xref: /linux/drivers/net/ethernet/intel/i40e/i40e_main.c (revision 83a37b3292f4aca799b355179ad6fbdd78a08e10)
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_stat_update_and_clear32 - read and clear hw reg, update a 32 bit stat
604  * @hw: ptr to the hardware info
605  * @reg: the hw reg to read and clear
606  * @stat: ptr to the stat
607  **/
608 static void i40e_stat_update_and_clear32(struct i40e_hw *hw, u32 reg, u64 *stat)
609 {
610 	u32 new_data = rd32(hw, reg);
611 
612 	wr32(hw, reg, 1); /* must write a nonzero value to clear register */
613 	*stat += new_data;
614 }
615 
616 /**
617  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
618  * @vsi: the VSI to be updated
619  **/
620 void i40e_update_eth_stats(struct i40e_vsi *vsi)
621 {
622 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
623 	struct i40e_pf *pf = vsi->back;
624 	struct i40e_hw *hw = &pf->hw;
625 	struct i40e_eth_stats *oes;
626 	struct i40e_eth_stats *es;     /* device's eth stats */
627 
628 	es = &vsi->eth_stats;
629 	oes = &vsi->eth_stats_offsets;
630 
631 	/* Gather up the stats that the hw collects */
632 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
633 			   vsi->stat_offsets_loaded,
634 			   &oes->tx_errors, &es->tx_errors);
635 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
636 			   vsi->stat_offsets_loaded,
637 			   &oes->rx_discards, &es->rx_discards);
638 	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
639 			   vsi->stat_offsets_loaded,
640 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
641 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
642 			   vsi->stat_offsets_loaded,
643 			   &oes->tx_errors, &es->tx_errors);
644 
645 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
646 			   I40E_GLV_GORCL(stat_idx),
647 			   vsi->stat_offsets_loaded,
648 			   &oes->rx_bytes, &es->rx_bytes);
649 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
650 			   I40E_GLV_UPRCL(stat_idx),
651 			   vsi->stat_offsets_loaded,
652 			   &oes->rx_unicast, &es->rx_unicast);
653 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
654 			   I40E_GLV_MPRCL(stat_idx),
655 			   vsi->stat_offsets_loaded,
656 			   &oes->rx_multicast, &es->rx_multicast);
657 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
658 			   I40E_GLV_BPRCL(stat_idx),
659 			   vsi->stat_offsets_loaded,
660 			   &oes->rx_broadcast, &es->rx_broadcast);
661 
662 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
663 			   I40E_GLV_GOTCL(stat_idx),
664 			   vsi->stat_offsets_loaded,
665 			   &oes->tx_bytes, &es->tx_bytes);
666 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
667 			   I40E_GLV_UPTCL(stat_idx),
668 			   vsi->stat_offsets_loaded,
669 			   &oes->tx_unicast, &es->tx_unicast);
670 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
671 			   I40E_GLV_MPTCL(stat_idx),
672 			   vsi->stat_offsets_loaded,
673 			   &oes->tx_multicast, &es->tx_multicast);
674 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
675 			   I40E_GLV_BPTCL(stat_idx),
676 			   vsi->stat_offsets_loaded,
677 			   &oes->tx_broadcast, &es->tx_broadcast);
678 	vsi->stat_offsets_loaded = true;
679 }
680 
681 /**
682  * i40e_update_veb_stats - Update Switch component statistics
683  * @veb: the VEB being updated
684  **/
685 static void i40e_update_veb_stats(struct i40e_veb *veb)
686 {
687 	struct i40e_pf *pf = veb->pf;
688 	struct i40e_hw *hw = &pf->hw;
689 	struct i40e_eth_stats *oes;
690 	struct i40e_eth_stats *es;     /* device's eth stats */
691 	struct i40e_veb_tc_stats *veb_oes;
692 	struct i40e_veb_tc_stats *veb_es;
693 	int i, idx = 0;
694 
695 	idx = veb->stats_idx;
696 	es = &veb->stats;
697 	oes = &veb->stats_offsets;
698 	veb_es = &veb->tc_stats;
699 	veb_oes = &veb->tc_stats_offsets;
700 
701 	/* Gather up the stats that the hw collects */
702 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
703 			   veb->stat_offsets_loaded,
704 			   &oes->tx_discards, &es->tx_discards);
705 	if (hw->revision_id > 0)
706 		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
707 				   veb->stat_offsets_loaded,
708 				   &oes->rx_unknown_protocol,
709 				   &es->rx_unknown_protocol);
710 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
711 			   veb->stat_offsets_loaded,
712 			   &oes->rx_bytes, &es->rx_bytes);
713 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
714 			   veb->stat_offsets_loaded,
715 			   &oes->rx_unicast, &es->rx_unicast);
716 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
717 			   veb->stat_offsets_loaded,
718 			   &oes->rx_multicast, &es->rx_multicast);
719 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
720 			   veb->stat_offsets_loaded,
721 			   &oes->rx_broadcast, &es->rx_broadcast);
722 
723 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
724 			   veb->stat_offsets_loaded,
725 			   &oes->tx_bytes, &es->tx_bytes);
726 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
727 			   veb->stat_offsets_loaded,
728 			   &oes->tx_unicast, &es->tx_unicast);
729 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
730 			   veb->stat_offsets_loaded,
731 			   &oes->tx_multicast, &es->tx_multicast);
732 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
733 			   veb->stat_offsets_loaded,
734 			   &oes->tx_broadcast, &es->tx_broadcast);
735 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
736 		i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
737 				   I40E_GLVEBTC_RPCL(i, idx),
738 				   veb->stat_offsets_loaded,
739 				   &veb_oes->tc_rx_packets[i],
740 				   &veb_es->tc_rx_packets[i]);
741 		i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
742 				   I40E_GLVEBTC_RBCL(i, idx),
743 				   veb->stat_offsets_loaded,
744 				   &veb_oes->tc_rx_bytes[i],
745 				   &veb_es->tc_rx_bytes[i]);
746 		i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
747 				   I40E_GLVEBTC_TPCL(i, idx),
748 				   veb->stat_offsets_loaded,
749 				   &veb_oes->tc_tx_packets[i],
750 				   &veb_es->tc_tx_packets[i]);
751 		i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
752 				   I40E_GLVEBTC_TBCL(i, idx),
753 				   veb->stat_offsets_loaded,
754 				   &veb_oes->tc_tx_bytes[i],
755 				   &veb_es->tc_tx_bytes[i]);
756 	}
757 	veb->stat_offsets_loaded = true;
758 }
759 
760 /**
761  * i40e_update_vsi_stats - Update the vsi statistics counters.
762  * @vsi: the VSI to be updated
763  *
764  * There are a few instances where we store the same stat in a
765  * couple of different structs.  This is partly because we have
766  * the netdev stats that need to be filled out, which is slightly
767  * different from the "eth_stats" defined by the chip and used in
768  * VF communications.  We sort it out here.
769  **/
770 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
771 {
772 	struct i40e_pf *pf = vsi->back;
773 	struct rtnl_link_stats64 *ons;
774 	struct rtnl_link_stats64 *ns;   /* netdev stats */
775 	struct i40e_eth_stats *oes;
776 	struct i40e_eth_stats *es;     /* device's eth stats */
777 	u32 tx_restart, tx_busy;
778 	struct i40e_ring *p;
779 	u32 rx_page, rx_buf;
780 	u64 bytes, packets;
781 	unsigned int start;
782 	u64 tx_linearize;
783 	u64 tx_force_wb;
784 	u64 rx_p, rx_b;
785 	u64 tx_p, tx_b;
786 	u16 q;
787 
788 	if (test_bit(__I40E_VSI_DOWN, vsi->state) ||
789 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
790 		return;
791 
792 	ns = i40e_get_vsi_stats_struct(vsi);
793 	ons = &vsi->net_stats_offsets;
794 	es = &vsi->eth_stats;
795 	oes = &vsi->eth_stats_offsets;
796 
797 	/* Gather up the netdev and vsi stats that the driver collects
798 	 * on the fly during packet processing
799 	 */
800 	rx_b = rx_p = 0;
801 	tx_b = tx_p = 0;
802 	tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
803 	rx_page = 0;
804 	rx_buf = 0;
805 	rcu_read_lock();
806 	for (q = 0; q < vsi->num_queue_pairs; q++) {
807 		/* locate Tx ring */
808 		p = ACCESS_ONCE(vsi->tx_rings[q]);
809 
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 		tx_b += bytes;
816 		tx_p += packets;
817 		tx_restart += p->tx_stats.restart_queue;
818 		tx_busy += p->tx_stats.tx_busy;
819 		tx_linearize += p->tx_stats.tx_linearize;
820 		tx_force_wb += p->tx_stats.tx_force_wb;
821 
822 		/* Rx queue is part of the same block as Tx queue */
823 		p = &p[1];
824 		do {
825 			start = u64_stats_fetch_begin_irq(&p->syncp);
826 			packets = p->stats.packets;
827 			bytes = p->stats.bytes;
828 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
829 		rx_b += bytes;
830 		rx_p += packets;
831 		rx_buf += p->rx_stats.alloc_buff_failed;
832 		rx_page += p->rx_stats.alloc_page_failed;
833 	}
834 	rcu_read_unlock();
835 	vsi->tx_restart = tx_restart;
836 	vsi->tx_busy = tx_busy;
837 	vsi->tx_linearize = tx_linearize;
838 	vsi->tx_force_wb = tx_force_wb;
839 	vsi->rx_page_failed = rx_page;
840 	vsi->rx_buf_failed = rx_buf;
841 
842 	ns->rx_packets = rx_p;
843 	ns->rx_bytes = rx_b;
844 	ns->tx_packets = tx_p;
845 	ns->tx_bytes = tx_b;
846 
847 	/* update netdev stats from eth stats */
848 	i40e_update_eth_stats(vsi);
849 	ons->tx_errors = oes->tx_errors;
850 	ns->tx_errors = es->tx_errors;
851 	ons->multicast = oes->rx_multicast;
852 	ns->multicast = es->rx_multicast;
853 	ons->rx_dropped = oes->rx_discards;
854 	ns->rx_dropped = es->rx_discards;
855 	ons->tx_dropped = oes->tx_discards;
856 	ns->tx_dropped = es->tx_discards;
857 
858 	/* pull in a couple PF stats if this is the main vsi */
859 	if (vsi == pf->vsi[pf->lan_vsi]) {
860 		ns->rx_crc_errors = pf->stats.crc_errors;
861 		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
862 		ns->rx_length_errors = pf->stats.rx_length_errors;
863 	}
864 }
865 
866 /**
867  * i40e_update_pf_stats - Update the PF statistics counters.
868  * @pf: the PF to be updated
869  **/
870 static void i40e_update_pf_stats(struct i40e_pf *pf)
871 {
872 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
873 	struct i40e_hw_port_stats *nsd = &pf->stats;
874 	struct i40e_hw *hw = &pf->hw;
875 	u32 val;
876 	int i;
877 
878 	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
879 			   I40E_GLPRT_GORCL(hw->port),
880 			   pf->stat_offsets_loaded,
881 			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
882 	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
883 			   I40E_GLPRT_GOTCL(hw->port),
884 			   pf->stat_offsets_loaded,
885 			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
886 	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
887 			   pf->stat_offsets_loaded,
888 			   &osd->eth.rx_discards,
889 			   &nsd->eth.rx_discards);
890 	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
891 			   I40E_GLPRT_UPRCL(hw->port),
892 			   pf->stat_offsets_loaded,
893 			   &osd->eth.rx_unicast,
894 			   &nsd->eth.rx_unicast);
895 	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
896 			   I40E_GLPRT_MPRCL(hw->port),
897 			   pf->stat_offsets_loaded,
898 			   &osd->eth.rx_multicast,
899 			   &nsd->eth.rx_multicast);
900 	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
901 			   I40E_GLPRT_BPRCL(hw->port),
902 			   pf->stat_offsets_loaded,
903 			   &osd->eth.rx_broadcast,
904 			   &nsd->eth.rx_broadcast);
905 	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
906 			   I40E_GLPRT_UPTCL(hw->port),
907 			   pf->stat_offsets_loaded,
908 			   &osd->eth.tx_unicast,
909 			   &nsd->eth.tx_unicast);
910 	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
911 			   I40E_GLPRT_MPTCL(hw->port),
912 			   pf->stat_offsets_loaded,
913 			   &osd->eth.tx_multicast,
914 			   &nsd->eth.tx_multicast);
915 	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
916 			   I40E_GLPRT_BPTCL(hw->port),
917 			   pf->stat_offsets_loaded,
918 			   &osd->eth.tx_broadcast,
919 			   &nsd->eth.tx_broadcast);
920 
921 	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
922 			   pf->stat_offsets_loaded,
923 			   &osd->tx_dropped_link_down,
924 			   &nsd->tx_dropped_link_down);
925 
926 	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
927 			   pf->stat_offsets_loaded,
928 			   &osd->crc_errors, &nsd->crc_errors);
929 
930 	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
931 			   pf->stat_offsets_loaded,
932 			   &osd->illegal_bytes, &nsd->illegal_bytes);
933 
934 	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
935 			   pf->stat_offsets_loaded,
936 			   &osd->mac_local_faults,
937 			   &nsd->mac_local_faults);
938 	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
939 			   pf->stat_offsets_loaded,
940 			   &osd->mac_remote_faults,
941 			   &nsd->mac_remote_faults);
942 
943 	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
944 			   pf->stat_offsets_loaded,
945 			   &osd->rx_length_errors,
946 			   &nsd->rx_length_errors);
947 
948 	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
949 			   pf->stat_offsets_loaded,
950 			   &osd->link_xon_rx, &nsd->link_xon_rx);
951 	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
952 			   pf->stat_offsets_loaded,
953 			   &osd->link_xon_tx, &nsd->link_xon_tx);
954 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
955 			   pf->stat_offsets_loaded,
956 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
957 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
958 			   pf->stat_offsets_loaded,
959 			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
960 
961 	for (i = 0; i < 8; i++) {
962 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
963 				   pf->stat_offsets_loaded,
964 				   &osd->priority_xoff_rx[i],
965 				   &nsd->priority_xoff_rx[i]);
966 		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
967 				   pf->stat_offsets_loaded,
968 				   &osd->priority_xon_rx[i],
969 				   &nsd->priority_xon_rx[i]);
970 		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
971 				   pf->stat_offsets_loaded,
972 				   &osd->priority_xon_tx[i],
973 				   &nsd->priority_xon_tx[i]);
974 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
975 				   pf->stat_offsets_loaded,
976 				   &osd->priority_xoff_tx[i],
977 				   &nsd->priority_xoff_tx[i]);
978 		i40e_stat_update32(hw,
979 				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
980 				   pf->stat_offsets_loaded,
981 				   &osd->priority_xon_2_xoff[i],
982 				   &nsd->priority_xon_2_xoff[i]);
983 	}
984 
985 	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
986 			   I40E_GLPRT_PRC64L(hw->port),
987 			   pf->stat_offsets_loaded,
988 			   &osd->rx_size_64, &nsd->rx_size_64);
989 	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
990 			   I40E_GLPRT_PRC127L(hw->port),
991 			   pf->stat_offsets_loaded,
992 			   &osd->rx_size_127, &nsd->rx_size_127);
993 	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
994 			   I40E_GLPRT_PRC255L(hw->port),
995 			   pf->stat_offsets_loaded,
996 			   &osd->rx_size_255, &nsd->rx_size_255);
997 	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
998 			   I40E_GLPRT_PRC511L(hw->port),
999 			   pf->stat_offsets_loaded,
1000 			   &osd->rx_size_511, &nsd->rx_size_511);
1001 	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1002 			   I40E_GLPRT_PRC1023L(hw->port),
1003 			   pf->stat_offsets_loaded,
1004 			   &osd->rx_size_1023, &nsd->rx_size_1023);
1005 	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1006 			   I40E_GLPRT_PRC1522L(hw->port),
1007 			   pf->stat_offsets_loaded,
1008 			   &osd->rx_size_1522, &nsd->rx_size_1522);
1009 	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1010 			   I40E_GLPRT_PRC9522L(hw->port),
1011 			   pf->stat_offsets_loaded,
1012 			   &osd->rx_size_big, &nsd->rx_size_big);
1013 
1014 	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1015 			   I40E_GLPRT_PTC64L(hw->port),
1016 			   pf->stat_offsets_loaded,
1017 			   &osd->tx_size_64, &nsd->tx_size_64);
1018 	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1019 			   I40E_GLPRT_PTC127L(hw->port),
1020 			   pf->stat_offsets_loaded,
1021 			   &osd->tx_size_127, &nsd->tx_size_127);
1022 	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1023 			   I40E_GLPRT_PTC255L(hw->port),
1024 			   pf->stat_offsets_loaded,
1025 			   &osd->tx_size_255, &nsd->tx_size_255);
1026 	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1027 			   I40E_GLPRT_PTC511L(hw->port),
1028 			   pf->stat_offsets_loaded,
1029 			   &osd->tx_size_511, &nsd->tx_size_511);
1030 	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1031 			   I40E_GLPRT_PTC1023L(hw->port),
1032 			   pf->stat_offsets_loaded,
1033 			   &osd->tx_size_1023, &nsd->tx_size_1023);
1034 	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1035 			   I40E_GLPRT_PTC1522L(hw->port),
1036 			   pf->stat_offsets_loaded,
1037 			   &osd->tx_size_1522, &nsd->tx_size_1522);
1038 	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1039 			   I40E_GLPRT_PTC9522L(hw->port),
1040 			   pf->stat_offsets_loaded,
1041 			   &osd->tx_size_big, &nsd->tx_size_big);
1042 
1043 	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1044 			   pf->stat_offsets_loaded,
1045 			   &osd->rx_undersize, &nsd->rx_undersize);
1046 	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1047 			   pf->stat_offsets_loaded,
1048 			   &osd->rx_fragments, &nsd->rx_fragments);
1049 	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1050 			   pf->stat_offsets_loaded,
1051 			   &osd->rx_oversize, &nsd->rx_oversize);
1052 	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1053 			   pf->stat_offsets_loaded,
1054 			   &osd->rx_jabber, &nsd->rx_jabber);
1055 
1056 	/* FDIR stats */
1057 	i40e_stat_update_and_clear32(hw,
1058 			I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(hw->pf_id)),
1059 			&nsd->fd_atr_match);
1060 	i40e_stat_update_and_clear32(hw,
1061 			I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(hw->pf_id)),
1062 			&nsd->fd_sb_match);
1063 	i40e_stat_update_and_clear32(hw,
1064 			I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(hw->pf_id)),
1065 			&nsd->fd_atr_tunnel_match);
1066 
1067 	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1068 	nsd->tx_lpi_status =
1069 		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1070 			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1071 	nsd->rx_lpi_status =
1072 		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1073 			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1074 	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1075 			   pf->stat_offsets_loaded,
1076 			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1077 	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1078 			   pf->stat_offsets_loaded,
1079 			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1080 
1081 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1082 	    !(pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED))
1083 		nsd->fd_sb_status = true;
1084 	else
1085 		nsd->fd_sb_status = false;
1086 
1087 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1088 	    !(pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED))
1089 		nsd->fd_atr_status = true;
1090 	else
1091 		nsd->fd_atr_status = false;
1092 
1093 	pf->stat_offsets_loaded = true;
1094 }
1095 
1096 /**
1097  * i40e_update_stats - Update the various statistics counters.
1098  * @vsi: the VSI to be updated
1099  *
1100  * Update the various stats for this VSI and its related entities.
1101  **/
1102 void i40e_update_stats(struct i40e_vsi *vsi)
1103 {
1104 	struct i40e_pf *pf = vsi->back;
1105 
1106 	if (vsi == pf->vsi[pf->lan_vsi])
1107 		i40e_update_pf_stats(pf);
1108 
1109 	i40e_update_vsi_stats(vsi);
1110 }
1111 
1112 /**
1113  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1114  * @vsi: the VSI to be searched
1115  * @macaddr: the MAC address
1116  * @vlan: the vlan
1117  *
1118  * Returns ptr to the filter object or NULL
1119  **/
1120 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1121 						const u8 *macaddr, s16 vlan)
1122 {
1123 	struct i40e_mac_filter *f;
1124 	u64 key;
1125 
1126 	if (!vsi || !macaddr)
1127 		return NULL;
1128 
1129 	key = i40e_addr_to_hkey(macaddr);
1130 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1131 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1132 		    (vlan == f->vlan))
1133 			return f;
1134 	}
1135 	return NULL;
1136 }
1137 
1138 /**
1139  * i40e_find_mac - Find a mac addr in the macvlan filters list
1140  * @vsi: the VSI to be searched
1141  * @macaddr: the MAC address we are searching for
1142  *
1143  * Returns the first filter with the provided MAC address or NULL if
1144  * MAC address was not found
1145  **/
1146 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1147 {
1148 	struct i40e_mac_filter *f;
1149 	u64 key;
1150 
1151 	if (!vsi || !macaddr)
1152 		return NULL;
1153 
1154 	key = i40e_addr_to_hkey(macaddr);
1155 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1156 		if ((ether_addr_equal(macaddr, f->macaddr)))
1157 			return f;
1158 	}
1159 	return NULL;
1160 }
1161 
1162 /**
1163  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1164  * @vsi: the VSI to be searched
1165  *
1166  * Returns true if VSI is in vlan mode or false otherwise
1167  **/
1168 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1169 {
1170 	/* If we have a PVID, always operate in VLAN mode */
1171 	if (vsi->info.pvid)
1172 		return true;
1173 
1174 	/* We need to operate in VLAN mode whenever we have any filters with
1175 	 * a VLAN other than I40E_VLAN_ALL. We could check the table each
1176 	 * time, incurring search cost repeatedly. However, we can notice two
1177 	 * things:
1178 	 *
1179 	 * 1) the only place where we can gain a VLAN filter is in
1180 	 *    i40e_add_filter.
1181 	 *
1182 	 * 2) the only place where filters are actually removed is in
1183 	 *    i40e_sync_filters_subtask.
1184 	 *
1185 	 * Thus, we can simply use a boolean value, has_vlan_filters which we
1186 	 * will set to true when we add a VLAN filter in i40e_add_filter. Then
1187 	 * we have to perform the full search after deleting filters in
1188 	 * i40e_sync_filters_subtask, but we already have to search
1189 	 * filters here and can perform the check at the same time. This
1190 	 * results in avoiding embedding a loop for VLAN mode inside another
1191 	 * loop over all the filters, and should maintain correctness as noted
1192 	 * above.
1193 	 */
1194 	return vsi->has_vlan_filter;
1195 }
1196 
1197 /**
1198  * i40e_correct_mac_vlan_filters - Correct non-VLAN filters if necessary
1199  * @vsi: the VSI to configure
1200  * @tmp_add_list: list of filters ready to be added
1201  * @tmp_del_list: list of filters ready to be deleted
1202  * @vlan_filters: the number of active VLAN filters
1203  *
1204  * Update VLAN=0 and VLAN=-1 (I40E_VLAN_ANY) filters properly so that they
1205  * behave as expected. If we have any active VLAN filters remaining or about
1206  * to be added then we need to update non-VLAN filters to be marked as VLAN=0
1207  * so that they only match against untagged traffic. If we no longer have any
1208  * active VLAN filters, we need to make all non-VLAN filters marked as VLAN=-1
1209  * so that they match against both tagged and untagged traffic. In this way,
1210  * we ensure that we correctly receive the desired traffic. This ensures that
1211  * when we have an active VLAN we will receive only untagged traffic and
1212  * traffic matching active VLANs. If we have no active VLANs then we will
1213  * operate in non-VLAN mode and receive all traffic, tagged or untagged.
1214  *
1215  * Finally, in a similar fashion, this function also corrects filters when
1216  * there is an active PVID assigned to this VSI.
1217  *
1218  * In case of memory allocation failure return -ENOMEM. Otherwise, return 0.
1219  *
1220  * This function is only expected to be called from within
1221  * i40e_sync_vsi_filters.
1222  *
1223  * NOTE: This function expects to be called while under the
1224  * mac_filter_hash_lock
1225  */
1226 static int i40e_correct_mac_vlan_filters(struct i40e_vsi *vsi,
1227 					 struct hlist_head *tmp_add_list,
1228 					 struct hlist_head *tmp_del_list,
1229 					 int vlan_filters)
1230 {
1231 	s16 pvid = le16_to_cpu(vsi->info.pvid);
1232 	struct i40e_mac_filter *f, *add_head;
1233 	struct i40e_new_mac_filter *new;
1234 	struct hlist_node *h;
1235 	int bkt, new_vlan;
1236 
1237 	/* To determine if a particular filter needs to be replaced we
1238 	 * have the three following conditions:
1239 	 *
1240 	 * a) if we have a PVID assigned, then all filters which are
1241 	 *    not marked as VLAN=PVID must be replaced with filters that
1242 	 *    are.
1243 	 * b) otherwise, if we have any active VLANS, all filters
1244 	 *    which are marked as VLAN=-1 must be replaced with
1245 	 *    filters marked as VLAN=0
1246 	 * c) finally, if we do not have any active VLANS, all filters
1247 	 *    which are marked as VLAN=0 must be replaced with filters
1248 	 *    marked as VLAN=-1
1249 	 */
1250 
1251 	/* Update the filters about to be added in place */
1252 	hlist_for_each_entry(new, tmp_add_list, hlist) {
1253 		if (pvid && new->f->vlan != pvid)
1254 			new->f->vlan = pvid;
1255 		else if (vlan_filters && new->f->vlan == I40E_VLAN_ANY)
1256 			new->f->vlan = 0;
1257 		else if (!vlan_filters && new->f->vlan == 0)
1258 			new->f->vlan = I40E_VLAN_ANY;
1259 	}
1260 
1261 	/* Update the remaining active filters */
1262 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1263 		/* Combine the checks for whether a filter needs to be changed
1264 		 * and then determine the new VLAN inside the if block, in
1265 		 * order to avoid duplicating code for adding the new filter
1266 		 * then deleting the old filter.
1267 		 */
1268 		if ((pvid && f->vlan != pvid) ||
1269 		    (vlan_filters && f->vlan == I40E_VLAN_ANY) ||
1270 		    (!vlan_filters && f->vlan == 0)) {
1271 			/* Determine the new vlan we will be adding */
1272 			if (pvid)
1273 				new_vlan = pvid;
1274 			else if (vlan_filters)
1275 				new_vlan = 0;
1276 			else
1277 				new_vlan = I40E_VLAN_ANY;
1278 
1279 			/* Create the new filter */
1280 			add_head = i40e_add_filter(vsi, f->macaddr, new_vlan);
1281 			if (!add_head)
1282 				return -ENOMEM;
1283 
1284 			/* Create a temporary i40e_new_mac_filter */
1285 			new = kzalloc(sizeof(*new), GFP_ATOMIC);
1286 			if (!new)
1287 				return -ENOMEM;
1288 
1289 			new->f = add_head;
1290 			new->state = add_head->state;
1291 
1292 			/* Add the new filter to the tmp list */
1293 			hlist_add_head(&new->hlist, tmp_add_list);
1294 
1295 			/* Put the original filter into the delete list */
1296 			f->state = I40E_FILTER_REMOVE;
1297 			hash_del(&f->hlist);
1298 			hlist_add_head(&f->hlist, tmp_del_list);
1299 		}
1300 	}
1301 
1302 	vsi->has_vlan_filter = !!vlan_filters;
1303 
1304 	return 0;
1305 }
1306 
1307 /**
1308  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1309  * @vsi: the PF Main VSI - inappropriate for any other VSI
1310  * @macaddr: the MAC address
1311  *
1312  * Remove whatever filter the firmware set up so the driver can manage
1313  * its own filtering intelligently.
1314  **/
1315 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1316 {
1317 	struct i40e_aqc_remove_macvlan_element_data element;
1318 	struct i40e_pf *pf = vsi->back;
1319 
1320 	/* Only appropriate for the PF main VSI */
1321 	if (vsi->type != I40E_VSI_MAIN)
1322 		return;
1323 
1324 	memset(&element, 0, sizeof(element));
1325 	ether_addr_copy(element.mac_addr, macaddr);
1326 	element.vlan_tag = 0;
1327 	/* Ignore error returns, some firmware does it this way... */
1328 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1329 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1330 
1331 	memset(&element, 0, sizeof(element));
1332 	ether_addr_copy(element.mac_addr, macaddr);
1333 	element.vlan_tag = 0;
1334 	/* ...and some firmware does it this way. */
1335 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1336 			I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1337 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1338 }
1339 
1340 /**
1341  * i40e_add_filter - Add a mac/vlan filter to the VSI
1342  * @vsi: the VSI to be searched
1343  * @macaddr: the MAC address
1344  * @vlan: the vlan
1345  *
1346  * Returns ptr to the filter object or NULL when no memory available.
1347  *
1348  * NOTE: This function is expected to be called with mac_filter_hash_lock
1349  * being held.
1350  **/
1351 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1352 					const u8 *macaddr, s16 vlan)
1353 {
1354 	struct i40e_mac_filter *f;
1355 	u64 key;
1356 
1357 	if (!vsi || !macaddr)
1358 		return NULL;
1359 
1360 	f = i40e_find_filter(vsi, macaddr, vlan);
1361 	if (!f) {
1362 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1363 		if (!f)
1364 			return NULL;
1365 
1366 		/* Update the boolean indicating if we need to function in
1367 		 * VLAN mode.
1368 		 */
1369 		if (vlan >= 0)
1370 			vsi->has_vlan_filter = true;
1371 
1372 		ether_addr_copy(f->macaddr, macaddr);
1373 		f->vlan = vlan;
1374 		/* If we're in overflow promisc mode, set the state directly
1375 		 * to failed, so we don't bother to try sending the filter
1376 		 * to the hardware.
1377 		 */
1378 		if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state))
1379 			f->state = I40E_FILTER_FAILED;
1380 		else
1381 			f->state = I40E_FILTER_NEW;
1382 		INIT_HLIST_NODE(&f->hlist);
1383 
1384 		key = i40e_addr_to_hkey(macaddr);
1385 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1386 
1387 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1388 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1389 	}
1390 
1391 	/* If we're asked to add a filter that has been marked for removal, it
1392 	 * is safe to simply restore it to active state. __i40e_del_filter
1393 	 * will have simply deleted any filters which were previously marked
1394 	 * NEW or FAILED, so if it is currently marked REMOVE it must have
1395 	 * previously been ACTIVE. Since we haven't yet run the sync filters
1396 	 * task, just restore this filter to the ACTIVE state so that the
1397 	 * sync task leaves it in place
1398 	 */
1399 	if (f->state == I40E_FILTER_REMOVE)
1400 		f->state = I40E_FILTER_ACTIVE;
1401 
1402 	return f;
1403 }
1404 
1405 /**
1406  * __i40e_del_filter - Remove a specific filter from the VSI
1407  * @vsi: VSI to remove from
1408  * @f: the filter to remove from the list
1409  *
1410  * This function should be called instead of i40e_del_filter only if you know
1411  * the exact filter you will remove already, such as via i40e_find_filter or
1412  * i40e_find_mac.
1413  *
1414  * NOTE: This function is expected to be called with mac_filter_hash_lock
1415  * being held.
1416  * ANOTHER NOTE: This function MUST be called from within the context of
1417  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1418  * instead of list_for_each_entry().
1419  **/
1420 void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1421 {
1422 	if (!f)
1423 		return;
1424 
1425 	/* If the filter was never added to firmware then we can just delete it
1426 	 * directly and we don't want to set the status to remove or else an
1427 	 * admin queue command will unnecessarily fire.
1428 	 */
1429 	if ((f->state == I40E_FILTER_FAILED) ||
1430 	    (f->state == I40E_FILTER_NEW)) {
1431 		hash_del(&f->hlist);
1432 		kfree(f);
1433 	} else {
1434 		f->state = I40E_FILTER_REMOVE;
1435 	}
1436 
1437 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1438 	vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1439 }
1440 
1441 /**
1442  * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1443  * @vsi: the VSI to be searched
1444  * @macaddr: the MAC address
1445  * @vlan: the VLAN
1446  *
1447  * NOTE: This function is expected to be called with mac_filter_hash_lock
1448  * being held.
1449  * ANOTHER NOTE: This function MUST be called from within the context of
1450  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1451  * instead of list_for_each_entry().
1452  **/
1453 void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1454 {
1455 	struct i40e_mac_filter *f;
1456 
1457 	if (!vsi || !macaddr)
1458 		return;
1459 
1460 	f = i40e_find_filter(vsi, macaddr, vlan);
1461 	__i40e_del_filter(vsi, f);
1462 }
1463 
1464 /**
1465  * i40e_add_mac_filter - Add a MAC filter for all active VLANs
1466  * @vsi: the VSI to be searched
1467  * @macaddr: the mac address to be filtered
1468  *
1469  * If we're not in VLAN mode, just add the filter to I40E_VLAN_ANY. Otherwise,
1470  * go through all the macvlan filters and add a macvlan filter for each
1471  * unique vlan that already exists. If a PVID has been assigned, instead only
1472  * add the macaddr to that VLAN.
1473  *
1474  * Returns last filter added on success, else NULL
1475  **/
1476 struct i40e_mac_filter *i40e_add_mac_filter(struct i40e_vsi *vsi,
1477 					    const u8 *macaddr)
1478 {
1479 	struct i40e_mac_filter *f, *add = NULL;
1480 	struct hlist_node *h;
1481 	int bkt;
1482 
1483 	if (vsi->info.pvid)
1484 		return i40e_add_filter(vsi, macaddr,
1485 				       le16_to_cpu(vsi->info.pvid));
1486 
1487 	if (!i40e_is_vsi_in_vlan(vsi))
1488 		return i40e_add_filter(vsi, macaddr, I40E_VLAN_ANY);
1489 
1490 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1491 		if (f->state == I40E_FILTER_REMOVE)
1492 			continue;
1493 		add = i40e_add_filter(vsi, macaddr, f->vlan);
1494 		if (!add)
1495 			return NULL;
1496 	}
1497 
1498 	return add;
1499 }
1500 
1501 /**
1502  * i40e_del_mac_filter - Remove a MAC filter from all VLANs
1503  * @vsi: the VSI to be searched
1504  * @macaddr: the mac address to be removed
1505  *
1506  * Removes a given MAC address from a VSI regardless of what VLAN it has been
1507  * associated with.
1508  *
1509  * Returns 0 for success, or error
1510  **/
1511 int i40e_del_mac_filter(struct i40e_vsi *vsi, const u8 *macaddr)
1512 {
1513 	struct i40e_mac_filter *f;
1514 	struct hlist_node *h;
1515 	bool found = false;
1516 	int bkt;
1517 
1518 	WARN(!spin_is_locked(&vsi->mac_filter_hash_lock),
1519 	     "Missing mac_filter_hash_lock\n");
1520 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1521 		if (ether_addr_equal(macaddr, f->macaddr)) {
1522 			__i40e_del_filter(vsi, f);
1523 			found = true;
1524 		}
1525 	}
1526 
1527 	if (found)
1528 		return 0;
1529 	else
1530 		return -ENOENT;
1531 }
1532 
1533 /**
1534  * i40e_set_mac - NDO callback to set mac address
1535  * @netdev: network interface device structure
1536  * @p: pointer to an address structure
1537  *
1538  * Returns 0 on success, negative on failure
1539  **/
1540 static int i40e_set_mac(struct net_device *netdev, void *p)
1541 {
1542 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1543 	struct i40e_vsi *vsi = np->vsi;
1544 	struct i40e_pf *pf = vsi->back;
1545 	struct i40e_hw *hw = &pf->hw;
1546 	struct sockaddr *addr = p;
1547 
1548 	if (!is_valid_ether_addr(addr->sa_data))
1549 		return -EADDRNOTAVAIL;
1550 
1551 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1552 		netdev_info(netdev, "already using mac address %pM\n",
1553 			    addr->sa_data);
1554 		return 0;
1555 	}
1556 
1557 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
1558 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
1559 		return -EADDRNOTAVAIL;
1560 
1561 	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1562 		netdev_info(netdev, "returning to hw mac address %pM\n",
1563 			    hw->mac.addr);
1564 	else
1565 		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1566 
1567 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1568 	i40e_del_mac_filter(vsi, netdev->dev_addr);
1569 	i40e_add_mac_filter(vsi, addr->sa_data);
1570 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1571 	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1572 	if (vsi->type == I40E_VSI_MAIN) {
1573 		i40e_status ret;
1574 
1575 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1576 						I40E_AQC_WRITE_TYPE_LAA_WOL,
1577 						addr->sa_data, NULL);
1578 		if (ret)
1579 			netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1580 				    i40e_stat_str(hw, ret),
1581 				    i40e_aq_str(hw, hw->aq.asq_last_status));
1582 	}
1583 
1584 	/* schedule our worker thread which will take care of
1585 	 * applying the new filter changes
1586 	 */
1587 	i40e_service_event_schedule(vsi->back);
1588 	return 0;
1589 }
1590 
1591 /**
1592  * i40e_config_rss_aq - Prepare for RSS using AQ commands
1593  * @vsi: vsi structure
1594  * @seed: RSS hash seed
1595  **/
1596 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
1597 			      u8 *lut, u16 lut_size)
1598 {
1599 	struct i40e_pf *pf = vsi->back;
1600 	struct i40e_hw *hw = &pf->hw;
1601 	int ret = 0;
1602 
1603 	if (seed) {
1604 		struct i40e_aqc_get_set_rss_key_data *seed_dw =
1605 			(struct i40e_aqc_get_set_rss_key_data *)seed;
1606 		ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
1607 		if (ret) {
1608 			dev_info(&pf->pdev->dev,
1609 				 "Cannot set RSS key, err %s aq_err %s\n",
1610 				 i40e_stat_str(hw, ret),
1611 				 i40e_aq_str(hw, hw->aq.asq_last_status));
1612 			return ret;
1613 		}
1614 	}
1615 	if (lut) {
1616 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
1617 
1618 		ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
1619 		if (ret) {
1620 			dev_info(&pf->pdev->dev,
1621 				 "Cannot set RSS lut, err %s aq_err %s\n",
1622 				 i40e_stat_str(hw, ret),
1623 				 i40e_aq_str(hw, hw->aq.asq_last_status));
1624 			return ret;
1625 		}
1626 	}
1627 	return ret;
1628 }
1629 
1630 /**
1631  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
1632  * @vsi: VSI structure
1633  **/
1634 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
1635 {
1636 	struct i40e_pf *pf = vsi->back;
1637 	u8 seed[I40E_HKEY_ARRAY_SIZE];
1638 	u8 *lut;
1639 	int ret;
1640 
1641 	if (!(pf->hw_features & I40E_HW_RSS_AQ_CAPABLE))
1642 		return 0;
1643 	if (!vsi->rss_size)
1644 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
1645 				      vsi->num_queue_pairs);
1646 	if (!vsi->rss_size)
1647 		return -EINVAL;
1648 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1649 	if (!lut)
1650 		return -ENOMEM;
1651 
1652 	/* Use the user configured hash keys and lookup table if there is one,
1653 	 * otherwise use default
1654 	 */
1655 	if (vsi->rss_lut_user)
1656 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1657 	else
1658 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
1659 	if (vsi->rss_hkey_user)
1660 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
1661 	else
1662 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
1663 	ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
1664 	kfree(lut);
1665 	return ret;
1666 }
1667 
1668 /**
1669  * i40e_vsi_setup_queue_map_mqprio - Prepares mqprio based tc_config
1670  * @vsi: the VSI being configured,
1671  * @ctxt: VSI context structure
1672  * @enabled_tc: number of traffic classes to enable
1673  *
1674  * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
1675  **/
1676 static int i40e_vsi_setup_queue_map_mqprio(struct i40e_vsi *vsi,
1677 					   struct i40e_vsi_context *ctxt,
1678 					   u8 enabled_tc)
1679 {
1680 	u16 qcount = 0, max_qcount, qmap, sections = 0;
1681 	int i, override_q, pow, num_qps, ret;
1682 	u8 netdev_tc = 0, offset = 0;
1683 
1684 	if (vsi->type != I40E_VSI_MAIN)
1685 		return -EINVAL;
1686 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1687 	sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1688 	vsi->tc_config.numtc = vsi->mqprio_qopt.qopt.num_tc;
1689 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1690 	num_qps = vsi->mqprio_qopt.qopt.count[0];
1691 
1692 	/* find the next higher power-of-2 of num queue pairs */
1693 	pow = ilog2(num_qps);
1694 	if (!is_power_of_2(num_qps))
1695 		pow++;
1696 	qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1697 		(pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1698 
1699 	/* Setup queue offset/count for all TCs for given VSI */
1700 	max_qcount = vsi->mqprio_qopt.qopt.count[0];
1701 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1702 		/* See if the given TC is enabled for the given VSI */
1703 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1704 			offset = vsi->mqprio_qopt.qopt.offset[i];
1705 			qcount = vsi->mqprio_qopt.qopt.count[i];
1706 			if (qcount > max_qcount)
1707 				max_qcount = qcount;
1708 			vsi->tc_config.tc_info[i].qoffset = offset;
1709 			vsi->tc_config.tc_info[i].qcount = qcount;
1710 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1711 		} else {
1712 			/* TC is not enabled so set the offset to
1713 			 * default queue and allocate one queue
1714 			 * for the given TC.
1715 			 */
1716 			vsi->tc_config.tc_info[i].qoffset = 0;
1717 			vsi->tc_config.tc_info[i].qcount = 1;
1718 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1719 		}
1720 	}
1721 
1722 	/* Set actual Tx/Rx queue pairs */
1723 	vsi->num_queue_pairs = offset + qcount;
1724 
1725 	/* Setup queue TC[0].qmap for given VSI context */
1726 	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1727 	ctxt->info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1728 	ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1729 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1730 
1731 	/* Reconfigure RSS for main VSI with max queue count */
1732 	vsi->rss_size = max_qcount;
1733 	ret = i40e_vsi_config_rss(vsi);
1734 	if (ret) {
1735 		dev_info(&vsi->back->pdev->dev,
1736 			 "Failed to reconfig rss for num_queues (%u)\n",
1737 			 max_qcount);
1738 		return ret;
1739 	}
1740 	vsi->reconfig_rss = true;
1741 	dev_dbg(&vsi->back->pdev->dev,
1742 		"Reconfigured rss with num_queues (%u)\n", max_qcount);
1743 
1744 	/* Find queue count available for channel VSIs and starting offset
1745 	 * for channel VSIs
1746 	 */
1747 	override_q = vsi->mqprio_qopt.qopt.count[0];
1748 	if (override_q && override_q < vsi->num_queue_pairs) {
1749 		vsi->cnt_q_avail = vsi->num_queue_pairs - override_q;
1750 		vsi->next_base_queue = override_q;
1751 	}
1752 	return 0;
1753 }
1754 
1755 /**
1756  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1757  * @vsi: the VSI being setup
1758  * @ctxt: VSI context structure
1759  * @enabled_tc: Enabled TCs bitmap
1760  * @is_add: True if called before Add VSI
1761  *
1762  * Setup VSI queue mapping for enabled traffic classes.
1763  **/
1764 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1765 				     struct i40e_vsi_context *ctxt,
1766 				     u8 enabled_tc,
1767 				     bool is_add)
1768 {
1769 	struct i40e_pf *pf = vsi->back;
1770 	u16 sections = 0;
1771 	u8 netdev_tc = 0;
1772 	u16 numtc = 0;
1773 	u16 qcount;
1774 	u8 offset;
1775 	u16 qmap;
1776 	int i;
1777 	u16 num_tc_qps = 0;
1778 
1779 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1780 	offset = 0;
1781 
1782 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1783 		/* Find numtc from enabled TC bitmap */
1784 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1785 			if (enabled_tc & BIT(i)) /* TC is enabled */
1786 				numtc++;
1787 		}
1788 		if (!numtc) {
1789 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1790 			numtc = 1;
1791 		}
1792 	} else {
1793 		/* At least TC0 is enabled in non-DCB, non-MQPRIO case */
1794 		numtc = 1;
1795 	}
1796 
1797 	vsi->tc_config.numtc = numtc;
1798 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1799 	/* Number of queues per enabled TC */
1800 	qcount = vsi->alloc_queue_pairs;
1801 
1802 	num_tc_qps = qcount / numtc;
1803 	num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1804 
1805 	/* Setup queue offset/count for all TCs for given VSI */
1806 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1807 		/* See if the given TC is enabled for the given VSI */
1808 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1809 			/* TC is enabled */
1810 			int pow, num_qps;
1811 
1812 			switch (vsi->type) {
1813 			case I40E_VSI_MAIN:
1814 				qcount = min_t(int, pf->alloc_rss_size,
1815 					       num_tc_qps);
1816 				break;
1817 			case I40E_VSI_FDIR:
1818 			case I40E_VSI_SRIOV:
1819 			case I40E_VSI_VMDQ2:
1820 			default:
1821 				qcount = num_tc_qps;
1822 				WARN_ON(i != 0);
1823 				break;
1824 			}
1825 			vsi->tc_config.tc_info[i].qoffset = offset;
1826 			vsi->tc_config.tc_info[i].qcount = qcount;
1827 
1828 			/* find the next higher power-of-2 of num queue pairs */
1829 			num_qps = qcount;
1830 			pow = 0;
1831 			while (num_qps && (BIT_ULL(pow) < qcount)) {
1832 				pow++;
1833 				num_qps >>= 1;
1834 			}
1835 
1836 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1837 			qmap =
1838 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1839 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1840 
1841 			offset += qcount;
1842 		} else {
1843 			/* TC is not enabled so set the offset to
1844 			 * default queue and allocate one queue
1845 			 * for the given TC.
1846 			 */
1847 			vsi->tc_config.tc_info[i].qoffset = 0;
1848 			vsi->tc_config.tc_info[i].qcount = 1;
1849 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1850 
1851 			qmap = 0;
1852 		}
1853 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1854 	}
1855 
1856 	/* Set actual Tx/Rx queue pairs */
1857 	vsi->num_queue_pairs = offset;
1858 	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1859 		if (vsi->req_queue_pairs > 0)
1860 			vsi->num_queue_pairs = vsi->req_queue_pairs;
1861 		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1862 			vsi->num_queue_pairs = pf->num_lan_msix;
1863 	}
1864 
1865 	/* Scheduler section valid can only be set for ADD VSI */
1866 	if (is_add) {
1867 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1868 
1869 		ctxt->info.up_enable_bits = enabled_tc;
1870 	}
1871 	if (vsi->type == I40E_VSI_SRIOV) {
1872 		ctxt->info.mapping_flags |=
1873 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1874 		for (i = 0; i < vsi->num_queue_pairs; i++)
1875 			ctxt->info.queue_mapping[i] =
1876 					       cpu_to_le16(vsi->base_queue + i);
1877 	} else {
1878 		ctxt->info.mapping_flags |=
1879 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1880 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1881 	}
1882 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1883 }
1884 
1885 /**
1886  * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1887  * @netdev: the netdevice
1888  * @addr: address to add
1889  *
1890  * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1891  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1892  */
1893 static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1894 {
1895 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1896 	struct i40e_vsi *vsi = np->vsi;
1897 
1898 	if (i40e_add_mac_filter(vsi, addr))
1899 		return 0;
1900 	else
1901 		return -ENOMEM;
1902 }
1903 
1904 /**
1905  * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1906  * @netdev: the netdevice
1907  * @addr: address to add
1908  *
1909  * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1910  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1911  */
1912 static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1913 {
1914 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1915 	struct i40e_vsi *vsi = np->vsi;
1916 
1917 	i40e_del_mac_filter(vsi, addr);
1918 
1919 	return 0;
1920 }
1921 
1922 /**
1923  * i40e_set_rx_mode - NDO callback to set the netdev filters
1924  * @netdev: network interface device structure
1925  **/
1926 static void i40e_set_rx_mode(struct net_device *netdev)
1927 {
1928 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1929 	struct i40e_vsi *vsi = np->vsi;
1930 
1931 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1932 
1933 	__dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1934 	__dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1935 
1936 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1937 
1938 	/* check for other flag changes */
1939 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1940 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1941 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1942 	}
1943 }
1944 
1945 /**
1946  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1947  * @vsi: Pointer to VSI struct
1948  * @from: Pointer to list which contains MAC filter entries - changes to
1949  *        those entries needs to be undone.
1950  *
1951  * MAC filter entries from this list were slated for deletion.
1952  **/
1953 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1954 					 struct hlist_head *from)
1955 {
1956 	struct i40e_mac_filter *f;
1957 	struct hlist_node *h;
1958 
1959 	hlist_for_each_entry_safe(f, h, from, hlist) {
1960 		u64 key = i40e_addr_to_hkey(f->macaddr);
1961 
1962 		/* Move the element back into MAC filter list*/
1963 		hlist_del(&f->hlist);
1964 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1965 	}
1966 }
1967 
1968 /**
1969  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1970  * @vsi: Pointer to vsi struct
1971  * @from: Pointer to list which contains MAC filter entries - changes to
1972  *        those entries needs to be undone.
1973  *
1974  * MAC filter entries from this list were slated for addition.
1975  **/
1976 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi,
1977 					 struct hlist_head *from)
1978 {
1979 	struct i40e_new_mac_filter *new;
1980 	struct hlist_node *h;
1981 
1982 	hlist_for_each_entry_safe(new, h, from, hlist) {
1983 		/* We can simply free the wrapper structure */
1984 		hlist_del(&new->hlist);
1985 		kfree(new);
1986 	}
1987 }
1988 
1989 /**
1990  * i40e_next_entry - Get the next non-broadcast filter from a list
1991  * @next: pointer to filter in list
1992  *
1993  * Returns the next non-broadcast filter in the list. Required so that we
1994  * ignore broadcast filters within the list, since these are not handled via
1995  * the normal firmware update path.
1996  */
1997 static
1998 struct i40e_new_mac_filter *i40e_next_filter(struct i40e_new_mac_filter *next)
1999 {
2000 	hlist_for_each_entry_continue(next, hlist) {
2001 		if (!is_broadcast_ether_addr(next->f->macaddr))
2002 			return next;
2003 	}
2004 
2005 	return NULL;
2006 }
2007 
2008 /**
2009  * i40e_update_filter_state - Update filter state based on return data
2010  * from firmware
2011  * @count: Number of filters added
2012  * @add_list: return data from fw
2013  * @head: pointer to first filter in current batch
2014  *
2015  * MAC filter entries from list were slated to be added to device. Returns
2016  * number of successful filters. Note that 0 does NOT mean success!
2017  **/
2018 static int
2019 i40e_update_filter_state(int count,
2020 			 struct i40e_aqc_add_macvlan_element_data *add_list,
2021 			 struct i40e_new_mac_filter *add_head)
2022 {
2023 	int retval = 0;
2024 	int i;
2025 
2026 	for (i = 0; i < count; i++) {
2027 		/* Always check status of each filter. We don't need to check
2028 		 * the firmware return status because we pre-set the filter
2029 		 * status to I40E_AQC_MM_ERR_NO_RES when sending the filter
2030 		 * request to the adminq. Thus, if it no longer matches then
2031 		 * we know the filter is active.
2032 		 */
2033 		if (add_list[i].match_method == I40E_AQC_MM_ERR_NO_RES) {
2034 			add_head->state = I40E_FILTER_FAILED;
2035 		} else {
2036 			add_head->state = I40E_FILTER_ACTIVE;
2037 			retval++;
2038 		}
2039 
2040 		add_head = i40e_next_filter(add_head);
2041 		if (!add_head)
2042 			break;
2043 	}
2044 
2045 	return retval;
2046 }
2047 
2048 /**
2049  * i40e_aqc_del_filters - Request firmware to delete a set of filters
2050  * @vsi: ptr to the VSI
2051  * @vsi_name: name to display in messages
2052  * @list: the list of filters to send to firmware
2053  * @num_del: the number of filters to delete
2054  * @retval: Set to -EIO on failure to delete
2055  *
2056  * Send a request to firmware via AdminQ to delete a set of filters. Uses
2057  * *retval instead of a return value so that success does not force ret_val to
2058  * be set to 0. This ensures that a sequence of calls to this function
2059  * preserve the previous value of *retval on successful delete.
2060  */
2061 static
2062 void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
2063 			  struct i40e_aqc_remove_macvlan_element_data *list,
2064 			  int num_del, int *retval)
2065 {
2066 	struct i40e_hw *hw = &vsi->back->hw;
2067 	i40e_status aq_ret;
2068 	int aq_err;
2069 
2070 	aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
2071 	aq_err = hw->aq.asq_last_status;
2072 
2073 	/* Explicitly ignore and do not report when firmware returns ENOENT */
2074 	if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
2075 		*retval = -EIO;
2076 		dev_info(&vsi->back->pdev->dev,
2077 			 "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
2078 			 vsi_name, i40e_stat_str(hw, aq_ret),
2079 			 i40e_aq_str(hw, aq_err));
2080 	}
2081 }
2082 
2083 /**
2084  * i40e_aqc_add_filters - Request firmware to add a set of filters
2085  * @vsi: ptr to the VSI
2086  * @vsi_name: name to display in messages
2087  * @list: the list of filters to send to firmware
2088  * @add_head: Position in the add hlist
2089  * @num_add: the number of filters to add
2090  * @promisc_change: set to true on exit if promiscuous mode was forced on
2091  *
2092  * Send a request to firmware via AdminQ to add a chunk of filters. Will set
2093  * promisc_changed to true if the firmware has run out of space for more
2094  * filters.
