xref: /openbmc/linux/drivers/net/ethernet/intel/i40e/i40e_main.c (revision 2eb3ed33e55d003d721d4d1a5e72fe323c12b4c0)
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 static int i40e_add_del_cloud_filter(struct i40e_vsi *vsi,
73 				     struct i40e_cloud_filter *filter,
74 				     bool add);
75 static int i40e_add_del_cloud_filter_big_buf(struct i40e_vsi *vsi,
76 					     struct i40e_cloud_filter *filter,
77 					     bool add);
78 static int i40e_get_capabilities(struct i40e_pf *pf,
79 				 enum i40e_admin_queue_opc list_type);
80 
81 
82 /* i40e_pci_tbl - PCI Device ID Table
83  *
84  * Last entry must be all 0s
85  *
86  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
87  *   Class, Class Mask, private data (not used) }
88  */
89 static const struct pci_device_id i40e_pci_tbl[] = {
90 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
91 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
92 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
93 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
94 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
95 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
96 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
97 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
98 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
99 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
100 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
101 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
102 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
103 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
104 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
105 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
106 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
107 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_B), 0},
108 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_SFP28), 0},
109 	/* required last entry */
110 	{0, }
111 };
112 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
113 
114 #define I40E_MAX_VF_COUNT 128
115 static int debug = -1;
116 module_param(debug, uint, 0);
117 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
118 
119 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
120 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
121 MODULE_LICENSE("GPL");
122 MODULE_VERSION(DRV_VERSION);
123 
124 static struct workqueue_struct *i40e_wq;
125 
126 /**
127  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
128  * @hw:   pointer to the HW structure
129  * @mem:  ptr to mem struct to fill out
130  * @size: size of memory requested
131  * @alignment: what to align the allocation to
132  **/
133 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
134 			    u64 size, u32 alignment)
135 {
136 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
137 
138 	mem->size = ALIGN(size, alignment);
139 	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
140 				      &mem->pa, GFP_KERNEL);
141 	if (!mem->va)
142 		return -ENOMEM;
143 
144 	return 0;
145 }
146 
147 /**
148  * i40e_free_dma_mem_d - OS specific memory free for shared code
149  * @hw:   pointer to the HW structure
150  * @mem:  ptr to mem struct to free
151  **/
152 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
153 {
154 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
155 
156 	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
157 	mem->va = NULL;
158 	mem->pa = 0;
159 	mem->size = 0;
160 
161 	return 0;
162 }
163 
164 /**
165  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
166  * @hw:   pointer to the HW structure
167  * @mem:  ptr to mem struct to fill out
168  * @size: size of memory requested
169  **/
170 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
171 			     u32 size)
172 {
173 	mem->size = size;
174 	mem->va = kzalloc(size, GFP_KERNEL);
175 
176 	if (!mem->va)
177 		return -ENOMEM;
178 
179 	return 0;
180 }
181 
182 /**
183  * i40e_free_virt_mem_d - OS specific memory free for shared code
184  * @hw:   pointer to the HW structure
185  * @mem:  ptr to mem struct to free
186  **/
187 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
188 {
189 	/* it's ok to kfree a NULL pointer */
190 	kfree(mem->va);
191 	mem->va = NULL;
192 	mem->size = 0;
193 
194 	return 0;
195 }
196 
197 /**
198  * i40e_get_lump - find a lump of free generic resource
199  * @pf: board private structure
200  * @pile: the pile of resource to search
201  * @needed: the number of items needed
202  * @id: an owner id to stick on the items assigned
203  *
204  * Returns the base item index of the lump, or negative for error
205  *
206  * The search_hint trick and lack of advanced fit-finding only work
207  * because we're highly likely to have all the same size lump requests.
208  * Linear search time and any fragmentation should be minimal.
209  **/
210 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
211 			 u16 needed, u16 id)
212 {
213 	int ret = -ENOMEM;
214 	int i, j;
215 
216 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
217 		dev_info(&pf->pdev->dev,
218 			 "param err: pile=%p needed=%d id=0x%04x\n",
219 			 pile, needed, id);
220 		return -EINVAL;
221 	}
222 
223 	/* start the linear search with an imperfect hint */
224 	i = pile->search_hint;
225 	while (i < pile->num_entries) {
226 		/* skip already allocated entries */
227 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
228 			i++;
229 			continue;
230 		}
231 
232 		/* do we have enough in this lump? */
233 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
234 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
235 				break;
236 		}
237 
238 		if (j == needed) {
239 			/* there was enough, so assign it to the requestor */
240 			for (j = 0; j < needed; j++)
241 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
242 			ret = i;
243 			pile->search_hint = i + j;
244 			break;
245 		}
246 
247 		/* not enough, so skip over it and continue looking */
248 		i += j;
249 	}
250 
251 	return ret;
252 }
253 
254 /**
255  * i40e_put_lump - return a lump of generic resource
256  * @pile: the pile of resource to search
257  * @index: the base item index
258  * @id: the owner id of the items assigned
259  *
260  * Returns the count of items in the lump
261  **/
262 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
263 {
264 	int valid_id = (id | I40E_PILE_VALID_BIT);
265 	int count = 0;
266 	int i;
267 
268 	if (!pile || index >= pile->num_entries)
269 		return -EINVAL;
270 
271 	for (i = index;
272 	     i < pile->num_entries && pile->list[i] == valid_id;
273 	     i++) {
274 		pile->list[i] = 0;
275 		count++;
276 	}
277 
278 	if (count && index < pile->search_hint)
279 		pile->search_hint = index;
280 
281 	return count;
282 }
283 
284 /**
285  * i40e_find_vsi_from_id - searches for the vsi with the given id
286  * @pf - the pf structure to search for the vsi
287  * @id - id of the vsi it is searching for
288  **/
289 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
290 {
291 	int i;
292 
293 	for (i = 0; i < pf->num_alloc_vsi; i++)
294 		if (pf->vsi[i] && (pf->vsi[i]->id == id))
295 			return pf->vsi[i];
296 
297 	return NULL;
298 }
299 
300 /**
301  * i40e_service_event_schedule - Schedule the service task to wake up
302  * @pf: board private structure
303  *
304  * If not already scheduled, this puts the task into the work queue
305  **/
306 void i40e_service_event_schedule(struct i40e_pf *pf)
307 {
308 	if (!test_bit(__I40E_DOWN, pf->state) &&
309 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
310 		queue_work(i40e_wq, &pf->service_task);
311 }
312 
313 /**
314  * i40e_tx_timeout - Respond to a Tx Hang
315  * @netdev: network interface device structure
316  *
317  * If any port has noticed a Tx timeout, it is likely that the whole
318  * device is munged, not just the one netdev port, so go for the full
319  * reset.
320  **/
321 static void i40e_tx_timeout(struct net_device *netdev)
322 {
323 	struct i40e_netdev_priv *np = netdev_priv(netdev);
324 	struct i40e_vsi *vsi = np->vsi;
325 	struct i40e_pf *pf = vsi->back;
326 	struct i40e_ring *tx_ring = NULL;
327 	unsigned int i, hung_queue = 0;
328 	u32 head, val;
329 
330 	pf->tx_timeout_count++;
331 
332 	/* find the stopped queue the same way the stack does */
333 	for (i = 0; i < netdev->num_tx_queues; i++) {
334 		struct netdev_queue *q;
335 		unsigned long trans_start;
336 
337 		q = netdev_get_tx_queue(netdev, i);
338 		trans_start = q->trans_start;
339 		if (netif_xmit_stopped(q) &&
340 		    time_after(jiffies,
341 			       (trans_start + netdev->watchdog_timeo))) {
342 			hung_queue = i;
343 			break;
344 		}
345 	}
346 
347 	if (i == netdev->num_tx_queues) {
348 		netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
349 	} else {
350 		/* now that we have an index, find the tx_ring struct */
351 		for (i = 0; i < vsi->num_queue_pairs; i++) {
352 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
353 				if (hung_queue ==
354 				    vsi->tx_rings[i]->queue_index) {
355 					tx_ring = vsi->tx_rings[i];
356 					break;
357 				}
358 			}
359 		}
360 	}
361 
362 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
363 		pf->tx_timeout_recovery_level = 1;  /* reset after some time */
364 	else if (time_before(jiffies,
365 		      (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
366 		return;   /* don't do any new action before the next timeout */
367 
368 	if (tx_ring) {
369 		head = i40e_get_head(tx_ring);
370 		/* Read interrupt register */
371 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
372 			val = rd32(&pf->hw,
373 			     I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
374 						tx_ring->vsi->base_vector - 1));
375 		else
376 			val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
377 
378 		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",
379 			    vsi->seid, hung_queue, tx_ring->next_to_clean,
380 			    head, tx_ring->next_to_use,
381 			    readl(tx_ring->tail), val);
382 	}
383 
384 	pf->tx_timeout_last_recovery = jiffies;
385 	netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
386 		    pf->tx_timeout_recovery_level, hung_queue);
387 
388 	switch (pf->tx_timeout_recovery_level) {
389 	case 1:
390 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
391 		break;
392 	case 2:
393 		set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
394 		break;
395 	case 3:
396 		set_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
397 		break;
398 	default:
399 		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
400 		break;
401 	}
402 
403 	i40e_service_event_schedule(pf);
404 	pf->tx_timeout_recovery_level++;
405 }
406 
407 /**
408  * i40e_get_vsi_stats_struct - Get System Network Statistics
409  * @vsi: the VSI we care about
410  *
411  * Returns the address of the device statistics structure.
412  * The statistics are actually updated from the service task.
413  **/
414 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
415 {
416 	return &vsi->net_stats;
417 }
418 
419 /**
420  * i40e_get_netdev_stats_struct_tx - populate stats from a Tx ring
421  * @ring: Tx ring to get statistics from
422  * @stats: statistics entry to be updated
423  **/
424 static void i40e_get_netdev_stats_struct_tx(struct i40e_ring *ring,
425 					    struct rtnl_link_stats64 *stats)
426 {
427 	u64 bytes, packets;
428 	unsigned int start;
429 
430 	do {
431 		start = u64_stats_fetch_begin_irq(&ring->syncp);
432 		packets = ring->stats.packets;
433 		bytes   = ring->stats.bytes;
434 	} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
435 
436 	stats->tx_packets += packets;
437 	stats->tx_bytes   += bytes;
438 }
439 
440 /**
441  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
442  * @netdev: network interface device structure
443  *
444  * Returns the address of the device statistics structure.
445  * The statistics are actually updated from the service task.
446  **/
447 static void i40e_get_netdev_stats_struct(struct net_device *netdev,
448 				  struct rtnl_link_stats64 *stats)
449 {
450 	struct i40e_netdev_priv *np = netdev_priv(netdev);
451 	struct i40e_ring *tx_ring, *rx_ring;
452 	struct i40e_vsi *vsi = np->vsi;
453 	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
454 	int i;
455 
456 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
457 		return;
458 
459 	if (!vsi->tx_rings)
460 		return;
461 
462 	rcu_read_lock();
463 	for (i = 0; i < vsi->num_queue_pairs; i++) {
464 		u64 bytes, packets;
465 		unsigned int start;
466 
467 		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
468 		if (!tx_ring)
469 			continue;
470 		i40e_get_netdev_stats_struct_tx(tx_ring, stats);
471 
472 		rx_ring = &tx_ring[1];
473 
474 		do {
475 			start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
476 			packets = rx_ring->stats.packets;
477 			bytes   = rx_ring->stats.bytes;
478 		} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
479 
480 		stats->rx_packets += packets;
481 		stats->rx_bytes   += bytes;
482 
483 		if (i40e_enabled_xdp_vsi(vsi))
484 			i40e_get_netdev_stats_struct_tx(&rx_ring[1], stats);
485 	}
486 	rcu_read_unlock();
487 
488 	/* following stats updated by i40e_watchdog_subtask() */
489 	stats->multicast	= vsi_stats->multicast;
490 	stats->tx_errors	= vsi_stats->tx_errors;
491 	stats->tx_dropped	= vsi_stats->tx_dropped;
492 	stats->rx_errors	= vsi_stats->rx_errors;
493 	stats->rx_dropped	= vsi_stats->rx_dropped;
494 	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
495 	stats->rx_length_errors	= vsi_stats->rx_length_errors;
496 }
497 
498 /**
499  * i40e_vsi_reset_stats - Resets all stats of the given vsi
500  * @vsi: the VSI to have its stats reset
501  **/
502 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
503 {
504 	struct rtnl_link_stats64 *ns;
505 	int i;
506 
507 	if (!vsi)
508 		return;
509 
510 	ns = i40e_get_vsi_stats_struct(vsi);
511 	memset(ns, 0, sizeof(*ns));
512 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
513 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
514 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
515 	if (vsi->rx_rings && vsi->rx_rings[0]) {
516 		for (i = 0; i < vsi->num_queue_pairs; i++) {
517 			memset(&vsi->rx_rings[i]->stats, 0,
518 			       sizeof(vsi->rx_rings[i]->stats));
519 			memset(&vsi->rx_rings[i]->rx_stats, 0,
520 			       sizeof(vsi->rx_rings[i]->rx_stats));
521 			memset(&vsi->tx_rings[i]->stats, 0,
522 			       sizeof(vsi->tx_rings[i]->stats));
523 			memset(&vsi->tx_rings[i]->tx_stats, 0,
524 			       sizeof(vsi->tx_rings[i]->tx_stats));
525 		}
526 	}
527 	vsi->stat_offsets_loaded = false;
528 }
529 
530 /**
531  * i40e_pf_reset_stats - Reset all of the stats for the given PF
532  * @pf: the PF to be reset
533  **/
534 void i40e_pf_reset_stats(struct i40e_pf *pf)
535 {
536 	int i;
537 
538 	memset(&pf->stats, 0, sizeof(pf->stats));
539 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
540 	pf->stat_offsets_loaded = false;
541 
542 	for (i = 0; i < I40E_MAX_VEB; i++) {
543 		if (pf->veb[i]) {
544 			memset(&pf->veb[i]->stats, 0,
545 			       sizeof(pf->veb[i]->stats));
546 			memset(&pf->veb[i]->stats_offsets, 0,
547 			       sizeof(pf->veb[i]->stats_offsets));
548 			pf->veb[i]->stat_offsets_loaded = false;
549 		}
550 	}
551 	pf->hw_csum_rx_error = 0;
552 }
553 
554 /**
555  * i40e_stat_update48 - read and update a 48 bit stat from the chip
556  * @hw: ptr to the hardware info
557  * @hireg: the high 32 bit reg to read
558  * @loreg: the low 32 bit reg to read
559  * @offset_loaded: has the initial offset been loaded yet
560  * @offset: ptr to current offset value
561  * @stat: ptr to the stat
562  *
563  * Since the device stats are not reset at PFReset, they likely will not
564  * be zeroed when the driver starts.  We'll save the first values read
565  * and use them as offsets to be subtracted from the raw values in order
566  * to report stats that count from zero.  In the process, we also manage
567  * the potential roll-over.
568  **/
569 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
570 			       bool offset_loaded, u64 *offset, u64 *stat)
571 {
572 	u64 new_data;
573 
574 	if (hw->device_id == I40E_DEV_ID_QEMU) {
575 		new_data = rd32(hw, loreg);
576 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
577 	} else {
578 		new_data = rd64(hw, loreg);
579 	}
580 	if (!offset_loaded)
581 		*offset = new_data;
582 	if (likely(new_data >= *offset))
583 		*stat = new_data - *offset;
584 	else
585 		*stat = (new_data + BIT_ULL(48)) - *offset;
586 	*stat &= 0xFFFFFFFFFFFFULL;
587 }
588 
589 /**
590  * i40e_stat_update32 - read and update a 32 bit stat from the chip
591  * @hw: ptr to the hardware info
592  * @reg: the hw reg to read
593  * @offset_loaded: has the initial offset been loaded yet
594  * @offset: ptr to current offset value
595  * @stat: ptr to the stat
596  **/
597 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
598 			       bool offset_loaded, u64 *offset, u64 *stat)
599 {
600 	u32 new_data;
601 
602 	new_data = rd32(hw, reg);
603 	if (!offset_loaded)
604 		*offset = new_data;
605 	if (likely(new_data >= *offset))
606 		*stat = (u32)(new_data - *offset);
607 	else
608 		*stat = (u32)((new_data + BIT_ULL(32)) - *offset);
609 }
610 
611 /**
612  * i40e_stat_update_and_clear32 - read and clear hw reg, update a 32 bit stat
613  * @hw: ptr to the hardware info
614  * @reg: the hw reg to read and clear
615  * @stat: ptr to the stat
616  **/
617 static void i40e_stat_update_and_clear32(struct i40e_hw *hw, u32 reg, u64 *stat)
618 {
619 	u32 new_data = rd32(hw, reg);
620 
621 	wr32(hw, reg, 1); /* must write a nonzero value to clear register */
622 	*stat += new_data;
623 }
624 
625 /**
626  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
627  * @vsi: the VSI to be updated
628  **/
629 void i40e_update_eth_stats(struct i40e_vsi *vsi)
630 {
631 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
632 	struct i40e_pf *pf = vsi->back;
633 	struct i40e_hw *hw = &pf->hw;
634 	struct i40e_eth_stats *oes;
635 	struct i40e_eth_stats *es;     /* device's eth stats */
636 
637 	es = &vsi->eth_stats;
638 	oes = &vsi->eth_stats_offsets;
639 
640 	/* Gather up the stats that the hw collects */
641 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
642 			   vsi->stat_offsets_loaded,
643 			   &oes->tx_errors, &es->tx_errors);
644 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
645 			   vsi->stat_offsets_loaded,
646 			   &oes->rx_discards, &es->rx_discards);
647 	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
648 			   vsi->stat_offsets_loaded,
649 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
650 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
651 			   vsi->stat_offsets_loaded,
652 			   &oes->tx_errors, &es->tx_errors);
653 
654 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
655 			   I40E_GLV_GORCL(stat_idx),
656 			   vsi->stat_offsets_loaded,
657 			   &oes->rx_bytes, &es->rx_bytes);
658 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
659 			   I40E_GLV_UPRCL(stat_idx),
660 			   vsi->stat_offsets_loaded,
661 			   &oes->rx_unicast, &es->rx_unicast);
662 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
663 			   I40E_GLV_MPRCL(stat_idx),
664 			   vsi->stat_offsets_loaded,
665 			   &oes->rx_multicast, &es->rx_multicast);
666 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
667 			   I40E_GLV_BPRCL(stat_idx),
668 			   vsi->stat_offsets_loaded,
669 			   &oes->rx_broadcast, &es->rx_broadcast);
670 
671 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
672 			   I40E_GLV_GOTCL(stat_idx),
673 			   vsi->stat_offsets_loaded,
674 			   &oes->tx_bytes, &es->tx_bytes);
675 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
676 			   I40E_GLV_UPTCL(stat_idx),
677 			   vsi->stat_offsets_loaded,
678 			   &oes->tx_unicast, &es->tx_unicast);
679 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
680 			   I40E_GLV_MPTCL(stat_idx),
681 			   vsi->stat_offsets_loaded,
682 			   &oes->tx_multicast, &es->tx_multicast);
683 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
684 			   I40E_GLV_BPTCL(stat_idx),
685 			   vsi->stat_offsets_loaded,
686 			   &oes->tx_broadcast, &es->tx_broadcast);
687 	vsi->stat_offsets_loaded = true;
688 }
689 
690 /**
691  * i40e_update_veb_stats - Update Switch component statistics
692  * @veb: the VEB being updated
693  **/
694 static void i40e_update_veb_stats(struct i40e_veb *veb)
695 {
696 	struct i40e_pf *pf = veb->pf;
697 	struct i40e_hw *hw = &pf->hw;
698 	struct i40e_eth_stats *oes;
699 	struct i40e_eth_stats *es;     /* device's eth stats */
700 	struct i40e_veb_tc_stats *veb_oes;
701 	struct i40e_veb_tc_stats *veb_es;
702 	int i, idx = 0;
703 
704 	idx = veb->stats_idx;
705 	es = &veb->stats;
706 	oes = &veb->stats_offsets;
707 	veb_es = &veb->tc_stats;
708 	veb_oes = &veb->tc_stats_offsets;
709 
710 	/* Gather up the stats that the hw collects */
711 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
712 			   veb->stat_offsets_loaded,
713 			   &oes->tx_discards, &es->tx_discards);
714 	if (hw->revision_id > 0)
715 		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
716 				   veb->stat_offsets_loaded,
717 				   &oes->rx_unknown_protocol,
718 				   &es->rx_unknown_protocol);
719 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
720 			   veb->stat_offsets_loaded,
721 			   &oes->rx_bytes, &es->rx_bytes);
722 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
723 			   veb->stat_offsets_loaded,
724 			   &oes->rx_unicast, &es->rx_unicast);
725 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
726 			   veb->stat_offsets_loaded,
727 			   &oes->rx_multicast, &es->rx_multicast);
728 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
729 			   veb->stat_offsets_loaded,
730 			   &oes->rx_broadcast, &es->rx_broadcast);
731 
732 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
733 			   veb->stat_offsets_loaded,
734 			   &oes->tx_bytes, &es->tx_bytes);
735 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
736 			   veb->stat_offsets_loaded,
737 			   &oes->tx_unicast, &es->tx_unicast);
738 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
739 			   veb->stat_offsets_loaded,
740 			   &oes->tx_multicast, &es->tx_multicast);
741 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
742 			   veb->stat_offsets_loaded,
743 			   &oes->tx_broadcast, &es->tx_broadcast);
744 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
745 		i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
746 				   I40E_GLVEBTC_RPCL(i, idx),
747 				   veb->stat_offsets_loaded,
748 				   &veb_oes->tc_rx_packets[i],
749 				   &veb_es->tc_rx_packets[i]);
750 		i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
751 				   I40E_GLVEBTC_RBCL(i, idx),
752 				   veb->stat_offsets_loaded,
753 				   &veb_oes->tc_rx_bytes[i],
754 				   &veb_es->tc_rx_bytes[i]);
755 		i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
756 				   I40E_GLVEBTC_TPCL(i, idx),
757 				   veb->stat_offsets_loaded,
758 				   &veb_oes->tc_tx_packets[i],
759 				   &veb_es->tc_tx_packets[i]);
760 		i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
761 				   I40E_GLVEBTC_TBCL(i, idx),
762 				   veb->stat_offsets_loaded,
763 				   &veb_oes->tc_tx_bytes[i],
764 				   &veb_es->tc_tx_bytes[i]);
765 	}
766 	veb->stat_offsets_loaded = true;
767 }
768 
769 /**
770  * i40e_update_vsi_stats - Update the vsi statistics counters.
771  * @vsi: the VSI to be updated
772  *
773  * There are a few instances where we store the same stat in a
774  * couple of different structs.  This is partly because we have
775  * the netdev stats that need to be filled out, which is slightly
776  * different from the "eth_stats" defined by the chip and used in
777  * VF communications.  We sort it out here.
778  **/
779 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
780 {
781 	struct i40e_pf *pf = vsi->back;
782 	struct rtnl_link_stats64 *ons;
783 	struct rtnl_link_stats64 *ns;   /* netdev stats */
784 	struct i40e_eth_stats *oes;
785 	struct i40e_eth_stats *es;     /* device's eth stats */
786 	u32 tx_restart, tx_busy;
787 	struct i40e_ring *p;
788 	u32 rx_page, rx_buf;
789 	u64 bytes, packets;
790 	unsigned int start;
791 	u64 tx_linearize;
792 	u64 tx_force_wb;
793 	u64 rx_p, rx_b;
794 	u64 tx_p, tx_b;
795 	u16 q;
796 
797 	if (test_bit(__I40E_VSI_DOWN, vsi->state) ||
798 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
799 		return;
800 
801 	ns = i40e_get_vsi_stats_struct(vsi);
802 	ons = &vsi->net_stats_offsets;
803 	es = &vsi->eth_stats;
804 	oes = &vsi->eth_stats_offsets;
805 
806 	/* Gather up the netdev and vsi stats that the driver collects
807 	 * on the fly during packet processing
808 	 */
809 	rx_b = rx_p = 0;
810 	tx_b = tx_p = 0;
811 	tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
812 	rx_page = 0;
813 	rx_buf = 0;
814 	rcu_read_lock();
815 	for (q = 0; q < vsi->num_queue_pairs; q++) {
816 		/* locate Tx ring */
817 		p = ACCESS_ONCE(vsi->tx_rings[q]);
818 
819 		do {
820 			start = u64_stats_fetch_begin_irq(&p->syncp);
821 			packets = p->stats.packets;
822 			bytes = p->stats.bytes;
823 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
824 		tx_b += bytes;
825 		tx_p += packets;
826 		tx_restart += p->tx_stats.restart_queue;
827 		tx_busy += p->tx_stats.tx_busy;
828 		tx_linearize += p->tx_stats.tx_linearize;
829 		tx_force_wb += p->tx_stats.tx_force_wb;
830 
831 		/* Rx queue is part of the same block as Tx queue */
832 		p = &p[1];
833 		do {
834 			start = u64_stats_fetch_begin_irq(&p->syncp);
835 			packets = p->stats.packets;
836 			bytes = p->stats.bytes;
837 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
838 		rx_b += bytes;
839 		rx_p += packets;
840 		rx_buf += p->rx_stats.alloc_buff_failed;
841 		rx_page += p->rx_stats.alloc_page_failed;
842 	}
843 	rcu_read_unlock();
844 	vsi->tx_restart = tx_restart;
845 	vsi->tx_busy = tx_busy;
846 	vsi->tx_linearize = tx_linearize;
847 	vsi->tx_force_wb = tx_force_wb;
848 	vsi->rx_page_failed = rx_page;
849 	vsi->rx_buf_failed = rx_buf;
850 
851 	ns->rx_packets = rx_p;
852 	ns->rx_bytes = rx_b;
853 	ns->tx_packets = tx_p;
854 	ns->tx_bytes = tx_b;
855 
856 	/* update netdev stats from eth stats */
857 	i40e_update_eth_stats(vsi);
858 	ons->tx_errors = oes->tx_errors;
859 	ns->tx_errors = es->tx_errors;
860 	ons->multicast = oes->rx_multicast;
861 	ns->multicast = es->rx_multicast;
862 	ons->rx_dropped = oes->rx_discards;
863 	ns->rx_dropped = es->rx_discards;
864 	ons->tx_dropped = oes->tx_discards;
865 	ns->tx_dropped = es->tx_discards;
866 
867 	/* pull in a couple PF stats if this is the main vsi */
868 	if (vsi == pf->vsi[pf->lan_vsi]) {
869 		ns->rx_crc_errors = pf->stats.crc_errors;
870 		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
871 		ns->rx_length_errors = pf->stats.rx_length_errors;
872 	}
873 }
874 
875 /**
876  * i40e_update_pf_stats - Update the PF statistics counters.
