			}
9881 			/* update the local VSI info queue map */
9882 			i40e_vsi_update_queue_map(vsi, &ctxt);
9883 			vsi->info.valid_sections = 0;
9884 		} else {
9885 			/* Default/Main VSI is only enabled for TC0
9886 			 * reconfigure it to enable all TCs that are
9887 			 * available on the port in SFP mode.
9888 			 * For MFP case the iSCSI PF would use this
9889 			 * flow to enable LAN+iSCSI TC.
9890 			 */
9891 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
9892 			if (ret) {
9893 				dev_info(&pf->pdev->dev,
9894 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9895 					 enabled_tc,
9896 					 i40e_stat_str(&pf->hw, ret),
9897 					 i40e_aq_str(&pf->hw,
9898 						    pf->hw.aq.asq_last_status));
9899 				ret = -ENOENT;
9900 			}
9901 		}
9902 		break;
9903 
9904 	case I40E_VSI_FDIR:
9905 		ctxt.pf_num = hw->pf_id;
9906 		ctxt.vf_num = 0;
9907 		ctxt.uplink_seid = vsi->uplink_seid;
9908 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9909 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9910 		if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9911 		    (i40e_is_vsi_uplink_mode_veb(vsi))) {
9912 			ctxt.info.valid_sections |=
9913 			     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9914 			ctxt.info.switch_id =
9915 			   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9916 		}
9917 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9918 		break;
9919 
9920 	case I40E_VSI_VMDQ2:
9921 		ctxt.pf_num = hw->pf_id;
9922 		ctxt.vf_num = 0;
9923 		ctxt.uplink_seid = vsi->uplink_seid;
9924 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9925 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9926 
9927 		/* This VSI is connected to VEB so the switch_id
9928 		 * should be set to zero by default.
9929 		 */
9930 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9931 			ctxt.info.valid_sections |=
9932 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9933 			ctxt.info.switch_id =
9934 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9935 		}
9936 
9937 		/* Setup the VSI tx/rx queue map for TC0 only for now */
9938 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9939 		break;
9940 
9941 	case I40E_VSI_SRIOV:
9942 		ctxt.pf_num = hw->pf_id;
9943 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9944 		ctxt.uplink_seid = vsi->uplink_seid;
9945 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9946 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9947 
9948 		/* This VSI is connected to VEB so the switch_id
9949 		 * should be set to zero by default.
9950 		 */
9951 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9952 			ctxt.info.valid_sections |=
9953 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9954 			ctxt.info.switch_id =
9955 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9956 		}
9957 
9958 		if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9959 			ctxt.info.valid_sections |=
9960 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9961 			ctxt.info.queueing_opt_flags |=
9962 				(I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9963 				 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9964 		}
9965 
9966 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9967 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9968 		if (pf->vf[vsi->vf_id].spoofchk) {
9969 			ctxt.info.valid_sections |=
9970 				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9971 			ctxt.info.sec_flags |=
9972 				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9973 				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9974 		}
9975 		/* Setup the VSI tx/rx queue map for TC0 only for now */
9976 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9977 		break;
9978 
9979 	case I40E_VSI_IWARP:
9980 		/* send down message to iWARP */
9981 		break;
9982 
9983 	default:
9984 		return -ENODEV;
9985 	}
9986 
9987 	if (vsi->type != I40E_VSI_MAIN) {
9988 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9989 		if (ret) {
9990 			dev_info(&vsi->back->pdev->dev,
9991 				 "add vsi failed, err %s aq_err %s\n",
9992 				 i40e_stat_str(&pf->hw, ret),
9993 				 i40e_aq_str(&pf->hw,
9994 					     pf->hw.aq.asq_last_status));
9995 			ret = -ENOENT;
9996 			goto err;
9997 		}
9998 		vsi->info = ctxt.info;
9999 		vsi->info.valid_sections = 0;
10000 		vsi->seid = ctxt.seid;
10001 		vsi->id = ctxt.vsi_number;
10002 	}
10003 
10004 	vsi->active_filters = 0;
10005 	clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
10006 	spin_lock_bh(&vsi->mac_filter_hash_lock);
10007 	/* If macvlan filters already exist, force them to get loaded */
10008 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
10009 		f->state = I40E_FILTER_NEW;
10010 		f_count++;
10011 	}
10012 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
10013 
10014 	if (f_count) {
10015 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
10016 		pf->flags |= I40E_FLAG_FILTER_SYNC;
10017 	}
10018 
10019 	/* Update VSI BW information */
10020 	ret = i40e_vsi_get_bw_info(vsi);
10021 	if (ret) {
10022 		dev_info(&pf->pdev->dev,
10023 			 "couldn't get vsi bw info, err %s aq_err %s\n",
10024 			 i40e_stat_str(&pf->hw, ret),
10025 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10026 		/* VSI is already added so not tearing that up */
10027 		ret = 0;
10028 	}
10029 
10030 err:
10031 	return ret;
10032 }
10033 
10034 /**
10035  * i40e_vsi_release - Delete a VSI and free its resources
10036  * @vsi: the VSI being removed
10037  *
10038  * Returns 0 on success or < 0 on error
10039  **/
10040 int i40e_vsi_release(struct i40e_vsi *vsi)
10041 {
10042 	struct i40e_mac_filter *f;
10043 	struct hlist_node *h;
10044 	struct i40e_veb *veb = NULL;
10045 	struct i40e_pf *pf;
10046 	u16 uplink_seid;
10047 	int i, n, bkt;
10048 
10049 	pf = vsi->back;
10050 
10051 	/* release of a VEB-owner or last VSI is not allowed */
10052 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
10053 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
10054 			 vsi->seid, vsi->uplink_seid);
10055 		return -ENODEV;
10056 	}
10057 	if (vsi == pf->vsi[pf->lan_vsi] &&
10058 	    !test_bit(__I40E_DOWN, pf->state)) {
10059 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
10060 		return -ENODEV;
10061 	}
10062 
10063 	uplink_seid = vsi->uplink_seid;
10064 	if (vsi->type != I40E_VSI_SRIOV) {
10065 		if (vsi->netdev_registered) {
10066 			vsi->netdev_registered = false;
10067 			if (vsi->netdev) {
10068 				/* results in a call to i40e_close() */
10069 				unregister_netdev(vsi->netdev);
10070 			}
10071 		} else {
10072 			i40e_vsi_close(vsi);
10073 		}
10074 		i40e_vsi_disable_irq(vsi);
10075 	}
10076 
10077 	spin_lock_bh(&vsi->mac_filter_hash_lock);
10078 
10079 	/* clear the sync flag on all filters */
10080 	if (vsi->netdev) {
10081 		__dev_uc_unsync(vsi->netdev, NULL);
10082 		__dev_mc_unsync(vsi->netdev, NULL);
10083 	}
10084 
10085 	/* make sure any remaining filters are marked for deletion */
10086 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
10087 		__i40e_del_filter(vsi, f);
10088 
10089 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
10090 
10091 	i40e_sync_vsi_filters(vsi);
10092 
10093 	i40e_vsi_delete(vsi);
10094 	i40e_vsi_free_q_vectors(vsi);
10095 	if (vsi->netdev) {
10096 		free_netdev(vsi->netdev);
10097 		vsi->netdev = NULL;
10098 	}
10099 	i40e_vsi_clear_rings(vsi);
10100 	i40e_vsi_clear(vsi);
10101 
10102 	/* If this was the last thing on the VEB, except for the
10103 	 * controlling VSI, remove the VEB, which puts the controlling
10104 	 * VSI onto the next level down in the switch.
10105 	 *
10106 	 * Well, okay, there's one more exception here: don't remove
10107 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
10108 	 * from up the network stack.
10109 	 */
10110 	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
10111 		if (pf->vsi[i] &&
10112 		    pf->vsi[i]->uplink_seid == uplink_seid &&
10113 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10114 			n++;      /* count the VSIs */
10115 		}
10116 	}
10117 	for (i = 0; i < I40E_MAX_VEB; i++) {
10118 		if (!pf->veb[i])
10119 			continue;
10120 		if (pf->veb[i]->uplink_seid == uplink_seid)
10121 			n++;     /* count the VEBs */
10122 		if (pf->veb[i]->seid == uplink_seid)
10123 			veb = pf->veb[i];
10124 	}
10125 	if (n == 0 && veb && veb->uplink_seid != 0)
10126 		i40e_veb_release(veb);
10127 
10128 	return 0;
10129 }
10130 
10131 /**
10132  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
10133  * @vsi: ptr to the VSI
10134  *
10135  * This should only be called after i40e_vsi_mem_alloc() which allocates the
10136  * corresponding SW VSI structure and initializes num_queue_pairs for the
10137  * newly allocated VSI.
10138  *
10139  * Returns 0 on success or negative on failure
10140  **/
10141 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
10142 {
10143 	int ret = -ENOENT;
10144 	struct i40e_pf *pf = vsi->back;
10145 
10146 	if (vsi->q_vectors[0]) {
10147 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
10148 			 vsi->seid);
10149 		return -EEXIST;
10150 	}
10151 
10152 	if (vsi->base_vector) {
10153 		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
10154 			 vsi->seid, vsi->base_vector);
10155 		return -EEXIST;
10156 	}
10157 
10158 	ret = i40e_vsi_alloc_q_vectors(vsi);
10159 	if (ret) {
10160 		dev_info(&pf->pdev->dev,
10161 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
10162 			 vsi->num_q_vectors, vsi->seid, ret);
10163 		vsi->num_q_vectors = 0;
10164 		goto vector_setup_out;
10165 	}
10166 
10167 	/* In Legacy mode, we do not have to get any other vector since we
10168 	 * piggyback on the misc/ICR0 for queue interrupts.