2095  */
2096 static
2097 void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
2098 			  struct i40e_aqc_add_macvlan_element_data *list,
2099 			  struct i40e_new_mac_filter *add_head,
2100 			  int num_add, bool *promisc_changed)
2101 {
2102 	struct i40e_hw *hw = &vsi->back->hw;
2103 	int aq_err, fcnt;
2104 
2105 	i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
2106 	aq_err = hw->aq.asq_last_status;
2107 	fcnt = i40e_update_filter_state(num_add, list, add_head);
2108 
2109 	if (fcnt != num_add) {
2110 		*promisc_changed = true;
2111 		set_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2112 		dev_warn(&vsi->back->pdev->dev,
2113 			 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2114 			 i40e_aq_str(hw, aq_err),
2115 			 vsi_name);
2116 	}
2117 }
2118 
2119 /**
2120  * i40e_aqc_broadcast_filter - Set promiscuous broadcast flags
2121  * @vsi: pointer to the VSI
2122  * @f: filter data
2123  *
2124  * This function sets or clears the promiscuous broadcast flags for VLAN
2125  * filters in order to properly receive broadcast frames. Assumes that only
2126  * broadcast filters are passed.
2127  *
2128  * Returns status indicating success or failure;
2129  **/
2130 static i40e_status
2131 i40e_aqc_broadcast_filter(struct i40e_vsi *vsi, const char *vsi_name,
2132 			  struct i40e_mac_filter *f)
2133 {
2134 	bool enable = f->state == I40E_FILTER_NEW;
2135 	struct i40e_hw *hw = &vsi->back->hw;
2136 	i40e_status aq_ret;
2137 
2138 	if (f->vlan == I40E_VLAN_ANY) {
2139 		aq_ret = i40e_aq_set_vsi_broadcast(hw,
2140 						   vsi->seid,
2141 						   enable,
2142 						   NULL);
2143 	} else {
2144 		aq_ret = i40e_aq_set_vsi_bc_promisc_on_vlan(hw,
2145 							    vsi->seid,
2146 							    enable,
2147 							    f->vlan,
2148 							    NULL);
2149 	}
2150 
2151 	if (aq_ret)
2152 		dev_warn(&vsi->back->pdev->dev,
2153 			 "Error %s setting broadcast promiscuous mode on %s\n",
2154 			 i40e_aq_str(hw, hw->aq.asq_last_status),
2155 			 vsi_name);
2156 
2157 	return aq_ret;
2158 }
2159 
2160 /**
2161  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
2162  * @vsi: ptr to the VSI
2163  *
2164  * Push any outstanding VSI filter changes through the AdminQ.
2165  *
2166  * Returns 0 or error value
2167  **/
2168 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
2169 {
2170 	struct hlist_head tmp_add_list, tmp_del_list;
2171 	struct i40e_mac_filter *f;
2172 	struct i40e_new_mac_filter *new, *add_head = NULL;
2173 	struct i40e_hw *hw = &vsi->back->hw;
2174 	unsigned int failed_filters = 0;
2175 	unsigned int vlan_filters = 0;
2176 	bool promisc_changed = false;
2177 	char vsi_name[16] = "PF";
2178 	int filter_list_len = 0;
2179 	i40e_status aq_ret = 0;
2180 	u32 changed_flags = 0;
2181 	struct hlist_node *h;
2182 	struct i40e_pf *pf;
2183 	int num_add = 0;
2184 	int num_del = 0;
2185 	int retval = 0;
2186 	u16 cmd_flags;
2187 	int list_size;
2188 	int bkt;
2189 
2190 	/* empty array typed pointers, kcalloc later */
2191 	struct i40e_aqc_add_macvlan_element_data *add_list;
2192 	struct i40e_aqc_remove_macvlan_element_data *del_list;
2193 
2194 	while (test_and_set_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state))
2195 		usleep_range(1000, 2000);
2196 	pf = vsi->back;
2197 
2198 	if (vsi->netdev) {
2199 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
2200 		vsi->current_netdev_flags = vsi->netdev->flags;
2201 	}
2202 
2203 	INIT_HLIST_HEAD(&tmp_add_list);
2204 	INIT_HLIST_HEAD(&tmp_del_list);
2205 
2206 	if (vsi->type == I40E_VSI_SRIOV)
2207 		snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
2208 	else if (vsi->type != I40E_VSI_MAIN)
2209 		snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
2210 
2211 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
2212 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
2213 
2214 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2215 		/* Create a list of filters to delete. */
2216 		hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2217 			if (f->state == I40E_FILTER_REMOVE) {
2218 				/* Move the element into temporary del_list */
2219 				hash_del(&f->hlist);
2220 				hlist_add_head(&f->hlist, &tmp_del_list);
2221 
2222 				/* Avoid counting removed filters */
2223 				continue;
2224 			}
2225 			if (f->state == I40E_FILTER_NEW) {
2226 				/* Create a temporary i40e_new_mac_filter */
2227 				new = kzalloc(sizeof(*new), GFP_ATOMIC);
2228 				if (!new)
2229 					goto err_no_memory_locked;
2230 
2231 				/* Store pointer to the real filter */
2232 				new->f = f;
2233 				new->state = f->state;
2234 
2235 				/* Add it to the hash list */
2236 				hlist_add_head(&new->hlist, &tmp_add_list);
2237 			}
2238 
2239 			/* Count the number of active (current and new) VLAN
2240 			 * filters we have now. Does not count filters which
2241 			 * are marked for deletion.
2242 			 */
2243 			if (f->vlan > 0)
2244 				vlan_filters++;
2245 		}
2246 
2247 		retval = i40e_correct_mac_vlan_filters(vsi,
2248 						       &tmp_add_list,
2249 						       &tmp_del_list,
2250 						       vlan_filters);
2251 		if (retval)
2252 			goto err_no_memory_locked;
2253 
2254 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2255 	}
2256 
2257 	/* Now process 'del_list' outside the lock */
2258 	if (!hlist_empty(&tmp_del_list)) {
2259 		filter_list_len = hw->aq.asq_buf_size /
2260 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2261 		list_size = filter_list_len *
2262 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2263 		del_list = kzalloc(list_size, GFP_ATOMIC);
2264 		if (!del_list)
2265 			goto err_no_memory;
2266 
2267 		hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2268 			cmd_flags = 0;
2269 
2270 			/* handle broadcast filters by updating the broadcast
2271 			 * promiscuous flag and release filter list.
2272 			 */
2273 			if (is_broadcast_ether_addr(f->macaddr)) {
2274 				i40e_aqc_broadcast_filter(vsi, vsi_name, f);
2275 
2276 				hlist_del(&f->hlist);
2277 				kfree(f);
2278 				continue;
2279 			}
2280 
2281 			/* add to delete list */
2282 			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2283 			if (f->vlan == I40E_VLAN_ANY) {
2284 				del_list[num_del].vlan_tag = 0;
2285 				cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2286 			} else {
2287 				del_list[num_del].vlan_tag =
2288 					cpu_to_le16((u16)(f->vlan));
2289 			}
2290 
2291 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2292 			del_list[num_del].flags = cmd_flags;
2293 			num_del++;
2294 
2295 			/* flush a full buffer */
2296 			if (num_del == filter_list_len) {
2297 				i40e_aqc_del_filters(vsi, vsi_name, del_list,
2298 						     num_del, &retval);
2299 				memset(del_list, 0, list_size);
2300 				num_del = 0;
2301 			}
2302 			/* Release memory for MAC filter entries which were
2303 			 * synced up with HW.
2304 			 */
2305 			hlist_del(&f->hlist);
2306 			kfree(f);
2307 		}
2308 
2309 		if (num_del) {
2310 			i40e_aqc_del_filters(vsi, vsi_name, del_list,
2311 					     num_del, &retval);
2312 		}
2313 
2314 		kfree(del_list);
2315 		del_list = NULL;
2316 	}
2317 
2318 	if (!hlist_empty(&tmp_add_list)) {
2319 		/* Do all the adds now. */
2320 		filter_list_len = hw->aq.asq_buf_size /
2321 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2322 		list_size = filter_list_len *
2323 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2324 		add_list = kzalloc(list_size, GFP_ATOMIC);
2325 		if (!add_list)
2326 			goto err_no_memory;
2327 
2328 		num_add = 0;
2329 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2330 			if (test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2331 				     vsi->state)) {
2332 				new->state = I40E_FILTER_FAILED;
2333 				continue;
2334 			}
2335 
2336 			/* handle broadcast filters by updating the broadcast
2337 			 * promiscuous flag instead of adding a MAC filter.
2338 			 */
2339 			if (is_broadcast_ether_addr(new->f->macaddr)) {
2340 				if (i40e_aqc_broadcast_filter(vsi, vsi_name,
2341 							      new->f))
2342 					new->state = I40E_FILTER_FAILED;
2343 				else
2344 					new->state = I40E_FILTER_ACTIVE;
2345 				continue;
2346 			}
2347 
2348 			/* add to add array */
2349 			if (num_add == 0)
2350 				add_head = new;
2351 			cmd_flags = 0;
2352 			ether_addr_copy(add_list[num_add].mac_addr,
2353 					new->f->macaddr);
2354 			if (new->f->vlan == I40E_VLAN_ANY) {
2355 				add_list[num_add].vlan_tag = 0;
2356 				cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2357 			} else {
2358 				add_list[num_add].vlan_tag =
2359 					cpu_to_le16((u16)(new->f->vlan));
2360 			}
2361 			add_list[num_add].queue_number = 0;
2362 			/* set invalid match method for later detection */
2363 			add_list[num_add].match_method = I40E_AQC_MM_ERR_NO_RES;
2364 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2365 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
2366 			num_add++;
2367 
2368 			/* flush a full buffer */
2369 			if (num_add == filter_list_len) {
2370 				i40e_aqc_add_filters(vsi, vsi_name, add_list,
2371 						     add_head, num_add,
2372 						     &promisc_changed);
2373 				memset(add_list, 0, list_size);
2374 				num_add = 0;
2375 			}
2376 		}
2377 		if (num_add) {
2378 			i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2379 					     num_add, &promisc_changed);
2380 		}
2381 		/* Now move all of the filters from the temp add list back to
2382 		 * the VSI's list.
2383 		 */
2384 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2385 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2386 			/* Only update the state if we're still NEW */
2387 			if (new->f->state == I40E_FILTER_NEW)
2388 				new->f->state = new->state;
2389 			hlist_del(&new->hlist);
2390 			kfree(new);
2391 		}
2392 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2393 		kfree(add_list);
2394 		add_list = NULL;
2395 	}
2396 
2397 	/* Determine the number of active and failed filters. */
2398 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2399 	vsi->active_filters = 0;
2400 	hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2401 		if (f->state == I40E_FILTER_ACTIVE)
2402 			vsi->active_filters++;
2403 		else if (f->state == I40E_FILTER_FAILED)
2404 			failed_filters++;
2405 	}
2406 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2407 
2408 	/* If promiscuous mode has changed, we need to calculate a new
2409 	 * threshold for when we are safe to exit
2410 	 */
2411 	if (promisc_changed)
2412 		vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
2413 
2414 	/* Check if we are able to exit overflow promiscuous mode. We can
2415 	 * safely exit if we didn't just enter, we no longer have any failed
2416 	 * filters, and we have reduced filters below the threshold value.
2417 	 */
2418 	if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state) &&
2419 	    !promisc_changed && !failed_filters &&
2420 	    (vsi->active_filters < vsi->promisc_threshold)) {
2421 		dev_info(&pf->pdev->dev,
2422 			 "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2423 			 vsi_name);
2424 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2425 		promisc_changed = true;
2426 		vsi->promisc_threshold = 0;
2427 	}
2428 
2429 	/* if the VF is not trusted do not do promisc */
2430 	if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2431 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2432 		goto out;
2433 	}
2434 
2435 	/* check for changes in promiscuous modes */
2436 	if (changed_flags & IFF_ALLMULTI) {
2437 		bool cur_multipromisc;
2438 
2439 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2440 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2441 							       vsi->seid,
2442 							       cur_multipromisc,
2443 							       NULL);
2444 		if (aq_ret) {
2445 			retval = i40e_aq_rc_to_posix(aq_ret,
2446 						     hw->aq.asq_last_status);
2447 			dev_info(&pf->pdev->dev,
2448 				 "set multi promisc failed on %s, err %s aq_err %s\n",
2449 				 vsi_name,
2450 				 i40e_stat_str(hw, aq_ret),
2451 				 i40e_aq_str(hw, hw->aq.asq_last_status));
2452 		}
2453 	}
2454 
2455 	if ((changed_flags & IFF_PROMISC) || promisc_changed) {
2456 		bool cur_promisc;
2457 
2458 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2459 			       test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2460 					vsi->state));
2461 		if ((vsi->type == I40E_VSI_MAIN) &&
2462 		    (pf->lan_veb != I40E_NO_VEB) &&
2463 		    !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2464 			/* set defport ON for Main VSI instead of true promisc
2465 			 * this way we will get all unicast/multicast and VLAN
2466 			 * promisc behavior but will not get VF or VMDq traffic
2467 			 * replicated on the Main VSI.
2468 			 */
2469 			if (pf->cur_promisc != cur_promisc) {
2470 				pf->cur_promisc = cur_promisc;
2471 				if (cur_promisc)
2472 					aq_ret =
2473 					      i40e_aq_set_default_vsi(hw,
2474 								      vsi->seid,
2475 								      NULL);
2476 				else
2477 					aq_ret =
2478 					    i40e_aq_clear_default_vsi(hw,
2479 								      vsi->seid,
2480 								      NULL);
2481 				if (aq_ret) {
2482 					retval = i40e_aq_rc_to_posix(aq_ret,
2483 							hw->aq.asq_last_status);
2484 					dev_info(&pf->pdev->dev,
2485 						 "Set default VSI failed on %s, err %s, aq_err %s\n",
2486 						 vsi_name,
2487 						 i40e_stat_str(hw, aq_ret),
2488 						 i40e_aq_str(hw,
2489 						     hw->aq.asq_last_status));
2490 				}
2491 			}
2492 		} else {
2493 			aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2494 							  hw,
2495 							  vsi->seid,
2496 							  cur_promisc, NULL,
2497 							  true);
2498 			if (aq_ret) {
2499 				retval =
2500 				i40e_aq_rc_to_posix(aq_ret,
2501 						    hw->aq.asq_last_status);
2502 				dev_info(&pf->pdev->dev,
2503 					 "set unicast promisc failed on %s, err %s, aq_err %s\n",
2504 					 vsi_name,
2505 					 i40e_stat_str(hw, aq_ret),
2506 					 i40e_aq_str(hw,
2507 						     hw->aq.asq_last_status));
2508 			}
2509 			aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2510 							  hw,
2511 							  vsi->seid,
2512 							  cur_promisc, NULL);
2513 			if (aq_ret) {
2514 				retval =
2515 				i40e_aq_rc_to_posix(aq_ret,
2516 						    hw->aq.asq_last_status);
2517 				dev_info(&pf->pdev->dev,
2518 					 "set multicast promisc failed on %s, err %s, aq_err %s\n",
2519 					 vsi_name,
2520 					 i40e_stat_str(hw, aq_ret),
2521 					 i40e_aq_str(hw,
2522 						     hw->aq.asq_last_status));
2523 			}
2524 		}
2525 		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2526 						   vsi->seid,
2527 						   cur_promisc, NULL);
2528 		if (aq_ret) {
2529 			retval = i40e_aq_rc_to_posix(aq_ret,
2530 						     pf->hw.aq.asq_last_status);
2531 			dev_info(&pf->pdev->dev,
2532 				 "set brdcast promisc failed, err %s, aq_err %s\n",
2533 					 i40e_stat_str(hw, aq_ret),
2534 					 i40e_aq_str(hw,
2535 						     hw->aq.asq_last_status));
2536 		}
2537 	}
2538 out:
2539 	/* if something went wrong then set the changed flag so we try again */
2540 	if (retval)
2541 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2542 
2543 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2544 	return retval;
2545 
2546 err_no_memory:
2547 	/* Restore elements on the temporary add and delete lists */
2548 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2549 err_no_memory_locked:
2550 	i40e_undo_del_filter_entries(vsi, &tmp_del_list);
2551 	i40e_undo_add_filter_entries(vsi, &tmp_add_list);
2552 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2553 
2554 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2555 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2556 	return -ENOMEM;
2557 }
2558 
2559 /**
2560  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2561  * @pf: board private structure
2562  **/
2563 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2564 {
2565 	int v;
2566 
2567 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2568 		return;
2569 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2570 
2571 	for (v = 0; v < pf->num_alloc_vsi; v++) {
2572 		if (pf->vsi[v] &&
2573 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2574 			int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2575 
2576 			if (ret) {
2577 				/* come back and try again later */
2578 				pf->flags |= I40E_FLAG_FILTER_SYNC;
2579 				break;
2580 			}
2581 		}
2582 	}
2583 }
2584 
2585 /**
2586  * i40e_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2587  * @vsi: the vsi
2588  **/
2589 static int i40e_max_xdp_frame_size(struct i40e_vsi *vsi)
2590 {
2591 	if (PAGE_SIZE >= 8192 || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
2592 		return I40E_RXBUFFER_2048;
2593 	else
2594 		return I40E_RXBUFFER_3072;
2595 }
2596 
2597 /**
2598  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2599  * @netdev: network interface device structure
2600  * @new_mtu: new value for maximum frame size
2601  *
2602  * Returns 0 on success, negative on failure
2603  **/
2604 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2605 {
2606 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2607 	struct i40e_vsi *vsi = np->vsi;
2608 	struct i40e_pf *pf = vsi->back;
2609 
2610 	if (i40e_enabled_xdp_vsi(vsi)) {
2611 		int frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2612 
2613 		if (frame_size > i40e_max_xdp_frame_size(vsi))
2614 			return -EINVAL;
2615 	}
2616 
2617 	netdev_info(netdev, "changing MTU from %d to %d\n",
2618 		    netdev->mtu, new_mtu);
2619 	netdev->mtu = new_mtu;
2620 	if (netif_running(netdev))
2621 		i40e_vsi_reinit_locked(vsi);
2622 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
2623 		      I40E_FLAG_CLIENT_L2_CHANGE);
2624 	return 0;
2625 }
2626 
2627 /**
2628  * i40e_ioctl - Access the hwtstamp interface
2629  * @netdev: network interface device structure
2630  * @ifr: interface request data
2631  * @cmd: ioctl command
2632  **/
2633 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2634 {
2635 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2636 	struct i40e_pf *pf = np->vsi->back;
2637 
2638 	switch (cmd) {
2639 	case SIOCGHWTSTAMP:
2640 		return i40e_ptp_get_ts_config(pf, ifr);
2641 	case SIOCSHWTSTAMP:
2642 		return i40e_ptp_set_ts_config(pf, ifr);
2643 	default:
2644 		return -EOPNOTSUPP;
2645 	}
2646 }
2647 
2648 /**
2649  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2650  * @vsi: the vsi being adjusted
2651  **/
2652 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2653 {
2654 	struct i40e_vsi_context ctxt;
2655 	i40e_status ret;
2656 
2657 	if ((vsi->info.valid_sections &
2658 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2659 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2660 		return;  /* already enabled */
2661 
2662 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2663 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2664 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2665 
2666 	ctxt.seid = vsi->seid;
2667 	ctxt.info = vsi->info;
2668 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2669 	if (ret) {
2670 		dev_info(&vsi->back->pdev->dev,
2671 			 "update vlan stripping failed, err %s aq_err %s\n",
2672 			 i40e_stat_str(&vsi->back->hw, ret),
2673 			 i40e_aq_str(&vsi->back->hw,
2674 				     vsi->back->hw.aq.asq_last_status));
2675 	}
2676 }
2677 
2678 /**
2679  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2680  * @vsi: the vsi being adjusted
2681  **/
2682 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2683 {
2684 	struct i40e_vsi_context ctxt;
2685 	i40e_status ret;
2686 
2687 	if ((vsi->info.valid_sections &
2688 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2689 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2690 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2691 		return;  /* already disabled */
2692 
2693 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2694 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2695 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2696 
2697 	ctxt.seid = vsi->seid;
2698 	ctxt.info = vsi->info;
2699 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2700 	if (ret) {
2701 		dev_info(&vsi->back->pdev->dev,
2702 			 "update vlan stripping failed, err %s aq_err %s\n",
2703 			 i40e_stat_str(&vsi->back->hw, ret),
2704 			 i40e_aq_str(&vsi->back->hw,
2705 				     vsi->back->hw.aq.asq_last_status));
2706 	}
2707 }
2708 
2709 /**
2710  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2711  * @netdev: network interface to be adjusted
2712  * @features: netdev features to test if VLAN offload is enabled or not
2713  **/
2714 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2715 {
2716 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2717 	struct i40e_vsi *vsi = np->vsi;
2718 
2719 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2720 		i40e_vlan_stripping_enable(vsi);
2721 	else
2722 		i40e_vlan_stripping_disable(vsi);
2723 }
2724 
2725 /**
2726  * i40e_add_vlan_all_mac - Add a MAC/VLAN filter for each existing MAC address
2727  * @vsi: the vsi being configured
2728  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2729  *
2730  * This is a helper function for adding a new MAC/VLAN filter with the
2731  * specified VLAN for each existing MAC address already in the hash table.
2732  * This function does *not* perform any accounting to update filters based on
2733  * VLAN mode.
2734  *
2735  * NOTE: this function expects to be called while under the
2736  * mac_filter_hash_lock
2737  **/
2738 int i40e_add_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2739 {
2740 	struct i40e_mac_filter *f, *add_f;
2741 	struct hlist_node *h;
2742 	int bkt;
2743 
2744 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2745 		if (f->state == I40E_FILTER_REMOVE)
2746 			continue;
2747 		add_f = i40e_add_filter(vsi, f->macaddr, vid);
2748 		if (!add_f) {
2749 			dev_info(&vsi->back->pdev->dev,
2750 				 "Could not add vlan filter %d for %pM\n",
2751 				 vid, f->macaddr);
2752 			return -ENOMEM;
2753 		}
2754 	}
2755 
2756 	return 0;
2757 }
2758 
2759 /**
2760  * i40e_vsi_add_vlan - Add VSI membership for given VLAN
2761  * @vsi: the VSI being configured
2762  * @vid: VLAN id to be added
2763  **/
2764 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, u16 vid)
2765 {
2766 	int err;
2767 
2768 	if (vsi->info.pvid)
2769 		return -EINVAL;
2770 
2771 	/* The network stack will attempt to add VID=0, with the intention to
2772 	 * receive priority tagged packets with a VLAN of 0. Our HW receives
2773 	 * these packets by default when configured to receive untagged
2774 	 * packets, so we don't need to add a filter for this case.
2775 	 * Additionally, HW interprets adding a VID=0 filter as meaning to
2776 	 * receive *only* tagged traffic and stops receiving untagged traffic.
2777 	 * Thus, we do not want to actually add a filter for VID=0
2778 	 */
2779 	if (!vid)
2780 		return 0;
2781 
2782 	/* Locked once because all functions invoked below iterates list*/
2783 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2784 	err = i40e_add_vlan_all_mac(vsi, vid);
2785 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2786 	if (err)
2787 		return err;
2788 
2789 	/* schedule our worker thread which will take care of
2790 	 * applying the new filter changes
2791 	 */
2792 	i40e_service_event_schedule(vsi->back);
2793 	return 0;
2794 }
2795 
2796 /**
2797  * i40e_rm_vlan_all_mac - Remove MAC/VLAN pair for all MAC with the given VLAN
2798  * @vsi: the vsi being configured
2799  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2800  *
2801  * This function should be used to remove all VLAN filters which match the
2802  * given VID. It does not schedule the service event and does not take the
2803  * mac_filter_hash_lock so it may be combined with other operations under
2804  * a single invocation of the mac_filter_hash_lock.
2805  *
2806  * NOTE: this function expects to be called while under the
2807  * mac_filter_hash_lock
2808  */
2809 void i40e_rm_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2810 {
2811 	struct i40e_mac_filter *f;
2812 	struct hlist_node *h;
2813 	int bkt;
2814 
2815 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2816 		if (f->vlan == vid)
2817 			__i40e_del_filter(vsi, f);
2818 	}
2819 }
2820 
2821 /**
2822  * i40e_vsi_kill_vlan - Remove VSI membership for given VLAN
2823  * @vsi: the VSI being configured
2824  * @vid: VLAN id to be removed
2825  **/
2826 void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, u16 vid)
2827 {
2828 	if (!vid || vsi->info.pvid)
2829 		return;
2830 
2831 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2832 	i40e_rm_vlan_all_mac(vsi, vid);
2833 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2834 
2835 	/* schedule our worker thread which will take care of
2836 	 * applying the new filter changes
2837 	 */
2838 	i40e_service_event_schedule(vsi->back);
2839 }
2840 
2841 /**
2842  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2843  * @netdev: network interface to be adjusted
2844  * @vid: vlan id to be added
2845  *
2846  * net_device_ops implementation for adding vlan ids
2847  **/
2848 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2849 				__always_unused __be16 proto, u16 vid)
2850 {
2851 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2852 	struct i40e_vsi *vsi = np->vsi;
2853 	int ret = 0;
2854 
2855 	if (vid >= VLAN_N_VID)
2856 		return -EINVAL;
2857 
2858 	ret = i40e_vsi_add_vlan(vsi, vid);
2859 	if (!ret)
2860 		set_bit(vid, vsi->active_vlans);
2861 
2862 	return ret;
2863 }
2864 
2865 /**
2866  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2867  * @netdev: network interface to be adjusted
2868  * @vid: vlan id to be removed
2869  *
2870  * net_device_ops implementation for removing vlan ids
2871  **/
2872 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2873 				 __always_unused __be16 proto, u16 vid)
2874 {
2875 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2876 	struct i40e_vsi *vsi = np->vsi;
2877 
2878 	/* return code is ignored as there is nothing a user
2879 	 * can do about failure to remove and a log message was
2880 	 * already printed from the other function
2881 	 */
2882 	i40e_vsi_kill_vlan(vsi, vid);
2883 
2884 	clear_bit(vid, vsi->active_vlans);
2885 
2886 	return 0;
2887 }
2888 
2889 /**
2890  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2891  * @vsi: the vsi being brought back up
2892  **/
2893 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2894 {
2895 	u16 vid;
2896 
2897 	if (!vsi->netdev)
2898 		return;
2899 
2900 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2901 
2902 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2903 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2904 				     vid);
2905 }
2906 
2907 /**
2908  * i40e_vsi_add_pvid - Add pvid for the VSI
2909  * @vsi: the vsi being adjusted
2910  * @vid: the vlan id to set as a PVID
2911  **/
2912 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2913 {
2914 	struct i40e_vsi_context ctxt;
2915 	i40e_status ret;
2916 
2917 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2918 	vsi->info.pvid = cpu_to_le16(vid);
2919 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2920 				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2921 				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2922 
2923 	ctxt.seid = vsi->seid;
2924 	ctxt.info = vsi->info;
2925 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2926 	if (ret) {
2927 		dev_info(&vsi->back->pdev->dev,
2928 			 "add pvid failed, err %s aq_err %s\n",
2929 			 i40e_stat_str(&vsi->back->hw, ret),
2930 			 i40e_aq_str(&vsi->back->hw,
2931 				     vsi->back->hw.aq.asq_last_status));
2932 		return -ENOENT;
2933 	}
2934 
2935 	return 0;
2936 }
2937 
2938 /**
2939  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2940  * @vsi: the vsi being adjusted
2941  *
2942  * Just use the vlan_rx_register() service to put it back to normal
2943  **/
2944 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2945 {
2946 	i40e_vlan_stripping_disable(vsi);
2947 
2948 	vsi->info.pvid = 0;
2949 }
2950 
2951 /**
2952  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2953  * @vsi: ptr to the VSI
2954  *
2955  * If this function returns with an error, then it's possible one or
2956  * more of the rings is populated (while the rest are not).  It is the
2957  * callers duty to clean those orphaned rings.
2958  *
2959  * Return 0 on success, negative on failure
2960  **/
2961 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2962 {
2963 	int i, err = 0;
2964 
2965 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2966 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2967 
2968 	if (!i40e_enabled_xdp_vsi(vsi))
2969 		return err;
2970 
2971 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2972 		err = i40e_setup_tx_descriptors(vsi->xdp_rings[i]);
2973 
2974 	return err;
2975 }
2976 
2977 /**
2978  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2979  * @vsi: ptr to the VSI
2980  *
2981  * Free VSI's transmit software resources
2982  **/
2983 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2984 {
2985 	int i;
2986 
2987 	if (vsi->tx_rings) {
2988 		for (i = 0; i < vsi->num_queue_pairs; i++)
2989 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2990 				i40e_free_tx_resources(vsi->tx_rings[i]);
2991 	}
2992 
2993 	if (vsi->xdp_rings) {
2994 		for (i = 0; i < vsi->num_queue_pairs; i++)
2995 			if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2996 				i40e_free_tx_resources(vsi->xdp_rings[i]);
2997 	}
2998 }
2999 
3000 /**
3001  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
3002  * @vsi: ptr to the VSI
3003  *
3004  * If this function returns with an error, then it's possible one or
3005  * more of the rings is populated (while the rest are not).  It is the
3006  * callers duty to clean those orphaned rings.
3007  *
3008  * Return 0 on success, negative on failure
3009  **/
3010 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
3011 {
3012 	int i, err = 0;
3013 
3014 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3015 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
3016 	return err;
3017 }
3018 
3019 /**
3020  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
3021  * @vsi: ptr to the VSI
3022  *
3023  * Free all receive software resources
3024  **/
3025 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
3026 {
3027 	int i;
3028 
3029 	if (!vsi->rx_rings)
3030 		return;
3031 
3032 	for (i = 0; i < vsi->num_queue_pairs; i++)
3033 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
3034 			i40e_free_rx_resources(vsi->rx_rings[i]);
3035 }
3036 
3037 /**
3038  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
3039  * @ring: The Tx ring to configure
3040  *
3041  * This enables/disables XPS for a given Tx descriptor ring
3042  * based on the TCs enabled for the VSI that ring belongs to.
3043  **/
3044 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
3045 {
3046 	int cpu;
3047 
3048 	if (!ring->q_vector || !ring->netdev || ring->ch)
3049 		return;
3050 
3051 	/* We only initialize XPS once, so as not to overwrite user settings */
3052 	if (test_and_set_bit(__I40E_TX_XPS_INIT_DONE, ring->state))
3053 		return;
3054 
3055 	cpu = cpumask_local_spread(ring->q_vector->v_idx, -1);
3056 	netif_set_xps_queue(ring->netdev, get_cpu_mask(cpu),
3057 			    ring->queue_index);
3058 }
3059 
3060 /**
3061  * i40e_configure_tx_ring - Configure a transmit ring context and rest
3062  * @ring: The Tx ring to configure
3063  *
3064  * Configure the Tx descriptor ring in the HMC context.
3065  **/
3066 static int i40e_configure_tx_ring(struct i40e_ring *ring)
3067 {
3068 	struct i40e_vsi *vsi = ring->vsi;
3069 	u16 pf_q = vsi->base_queue + ring->queue_index;
3070 	struct i40e_hw *hw = &vsi->back->hw;
3071 	struct i40e_hmc_obj_txq tx_ctx;
3072 	i40e_status err = 0;
3073 	u32 qtx_ctl = 0;
3074 
3075 	/* some ATR related tx ring init */
3076 	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
3077 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
3078 		ring->atr_count = 0;
3079 	} else {
3080 		ring->atr_sample_rate = 0;
3081 	}
3082 
3083 	/* configure XPS */
3084 	i40e_config_xps_tx_ring(ring);
3085 
3086 	/* clear the context structure first */
3087 	memset(&tx_ctx, 0, sizeof(tx_ctx));
3088 
3089 	tx_ctx.new_context = 1;
3090 	tx_ctx.base = (ring->dma / 128);
3091 	tx_ctx.qlen = ring->count;
3092 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
3093 					       I40E_FLAG_FD_ATR_ENABLED));
3094 	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
3095 	/* FDIR VSI tx ring can still use RS bit and writebacks */
3096 	if (vsi->type != I40E_VSI_FDIR)
3097 		tx_ctx.head_wb_ena = 1;
3098 	tx_ctx.head_wb_addr = ring->dma +
3099 			      (ring->count * sizeof(struct i40e_tx_desc));
3100 
3101 	/* As part of VSI creation/update, FW allocates certain
3102 	 * Tx arbitration queue sets for each TC enabled for
3103 	 * the VSI. The FW returns the handles to these queue
3104 	 * sets as part of the response buffer to Add VSI,
3105 	 * Update VSI, etc. AQ commands. It is expected that
3106 	 * these queue set handles be associated with the Tx
3107 	 * queues by the driver as part of the TX queue context
3108 	 * initialization. This has to be done regardless of
3109 	 * DCB as by default everything is mapped to TC0.
3110 	 */
3111 
3112 	if (ring->ch)
3113 		tx_ctx.rdylist =
3114 			le16_to_cpu(ring->ch->info.qs_handle[ring->dcb_tc]);
3115 
3116 	else
3117 		tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
3118 
3119 	tx_ctx.rdylist_act = 0;
3120 
3121 	/* clear the context in the HMC */
3122 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
3123 	if (err) {
3124 		dev_info(&vsi->back->pdev->dev,
3125 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
3126 			 ring->queue_index, pf_q, err);
3127 		return -ENOMEM;
3128 	}
3129 
3130 	/* set the context in the HMC */
3131 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
3132 	if (err) {
3133 		dev_info(&vsi->back->pdev->dev,
3134 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
3135 			 ring->queue_index, pf_q, err);
3136 		return -ENOMEM;
3137 	}
3138 
3139 	/* Now associate this queue with this PCI function */
3140 	if (ring->ch) {
3141 		if (ring->ch->type == I40E_VSI_VMDQ2)
3142 			qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
3143 		else
3144 			return -EINVAL;
3145 
3146 		qtx_ctl |= (ring->ch->vsi_number <<
3147 			    I40E_QTX_CTL_VFVM_INDX_SHIFT) &
3148 			    I40E_QTX_CTL_VFVM_INDX_MASK;
3149 	} else {
3150 		if (vsi->type == I40E_VSI_VMDQ2) {
3151 			qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
3152 			qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
3153 				    I40E_QTX_CTL_VFVM_INDX_MASK;
3154 		} else {
3155 			qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
3156 		}
3157 	}
3158 
3159 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
3160 		    I40E_QTX_CTL_PF_INDX_MASK);
3161 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
3162 	i40e_flush(hw);
3163 
3164 	/* cache tail off for easier writes later */
3165 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
3166 
3167 	return 0;
3168 }
3169 
3170 /**
3171  * i40e_configure_rx_ring - Configure a receive ring context
3172  * @ring: The Rx ring to configure
3173  *
3174  * Configure the Rx descriptor ring in the HMC context.
3175  **/
3176 static int i40e_configure_rx_ring(struct i40e_ring *ring)
3177 {
3178 	struct i40e_vsi *vsi = ring->vsi;
3179 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
3180 	u16 pf_q = vsi->base_queue + ring->queue_index;
3181 	struct i40e_hw *hw = &vsi->back->hw;
3182 	struct i40e_hmc_obj_rxq rx_ctx;
3183 	i40e_status err = 0;
3184 
3185 	bitmap_zero(ring->state, __I40E_RING_STATE_NBITS);
3186 
3187 	/* clear the context structure first */
3188 	memset(&rx_ctx, 0, sizeof(rx_ctx));
3189 
3190 	ring->rx_buf_len = vsi->rx_buf_len;
3191 
3192 	rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3193 				    BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3194 
3195 	rx_ctx.base = (ring->dma / 128);
3196 	rx_ctx.qlen = ring->count;
3197 
3198 	/* use 32 byte descriptors */
3199 	rx_ctx.dsize = 1;
3200 
3201 	/* descriptor type is always zero
3202 	 * rx_ctx.dtype = 0;
3203 	 */
3204 	rx_ctx.hsplit_0 = 0;
3205 
3206 	rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3207 	if (hw->revision_id == 0)
3208 		rx_ctx.lrxqthresh = 0;
3209 	else
3210 		rx_ctx.lrxqthresh = 1;
3211 	rx_ctx.crcstrip = 1;
3212 	rx_ctx.l2tsel = 1;
3213 	/* this controls whether VLAN is stripped from inner headers */
3214 	rx_ctx.showiv = 0;
3215 	/* set the prefena field to 1 because the manual says to */
3216 	rx_ctx.prefena = 1;
3217 
3218 	/* clear the context in the HMC */
3219 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3220 	if (err) {
3221 		dev_info(&vsi->back->pdev->dev,
3222 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3223 			 ring->queue_index, pf_q, err);
3224 		return -ENOMEM;
3225 	}
3226 
3227 	/* set the context in the HMC */
3228 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3229 	if (err) {
3230 		dev_info(&vsi->back->pdev->dev,
3231 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3232 			 ring->queue_index, pf_q, err);
3233 		return -ENOMEM;
3234 	}
3235 
3236 	/* configure Rx buffer alignment */
3237 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3238 		clear_ring_build_skb_enabled(ring);
3239 	else
3240 		set_ring_build_skb_enabled(ring);
3241 
3242 	/* cache tail for quicker writes, and clear the reg before use */
3243 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3244 	writel(0, ring->tail);
3245 
3246 	i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3247 
3248 	return 0;
3249 }
3250 
3251 /**
3252  * i40e_vsi_configure_tx - Configure the VSI for Tx
3253  * @vsi: VSI structure describing this set of rings and resources
3254  *
3255  * Configure the Tx VSI for operation.
3256  **/
3257 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3258 {
3259 	int err = 0;
3260 	u16 i;
3261 
3262 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3263 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3264 
3265 	if (!i40e_enabled_xdp_vsi(vsi))
3266 		return err;
3267 
3268 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3269 		err = i40e_configure_tx_ring(vsi->xdp_rings[i]);
3270 
3271 	return err;
3272 }
3273 
3274 /**
3275  * i40e_vsi_configure_rx - Configure the VSI for Rx
3276  * @vsi: the VSI being configured
3277  *
3278  * Configure the Rx VSI for operation.
3279  **/
3280 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3281 {
3282 	int err = 0;
3283 	u16 i;
3284 
3285 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3286 		vsi->max_frame = I40E_MAX_RXBUFFER;
3287 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
3288 #if (PAGE_SIZE < 8192)
3289 	} else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3290 		   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3291 		vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3292 		vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3293 #endif
3294 	} else {
3295 		vsi->max_frame = I40E_MAX_RXBUFFER;
3296 		vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3297 						       I40E_RXBUFFER_2048;
3298 	}
3299 
3300 	/* set up individual rings */
3301 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3302 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3303 
3304 	return err;
3305 }
3306 
3307 /**
3308  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3309  * @vsi: ptr to the VSI
3310  **/
3311 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3312 {
3313 	struct i40e_ring *tx_ring, *rx_ring;
3314 	u16 qoffset, qcount;
3315 	int i, n;
3316 
3317 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3318 		/* Reset the TC information */
3319 		for (i = 0; i < vsi->num_queue_pairs; i++) {
3320 			rx_ring = vsi->rx_rings[i];
3321 			tx_ring = vsi->tx_rings[i];
3322 			rx_ring->dcb_tc = 0;
3323 			tx_ring->dcb_tc = 0;
3324 		}
3325 		return;
3326 	}
3327 
3328 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3329 		if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3330 			continue;
3331 
3332 		qoffset = vsi->tc_config.tc_info[n].qoffset;
3333 		qcount = vsi->tc_config.tc_info[n].qcount;
3334 		for (i = qoffset; i < (qoffset + qcount); i++) {
3335 			rx_ring = vsi->rx_rings[i];
3336 			tx_ring = vsi->tx_rings[i];
3337 			rx_ring->dcb_tc = n;
3338 			tx_ring->dcb_tc = n;
3339 		}
3340 	}
3341 }
3342 
3343 /**
3344  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3345  * @vsi: ptr to the VSI
3346  **/
3347 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3348 {
3349 	if (vsi->netdev)
3350 		i40e_set_rx_mode(vsi->netdev);
3351 }
3352 
3353 /**
3354  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3355  * @vsi: Pointer to the targeted VSI
3356  *
3357  * This function replays the hlist on the hw where all the SB Flow Director
3358  * filters were saved.
3359  **/
3360 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3361 {
3362 	struct i40e_fdir_filter *filter;
3363 	struct i40e_pf *pf = vsi->back;
3364 	struct hlist_node *node;
3365 
3366 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3367 		return;
3368 
3369 	/* Reset FDir counters as we're replaying all existing filters */
3370 	pf->fd_tcp4_filter_cnt = 0;
3371 	pf->fd_udp4_filter_cnt = 0;
3372 	pf->fd_sctp4_filter_cnt = 0;
3373 	pf->fd_ip4_filter_cnt = 0;
3374 
3375 	hlist_for_each_entry_safe(filter, node,
3376 				  &pf->fdir_filter_list, fdir_node) {
3377 		i40e_add_del_fdir(vsi, filter, true);
3378 	}
3379 }
3380 
3381 /**
3382  * i40e_vsi_configure - Set up the VSI for action
3383  * @vsi: the VSI being configured
3384  **/
3385 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3386 {
3387 	int err;
3388 
3389 	i40e_set_vsi_rx_mode(vsi);
3390 	i40e_restore_vlan(vsi);
3391 	i40e_vsi_config_dcb_rings(vsi);
3392 	err = i40e_vsi_configure_tx(vsi);
3393 	if (!err)
3394 		err = i40e_vsi_configure_rx(vsi);
3395 
3396 	return err;
3397 }
3398 
3399 /**
3400  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3401  * @vsi: the VSI being configured
3402  **/
3403 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3404 {
3405 	bool has_xdp = i40e_enabled_xdp_vsi(vsi);
3406 	struct i40e_pf *pf = vsi->back;
3407 	struct i40e_hw *hw = &pf->hw;
3408 	u16 vector;
3409 	int i, q;
3410 	u32 qp;
3411 
3412 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
3413 	 * and PFINT_LNKLSTn registers, e.g.:
3414 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3415 	 */
3416 	qp = vsi->base_queue;
3417 	vector = vsi->base_vector;
3418 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3419 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3420 
3421 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
3422 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3423 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
3424 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3425 		     q_vector->rx.itr);
3426 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3427 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
3428 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3429 		     q_vector->tx.itr);
3430 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
3431 		     i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3432 
3433 		/* Linked list for the queuepairs assigned to this vector */
3434 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3435 		for (q = 0; q < q_vector->num_ringpairs; q++) {
3436 			u32 nextqp = has_xdp ? qp + vsi->alloc_queue_pairs : qp;
3437 			u32 val;
3438 
3439 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3440 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3441 			      (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3442 			      (nextqp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT) |
3443 			      (I40E_QUEUE_TYPE_TX <<
3444 			       I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3445 
3446 			wr32(hw, I40E_QINT_RQCTL(qp), val);
3447 
3448 			if (has_xdp) {
3449 				val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3450 				      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3451 				      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3452 				      (qp << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3453 				      (I40E_QUEUE_TYPE_TX <<
3454 				       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3455 
3456 				wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3457 			}
3458 
3459 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3460 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3461 			      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3462 			      ((qp + 1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3463 			      (I40E_QUEUE_TYPE_RX <<
3464 			       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3465 
3466 			/* Terminate the linked list */
3467 			if (q == (q_vector->num_ringpairs - 1))
3468 				val |= (I40E_QUEUE_END_OF_LIST <<
3469 					I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3470 
3471 			wr32(hw, I40E_QINT_TQCTL(qp), val);
3472 			qp++;
3473 		}
3474 	}
3475 
3476 	i40e_flush(hw);
3477 }
3478 
3479 /**
3480  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3481  * @hw: ptr to the hardware info
3482  **/
3483 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3484 {
3485 	struct i40e_hw *hw = &pf->hw;
3486 	u32 val;
3487 
3488 	/* clear things first */
3489 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3490 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3491 
3492 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3493 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3494 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
3495 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3496 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3497 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3498 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3499 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3500 
3501 	if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3502 		val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3503 
3504 	if (pf->flags & I40E_FLAG_PTP)
3505 		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3506 
3507 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
3508 
3509 	/* SW_ITR_IDX = 0, but don't change INTENA */
3510 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3511 					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3512 
3513 	/* OTHER_ITR_IDX = 0 */
3514 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3515 }
3516 
3517 /**
3518  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3519  * @vsi: the VSI being configured
3520  **/
3521 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3522 {
3523 	u32 nextqp = i40e_enabled_xdp_vsi(vsi) ? vsi->alloc_queue_pairs : 0;
3524 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3525 	struct i40e_pf *pf = vsi->back;
3526 	struct i40e_hw *hw = &pf->hw;
3527 	u32 val;
3528 
3529 	/* set the ITR configuration */
3530 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
3531 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3532 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
3533 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3534 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3535 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
3536 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3537 
3538 	i40e_enable_misc_int_causes(pf);
3539 
3540 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3541 	wr32(hw, I40E_PFINT_LNKLST0, 0);
3542 
3543 	/* Associate the queue pair to the vector and enable the queue int */
3544 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		       |
3545 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3546 	      (nextqp	   << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3547 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3548 
3549 	wr32(hw, I40E_QINT_RQCTL(0), val);
3550 
3551 	if (i40e_enabled_xdp_vsi(vsi)) {
3552 		val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		     |
3553 		      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)|
3554 		      (I40E_QUEUE_TYPE_TX
3555 		       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3556 
3557 	       wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3558 	}
3559 
3560 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
3561 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3562 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3563 
3564 	wr32(hw, I40E_QINT_TQCTL(0), val);
3565 	i40e_flush(hw);
3566 }
3567 
3568 /**
3569  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3570  * @pf: board private structure
3571  **/
3572 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3573 {
3574 	struct i40e_hw *hw = &pf->hw;
3575 
3576 	wr32(hw, I40E_PFINT_DYN_CTL0,
3577 	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3578 	i40e_flush(hw);
3579 }
3580 
3581 /**
3582  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3583  * @pf: board private structure
3584  **/
3585 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3586 {
3587 	struct i40e_hw *hw = &pf->hw;
3588 	u32 val;
3589 
3590 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3591 	      I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3592 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3593 
3594 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
3595 	i40e_flush(hw);
3596 }
3597 
3598 /**
3599  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3600  * @irq: interrupt number
3601  * @data: pointer to a q_vector
3602  **/
3603 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3604 {
3605 	struct i40e_q_vector *q_vector = data;
3606 
3607 	if (!q_vector->tx.ring && !q_vector->rx.ring)
3608 		return IRQ_HANDLED;
3609 
3610 	napi_schedule_irqoff(&q_vector->napi);
3611 
3612 	return IRQ_HANDLED;
3613 }
3614 
3615 /**
3616  * i40e_irq_affinity_notify - Callback for affinity changes
3617  * @notify: context as to what irq was changed
3618  * @mask: the new affinity mask
3619  *
3620  * This is a callback function used by the irq_set_affinity_notifier function
3621  * so that we may register to receive changes to the irq affinity masks.
3622  **/
3623 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3624 				     const cpumask_t *mask)
3625 {
3626 	struct i40e_q_vector *q_vector =
3627 		container_of(notify, struct i40e_q_vector, affinity_notify);
3628 
3629 	cpumask_copy(&q_vector->affinity_mask, mask);
3630 }
3631 
3632 /**
3633  * i40e_irq_affinity_release - Callback for affinity notifier release
3634  * @ref: internal core kernel usage
3635  *
3636  * This is a callback function used by the irq_set_affinity_notifier function
3637  * to inform the current notification subscriber that they will no longer
3638  * receive notifications.
3639  **/
3640 static void i40e_irq_affinity_release(struct kref *ref) {}
3641 
3642 /**
3643  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3644  * @vsi: the VSI being configured
3645  * @basename: name for the vector
3646  *
3647  * Allocates MSI-X vectors and requests interrupts from the kernel.
3648  **/
3649 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3650 {
3651 	int q_vectors = vsi->num_q_vectors;
3652 	struct i40e_pf *pf = vsi->back;
3653 	int base = vsi->base_vector;
3654 	int rx_int_idx = 0;
3655 	int tx_int_idx = 0;
3656 	int vector, err;
3657 	int irq_num;
3658 	int cpu;
3659 
3660 	for (vector = 0; vector < q_vectors; vector++) {
3661 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3662 
3663 		irq_num = pf->msix_entries[base + vector].vector;
3664 
3665 		if (q_vector->tx.ring && q_vector->rx.ring) {
3666 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3667 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3668 			tx_int_idx++;
3669 		} else if (q_vector->rx.ring) {
3670 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3671 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3672 		} else if (q_vector->tx.ring) {
3673 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3674 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3675 		} else {
3676 			/* skip this unused q_vector */
3677 			continue;
3678 		}
3679 		err = request_irq(irq_num,
3680 				  vsi->irq_handler,
3681 				  0,
3682 				  q_vector->name,
3683 				  q_vector);
3684 		if (err) {
3685 			dev_info(&pf->pdev->dev,
3686 				 "MSIX request_irq failed, error: %d\n", err);
3687 			goto free_queue_irqs;
3688 		}
3689 
3690 		/* register for affinity change notifications */
3691 		q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3692 		q_vector->affinity_notify.release = i40e_irq_affinity_release;
3693 		irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3694 		/* Spread affinity hints out across online CPUs.
3695 		 *
3696 		 * get_cpu_mask returns a static constant mask with
3697 		 * a permanent lifetime so it's ok to pass to
3698 		 * irq_set_affinity_hint without making a copy.
3699 		 */
3700 		cpu = cpumask_local_spread(q_vector->v_idx, -1);
3701 		irq_set_affinity_hint(irq_num, get_cpu_mask(cpu));
3702 	}
3703 
3704 	vsi->irqs_ready = true;
3705 	return 0;
3706 
3707 free_queue_irqs:
3708 	while (vector) {
3709 		vector--;
3710 		irq_num = pf->msix_entries[base + vector].vector;
3711 		irq_set_affinity_notifier(irq_num, NULL);
3712 		irq_set_affinity_hint(irq_num, NULL);
3713 		free_irq(irq_num, &vsi->q_vectors[vector]);
3714 	}
3715 	return err;
3716 }
3717 
3718 /**
3719  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3720  * @vsi: the VSI being un-configured
3721  **/
3722 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3723 {
3724 	struct i40e_pf *pf = vsi->back;
3725 	struct i40e_hw *hw = &pf->hw;
3726 	int base = vsi->base_vector;
3727 	int i;
3728 
3729 	/* disable interrupt causation from each queue */
3730 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3731 		u32 val;
3732 
3733 		val = rd32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx));
3734 		val &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3735 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), val);
3736 
3737 		val = rd32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx));
3738 		val &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3739 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), val);
3740 
3741 		if (!i40e_enabled_xdp_vsi(vsi))
3742 			continue;
3743 		wr32(hw, I40E_QINT_TQCTL(vsi->xdp_rings[i]->reg_idx), 0);
3744 	}
3745 
3746 	/* disable each interrupt */
3747 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3748 		for (i = vsi->base_vector;
3749 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3750 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3751 
3752 		i40e_flush(hw);
3753 		for (i = 0; i < vsi->num_q_vectors; i++)
3754 			synchronize_irq(pf->msix_entries[i + base].vector);
3755 	} else {
3756 		/* Legacy and MSI mode - this stops all interrupt handling */
3757 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3758 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3759 		i40e_flush(hw);
3760 		synchronize_irq(pf->pdev->irq);
3761 	}
3762 }
3763 
3764 /**
3765  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3766  * @vsi: the VSI being configured
3767  **/
3768 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3769 {
3770 	struct i40e_pf *pf = vsi->back;
3771 	int i;
3772 
3773 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3774 		for (i = 0; i < vsi->num_q_vectors; i++)
3775 			i40e_irq_dynamic_enable(vsi, i);
3776 	} else {
3777 		i40e_irq_dynamic_enable_icr0(pf);
3778 	}
3779 
3780 	i40e_flush(&pf->hw);
3781 	return 0;
3782 }
3783 
3784 /**
3785  * i40e_free_misc_vector - Free the vector that handles non-queue events
3786  * @pf: board private structure
3787  **/
3788 static void i40e_free_misc_vector(struct i40e_pf *pf)
3789 {
3790 	/* Disable ICR 0 */
3791 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3792 	i40e_flush(&pf->hw);
3793 
3794 	if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
3795 		synchronize_irq(pf->msix_entries[0].vector);
3796 		free_irq(pf->msix_entries[0].vector, pf);
3797 		clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
3798 	}
3799 }
3800 
3801 /**
3802  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3803  * @irq: interrupt number
3804  * @data: pointer to a q_vector
3805  *
3806  * This is the handler used for all MSI/Legacy interrupts, and deals
3807  * with both queue and non-queue interrupts.  This is also used in
3808  * MSIX mode to handle the non-queue interrupts.
3809  **/
3810 static irqreturn_t i40e_intr(int irq, void *data)
3811 {
3812 	struct i40e_pf *pf = (struct i40e_pf *)data;
3813 	struct i40e_hw *hw = &pf->hw;
3814 	irqreturn_t ret = IRQ_NONE;
3815 	u32 icr0, icr0_remaining;
3816 	u32 val, ena_mask;
3817 
3818 	icr0 = rd32(hw, I40E_PFINT_ICR0);
3819 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3820 
3821 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3822 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3823 		goto enable_intr;
3824 
3825 	/* if interrupt but no bits showing, must be SWINT */
3826 	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3827 	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3828 		pf->sw_int_count++;
3829 
3830 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3831 	    (icr0 & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3832 		ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3833 		dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3834 		set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
3835 	}
3836 
3837 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3838 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3839 		struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3840 		struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3841 
3842 		/* We do not have a way to disarm Queue causes while leaving
3843 		 * interrupt enabled for all other causes, ideally
3844 		 * interrupt should be disabled while we are in NAPI but
3845 		 * this is not a performance path and napi_schedule()
3846 		 * can deal with rescheduling.