877  * @pf: the PF to be updated
878  **/
879 static void i40e_update_pf_stats(struct i40e_pf *pf)
880 {
881 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
882 	struct i40e_hw_port_stats *nsd = &pf->stats;
883 	struct i40e_hw *hw = &pf->hw;
884 	u32 val;
885 	int i;
886 
887 	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
888 			   I40E_GLPRT_GORCL(hw->port),
889 			   pf->stat_offsets_loaded,
890 			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
891 	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
892 			   I40E_GLPRT_GOTCL(hw->port),
893 			   pf->stat_offsets_loaded,
894 			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
895 	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
896 			   pf->stat_offsets_loaded,
897 			   &osd->eth.rx_discards,
898 			   &nsd->eth.rx_discards);
899 	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
900 			   I40E_GLPRT_UPRCL(hw->port),
901 			   pf->stat_offsets_loaded,
902 			   &osd->eth.rx_unicast,
903 			   &nsd->eth.rx_unicast);
904 	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
905 			   I40E_GLPRT_MPRCL(hw->port),
906 			   pf->stat_offsets_loaded,
907 			   &osd->eth.rx_multicast,
908 			   &nsd->eth.rx_multicast);
909 	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
910 			   I40E_GLPRT_BPRCL(hw->port),
911 			   pf->stat_offsets_loaded,
912 			   &osd->eth.rx_broadcast,
913 			   &nsd->eth.rx_broadcast);
914 	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
915 			   I40E_GLPRT_UPTCL(hw->port),
916 			   pf->stat_offsets_loaded,
917 			   &osd->eth.tx_unicast,
918 			   &nsd->eth.tx_unicast);
919 	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
920 			   I40E_GLPRT_MPTCL(hw->port),
921 			   pf->stat_offsets_loaded,
922 			   &osd->eth.tx_multicast,
923 			   &nsd->eth.tx_multicast);
924 	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
925 			   I40E_GLPRT_BPTCL(hw->port),
926 			   pf->stat_offsets_loaded,
927 			   &osd->eth.tx_broadcast,
928 			   &nsd->eth.tx_broadcast);
929 
930 	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
931 			   pf->stat_offsets_loaded,
932 			   &osd->tx_dropped_link_down,
933 			   &nsd->tx_dropped_link_down);
934 
935 	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
936 			   pf->stat_offsets_loaded,
937 			   &osd->crc_errors, &nsd->crc_errors);
938 
939 	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
940 			   pf->stat_offsets_loaded,
941 			   &osd->illegal_bytes, &nsd->illegal_bytes);
942 
943 	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
944 			   pf->stat_offsets_loaded,
945 			   &osd->mac_local_faults,
946 			   &nsd->mac_local_faults);
947 	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
948 			   pf->stat_offsets_loaded,
949 			   &osd->mac_remote_faults,
950 			   &nsd->mac_remote_faults);
951 
952 	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
953 			   pf->stat_offsets_loaded,
954 			   &osd->rx_length_errors,
955 			   &nsd->rx_length_errors);
956 
957 	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
958 			   pf->stat_offsets_loaded,
959 			   &osd->link_xon_rx, &nsd->link_xon_rx);
960 	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
961 			   pf->stat_offsets_loaded,
962 			   &osd->link_xon_tx, &nsd->link_xon_tx);
963 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
964 			   pf->stat_offsets_loaded,
965 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
966 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
967 			   pf->stat_offsets_loaded,
968 			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
969 
970 	for (i = 0; i < 8; i++) {
971 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
972 				   pf->stat_offsets_loaded,
973 				   &osd->priority_xoff_rx[i],
974 				   &nsd->priority_xoff_rx[i]);
975 		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
976 				   pf->stat_offsets_loaded,
977 				   &osd->priority_xon_rx[i],
978 				   &nsd->priority_xon_rx[i]);
979 		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
980 				   pf->stat_offsets_loaded,
981 				   &osd->priority_xon_tx[i],
982 				   &nsd->priority_xon_tx[i]);
983 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
984 				   pf->stat_offsets_loaded,
985 				   &osd->priority_xoff_tx[i],
986 				   &nsd->priority_xoff_tx[i]);
987 		i40e_stat_update32(hw,
988 				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
989 				   pf->stat_offsets_loaded,
990 				   &osd->priority_xon_2_xoff[i],
991 				   &nsd->priority_xon_2_xoff[i]);
992 	}
993 
994 	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
995 			   I40E_GLPRT_PRC64L(hw->port),
996 			   pf->stat_offsets_loaded,
997 			   &osd->rx_size_64, &nsd->rx_size_64);
998 	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
999 			   I40E_GLPRT_PRC127L(hw->port),
1000 			   pf->stat_offsets_loaded,
1001 			   &osd->rx_size_127, &nsd->rx_size_127);
1002 	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1003 			   I40E_GLPRT_PRC255L(hw->port),
1004 			   pf->stat_offsets_loaded,
1005 			   &osd->rx_size_255, &nsd->rx_size_255);
1006 	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1007 			   I40E_GLPRT_PRC511L(hw->port),
1008 			   pf->stat_offsets_loaded,
1009 			   &osd->rx_size_511, &nsd->rx_size_511);
1010 	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1011 			   I40E_GLPRT_PRC1023L(hw->port),
1012 			   pf->stat_offsets_loaded,
1013 			   &osd->rx_size_1023, &nsd->rx_size_1023);
1014 	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1015 			   I40E_GLPRT_PRC1522L(hw->port),
1016 			   pf->stat_offsets_loaded,
1017 			   &osd->rx_size_1522, &nsd->rx_size_1522);
1018 	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1019 			   I40E_GLPRT_PRC9522L(hw->port),
1020 			   pf->stat_offsets_loaded,
1021 			   &osd->rx_size_big, &nsd->rx_size_big);
1022 
1023 	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1024 			   I40E_GLPRT_PTC64L(hw->port),
1025 			   pf->stat_offsets_loaded,
1026 			   &osd->tx_size_64, &nsd->tx_size_64);
1027 	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1028 			   I40E_GLPRT_PTC127L(hw->port),
1029 			   pf->stat_offsets_loaded,
1030 			   &osd->tx_size_127, &nsd->tx_size_127);
1031 	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1032 			   I40E_GLPRT_PTC255L(hw->port),
1033 			   pf->stat_offsets_loaded,
1034 			   &osd->tx_size_255, &nsd->tx_size_255);
1035 	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1036 			   I40E_GLPRT_PTC511L(hw->port),
1037 			   pf->stat_offsets_loaded,
1038 			   &osd->tx_size_511, &nsd->tx_size_511);
1039 	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1040 			   I40E_GLPRT_PTC1023L(hw->port),
1041 			   pf->stat_offsets_loaded,
1042 			   &osd->tx_size_1023, &nsd->tx_size_1023);
1043 	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1044 			   I40E_GLPRT_PTC1522L(hw->port),
1045 			   pf->stat_offsets_loaded,
1046 			   &osd->tx_size_1522, &nsd->tx_size_1522);
1047 	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1048 			   I40E_GLPRT_PTC9522L(hw->port),
1049 			   pf->stat_offsets_loaded,
1050 			   &osd->tx_size_big, &nsd->tx_size_big);
1051 
1052 	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1053 			   pf->stat_offsets_loaded,
1054 			   &osd->rx_undersize, &nsd->rx_undersize);
1055 	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1056 			   pf->stat_offsets_loaded,
1057 			   &osd->rx_fragments, &nsd->rx_fragments);
1058 	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1059 			   pf->stat_offsets_loaded,
1060 			   &osd->rx_oversize, &nsd->rx_oversize);
1061 	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1062 			   pf->stat_offsets_loaded,
1063 			   &osd->rx_jabber, &nsd->rx_jabber);
1064 
1065 	/* FDIR stats */
1066 	i40e_stat_update_and_clear32(hw,
1067 			I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(hw->pf_id)),
1068 			&nsd->fd_atr_match);
1069 	i40e_stat_update_and_clear32(hw,
1070 			I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(hw->pf_id)),
1071 			&nsd->fd_sb_match);
1072 	i40e_stat_update_and_clear32(hw,
1073 			I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(hw->pf_id)),
1074 			&nsd->fd_atr_tunnel_match);
1075 
1076 	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1077 	nsd->tx_lpi_status =
1078 		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1079 			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1080 	nsd->rx_lpi_status =
1081 		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1082 			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1083 	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1084 			   pf->stat_offsets_loaded,
1085 			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1086 	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1087 			   pf->stat_offsets_loaded,
1088 			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1089 
1090 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1091 	    !(pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED))
1092 		nsd->fd_sb_status = true;
1093 	else
1094 		nsd->fd_sb_status = false;
1095 
1096 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1097 	    !(pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED))
1098 		nsd->fd_atr_status = true;
1099 	else
1100 		nsd->fd_atr_status = false;
1101 
1102 	pf->stat_offsets_loaded = true;
1103 }
1104 
1105 /**
1106  * i40e_update_stats - Update the various statistics counters.
1107  * @vsi: the VSI to be updated
1108  *
1109  * Update the various stats for this VSI and its related entities.
1110  **/
1111 void i40e_update_stats(struct i40e_vsi *vsi)
1112 {
1113 	struct i40e_pf *pf = vsi->back;
1114 
1115 	if (vsi == pf->vsi[pf->lan_vsi])
1116 		i40e_update_pf_stats(pf);
1117 
1118 	i40e_update_vsi_stats(vsi);
1119 }
1120 
1121 /**
1122  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1123  * @vsi: the VSI to be searched
1124  * @macaddr: the MAC address
1125  * @vlan: the vlan
1126  *
1127  * Returns ptr to the filter object or NULL
1128  **/
1129 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1130 						const u8 *macaddr, s16 vlan)
1131 {
1132 	struct i40e_mac_filter *f;
1133 	u64 key;
1134 
1135 	if (!vsi || !macaddr)
1136 		return NULL;
1137 
1138 	key = i40e_addr_to_hkey(macaddr);
1139 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1140 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1141 		    (vlan == f->vlan))
1142 			return f;
1143 	}
1144 	return NULL;
1145 }
1146 
1147 /**
1148  * i40e_find_mac - Find a mac addr in the macvlan filters list
1149  * @vsi: the VSI to be searched
1150  * @macaddr: the MAC address we are searching for
1151  *
1152  * Returns the first filter with the provided MAC address or NULL if
1153  * MAC address was not found
1154  **/
1155 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1156 {
1157 	struct i40e_mac_filter *f;
1158 	u64 key;
1159 
1160 	if (!vsi || !macaddr)
1161 		return NULL;
1162 
1163 	key = i40e_addr_to_hkey(macaddr);
1164 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1165 		if ((ether_addr_equal(macaddr, f->macaddr)))
1166 			return f;
1167 	}
1168 	return NULL;
1169 }
1170 
1171 /**
1172  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1173  * @vsi: the VSI to be searched
1174  *
1175  * Returns true if VSI is in vlan mode or false otherwise
1176  **/
1177 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1178 {
1179 	/* If we have a PVID, always operate in VLAN mode */
1180 	if (vsi->info.pvid)
1181 		return true;
1182 
1183 	/* We need to operate in VLAN mode whenever we have any filters with
1184 	 * a VLAN other than I40E_VLAN_ALL. We could check the table each
1185 	 * time, incurring search cost repeatedly. However, we can notice two
1186 	 * things:
1187 	 *
1188 	 * 1) the only place where we can gain a VLAN filter is in
1189 	 *    i40e_add_filter.
1190 	 *
1191 	 * 2) the only place where filters are actually removed is in
1192 	 *    i40e_sync_filters_subtask.
1193 	 *
1194 	 * Thus, we can simply use a boolean value, has_vlan_filters which we
1195 	 * will set to true when we add a VLAN filter in i40e_add_filter. Then
1196 	 * we have to perform the full search after deleting filters in
1197 	 * i40e_sync_filters_subtask, but we already have to search
1198 	 * filters here and can perform the check at the same time. This
1199 	 * results in avoiding embedding a loop for VLAN mode inside another
1200 	 * loop over all the filters, and should maintain correctness as noted
1201 	 * above.
1202 	 */
1203 	return vsi->has_vlan_filter;
1204 }
1205 
1206 /**
1207  * i40e_correct_mac_vlan_filters - Correct non-VLAN filters if necessary
1208  * @vsi: the VSI to configure
1209  * @tmp_add_list: list of filters ready to be added
1210  * @tmp_del_list: list of filters ready to be deleted
1211  * @vlan_filters: the number of active VLAN filters
1212  *
1213  * Update VLAN=0 and VLAN=-1 (I40E_VLAN_ANY) filters properly so that they
1214  * behave as expected. If we have any active VLAN filters remaining or about
1215  * to be added then we need to update non-VLAN filters to be marked as VLAN=0
1216  * so that they only match against untagged traffic. If we no longer have any
1217  * active VLAN filters, we need to make all non-VLAN filters marked as VLAN=-1
1218  * so that they match against both tagged and untagged traffic. In this way,
1219  * we ensure that we correctly receive the desired traffic. This ensures that
1220  * when we have an active VLAN we will receive only untagged traffic and
1221  * traffic matching active VLANs. If we have no active VLANs then we will
1222  * operate in non-VLAN mode and receive all traffic, tagged or untagged.
1223  *
1224  * Finally, in a similar fashion, this function also corrects filters when
1225  * there is an active PVID assigned to this VSI.
1226  *
1227  * In case of memory allocation failure return -ENOMEM. Otherwise, return 0.
1228  *
1229  * This function is only expected to be called from within
1230  * i40e_sync_vsi_filters.
1231  *
1232  * NOTE: This function expects to be called while under the
1233  * mac_filter_hash_lock
1234  */
1235 static int i40e_correct_mac_vlan_filters(struct i40e_vsi *vsi,
1236 					 struct hlist_head *tmp_add_list,
1237 					 struct hlist_head *tmp_del_list,
1238 					 int vlan_filters)
1239 {
1240 	s16 pvid = le16_to_cpu(vsi->info.pvid);
1241 	struct i40e_mac_filter *f, *add_head;
1242 	struct i40e_new_mac_filter *new;
1243 	struct hlist_node *h;
1244 	int bkt, new_vlan;
1245 
1246 	/* To determine if a particular filter needs to be replaced we
1247 	 * have the three following conditions:
1248 	 *
1249 	 * a) if we have a PVID assigned, then all filters which are
1250 	 *    not marked as VLAN=PVID must be replaced with filters that
1251 	 *    are.
1252 	 * b) otherwise, if we have any active VLANS, all filters
1253 	 *    which are marked as VLAN=-1 must be replaced with
1254 	 *    filters marked as VLAN=0
1255 	 * c) finally, if we do not have any active VLANS, all filters
1256 	 *    which are marked as VLAN=0 must be replaced with filters
1257 	 *    marked as VLAN=-1
1258 	 */
1259 
1260 	/* Update the filters about to be added in place */
1261 	hlist_for_each_entry(new, tmp_add_list, hlist) {
1262 		if (pvid && new->f->vlan != pvid)
1263 			new->f->vlan = pvid;
1264 		else if (vlan_filters && new->f->vlan == I40E_VLAN_ANY)
1265 			new->f->vlan = 0;
1266 		else if (!vlan_filters && new->f->vlan == 0)
1267 			new->f->vlan = I40E_VLAN_ANY;
1268 	}
1269 
1270 	/* Update the remaining active filters */
1271 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1272 		/* Combine the checks for whether a filter needs to be changed
1273 		 * and then determine the new VLAN inside the if block, in
1274 		 * order to avoid duplicating code for adding the new filter
1275 		 * then deleting the old filter.
1276 		 */
1277 		if ((pvid && f->vlan != pvid) ||
1278 		    (vlan_filters && f->vlan == I40E_VLAN_ANY) ||
1279 		    (!vlan_filters && f->vlan == 0)) {
1280 			/* Determine the new vlan we will be adding */
1281 			if (pvid)
1282 				new_vlan = pvid;
1283 			else if (vlan_filters)
1284 				new_vlan = 0;
1285 			else
1286 				new_vlan = I40E_VLAN_ANY;
1287 
1288 			/* Create the new filter */
1289 			add_head = i40e_add_filter(vsi, f->macaddr, new_vlan);
1290 			if (!add_head)
1291 				return -ENOMEM;
1292 
1293 			/* Create a temporary i40e_new_mac_filter */
1294 			new = kzalloc(sizeof(*new), GFP_ATOMIC);
1295 			if (!new)
1296 				return -ENOMEM;
1297 
1298 			new->f = add_head;
1299 			new->state = add_head->state;
1300 
1301 			/* Add the new filter to the tmp list */
1302 			hlist_add_head(&new->hlist, tmp_add_list);
1303 
1304 			/* Put the original filter into the delete list */
1305 			f->state = I40E_FILTER_REMOVE;
1306 			hash_del(&f->hlist);
1307 			hlist_add_head(&f->hlist, tmp_del_list);
1308 		}
1309 	}
1310 
1311 	vsi->has_vlan_filter = !!vlan_filters;
1312 
1313 	return 0;
1314 }
1315 
1316 /**
1317  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1318  * @vsi: the PF Main VSI - inappropriate for any other VSI
1319  * @macaddr: the MAC address
1320  *
1321  * Remove whatever filter the firmware set up so the driver can manage
1322  * its own filtering intelligently.
1323  **/
1324 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1325 {
1326 	struct i40e_aqc_remove_macvlan_element_data element;
1327 	struct i40e_pf *pf = vsi->back;
1328 
1329 	/* Only appropriate for the PF main VSI */
1330 	if (vsi->type != I40E_VSI_MAIN)
1331 		return;
1332 
1333 	memset(&element, 0, sizeof(element));
1334 	ether_addr_copy(element.mac_addr, macaddr);
1335 	element.vlan_tag = 0;
1336 	/* Ignore error returns, some firmware does it this way... */
1337 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1338 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1339 
1340 	memset(&element, 0, sizeof(element));
1341 	ether_addr_copy(element.mac_addr, macaddr);
1342 	element.vlan_tag = 0;
1343 	/* ...and some firmware does it this way. */
1344 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1345 			I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1346 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1347 }
1348 
1349 /**
1350  * i40e_add_filter - Add a mac/vlan filter to the VSI
1351  * @vsi: the VSI to be searched
1352  * @macaddr: the MAC address
1353  * @vlan: the vlan
1354  *
1355  * Returns ptr to the filter object or NULL when no memory available.
1356  *
1357  * NOTE: This function is expected to be called with mac_filter_hash_lock
1358  * being held.
1359  **/
1360 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1361 					const u8 *macaddr, s16 vlan)
1362 {
1363 	struct i40e_mac_filter *f;
1364 	u64 key;
1365 
1366 	if (!vsi || !macaddr)
1367 		return NULL;
1368 
1369 	f = i40e_find_filter(vsi, macaddr, vlan);
1370 	if (!f) {
1371 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1372 		if (!f)
1373 			return NULL;
1374 
1375 		/* Update the boolean indicating if we need to function in
1376 		 * VLAN mode.
1377 		 */
1378 		if (vlan >= 0)
1379 			vsi->has_vlan_filter = true;
1380 
1381 		ether_addr_copy(f->macaddr, macaddr);
1382 		f->vlan = vlan;
1383 		/* If we're in overflow promisc mode, set the state directly
1384 		 * to failed, so we don't bother to try sending the filter
1385 		 * to the hardware.
1386 		 */
1387 		if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state))
1388 			f->state = I40E_FILTER_FAILED;
1389 		else
1390 			f->state = I40E_FILTER_NEW;
1391 		INIT_HLIST_NODE(&f->hlist);
1392 
1393 		key = i40e_addr_to_hkey(macaddr);
1394 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1395 
1396 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1397 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1398 	}
1399 
1400 	/* If we're asked to add a filter that has been marked for removal, it
1401 	 * is safe to simply restore it to active state. __i40e_del_filter
1402 	 * will have simply deleted any filters which were previously marked
1403 	 * NEW or FAILED, so if it is currently marked REMOVE it must have
1404 	 * previously been ACTIVE. Since we haven't yet run the sync filters
1405 	 * task, just restore this filter to the ACTIVE state so that the
1406 	 * sync task leaves it in place
1407 	 */
1408 	if (f->state == I40E_FILTER_REMOVE)
1409 		f->state = I40E_FILTER_ACTIVE;
1410 
1411 	return f;
1412 }
1413 
1414 /**
1415  * __i40e_del_filter - Remove a specific filter from the VSI
1416  * @vsi: VSI to remove from
1417  * @f: the filter to remove from the list
1418  *
1419  * This function should be called instead of i40e_del_filter only if you know
1420  * the exact filter you will remove already, such as via i40e_find_filter or
1421  * i40e_find_mac.
1422  *
1423  * NOTE: This function is expected to be called with mac_filter_hash_lock
1424  * being held.
1425  * ANOTHER NOTE: This function MUST be called from within the context of
1426  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1427  * instead of list_for_each_entry().
1428  **/
1429 void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1430 {
1431 	if (!f)
1432 		return;
1433 
1434 	/* If the filter was never added to firmware then we can just delete it
1435 	 * directly and we don't want to set the status to remove or else an
1436 	 * admin queue command will unnecessarily fire.
1437 	 */
1438 	if ((f->state == I40E_FILTER_FAILED) ||
1439 	    (f->state == I40E_FILTER_NEW)) {
1440 		hash_del(&f->hlist);
1441 		kfree(f);
1442 	} else {
1443 		f->state = I40E_FILTER_REMOVE;
1444 	}
1445 
1446 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1447 	vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1448 }
1449 
1450 /**
1451  * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1452  * @vsi: the VSI to be searched
1453  * @macaddr: the MAC address
1454  * @vlan: the VLAN
1455  *
1456  * NOTE: This function is expected to be called with mac_filter_hash_lock
1457  * being held.
1458  * ANOTHER NOTE: This function MUST be called from within the context of
1459  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1460  * instead of list_for_each_entry().
1461  **/
1462 void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1463 {
1464 	struct i40e_mac_filter *f;
1465 
1466 	if (!vsi || !macaddr)
1467 		return;
1468 
1469 	f = i40e_find_filter(vsi, macaddr, vlan);
1470 	__i40e_del_filter(vsi, f);
1471 }
1472 
1473 /**
1474  * i40e_add_mac_filter - Add a MAC filter for all active VLANs
1475  * @vsi: the VSI to be searched
1476  * @macaddr: the mac address to be filtered
1477  *
1478  * If we're not in VLAN mode, just add the filter to I40E_VLAN_ANY. Otherwise,
1479  * go through all the macvlan filters and add a macvlan filter for each
1480  * unique vlan that already exists. If a PVID has been assigned, instead only
1481  * add the macaddr to that VLAN.
1482  *
1483  * Returns last filter added on success, else NULL
1484  **/
1485 struct i40e_mac_filter *i40e_add_mac_filter(struct i40e_vsi *vsi,
1486 					    const u8 *macaddr)
1487 {
1488 	struct i40e_mac_filter *f, *add = NULL;
1489 	struct hlist_node *h;
1490 	int bkt;
1491 
1492 	if (vsi->info.pvid)
1493 		return i40e_add_filter(vsi, macaddr,
1494 				       le16_to_cpu(vsi->info.pvid));
1495 
1496 	if (!i40e_is_vsi_in_vlan(vsi))
1497 		return i40e_add_filter(vsi, macaddr, I40E_VLAN_ANY);
1498 
1499 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1500 		if (f->state == I40E_FILTER_REMOVE)
1501 			continue;
1502 		add = i40e_add_filter(vsi, macaddr, f->vlan);
1503 		if (!add)
1504 			return NULL;
1505 	}
1506 
1507 	return add;
1508 }
1509 
1510 /**
1511  * i40e_del_mac_filter - Remove a MAC filter from all VLANs
1512  * @vsi: the VSI to be searched
1513  * @macaddr: the mac address to be removed
1514  *
1515  * Removes a given MAC address from a VSI regardless of what VLAN it has been
1516  * associated with.
1517  *
1518  * Returns 0 for success, or error
1519  **/
1520 int i40e_del_mac_filter(struct i40e_vsi *vsi, const u8 *macaddr)
1521 {
1522 	struct i40e_mac_filter *f;
1523 	struct hlist_node *h;
1524 	bool found = false;
1525 	int bkt;
1526 
1527 	WARN(!spin_is_locked(&vsi->mac_filter_hash_lock),
1528 	     "Missing mac_filter_hash_lock\n");
1529 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1530 		if (ether_addr_equal(macaddr, f->macaddr)) {
1531 			__i40e_del_filter(vsi, f);
1532 			found = true;
1533 		}
1534 	}
1535 
1536 	if (found)
1537 		return 0;
1538 	else
1539 		return -ENOENT;
1540 }
1541 
1542 /**
1543  * i40e_set_mac - NDO callback to set mac address
1544  * @netdev: network interface device structure
1545  * @p: pointer to an address structure
1546  *
1547  * Returns 0 on success, negative on failure
1548  **/
1549 static int i40e_set_mac(struct net_device *netdev, void *p)
1550 {
1551 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1552 	struct i40e_vsi *vsi = np->vsi;
1553 	struct i40e_pf *pf = vsi->back;
1554 	struct i40e_hw *hw = &pf->hw;
1555 	struct sockaddr *addr = p;
1556 
1557 	if (!is_valid_ether_addr(addr->sa_data))
1558 		return -EADDRNOTAVAIL;
1559 
1560 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1561 		netdev_info(netdev, "already using mac address %pM\n",
1562 			    addr->sa_data);
1563 		return 0;
1564 	}
1565 
1566 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
1567 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
1568 		return -EADDRNOTAVAIL;
1569 
1570 	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1571 		netdev_info(netdev, "returning to hw mac address %pM\n",
1572 			    hw->mac.addr);
1573 	else
1574 		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1575 
1576 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1577 	i40e_del_mac_filter(vsi, netdev->dev_addr);
1578 	i40e_add_mac_filter(vsi, addr->sa_data);
1579 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1580 	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1581 	if (vsi->type == I40E_VSI_MAIN) {
1582 		i40e_status ret;
1583 
1584 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1585 						I40E_AQC_WRITE_TYPE_LAA_WOL,
1586 						addr->sa_data, NULL);
1587 		if (ret)
1588 			netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1589 				    i40e_stat_str(hw, ret),
1590 				    i40e_aq_str(hw, hw->aq.asq_last_status));
1591 	}
1592 
1593 	/* schedule our worker thread which will take care of
1594 	 * applying the new filter changes
1595 	 */
1596 	i40e_service_event_schedule(vsi->back);
1597 	return 0;
1598 }
1599 
1600 /**
1601  * i40e_config_rss_aq - Prepare for RSS using AQ commands
1602  * @vsi: vsi structure
1603  * @seed: RSS hash seed
1604  **/
1605 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
1606 			      u8 *lut, u16 lut_size)
1607 {
1608 	struct i40e_pf *pf = vsi->back;
1609 	struct i40e_hw *hw = &pf->hw;
1610 	int ret = 0;
1611 
1612 	if (seed) {
1613 		struct i40e_aqc_get_set_rss_key_data *seed_dw =
1614 			(struct i40e_aqc_get_set_rss_key_data *)seed;
1615 		ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
1616 		if (ret) {
1617 			dev_info(&pf->pdev->dev,
1618 				 "Cannot set RSS key, err %s aq_err %s\n",
1619 				 i40e_stat_str(hw, ret),
1620 				 i40e_aq_str(hw, hw->aq.asq_last_status));
1621 			return ret;
1622 		}
1623 	}
1624 	if (lut) {
1625 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
1626 
1627 		ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
1628 		if (ret) {
1629 			dev_info(&pf->pdev->dev,
1630 				 "Cannot set RSS lut, err %s aq_err %s\n",
1631 				 i40e_stat_str(hw, ret),
1632 				 i40e_aq_str(hw, hw->aq.asq_last_status));
1633 			return ret;
1634 		}
1635 	}
1636 	return ret;
1637 }
1638 
1639 /**
1640  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
1641  * @vsi: VSI structure
1642  **/
1643 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
1644 {
1645 	struct i40e_pf *pf = vsi->back;
1646 	u8 seed[I40E_HKEY_ARRAY_SIZE];
1647 	u8 *lut;
1648 	int ret;
1649 
1650 	if (!(pf->hw_features & I40E_HW_RSS_AQ_CAPABLE))
1651 		return 0;
1652 	if (!vsi->rss_size)
1653 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
1654 				      vsi->num_queue_pairs);
1655 	if (!vsi->rss_size)
1656 		return -EINVAL;
1657 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1658 	if (!lut)
1659 		return -ENOMEM;
1660 
1661 	/* Use the user configured hash keys and lookup table if there is one,
1662 	 * otherwise use default
1663 	 */
1664 	if (vsi->rss_lut_user)
1665 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1666 	else
1667 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
1668 	if (vsi->rss_hkey_user)
1669 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
1670 	else
1671 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
1672 	ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
1673 	kfree(lut);
1674 	return ret;
1675 }
1676 
1677 /**
1678  * i40e_vsi_setup_queue_map_mqprio - Prepares mqprio based tc_config
1679  * @vsi: the VSI being configured,
1680  * @ctxt: VSI context structure
1681  * @enabled_tc: number of traffic classes to enable
1682  *
1683  * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
1684  **/
1685 static int i40e_vsi_setup_queue_map_mqprio(struct i40e_vsi *vsi,
1686 					   struct i40e_vsi_context *ctxt,
1687 					   u8 enabled_tc)
1688 {
1689 	u16 qcount = 0, max_qcount, qmap, sections = 0;
1690 	int i, override_q, pow, num_qps, ret;
1691 	u8 netdev_tc = 0, offset = 0;
1692 
1693 	if (vsi->type != I40E_VSI_MAIN)
1694 		return -EINVAL;
1695 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1696 	sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1697 	vsi->tc_config.numtc = vsi->mqprio_qopt.qopt.num_tc;
1698 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1699 	num_qps = vsi->mqprio_qopt.qopt.count[0];
1700 
1701 	/* find the next higher power-of-2 of num queue pairs */
1702 	pow = ilog2(num_qps);
1703 	if (!is_power_of_2(num_qps))
1704 		pow++;
1705 	qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1706 		(pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1707 
1708 	/* Setup queue offset/count for all TCs for given VSI */
1709 	max_qcount = vsi->mqprio_qopt.qopt.count[0];
1710 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1711 		/* See if the given TC is enabled for the given VSI */
1712 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1713 			offset = vsi->mqprio_qopt.qopt.offset[i];
1714 			qcount = vsi->mqprio_qopt.qopt.count[i];
1715 			if (qcount > max_qcount)
1716 				max_qcount = qcount;
1717 			vsi->tc_config.tc_info[i].qoffset = offset;
1718 			vsi->tc_config.tc_info[i].qcount = qcount;
1719 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1720 		} else {
1721 			/* TC is not enabled so set the offset to
1722 			 * default queue and allocate one queue
1723 			 * for the given TC.