10169 	*/
10170 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
10171 		return ret;
10172 	if (vsi->num_q_vectors)
10173 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
10174 						 vsi->num_q_vectors, vsi->idx);
10175 	if (vsi->base_vector < 0) {
10176 		dev_info(&pf->pdev->dev,
10177 			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
10178 			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
10179 		i40e_vsi_free_q_vectors(vsi);
10180 		ret = -ENOENT;
10181 		goto vector_setup_out;
10182 	}
10183 
10184 vector_setup_out:
10185 	return ret;
10186 }
10187 
10188 /**
10189  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
10190  * @vsi: pointer to the vsi.
10191  *
10192  * This re-allocates a vsi's queue resources.
10193  *
10194  * Returns pointer to the successfully allocated and configured VSI sw struct
10195  * on success, otherwise returns NULL on failure.
10196  **/
10197 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
10198 {
10199 	u16 alloc_queue_pairs;
10200 	struct i40e_pf *pf;
10201 	u8 enabled_tc;
10202 	int ret;
10203 
10204 	if (!vsi)
10205 		return NULL;
10206 
10207 	pf = vsi->back;
10208 
10209 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
10210 	i40e_vsi_clear_rings(vsi);
10211 
10212 	i40e_vsi_free_arrays(vsi, false);
10213 	i40e_set_num_rings_in_vsi(vsi);
10214 	ret = i40e_vsi_alloc_arrays(vsi, false);
10215 	if (ret)
10216 		goto err_vsi;
10217 
10218 	alloc_queue_pairs = vsi->alloc_queue_pairs *
10219 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
10220 
10221 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
10222 	if (ret < 0) {
10223 		dev_info(&pf->pdev->dev,
10224 			 "failed to get tracking for %d queues for VSI %d err %d\n",
10225 			 alloc_queue_pairs, vsi->seid, ret);
10226 		goto err_vsi;
10227 	}
10228 	vsi->base_queue = ret;
10229 
10230 	/* Update the FW view of the VSI. Force a reset of TC and queue
10231 	 * layout configurations.
10232 	 */
10233 	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10234 	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10235 	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10236 	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10237 	if (vsi->type == I40E_VSI_MAIN)
10238 		i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
10239 
10240 	/* assign it some queues */
10241 	ret = i40e_alloc_rings(vsi);
10242 	if (ret)
10243 		goto err_rings;
10244 
10245 	/* map all of the rings to the q_vectors */
10246 	i40e_vsi_map_rings_to_vectors(vsi);
10247 	return vsi;
10248 
10249 err_rings:
10250 	i40e_vsi_free_q_vectors(vsi);
10251 	if (vsi->netdev_registered) {
10252 		vsi->netdev_registered = false;
10253 		unregister_netdev(vsi->netdev);
10254 		free_netdev(vsi->netdev);
10255 		vsi->netdev = NULL;
10256 	}
10257 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
10258 err_vsi:
10259 	i40e_vsi_clear(vsi);
10260 	return NULL;
10261 }
10262 
10263 /**
10264  * i40e_vsi_setup - Set up a VSI by a given type
10265  * @pf: board private structure
10266  * @type: VSI type
10267  * @uplink_seid: the switch element to link to
10268  * @param1: usage depends upon VSI type. For VF types, indicates VF id
10269  *
10270  * This allocates the sw VSI structure and its queue resources, then add a VSI
10271  * to the identified VEB.
10272  *
10273  * Returns pointer to the successfully allocated and configure VSI sw struct on
10274  * success, otherwise returns NULL on failure.
10275  **/
10276 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
10277 				u16 uplink_seid, u32 param1)
10278 {
10279 	struct i40e_vsi *vsi = NULL;
10280 	struct i40e_veb *veb = NULL;
10281 	u16 alloc_queue_pairs;
10282 	int ret, i;
10283 	int v_idx;
10284 
10285 	/* The requested uplink_seid must be either
10286 	 *     - the PF's port seid
10287 	 *              no VEB is needed because this is the PF
10288 	 *              or this is a Flow Director special case VSI
10289 	 *     - seid of an existing VEB
10290 	 *     - seid of a VSI that owns an existing VEB
10291 	 *     - seid of a VSI that doesn't own a VEB
10292 	 *              a new VEB is created and the VSI becomes the owner
10293 	 *     - seid of the PF VSI, which is what creates the first VEB
10294 	 *              this is a special case of the previous
10295 	 *
10296 	 * Find which uplink_seid we were given and create a new VEB if needed
10297 	 */
10298 	for (i = 0; i < I40E_MAX_VEB; i++) {
10299 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
10300 			veb = pf->veb[i];
10301 			break;
10302 		}
10303 	}
10304 
10305 	if (!veb && uplink_seid != pf->mac_seid) {
10306 
10307 		for (i = 0; i < pf->num_alloc_vsi; i++) {
10308 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
10309 				vsi = pf->vsi[i];
10310 				break;
10311 			}
10312 		}
10313 		if (!vsi) {
10314 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
10315 				 uplink_seid);
10316 			return NULL;
10317 		}
10318 
10319 		if (vsi->uplink_seid == pf->mac_seid)
10320 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
10321 					     vsi->tc_config.enabled_tc);
10322 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
10323 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
10324 					     vsi->tc_config.enabled_tc);
10325 		if (veb) {
10326 			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
10327 				dev_info(&vsi->back->pdev->dev,
10328 					 "New VSI creation error, uplink seid of LAN VSI expected.\n");
10329 				return NULL;
10330 			}
10331 			/* We come up by default in VEPA mode if SRIOV is not
10332 			 * already enabled, in which case we can't force VEPA
10333 			 * mode.
10334 			 */
10335 			if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
10336 				veb->bridge_mode = BRIDGE_MODE_VEPA;
10337 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
10338 			}
10339 			i40e_config_bridge_mode(veb);
10340 		}
10341 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
10342 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
10343 				veb = pf->veb[i];
10344 		}
10345 		if (!veb) {
10346 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
10347 			return NULL;
10348 		}
10349 
10350 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10351 		uplink_seid = veb->seid;
10352 	}
10353 
10354 	/* get vsi sw struct */
10355 	v_idx = i40e_vsi_mem_alloc(pf, type);
10356 	if (v_idx < 0)
10357 		goto err_alloc;
10358 	vsi = pf->vsi[v_idx];
10359 	if (!vsi)
10360 		goto err_alloc;
10361 	vsi->type = type;
10362 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
10363 
10364 	if (type == I40E_VSI_MAIN)
10365 		pf->lan_vsi = v_idx;
10366 	else if (type == I40E_VSI_SRIOV)
10367 		vsi->vf_id = param1;
10368 	/* assign it some queues */
10369 	alloc_queue_pairs = vsi->alloc_queue_pairs *
10370 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
10371 
10372 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
10373 	if (ret < 0) {
10374 		dev_info(&pf->pdev->dev,
10375 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
10376 			 alloc_queue_pairs, vsi->seid, ret);
10377 		goto err_vsi;
10378 	}
10379 	vsi->base_queue = ret;
10380 
10381 	/* get a VSI from the hardware */
10382 	vsi->uplink_seid = uplink_seid;
10383 	ret = i40e_add_vsi(vsi);
10384 	if (ret)
10385 		goto err_vsi;
10386 
10387 	switch (vsi->type) {
10388 	/* setup the netdev if needed */
10389 	case I40E_VSI_MAIN:
10390 		/* Apply relevant filters if a platform-specific mac
10391 		 * address was selected.
10392 		 */
10393 		if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
10394 			ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
10395 			if (ret) {
10396 				dev_warn(&pf->pdev->dev,
10397 					 "could not set up macaddr; err %d\n",
10398 					 ret);
10399 			}
10400 		}
10401 	case I40E_VSI_VMDQ2:
10402 		ret = i40e_config_netdev(vsi);
10403 		if (ret)
10404 			goto err_netdev;
10405 		ret = register_netdev(vsi->netdev);
10406 		if (ret)
10407 			goto err_netdev;
10408 		vsi->netdev_registered = true;
10409 		netif_carrier_off(vsi->netdev);
10410 #ifdef CONFIG_I40E_DCB
10411 		/* Setup DCB netlink interface */
10412 		i40e_dcbnl_setup(vsi);
10413 #endif /* CONFIG_I40E_DCB */
10414 		/* fall through */
10415 
10416 	case I40E_VSI_FDIR:
10417 		/* set up vectors and rings if needed */
10418 		ret = i40e_vsi_setup_vectors(vsi);
10419 		if (ret)
10420 			goto err_msix;
10421 
10422 		ret = i40e_alloc_rings(vsi);
10423 		if (ret)
10424 			goto err_rings;
10425 
10426 		/* map all of the rings to the q_vectors */
10427 		i40e_vsi_map_rings_to_vectors(vsi);
10428 
10429 		i40e_vsi_reset_stats(vsi);
10430 		break;
10431 
10432 	default:
10433 		/* no netdev or rings for the other VSI types */
10434 		break;
10435 	}
10436 
10437 	if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
10438 	    (vsi->type == I40E_VSI_VMDQ2)) {
10439 		ret = i40e_vsi_config_rss(vsi);
10440 	}
10441 	return vsi;
10442 
10443 err_rings:
10444 	i40e_vsi_free_q_vectors(vsi);
10445 err_msix:
10446 	if (vsi->netdev_registered) {
10447 		vsi->netdev_registered = false;
10448 		unregister_netdev(vsi->netdev);
10449 		free_netdev(vsi->netdev);
10450 		vsi->netdev = NULL;
10451 	}
10452 err_netdev:
10453 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
10454 err_vsi:
10455 	i40e_vsi_clear(vsi);
10456 err_alloc:
10457 	return NULL;
10458 }
10459 
10460 /**
10461  * i40e_veb_get_bw_info - Query VEB BW information
10462  * @veb: the veb to query
10463  *
10464  * Query the Tx scheduler BW configuration data for given VEB
10465  **/
10466 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
10467 {
10468 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
10469 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
10470 	struct i40e_pf *pf = veb->pf;
10471 	struct i40e_hw *hw = &pf->hw;
10472 	u32 tc_bw_max;
10473 	int ret = 0;
10474 	int i;
10475 
10476 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
10477 						  &bw_data, NULL);
10478 	if (ret) {
10479 		dev_info(&pf->pdev->dev,
10480 			 "query veb bw config failed, err %s aq_err %s\n",
10481 			 i40e_stat_str(&pf->hw, ret),
10482 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10483 		goto out;
10484 	}
10485 
10486 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
10487 						   &ets_data, NULL);
10488 	if (ret) {
10489 		dev_info(&pf->pdev->dev,
10490 			 "query veb bw ets config failed, err %s aq_err %s\n",
10491 			 i40e_stat_str(&pf->hw, ret),
10492 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10493 		goto out;
10494 	}
10495 
10496 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
10497 	veb->bw_max_quanta = ets_data.tc_bw_max;
10498 	veb->is_abs_credits = bw_data.absolute_credits_enable;
10499 	veb->enabled_tc = ets_data.tc_valid_bits;
10500 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
10501 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
10502 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
10503 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
10504 		veb->bw_tc_limit_credits[i] =
10505 					le16_to_cpu(bw_data.tc_bw_limits[i]);
10506 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
10507 	}
10508 
10509 out:
10510 	return ret;
10511 }
10512 
10513 /**
10514  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
10515  * @pf: board private structure
10516  *
10517  * On error: returns error code (negative)
10518  * On success: returns vsi index in PF (positive)
10519  **/
10520 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
10521 {
10522 	int ret = -ENOENT;
10523 	struct i40e_veb *veb;
10524 	int i;
10525 
10526 	/* Need to protect the allocation of switch elements at the PF level */
10527 	mutex_lock(&pf->switch_mutex);
10528 
10529 	/* VEB list may be fragmented if VEB creation/destruction has
10530 	 * been happening.  We can afford to do a quick scan to look
10531 	 * for any free slots in the list.