3847 		 */
3848 		if (!test_bit(__I40E_DOWN, pf->state))
3849 			napi_schedule_irqoff(&q_vector->napi);
3850 	}
3851 
3852 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3853 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3854 		set_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
3855 		i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3856 	}
3857 
3858 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3859 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3860 		set_bit(__I40E_MDD_EVENT_PENDING, pf->state);
3861 	}
3862 
3863 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3864 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3865 		set_bit(__I40E_VFLR_EVENT_PENDING, pf->state);
3866 	}
3867 
3868 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3869 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
3870 			set_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
3871 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3872 		val = rd32(hw, I40E_GLGEN_RSTAT);
3873 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3874 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3875 		if (val == I40E_RESET_CORER) {
3876 			pf->corer_count++;
3877 		} else if (val == I40E_RESET_GLOBR) {
3878 			pf->globr_count++;
3879 		} else if (val == I40E_RESET_EMPR) {
3880 			pf->empr_count++;
3881 			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state);
3882 		}
3883 	}
3884 
3885 	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3886 		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3887 		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3888 		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3889 			 rd32(hw, I40E_PFHMC_ERRORINFO),
3890 			 rd32(hw, I40E_PFHMC_ERRORDATA));
3891 	}
3892 
3893 	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3894 		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3895 
3896 		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3897 			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3898 			i40e_ptp_tx_hwtstamp(pf);
3899 		}
3900 	}
3901 
3902 	/* If a critical error is pending we have no choice but to reset the
3903 	 * device.
3904 	 * Report and mask out any remaining unexpected interrupts.
3905 	 */
3906 	icr0_remaining = icr0 & ena_mask;
3907 	if (icr0_remaining) {
3908 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3909 			 icr0_remaining);
3910 		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3911 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3912 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3913 			dev_info(&pf->pdev->dev, "device will be reset\n");
3914 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
3915 			i40e_service_event_schedule(pf);
3916 		}
3917 		ena_mask &= ~icr0_remaining;
3918 	}
3919 	ret = IRQ_HANDLED;
3920 
3921 enable_intr:
3922 	/* re-enable interrupt causes */
3923 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3924 	if (!test_bit(__I40E_DOWN, pf->state)) {
3925 		i40e_service_event_schedule(pf);
3926 		i40e_irq_dynamic_enable_icr0(pf);
3927 	}
3928 
3929 	return ret;
3930 }
3931 
3932 /**
3933  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3934  * @tx_ring:  tx ring to clean
3935  * @budget:   how many cleans we're allowed
3936  *
3937  * Returns true if there's any budget left (e.g. the clean is finished)
3938  **/
3939 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3940 {
3941 	struct i40e_vsi *vsi = tx_ring->vsi;
3942 	u16 i = tx_ring->next_to_clean;
3943 	struct i40e_tx_buffer *tx_buf;
3944 	struct i40e_tx_desc *tx_desc;
3945 
3946 	tx_buf = &tx_ring->tx_bi[i];
3947 	tx_desc = I40E_TX_DESC(tx_ring, i);
3948 	i -= tx_ring->count;
3949 
3950 	do {
3951 		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3952 
3953 		/* if next_to_watch is not set then there is no work pending */
3954 		if (!eop_desc)
3955 			break;
3956 
3957 		/* prevent any other reads prior to eop_desc */
3958 		read_barrier_depends();
3959 
3960 		/* if the descriptor isn't done, no work yet to do */
3961 		if (!(eop_desc->cmd_type_offset_bsz &
3962 		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3963 			break;
3964 
3965 		/* clear next_to_watch to prevent false hangs */
3966 		tx_buf->next_to_watch = NULL;
3967 
3968 		tx_desc->buffer_addr = 0;
3969 		tx_desc->cmd_type_offset_bsz = 0;
3970 		/* move past filter desc */
3971 		tx_buf++;
3972 		tx_desc++;
3973 		i++;
3974 		if (unlikely(!i)) {
3975 			i -= tx_ring->count;
3976 			tx_buf = tx_ring->tx_bi;
3977 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3978 		}
3979 		/* unmap skb header data */
3980 		dma_unmap_single(tx_ring->dev,
3981 				 dma_unmap_addr(tx_buf, dma),
3982 				 dma_unmap_len(tx_buf, len),
3983 				 DMA_TO_DEVICE);
3984 		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3985 			kfree(tx_buf->raw_buf);
3986 
3987 		tx_buf->raw_buf = NULL;
3988 		tx_buf->tx_flags = 0;
3989 		tx_buf->next_to_watch = NULL;
3990 		dma_unmap_len_set(tx_buf, len, 0);
3991 		tx_desc->buffer_addr = 0;
3992 		tx_desc->cmd_type_offset_bsz = 0;
3993 
3994 		/* move us past the eop_desc for start of next FD desc */
3995 		tx_buf++;
3996 		tx_desc++;
3997 		i++;
3998 		if (unlikely(!i)) {
3999 			i -= tx_ring->count;
4000 			tx_buf = tx_ring->tx_bi;
4001 			tx_desc = I40E_TX_DESC(tx_ring, 0);
4002 		}
4003 
4004 		/* update budget accounting */
4005 		budget--;
4006 	} while (likely(budget));
4007 
4008 	i += tx_ring->count;
4009 	tx_ring->next_to_clean = i;
4010 
4011 	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
4012 		i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
4013 
4014 	return budget > 0;
4015 }
4016 
4017 /**
4018  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
4019  * @irq: interrupt number
4020  * @data: pointer to a q_vector
4021  **/
4022 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
4023 {
4024 	struct i40e_q_vector *q_vector = data;
4025 	struct i40e_vsi *vsi;
4026 
4027 	if (!q_vector->tx.ring)
4028 		return IRQ_HANDLED;
4029 
4030 	vsi = q_vector->tx.ring->vsi;
4031 	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
4032 
4033 	return IRQ_HANDLED;
4034 }
4035 
4036 /**
4037  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
4038  * @vsi: the VSI being configured
4039  * @v_idx: vector index
4040  * @qp_idx: queue pair index
4041  **/
4042 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
4043 {
4044 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4045 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
4046 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
4047 
4048 	tx_ring->q_vector = q_vector;
4049 	tx_ring->next = q_vector->tx.ring;
4050 	q_vector->tx.ring = tx_ring;
4051 	q_vector->tx.count++;
4052 
4053 	/* Place XDP Tx ring in the same q_vector ring list as regular Tx */
4054 	if (i40e_enabled_xdp_vsi(vsi)) {
4055 		struct i40e_ring *xdp_ring = vsi->xdp_rings[qp_idx];
4056 
4057 		xdp_ring->q_vector = q_vector;
4058 		xdp_ring->next = q_vector->tx.ring;
4059 		q_vector->tx.ring = xdp_ring;
4060 		q_vector->tx.count++;
4061 	}
4062 
4063 	rx_ring->q_vector = q_vector;
4064 	rx_ring->next = q_vector->rx.ring;
4065 	q_vector->rx.ring = rx_ring;
4066 	q_vector->rx.count++;
4067 }
4068 
4069 /**
4070  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
4071  * @vsi: the VSI being configured
4072  *
4073  * This function maps descriptor rings to the queue-specific vectors
4074  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
4075  * one vector per queue pair, but on a constrained vector budget, we
4076  * group the queue pairs as "efficiently" as possible.
4077  **/
4078 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
4079 {
4080 	int qp_remaining = vsi->num_queue_pairs;
4081 	int q_vectors = vsi->num_q_vectors;
4082 	int num_ringpairs;
4083 	int v_start = 0;
4084 	int qp_idx = 0;
4085 
4086 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
4087 	 * group them so there are multiple queues per vector.
4088 	 * It is also important to go through all the vectors available to be
4089 	 * sure that if we don't use all the vectors, that the remaining vectors
4090 	 * are cleared. This is especially important when decreasing the
4091 	 * number of queues in use.
4092 	 */
4093 	for (; v_start < q_vectors; v_start++) {
4094 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
4095 
4096 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
4097 
4098 		q_vector->num_ringpairs = num_ringpairs;
4099 
4100 		q_vector->rx.count = 0;
4101 		q_vector->tx.count = 0;
4102 		q_vector->rx.ring = NULL;
4103 		q_vector->tx.ring = NULL;
4104 
4105 		while (num_ringpairs--) {
4106 			i40e_map_vector_to_qp(vsi, v_start, qp_idx);
4107 			qp_idx++;
4108 			qp_remaining--;
4109 		}
4110 	}
4111 }
4112 
4113 /**
4114  * i40e_vsi_request_irq - Request IRQ from the OS
4115  * @vsi: the VSI being configured
4116  * @basename: name for the vector
4117  **/
4118 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
4119 {
4120 	struct i40e_pf *pf = vsi->back;
4121 	int err;
4122 
4123 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4124 		err = i40e_vsi_request_irq_msix(vsi, basename);
4125 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
4126 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
4127 				  pf->int_name, pf);
4128 	else
4129 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
4130 				  pf->int_name, pf);
4131 
4132 	if (err)
4133 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
4134 
4135 	return err;
4136 }
4137 
4138 #ifdef CONFIG_NET_POLL_CONTROLLER
4139 /**
4140  * i40e_netpoll - A Polling 'interrupt' handler
4141  * @netdev: network interface device structure
4142  *
4143  * This is used by netconsole to send skbs without having to re-enable
4144  * interrupts.  It's not called while the normal interrupt routine is executing.
4145  **/
4146 static void i40e_netpoll(struct net_device *netdev)
4147 {
4148 	struct i40e_netdev_priv *np = netdev_priv(netdev);
4149 	struct i40e_vsi *vsi = np->vsi;
4150 	struct i40e_pf *pf = vsi->back;
4151 	int i;
4152 
4153 	/* if interface is down do nothing */
4154 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
4155 		return;
4156 
4157 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4158 		for (i = 0; i < vsi->num_q_vectors; i++)
4159 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
4160 	} else {
4161 		i40e_intr(pf->pdev->irq, netdev);
4162 	}
4163 }
4164 #endif
4165 
4166 #define I40E_QTX_ENA_WAIT_COUNT 50
4167 
4168 /**
4169  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
4170  * @pf: the PF being configured
4171  * @pf_q: the PF queue
4172  * @enable: enable or disable state of the queue
4173  *
4174  * This routine will wait for the given Tx queue of the PF to reach the
4175  * enabled or disabled state.
4176  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4177  * multiple retries; else will return 0 in case of success.
4178  **/
4179 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4180 {
4181 	int i;
4182 	u32 tx_reg;
4183 
4184 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4185 		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
4186 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4187 			break;
4188 
4189 		usleep_range(10, 20);
4190 	}
4191 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4192 		return -ETIMEDOUT;
4193 
4194 	return 0;
4195 }
4196 
4197 /**
4198  * i40e_control_tx_q - Start or stop a particular Tx queue
4199  * @pf: the PF structure
4200  * @pf_q: the PF queue to configure
4201  * @enable: start or stop the queue
4202  *
4203  * This function enables or disables a single queue. Note that any delay
4204  * required after the operation is expected to be handled by the caller of
4205  * this function.
4206  **/
4207 static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
4208 {
4209 	struct i40e_hw *hw = &pf->hw;
4210 	u32 tx_reg;
4211 	int i;
4212 
4213 	/* warn the TX unit of coming changes */
4214 	i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
4215 	if (!enable)
4216 		usleep_range(10, 20);
4217 
4218 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4219 		tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
4220 		if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
4221 		    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
4222 			break;
4223 		usleep_range(1000, 2000);
4224 	}
4225 
4226 	/* Skip if the queue is already in the requested state */
4227 	if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4228 		return;
4229 
4230 	/* turn on/off the queue */
4231 	if (enable) {
4232 		wr32(hw, I40E_QTX_HEAD(pf_q), 0);
4233 		tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
4234 	} else {
4235 		tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
4236 	}
4237 
4238 	wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
4239 }
4240 
4241 /**
4242  * i40e_control_wait_tx_q - Start/stop Tx queue and wait for completion
4243  * @seid: VSI SEID
4244  * @pf: the PF structure
4245  * @pf_q: the PF queue to configure
4246  * @is_xdp: true if the queue is used for XDP
4247  * @enable: start or stop the queue
4248  **/
4249 static int i40e_control_wait_tx_q(int seid, struct i40e_pf *pf, int pf_q,
4250 				  bool is_xdp, bool enable)
4251 {
4252 	int ret;
4253 
4254 	i40e_control_tx_q(pf, pf_q, enable);
4255 
4256 	/* wait for the change to finish */
4257 	ret = i40e_pf_txq_wait(pf, pf_q, enable);
4258 	if (ret) {
4259 		dev_info(&pf->pdev->dev,
4260 			 "VSI seid %d %sTx ring %d %sable timeout\n",
4261 			 seid, (is_xdp ? "XDP " : ""), pf_q,
4262 			 (enable ? "en" : "dis"));
4263 	}
4264 
4265 	return ret;
4266 }
4267 
4268 /**
4269  * i40e_vsi_control_tx - Start or stop a VSI's rings
4270  * @vsi: the VSI being configured
4271  * @enable: start or stop the rings
4272  **/
4273 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4274 {
4275 	struct i40e_pf *pf = vsi->back;
4276 	int i, pf_q, ret = 0;
4277 
4278 	pf_q = vsi->base_queue;
4279 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4280 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4281 					     pf_q,
4282 					     false /*is xdp*/, enable);
4283 		if (ret)
4284 			break;
4285 
4286 		if (!i40e_enabled_xdp_vsi(vsi))
4287 			continue;
4288 
4289 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4290 					     pf_q + vsi->alloc_queue_pairs,
4291 					     true /*is xdp*/, enable);
4292 		if (ret)
4293 			break;
4294 	}
4295 
4296 	return ret;
4297 }
4298 
4299 /**
4300  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4301  * @pf: the PF being configured
4302  * @pf_q: the PF queue
4303  * @enable: enable or disable state of the queue
4304  *
4305  * This routine will wait for the given Rx queue of the PF to reach the
4306  * enabled or disabled state.
4307  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4308  * multiple retries; else will return 0 in case of success.
4309  **/
4310 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4311 {
4312 	int i;
4313 	u32 rx_reg;
4314 
4315 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4316 		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4317 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4318 			break;
4319 
4320 		usleep_range(10, 20);
4321 	}
4322 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4323 		return -ETIMEDOUT;
4324 
4325 	return 0;
4326 }
4327 
4328 /**
4329  * i40e_control_rx_q - Start or stop a particular Rx queue
4330  * @pf: the PF structure
4331  * @pf_q: the PF queue to configure
4332  * @enable: start or stop the queue
4333  *
4334  * This function enables or disables a single queue. Note that any delay
4335  * required after the operation is expected to be handled by the caller of
4336  * this function.
4337  **/
4338 static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4339 {
4340 	struct i40e_hw *hw = &pf->hw;
4341 	u32 rx_reg;
4342 	int i;
4343 
4344 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4345 		rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4346 		if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4347 		    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4348 			break;
4349 		usleep_range(1000, 2000);
4350 	}
4351 
4352 	/* Skip if the queue is already in the requested state */
4353 	if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4354 		return;
4355 
4356 	/* turn on/off the queue */
4357 	if (enable)
4358 		rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4359 	else
4360 		rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4361 
4362 	wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4363 }
4364 
4365 /**
4366  * i40e_vsi_control_rx - Start or stop a VSI's rings
4367  * @vsi: the VSI being configured
4368  * @enable: start or stop the rings
4369  **/
4370 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4371 {
4372 	struct i40e_pf *pf = vsi->back;
4373 	int i, pf_q, ret = 0;
4374 
4375 	pf_q = vsi->base_queue;
4376 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4377 		i40e_control_rx_q(pf, pf_q, enable);
4378 
4379 		/* wait for the change to finish */
4380 		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4381 		if (ret) {
4382 			dev_info(&pf->pdev->dev,
4383 				 "VSI seid %d Rx ring %d %sable timeout\n",
4384 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
4385 			break;
4386 		}
4387 	}
4388 
4389 	/* Due to HW errata, on Rx disable only, the register can indicate done
4390 	 * before it really is. Needs 50ms to be sure
4391 	 */
4392 	if (!enable)
4393 		mdelay(50);
4394 
4395 	return ret;
4396 }
4397 
4398 /**
4399  * i40e_vsi_start_rings - Start a VSI's rings
4400  * @vsi: the VSI being configured
4401  **/
4402 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4403 {
4404 	int ret = 0;
4405 
4406 	/* do rx first for enable and last for disable */
4407 	ret = i40e_vsi_control_rx(vsi, true);
4408 	if (ret)
4409 		return ret;
4410 	ret = i40e_vsi_control_tx(vsi, true);
4411 
4412 	return ret;
4413 }
4414 
4415 /**
4416  * i40e_vsi_stop_rings - Stop a VSI's rings
4417  * @vsi: the VSI being configured
4418  **/
4419 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4420 {
4421 	/* When port TX is suspended, don't wait */
4422 	if (test_bit(__I40E_PORT_SUSPENDED, vsi->back->state))
4423 		return i40e_vsi_stop_rings_no_wait(vsi);
4424 
4425 	/* do rx first for enable and last for disable
4426 	 * Ignore return value, we need to shutdown whatever we can
4427 	 */
4428 	i40e_vsi_control_tx(vsi, false);
4429 	i40e_vsi_control_rx(vsi, false);
4430 }
4431 
4432 /**
4433  * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4434  * @vsi: the VSI being shutdown
4435  *
4436  * This function stops all the rings for a VSI but does not delay to verify
4437  * that rings have been disabled. It is expected that the caller is shutting
4438  * down multiple VSIs at once and will delay together for all the VSIs after
4439  * initiating the shutdown. This is particularly useful for shutting down lots
4440  * of VFs together. Otherwise, a large delay can be incurred while configuring
4441  * each VSI in serial.
4442  **/
4443 void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4444 {
4445 	struct i40e_pf *pf = vsi->back;
4446 	int i, pf_q;
4447 
4448 	pf_q = vsi->base_queue;
4449 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4450 		i40e_control_tx_q(pf, pf_q, false);
4451 		i40e_control_rx_q(pf, pf_q, false);
4452 	}
4453 }
4454 
4455 /**
4456  * i40e_vsi_free_irq - Free the irq association with the OS
4457  * @vsi: the VSI being configured
4458  **/
4459 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4460 {
4461 	struct i40e_pf *pf = vsi->back;
4462 	struct i40e_hw *hw = &pf->hw;
4463 	int base = vsi->base_vector;
4464 	u32 val, qp;
4465 	int i;
4466 
4467 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4468 		if (!vsi->q_vectors)
4469 			return;
4470 
4471 		if (!vsi->irqs_ready)
4472 			return;
4473 
4474 		vsi->irqs_ready = false;
4475 		for (i = 0; i < vsi->num_q_vectors; i++) {
4476 			int irq_num;
4477 			u16 vector;
4478 
4479 			vector = i + base;
4480 			irq_num = pf->msix_entries[vector].vector;
4481 
4482 			/* free only the irqs that were actually requested */
4483 			if (!vsi->q_vectors[i] ||
4484 			    !vsi->q_vectors[i]->num_ringpairs)
4485 				continue;
4486 
4487 			/* clear the affinity notifier in the IRQ descriptor */
4488 			irq_set_affinity_notifier(irq_num, NULL);
4489 			/* remove our suggested affinity mask for this IRQ */
4490 			irq_set_affinity_hint(irq_num, NULL);
4491 			synchronize_irq(irq_num);
4492 			free_irq(irq_num, vsi->q_vectors[i]);
4493 
4494 			/* Tear down the interrupt queue link list
4495 			 *
4496 			 * We know that they come in pairs and always
4497 			 * the Rx first, then the Tx.  To clear the
4498 			 * link list, stick the EOL value into the
4499 			 * next_q field of the registers.
4500 			 */
4501 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4502 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4503 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4504 			val |= I40E_QUEUE_END_OF_LIST
4505 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4506 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4507 
4508 			while (qp != I40E_QUEUE_END_OF_LIST) {
4509 				u32 next;
4510 
4511 				val = rd32(hw, I40E_QINT_RQCTL(qp));
4512 
4513 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4514 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4515 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4516 					 I40E_QINT_RQCTL_INTEVENT_MASK);
4517 
4518 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4519 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4520 
4521 				wr32(hw, I40E_QINT_RQCTL(qp), val);
4522 
4523 				val = rd32(hw, I40E_QINT_TQCTL(qp));
4524 
4525 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4526 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4527 
4528 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4529 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4530 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4531 					 I40E_QINT_TQCTL_INTEVENT_MASK);
4532 
4533 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4534 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4535 
4536 				wr32(hw, I40E_QINT_TQCTL(qp), val);
4537 				qp = next;
4538 			}
4539 		}
4540 	} else {
4541 		free_irq(pf->pdev->irq, pf);
4542 
4543 		val = rd32(hw, I40E_PFINT_LNKLST0);
4544 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4545 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4546 		val |= I40E_QUEUE_END_OF_LIST
4547 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4548 		wr32(hw, I40E_PFINT_LNKLST0, val);
4549 
4550 		val = rd32(hw, I40E_QINT_RQCTL(qp));
4551 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4552 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4553 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4554 			 I40E_QINT_RQCTL_INTEVENT_MASK);
4555 
4556 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4557 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4558 
4559 		wr32(hw, I40E_QINT_RQCTL(qp), val);
4560 
4561 		val = rd32(hw, I40E_QINT_TQCTL(qp));
4562 
4563 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4564 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4565 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4566 			 I40E_QINT_TQCTL_INTEVENT_MASK);
4567 
4568 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4569 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4570 
4571 		wr32(hw, I40E_QINT_TQCTL(qp), val);
4572 	}
4573 }
4574 
4575 /**
4576  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4577  * @vsi: the VSI being configured
4578  * @v_idx: Index of vector to be freed
4579  *
4580  * This function frees the memory allocated to the q_vector.  In addition if
4581  * NAPI is enabled it will delete any references to the NAPI struct prior
4582  * to freeing the q_vector.
4583  **/
4584 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4585 {
4586 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4587 	struct i40e_ring *ring;
4588 
4589 	if (!q_vector)
4590 		return;
4591 
4592 	/* disassociate q_vector from rings */
4593 	i40e_for_each_ring(ring, q_vector->tx)
4594 		ring->q_vector = NULL;
4595 
4596 	i40e_for_each_ring(ring, q_vector->rx)
4597 		ring->q_vector = NULL;
4598 
4599 	/* only VSI w/ an associated netdev is set up w/ NAPI */
4600 	if (vsi->netdev)
4601 		netif_napi_del(&q_vector->napi);
4602 
4603 	vsi->q_vectors[v_idx] = NULL;
4604 
4605 	kfree_rcu(q_vector, rcu);
4606 }
4607 
4608 /**
4609  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4610  * @vsi: the VSI being un-configured
4611  *
4612  * This frees the memory allocated to the q_vectors and
4613  * deletes references to the NAPI struct.
4614  **/
4615 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4616 {
4617 	int v_idx;
4618 
4619 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4620 		i40e_free_q_vector(vsi, v_idx);
4621 }
4622 
4623 /**
4624  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4625  * @pf: board private structure
4626  **/
4627 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4628 {
4629 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4630 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4631 		pci_disable_msix(pf->pdev);
4632 		kfree(pf->msix_entries);
4633 		pf->msix_entries = NULL;
4634 		kfree(pf->irq_pile);
4635 		pf->irq_pile = NULL;
4636 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4637 		pci_disable_msi(pf->pdev);
4638 	}
4639 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4640 }
4641 
4642 /**
4643  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4644  * @pf: board private structure
4645  *
4646  * We go through and clear interrupt specific resources and reset the structure
4647  * to pre-load conditions
4648  **/
4649 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4650 {
4651 	int i;
4652 
4653 	i40e_free_misc_vector(pf);
4654 
4655 	i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4656 		      I40E_IWARP_IRQ_PILE_ID);
4657 
4658 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4659 	for (i = 0; i < pf->num_alloc_vsi; i++)
4660 		if (pf->vsi[i])
4661 			i40e_vsi_free_q_vectors(pf->vsi[i]);
4662 	i40e_reset_interrupt_capability(pf);
4663 }
4664 
4665 /**
4666  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4667  * @vsi: the VSI being configured
4668  **/
4669 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4670 {
4671 	int q_idx;
4672 
4673 	if (!vsi->netdev)
4674 		return;
4675 
4676 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4677 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4678 
4679 		if (q_vector->rx.ring || q_vector->tx.ring)
4680 			napi_enable(&q_vector->napi);
4681 	}
4682 }
4683 
4684 /**
4685  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4686  * @vsi: the VSI being configured
4687  **/
4688 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4689 {
4690 	int q_idx;
4691 
4692 	if (!vsi->netdev)
4693 		return;
4694 
4695 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4696 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4697 
4698 		if (q_vector->rx.ring || q_vector->tx.ring)
4699 			napi_disable(&q_vector->napi);
4700 	}
4701 }
4702 
4703 /**
4704  * i40e_vsi_close - Shut down a VSI
4705  * @vsi: the vsi to be quelled
4706  **/
4707 static void i40e_vsi_close(struct i40e_vsi *vsi)
4708 {
4709 	struct i40e_pf *pf = vsi->back;
4710 	if (!test_and_set_bit(__I40E_VSI_DOWN, vsi->state))
4711 		i40e_down(vsi);
4712 	i40e_vsi_free_irq(vsi);
4713 	i40e_vsi_free_tx_resources(vsi);
4714 	i40e_vsi_free_rx_resources(vsi);
4715 	vsi->current_netdev_flags = 0;
4716 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
4717 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
4718 		pf->flags |=  I40E_FLAG_CLIENT_RESET;
4719 }
4720 
4721 /**
4722  * i40e_quiesce_vsi - Pause a given VSI
4723  * @vsi: the VSI being paused
4724  **/
4725 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4726 {
4727 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
4728 		return;
4729 
4730 	set_bit(__I40E_VSI_NEEDS_RESTART, vsi->state);
4731 	if (vsi->netdev && netif_running(vsi->netdev))
4732 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4733 	else
4734 		i40e_vsi_close(vsi);
4735 }
4736 
4737 /**
4738  * i40e_unquiesce_vsi - Resume a given VSI
4739  * @vsi: the VSI being resumed
4740  **/
4741 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4742 {
4743 	if (!test_and_clear_bit(__I40E_VSI_NEEDS_RESTART, vsi->state))
4744 		return;
4745 
4746 	if (vsi->netdev && netif_running(vsi->netdev))
4747 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4748 	else
4749 		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4750 }
4751 
4752 /**
4753  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4754  * @pf: the PF
4755  **/
4756 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4757 {
4758 	int v;
4759 
4760 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4761 		if (pf->vsi[v])
4762 			i40e_quiesce_vsi(pf->vsi[v]);
4763 	}
4764 }
4765 
4766 /**
4767  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4768  * @pf: the PF
4769  **/
4770 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4771 {
4772 	int v;
4773 
4774 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4775 		if (pf->vsi[v])
4776 			i40e_unquiesce_vsi(pf->vsi[v]);
4777 	}
4778 }
4779 
4780 /**
4781  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4782  * @vsi: the VSI being configured
4783  *
4784  * Wait until all queues on a given VSI have been disabled.
4785  **/
4786 int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4787 {
4788 	struct i40e_pf *pf = vsi->back;
4789 	int i, pf_q, ret;
4790 
4791 	pf_q = vsi->base_queue;
4792 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4793 		/* Check and wait for the Tx queue */
4794 		ret = i40e_pf_txq_wait(pf, pf_q, false);
4795 		if (ret) {
4796 			dev_info(&pf->pdev->dev,
4797 				 "VSI seid %d Tx ring %d disable timeout\n",
4798 				 vsi->seid, pf_q);
4799 			return ret;
4800 		}
4801 
4802 		if (!i40e_enabled_xdp_vsi(vsi))
4803 			goto wait_rx;
4804 
4805 		/* Check and wait for the XDP Tx queue */
4806 		ret = i40e_pf_txq_wait(pf, pf_q + vsi->alloc_queue_pairs,
4807 				       false);
4808 		if (ret) {
4809 			dev_info(&pf->pdev->dev,
4810 				 "VSI seid %d XDP Tx ring %d disable timeout\n",
4811 				 vsi->seid, pf_q);
4812 			return ret;
4813 		}
4814 wait_rx:
4815 		/* Check and wait for the Rx queue */
4816 		ret = i40e_pf_rxq_wait(pf, pf_q, false);
4817 		if (ret) {
4818 			dev_info(&pf->pdev->dev,
4819 				 "VSI seid %d Rx ring %d disable timeout\n",
4820 				 vsi->seid, pf_q);
4821 			return ret;
4822 		}
4823 	}
4824 
4825 	return 0;
4826 }
4827 
4828 #ifdef CONFIG_I40E_DCB
4829 /**
4830  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4831  * @pf: the PF
4832  *
4833  * This function waits for the queues to be in disabled state for all the
4834  * VSIs that are managed by this PF.
4835  **/
4836 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4837 {
4838 	int v, ret = 0;
4839 
4840 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4841 		if (pf->vsi[v]) {
4842 			ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4843 			if (ret)
4844 				break;
4845 		}
4846 	}
4847 
4848 	return ret;
4849 }
4850 
4851 #endif
4852 
4853 /**
4854  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4855  * @q_idx: TX queue number
4856  * @vsi: Pointer to VSI struct
4857  *
4858  * This function checks specified queue for given VSI. Detects hung condition.
4859  * We proactively detect hung TX queues by checking if interrupts are disabled
4860  * but there are pending descriptors.  If it appears hung, attempt to recover
4861  * by triggering a SW interrupt.
4862  **/
4863 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4864 {
4865 	struct i40e_ring *tx_ring = NULL;
4866 	struct i40e_pf	*pf;
4867 	u32 val, tx_pending;
4868 	int i;
4869 
4870 	pf = vsi->back;
4871 
4872 	/* now that we have an index, find the tx_ring struct */
4873 	for (i = 0; i < vsi->num_queue_pairs; i++) {
4874 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4875 			if (q_idx == vsi->tx_rings[i]->queue_index) {
4876 				tx_ring = vsi->tx_rings[i];
4877 				break;
4878 			}
4879 		}
4880 	}
4881 
4882 	if (!tx_ring)
4883 		return;
4884 
4885 	/* Read interrupt register */
4886 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4887 		val = rd32(&pf->hw,
4888 			   I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4889 					       tx_ring->vsi->base_vector - 1));
4890 	else
4891 		val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4892 
4893 	tx_pending = i40e_get_tx_pending(tx_ring);
4894 
4895 	/* Interrupts are disabled and TX pending is non-zero,
4896 	 * trigger the SW interrupt (don't wait). Worst case
4897 	 * there will be one extra interrupt which may result
4898 	 * into not cleaning any queues because queues are cleaned.
4899 	 */
4900 	if (tx_pending && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)))
4901 		i40e_force_wb(vsi, tx_ring->q_vector);
4902 }
4903 
4904 /**
4905  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4906  * @pf:  pointer to PF struct
4907  *
4908  * LAN VSI has netdev and netdev has TX queues. This function is to check
4909  * each of those TX queues if they are hung, trigger recovery by issuing
4910  * SW interrupt.
4911  **/
4912 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4913 {
4914 	struct net_device *netdev;
4915 	struct i40e_vsi *vsi;
4916 	unsigned int i;
4917 
4918 	/* Only for LAN VSI */
4919 	vsi = pf->vsi[pf->lan_vsi];
4920 
4921 	if (!vsi)
4922 		return;
4923 
4924 	/* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4925 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
4926 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
4927 		return;
4928 
4929 	/* Make sure type is MAIN VSI */
4930 	if (vsi->type != I40E_VSI_MAIN)
4931 		return;
4932 
4933 	netdev = vsi->netdev;
4934 	if (!netdev)
4935 		return;
4936 
4937 	/* Bail out if netif_carrier is not OK */
4938 	if (!netif_carrier_ok(netdev))
4939 		return;
4940 
4941 	/* Go thru' TX queues for netdev */
4942 	for (i = 0; i < netdev->num_tx_queues; i++) {
4943 		struct netdev_queue *q;
4944 
4945 		q = netdev_get_tx_queue(netdev, i);
4946 		if (q)
4947 			i40e_detect_recover_hung_queue(i, vsi);
4948 	}
4949 }
4950 
4951 /**
4952  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4953  * @pf: pointer to PF
4954  *
4955  * Get TC map for ISCSI PF type that will include iSCSI TC
4956  * and LAN TC.
4957  **/
4958 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4959 {
4960 	struct i40e_dcb_app_priority_table app;
4961 	struct i40e_hw *hw = &pf->hw;
4962 	u8 enabled_tc = 1; /* TC0 is always enabled */
4963 	u8 tc, i;
4964 	/* Get the iSCSI APP TLV */
4965 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4966 
4967 	for (i = 0; i < dcbcfg->numapps; i++) {
4968 		app = dcbcfg->app[i];
4969 		if (app.selector == I40E_APP_SEL_TCPIP &&
4970 		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4971 			tc = dcbcfg->etscfg.prioritytable[app.priority];
4972 			enabled_tc |= BIT(tc);
4973 			break;
4974 		}
4975 	}
4976 
4977 	return enabled_tc;
4978 }
4979 
4980 /**
4981  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4982  * @dcbcfg: the corresponding DCBx configuration structure
4983  *
4984  * Return the number of TCs from given DCBx configuration
4985  **/
4986 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4987 {
4988 	int i, tc_unused = 0;
4989 	u8 num_tc = 0;
4990 	u8 ret = 0;
4991 
4992 	/* Scan the ETS Config Priority Table to find
4993 	 * traffic class enabled for a given priority
4994 	 * and create a bitmask of enabled TCs
4995 	 */
4996 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4997 		num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4998 
4999 	/* Now scan the bitmask to check for
5000 	 * contiguous TCs starting with TC0
5001 	 */
5002 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5003 		if (num_tc & BIT(i)) {
5004 			if (!tc_unused) {
5005 				ret++;
5006 			} else {
5007 				pr_err("Non-contiguous TC - Disabling DCB\n");
5008 				return 1;
5009 			}
5010 		} else {
5011 			tc_unused = 1;
5012 		}
5013 	}
5014 
5015 	/* There is always at least TC0 */
5016 	if (!ret)
5017 		ret = 1;
5018 
5019 	return ret;
5020 }
5021 
5022 /**
5023  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
5024  * @dcbcfg: the corresponding DCBx configuration structure
5025  *
5026  * Query the current DCB configuration and return the number of
5027  * traffic classes enabled from the given DCBX config
5028  **/
5029 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
5030 {
5031 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
5032 	u8 enabled_tc = 1;
5033 	u8 i;
5034 
5035 	for (i = 0; i < num_tc; i++)
5036 		enabled_tc |= BIT(i);
5037 
5038 	return enabled_tc;
5039 }
5040 
5041 /**
5042  * i40e_mqprio_get_enabled_tc - Get enabled traffic classes
5043  * @pf: PF being queried
5044  *
5045  * Query the current MQPRIO configuration and return the number of
5046  * traffic classes enabled.
5047  **/
5048 static u8 i40e_mqprio_get_enabled_tc(struct i40e_pf *pf)
5049 {
5050 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5051 	u8 num_tc = vsi->mqprio_qopt.qopt.num_tc;
5052 	u8 enabled_tc = 1, i;
5053 
5054 	for (i = 1; i < num_tc; i++)
5055 		enabled_tc |= BIT(i);
5056 	return enabled_tc;
5057 }
5058 
5059 /**
5060  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
5061  * @pf: PF being queried
5062  *
5063  * Return number of traffic classes enabled for the given PF
5064  **/
5065 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
5066 {
5067 	struct i40e_hw *hw = &pf->hw;
5068 	u8 i, enabled_tc = 1;
5069 	u8 num_tc = 0;
5070 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5071 
5072 	if (pf->flags & I40E_FLAG_TC_MQPRIO)
5073 		return pf->vsi[pf->lan_vsi]->mqprio_qopt.qopt.num_tc;
5074 
5075 	/* If neither MQPRIO nor DCB is enabled, then always use single TC */
5076 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5077 		return 1;
5078 
5079 	/* SFP mode will be enabled for all TCs on port */
5080 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5081 		return i40e_dcb_get_num_tc(dcbcfg);
5082 
5083 	/* MFP mode return count of enabled TCs for this PF */
5084 	if (pf->hw.func_caps.iscsi)
5085 		enabled_tc =  i40e_get_iscsi_tc_map(pf);
5086 	else
5087 		return 1; /* Only TC0 */
5088 
5089 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5090 		if (enabled_tc & BIT(i))
5091 			num_tc++;
5092 	}
5093 	return num_tc;
5094 }
5095 
5096 /**
5097  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
5098  * @pf: PF being queried
5099  *
5100  * Return a bitmap for enabled traffic classes for this PF.
5101  **/
5102 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
5103 {
5104 	if (pf->flags & I40E_FLAG_TC_MQPRIO)
5105 		return i40e_mqprio_get_enabled_tc(pf);
5106 
5107 	/* If neither MQPRIO nor DCB is enabled for this PF then just return
5108 	 * default TC
5109 	 */
5110 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5111 		return I40E_DEFAULT_TRAFFIC_CLASS;
5112 
5113 	/* SFP mode we want PF to be enabled for all TCs */
5114 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5115 		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
5116 
5117 	/* MFP enabled and iSCSI PF type */
5118 	if (pf->hw.func_caps.iscsi)
5119 		return i40e_get_iscsi_tc_map(pf);
5120 	else
5121 		return I40E_DEFAULT_TRAFFIC_CLASS;
5122 }
5123 
5124 /**
5125  * i40e_vsi_get_bw_info - Query VSI BW Information
5126  * @vsi: the VSI being queried
5127  *
5128  * Returns 0 on success, negative value on failure
5129  **/
5130 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
5131 {
5132 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
5133 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
5134 	struct i40e_pf *pf = vsi->back;
5135 	struct i40e_hw *hw = &pf->hw;
5136 	i40e_status ret;
5137 	u32 tc_bw_max;
5138 	int i;
5139 
5140 	/* Get the VSI level BW configuration */
5141 	ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
5142 	if (ret) {
5143 		dev_info(&pf->pdev->dev,
5144 			 "couldn't get PF vsi bw config, err %s aq_err %s\n",
5145 			 i40e_stat_str(&pf->hw, ret),
5146 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5147 		return -EINVAL;
5148 	}
5149 
5150 	/* Get the VSI level BW configuration per TC */
5151 	ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
5152 					       NULL);
5153 	if (ret) {
5154 		dev_info(&pf->pdev->dev,
5155 			 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
5156 			 i40e_stat_str(&pf->hw, ret),
5157 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5158 		return -EINVAL;
5159 	}
5160 
5161 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
5162 		dev_info(&pf->pdev->dev,
5163 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
5164 			 bw_config.tc_valid_bits,
5165 			 bw_ets_config.tc_valid_bits);
5166 		/* Still continuing */
5167 	}
5168 
5169 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
5170 	vsi->bw_max_quanta = bw_config.max_bw;
5171 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
5172 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
5173 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5174 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
5175 		vsi->bw_ets_limit_credits[i] =
5176 					le16_to_cpu(bw_ets_config.credits[i]);
5177 		/* 3 bits out of 4 for each TC */
5178 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
5179 	}
5180 
5181 	return 0;
5182 }
5183 
5184 /**
5185  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
5186  * @vsi: the VSI being configured
5187  * @enabled_tc: TC bitmap
5188  * @bw_credits: BW shared credits per TC
5189  *
5190  * Returns 0 on success, negative value on failure
5191  **/
5192 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
5193 				       u8 *bw_share)
5194 {
5195 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5196 	i40e_status ret;
5197 	int i;
5198 
5199 	if (vsi->back->flags & I40E_FLAG_TC_MQPRIO)
5200 		return 0;
5201 	if (!vsi->mqprio_qopt.qopt.hw) {
5202 		ret = i40e_set_bw_limit(vsi, vsi->seid, 0);
5203 		if (ret)
5204 			dev_info(&vsi->back->pdev->dev,
5205 				 "Failed to reset tx rate for vsi->seid %u\n",
5206 				 vsi->seid);
5207 		return ret;
5208 	}
5209 	bw_data.tc_valid_bits = enabled_tc;
5210 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5211 		bw_data.tc_bw_credits[i] = bw_share[i];
5212 
5213 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
5214 				       NULL);
5215 	if (ret) {
5216 		dev_info(&vsi->back->pdev->dev,
5217 			 "AQ command Config VSI BW allocation per TC failed = %d\n",
5218 			 vsi->back->hw.aq.asq_last_status);
5219 		return -EINVAL;
5220 	}
5221 
5222 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5223 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
5224 
5225 	return 0;
5226 }
5227 
5228 /**
5229  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
5230  * @vsi: the VSI being configured
5231  * @enabled_tc: TC map to be enabled
5232  *
5233  **/
5234 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5235 {
5236 	struct net_device *netdev = vsi->netdev;
5237 	struct i40e_pf *pf = vsi->back;
5238 	struct i40e_hw *hw = &pf->hw;
5239 	u8 netdev_tc = 0;
5240 	int i;
5241 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5242 
5243 	if (!netdev)
5244 		return;
5245 
5246 	if (!enabled_tc) {
5247 		netdev_reset_tc(netdev);
5248 		return;
5249 	}
5250 
5251 	/* Set up actual enabled TCs on the VSI */
5252 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
5253 		return;
5254 
5255 	/* set per TC queues for the VSI */
5256 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5257 		/* Only set TC queues for enabled tcs
5258 		 *
5259 		 * e.g. For a VSI that has TC0 and TC3 enabled the
5260 		 * enabled_tc bitmap would be 0x00001001; the driver
5261 		 * will set the numtc for netdev as 2 that will be
5262 		 * referenced by the netdev layer as TC 0 and 1.
5263 		 */
5264 		if (vsi->tc_config.enabled_tc & BIT(i))
5265 			netdev_set_tc_queue(netdev,
5266 					vsi->tc_config.tc_info[i].netdev_tc,
5267 					vsi->tc_config.tc_info[i].qcount,
5268 					vsi->tc_config.tc_info[i].qoffset);
5269 	}
5270 
5271 	if (pf->flags & I40E_FLAG_TC_MQPRIO)
5272 		return;
5273 
5274 	/* Assign UP2TC map for the VSI */
5275 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
5276 		/* Get the actual TC# for the UP */
5277 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
5278 		/* Get the mapped netdev TC# for the UP */
5279 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
5280 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
5281 	}
5282 }
5283 
5284 /**
5285  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
5286  * @vsi: the VSI being configured
5287  * @ctxt: the ctxt buffer returned from AQ VSI update param command
5288  **/
5289 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
5290 				      struct i40e_vsi_context *ctxt)
5291 {
5292 	/* copy just the sections touched not the entire info
5293 	 * since not all sections are valid as returned by
5294 	 * update vsi params
5295 	 */
5296 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
5297 	memcpy(&vsi->info.queue_mapping,
5298 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
5299 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
5300 	       sizeof(vsi->info.tc_mapping));
5301 }
5302 
5303 /**
5304  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
5305  * @vsi: VSI to be configured
5306  * @enabled_tc: TC bitmap
5307  *
5308  * This configures a particular VSI for TCs that are mapped to the
5309  * given TC bitmap. It uses default bandwidth share for TCs across
5310  * VSIs to configure TC for a particular VSI.
5311  *
5312  * NOTE:
5313  * It is expected that the VSI queues have been quisced before calling
5314  * this function.
5315  **/
5316 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5317 {
5318 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5319 	struct i40e_vsi_context ctxt;
5320 	int ret = 0;
5321 	int i;
5322 
5323 	/* Check if enabled_tc is same as existing or new TCs */
5324 	if (vsi->tc_config.enabled_tc == enabled_tc &&
5325 	    vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
5326 		return ret;
5327 
5328 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
5329 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5330 		if (enabled_tc & BIT(i))
5331 			bw_share[i] = 1;
5332 	}
5333 
5334 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5335 	if (ret) {
5336 		dev_info(&vsi->back->pdev->dev,
5337 			 "Failed configuring TC map %d for VSI %d\n",
5338 			 enabled_tc, vsi->seid);
5339 		goto out;
5340 	}
5341 
5342 	/* Update Queue Pairs Mapping for currently enabled UPs */
5343 	ctxt.seid = vsi->seid;
5344 	ctxt.pf_num = vsi->back->hw.pf_id;
5345 	ctxt.vf_num = 0;
5346 	ctxt.uplink_seid = vsi->uplink_seid;
5347 	ctxt.info = vsi->info;
5348 	if (vsi->back->flags & I40E_FLAG_TC_MQPRIO) {
5349 		ret = i40e_vsi_setup_queue_map_mqprio(vsi, &ctxt, enabled_tc);
5350 		if (ret)
5351 			goto out;
5352 	} else {
5353 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5354 	}
5355 
5356 	/* On destroying the qdisc, reset vsi->rss_size, as number of enabled
5357 	 * queues changed.
5358 	 */
5359 	if (!vsi->mqprio_qopt.qopt.hw && vsi->reconfig_rss) {
5360 		vsi->rss_size = min_t(int, vsi->back->alloc_rss_size,
5361 				      vsi->num_queue_pairs);
5362 		ret = i40e_vsi_config_rss(vsi);
5363 		if (ret) {
5364 			dev_info(&vsi->back->pdev->dev,
5365 				 "Failed to reconfig rss for num_queues\n");
5366 			return ret;
5367 		}
5368 		vsi->reconfig_rss = false;
5369 	}
5370 	if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5371 		ctxt.info.valid_sections |=
5372 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5373 		ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5374 	}
5375 
5376 	/* Update the VSI after updating the VSI queue-mapping
5377 	 * information
5378 	 */
5379 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5380 	if (ret) {
5381 		dev_info(&vsi->back->pdev->dev,
5382 			 "Update vsi tc config failed, err %s aq_err %s\n",
5383 			 i40e_stat_str(&vsi->back->hw, ret),
5384 			 i40e_aq_str(&vsi->back->hw,
5385 				     vsi->back->hw.aq.asq_last_status));
5386 		goto out;
5387 	}
5388 	/* update the local VSI info with updated queue map */
5389 	i40e_vsi_update_queue_map(vsi, &ctxt);
5390 	vsi->info.valid_sections = 0;
5391 
5392 	/* Update current VSI BW information */
5393 	ret = i40e_vsi_get_bw_info(vsi);
5394 	if (ret) {
5395 		dev_info(&vsi->back->pdev->dev,
5396 			 "Failed updating vsi bw info, err %s aq_err %s\n",
5397 			 i40e_stat_str(&vsi->back->hw, ret),
5398 			 i40e_aq_str(&vsi->back->hw,
5399 				     vsi->back->hw.aq.asq_last_status));
5400 		goto out;
5401 	}
5402 
5403 	/* Update the netdev TC setup */
5404 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5405 out:
5406 	return ret;
5407 }
5408 
5409 /**
5410  * i40e_get_link_speed - Returns link speed for the interface
5411  * @vsi: VSI to be configured
5412  *
5413  **/
5414 int i40e_get_link_speed(struct i40e_vsi *vsi)
5415 {
5416 	struct i40e_pf *pf = vsi->back;
5417 
5418 	switch (pf->hw.phy.link_info.link_speed) {
5419 	case I40E_LINK_SPEED_40GB:
5420 		return 40000;
5421 	case I40E_LINK_SPEED_25GB:
5422 		return 25000;
5423 	case I40E_LINK_SPEED_20GB:
5424 		return 20000;
5425 	case I40E_LINK_SPEED_10GB:
5426 		return 10000;
5427 	case I40E_LINK_SPEED_1GB:
5428 		return 1000;
5429 	default:
5430 		return -EINVAL;
5431 	}
5432 }
5433 
5434 /**
5435  * i40e_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
5436  * @vsi: VSI to be configured
5437  * @seid: seid of the channel/VSI
5438  * @max_tx_rate: max TX rate to be configured as BW limit
5439  *
5440  * Helper function to set BW limit for a given VSI
5441  **/
5442 int i40e_set_bw_limit(struct i40e_vsi *vsi, u16 seid, u64 max_tx_rate)
5443 {
5444 	struct i40e_pf *pf = vsi->back;
5445 	int speed = 0;
5446 	int ret = 0;
5447 
5448 	speed = i40e_get_link_speed(vsi);
5449 	if (max_tx_rate > speed) {
5450 		dev_err(&pf->pdev->dev,
5451 			"Invalid max tx rate %llu specified for VSI seid %d.",
5452 			max_tx_rate, seid);
5453 		return -EINVAL;
5454 	}
5455 	if (max_tx_rate && max_tx_rate < 50) {
5456 		dev_warn(&pf->pdev->dev,
5457 			 "Setting max tx rate to minimum usable value of 50Mbps.\n");
5458 		max_tx_rate = 50;
5459 	}
5460 
5461 	/* Tx rate credits are in values of 50Mbps, 0 is disabled */
5462 	ret = i40e_aq_config_vsi_bw_limit(&pf->hw, seid,
5463 					  max_tx_rate / I40E_BW_CREDIT_DIVISOR,
5464 					  I40E_MAX_BW_INACTIVE_ACCUM, NULL);
5465 	if (ret)
5466 		dev_err(&pf->pdev->dev,
5467 			"Failed set tx rate (%llu Mbps) for vsi->seid %u, err %s aq_err %s\n",
5468 			max_tx_rate, seid, i40e_stat_str(&pf->hw, ret),
5469 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5470 	return ret;
5471 }
5472 
5473 /**
5474  * i40e_remove_queue_channels - Remove queue channels for the TCs
5475  * @vsi: VSI to be configured
5476  *
5477  * Remove queue channels for the TCs
5478  **/
5479 static void i40e_remove_queue_channels(struct i40e_vsi *vsi)
5480 {
5481 	struct i40e_channel *ch, *ch_tmp;
5482 	int ret, i;
5483 
5484 	/* Reset rss size that was stored when reconfiguring rss for
5485 	 * channel VSIs with non-power-of-2 queue count.
5486 	 */
5487 	vsi->current_rss_size = 0;
5488 
5489 	/* perform cleanup for channels if they exist */
5490 	if (list_empty(&vsi->ch_list))
5491 		return;
5492 
5493 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5494 		struct i40e_vsi *p_vsi;
5495 
5496 		list_del(&ch->list);
5497 		p_vsi = ch->parent_vsi;
5498 		if (!p_vsi || !ch->initialized) {
5499 			kfree(ch);
5500 			continue;
5501 		}
5502 		/* Reset queue contexts */
5503 		for (i = 0; i < ch->num_queue_pairs; i++) {
5504 			struct i40e_ring *tx_ring, *rx_ring;
5505 			u16 pf_q;
5506 
5507 			pf_q = ch->base_queue + i;
5508 			tx_ring = vsi->tx_rings[pf_q];
5509 			tx_ring->ch = NULL;
5510 
5511 			rx_ring = vsi->rx_rings[pf_q];
5512 			rx_ring->ch = NULL;
5513 		}
5514 
5515 		/* Reset BW configured for this VSI via mqprio */
5516 		ret = i40e_set_bw_limit(vsi, ch->seid, 0);
5517 		if (ret)
5518 			dev_info(&vsi->back->pdev->dev,
5519 				 "Failed to reset tx rate for ch->seid %u\n",
5520 				 ch->seid);
5521 
5522 		/* delete VSI from FW */
5523 		ret = i40e_aq_delete_element(&vsi->back->hw, ch->seid,
5524 					     NULL);
5525 		if (ret)
5526 			dev_err(&vsi->back->pdev->dev,
5527 				"unable to remove channel (%d) for parent VSI(%d)\n",
5528 				ch->seid, p_vsi->seid);
5529 		kfree(ch);
5530 	}
5531 	INIT_LIST_HEAD(&vsi->ch_list);
5532 }
5533 
5534 /**
5535  * i40e_is_any_channel - channel exist or not
5536  * @vsi: ptr to VSI to which channels are associated with
5537  *
5538  * Returns true or false if channel(s) exist for associated VSI or not
5539  **/
5540 static bool i40e_is_any_channel(struct i40e_vsi *vsi)
5541 {
5542 	struct i40e_channel *ch, *ch_tmp;
5543 
5544 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5545 		if (ch->initialized)
5546 			return true;
5547 	}
5548 
5549 	return false;
5550 }
5551 
5552 /**
5553  * i40e_get_max_queues_for_channel
5554  * @vsi: ptr to VSI to which channels are associated with
5555  *
5556  * Helper function which returns max value among the queue counts set on the
5557  * channels/TCs created.
5558  **/
5559 static int i40e_get_max_queues_for_channel(struct i40e_vsi *vsi)
5560 {
5561 	struct i40e_channel *ch, *ch_tmp;
5562 	int max = 0;
5563 
5564 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5565 		if (!ch->initialized)
5566 			continue;
5567 		if (ch->num_queue_pairs > max)
5568 			max = ch->num_queue_pairs;
5569 	}
5570 
5571 	return max;
5572 }
5573 
5574 /**
5575  * i40e_validate_num_queues - validate num_queues w.r.t channel
5576  * @pf: ptr to PF device
5577  * @num_queues: number of queues
5578  * @vsi: the parent VSI
5579  * @reconfig_rss: indicates should the RSS be reconfigured or not
5580  *
5581  * This function validates number of queues in the context of new channel
5582  * which is being established and determines if RSS should be reconfigured
5583  * or not for parent VSI.