1724 			 */
1725 			vsi->tc_config.tc_info[i].qoffset = 0;
1726 			vsi->tc_config.tc_info[i].qcount = 1;
1727 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1728 		}
1729 	}
1730 
1731 	/* Set actual Tx/Rx queue pairs */
1732 	vsi->num_queue_pairs = offset + qcount;
1733 
1734 	/* Setup queue TC[0].qmap for given VSI context */
1735 	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1736 	ctxt->info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1737 	ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1738 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1739 
1740 	/* Reconfigure RSS for main VSI with max queue count */
1741 	vsi->rss_size = max_qcount;
1742 	ret = i40e_vsi_config_rss(vsi);
1743 	if (ret) {
1744 		dev_info(&vsi->back->pdev->dev,
1745 			 "Failed to reconfig rss for num_queues (%u)\n",
1746 			 max_qcount);
1747 		return ret;
1748 	}
1749 	vsi->reconfig_rss = true;
1750 	dev_dbg(&vsi->back->pdev->dev,
1751 		"Reconfigured rss with num_queues (%u)\n", max_qcount);
1752 
1753 	/* Find queue count available for channel VSIs and starting offset
1754 	 * for channel VSIs
1755 	 */
1756 	override_q = vsi->mqprio_qopt.qopt.count[0];
1757 	if (override_q && override_q < vsi->num_queue_pairs) {
1758 		vsi->cnt_q_avail = vsi->num_queue_pairs - override_q;
1759 		vsi->next_base_queue = override_q;
1760 	}
1761 	return 0;
1762 }
1763 
1764 /**
1765  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1766  * @vsi: the VSI being setup
1767  * @ctxt: VSI context structure
1768  * @enabled_tc: Enabled TCs bitmap
1769  * @is_add: True if called before Add VSI
1770  *
1771  * Setup VSI queue mapping for enabled traffic classes.
1772  **/
1773 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1774 				     struct i40e_vsi_context *ctxt,
1775 				     u8 enabled_tc,
1776 				     bool is_add)
1777 {
1778 	struct i40e_pf *pf = vsi->back;
1779 	u16 sections = 0;
1780 	u8 netdev_tc = 0;
1781 	u16 numtc = 0;
1782 	u16 qcount;
1783 	u8 offset;
1784 	u16 qmap;
1785 	int i;
1786 	u16 num_tc_qps = 0;
1787 
1788 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1789 	offset = 0;
1790 
1791 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1792 		/* Find numtc from enabled TC bitmap */
1793 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1794 			if (enabled_tc & BIT(i)) /* TC is enabled */
1795 				numtc++;
1796 		}
1797 		if (!numtc) {
1798 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1799 			numtc = 1;
1800 		}
1801 	} else {
1802 		/* At least TC0 is enabled in non-DCB, non-MQPRIO case */
1803 		numtc = 1;
1804 	}
1805 
1806 	vsi->tc_config.numtc = numtc;
1807 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1808 	/* Number of queues per enabled TC */
1809 	qcount = vsi->alloc_queue_pairs;
1810 
1811 	num_tc_qps = qcount / numtc;
1812 	num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1813 
1814 	/* Setup queue offset/count for all TCs for given VSI */
1815 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1816 		/* See if the given TC is enabled for the given VSI */
1817 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1818 			/* TC is enabled */
1819 			int pow, num_qps;
1820 
1821 			switch (vsi->type) {
1822 			case I40E_VSI_MAIN:
1823 				qcount = min_t(int, pf->alloc_rss_size,
1824 					       num_tc_qps);
1825 				break;
1826 			case I40E_VSI_FDIR:
1827 			case I40E_VSI_SRIOV:
1828 			case I40E_VSI_VMDQ2:
1829 			default:
1830 				qcount = num_tc_qps;
1831 				WARN_ON(i != 0);
1832 				break;
1833 			}
1834 			vsi->tc_config.tc_info[i].qoffset = offset;
1835 			vsi->tc_config.tc_info[i].qcount = qcount;
1836 
1837 			/* find the next higher power-of-2 of num queue pairs */
1838 			num_qps = qcount;
1839 			pow = 0;
1840 			while (num_qps && (BIT_ULL(pow) < qcount)) {
1841 				pow++;
1842 				num_qps >>= 1;
1843 			}
1844 
1845 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1846 			qmap =
1847 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1848 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1849 
1850 			offset += qcount;
1851 		} else {
1852 			/* TC is not enabled so set the offset to
1853 			 * default queue and allocate one queue
1854 			 * for the given TC.
1855 			 */
1856 			vsi->tc_config.tc_info[i].qoffset = 0;
1857 			vsi->tc_config.tc_info[i].qcount = 1;
1858 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1859 
1860 			qmap = 0;
1861 		}
1862 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1863 	}
1864 
1865 	/* Set actual Tx/Rx queue pairs */
1866 	vsi->num_queue_pairs = offset;
1867 	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1868 		if (vsi->req_queue_pairs > 0)
1869 			vsi->num_queue_pairs = vsi->req_queue_pairs;
1870 		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1871 			vsi->num_queue_pairs = pf->num_lan_msix;
1872 	}
1873 
1874 	/* Scheduler section valid can only be set for ADD VSI */
1875 	if (is_add) {
1876 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1877 
1878 		ctxt->info.up_enable_bits = enabled_tc;
1879 	}
1880 	if (vsi->type == I40E_VSI_SRIOV) {
1881 		ctxt->info.mapping_flags |=
1882 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1883 		for (i = 0; i < vsi->num_queue_pairs; i++)
1884 			ctxt->info.queue_mapping[i] =
1885 					       cpu_to_le16(vsi->base_queue + i);
1886 	} else {
1887 		ctxt->info.mapping_flags |=
1888 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1889 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1890 	}
1891 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1892 }
1893 
1894 /**
1895  * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1896  * @netdev: the netdevice
1897  * @addr: address to add
1898  *
1899  * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1900  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1901  */
1902 static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1903 {
1904 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1905 	struct i40e_vsi *vsi = np->vsi;
1906 
1907 	if (i40e_add_mac_filter(vsi, addr))
1908 		return 0;
1909 	else
1910 		return -ENOMEM;
1911 }
1912 
1913 /**
1914  * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1915  * @netdev: the netdevice
1916  * @addr: address to add
1917  *
1918  * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1919  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1920  */
1921 static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1922 {
1923 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1924 	struct i40e_vsi *vsi = np->vsi;
1925 
1926 	i40e_del_mac_filter(vsi, addr);
1927 
1928 	return 0;
1929 }
1930 
1931 /**
1932  * i40e_set_rx_mode - NDO callback to set the netdev filters
1933  * @netdev: network interface device structure
1934  **/
1935 static void i40e_set_rx_mode(struct net_device *netdev)
1936 {
1937 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1938 	struct i40e_vsi *vsi = np->vsi;
1939 
1940 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1941 
1942 	__dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1943 	__dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1944 
1945 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1946 
1947 	/* check for other flag changes */
1948 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1949 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1950 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1951 	}
1952 }
1953 
1954 /**
1955  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1956  * @vsi: Pointer to VSI struct
1957  * @from: Pointer to list which contains MAC filter entries - changes to
1958  *        those entries needs to be undone.
1959  *
1960  * MAC filter entries from this list were slated for deletion.
1961  **/
1962 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1963 					 struct hlist_head *from)
1964 {
1965 	struct i40e_mac_filter *f;
1966 	struct hlist_node *h;
1967 
1968 	hlist_for_each_entry_safe(f, h, from, hlist) {
1969 		u64 key = i40e_addr_to_hkey(f->macaddr);
1970 
1971 		/* Move the element back into MAC filter list*/
1972 		hlist_del(&f->hlist);
1973 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1974 	}
1975 }
1976 
1977 /**
1978  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1979  * @vsi: Pointer to vsi struct
1980  * @from: Pointer to list which contains MAC filter entries - changes to
1981  *        those entries needs to be undone.
1982  *
1983  * MAC filter entries from this list were slated for addition.
1984  **/
1985 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi,
1986 					 struct hlist_head *from)
1987 {
1988 	struct i40e_new_mac_filter *new;
1989 	struct hlist_node *h;
1990 
1991 	hlist_for_each_entry_safe(new, h, from, hlist) {
1992 		/* We can simply free the wrapper structure */
1993 		hlist_del(&new->hlist);
1994 		kfree(new);
1995 	}
1996 }
1997 
1998 /**
1999  * i40e_next_entry - Get the next non-broadcast filter from a list
2000  * @next: pointer to filter in list
2001  *
2002  * Returns the next non-broadcast filter in the list. Required so that we
2003  * ignore broadcast filters within the list, since these are not handled via
2004  * the normal firmware update path.
2005  */
2006 static
2007 struct i40e_new_mac_filter *i40e_next_filter(struct i40e_new_mac_filter *next)
2008 {
2009 	hlist_for_each_entry_continue(next, hlist) {
2010 		if (!is_broadcast_ether_addr(next->f->macaddr))
2011 			return next;
2012 	}
2013 
2014 	return NULL;
2015 }
2016 
2017 /**
2018  * i40e_update_filter_state - Update filter state based on return data
2019  * from firmware
2020  * @count: Number of filters added
2021  * @add_list: return data from fw
2022  * @head: pointer to first filter in current batch
2023  *
2024  * MAC filter entries from list were slated to be added to device. Returns
2025  * number of successful filters. Note that 0 does NOT mean success!
2026  **/
2027 static int
2028 i40e_update_filter_state(int count,
2029 			 struct i40e_aqc_add_macvlan_element_data *add_list,
2030 			 struct i40e_new_mac_filter *add_head)
2031 {
2032 	int retval = 0;
2033 	int i;
2034 
2035 	for (i = 0; i < count; i++) {
2036 		/* Always check status of each filter. We don't need to check
2037 		 * the firmware return status because we pre-set the filter
2038 		 * status to I40E_AQC_MM_ERR_NO_RES when sending the filter
2039 		 * request to the adminq. Thus, if it no longer matches then
2040 		 * we know the filter is active.
2041 		 */
2042 		if (add_list[i].match_method == I40E_AQC_MM_ERR_NO_RES) {
2043 			add_head->state = I40E_FILTER_FAILED;
2044 		} else {
2045 			add_head->state = I40E_FILTER_ACTIVE;
2046 			retval++;
2047 		}
2048 
2049 		add_head = i40e_next_filter(add_head);
2050 		if (!add_head)
2051 			break;
2052 	}
2053 
2054 	return retval;
2055 }
2056 
2057 /**
2058  * i40e_aqc_del_filters - Request firmware to delete a set of filters
2059  * @vsi: ptr to the VSI
2060  * @vsi_name: name to display in messages
2061  * @list: the list of filters to send to firmware
2062  * @num_del: the number of filters to delete
2063  * @retval: Set to -EIO on failure to delete
2064  *
2065  * Send a request to firmware via AdminQ to delete a set of filters. Uses
2066  * *retval instead of a return value so that success does not force ret_val to
2067  * be set to 0. This ensures that a sequence of calls to this function
2068  * preserve the previous value of *retval on successful delete.
2069  */
2070 static
2071 void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
2072 			  struct i40e_aqc_remove_macvlan_element_data *list,
2073 			  int num_del, int *retval)
2074 {
2075 	struct i40e_hw *hw = &vsi->back->hw;
2076 	i40e_status aq_ret;
2077 	int aq_err;
2078 
2079 	aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
2080 	aq_err = hw->aq.asq_last_status;
2081 
2082 	/* Explicitly ignore and do not report when firmware returns ENOENT */
2083 	if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
2084 		*retval = -EIO;
2085 		dev_info(&vsi->back->pdev->dev,
2086 			 "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
2087 			 vsi_name, i40e_stat_str(hw, aq_ret),
2088 			 i40e_aq_str(hw, aq_err));
2089 	}
2090 }
2091 
2092 /**
2093  * i40e_aqc_add_filters - Request firmware to add a set of filters
2094  * @vsi: ptr to the VSI
2095  * @vsi_name: name to display in messages
2096  * @list: the list of filters to send to firmware
2097  * @add_head: Position in the add hlist
2098  * @num_add: the number of filters to add
2099  * @promisc_change: set to true on exit if promiscuous mode was forced on
2100  *
2101  * Send a request to firmware via AdminQ to add a chunk of filters. Will set
2102  * promisc_changed to true if the firmware has run out of space for more
2103  * filters.
2104  */
2105 static
2106 void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
2107 			  struct i40e_aqc_add_macvlan_element_data *list,
2108 			  struct i40e_new_mac_filter *add_head,
2109 			  int num_add, bool *promisc_changed)
2110 {
2111 	struct i40e_hw *hw = &vsi->back->hw;
2112 	int aq_err, fcnt;
2113 
2114 	i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
2115 	aq_err = hw->aq.asq_last_status;
2116 	fcnt = i40e_update_filter_state(num_add, list, add_head);
2117 
2118 	if (fcnt != num_add) {
2119 		*promisc_changed = true;
2120 		set_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2121 		dev_warn(&vsi->back->pdev->dev,
2122 			 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2123 			 i40e_aq_str(hw, aq_err),
2124 			 vsi_name);
2125 	}
2126 }
2127 
2128 /**
2129  * i40e_aqc_broadcast_filter - Set promiscuous broadcast flags
2130  * @vsi: pointer to the VSI
2131  * @f: filter data
2132  *
2133  * This function sets or clears the promiscuous broadcast flags for VLAN
2134  * filters in order to properly receive broadcast frames. Assumes that only
2135  * broadcast filters are passed.
2136  *
2137  * Returns status indicating success or failure;
2138  **/
2139 static i40e_status
2140 i40e_aqc_broadcast_filter(struct i40e_vsi *vsi, const char *vsi_name,
2141 			  struct i40e_mac_filter *f)
2142 {
2143 	bool enable = f->state == I40E_FILTER_NEW;
2144 	struct i40e_hw *hw = &vsi->back->hw;
2145 	i40e_status aq_ret;
2146 
2147 	if (f->vlan == I40E_VLAN_ANY) {
2148 		aq_ret = i40e_aq_set_vsi_broadcast(hw,
2149 						   vsi->seid,
2150 						   enable,
2151 						   NULL);
2152 	} else {
2153 		aq_ret = i40e_aq_set_vsi_bc_promisc_on_vlan(hw,
2154 							    vsi->seid,
2155 							    enable,
2156 							    f->vlan,
2157 							    NULL);
2158 	}
2159 
2160 	if (aq_ret)
2161 		dev_warn(&vsi->back->pdev->dev,
2162 			 "Error %s setting broadcast promiscuous mode on %s\n",
2163 			 i40e_aq_str(hw, hw->aq.asq_last_status),
2164 			 vsi_name);
2165 
2166 	return aq_ret;
2167 }
2168 
2169 /**
2170  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
2171  * @vsi: ptr to the VSI
2172  *
2173  * Push any outstanding VSI filter changes through the AdminQ.
2174  *
2175  * Returns 0 or error value
2176  **/
2177 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
2178 {
2179 	struct hlist_head tmp_add_list, tmp_del_list;
2180 	struct i40e_mac_filter *f;
2181 	struct i40e_new_mac_filter *new, *add_head = NULL;
2182 	struct i40e_hw *hw = &vsi->back->hw;
2183 	unsigned int failed_filters = 0;
2184 	unsigned int vlan_filters = 0;
2185 	bool promisc_changed = false;
2186 	char vsi_name[16] = "PF";
2187 	int filter_list_len = 0;
2188 	i40e_status aq_ret = 0;
2189 	u32 changed_flags = 0;
2190 	struct hlist_node *h;
2191 	struct i40e_pf *pf;
2192 	int num_add = 0;
2193 	int num_del = 0;
2194 	int retval = 0;
2195 	u16 cmd_flags;
2196 	int list_size;
2197 	int bkt;
2198 
2199 	/* empty array typed pointers, kcalloc later */
2200 	struct i40e_aqc_add_macvlan_element_data *add_list;
2201 	struct i40e_aqc_remove_macvlan_element_data *del_list;
2202 
2203 	while (test_and_set_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state))
2204 		usleep_range(1000, 2000);
2205 	pf = vsi->back;
2206 
2207 	if (vsi->netdev) {
2208 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
2209 		vsi->current_netdev_flags = vsi->netdev->flags;
2210 	}
2211 
2212 	INIT_HLIST_HEAD(&tmp_add_list);
2213 	INIT_HLIST_HEAD(&tmp_del_list);
2214 
2215 	if (vsi->type == I40E_VSI_SRIOV)
2216 		snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
2217 	else if (vsi->type != I40E_VSI_MAIN)
2218 		snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
2219 
2220 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
2221 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
2222 
2223 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2224 		/* Create a list of filters to delete. */
2225 		hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2226 			if (f->state == I40E_FILTER_REMOVE) {
2227 				/* Move the element into temporary del_list */
2228 				hash_del(&f->hlist);
2229 				hlist_add_head(&f->hlist, &tmp_del_list);
2230 
2231 				/* Avoid counting removed filters */
2232 				continue;
2233 			}
2234 			if (f->state == I40E_FILTER_NEW) {
2235 				/* Create a temporary i40e_new_mac_filter */
2236 				new = kzalloc(sizeof(*new), GFP_ATOMIC);
2237 				if (!new)
2238 					goto err_no_memory_locked;
2239 
2240 				/* Store pointer to the real filter */
2241 				new->f = f;
2242 				new->state = f->state;
2243 
2244 				/* Add it to the hash list */
2245 				hlist_add_head(&new->hlist, &tmp_add_list);
2246 			}
2247 
2248 			/* Count the number of active (current and new) VLAN
2249 			 * filters we have now. Does not count filters which
2250 			 * are marked for deletion.
2251 			 */
2252 			if (f->vlan > 0)
2253 				vlan_filters++;
2254 		}
2255 
2256 		retval = i40e_correct_mac_vlan_filters(vsi,
2257 						       &tmp_add_list,
2258 						       &tmp_del_list,
2259 						       vlan_filters);
2260 		if (retval)
2261 			goto err_no_memory_locked;
2262 
2263 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2264 	}
2265 
2266 	/* Now process 'del_list' outside the lock */
2267 	if (!hlist_empty(&tmp_del_list)) {
2268 		filter_list_len = hw->aq.asq_buf_size /
2269 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2270 		list_size = filter_list_len *
2271 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2272 		del_list = kzalloc(list_size, GFP_ATOMIC);
2273 		if (!del_list)
2274 			goto err_no_memory;
2275 
2276 		hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2277 			cmd_flags = 0;
2278 
2279 			/* handle broadcast filters by updating the broadcast
2280 			 * promiscuous flag and release filter list.
2281 			 */
2282 			if (is_broadcast_ether_addr(f->macaddr)) {
2283 				i40e_aqc_broadcast_filter(vsi, vsi_name, f);
2284 
2285 				hlist_del(&f->hlist);
2286 				kfree(f);
2287 				continue;
2288 			}
2289 
2290 			/* add to delete list */
2291 			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2292 			if (f->vlan == I40E_VLAN_ANY) {
2293 				del_list[num_del].vlan_tag = 0;
2294 				cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2295 			} else {
2296 				del_list[num_del].vlan_tag =
2297 					cpu_to_le16((u16)(f->vlan));
2298 			}
2299 
2300 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2301 			del_list[num_del].flags = cmd_flags;
2302 			num_del++;
2303 
2304 			/* flush a full buffer */
2305 			if (num_del == filter_list_len) {
2306 				i40e_aqc_del_filters(vsi, vsi_name, del_list,
2307 						     num_del, &retval);
2308 				memset(del_list, 0, list_size);
2309 				num_del = 0;
2310 			}
2311 			/* Release memory for MAC filter entries which were
2312 			 * synced up with HW.
2313 			 */
2314 			hlist_del(&f->hlist);
2315 			kfree(f);
2316 		}
2317 
2318 		if (num_del) {
2319 			i40e_aqc_del_filters(vsi, vsi_name, del_list,
2320 					     num_del, &retval);
2321 		}
2322 
2323 		kfree(del_list);
2324 		del_list = NULL;
2325 	}
2326 
2327 	if (!hlist_empty(&tmp_add_list)) {
2328 		/* Do all the adds now. */
2329 		filter_list_len = hw->aq.asq_buf_size /
2330 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2331 		list_size = filter_list_len *
2332 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2333 		add_list = kzalloc(list_size, GFP_ATOMIC);
2334 		if (!add_list)
2335 			goto err_no_memory;
2336 
2337 		num_add = 0;
2338 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2339 			if (test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2340 				     vsi->state)) {
2341 				new->state = I40E_FILTER_FAILED;
2342 				continue;
2343 			}
2344 
2345 			/* handle broadcast filters by updating the broadcast
2346 			 * promiscuous flag instead of adding a MAC filter.
2347 			 */
2348 			if (is_broadcast_ether_addr(new->f->macaddr)) {
2349 				if (i40e_aqc_broadcast_filter(vsi, vsi_name,
2350 							      new->f))
2351 					new->state = I40E_FILTER_FAILED;
2352 				else
2353 					new->state = I40E_FILTER_ACTIVE;
2354 				continue;
2355 			}
2356 
2357 			/* add to add array */
2358 			if (num_add == 0)
2359 				add_head = new;
2360 			cmd_flags = 0;
2361 			ether_addr_copy(add_list[num_add].mac_addr,
2362 					new->f->macaddr);
2363 			if (new->f->vlan == I40E_VLAN_ANY) {
2364 				add_list[num_add].vlan_tag = 0;
2365 				cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2366 			} else {
2367 				add_list[num_add].vlan_tag =
2368 					cpu_to_le16((u16)(new->f->vlan));
2369 			}
2370 			add_list[num_add].queue_number = 0;
2371 			/* set invalid match method for later detection */
2372 			add_list[num_add].match_method = I40E_AQC_MM_ERR_NO_RES;
2373 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2374 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
2375 			num_add++;
2376 
2377 			/* flush a full buffer */
2378 			if (num_add == filter_list_len) {
2379 				i40e_aqc_add_filters(vsi, vsi_name, add_list,
2380 						     add_head, num_add,
2381 						     &promisc_changed);
2382 				memset(add_list, 0, list_size);
2383 				num_add = 0;
2384 			}
2385 		}
2386 		if (num_add) {
2387 			i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2388 					     num_add, &promisc_changed);
2389 		}
2390 		/* Now move all of the filters from the temp add list back to
2391 		 * the VSI's list.
2392 		 */
2393 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2394 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2395 			/* Only update the state if we're still NEW */
2396 			if (new->f->state == I40E_FILTER_NEW)
2397 				new->f->state = new->state;
2398 			hlist_del(&new->hlist);
2399 			kfree(new);
2400 		}
2401 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2402 		kfree(add_list);
2403 		add_list = NULL;
2404 	}
2405 
2406 	/* Determine the number of active and failed filters. */
2407 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2408 	vsi->active_filters = 0;
2409 	hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2410 		if (f->state == I40E_FILTER_ACTIVE)
2411 			vsi->active_filters++;
2412 		else if (f->state == I40E_FILTER_FAILED)
2413 			failed_filters++;
2414 	}
2415 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2416 
2417 	/* If promiscuous mode has changed, we need to calculate a new
2418 	 * threshold for when we are safe to exit
2419 	 */
2420 	if (promisc_changed)
2421 		vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
2422 
2423 	/* Check if we are able to exit overflow promiscuous mode. We can
2424 	 * safely exit if we didn't just enter, we no longer have any failed
2425 	 * filters, and we have reduced filters below the threshold value.
2426 	 */
2427 	if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state) &&
2428 	    !promisc_changed && !failed_filters &&
2429 	    (vsi->active_filters < vsi->promisc_threshold)) {
2430 		dev_info(&pf->pdev->dev,
2431 			 "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2432 			 vsi_name);
2433 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2434 		promisc_changed = true;
2435 		vsi->promisc_threshold = 0;
2436 	}
2437 
2438 	/* if the VF is not trusted do not do promisc */
2439 	if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2440 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2441 		goto out;
2442 	}
2443 
2444 	/* check for changes in promiscuous modes */
2445 	if (changed_flags & IFF_ALLMULTI) {
2446 		bool cur_multipromisc;
2447 
2448 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2449 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2450 							       vsi->seid,
2451 							       cur_multipromisc,
2452 							       NULL);
2453 		if (aq_ret) {
2454 			retval = i40e_aq_rc_to_posix(aq_ret,
2455 						     hw->aq.asq_last_status);
2456 			dev_info(&pf->pdev->dev,
2457 				 "set multi promisc failed on %s, err %s aq_err %s\n",
2458 				 vsi_name,
2459 				 i40e_stat_str(hw, aq_ret),
2460 				 i40e_aq_str(hw, hw->aq.asq_last_status));
2461 		}
2462 	}
2463 
2464 	if ((changed_flags & IFF_PROMISC) || promisc_changed) {
2465 		bool cur_promisc;
2466 
2467 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2468 			       test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2469 					vsi->state));
2470 		if ((vsi->type == I40E_VSI_MAIN) &&
2471 		    (pf->lan_veb != I40E_NO_VEB) &&
2472 		    !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2473 			/* set defport ON for Main VSI instead of true promisc
2474 			 * this way we will get all unicast/multicast and VLAN
2475 			 * promisc behavior but will not get VF or VMDq traffic
2476 			 * replicated on the Main VSI.