10532 	 *
10533 	 * find next empty veb slot, looping back around if necessary
10534 	 */
10535 	i = 0;
10536 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
10537 		i++;
10538 	if (i >= I40E_MAX_VEB) {
10539 		ret = -ENOMEM;
10540 		goto err_alloc_veb;  /* out of VEB slots! */
10541 	}
10542 
10543 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
10544 	if (!veb) {
10545 		ret = -ENOMEM;
10546 		goto err_alloc_veb;
10547 	}
10548 	veb->pf = pf;
10549 	veb->idx = i;
10550 	veb->enabled_tc = 1;
10551 
10552 	pf->veb[i] = veb;
10553 	ret = i;
10554 err_alloc_veb:
10555 	mutex_unlock(&pf->switch_mutex);
10556 	return ret;
10557 }
10558 
10559 /**
10560  * i40e_switch_branch_release - Delete a branch of the switch tree
10561  * @branch: where to start deleting
10562  *
10563  * This uses recursion to find the tips of the branch to be
10564  * removed, deleting until we get back to and can delete this VEB.
10565  **/
10566 static void i40e_switch_branch_release(struct i40e_veb *branch)
10567 {
10568 	struct i40e_pf *pf = branch->pf;
10569 	u16 branch_seid = branch->seid;
10570 	u16 veb_idx = branch->idx;
10571 	int i;
10572 
10573 	/* release any VEBs on this VEB - RECURSION */
10574 	for (i = 0; i < I40E_MAX_VEB; i++) {
10575 		if (!pf->veb[i])
10576 			continue;
10577 		if (pf->veb[i]->uplink_seid == branch->seid)
10578 			i40e_switch_branch_release(pf->veb[i]);
10579 	}
10580 
10581 	/* Release the VSIs on this VEB, but not the owner VSI.
10582 	 *
10583 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10584 	 *       the VEB itself, so don't use (*branch) after this loop.
10585 	 */
10586 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10587 		if (!pf->vsi[i])
10588 			continue;
10589 		if (pf->vsi[i]->uplink_seid == branch_seid &&
10590 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10591 			i40e_vsi_release(pf->vsi[i]);
10592 		}
10593 	}
10594 
10595 	/* There's one corner case where the VEB might not have been
10596 	 * removed, so double check it here and remove it if needed.
10597 	 * This case happens if the veb was created from the debugfs
10598 	 * commands and no VSIs were added to it.
10599 	 */
10600 	if (pf->veb[veb_idx])
10601 		i40e_veb_release(pf->veb[veb_idx]);
10602 }
10603 
10604 /**
10605  * i40e_veb_clear - remove veb struct
10606  * @veb: the veb to remove
10607  **/
10608 static void i40e_veb_clear(struct i40e_veb *veb)
10609 {
10610 	if (!veb)
10611 		return;
10612 
10613 	if (veb->pf) {
10614 		struct i40e_pf *pf = veb->pf;
10615 
10616 		mutex_lock(&pf->switch_mutex);
10617 		if (pf->veb[veb->idx] == veb)
10618 			pf->veb[veb->idx] = NULL;
10619 		mutex_unlock(&pf->switch_mutex);
10620 	}
10621 
10622 	kfree(veb);
10623 }
10624 
10625 /**
10626  * i40e_veb_release - Delete a VEB and free its resources
10627  * @veb: the VEB being removed
10628  **/
10629 void i40e_veb_release(struct i40e_veb *veb)
10630 {
10631 	struct i40e_vsi *vsi = NULL;
10632 	struct i40e_pf *pf;
10633 	int i, n = 0;
10634 
10635 	pf = veb->pf;
10636 
10637 	/* find the remaining VSI and check for extras */
10638 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10639 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10640 			n++;
10641 			vsi = pf->vsi[i];
10642 		}
10643 	}
10644 	if (n != 1) {
10645 		dev_info(&pf->pdev->dev,
10646 			 "can't remove VEB %d with %d VSIs left\n",
10647 			 veb->seid, n);
10648 		return;
10649 	}
10650 
10651 	/* move the remaining VSI to uplink veb */
10652 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10653 	if (veb->uplink_seid) {
10654 		vsi->uplink_seid = veb->uplink_seid;
10655 		if (veb->uplink_seid == pf->mac_seid)
10656 			vsi->veb_idx = I40E_NO_VEB;
10657 		else
10658 			vsi->veb_idx = veb->veb_idx;
10659 	} else {
10660 		/* floating VEB */
10661 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10662 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10663 	}
10664 
10665 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10666 	i40e_veb_clear(veb);
10667 }
10668 
10669 /**
10670  * i40e_add_veb - create the VEB in the switch
10671  * @veb: the VEB to be instantiated
10672  * @vsi: the controlling VSI
10673  **/
10674 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10675 {
10676 	struct i40e_pf *pf = veb->pf;
10677 	bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10678 	int ret;
10679 
10680 	ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10681 			      veb->enabled_tc, false,
10682 			      &veb->seid, enable_stats, NULL);
10683 
10684 	/* get a VEB from the hardware */
10685 	if (ret) {
10686 		dev_info(&pf->pdev->dev,
10687 			 "couldn't add VEB, err %s aq_err %s\n",
10688 			 i40e_stat_str(&pf->hw, ret),
10689 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10690 		return -EPERM;
10691 	}
10692 
10693 	/* get statistics counter */
10694 	ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10695 					 &veb->stats_idx, NULL, NULL, NULL);
10696 	if (ret) {
10697 		dev_info(&pf->pdev->dev,
10698 			 "couldn't get VEB statistics idx, err %s aq_err %s\n",
10699 			 i40e_stat_str(&pf->hw, ret),
10700 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10701 		return -EPERM;
10702 	}
10703 	ret = i40e_veb_get_bw_info(veb);
10704 	if (ret) {
10705 		dev_info(&pf->pdev->dev,
10706 			 "couldn't get VEB bw info, err %s aq_err %s\n",
10707 			 i40e_stat_str(&pf->hw, ret),
10708 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10709 		i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10710 		return -ENOENT;
10711 	}
10712 
10713 	vsi->uplink_seid = veb->seid;
10714 	vsi->veb_idx = veb->idx;
10715 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10716 
10717 	return 0;
10718 }
10719 
10720 /**
10721  * i40e_veb_setup - Set up a VEB
10722  * @pf: board private structure
10723  * @flags: VEB setup flags
10724  * @uplink_seid: the switch element to link to
10725  * @vsi_seid: the initial VSI seid
10726  * @enabled_tc: Enabled TC bit-map
10727  *
10728  * This allocates the sw VEB structure and links it into the switch
10729  * It is possible and legal for this to be a duplicate of an already
10730  * existing VEB.  It is also possible for both uplink and vsi seids
10731  * to be zero, in order to create a floating VEB.
10732  *
10733  * Returns pointer to the successfully allocated VEB sw struct on
10734  * success, otherwise returns NULL on failure.