5584  **/
5585 static int i40e_validate_num_queues(struct i40e_pf *pf, int num_queues,
5586 				    struct i40e_vsi *vsi, bool *reconfig_rss)
5587 {
5588 	int max_ch_queues;
5589 
5590 	if (!reconfig_rss)
5591 		return -EINVAL;
5592 
5593 	*reconfig_rss = false;
5594 
5595 	if (num_queues > I40E_MAX_QUEUES_PER_CH) {
5596 		dev_err(&pf->pdev->dev,
5597 			"Failed to create VMDq VSI. User requested num_queues (%d) > I40E_MAX_QUEUES_PER_VSI (%u)\n",
5598 			num_queues, I40E_MAX_QUEUES_PER_CH);
5599 		return -EINVAL;
5600 	}
5601 
5602 	if (vsi->current_rss_size) {
5603 		if (num_queues > vsi->current_rss_size) {
5604 			dev_dbg(&pf->pdev->dev,
5605 				"Error: num_queues (%d) > vsi's current_size(%d)\n",
5606 				num_queues, vsi->current_rss_size);
5607 			return -EINVAL;
5608 		} else if ((num_queues < vsi->current_rss_size) &&
5609 			   (!is_power_of_2(num_queues))) {
5610 			dev_dbg(&pf->pdev->dev,
5611 				"Error: num_queues (%d) < vsi's current_size(%d), but not power of 2\n",
5612 				num_queues, vsi->current_rss_size);
5613 			return -EINVAL;
5614 		}
5615 	}
5616 
5617 	if (!is_power_of_2(num_queues)) {
5618 		/* Find the max num_queues configured for channel if channel
5619 		 * exist.
5620 		 * if channel exist, then enforce 'num_queues' to be more than
5621 		 * max ever queues configured for channel.
5622 		 */
5623 		max_ch_queues = i40e_get_max_queues_for_channel(vsi);
5624 		if (num_queues < max_ch_queues) {
5625 			dev_dbg(&pf->pdev->dev,
5626 				"Error: num_queues (%d) < max queues configured for channel(%d)\n",
5627 				num_queues, max_ch_queues);
5628 			return -EINVAL;
5629 		}
5630 		*reconfig_rss = true;
5631 	}
5632 
5633 	return 0;
5634 }
5635 
5636 /**
5637  * i40e_vsi_reconfig_rss - reconfig RSS based on specified rss_size
5638  * @vsi: the VSI being setup
5639  * @rss_size: size of RSS, accordingly LUT gets reprogrammed
5640  *
5641  * This function reconfigures RSS by reprogramming LUTs using 'rss_size'
5642  **/
5643 static int i40e_vsi_reconfig_rss(struct i40e_vsi *vsi, u16 rss_size)
5644 {
5645 	struct i40e_pf *pf = vsi->back;
5646 	u8 seed[I40E_HKEY_ARRAY_SIZE];
5647 	struct i40e_hw *hw = &pf->hw;
5648 	int local_rss_size;
5649 	u8 *lut;
5650 	int ret;
5651 
5652 	if (!vsi->rss_size)
5653 		return -EINVAL;
5654 
5655 	if (rss_size > vsi->rss_size)
5656 		return -EINVAL;
5657 
5658 	local_rss_size = min_t(int, vsi->rss_size, rss_size);
5659 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
5660 	if (!lut)
5661 		return -ENOMEM;
5662 
5663 	/* Ignoring user configured lut if there is one */
5664 	i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, local_rss_size);
5665 
5666 	/* Use user configured hash key if there is one, otherwise
5667 	 * use default.
5668 	 */
5669 	if (vsi->rss_hkey_user)
5670 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
5671 	else
5672 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
5673 
5674 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
5675 	if (ret) {
5676 		dev_info(&pf->pdev->dev,
5677 			 "Cannot set RSS lut, err %s aq_err %s\n",
5678 			 i40e_stat_str(hw, ret),
5679 			 i40e_aq_str(hw, hw->aq.asq_last_status));
5680 		kfree(lut);
5681 		return ret;
5682 	}
5683 	kfree(lut);
5684 
5685 	/* Do the update w.r.t. storing rss_size */
5686 	if (!vsi->orig_rss_size)
5687 		vsi->orig_rss_size = vsi->rss_size;
5688 	vsi->current_rss_size = local_rss_size;
5689 
5690 	return ret;
5691 }
5692 
5693 /**
5694  * i40e_channel_setup_queue_map - Setup a channel queue map
5695  * @pf: ptr to PF device
5696  * @vsi: the VSI being setup
5697  * @ctxt: VSI context structure
5698  * @ch: ptr to channel structure
5699  *
5700  * Setup queue map for a specific channel
5701  **/
5702 static void i40e_channel_setup_queue_map(struct i40e_pf *pf,
5703 					 struct i40e_vsi_context *ctxt,
5704 					 struct i40e_channel *ch)
5705 {
5706 	u16 qcount, qmap, sections = 0;
5707 	u8 offset = 0;
5708 	int pow;
5709 
5710 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
5711 	sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
5712 
5713 	qcount = min_t(int, ch->num_queue_pairs, pf->num_lan_msix);
5714 	ch->num_queue_pairs = qcount;
5715 
5716 	/* find the next higher power-of-2 of num queue pairs */
5717 	pow = ilog2(qcount);
5718 	if (!is_power_of_2(qcount))
5719 		pow++;
5720 
5721 	qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
5722 		(pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
5723 
5724 	/* Setup queue TC[0].qmap for given VSI context */
5725 	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
5726 
5727 	ctxt->info.up_enable_bits = 0x1; /* TC0 enabled */
5728 	ctxt->info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
5729 	ctxt->info.queue_mapping[0] = cpu_to_le16(ch->base_queue);
5730 	ctxt->info.valid_sections |= cpu_to_le16(sections);
5731 }
5732 
5733 /**
5734  * i40e_add_channel - add a channel by adding VSI
5735  * @pf: ptr to PF device
5736  * @uplink_seid: underlying HW switching element (VEB) ID
5737  * @ch: ptr to channel structure
5738  *
5739  * Add a channel (VSI) using add_vsi and queue_map
5740  **/
5741 static int i40e_add_channel(struct i40e_pf *pf, u16 uplink_seid,
5742 			    struct i40e_channel *ch)
5743 {
5744 	struct i40e_hw *hw = &pf->hw;
5745 	struct i40e_vsi_context ctxt;
5746 	u8 enabled_tc = 0x1; /* TC0 enabled */
5747 	int ret;
5748 
5749 	if (ch->type != I40E_VSI_VMDQ2) {
5750 		dev_info(&pf->pdev->dev,
5751 			 "add new vsi failed, ch->type %d\n", ch->type);
5752 		return -EINVAL;
5753 	}
5754 
5755 	memset(&ctxt, 0, sizeof(ctxt));
5756 	ctxt.pf_num = hw->pf_id;
5757 	ctxt.vf_num = 0;
5758 	ctxt.uplink_seid = uplink_seid;
5759 	ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
5760 	if (ch->type == I40E_VSI_VMDQ2)
5761 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
5762 
5763 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED) {
5764 		ctxt.info.valid_sections |=
5765 		     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5766 		ctxt.info.switch_id =
5767 		   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5768 	}
5769 
5770 	/* Set queue map for a given VSI context */
5771 	i40e_channel_setup_queue_map(pf, &ctxt, ch);
5772 
5773 	/* Now time to create VSI */
5774 	ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
5775 	if (ret) {
5776 		dev_info(&pf->pdev->dev,
5777 			 "add new vsi failed, err %s aq_err %s\n",
5778 			 i40e_stat_str(&pf->hw, ret),
5779 			 i40e_aq_str(&pf->hw,
5780 				     pf->hw.aq.asq_last_status));
5781 		return -ENOENT;
5782 	}
5783 
5784 	/* Success, update channel */
5785 	ch->enabled_tc = enabled_tc;
5786 	ch->seid = ctxt.seid;
5787 	ch->vsi_number = ctxt.vsi_number;
5788 	ch->stat_counter_idx = cpu_to_le16(ctxt.info.stat_counter_idx);
5789 
5790 	/* copy just the sections touched not the entire info
5791 	 * since not all sections are valid as returned by
5792 	 * update vsi params
5793 	 */
5794 	ch->info.mapping_flags = ctxt.info.mapping_flags;
5795 	memcpy(&ch->info.queue_mapping,
5796 	       &ctxt.info.queue_mapping, sizeof(ctxt.info.queue_mapping));
5797 	memcpy(&ch->info.tc_mapping, ctxt.info.tc_mapping,
5798 	       sizeof(ctxt.info.tc_mapping));
5799 
5800 	return 0;
5801 }
5802 
5803 static int i40e_channel_config_bw(struct i40e_vsi *vsi, struct i40e_channel *ch,
5804 				  u8 *bw_share)
5805 {
5806 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5807 	i40e_status ret;
5808 	int i;
5809 
5810 	bw_data.tc_valid_bits = ch->enabled_tc;
5811 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5812 		bw_data.tc_bw_credits[i] = bw_share[i];
5813 
5814 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, ch->seid,
5815 				       &bw_data, NULL);
5816 	if (ret) {
5817 		dev_info(&vsi->back->pdev->dev,
5818 			 "Config VSI BW allocation per TC failed, aq_err: %d for new_vsi->seid %u\n",
5819 			 vsi->back->hw.aq.asq_last_status, ch->seid);
5820 		return -EINVAL;
5821 	}
5822 
5823 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5824 		ch->info.qs_handle[i] = bw_data.qs_handles[i];
5825 
5826 	return 0;
5827 }
5828 
5829 /**
5830  * i40e_channel_config_tx_ring - config TX ring associated with new channel
5831  * @pf: ptr to PF device
5832  * @vsi: the VSI being setup
5833  * @ch: ptr to channel structure
5834  *
5835  * Configure TX rings associated with channel (VSI) since queues are being
5836  * from parent VSI.
5837  **/
5838 static int i40e_channel_config_tx_ring(struct i40e_pf *pf,
5839 				       struct i40e_vsi *vsi,
5840 				       struct i40e_channel *ch)
5841 {
5842 	i40e_status ret;
5843 	int i;
5844 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5845 
5846 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
5847 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5848 		if (ch->enabled_tc & BIT(i))
5849 			bw_share[i] = 1;
5850 	}
5851 
5852 	/* configure BW for new VSI */
5853 	ret = i40e_channel_config_bw(vsi, ch, bw_share);
5854 	if (ret) {
5855 		dev_info(&vsi->back->pdev->dev,
5856 			 "Failed configuring TC map %d for channel (seid %u)\n",
5857 			 ch->enabled_tc, ch->seid);
5858 		return ret;
5859 	}
5860 
5861 	for (i = 0; i < ch->num_queue_pairs; i++) {
5862 		struct i40e_ring *tx_ring, *rx_ring;
5863 		u16 pf_q;
5864 
5865 		pf_q = ch->base_queue + i;
5866 
5867 		/* Get to TX ring ptr of main VSI, for re-setup TX queue
5868 		 * context
5869 		 */
5870 		tx_ring = vsi->tx_rings[pf_q];
5871 		tx_ring->ch = ch;
5872 
5873 		/* Get the RX ring ptr */
5874 		rx_ring = vsi->rx_rings[pf_q];
5875 		rx_ring->ch = ch;
5876 	}
5877 
5878 	return 0;
5879 }
5880 
5881 /**
5882  * i40e_setup_hw_channel - setup new channel
5883  * @pf: ptr to PF device
5884  * @vsi: the VSI being setup
5885  * @ch: ptr to channel structure
5886  * @uplink_seid: underlying HW switching element (VEB) ID
5887  * @type: type of channel to be created (VMDq2/VF)
5888  *
5889  * Setup new channel (VSI) based on specified type (VMDq2/VF)
5890  * and configures TX rings accordingly
5891  **/
5892 static inline int i40e_setup_hw_channel(struct i40e_pf *pf,
5893 					struct i40e_vsi *vsi,
5894 					struct i40e_channel *ch,
5895 					u16 uplink_seid, u8 type)
5896 {
5897 	int ret;
5898 
5899 	ch->initialized = false;
5900 	ch->base_queue = vsi->next_base_queue;
5901 	ch->type = type;
5902 
5903 	/* Proceed with creation of channel (VMDq2) VSI */
5904 	ret = i40e_add_channel(pf, uplink_seid, ch);
5905 	if (ret) {
5906 		dev_info(&pf->pdev->dev,
5907 			 "failed to add_channel using uplink_seid %u\n",
5908 			 uplink_seid);
5909 		return ret;
5910 	}
5911 
5912 	/* Mark the successful creation of channel */
5913 	ch->initialized = true;
5914 
5915 	/* Reconfigure TX queues using QTX_CTL register */
5916 	ret = i40e_channel_config_tx_ring(pf, vsi, ch);
5917 	if (ret) {
5918 		dev_info(&pf->pdev->dev,
5919 			 "failed to configure TX rings for channel %u\n",
5920 			 ch->seid);
5921 		return ret;
5922 	}
5923 
5924 	/* update 'next_base_queue' */
5925 	vsi->next_base_queue = vsi->next_base_queue + ch->num_queue_pairs;
5926 	dev_dbg(&pf->pdev->dev,
5927 		"Added channel: vsi_seid %u, vsi_number %u, stat_counter_idx %u, num_queue_pairs %u, pf->next_base_queue %d\n",
5928 		ch->seid, ch->vsi_number, ch->stat_counter_idx,
5929 		ch->num_queue_pairs,
5930 		vsi->next_base_queue);
5931 	return ret;
5932 }
5933 
5934 /**
5935  * i40e_setup_channel - setup new channel using uplink element
5936  * @pf: ptr to PF device
5937  * @type: type of channel to be created (VMDq2/VF)
5938  * @uplink_seid: underlying HW switching element (VEB) ID
5939  * @ch: ptr to channel structure
5940  *
5941  * Setup new channel (VSI) based on specified type (VMDq2/VF)
5942  * and uplink switching element (uplink_seid)
5943  **/
5944 static bool i40e_setup_channel(struct i40e_pf *pf, struct i40e_vsi *vsi,
5945 			       struct i40e_channel *ch)
5946 {
5947 	u8 vsi_type;
5948 	u16 seid;
5949 	int ret;
5950 
5951 	if (vsi->type == I40E_VSI_MAIN) {
5952 		vsi_type = I40E_VSI_VMDQ2;
5953 	} else {
5954 		dev_err(&pf->pdev->dev, "unsupported parent vsi type(%d)\n",
5955 			vsi->type);
5956 		return false;
5957 	}
5958 
5959 	/* underlying switching element */
5960 	seid = pf->vsi[pf->lan_vsi]->uplink_seid;
5961 
5962 	/* create channel (VSI), configure TX rings */
5963 	ret = i40e_setup_hw_channel(pf, vsi, ch, seid, vsi_type);
5964 	if (ret) {
5965 		dev_err(&pf->pdev->dev, "failed to setup hw_channel\n");
5966 		return false;
5967 	}
5968 
5969 	return ch->initialized ? true : false;
5970 }
5971 
5972 /**
5973  * i40e_create_queue_channel - function to create channel
5974  * @vsi: VSI to be configured
5975  * @ch: ptr to channel (it contains channel specific params)
5976  *
5977  * This function creates channel (VSI) using num_queues specified by user,
5978  * reconfigs RSS if needed.
5979  **/
5980 int i40e_create_queue_channel(struct i40e_vsi *vsi,
5981 			      struct i40e_channel *ch)
5982 {
5983 	struct i40e_pf *pf = vsi->back;
5984 	bool reconfig_rss;
5985 	int err;
5986 
5987 	if (!ch)
5988 		return -EINVAL;
5989 
5990 	if (!ch->num_queue_pairs) {
5991 		dev_err(&pf->pdev->dev, "Invalid num_queues requested: %d\n",
5992 			ch->num_queue_pairs);
5993 		return -EINVAL;
5994 	}
5995 
5996 	/* validate user requested num_queues for channel */
5997 	err = i40e_validate_num_queues(pf, ch->num_queue_pairs, vsi,
5998 				       &reconfig_rss);
5999 	if (err) {
6000 		dev_info(&pf->pdev->dev, "Failed to validate num_queues (%d)\n",
6001 			 ch->num_queue_pairs);
6002 		return -EINVAL;
6003 	}
6004 
6005 	/* By default we are in VEPA mode, if this is the first VF/VMDq
6006 	 * VSI to be added switch to VEB mode.
6007 	 */
6008 	if ((!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) ||
6009 	    (!i40e_is_any_channel(vsi))) {
6010 		if (!is_power_of_2(vsi->tc_config.tc_info[0].qcount)) {
6011 			dev_dbg(&pf->pdev->dev,
6012 				"Failed to create channel. Override queues (%u) not power of 2\n",
6013 				vsi->tc_config.tc_info[0].qcount);
6014 			return -EINVAL;
6015 		}
6016 
6017 		if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
6018 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
6019 
6020 			if (vsi->type == I40E_VSI_MAIN) {
6021 				if (pf->flags & I40E_FLAG_TC_MQPRIO)
6022 					i40e_do_reset(pf, I40E_PF_RESET_FLAG,
6023 						      true);
6024 				else
6025 					i40e_do_reset_safe(pf,
6026 							   I40E_PF_RESET_FLAG);
6027 			}
6028 		}
6029 		/* now onwards for main VSI, number of queues will be value
6030 		 * of TC0's queue count
6031 		 */
6032 	}
6033 
6034 	/* By this time, vsi->cnt_q_avail shall be set to non-zero and
6035 	 * it should be more than num_queues
6036 	 */
6037 	if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_queue_pairs) {
6038 		dev_dbg(&pf->pdev->dev,
6039 			"Error: cnt_q_avail (%u) less than num_queues %d\n",
6040 			vsi->cnt_q_avail, ch->num_queue_pairs);
6041 		return -EINVAL;
6042 	}
6043 
6044 	/* reconfig_rss only if vsi type is MAIN_VSI */
6045 	if (reconfig_rss && (vsi->type == I40E_VSI_MAIN)) {
6046 		err = i40e_vsi_reconfig_rss(vsi, ch->num_queue_pairs);
6047 		if (err) {
6048 			dev_info(&pf->pdev->dev,
6049 				 "Error: unable to reconfig rss for num_queues (%u)\n",
6050 				 ch->num_queue_pairs);
6051 			return -EINVAL;
6052 		}
6053 	}
6054 
6055 	if (!i40e_setup_channel(pf, vsi, ch)) {
6056 		dev_info(&pf->pdev->dev, "Failed to setup channel\n");
6057 		return -EINVAL;
6058 	}
6059 
6060 	dev_info(&pf->pdev->dev,
6061 		 "Setup channel (id:%u) utilizing num_queues %d\n",
6062 		 ch->seid, ch->num_queue_pairs);
6063 
6064 	/* configure VSI for BW limit */
6065 	if (ch->max_tx_rate) {
6066 		if (i40e_set_bw_limit(vsi, ch->seid, ch->max_tx_rate))
6067 			return -EINVAL;
6068 
6069 		dev_dbg(&pf->pdev->dev,
6070 			"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
6071 			ch->max_tx_rate,
6072 			ch->max_tx_rate / I40E_BW_CREDIT_DIVISOR, ch->seid);
6073 	}
6074 
6075 	/* in case of VF, this will be main SRIOV VSI */
6076 	ch->parent_vsi = vsi;
6077 
6078 	/* and update main_vsi's count for queue_available to use */
6079 	vsi->cnt_q_avail -= ch->num_queue_pairs;
6080 
6081 	return 0;
6082 }
6083 
6084 /**
6085  * i40e_configure_queue_channels - Add queue channel for the given TCs
6086  * @vsi: VSI to be configured
6087  *
6088  * Configures queue channel mapping to the given TCs
6089  **/
6090 static int i40e_configure_queue_channels(struct i40e_vsi *vsi)
6091 {
6092 	struct i40e_channel *ch;
6093 	int ret = 0, i;
6094 
6095 	/* Create app vsi with the TCs. Main VSI with TC0 is already set up */
6096 	for (i = 1; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6097 		if (vsi->tc_config.enabled_tc & BIT(i)) {
6098 			ch = kzalloc(sizeof(*ch), GFP_KERNEL);
6099 			if (!ch) {
6100 				ret = -ENOMEM;
6101 				goto err_free;
6102 			}
6103 
6104 			INIT_LIST_HEAD(&ch->list);
6105 			ch->num_queue_pairs =
6106 				vsi->tc_config.tc_info[i].qcount;
6107 			ch->base_queue =
6108 				vsi->tc_config.tc_info[i].qoffset;
6109 
6110 			/* Bandwidth limit through tc interface is in bytes/s,
6111 			 * change to Mbit/s
6112 			 */
6113 			ch->max_tx_rate =
6114 				vsi->mqprio_qopt.max_rate[i] / (1000000 / 8);
6115 
6116 			list_add_tail(&ch->list, &vsi->ch_list);
6117 
6118 			ret = i40e_create_queue_channel(vsi, ch);
6119 			if (ret) {
6120 				dev_err(&vsi->back->pdev->dev,
6121 					"Failed creating queue channel with TC%d: queues %d\n",
6122 					i, ch->num_queue_pairs);
6123 				goto err_free;
6124 			}
6125 		}
6126 	}
6127 	return ret;
6128 
6129 err_free:
6130 	i40e_remove_queue_channels(vsi);
6131 	return ret;
6132 }
6133 
6134 /**
6135  * i40e_veb_config_tc - Configure TCs for given VEB
6136  * @veb: given VEB
6137  * @enabled_tc: TC bitmap
6138  *
6139  * Configures given TC bitmap for VEB (switching) element
6140  **/
6141 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
6142 {
6143 	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
6144 	struct i40e_pf *pf = veb->pf;
6145 	int ret = 0;
6146 	int i;
6147 
6148 	/* No TCs or already enabled TCs just return */
6149 	if (!enabled_tc || veb->enabled_tc == enabled_tc)
6150 		return ret;
6151 
6152 	bw_data.tc_valid_bits = enabled_tc;
6153 	/* bw_data.absolute_credits is not set (relative) */
6154 
6155 	/* Enable ETS TCs with equal BW Share for now */
6156 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6157 		if (enabled_tc & BIT(i))
6158 			bw_data.tc_bw_share_credits[i] = 1;
6159 	}
6160 
6161 	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
6162 						   &bw_data, NULL);
6163 	if (ret) {
6164 		dev_info(&pf->pdev->dev,
6165 			 "VEB bw config failed, err %s aq_err %s\n",
6166 			 i40e_stat_str(&pf->hw, ret),
6167 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6168 		goto out;
6169 	}
6170 
6171 	/* Update the BW information */
6172 	ret = i40e_veb_get_bw_info(veb);
6173 	if (ret) {
6174 		dev_info(&pf->pdev->dev,
6175 			 "Failed getting veb bw config, err %s aq_err %s\n",
6176 			 i40e_stat_str(&pf->hw, ret),
6177 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6178 	}
6179 
6180 out:
6181 	return ret;
6182 }
6183 
6184 #ifdef CONFIG_I40E_DCB
6185 /**
6186  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
6187  * @pf: PF struct
6188  *
6189  * Reconfigure VEB/VSIs on a given PF; it is assumed that
6190  * the caller would've quiesce all the VSIs before calling
6191  * this function
6192  **/
6193 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
6194 {
6195 	u8 tc_map = 0;
6196 	int ret;
6197 	u8 v;
6198 
6199 	/* Enable the TCs available on PF to all VEBs */
6200 	tc_map = i40e_pf_get_tc_map(pf);
6201 	for (v = 0; v < I40E_MAX_VEB; v++) {
6202 		if (!pf->veb[v])
6203 			continue;
6204 		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
6205 		if (ret) {
6206 			dev_info(&pf->pdev->dev,
6207 				 "Failed configuring TC for VEB seid=%d\n",
6208 				 pf->veb[v]->seid);
6209 			/* Will try to configure as many components */
6210 		}
6211 	}
6212 
6213 	/* Update each VSI */
6214 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6215 		if (!pf->vsi[v])
6216 			continue;
6217 
6218 		/* - Enable all TCs for the LAN VSI
6219 		 * - For all others keep them at TC0 for now
6220 		 */
6221 		if (v == pf->lan_vsi)
6222 			tc_map = i40e_pf_get_tc_map(pf);
6223 		else
6224 			tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
6225 
6226 		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
6227 		if (ret) {
6228 			dev_info(&pf->pdev->dev,
6229 				 "Failed configuring TC for VSI seid=%d\n",
6230 				 pf->vsi[v]->seid);
6231 			/* Will try to configure as many components */
6232 		} else {
6233 			/* Re-configure VSI vectors based on updated TC map */
6234 			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
6235 			if (pf->vsi[v]->netdev)
6236 				i40e_dcbnl_set_all(pf->vsi[v]);
6237 		}
6238 	}
6239 }
6240 
6241 /**
6242  * i40e_resume_port_tx - Resume port Tx
6243  * @pf: PF struct
6244  *
6245  * Resume a port's Tx and issue a PF reset in case of failure to
6246  * resume.
6247  **/
6248 static int i40e_resume_port_tx(struct i40e_pf *pf)
6249 {
6250 	struct i40e_hw *hw = &pf->hw;
6251 	int ret;
6252 
6253 	ret = i40e_aq_resume_port_tx(hw, NULL);
6254 	if (ret) {
6255 		dev_info(&pf->pdev->dev,
6256 			 "Resume Port Tx failed, err %s aq_err %s\n",
6257 			  i40e_stat_str(&pf->hw, ret),
6258 			  i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6259 		/* Schedule PF reset to recover */
6260 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6261 		i40e_service_event_schedule(pf);
6262 	}
6263 
6264 	return ret;
6265 }
6266 
6267 /**
6268  * i40e_init_pf_dcb - Initialize DCB configuration
6269  * @pf: PF being configured
6270  *
6271  * Query the current DCB configuration and cache it
6272  * in the hardware structure
6273  **/
6274 static int i40e_init_pf_dcb(struct i40e_pf *pf)
6275 {
6276 	struct i40e_hw *hw = &pf->hw;
6277 	int err = 0;
6278 
6279 	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
6280 	if (pf->hw_features & I40E_HW_NO_DCB_SUPPORT)
6281 		goto out;
6282 
6283 	/* Get the initial DCB configuration */
6284 	err = i40e_init_dcb(hw);
6285 	if (!err) {
6286 		/* Device/Function is not DCBX capable */
6287 		if ((!hw->func_caps.dcb) ||
6288 		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
6289 			dev_info(&pf->pdev->dev,
6290 				 "DCBX offload is not supported or is disabled for this PF.\n");
6291 		} else {
6292 			/* When status is not DISABLED then DCBX in FW */
6293 			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
6294 				       DCB_CAP_DCBX_VER_IEEE;
6295 
6296 			pf->flags |= I40E_FLAG_DCB_CAPABLE;
6297 			/* Enable DCB tagging only when more than one TC
6298 			 * or explicitly disable if only one TC
6299 			 */
6300 			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
6301 				pf->flags |= I40E_FLAG_DCB_ENABLED;
6302 			else
6303 				pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6304 			dev_dbg(&pf->pdev->dev,
6305 				"DCBX offload is supported for this PF.\n");
6306 		}
6307 	} else {
6308 		dev_info(&pf->pdev->dev,
6309 			 "Query for DCB configuration failed, err %s aq_err %s\n",
6310 			 i40e_stat_str(&pf->hw, err),
6311 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6312 	}
6313 
6314 out:
6315 	return err;
6316 }
6317 #endif /* CONFIG_I40E_DCB */
6318 #define SPEED_SIZE 14
6319 #define FC_SIZE 8
6320 /**
6321  * i40e_print_link_message - print link up or down
6322  * @vsi: the VSI for which link needs a message
6323  */
6324 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
6325 {
6326 	enum i40e_aq_link_speed new_speed;
6327 	struct i40e_pf *pf = vsi->back;
6328 	char *speed = "Unknown";
6329 	char *fc = "Unknown";
6330 	char *fec = "";
6331 	char *req_fec = "";
6332 	char *an = "";
6333 
6334 	new_speed = pf->hw.phy.link_info.link_speed;
6335 
6336 	if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
6337 		return;
6338 	vsi->current_isup = isup;
6339 	vsi->current_speed = new_speed;
6340 	if (!isup) {
6341 		netdev_info(vsi->netdev, "NIC Link is Down\n");
6342 		return;
6343 	}
6344 
6345 	/* Warn user if link speed on NPAR enabled partition is not at
6346 	 * least 10GB
6347 	 */
6348 	if (pf->hw.func_caps.npar_enable &&
6349 	    (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
6350 	     pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
6351 		netdev_warn(vsi->netdev,
6352 			    "The partition detected link speed that is less than 10Gbps\n");
6353 
6354 	switch (pf->hw.phy.link_info.link_speed) {
6355 	case I40E_LINK_SPEED_40GB:
6356 		speed = "40 G";
6357 		break;
6358 	case I40E_LINK_SPEED_20GB:
6359 		speed = "20 G";
6360 		break;
6361 	case I40E_LINK_SPEED_25GB:
6362 		speed = "25 G";
6363 		break;
6364 	case I40E_LINK_SPEED_10GB:
6365 		speed = "10 G";
6366 		break;
6367 	case I40E_LINK_SPEED_1GB:
6368 		speed = "1000 M";
6369 		break;
6370 	case I40E_LINK_SPEED_100MB:
6371 		speed = "100 M";
6372 		break;
6373 	default:
6374 		break;
6375 	}
6376 
6377 	switch (pf->hw.fc.current_mode) {
6378 	case I40E_FC_FULL:
6379 		fc = "RX/TX";
6380 		break;
6381 	case I40E_FC_TX_PAUSE:
6382 		fc = "TX";
6383 		break;
6384 	case I40E_FC_RX_PAUSE:
6385 		fc = "RX";
6386 		break;
6387 	default:
6388 		fc = "None";
6389 		break;
6390 	}
6391 
6392 	if (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
6393 		req_fec = ", Requested FEC: None";
6394 		fec = ", FEC: None";
6395 		an = ", Autoneg: False";
6396 
6397 		if (pf->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
6398 			an = ", Autoneg: True";
6399 
6400 		if (pf->hw.phy.link_info.fec_info &
6401 		    I40E_AQ_CONFIG_FEC_KR_ENA)
6402 			fec = ", FEC: CL74 FC-FEC/BASE-R";
6403 		else if (pf->hw.phy.link_info.fec_info &
6404 			 I40E_AQ_CONFIG_FEC_RS_ENA)
6405 			fec = ", FEC: CL108 RS-FEC";
6406 
6407 		/* 'CL108 RS-FEC' should be displayed when RS is requested, or
6408 		 * both RS and FC are requested
6409 		 */
6410 		if (vsi->back->hw.phy.link_info.req_fec_info &
6411 		    (I40E_AQ_REQUEST_FEC_KR | I40E_AQ_REQUEST_FEC_RS)) {
6412 			if (vsi->back->hw.phy.link_info.req_fec_info &
6413 			    I40E_AQ_REQUEST_FEC_RS)
6414 				req_fec = ", Requested FEC: CL108 RS-FEC";
6415 			else
6416 				req_fec = ", Requested FEC: CL74 FC-FEC/BASE-R";
6417 		}
6418 	}
6419 
6420 	netdev_info(vsi->netdev, "NIC Link is Up, %sbps Full Duplex%s%s%s, Flow Control: %s\n",
6421 		    speed, req_fec, fec, an, fc);
6422 }
6423 
6424 /**
6425  * i40e_up_complete - Finish the last steps of bringing up a connection
6426  * @vsi: the VSI being configured
6427  **/
6428 static int i40e_up_complete(struct i40e_vsi *vsi)
6429 {
6430 	struct i40e_pf *pf = vsi->back;
6431 	int err;
6432 
6433 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6434 		i40e_vsi_configure_msix(vsi);
6435 	else
6436 		i40e_configure_msi_and_legacy(vsi);
6437 
6438 	/* start rings */
6439 	err = i40e_vsi_start_rings(vsi);
6440 	if (err)
6441 		return err;
6442 
6443 	clear_bit(__I40E_VSI_DOWN, vsi->state);
6444 	i40e_napi_enable_all(vsi);
6445 	i40e_vsi_enable_irq(vsi);
6446 
6447 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
6448 	    (vsi->netdev)) {
6449 		i40e_print_link_message(vsi, true);
6450 		netif_tx_start_all_queues(vsi->netdev);
6451 		netif_carrier_on(vsi->netdev);
6452 	}
6453 
6454 	/* replay FDIR SB filters */
6455 	if (vsi->type == I40E_VSI_FDIR) {
6456 		/* reset fd counters */
6457 		pf->fd_add_err = 0;
6458 		pf->fd_atr_cnt = 0;
6459 		i40e_fdir_filter_restore(vsi);
6460 	}
6461 
6462 	/* On the next run of the service_task, notify any clients of the new
6463 	 * opened netdev
6464 	 */
6465 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
6466 	i40e_service_event_schedule(pf);
6467 
6468 	return 0;
6469 }
6470 
6471 /**
6472  * i40e_vsi_reinit_locked - Reset the VSI
6473  * @vsi: the VSI being configured
6474  *
6475  * Rebuild the ring structs after some configuration
6476  * has changed, e.g. MTU size.
6477  **/
6478 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
6479 {
6480 	struct i40e_pf *pf = vsi->back;
6481 
6482 	WARN_ON(in_interrupt());
6483 	while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state))
6484 		usleep_range(1000, 2000);
6485 	i40e_down(vsi);
6486 
6487 	i40e_up(vsi);
6488 	clear_bit(__I40E_CONFIG_BUSY, pf->state);
6489 }
6490 
6491 /**
6492  * i40e_up - Bring the connection back up after being down
6493  * @vsi: the VSI being configured
6494  **/
6495 int i40e_up(struct i40e_vsi *vsi)
6496 {
6497 	int err;
6498 
6499 	err = i40e_vsi_configure(vsi);
6500 	if (!err)
6501 		err = i40e_up_complete(vsi);
6502 
6503 	return err;
6504 }
6505 
6506 /**
6507  * i40e_down - Shutdown the connection processing
6508  * @vsi: the VSI being stopped
6509  **/
6510 void i40e_down(struct i40e_vsi *vsi)
6511 {
6512 	int i;
6513 
6514 	/* It is assumed that the caller of this function
6515 	 * sets the vsi->state __I40E_VSI_DOWN bit.
6516 	 */
6517 	if (vsi->netdev) {
6518 		netif_carrier_off(vsi->netdev);
6519 		netif_tx_disable(vsi->netdev);
6520 	}
6521 	i40e_vsi_disable_irq(vsi);
6522 	i40e_vsi_stop_rings(vsi);
6523 	i40e_napi_disable_all(vsi);
6524 
6525 	for (i = 0; i < vsi->num_queue_pairs; i++) {
6526 		i40e_clean_tx_ring(vsi->tx_rings[i]);
6527 		if (i40e_enabled_xdp_vsi(vsi))
6528 			i40e_clean_tx_ring(vsi->xdp_rings[i]);
6529 		i40e_clean_rx_ring(vsi->rx_rings[i]);
6530 	}
6531 
6532 }
6533 
6534 /**
6535  * i40e_validate_mqprio_qopt- validate queue mapping info
6536  * @vsi: the VSI being configured
6537  * @mqprio_qopt: queue parametrs
6538  **/
6539 static int i40e_validate_mqprio_qopt(struct i40e_vsi *vsi,
6540 				     struct tc_mqprio_qopt_offload *mqprio_qopt)
6541 {
6542 	u64 sum_max_rate = 0;
6543 	int i;
6544 
6545 	if (mqprio_qopt->qopt.offset[0] != 0 ||
6546 	    mqprio_qopt->qopt.num_tc < 1 ||
6547 	    mqprio_qopt->qopt.num_tc > I40E_MAX_TRAFFIC_CLASS)
6548 		return -EINVAL;
6549 	for (i = 0; ; i++) {
6550 		if (!mqprio_qopt->qopt.count[i])
6551 			return -EINVAL;
6552 		if (mqprio_qopt->min_rate[i]) {
6553 			dev_err(&vsi->back->pdev->dev,
6554 				"Invalid min tx rate (greater than 0) specified\n");
6555 			return -EINVAL;
6556 		}
6557 		sum_max_rate += (mqprio_qopt->max_rate[i] / (1000000 / 8));
6558 
6559 		if (i >= mqprio_qopt->qopt.num_tc - 1)
6560 			break;
6561 		if (mqprio_qopt->qopt.offset[i + 1] !=
6562 		    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
6563 			return -EINVAL;
6564 	}
6565 	if (vsi->num_queue_pairs <
6566 	    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i])) {
6567 		return -EINVAL;
6568 	}
6569 	if (sum_max_rate > i40e_get_link_speed(vsi)) {
6570 		dev_err(&vsi->back->pdev->dev,
6571 			"Invalid max tx rate specified\n");
6572 		return -EINVAL;
6573 	}
6574 	return 0;
6575 }
6576 
6577 /**
6578  * i40e_vsi_set_default_tc_config - set default values for tc configuration
6579  * @vsi: the VSI being configured
6580  **/
6581 static void i40e_vsi_set_default_tc_config(struct i40e_vsi *vsi)
6582 {
6583 	u16 qcount;
6584 	int i;
6585 
6586 	/* Only TC0 is enabled */
6587 	vsi->tc_config.numtc = 1;
6588 	vsi->tc_config.enabled_tc = 1;
6589 	qcount = min_t(int, vsi->alloc_queue_pairs,
6590 		       i40e_pf_get_max_q_per_tc(vsi->back));
6591 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6592 		/* For the TC that is not enabled set the offset to to default
6593 		 * queue and allocate one queue for the given TC.
6594 		 */
6595 		vsi->tc_config.tc_info[i].qoffset = 0;
6596 		if (i == 0)
6597 			vsi->tc_config.tc_info[i].qcount = qcount;
6598 		else
6599 			vsi->tc_config.tc_info[i].qcount = 1;
6600 		vsi->tc_config.tc_info[i].netdev_tc = 0;
6601 	}
6602 }
6603 
6604 /**
6605  * i40e_setup_tc - configure multiple traffic classes
6606  * @netdev: net device to configure
6607  * @type_data: tc offload data
6608  **/
6609 static int i40e_setup_tc(struct net_device *netdev, void *type_data)
6610 {
6611 	struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
6612 	struct i40e_netdev_priv *np = netdev_priv(netdev);
6613 	struct i40e_vsi *vsi = np->vsi;
6614 	struct i40e_pf *pf = vsi->back;
6615 	u8 enabled_tc = 0, num_tc, hw;
6616 	bool need_reset = false;
6617 	int ret = -EINVAL;
6618 	u16 mode;
6619 	int i;
6620 
6621 	num_tc = mqprio_qopt->qopt.num_tc;
6622 	hw = mqprio_qopt->qopt.hw;
6623 	mode = mqprio_qopt->mode;
6624 	if (!hw) {
6625 		pf->flags &= ~I40E_FLAG_TC_MQPRIO;
6626 		memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
6627 		goto config_tc;
6628 	}
6629 
6630 	/* Check if MFP enabled */
6631 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
6632 		netdev_info(netdev,
6633 			    "Configuring TC not supported in MFP mode\n");
6634 		return ret;
6635 	}
6636 	switch (mode) {
6637 	case TC_MQPRIO_MODE_DCB:
6638 		pf->flags &= ~I40E_FLAG_TC_MQPRIO;
6639 
6640 		/* Check if DCB enabled to continue */
6641 		if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
6642 			netdev_info(netdev,
6643 				    "DCB is not enabled for adapter\n");
6644 			return ret;
6645 		}
6646 
6647 		/* Check whether tc count is within enabled limit */
6648 		if (num_tc > i40e_pf_get_num_tc(pf)) {
6649 			netdev_info(netdev,
6650 				    "TC count greater than enabled on link for adapter\n");
6651 			return ret;
6652 		}
6653 		break;
6654 	case TC_MQPRIO_MODE_CHANNEL:
6655 		if (pf->flags & I40E_FLAG_DCB_ENABLED) {
6656 			netdev_info(netdev,
6657 				    "Full offload of TC Mqprio options is not supported when DCB is enabled\n");
6658 			return ret;
6659 		}
6660 		if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
6661 			return ret;
6662 		ret = i40e_validate_mqprio_qopt(vsi, mqprio_qopt);
6663 		if (ret)
6664 			return ret;
6665 		memcpy(&vsi->mqprio_qopt, mqprio_qopt,
6666 		       sizeof(*mqprio_qopt));
6667 		pf->flags |= I40E_FLAG_TC_MQPRIO;
6668 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6669 		break;
6670 	default:
6671 		return -EINVAL;
6672 	}
6673 
6674 config_tc:
6675 	/* Generate TC map for number of tc requested */
6676 	for (i = 0; i < num_tc; i++)
6677 		enabled_tc |= BIT(i);
6678 
6679 	/* Requesting same TC configuration as already enabled */
6680 	if (enabled_tc == vsi->tc_config.enabled_tc &&
6681 	    mode != TC_MQPRIO_MODE_CHANNEL)
6682 		return 0;
6683 
6684 	/* Quiesce VSI queues */
6685 	i40e_quiesce_vsi(vsi);
6686 
6687 	if (!hw && !(pf->flags & I40E_FLAG_TC_MQPRIO))
6688 		i40e_remove_queue_channels(vsi);
6689 
6690 	/* Configure VSI for enabled TCs */
6691 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
6692 	if (ret) {
6693 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
6694 			    vsi->seid);
6695 		need_reset = true;
6696 		goto exit;
6697 	}
6698 
6699 	if (pf->flags & I40E_FLAG_TC_MQPRIO) {
6700 		if (vsi->mqprio_qopt.max_rate[0]) {
6701 			u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0] /
6702 								(1000000 / 8);
6703 			ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
6704 			if (!ret) {
6705 				dev_dbg(&vsi->back->pdev->dev,
6706 					"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
6707 					max_tx_rate,
6708 					max_tx_rate / I40E_BW_CREDIT_DIVISOR,
6709 					vsi->seid);
6710 			} else {
6711 				need_reset = true;
6712 				goto exit;
6713 			}
6714 		}
6715 		ret = i40e_configure_queue_channels(vsi);
6716 		if (ret) {
6717 			netdev_info(netdev,
6718 				    "Failed configuring queue channels\n");
6719 			need_reset = true;
6720 			goto exit;
6721 		}
6722 	}
6723 
6724 exit:
6725 	/* Reset the configuration data to defaults, only TC0 is enabled */
6726 	if (need_reset) {
6727 		i40e_vsi_set_default_tc_config(vsi);
6728 		need_reset = false;
6729 	}
6730 
6731 	/* Unquiesce VSI */
6732 	i40e_unquiesce_vsi(vsi);
6733 	return ret;
6734 }
6735 
6736 static int __i40e_setup_tc(struct net_device *netdev, enum tc_setup_type type,
6737 			   void *type_data)
6738 {
6739 	if (type != TC_SETUP_MQPRIO)
6740 		return -EOPNOTSUPP;
6741 
6742 	return i40e_setup_tc(netdev, type_data);
6743 }
6744 
6745 /**
6746  * i40e_open - Called when a network interface is made active
6747  * @netdev: network interface device structure
6748  *
6749  * The open entry point is called when a network interface is made
6750  * active by the system (IFF_UP).  At this point all resources needed
6751  * for transmit and receive operations are allocated, the interrupt
6752  * handler is registered with the OS, the netdev watchdog subtask is
6753  * enabled, and the stack is notified that the interface is ready.
6754  *
6755  * Returns 0 on success, negative value on failure
6756  **/
6757 int i40e_open(struct net_device *netdev)
6758 {
6759 	struct i40e_netdev_priv *np = netdev_priv(netdev);
6760 	struct i40e_vsi *vsi = np->vsi;
6761 	struct i40e_pf *pf = vsi->back;
6762 	int err;
6763 
6764 	/* disallow open during test or if eeprom is broken */
6765 	if (test_bit(__I40E_TESTING, pf->state) ||
6766 	    test_bit(__I40E_BAD_EEPROM, pf->state))
6767 		return -EBUSY;
6768 
6769 	netif_carrier_off(netdev);
6770 
6771 	err = i40e_vsi_open(vsi);
6772 	if (err)
6773 		return err;
6774 
6775 	/* configure global TSO hardware offload settings */
6776 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
6777 						       TCP_FLAG_FIN) >> 16);
6778 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
6779 						       TCP_FLAG_FIN |
6780 						       TCP_FLAG_CWR) >> 16);
6781 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
6782 
6783 	udp_tunnel_get_rx_info(netdev);
6784 
6785 	return 0;
6786 }
6787 
6788 /**
6789  * i40e_vsi_open -
6790  * @vsi: the VSI to open
6791  *
6792  * Finish initialization of the VSI.
6793  *
6794  * Returns 0 on success, negative value on failure
6795  *
6796  * Note: expects to be called while under rtnl_lock()
6797  **/
6798 int i40e_vsi_open(struct i40e_vsi *vsi)
6799 {
6800 	struct i40e_pf *pf = vsi->back;
6801 	char int_name[I40E_INT_NAME_STR_LEN];
6802 	int err;
6803 
6804 	/* allocate descriptors */
6805 	err = i40e_vsi_setup_tx_resources(vsi);
6806 	if (err)
6807 		goto err_setup_tx;
6808 	err = i40e_vsi_setup_rx_resources(vsi);
6809 	if (err)
6810 		goto err_setup_rx;
6811 
6812 	err = i40e_vsi_configure(vsi);
6813 	if (err)
6814 		goto err_setup_rx;
6815 
6816 	if (vsi->netdev) {
6817 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6818 			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
6819 		err = i40e_vsi_request_irq(vsi, int_name);
6820 		if (err)
6821 			goto err_setup_rx;
6822 
6823 		/* Notify the stack of the actual queue counts. */
6824 		err = netif_set_real_num_tx_queues(vsi->netdev,
6825 						   vsi->num_queue_pairs);
6826 		if (err)
6827 			goto err_set_queues;
6828 
6829 		err = netif_set_real_num_rx_queues(vsi->netdev,
6830 						   vsi->num_queue_pairs);
6831 		if (err)
6832 			goto err_set_queues;
6833 
6834 	} else if (vsi->type == I40E_VSI_FDIR) {
6835 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
6836 			 dev_driver_string(&pf->pdev->dev),
6837 			 dev_name(&pf->pdev->dev));
6838 		err = i40e_vsi_request_irq(vsi, int_name);
6839 
6840 	} else {
6841 		err = -EINVAL;
6842 		goto err_setup_rx;
6843 	}
6844 
6845 	err = i40e_up_complete(vsi);
6846 	if (err)
6847 		goto err_up_complete;
6848 
6849 	return 0;
6850 
6851 err_up_complete:
6852 	i40e_down(vsi);
6853 err_set_queues:
6854 	i40e_vsi_free_irq(vsi);
6855 err_setup_rx:
6856 	i40e_vsi_free_rx_resources(vsi);
6857 err_setup_tx:
6858 	i40e_vsi_free_tx_resources(vsi);
6859 	if (vsi == pf->vsi[pf->lan_vsi])
6860 		i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
6861 
6862 	return err;
6863 }
6864 
6865 /**
6866  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
6867  * @pf: Pointer to PF
6868  *
6869  * This function destroys the hlist where all the Flow Director
6870  * filters were saved.
6871  **/
6872 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
6873 {
6874 	struct i40e_fdir_filter *filter;
6875 	struct i40e_flex_pit *pit_entry, *tmp;
6876 	struct hlist_node *node2;
6877 
6878 	hlist_for_each_entry_safe(filter, node2,
6879 				  &pf->fdir_filter_list, fdir_node) {
6880 		hlist_del(&filter->fdir_node);
6881 		kfree(filter);
6882 	}
6883 
6884 	list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
6885 		list_del(&pit_entry->list);
6886 		kfree(pit_entry);
6887 	}
6888 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
6889 
6890 	list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
6891 		list_del(&pit_entry->list);
6892 		kfree(pit_entry);
6893 	}
6894 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
6895 
6896 	pf->fdir_pf_active_filters = 0;
6897 	pf->fd_tcp4_filter_cnt = 0;
6898 	pf->fd_udp4_filter_cnt = 0;
6899 	pf->fd_sctp4_filter_cnt = 0;
6900 	pf->fd_ip4_filter_cnt = 0;
6901 
6902 	/* Reprogram the default input set for TCP/IPv4 */
6903 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
6904 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
6905 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
6906 
6907 	/* Reprogram the default input set for UDP/IPv4 */
6908 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
6909 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
6910 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
6911 
6912 	/* Reprogram the default input set for SCTP/IPv4 */
6913 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
6914 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
6915 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
6916 
6917 	/* Reprogram the default input set for Other/IPv4 */
6918 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
6919 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
6920 }
6921 
6922 /**
6923  * i40e_close - Disables a network interface
6924  * @netdev: network interface device structure
6925  *
6926  * The close entry point is called when an interface is de-activated
6927  * by the OS.  The hardware is still under the driver's control, but
6928  * this netdev interface is disabled.
6929  *
6930  * Returns 0, this is not allowed to fail
6931  **/
6932 int i40e_close(struct net_device *netdev)
6933 {
6934 	struct i40e_netdev_priv *np = netdev_priv(netdev);
6935 	struct i40e_vsi *vsi = np->vsi;
6936 
6937 	i40e_vsi_close(vsi);
6938 
6939 	return 0;
6940 }
6941 
6942 /**
6943  * i40e_do_reset - Start a PF or Core Reset sequence
6944  * @pf: board private structure
6945  * @reset_flags: which reset is requested
6946  * @lock_acquired: indicates whether or not the lock has been acquired
6947  * before this function was called.
6948  *
6949  * The essential difference in resets is that the PF Reset
6950  * doesn't clear the packet buffers, doesn't reset the PE
6951  * firmware, and doesn't bother the other PFs on the chip.
6952  **/
6953 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
6954 {
6955 	u32 val;
6956 
6957 	WARN_ON(in_interrupt());
6958 
6959 
6960 	/* do the biggest reset indicated */
6961 	if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
6962 
6963 		/* Request a Global Reset
6964 		 *
6965 		 * This will start the chip's countdown to the actual full
6966 		 * chip reset event, and a warning interrupt to be sent
6967 		 * to all PFs, including the requestor.  Our handler
6968 		 * for the warning interrupt will deal with the shutdown
6969 		 * and recovery of the switch setup.
6970 		 */
6971 		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
6972 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
6973 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
6974 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
6975 
6976 	} else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
6977 
6978 		/* Request a Core Reset
6979 		 *
6980 		 * Same as Global Reset, except does *not* include the MAC/PHY
6981 		 */
6982 		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
6983 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
6984 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
6985 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
6986 		i40e_flush(&pf->hw);
6987 
6988 	} else if (reset_flags & I40E_PF_RESET_FLAG) {
6989 
6990 		/* Request a PF Reset
6991 		 *
6992 		 * Resets only the PF-specific registers
6993 		 *
6994 		 * This goes directly to the tear-down and rebuild of
6995 		 * the switch, since we need to do all the recovery as
6996 		 * for the Core Reset.