2477 			 */
2478 			if (pf->cur_promisc != cur_promisc) {
2479 				pf->cur_promisc = cur_promisc;
2480 				if (cur_promisc)
2481 					aq_ret =
2482 					      i40e_aq_set_default_vsi(hw,
2483 								      vsi->seid,
2484 								      NULL);
2485 				else
2486 					aq_ret =
2487 					    i40e_aq_clear_default_vsi(hw,
2488 								      vsi->seid,
2489 								      NULL);
2490 				if (aq_ret) {
2491 					retval = i40e_aq_rc_to_posix(aq_ret,
2492 							hw->aq.asq_last_status);
2493 					dev_info(&pf->pdev->dev,
2494 						 "Set default VSI failed on %s, err %s, aq_err %s\n",
2495 						 vsi_name,
2496 						 i40e_stat_str(hw, aq_ret),
2497 						 i40e_aq_str(hw,
2498 						     hw->aq.asq_last_status));
2499 				}
2500 			}
2501 		} else {
2502 			aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2503 							  hw,
2504 							  vsi->seid,
2505 							  cur_promisc, NULL,
2506 							  true);
2507 			if (aq_ret) {
2508 				retval =
2509 				i40e_aq_rc_to_posix(aq_ret,
2510 						    hw->aq.asq_last_status);
2511 				dev_info(&pf->pdev->dev,
2512 					 "set unicast promisc failed on %s, err %s, aq_err %s\n",
2513 					 vsi_name,
2514 					 i40e_stat_str(hw, aq_ret),
2515 					 i40e_aq_str(hw,
2516 						     hw->aq.asq_last_status));
2517 			}
2518 			aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2519 							  hw,
2520 							  vsi->seid,
2521 							  cur_promisc, NULL);
2522 			if (aq_ret) {
2523 				retval =
2524 				i40e_aq_rc_to_posix(aq_ret,
2525 						    hw->aq.asq_last_status);
2526 				dev_info(&pf->pdev->dev,
2527 					 "set multicast promisc failed on %s, err %s, aq_err %s\n",
2528 					 vsi_name,
2529 					 i40e_stat_str(hw, aq_ret),
2530 					 i40e_aq_str(hw,
2531 						     hw->aq.asq_last_status));
2532 			}
2533 		}
2534 		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2535 						   vsi->seid,
2536 						   cur_promisc, NULL);
2537 		if (aq_ret) {
2538 			retval = i40e_aq_rc_to_posix(aq_ret,
2539 						     pf->hw.aq.asq_last_status);
2540 			dev_info(&pf->pdev->dev,
2541 				 "set brdcast promisc failed, err %s, aq_err %s\n",
2542 					 i40e_stat_str(hw, aq_ret),
2543 					 i40e_aq_str(hw,
2544 						     hw->aq.asq_last_status));
2545 		}
2546 	}
2547 out:
2548 	/* if something went wrong then set the changed flag so we try again */
2549 	if (retval)
2550 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2551 
2552 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2553 	return retval;
2554 
2555 err_no_memory:
2556 	/* Restore elements on the temporary add and delete lists */
2557 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2558 err_no_memory_locked:
2559 	i40e_undo_del_filter_entries(vsi, &tmp_del_list);
2560 	i40e_undo_add_filter_entries(vsi, &tmp_add_list);
2561 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2562 
2563 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2564 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2565 	return -ENOMEM;
2566 }
2567 
2568 /**
2569  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2570  * @pf: board private structure
2571  **/
2572 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2573 {
2574 	int v;
2575 
2576 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2577 		return;
2578 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2579 
2580 	for (v = 0; v < pf->num_alloc_vsi; v++) {
2581 		if (pf->vsi[v] &&
2582 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2583 			int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2584 
2585 			if (ret) {
2586 				/* come back and try again later */
2587 				pf->flags |= I40E_FLAG_FILTER_SYNC;
2588 				break;
2589 			}
2590 		}
2591 	}
2592 }
2593 
2594 /**
2595  * i40e_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2596  * @vsi: the vsi
2597  **/
2598 static int i40e_max_xdp_frame_size(struct i40e_vsi *vsi)
2599 {
2600 	if (PAGE_SIZE >= 8192 || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
2601 		return I40E_RXBUFFER_2048;
2602 	else
2603 		return I40E_RXBUFFER_3072;
2604 }
2605 
2606 /**
2607  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2608  * @netdev: network interface device structure
2609  * @new_mtu: new value for maximum frame size
2610  *
2611  * Returns 0 on success, negative on failure
2612  **/
2613 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2614 {
2615 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2616 	struct i40e_vsi *vsi = np->vsi;
2617 	struct i40e_pf *pf = vsi->back;
2618 
2619 	if (i40e_enabled_xdp_vsi(vsi)) {
2620 		int frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2621 
2622 		if (frame_size > i40e_max_xdp_frame_size(vsi))
2623 			return -EINVAL;
2624 	}
2625 
2626 	netdev_info(netdev, "changing MTU from %d to %d\n",
2627 		    netdev->mtu, new_mtu);
2628 	netdev->mtu = new_mtu;
2629 	if (netif_running(netdev))
2630 		i40e_vsi_reinit_locked(vsi);
2631 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
2632 		      I40E_FLAG_CLIENT_L2_CHANGE);
2633 	return 0;
2634 }
2635 
2636 /**
2637  * i40e_ioctl - Access the hwtstamp interface
2638  * @netdev: network interface device structure
2639  * @ifr: interface request data
2640  * @cmd: ioctl command
2641  **/
2642 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2643 {
2644 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2645 	struct i40e_pf *pf = np->vsi->back;
2646 
2647 	switch (cmd) {
2648 	case SIOCGHWTSTAMP:
2649 		return i40e_ptp_get_ts_config(pf, ifr);
2650 	case SIOCSHWTSTAMP:
2651 		return i40e_ptp_set_ts_config(pf, ifr);
2652 	default:
2653 		return -EOPNOTSUPP;
2654 	}
2655 }
2656 
2657 /**
2658  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2659  * @vsi: the vsi being adjusted
2660  **/
2661 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2662 {
2663 	struct i40e_vsi_context ctxt;
2664 	i40e_status ret;
2665 
2666 	if ((vsi->info.valid_sections &
2667 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2668 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2669 		return;  /* already enabled */
2670 
2671 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2672 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2673 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2674 
2675 	ctxt.seid = vsi->seid;
2676 	ctxt.info = vsi->info;
2677 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2678 	if (ret) {
2679 		dev_info(&vsi->back->pdev->dev,
2680 			 "update vlan stripping failed, err %s aq_err %s\n",
2681 			 i40e_stat_str(&vsi->back->hw, ret),
2682 			 i40e_aq_str(&vsi->back->hw,
2683 				     vsi->back->hw.aq.asq_last_status));
2684 	}
2685 }
2686 
2687 /**
2688  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2689  * @vsi: the vsi being adjusted
2690  **/
2691 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2692 {
2693 	struct i40e_vsi_context ctxt;
2694 	i40e_status ret;
2695 
2696 	if ((vsi->info.valid_sections &
2697 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2698 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2699 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2700 		return;  /* already disabled */
2701 
2702 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2703 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2704 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2705 
2706 	ctxt.seid = vsi->seid;
2707 	ctxt.info = vsi->info;
2708 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2709 	if (ret) {
2710 		dev_info(&vsi->back->pdev->dev,
2711 			 "update vlan stripping failed, err %s aq_err %s\n",
2712 			 i40e_stat_str(&vsi->back->hw, ret),
2713 			 i40e_aq_str(&vsi->back->hw,
2714 				     vsi->back->hw.aq.asq_last_status));
2715 	}
2716 }
2717 
2718 /**
2719  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2720  * @netdev: network interface to be adjusted
2721  * @features: netdev features to test if VLAN offload is enabled or not
2722  **/
2723 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2724 {
2725 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2726 	struct i40e_vsi *vsi = np->vsi;
2727 
2728 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2729 		i40e_vlan_stripping_enable(vsi);
2730 	else
2731 		i40e_vlan_stripping_disable(vsi);
2732 }
2733 
2734 /**
2735  * i40e_add_vlan_all_mac - Add a MAC/VLAN filter for each existing MAC address
2736  * @vsi: the vsi being configured
2737  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2738  *
2739  * This is a helper function for adding a new MAC/VLAN filter with the
2740  * specified VLAN for each existing MAC address already in the hash table.
2741  * This function does *not* perform any accounting to update filters based on
2742  * VLAN mode.
2743  *
2744  * NOTE: this function expects to be called while under the
2745  * mac_filter_hash_lock
2746  **/
2747 int i40e_add_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2748 {
2749 	struct i40e_mac_filter *f, *add_f;
2750 	struct hlist_node *h;
2751 	int bkt;
2752 
2753 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2754 		if (f->state == I40E_FILTER_REMOVE)
2755 			continue;
2756 		add_f = i40e_add_filter(vsi, f->macaddr, vid);
2757 		if (!add_f) {
2758 			dev_info(&vsi->back->pdev->dev,
2759 				 "Could not add vlan filter %d for %pM\n",
2760 				 vid, f->macaddr);
2761 			return -ENOMEM;
2762 		}
2763 	}
2764 
2765 	return 0;
2766 }
2767 
2768 /**
2769  * i40e_vsi_add_vlan - Add VSI membership for given VLAN
2770  * @vsi: the VSI being configured
2771  * @vid: VLAN id to be added
2772  **/
2773 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, u16 vid)
2774 {
2775 	int err;
2776 
2777 	if (vsi->info.pvid)
2778 		return -EINVAL;
2779 
2780 	/* The network stack will attempt to add VID=0, with the intention to
2781 	 * receive priority tagged packets with a VLAN of 0. Our HW receives
2782 	 * these packets by default when configured to receive untagged
2783 	 * packets, so we don't need to add a filter for this case.
2784 	 * Additionally, HW interprets adding a VID=0 filter as meaning to
2785 	 * receive *only* tagged traffic and stops receiving untagged traffic.
2786 	 * Thus, we do not want to actually add a filter for VID=0
2787 	 */
2788 	if (!vid)
2789 		return 0;
2790 
2791 	/* Locked once because all functions invoked below iterates list*/
2792 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2793 	err = i40e_add_vlan_all_mac(vsi, vid);
2794 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2795 	if (err)
2796 		return err;
2797 
2798 	/* schedule our worker thread which will take care of
2799 	 * applying the new filter changes
2800 	 */
2801 	i40e_service_event_schedule(vsi->back);
2802 	return 0;
2803 }
2804 
2805 /**
2806  * i40e_rm_vlan_all_mac - Remove MAC/VLAN pair for all MAC with the given VLAN
2807  * @vsi: the vsi being configured
2808  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2809  *
2810  * This function should be used to remove all VLAN filters which match the
2811  * given VID. It does not schedule the service event and does not take the
2812  * mac_filter_hash_lock so it may be combined with other operations under
2813  * a single invocation of the mac_filter_hash_lock.
2814  *
2815  * NOTE: this function expects to be called while under the
2816  * mac_filter_hash_lock
2817  */
2818 void i40e_rm_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2819 {
2820 	struct i40e_mac_filter *f;
2821 	struct hlist_node *h;
2822 	int bkt;
2823 
2824 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2825 		if (f->vlan == vid)
2826 			__i40e_del_filter(vsi, f);
2827 	}
2828 }
2829 
2830 /**
2831  * i40e_vsi_kill_vlan - Remove VSI membership for given VLAN
2832  * @vsi: the VSI being configured
2833  * @vid: VLAN id to be removed
2834  **/
2835 void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, u16 vid)
2836 {
2837 	if (!vid || vsi->info.pvid)
2838 		return;
2839 
2840 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2841 	i40e_rm_vlan_all_mac(vsi, vid);
2842 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2843 
2844 	/* schedule our worker thread which will take care of
2845 	 * applying the new filter changes
2846 	 */
2847 	i40e_service_event_schedule(vsi->back);
2848 }
2849 
2850 /**
2851  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2852  * @netdev: network interface to be adjusted
2853  * @vid: vlan id to be added
2854  *
2855  * net_device_ops implementation for adding vlan ids
2856  **/
2857 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2858 				__always_unused __be16 proto, u16 vid)
2859 {
2860 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2861 	struct i40e_vsi *vsi = np->vsi;
2862 	int ret = 0;
2863 
2864 	if (vid >= VLAN_N_VID)
2865 		return -EINVAL;
2866 
2867 	ret = i40e_vsi_add_vlan(vsi, vid);
2868 	if (!ret)
2869 		set_bit(vid, vsi->active_vlans);
2870 
2871 	return ret;
2872 }
2873 
2874 /**
2875  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2876  * @netdev: network interface to be adjusted
2877  * @vid: vlan id to be removed
2878  *
2879  * net_device_ops implementation for removing vlan ids
2880  **/
2881 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2882 				 __always_unused __be16 proto, u16 vid)
2883 {
2884 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2885 	struct i40e_vsi *vsi = np->vsi;
2886 
2887 	/* return code is ignored as there is nothing a user
2888 	 * can do about failure to remove and a log message was
2889 	 * already printed from the other function
2890 	 */
2891 	i40e_vsi_kill_vlan(vsi, vid);
2892 
2893 	clear_bit(vid, vsi->active_vlans);
2894 
2895 	return 0;
2896 }
2897 
2898 /**
2899  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2900  * @vsi: the vsi being brought back up
2901  **/
2902 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2903 {
2904 	u16 vid;
2905 
2906 	if (!vsi->netdev)
2907 		return;
2908 
2909 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2910 
2911 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2912 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2913 				     vid);
2914 }
2915 
2916 /**
2917  * i40e_vsi_add_pvid - Add pvid for the VSI
2918  * @vsi: the vsi being adjusted
2919  * @vid: the vlan id to set as a PVID
2920  **/
2921 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2922 {
2923 	struct i40e_vsi_context ctxt;
2924 	i40e_status ret;
2925 
2926 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2927 	vsi->info.pvid = cpu_to_le16(vid);
2928 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2929 				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2930 				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2931 
2932 	ctxt.seid = vsi->seid;
2933 	ctxt.info = vsi->info;
2934 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2935 	if (ret) {
2936 		dev_info(&vsi->back->pdev->dev,
2937 			 "add pvid failed, err %s aq_err %s\n",
2938 			 i40e_stat_str(&vsi->back->hw, ret),
2939 			 i40e_aq_str(&vsi->back->hw,
2940 				     vsi->back->hw.aq.asq_last_status));
2941 		return -ENOENT;
2942 	}
2943 
2944 	return 0;
2945 }
2946 
2947 /**
2948  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2949  * @vsi: the vsi being adjusted
2950  *
2951  * Just use the vlan_rx_register() service to put it back to normal
2952  **/
2953 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2954 {
2955 	i40e_vlan_stripping_disable(vsi);
2956 
2957 	vsi->info.pvid = 0;
2958 }
2959 
2960 /**
2961  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2962  * @vsi: ptr to the VSI
2963  *
2964  * If this function returns with an error, then it's possible one or
2965  * more of the rings is populated (while the rest are not).  It is the
2966  * callers duty to clean those orphaned rings.
2967  *
2968  * Return 0 on success, negative on failure
2969  **/
2970 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2971 {
2972 	int i, err = 0;
2973 
2974 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2975 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2976 
2977 	if (!i40e_enabled_xdp_vsi(vsi))
2978 		return err;
2979 
2980 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2981 		err = i40e_setup_tx_descriptors(vsi->xdp_rings[i]);
2982 
2983 	return err;
2984 }
2985 
2986 /**
2987  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2988  * @vsi: ptr to the VSI
2989  *
2990  * Free VSI's transmit software resources
2991  **/
2992 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2993 {
2994 	int i;
2995 
2996 	if (vsi->tx_rings) {
2997 		for (i = 0; i < vsi->num_queue_pairs; i++)
2998 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2999 				i40e_free_tx_resources(vsi->tx_rings[i]);
3000 	}
3001 
3002 	if (vsi->xdp_rings) {
3003 		for (i = 0; i < vsi->num_queue_pairs; i++)
3004 			if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
3005 				i40e_free_tx_resources(vsi->xdp_rings[i]);
3006 	}
3007 }
3008 
3009 /**
3010  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
3011  * @vsi: ptr to the VSI
3012  *
3013  * If this function returns with an error, then it's possible one or
3014  * more of the rings is populated (while the rest are not).  It is the
3015  * callers duty to clean those orphaned rings.
3016  *
3017  * Return 0 on success, negative on failure
3018  **/
3019 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
3020 {
3021 	int i, err = 0;
3022 
3023 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3024 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
3025 	return err;
3026 }
3027 
3028 /**
3029  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
3030  * @vsi: ptr to the VSI
3031  *
3032  * Free all receive software resources
3033  **/
3034 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
3035 {
3036 	int i;
3037 
3038 	if (!vsi->rx_rings)
3039 		return;
3040 
3041 	for (i = 0; i < vsi->num_queue_pairs; i++)
3042 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
3043 			i40e_free_rx_resources(vsi->rx_rings[i]);
3044 }
3045 
3046 /**
3047  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
3048  * @ring: The Tx ring to configure
3049  *
3050  * This enables/disables XPS for a given Tx descriptor ring
3051  * based on the TCs enabled for the VSI that ring belongs to.
3052  **/
3053 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
3054 {
3055 	int cpu;
3056 
3057 	if (!ring->q_vector || !ring->netdev || ring->ch)
3058 		return;
3059 
3060 	/* We only initialize XPS once, so as not to overwrite user settings */
3061 	if (test_and_set_bit(__I40E_TX_XPS_INIT_DONE, ring->state))
3062 		return;
3063 
3064 	cpu = cpumask_local_spread(ring->q_vector->v_idx, -1);
3065 	netif_set_xps_queue(ring->netdev, get_cpu_mask(cpu),
3066 			    ring->queue_index);
3067 }
3068 
3069 /**
3070  * i40e_configure_tx_ring - Configure a transmit ring context and rest
3071  * @ring: The Tx ring to configure
3072  *
3073  * Configure the Tx descriptor ring in the HMC context.
3074  **/
3075 static int i40e_configure_tx_ring(struct i40e_ring *ring)
3076 {
3077 	struct i40e_vsi *vsi = ring->vsi;
3078 	u16 pf_q = vsi->base_queue + ring->queue_index;
3079 	struct i40e_hw *hw = &vsi->back->hw;
3080 	struct i40e_hmc_obj_txq tx_ctx;
3081 	i40e_status err = 0;
3082 	u32 qtx_ctl = 0;
3083 
3084 	/* some ATR related tx ring init */
3085 	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
3086 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
3087 		ring->atr_count = 0;
3088 	} else {
3089 		ring->atr_sample_rate = 0;
3090 	}
3091 
3092 	/* configure XPS */
3093 	i40e_config_xps_tx_ring(ring);
3094 
3095 	/* clear the context structure first */
3096 	memset(&tx_ctx, 0, sizeof(tx_ctx));
3097 
3098 	tx_ctx.new_context = 1;
3099 	tx_ctx.base = (ring->dma / 128);
3100 	tx_ctx.qlen = ring->count;
3101 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
3102 					       I40E_FLAG_FD_ATR_ENABLED));
3103 	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
3104 	/* FDIR VSI tx ring can still use RS bit and writebacks */
3105 	if (vsi->type != I40E_VSI_FDIR)
3106 		tx_ctx.head_wb_ena = 1;
3107 	tx_ctx.head_wb_addr = ring->dma +
3108 			      (ring->count * sizeof(struct i40e_tx_desc));
3109 
3110 	/* As part of VSI creation/update, FW allocates certain
3111 	 * Tx arbitration queue sets for each TC enabled for
3112 	 * the VSI. The FW returns the handles to these queue
3113 	 * sets as part of the response buffer to Add VSI,
3114 	 * Update VSI, etc. AQ commands. It is expected that
3115 	 * these queue set handles be associated with the Tx
3116 	 * queues by the driver as part of the TX queue context
3117 	 * initialization. This has to be done regardless of
3118 	 * DCB as by default everything is mapped to TC0.
3119 	 */
3120 
3121 	if (ring->ch)
3122 		tx_ctx.rdylist =
3123 			le16_to_cpu(ring->ch->info.qs_handle[ring->dcb_tc]);
3124 
3125 	else
3126 		tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
3127 
3128 	tx_ctx.rdylist_act = 0;
3129 
3130 	/* clear the context in the HMC */
3131 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
3132 	if (err) {
3133 		dev_info(&vsi->back->pdev->dev,
3134 			 "Failed to clear 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 	/* set the context in the HMC */
3140 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
3141 	if (err) {
3142 		dev_info(&vsi->back->pdev->dev,
3143 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
3144 			 ring->queue_index, pf_q, err);
3145 		return -ENOMEM;
3146 	}
3147 
3148 	/* Now associate this queue with this PCI function */
3149 	if (ring->ch) {
3150 		if (ring->ch->type == I40E_VSI_VMDQ2)
3151 			qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
3152 		else
3153 			return -EINVAL;
3154 
3155 		qtx_ctl |= (ring->ch->vsi_number <<
3156 			    I40E_QTX_CTL_VFVM_INDX_SHIFT) &
3157 			    I40E_QTX_CTL_VFVM_INDX_MASK;
3158 	} else {
3159 		if (vsi->type == I40E_VSI_VMDQ2) {
3160 			qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
3161 			qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
3162 				    I40E_QTX_CTL_VFVM_INDX_MASK;
3163 		} else {
3164 			qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
3165 		}
3166 	}
3167 
3168 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
3169 		    I40E_QTX_CTL_PF_INDX_MASK);
3170 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
3171 	i40e_flush(hw);
3172 
3173 	/* cache tail off for easier writes later */
3174 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
3175 
3176 	return 0;
3177 }
3178 
3179 /**
3180  * i40e_configure_rx_ring - Configure a receive ring context
3181  * @ring: The Rx ring to configure
3182  *
3183  * Configure the Rx descriptor ring in the HMC context.
3184  **/
3185 static int i40e_configure_rx_ring(struct i40e_ring *ring)
3186 {
3187 	struct i40e_vsi *vsi = ring->vsi;
3188 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
3189 	u16 pf_q = vsi->base_queue + ring->queue_index;
3190 	struct i40e_hw *hw = &vsi->back->hw;
3191 	struct i40e_hmc_obj_rxq rx_ctx;
3192 	i40e_status err = 0;
3193 
3194 	bitmap_zero(ring->state, __I40E_RING_STATE_NBITS);
3195 
3196 	/* clear the context structure first */
3197 	memset(&rx_ctx, 0, sizeof(rx_ctx));
3198 
3199 	ring->rx_buf_len = vsi->rx_buf_len;
3200 
3201 	rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3202 				    BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3203 
3204 	rx_ctx.base = (ring->dma / 128);
3205 	rx_ctx.qlen = ring->count;
3206 
3207 	/* use 32 byte descriptors */
3208 	rx_ctx.dsize = 1;
3209 
3210 	/* descriptor type is always zero
3211 	 * rx_ctx.dtype = 0;
3212 	 */
3213 	rx_ctx.hsplit_0 = 0;
3214 
3215 	rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3216 	if (hw->revision_id == 0)
3217 		rx_ctx.lrxqthresh = 0;
3218 	else
3219 		rx_ctx.lrxqthresh = 1;
3220 	rx_ctx.crcstrip = 1;
3221 	rx_ctx.l2tsel = 1;
3222 	/* this controls whether VLAN is stripped from inner headers */
3223 	rx_ctx.showiv = 0;
3224 	/* set the prefena field to 1 because the manual says to */
3225 	rx_ctx.prefena = 1;
3226 
3227 	/* clear the context in the HMC */
3228 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3229 	if (err) {
3230 		dev_info(&vsi->back->pdev->dev,
3231 			 "Failed to clear 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 	/* set the context in the HMC */
3237 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3238 	if (err) {
3239 		dev_info(&vsi->back->pdev->dev,
3240 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3241 			 ring->queue_index, pf_q, err);
3242 		return -ENOMEM;
3243 	}
3244 
3245 	/* configure Rx buffer alignment */
3246 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3247 		clear_ring_build_skb_enabled(ring);
3248 	else
3249 		set_ring_build_skb_enabled(ring);
3250 
3251 	/* cache tail for quicker writes, and clear the reg before use */
3252 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3253 	writel(0, ring->tail);
3254 
3255 	i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3256 
3257 	return 0;
3258 }
3259 
3260 /**
3261  * i40e_vsi_configure_tx - Configure the VSI for Tx
3262  * @vsi: VSI structure describing this set of rings and resources
3263  *
3264  * Configure the Tx VSI for operation.
3265  **/
3266 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3267 {
3268 	int err = 0;
3269 	u16 i;
3270 
3271 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3272 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3273 
3274 	if (!i40e_enabled_xdp_vsi(vsi))
3275 		return err;
3276 
3277 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3278 		err = i40e_configure_tx_ring(vsi->xdp_rings[i]);
3279 
3280 	return err;
3281 }
3282 
3283 /**
3284  * i40e_vsi_configure_rx - Configure the VSI for Rx
3285  * @vsi: the VSI being configured
3286  *
3287  * Configure the Rx VSI for operation.
3288  **/
3289 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3290 {
3291 	int err = 0;
3292 	u16 i;
3293 
3294 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3295 		vsi->max_frame = I40E_MAX_RXBUFFER;
3296 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
3297 #if (PAGE_SIZE < 8192)
3298 	} else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3299 		   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3300 		vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3301 		vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3302 #endif
3303 	} else {
3304 		vsi->max_frame = I40E_MAX_RXBUFFER;
3305 		vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3306 						       I40E_RXBUFFER_2048;
3307 	}
3308 
3309 	/* set up individual rings */
3310 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3311 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3312 
3313 	return err;
3314 }
3315 
3316 /**
3317  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3318  * @vsi: ptr to the VSI
3319  **/
3320 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3321 {
3322 	struct i40e_ring *tx_ring, *rx_ring;
3323 	u16 qoffset, qcount;
3324 	int i, n;
3325 
3326 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3327 		/* Reset the TC information */
3328 		for (i = 0; i < vsi->num_queue_pairs; i++) {
3329 			rx_ring = vsi->rx_rings[i];
3330 			tx_ring = vsi->tx_rings[i];
3331 			rx_ring->dcb_tc = 0;
3332 			tx_ring->dcb_tc = 0;
3333 		}
3334 		return;
3335 	}
3336 
3337 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3338 		if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3339 			continue;
3340 
3341 		qoffset = vsi->tc_config.tc_info[n].qoffset;
3342 		qcount = vsi->tc_config.tc_info[n].qcount;
3343 		for (i = qoffset; i < (qoffset + qcount); i++) {
3344 			rx_ring = vsi->rx_rings[i];
3345 			tx_ring = vsi->tx_rings[i];
3346 			rx_ring->dcb_tc = n;
3347 			tx_ring->dcb_tc = n;
3348 		}
3349 	}
3350 }
3351 
3352 /**
3353  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3354  * @vsi: ptr to the VSI
3355  **/
3356 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3357 {
3358 	if (vsi->netdev)
3359 		i40e_set_rx_mode(vsi->netdev);
3360 }
3361 
3362 /**
3363  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3364  * @vsi: Pointer to the targeted VSI
3365  *
3366  * This function replays the hlist on the hw where all the SB Flow Director
3367  * filters were saved.