10735  **/
10736 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10737 				u16 uplink_seid, u16 vsi_seid,
10738 				u8 enabled_tc)
10739 {
10740 	struct i40e_veb *veb, *uplink_veb = NULL;
10741 	int vsi_idx, veb_idx;
10742 	int ret;
10743 
10744 	/* if one seid is 0, the other must be 0 to create a floating relay */
10745 	if ((uplink_seid == 0 || vsi_seid == 0) &&
10746 	    (uplink_seid + vsi_seid != 0)) {
10747 		dev_info(&pf->pdev->dev,
10748 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
10749 			 uplink_seid, vsi_seid);
10750 		return NULL;
10751 	}
10752 
10753 	/* make sure there is such a vsi and uplink */
10754 	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10755 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10756 			break;
10757 	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10758 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10759 			 vsi_seid);
10760 		return NULL;
10761 	}
10762 
10763 	if (uplink_seid && uplink_seid != pf->mac_seid) {
10764 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10765 			if (pf->veb[veb_idx] &&
10766 			    pf->veb[veb_idx]->seid == uplink_seid) {
10767 				uplink_veb = pf->veb[veb_idx];
10768 				break;
10769 			}
10770 		}
10771 		if (!uplink_veb) {
10772 			dev_info(&pf->pdev->dev,
10773 				 "uplink seid %d not found\n", uplink_seid);
10774 			return NULL;
10775 		}
10776 	}
10777 
10778 	/* get veb sw struct */
10779 	veb_idx = i40e_veb_mem_alloc(pf);
10780 	if (veb_idx < 0)
10781 		goto err_alloc;
10782 	veb = pf->veb[veb_idx];
10783 	veb->flags = flags;
10784 	veb->uplink_seid = uplink_seid;
10785 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10786 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10787 
10788 	/* create the VEB in the switch */
10789 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10790 	if (ret)
10791 		goto err_veb;
10792 	if (vsi_idx == pf->lan_vsi)
10793 		pf->lan_veb = veb->idx;
10794 
10795 	return veb;
10796 
10797 err_veb:
10798 	i40e_veb_clear(veb);
10799 err_alloc:
10800 	return NULL;
10801 }
10802 
10803 /**
10804  * i40e_setup_pf_switch_element - set PF vars based on switch type
10805  * @pf: board private structure
10806  * @ele: element we are building info from
10807  * @num_reported: total number of elements
10808  * @printconfig: should we print the contents
10809  *
10810  * helper function to assist in extracting a few useful SEID values.
10811  **/
10812 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10813 				struct i40e_aqc_switch_config_element_resp *ele,
10814 				u16 num_reported, bool printconfig)
10815 {
10816 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10817 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10818 	u8 element_type = ele->element_type;
10819 	u16 seid = le16_to_cpu(ele->seid);
10820 
10821 	if (printconfig)
10822 		dev_info(&pf->pdev->dev,
10823 			 "type=%d seid=%d uplink=%d downlink=%d\n",
10824 			 element_type, seid, uplink_seid, downlink_seid);
10825 
10826 	switch (element_type) {
10827 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
10828 		pf->mac_seid = seid;
10829 		break;
10830 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
10831 		/* Main VEB? */
10832 		if (uplink_seid != pf->mac_seid)
10833 			break;
10834 		if (pf->lan_veb == I40E_NO_VEB) {
10835 			int v;
10836 
10837 			/* find existing or else empty VEB */
10838 			for (v = 0; v < I40E_MAX_VEB; v++) {
10839 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10840 					pf->lan_veb = v;
10841 					break;
10842 				}
10843 			}
10844 			if (pf->lan_veb == I40E_NO_VEB) {
10845 				v = i40e_veb_mem_alloc(pf);
10846 				if (v < 0)
10847 					break;
10848 				pf->lan_veb = v;
10849 			}
10850 		}
10851 
10852 		pf->veb[pf->lan_veb]->seid = seid;
10853 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10854 		pf->veb[pf->lan_veb]->pf = pf;
10855 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10856 		break;
10857 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
10858 		if (num_reported != 1)
10859 			break;
10860 		/* This is immediately after a reset so we can assume this is
10861 		 * the PF's VSI
10862 		 */
10863 		pf->mac_seid = uplink_seid;
10864 		pf->pf_seid = downlink_seid;
10865 		pf->main_vsi_seid = seid;
10866 		if (printconfig)
10867 			dev_info(&pf->pdev->dev,
10868 				 "pf_seid=%d main_vsi_seid=%d\n",
10869 				 pf->pf_seid, pf->main_vsi_seid);
10870 		break;
10871 	case I40E_SWITCH_ELEMENT_TYPE_PF:
10872 	case I40E_SWITCH_ELEMENT_TYPE_VF:
10873 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
10874 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
10875 	case I40E_SWITCH_ELEMENT_TYPE_PE:
10876 	case I40E_SWITCH_ELEMENT_TYPE_PA:
10877 		/* ignore these for now */
10878 		break;
10879 	default:
10880 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10881 			 element_type, seid);
10882 		break;
10883 	}
10884 }
10885 
10886 /**
10887  * i40e_fetch_switch_configuration - Get switch config from firmware
10888  * @pf: board private structure
10889  * @printconfig: should we print the contents
10890  *
10891  * Get the current switch configuration from the device and
10892  * extract a few useful SEID values.
10893  **/
10894 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10895 {
10896 	struct i40e_aqc_get_switch_config_resp *sw_config;
10897 	u16 next_seid = 0;
10898 	int ret = 0;
10899 	u8 *aq_buf;
10900 	int i;
10901 
10902 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10903 	if (!aq_buf)
10904 		return -ENOMEM;
10905 
10906 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10907 	do {
10908 		u16 num_reported, num_total;
10909 
10910 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10911 						I40E_AQ_LARGE_BUF,
10912 						&next_seid, NULL);
10913 		if (ret) {
10914 			dev_info(&pf->pdev->dev,
10915 				 "get switch config failed err %s aq_err %s\n",
10916 				 i40e_stat_str(&pf->hw, ret),
10917 				 i40e_aq_str(&pf->hw,
10918 					     pf->hw.aq.asq_last_status));
10919 			kfree(aq_buf);
10920 			return -ENOENT;
10921 		}
10922 
10923 		num_reported = le16_to_cpu(sw_config->header.num_reported);
10924 		num_total = le16_to_cpu(sw_config->header.num_total);
10925 
10926 		if (printconfig)
10927 			dev_info(&pf->pdev->dev,
10928 				 "header: %d reported %d total\n",
10929 				 num_reported, num_total);
10930 
10931 		for (i = 0; i < num_reported; i++) {
10932 			struct i40e_aqc_switch_config_element_resp *ele =
10933 				&sw_config->element[i];
10934 
10935 			i40e_setup_pf_switch_element(pf, ele, num_reported,
10936 						     printconfig);
10937 		}
10938 	} while (next_seid != 0);
10939 
10940 	kfree(aq_buf);
10941 	return ret;
10942 }
10943 
10944 /**
10945  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10946  * @pf: board private structure
10947  * @reinit: if the Main VSI needs to re-initialized.
10948  *
10949  * Returns 0 on success, negative value on failure
10950  **/
10951 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10952 {
10953 	u16 flags = 0;
10954 	int ret;
10955 
10956 	/* find out what's out there already */
10957 	ret = i40e_fetch_switch_configuration(pf, false);
10958 	if (ret) {
10959 		dev_info(&pf->pdev->dev,
10960 			 "couldn't fetch switch config, err %s aq_err %s\n",
10961 			 i40e_stat_str(&pf->hw, ret),
10962 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10963 		return ret;
10964 	}
10965 	i40e_pf_reset_stats(pf);
10966 
10967 	/* set the switch config bit for the whole device to
10968 	 * support limited promisc or true promisc
10969 	 * when user requests promisc. The default is limited
10970 	 * promisc.
10971 	*/
10972 
10973 	if ((pf->hw.pf_id == 0) &&
10974 	    !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10975 		flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10976 
10977 	if (pf->hw.pf_id == 0) {
10978 		u16 valid_flags;
10979 
10980 		valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10981 		ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10982 						NULL);
10983 		if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10984 			dev_info(&pf->pdev->dev,
10985 				 "couldn't set switch config bits, err %s aq_err %s\n",
10986 				 i40e_stat_str(&pf->hw, ret),
10987 				 i40e_aq_str(&pf->hw,
10988 					     pf->hw.aq.asq_last_status));
10989 			/* not a fatal problem, just keep going */
10990 		}
10991 	}
10992 
10993 	/* first time setup */
10994 	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10995 		struct i40e_vsi *vsi = NULL;
10996 		u16 uplink_seid;
10997 
10998 		/* Set up the PF VSI associated with the PF's main VSI
10999 		 * that is already in the HW switch
11000 		 */
11001 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
11002 			uplink_seid = pf->veb[pf->lan_veb]->seid;
11003 		else
11004 			uplink_seid = pf->mac_seid;
11005 		if (pf->lan_vsi == I40E_NO_VSI)
11006 			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
11007 		else if (reinit)
11008 			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
11009 		if (!vsi) {
11010 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
11011 			i40e_fdir_teardown(pf);
11012 			return -EAGAIN;
11013 		}
11014 	} else {
11015 		/* force a reset of TC and queue layout configurations */
11016 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
11017 
11018 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
11019 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
11020 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
11021 	}
11022 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
11023 
11024 	i40e_fdir_sb_setup(pf);
11025 
11026 	/* Setup static PF queue filter control settings */
11027 	ret = i40e_setup_pf_filter_control(pf);
11028 	if (ret) {
11029 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
11030 			 ret);
11031 		/* Failure here should not stop continuing other steps */
11032 	}
11033 
11034 	/* enable RSS in the HW, even for only one queue, as the stack can use
11035 	 * the hash
11036 	 */
11037 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
11038 		i40e_pf_config_rss(pf);
11039 
11040 	/* fill in link information and enable LSE reporting */
11041 	i40e_link_event(pf);
11042 
11043 	/* Initialize user-specific link properties */
11044 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
11045 				  I40E_AQ_AN_COMPLETED) ? true : false);
11046 
11047 	i40e_ptp_init(pf);
11048 
11049 	/* repopulate tunnel port filters */
11050 	i40e_sync_udp_filters(pf);
11051 
11052 	return ret;
11053 }
11054 
11055 /**
11056  * i40e_determine_queue_usage - Work out queue distribution
11057  * @pf: board private structure
11058  **/
11059 static void i40e_determine_queue_usage(struct i40e_pf *pf)
11060 {
11061 	int queues_left;
11062 
11063 	pf->num_lan_qps = 0;
11064 
11065 	/* Find the max queues to be put into basic use.  We'll always be
11066 	 * using TC0, whether or not DCB is running, and TC0 will get the
11067 	 * big RSS set.