6997 		 */
6998 		dev_dbg(&pf->pdev->dev, "PFR requested\n");
6999 		i40e_handle_reset_warning(pf, lock_acquired);
7000 
7001 	} else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
7002 		int v;
7003 
7004 		/* Find the VSI(s) that requested a re-init */
7005 		dev_info(&pf->pdev->dev,
7006 			 "VSI reinit requested\n");
7007 		for (v = 0; v < pf->num_alloc_vsi; v++) {
7008 			struct i40e_vsi *vsi = pf->vsi[v];
7009 
7010 			if (vsi != NULL &&
7011 			    test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
7012 					       vsi->state))
7013 				i40e_vsi_reinit_locked(pf->vsi[v]);
7014 		}
7015 	} else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
7016 		int v;
7017 
7018 		/* Find the VSI(s) that needs to be brought down */
7019 		dev_info(&pf->pdev->dev, "VSI down requested\n");
7020 		for (v = 0; v < pf->num_alloc_vsi; v++) {
7021 			struct i40e_vsi *vsi = pf->vsi[v];
7022 
7023 			if (vsi != NULL &&
7024 			    test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
7025 					       vsi->state)) {
7026 				set_bit(__I40E_VSI_DOWN, vsi->state);
7027 				i40e_down(vsi);
7028 			}
7029 		}
7030 	} else {
7031 		dev_info(&pf->pdev->dev,
7032 			 "bad reset request 0x%08x\n", reset_flags);
7033 	}
7034 }
7035 
7036 #ifdef CONFIG_I40E_DCB
7037 /**
7038  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
7039  * @pf: board private structure
7040  * @old_cfg: current DCB config
7041  * @new_cfg: new DCB config
7042  **/
7043 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
7044 			    struct i40e_dcbx_config *old_cfg,
7045 			    struct i40e_dcbx_config *new_cfg)
7046 {
7047 	bool need_reconfig = false;
7048 
7049 	/* Check if ETS configuration has changed */
7050 	if (memcmp(&new_cfg->etscfg,
7051 		   &old_cfg->etscfg,
7052 		   sizeof(new_cfg->etscfg))) {
7053 		/* If Priority Table has changed reconfig is needed */
7054 		if (memcmp(&new_cfg->etscfg.prioritytable,
7055 			   &old_cfg->etscfg.prioritytable,
7056 			   sizeof(new_cfg->etscfg.prioritytable))) {
7057 			need_reconfig = true;
7058 			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
7059 		}
7060 
7061 		if (memcmp(&new_cfg->etscfg.tcbwtable,
7062 			   &old_cfg->etscfg.tcbwtable,
7063 			   sizeof(new_cfg->etscfg.tcbwtable)))
7064 			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
7065 
7066 		if (memcmp(&new_cfg->etscfg.tsatable,
7067 			   &old_cfg->etscfg.tsatable,
7068 			   sizeof(new_cfg->etscfg.tsatable)))
7069 			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
7070 	}
7071 
7072 	/* Check if PFC configuration has changed */
7073 	if (memcmp(&new_cfg->pfc,
7074 		   &old_cfg->pfc,
7075 		   sizeof(new_cfg->pfc))) {
7076 		need_reconfig = true;
7077 		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
7078 	}
7079 
7080 	/* Check if APP Table has changed */
7081 	if (memcmp(&new_cfg->app,
7082 		   &old_cfg->app,
7083 		   sizeof(new_cfg->app))) {
7084 		need_reconfig = true;
7085 		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
7086 	}
7087 
7088 	dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
7089 	return need_reconfig;
7090 }
7091 
7092 /**
7093  * i40e_handle_lldp_event - Handle LLDP Change MIB event
7094  * @pf: board private structure
7095  * @e: event info posted on ARQ
7096  **/
7097 static int i40e_handle_lldp_event(struct i40e_pf *pf,
7098 				  struct i40e_arq_event_info *e)
7099 {
7100 	struct i40e_aqc_lldp_get_mib *mib =
7101 		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
7102 	struct i40e_hw *hw = &pf->hw;
7103 	struct i40e_dcbx_config tmp_dcbx_cfg;
7104 	bool need_reconfig = false;
7105 	int ret = 0;
7106 	u8 type;
7107 
7108 	/* Not DCB capable or capability disabled */
7109 	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
7110 		return ret;
7111 
7112 	/* Ignore if event is not for Nearest Bridge */
7113 	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
7114 		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
7115 	dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
7116 	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
7117 		return ret;
7118 
7119 	/* Check MIB Type and return if event for Remote MIB update */
7120 	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
7121 	dev_dbg(&pf->pdev->dev,
7122 		"LLDP event mib type %s\n", type ? "remote" : "local");
7123 	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
7124 		/* Update the remote cached instance and return */
7125 		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
7126 				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
7127 				&hw->remote_dcbx_config);
7128 		goto exit;
7129 	}
7130 
7131 	/* Store the old configuration */
7132 	tmp_dcbx_cfg = hw->local_dcbx_config;
7133 
7134 	/* Reset the old DCBx configuration data */
7135 	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
7136 	/* Get updated DCBX data from firmware */
7137 	ret = i40e_get_dcb_config(&pf->hw);
7138 	if (ret) {
7139 		dev_info(&pf->pdev->dev,
7140 			 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
7141 			 i40e_stat_str(&pf->hw, ret),
7142 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7143 		goto exit;
7144 	}
7145 
7146 	/* No change detected in DCBX configs */
7147 	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
7148 		    sizeof(tmp_dcbx_cfg))) {
7149 		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
7150 		goto exit;
7151 	}
7152 
7153 	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
7154 					       &hw->local_dcbx_config);
7155 
7156 	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
7157 
7158 	if (!need_reconfig)
7159 		goto exit;
7160 
7161 	/* Enable DCB tagging only when more than one TC */
7162 	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
7163 		pf->flags |= I40E_FLAG_DCB_ENABLED;
7164 	else
7165 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
7166 
7167 	set_bit(__I40E_PORT_SUSPENDED, pf->state);
7168 	/* Reconfiguration needed quiesce all VSIs */
7169 	i40e_pf_quiesce_all_vsi(pf);
7170 
7171 	/* Changes in configuration update VEB/VSI */
7172 	i40e_dcb_reconfigure(pf);
7173 
7174 	ret = i40e_resume_port_tx(pf);
7175 
7176 	clear_bit(__I40E_PORT_SUSPENDED, pf->state);
7177 	/* In case of error no point in resuming VSIs */
7178 	if (ret)
7179 		goto exit;
7180 
7181 	/* Wait for the PF's queues to be disabled */
7182 	ret = i40e_pf_wait_queues_disabled(pf);
7183 	if (ret) {
7184 		/* Schedule PF reset to recover */
7185 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
7186 		i40e_service_event_schedule(pf);
7187 	} else {
7188 		i40e_pf_unquiesce_all_vsi(pf);
7189 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
7190 		      I40E_FLAG_CLIENT_L2_CHANGE);
7191 	}
7192 
7193 exit:
7194 	return ret;
7195 }
7196 #endif /* CONFIG_I40E_DCB */
7197 
7198 /**
7199  * i40e_do_reset_safe - Protected reset path for userland calls.
7200  * @pf: board private structure
7201  * @reset_flags: which reset is requested
7202  *
7203  **/
7204 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
7205 {
7206 	rtnl_lock();
7207 	i40e_do_reset(pf, reset_flags, true);
7208 	rtnl_unlock();
7209 }
7210 
7211 /**
7212  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
7213  * @pf: board private structure
7214  * @e: event info posted on ARQ
7215  *
7216  * Handler for LAN Queue Overflow Event generated by the firmware for PF
7217  * and VF queues
7218  **/
7219 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
7220 					   struct i40e_arq_event_info *e)
7221 {
7222 	struct i40e_aqc_lan_overflow *data =
7223 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
7224 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
7225 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
7226 	struct i40e_hw *hw = &pf->hw;
7227 	struct i40e_vf *vf;
7228 	u16 vf_id;
7229 
7230 	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
7231 		queue, qtx_ctl);
7232 
7233 	/* Queue belongs to VF, find the VF and issue VF reset */
7234 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
7235 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
7236 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
7237 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
7238 		vf_id -= hw->func_caps.vf_base_id;
7239 		vf = &pf->vf[vf_id];
7240 		i40e_vc_notify_vf_reset(vf);
7241 		/* Allow VF to process pending reset notification */
7242 		msleep(20);
7243 		i40e_reset_vf(vf, false);
7244 	}
7245 }
7246 
7247 /**
7248  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
7249  * @pf: board private structure
7250  **/
7251 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
7252 {
7253 	u32 val, fcnt_prog;
7254 
7255 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
7256 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
7257 	return fcnt_prog;
7258 }
7259 
7260 /**
7261  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
7262  * @pf: board private structure
7263  **/
7264 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
7265 {
7266 	u32 val, fcnt_prog;
7267 
7268 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
7269 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
7270 		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
7271 		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
7272 	return fcnt_prog;
7273 }
7274 
7275 /**
7276  * i40e_get_global_fd_count - Get total FD filters programmed on device
7277  * @pf: board private structure
7278  **/
7279 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
7280 {
7281 	u32 val, fcnt_prog;
7282 
7283 	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
7284 	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
7285 		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
7286 		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
7287 	return fcnt_prog;
7288 }
7289 
7290 /**
7291  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
7292  * @pf: board private structure
7293  **/
7294 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
7295 {
7296 	struct i40e_fdir_filter *filter;
7297 	u32 fcnt_prog, fcnt_avail;
7298 	struct hlist_node *node;
7299 
7300 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
7301 		return;
7302 
7303 	/* Check if we have enough room to re-enable FDir SB capability. */
7304 	fcnt_prog = i40e_get_global_fd_count(pf);
7305 	fcnt_avail = pf->fdir_pf_filter_count;
7306 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
7307 	    (pf->fd_add_err == 0) ||
7308 	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
7309 		if (pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED) {
7310 			pf->flags &= ~I40E_FLAG_FD_SB_AUTO_DISABLED;
7311 			if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
7312 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
7313 				dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
7314 		}
7315 	}
7316 
7317 	/* We should wait for even more space before re-enabling ATR.
7318 	 * Additionally, we cannot enable ATR as long as we still have TCP SB
7319 	 * rules active.
7320 	 */
7321 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
7322 	    (pf->fd_tcp4_filter_cnt == 0)) {
7323 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
7324 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
7325 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
7326 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
7327 				dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
7328 		}
7329 	}
7330 
7331 	/* if hw had a problem adding a filter, delete it */
7332 	if (pf->fd_inv > 0) {
7333 		hlist_for_each_entry_safe(filter, node,
7334 					  &pf->fdir_filter_list, fdir_node) {
7335 			if (filter->fd_id == pf->fd_inv) {
7336 				hlist_del(&filter->fdir_node);
7337 				kfree(filter);
7338 				pf->fdir_pf_active_filters--;
7339 				pf->fd_inv = 0;
7340 			}
7341 		}
7342 	}
7343 }
7344 
7345 #define I40E_MIN_FD_FLUSH_INTERVAL 10
7346 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
7347 /**
7348  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
7349  * @pf: board private structure
7350  **/
7351 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
7352 {
7353 	unsigned long min_flush_time;
7354 	int flush_wait_retry = 50;
7355 	bool disable_atr = false;
7356 	int fd_room;
7357 	int reg;
7358 
7359 	if (!time_after(jiffies, pf->fd_flush_timestamp +
7360 				 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
7361 		return;
7362 
7363 	/* If the flush is happening too quick and we have mostly SB rules we
7364 	 * should not re-enable ATR for some time.
7365 	 */
7366 	min_flush_time = pf->fd_flush_timestamp +
7367 			 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
7368 	fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
7369 
7370 	if (!(time_after(jiffies, min_flush_time)) &&
7371 	    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
7372 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
7373 			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
7374 		disable_atr = true;
7375 	}
7376 
7377 	pf->fd_flush_timestamp = jiffies;
7378 	pf->flags |= I40E_FLAG_FD_ATR_AUTO_DISABLED;
7379 	/* flush all filters */
7380 	wr32(&pf->hw, I40E_PFQF_CTL_1,
7381 	     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
7382 	i40e_flush(&pf->hw);
7383 	pf->fd_flush_cnt++;
7384 	pf->fd_add_err = 0;
7385 	do {
7386 		/* Check FD flush status every 5-6msec */
7387 		usleep_range(5000, 6000);
7388 		reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
7389 		if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
7390 			break;
7391 	} while (flush_wait_retry--);
7392 	if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
7393 		dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
7394 	} else {
7395 		/* replay sideband filters */
7396 		i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
7397 		if (!disable_atr && !pf->fd_tcp4_filter_cnt)
7398 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
7399 		clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
7400 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
7401 			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
7402 	}
7403 }
7404 
7405 /**
7406  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
7407  * @pf: board private structure
7408  **/
7409 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
7410 {
7411 	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
7412 }
7413 
7414 /* We can see up to 256 filter programming desc in transit if the filters are
7415  * being applied really fast; before we see the first
7416  * filter miss error on Rx queue 0. Accumulating enough error messages before
7417  * reacting will make sure we don't cause flush too often.
7418  */
7419 #define I40E_MAX_FD_PROGRAM_ERROR 256
7420 
7421 /**
7422  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
7423  * @pf: board private structure
7424  **/
7425 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
7426 {
7427 
7428 	/* if interface is down do nothing */
7429 	if (test_bit(__I40E_DOWN, pf->state))
7430 		return;
7431 
7432 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
7433 		i40e_fdir_flush_and_replay(pf);
7434 
7435 	i40e_fdir_check_and_reenable(pf);
7436 
7437 }
7438 
7439 /**
7440  * i40e_vsi_link_event - notify VSI of a link event
7441  * @vsi: vsi to be notified
7442  * @link_up: link up or down
7443  **/
7444 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
7445 {
7446 	if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
7447 		return;
7448 
7449 	switch (vsi->type) {
7450 	case I40E_VSI_MAIN:
7451 		if (!vsi->netdev || !vsi->netdev_registered)
7452 			break;
7453 
7454 		if (link_up) {
7455 			netif_carrier_on(vsi->netdev);
7456 			netif_tx_wake_all_queues(vsi->netdev);
7457 		} else {
7458 			netif_carrier_off(vsi->netdev);
7459 			netif_tx_stop_all_queues(vsi->netdev);
7460 		}
7461 		break;
7462 
7463 	case I40E_VSI_SRIOV:
7464 	case I40E_VSI_VMDQ2:
7465 	case I40E_VSI_CTRL:
7466 	case I40E_VSI_IWARP:
7467 	case I40E_VSI_MIRROR:
7468 	default:
7469 		/* there is no notification for other VSIs */
7470 		break;
7471 	}
7472 }
7473 
7474 /**
7475  * i40e_veb_link_event - notify elements on the veb of a link event
7476  * @veb: veb to be notified
7477  * @link_up: link up or down
7478  **/
7479 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
7480 {
7481 	struct i40e_pf *pf;
7482 	int i;
7483 
7484 	if (!veb || !veb->pf)
7485 		return;
7486 	pf = veb->pf;
7487 
7488 	/* depth first... */
7489 	for (i = 0; i < I40E_MAX_VEB; i++)
7490 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
7491 			i40e_veb_link_event(pf->veb[i], link_up);
7492 
7493 	/* ... now the local VSIs */
7494 	for (i = 0; i < pf->num_alloc_vsi; i++)
7495 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
7496 			i40e_vsi_link_event(pf->vsi[i], link_up);
7497 }
7498 
7499 /**
7500  * i40e_link_event - Update netif_carrier status
7501  * @pf: board private structure
7502  **/
7503 static void i40e_link_event(struct i40e_pf *pf)
7504 {
7505 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7506 	u8 new_link_speed, old_link_speed;
7507 	i40e_status status;
7508 	bool new_link, old_link;
7509 
7510 	/* save off old link status information */
7511 	pf->hw.phy.link_info_old = pf->hw.phy.link_info;
7512 
7513 	/* set this to force the get_link_status call to refresh state */
7514 	pf->hw.phy.get_link_info = true;
7515 
7516 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
7517 
7518 	status = i40e_get_link_status(&pf->hw, &new_link);
7519 
7520 	/* On success, disable temp link polling */
7521 	if (status == I40E_SUCCESS) {
7522 		if (pf->flags & I40E_FLAG_TEMP_LINK_POLLING)
7523 			pf->flags &= ~I40E_FLAG_TEMP_LINK_POLLING;
7524 	} else {
7525 		/* Enable link polling temporarily until i40e_get_link_status
7526 		 * returns I40E_SUCCESS
7527 		 */
7528 		pf->flags |= I40E_FLAG_TEMP_LINK_POLLING;
7529 		dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
7530 			status);
7531 		return;
7532 	}
7533 
7534 	old_link_speed = pf->hw.phy.link_info_old.link_speed;
7535 	new_link_speed = pf->hw.phy.link_info.link_speed;
7536 
7537 	if (new_link == old_link &&
7538 	    new_link_speed == old_link_speed &&
7539 	    (test_bit(__I40E_VSI_DOWN, vsi->state) ||
7540 	     new_link == netif_carrier_ok(vsi->netdev)))
7541 		return;
7542 
7543 	i40e_print_link_message(vsi, new_link);
7544 
7545 	/* Notify the base of the switch tree connected to
7546 	 * the link.  Floating VEBs are not notified.
7547 	 */
7548 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
7549 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
7550 	else
7551 		i40e_vsi_link_event(vsi, new_link);
7552 
7553 	if (pf->vf)
7554 		i40e_vc_notify_link_state(pf);
7555 
7556 	if (pf->flags & I40E_FLAG_PTP)
7557 		i40e_ptp_set_increment(pf);
7558 }
7559 
7560 /**
7561  * i40e_watchdog_subtask - periodic checks not using event driven response
7562  * @pf: board private structure
7563  **/
7564 static void i40e_watchdog_subtask(struct i40e_pf *pf)
7565 {
7566 	int i;
7567 
7568 	/* if interface is down do nothing */
7569 	if (test_bit(__I40E_DOWN, pf->state) ||
7570 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
7571 		return;
7572 
7573 	/* make sure we don't do these things too often */
7574 	if (time_before(jiffies, (pf->service_timer_previous +
7575 				  pf->service_timer_period)))
7576 		return;
7577 	pf->service_timer_previous = jiffies;
7578 
7579 	if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
7580 	    (pf->flags & I40E_FLAG_TEMP_LINK_POLLING))
7581 		i40e_link_event(pf);
7582 
7583 	/* Update the stats for active netdevs so the network stack
7584 	 * can look at updated numbers whenever it cares to
7585 	 */
7586 	for (i = 0; i < pf->num_alloc_vsi; i++)
7587 		if (pf->vsi[i] && pf->vsi[i]->netdev)
7588 			i40e_update_stats(pf->vsi[i]);
7589 
7590 	if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
7591 		/* Update the stats for the active switching components */
7592 		for (i = 0; i < I40E_MAX_VEB; i++)
7593 			if (pf->veb[i])
7594 				i40e_update_veb_stats(pf->veb[i]);
7595 	}
7596 
7597 	i40e_ptp_rx_hang(pf);
7598 	i40e_ptp_tx_hang(pf);
7599 }
7600 
7601 /**
7602  * i40e_reset_subtask - Set up for resetting the device and driver
7603  * @pf: board private structure
7604  **/
7605 static void i40e_reset_subtask(struct i40e_pf *pf)
7606 {
7607 	u32 reset_flags = 0;
7608 
7609 	if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
7610 		reset_flags |= BIT(__I40E_REINIT_REQUESTED);
7611 		clear_bit(__I40E_REINIT_REQUESTED, pf->state);
7612 	}
7613 	if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
7614 		reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
7615 		clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
7616 	}
7617 	if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
7618 		reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
7619 		clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
7620 	}
7621 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
7622 		reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
7623 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
7624 	}
7625 	if (test_bit(__I40E_DOWN_REQUESTED, pf->state)) {
7626 		reset_flags |= BIT(__I40E_DOWN_REQUESTED);
7627 		clear_bit(__I40E_DOWN_REQUESTED, pf->state);
7628 	}
7629 
7630 	/* If there's a recovery already waiting, it takes
7631 	 * precedence before starting a new reset sequence.
7632 	 */
7633 	if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
7634 		i40e_prep_for_reset(pf, false);
7635 		i40e_reset(pf);
7636 		i40e_rebuild(pf, false, false);
7637 	}
7638 
7639 	/* If we're already down or resetting, just bail */
7640 	if (reset_flags &&
7641 	    !test_bit(__I40E_DOWN, pf->state) &&
7642 	    !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
7643 		i40e_do_reset(pf, reset_flags, false);
7644 	}
7645 }
7646 
7647 /**
7648  * i40e_handle_link_event - Handle link event
7649  * @pf: board private structure
7650  * @e: event info posted on ARQ
7651  **/
7652 static void i40e_handle_link_event(struct i40e_pf *pf,
7653 				   struct i40e_arq_event_info *e)
7654 {
7655 	struct i40e_aqc_get_link_status *status =
7656 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
7657 
7658 	/* Do a new status request to re-enable LSE reporting
7659 	 * and load new status information into the hw struct
7660 	 * This completely ignores any state information
7661 	 * in the ARQ event info, instead choosing to always
7662 	 * issue the AQ update link status command.
7663 	 */
7664 	i40e_link_event(pf);
7665 
7666 	/* Check if module meets thermal requirements */
7667 	if (status->phy_type == I40E_PHY_TYPE_NOT_SUPPORTED_HIGH_TEMP) {
7668 		dev_err(&pf->pdev->dev,
7669 			"Rx/Tx is disabled on this device because the module does not meet thermal requirements.\n");
7670 		dev_err(&pf->pdev->dev,
7671 			"Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
7672 	} else {
7673 		/* check for unqualified module, if link is down, suppress
7674 		 * the message if link was forced to be down.
7675 		 */
7676 		if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
7677 		    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
7678 		    (!(status->link_info & I40E_AQ_LINK_UP)) &&
7679 		    (!(pf->flags & I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED))) {
7680 			dev_err(&pf->pdev->dev,
7681 				"Rx/Tx is disabled on this device because an unsupported SFP module type was detected.\n");
7682 			dev_err(&pf->pdev->dev,
7683 				"Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
7684 		}
7685 	}
7686 }
7687 
7688 /**
7689  * i40e_clean_adminq_subtask - Clean the AdminQ rings
7690  * @pf: board private structure
7691  **/
7692 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
7693 {
7694 	struct i40e_arq_event_info event;
7695 	struct i40e_hw *hw = &pf->hw;
7696 	u16 pending, i = 0;
7697 	i40e_status ret;
7698 	u16 opcode;
7699 	u32 oldval;
7700 	u32 val;
7701 
7702 	/* Do not run clean AQ when PF reset fails */
7703 	if (test_bit(__I40E_RESET_FAILED, pf->state))
7704 		return;
7705 
7706 	/* check for error indications */
7707 	val = rd32(&pf->hw, pf->hw.aq.arq.len);
7708 	oldval = val;
7709 	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
7710 		if (hw->debug_mask & I40E_DEBUG_AQ)
7711 			dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
7712 		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
7713 	}
7714 	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
7715 		if (hw->debug_mask & I40E_DEBUG_AQ)
7716 			dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
7717 		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
7718 		pf->arq_overflows++;
7719 	}
7720 	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
7721 		if (hw->debug_mask & I40E_DEBUG_AQ)
7722 			dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
7723 		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
7724 	}
7725 	if (oldval != val)
7726 		wr32(&pf->hw, pf->hw.aq.arq.len, val);
7727 
7728 	val = rd32(&pf->hw, pf->hw.aq.asq.len);
7729 	oldval = val;
7730 	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
7731 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
7732 			dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
7733 		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
7734 	}
7735 	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
7736 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
7737 			dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
7738 		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
7739 	}
7740 	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
7741 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
7742 			dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
7743 		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
7744 	}
7745 	if (oldval != val)
7746 		wr32(&pf->hw, pf->hw.aq.asq.len, val);
7747 
7748 	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
7749 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
7750 	if (!event.msg_buf)
7751 		return;
7752 
7753 	do {
7754 		ret = i40e_clean_arq_element(hw, &event, &pending);
7755 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
7756 			break;
7757 		else if (ret) {
7758 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
7759 			break;
7760 		}
7761 
7762 		opcode = le16_to_cpu(event.desc.opcode);
7763 		switch (opcode) {
7764 
7765 		case i40e_aqc_opc_get_link_status:
7766 			i40e_handle_link_event(pf, &event);
7767 			break;
7768 		case i40e_aqc_opc_send_msg_to_pf:
7769 			ret = i40e_vc_process_vf_msg(pf,
7770 					le16_to_cpu(event.desc.retval),
7771 					le32_to_cpu(event.desc.cookie_high),
7772 					le32_to_cpu(event.desc.cookie_low),
7773 					event.msg_buf,
7774 					event.msg_len);
7775 			break;
7776 		case i40e_aqc_opc_lldp_update_mib:
7777 			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
7778 #ifdef CONFIG_I40E_DCB
7779 			rtnl_lock();
7780 			ret = i40e_handle_lldp_event(pf, &event);
7781 			rtnl_unlock();
7782 #endif /* CONFIG_I40E_DCB */
7783 			break;
7784 		case i40e_aqc_opc_event_lan_overflow:
7785 			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
7786 			i40e_handle_lan_overflow_event(pf, &event);
7787 			break;
7788 		case i40e_aqc_opc_send_msg_to_peer:
7789 			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
7790 			break;
7791 		case i40e_aqc_opc_nvm_erase:
7792 		case i40e_aqc_opc_nvm_update:
7793 		case i40e_aqc_opc_oem_post_update:
7794 			i40e_debug(&pf->hw, I40E_DEBUG_NVM,
7795 				   "ARQ NVM operation 0x%04x completed\n",
7796 				   opcode);
7797 			break;
7798 		default:
7799 			dev_info(&pf->pdev->dev,
7800 				 "ARQ: Unknown event 0x%04x ignored\n",
7801 				 opcode);
7802 			break;
7803 		}
7804 	} while (i++ < pf->adminq_work_limit);
7805 
7806 	if (i < pf->adminq_work_limit)
7807 		clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
7808 
7809 	/* re-enable Admin queue interrupt cause */
7810 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
7811 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
7812 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
7813 	i40e_flush(hw);
7814 
7815 	kfree(event.msg_buf);
7816 }
7817 
7818 /**
7819  * i40e_verify_eeprom - make sure eeprom is good to use
7820  * @pf: board private structure
7821  **/
7822 static void i40e_verify_eeprom(struct i40e_pf *pf)
7823 {
7824 	int err;
7825 
7826 	err = i40e_diag_eeprom_test(&pf->hw);
7827 	if (err) {
7828 		/* retry in case of garbage read */
7829 		err = i40e_diag_eeprom_test(&pf->hw);
7830 		if (err) {
7831 			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
7832 				 err);
7833 			set_bit(__I40E_BAD_EEPROM, pf->state);
7834 		}
7835 	}
7836 
7837 	if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
7838 		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
7839 		clear_bit(__I40E_BAD_EEPROM, pf->state);
7840 	}
7841 }
7842 
7843 /**
7844  * i40e_enable_pf_switch_lb
7845  * @pf: pointer to the PF structure
7846  *
7847  * enable switch loop back or die - no point in a return value
7848  **/
7849 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
7850 {
7851 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7852 	struct i40e_vsi_context ctxt;
7853 	int ret;
7854 
7855 	ctxt.seid = pf->main_vsi_seid;
7856 	ctxt.pf_num = pf->hw.pf_id;
7857 	ctxt.vf_num = 0;
7858 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
7859 	if (ret) {
7860 		dev_info(&pf->pdev->dev,
7861 			 "couldn't get PF vsi config, err %s aq_err %s\n",
7862 			 i40e_stat_str(&pf->hw, ret),
7863 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7864 		return;
7865 	}
7866 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
7867 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
7868 	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
7869 
7870 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
7871 	if (ret) {
7872 		dev_info(&pf->pdev->dev,
7873 			 "update vsi switch failed, err %s aq_err %s\n",
7874 			 i40e_stat_str(&pf->hw, ret),
7875 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7876 	}
7877 }
7878 
7879 /**
7880  * i40e_disable_pf_switch_lb
7881  * @pf: pointer to the PF structure
7882  *
7883  * disable switch loop back or die - no point in a return value
7884  **/
7885 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
7886 {
7887 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7888 	struct i40e_vsi_context ctxt;
7889 	int ret;
7890 
7891 	ctxt.seid = pf->main_vsi_seid;
7892 	ctxt.pf_num = pf->hw.pf_id;
7893 	ctxt.vf_num = 0;
7894 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
7895 	if (ret) {
7896 		dev_info(&pf->pdev->dev,
7897 			 "couldn't get PF vsi config, err %s aq_err %s\n",
7898 			 i40e_stat_str(&pf->hw, ret),
7899 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7900 		return;
7901 	}
7902 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
7903 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
7904 	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
7905 
7906 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
7907 	if (ret) {
7908 		dev_info(&pf->pdev->dev,
7909 			 "update vsi switch failed, err %s aq_err %s\n",
7910 			 i40e_stat_str(&pf->hw, ret),
7911 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7912 	}
7913 }
7914 
7915 /**
7916  * i40e_config_bridge_mode - Configure the HW bridge mode
7917  * @veb: pointer to the bridge instance
7918  *
7919  * Configure the loop back mode for the LAN VSI that is downlink to the
7920  * specified HW bridge instance. It is expected this function is called
7921  * when a new HW bridge is instantiated.
7922  **/
7923 static void i40e_config_bridge_mode(struct i40e_veb *veb)
7924 {
7925 	struct i40e_pf *pf = veb->pf;
7926 
7927 	if (pf->hw.debug_mask & I40E_DEBUG_LAN)
7928 		dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
7929 			 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
7930 	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
7931 		i40e_disable_pf_switch_lb(pf);
7932 	else
7933 		i40e_enable_pf_switch_lb(pf);
7934 }
7935 
7936 /**
7937  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
7938  * @veb: pointer to the VEB instance
7939  *
7940  * This is a recursive function that first builds the attached VSIs then
7941  * recurses in to build the next layer of VEB.  We track the connections
7942  * through our own index numbers because the seid's from the HW could
7943  * change across the reset.
7944  **/
7945 static int i40e_reconstitute_veb(struct i40e_veb *veb)
7946 {
7947 	struct i40e_vsi *ctl_vsi = NULL;
7948 	struct i40e_pf *pf = veb->pf;
7949 	int v, veb_idx;
7950 	int ret;
7951 
7952 	/* build VSI that owns this VEB, temporarily attached to base VEB */
7953 	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
7954 		if (pf->vsi[v] &&
7955 		    pf->vsi[v]->veb_idx == veb->idx &&
7956 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
7957 			ctl_vsi = pf->vsi[v];
7958 			break;
7959 		}
7960 	}
7961 	if (!ctl_vsi) {
7962 		dev_info(&pf->pdev->dev,
7963 			 "missing owner VSI for veb_idx %d\n", veb->idx);
7964 		ret = -ENOENT;
7965 		goto end_reconstitute;
7966 	}
7967 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
7968 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
7969 	ret = i40e_add_vsi(ctl_vsi);
7970 	if (ret) {
7971 		dev_info(&pf->pdev->dev,
7972 			 "rebuild of veb_idx %d owner VSI failed: %d\n",
7973 			 veb->idx, ret);
7974 		goto end_reconstitute;
7975 	}
7976 	i40e_vsi_reset_stats(ctl_vsi);
7977 
7978 	/* create the VEB in the switch and move the VSI onto the VEB */
7979 	ret = i40e_add_veb(veb, ctl_vsi);
7980 	if (ret)
7981 		goto end_reconstitute;
7982 
7983 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
7984 		veb->bridge_mode = BRIDGE_MODE_VEB;
7985 	else
7986 		veb->bridge_mode = BRIDGE_MODE_VEPA;
7987 	i40e_config_bridge_mode(veb);
7988 
7989 	/* create the remaining VSIs attached to this VEB */
7990 	for (v = 0; v < pf->num_alloc_vsi; v++) {
7991 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
7992 			continue;
7993 
7994 		if (pf->vsi[v]->veb_idx == veb->idx) {
7995 			struct i40e_vsi *vsi = pf->vsi[v];
7996 
7997 			vsi->uplink_seid = veb->seid;
7998 			ret = i40e_add_vsi(vsi);
7999 			if (ret) {
8000 				dev_info(&pf->pdev->dev,
8001 					 "rebuild of vsi_idx %d failed: %d\n",
8002 					 v, ret);
8003 				goto end_reconstitute;
8004 			}
8005 			i40e_vsi_reset_stats(vsi);
8006 		}
8007 	}
8008 
8009 	/* create any VEBs attached to this VEB - RECURSION */
8010 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
8011 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
8012 			pf->veb[veb_idx]->uplink_seid = veb->seid;
8013 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
8014 			if (ret)
8015 				break;
8016 		}
8017 	}
8018 
8019 end_reconstitute:
8020 	return ret;
8021 }
8022 
8023 /**
8024  * i40e_get_capabilities - get info about the HW
8025  * @pf: the PF struct
8026  **/
8027 static int i40e_get_capabilities(struct i40e_pf *pf)
8028 {
8029 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
8030 	u16 data_size;
8031 	int buf_len;
8032 	int err;
8033 
8034 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
8035 	do {
8036 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
8037 		if (!cap_buf)
8038 			return -ENOMEM;
8039 
8040 		/* this loads the data into the hw struct for us */
8041 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
8042 					    &data_size,
8043 					    i40e_aqc_opc_list_func_capabilities,
8044 					    NULL);
8045 		/* data loaded, buffer no longer needed */
8046 		kfree(cap_buf);
8047 
8048 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
8049 			/* retry with a larger buffer */
8050 			buf_len = data_size;
8051 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
8052 			dev_info(&pf->pdev->dev,
8053 				 "capability discovery failed, err %s aq_err %s\n",
8054 				 i40e_stat_str(&pf->hw, err),
8055 				 i40e_aq_str(&pf->hw,
8056 					     pf->hw.aq.asq_last_status));
8057 			return -ENODEV;
8058 		}
8059 	} while (err);
8060 
8061 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
8062 		dev_info(&pf->pdev->dev,
8063 			 "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",
8064 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
8065 			 pf->hw.func_caps.num_msix_vectors,
8066 			 pf->hw.func_caps.num_msix_vectors_vf,
8067 			 pf->hw.func_caps.fd_filters_guaranteed,
8068 			 pf->hw.func_caps.fd_filters_best_effort,
8069 			 pf->hw.func_caps.num_tx_qp,
8070 			 pf->hw.func_caps.num_vsis);
8071 
8072 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
8073 		       + pf->hw.func_caps.num_vfs)
8074 	if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
8075 		dev_info(&pf->pdev->dev,
8076 			 "got num_vsis %d, setting num_vsis to %d\n",
8077 			 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
8078 		pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
8079 	}
8080 
8081 	return 0;
8082 }
8083 
8084 static int i40e_vsi_clear(struct i40e_vsi *vsi);
8085 
8086 /**
8087  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
8088  * @pf: board private structure
8089  **/
8090 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
8091 {
8092 	struct i40e_vsi *vsi;
8093 
8094 	/* quick workaround for an NVM issue that leaves a critical register
8095 	 * uninitialized
8096 	 */
8097 	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
8098 		static const u32 hkey[] = {
8099 			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
8100 			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
8101 			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
8102 			0x95b3a76d};
8103 		int i;
8104 
8105 		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
8106 			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
8107 	}
8108 
8109 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8110 		return;
8111 
8112 	/* find existing VSI and see if it needs configuring */
8113 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
8114 
8115 	/* create a new VSI if none exists */
8116 	if (!vsi) {
8117 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
8118 				     pf->vsi[pf->lan_vsi]->seid, 0);
8119 		if (!vsi) {
8120 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
8121 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8122 			return;
8123 		}
8124 	}
8125 
8126 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
8127 }
8128 
8129 /**
8130  * i40e_fdir_teardown - release the Flow Director resources
8131  * @pf: board private structure
8132  **/
8133 static void i40e_fdir_teardown(struct i40e_pf *pf)
8134 {
8135 	struct i40e_vsi *vsi;
8136 
8137 	i40e_fdir_filter_exit(pf);
8138 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
8139 	if (vsi)
8140 		i40e_vsi_release(vsi);
8141 }
8142 
8143 /**
8144  * i40e_rebuild_channels - Rebuilds channel VSIs if they existed before reset
8145  * @vsi: PF main vsi
8146  *
8147  * Rebuilds channel VSIs if they existed before reset
8148  **/
8149 static int i40e_rebuild_channels(struct i40e_vsi *vsi)
8150 {
8151 	struct i40e_channel *ch, *ch_tmp;
8152 	i40e_status ret;
8153 
8154 	if (list_empty(&vsi->ch_list))
8155 		return 0;
8156 
8157 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8158 		if (!ch->initialized)
8159 			break;
8160 		/* Proceed with creation of channel (VMDq2) VSI */
8161 		ret = i40e_add_channel(vsi->back, vsi->uplink_seid, ch);
8162 		if (ret) {
8163 			dev_info(&vsi->back->pdev->dev,
8164 				 "failed to rebuild channels using uplink_seid %u\n",
8165 				 vsi->uplink_seid);
8166 			return ret;
8167 		}
8168 		if (ch->max_tx_rate) {
8169 			if (i40e_set_bw_limit(vsi, ch->seid,
8170 					      ch->max_tx_rate))
8171 				return -EINVAL;
8172 
8173 			dev_dbg(&vsi->back->pdev->dev,
8174 				"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
8175 				ch->max_tx_rate,
8176 				ch->max_tx_rate / I40E_BW_CREDIT_DIVISOR,
8177 				ch->seid);
8178 		}
8179 	}
8180 	return 0;
8181 }
8182 
8183 /**
8184  * i40e_prep_for_reset - prep for the core to reset
8185  * @pf: board private structure
8186  * @lock_acquired: indicates whether or not the lock has been acquired
8187  * before this function was called.
8188  *
8189  * Close up the VFs and other things in prep for PF Reset.
8190   **/
8191 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
8192 {
8193 	struct i40e_hw *hw = &pf->hw;
8194 	i40e_status ret = 0;
8195 	u32 v;
8196 
8197 	clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
8198 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
8199 		return;
8200 	if (i40e_check_asq_alive(&pf->hw))
8201 		i40e_vc_notify_reset(pf);
8202 
8203 	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
8204 
8205 	/* quiesce the VSIs and their queues that are not already DOWN */
8206 	/* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
8207 	if (!lock_acquired)
8208 		rtnl_lock();
8209 	i40e_pf_quiesce_all_vsi(pf);
8210 	if (!lock_acquired)
8211 		rtnl_unlock();
8212 
8213 	for (v = 0; v < pf->num_alloc_vsi; v++) {
8214 		if (pf->vsi[v])
8215 			pf->vsi[v]->seid = 0;
8216 	}
8217 
8218 	i40e_shutdown_adminq(&pf->hw);
8219 
8220 	/* call shutdown HMC */
8221 	if (hw->hmc.hmc_obj) {
8222 		ret = i40e_shutdown_lan_hmc(hw);
8223 		if (ret)
8224 			dev_warn(&pf->pdev->dev,
8225 				 "shutdown_lan_hmc failed: %d\n", ret);
8226 	}
8227 }
8228 
8229 /**
8230  * i40e_send_version - update firmware with driver version
8231  * @pf: PF struct
8232  */
8233 static void i40e_send_version(struct i40e_pf *pf)
8234 {
8235 	struct i40e_driver_version dv;
8236 
8237 	dv.major_version = DRV_VERSION_MAJOR;
8238 	dv.minor_version = DRV_VERSION_MINOR;
8239 	dv.build_version = DRV_VERSION_BUILD;
8240 	dv.subbuild_version = 0;
8241 	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
8242 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
8243 }
8244 
8245 /**
8246  * i40e_get_oem_version - get OEM specific version information
8247  * @hw: pointer to the hardware structure
8248  **/
8249 static void i40e_get_oem_version(struct i40e_hw *hw)
8250 {
8251 	u16 block_offset = 0xffff;
8252 	u16 block_length = 0;
8253 	u16 capabilities = 0;
8254 	u16 gen_snap = 0;
8255 	u16 release = 0;
8256 
8257 #define I40E_SR_NVM_OEM_VERSION_PTR		0x1B
8258 #define I40E_NVM_OEM_LENGTH_OFFSET		0x00
8259 #define I40E_NVM_OEM_CAPABILITIES_OFFSET	0x01
8260 #define I40E_NVM_OEM_GEN_OFFSET			0x02
8261 #define I40E_NVM_OEM_RELEASE_OFFSET		0x03
8262 #define I40E_NVM_OEM_CAPABILITIES_MASK		0x000F
8263 #define I40E_NVM_OEM_LENGTH			3
8264 
8265 	/* Check if pointer to OEM version block is valid. */
8266 	i40e_read_nvm_word(hw, I40E_SR_NVM_OEM_VERSION_PTR, &block_offset);
8267 	if (block_offset == 0xffff)
8268 		return;
8269 
8270 	/* Check if OEM version block has correct length. */
8271 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_LENGTH_OFFSET,
8272 			   &block_length);
8273 	if (block_length < I40E_NVM_OEM_LENGTH)
8274 		return;
8275 
8276 	/* Check if OEM version format is as expected. */
8277 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_CAPABILITIES_OFFSET,
8278 			   &capabilities);
8279 	if ((capabilities & I40E_NVM_OEM_CAPABILITIES_MASK) != 0)
8280 		return;
8281 
8282 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_GEN_OFFSET,
8283 			   &gen_snap);
8284 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_RELEASE_OFFSET,
8285 			   &release);
8286 	hw->nvm.oem_ver = (gen_snap << I40E_OEM_SNAP_SHIFT) | release;
8287 	hw->nvm.eetrack = I40E_OEM_EETRACK_ID;
8288 }
8289 
8290 /**
8291  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
8292  * @pf: board private structure
8293  **/
8294 static int i40e_reset(struct i40e_pf *pf)
8295 {
8296 	struct i40e_hw *hw = &pf->hw;
8297 	i40e_status ret;
8298 
8299 	ret = i40e_pf_reset(hw);
8300 	if (ret) {
8301 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
8302 		set_bit(__I40E_RESET_FAILED, pf->state);
8303 		clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
8304 	} else {
8305 		pf->pfr_count++;
8306 	}
8307 	return ret;
8308 }
8309 
8310 /**
8311  * i40e_rebuild - rebuild using a saved config
8312  * @pf: board private structure
8313  * @reinit: if the Main VSI needs to re-initialized.
8314  * @lock_acquired: indicates whether or not the lock has been acquired
8315  * before this function was called.
8316  **/
8317 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
8318 {
8319 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8320 	struct i40e_hw *hw = &pf->hw;
8321 	u8 set_fc_aq_fail = 0;
8322 	i40e_status ret;
8323 	u32 val;
8324 	int v;
8325 
8326 	if (test_bit(__I40E_DOWN, pf->state))
8327 		goto clear_recovery;
8328 	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
8329 
8330 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
8331 	ret = i40e_init_adminq(&pf->hw);
8332 	if (ret) {
8333 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
8334 			 i40e_stat_str(&pf->hw, ret),
8335 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8336 		goto clear_recovery;
8337 	}
8338 	i40e_get_oem_version(&pf->hw);
8339 
8340 	/* re-verify the eeprom if we just had an EMP reset */
8341 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
8342 		i40e_verify_eeprom(pf);
8343 
8344 	i40e_clear_pxe_mode(hw);
8345 	ret = i40e_get_capabilities(pf);
8346 	if (ret)
8347 		goto end_core_reset;
8348 
8349 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
8350 				hw->func_caps.num_rx_qp, 0, 0);
8351 	if (ret) {
8352 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
8353 		goto end_core_reset;
8354 	}
8355 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
8356 	if (ret) {
8357 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
8358 		goto end_core_reset;
8359 	}
8360 
8361 #ifdef CONFIG_I40E_DCB
8362 	ret = i40e_init_pf_dcb(pf);
8363 	if (ret) {
8364 		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
8365 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
8366 		/* Continue without DCB enabled */
8367 	}
8368 #endif /* CONFIG_I40E_DCB */
8369 	/* do basic switch setup */
8370 	if (!lock_acquired)
8371 		rtnl_lock();
8372 	ret = i40e_setup_pf_switch(pf, reinit);
8373 	if (ret)
8374 		goto end_unlock;
8375 
8376 	/* The driver only wants link up/down and module qualification
8377 	 * reports from firmware.  Note the negative logic.
8378 	 */
8379 	ret = i40e_aq_set_phy_int_mask(&pf->hw,
8380 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
8381 					 I40E_AQ_EVENT_MEDIA_NA |
8382 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
8383 	if (ret)
8384 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
8385 			 i40e_stat_str(&pf->hw, ret),
8386 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8387 
8388 	/* make sure our flow control settings are restored */
8389 	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
8390 	if (ret)
8391 		dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
8392 			i40e_stat_str(&pf->hw, ret),
8393 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8394 
8395 	/* Rebuild the VSIs and VEBs that existed before reset.
8396 	 * They are still in our local switch element arrays, so only
8397 	 * need to rebuild the switch model in the HW.
8398 	 *
8399 	 * If there were VEBs but the reconstitution failed, we'll try
8400 	 * try to recover minimal use by getting the basic PF VSI working.
8401 	 */
8402 	if (vsi->uplink_seid != pf->mac_seid) {
8403 		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
8404 		/* find the one VEB connected to the MAC, and find orphans */
8405 		for (v = 0; v < I40E_MAX_VEB; v++) {
8406 			if (!pf->veb[v])
8407 				continue;
8408 
8409 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
8410 			    pf->veb[v]->uplink_seid == 0) {
8411 				ret = i40e_reconstitute_veb(pf->veb[v]);
8412 
8413 				if (!ret)
8414 					continue;
8415 
8416 				/* If Main VEB failed, we're in deep doodoo,
8417 				 * so give up rebuilding the switch and set up
8418 				 * for minimal rebuild of PF VSI.
8419 				 * If orphan failed, we'll report the error
8420 				 * but try to keep going.
8421 				 */
8422 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
8423 					dev_info(&pf->pdev->dev,
8424 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
8425 						 ret);
8426 					vsi->uplink_seid = pf->mac_seid;
8427 					break;
8428 				} else if (pf->veb[v]->uplink_seid == 0) {
8429 					dev_info(&pf->pdev->dev,
8430 						 "rebuild of orphan VEB failed: %d\n",
8431 						 ret);
8432 				}
8433 			}
8434 		}
8435 	}
8436 
8437 	if (vsi->uplink_seid == pf->mac_seid) {
8438 		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
8439 		/* no VEB, so rebuild only the Main VSI */
8440 		ret = i40e_add_vsi(vsi);
8441 		if (ret) {
8442 			dev_info(&pf->pdev->dev,
8443 				 "rebuild of Main VSI failed: %d\n", ret);
8444 			goto end_unlock;
8445 		}
8446 	}
8447 
8448 	if (vsi->mqprio_qopt.max_rate[0]) {
8449 		u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0] / (1000000 / 8);
8450 
8451 		ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
8452 		if (!ret)
8453 			dev_dbg(&vsi->back->pdev->dev,
8454 				"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
8455 				max_tx_rate,
8456 				max_tx_rate / I40E_BW_CREDIT_DIVISOR,
8457 				vsi->seid);
8458 		else
8459 			goto end_unlock;
8460 	}
8461 
8462 	/* PF Main VSI is rebuild by now, go ahead and rebuild channel VSIs
8463 	 * for this main VSI if they exist
8464 	 */
8465 	ret = i40e_rebuild_channels(vsi);
8466 	if (ret)
8467 		goto end_unlock;
8468 
8469 	/* Reconfigure hardware for allowing smaller MSS in the case
8470 	 * of TSO, so that we avoid the MDD being fired and causing
8471 	 * a reset in the case of small MSS+TSO.
8472 	 */
8473 #define I40E_REG_MSS          0x000E64DC
8474 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
8475 #define I40E_64BYTE_MSS       0x400000
8476 	val = rd32(hw, I40E_REG_MSS);
8477 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
8478 		val &= ~I40E_REG_MSS_MIN_MASK;
8479 		val |= I40E_64BYTE_MSS;
8480 		wr32(hw, I40E_REG_MSS, val);
8481 	}
8482 
8483 	if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
8484 		msleep(75);
8485 		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
8486 		if (ret)
8487 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
8488 				 i40e_stat_str(&pf->hw, ret),
8489 				 i40e_aq_str(&pf->hw,
8490 					     pf->hw.aq.asq_last_status));
8491 	}
8492 	/* reinit the misc interrupt */
8493 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
8494 		ret = i40e_setup_misc_vector(pf);
8495 
8496 	/* Add a filter to drop all Flow control frames from any VSI from being
8497 	 * transmitted. By doing so we stop a malicious VF from sending out
8498 	 * PAUSE or PFC frames and potentially controlling traffic for other
8499 	 * PF/VF VSIs.
8500 	 * The FW can still send Flow control frames if enabled.
8501 	 */
8502 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
8503 						       pf->main_vsi_seid);
8504 
8505 	/* restart the VSIs that were rebuilt and running before the reset */
8506 	i40e_pf_unquiesce_all_vsi(pf);
8507 
8508 	/* Release the RTNL lock before we start resetting VFs */
8509 	if (!lock_acquired)
8510 		rtnl_unlock();
8511 
8512 	i40e_reset_all_vfs(pf, true);
8513 
8514 	/* tell the firmware that we're starting */
8515 	i40e_send_version(pf);
8516 
8517 	/* We've already released the lock, so don't do it again */
8518 	goto end_core_reset;
8519 
8520 end_unlock:
8521 	if (!lock_acquired)
8522 		rtnl_unlock();
8523 end_core_reset:
8524 	clear_bit(__I40E_RESET_FAILED, pf->state);
8525 clear_recovery:
8526 	clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
8527 }
8528 
8529 /**
8530  * i40e_reset_and_rebuild - reset and rebuild using a saved config
8531  * @pf: board private structure
8532  * @reinit: if the Main VSI needs to re-initialized.
8533  * @lock_acquired: indicates whether or not the lock has been acquired
8534  * before this function was called.
8535  **/
8536 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
8537 				   bool lock_acquired)
8538 {
8539 	int ret;
8540 	/* Now we wait for GRST to settle out.
8541 	 * We don't have to delete the VEBs or VSIs from the hw switch
8542 	 * because the reset will make them disappear.
8543 	 */
8544 	ret = i40e_reset(pf);
8545 	if (!ret)
8546 		i40e_rebuild(pf, reinit, lock_acquired);
8547 }
8548 
8549 /**
8550  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
8551  * @pf: board private structure
8552  *
8553  * Close up the VFs and other things in prep for a Core Reset,
8554  * then get ready to rebuild the world.
8555  * @lock_acquired: indicates whether or not the lock has been acquired
8556  * before this function was called.
8557  **/
8558 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
8559 {
8560 	i40e_prep_for_reset(pf, lock_acquired);
8561 	i40e_reset_and_rebuild(pf, false, lock_acquired);
8562 }
8563 
8564 /**
8565  * i40e_handle_mdd_event
8566  * @pf: pointer to the PF structure
8567  *
8568  * Called from the MDD irq handler to identify possibly malicious vfs
8569  **/
8570 static void i40e_handle_mdd_event(struct i40e_pf *pf)
8571 {
8572 	struct i40e_hw *hw = &pf->hw;
8573 	bool mdd_detected = false;
8574 	bool pf_mdd_detected = false;
8575 	struct i40e_vf *vf;
8576 	u32 reg;
8577 	int i;
8578 
8579 	if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
8580 		return;
8581 
8582 	/* find what triggered the MDD event */
8583 	reg = rd32(hw, I40E_GL_MDET_TX);
8584 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
8585 		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
8586 				I40E_GL_MDET_TX_PF_NUM_SHIFT;
8587 		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
8588 				I40E_GL_MDET_TX_VF_NUM_SHIFT;
8589 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
8590 				I40E_GL_MDET_TX_EVENT_SHIFT;
8591 		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
8592 				I40E_GL_MDET_TX_QUEUE_SHIFT) -
8593 				pf->hw.func_caps.base_queue;
8594 		if (netif_msg_tx_err(pf))
8595 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
8596 				 event, queue, pf_num, vf_num);
8597 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
8598 		mdd_detected = true;
8599 	}
8600 	reg = rd32(hw, I40E_GL_MDET_RX);
8601 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
8602 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
8603 				I40E_GL_MDET_RX_FUNCTION_SHIFT;
8604 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
8605 				I40E_GL_MDET_RX_EVENT_SHIFT;
8606 		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
8607 				I40E_GL_MDET_RX_QUEUE_SHIFT) -
8608 				pf->hw.func_caps.base_queue;
8609 		if (netif_msg_rx_err(pf))
8610 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
8611 				 event, queue, func);
8612 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
8613 		mdd_detected = true;
8614 	}
8615 
8616 	if (mdd_detected) {
8617 		reg = rd32(hw, I40E_PF_MDET_TX);
8618 		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
8619 			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
8620 			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
8621 			pf_mdd_detected = true;
8622 		}
8623 		reg = rd32(hw, I40E_PF_MDET_RX);
8624 		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
8625 			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
8626 			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
8627 			pf_mdd_detected = true;
8628 		}
8629 		/* Queue belongs to the PF, initiate a reset */
8630 		if (pf_mdd_detected) {
8631 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
8632 			i40e_service_event_schedule(pf);
8633 		}
8634 	}
8635 
8636 	/* see if one of the VFs needs its hand slapped */
8637 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
8638 		vf = &(pf->vf[i]);
8639 		reg = rd32(hw, I40E_VP_MDET_TX(i));
8640 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
8641 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
8642 			vf->num_mdd_events++;
8643 			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
8644 				 i);
8645 		}
8646 
8647 		reg = rd32(hw, I40E_VP_MDET_RX(i));
8648 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
8649 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
8650 			vf->num_mdd_events++;
8651 			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
8652 				 i);
8653 		}
8654 
8655 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
8656 			dev_info(&pf->pdev->dev,
8657 				 "Too many MDD events on VF %d, disabled\n", i);
8658 			dev_info(&pf->pdev->dev,
8659 				 "Use PF Control I/F to re-enable the VF\n");
8660 			set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
8661 		}
8662 	}
8663 
8664 	/* re-enable mdd interrupt cause */
8665 	clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
8666 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
8667 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
8668 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
8669 	i40e_flush(hw);
8670 }
8671 
8672 static const char *i40e_tunnel_name(struct i40e_udp_port_config *port)
8673 {
8674 	switch (port->type) {
8675 	case UDP_TUNNEL_TYPE_VXLAN:
8676 		return "vxlan";
8677 	case UDP_TUNNEL_TYPE_GENEVE:
8678 		return "geneve";
8679 	default:
8680 		return "unknown";
8681 	}
8682 }
8683 
8684 /**
8685  * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
8686  * @pf: board private structure
8687  **/
8688 static void i40e_sync_udp_filters(struct i40e_pf *pf)
8689 {
8690 	int i;
8691 
8692 	/* loop through and set pending bit for all active UDP filters */
8693 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8694 		if (pf->udp_ports[i].port)
8695 			pf->pending_udp_bitmap |= BIT_ULL(i);
8696 	}
8697 
8698 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8699 }
8700 
8701 /**
8702  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
8703  * @pf: board private structure
8704  **/
8705 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
8706 {
8707 	struct i40e_hw *hw = &pf->hw;
8708 	i40e_status ret;
8709 	u16 port;
8710 	int i;
8711 
8712 	if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
8713 		return;
8714 
8715 	pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
8716 
8717 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8718 		if (pf->pending_udp_bitmap & BIT_ULL(i)) {
8719 			pf->pending_udp_bitmap &= ~BIT_ULL(i);
8720 			port = pf->udp_ports[i].port;
8721 			if (port)
8722 				ret = i40e_aq_add_udp_tunnel(hw, port,
8723 							pf->udp_ports[i].type,
8724 							NULL, NULL);
8725 			else
8726 				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
8727 
8728 			if (ret) {
8729 				dev_info(&pf->pdev->dev,
8730 					 "%s %s port %d, index %d failed, err %s aq_err %s\n",
8731 					 i40e_tunnel_name(&pf->udp_ports[i]),
8732 					 port ? "add" : "delete",
8733 					 port, i,
8734 					 i40e_stat_str(&pf->hw, ret),
8735 					 i40e_aq_str(&pf->hw,
8736 						     pf->hw.aq.asq_last_status));
8737 				pf->udp_ports[i].port = 0;
8738 			}
8739 		}
8740 	}
8741 }
8742 
8743 /**
8744  * i40e_service_task - Run the driver's async subtasks
8745  * @work: pointer to work_struct containing our data
8746  **/
8747 static void i40e_service_task(struct work_struct *work)
8748 {
8749 	struct i40e_pf *pf = container_of(work,
8750 					  struct i40e_pf,
8751 					  service_task);
8752 	unsigned long start_time = jiffies;
8753 
8754 	/* don't bother with service tasks if a reset is in progress */
8755 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
8756 		return;
8757 
8758 	if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
8759 		return;
8760 
8761 	i40e_detect_recover_hung(pf);
8762 	i40e_sync_filters_subtask(pf);
8763 	i40e_reset_subtask(pf);
8764 	i40e_handle_mdd_event(pf);
8765 	i40e_vc_process_vflr_event(pf);
8766 	i40e_watchdog_subtask(pf);
8767 	i40e_fdir_reinit_subtask(pf);
8768 	if (pf->flags & I40E_FLAG_CLIENT_RESET) {
8769 		/* Client subtask will reopen next time through. */
8770 		i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], true);
8771 		pf->flags &= ~I40E_FLAG_CLIENT_RESET;
8772 	} else {
8773 		i40e_client_subtask(pf);
8774 		if (pf->flags & I40E_FLAG_CLIENT_L2_CHANGE) {
8775 			i40e_notify_client_of_l2_param_changes(
8776 							pf->vsi[pf->lan_vsi]);
8777 			pf->flags &= ~I40E_FLAG_CLIENT_L2_CHANGE;
8778 		}
8779 	}
8780 	i40e_sync_filters_subtask(pf);
8781 	i40e_sync_udp_filters_subtask(pf);
8782 	i40e_clean_adminq_subtask(pf);
8783 
8784 	/* flush memory to make sure state is correct before next watchdog */
8785 	smp_mb__before_atomic();
8786 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
8787 
8788 	/* If the tasks have taken longer than one timer cycle or there
8789 	 * is more work to be done, reschedule the service task now
8790 	 * rather than wait for the timer to tick again.