3368  **/
3369 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3370 {
3371 	struct i40e_fdir_filter *filter;
3372 	struct i40e_pf *pf = vsi->back;
3373 	struct hlist_node *node;
3374 
3375 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3376 		return;
3377 
3378 	/* Reset FDir counters as we're replaying all existing filters */
3379 	pf->fd_tcp4_filter_cnt = 0;
3380 	pf->fd_udp4_filter_cnt = 0;
3381 	pf->fd_sctp4_filter_cnt = 0;
3382 	pf->fd_ip4_filter_cnt = 0;
3383 
3384 	hlist_for_each_entry_safe(filter, node,
3385 				  &pf->fdir_filter_list, fdir_node) {
3386 		i40e_add_del_fdir(vsi, filter, true);
3387 	}
3388 }
3389 
3390 /**
3391  * i40e_vsi_configure - Set up the VSI for action
3392  * @vsi: the VSI being configured
3393  **/
3394 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3395 {
3396 	int err;
3397 
3398 	i40e_set_vsi_rx_mode(vsi);
3399 	i40e_restore_vlan(vsi);
3400 	i40e_vsi_config_dcb_rings(vsi);
3401 	err = i40e_vsi_configure_tx(vsi);
3402 	if (!err)
3403 		err = i40e_vsi_configure_rx(vsi);
3404 
3405 	return err;
3406 }
3407 
3408 /**
3409  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3410  * @vsi: the VSI being configured
3411  **/
3412 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3413 {
3414 	bool has_xdp = i40e_enabled_xdp_vsi(vsi);
3415 	struct i40e_pf *pf = vsi->back;
3416 	struct i40e_hw *hw = &pf->hw;
3417 	u16 vector;
3418 	int i, q;
3419 	u32 qp;
3420 
3421 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
3422 	 * and PFINT_LNKLSTn registers, e.g.:
3423 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3424 	 */
3425 	qp = vsi->base_queue;
3426 	vector = vsi->base_vector;
3427 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3428 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3429 
3430 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
3431 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3432 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
3433 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3434 		     q_vector->rx.itr);
3435 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3436 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
3437 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3438 		     q_vector->tx.itr);
3439 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
3440 		     i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3441 
3442 		/* Linked list for the queuepairs assigned to this vector */
3443 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3444 		for (q = 0; q < q_vector->num_ringpairs; q++) {
3445 			u32 nextqp = has_xdp ? qp + vsi->alloc_queue_pairs : qp;
3446 			u32 val;
3447 
3448 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3449 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3450 			      (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3451 			      (nextqp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT) |
3452 			      (I40E_QUEUE_TYPE_TX <<
3453 			       I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3454 
3455 			wr32(hw, I40E_QINT_RQCTL(qp), val);
3456 
3457 			if (has_xdp) {
3458 				val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3459 				      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3460 				      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3461 				      (qp << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3462 				      (I40E_QUEUE_TYPE_TX <<
3463 				       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3464 
3465 				wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3466 			}
3467 
3468 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3469 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3470 			      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3471 			      ((qp + 1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3472 			      (I40E_QUEUE_TYPE_RX <<
3473 			       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3474 
3475 			/* Terminate the linked list */
3476 			if (q == (q_vector->num_ringpairs - 1))
3477 				val |= (I40E_QUEUE_END_OF_LIST <<
3478 					I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3479 
3480 			wr32(hw, I40E_QINT_TQCTL(qp), val);
3481 			qp++;
3482 		}
3483 	}
3484 
3485 	i40e_flush(hw);
3486 }
3487 
3488 /**
3489  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3490  * @hw: ptr to the hardware info
3491  **/
3492 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3493 {
3494 	struct i40e_hw *hw = &pf->hw;
3495 	u32 val;
3496 
3497 	/* clear things first */
3498 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3499 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3500 
3501 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3502 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3503 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
3504 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3505 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3506 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3507 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3508 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3509 
3510 	if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3511 		val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3512 
3513 	if (pf->flags & I40E_FLAG_PTP)
3514 		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3515 
3516 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
3517 
3518 	/* SW_ITR_IDX = 0, but don't change INTENA */
3519 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3520 					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3521 
3522 	/* OTHER_ITR_IDX = 0 */
3523 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3524 }
3525 
3526 /**
3527  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3528  * @vsi: the VSI being configured
3529  **/
3530 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3531 {
3532 	u32 nextqp = i40e_enabled_xdp_vsi(vsi) ? vsi->alloc_queue_pairs : 0;
3533 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3534 	struct i40e_pf *pf = vsi->back;
3535 	struct i40e_hw *hw = &pf->hw;
3536 	u32 val;
3537 
3538 	/* set the ITR configuration */
3539 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
3540 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3541 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
3542 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3543 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3544 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
3545 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3546 
3547 	i40e_enable_misc_int_causes(pf);
3548 
3549 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3550 	wr32(hw, I40E_PFINT_LNKLST0, 0);
3551 
3552 	/* Associate the queue pair to the vector and enable the queue int */
3553 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		       |
3554 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3555 	      (nextqp	   << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3556 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3557 
3558 	wr32(hw, I40E_QINT_RQCTL(0), val);
3559 
3560 	if (i40e_enabled_xdp_vsi(vsi)) {
3561 		val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		     |
3562 		      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)|
3563 		      (I40E_QUEUE_TYPE_TX
3564 		       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3565 
3566 	       wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3567 	}
3568 
3569 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
3570 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3571 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3572 
3573 	wr32(hw, I40E_QINT_TQCTL(0), val);
3574 	i40e_flush(hw);
3575 }
3576 
3577 /**
3578  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3579  * @pf: board private structure
3580  **/
3581 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3582 {
3583 	struct i40e_hw *hw = &pf->hw;
3584 
3585 	wr32(hw, I40E_PFINT_DYN_CTL0,
3586 	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3587 	i40e_flush(hw);
3588 }
3589 
3590 /**
3591  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3592  * @pf: board private structure
3593  **/
3594 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3595 {
3596 	struct i40e_hw *hw = &pf->hw;
3597 	u32 val;
3598 
3599 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3600 	      I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3601 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3602 
3603 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
3604 	i40e_flush(hw);
3605 }
3606 
3607 /**
3608  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3609  * @irq: interrupt number
3610  * @data: pointer to a q_vector
3611  **/
3612 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3613 {
3614 	struct i40e_q_vector *q_vector = data;
3615 
3616 	if (!q_vector->tx.ring && !q_vector->rx.ring)
3617 		return IRQ_HANDLED;
3618 
3619 	napi_schedule_irqoff(&q_vector->napi);
3620 
3621 	return IRQ_HANDLED;
3622 }
3623 
3624 /**
3625  * i40e_irq_affinity_notify - Callback for affinity changes
3626  * @notify: context as to what irq was changed
3627  * @mask: the new affinity mask
3628  *
3629  * This is a callback function used by the irq_set_affinity_notifier function
3630  * so that we may register to receive changes to the irq affinity masks.
3631  **/
3632 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3633 				     const cpumask_t *mask)
3634 {
3635 	struct i40e_q_vector *q_vector =
3636 		container_of(notify, struct i40e_q_vector, affinity_notify);
3637 
3638 	cpumask_copy(&q_vector->affinity_mask, mask);
3639 }
3640 
3641 /**
3642  * i40e_irq_affinity_release - Callback for affinity notifier release
3643  * @ref: internal core kernel usage
3644  *
3645  * This is a callback function used by the irq_set_affinity_notifier function
3646  * to inform the current notification subscriber that they will no longer
3647  * receive notifications.
3648  **/
3649 static void i40e_irq_affinity_release(struct kref *ref) {}
3650 
3651 /**
3652  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3653  * @vsi: the VSI being configured
3654  * @basename: name for the vector
3655  *
3656  * Allocates MSI-X vectors and requests interrupts from the kernel.
3657  **/
3658 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3659 {
3660 	int q_vectors = vsi->num_q_vectors;
3661 	struct i40e_pf *pf = vsi->back;
3662 	int base = vsi->base_vector;
3663 	int rx_int_idx = 0;
3664 	int tx_int_idx = 0;
3665 	int vector, err;
3666 	int irq_num;
3667 	int cpu;
3668 
3669 	for (vector = 0; vector < q_vectors; vector++) {
3670 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3671 
3672 		irq_num = pf->msix_entries[base + vector].vector;
3673 
3674 		if (q_vector->tx.ring && q_vector->rx.ring) {
3675 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3676 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3677 			tx_int_idx++;
3678 		} else if (q_vector->rx.ring) {
3679 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3680 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3681 		} else if (q_vector->tx.ring) {
3682 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3683 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3684 		} else {
3685 			/* skip this unused q_vector */
3686 			continue;
3687 		}
3688 		err = request_irq(irq_num,
3689 				  vsi->irq_handler,
3690 				  0,
3691 				  q_vector->name,
3692 				  q_vector);
3693 		if (err) {
3694 			dev_info(&pf->pdev->dev,
3695 				 "MSIX request_irq failed, error: %d\n", err);
3696 			goto free_queue_irqs;
3697 		}
3698 
3699 		/* register for affinity change notifications */
3700 		q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3701 		q_vector->affinity_notify.release = i40e_irq_affinity_release;
3702 		irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3703 		/* Spread affinity hints out across online CPUs.
3704 		 *
3705 		 * get_cpu_mask returns a static constant mask with
3706 		 * a permanent lifetime so it's ok to pass to
3707 		 * irq_set_affinity_hint without making a copy.
3708 		 */
3709 		cpu = cpumask_local_spread(q_vector->v_idx, -1);
3710 		irq_set_affinity_hint(irq_num, get_cpu_mask(cpu));
3711 	}
3712 
3713 	vsi->irqs_ready = true;
3714 	return 0;
3715 
3716 free_queue_irqs:
3717 	while (vector) {
3718 		vector--;
3719 		irq_num = pf->msix_entries[base + vector].vector;
3720 		irq_set_affinity_notifier(irq_num, NULL);
3721 		irq_set_affinity_hint(irq_num, NULL);
3722 		free_irq(irq_num, &vsi->q_vectors[vector]);
3723 	}
3724 	return err;
3725 }
3726 
3727 /**
3728  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3729  * @vsi: the VSI being un-configured
3730  **/
3731 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3732 {
3733 	struct i40e_pf *pf = vsi->back;
3734 	struct i40e_hw *hw = &pf->hw;
3735 	int base = vsi->base_vector;
3736 	int i;
3737 
3738 	/* disable interrupt causation from each queue */
3739 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3740 		u32 val;
3741 
3742 		val = rd32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx));
3743 		val &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3744 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), val);
3745 
3746 		val = rd32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx));
3747 		val &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3748 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), val);
3749 
3750 		if (!i40e_enabled_xdp_vsi(vsi))
3751 			continue;
3752 		wr32(hw, I40E_QINT_TQCTL(vsi->xdp_rings[i]->reg_idx), 0);
3753 	}
3754 
3755 	/* disable each interrupt */
3756 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3757 		for (i = vsi->base_vector;
3758 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3759 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3760 
3761 		i40e_flush(hw);
3762 		for (i = 0; i < vsi->num_q_vectors; i++)
3763 			synchronize_irq(pf->msix_entries[i + base].vector);
3764 	} else {
3765 		/* Legacy and MSI mode - this stops all interrupt handling */
3766 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3767 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3768 		i40e_flush(hw);
3769 		synchronize_irq(pf->pdev->irq);
3770 	}
3771 }
3772 
3773 /**
3774  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3775  * @vsi: the VSI being configured
3776  **/
3777 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3778 {
3779 	struct i40e_pf *pf = vsi->back;
3780 	int i;
3781 
3782 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3783 		for (i = 0; i < vsi->num_q_vectors; i++)
3784 			i40e_irq_dynamic_enable(vsi, i);
3785 	} else {
3786 		i40e_irq_dynamic_enable_icr0(pf);
3787 	}
3788 
3789 	i40e_flush(&pf->hw);
3790 	return 0;
3791 }
3792 
3793 /**
3794  * i40e_free_misc_vector - Free the vector that handles non-queue events
3795  * @pf: board private structure
3796  **/
3797 static void i40e_free_misc_vector(struct i40e_pf *pf)
3798 {
3799 	/* Disable ICR 0 */
3800 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3801 	i40e_flush(&pf->hw);
3802 
3803 	if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
3804 		synchronize_irq(pf->msix_entries[0].vector);
3805 		free_irq(pf->msix_entries[0].vector, pf);
3806 		clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
3807 	}
3808 }
3809 
3810 /**
3811  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3812  * @irq: interrupt number
3813  * @data: pointer to a q_vector
3814  *
3815  * This is the handler used for all MSI/Legacy interrupts, and deals
3816  * with both queue and non-queue interrupts.  This is also used in
3817  * MSIX mode to handle the non-queue interrupts.
3818  **/
3819 static irqreturn_t i40e_intr(int irq, void *data)
3820 {
3821 	struct i40e_pf *pf = (struct i40e_pf *)data;
3822 	struct i40e_hw *hw = &pf->hw;
3823 	irqreturn_t ret = IRQ_NONE;
3824 	u32 icr0, icr0_remaining;
3825 	u32 val, ena_mask;
3826 
3827 	icr0 = rd32(hw, I40E_PFINT_ICR0);
3828 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3829 
3830 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3831 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3832 		goto enable_intr;
3833 
3834 	/* if interrupt but no bits showing, must be SWINT */
3835 	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3836 	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3837 		pf->sw_int_count++;
3838 
3839 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3840 	    (icr0 & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3841 		ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3842 		dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3843 		set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
3844 	}
3845 
3846 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3847 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3848 		struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3849 		struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3850 
3851 		/* We do not have a way to disarm Queue causes while leaving
3852 		 * interrupt enabled for all other causes, ideally
3853 		 * interrupt should be disabled while we are in NAPI but
3854 		 * this is not a performance path and napi_schedule()
3855 		 * can deal with rescheduling.
3856 		 */
3857 		if (!test_bit(__I40E_DOWN, pf->state))
3858 			napi_schedule_irqoff(&q_vector->napi);
3859 	}
3860 
3861 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3862 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3863 		set_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
3864 		i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3865 	}
3866 
3867 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3868 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3869 		set_bit(__I40E_MDD_EVENT_PENDING, pf->state);
3870 	}
3871 
3872 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3873 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3874 		set_bit(__I40E_VFLR_EVENT_PENDING, pf->state);
3875 	}
3876 
3877 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3878 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
3879 			set_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
3880 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3881 		val = rd32(hw, I40E_GLGEN_RSTAT);
3882 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3883 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3884 		if (val == I40E_RESET_CORER) {
3885 			pf->corer_count++;
3886 		} else if (val == I40E_RESET_GLOBR) {
3887 			pf->globr_count++;
3888 		} else if (val == I40E_RESET_EMPR) {
3889 			pf->empr_count++;
3890 			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state);
3891 		}
3892 	}
3893 
3894 	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3895 		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3896 		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3897 		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3898 			 rd32(hw, I40E_PFHMC_ERRORINFO),
3899 			 rd32(hw, I40E_PFHMC_ERRORDATA));
3900 	}
3901 
3902 	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3903 		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3904 
3905 		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3906 			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3907 			i40e_ptp_tx_hwtstamp(pf);
3908 		}
3909 	}
3910 
3911 	/* If a critical error is pending we have no choice but to reset the
3912 	 * device.
3913 	 * Report and mask out any remaining unexpected interrupts.
3914 	 */
3915 	icr0_remaining = icr0 & ena_mask;
3916 	if (icr0_remaining) {
3917 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3918 			 icr0_remaining);
3919 		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3920 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3921 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3922 			dev_info(&pf->pdev->dev, "device will be reset\n");
3923 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
3924 			i40e_service_event_schedule(pf);
3925 		}
3926 		ena_mask &= ~icr0_remaining;
3927 	}
3928 	ret = IRQ_HANDLED;
3929 
3930 enable_intr:
3931 	/* re-enable interrupt causes */
3932 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3933 	if (!test_bit(__I40E_DOWN, pf->state)) {
3934 		i40e_service_event_schedule(pf);
3935 		i40e_irq_dynamic_enable_icr0(pf);
3936 	}
3937 
3938 	return ret;
3939 }
3940 
3941 /**
3942  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3943  * @tx_ring:  tx ring to clean
3944  * @budget:   how many cleans we're allowed
3945  *
3946  * Returns true if there's any budget left (e.g. the clean is finished)
3947  **/
3948 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3949 {
3950 	struct i40e_vsi *vsi = tx_ring->vsi;
3951 	u16 i = tx_ring->next_to_clean;
3952 	struct i40e_tx_buffer *tx_buf;
3953 	struct i40e_tx_desc *tx_desc;
3954 
3955 	tx_buf = &tx_ring->tx_bi[i];
3956 	tx_desc = I40E_TX_DESC(tx_ring, i);
3957 	i -= tx_ring->count;
3958 
3959 	do {
3960 		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3961 
3962 		/* if next_to_watch is not set then there is no work pending */
3963 		if (!eop_desc)
3964 			break;
3965 
3966 		/* prevent any other reads prior to eop_desc */
3967 		read_barrier_depends();
3968 
3969 		/* if the descriptor isn't done, no work yet to do */
3970 		if (!(eop_desc->cmd_type_offset_bsz &
3971 		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3972 			break;
3973 
3974 		/* clear next_to_watch to prevent false hangs */
3975 		tx_buf->next_to_watch = NULL;
3976 
3977 		tx_desc->buffer_addr = 0;
3978 		tx_desc->cmd_type_offset_bsz = 0;
3979 		/* move past filter desc */
3980 		tx_buf++;
3981 		tx_desc++;
3982 		i++;
3983 		if (unlikely(!i)) {
3984 			i -= tx_ring->count;
3985 			tx_buf = tx_ring->tx_bi;
3986 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3987 		}
3988 		/* unmap skb header data */
3989 		dma_unmap_single(tx_ring->dev,
3990 				 dma_unmap_addr(tx_buf, dma),
3991 				 dma_unmap_len(tx_buf, len),
3992 				 DMA_TO_DEVICE);
3993 		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3994 			kfree(tx_buf->raw_buf);
3995 
3996 		tx_buf->raw_buf = NULL;
3997 		tx_buf->tx_flags = 0;
3998 		tx_buf->next_to_watch = NULL;
3999 		dma_unmap_len_set(tx_buf, len, 0);
4000 		tx_desc->buffer_addr = 0;
4001 		tx_desc->cmd_type_offset_bsz = 0;
4002 
4003 		/* move us past the eop_desc for start of next FD desc */
4004 		tx_buf++;
4005 		tx_desc++;
4006 		i++;
4007 		if (unlikely(!i)) {
4008 			i -= tx_ring->count;
4009 			tx_buf = tx_ring->tx_bi;
4010 			tx_desc = I40E_TX_DESC(tx_ring, 0);
4011 		}
4012 
4013 		/* update budget accounting */
4014 		budget--;
4015 	} while (likely(budget));
4016 
4017 	i += tx_ring->count;
4018 	tx_ring->next_to_clean = i;
4019 
4020 	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
4021 		i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
4022 
4023 	return budget > 0;
4024 }
4025 
4026 /**
4027  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
4028  * @irq: interrupt number
4029  * @data: pointer to a q_vector
4030  **/
4031 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
4032 {
4033 	struct i40e_q_vector *q_vector = data;
4034 	struct i40e_vsi *vsi;
4035 
4036 	if (!q_vector->tx.ring)
4037 		return IRQ_HANDLED;
4038 
4039 	vsi = q_vector->tx.ring->vsi;
4040 	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
4041 
4042 	return IRQ_HANDLED;
4043 }
4044 
4045 /**
4046  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
4047  * @vsi: the VSI being configured
4048  * @v_idx: vector index
4049  * @qp_idx: queue pair index
4050  **/
4051 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
4052 {
4053 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4054 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
4055 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
4056 
4057 	tx_ring->q_vector = q_vector;
4058 	tx_ring->next = q_vector->tx.ring;
4059 	q_vector->tx.ring = tx_ring;
4060 	q_vector->tx.count++;
4061 
4062 	/* Place XDP Tx ring in the same q_vector ring list as regular Tx */
4063 	if (i40e_enabled_xdp_vsi(vsi)) {
4064 		struct i40e_ring *xdp_ring = vsi->xdp_rings[qp_idx];
4065 
4066 		xdp_ring->q_vector = q_vector;
4067 		xdp_ring->next = q_vector->tx.ring;
4068 		q_vector->tx.ring = xdp_ring;
4069 		q_vector->tx.count++;
4070 	}
4071 
4072 	rx_ring->q_vector = q_vector;
4073 	rx_ring->next = q_vector->rx.ring;
4074 	q_vector->rx.ring = rx_ring;
4075 	q_vector->rx.count++;
4076 }
4077 
4078 /**
4079  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
4080  * @vsi: the VSI being configured
4081  *
4082  * This function maps descriptor rings to the queue-specific vectors
4083  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
4084  * one vector per queue pair, but on a constrained vector budget, we
4085  * group the queue pairs as "efficiently" as possible.
4086  **/
4087 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
4088 {
4089 	int qp_remaining = vsi->num_queue_pairs;
4090 	int q_vectors = vsi->num_q_vectors;
4091 	int num_ringpairs;
4092 	int v_start = 0;
4093 	int qp_idx = 0;
4094 
4095 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
4096 	 * group them so there are multiple queues per vector.
4097 	 * It is also important to go through all the vectors available to be
4098 	 * sure that if we don't use all the vectors, that the remaining vectors
4099 	 * are cleared. This is especially important when decreasing the
4100 	 * number of queues in use.
4101 	 */
4102 	for (; v_start < q_vectors; v_start++) {
4103 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
4104 
4105 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
4106 
4107 		q_vector->num_ringpairs = num_ringpairs;
4108 
4109 		q_vector->rx.count = 0;
4110 		q_vector->tx.count = 0;
4111 		q_vector->rx.ring = NULL;
4112 		q_vector->tx.ring = NULL;
4113 
4114 		while (num_ringpairs--) {
4115 			i40e_map_vector_to_qp(vsi, v_start, qp_idx);
4116 			qp_idx++;
4117 			qp_remaining--;
4118 		}
4119 	}
4120 }
4121 
4122 /**
4123  * i40e_vsi_request_irq - Request IRQ from the OS
4124  * @vsi: the VSI being configured
4125  * @basename: name for the vector
4126  **/
4127 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
4128 {
4129 	struct i40e_pf *pf = vsi->back;
4130 	int err;
4131 
4132 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4133 		err = i40e_vsi_request_irq_msix(vsi, basename);
4134 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
4135 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
4136 				  pf->int_name, pf);
4137 	else
4138 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
4139 				  pf->int_name, pf);
4140 
4141 	if (err)
4142 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
4143 
4144 	return err;
4145 }
4146 
4147 #ifdef CONFIG_NET_POLL_CONTROLLER
4148 /**
4149  * i40e_netpoll - A Polling 'interrupt' handler
4150  * @netdev: network interface device structure
4151  *
4152  * This is used by netconsole to send skbs without having to re-enable
4153  * interrupts.  It's not called while the normal interrupt routine is executing.
4154  **/
4155 static void i40e_netpoll(struct net_device *netdev)
4156 {
4157 	struct i40e_netdev_priv *np = netdev_priv(netdev);
4158 	struct i40e_vsi *vsi = np->vsi;
4159 	struct i40e_pf *pf = vsi->back;
4160 	int i;
4161 
4162 	/* if interface is down do nothing */
4163 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
4164 		return;
4165 
4166 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4167 		for (i = 0; i < vsi->num_q_vectors; i++)
4168 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
4169 	} else {
4170 		i40e_intr(pf->pdev->irq, netdev);
4171 	}
4172 }
4173 #endif
4174 
4175 #define I40E_QTX_ENA_WAIT_COUNT 50
4176 
4177 /**
4178  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
4179  * @pf: the PF being configured
4180  * @pf_q: the PF queue
4181  * @enable: enable or disable state of the queue
4182  *
4183  * This routine will wait for the given Tx queue of the PF to reach the
4184  * enabled or disabled state.
4185  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4186  * multiple retries; else will return 0 in case of success.
4187  **/
4188 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4189 {
4190 	int i;
4191 	u32 tx_reg;
4192 
4193 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4194 		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
4195 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4196 			break;
4197 
4198 		usleep_range(10, 20);
4199 	}
4200 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4201 		return -ETIMEDOUT;
4202 
4203 	return 0;
4204 }
4205 
4206 /**
4207  * i40e_control_tx_q - Start or stop a particular Tx queue
4208  * @pf: the PF structure
4209  * @pf_q: the PF queue to configure
4210  * @enable: start or stop the queue
4211  *
4212  * This function enables or disables a single queue. Note that any delay
4213  * required after the operation is expected to be handled by the caller of
4214  * this function.
4215  **/
4216 static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
4217 {
4218 	struct i40e_hw *hw = &pf->hw;
4219 	u32 tx_reg;
4220 	int i;
4221 
4222 	/* warn the TX unit of coming changes */
4223 	i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
4224 	if (!enable)
4225 		usleep_range(10, 20);
4226 
4227 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4228 		tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
4229 		if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
4230 		    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
4231 			break;
4232 		usleep_range(1000, 2000);
4233 	}
4234 
4235 	/* Skip if the queue is already in the requested state */
4236 	if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4237 		return;
4238 
4239 	/* turn on/off the queue */
4240 	if (enable) {
4241 		wr32(hw, I40E_QTX_HEAD(pf_q), 0);
4242 		tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
4243 	} else {
4244 		tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
4245 	}
4246 
4247 	wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
4248 }
4249 
4250 /**
4251  * i40e_control_wait_tx_q - Start/stop Tx queue and wait for completion
4252  * @seid: VSI SEID
4253  * @pf: the PF structure
4254  * @pf_q: the PF queue to configure
4255  * @is_xdp: true if the queue is used for XDP
4256  * @enable: start or stop the queue
4257  **/
4258 static int i40e_control_wait_tx_q(int seid, struct i40e_pf *pf, int pf_q,
4259 				  bool is_xdp, bool enable)
4260 {
4261 	int ret;
4262 
4263 	i40e_control_tx_q(pf, pf_q, enable);
4264 
4265 	/* wait for the change to finish */
4266 	ret = i40e_pf_txq_wait(pf, pf_q, enable);
4267 	if (ret) {
4268 		dev_info(&pf->pdev->dev,
4269 			 "VSI seid %d %sTx ring %d %sable timeout\n",
4270 			 seid, (is_xdp ? "XDP " : ""), pf_q,
4271 			 (enable ? "en" : "dis"));
4272 	}
4273 
4274 	return ret;
4275 }
4276 
4277 /**
4278  * i40e_vsi_control_tx - Start or stop a VSI's rings
4279  * @vsi: the VSI being configured
4280  * @enable: start or stop the rings
4281  **/
4282 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4283 {
4284 	struct i40e_pf *pf = vsi->back;
4285 	int i, pf_q, ret = 0;
4286 
4287 	pf_q = vsi->base_queue;
4288 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4289 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4290 					     pf_q,
4291 					     false /*is xdp*/, enable);
4292 		if (ret)
4293 			break;
4294 
4295 		if (!i40e_enabled_xdp_vsi(vsi))
4296 			continue;
4297 
4298 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4299 					     pf_q + vsi->alloc_queue_pairs,
4300 					     true /*is xdp*/, enable);
4301 		if (ret)
4302 			break;
4303 	}
4304 
4305 	return ret;
4306 }
4307 
4308 /**
4309  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4310  * @pf: the PF being configured
4311  * @pf_q: the PF queue
4312  * @enable: enable or disable state of the queue
4313  *
4314  * This routine will wait for the given Rx queue of the PF to reach the
4315  * enabled or disabled state.
4316  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4317  * multiple retries; else will return 0 in case of success.
4318  **/
4319 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4320 {
4321 	int i;
4322 	u32 rx_reg;
4323 
4324 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4325 		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4326 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4327 			break;
4328 
4329 		usleep_range(10, 20);
4330 	}
4331 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4332 		return -ETIMEDOUT;
4333 
4334 	return 0;
4335 }
4336 
4337 /**
4338  * i40e_control_rx_q - Start or stop a particular Rx queue
4339  * @pf: the PF structure
4340  * @pf_q: the PF queue to configure
4341  * @enable: start or stop the queue
4342  *
4343  * This function enables or disables a single queue. Note that any delay
4344  * required after the operation is expected to be handled by the caller of
4345  * this function.