11068 	 */
11069 	queues_left = pf->hw.func_caps.num_tx_qp;
11070 
11071 	if ((queues_left == 1) ||
11072 	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
11073 		/* one qp for PF, no queues for anything else */
11074 		queues_left = 0;
11075 		pf->alloc_rss_size = pf->num_lan_qps = 1;
11076 
11077 		/* make sure all the fancies are disabled */
11078 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
11079 			       I40E_FLAG_IWARP_ENABLED	|
11080 			       I40E_FLAG_FD_SB_ENABLED	|
11081 			       I40E_FLAG_FD_ATR_ENABLED	|
11082 			       I40E_FLAG_DCB_CAPABLE	|
11083 			       I40E_FLAG_DCB_ENABLED	|
11084 			       I40E_FLAG_SRIOV_ENABLED	|
11085 			       I40E_FLAG_VMDQ_ENABLED);
11086 	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
11087 				  I40E_FLAG_FD_SB_ENABLED |
11088 				  I40E_FLAG_FD_ATR_ENABLED |
11089 				  I40E_FLAG_DCB_CAPABLE))) {
11090 		/* one qp for PF */
11091 		pf->alloc_rss_size = pf->num_lan_qps = 1;
11092 		queues_left -= pf->num_lan_qps;
11093 
11094 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
11095 			       I40E_FLAG_IWARP_ENABLED	|
11096 			       I40E_FLAG_FD_SB_ENABLED	|
11097 			       I40E_FLAG_FD_ATR_ENABLED	|
11098 			       I40E_FLAG_DCB_ENABLED	|
11099 			       I40E_FLAG_VMDQ_ENABLED);
11100 	} else {
11101 		/* Not enough queues for all TCs */
11102 		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
11103 		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
11104 			pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
11105 					I40E_FLAG_DCB_ENABLED);
11106 			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
11107 		}
11108 		pf->num_lan_qps = max_t(int, pf->rss_size_max,
11109 					num_online_cpus());
11110 		pf->num_lan_qps = min_t(int, pf->num_lan_qps,
11111 					pf->hw.func_caps.num_tx_qp);
11112 
11113 		queues_left -= pf->num_lan_qps;
11114 	}
11115 
11116 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11117 		if (queues_left > 1) {
11118 			queues_left -= 1; /* save 1 queue for FD */
11119 		} else {
11120 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
11121 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
11122 		}
11123 	}
11124 
11125 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11126 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
11127 		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
11128 					(queues_left / pf->num_vf_qps));
11129 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
11130 	}
11131 
11132 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
11133 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
11134 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
11135 					  (queues_left / pf->num_vmdq_qps));
11136 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
11137 	}
11138 
11139 	pf->queues_left = queues_left;
11140 	dev_dbg(&pf->pdev->dev,
11141 		"qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
11142 		pf->hw.func_caps.num_tx_qp,
11143 		!!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
11144 		pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
11145 		pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
11146 		queues_left);
11147 }
11148 
11149 /**
11150  * i40e_setup_pf_filter_control - Setup PF static filter control
11151  * @pf: PF to be setup
11152  *
11153  * i40e_setup_pf_filter_control sets up a PF's initial filter control
11154  * settings. If PE/FCoE are enabled then it will also set the per PF
11155  * based filter sizes required for them. It also enables Flow director,
11156  * ethertype and macvlan type filter settings for the pf.
11157  *
11158  * Returns 0 on success, negative on failure
11159  **/
11160 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
11161 {
11162 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
11163 
11164 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
11165 
11166 	/* Flow Director is enabled */
11167 	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
11168 		settings->enable_fdir = true;
11169 
11170 	/* Ethtype and MACVLAN filters enabled for PF */
11171 	settings->enable_ethtype = true;
11172 	settings->enable_macvlan = true;
11173 
11174 	if (i40e_set_filter_control(&pf->hw, settings))
11175 		return -ENOENT;
11176 
11177 	return 0;
11178 }
11179 
11180 #define INFO_STRING_LEN 255
11181 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
11182 static void i40e_print_features(struct i40e_pf *pf)
11183 {
11184 	struct i40e_hw *hw = &pf->hw;
11185 	char *buf;
11186 	int i;
11187 
11188 	buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
11189 	if (!buf)
11190 		return;
11191 
11192 	i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
11193 #ifdef CONFIG_PCI_IOV
11194 	i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
11195 #endif
11196 	i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
11197 		      pf->hw.func_caps.num_vsis,
11198 		      pf->vsi[pf->lan_vsi]->num_queue_pairs);
11199 	if (pf->flags & I40E_FLAG_RSS_ENABLED)
11200 		i += snprintf(&buf[i], REMAIN(i), " RSS");
11201 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
11202 		i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
11203 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11204 		i += snprintf(&buf[i], REMAIN(i), " FD_SB");
11205 		i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
11206 	}
11207 	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
11208 		i += snprintf(&buf[i], REMAIN(i), " DCB");
11209 	i += snprintf(&buf[i], REMAIN(i), " VxLAN");
11210 	i += snprintf(&buf[i], REMAIN(i), " Geneve");
11211 	if (pf->flags & I40E_FLAG_PTP)
11212 		i += snprintf(&buf[i], REMAIN(i), " PTP");
11213 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
11214 		i += snprintf(&buf[i], REMAIN(i), " VEB");
11215 	else
11216 		i += snprintf(&buf[i], REMAIN(i), " VEPA");
11217 
11218 	dev_info(&pf->pdev->dev, "%s\n", buf);
11219 	kfree(buf);
11220 	WARN_ON(i > INFO_STRING_LEN);
11221 }
11222 
11223 /**
11224  * i40e_get_platform_mac_addr - get platform-specific MAC address
11225  * @pdev: PCI device information struct
11226  * @pf: board private structure
11227  *
11228  * Look up the MAC address for the device. First we'll try
11229  * eth_platform_get_mac_address, which will check Open Firmware, or arch
11230  * specific fallback. Otherwise, we'll default to the stored value in
11231  * firmware.
11232  **/
11233 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
11234 {
11235 	if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
11236 		i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
11237 }
11238 
11239 /**
11240  * i40e_probe - Device initialization routine
11241  * @pdev: PCI device information struct
11242  * @ent: entry in i40e_pci_tbl
11243  *
11244  * i40e_probe initializes a PF identified by a pci_dev structure.
11245  * The OS initialization, configuring of the PF private structure,
11246  * and a hardware reset occur.
11247  *
11248  * Returns 0 on success, negative on failure
11249  **/
11250 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
11251 {
11252 	struct i40e_aq_get_phy_abilities_resp abilities;
11253 	struct i40e_pf *pf;
11254 	struct i40e_hw *hw;
11255 	static u16 pfs_found;
11256 	u16 wol_nvm_bits;
11257 	u16 link_status;
11258 	int err;
11259 	u32 val;
11260 	u32 i;
11261 	u8 set_fc_aq_fail;
11262 
11263 	err = pci_enable_device_mem(pdev);
11264 	if (err)
11265 		return err;
11266 
11267 	/* set up for high or low dma */
11268 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
11269 	if (err) {
11270 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
11271 		if (err) {
11272 			dev_err(&pdev->dev,
11273 				"DMA configuration failed: 0x%x\n", err);
11274 			goto err_dma;
11275 		}
11276 	}
11277 
11278 	/* set up pci connections */
11279 	err = pci_request_mem_regions(pdev, i40e_driver_name);
11280 	if (err) {
11281 		dev_info(&pdev->dev,
11282 			 "pci_request_selected_regions failed %d\n", err);
11283 		goto err_pci_reg;
11284 	}
11285 
11286 	pci_enable_pcie_error_reporting(pdev);
11287 	pci_set_master(pdev);
11288 
11289 	/* Now that we have a PCI connection, we need to do the
11290 	 * low level device setup.  This is primarily setting up
11291 	 * the Admin Queue structures and then querying for the
11292 	 * device's current profile information.