8791 	 */
8792 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
8793 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)		 ||
8794 	    test_bit(__I40E_MDD_EVENT_PENDING, pf->state)		 ||
8795 	    test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
8796 		i40e_service_event_schedule(pf);
8797 }
8798 
8799 /**
8800  * i40e_service_timer - timer callback
8801  * @data: pointer to PF struct
8802  **/
8803 static void i40e_service_timer(unsigned long data)
8804 {
8805 	struct i40e_pf *pf = (struct i40e_pf *)data;
8806 
8807 	mod_timer(&pf->service_timer,
8808 		  round_jiffies(jiffies + pf->service_timer_period));
8809 	i40e_service_event_schedule(pf);
8810 }
8811 
8812 /**
8813  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
8814  * @vsi: the VSI being configured
8815  **/
8816 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
8817 {
8818 	struct i40e_pf *pf = vsi->back;
8819 
8820 	switch (vsi->type) {
8821 	case I40E_VSI_MAIN:
8822 		vsi->alloc_queue_pairs = pf->num_lan_qps;
8823 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
8824 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
8825 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
8826 			vsi->num_q_vectors = pf->num_lan_msix;
8827 		else
8828 			vsi->num_q_vectors = 1;
8829 
8830 		break;
8831 
8832 	case I40E_VSI_FDIR:
8833 		vsi->alloc_queue_pairs = 1;
8834 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
8835 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
8836 		vsi->num_q_vectors = pf->num_fdsb_msix;
8837 		break;
8838 
8839 	case I40E_VSI_VMDQ2:
8840 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
8841 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
8842 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
8843 		vsi->num_q_vectors = pf->num_vmdq_msix;
8844 		break;
8845 
8846 	case I40E_VSI_SRIOV:
8847 		vsi->alloc_queue_pairs = pf->num_vf_qps;
8848 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
8849 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
8850 		break;
8851 
8852 	default:
8853 		WARN_ON(1);
8854 		return -ENODATA;
8855 	}
8856 
8857 	return 0;
8858 }
8859 
8860 /**
8861  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
8862  * @vsi: VSI pointer
8863  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
8864  *
8865  * On error: returns error code (negative)
8866  * On success: returns 0
8867  **/
8868 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
8869 {
8870 	struct i40e_ring **next_rings;
8871 	int size;
8872 	int ret = 0;
8873 
8874 	/* allocate memory for both Tx, XDP Tx and Rx ring pointers */
8875 	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs *
8876 	       (i40e_enabled_xdp_vsi(vsi) ? 3 : 2);
8877 	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
8878 	if (!vsi->tx_rings)
8879 		return -ENOMEM;
8880 	next_rings = vsi->tx_rings + vsi->alloc_queue_pairs;
8881 	if (i40e_enabled_xdp_vsi(vsi)) {
8882 		vsi->xdp_rings = next_rings;
8883 		next_rings += vsi->alloc_queue_pairs;
8884 	}
8885 	vsi->rx_rings = next_rings;
8886 
8887 	if (alloc_qvectors) {
8888 		/* allocate memory for q_vector pointers */
8889 		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
8890 		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
8891 		if (!vsi->q_vectors) {
8892 			ret = -ENOMEM;
8893 			goto err_vectors;
8894 		}
8895 	}
8896 	return ret;
8897 
8898 err_vectors:
8899 	kfree(vsi->tx_rings);
8900 	return ret;
8901 }
8902 
8903 /**
8904  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
8905  * @pf: board private structure
8906  * @type: type of VSI
8907  *
8908  * On error: returns error code (negative)
8909  * On success: returns vsi index in PF (positive)
8910  **/
8911 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
8912 {
8913 	int ret = -ENODEV;
8914 	struct i40e_vsi *vsi;
8915 	int vsi_idx;
8916 	int i;
8917 
8918 	/* Need to protect the allocation of the VSIs at the PF level */
8919 	mutex_lock(&pf->switch_mutex);
8920 
8921 	/* VSI list may be fragmented if VSI creation/destruction has
8922 	 * been happening.  We can afford to do a quick scan to look
8923 	 * for any free VSIs in the list.
8924 	 *
8925 	 * find next empty vsi slot, looping back around if necessary
8926 	 */
8927 	i = pf->next_vsi;
8928 	while (i < pf->num_alloc_vsi && pf->vsi[i])
8929 		i++;
8930 	if (i >= pf->num_alloc_vsi) {
8931 		i = 0;
8932 		while (i < pf->next_vsi && pf->vsi[i])
8933 			i++;
8934 	}
8935 
8936 	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
8937 		vsi_idx = i;             /* Found one! */
8938 	} else {
8939 		ret = -ENODEV;
8940 		goto unlock_pf;  /* out of VSI slots! */
8941 	}
8942 	pf->next_vsi = ++i;
8943 
8944 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
8945 	if (!vsi) {
8946 		ret = -ENOMEM;
8947 		goto unlock_pf;
8948 	}
8949 	vsi->type = type;
8950 	vsi->back = pf;
8951 	set_bit(__I40E_VSI_DOWN, vsi->state);
8952 	vsi->flags = 0;
8953 	vsi->idx = vsi_idx;
8954 	vsi->int_rate_limit = 0;
8955 	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
8956 				pf->rss_table_size : 64;
8957 	vsi->netdev_registered = false;
8958 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
8959 	hash_init(vsi->mac_filter_hash);
8960 	vsi->irqs_ready = false;
8961 
8962 	ret = i40e_set_num_rings_in_vsi(vsi);
8963 	if (ret)
8964 		goto err_rings;
8965 
8966 	ret = i40e_vsi_alloc_arrays(vsi, true);
8967 	if (ret)
8968 		goto err_rings;
8969 
8970 	/* Setup default MSIX irq handler for VSI */
8971 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
8972 
8973 	/* Initialize VSI lock */
8974 	spin_lock_init(&vsi->mac_filter_hash_lock);
8975 	pf->vsi[vsi_idx] = vsi;
8976 	ret = vsi_idx;
8977 	goto unlock_pf;
8978 
8979 err_rings:
8980 	pf->next_vsi = i - 1;
8981 	kfree(vsi);
8982 unlock_pf:
8983 	mutex_unlock(&pf->switch_mutex);
8984 	return ret;
8985 }
8986 
8987 /**
8988  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
8989  * @type: VSI pointer
8990  * @free_qvectors: a bool to specify if q_vectors need to be freed.
8991  *
8992  * On error: returns error code (negative)
8993  * On success: returns 0
8994  **/
8995 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
8996 {
8997 	/* free the ring and vector containers */
8998 	if (free_qvectors) {
8999 		kfree(vsi->q_vectors);
9000 		vsi->q_vectors = NULL;
9001 	}
9002 	kfree(vsi->tx_rings);
9003 	vsi->tx_rings = NULL;
9004 	vsi->rx_rings = NULL;
9005 	vsi->xdp_rings = NULL;
9006 }
9007 
9008 /**
9009  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
9010  * and lookup table
9011  * @vsi: Pointer to VSI structure
9012  */
9013 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
9014 {
9015 	if (!vsi)
9016 		return;
9017 
9018 	kfree(vsi->rss_hkey_user);
9019 	vsi->rss_hkey_user = NULL;
9020 
9021 	kfree(vsi->rss_lut_user);
9022 	vsi->rss_lut_user = NULL;
9023 }
9024 
9025 /**
9026  * i40e_vsi_clear - Deallocate the VSI provided
9027  * @vsi: the VSI being un-configured
9028  **/
9029 static int i40e_vsi_clear(struct i40e_vsi *vsi)
9030 {
9031 	struct i40e_pf *pf;
9032 
9033 	if (!vsi)
9034 		return 0;
9035 
9036 	if (!vsi->back)
9037 		goto free_vsi;
9038 	pf = vsi->back;
9039 
9040 	mutex_lock(&pf->switch_mutex);
9041 	if (!pf->vsi[vsi->idx]) {
9042 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
9043 			vsi->idx, vsi->idx, vsi, vsi->type);
9044 		goto unlock_vsi;
9045 	}
9046 
9047 	if (pf->vsi[vsi->idx] != vsi) {
9048 		dev_err(&pf->pdev->dev,
9049 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
9050 			pf->vsi[vsi->idx]->idx,
9051 			pf->vsi[vsi->idx],
9052 			pf->vsi[vsi->idx]->type,
9053 			vsi->idx, vsi, vsi->type);
9054 		goto unlock_vsi;
9055 	}
9056 
9057 	/* updates the PF for this cleared vsi */
9058 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9059 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
9060 
9061 	i40e_vsi_free_arrays(vsi, true);
9062 	i40e_clear_rss_config_user(vsi);
9063 
9064 	pf->vsi[vsi->idx] = NULL;
9065 	if (vsi->idx < pf->next_vsi)
9066 		pf->next_vsi = vsi->idx;
9067 
9068 unlock_vsi:
9069 	mutex_unlock(&pf->switch_mutex);
9070 free_vsi:
9071 	kfree(vsi);
9072 
9073 	return 0;
9074 }
9075 
9076 /**
9077  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
9078  * @vsi: the VSI being cleaned
9079  **/
9080 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
9081 {
9082 	int i;
9083 
9084 	if (vsi->tx_rings && vsi->tx_rings[0]) {
9085 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
9086 			kfree_rcu(vsi->tx_rings[i], rcu);
9087 			vsi->tx_rings[i] = NULL;
9088 			vsi->rx_rings[i] = NULL;
9089 			if (vsi->xdp_rings)
9090 				vsi->xdp_rings[i] = NULL;
9091 		}
9092 	}
9093 }
9094 
9095 /**
9096  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
9097  * @vsi: the VSI being configured
9098  **/
9099 static int i40e_alloc_rings(struct i40e_vsi *vsi)
9100 {
9101 	int i, qpv = i40e_enabled_xdp_vsi(vsi) ? 3 : 2;
9102 	struct i40e_pf *pf = vsi->back;
9103 	struct i40e_ring *ring;
9104 
9105 	/* Set basic values in the rings to be used later during open() */
9106 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
9107 		/* allocate space for both Tx and Rx in one shot */
9108 		ring = kcalloc(qpv, sizeof(struct i40e_ring), GFP_KERNEL);
9109 		if (!ring)
9110 			goto err_out;
9111 
9112 		ring->queue_index = i;
9113 		ring->reg_idx = vsi->base_queue + i;
9114 		ring->ring_active = false;
9115 		ring->vsi = vsi;
9116 		ring->netdev = vsi->netdev;
9117 		ring->dev = &pf->pdev->dev;
9118 		ring->count = vsi->num_desc;
9119 		ring->size = 0;
9120 		ring->dcb_tc = 0;
9121 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
9122 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
9123 		ring->tx_itr_setting = pf->tx_itr_default;
9124 		vsi->tx_rings[i] = ring++;
9125 
9126 		if (!i40e_enabled_xdp_vsi(vsi))
9127 			goto setup_rx;
9128 
9129 		ring->queue_index = vsi->alloc_queue_pairs + i;
9130 		ring->reg_idx = vsi->base_queue + ring->queue_index;
9131 		ring->ring_active = false;
9132 		ring->vsi = vsi;
9133 		ring->netdev = NULL;
9134 		ring->dev = &pf->pdev->dev;
9135 		ring->count = vsi->num_desc;
9136 		ring->size = 0;
9137 		ring->dcb_tc = 0;
9138 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
9139 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
9140 		set_ring_xdp(ring);
9141 		ring->tx_itr_setting = pf->tx_itr_default;
9142 		vsi->xdp_rings[i] = ring++;
9143 
9144 setup_rx:
9145 		ring->queue_index = i;
9146 		ring->reg_idx = vsi->base_queue + i;
9147 		ring->ring_active = false;
9148 		ring->vsi = vsi;
9149 		ring->netdev = vsi->netdev;
9150 		ring->dev = &pf->pdev->dev;
9151 		ring->count = vsi->num_desc;
9152 		ring->size = 0;
9153 		ring->dcb_tc = 0;
9154 		ring->rx_itr_setting = pf->rx_itr_default;
9155 		vsi->rx_rings[i] = ring;
9156 	}
9157 
9158 	return 0;
9159 
9160 err_out:
9161 	i40e_vsi_clear_rings(vsi);
9162 	return -ENOMEM;
9163 }
9164 
9165 /**
9166  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
9167  * @pf: board private structure
9168  * @vectors: the number of MSI-X vectors to request
9169  *
9170  * Returns the number of vectors reserved, or error
9171  **/
9172 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
9173 {
9174 	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
9175 					I40E_MIN_MSIX, vectors);
9176 	if (vectors < 0) {
9177 		dev_info(&pf->pdev->dev,
9178 			 "MSI-X vector reservation failed: %d\n", vectors);
9179 		vectors = 0;
9180 	}
9181 
9182 	return vectors;
9183 }
9184 
9185 /**
9186  * i40e_init_msix - Setup the MSIX capability
9187  * @pf: board private structure
9188  *
9189  * Work with the OS to set up the MSIX vectors needed.
9190  *
9191  * Returns the number of vectors reserved or negative on failure
9192  **/
9193 static int i40e_init_msix(struct i40e_pf *pf)
9194 {
9195 	struct i40e_hw *hw = &pf->hw;
9196 	int cpus, extra_vectors;
9197 	int vectors_left;
9198 	int v_budget, i;
9199 	int v_actual;
9200 	int iwarp_requested = 0;
9201 
9202 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9203 		return -ENODEV;
9204 
9205 	/* The number of vectors we'll request will be comprised of:
9206 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
9207 	 *   - The number of LAN queue pairs
9208 	 *	- Queues being used for RSS.
9209 	 *		We don't need as many as max_rss_size vectors.
9210 	 *		use rss_size instead in the calculation since that
9211 	 *		is governed by number of cpus in the system.
9212 	 *	- assumes symmetric Tx/Rx pairing
9213 	 *   - The number of VMDq pairs
9214 	 *   - The CPU count within the NUMA node if iWARP is enabled
9215 	 * Once we count this up, try the request.
9216 	 *
9217 	 * If we can't get what we want, we'll simplify to nearly nothing
9218 	 * and try again.  If that still fails, we punt.
9219 	 */
9220 	vectors_left = hw->func_caps.num_msix_vectors;
9221 	v_budget = 0;
9222 
9223 	/* reserve one vector for miscellaneous handler */
9224 	if (vectors_left) {
9225 		v_budget++;
9226 		vectors_left--;
9227 	}
9228 
9229 	/* reserve some vectors for the main PF traffic queues. Initially we
9230 	 * only reserve at most 50% of the available vectors, in the case that
9231 	 * the number of online CPUs is large. This ensures that we can enable
9232 	 * extra features as well. Once we've enabled the other features, we
9233 	 * will use any remaining vectors to reach as close as we can to the
9234 	 * number of online CPUs.
9235 	 */
9236 	cpus = num_online_cpus();
9237 	pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
9238 	vectors_left -= pf->num_lan_msix;
9239 
9240 	/* reserve one vector for sideband flow director */
9241 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9242 		if (vectors_left) {
9243 			pf->num_fdsb_msix = 1;
9244 			v_budget++;
9245 			vectors_left--;
9246 		} else {
9247 			pf->num_fdsb_msix = 0;
9248 		}
9249 	}
9250 
9251 	/* can we reserve enough for iWARP? */
9252 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
9253 		iwarp_requested = pf->num_iwarp_msix;
9254 
9255 		if (!vectors_left)
9256 			pf->num_iwarp_msix = 0;
9257 		else if (vectors_left < pf->num_iwarp_msix)
9258 			pf->num_iwarp_msix = 1;
9259 		v_budget += pf->num_iwarp_msix;
9260 		vectors_left -= pf->num_iwarp_msix;
9261 	}
9262 
9263 	/* any vectors left over go for VMDq support */
9264 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
9265 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
9266 		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
9267 
9268 		if (!vectors_left) {
9269 			pf->num_vmdq_msix = 0;
9270 			pf->num_vmdq_qps = 0;
9271 		} else {
9272 			/* if we're short on vectors for what's desired, we limit
9273 			 * the queues per vmdq.  If this is still more than are
9274 			 * available, the user will need to change the number of
9275 			 * queues/vectors used by the PF later with the ethtool
9276 			 * channels command
9277 			 */
9278 			if (vmdq_vecs < vmdq_vecs_wanted)
9279 				pf->num_vmdq_qps = 1;
9280 			pf->num_vmdq_msix = pf->num_vmdq_qps;
9281 
9282 			v_budget += vmdq_vecs;
9283 			vectors_left -= vmdq_vecs;
9284 		}
9285 	}
9286 
9287 	/* On systems with a large number of SMP cores, we previously limited
9288 	 * the number of vectors for num_lan_msix to be at most 50% of the
9289 	 * available vectors, to allow for other features. Now, we add back
9290 	 * the remaining vectors. However, we ensure that the total
9291 	 * num_lan_msix will not exceed num_online_cpus(). To do this, we
9292 	 * calculate the number of vectors we can add without going over the
9293 	 * cap of CPUs. For systems with a small number of CPUs this will be
9294 	 * zero.
9295 	 */
9296 	extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
9297 	pf->num_lan_msix += extra_vectors;
9298 	vectors_left -= extra_vectors;
9299 
9300 	WARN(vectors_left < 0,
9301 	     "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
9302 
9303 	v_budget += pf->num_lan_msix;
9304 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
9305 				   GFP_KERNEL);
9306 	if (!pf->msix_entries)
9307 		return -ENOMEM;
9308 
9309 	for (i = 0; i < v_budget; i++)
9310 		pf->msix_entries[i].entry = i;
9311 	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
9312 
9313 	if (v_actual < I40E_MIN_MSIX) {
9314 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
9315 		kfree(pf->msix_entries);
9316 		pf->msix_entries = NULL;
9317 		pci_disable_msix(pf->pdev);
9318 		return -ENODEV;
9319 
9320 	} else if (v_actual == I40E_MIN_MSIX) {
9321 		/* Adjust for minimal MSIX use */
9322 		pf->num_vmdq_vsis = 0;
9323 		pf->num_vmdq_qps = 0;
9324 		pf->num_lan_qps = 1;
9325 		pf->num_lan_msix = 1;
9326 
9327 	} else if (!vectors_left) {
9328 		/* If we have limited resources, we will start with no vectors
9329 		 * for the special features and then allocate vectors to some
9330 		 * of these features based on the policy and at the end disable
9331 		 * the features that did not get any vectors.
9332 		 */
9333 		int vec;
9334 
9335 		dev_info(&pf->pdev->dev,
9336 			 "MSI-X vector limit reached, attempting to redistribute vectors\n");
9337 		/* reserve the misc vector */
9338 		vec = v_actual - 1;
9339 
9340 		/* Scale vector usage down */
9341 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
9342 		pf->num_vmdq_vsis = 1;
9343 		pf->num_vmdq_qps = 1;
9344 
9345 		/* partition out the remaining vectors */
9346 		switch (vec) {
9347 		case 2:
9348 			pf->num_lan_msix = 1;
9349 			break;
9350 		case 3:
9351 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
9352 				pf->num_lan_msix = 1;
9353 				pf->num_iwarp_msix = 1;
9354 			} else {
9355 				pf->num_lan_msix = 2;
9356 			}
9357 			break;
9358 		default:
9359 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
9360 				pf->num_iwarp_msix = min_t(int, (vec / 3),
9361 						 iwarp_requested);
9362 				pf->num_vmdq_vsis = min_t(int, (vec / 3),
9363 						  I40E_DEFAULT_NUM_VMDQ_VSI);
9364 			} else {
9365 				pf->num_vmdq_vsis = min_t(int, (vec / 2),
9366 						  I40E_DEFAULT_NUM_VMDQ_VSI);
9367 			}
9368 			if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9369 				pf->num_fdsb_msix = 1;
9370 				vec--;
9371 			}
9372 			pf->num_lan_msix = min_t(int,
9373 			       (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
9374 							      pf->num_lan_msix);
9375 			pf->num_lan_qps = pf->num_lan_msix;
9376 			break;
9377 		}
9378 	}
9379 
9380 	if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
9381 	    (pf->num_fdsb_msix == 0)) {
9382 		dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
9383 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
9384 	}
9385 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
9386 	    (pf->num_vmdq_msix == 0)) {
9387 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
9388 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
9389 	}
9390 
9391 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
9392 	    (pf->num_iwarp_msix == 0)) {
9393 		dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
9394 		pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
9395 	}
9396 	i40e_debug(&pf->hw, I40E_DEBUG_INIT,
9397 		   "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
9398 		   pf->num_lan_msix,
9399 		   pf->num_vmdq_msix * pf->num_vmdq_vsis,
9400 		   pf->num_fdsb_msix,
9401 		   pf->num_iwarp_msix);
9402 
9403 	return v_actual;
9404 }
9405 
9406 /**
9407  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
9408  * @vsi: the VSI being configured
9409  * @v_idx: index of the vector in the vsi struct
9410  * @cpu: cpu to be used on affinity_mask
9411  *
9412  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
9413  **/
9414 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
9415 {
9416 	struct i40e_q_vector *q_vector;
9417 
9418 	/* allocate q_vector */
9419 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
9420 	if (!q_vector)
9421 		return -ENOMEM;
9422 
9423 	q_vector->vsi = vsi;
9424 	q_vector->v_idx = v_idx;
9425 	cpumask_copy(&q_vector->affinity_mask, cpu_possible_mask);
9426 
9427 	if (vsi->netdev)
9428 		netif_napi_add(vsi->netdev, &q_vector->napi,
9429 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
9430 
9431 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
9432 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
9433 
9434 	/* tie q_vector and vsi together */
9435 	vsi->q_vectors[v_idx] = q_vector;
9436 
9437 	return 0;
9438 }
9439 
9440 /**
9441  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
9442  * @vsi: the VSI being configured
9443  *
9444  * We allocate one q_vector per queue interrupt.  If allocation fails we
9445  * return -ENOMEM.
9446  **/
9447 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
9448 {
9449 	struct i40e_pf *pf = vsi->back;
9450 	int err, v_idx, num_q_vectors, current_cpu;
9451 
9452 	/* if not MSIX, give the one vector only to the LAN VSI */
9453 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
9454 		num_q_vectors = vsi->num_q_vectors;
9455 	else if (vsi == pf->vsi[pf->lan_vsi])
9456 		num_q_vectors = 1;
9457 	else
9458 		return -EINVAL;
9459 
9460 	current_cpu = cpumask_first(cpu_online_mask);
9461 
9462 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
9463 		err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
9464 		if (err)
9465 			goto err_out;
9466 		current_cpu = cpumask_next(current_cpu, cpu_online_mask);
9467 		if (unlikely(current_cpu >= nr_cpu_ids))
9468 			current_cpu = cpumask_first(cpu_online_mask);
9469 	}
9470 
9471 	return 0;
9472 
9473 err_out:
9474 	while (v_idx--)
9475 		i40e_free_q_vector(vsi, v_idx);
9476 
9477 	return err;
9478 }
9479 
9480 /**
9481  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
9482  * @pf: board private structure to initialize
9483  **/
9484 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
9485 {
9486 	int vectors = 0;
9487 	ssize_t size;
9488 
9489 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
9490 		vectors = i40e_init_msix(pf);
9491 		if (vectors < 0) {
9492 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
9493 				       I40E_FLAG_IWARP_ENABLED	|
9494 				       I40E_FLAG_RSS_ENABLED	|
9495 				       I40E_FLAG_DCB_CAPABLE	|
9496 				       I40E_FLAG_DCB_ENABLED	|
9497 				       I40E_FLAG_SRIOV_ENABLED	|
9498 				       I40E_FLAG_FD_SB_ENABLED	|
9499 				       I40E_FLAG_FD_ATR_ENABLED	|
9500 				       I40E_FLAG_VMDQ_ENABLED);
9501 
9502 			/* rework the queue expectations without MSIX */
9503 			i40e_determine_queue_usage(pf);
9504 		}
9505 	}
9506 
9507 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
9508 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
9509 		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
9510 		vectors = pci_enable_msi(pf->pdev);
9511 		if (vectors < 0) {
9512 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
9513 				 vectors);
9514 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
9515 		}
9516 		vectors = 1;  /* one MSI or Legacy vector */
9517 	}
9518 
9519 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
9520 		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
9521 
9522 	/* set up vector assignment tracking */
9523 	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
9524 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
9525 	if (!pf->irq_pile) {
9526 		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
9527 		return -ENOMEM;
9528 	}
9529 	pf->irq_pile->num_entries = vectors;
9530 	pf->irq_pile->search_hint = 0;
9531 
9532 	/* track first vector for misc interrupts, ignore return */
9533 	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
9534 
9535 	return 0;
9536 }
9537 
9538 #ifdef CONFIG_PM
9539 /**
9540  * i40e_restore_interrupt_scheme - Restore the interrupt scheme
9541  * @pf: private board data structure
9542  *
9543  * Restore the interrupt scheme that was cleared when we suspended the
9544  * device. This should be called during resume to re-allocate the q_vectors
9545  * and reacquire IRQs.
9546  */
9547 static int i40e_restore_interrupt_scheme(struct i40e_pf *pf)
9548 {
9549 	int err, i;
9550 
9551 	/* We cleared the MSI and MSI-X flags when disabling the old interrupt
9552 	 * scheme. We need to re-enabled them here in order to attempt to
9553 	 * re-acquire the MSI or MSI-X vectors
9554 	 */
9555 	pf->flags |= (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
9556 
9557 	err = i40e_init_interrupt_scheme(pf);
9558 	if (err)
9559 		return err;
9560 
9561 	/* Now that we've re-acquired IRQs, we need to remap the vectors and
9562 	 * rings together again.
9563 	 */
9564 	for (i = 0; i < pf->num_alloc_vsi; i++) {
9565 		if (pf->vsi[i]) {
9566 			err = i40e_vsi_alloc_q_vectors(pf->vsi[i]);
9567 			if (err)
9568 				goto err_unwind;
9569 			i40e_vsi_map_rings_to_vectors(pf->vsi[i]);
9570 		}
9571 	}
9572 
9573 	err = i40e_setup_misc_vector(pf);
9574 	if (err)
9575 		goto err_unwind;
9576 
9577 	return 0;
9578 
9579 err_unwind:
9580 	while (i--) {
9581 		if (pf->vsi[i])
9582 			i40e_vsi_free_q_vectors(pf->vsi[i]);
9583 	}
9584 
9585 	return err;
9586 }
9587 #endif /* CONFIG_PM */
9588 
9589 /**
9590  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
9591  * @pf: board private structure
9592  *
9593  * This sets up the handler for MSIX 0, which is used to manage the
9594  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
9595  * when in MSI or Legacy interrupt mode.
9596  **/
9597 static int i40e_setup_misc_vector(struct i40e_pf *pf)
9598 {
9599 	struct i40e_hw *hw = &pf->hw;
9600 	int err = 0;
9601 
9602 	/* Only request the IRQ once, the first time through. */
9603 	if (!test_and_set_bit(__I40E_MISC_IRQ_REQUESTED, pf->state)) {
9604 		err = request_irq(pf->msix_entries[0].vector,
9605 				  i40e_intr, 0, pf->int_name, pf);
9606 		if (err) {
9607 			clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
9608 			dev_info(&pf->pdev->dev,
9609 				 "request_irq for %s failed: %d\n",
9610 				 pf->int_name, err);
9611 			return -EFAULT;
9612 		}
9613 	}
9614 
9615 	i40e_enable_misc_int_causes(pf);
9616 
9617 	/* associate no queues to the misc vector */
9618 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
9619 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
9620 
9621 	i40e_flush(hw);
9622 
9623 	i40e_irq_dynamic_enable_icr0(pf);
9624 
9625 	return err;
9626 }
9627 
9628 /**
9629  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
9630  * @vsi: Pointer to vsi structure
9631  * @seed: Buffter to store the hash keys
9632  * @lut: Buffer to store the lookup table entries
9633  * @lut_size: Size of buffer to store the lookup table entries
9634  *
9635  * Return 0 on success, negative on failure
9636  */
9637 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
9638 			   u8 *lut, u16 lut_size)
9639 {
9640 	struct i40e_pf *pf = vsi->back;
9641 	struct i40e_hw *hw = &pf->hw;
9642 	int ret = 0;
9643 
9644 	if (seed) {
9645 		ret = i40e_aq_get_rss_key(hw, vsi->id,
9646 			(struct i40e_aqc_get_set_rss_key_data *)seed);
9647 		if (ret) {
9648 			dev_info(&pf->pdev->dev,
9649 				 "Cannot get RSS key, err %s aq_err %s\n",
9650 				 i40e_stat_str(&pf->hw, ret),
9651 				 i40e_aq_str(&pf->hw,
9652 					     pf->hw.aq.asq_last_status));
9653 			return ret;
9654 		}
9655 	}
9656 
9657 	if (lut) {
9658 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
9659 
9660 		ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
9661 		if (ret) {
9662 			dev_info(&pf->pdev->dev,
9663 				 "Cannot get RSS lut, err %s aq_err %s\n",
9664 				 i40e_stat_str(&pf->hw, ret),
9665 				 i40e_aq_str(&pf->hw,
9666 					     pf->hw.aq.asq_last_status));
9667 			return ret;
9668 		}
9669 	}
9670 
9671 	return ret;
9672 }
9673 
9674 /**
9675  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
9676  * @vsi: Pointer to vsi structure
9677  * @seed: RSS hash seed
9678  * @lut: Lookup table
9679  * @lut_size: Lookup table size
9680  *
9681  * Returns 0 on success, negative on failure
9682  **/
9683 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
9684 			       const u8 *lut, u16 lut_size)
9685 {
9686 	struct i40e_pf *pf = vsi->back;
9687 	struct i40e_hw *hw = &pf->hw;
9688 	u16 vf_id = vsi->vf_id;
9689 	u8 i;
9690 
9691 	/* Fill out hash function seed */
9692 	if (seed) {
9693 		u32 *seed_dw = (u32 *)seed;
9694 
9695 		if (vsi->type == I40E_VSI_MAIN) {
9696 			for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
9697 				wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
9698 		} else if (vsi->type == I40E_VSI_SRIOV) {
9699 			for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
9700 				wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
9701 		} else {
9702 			dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
9703 		}
9704 	}
9705 
9706 	if (lut) {
9707 		u32 *lut_dw = (u32 *)lut;
9708 
9709 		if (vsi->type == I40E_VSI_MAIN) {
9710 			if (lut_size != I40E_HLUT_ARRAY_SIZE)
9711 				return -EINVAL;
9712 			for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
9713 				wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
9714 		} else if (vsi->type == I40E_VSI_SRIOV) {
9715 			if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
9716 				return -EINVAL;
9717 			for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
9718 				wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
9719 		} else {
9720 			dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
9721 		}
9722 	}
9723 	i40e_flush(hw);
9724 
9725 	return 0;
9726 }
9727 
9728 /**
9729  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
9730  * @vsi: Pointer to VSI structure
9731  * @seed: Buffer to store the keys
9732  * @lut: Buffer to store the lookup table entries
9733  * @lut_size: Size of buffer to store the lookup table entries
9734  *
9735  * Returns 0 on success, negative on failure
9736  */
9737 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
9738 			    u8 *lut, u16 lut_size)
9739 {
9740 	struct i40e_pf *pf = vsi->back;
9741 	struct i40e_hw *hw = &pf->hw;
9742 	u16 i;
9743 
9744 	if (seed) {
9745 		u32 *seed_dw = (u32 *)seed;
9746 
9747 		for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
9748 			seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
9749 	}
9750 	if (lut) {
9751 		u32 *lut_dw = (u32 *)lut;
9752 
9753 		if (lut_size != I40E_HLUT_ARRAY_SIZE)
9754 			return -EINVAL;
9755 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
9756 			lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
9757 	}
9758 
9759 	return 0;
9760 }
9761 
9762 /**
9763  * i40e_config_rss - Configure RSS keys and lut
9764  * @vsi: Pointer to VSI structure
9765  * @seed: RSS hash seed
9766  * @lut: Lookup table
9767  * @lut_size: Lookup table size
9768  *
9769  * Returns 0 on success, negative on failure
9770  */
9771 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
9772 {
9773 	struct i40e_pf *pf = vsi->back;
9774 
9775 	if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
9776 		return i40e_config_rss_aq(vsi, seed, lut, lut_size);
9777 	else
9778 		return i40e_config_rss_reg(vsi, seed, lut, lut_size);
9779 }
9780 
9781 /**
9782  * i40e_get_rss - Get RSS keys and lut
9783  * @vsi: Pointer to VSI structure
9784  * @seed: Buffer to store the keys
9785  * @lut: Buffer to store the lookup table entries
9786  * lut_size: Size of buffer to store the lookup table entries
9787  *
9788  * Returns 0 on success, negative on failure
9789  */
9790 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
9791 {
9792 	struct i40e_pf *pf = vsi->back;
9793 
9794 	if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
9795 		return i40e_get_rss_aq(vsi, seed, lut, lut_size);
9796 	else
9797 		return i40e_get_rss_reg(vsi, seed, lut, lut_size);
9798 }
9799 
9800 /**
9801  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
9802  * @pf: Pointer to board private structure
9803  * @lut: Lookup table
9804  * @rss_table_size: Lookup table size
9805  * @rss_size: Range of queue number for hashing
9806  */
9807 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
9808 		       u16 rss_table_size, u16 rss_size)
9809 {
9810 	u16 i;
9811 
9812 	for (i = 0; i < rss_table_size; i++)
9813 		lut[i] = i % rss_size;
9814 }
9815 
9816 /**
9817  * i40e_pf_config_rss - Prepare for RSS if used
9818  * @pf: board private structure
9819  **/
9820 static int i40e_pf_config_rss(struct i40e_pf *pf)
9821 {
9822 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9823 	u8 seed[I40E_HKEY_ARRAY_SIZE];
9824 	u8 *lut;
9825 	struct i40e_hw *hw = &pf->hw;
9826 	u32 reg_val;
9827 	u64 hena;
9828 	int ret;
9829 
9830 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
9831 	hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
9832 		((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
9833 	hena |= i40e_pf_get_default_rss_hena(pf);
9834 
9835 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
9836 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
9837 
9838 	/* Determine the RSS table size based on the hardware capabilities */
9839 	reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
9840 	reg_val = (pf->rss_table_size == 512) ?
9841 			(reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
9842 			(reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
9843 	i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
9844 
9845 	/* Determine the RSS size of the VSI */
9846 	if (!vsi->rss_size) {
9847 		u16 qcount;
9848 
9849 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
9850 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
9851 	}
9852 	if (!vsi->rss_size)
9853 		return -EINVAL;
9854 
9855 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
9856 	if (!lut)
9857 		return -ENOMEM;
9858 
9859 	/* Use user configured lut if there is one, otherwise use default */
9860 	if (vsi->rss_lut_user)
9861 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
9862 	else
9863 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
9864 
9865 	/* Use user configured hash key if there is one, otherwise
9866 	 * use default.
9867 	 */
9868 	if (vsi->rss_hkey_user)
9869 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
9870 	else
9871 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
9872 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
9873 	kfree(lut);
9874 
9875 	return ret;
9876 }
9877 
9878 /**
9879  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
9880  * @pf: board private structure
9881  * @queue_count: the requested queue count for rss.
9882  *
9883  * returns 0 if rss is not enabled, if enabled returns the final rss queue
9884  * count which may be different from the requested queue count.
9885  * Note: expects to be called while under rtnl_lock()
9886  **/
9887 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
9888 {
9889 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9890 	int new_rss_size;
9891 
9892 	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
9893 		return 0;
9894 
9895 	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
9896 
9897 	if (queue_count != vsi->num_queue_pairs) {
9898 		u16 qcount;
9899 
9900 		vsi->req_queue_pairs = queue_count;
9901 		i40e_prep_for_reset(pf, true);
9902 
9903 		pf->alloc_rss_size = new_rss_size;
9904 
9905 		i40e_reset_and_rebuild(pf, true, true);
9906 
9907 		/* Discard the user configured hash keys and lut, if less
9908 		 * queues are enabled.
9909 		 */
9910 		if (queue_count < vsi->rss_size) {
9911 			i40e_clear_rss_config_user(vsi);
9912 			dev_dbg(&pf->pdev->dev,
9913 				"discard user configured hash keys and lut\n");
9914 		}
9915 
9916 		/* Reset vsi->rss_size, as number of enabled queues changed */
9917 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
9918 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
9919 
9920 		i40e_pf_config_rss(pf);
9921 	}
9922 	dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
9923 		 vsi->req_queue_pairs, pf->rss_size_max);
9924 	return pf->alloc_rss_size;
9925 }
9926 
9927 /**
9928  * i40e_get_partition_bw_setting - Retrieve BW settings for this PF partition
9929  * @pf: board private structure
9930  **/
9931 i40e_status i40e_get_partition_bw_setting(struct i40e_pf *pf)
9932 {
9933 	i40e_status status;
9934 	bool min_valid, max_valid;
9935 	u32 max_bw, min_bw;
9936 
9937 	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
9938 					   &min_valid, &max_valid);
9939 
9940 	if (!status) {
9941 		if (min_valid)
9942 			pf->min_bw = min_bw;
9943 		if (max_valid)
9944 			pf->max_bw = max_bw;
9945 	}
9946 
9947 	return status;
9948 }
9949 
9950 /**
9951  * i40e_set_partition_bw_setting - Set BW settings for this PF partition
9952  * @pf: board private structure
9953  **/
9954 i40e_status i40e_set_partition_bw_setting(struct i40e_pf *pf)
9955 {
9956 	struct i40e_aqc_configure_partition_bw_data bw_data;
9957 	i40e_status status;
9958 
9959 	/* Set the valid bit for this PF */
9960 	bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
9961 	bw_data.max_bw[pf->hw.pf_id] = pf->max_bw & I40E_ALT_BW_VALUE_MASK;
9962 	bw_data.min_bw[pf->hw.pf_id] = pf->min_bw & I40E_ALT_BW_VALUE_MASK;
9963 
9964 	/* Set the new bandwidths */
9965 	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
9966 
9967 	return status;
9968 }
9969 
9970 /**
9971  * i40e_commit_partition_bw_setting - Commit BW settings for this PF partition
9972  * @pf: board private structure
9973  **/
9974 i40e_status i40e_commit_partition_bw_setting(struct i40e_pf *pf)
9975 {
9976 	/* Commit temporary BW setting to permanent NVM image */
9977 	enum i40e_admin_queue_err last_aq_status;
9978 	i40e_status ret;
9979 	u16 nvm_word;
9980 
9981 	if (pf->hw.partition_id != 1) {
9982 		dev_info(&pf->pdev->dev,
9983 			 "Commit BW only works on partition 1! This is partition %d",
9984 			 pf->hw.partition_id);
9985 		ret = I40E_NOT_SUPPORTED;
9986 		goto bw_commit_out;
9987 	}
9988 
9989 	/* Acquire NVM for read access */
9990 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
9991 	last_aq_status = pf->hw.aq.asq_last_status;
9992 	if (ret) {
9993 		dev_info(&pf->pdev->dev,
9994 			 "Cannot acquire NVM for read access, err %s aq_err %s\n",
9995 			 i40e_stat_str(&pf->hw, ret),
9996 			 i40e_aq_str(&pf->hw, last_aq_status));
9997 		goto bw_commit_out;
9998 	}
9999 
10000 	/* Read word 0x10 of NVM - SW compatibility word 1 */
10001 	ret = i40e_aq_read_nvm(&pf->hw,
10002 			       I40E_SR_NVM_CONTROL_WORD,
10003 			       0x10, sizeof(nvm_word), &nvm_word,
10004 			       false, NULL);
10005 	/* Save off last admin queue command status before releasing
10006 	 * the NVM
10007 	 */
10008 	last_aq_status = pf->hw.aq.asq_last_status;
10009 	i40e_release_nvm(&pf->hw);
10010 	if (ret) {
10011 		dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
10012 			 i40e_stat_str(&pf->hw, ret),
10013 			 i40e_aq_str(&pf->hw, last_aq_status));
10014 		goto bw_commit_out;
10015 	}
10016 
10017 	/* Wait a bit for NVM release to complete */
10018 	msleep(50);
10019 
10020 	/* Acquire NVM for write access */
10021 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
10022 	last_aq_status = pf->hw.aq.asq_last_status;
10023 	if (ret) {
10024 		dev_info(&pf->pdev->dev,
10025 			 "Cannot acquire NVM for write access, err %s aq_err %s\n",
10026 			 i40e_stat_str(&pf->hw, ret),
10027 			 i40e_aq_str(&pf->hw, last_aq_status));
10028 		goto bw_commit_out;
10029 	}
10030 	/* Write it back out unchanged to initiate update NVM,
10031 	 * which will force a write of the shadow (alt) RAM to
10032 	 * the NVM - thus storing the bandwidth values permanently.
10033 	 */
10034 	ret = i40e_aq_update_nvm(&pf->hw,
10035 				 I40E_SR_NVM_CONTROL_WORD,
10036 				 0x10, sizeof(nvm_word),
10037 				 &nvm_word, true, NULL);
10038 	/* Save off last admin queue command status before releasing
10039 	 * the NVM
10040 	 */
10041 	last_aq_status = pf->hw.aq.asq_last_status;
10042 	i40e_release_nvm(&pf->hw);
10043 	if (ret)
10044 		dev_info(&pf->pdev->dev,
10045 			 "BW settings NOT SAVED, err %s aq_err %s\n",
10046 			 i40e_stat_str(&pf->hw, ret),
10047 			 i40e_aq_str(&pf->hw, last_aq_status));
10048 bw_commit_out:
10049 
10050 	return ret;
10051 }
10052 
10053 /**
10054  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
10055  * @pf: board private structure to initialize
10056  *
10057  * i40e_sw_init initializes the Adapter private data structure.
10058  * Fields are initialized based on PCI device information and
10059  * OS network device settings (MTU size).
10060  **/
10061 static int i40e_sw_init(struct i40e_pf *pf)
10062 {
10063 	int err = 0;
10064 	int size;
10065 
10066 	/* Set default capability flags */
10067 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
10068 		    I40E_FLAG_MSI_ENABLED     |
10069 		    I40E_FLAG_MSIX_ENABLED;
10070 
10071 	/* Set default ITR */
10072 	pf->rx_itr_default = I40E_ITR_RX_DEF;
10073 	pf->tx_itr_default = I40E_ITR_TX_DEF;
10074 
10075 	/* Depending on PF configurations, it is possible that the RSS
10076 	 * maximum might end up larger than the available queues
10077 	 */
10078 	pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
10079 	pf->alloc_rss_size = 1;
10080 	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
10081 	pf->rss_size_max = min_t(int, pf->rss_size_max,
10082 				 pf->hw.func_caps.num_tx_qp);
10083 	if (pf->hw.func_caps.rss) {
10084 		pf->flags |= I40E_FLAG_RSS_ENABLED;
10085 		pf->alloc_rss_size = min_t(int, pf->rss_size_max,
10086 					   num_online_cpus());
10087 	}
10088 
10089 	/* MFP mode enabled */
10090 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
10091 		pf->flags |= I40E_FLAG_MFP_ENABLED;
10092 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
10093 		if (i40e_get_partition_bw_setting(pf)) {
10094 			dev_warn(&pf->pdev->dev,
10095 				 "Could not get partition bw settings\n");
10096 		} else {
10097 			dev_info(&pf->pdev->dev,
10098 				 "Partition BW Min = %8.8x, Max = %8.8x\n",
10099 				 pf->min_bw, pf->max_bw);
10100 
10101 			/* nudge the Tx scheduler */
10102 			i40e_set_partition_bw_setting(pf);
10103 		}
10104 	}
10105 
10106 	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
10107 	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
10108 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
10109 		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
10110 		if (pf->flags & I40E_FLAG_MFP_ENABLED &&
10111 		    pf->hw.num_partitions > 1)
10112 			dev_info(&pf->pdev->dev,
10113 				 "Flow Director Sideband mode Disabled in MFP mode\n");
10114 		else
10115 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
10116 		pf->fdir_pf_filter_count =
10117 				 pf->hw.func_caps.fd_filters_guaranteed;
10118 		pf->hw.fdir_shared_filter_count =
10119 				 pf->hw.func_caps.fd_filters_best_effort;
10120 	}
10121 
10122 	if (pf->hw.mac.type == I40E_MAC_X722) {
10123 		pf->hw_features |= (I40E_HW_RSS_AQ_CAPABLE |
10124 				    I40E_HW_128_QP_RSS_CAPABLE |
10125 				    I40E_HW_ATR_EVICT_CAPABLE |
10126 				    I40E_HW_WB_ON_ITR_CAPABLE |
10127 				    I40E_HW_MULTIPLE_TCP_UDP_RSS_PCTYPE |
10128 				    I40E_HW_NO_PCI_LINK_CHECK |
10129 				    I40E_HW_USE_SET_LLDP_MIB |
10130 				    I40E_HW_GENEVE_OFFLOAD_CAPABLE |
10131 				    I40E_HW_PTP_L4_CAPABLE |
10132 				    I40E_HW_WOL_MC_MAGIC_PKT_WAKE |
10133 				    I40E_HW_OUTER_UDP_CSUM_CAPABLE);
10134 
10135 #define I40E_FDEVICT_PCTYPE_DEFAULT 0xc03
10136 		if (rd32(&pf->hw, I40E_GLQF_FDEVICTENA(1)) !=
10137 		    I40E_FDEVICT_PCTYPE_DEFAULT) {
10138 			dev_warn(&pf->pdev->dev,
10139 				 "FD EVICT PCTYPES are not right, disable FD HW EVICT\n");
10140 			pf->hw_features &= ~I40E_HW_ATR_EVICT_CAPABLE;
10141 		}
10142 	} else if ((pf->hw.aq.api_maj_ver > 1) ||
10143 		   ((pf->hw.aq.api_maj_ver == 1) &&
10144 		    (pf->hw.aq.api_min_ver > 4))) {
10145 		/* Supported in FW API version higher than 1.4 */
10146 		pf->hw_features |= I40E_HW_GENEVE_OFFLOAD_CAPABLE;
10147 	}
10148 
10149 	/* Enable HW ATR eviction if possible */
10150 	if (pf->hw_features & I40E_HW_ATR_EVICT_CAPABLE)
10151 		pf->flags |= I40E_FLAG_HW_ATR_EVICT_ENABLED;
10152 
10153 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
10154 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
10155 	    (pf->hw.aq.fw_maj_ver < 4))) {
10156 		pf->hw_features |= I40E_HW_RESTART_AUTONEG;
10157 		/* No DCB support  for FW < v4.33 */
10158 		pf->hw_features |= I40E_HW_NO_DCB_SUPPORT;
10159 	}
10160 
10161 	/* Disable FW LLDP if FW < v4.3 */
10162 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
10163 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
10164 	    (pf->hw.aq.fw_maj_ver < 4)))
10165 		pf->hw_features |= I40E_HW_STOP_FW_LLDP;
10166 
10167 	/* Use the FW Set LLDP MIB API if FW > v4.40 */
10168 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
10169 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
10170 	    (pf->hw.aq.fw_maj_ver >= 5)))
10171 		pf->hw_features |= I40E_HW_USE_SET_LLDP_MIB;
10172 
10173 	/* Enable PTP L4 if FW > v6.0 */
10174 	if (pf->hw.mac.type == I40E_MAC_XL710 &&
10175 	    pf->hw.aq.fw_maj_ver >= 6)
10176 		pf->hw_features |= I40E_HW_PTP_L4_CAPABLE;
10177 
10178 	if (pf->hw.func_caps.vmdq) {
10179 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
10180 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
10181 		pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
10182 	}
10183 
10184 	if (pf->hw.func_caps.iwarp) {
10185 		pf->flags |= I40E_FLAG_IWARP_ENABLED;
10186 		/* IWARP needs one extra vector for CQP just like MISC.*/
10187 		pf->num_iwarp_msix = (int)num_online_cpus() + 1;
10188 	}
10189 
10190 #ifdef CONFIG_PCI_IOV
10191 	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
10192 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
10193 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
10194 		pf->num_req_vfs = min_t(int,
10195 					pf->hw.func_caps.num_vfs,
10196 					I40E_MAX_VF_COUNT);
10197 	}
10198 #endif /* CONFIG_PCI_IOV */
10199 	pf->eeprom_version = 0xDEAD;
10200 	pf->lan_veb = I40E_NO_VEB;
10201 	pf->lan_vsi = I40E_NO_VSI;
10202 
10203 	/* By default FW has this off for performance reasons */
10204 	pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
10205 
10206 	/* set up queue assignment tracking */
10207 	size = sizeof(struct i40e_lump_tracking)
10208 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
10209 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
10210 	if (!pf->qp_pile) {
10211 		err = -ENOMEM;
10212 		goto sw_init_done;
10213 	}
10214 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
10215 	pf->qp_pile->search_hint = 0;
10216 
10217 	pf->tx_timeout_recovery_level = 1;
10218 
10219 	mutex_init(&pf->switch_mutex);
10220 
10221 sw_init_done:
10222 	return err;
10223 }
10224 
10225 /**
10226  * i40e_set_ntuple - set the ntuple feature flag and take action
10227  * @pf: board private structure to initialize
10228  * @features: the feature set that the stack is suggesting
10229  *
10230  * returns a bool to indicate if reset needs to happen
10231  **/
10232 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
10233 {
10234 	bool need_reset = false;
10235 
10236 	/* Check if Flow Director n-tuple support was enabled or disabled.  If
10237 	 * the state changed, we need to reset.