4346  **/
4347 static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4348 {
4349 	struct i40e_hw *hw = &pf->hw;
4350 	u32 rx_reg;
4351 	int i;
4352 
4353 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4354 		rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4355 		if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4356 		    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4357 			break;
4358 		usleep_range(1000, 2000);
4359 	}
4360 
4361 	/* Skip if the queue is already in the requested state */
4362 	if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4363 		return;
4364 
4365 	/* turn on/off the queue */
4366 	if (enable)
4367 		rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4368 	else
4369 		rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4370 
4371 	wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4372 }
4373 
4374 /**
4375  * i40e_vsi_control_rx - Start or stop a VSI's rings
4376  * @vsi: the VSI being configured
4377  * @enable: start or stop the rings
4378  **/
4379 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4380 {
4381 	struct i40e_pf *pf = vsi->back;
4382 	int i, pf_q, ret = 0;
4383 
4384 	pf_q = vsi->base_queue;
4385 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4386 		i40e_control_rx_q(pf, pf_q, enable);
4387 
4388 		/* wait for the change to finish */
4389 		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4390 		if (ret) {
4391 			dev_info(&pf->pdev->dev,
4392 				 "VSI seid %d Rx ring %d %sable timeout\n",
4393 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
4394 			break;
4395 		}
4396 	}
4397 
4398 	/* Due to HW errata, on Rx disable only, the register can indicate done
4399 	 * before it really is. Needs 50ms to be sure
4400 	 */
4401 	if (!enable)
4402 		mdelay(50);
4403 
4404 	return ret;
4405 }
4406 
4407 /**
4408  * i40e_vsi_start_rings - Start a VSI's rings
4409  * @vsi: the VSI being configured
4410  **/
4411 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4412 {
4413 	int ret = 0;
4414 
4415 	/* do rx first for enable and last for disable */
4416 	ret = i40e_vsi_control_rx(vsi, true);
4417 	if (ret)
4418 		return ret;
4419 	ret = i40e_vsi_control_tx(vsi, true);
4420 
4421 	return ret;
4422 }
4423 
4424 /**
4425  * i40e_vsi_stop_rings - Stop a VSI's rings
4426  * @vsi: the VSI being configured
4427  **/
4428 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4429 {
4430 	/* When port TX is suspended, don't wait */
4431 	if (test_bit(__I40E_PORT_SUSPENDED, vsi->back->state))
4432 		return i40e_vsi_stop_rings_no_wait(vsi);
4433 
4434 	/* do rx first for enable and last for disable
4435 	 * Ignore return value, we need to shutdown whatever we can
4436 	 */
4437 	i40e_vsi_control_tx(vsi, false);
4438 	i40e_vsi_control_rx(vsi, false);
4439 }
4440 
4441 /**
4442  * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4443  * @vsi: the VSI being shutdown
4444  *
4445  * This function stops all the rings for a VSI but does not delay to verify
4446  * that rings have been disabled. It is expected that the caller is shutting
4447  * down multiple VSIs at once and will delay together for all the VSIs after
4448  * initiating the shutdown. This is particularly useful for shutting down lots
4449  * of VFs together. Otherwise, a large delay can be incurred while configuring
4450  * each VSI in serial.
4451  **/
4452 void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4453 {
4454 	struct i40e_pf *pf = vsi->back;
4455 	int i, pf_q;
4456 
4457 	pf_q = vsi->base_queue;
4458 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4459 		i40e_control_tx_q(pf, pf_q, false);
4460 		i40e_control_rx_q(pf, pf_q, false);
4461 	}
4462 }
4463 
4464 /**
4465  * i40e_vsi_free_irq - Free the irq association with the OS
4466  * @vsi: the VSI being configured
4467  **/
4468 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4469 {
4470 	struct i40e_pf *pf = vsi->back;
4471 	struct i40e_hw *hw = &pf->hw;
4472 	int base = vsi->base_vector;
4473 	u32 val, qp;
4474 	int i;
4475 
4476 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4477 		if (!vsi->q_vectors)
4478 			return;
4479 
4480 		if (!vsi->irqs_ready)
4481 			return;
4482 
4483 		vsi->irqs_ready = false;
4484 		for (i = 0; i < vsi->num_q_vectors; i++) {
4485 			int irq_num;
4486 			u16 vector;
4487 
4488 			vector = i + base;
4489 			irq_num = pf->msix_entries[vector].vector;
4490 
4491 			/* free only the irqs that were actually requested */
4492 			if (!vsi->q_vectors[i] ||
4493 			    !vsi->q_vectors[i]->num_ringpairs)
4494 				continue;
4495 
4496 			/* clear the affinity notifier in the IRQ descriptor */
4497 			irq_set_affinity_notifier(irq_num, NULL);
4498 			/* remove our suggested affinity mask for this IRQ */
4499 			irq_set_affinity_hint(irq_num, NULL);
4500 			synchronize_irq(irq_num);
4501 			free_irq(irq_num, vsi->q_vectors[i]);
4502 
4503 			/* Tear down the interrupt queue link list
4504 			 *
4505 			 * We know that they come in pairs and always
4506 			 * the Rx first, then the Tx.  To clear the
4507 			 * link list, stick the EOL value into the
4508 			 * next_q field of the registers.
4509 			 */
4510 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4511 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4512 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4513 			val |= I40E_QUEUE_END_OF_LIST
4514 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4515 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4516 
4517 			while (qp != I40E_QUEUE_END_OF_LIST) {
4518 				u32 next;
4519 
4520 				val = rd32(hw, I40E_QINT_RQCTL(qp));
4521 
4522 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4523 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4524 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4525 					 I40E_QINT_RQCTL_INTEVENT_MASK);
4526 
4527 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4528 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4529 
4530 				wr32(hw, I40E_QINT_RQCTL(qp), val);
4531 
4532 				val = rd32(hw, I40E_QINT_TQCTL(qp));
4533 
4534 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4535 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4536 
4537 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4538 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4539 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4540 					 I40E_QINT_TQCTL_INTEVENT_MASK);
4541 
4542 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4543 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4544 
4545 				wr32(hw, I40E_QINT_TQCTL(qp), val);
4546 				qp = next;
4547 			}
4548 		}
4549 	} else {
4550 		free_irq(pf->pdev->irq, pf);
4551 
4552 		val = rd32(hw, I40E_PFINT_LNKLST0);
4553 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4554 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4555 		val |= I40E_QUEUE_END_OF_LIST
4556 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4557 		wr32(hw, I40E_PFINT_LNKLST0, val);
4558 
4559 		val = rd32(hw, I40E_QINT_RQCTL(qp));
4560 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4561 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4562 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4563 			 I40E_QINT_RQCTL_INTEVENT_MASK);
4564 
4565 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4566 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4567 
4568 		wr32(hw, I40E_QINT_RQCTL(qp), val);
4569 
4570 		val = rd32(hw, I40E_QINT_TQCTL(qp));
4571 
4572 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4573 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4574 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4575 			 I40E_QINT_TQCTL_INTEVENT_MASK);
4576 
4577 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4578 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4579 
4580 		wr32(hw, I40E_QINT_TQCTL(qp), val);
4581 	}
4582 }
4583 
4584 /**
4585  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4586  * @vsi: the VSI being configured
4587  * @v_idx: Index of vector to be freed
4588  *
4589  * This function frees the memory allocated to the q_vector.  In addition if
4590  * NAPI is enabled it will delete any references to the NAPI struct prior
4591  * to freeing the q_vector.
4592  **/
4593 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4594 {
4595 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4596 	struct i40e_ring *ring;
4597 
4598 	if (!q_vector)
4599 		return;
4600 
4601 	/* disassociate q_vector from rings */
4602 	i40e_for_each_ring(ring, q_vector->tx)
4603 		ring->q_vector = NULL;
4604 
4605 	i40e_for_each_ring(ring, q_vector->rx)
4606 		ring->q_vector = NULL;
4607 
4608 	/* only VSI w/ an associated netdev is set up w/ NAPI */
4609 	if (vsi->netdev)
4610 		netif_napi_del(&q_vector->napi);
4611 
4612 	vsi->q_vectors[v_idx] = NULL;
4613 
4614 	kfree_rcu(q_vector, rcu);
4615 }
4616 
4617 /**
4618  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4619  * @vsi: the VSI being un-configured
4620  *
4621  * This frees the memory allocated to the q_vectors and
4622  * deletes references to the NAPI struct.
4623  **/
4624 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4625 {
4626 	int v_idx;
4627 
4628 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4629 		i40e_free_q_vector(vsi, v_idx);
4630 }
4631 
4632 /**
4633  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4634  * @pf: board private structure
4635  **/
4636 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4637 {
4638 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4639 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4640 		pci_disable_msix(pf->pdev);
4641 		kfree(pf->msix_entries);
4642 		pf->msix_entries = NULL;
4643 		kfree(pf->irq_pile);
4644 		pf->irq_pile = NULL;
4645 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4646 		pci_disable_msi(pf->pdev);
4647 	}
4648 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4649 }
4650 
4651 /**
4652  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4653  * @pf: board private structure
4654  *
4655  * We go through and clear interrupt specific resources and reset the structure
4656  * to pre-load conditions
4657  **/
4658 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4659 {
4660 	int i;
4661 
4662 	i40e_free_misc_vector(pf);
4663 
4664 	i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4665 		      I40E_IWARP_IRQ_PILE_ID);
4666 
4667 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4668 	for (i = 0; i < pf->num_alloc_vsi; i++)
4669 		if (pf->vsi[i])
4670 			i40e_vsi_free_q_vectors(pf->vsi[i]);
4671 	i40e_reset_interrupt_capability(pf);
4672 }
4673 
4674 /**
4675  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4676  * @vsi: the VSI being configured
4677  **/
4678 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4679 {
4680 	int q_idx;
4681 
4682 	if (!vsi->netdev)
4683 		return;
4684 
4685 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4686 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4687 
4688 		if (q_vector->rx.ring || q_vector->tx.ring)
4689 			napi_enable(&q_vector->napi);
4690 	}
4691 }
4692 
4693 /**
4694  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4695  * @vsi: the VSI being configured
4696  **/
4697 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4698 {
4699 	int q_idx;
4700 
4701 	if (!vsi->netdev)
4702 		return;
4703 
4704 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4705 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4706 
4707 		if (q_vector->rx.ring || q_vector->tx.ring)
4708 			napi_disable(&q_vector->napi);
4709 	}
4710 }
4711 
4712 /**
4713  * i40e_vsi_close - Shut down a VSI
4714  * @vsi: the vsi to be quelled
4715  **/
4716 static void i40e_vsi_close(struct i40e_vsi *vsi)
4717 {
4718 	struct i40e_pf *pf = vsi->back;
4719 	if (!test_and_set_bit(__I40E_VSI_DOWN, vsi->state))
4720 		i40e_down(vsi);
4721 	i40e_vsi_free_irq(vsi);
4722 	i40e_vsi_free_tx_resources(vsi);
4723 	i40e_vsi_free_rx_resources(vsi);
4724 	vsi->current_netdev_flags = 0;
4725 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
4726 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
4727 		pf->flags |=  I40E_FLAG_CLIENT_RESET;
4728 }
4729 
4730 /**
4731  * i40e_quiesce_vsi - Pause a given VSI
4732  * @vsi: the VSI being paused
4733  **/
4734 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4735 {
4736 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
4737 		return;
4738 
4739 	set_bit(__I40E_VSI_NEEDS_RESTART, vsi->state);
4740 	if (vsi->netdev && netif_running(vsi->netdev))
4741 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4742 	else
4743 		i40e_vsi_close(vsi);
4744 }
4745 
4746 /**
4747  * i40e_unquiesce_vsi - Resume a given VSI
4748  * @vsi: the VSI being resumed
4749  **/
4750 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4751 {
4752 	if (!test_and_clear_bit(__I40E_VSI_NEEDS_RESTART, vsi->state))
4753 		return;
4754 
4755 	if (vsi->netdev && netif_running(vsi->netdev))
4756 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4757 	else
4758 		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4759 }
4760 
4761 /**
4762  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4763  * @pf: the PF
4764  **/
4765 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4766 {
4767 	int v;
4768 
4769 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4770 		if (pf->vsi[v])
4771 			i40e_quiesce_vsi(pf->vsi[v]);
4772 	}
4773 }
4774 
4775 /**
4776  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4777  * @pf: the PF
4778  **/
4779 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4780 {
4781 	int v;
4782 
4783 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4784 		if (pf->vsi[v])
4785 			i40e_unquiesce_vsi(pf->vsi[v]);
4786 	}
4787 }
4788 
4789 /**
4790  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4791  * @vsi: the VSI being configured
4792  *
4793  * Wait until all queues on a given VSI have been disabled.
4794  **/
4795 int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4796 {
4797 	struct i40e_pf *pf = vsi->back;
4798 	int i, pf_q, ret;
4799 
4800 	pf_q = vsi->base_queue;
4801 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4802 		/* Check and wait for the Tx queue */
4803 		ret = i40e_pf_txq_wait(pf, pf_q, false);
4804 		if (ret) {
4805 			dev_info(&pf->pdev->dev,
4806 				 "VSI seid %d Tx ring %d disable timeout\n",
4807 				 vsi->seid, pf_q);
4808 			return ret;
4809 		}
4810 
4811 		if (!i40e_enabled_xdp_vsi(vsi))
4812 			goto wait_rx;
4813 
4814 		/* Check and wait for the XDP Tx queue */
4815 		ret = i40e_pf_txq_wait(pf, pf_q + vsi->alloc_queue_pairs,
4816 				       false);
4817 		if (ret) {
4818 			dev_info(&pf->pdev->dev,
4819 				 "VSI seid %d XDP Tx ring %d disable timeout\n",
4820 				 vsi->seid, pf_q);
4821 			return ret;
4822 		}
4823 wait_rx:
4824 		/* Check and wait for the Rx queue */
4825 		ret = i40e_pf_rxq_wait(pf, pf_q, false);
4826 		if (ret) {
4827 			dev_info(&pf->pdev->dev,
4828 				 "VSI seid %d Rx ring %d disable timeout\n",
4829 				 vsi->seid, pf_q);
4830 			return ret;
4831 		}
4832 	}
4833 
4834 	return 0;
4835 }
4836 
4837 #ifdef CONFIG_I40E_DCB
4838 /**
4839  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4840  * @pf: the PF
4841  *
4842  * This function waits for the queues to be in disabled state for all the
4843  * VSIs that are managed by this PF.
4844  **/
4845 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4846 {
4847 	int v, ret = 0;
4848 
4849 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4850 		if (pf->vsi[v]) {
4851 			ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4852 			if (ret)
4853 				break;
4854 		}
4855 	}
4856 
4857 	return ret;
4858 }
4859 
4860 #endif
4861 
4862 /**
4863  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4864  * @q_idx: TX queue number
4865  * @vsi: Pointer to VSI struct
4866  *
4867  * This function checks specified queue for given VSI. Detects hung condition.
4868  * We proactively detect hung TX queues by checking if interrupts are disabled
4869  * but there are pending descriptors.  If it appears hung, attempt to recover
4870  * by triggering a SW interrupt.
4871  **/
4872 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4873 {
4874 	struct i40e_ring *tx_ring = NULL;
4875 	struct i40e_pf	*pf;
4876 	u32 val, tx_pending;
4877 	int i;
4878 
4879 	pf = vsi->back;
4880 
4881 	/* now that we have an index, find the tx_ring struct */
4882 	for (i = 0; i < vsi->num_queue_pairs; i++) {
4883 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4884 			if (q_idx == vsi->tx_rings[i]->queue_index) {
4885 				tx_ring = vsi->tx_rings[i];
4886 				break;
4887 			}
4888 		}
4889 	}
4890 
4891 	if (!tx_ring)
4892 		return;
4893 
4894 	/* Read interrupt register */
4895 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4896 		val = rd32(&pf->hw,
4897 			   I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4898 					       tx_ring->vsi->base_vector - 1));
4899 	else
4900 		val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4901 
4902 	tx_pending = i40e_get_tx_pending(tx_ring);
4903 
4904 	/* Interrupts are disabled and TX pending is non-zero,
4905 	 * trigger the SW interrupt (don't wait). Worst case
4906 	 * there will be one extra interrupt which may result
4907 	 * into not cleaning any queues because queues are cleaned.
4908 	 */
4909 	if (tx_pending && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)))
4910 		i40e_force_wb(vsi, tx_ring->q_vector);
4911 }
4912 
4913 /**
4914  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4915  * @pf:  pointer to PF struct
4916  *
4917  * LAN VSI has netdev and netdev has TX queues. This function is to check
4918  * each of those TX queues if they are hung, trigger recovery by issuing
4919  * SW interrupt.
4920  **/
4921 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4922 {
4923 	struct net_device *netdev;
4924 	struct i40e_vsi *vsi;
4925 	unsigned int i;
4926 
4927 	/* Only for LAN VSI */
4928 	vsi = pf->vsi[pf->lan_vsi];
4929 
4930 	if (!vsi)
4931 		return;
4932 
4933 	/* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4934 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
4935 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
4936 		return;
4937 
4938 	/* Make sure type is MAIN VSI */
4939 	if (vsi->type != I40E_VSI_MAIN)
4940 		return;
4941 
4942 	netdev = vsi->netdev;
4943 	if (!netdev)
4944 		return;
4945 
4946 	/* Bail out if netif_carrier is not OK */
4947 	if (!netif_carrier_ok(netdev))
4948 		return;
4949 
4950 	/* Go thru' TX queues for netdev */
4951 	for (i = 0; i < netdev->num_tx_queues; i++) {
4952 		struct netdev_queue *q;
4953 
4954 		q = netdev_get_tx_queue(netdev, i);
4955 		if (q)
4956 			i40e_detect_recover_hung_queue(i, vsi);
4957 	}
4958 }
4959 
4960 /**
4961  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4962  * @pf: pointer to PF
4963  *
4964  * Get TC map for ISCSI PF type that will include iSCSI TC
4965  * and LAN TC.
4966  **/
4967 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4968 {
4969 	struct i40e_dcb_app_priority_table app;
4970 	struct i40e_hw *hw = &pf->hw;
4971 	u8 enabled_tc = 1; /* TC0 is always enabled */
4972 	u8 tc, i;
4973 	/* Get the iSCSI APP TLV */
4974 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4975 
4976 	for (i = 0; i < dcbcfg->numapps; i++) {
4977 		app = dcbcfg->app[i];
4978 		if (app.selector == I40E_APP_SEL_TCPIP &&
4979 		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4980 			tc = dcbcfg->etscfg.prioritytable[app.priority];
4981 			enabled_tc |= BIT(tc);
4982 			break;
4983 		}
4984 	}
4985 
4986 	return enabled_tc;
4987 }
4988 
4989 /**
4990  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4991  * @dcbcfg: the corresponding DCBx configuration structure
4992  *
4993  * Return the number of TCs from given DCBx configuration
4994  **/
4995 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4996 {
4997 	int i, tc_unused = 0;
4998 	u8 num_tc = 0;
4999 	u8 ret = 0;
5000 
5001 	/* Scan the ETS Config Priority Table to find
5002 	 * traffic class enabled for a given priority
5003 	 * and create a bitmask of enabled TCs
5004 	 */
5005 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
5006 		num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
5007 
5008 	/* Now scan the bitmask to check for
5009 	 * contiguous TCs starting with TC0
5010 	 */
5011 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5012 		if (num_tc & BIT(i)) {
5013 			if (!tc_unused) {
5014 				ret++;
5015 			} else {
5016 				pr_err("Non-contiguous TC - Disabling DCB\n");
5017 				return 1;
5018 			}
5019 		} else {
5020 			tc_unused = 1;
5021 		}
5022 	}
5023 
5024 	/* There is always at least TC0 */
5025 	if (!ret)
5026 		ret = 1;
5027 
5028 	return ret;
5029 }
5030 
5031 /**
5032  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
5033  * @dcbcfg: the corresponding DCBx configuration structure
5034  *
5035  * Query the current DCB configuration and return the number of
5036  * traffic classes enabled from the given DCBX config
5037  **/
5038 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
5039 {
5040 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
5041 	u8 enabled_tc = 1;
5042 	u8 i;
5043 
5044 	for (i = 0; i < num_tc; i++)
5045 		enabled_tc |= BIT(i);
5046 
5047 	return enabled_tc;
5048 }
5049 
5050 /**
5051  * i40e_mqprio_get_enabled_tc - Get enabled traffic classes
5052  * @pf: PF being queried
5053  *
5054  * Query the current MQPRIO configuration and return the number of
5055  * traffic classes enabled.
5056  **/
5057 static u8 i40e_mqprio_get_enabled_tc(struct i40e_pf *pf)
5058 {
5059 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5060 	u8 num_tc = vsi->mqprio_qopt.qopt.num_tc;
5061 	u8 enabled_tc = 1, i;
5062 
5063 	for (i = 1; i < num_tc; i++)
5064 		enabled_tc |= BIT(i);
5065 	return enabled_tc;
5066 }
5067 
5068 /**
5069  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
5070  * @pf: PF being queried
5071  *
5072  * Return number of traffic classes enabled for the given PF
5073  **/
5074 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
5075 {
5076 	struct i40e_hw *hw = &pf->hw;
5077 	u8 i, enabled_tc = 1;
5078 	u8 num_tc = 0;
5079 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5080 
5081 	if (pf->flags & I40E_FLAG_TC_MQPRIO)
5082 		return pf->vsi[pf->lan_vsi]->mqprio_qopt.qopt.num_tc;
5083 
5084 	/* If neither MQPRIO nor DCB is enabled, then always use single TC */
5085 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5086 		return 1;
5087 
5088 	/* SFP mode will be enabled for all TCs on port */
5089 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5090 		return i40e_dcb_get_num_tc(dcbcfg);
5091 
5092 	/* MFP mode return count of enabled TCs for this PF */
5093 	if (pf->hw.func_caps.iscsi)
5094 		enabled_tc =  i40e_get_iscsi_tc_map(pf);
5095 	else
5096 		return 1; /* Only TC0 */
5097 
5098 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5099 		if (enabled_tc & BIT(i))
5100 			num_tc++;
5101 	}
5102 	return num_tc;
5103 }
5104 
5105 /**
5106  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
5107  * @pf: PF being queried
5108  *
5109  * Return a bitmap for enabled traffic classes for this PF.
5110  **/
5111 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
5112 {
5113 	if (pf->flags & I40E_FLAG_TC_MQPRIO)
5114 		return i40e_mqprio_get_enabled_tc(pf);
5115 
5116 	/* If neither MQPRIO nor DCB is enabled for this PF then just return
5117 	 * default TC
5118 	 */
5119 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5120 		return I40E_DEFAULT_TRAFFIC_CLASS;
5121 
5122 	/* SFP mode we want PF to be enabled for all TCs */
5123 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5124 		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
5125 
5126 	/* MFP enabled and iSCSI PF type */
5127 	if (pf->hw.func_caps.iscsi)
5128 		return i40e_get_iscsi_tc_map(pf);
5129 	else
5130 		return I40E_DEFAULT_TRAFFIC_CLASS;
5131 }
5132 
5133 /**
5134  * i40e_vsi_get_bw_info - Query VSI BW Information
5135  * @vsi: the VSI being queried
5136  *
5137  * Returns 0 on success, negative value on failure
5138  **/
5139 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
5140 {
5141 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
5142 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
5143 	struct i40e_pf *pf = vsi->back;
5144 	struct i40e_hw *hw = &pf->hw;
5145 	i40e_status ret;
5146 	u32 tc_bw_max;
5147 	int i;
5148 
5149 	/* Get the VSI level BW configuration */
5150 	ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
5151 	if (ret) {
5152 		dev_info(&pf->pdev->dev,
5153 			 "couldn't get PF vsi bw config, err %s aq_err %s\n",
5154 			 i40e_stat_str(&pf->hw, ret),
5155 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5156 		return -EINVAL;
5157 	}
5158 
5159 	/* Get the VSI level BW configuration per TC */
5160 	ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
5161 					       NULL);
5162 	if (ret) {
5163 		dev_info(&pf->pdev->dev,
5164 			 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
5165 			 i40e_stat_str(&pf->hw, ret),
5166 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5167 		return -EINVAL;
5168 	}
5169 
5170 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
5171 		dev_info(&pf->pdev->dev,
5172 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
5173 			 bw_config.tc_valid_bits,
5174 			 bw_ets_config.tc_valid_bits);
5175 		/* Still continuing */
5176 	}
5177 
5178 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
5179 	vsi->bw_max_quanta = bw_config.max_bw;
5180 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
5181 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
5182 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5183 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
5184 		vsi->bw_ets_limit_credits[i] =
5185 					le16_to_cpu(bw_ets_config.credits[i]);
5186 		/* 3 bits out of 4 for each TC */
5187 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
5188 	}
5189 
5190 	return 0;
5191 }
5192 
5193 /**
5194  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
5195  * @vsi: the VSI being configured
5196  * @enabled_tc: TC bitmap
5197  * @bw_credits: BW shared credits per TC
5198  *
5199  * Returns 0 on success, negative value on failure
5200  **/
5201 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
5202 				       u8 *bw_share)
5203 {
5204 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5205 	i40e_status ret;
5206 	int i;
5207 
5208 	if (vsi->back->flags & I40E_FLAG_TC_MQPRIO)
5209 		return 0;
5210 	if (!vsi->mqprio_qopt.qopt.hw) {
5211 		ret = i40e_set_bw_limit(vsi, vsi->seid, 0);
5212 		if (ret)
5213 			dev_info(&vsi->back->pdev->dev,
5214 				 "Failed to reset tx rate for vsi->seid %u\n",
5215 				 vsi->seid);
5216 		return ret;
5217 	}
5218 	bw_data.tc_valid_bits = enabled_tc;
5219 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5220 		bw_data.tc_bw_credits[i] = bw_share[i];
5221 
5222 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
5223 				       NULL);
5224 	if (ret) {
5225 		dev_info(&vsi->back->pdev->dev,
5226 			 "AQ command Config VSI BW allocation per TC failed = %d\n",
5227 			 vsi->back->hw.aq.asq_last_status);
5228 		return -EINVAL;
5229 	}
5230 
5231 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5232 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
5233 
5234 	return 0;
5235 }
5236 
5237 /**
5238  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
5239  * @vsi: the VSI being configured
5240  * @enabled_tc: TC map to be enabled
5241  *
5242  **/
5243 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5244 {
5245 	struct net_device *netdev = vsi->netdev;
5246 	struct i40e_pf *pf = vsi->back;
5247 	struct i40e_hw *hw = &pf->hw;
5248 	u8 netdev_tc = 0;
5249 	int i;
5250 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5251 
5252 	if (!netdev)
5253 		return;
5254 
5255 	if (!enabled_tc) {
5256 		netdev_reset_tc(netdev);
5257 		return;
5258 	}
5259 
5260 	/* Set up actual enabled TCs on the VSI */
5261 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
5262 		return;
5263 
5264 	/* set per TC queues for the VSI */
5265 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5266 		/* Only set TC queues for enabled tcs
5267 		 *
5268 		 * e.g. For a VSI that has TC0 and TC3 enabled the
5269 		 * enabled_tc bitmap would be 0x00001001; the driver
5270 		 * will set the numtc for netdev as 2 that will be
5271 		 * referenced by the netdev layer as TC 0 and 1.