11293 	 */
11294 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
11295 	if (!pf) {
11296 		err = -ENOMEM;
11297 		goto err_pf_alloc;
11298 	}
11299 	pf->next_vsi = 0;
11300 	pf->pdev = pdev;
11301 	set_bit(__I40E_DOWN, pf->state);
11302 
11303 	hw = &pf->hw;
11304 	hw->back = pf;
11305 
11306 	pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
11307 				I40E_MAX_CSR_SPACE);
11308 
11309 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
11310 	if (!hw->hw_addr) {
11311 		err = -EIO;
11312 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
11313 			 (unsigned int)pci_resource_start(pdev, 0),
11314 			 pf->ioremap_len, err);
11315 		goto err_ioremap;
11316 	}
11317 	hw->vendor_id = pdev->vendor;
11318 	hw->device_id = pdev->device;
11319 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
11320 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
11321 	hw->subsystem_device_id = pdev->subsystem_device;
11322 	hw->bus.device = PCI_SLOT(pdev->devfn);
11323 	hw->bus.func = PCI_FUNC(pdev->devfn);
11324 	hw->bus.bus_id = pdev->bus->number;
11325 	pf->instance = pfs_found;
11326 
11327 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
11328 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
11329 
11330 	/* set up the locks for the AQ, do this only once in probe
11331 	 * and destroy them only once in remove
11332 	 */
11333 	mutex_init(&hw->aq.asq_mutex);
11334 	mutex_init(&hw->aq.arq_mutex);
11335 
11336 	pf->msg_enable = netif_msg_init(debug,
11337 					NETIF_MSG_DRV |
11338 					NETIF_MSG_PROBE |
11339 					NETIF_MSG_LINK);
11340 	if (debug < -1)
11341 		pf->hw.debug_mask = debug;
11342 
11343 	/* do a special CORER for clearing PXE mode once at init */
11344 	if (hw->revision_id == 0 &&
11345 	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
11346 		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
11347 		i40e_flush(hw);
11348 		msleep(200);
11349 		pf->corer_count++;
11350 
11351 		i40e_clear_pxe_mode(hw);
11352 	}
11353 
11354 	/* Reset here to make sure all is clean and to define PF 'n' */
11355 	i40e_clear_hw(hw);
11356 	err = i40e_pf_reset(hw);
11357 	if (err) {
11358 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
11359 		goto err_pf_reset;
11360 	}
11361 	pf->pfr_count++;
11362 
11363 	hw->aq.num_arq_entries = I40E_AQ_LEN;
11364 	hw->aq.num_asq_entries = I40E_AQ_LEN;
11365 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11366 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11367 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
11368 
11369 	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
11370 		 "%s-%s:misc",
11371 		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
11372 
11373 	err = i40e_init_shared_code(hw);
11374 	if (err) {
11375 		dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
11376 			 err);
11377 		goto err_pf_reset;
11378 	}
11379 
11380 	/* set up a default setting for link flow control */
11381 	pf->hw.fc.requested_mode = I40E_FC_NONE;
11382 
11383 	err = i40e_init_adminq(hw);
11384 	if (err) {
11385 		if (err == I40E_ERR_FIRMWARE_API_VERSION)
11386 			dev_info(&pdev->dev,
11387 				 "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");
11388 		else
11389 			dev_info(&pdev->dev,
11390 				 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
11391 
11392 		goto err_pf_reset;
11393 	}
11394 	i40e_get_oem_version(hw);
11395 
11396 	/* provide nvm, fw, api versions */
11397 	dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
11398 		 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
11399 		 hw->aq.api_maj_ver, hw->aq.api_min_ver,
11400 		 i40e_nvm_version_str(hw));
11401 
11402 	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
11403 	    hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
11404 		dev_info(&pdev->dev,
11405 			 "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");
11406 	else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
11407 		 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
11408 		dev_info(&pdev->dev,
11409 			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
11410 
11411 	i40e_verify_eeprom(pf);
11412 
11413 	/* Rev 0 hardware was never productized */
11414 	if (hw->revision_id < 1)
11415 		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");
11416 
11417 	i40e_clear_pxe_mode(hw);
11418 	err = i40e_get_capabilities(pf);
11419 	if (err)
11420 		goto err_adminq_setup;
11421 
11422 	err = i40e_sw_init(pf);
11423 	if (err) {
11424 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
11425 		goto err_sw_init;
11426 	}
11427 
11428 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
11429 				hw->func_caps.num_rx_qp, 0, 0);
11430 	if (err) {
11431 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
11432 		goto err_init_lan_hmc;
11433 	}
11434 
11435 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
11436 	if (err) {
11437 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
11438 		err = -ENOENT;
11439 		goto err_configure_lan_hmc;
11440 	}
11441 
11442 	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
11443 	 * Ignore error return codes because if it was already disabled via
11444 	 * hardware settings this will fail
11445 	 */
11446 	if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
11447 		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
11448 		i40e_aq_stop_lldp(hw, true, NULL);
11449 	}
11450 
11451 	/* allow a platform config to override the HW addr */
11452 	i40e_get_platform_mac_addr(pdev, pf);
11453 
11454 	if (!is_valid_ether_addr(hw->mac.addr)) {
11455 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
11456 		err = -EIO;
11457 		goto err_mac_addr;
11458 	}
11459 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
11460 	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
11461 	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
11462 	if (is_valid_ether_addr(hw->mac.port_addr))
11463 		pf->flags |= I40E_FLAG_PORT_ID_VALID;
11464 
11465 	pci_set_drvdata(pdev, pf);
11466 	pci_save_state(pdev);
11467 #ifdef CONFIG_I40E_DCB
11468 	err = i40e_init_pf_dcb(pf);
11469 	if (err) {
11470 		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
11471 		pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
11472 		/* Continue without DCB enabled */
11473 	}
11474 #endif /* CONFIG_I40E_DCB */
11475 
11476 	/* set up periodic task facility */
11477 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
11478 	pf->service_timer_period = HZ;
11479 
11480 	INIT_WORK(&pf->service_task, i40e_service_task);
11481 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
11482 
11483 	/* NVM bit on means WoL disabled for the port */
11484 	i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
11485 	if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
11486 		pf->wol_en = false;
11487 	else
11488 		pf->wol_en = true;
11489 	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
11490 
11491 	/* set up the main switch operations */
11492 	i40e_determine_queue_usage(pf);
11493 	err = i40e_init_interrupt_scheme(pf);
11494 	if (err)
11495 		goto err_switch_setup;
11496 
11497 	/* The number of VSIs reported by the FW is the minimum guaranteed
11498 	 * to us; HW supports far more and we share the remaining pool with
11499 	 * the other PFs. We allocate space for more than the guarantee with
11500 	 * the understanding that we might not get them all later.
11501 	 */
11502 	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
11503 		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
11504 	else
11505 		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
11506 
11507 	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11508 	pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11509 			  GFP_KERNEL);
11510 	if (!pf->vsi) {
11511 		err = -ENOMEM;
11512 		goto err_switch_setup;
11513 	}
11514 
11515 #ifdef CONFIG_PCI_IOV
11516 	/* prep for VF support */
11517 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11518 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11519 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
11520 		if (pci_num_vf(pdev))
11521 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11522 	}
11523 #endif
11524 	err = i40e_setup_pf_switch(pf, false);
11525 	if (err) {
11526 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11527 		goto err_vsis;
11528 	}
11529 
11530 	/* Make sure flow control is set according to current settings */
11531 	err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11532 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11533 		dev_dbg(&pf->pdev->dev,
11534 			"Set fc with err %s aq_err %s on get_phy_cap\n",
11535 			i40e_stat_str(hw, err),
11536 			i40e_aq_str(hw, hw->aq.asq_last_status));
11537 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11538 		dev_dbg(&pf->pdev->dev,
11539 			"Set fc with err %s aq_err %s on set_phy_config\n",
11540 			i40e_stat_str(hw, err),
11541 			i40e_aq_str(hw, hw->aq.asq_last_status));
11542 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11543 		dev_dbg(&pf->pdev->dev,
11544 			"Set fc with err %s aq_err %s on get_link_info\n",
11545 			i40e_stat_str(hw, err),
11546 			i40e_aq_str(hw, hw->aq.asq_last_status));
11547 
11548 	/* if FDIR VSI was set up, start it now */
11549 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11550 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11551 			i40e_vsi_open(pf->vsi[i]);
11552 			break;
11553 		}
11554 	}
11555 
11556 	/* The driver only wants link up/down and module qualification
11557 	 * reports from firmware.  Note the negative logic.
11558 	 */
11559 	err = i40e_aq_set_phy_int_mask(&pf->hw,
11560 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
11561 					 I40E_AQ_EVENT_MEDIA_NA |
11562 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11563 	if (err)
11564 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11565 			 i40e_stat_str(&pf->hw, err),
11566 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11567 
11568 	/* Reconfigure hardware for allowing smaller MSS in the case
11569 	 * of TSO, so that we avoid the MDD being fired and causing
11570 	 * a reset in the case of small MSS+TSO.
11571 	 */
11572 	val = rd32(hw, I40E_REG_MSS);
11573 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11574 		val &= ~I40E_REG_MSS_MIN_MASK;
11575 		val |= I40E_64BYTE_MSS;
11576 		wr32(hw, I40E_REG_MSS, val);
11577 	}
11578 
11579 	if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11580 		msleep(75);
11581 		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11582 		if (err)
11583 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11584 				 i40e_stat_str(&pf->hw, err),
11585 				 i40e_aq_str(&pf->hw,
11586 					     pf->hw.aq.asq_last_status));
11587 	}
11588 	/* The main driver is (mostly) up and happy. We need to set this state
11589 	 * before setting up the misc vector or we get a race and the vector
11590 	 * ends up disabled forever.
11591 	 */
11592 	clear_bit(__I40E_DOWN, pf->state);
11593 
11594 	/* In case of MSIX we are going to setup the misc vector right here
11595 	 * to handle admin queue events etc. In case of legacy and MSI
11596 	 * the misc functionality and queue processing is combined in
11597 	 * the same vector and that gets setup at open.