10238 	 */
10239 	if (features & NETIF_F_NTUPLE) {
10240 		/* Enable filters and mark for reset */
10241 		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
10242 			need_reset = true;
10243 		/* enable FD_SB only if there is MSI-X vector */
10244 		if (pf->num_fdsb_msix > 0)
10245 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
10246 	} else {
10247 		/* turn off filters, mark for reset and clear SW filter list */
10248 		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10249 			need_reset = true;
10250 			i40e_fdir_filter_exit(pf);
10251 		}
10252 		pf->flags &= ~(I40E_FLAG_FD_SB_ENABLED |
10253 			       I40E_FLAG_FD_SB_AUTO_DISABLED);
10254 		/* reset fd counters */
10255 		pf->fd_add_err = 0;
10256 		pf->fd_atr_cnt = 0;
10257 		/* if ATR was auto disabled it can be re-enabled. */
10258 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
10259 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
10260 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
10261 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
10262 				dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
10263 		}
10264 	}
10265 	return need_reset;
10266 }
10267 
10268 /**
10269  * i40e_clear_rss_lut - clear the rx hash lookup table
10270  * @vsi: the VSI being configured
10271  **/
10272 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
10273 {
10274 	struct i40e_pf *pf = vsi->back;
10275 	struct i40e_hw *hw = &pf->hw;
10276 	u16 vf_id = vsi->vf_id;
10277 	u8 i;
10278 
10279 	if (vsi->type == I40E_VSI_MAIN) {
10280 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
10281 			wr32(hw, I40E_PFQF_HLUT(i), 0);
10282 	} else if (vsi->type == I40E_VSI_SRIOV) {
10283 		for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
10284 			i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
10285 	} else {
10286 		dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
10287 	}
10288 }
10289 
10290 /**
10291  * i40e_set_features - set the netdev feature flags
10292  * @netdev: ptr to the netdev being adjusted
10293  * @features: the feature set that the stack is suggesting
10294  * Note: expects to be called while under rtnl_lock()
10295  **/
10296 static int i40e_set_features(struct net_device *netdev,
10297 			     netdev_features_t features)
10298 {
10299 	struct i40e_netdev_priv *np = netdev_priv(netdev);
10300 	struct i40e_vsi *vsi = np->vsi;
10301 	struct i40e_pf *pf = vsi->back;
10302 	bool need_reset;
10303 
10304 	if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
10305 		i40e_pf_config_rss(pf);
10306 	else if (!(features & NETIF_F_RXHASH) &&
10307 		 netdev->features & NETIF_F_RXHASH)
10308 		i40e_clear_rss_lut(vsi);
10309 
10310 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
10311 		i40e_vlan_stripping_enable(vsi);
10312 	else
10313 		i40e_vlan_stripping_disable(vsi);
10314 
10315 	need_reset = i40e_set_ntuple(pf, features);
10316 
10317 	if (need_reset)
10318 		i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
10319 
10320 	return 0;
10321 }
10322 
10323 /**
10324  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
10325  * @pf: board private structure
10326  * @port: The UDP port to look up
10327  *
10328  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
10329  **/
10330 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
10331 {
10332 	u8 i;
10333 
10334 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
10335 		if (pf->udp_ports[i].port == port)
10336 			return i;
10337 	}
10338 
10339 	return i;
10340 }
10341 
10342 /**
10343  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
10344  * @netdev: This physical port's netdev
10345  * @ti: Tunnel endpoint information
10346  **/
10347 static void i40e_udp_tunnel_add(struct net_device *netdev,
10348 				struct udp_tunnel_info *ti)
10349 {
10350 	struct i40e_netdev_priv *np = netdev_priv(netdev);
10351 	struct i40e_vsi *vsi = np->vsi;
10352 	struct i40e_pf *pf = vsi->back;
10353 	u16 port = ntohs(ti->port);
10354 	u8 next_idx;
10355 	u8 idx;
10356 
10357 	idx = i40e_get_udp_port_idx(pf, port);
10358 
10359 	/* Check if port already exists */
10360 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
10361 		netdev_info(netdev, "port %d already offloaded\n", port);
10362 		return;
10363 	}
10364 
10365 	/* Now check if there is space to add the new port */
10366 	next_idx = i40e_get_udp_port_idx(pf, 0);
10367 
10368 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
10369 		netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
10370 			    port);
10371 		return;
10372 	}
10373 
10374 	switch (ti->type) {
10375 	case UDP_TUNNEL_TYPE_VXLAN:
10376 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
10377 		break;
10378 	case UDP_TUNNEL_TYPE_GENEVE:
10379 		if (!(pf->hw_features & I40E_HW_GENEVE_OFFLOAD_CAPABLE))
10380 			return;
10381 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
10382 		break;
10383 	default:
10384 		return;
10385 	}
10386 
10387 	/* New port: add it and mark its index in the bitmap */
10388 	pf->udp_ports[next_idx].port = port;
10389 	pf->pending_udp_bitmap |= BIT_ULL(next_idx);
10390 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
10391 }
10392 
10393 /**
10394  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
10395  * @netdev: This physical port's netdev
10396  * @ti: Tunnel endpoint information
10397  **/
10398 static void i40e_udp_tunnel_del(struct net_device *netdev,
10399 				struct udp_tunnel_info *ti)
10400 {
10401 	struct i40e_netdev_priv *np = netdev_priv(netdev);
10402 	struct i40e_vsi *vsi = np->vsi;
10403 	struct i40e_pf *pf = vsi->back;
10404 	u16 port = ntohs(ti->port);
10405 	u8 idx;
10406 
10407 	idx = i40e_get_udp_port_idx(pf, port);
10408 
10409 	/* Check if port already exists */
10410 	if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
10411 		goto not_found;
10412 
10413 	switch (ti->type) {
10414 	case UDP_TUNNEL_TYPE_VXLAN:
10415 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
10416 			goto not_found;
10417 		break;
10418 	case UDP_TUNNEL_TYPE_GENEVE:
10419 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
10420 			goto not_found;
10421 		break;
10422 	default:
10423 		goto not_found;
10424 	}
10425 
10426 	/* if port exists, set it to 0 (mark for deletion)
10427 	 * and make it pending
10428 	 */
10429 	pf->udp_ports[idx].port = 0;
10430 	pf->pending_udp_bitmap |= BIT_ULL(idx);
10431 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
10432 
10433 	return;
10434 not_found:
10435 	netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
10436 		    port);
10437 }
10438 
10439 static int i40e_get_phys_port_id(struct net_device *netdev,
10440 				 struct netdev_phys_item_id *ppid)
10441 {
10442 	struct i40e_netdev_priv *np = netdev_priv(netdev);
10443 	struct i40e_pf *pf = np->vsi->back;
10444 	struct i40e_hw *hw = &pf->hw;
10445 
10446 	if (!(pf->hw_features & I40E_HW_PORT_ID_VALID))
10447 		return -EOPNOTSUPP;
10448 
10449 	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
10450 	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
10451 
10452 	return 0;
10453 }
10454 
10455 /**
10456  * i40e_ndo_fdb_add - add an entry to the hardware database
10457  * @ndm: the input from the stack
10458  * @tb: pointer to array of nladdr (unused)
10459  * @dev: the net device pointer
10460  * @addr: the MAC address entry being added
10461  * @flags: instructions from stack about fdb operation
10462  */
10463 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
10464 			    struct net_device *dev,
10465 			    const unsigned char *addr, u16 vid,
10466 			    u16 flags)
10467 {
10468 	struct i40e_netdev_priv *np = netdev_priv(dev);
10469 	struct i40e_pf *pf = np->vsi->back;
10470 	int err = 0;
10471 
10472 	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
10473 		return -EOPNOTSUPP;
10474 
10475 	if (vid) {
10476 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
10477 		return -EINVAL;
10478 	}
10479 
10480 	/* Hardware does not support aging addresses so if a
10481 	 * ndm_state is given only allow permanent addresses
10482 	 */
10483 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
10484 		netdev_info(dev, "FDB only supports static addresses\n");
10485 		return -EINVAL;
10486 	}
10487 
10488 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
10489 		err = dev_uc_add_excl(dev, addr);
10490 	else if (is_multicast_ether_addr(addr))
10491 		err = dev_mc_add_excl(dev, addr);
10492 	else
10493 		err = -EINVAL;
10494 
10495 	/* Only return duplicate errors if NLM_F_EXCL is set */
10496 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
10497 		err = 0;
10498 
10499 	return err;
10500 }
10501 
10502 /**
10503  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
10504  * @dev: the netdev being configured
10505  * @nlh: RTNL message
10506  *
10507  * Inserts a new hardware bridge if not already created and
10508  * enables the bridging mode requested (VEB or VEPA). If the
10509  * hardware bridge has already been inserted and the request
10510  * is to change the mode then that requires a PF reset to
10511  * allow rebuild of the components with required hardware
10512  * bridge mode enabled.
10513  *
10514  * Note: expects to be called while under rtnl_lock()
10515  **/
10516 static int i40e_ndo_bridge_setlink(struct net_device *dev,
10517 				   struct nlmsghdr *nlh,
10518 				   u16 flags)
10519 {
10520 	struct i40e_netdev_priv *np = netdev_priv(dev);
10521 	struct i40e_vsi *vsi = np->vsi;
10522 	struct i40e_pf *pf = vsi->back;
10523 	struct i40e_veb *veb = NULL;
10524 	struct nlattr *attr, *br_spec;
10525 	int i, rem;
10526 
10527 	/* Only for PF VSI for now */
10528 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
10529 		return -EOPNOTSUPP;
10530 
10531 	/* Find the HW bridge for PF VSI */
10532 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
10533 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
10534 			veb = pf->veb[i];
10535 	}
10536 
10537 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
10538 
10539 	nla_for_each_nested(attr, br_spec, rem) {
10540 		__u16 mode;
10541 
10542 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
10543 			continue;
10544 
10545 		mode = nla_get_u16(attr);
10546 		if ((mode != BRIDGE_MODE_VEPA) &&
10547 		    (mode != BRIDGE_MODE_VEB))
10548 			return -EINVAL;
10549 
10550 		/* Insert a new HW bridge */
10551 		if (!veb) {
10552 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
10553 					     vsi->tc_config.enabled_tc);
10554 			if (veb) {
10555 				veb->bridge_mode = mode;
10556 				i40e_config_bridge_mode(veb);
10557 			} else {
10558 				/* No Bridge HW offload available */
10559 				return -ENOENT;
10560 			}
10561 			break;
10562 		} else if (mode != veb->bridge_mode) {
10563 			/* Existing HW bridge but different mode needs reset */
10564 			veb->bridge_mode = mode;
10565 			/* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
10566 			if (mode == BRIDGE_MODE_VEB)
10567 				pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
10568 			else
10569 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
10570 			i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
10571 			break;
10572 		}
10573 	}
10574 
10575 	return 0;
10576 }
10577 
10578 /**
10579  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
10580  * @skb: skb buff
10581  * @pid: process id
10582  * @seq: RTNL message seq #
10583  * @dev: the netdev being configured
10584  * @filter_mask: unused
10585  * @nlflags: netlink flags passed in
10586  *
10587  * Return the mode in which the hardware bridge is operating in
10588  * i.e VEB or VEPA.
10589  **/
10590 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
10591 				   struct net_device *dev,
10592 				   u32 __always_unused filter_mask,
10593 				   int nlflags)
10594 {
10595 	struct i40e_netdev_priv *np = netdev_priv(dev);
10596 	struct i40e_vsi *vsi = np->vsi;
10597 	struct i40e_pf *pf = vsi->back;
10598 	struct i40e_veb *veb = NULL;
10599 	int i;
10600 
10601 	/* Only for PF VSI for now */
10602 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
10603 		return -EOPNOTSUPP;
10604 
10605 	/* Find the HW bridge for the PF VSI */
10606 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
10607 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
10608 			veb = pf->veb[i];
10609 	}
10610 
10611 	if (!veb)
10612 		return 0;
10613 
10614 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
10615 				       0, 0, nlflags, filter_mask, NULL);
10616 }
10617 
10618 /**
10619  * i40e_features_check - Validate encapsulated packet conforms to limits
10620  * @skb: skb buff
10621  * @dev: This physical port's netdev
10622  * @features: Offload features that the stack believes apply
10623  **/
10624 static netdev_features_t i40e_features_check(struct sk_buff *skb,
10625 					     struct net_device *dev,
10626 					     netdev_features_t features)
10627 {
10628 	size_t len;
10629 
10630 	/* No point in doing any of this if neither checksum nor GSO are
10631 	 * being requested for this frame.  We can rule out both by just
10632 	 * checking for CHECKSUM_PARTIAL
10633 	 */
10634 	if (skb->ip_summed != CHECKSUM_PARTIAL)
10635 		return features;
10636 
10637 	/* We cannot support GSO if the MSS is going to be less than
10638 	 * 64 bytes.  If it is then we need to drop support for GSO.
10639 	 */
10640 	if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
10641 		features &= ~NETIF_F_GSO_MASK;
10642 
10643 	/* MACLEN can support at most 63 words */
10644 	len = skb_network_header(skb) - skb->data;
10645 	if (len & ~(63 * 2))
10646 		goto out_err;
10647 
10648 	/* IPLEN and EIPLEN can support at most 127 dwords */
10649 	len = skb_transport_header(skb) - skb_network_header(skb);
10650 	if (len & ~(127 * 4))
10651 		goto out_err;
10652 
10653 	if (skb->encapsulation) {
10654 		/* L4TUNLEN can support 127 words */
10655 		len = skb_inner_network_header(skb) - skb_transport_header(skb);
10656 		if (len & ~(127 * 2))
10657 			goto out_err;
10658 
10659 		/* IPLEN can support at most 127 dwords */
10660 		len = skb_inner_transport_header(skb) -
10661 		      skb_inner_network_header(skb);
10662 		if (len & ~(127 * 4))
10663 			goto out_err;
10664 	}
10665 
10666 	/* No need to validate L4LEN as TCP is the only protocol with a
10667 	 * a flexible value and we support all possible values supported
10668 	 * by TCP, which is at most 15 dwords
10669 	 */
10670 
10671 	return features;
10672 out_err:
10673 	return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
10674 }
10675 
10676 /**
10677  * i40e_xdp_setup - add/remove an XDP program
10678  * @vsi: VSI to changed
10679  * @prog: XDP program
10680  **/
10681 static int i40e_xdp_setup(struct i40e_vsi *vsi,
10682 			  struct bpf_prog *prog)
10683 {
10684 	int frame_size = vsi->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
10685 	struct i40e_pf *pf = vsi->back;
10686 	struct bpf_prog *old_prog;
10687 	bool need_reset;
10688 	int i;
10689 
10690 	/* Don't allow frames that span over multiple buffers */
10691 	if (frame_size > vsi->rx_buf_len)
10692 		return -EINVAL;
10693 
10694 	if (!i40e_enabled_xdp_vsi(vsi) && !prog)
10695 		return 0;
10696 
10697 	/* When turning XDP on->off/off->on we reset and rebuild the rings. */
10698 	need_reset = (i40e_enabled_xdp_vsi(vsi) != !!prog);
10699 
10700 	if (need_reset)
10701 		i40e_prep_for_reset(pf, true);
10702 
10703 	old_prog = xchg(&vsi->xdp_prog, prog);
10704 
10705 	if (need_reset)
10706 		i40e_reset_and_rebuild(pf, true, true);
10707 
10708 	for (i = 0; i < vsi->num_queue_pairs; i++)
10709 		WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
10710 
10711 	if (old_prog)
10712 		bpf_prog_put(old_prog);
10713 
10714 	return 0;
10715 }
10716 
10717 /**
10718  * i40e_xdp - implements ndo_xdp for i40e
10719  * @dev: netdevice
10720  * @xdp: XDP command
10721  **/
10722 static int i40e_xdp(struct net_device *dev,
10723 		    struct netdev_xdp *xdp)
10724 {
10725 	struct i40e_netdev_priv *np = netdev_priv(dev);
10726 	struct i40e_vsi *vsi = np->vsi;
10727 
10728 	if (vsi->type != I40E_VSI_MAIN)
10729 		return -EINVAL;
10730 
10731 	switch (xdp->command) {
10732 	case XDP_SETUP_PROG:
10733 		return i40e_xdp_setup(vsi, xdp->prog);
10734 	case XDP_QUERY_PROG:
10735 		xdp->prog_attached = i40e_enabled_xdp_vsi(vsi);
10736 		xdp->prog_id = vsi->xdp_prog ? vsi->xdp_prog->aux->id : 0;
10737 		return 0;
10738 	default:
10739 		return -EINVAL;
10740 	}
10741 }
10742 
10743 static const struct net_device_ops i40e_netdev_ops = {
10744 	.ndo_open		= i40e_open,
10745 	.ndo_stop		= i40e_close,
10746 	.ndo_start_xmit		= i40e_lan_xmit_frame,
10747 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
10748 	.ndo_set_rx_mode	= i40e_set_rx_mode,
10749 	.ndo_validate_addr	= eth_validate_addr,
10750 	.ndo_set_mac_address	= i40e_set_mac,
10751 	.ndo_change_mtu		= i40e_change_mtu,
10752 	.ndo_do_ioctl		= i40e_ioctl,
10753 	.ndo_tx_timeout		= i40e_tx_timeout,
10754 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
10755 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
10756 #ifdef CONFIG_NET_POLL_CONTROLLER
10757 	.ndo_poll_controller	= i40e_netpoll,
10758 #endif
10759 	.ndo_setup_tc		= __i40e_setup_tc,
10760 	.ndo_set_features	= i40e_set_features,
10761 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
10762 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
10763 	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
10764 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
10765 	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
10766 	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
10767 	.ndo_set_vf_trust	= i40e_ndo_set_vf_trust,
10768 	.ndo_udp_tunnel_add	= i40e_udp_tunnel_add,
10769 	.ndo_udp_tunnel_del	= i40e_udp_tunnel_del,
10770 	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
10771 	.ndo_fdb_add		= i40e_ndo_fdb_add,
10772 	.ndo_features_check	= i40e_features_check,
10773 	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
10774 	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
10775 	.ndo_xdp		= i40e_xdp,
10776 };
10777 
10778 /**
10779  * i40e_config_netdev - Setup the netdev flags
10780  * @vsi: the VSI being configured
10781  *
10782  * Returns 0 on success, negative value on failure
10783  **/
10784 static int i40e_config_netdev(struct i40e_vsi *vsi)
10785 {
10786 	struct i40e_pf *pf = vsi->back;
10787 	struct i40e_hw *hw = &pf->hw;
10788 	struct i40e_netdev_priv *np;
10789 	struct net_device *netdev;
10790 	u8 broadcast[ETH_ALEN];
10791 	u8 mac_addr[ETH_ALEN];
10792 	int etherdev_size;
10793 	netdev_features_t hw_enc_features;
10794 	netdev_features_t hw_features;
10795 
10796 	etherdev_size = sizeof(struct i40e_netdev_priv);
10797 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
10798 	if (!netdev)
10799 		return -ENOMEM;
10800 
10801 	vsi->netdev = netdev;
10802 	np = netdev_priv(netdev);
10803 	np->vsi = vsi;
10804 
10805 	hw_enc_features = NETIF_F_SG			|
10806 			  NETIF_F_IP_CSUM		|
10807 			  NETIF_F_IPV6_CSUM		|
10808 			  NETIF_F_HIGHDMA		|
10809 			  NETIF_F_SOFT_FEATURES		|
10810 			  NETIF_F_TSO			|
10811 			  NETIF_F_TSO_ECN		|
10812 			  NETIF_F_TSO6			|
10813 			  NETIF_F_GSO_GRE		|
10814 			  NETIF_F_GSO_GRE_CSUM		|
10815 			  NETIF_F_GSO_PARTIAL		|
10816 			  NETIF_F_GSO_UDP_TUNNEL	|
10817 			  NETIF_F_GSO_UDP_TUNNEL_CSUM	|
10818 			  NETIF_F_SCTP_CRC		|
10819 			  NETIF_F_RXHASH		|
10820 			  NETIF_F_RXCSUM		|
10821 			  0;
10822 
10823 	if (!(pf->hw_features & I40E_HW_OUTER_UDP_CSUM_CAPABLE))
10824 		netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
10825 
10826 	netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
10827 
10828 	netdev->hw_enc_features |= hw_enc_features;
10829 
10830 	/* record features VLANs can make use of */
10831 	netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
10832 
10833 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
10834 		netdev->hw_features |= NETIF_F_NTUPLE;
10835 	hw_features = hw_enc_features		|
10836 		      NETIF_F_HW_VLAN_CTAG_TX	|
10837 		      NETIF_F_HW_VLAN_CTAG_RX;
10838 
10839 	netdev->hw_features |= hw_features;
10840 
10841 	netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
10842 	netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
10843 
10844 	if (vsi->type == I40E_VSI_MAIN) {
10845 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
10846 		ether_addr_copy(mac_addr, hw->mac.perm_addr);
10847 		/* The following steps are necessary for two reasons. First,
10848 		 * some older NVM configurations load a default MAC-VLAN
10849 		 * filter that will accept any tagged packet, and we want to
10850 		 * replace this with a normal filter. Additionally, it is
10851 		 * possible our MAC address was provided by the platform using
10852 		 * Open Firmware or similar.
10853 		 *
10854 		 * Thus, we need to remove the default filter and install one
10855 		 * specific to the MAC address.
10856 		 */
10857 		i40e_rm_default_mac_filter(vsi, mac_addr);
10858 		spin_lock_bh(&vsi->mac_filter_hash_lock);
10859 		i40e_add_mac_filter(vsi, mac_addr);
10860 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
10861 	} else {
10862 		/* Relate the VSI_VMDQ name to the VSI_MAIN name. Note that we
10863 		 * are still limited by IFNAMSIZ, but we're adding 'v%d\0' to
10864 		 * the end, which is 4 bytes long, so force truncation of the
10865 		 * original name by IFNAMSIZ - 4
10866 		 */
10867 		snprintf(netdev->name, IFNAMSIZ, "%.*sv%%d",
10868 			 IFNAMSIZ - 4,
10869 			 pf->vsi[pf->lan_vsi]->netdev->name);
10870 		random_ether_addr(mac_addr);
10871 
10872 		spin_lock_bh(&vsi->mac_filter_hash_lock);
10873 		i40e_add_mac_filter(vsi, mac_addr);
10874 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
10875 	}
10876 
10877 	/* Add the broadcast filter so that we initially will receive
10878 	 * broadcast packets. Note that when a new VLAN is first added the
10879 	 * driver will convert all filters marked I40E_VLAN_ANY into VLAN
10880 	 * specific filters as part of transitioning into "vlan" operation.
10881 	 * When more VLANs are added, the driver will copy each existing MAC
10882 	 * filter and add it for the new VLAN.
10883 	 *
10884 	 * Broadcast filters are handled specially by
10885 	 * i40e_sync_filters_subtask, as the driver must to set the broadcast
10886 	 * promiscuous bit instead of adding this directly as a MAC/VLAN
10887 	 * filter. The subtask will update the correct broadcast promiscuous
10888 	 * bits as VLANs become active or inactive.
10889 	 */
10890 	eth_broadcast_addr(broadcast);
10891 	spin_lock_bh(&vsi->mac_filter_hash_lock);
10892 	i40e_add_mac_filter(vsi, broadcast);
10893 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
10894 
10895 	ether_addr_copy(netdev->dev_addr, mac_addr);
10896 	ether_addr_copy(netdev->perm_addr, mac_addr);
10897 
10898 	netdev->priv_flags |= IFF_UNICAST_FLT;
10899 	netdev->priv_flags |= IFF_SUPP_NOFCS;
10900 	/* Setup netdev TC information */
10901 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
10902 
10903 	netdev->netdev_ops = &i40e_netdev_ops;
10904 	netdev->watchdog_timeo = 5 * HZ;
10905 	i40e_set_ethtool_ops(netdev);
10906 
10907 	/* MTU range: 68 - 9706 */
10908 	netdev->min_mtu = ETH_MIN_MTU;
10909 	netdev->max_mtu = I40E_MAX_RXBUFFER - I40E_PACKET_HDR_PAD;
10910 
10911 	return 0;
10912 }
10913 
10914 /**
10915  * i40e_vsi_delete - Delete a VSI from the switch
10916  * @vsi: the VSI being removed
10917  *
10918  * Returns 0 on success, negative value on failure
10919  **/
10920 static void i40e_vsi_delete(struct i40e_vsi *vsi)
10921 {
10922 	/* remove default VSI is not allowed */
10923 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
10924 		return;
10925 
10926 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
10927 }
10928 
10929 /**
10930  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
10931  * @vsi: the VSI being queried
10932  *
10933  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
10934  **/
10935 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
10936 {
10937 	struct i40e_veb *veb;
10938 	struct i40e_pf *pf = vsi->back;
10939 
10940 	/* Uplink is not a bridge so default to VEB */
10941 	if (vsi->veb_idx == I40E_NO_VEB)
10942 		return 1;
10943 
10944 	veb = pf->veb[vsi->veb_idx];
10945 	if (!veb) {
10946 		dev_info(&pf->pdev->dev,
10947 			 "There is no veb associated with the bridge\n");
10948 		return -ENOENT;
10949 	}
10950 
10951 	/* Uplink is a bridge in VEPA mode */
10952 	if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
10953 		return 0;
10954 	} else {
10955 		/* Uplink is a bridge in VEB mode */
10956 		return 1;
10957 	}
10958 
10959 	/* VEPA is now default bridge, so return 0 */
10960 	return 0;
10961 }
10962 
10963 /**
10964  * i40e_add_vsi - Add a VSI to the switch
10965  * @vsi: the VSI being configured
10966  *
10967  * This initializes a VSI context depending on the VSI type to be added and
10968  * passes it down to the add_vsi aq command.
10969  **/
10970 static int i40e_add_vsi(struct i40e_vsi *vsi)
10971 {
10972 	int ret = -ENODEV;
10973 	struct i40e_pf *pf = vsi->back;
10974 	struct i40e_hw *hw = &pf->hw;
10975 	struct i40e_vsi_context ctxt;
10976 	struct i40e_mac_filter *f;
10977 	struct hlist_node *h;
10978 	int bkt;
10979 
10980 	u8 enabled_tc = 0x1; /* TC0 enabled */
10981 	int f_count = 0;
10982 
10983 	memset(&ctxt, 0, sizeof(ctxt));
10984 	switch (vsi->type) {
10985 	case I40E_VSI_MAIN:
10986 		/* The PF's main VSI is already setup as part of the
10987 		 * device initialization, so we'll not bother with
10988 		 * the add_vsi call, but we will retrieve the current
10989 		 * VSI context.
10990 		 */
10991 		ctxt.seid = pf->main_vsi_seid;
10992 		ctxt.pf_num = pf->hw.pf_id;
10993 		ctxt.vf_num = 0;
10994 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
10995 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
10996 		if (ret) {
10997 			dev_info(&pf->pdev->dev,
10998 				 "couldn't get PF vsi config, err %s aq_err %s\n",
10999 				 i40e_stat_str(&pf->hw, ret),
11000 				 i40e_aq_str(&pf->hw,
11001 					     pf->hw.aq.asq_last_status));
11002 			return -ENOENT;
11003 		}
11004 		vsi->info = ctxt.info;
11005 		vsi->info.valid_sections = 0;
11006 
11007 		vsi->seid = ctxt.seid;
11008 		vsi->id = ctxt.vsi_number;
11009 
11010 		enabled_tc = i40e_pf_get_tc_map(pf);
11011 
11012 		/* Source pruning is enabled by default, so the flag is
11013 		 * negative logic - if it's set, we need to fiddle with
11014 		 * the VSI to disable source pruning.
11015 		 */
11016 		if (pf->flags & I40E_FLAG_SOURCE_PRUNING_DISABLED) {
11017 			memset(&ctxt, 0, sizeof(ctxt));
11018 			ctxt.seid = pf->main_vsi_seid;
11019 			ctxt.pf_num = pf->hw.pf_id;
11020 			ctxt.vf_num = 0;
11021 			ctxt.info.valid_sections |=
11022 				     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
11023 			ctxt.info.switch_id =
11024 				   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
11025 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
11026 			if (ret) {
11027 				dev_info(&pf->pdev->dev,
11028 					 "update vsi failed, err %s aq_err %s\n",
11029 					 i40e_stat_str(&pf->hw, ret),
11030 					 i40e_aq_str(&pf->hw,
11031 						     pf->hw.aq.asq_last_status));
11032 				ret = -ENOENT;
11033 				goto err;
11034 			}
11035 		}
11036 
11037 		/* MFP mode setup queue map and update VSI */
11038 		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
11039 		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
11040 			memset(&ctxt, 0, sizeof(ctxt));
11041 			ctxt.seid = pf->main_vsi_seid;
11042 			ctxt.pf_num = pf->hw.pf_id;
11043 			ctxt.vf_num = 0;
11044 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
11045 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
11046 			if (ret) {
11047 				dev_info(&pf->pdev->dev,
11048 					 "update vsi failed, err %s aq_err %s\n",
11049 					 i40e_stat_str(&pf->hw, ret),
11050 					 i40e_aq_str(&pf->hw,
11051 						    pf->hw.aq.asq_last_status));
11052 				ret = -ENOENT;
11053 				goto err;
11054 			}
11055 			/* update the local VSI info queue map */
11056 			i40e_vsi_update_queue_map(vsi, &ctxt);
11057 			vsi->info.valid_sections = 0;
11058 		} else {
11059 			/* Default/Main VSI is only enabled for TC0
11060 			 * reconfigure it to enable all TCs that are
11061 			 * available on the port in SFP mode.
11062 			 * For MFP case the iSCSI PF would use this
11063 			 * flow to enable LAN+iSCSI TC.
11064 			 */
11065 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
11066 			if (ret) {
11067 				/* Single TC condition is not fatal,
11068 				 * message and continue
11069 				 */
11070 				dev_info(&pf->pdev->dev,
11071 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
11072 					 enabled_tc,
11073 					 i40e_stat_str(&pf->hw, ret),
11074 					 i40e_aq_str(&pf->hw,
11075 						    pf->hw.aq.asq_last_status));
11076 			}
11077 		}
11078 		break;
11079 
11080 	case I40E_VSI_FDIR:
11081 		ctxt.pf_num = hw->pf_id;
11082 		ctxt.vf_num = 0;
11083 		ctxt.uplink_seid = vsi->uplink_seid;
11084 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
11085 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
11086 		if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
11087 		    (i40e_is_vsi_uplink_mode_veb(vsi))) {
11088 			ctxt.info.valid_sections |=
11089 			     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
11090 			ctxt.info.switch_id =
11091 			   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
11092 		}
11093 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
11094 		break;
11095 
11096 	case I40E_VSI_VMDQ2:
11097 		ctxt.pf_num = hw->pf_id;
11098 		ctxt.vf_num = 0;
11099 		ctxt.uplink_seid = vsi->uplink_seid;
11100 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
11101 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
11102 
11103 		/* This VSI is connected to VEB so the switch_id
11104 		 * should be set to zero by default.
11105 		 */
11106 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
11107 			ctxt.info.valid_sections |=
11108 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
11109 			ctxt.info.switch_id =
11110 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
11111 		}
11112 
11113 		/* Setup the VSI tx/rx queue map for TC0 only for now */
11114 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
11115 		break;
11116 
11117 	case I40E_VSI_SRIOV:
11118 		ctxt.pf_num = hw->pf_id;
11119 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
11120 		ctxt.uplink_seid = vsi->uplink_seid;
11121 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
11122 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
11123 
11124 		/* This VSI is connected to VEB so the switch_id
11125 		 * should be set to zero by default.
11126 		 */
11127 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
11128 			ctxt.info.valid_sections |=
11129 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
11130 			ctxt.info.switch_id =
11131 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
11132 		}
11133 
11134 		if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
11135 			ctxt.info.valid_sections |=
11136 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
11137 			ctxt.info.queueing_opt_flags |=
11138 				(I40E_AQ_VSI_QUE_OPT_TCP_ENA |
11139 				 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
11140 		}
11141 
11142 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
11143 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
11144 		if (pf->vf[vsi->vf_id].spoofchk) {
11145 			ctxt.info.valid_sections |=
11146 				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
11147 			ctxt.info.sec_flags |=
11148 				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
11149 				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
11150 		}
11151 		/* Setup the VSI tx/rx queue map for TC0 only for now */
11152 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
11153 		break;
11154 
11155 	case I40E_VSI_IWARP:
11156 		/* send down message to iWARP */
11157 		break;
11158 
11159 	default:
11160 		return -ENODEV;
11161 	}
11162 
11163 	if (vsi->type != I40E_VSI_MAIN) {
11164 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
11165 		if (ret) {
11166 			dev_info(&vsi->back->pdev->dev,
11167 				 "add vsi failed, err %s aq_err %s\n",
11168 				 i40e_stat_str(&pf->hw, ret),
11169 				 i40e_aq_str(&pf->hw,
11170 					     pf->hw.aq.asq_last_status));
11171 			ret = -ENOENT;
11172 			goto err;
11173 		}
11174 		vsi->info = ctxt.info;
11175 		vsi->info.valid_sections = 0;
11176 		vsi->seid = ctxt.seid;
11177 		vsi->id = ctxt.vsi_number;
11178 	}
11179 
11180 	vsi->active_filters = 0;
11181 	clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
11182 	spin_lock_bh(&vsi->mac_filter_hash_lock);
11183 	/* If macvlan filters already exist, force them to get loaded */
11184 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
11185 		f->state = I40E_FILTER_NEW;
11186 		f_count++;
11187 	}
11188 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
11189 
11190 	if (f_count) {
11191 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
11192 		pf->flags |= I40E_FLAG_FILTER_SYNC;
11193 	}
11194 
11195 	/* Update VSI BW information */
11196 	ret = i40e_vsi_get_bw_info(vsi);
11197 	if (ret) {
11198 		dev_info(&pf->pdev->dev,
11199 			 "couldn't get vsi bw info, err %s aq_err %s\n",
11200 			 i40e_stat_str(&pf->hw, ret),
11201 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11202 		/* VSI is already added so not tearing that up */
11203 		ret = 0;
11204 	}
11205 
11206 err:
11207 	return ret;
11208 }
11209 
11210 /**
11211  * i40e_vsi_release - Delete a VSI and free its resources
11212  * @vsi: the VSI being removed
11213  *
11214  * Returns 0 on success or < 0 on error
11215  **/
11216 int i40e_vsi_release(struct i40e_vsi *vsi)
11217 {
11218 	struct i40e_mac_filter *f;
11219 	struct hlist_node *h;
11220 	struct i40e_veb *veb = NULL;
11221 	struct i40e_pf *pf;
11222 	u16 uplink_seid;
11223 	int i, n, bkt;
11224 
11225 	pf = vsi->back;
11226 
11227 	/* release of a VEB-owner or last VSI is not allowed */
11228 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
11229 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
11230 			 vsi->seid, vsi->uplink_seid);
11231 		return -ENODEV;
11232 	}
11233 	if (vsi == pf->vsi[pf->lan_vsi] &&
11234 	    !test_bit(__I40E_DOWN, pf->state)) {
11235 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
11236 		return -ENODEV;
11237 	}
11238 
11239 	uplink_seid = vsi->uplink_seid;
11240 	if (vsi->type != I40E_VSI_SRIOV) {
11241 		if (vsi->netdev_registered) {
11242 			vsi->netdev_registered = false;
11243 			if (vsi->netdev) {
11244 				/* results in a call to i40e_close() */
11245 				unregister_netdev(vsi->netdev);
11246 			}
11247 		} else {
11248 			i40e_vsi_close(vsi);
11249 		}
11250 		i40e_vsi_disable_irq(vsi);
11251 	}
11252 
11253 	spin_lock_bh(&vsi->mac_filter_hash_lock);
11254 
11255 	/* clear the sync flag on all filters */
11256 	if (vsi->netdev) {
11257 		__dev_uc_unsync(vsi->netdev, NULL);
11258 		__dev_mc_unsync(vsi->netdev, NULL);
11259 	}
11260 
11261 	/* make sure any remaining filters are marked for deletion */
11262 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
11263 		__i40e_del_filter(vsi, f);
11264 
11265 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
11266 
11267 	i40e_sync_vsi_filters(vsi);
11268 
11269 	i40e_vsi_delete(vsi);
11270 	i40e_vsi_free_q_vectors(vsi);
11271 	if (vsi->netdev) {
11272 		free_netdev(vsi->netdev);
11273 		vsi->netdev = NULL;
11274 	}
11275 	i40e_vsi_clear_rings(vsi);
11276 	i40e_vsi_clear(vsi);
11277 
11278 	/* If this was the last thing on the VEB, except for the
11279 	 * controlling VSI, remove the VEB, which puts the controlling
11280 	 * VSI onto the next level down in the switch.
11281 	 *
11282 	 * Well, okay, there's one more exception here: don't remove
11283 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
11284 	 * from up the network stack.
11285 	 */
11286 	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
11287 		if (pf->vsi[i] &&
11288 		    pf->vsi[i]->uplink_seid == uplink_seid &&
11289 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
11290 			n++;      /* count the VSIs */
11291 		}
11292 	}
11293 	for (i = 0; i < I40E_MAX_VEB; i++) {
11294 		if (!pf->veb[i])
11295 			continue;
11296 		if (pf->veb[i]->uplink_seid == uplink_seid)
11297 			n++;     /* count the VEBs */
11298 		if (pf->veb[i]->seid == uplink_seid)
11299 			veb = pf->veb[i];
11300 	}
11301 	if (n == 0 && veb && veb->uplink_seid != 0)
11302 		i40e_veb_release(veb);
11303 
11304 	return 0;
11305 }
11306 
11307 /**
11308  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
11309  * @vsi: ptr to the VSI
11310  *
11311  * This should only be called after i40e_vsi_mem_alloc() which allocates the
11312  * corresponding SW VSI structure and initializes num_queue_pairs for the
11313  * newly allocated VSI.
11314  *
11315  * Returns 0 on success or negative on failure
11316  **/
11317 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
11318 {
11319 	int ret = -ENOENT;
11320 	struct i40e_pf *pf = vsi->back;
11321 
11322 	if (vsi->q_vectors[0]) {
11323 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
11324 			 vsi->seid);
11325 		return -EEXIST;
11326 	}
11327 
11328 	if (vsi->base_vector) {
11329 		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
11330 			 vsi->seid, vsi->base_vector);
11331 		return -EEXIST;
11332 	}
11333 
11334 	ret = i40e_vsi_alloc_q_vectors(vsi);
11335 	if (ret) {
11336 		dev_info(&pf->pdev->dev,
11337 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
11338 			 vsi->num_q_vectors, vsi->seid, ret);
11339 		vsi->num_q_vectors = 0;
11340 		goto vector_setup_out;
11341 	}
11342 
11343 	/* In Legacy mode, we do not have to get any other vector since we
11344 	 * piggyback on the misc/ICR0 for queue interrupts.
11345 	*/
11346 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
11347 		return ret;
11348 	if (vsi->num_q_vectors)
11349 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
11350 						 vsi->num_q_vectors, vsi->idx);
11351 	if (vsi->base_vector < 0) {
11352 		dev_info(&pf->pdev->dev,
11353 			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
11354 			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
11355 		i40e_vsi_free_q_vectors(vsi);
11356 		ret = -ENOENT;
11357 		goto vector_setup_out;
11358 	}
11359 
11360 vector_setup_out:
11361 	return ret;
11362 }
11363 
11364 /**
11365  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
11366  * @vsi: pointer to the vsi.
11367  *
11368  * This re-allocates a vsi's queue resources.
11369  *
11370  * Returns pointer to the successfully allocated and configured VSI sw struct
11371  * on success, otherwise returns NULL on failure.
11372  **/
11373 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
11374 {
11375 	u16 alloc_queue_pairs;
11376 	struct i40e_pf *pf;
11377 	u8 enabled_tc;
11378 	int ret;
11379 
11380 	if (!vsi)
11381 		return NULL;
11382 
11383 	pf = vsi->back;
11384 
11385 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
11386 	i40e_vsi_clear_rings(vsi);
11387 
11388 	i40e_vsi_free_arrays(vsi, false);
11389 	i40e_set_num_rings_in_vsi(vsi);
11390 	ret = i40e_vsi_alloc_arrays(vsi, false);
11391 	if (ret)
11392 		goto err_vsi;
11393 
11394 	alloc_queue_pairs = vsi->alloc_queue_pairs *
11395 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
11396 
11397 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
11398 	if (ret < 0) {
11399 		dev_info(&pf->pdev->dev,
11400 			 "failed to get tracking for %d queues for VSI %d err %d\n",
11401 			 alloc_queue_pairs, vsi->seid, ret);
11402 		goto err_vsi;
11403 	}
11404 	vsi->base_queue = ret;
11405 
11406 	/* Update the FW view of the VSI. Force a reset of TC and queue
11407 	 * layout configurations.
11408 	 */
11409 	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
11410 	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
11411 	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
11412 	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
11413 	if (vsi->type == I40E_VSI_MAIN)
11414 		i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
11415 
11416 	/* assign it some queues */
11417 	ret = i40e_alloc_rings(vsi);
11418 	if (ret)
11419 		goto err_rings;
11420 
11421 	/* map all of the rings to the q_vectors */
11422 	i40e_vsi_map_rings_to_vectors(vsi);
11423 	return vsi;
11424 
11425 err_rings:
11426 	i40e_vsi_free_q_vectors(vsi);
11427 	if (vsi->netdev_registered) {
11428 		vsi->netdev_registered = false;
11429 		unregister_netdev(vsi->netdev);
11430 		free_netdev(vsi->netdev);
11431 		vsi->netdev = NULL;
11432 	}
11433 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
11434 err_vsi:
11435 	i40e_vsi_clear(vsi);
11436 	return NULL;
11437 }
11438 
11439 /**
11440  * i40e_vsi_setup - Set up a VSI by a given type
11441  * @pf: board private structure
11442  * @type: VSI type
11443  * @uplink_seid: the switch element to link to
11444  * @param1: usage depends upon VSI type. For VF types, indicates VF id
11445  *
11446  * This allocates the sw VSI structure and its queue resources, then add a VSI
11447  * to the identified VEB.
11448  *
11449  * Returns pointer to the successfully allocated and configure VSI sw struct on
11450  * success, otherwise returns NULL on failure.
11451  **/
11452 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
11453 				u16 uplink_seid, u32 param1)
11454 {
11455 	struct i40e_vsi *vsi = NULL;
11456 	struct i40e_veb *veb = NULL;
11457 	u16 alloc_queue_pairs;
11458 	int ret, i;
11459 	int v_idx;
11460 
11461 	/* The requested uplink_seid must be either
11462 	 *     - the PF's port seid
11463 	 *              no VEB is needed because this is the PF
11464 	 *              or this is a Flow Director special case VSI
11465 	 *     - seid of an existing VEB
11466 	 *     - seid of a VSI that owns an existing VEB
11467 	 *     - seid of a VSI that doesn't own a VEB
11468 	 *              a new VEB is created and the VSI becomes the owner
11469 	 *     - seid of the PF VSI, which is what creates the first VEB
11470 	 *              this is a special case of the previous
11471 	 *
11472 	 * Find which uplink_seid we were given and create a new VEB if needed
11473 	 */
11474 	for (i = 0; i < I40E_MAX_VEB; i++) {
11475 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
11476 			veb = pf->veb[i];
11477 			break;
11478 		}
11479 	}
11480 
11481 	if (!veb && uplink_seid != pf->mac_seid) {
11482 
11483 		for (i = 0; i < pf->num_alloc_vsi; i++) {
11484 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
11485 				vsi = pf->vsi[i];
11486 				break;
11487 			}
11488 		}
11489 		if (!vsi) {
11490 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
11491 				 uplink_seid);
11492 			return NULL;
11493 		}
11494 
11495 		if (vsi->uplink_seid == pf->mac_seid)
11496 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
11497 					     vsi->tc_config.enabled_tc);
11498 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
11499 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
11500 					     vsi->tc_config.enabled_tc);
11501 		if (veb) {
11502 			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
11503 				dev_info(&vsi->back->pdev->dev,
11504 					 "New VSI creation error, uplink seid of LAN VSI expected.\n");
11505 				return NULL;
11506 			}
11507 			/* We come up by default in VEPA mode if SRIOV is not
11508 			 * already enabled, in which case we can't force VEPA
11509 			 * mode.
11510 			 */
11511 			if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
11512 				veb->bridge_mode = BRIDGE_MODE_VEPA;
11513 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
11514 			}
11515 			i40e_config_bridge_mode(veb);
11516 		}
11517 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
11518 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
11519 				veb = pf->veb[i];
11520 		}
11521 		if (!veb) {
11522 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
11523 			return NULL;
11524 		}
11525 
11526 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
11527 		uplink_seid = veb->seid;
11528 	}
11529 
11530 	/* get vsi sw struct */
11531 	v_idx = i40e_vsi_mem_alloc(pf, type);
11532 	if (v_idx < 0)
11533 		goto err_alloc;
11534 	vsi = pf->vsi[v_idx];
11535 	if (!vsi)
11536 		goto err_alloc;
11537 	vsi->type = type;
11538 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
11539 
11540 	if (type == I40E_VSI_MAIN)
11541 		pf->lan_vsi = v_idx;
11542 	else if (type == I40E_VSI_SRIOV)
11543 		vsi->vf_id = param1;
11544 	/* assign it some queues */
11545 	alloc_queue_pairs = vsi->alloc_queue_pairs *
11546 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
11547 
11548 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
11549 	if (ret < 0) {
11550 		dev_info(&pf->pdev->dev,
11551 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
11552 			 alloc_queue_pairs, vsi->seid, ret);
11553 		goto err_vsi;
11554 	}
11555 	vsi->base_queue = ret;
11556 
11557 	/* get a VSI from the hardware */
11558 	vsi->uplink_seid = uplink_seid;
11559 	ret = i40e_add_vsi(vsi);
11560 	if (ret)
11561 		goto err_vsi;
11562 
11563 	switch (vsi->type) {
11564 	/* setup the netdev if needed */
11565 	case I40E_VSI_MAIN:
11566 	case I40E_VSI_VMDQ2:
11567 		ret = i40e_config_netdev(vsi);
11568 		if (ret)
11569 			goto err_netdev;
11570 		ret = register_netdev(vsi->netdev);
11571 		if (ret)
11572 			goto err_netdev;
11573 		vsi->netdev_registered = true;
11574 		netif_carrier_off(vsi->netdev);
11575 #ifdef CONFIG_I40E_DCB
11576 		/* Setup DCB netlink interface */
11577 		i40e_dcbnl_setup(vsi);
11578 #endif /* CONFIG_I40E_DCB */
11579 		/* fall through */
11580 
11581 	case I40E_VSI_FDIR:
11582 		/* set up vectors and rings if needed */
11583 		ret = i40e_vsi_setup_vectors(vsi);
11584 		if (ret)
11585 			goto err_msix;
11586 
11587 		ret = i40e_alloc_rings(vsi);
11588 		if (ret)
11589 			goto err_rings;
11590 
11591 		/* map all of the rings to the q_vectors */
11592 		i40e_vsi_map_rings_to_vectors(vsi);
11593 
11594 		i40e_vsi_reset_stats(vsi);
11595 		break;
11596 
11597 	default:
11598 		/* no netdev or rings for the other VSI types */
11599 		break;
11600 	}
11601 
11602 	if ((pf->hw_features & I40E_HW_RSS_AQ_CAPABLE) &&
11603 	    (vsi->type == I40E_VSI_VMDQ2)) {
11604 		ret = i40e_vsi_config_rss(vsi);
11605 	}
11606 	return vsi;
11607 
11608 err_rings:
11609 	i40e_vsi_free_q_vectors(vsi);
11610 err_msix:
11611 	if (vsi->netdev_registered) {
11612 		vsi->netdev_registered = false;
11613 		unregister_netdev(vsi->netdev);
11614 		free_netdev(vsi->netdev);
11615 		vsi->netdev = NULL;
11616 	}
11617 err_netdev:
11618 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
11619 err_vsi:
11620 	i40e_vsi_clear(vsi);
11621 err_alloc:
11622 	return NULL;
11623 }
11624 
11625 /**
11626  * i40e_veb_get_bw_info - Query VEB BW information
11627  * @veb: the veb to query
11628  *
11629  * Query the Tx scheduler BW configuration data for given VEB
11630  **/
11631 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
11632 {
11633 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
11634 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
11635 	struct i40e_pf *pf = veb->pf;
11636 	struct i40e_hw *hw = &pf->hw;
11637 	u32 tc_bw_max;
11638 	int ret = 0;
11639 	int i;
11640 
11641 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
11642 						  &bw_data, NULL);
11643 	if (ret) {
11644 		dev_info(&pf->pdev->dev,
11645 			 "query veb bw config failed, err %s aq_err %s\n",
11646 			 i40e_stat_str(&pf->hw, ret),
11647 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
11648 		goto out;
11649 	}
11650 
11651 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
11652 						   &ets_data, NULL);
11653 	if (ret) {
11654 		dev_info(&pf->pdev->dev,
11655 			 "query veb bw ets config failed, err %s aq_err %s\n",
11656 			 i40e_stat_str(&pf->hw, ret),
11657 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
11658 		goto out;
11659 	}
11660 
11661 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
11662 	veb->bw_max_quanta = ets_data.tc_bw_max;
11663 	veb->is_abs_credits = bw_data.absolute_credits_enable;
11664 	veb->enabled_tc = ets_data.tc_valid_bits;
11665 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
11666 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
11667 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
11668 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
11669 		veb->bw_tc_limit_credits[i] =
11670 					le16_to_cpu(bw_data.tc_bw_limits[i]);
11671 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
11672 	}
11673 
11674 out:
11675 	return ret;
11676 }
11677 
11678 /**
11679  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
11680  * @pf: board private structure
11681  *
11682  * On error: returns error code (negative)
11683  * On success: returns vsi index in PF (positive)
11684  **/
11685 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
11686 {
11687 	int ret = -ENOENT;
11688 	struct i40e_veb *veb;
11689 	int i;
11690 
11691 	/* Need to protect the allocation of switch elements at the PF level */
11692 	mutex_lock(&pf->switch_mutex);
11693 
11694 	/* VEB list may be fragmented if VEB creation/destruction has
11695 	 * been happening.  We can afford to do a quick scan to look
11696 	 * for any free slots in the list.
11697 	 *
11698 	 * find next empty veb slot, looping back around if necessary
11699 	 */
11700 	i = 0;
11701 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
11702 		i++;
11703 	if (i >= I40E_MAX_VEB) {
11704 		ret = -ENOMEM;
11705 		goto err_alloc_veb;  /* out of VEB slots! */
11706 	}
11707 
11708 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
11709 	if (!veb) {
11710 		ret = -ENOMEM;
11711 		goto err_alloc_veb;
11712 	}
11713 	veb->pf = pf;
11714 	veb->idx = i;
11715 	veb->enabled_tc = 1;
11716 
11717 	pf->veb[i] = veb;
11718 	ret = i;
11719 err_alloc_veb:
11720 	mutex_unlock(&pf->switch_mutex);
11721 	return ret;
11722 }
11723 
11724 /**
11725  * i40e_switch_branch_release - Delete a branch of the switch tree
11726  * @branch: where to start deleting
11727  *
11728  * This uses recursion to find the tips of the branch to be
11729  * removed, deleting until we get back to and can delete this VEB.
11730  **/
11731 static void i40e_switch_branch_release(struct i40e_veb *branch)
11732 {
11733 	struct i40e_pf *pf = branch->pf;
11734 	u16 branch_seid = branch->seid;
11735 	u16 veb_idx = branch->idx;
11736 	int i;
11737 
11738 	/* release any VEBs on this VEB - RECURSION */
11739 	for (i = 0; i < I40E_MAX_VEB; i++) {
11740 		if (!pf->veb[i])
11741 			continue;
11742 		if (pf->veb[i]->uplink_seid == branch->seid)
11743 			i40e_switch_branch_release(pf->veb[i]);
11744 	}
11745 
11746 	/* Release the VSIs on this VEB, but not the owner VSI.
11747 	 *
11748 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
11749 	 *       the VEB itself, so don't use (*branch) after this loop.
11750 	 */
11751 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11752 		if (!pf->vsi[i])
11753 			continue;
11754 		if (pf->vsi[i]->uplink_seid == branch_seid &&
11755 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
11756 			i40e_vsi_release(pf->vsi[i]);
11757 		}
11758 	}
11759 
11760 	/* There's one corner case where the VEB might not have been
11761 	 * removed, so double check it here and remove it if needed.
11762 	 * This case happens if the veb was created from the debugfs
11763 	 * commands and no VSIs were added to it.