5272 		 */
5273 		if (vsi->tc_config.enabled_tc & BIT(i))
5274 			netdev_set_tc_queue(netdev,
5275 					vsi->tc_config.tc_info[i].netdev_tc,
5276 					vsi->tc_config.tc_info[i].qcount,
5277 					vsi->tc_config.tc_info[i].qoffset);
5278 	}
5279 
5280 	if (pf->flags & I40E_FLAG_TC_MQPRIO)
5281 		return;
5282 
5283 	/* Assign UP2TC map for the VSI */
5284 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
5285 		/* Get the actual TC# for the UP */
5286 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
5287 		/* Get the mapped netdev TC# for the UP */
5288 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
5289 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
5290 	}
5291 }
5292 
5293 /**
5294  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
5295  * @vsi: the VSI being configured
5296  * @ctxt: the ctxt buffer returned from AQ VSI update param command
5297  **/
5298 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
5299 				      struct i40e_vsi_context *ctxt)
5300 {
5301 	/* copy just the sections touched not the entire info
5302 	 * since not all sections are valid as returned by
5303 	 * update vsi params
5304 	 */
5305 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
5306 	memcpy(&vsi->info.queue_mapping,
5307 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
5308 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
5309 	       sizeof(vsi->info.tc_mapping));
5310 }
5311 
5312 /**
5313  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
5314  * @vsi: VSI to be configured
5315  * @enabled_tc: TC bitmap
5316  *
5317  * This configures a particular VSI for TCs that are mapped to the
5318  * given TC bitmap. It uses default bandwidth share for TCs across
5319  * VSIs to configure TC for a particular VSI.
5320  *
5321  * NOTE:
5322  * It is expected that the VSI queues have been quisced before calling
5323  * this function.
5324  **/
5325 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5326 {
5327 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5328 	struct i40e_vsi_context ctxt;
5329 	int ret = 0;
5330 	int i;
5331 
5332 	/* Check if enabled_tc is same as existing or new TCs */
5333 	if (vsi->tc_config.enabled_tc == enabled_tc &&
5334 	    vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
5335 		return ret;
5336 
5337 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
5338 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5339 		if (enabled_tc & BIT(i))
5340 			bw_share[i] = 1;
5341 	}
5342 
5343 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5344 	if (ret) {
5345 		dev_info(&vsi->back->pdev->dev,
5346 			 "Failed configuring TC map %d for VSI %d\n",
5347 			 enabled_tc, vsi->seid);
5348 		goto out;
5349 	}
5350 
5351 	/* Update Queue Pairs Mapping for currently enabled UPs */
5352 	ctxt.seid = vsi->seid;
5353 	ctxt.pf_num = vsi->back->hw.pf_id;
5354 	ctxt.vf_num = 0;
5355 	ctxt.uplink_seid = vsi->uplink_seid;
5356 	ctxt.info = vsi->info;
5357 	if (vsi->back->flags & I40E_FLAG_TC_MQPRIO) {
5358 		ret = i40e_vsi_setup_queue_map_mqprio(vsi, &ctxt, enabled_tc);
5359 		if (ret)
5360 			goto out;
5361 	} else {
5362 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5363 	}
5364 
5365 	/* On destroying the qdisc, reset vsi->rss_size, as number of enabled
5366 	 * queues changed.
5367 	 */
5368 	if (!vsi->mqprio_qopt.qopt.hw && vsi->reconfig_rss) {
5369 		vsi->rss_size = min_t(int, vsi->back->alloc_rss_size,
5370 				      vsi->num_queue_pairs);
5371 		ret = i40e_vsi_config_rss(vsi);
5372 		if (ret) {
5373 			dev_info(&vsi->back->pdev->dev,
5374 				 "Failed to reconfig rss for num_queues\n");
5375 			return ret;
5376 		}
5377 		vsi->reconfig_rss = false;
5378 	}
5379 	if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5380 		ctxt.info.valid_sections |=
5381 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5382 		ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5383 	}
5384 
5385 	/* Update the VSI after updating the VSI queue-mapping
5386 	 * information
5387 	 */
5388 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5389 	if (ret) {
5390 		dev_info(&vsi->back->pdev->dev,
5391 			 "Update vsi tc config failed, err %s aq_err %s\n",
5392 			 i40e_stat_str(&vsi->back->hw, ret),
5393 			 i40e_aq_str(&vsi->back->hw,
5394 				     vsi->back->hw.aq.asq_last_status));
5395 		goto out;
5396 	}
5397 	/* update the local VSI info with updated queue map */
5398 	i40e_vsi_update_queue_map(vsi, &ctxt);
5399 	vsi->info.valid_sections = 0;
5400 
5401 	/* Update current VSI BW information */
5402 	ret = i40e_vsi_get_bw_info(vsi);
5403 	if (ret) {
5404 		dev_info(&vsi->back->pdev->dev,
5405 			 "Failed updating vsi bw info, err %s aq_err %s\n",
5406 			 i40e_stat_str(&vsi->back->hw, ret),
5407 			 i40e_aq_str(&vsi->back->hw,
5408 				     vsi->back->hw.aq.asq_last_status));
5409 		goto out;
5410 	}
5411 
5412 	/* Update the netdev TC setup */
5413 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5414 out:
5415 	return ret;
5416 }
5417 
5418 /**
5419  * i40e_get_link_speed - Returns link speed for the interface
5420  * @vsi: VSI to be configured
5421  *
5422  **/
5423 int i40e_get_link_speed(struct i40e_vsi *vsi)
5424 {
5425 	struct i40e_pf *pf = vsi->back;
5426 
5427 	switch (pf->hw.phy.link_info.link_speed) {
5428 	case I40E_LINK_SPEED_40GB:
5429 		return 40000;
5430 	case I40E_LINK_SPEED_25GB:
5431 		return 25000;
5432 	case I40E_LINK_SPEED_20GB:
5433 		return 20000;
5434 	case I40E_LINK_SPEED_10GB:
5435 		return 10000;
5436 	case I40E_LINK_SPEED_1GB:
5437 		return 1000;
5438 	default:
5439 		return -EINVAL;
5440 	}
5441 }
5442 
5443 /**
5444  * i40e_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
5445  * @vsi: VSI to be configured
5446  * @seid: seid of the channel/VSI
5447  * @max_tx_rate: max TX rate to be configured as BW limit
5448  *
5449  * Helper function to set BW limit for a given VSI
5450  **/
5451 int i40e_set_bw_limit(struct i40e_vsi *vsi, u16 seid, u64 max_tx_rate)
5452 {
5453 	struct i40e_pf *pf = vsi->back;
5454 	u64 credits = 0;
5455 	int speed = 0;
5456 	int ret = 0;
5457 
5458 	speed = i40e_get_link_speed(vsi);
5459 	if (max_tx_rate > speed) {
5460 		dev_err(&pf->pdev->dev,
5461 			"Invalid max tx rate %llu specified for VSI seid %d.",
5462 			max_tx_rate, seid);
5463 		return -EINVAL;
5464 	}
5465 	if (max_tx_rate && max_tx_rate < 50) {
5466 		dev_warn(&pf->pdev->dev,
5467 			 "Setting max tx rate to minimum usable value of 50Mbps.\n");
5468 		max_tx_rate = 50;
5469 	}
5470 
5471 	/* Tx rate credits are in values of 50Mbps, 0 is disabled */
5472 	credits = max_tx_rate;
5473 	do_div(credits, I40E_BW_CREDIT_DIVISOR);
5474 	ret = i40e_aq_config_vsi_bw_limit(&pf->hw, seid, credits,
5475 					  I40E_MAX_BW_INACTIVE_ACCUM, NULL);
5476 	if (ret)
5477 		dev_err(&pf->pdev->dev,
5478 			"Failed set tx rate (%llu Mbps) for vsi->seid %u, err %s aq_err %s\n",
5479 			max_tx_rate, seid, i40e_stat_str(&pf->hw, ret),
5480 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5481 	return ret;
5482 }
5483 
5484 /**
5485  * i40e_remove_queue_channels - Remove queue channels for the TCs
5486  * @vsi: VSI to be configured
5487  *
5488  * Remove queue channels for the TCs
5489  **/
5490 static void i40e_remove_queue_channels(struct i40e_vsi *vsi)
5491 {
5492 	enum i40e_admin_queue_err last_aq_status;
5493 	struct i40e_cloud_filter *cfilter;
5494 	struct i40e_channel *ch, *ch_tmp;
5495 	struct i40e_pf *pf = vsi->back;
5496 	struct hlist_node *node;
5497 	int ret, i;
5498 
5499 	/* Reset rss size that was stored when reconfiguring rss for
5500 	 * channel VSIs with non-power-of-2 queue count.
5501 	 */
5502 	vsi->current_rss_size = 0;
5503 
5504 	/* perform cleanup for channels if they exist */
5505 	if (list_empty(&vsi->ch_list))
5506 		return;
5507 
5508 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5509 		struct i40e_vsi *p_vsi;
5510 
5511 		list_del(&ch->list);
5512 		p_vsi = ch->parent_vsi;
5513 		if (!p_vsi || !ch->initialized) {
5514 			kfree(ch);
5515 			continue;
5516 		}
5517 		/* Reset queue contexts */
5518 		for (i = 0; i < ch->num_queue_pairs; i++) {
5519 			struct i40e_ring *tx_ring, *rx_ring;
5520 			u16 pf_q;
5521 
5522 			pf_q = ch->base_queue + i;
5523 			tx_ring = vsi->tx_rings[pf_q];
5524 			tx_ring->ch = NULL;
5525 
5526 			rx_ring = vsi->rx_rings[pf_q];
5527 			rx_ring->ch = NULL;
5528 		}
5529 
5530 		/* Reset BW configured for this VSI via mqprio */
5531 		ret = i40e_set_bw_limit(vsi, ch->seid, 0);
5532 		if (ret)
5533 			dev_info(&vsi->back->pdev->dev,
5534 				 "Failed to reset tx rate for ch->seid %u\n",
5535 				 ch->seid);
5536 
5537 		/* delete cloud filters associated with this channel */
5538 		hlist_for_each_entry_safe(cfilter, node,
5539 					  &pf->cloud_filter_list, cloud_node) {
5540 			if (cfilter->seid != ch->seid)
5541 				continue;
5542 
5543 			hash_del(&cfilter->cloud_node);
5544 			if (cfilter->dst_port)
5545 				ret = i40e_add_del_cloud_filter_big_buf(vsi,
5546 									cfilter,
5547 									false);
5548 			else
5549 				ret = i40e_add_del_cloud_filter(vsi, cfilter,
5550 								false);
5551 			last_aq_status = pf->hw.aq.asq_last_status;
5552 			if (ret)
5553 				dev_info(&pf->pdev->dev,
5554 					 "Failed to delete cloud filter, err %s aq_err %s\n",
5555 					 i40e_stat_str(&pf->hw, ret),
5556 					 i40e_aq_str(&pf->hw, last_aq_status));
5557 			kfree(cfilter);
5558 		}
5559 
5560 		/* delete VSI from FW */
5561 		ret = i40e_aq_delete_element(&vsi->back->hw, ch->seid,
5562 					     NULL);
5563 		if (ret)
5564 			dev_err(&vsi->back->pdev->dev,
5565 				"unable to remove channel (%d) for parent VSI(%d)\n",
5566 				ch->seid, p_vsi->seid);
5567 		kfree(ch);
5568 	}
5569 	INIT_LIST_HEAD(&vsi->ch_list);
5570 }
5571 
5572 /**
5573  * i40e_is_any_channel - channel exist or not
5574  * @vsi: ptr to VSI to which channels are associated with
5575  *
5576  * Returns true or false if channel(s) exist for associated VSI or not
5577  **/
5578 static bool i40e_is_any_channel(struct i40e_vsi *vsi)
5579 {
5580 	struct i40e_channel *ch, *ch_tmp;
5581 
5582 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5583 		if (ch->initialized)
5584 			return true;
5585 	}
5586 
5587 	return false;
5588 }
5589 
5590 /**
5591  * i40e_get_max_queues_for_channel
5592  * @vsi: ptr to VSI to which channels are associated with
5593  *
5594  * Helper function which returns max value among the queue counts set on the
5595  * channels/TCs created.
5596  **/
5597 static int i40e_get_max_queues_for_channel(struct i40e_vsi *vsi)
5598 {
5599 	struct i40e_channel *ch, *ch_tmp;
5600 	int max = 0;
5601 
5602 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5603 		if (!ch->initialized)
5604 			continue;
5605 		if (ch->num_queue_pairs > max)
5606 			max = ch->num_queue_pairs;
5607 	}
5608 
5609 	return max;
5610 }
5611 
5612 /**
5613  * i40e_validate_num_queues - validate num_queues w.r.t channel
5614  * @pf: ptr to PF device
5615  * @num_queues: number of queues
5616  * @vsi: the parent VSI
5617  * @reconfig_rss: indicates should the RSS be reconfigured or not
5618  *
5619  * This function validates number of queues in the context of new channel
5620  * which is being established and determines if RSS should be reconfigured
5621  * or not for parent VSI.
5622  **/
5623 static int i40e_validate_num_queues(struct i40e_pf *pf, int num_queues,
5624 				    struct i40e_vsi *vsi, bool *reconfig_rss)
5625 {
5626 	int max_ch_queues;
5627 
5628 	if (!reconfig_rss)
5629 		return -EINVAL;
5630 
5631 	*reconfig_rss = false;
5632 
5633 	if (num_queues > I40E_MAX_QUEUES_PER_CH) {
5634 		dev_err(&pf->pdev->dev,
5635 			"Failed to create VMDq VSI. User requested num_queues (%d) > I40E_MAX_QUEUES_PER_VSI (%u)\n",
5636 			num_queues, I40E_MAX_QUEUES_PER_CH);
5637 		return -EINVAL;
5638 	}
5639 
5640 	if (vsi->current_rss_size) {
5641 		if (num_queues > vsi->current_rss_size) {
5642 			dev_dbg(&pf->pdev->dev,
5643 				"Error: num_queues (%d) > vsi's current_size(%d)\n",
5644 				num_queues, vsi->current_rss_size);
5645 			return -EINVAL;
5646 		} else if ((num_queues < vsi->current_rss_size) &&
5647 			   (!is_power_of_2(num_queues))) {
5648 			dev_dbg(&pf->pdev->dev,
5649 				"Error: num_queues (%d) < vsi's current_size(%d), but not power of 2\n",
5650 				num_queues, vsi->current_rss_size);
5651 			return -EINVAL;
5652 		}
5653 	}
5654 
5655 	if (!is_power_of_2(num_queues)) {
5656 		/* Find the max num_queues configured for channel if channel
5657 		 * exist.
5658 		 * if channel exist, then enforce 'num_queues' to be more than
5659 		 * max ever queues configured for channel.
5660 		 */
5661 		max_ch_queues = i40e_get_max_queues_for_channel(vsi);
5662 		if (num_queues < max_ch_queues) {
5663 			dev_dbg(&pf->pdev->dev,
5664 				"Error: num_queues (%d) < max queues configured for channel(%d)\n",
5665 				num_queues, max_ch_queues);
5666 			return -EINVAL;
5667 		}
5668 		*reconfig_rss = true;
5669 	}
5670 
5671 	return 0;
5672 }
5673 
5674 /**
5675  * i40e_vsi_reconfig_rss - reconfig RSS based on specified rss_size
5676  * @vsi: the VSI being setup
5677  * @rss_size: size of RSS, accordingly LUT gets reprogrammed
5678  *
5679  * This function reconfigures RSS by reprogramming LUTs using 'rss_size'
5680  **/
5681 static int i40e_vsi_reconfig_rss(struct i40e_vsi *vsi, u16 rss_size)
5682 {
5683 	struct i40e_pf *pf = vsi->back;
5684 	u8 seed[I40E_HKEY_ARRAY_SIZE];
5685 	struct i40e_hw *hw = &pf->hw;
5686 	int local_rss_size;
5687 	u8 *lut;
5688 	int ret;
5689 
5690 	if (!vsi->rss_size)
5691 		return -EINVAL;
5692 
5693 	if (rss_size > vsi->rss_size)
5694 		return -EINVAL;
5695 
5696 	local_rss_size = min_t(int, vsi->rss_size, rss_size);
5697 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
5698 	if (!lut)
5699 		return -ENOMEM;
5700 
5701 	/* Ignoring user configured lut if there is one */
5702 	i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, local_rss_size);
5703 
5704 	/* Use user configured hash key if there is one, otherwise
5705 	 * use default.
5706 	 */
5707 	if (vsi->rss_hkey_user)
5708 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
5709 	else
5710 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
5711 
5712 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
5713 	if (ret) {
5714 		dev_info(&pf->pdev->dev,
5715 			 "Cannot set RSS lut, err %s aq_err %s\n",
5716 			 i40e_stat_str(hw, ret),
5717 			 i40e_aq_str(hw, hw->aq.asq_last_status));
5718 		kfree(lut);
5719 		return ret;
5720 	}
5721 	kfree(lut);
5722 
5723 	/* Do the update w.r.t. storing rss_size */
5724 	if (!vsi->orig_rss_size)
5725 		vsi->orig_rss_size = vsi->rss_size;
5726 	vsi->current_rss_size = local_rss_size;
5727 
5728 	return ret;
5729 }
5730 
5731 /**
5732  * i40e_channel_setup_queue_map - Setup a channel queue map
5733  * @pf: ptr to PF device
5734  * @vsi: the VSI being setup
5735  * @ctxt: VSI context structure
5736  * @ch: ptr to channel structure
5737  *
5738  * Setup queue map for a specific channel
5739  **/
5740 static void i40e_channel_setup_queue_map(struct i40e_pf *pf,
5741 					 struct i40e_vsi_context *ctxt,
5742 					 struct i40e_channel *ch)
5743 {
5744 	u16 qcount, qmap, sections = 0;
5745 	u8 offset = 0;
5746 	int pow;
5747 
5748 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
5749 	sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
5750 
5751 	qcount = min_t(int, ch->num_queue_pairs, pf->num_lan_msix);
5752 	ch->num_queue_pairs = qcount;
5753 
5754 	/* find the next higher power-of-2 of num queue pairs */
5755 	pow = ilog2(qcount);
5756 	if (!is_power_of_2(qcount))
5757 		pow++;
5758 
5759 	qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
5760 		(pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
5761 
5762 	/* Setup queue TC[0].qmap for given VSI context */
5763 	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
5764 
5765 	ctxt->info.up_enable_bits = 0x1; /* TC0 enabled */
5766 	ctxt->info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
5767 	ctxt->info.queue_mapping[0] = cpu_to_le16(ch->base_queue);
5768 	ctxt->info.valid_sections |= cpu_to_le16(sections);
5769 }
5770 
5771 /**
5772  * i40e_add_channel - add a channel by adding VSI
5773  * @pf: ptr to PF device
5774  * @uplink_seid: underlying HW switching element (VEB) ID
5775  * @ch: ptr to channel structure
5776  *
5777  * Add a channel (VSI) using add_vsi and queue_map
5778  **/
5779 static int i40e_add_channel(struct i40e_pf *pf, u16 uplink_seid,
5780 			    struct i40e_channel *ch)
5781 {
5782 	struct i40e_hw *hw = &pf->hw;
5783 	struct i40e_vsi_context ctxt;
5784 	u8 enabled_tc = 0x1; /* TC0 enabled */
5785 	int ret;
5786 
5787 	if (ch->type != I40E_VSI_VMDQ2) {
5788 		dev_info(&pf->pdev->dev,
5789 			 "add new vsi failed, ch->type %d\n", ch->type);
5790 		return -EINVAL;
5791 	}
5792 
5793 	memset(&ctxt, 0, sizeof(ctxt));
5794 	ctxt.pf_num = hw->pf_id;
5795 	ctxt.vf_num = 0;
5796 	ctxt.uplink_seid = uplink_seid;
5797 	ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
5798 	if (ch->type == I40E_VSI_VMDQ2)
5799 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
5800 
5801 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED) {
5802 		ctxt.info.valid_sections |=
5803 		     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5804 		ctxt.info.switch_id =
5805 		   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5806 	}
5807 
5808 	/* Set queue map for a given VSI context */
5809 	i40e_channel_setup_queue_map(pf, &ctxt, ch);
5810 
5811 	/* Now time to create VSI */
5812 	ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
5813 	if (ret) {
5814 		dev_info(&pf->pdev->dev,
5815 			 "add new vsi failed, err %s aq_err %s\n",
5816 			 i40e_stat_str(&pf->hw, ret),
5817 			 i40e_aq_str(&pf->hw,
5818 				     pf->hw.aq.asq_last_status));
5819 		return -ENOENT;
5820 	}
5821 
5822 	/* Success, update channel */
5823 	ch->enabled_tc = enabled_tc;
5824 	ch->seid = ctxt.seid;
5825 	ch->vsi_number = ctxt.vsi_number;
5826 	ch->stat_counter_idx = cpu_to_le16(ctxt.info.stat_counter_idx);
5827 
5828 	/* copy just the sections touched not the entire info
5829 	 * since not all sections are valid as returned by
5830 	 * update vsi params
5831 	 */
5832 	ch->info.mapping_flags = ctxt.info.mapping_flags;
5833 	memcpy(&ch->info.queue_mapping,
5834 	       &ctxt.info.queue_mapping, sizeof(ctxt.info.queue_mapping));
5835 	memcpy(&ch->info.tc_mapping, ctxt.info.tc_mapping,
5836 	       sizeof(ctxt.info.tc_mapping));
5837 
5838 	return 0;
5839 }
5840 
5841 static int i40e_channel_config_bw(struct i40e_vsi *vsi, struct i40e_channel *ch,
5842 				  u8 *bw_share)
5843 {
5844 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5845 	i40e_status ret;
5846 	int i;
5847 
5848 	bw_data.tc_valid_bits = ch->enabled_tc;
5849 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5850 		bw_data.tc_bw_credits[i] = bw_share[i];
5851 
5852 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, ch->seid,
5853 				       &bw_data, NULL);
5854 	if (ret) {
5855 		dev_info(&vsi->back->pdev->dev,
5856 			 "Config VSI BW allocation per TC failed, aq_err: %d for new_vsi->seid %u\n",
5857 			 vsi->back->hw.aq.asq_last_status, ch->seid);
5858 		return -EINVAL;
5859 	}
5860 
5861 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5862 		ch->info.qs_handle[i] = bw_data.qs_handles[i];
5863 
5864 	return 0;
5865 }
5866 
5867 /**
5868  * i40e_channel_config_tx_ring - config TX ring associated with new channel
5869  * @pf: ptr to PF device
5870  * @vsi: the VSI being setup
5871  * @ch: ptr to channel structure
5872  *
5873  * Configure TX rings associated with channel (VSI) since queues are being
5874  * from parent VSI.
5875  **/
5876 static int i40e_channel_config_tx_ring(struct i40e_pf *pf,
5877 				       struct i40e_vsi *vsi,
5878 				       struct i40e_channel *ch)
5879 {
5880 	i40e_status ret;
5881 	int i;
5882 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5883 
5884 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
5885 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5886 		if (ch->enabled_tc & BIT(i))
5887 			bw_share[i] = 1;
5888 	}
5889 
5890 	/* configure BW for new VSI */
5891 	ret = i40e_channel_config_bw(vsi, ch, bw_share);
5892 	if (ret) {
5893 		dev_info(&vsi->back->pdev->dev,
5894 			 "Failed configuring TC map %d for channel (seid %u)\n",
5895 			 ch->enabled_tc, ch->seid);
5896 		return ret;
5897 	}
5898 
5899 	for (i = 0; i < ch->num_queue_pairs; i++) {
5900 		struct i40e_ring *tx_ring, *rx_ring;
5901 		u16 pf_q;
5902 
5903 		pf_q = ch->base_queue + i;
5904 
5905 		/* Get to TX ring ptr of main VSI, for re-setup TX queue
5906 		 * context
5907 		 */
5908 		tx_ring = vsi->tx_rings[pf_q];
5909 		tx_ring->ch = ch;
5910 
5911 		/* Get the RX ring ptr */
5912 		rx_ring = vsi->rx_rings[pf_q];
5913 		rx_ring->ch = ch;
5914 	}
5915 
5916 	return 0;
5917 }
5918 
5919 /**
5920  * i40e_setup_hw_channel - setup new channel
5921  * @pf: ptr to PF device
5922  * @vsi: the VSI being setup
5923  * @ch: ptr to channel structure
5924  * @uplink_seid: underlying HW switching element (VEB) ID
5925  * @type: type of channel to be created (VMDq2/VF)
5926  *
5927  * Setup new channel (VSI) based on specified type (VMDq2/VF)
5928  * and configures TX rings accordingly
5929  **/
5930 static inline int i40e_setup_hw_channel(struct i40e_pf *pf,
5931 					struct i40e_vsi *vsi,
5932 					struct i40e_channel *ch,
5933 					u16 uplink_seid, u8 type)
5934 {
5935 	int ret;
5936 
5937 	ch->initialized = false;
5938 	ch->base_queue = vsi->next_base_queue;
5939 	ch->type = type;
5940 
5941 	/* Proceed with creation of channel (VMDq2) VSI */
5942 	ret = i40e_add_channel(pf, uplink_seid, ch);
5943 	if (ret) {
5944 		dev_info(&pf->pdev->dev,
5945 			 "failed to add_channel using uplink_seid %u\n",
5946 			 uplink_seid);
5947 		return ret;
5948 	}
5949 
5950 	/* Mark the successful creation of channel */
5951 	ch->initialized = true;
5952 
5953 	/* Reconfigure TX queues using QTX_CTL register */
5954 	ret = i40e_channel_config_tx_ring(pf, vsi, ch);
5955 	if (ret) {
5956 		dev_info(&pf->pdev->dev,
5957 			 "failed to configure TX rings for channel %u\n",
5958 			 ch->seid);
5959 		return ret;
5960 	}
5961 
5962 	/* update 'next_base_queue' */
5963 	vsi->next_base_queue = vsi->next_base_queue + ch->num_queue_pairs;
5964 	dev_dbg(&pf->pdev->dev,
5965 		"Added channel: vsi_seid %u, vsi_number %u, stat_counter_idx %u, num_queue_pairs %u, pf->next_base_queue %d\n",
5966 		ch->seid, ch->vsi_number, ch->stat_counter_idx,
5967 		ch->num_queue_pairs,
5968 		vsi->next_base_queue);
5969 	return ret;
5970 }
5971 
5972 /**
5973  * i40e_setup_channel - setup new channel using uplink element
5974  * @pf: ptr to PF device
5975  * @type: type of channel to be created (VMDq2/VF)
5976  * @uplink_seid: underlying HW switching element (VEB) ID
5977  * @ch: ptr to channel structure
5978  *
5979  * Setup new channel (VSI) based on specified type (VMDq2/VF)
5980  * and uplink switching element (uplink_seid)
5981  **/
5982 static bool i40e_setup_channel(struct i40e_pf *pf, struct i40e_vsi *vsi,
5983 			       struct i40e_channel *ch)
5984 {
5985 	u8 vsi_type;
5986 	u16 seid;
5987 	int ret;
5988 
5989 	if (vsi->type == I40E_VSI_MAIN) {
5990 		vsi_type = I40E_VSI_VMDQ2;
5991 	} else {
5992 		dev_err(&pf->pdev->dev, "unsupported parent vsi type(%d)\n",
5993 			vsi->type);
5994 		return false;
5995 	}
5996 
5997 	/* underlying switching element */
5998 	seid = pf->vsi[pf->lan_vsi]->uplink_seid;
5999 
6000 	/* create channel (VSI), configure TX rings */
6001 	ret = i40e_setup_hw_channel(pf, vsi, ch, seid, vsi_type);
6002 	if (ret) {
6003 		dev_err(&pf->pdev->dev, "failed to setup hw_channel\n");
6004 		return false;
6005 	}
6006 
6007 	return ch->initialized ? true : false;
6008 }
6009 
6010 /**
6011  * i40e_validate_and_set_switch_mode - sets up switch mode correctly
6012  * @vsi: ptr to VSI which has PF backing
6013  *
6014  * Sets up switch mode correctly if it needs to be changed and perform
6015  * what are allowed modes.