11598 	 */
11599 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11600 		err = i40e_setup_misc_vector(pf);
11601 		if (err) {
11602 			dev_info(&pdev->dev,
11603 				 "setup of misc vector failed: %d\n", err);
11604 			goto err_vsis;
11605 		}
11606 	}
11607 
11608 #ifdef CONFIG_PCI_IOV
11609 	/* prep for VF support */
11610 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11611 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11612 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
11613 		/* disable link interrupts for VFs */
11614 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11615 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11616 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11617 		i40e_flush(hw);
11618 
11619 		if (pci_num_vf(pdev)) {
11620 			dev_info(&pdev->dev,
11621 				 "Active VFs found, allocating resources.\n");
11622 			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11623 			if (err)
11624 				dev_info(&pdev->dev,
11625 					 "Error %d allocating resources for existing VFs\n",
11626 					 err);
11627 		}
11628 	}
11629 #endif /* CONFIG_PCI_IOV */
11630 
11631 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11632 		pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11633 						      pf->num_iwarp_msix,
11634 						      I40E_IWARP_IRQ_PILE_ID);
11635 		if (pf->iwarp_base_vector < 0) {
11636 			dev_info(&pdev->dev,
11637 				 "failed to get tracking for %d vectors for IWARP err=%d\n",
11638 				 pf->num_iwarp_msix, pf->iwarp_base_vector);
11639 			pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11640 		}
11641 	}
11642 
11643 	i40e_dbg_pf_init(pf);
11644 
11645 	/* tell the firmware that we're starting */
11646 	i40e_send_version(pf);
11647 
11648 	/* since everything's happy, start the service_task timer */
11649 	mod_timer(&pf->service_timer,
11650 		  round_jiffies(jiffies + pf->service_timer_period));
11651 
11652 	/* add this PF to client device list and launch a client service task */
11653 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11654 		err = i40e_lan_add_device(pf);
11655 		if (err)
11656 			dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11657 				 err);
11658 	}
11659 
11660 #define PCI_SPEED_SIZE 8
11661 #define PCI_WIDTH_SIZE 8
11662 	/* Devices on the IOSF bus do not have this information
11663 	 * and will report PCI Gen 1 x 1 by default so don't bother
11664 	 * checking them.
11665 	 */
11666 	if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11667 		char speed[PCI_SPEED_SIZE] = "Unknown";
11668 		char width[PCI_WIDTH_SIZE] = "Unknown";
11669 
11670 		/* Get the negotiated link width and speed from PCI config
11671 		 * space
11672 		 */
11673 		pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11674 					  &link_status);
11675 
11676 		i40e_set_pci_config_data(hw, link_status);
11677 
11678 		switch (hw->bus.speed) {
11679 		case i40e_bus_speed_8000:
11680 			strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11681 		case i40e_bus_speed_5000:
11682 			strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11683 		case i40e_bus_speed_2500:
11684 			strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11685 		default:
11686 			break;
11687 		}
11688 		switch (hw->bus.width) {
11689 		case i40e_bus_width_pcie_x8:
11690 			strncpy(width, "8", PCI_WIDTH_SIZE); break;
11691 		case i40e_bus_width_pcie_x4:
11692 			strncpy(width, "4", PCI_WIDTH_SIZE); break;
11693 		case i40e_bus_width_pcie_x2:
11694 			strncpy(width, "2", PCI_WIDTH_SIZE); break;
11695 		case i40e_bus_width_pcie_x1:
11696 			strncpy(width, "1", PCI_WIDTH_SIZE); break;
11697 		default:
11698 			break;
11699 		}
11700 
11701 		dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11702 			 speed, width);
11703 
11704 		if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11705 		    hw->bus.speed < i40e_bus_speed_8000) {
11706 			dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11707 			dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11708 		}
11709 	}
11710 
11711 	/* get the requested speeds from the fw */
11712 	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11713 	if (err)
11714 		dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11715 			i40e_stat_str(&pf->hw, err),
11716 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11717 	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11718 
11719 	/* get the supported phy types from the fw */
11720 	err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11721 	if (err)
11722 		dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11723 			i40e_stat_str(&pf->hw, err),
11724 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11725 
11726 	/* Add a filter to drop all Flow control frames from any VSI from being
11727 	 * transmitted. By doing so we stop a malicious VF from sending out
11728 	 * PAUSE or PFC frames and potentially controlling traffic for other
11729 	 * PF/VF VSIs.
11730 	 * The FW can still send Flow control frames if enabled.
11731 	 */
11732 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11733 						       pf->main_vsi_seid);
11734 
11735 	if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11736 		(pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11737 		pf->flags |= I40E_FLAG_PHY_CONTROLS_LEDS;
11738 	if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
11739 		pf->flags |= I40E_FLAG_HAVE_CRT_RETIMER;
11740 	/* print a string summarizing features */
11741 	i40e_print_features(pf);
11742 
11743 	return 0;
11744 
11745 	/* Unwind what we've done if something failed in the setup */
11746 err_vsis:
11747 	set_bit(__I40E_DOWN, pf->state);
11748 	i40e_clear_interrupt_scheme(pf);
11749 	kfree(pf->vsi);
11750 err_switch_setup:
11751 	i40e_reset_interrupt_capability(pf);
11752 	del_timer_sync(&pf->service_timer);
11753 err_mac_addr:
11754 err_configure_lan_hmc:
11755 	(void)i40e_shutdown_lan_hmc(hw);
11756 err_init_lan_hmc:
11757 	kfree(pf->qp_pile);
11758 err_sw_init:
11759 err_adminq_setup:
11760 err_pf_reset:
11761 	iounmap(hw->hw_addr);
11762 err_ioremap:
11763 	kfree(pf);
11764 err_pf_alloc:
11765 	pci_disable_pcie_error_reporting(pdev);
11766 	pci_release_mem_regions(pdev);
11767 err_pci_reg:
11768 err_dma:
11769 	pci_disable_device(pdev);
11770 	return err;
11771 }
11772 
11773 /**
11774  * i40e_remove - Device removal routine
11775  * @pdev: PCI device information struct
11776  *
11777  * i40e_remove is called by the PCI subsystem to alert the driver
11778  * that is should release a PCI device.  This could be caused by a
11779  * Hot-Plug event, or because the driver is going to be removed from
11780  * memory.
11781  **/
11782 static void i40e_remove(struct pci_dev *pdev)
11783 {
11784 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11785 	struct i40e_hw *hw = &pf->hw;
11786 	i40e_status ret_code;
11787 	int i;
11788 
11789 	i40e_dbg_pf_exit(pf);
11790 
11791 	i40e_ptp_stop(pf);
11792 
11793 	/* Disable RSS in hw */
11794 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11795 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11796 
11797 	/* no more scheduling of any task */
11798 	set_bit(__I40E_SUSPENDED, pf->state);
11799 	set_bit(__I40E_DOWN, pf->state);
11800 	if (pf->service_timer.data)
11801 		del_timer_sync(&pf->service_timer);
11802 	if (pf->service_task.func)
11803 		cancel_work_sync(&pf->service_task);
11804 
11805 	/* Client close must be called explicitly here because the timer
11806 	 * has been stopped.
11807 	 */
11808 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
11809 
11810 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11811 		i40e_free_vfs(pf);
11812 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11813 	}
11814 
11815 	i40e_fdir_teardown(pf);
11816 
11817 	/* If there is a switch structure or any orphans, remove them.
11818 	 * This will leave only the PF's VSI remaining.
11819 	 */
11820 	for (i = 0; i < I40E_MAX_VEB; i++) {
11821 		if (!pf->veb[i])
11822 			continue;
11823 
11824 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11825 		    pf->veb[i]->uplink_seid == 0)
11826 			i40e_switch_branch_release(pf->veb[i]);
11827 	}
11828 
11829 	/* Now we can shutdown the PF's VSI, just before we kill
11830 	 * adminq and hmc.
11831 	 */
11832 	if (pf->vsi[pf->lan_vsi])
11833 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11834 
11835 	/* remove attached clients */
11836 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11837 		ret_code = i40e_lan_del_device(pf);
11838 		if (ret_code)
11839 			dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11840 				 ret_code);
11841 	}
11842 
11843 	/* shutdown and destroy the HMC */
11844 	if (hw->hmc.hmc_obj) {
11845 		ret_code = i40e_shutdown_lan_hmc(hw);
11846 		if (ret_code)
11847 			dev_warn(&pdev->dev,
11848 				 "Failed to destroy the HMC resources: %d\n",
11849 				 ret_code);
11850 	}
11851 
11852 	/* shutdown the adminq */
11853 	i40e_shutdown_adminq(hw);
11854 
11855 	/* destroy the locks only once, here */
11856 	mutex_destroy(&hw->aq.arq_mutex);
11857 	mutex_destroy(&hw->aq.asq_mutex);
11858 
11859 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11860 	i40e_clear_interrupt_scheme(pf);
11861 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11862 		if (pf->vsi[i]) {
11863 			i40e_vsi_clear_rings(pf->vsi[i]);
11864 			i40e_vsi_clear(pf->vsi[i]);
11865 			pf->vsi[i] = NULL;
11866 		}
11867 	}
11868 
11869 	for (i = 0; i < I40E_MAX_VEB; i++) {
11870 		kfree(pf->veb[i]);
11871 		pf->veb[i] = NULL;
11872 	}
11873 
11874 	kfree(pf->qp_pile);
11875 	kfree(pf->vsi);
11876 
11877 	iounmap(hw->hw_addr);
11878 	kfree(pf);
11879 	pci_release_mem_regions(pdev);
11880 
11881 	pci_disable_pcie_error_reporting(pdev);
11882 	pci_disable_device(pdev);
11883 }
11884 
11885 /**
11886  * i40e_pci_error_detected - warning that something funky happened in PCI land
11887  * @pdev: PCI device information struct
11888  *
11889  * Called to warn that something happened and the error handling steps
11890  * are in progress.  Allows the driver to quiesce things, be ready for
11891  * remediation.
11892  **/
11893 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11894 						enum pci_channel_state error)
11895 {
11896 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11897 
11898 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11899 
11900 	if (!pf) {
11901 		dev_info(&pdev->dev,
11902 			 "Cannot recover - error happened during device probe\n");
11903 		return PCI_ERS_RESULT_DISCONNECT;
11904 	}
11905 
11906 	/* shutdown all operations */
11907 	if (!test_bit(__I40E_SUSPENDED, pf->state))
11908 		i40e_prep_for_reset(pf, false);
11909 
11910 	/* Request a slot reset */
11911 	return PCI_ERS_RESULT_NEED_RESET;
11912 }
11913 
11914 /**
11915  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11916  * @pdev: PCI device information struct
11917  *
11918  * Called to find if the driver can work with the device now that
11919  * the pci slot has been reset.  If a basic connection seems good
11920  * (registers are readable and have sane content) then return a
11921  * happy little PCI_ERS_RESULT_xxx.