11764 	 */
11765 	if (pf->veb[veb_idx])
11766 		i40e_veb_release(pf->veb[veb_idx]);
11767 }
11768 
11769 /**
11770  * i40e_veb_clear - remove veb struct
11771  * @veb: the veb to remove
11772  **/
11773 static void i40e_veb_clear(struct i40e_veb *veb)
11774 {
11775 	if (!veb)
11776 		return;
11777 
11778 	if (veb->pf) {
11779 		struct i40e_pf *pf = veb->pf;
11780 
11781 		mutex_lock(&pf->switch_mutex);
11782 		if (pf->veb[veb->idx] == veb)
11783 			pf->veb[veb->idx] = NULL;
11784 		mutex_unlock(&pf->switch_mutex);
11785 	}
11786 
11787 	kfree(veb);
11788 }
11789 
11790 /**
11791  * i40e_veb_release - Delete a VEB and free its resources
11792  * @veb: the VEB being removed
11793  **/
11794 void i40e_veb_release(struct i40e_veb *veb)
11795 {
11796 	struct i40e_vsi *vsi = NULL;
11797 	struct i40e_pf *pf;
11798 	int i, n = 0;
11799 
11800 	pf = veb->pf;
11801 
11802 	/* find the remaining VSI and check for extras */
11803 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11804 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
11805 			n++;
11806 			vsi = pf->vsi[i];
11807 		}
11808 	}
11809 	if (n != 1) {
11810 		dev_info(&pf->pdev->dev,
11811 			 "can't remove VEB %d with %d VSIs left\n",
11812 			 veb->seid, n);
11813 		return;
11814 	}
11815 
11816 	/* move the remaining VSI to uplink veb */
11817 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
11818 	if (veb->uplink_seid) {
11819 		vsi->uplink_seid = veb->uplink_seid;
11820 		if (veb->uplink_seid == pf->mac_seid)
11821 			vsi->veb_idx = I40E_NO_VEB;
11822 		else
11823 			vsi->veb_idx = veb->veb_idx;
11824 	} else {
11825 		/* floating VEB */
11826 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
11827 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
11828 	}
11829 
11830 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
11831 	i40e_veb_clear(veb);
11832 }
11833 
11834 /**
11835  * i40e_add_veb - create the VEB in the switch
11836  * @veb: the VEB to be instantiated
11837  * @vsi: the controlling VSI
11838  **/
11839 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
11840 {
11841 	struct i40e_pf *pf = veb->pf;
11842 	bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
11843 	int ret;
11844 
11845 	ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
11846 			      veb->enabled_tc, false,
11847 			      &veb->seid, enable_stats, NULL);
11848 
11849 	/* get a VEB from the hardware */
11850 	if (ret) {
11851 		dev_info(&pf->pdev->dev,
11852 			 "couldn't add VEB, err %s aq_err %s\n",
11853 			 i40e_stat_str(&pf->hw, ret),
11854 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11855 		return -EPERM;
11856 	}
11857 
11858 	/* get statistics counter */
11859 	ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
11860 					 &veb->stats_idx, NULL, NULL, NULL);
11861 	if (ret) {
11862 		dev_info(&pf->pdev->dev,
11863 			 "couldn't get VEB statistics idx, err %s aq_err %s\n",
11864 			 i40e_stat_str(&pf->hw, ret),
11865 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11866 		return -EPERM;
11867 	}
11868 	ret = i40e_veb_get_bw_info(veb);
11869 	if (ret) {
11870 		dev_info(&pf->pdev->dev,
11871 			 "couldn't get VEB bw info, err %s aq_err %s\n",
11872 			 i40e_stat_str(&pf->hw, ret),
11873 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11874 		i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
11875 		return -ENOENT;
11876 	}
11877 
11878 	vsi->uplink_seid = veb->seid;
11879 	vsi->veb_idx = veb->idx;
11880 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
11881 
11882 	return 0;
11883 }
11884 
11885 /**
11886  * i40e_veb_setup - Set up a VEB
11887  * @pf: board private structure
11888  * @flags: VEB setup flags
11889  * @uplink_seid: the switch element to link to
11890  * @vsi_seid: the initial VSI seid
11891  * @enabled_tc: Enabled TC bit-map
11892  *
11893  * This allocates the sw VEB structure and links it into the switch
11894  * It is possible and legal for this to be a duplicate of an already
11895  * existing VEB.  It is also possible for both uplink and vsi seids
11896  * to be zero, in order to create a floating VEB.
11897  *
11898  * Returns pointer to the successfully allocated VEB sw struct on
11899  * success, otherwise returns NULL on failure.
11900  **/
11901 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
11902 				u16 uplink_seid, u16 vsi_seid,
11903 				u8 enabled_tc)
11904 {
11905 	struct i40e_veb *veb, *uplink_veb = NULL;
11906 	int vsi_idx, veb_idx;
11907 	int ret;
11908 
11909 	/* if one seid is 0, the other must be 0 to create a floating relay */
11910 	if ((uplink_seid == 0 || vsi_seid == 0) &&
11911 	    (uplink_seid + vsi_seid != 0)) {
11912 		dev_info(&pf->pdev->dev,
11913 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
11914 			 uplink_seid, vsi_seid);
11915 		return NULL;
11916 	}
11917 
11918 	/* make sure there is such a vsi and uplink */
11919 	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
11920 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
11921 			break;
11922 	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
11923 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
11924 			 vsi_seid);
11925 		return NULL;
11926 	}
11927 
11928 	if (uplink_seid && uplink_seid != pf->mac_seid) {
11929 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
11930 			if (pf->veb[veb_idx] &&
11931 			    pf->veb[veb_idx]->seid == uplink_seid) {
11932 				uplink_veb = pf->veb[veb_idx];
11933 				break;
11934 			}
11935 		}
11936 		if (!uplink_veb) {
11937 			dev_info(&pf->pdev->dev,
11938 				 "uplink seid %d not found\n", uplink_seid);
11939 			return NULL;
11940 		}
11941 	}
11942 
11943 	/* get veb sw struct */
11944 	veb_idx = i40e_veb_mem_alloc(pf);
11945 	if (veb_idx < 0)
11946 		goto err_alloc;
11947 	veb = pf->veb[veb_idx];
11948 	veb->flags = flags;
11949 	veb->uplink_seid = uplink_seid;
11950 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
11951 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
11952 
11953 	/* create the VEB in the switch */
11954 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
11955 	if (ret)
11956 		goto err_veb;
11957 	if (vsi_idx == pf->lan_vsi)
11958 		pf->lan_veb = veb->idx;
11959 
11960 	return veb;
11961 
11962 err_veb:
11963 	i40e_veb_clear(veb);
11964 err_alloc:
11965 	return NULL;
11966 }
11967 
11968 /**
11969  * i40e_setup_pf_switch_element - set PF vars based on switch type
11970  * @pf: board private structure
11971  * @ele: element we are building info from
11972  * @num_reported: total number of elements
11973  * @printconfig: should we print the contents
11974  *
11975  * helper function to assist in extracting a few useful SEID values.
11976  **/
11977 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
11978 				struct i40e_aqc_switch_config_element_resp *ele,
11979 				u16 num_reported, bool printconfig)
11980 {
11981 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
11982 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
11983 	u8 element_type = ele->element_type;
11984 	u16 seid = le16_to_cpu(ele->seid);
11985 
11986 	if (printconfig)
11987 		dev_info(&pf->pdev->dev,
11988 			 "type=%d seid=%d uplink=%d downlink=%d\n",
11989 			 element_type, seid, uplink_seid, downlink_seid);
11990 
11991 	switch (element_type) {
11992 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
11993 		pf->mac_seid = seid;
11994 		break;
11995 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
11996 		/* Main VEB? */
11997 		if (uplink_seid != pf->mac_seid)
11998 			break;
11999 		if (pf->lan_veb == I40E_NO_VEB) {
12000 			int v;
12001 
12002 			/* find existing or else empty VEB */
12003 			for (v = 0; v < I40E_MAX_VEB; v++) {
12004 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
12005 					pf->lan_veb = v;
12006 					break;
12007 				}
12008 			}
12009 			if (pf->lan_veb == I40E_NO_VEB) {
12010 				v = i40e_veb_mem_alloc(pf);
12011 				if (v < 0)
12012 					break;
12013 				pf->lan_veb = v;
12014 			}
12015 		}
12016 
12017 		pf->veb[pf->lan_veb]->seid = seid;
12018 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
12019 		pf->veb[pf->lan_veb]->pf = pf;
12020 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
12021 		break;
12022 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
12023 		if (num_reported != 1)
12024 			break;
12025 		/* This is immediately after a reset so we can assume this is
12026 		 * the PF's VSI
12027 		 */
12028 		pf->mac_seid = uplink_seid;
12029 		pf->pf_seid = downlink_seid;
12030 		pf->main_vsi_seid = seid;
12031 		if (printconfig)
12032 			dev_info(&pf->pdev->dev,
12033 				 "pf_seid=%d main_vsi_seid=%d\n",
12034 				 pf->pf_seid, pf->main_vsi_seid);
12035 		break;
12036 	case I40E_SWITCH_ELEMENT_TYPE_PF:
12037 	case I40E_SWITCH_ELEMENT_TYPE_VF:
12038 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
12039 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
12040 	case I40E_SWITCH_ELEMENT_TYPE_PE:
12041 	case I40E_SWITCH_ELEMENT_TYPE_PA:
12042 		/* ignore these for now */
12043 		break;
12044 	default:
12045 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
12046 			 element_type, seid);
12047 		break;
12048 	}
12049 }
12050 
12051 /**
12052  * i40e_fetch_switch_configuration - Get switch config from firmware
12053  * @pf: board private structure
12054  * @printconfig: should we print the contents
12055  *
12056  * Get the current switch configuration from the device and
12057  * extract a few useful SEID values.
12058  **/
12059 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
12060 {
12061 	struct i40e_aqc_get_switch_config_resp *sw_config;
12062 	u16 next_seid = 0;
12063 	int ret = 0;
12064 	u8 *aq_buf;
12065 	int i;
12066 
12067 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
12068 	if (!aq_buf)
12069 		return -ENOMEM;
12070 
12071 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
12072 	do {
12073 		u16 num_reported, num_total;
12074 
12075 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
12076 						I40E_AQ_LARGE_BUF,
12077 						&next_seid, NULL);
12078 		if (ret) {
12079 			dev_info(&pf->pdev->dev,
12080 				 "get switch config failed err %s aq_err %s\n",
12081 				 i40e_stat_str(&pf->hw, ret),
12082 				 i40e_aq_str(&pf->hw,
12083 					     pf->hw.aq.asq_last_status));
12084 			kfree(aq_buf);
12085 			return -ENOENT;
12086 		}
12087 
12088 		num_reported = le16_to_cpu(sw_config->header.num_reported);
12089 		num_total = le16_to_cpu(sw_config->header.num_total);
12090 
12091 		if (printconfig)
12092 			dev_info(&pf->pdev->dev,
12093 				 "header: %d reported %d total\n",
12094 				 num_reported, num_total);
12095 
12096 		for (i = 0; i < num_reported; i++) {
12097 			struct i40e_aqc_switch_config_element_resp *ele =
12098 				&sw_config->element[i];
12099 
12100 			i40e_setup_pf_switch_element(pf, ele, num_reported,
12101 						     printconfig);
12102 		}
12103 	} while (next_seid != 0);
12104 
12105 	kfree(aq_buf);
12106 	return ret;
12107 }
12108 
12109 /**
12110  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
12111  * @pf: board private structure
12112  * @reinit: if the Main VSI needs to re-initialized.
12113  *
12114  * Returns 0 on success, negative value on failure
12115  **/
12116 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
12117 {
12118 	u16 flags = 0;
12119 	int ret;
12120 
12121 	/* find out what's out there already */
12122 	ret = i40e_fetch_switch_configuration(pf, false);
12123 	if (ret) {
12124 		dev_info(&pf->pdev->dev,
12125 			 "couldn't fetch switch config, err %s aq_err %s\n",
12126 			 i40e_stat_str(&pf->hw, ret),
12127 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12128 		return ret;
12129 	}
12130 	i40e_pf_reset_stats(pf);
12131 
12132 	/* set the switch config bit for the whole device to
12133 	 * support limited promisc or true promisc
12134 	 * when user requests promisc. The default is limited
12135 	 * promisc.
12136 	*/
12137 
12138 	if ((pf->hw.pf_id == 0) &&
12139 	    !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
12140 		flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
12141 
12142 	if (pf->hw.pf_id == 0) {
12143 		u16 valid_flags;
12144 
12145 		valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
12146 		ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
12147 						NULL);
12148 		if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
12149 			dev_info(&pf->pdev->dev,
12150 				 "couldn't set switch config bits, err %s aq_err %s\n",
12151 				 i40e_stat_str(&pf->hw, ret),
12152 				 i40e_aq_str(&pf->hw,
12153 					     pf->hw.aq.asq_last_status));
12154 			/* not a fatal problem, just keep going */
12155 		}
12156 	}
12157 
12158 	/* first time setup */
12159 	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
12160 		struct i40e_vsi *vsi = NULL;
12161 		u16 uplink_seid;
12162 
12163 		/* Set up the PF VSI associated with the PF's main VSI
12164 		 * that is already in the HW switch
12165 		 */
12166 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
12167 			uplink_seid = pf->veb[pf->lan_veb]->seid;
12168 		else
12169 			uplink_seid = pf->mac_seid;
12170 		if (pf->lan_vsi == I40E_NO_VSI)
12171 			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
12172 		else if (reinit)
12173 			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
12174 		if (!vsi) {
12175 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
12176 			i40e_fdir_teardown(pf);
12177 			return -EAGAIN;
12178 		}
12179 	} else {
12180 		/* force a reset of TC and queue layout configurations */
12181 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
12182 
12183 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
12184 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
12185 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
12186 	}
12187 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
12188 
12189 	i40e_fdir_sb_setup(pf);
12190 
12191 	/* Setup static PF queue filter control settings */
12192 	ret = i40e_setup_pf_filter_control(pf);
12193 	if (ret) {
12194 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
12195 			 ret);
12196 		/* Failure here should not stop continuing other steps */
12197 	}
12198 
12199 	/* enable RSS in the HW, even for only one queue, as the stack can use
12200 	 * the hash
12201 	 */
12202 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
12203 		i40e_pf_config_rss(pf);
12204 
12205 	/* fill in link information and enable LSE reporting */
12206 	i40e_link_event(pf);
12207 
12208 	/* Initialize user-specific link properties */
12209 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
12210 				  I40E_AQ_AN_COMPLETED) ? true : false);
12211 
12212 	i40e_ptp_init(pf);
12213 
12214 	/* repopulate tunnel port filters */
12215 	i40e_sync_udp_filters(pf);
12216 
12217 	return ret;
12218 }
12219 
12220 /**
12221  * i40e_determine_queue_usage - Work out queue distribution
12222  * @pf: board private structure
12223  **/
12224 static void i40e_determine_queue_usage(struct i40e_pf *pf)
12225 {
12226 	int queues_left;
12227 	int q_max;
12228 
12229 	pf->num_lan_qps = 0;
12230 
12231 	/* Find the max queues to be put into basic use.  We'll always be
12232 	 * using TC0, whether or not DCB is running, and TC0 will get the
12233 	 * big RSS set.
12234 	 */
12235 	queues_left = pf->hw.func_caps.num_tx_qp;
12236 
12237 	if ((queues_left == 1) ||
12238 	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
12239 		/* one qp for PF, no queues for anything else */
12240 		queues_left = 0;
12241 		pf->alloc_rss_size = pf->num_lan_qps = 1;
12242 
12243 		/* make sure all the fancies are disabled */
12244 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
12245 			       I40E_FLAG_IWARP_ENABLED	|
12246 			       I40E_FLAG_FD_SB_ENABLED	|
12247 			       I40E_FLAG_FD_ATR_ENABLED	|
12248 			       I40E_FLAG_DCB_CAPABLE	|
12249 			       I40E_FLAG_DCB_ENABLED	|
12250 			       I40E_FLAG_SRIOV_ENABLED	|
12251 			       I40E_FLAG_VMDQ_ENABLED);
12252 	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
12253 				  I40E_FLAG_FD_SB_ENABLED |
12254 				  I40E_FLAG_FD_ATR_ENABLED |
12255 				  I40E_FLAG_DCB_CAPABLE))) {
12256 		/* one qp for PF */
12257 		pf->alloc_rss_size = pf->num_lan_qps = 1;
12258 		queues_left -= pf->num_lan_qps;
12259 
12260 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
12261 			       I40E_FLAG_IWARP_ENABLED	|
12262 			       I40E_FLAG_FD_SB_ENABLED	|
12263 			       I40E_FLAG_FD_ATR_ENABLED	|
12264 			       I40E_FLAG_DCB_ENABLED	|
12265 			       I40E_FLAG_VMDQ_ENABLED);
12266 	} else {
12267 		/* Not enough queues for all TCs */
12268 		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
12269 		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
12270 			pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
12271 					I40E_FLAG_DCB_ENABLED);
12272 			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
12273 		}
12274 
12275 		/* limit lan qps to the smaller of qps, cpus or msix */
12276 		q_max = max_t(int, pf->rss_size_max, num_online_cpus());
12277 		q_max = min_t(int, q_max, pf->hw.func_caps.num_tx_qp);
12278 		q_max = min_t(int, q_max, pf->hw.func_caps.num_msix_vectors);
12279 		pf->num_lan_qps = q_max;
12280 
12281 		queues_left -= pf->num_lan_qps;
12282 	}
12283 
12284 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
12285 		if (queues_left > 1) {
12286 			queues_left -= 1; /* save 1 queue for FD */
12287 		} else {
12288 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
12289 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
12290 		}
12291 	}
12292 
12293 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
12294 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
12295 		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
12296 					(queues_left / pf->num_vf_qps));
12297 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
12298 	}
12299 
12300 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
12301 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
12302 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
12303 					  (queues_left / pf->num_vmdq_qps));
12304 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
12305 	}
12306 
12307 	pf->queues_left = queues_left;
12308 	dev_dbg(&pf->pdev->dev,
12309 		"qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
12310 		pf->hw.func_caps.num_tx_qp,
12311 		!!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
12312 		pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
12313 		pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
12314 		queues_left);
12315 }
12316 
12317 /**
12318  * i40e_setup_pf_filter_control - Setup PF static filter control
12319  * @pf: PF to be setup
12320  *
12321  * i40e_setup_pf_filter_control sets up a PF's initial filter control
12322  * settings. If PE/FCoE are enabled then it will also set the per PF
12323  * based filter sizes required for them. It also enables Flow director,
12324  * ethertype and macvlan type filter settings for the pf.
12325  *
12326  * Returns 0 on success, negative on failure
12327  **/
12328 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
12329 {
12330 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
12331 
12332 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
12333 
12334 	/* Flow Director is enabled */
12335 	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
12336 		settings->enable_fdir = true;
12337 
12338 	/* Ethtype and MACVLAN filters enabled for PF */
12339 	settings->enable_ethtype = true;
12340 	settings->enable_macvlan = true;
12341 
12342 	if (i40e_set_filter_control(&pf->hw, settings))
12343 		return -ENOENT;
12344 
12345 	return 0;
12346 }
12347 
12348 #define INFO_STRING_LEN 255
12349 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
12350 static void i40e_print_features(struct i40e_pf *pf)
12351 {
12352 	struct i40e_hw *hw = &pf->hw;
12353 	char *buf;
12354 	int i;
12355 
12356 	buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
12357 	if (!buf)
12358 		return;
12359 
12360 	i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
12361 #ifdef CONFIG_PCI_IOV
12362 	i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
12363 #endif
12364 	i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
12365 		      pf->hw.func_caps.num_vsis,
12366 		      pf->vsi[pf->lan_vsi]->num_queue_pairs);
12367 	if (pf->flags & I40E_FLAG_RSS_ENABLED)
12368 		i += snprintf(&buf[i], REMAIN(i), " RSS");
12369 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
12370 		i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
12371 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
12372 		i += snprintf(&buf[i], REMAIN(i), " FD_SB");
12373 		i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
12374 	}
12375 	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
12376 		i += snprintf(&buf[i], REMAIN(i), " DCB");
12377 	i += snprintf(&buf[i], REMAIN(i), " VxLAN");
12378 	i += snprintf(&buf[i], REMAIN(i), " Geneve");
12379 	if (pf->flags & I40E_FLAG_PTP)
12380 		i += snprintf(&buf[i], REMAIN(i), " PTP");
12381 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
12382 		i += snprintf(&buf[i], REMAIN(i), " VEB");
12383 	else
12384 		i += snprintf(&buf[i], REMAIN(i), " VEPA");
12385 
12386 	dev_info(&pf->pdev->dev, "%s\n", buf);
12387 	kfree(buf);
12388 	WARN_ON(i > INFO_STRING_LEN);
12389 }
12390 
12391 /**
12392  * i40e_get_platform_mac_addr - get platform-specific MAC address
12393  * @pdev: PCI device information struct
12394  * @pf: board private structure
12395  *
12396  * Look up the MAC address for the device. First we'll try
12397  * eth_platform_get_mac_address, which will check Open Firmware, or arch
12398  * specific fallback. Otherwise, we'll default to the stored value in
12399  * firmware.
12400  **/
12401 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
12402 {
12403 	if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
12404 		i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
12405 }
12406 
12407 /**
12408  * i40e_probe - Device initialization routine
12409  * @pdev: PCI device information struct
12410  * @ent: entry in i40e_pci_tbl
12411  *
12412  * i40e_probe initializes a PF identified by a pci_dev structure.
12413  * The OS initialization, configuring of the PF private structure,
12414  * and a hardware reset occur.
12415  *
12416  * Returns 0 on success, negative on failure
12417  **/
12418 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
12419 {
12420 	struct i40e_aq_get_phy_abilities_resp abilities;
12421 	struct i40e_pf *pf;
12422 	struct i40e_hw *hw;
12423 	static u16 pfs_found;
12424 	u16 wol_nvm_bits;
12425 	u16 link_status;
12426 	int err;
12427 	u32 val;
12428 	u32 i;
12429 	u8 set_fc_aq_fail;
12430 
12431 	err = pci_enable_device_mem(pdev);
12432 	if (err)
12433 		return err;
12434 
12435 	/* set up for high or low dma */
12436 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
12437 	if (err) {
12438 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
12439 		if (err) {
12440 			dev_err(&pdev->dev,
12441 				"DMA configuration failed: 0x%x\n", err);
12442 			goto err_dma;
12443 		}
12444 	}
12445 
12446 	/* set up pci connections */
12447 	err = pci_request_mem_regions(pdev, i40e_driver_name);
12448 	if (err) {
12449 		dev_info(&pdev->dev,
12450 			 "pci_request_selected_regions failed %d\n", err);
12451 		goto err_pci_reg;
12452 	}
12453 
12454 	pci_enable_pcie_error_reporting(pdev);
12455 	pci_set_master(pdev);
12456 
12457 	/* Now that we have a PCI connection, we need to do the
12458 	 * low level device setup.  This is primarily setting up
12459 	 * the Admin Queue structures and then querying for the
12460 	 * device's current profile information.
12461 	 */
12462 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
12463 	if (!pf) {
12464 		err = -ENOMEM;
12465 		goto err_pf_alloc;
12466 	}
12467 	pf->next_vsi = 0;
12468 	pf->pdev = pdev;
12469 	set_bit(__I40E_DOWN, pf->state);
12470 
12471 	hw = &pf->hw;
12472 	hw->back = pf;
12473 
12474 	pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
12475 				I40E_MAX_CSR_SPACE);
12476 
12477 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
12478 	if (!hw->hw_addr) {
12479 		err = -EIO;
12480 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
12481 			 (unsigned int)pci_resource_start(pdev, 0),
12482 			 pf->ioremap_len, err);
12483 		goto err_ioremap;
12484 	}
12485 	hw->vendor_id = pdev->vendor;
12486 	hw->device_id = pdev->device;
12487 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
12488 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
12489 	hw->subsystem_device_id = pdev->subsystem_device;
12490 	hw->bus.device = PCI_SLOT(pdev->devfn);
12491 	hw->bus.func = PCI_FUNC(pdev->devfn);
12492 	hw->bus.bus_id = pdev->bus->number;
12493 	pf->instance = pfs_found;
12494 
12495 	/* Select something other than the 802.1ad ethertype for the
12496 	 * switch to use internally and drop on ingress.
12497 	 */
12498 	hw->switch_tag = 0xffff;
12499 	hw->first_tag = ETH_P_8021AD;
12500 	hw->second_tag = ETH_P_8021Q;
12501 
12502 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
12503 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
12504 
12505 	/* set up the locks for the AQ, do this only once in probe
12506 	 * and destroy them only once in remove
12507 	 */
12508 	mutex_init(&hw->aq.asq_mutex);
12509 	mutex_init(&hw->aq.arq_mutex);
12510 
12511 	pf->msg_enable = netif_msg_init(debug,
12512 					NETIF_MSG_DRV |
12513 					NETIF_MSG_PROBE |
12514 					NETIF_MSG_LINK);
12515 	if (debug < -1)
12516 		pf->hw.debug_mask = debug;
12517 
12518 	/* do a special CORER for clearing PXE mode once at init */
12519 	if (hw->revision_id == 0 &&
12520 	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
12521 		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
12522 		i40e_flush(hw);
12523 		msleep(200);
12524 		pf->corer_count++;
12525 
12526 		i40e_clear_pxe_mode(hw);
12527 	}
12528 
12529 	/* Reset here to make sure all is clean and to define PF 'n' */
12530 	i40e_clear_hw(hw);
12531 	err = i40e_pf_reset(hw);
12532 	if (err) {
12533 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
12534 		goto err_pf_reset;
12535 	}
12536 	pf->pfr_count++;
12537 
12538 	hw->aq.num_arq_entries = I40E_AQ_LEN;
12539 	hw->aq.num_asq_entries = I40E_AQ_LEN;
12540 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
12541 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
12542 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
12543 
12544 	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
12545 		 "%s-%s:misc",
12546 		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
12547 
12548 	err = i40e_init_shared_code(hw);
12549 	if (err) {
12550 		dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
12551 			 err);
12552 		goto err_pf_reset;
12553 	}
12554 
12555 	/* set up a default setting for link flow control */
12556 	pf->hw.fc.requested_mode = I40E_FC_NONE;
12557 
12558 	err = i40e_init_adminq(hw);
12559 	if (err) {
12560 		if (err == I40E_ERR_FIRMWARE_API_VERSION)
12561 			dev_info(&pdev->dev,
12562 				 "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");
12563 		else
12564 			dev_info(&pdev->dev,
12565 				 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
12566 
12567 		goto err_pf_reset;
12568 	}
12569 	i40e_get_oem_version(hw);
12570 
12571 	/* provide nvm, fw, api versions */
12572 	dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
12573 		 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
12574 		 hw->aq.api_maj_ver, hw->aq.api_min_ver,
12575 		 i40e_nvm_version_str(hw));
12576 
12577 	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
12578 	    hw->aq.api_min_ver > I40E_FW_MINOR_VERSION(hw))
12579 		dev_info(&pdev->dev,
12580 			 "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");
12581 	else if (hw->aq.api_maj_ver == 1 && hw->aq.api_min_ver < 4)
12582 		dev_info(&pdev->dev,
12583 			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
12584 
12585 	i40e_verify_eeprom(pf);
12586 
12587 	/* Rev 0 hardware was never productized */
12588 	if (hw->revision_id < 1)
12589 		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");
12590 
12591 	i40e_clear_pxe_mode(hw);
12592 	err = i40e_get_capabilities(pf);
12593 	if (err)
12594 		goto err_adminq_setup;
12595 
12596 	err = i40e_sw_init(pf);
12597 	if (err) {
12598 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
12599 		goto err_sw_init;
12600 	}
12601 
12602 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
12603 				hw->func_caps.num_rx_qp, 0, 0);
12604 	if (err) {
12605 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
12606 		goto err_init_lan_hmc;
12607 	}
12608 
12609 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
12610 	if (err) {
12611 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
12612 		err = -ENOENT;
12613 		goto err_configure_lan_hmc;
12614 	}
12615 
12616 	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
12617 	 * Ignore error return codes because if it was already disabled via
12618 	 * hardware settings this will fail
12619 	 */
12620 	if (pf->hw_features & I40E_HW_STOP_FW_LLDP) {
12621 		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
12622 		i40e_aq_stop_lldp(hw, true, NULL);
12623 	}
12624 
12625 	/* allow a platform config to override the HW addr */
12626 	i40e_get_platform_mac_addr(pdev, pf);
12627 
12628 	if (!is_valid_ether_addr(hw->mac.addr)) {
12629 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
12630 		err = -EIO;
12631 		goto err_mac_addr;
12632 	}
12633 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
12634 	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
12635 	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
12636 	if (is_valid_ether_addr(hw->mac.port_addr))
12637 		pf->hw_features |= I40E_HW_PORT_ID_VALID;
12638 
12639 	pci_set_drvdata(pdev, pf);
12640 	pci_save_state(pdev);
12641 #ifdef CONFIG_I40E_DCB
12642 	err = i40e_init_pf_dcb(pf);
12643 	if (err) {
12644 		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
12645 		pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
12646 		/* Continue without DCB enabled */
12647 	}
12648 #endif /* CONFIG_I40E_DCB */
12649 
12650 	/* set up periodic task facility */
12651 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
12652 	pf->service_timer_period = HZ;
12653 
12654 	INIT_WORK(&pf->service_task, i40e_service_task);
12655 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
12656 
12657 	/* NVM bit on means WoL disabled for the port */
12658 	i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
12659 	if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
12660 		pf->wol_en = false;
12661 	else
12662 		pf->wol_en = true;
12663 	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
12664 
12665 	/* set up the main switch operations */
12666 	i40e_determine_queue_usage(pf);
12667 	err = i40e_init_interrupt_scheme(pf);
12668 	if (err)
12669 		goto err_switch_setup;
12670 
12671 	/* The number of VSIs reported by the FW is the minimum guaranteed
12672 	 * to us; HW supports far more and we share the remaining pool with
12673 	 * the other PFs. We allocate space for more than the guarantee with
12674 	 * the understanding that we might not get them all later.
12675 	 */
12676 	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
12677 		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
12678 	else
12679 		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
12680 
12681 	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
12682 	pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
12683 			  GFP_KERNEL);
12684 	if (!pf->vsi) {
12685 		err = -ENOMEM;
12686 		goto err_switch_setup;
12687 	}
12688 
12689 #ifdef CONFIG_PCI_IOV
12690 	/* prep for VF support */
12691 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
12692 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
12693 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
12694 		if (pci_num_vf(pdev))
12695 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
12696 	}
12697 #endif
12698 	err = i40e_setup_pf_switch(pf, false);
12699 	if (err) {
12700 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
12701 		goto err_vsis;
12702 	}
12703 	INIT_LIST_HEAD(&pf->vsi[pf->lan_vsi]->ch_list);
12704 
12705 	/* Make sure flow control is set according to current settings */
12706 	err = i40e_set_fc(hw, &set_fc_aq_fail, true);
12707 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
12708 		dev_dbg(&pf->pdev->dev,
12709 			"Set fc with err %s aq_err %s on get_phy_cap\n",
12710 			i40e_stat_str(hw, err),
12711 			i40e_aq_str(hw, hw->aq.asq_last_status));
12712 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
12713 		dev_dbg(&pf->pdev->dev,
12714 			"Set fc with err %s aq_err %s on set_phy_config\n",
12715 			i40e_stat_str(hw, err),
12716 			i40e_aq_str(hw, hw->aq.asq_last_status));
12717 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
12718 		dev_dbg(&pf->pdev->dev,
12719 			"Set fc with err %s aq_err %s on get_link_info\n",
12720 			i40e_stat_str(hw, err),
12721 			i40e_aq_str(hw, hw->aq.asq_last_status));
12722 
12723 	/* if FDIR VSI was set up, start it now */
12724 	for (i = 0; i < pf->num_alloc_vsi; i++) {
12725 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
12726 			i40e_vsi_open(pf->vsi[i]);
12727 			break;
12728 		}
12729 	}
12730 
12731 	/* The driver only wants link up/down and module qualification
12732 	 * reports from firmware.  Note the negative logic.
12733 	 */
12734 	err = i40e_aq_set_phy_int_mask(&pf->hw,
12735 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
12736 					 I40E_AQ_EVENT_MEDIA_NA |
12737 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
12738 	if (err)
12739 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
12740 			 i40e_stat_str(&pf->hw, err),
12741 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12742 
12743 	/* Reconfigure hardware for allowing smaller MSS in the case
12744 	 * of TSO, so that we avoid the MDD being fired and causing
12745 	 * a reset in the case of small MSS+TSO.
12746 	 */
12747 	val = rd32(hw, I40E_REG_MSS);
12748 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
12749 		val &= ~I40E_REG_MSS_MIN_MASK;
12750 		val |= I40E_64BYTE_MSS;
12751 		wr32(hw, I40E_REG_MSS, val);
12752 	}
12753 
12754 	if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
12755 		msleep(75);
12756 		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
12757 		if (err)
12758 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
12759 				 i40e_stat_str(&pf->hw, err),
12760 				 i40e_aq_str(&pf->hw,
12761 					     pf->hw.aq.asq_last_status));
12762 	}
12763 	/* The main driver is (mostly) up and happy. We need to set this state
12764 	 * before setting up the misc vector or we get a race and the vector
12765 	 * ends up disabled forever.
12766 	 */
12767 	clear_bit(__I40E_DOWN, pf->state);
12768 
12769 	/* In case of MSIX we are going to setup the misc vector right here
12770 	 * to handle admin queue events etc. In case of legacy and MSI
12771 	 * the misc functionality and queue processing is combined in
12772 	 * the same vector and that gets setup at open.
12773 	 */
12774 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
12775 		err = i40e_setup_misc_vector(pf);
12776 		if (err) {
12777 			dev_info(&pdev->dev,
12778 				 "setup of misc vector failed: %d\n", err);
12779 			goto err_vsis;
12780 		}
12781 	}
12782 
12783 #ifdef CONFIG_PCI_IOV
12784 	/* prep for VF support */
12785 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
12786 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
12787 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
12788 		/* disable link interrupts for VFs */
12789 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
12790 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
12791 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
12792 		i40e_flush(hw);
12793 
12794 		if (pci_num_vf(pdev)) {
12795 			dev_info(&pdev->dev,
12796 				 "Active VFs found, allocating resources.\n");
12797 			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
12798 			if (err)
12799 				dev_info(&pdev->dev,
12800 					 "Error %d allocating resources for existing VFs\n",
12801 					 err);
12802 		}
12803 	}
12804 #endif /* CONFIG_PCI_IOV */
12805 
12806 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
12807 		pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
12808 						      pf->num_iwarp_msix,
12809 						      I40E_IWARP_IRQ_PILE_ID);
12810 		if (pf->iwarp_base_vector < 0) {
12811 			dev_info(&pdev->dev,
12812 				 "failed to get tracking for %d vectors for IWARP err=%d\n",
12813 				 pf->num_iwarp_msix, pf->iwarp_base_vector);
12814 			pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
12815 		}
12816 	}
12817 
12818 	i40e_dbg_pf_init(pf);
12819 
12820 	/* tell the firmware that we're starting */
12821 	i40e_send_version(pf);
12822 
12823 	/* since everything's happy, start the service_task timer */
12824 	mod_timer(&pf->service_timer,
12825 		  round_jiffies(jiffies + pf->service_timer_period));
12826 
12827 	/* add this PF to client device list and launch a client service task */
12828 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
12829 		err = i40e_lan_add_device(pf);
12830 		if (err)
12831 			dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
12832 				 err);
12833 	}
12834 
12835 #define PCI_SPEED_SIZE 8
12836 #define PCI_WIDTH_SIZE 8
12837 	/* Devices on the IOSF bus do not have this information
12838 	 * and will report PCI Gen 1 x 1 by default so don't bother
12839 	 * checking them.
12840 	 */
12841 	if (!(pf->hw_features & I40E_HW_NO_PCI_LINK_CHECK)) {
12842 		char speed[PCI_SPEED_SIZE] = "Unknown";
12843 		char width[PCI_WIDTH_SIZE] = "Unknown";
12844 
12845 		/* Get the negotiated link width and speed from PCI config
12846 		 * space
12847 		 */
12848 		pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
12849 					  &link_status);
12850 
12851 		i40e_set_pci_config_data(hw, link_status);
12852 
12853 		switch (hw->bus.speed) {
12854 		case i40e_bus_speed_8000:
12855 			strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
12856 		case i40e_bus_speed_5000:
12857 			strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
12858 		case i40e_bus_speed_2500:
12859 			strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
12860 		default:
12861 			break;
12862 		}
12863 		switch (hw->bus.width) {
12864 		case i40e_bus_width_pcie_x8:
12865 			strncpy(width, "8", PCI_WIDTH_SIZE); break;
12866 		case i40e_bus_width_pcie_x4:
12867 			strncpy(width, "4", PCI_WIDTH_SIZE); break;
12868 		case i40e_bus_width_pcie_x2:
12869 			strncpy(width, "2", PCI_WIDTH_SIZE); break;
12870 		case i40e_bus_width_pcie_x1:
12871 			strncpy(width, "1", PCI_WIDTH_SIZE); break;
12872 		default:
12873 			break;
12874 		}
12875 
12876 		dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
12877 			 speed, width);
12878 
12879 		if (hw->bus.width < i40e_bus_width_pcie_x8 ||
12880 		    hw->bus.speed < i40e_bus_speed_8000) {
12881 			dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
12882 			dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
12883 		}
12884 	}
12885 
12886 	/* get the requested speeds from the fw */
12887 	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
12888 	if (err)
12889 		dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
12890 			i40e_stat_str(&pf->hw, err),
12891 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12892 	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
12893 
12894 	/* get the supported phy types from the fw */
12895 	err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
12896 	if (err)
12897 		dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
12898 			i40e_stat_str(&pf->hw, err),
12899 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12900 
12901 	/* Add a filter to drop all Flow control frames from any VSI from being
12902 	 * transmitted. By doing so we stop a malicious VF from sending out
12903 	 * PAUSE or PFC frames and potentially controlling traffic for other
12904 	 * PF/VF VSIs.
12905 	 * The FW can still send Flow control frames if enabled.
12906 	 */
12907 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
12908 						       pf->main_vsi_seid);
12909 
12910 	if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
12911 		(pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
12912 		pf->hw_features |= I40E_HW_PHY_CONTROLS_LEDS;
12913 	if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
12914 		pf->hw_features |= I40E_HW_HAVE_CRT_RETIMER;
12915 	/* print a string summarizing features */
12916 	i40e_print_features(pf);
12917 
12918 	return 0;
12919 
12920 	/* Unwind what we've done if something failed in the setup */
12921 err_vsis:
12922 	set_bit(__I40E_DOWN, pf->state);
12923 	i40e_clear_interrupt_scheme(pf);
12924 	kfree(pf->vsi);
12925 err_switch_setup:
12926 	i40e_reset_interrupt_capability(pf);
12927 	del_timer_sync(&pf->service_timer);
12928 err_mac_addr:
12929 err_configure_lan_hmc:
12930 	(void)i40e_shutdown_lan_hmc(hw);
12931 err_init_lan_hmc:
12932 	kfree(pf->qp_pile);
12933 err_sw_init:
12934 err_adminq_setup:
12935 err_pf_reset:
12936 	iounmap(hw->hw_addr);
12937 err_ioremap:
12938 	kfree(pf);
12939 err_pf_alloc:
12940 	pci_disable_pcie_error_reporting(pdev);
12941 	pci_release_mem_regions(pdev);
12942 err_pci_reg:
12943 err_dma:
12944 	pci_disable_device(pdev);
12945 	return err;
12946 }
12947 
12948 /**
12949  * i40e_remove - Device removal routine
12950  * @pdev: PCI device information struct
12951  *
12952  * i40e_remove is called by the PCI subsystem to alert the driver
12953  * that is should release a PCI device.  This could be caused by a
12954  * Hot-Plug event, or because the driver is going to be removed from
12955  * memory.
12956  **/
12957 static void i40e_remove(struct pci_dev *pdev)
12958 {
12959 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12960 	struct i40e_hw *hw = &pf->hw;
12961 	i40e_status ret_code;
12962 	int i;
12963 
12964 	i40e_dbg_pf_exit(pf);
12965 
12966 	i40e_ptp_stop(pf);
12967 
12968 	/* Disable RSS in hw */
12969 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
12970 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
12971 
12972 	/* no more scheduling of any task */
12973 	set_bit(__I40E_SUSPENDED, pf->state);
12974 	set_bit(__I40E_DOWN, pf->state);
12975 	if (pf->service_timer.data)
12976 		del_timer_sync(&pf->service_timer);
12977 	if (pf->service_task.func)
12978 		cancel_work_sync(&pf->service_task);
12979 
12980 	/* Client close must be called explicitly here because the timer
12981 	 * has been stopped.
12982 	 */
12983 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
12984 
12985 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
12986 		i40e_free_vfs(pf);
12987 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
12988 	}
12989 
12990 	i40e_fdir_teardown(pf);
12991 
12992 	/* If there is a switch structure or any orphans, remove them.
12993 	 * This will leave only the PF's VSI remaining.
12994 	 */
12995 	for (i = 0; i < I40E_MAX_VEB; i++) {
12996 		if (!pf->veb[i])
12997 			continue;
12998 
12999 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
13000 		    pf->veb[i]->uplink_seid == 0)
13001 			i40e_switch_branch_release(pf->veb[i]);
13002 	}
13003 
13004 	/* Now we can shutdown the PF's VSI, just before we kill
13005 	 * adminq and hmc.
13006 	 */
13007 	if (pf->vsi[pf->lan_vsi])
13008 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
13009 
13010 	/* remove attached clients */
13011 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
13012 		ret_code = i40e_lan_del_device(pf);
13013 		if (ret_code)
13014 			dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
13015 				 ret_code);
13016 	}
13017 
13018 	/* shutdown and destroy the HMC */
13019 	if (hw->hmc.hmc_obj) {
13020 		ret_code = i40e_shutdown_lan_hmc(hw);
13021 		if (ret_code)
13022 			dev_warn(&pdev->dev,
13023 				 "Failed to destroy the HMC resources: %d\n",
13024 				 ret_code);
13025 	}
13026 
13027 	/* shutdown the adminq */
13028 	i40e_shutdown_adminq(hw);
13029 
13030 	/* destroy the locks only once, here */
13031 	mutex_destroy(&hw->aq.arq_mutex);
13032 	mutex_destroy(&hw->aq.asq_mutex);
13033 
13034 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
13035 	i40e_clear_interrupt_scheme(pf);
13036 	for (i = 0; i < pf->num_alloc_vsi; i++) {
13037 		if (pf->vsi[i]) {
13038 			i40e_vsi_clear_rings(pf->vsi[i]);
13039 			i40e_vsi_clear(pf->vsi[i]);
13040 			pf->vsi[i] = NULL;
13041 		}
13042 	}
13043 
13044 	for (i = 0; i < I40E_MAX_VEB; i++) {
13045 		kfree(pf->veb[i]);
13046 		pf->veb[i] = NULL;
13047 	}
13048 
13049 	kfree(pf->qp_pile);
13050 	kfree(pf->vsi);
13051 
13052 	iounmap(hw->hw_addr);
13053 	kfree(pf);
13054 	pci_release_mem_regions(pdev);
13055 
13056 	pci_disable_pcie_error_reporting(pdev);
13057 	pci_disable_device(pdev);
13058 }
13059 
13060 /**
13061  * i40e_pci_error_detected - warning that something funky happened in PCI land
13062  * @pdev: PCI device information struct
13063  *
13064  * Called to warn that something happened and the error handling steps
13065  * are in progress.  Allows the driver to quiesce things, be ready for
13066  * remediation.
13067  **/
13068 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
13069 						enum pci_channel_state error)
13070 {
13071 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13072 
13073 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
13074 
13075 	if (!pf) {
13076 		dev_info(&pdev->dev,
13077 			 "Cannot recover - error happened during device probe\n");
13078 		return PCI_ERS_RESULT_DISCONNECT;
13079 	}
13080 
13081 	/* shutdown all operations */
13082 	if (!test_bit(__I40E_SUSPENDED, pf->state))
13083 		i40e_prep_for_reset(pf, false);
13084 
13085 	/* Request a slot reset */
13086 	return PCI_ERS_RESULT_NEED_RESET;
13087 }
13088 
13089 /**
13090  * i40e_pci_error_slot_reset - a PCI slot reset just happened
13091  * @pdev: PCI device information struct
13092  *
13093  * Called to find if the driver can work with the device now that
13094  * the pci slot has been reset.  If a basic connection seems good
13095  * (registers are readable and have sane content) then return a
13096  * happy little PCI_ERS_RESULT_xxx.
13097  **/
13098 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
13099 {
13100 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13101 	pci_ers_result_t result;
13102 	int err;
13103 	u32 reg;
13104 
13105 	dev_dbg(&pdev->dev, "%s\n", __func__);
13106 	if (pci_enable_device_mem(pdev)) {
13107 		dev_info(&pdev->dev,
13108 			 "Cannot re-enable PCI device after reset.\n");
13109 		result = PCI_ERS_RESULT_DISCONNECT;
13110 	} else {
13111 		pci_set_master(pdev);
13112 		pci_restore_state(pdev);
13113 		pci_save_state(pdev);
13114 		pci_wake_from_d3(pdev, false);
13115 
13116 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
13117 		if (reg == 0)
13118 			result = PCI_ERS_RESULT_RECOVERED;
13119 		else
13120 			result = PCI_ERS_RESULT_DISCONNECT;
13121 	}
13122 
13123 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
13124 	if (err) {
13125 		dev_info(&pdev->dev,
13126 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
13127 			 err);
13128 		/* non-fatal, continue */
13129 	}
13130 
13131 	return result;
13132 }
13133 
13134 /**
13135  * i40e_pci_error_reset_prepare - prepare device driver for pci reset
13136  * @pdev: PCI device information struct
13137  */
13138 static void i40e_pci_error_reset_prepare(struct pci_dev *pdev)
13139 {
13140 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13141 
13142 	i40e_prep_for_reset(pf, false);
13143 }
13144 
13145 /**
13146  * i40e_pci_error_reset_done - pci reset done, device driver reset can begin
13147  * @pdev: PCI device information struct
13148  */
13149 static void i40e_pci_error_reset_done(struct pci_dev *pdev)
13150 {
13151 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13152 
13153 	i40e_reset_and_rebuild(pf, false, false);
13154 }
13155 
13156 /**
13157  * i40e_pci_error_resume - restart operations after PCI error recovery
13158  * @pdev: PCI device information struct
13159  *
13160  * Called to allow the driver to bring things back up after PCI error
13161  * and/or reset recovery has finished.
13162  **/
13163 static void i40e_pci_error_resume(struct pci_dev *pdev)
13164 {
13165 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13166 
13167 	dev_dbg(&pdev->dev, "%s\n", __func__);
13168 	if (test_bit(__I40E_SUSPENDED, pf->state))
13169 		return;
13170 
13171 	i40e_handle_reset_warning(pf, false);
13172 }
13173 
13174 /**
13175  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
13176  * using the mac_address_write admin q function
13177  * @pf: pointer to i40e_pf struct
13178  **/
13179 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
13180 {
13181 	struct i40e_hw *hw = &pf->hw;
13182 	i40e_status ret;
13183 	u8 mac_addr[6];
13184 	u16 flags = 0;
13185 
13186 	/* Get current MAC address in case it's an LAA */
13187 	if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
13188 		ether_addr_copy(mac_addr,
13189 				pf->vsi[pf->lan_vsi]->netdev->dev_addr);
13190 	} else {
13191 		dev_err(&pf->pdev->dev,
13192 			"Failed to retrieve MAC address; using default\n");
13193 		ether_addr_copy(mac_addr, hw->mac.addr);
13194 	}
13195 
13196 	/* The FW expects the mac address write cmd to first be called with
13197 	 * one of these flags before calling it again with the multicast
13198 	 * enable flags.
13199 	 */
13200 	flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
13201 
13202 	if (hw->func_caps.flex10_enable && hw->partition_id != 1)
13203 		flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
13204 
13205 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
13206 	if (ret) {
13207 		dev_err(&pf->pdev->dev,
13208 			"Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
13209 		return;
13210 	}
13211 
13212 	flags = I40E_AQC_MC_MAG_EN
13213 			| I40E_AQC_WOL_PRESERVE_ON_PFR
13214 			| I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
13215 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
13216 	if (ret)
13217 		dev_err(&pf->pdev->dev,
13218 			"Failed to enable Multicast Magic Packet wake up\n");
13219 }
13220 
13221 /**
13222  * i40e_shutdown - PCI callback for shutting down
13223  * @pdev: PCI device information struct
13224  **/
13225 static void i40e_shutdown(struct pci_dev *pdev)
13226 {
13227 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13228 	struct i40e_hw *hw = &pf->hw;
13229 
13230 	set_bit(__I40E_SUSPENDED, pf->state);
13231 	set_bit(__I40E_DOWN, pf->state);
13232 	rtnl_lock();
13233 	i40e_prep_for_reset(pf, true);
13234 	rtnl_unlock();
13235 
13236 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
13237 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
13238 
13239 	del_timer_sync(&pf->service_timer);
13240 	cancel_work_sync(&pf->service_task);
13241 	i40e_fdir_teardown(pf);
13242 
13243 	/* Client close must be called explicitly here because the timer
13244 	 * has been stopped.
13245 	 */
13246 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
13247 
13248 	if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
13249 		i40e_enable_mc_magic_wake(pf);
13250 
13251 	i40e_prep_for_reset(pf, false);
13252 
13253 	wr32(hw, I40E_PFPM_APM,
13254 	     (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
13255 	wr32(hw, I40E_PFPM_WUFC,
13256 	     (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
13257 
13258 	i40e_clear_interrupt_scheme(pf);
13259 
13260 	if (system_state == SYSTEM_POWER_OFF) {
13261 		pci_wake_from_d3(pdev, pf->wol_en);
13262 		pci_set_power_state(pdev, PCI_D3hot);
13263 	}
13264 }
13265 
13266 #ifdef CONFIG_PM
13267 /**
13268  * i40e_suspend - PM callback for moving to D3
13269  * @dev: generic device information structure
13270  **/
13271 static int i40e_suspend(struct device *dev)
13272 {
13273 	struct pci_dev *pdev = to_pci_dev(dev);
13274 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13275 	struct i40e_hw *hw = &pf->hw;
13276 
13277 	/* If we're already suspended, then there is nothing to do */
13278 	if (test_and_set_bit(__I40E_SUSPENDED, pf->state))
13279 		return 0;
13280 
13281 	set_bit(__I40E_DOWN, pf->state);
13282 
13283 	/* Ensure service task will not be running */
13284 	del_timer_sync(&pf->service_timer);
13285 	cancel_work_sync(&pf->service_task);
13286 
13287 	if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
13288 		i40e_enable_mc_magic_wake(pf);
13289 
13290 	i40e_prep_for_reset(pf, false);
13291 
13292 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
13293 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
13294 
13295 	/* Clear the interrupt scheme and release our IRQs so that the system
13296 	 * can safely hibernate even when there are a large number of CPUs.
13297 	 * Otherwise hibernation might fail when mapping all the vectors back
13298 	 * to CPU0.
13299 	 */
13300 	i40e_clear_interrupt_scheme(pf);
13301 
13302 	return 0;
13303 }
13304 
13305 /**
13306  * i40e_resume - PM callback for waking up from D3
13307  * @dev: generic device information structure
13308  **/
13309 static int i40e_resume(struct device *dev)
13310 {
13311 	struct pci_dev *pdev = to_pci_dev(dev);
13312 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13313 	int err;
13314 
13315 	/* If we're not suspended, then there is nothing to do */
13316 	if (!test_bit(__I40E_SUSPENDED, pf->state))
13317 		return 0;
13318 
13319 	/* We cleared the interrupt scheme when we suspended, so we need to
13320 	 * restore it now to resume device functionality.
13321 	 */
13322 	err = i40e_restore_interrupt_scheme(pf);
13323 	if (err) {
13324 		dev_err(&pdev->dev, "Cannot restore interrupt scheme: %d\n",
13325 			err);
13326 	}
13327 
13328 	clear_bit(__I40E_DOWN, pf->state);
13329 	i40e_reset_and_rebuild(pf, false, false);
13330 
13331 	/* Clear suspended state last after everything is recovered */
13332 	clear_bit(__I40E_SUSPENDED, pf->state);
13333 
13334 	/* Restart the service task */
13335 	mod_timer(&pf->service_timer,
13336 		  round_jiffies(jiffies + pf->service_timer_period));
13337 
13338 	return 0;
13339 }
13340 
13341 #endif /* CONFIG_PM */
13342 
13343 static const struct pci_error_handlers i40e_err_handler = {
13344 	.error_detected = i40e_pci_error_detected,
13345 	.slot_reset = i40e_pci_error_slot_reset,
13346 	.reset_prepare = i40e_pci_error_reset_prepare,
13347 	.reset_done = i40e_pci_error_reset_done,
13348 	.resume = i40e_pci_error_resume,
13349 };
13350 
13351 static SIMPLE_DEV_PM_OPS(i40e_pm_ops, i40e_suspend, i40e_resume);
13352 
13353 static struct pci_driver i40e_driver = {
13354 	.name     = i40e_driver_name,
13355 	.id_table = i40e_pci_tbl,
13356 	.probe    = i40e_probe,
13357 	.remove   = i40e_remove,
13358 #ifdef CONFIG_PM
13359 	.driver   = {
13360 		.pm = &i40e_pm_ops,
13361 	},
13362 #endif /* CONFIG_PM */
13363 	.shutdown = i40e_shutdown,
13364 	.err_handler = &i40e_err_handler,
13365 	.sriov_configure = i40e_pci_sriov_configure,
13366 };
13367 
13368 /**
13369  * i40e_init_module - Driver registration routine
13370  *
13371  * i40e_init_module is the first routine called when the driver is
13372  * loaded. All it does is register with the PCI subsystem.
13373  **/
13374 static int __init i40e_init_module(void)
13375 {
13376 	pr_info("%s: %s - version %s\n", i40e_driver_name,
13377 		i40e_driver_string, i40e_driver_version_str);
13378 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
13379 
13380 	/* There is no need to throttle the number of active tasks because
13381 	 * each device limits its own task using a state bit for scheduling
13382 	 * the service task, and the device tasks do not interfere with each
13383 	 * other, so we don't set a max task limit. We must set WQ_MEM_RECLAIM
13384 	 * since we need to be able to guarantee forward progress even under
13385 	 * memory pressure.
13386 	 */
13387 	i40e_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, i40e_driver_name);
13388 	if (!i40e_wq) {
13389 		pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
13390 		return -ENOMEM;
13391 	}
13392 
13393 	i40e_dbg_init();
13394 	return pci_register_driver(&i40e_driver);
13395 }
13396 module_init(i40e_init_module);
13397 
13398 /**
13399  * i40e_exit_module - Driver exit cleanup routine
13400  *
13401  * i40e_exit_module is called just before the driver is removed
13402  * from memory.
13403  **/
13404 static void __exit i40e_exit_module(void)
13405 {
13406 	pci_unregister_driver(&i40e_driver);
13407 	destroy_workqueue(i40e_wq);
13408 	i40e_dbg_exit();
13409 }
13410 module_exit(i40e_exit_module);
13411