11922  **/
11923 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11924 {
11925 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11926 	pci_ers_result_t result;
11927 	int err;
11928 	u32 reg;
11929 
11930 	dev_dbg(&pdev->dev, "%s\n", __func__);
11931 	if (pci_enable_device_mem(pdev)) {
11932 		dev_info(&pdev->dev,
11933 			 "Cannot re-enable PCI device after reset.\n");
11934 		result = PCI_ERS_RESULT_DISCONNECT;
11935 	} else {
11936 		pci_set_master(pdev);
11937 		pci_restore_state(pdev);
11938 		pci_save_state(pdev);
11939 		pci_wake_from_d3(pdev, false);
11940 
11941 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11942 		if (reg == 0)
11943 			result = PCI_ERS_RESULT_RECOVERED;
11944 		else
11945 			result = PCI_ERS_RESULT_DISCONNECT;
11946 	}
11947 
11948 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
11949 	if (err) {
11950 		dev_info(&pdev->dev,
11951 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11952 			 err);
11953 		/* non-fatal, continue */
11954 	}
11955 
11956 	return result;
11957 }
11958 
11959 /**
11960  * i40e_pci_error_resume - restart operations after PCI error recovery
11961  * @pdev: PCI device information struct
11962  *
11963  * Called to allow the driver to bring things back up after PCI error
11964  * and/or reset recovery has finished.
11965  **/
11966 static void i40e_pci_error_resume(struct pci_dev *pdev)
11967 {
11968 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11969 
11970 	dev_dbg(&pdev->dev, "%s\n", __func__);
11971 	if (test_bit(__I40E_SUSPENDED, pf->state))
11972 		return;
11973 
11974 	i40e_handle_reset_warning(pf, false);
11975 }
11976 
11977 /**
11978  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
11979  * using the mac_address_write admin q function
11980  * @pf: pointer to i40e_pf struct
11981  **/
11982 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
11983 {
11984 	struct i40e_hw *hw = &pf->hw;
11985 	i40e_status ret;
11986 	u8 mac_addr[6];
11987 	u16 flags = 0;
11988 
11989 	/* Get current MAC address in case it's an LAA */
11990 	if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
11991 		ether_addr_copy(mac_addr,
11992 				pf->vsi[pf->lan_vsi]->netdev->dev_addr);
11993 	} else {
11994 		dev_err(&pf->pdev->dev,
11995 			"Failed to retrieve MAC address; using default\n");
11996 		ether_addr_copy(mac_addr, hw->mac.addr);
11997 	}
11998 
11999 	/* The FW expects the mac address write cmd to first be called with
12000 	 * one of these flags before calling it again with the multicast
12001 	 * enable flags.
12002 	 */
12003 	flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
12004 
12005 	if (hw->func_caps.flex10_enable && hw->partition_id != 1)
12006 		flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
12007 
12008 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
12009 	if (ret) {
12010 		dev_err(&pf->pdev->dev,
12011 			"Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
12012 		return;
12013 	}
12014 
12015 	flags = I40E_AQC_MC_MAG_EN
12016 			| I40E_AQC_WOL_PRESERVE_ON_PFR
12017 			| I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
12018 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
12019 	if (ret)
12020 		dev_err(&pf->pdev->dev,
12021 			"Failed to enable Multicast Magic Packet wake up\n");
12022 }
12023 
12024 /**
12025  * i40e_shutdown - PCI callback for shutting down
12026  * @pdev: PCI device information struct
12027  **/
12028 static void i40e_shutdown(struct pci_dev *pdev)
12029 {
12030 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12031 	struct i40e_hw *hw = &pf->hw;
12032 
12033 	set_bit(__I40E_SUSPENDED, pf->state);
12034 	set_bit(__I40E_DOWN, pf->state);
12035 	rtnl_lock();
12036 	i40e_prep_for_reset(pf, true);
12037 	rtnl_unlock();
12038 
12039 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
12040 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
12041 
12042 	del_timer_sync(&pf->service_timer);
12043 	cancel_work_sync(&pf->service_task);
12044 	i40e_fdir_teardown(pf);
12045 
12046 	/* Client close must be called explicitly here because the timer
12047 	 * has been stopped.
12048 	 */
12049 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
12050 
12051 	if (pf->wol_en && (pf->flags & I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE))
12052 		i40e_enable_mc_magic_wake(pf);
12053 
12054 	i40e_prep_for_reset(pf, false);
12055 
12056 	wr32(hw, I40E_PFPM_APM,
12057 	     (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
12058 	wr32(hw, I40E_PFPM_WUFC,
12059 	     (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
12060 
12061 	i40e_clear_interrupt_scheme(pf);
12062 
12063 	if (system_state == SYSTEM_POWER_OFF) {
12064 		pci_wake_from_d3(pdev, pf->wol_en);
12065 		pci_set_power_state(pdev, PCI_D3hot);
12066 	}
12067 }
12068 
12069 #ifdef CONFIG_PM
12070 /**
12071  * i40e_suspend - PCI callback for moving to D3
12072  * @pdev: PCI device information struct
12073  **/
12074 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
12075 {
12076 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12077 	struct i40e_hw *hw = &pf->hw;
12078 	int retval = 0;
12079 
12080 	set_bit(__I40E_SUSPENDED, pf->state);
12081 	set_bit(__I40E_DOWN, pf->state);
12082 
12083 	if (pf->wol_en && (pf->flags & I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE))
12084 		i40e_enable_mc_magic_wake(pf);
12085 
12086 	i40e_prep_for_reset(pf, false);
12087 
12088 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
12089 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
12090 
12091 	i40e_stop_misc_vector(pf);
12092 
12093 	retval = pci_save_state(pdev);
12094 	if (retval)
12095 		return retval;
12096 
12097 	pci_wake_from_d3(pdev, pf->wol_en);
12098 	pci_set_power_state(pdev, PCI_D3hot);
12099 
12100 	return retval;
12101 }
12102 
12103 /**
12104  * i40e_resume - PCI callback for waking up from D3
12105  * @pdev: PCI device information struct
12106  **/
12107 static int i40e_resume(struct pci_dev *pdev)
12108 {
12109 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12110 	u32 err;
12111 
12112 	pci_set_power_state(pdev, PCI_D0);
12113 	pci_restore_state(pdev);
12114 	/* pci_restore_state() clears dev->state_saves, so
12115 	 * call pci_save_state() again to restore it.
12116 	 */
12117 	pci_save_state(pdev);
12118 
12119 	err = pci_enable_device_mem(pdev);
12120 	if (err) {
12121 		dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
12122 		return err;
12123 	}
12124 	pci_set_master(pdev);
12125 
12126 	/* no wakeup events while running */
12127 	pci_wake_from_d3(pdev, false);
12128 
12129 	/* handling the reset will rebuild the device state */
12130 	if (test_and_clear_bit(__I40E_SUSPENDED, pf->state)) {
12131 		clear_bit(__I40E_DOWN, pf->state);
12132 		i40e_reset_and_rebuild(pf, false, false);
12133 	}
12134 
12135 	return 0;
12136 }
12137 
12138 #endif
12139 static const struct pci_error_handlers i40e_err_handler = {
12140 	.error_detected = i40e_pci_error_detected,
12141 	.slot_reset = i40e_pci_error_slot_reset,
12142 	.resume = i40e_pci_error_resume,
12143 };
12144 
12145 static struct pci_driver i40e_driver = {
12146 	.name     = i40e_driver_name,
12147 	.id_table = i40e_pci_tbl,
12148 	.probe    = i40e_probe,
12149 	.remove   = i40e_remove,
12150 #ifdef CONFIG_PM
12151 	.suspend  = i40e_suspend,
12152 	.resume   = i40e_resume,
12153 #endif
12154 	.shutdown = i40e_shutdown,
12155 	.err_handler = &i40e_err_handler,
12156 	.sriov_configure = i40e_pci_sriov_configure,
12157 };
12158 
12159 /**
12160  * i40e_init_module - Driver registration routine
12161  *
12162  * i40e_init_module is the first routine called when the driver is
12163  * loaded. All it does is register with the PCI subsystem.
12164  **/
12165 static int __init i40e_init_module(void)
12166 {
12167 	pr_info("%s: %s - version %s\n", i40e_driver_name,
12168 		i40e_driver_string, i40e_driver_version_str);
12169 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
12170 
12171 	/* we will see if single thread per module is enough for now,
12172 	 * it can't be any worse than using the system workqueue which
12173 	 * was already single threaded
12174 	 */
12175 	i40e_wq = alloc_workqueue("%s", WQ_UNBOUND | WQ_MEM_RECLAIM, 1,
12176 				  i40e_driver_name);
12177 	if (!i40e_wq) {
12178 		pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
12179 		return -ENOMEM;
12180 	}
12181 
12182 	i40e_dbg_init();
12183 	return pci_register_driver(&i40e_driver);
12184 }
12185 module_init(i40e_init_module);
12186 
12187 /**
12188  * i40e_exit_module - Driver exit cleanup routine
12189  *
12190  * i40e_exit_module is called just before the driver is removed
12191  * from memory.
12192  **/
12193 static void __exit i40e_exit_module(void)
12194 {
12195 	pci_unregister_driver(&i40e_driver);
12196 	destroy_workqueue(i40e_wq);
12197 	i40e_dbg_exit();
12198 }
12199 module_exit(i40e_exit_module);
12200