6016  **/
6017 static int i40e_validate_and_set_switch_mode(struct i40e_vsi *vsi)
6018 {
6019 	u8 mode;
6020 	struct i40e_pf *pf = vsi->back;
6021 	struct i40e_hw *hw = &pf->hw;
6022 	int ret;
6023 
6024 	ret = i40e_get_capabilities(pf, i40e_aqc_opc_list_dev_capabilities);
6025 	if (ret)
6026 		return -EINVAL;
6027 
6028 	if (hw->dev_caps.switch_mode) {
6029 		/* if switch mode is set, support mode2 (non-tunneled for
6030 		 * cloud filter) for now
6031 		 */
6032 		u32 switch_mode = hw->dev_caps.switch_mode &
6033 				  I40E_SWITCH_MODE_MASK;
6034 		if (switch_mode >= I40E_CLOUD_FILTER_MODE1) {
6035 			if (switch_mode == I40E_CLOUD_FILTER_MODE2)
6036 				return 0;
6037 			dev_err(&pf->pdev->dev,
6038 				"Invalid switch_mode (%d), only non-tunneled mode for cloud filter is supported\n",
6039 				hw->dev_caps.switch_mode);
6040 			return -EINVAL;
6041 		}
6042 	}
6043 
6044 	/* Set Bit 7 to be valid */
6045 	mode = I40E_AQ_SET_SWITCH_BIT7_VALID;
6046 
6047 	/* Set L4type to both TCP and UDP support */
6048 	mode |= I40E_AQ_SET_SWITCH_L4_TYPE_BOTH;
6049 
6050 	/* Set cloud filter mode */
6051 	mode |= I40E_AQ_SET_SWITCH_MODE_NON_TUNNEL;
6052 
6053 	/* Prep mode field for set_switch_config */
6054 	ret = i40e_aq_set_switch_config(hw, pf->last_sw_conf_flags,
6055 					pf->last_sw_conf_valid_flags,
6056 					mode, NULL);
6057 	if (ret && hw->aq.asq_last_status != I40E_AQ_RC_ESRCH)
6058 		dev_err(&pf->pdev->dev,
6059 			"couldn't set switch config bits, err %s aq_err %s\n",
6060 			i40e_stat_str(hw, ret),
6061 			i40e_aq_str(hw,
6062 				    hw->aq.asq_last_status));
6063 
6064 	return ret;
6065 }
6066 
6067 /**
6068  * i40e_create_queue_channel - function to create channel
6069  * @vsi: VSI to be configured
6070  * @ch: ptr to channel (it contains channel specific params)
6071  *
6072  * This function creates channel (VSI) using num_queues specified by user,
6073  * reconfigs RSS if needed.
6074  **/
6075 int i40e_create_queue_channel(struct i40e_vsi *vsi,
6076 			      struct i40e_channel *ch)
6077 {
6078 	struct i40e_pf *pf = vsi->back;
6079 	bool reconfig_rss;
6080 	int err;
6081 
6082 	if (!ch)
6083 		return -EINVAL;
6084 
6085 	if (!ch->num_queue_pairs) {
6086 		dev_err(&pf->pdev->dev, "Invalid num_queues requested: %d\n",
6087 			ch->num_queue_pairs);
6088 		return -EINVAL;
6089 	}
6090 
6091 	/* validate user requested num_queues for channel */
6092 	err = i40e_validate_num_queues(pf, ch->num_queue_pairs, vsi,
6093 				       &reconfig_rss);
6094 	if (err) {
6095 		dev_info(&pf->pdev->dev, "Failed to validate num_queues (%d)\n",
6096 			 ch->num_queue_pairs);
6097 		return -EINVAL;
6098 	}
6099 
6100 	/* By default we are in VEPA mode, if this is the first VF/VMDq
6101 	 * VSI to be added switch to VEB mode.
6102 	 */
6103 	if ((!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) ||
6104 	    (!i40e_is_any_channel(vsi))) {
6105 		if (!is_power_of_2(vsi->tc_config.tc_info[0].qcount)) {
6106 			dev_dbg(&pf->pdev->dev,
6107 				"Failed to create channel. Override queues (%u) not power of 2\n",
6108 				vsi->tc_config.tc_info[0].qcount);
6109 			return -EINVAL;
6110 		}
6111 
6112 		if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
6113 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
6114 
6115 			if (vsi->type == I40E_VSI_MAIN) {
6116 				if (pf->flags & I40E_FLAG_TC_MQPRIO)
6117 					i40e_do_reset(pf, I40E_PF_RESET_FLAG,
6118 						      true);
6119 				else
6120 					i40e_do_reset_safe(pf,
6121 							   I40E_PF_RESET_FLAG);
6122 			}
6123 		}
6124 		/* now onwards for main VSI, number of queues will be value
6125 		 * of TC0's queue count
6126 		 */
6127 	}
6128 
6129 	/* By this time, vsi->cnt_q_avail shall be set to non-zero and
6130 	 * it should be more than num_queues
6131 	 */
6132 	if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_queue_pairs) {
6133 		dev_dbg(&pf->pdev->dev,
6134 			"Error: cnt_q_avail (%u) less than num_queues %d\n",
6135 			vsi->cnt_q_avail, ch->num_queue_pairs);
6136 		return -EINVAL;
6137 	}
6138 
6139 	/* reconfig_rss only if vsi type is MAIN_VSI */
6140 	if (reconfig_rss && (vsi->type == I40E_VSI_MAIN)) {
6141 		err = i40e_vsi_reconfig_rss(vsi, ch->num_queue_pairs);
6142 		if (err) {
6143 			dev_info(&pf->pdev->dev,
6144 				 "Error: unable to reconfig rss for num_queues (%u)\n",
6145 				 ch->num_queue_pairs);
6146 			return -EINVAL;
6147 		}
6148 	}
6149 
6150 	if (!i40e_setup_channel(pf, vsi, ch)) {
6151 		dev_info(&pf->pdev->dev, "Failed to setup channel\n");
6152 		return -EINVAL;
6153 	}
6154 
6155 	dev_info(&pf->pdev->dev,
6156 		 "Setup channel (id:%u) utilizing num_queues %d\n",
6157 		 ch->seid, ch->num_queue_pairs);
6158 
6159 	/* configure VSI for BW limit */
6160 	if (ch->max_tx_rate) {
6161 		u64 credits = ch->max_tx_rate;
6162 
6163 		if (i40e_set_bw_limit(vsi, ch->seid, ch->max_tx_rate))
6164 			return -EINVAL;
6165 
6166 		do_div(credits, I40E_BW_CREDIT_DIVISOR);
6167 		dev_dbg(&pf->pdev->dev,
6168 			"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
6169 			ch->max_tx_rate,
6170 			credits,
6171 			ch->seid);
6172 	}
6173 
6174 	/* in case of VF, this will be main SRIOV VSI */
6175 	ch->parent_vsi = vsi;
6176 
6177 	/* and update main_vsi's count for queue_available to use */
6178 	vsi->cnt_q_avail -= ch->num_queue_pairs;
6179 
6180 	return 0;
6181 }
6182 
6183 /**
6184  * i40e_configure_queue_channels - Add queue channel for the given TCs
6185  * @vsi: VSI to be configured
6186  *
6187  * Configures queue channel mapping to the given TCs
6188  **/
6189 static int i40e_configure_queue_channels(struct i40e_vsi *vsi)
6190 {
6191 	struct i40e_channel *ch;
6192 	u64 max_rate = 0;
6193 	int ret = 0, i;
6194 
6195 	/* Create app vsi with the TCs. Main VSI with TC0 is already set up */
6196 	vsi->tc_seid_map[0] = vsi->seid;
6197 	for (i = 1; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6198 		if (vsi->tc_config.enabled_tc & BIT(i)) {
6199 			ch = kzalloc(sizeof(*ch), GFP_KERNEL);
6200 			if (!ch) {
6201 				ret = -ENOMEM;
6202 				goto err_free;
6203 			}
6204 
6205 			INIT_LIST_HEAD(&ch->list);
6206 			ch->num_queue_pairs =
6207 				vsi->tc_config.tc_info[i].qcount;
6208 			ch->base_queue =
6209 				vsi->tc_config.tc_info[i].qoffset;
6210 
6211 			/* Bandwidth limit through tc interface is in bytes/s,
6212 			 * change to Mbit/s
6213 			 */
6214 			max_rate = vsi->mqprio_qopt.max_rate[i];
6215 			do_div(max_rate, I40E_BW_MBPS_DIVISOR);
6216 			ch->max_tx_rate = max_rate;
6217 
6218 			list_add_tail(&ch->list, &vsi->ch_list);
6219 
6220 			ret = i40e_create_queue_channel(vsi, ch);
6221 			if (ret) {
6222 				dev_err(&vsi->back->pdev->dev,
6223 					"Failed creating queue channel with TC%d: queues %d\n",
6224 					i, ch->num_queue_pairs);
6225 				goto err_free;
6226 			}
6227 			vsi->tc_seid_map[i] = ch->seid;
6228 		}
6229 	}
6230 	return ret;
6231 
6232 err_free:
6233 	i40e_remove_queue_channels(vsi);
6234 	return ret;
6235 }
6236 
6237 /**
6238  * i40e_veb_config_tc - Configure TCs for given VEB
6239  * @veb: given VEB
6240  * @enabled_tc: TC bitmap
6241  *
6242  * Configures given TC bitmap for VEB (switching) element
6243  **/
6244 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
6245 {
6246 	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
6247 	struct i40e_pf *pf = veb->pf;
6248 	int ret = 0;
6249 	int i;
6250 
6251 	/* No TCs or already enabled TCs just return */
6252 	if (!enabled_tc || veb->enabled_tc == enabled_tc)
6253 		return ret;
6254 
6255 	bw_data.tc_valid_bits = enabled_tc;
6256 	/* bw_data.absolute_credits is not set (relative) */
6257 
6258 	/* Enable ETS TCs with equal BW Share for now */
6259 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6260 		if (enabled_tc & BIT(i))
6261 			bw_data.tc_bw_share_credits[i] = 1;
6262 	}
6263 
6264 	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
6265 						   &bw_data, NULL);
6266 	if (ret) {
6267 		dev_info(&pf->pdev->dev,
6268 			 "VEB bw config failed, err %s aq_err %s\n",
6269 			 i40e_stat_str(&pf->hw, ret),
6270 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6271 		goto out;
6272 	}
6273 
6274 	/* Update the BW information */
6275 	ret = i40e_veb_get_bw_info(veb);
6276 	if (ret) {
6277 		dev_info(&pf->pdev->dev,
6278 			 "Failed getting veb bw config, err %s aq_err %s\n",
6279 			 i40e_stat_str(&pf->hw, ret),
6280 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6281 	}
6282 
6283 out:
6284 	return ret;
6285 }
6286 
6287 #ifdef CONFIG_I40E_DCB
6288 /**
6289  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
6290  * @pf: PF struct
6291  *
6292  * Reconfigure VEB/VSIs on a given PF; it is assumed that
6293  * the caller would've quiesce all the VSIs before calling
6294  * this function
6295  **/
6296 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
6297 {
6298 	u8 tc_map = 0;
6299 	int ret;
6300 	u8 v;
6301 
6302 	/* Enable the TCs available on PF to all VEBs */
6303 	tc_map = i40e_pf_get_tc_map(pf);
6304 	for (v = 0; v < I40E_MAX_VEB; v++) {
6305 		if (!pf->veb[v])
6306 			continue;
6307 		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
6308 		if (ret) {
6309 			dev_info(&pf->pdev->dev,
6310 				 "Failed configuring TC for VEB seid=%d\n",
6311 				 pf->veb[v]->seid);
6312 			/* Will try to configure as many components */
6313 		}
6314 	}
6315 
6316 	/* Update each VSI */
6317 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6318 		if (!pf->vsi[v])
6319 			continue;
6320 
6321 		/* - Enable all TCs for the LAN VSI
6322 		 * - For all others keep them at TC0 for now
6323 		 */
6324 		if (v == pf->lan_vsi)
6325 			tc_map = i40e_pf_get_tc_map(pf);
6326 		else
6327 			tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
6328 
6329 		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
6330 		if (ret) {
6331 			dev_info(&pf->pdev->dev,
6332 				 "Failed configuring TC for VSI seid=%d\n",
6333 				 pf->vsi[v]->seid);
6334 			/* Will try to configure as many components */
6335 		} else {
6336 			/* Re-configure VSI vectors based on updated TC map */
6337 			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
6338 			if (pf->vsi[v]->netdev)
6339 				i40e_dcbnl_set_all(pf->vsi[v]);
6340 		}
6341 	}
6342 }
6343 
6344 /**
6345  * i40e_resume_port_tx - Resume port Tx
6346  * @pf: PF struct
6347  *
6348  * Resume a port's Tx and issue a PF reset in case of failure to
6349  * resume.
6350  **/
6351 static int i40e_resume_port_tx(struct i40e_pf *pf)
6352 {
6353 	struct i40e_hw *hw = &pf->hw;
6354 	int ret;
6355 
6356 	ret = i40e_aq_resume_port_tx(hw, NULL);
6357 	if (ret) {
6358 		dev_info(&pf->pdev->dev,
6359 			 "Resume Port Tx failed, err %s aq_err %s\n",
6360 			  i40e_stat_str(&pf->hw, ret),
6361 			  i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6362 		/* Schedule PF reset to recover */
6363 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6364 		i40e_service_event_schedule(pf);
6365 	}
6366 
6367 	return ret;
6368 }
6369 
6370 /**
6371  * i40e_init_pf_dcb - Initialize DCB configuration
6372  * @pf: PF being configured
6373  *
6374  * Query the current DCB configuration and cache it
6375  * in the hardware structure
6376  **/
6377 static int i40e_init_pf_dcb(struct i40e_pf *pf)
6378 {
6379 	struct i40e_hw *hw = &pf->hw;
6380 	int err = 0;
6381 
6382 	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
6383 	if (pf->hw_features & I40E_HW_NO_DCB_SUPPORT)
6384 		goto out;
6385 
6386 	/* Get the initial DCB configuration */
6387 	err = i40e_init_dcb(hw);
6388 	if (!err) {
6389 		/* Device/Function is not DCBX capable */
6390 		if ((!hw->func_caps.dcb) ||
6391 		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
6392 			dev_info(&pf->pdev->dev,
6393 				 "DCBX offload is not supported or is disabled for this PF.\n");
6394 		} else {
6395 			/* When status is not DISABLED then DCBX in FW */
6396 			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
6397 				       DCB_CAP_DCBX_VER_IEEE;
6398 
6399 			pf->flags |= I40E_FLAG_DCB_CAPABLE;
6400 			/* Enable DCB tagging only when more than one TC
6401 			 * or explicitly disable if only one TC
6402 			 */
6403 			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
6404 				pf->flags |= I40E_FLAG_DCB_ENABLED;
6405 			else
6406 				pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6407 			dev_dbg(&pf->pdev->dev,
6408 				"DCBX offload is supported for this PF.\n");
6409 		}
6410 	} else {
6411 		dev_info(&pf->pdev->dev,
6412 			 "Query for DCB configuration failed, err %s aq_err %s\n",
6413 			 i40e_stat_str(&pf->hw, err),
6414 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6415 	}
6416 
6417 out:
6418 	return err;
6419 }
6420 #endif /* CONFIG_I40E_DCB */
6421 #define SPEED_SIZE 14
6422 #define FC_SIZE 8
6423 /**
6424  * i40e_print_link_message - print link up or down
6425  * @vsi: the VSI for which link needs a message
6426  */
6427 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
6428 {
6429 	enum i40e_aq_link_speed new_speed;
6430 	struct i40e_pf *pf = vsi->back;
6431 	char *speed = "Unknown";
6432 	char *fc = "Unknown";
6433 	char *fec = "";
6434 	char *req_fec = "";
6435 	char *an = "";
6436 
6437 	new_speed = pf->hw.phy.link_info.link_speed;
6438 
6439 	if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
6440 		return;
6441 	vsi->current_isup = isup;
6442 	vsi->current_speed = new_speed;
6443 	if (!isup) {
6444 		netdev_info(vsi->netdev, "NIC Link is Down\n");
6445 		return;
6446 	}
6447 
6448 	/* Warn user if link speed on NPAR enabled partition is not at
6449 	 * least 10GB
6450 	 */
6451 	if (pf->hw.func_caps.npar_enable &&
6452 	    (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
6453 	     pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
6454 		netdev_warn(vsi->netdev,
6455 			    "The partition detected link speed that is less than 10Gbps\n");
6456 
6457 	switch (pf->hw.phy.link_info.link_speed) {
6458 	case I40E_LINK_SPEED_40GB:
6459 		speed = "40 G";
6460 		break;
6461 	case I40E_LINK_SPEED_20GB:
6462 		speed = "20 G";
6463 		break;
6464 	case I40E_LINK_SPEED_25GB:
6465 		speed = "25 G";
6466 		break;
6467 	case I40E_LINK_SPEED_10GB:
6468 		speed = "10 G";
6469 		break;
6470 	case I40E_LINK_SPEED_1GB:
6471 		speed = "1000 M";
6472 		break;
6473 	case I40E_LINK_SPEED_100MB:
6474 		speed = "100 M";
6475 		break;
6476 	default:
6477 		break;
6478 	}
6479 
6480 	switch (pf->hw.fc.current_mode) {
6481 	case I40E_FC_FULL:
6482 		fc = "RX/TX";
6483 		break;
6484 	case I40E_FC_TX_PAUSE:
6485 		fc = "TX";
6486 		break;
6487 	case I40E_FC_RX_PAUSE:
6488 		fc = "RX";
6489 		break;
6490 	default:
6491 		fc = "None";
6492 		break;
6493 	}
6494 
6495 	if (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
6496 		req_fec = ", Requested FEC: None";
6497 		fec = ", FEC: None";
6498 		an = ", Autoneg: False";
6499 
6500 		if (pf->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
6501 			an = ", Autoneg: True";
6502 
6503 		if (pf->hw.phy.link_info.fec_info &
6504 		    I40E_AQ_CONFIG_FEC_KR_ENA)
6505 			fec = ", FEC: CL74 FC-FEC/BASE-R";
6506 		else if (pf->hw.phy.link_info.fec_info &
6507 			 I40E_AQ_CONFIG_FEC_RS_ENA)
6508 			fec = ", FEC: CL108 RS-FEC";
6509 
6510 		/* 'CL108 RS-FEC' should be displayed when RS is requested, or
6511 		 * both RS and FC are requested
6512 		 */
6513 		if (vsi->back->hw.phy.link_info.req_fec_info &
6514 		    (I40E_AQ_REQUEST_FEC_KR | I40E_AQ_REQUEST_FEC_RS)) {
6515 			if (vsi->back->hw.phy.link_info.req_fec_info &
6516 			    I40E_AQ_REQUEST_FEC_RS)
6517 				req_fec = ", Requested FEC: CL108 RS-FEC";
6518 			else
6519 				req_fec = ", Requested FEC: CL74 FC-FEC/BASE-R";
6520 		}
6521 	}
6522 
6523 	netdev_info(vsi->netdev, "NIC Link is Up, %sbps Full Duplex%s%s%s, Flow Control: %s\n",
6524 		    speed, req_fec, fec, an, fc);
6525 }
6526 
6527 /**
6528  * i40e_up_complete - Finish the last steps of bringing up a connection
6529  * @vsi: the VSI being configured
6530  **/
6531 static int i40e_up_complete(struct i40e_vsi *vsi)
6532 {
6533 	struct i40e_pf *pf = vsi->back;
6534 	int err;
6535 
6536 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6537 		i40e_vsi_configure_msix(vsi);
6538 	else
6539 		i40e_configure_msi_and_legacy(vsi);
6540 
6541 	/* start rings */
6542 	err = i40e_vsi_start_rings(vsi);
6543 	if (err)
6544 		return err;
6545 
6546 	clear_bit(__I40E_VSI_DOWN, vsi->state);
6547 	i40e_napi_enable_all(vsi);
6548 	i40e_vsi_enable_irq(vsi);
6549 
6550 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
6551 	    (vsi->netdev)) {
6552 		i40e_print_link_message(vsi, true);
6553 		netif_tx_start_all_queues(vsi->netdev);
6554 		netif_carrier_on(vsi->netdev);
6555 	}
6556 
6557 	/* replay FDIR SB filters */
6558 	if (vsi->type == I40E_VSI_FDIR) {
6559 		/* reset fd counters */
6560 		pf->fd_add_err = 0;
6561 		pf->fd_atr_cnt = 0;
6562 		i40e_fdir_filter_restore(vsi);
6563 	}
6564 
6565 	/* On the next run of the service_task, notify any clients of the new
6566 	 * opened netdev
6567 	 */
6568 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
6569 	i40e_service_event_schedule(pf);
6570 
6571 	return 0;
6572 }
6573 
6574 /**
6575  * i40e_vsi_reinit_locked - Reset the VSI
6576  * @vsi: the VSI being configured
6577  *
6578  * Rebuild the ring structs after some configuration
6579  * has changed, e.g. MTU size.
6580  **/
6581 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
6582 {
6583 	struct i40e_pf *pf = vsi->back;
6584 
6585 	WARN_ON(in_interrupt());
6586 	while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state))
6587 		usleep_range(1000, 2000);
6588 	i40e_down(vsi);
6589 
6590 	i40e_up(vsi);
6591 	clear_bit(__I40E_CONFIG_BUSY, pf->state);
6592 }
6593 
6594 /**
6595  * i40e_up - Bring the connection back up after being down
6596  * @vsi: the VSI being configured
6597  **/
6598 int i40e_up(struct i40e_vsi *vsi)
6599 {
6600 	int err;
6601 
6602 	err = i40e_vsi_configure(vsi);
6603 	if (!err)
6604 		err = i40e_up_complete(vsi);
6605 
6606 	return err;
6607 }
6608 
6609 /**
6610  * i40e_down - Shutdown the connection processing
6611  * @vsi: the VSI being stopped
6612  **/
6613 void i40e_down(struct i40e_vsi *vsi)
6614 {
6615 	int i;
6616 
6617 	/* It is assumed that the caller of this function
6618 	 * sets the vsi->state __I40E_VSI_DOWN bit.
6619 	 */
6620 	if (vsi->netdev) {
6621 		netif_carrier_off(vsi->netdev);
6622 		netif_tx_disable(vsi->netdev);
6623 	}
6624 	i40e_vsi_disable_irq(vsi);
6625 	i40e_vsi_stop_rings(vsi);
6626 	i40e_napi_disable_all(vsi);
6627 
6628 	for (i = 0; i < vsi->num_queue_pairs; i++) {
6629 		i40e_clean_tx_ring(vsi->tx_rings[i]);
6630 		if (i40e_enabled_xdp_vsi(vsi))
6631 			i40e_clean_tx_ring(vsi->xdp_rings[i]);
6632 		i40e_clean_rx_ring(vsi->rx_rings[i]);
6633 	}
6634 
6635 }
6636 
6637 /**
6638  * i40e_validate_mqprio_qopt- validate queue mapping info
6639  * @vsi: the VSI being configured
6640  * @mqprio_qopt: queue parametrs
6641  **/
6642 static int i40e_validate_mqprio_qopt(struct i40e_vsi *vsi,
6643 				     struct tc_mqprio_qopt_offload *mqprio_qopt)
6644 {
6645 	u64 sum_max_rate = 0;
6646 	u64 max_rate = 0;
6647 	int i;
6648 
6649 	if (mqprio_qopt->qopt.offset[0] != 0 ||
6650 	    mqprio_qopt->qopt.num_tc < 1 ||
6651 	    mqprio_qopt->qopt.num_tc > I40E_MAX_TRAFFIC_CLASS)
6652 		return -EINVAL;
6653 	for (i = 0; ; i++) {
6654 		if (!mqprio_qopt->qopt.count[i])
6655 			return -EINVAL;
6656 		if (mqprio_qopt->min_rate[i]) {
6657 			dev_err(&vsi->back->pdev->dev,
6658 				"Invalid min tx rate (greater than 0) specified\n");
6659 			return -EINVAL;
6660 		}
6661 		max_rate = mqprio_qopt->max_rate[i];
6662 		do_div(max_rate, I40E_BW_MBPS_DIVISOR);
6663 		sum_max_rate += max_rate;
6664 
6665 		if (i >= mqprio_qopt->qopt.num_tc - 1)
6666 			break;
6667 		if (mqprio_qopt->qopt.offset[i + 1] !=
6668 		    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
6669 			return -EINVAL;
6670 	}
6671 	if (vsi->num_queue_pairs <
6672 	    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i])) {
6673 		return -EINVAL;
6674 	}
6675 	if (sum_max_rate > i40e_get_link_speed(vsi)) {
6676 		dev_err(&vsi->back->pdev->dev,
6677 			"Invalid max tx rate specified\n");
6678 		return -EINVAL;
6679 	}
6680 	return 0;
6681 }
6682 
6683 /**
6684  * i40e_vsi_set_default_tc_config - set default values for tc configuration
6685  * @vsi: the VSI being configured
6686  **/
6687 static void i40e_vsi_set_default_tc_config(struct i40e_vsi *vsi)
6688 {
6689 	u16 qcount;
6690 	int i;
6691 
6692 	/* Only TC0 is enabled */
6693 	vsi->tc_config.numtc = 1;
6694 	vsi->tc_config.enabled_tc = 1;
6695 	qcount = min_t(int, vsi->alloc_queue_pairs,
6696 		       i40e_pf_get_max_q_per_tc(vsi->back));
6697 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6698 		/* For the TC that is not enabled set the offset to to default
6699 		 * queue and allocate one queue for the given TC.
6700 		 */
6701 		vsi->tc_config.tc_info[i].qoffset = 0;
6702 		if (i == 0)
6703 			vsi->tc_config.tc_info[i].qcount = qcount;
6704 		else
6705 			vsi->tc_config.tc_info[i].qcount = 1;
6706 		vsi->tc_config.tc_info[i].netdev_tc = 0;
6707 	}
6708 }
6709 
6710 /**
6711  * i40e_setup_tc - configure multiple traffic classes
6712  * @netdev: net device to configure
6713  * @type_data: tc offload data
6714  **/
6715 static int i40e_setup_tc(struct net_device *netdev, void *type_data)
6716 {
6717 	struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
6718 	struct i40e_netdev_priv *np = netdev_priv(netdev);
6719 	struct i40e_vsi *vsi = np->vsi;
6720 	struct i40e_pf *pf = vsi->back;
6721 	u8 enabled_tc = 0, num_tc, hw;
6722 	bool need_reset = false;
6723 	int ret = -EINVAL;
6724 	u16 mode;
6725 	int i;
6726 
6727 	num_tc = mqprio_qopt->qopt.num_tc;
6728 	hw = mqprio_qopt->qopt.hw;
6729 	mode = mqprio_qopt->mode;
6730 	if (!hw) {
6731 		pf->flags &= ~I40E_FLAG_TC_MQPRIO;
6732 		memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
6733 		goto config_tc;
6734 	}
6735 
6736 	/* Check if MFP enabled */
6737 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
6738 		netdev_info(netdev,
6739 			    "Configuring TC not supported in MFP mode\n");
6740 		return ret;
6741 	}
6742 	switch (mode) {
6743 	case TC_MQPRIO_MODE_DCB:
6744 		pf->flags &= ~I40E_FLAG_TC_MQPRIO;
6745 
6746 		/* Check if DCB enabled to continue */
6747 		if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
6748 			netdev_info(netdev,
6749 				    "DCB is not enabled for adapter\n");
6750 			return ret;
6751 		}
6752 
6753 		/* Check whether tc count is within enabled limit */
6754 		if (num_tc > i40e_pf_get_num_tc(pf)) {
6755 			netdev_info(netdev,
6756 				    "TC count greater than enabled on link for adapter\n");
6757 			return ret;
6758 		}
6759 		break;
6760 	case TC_MQPRIO_MODE_CHANNEL:
6761 		if (pf->flags & I40E_FLAG_DCB_ENABLED) {
6762 			netdev_info(netdev,
6763 				    "Full offload of TC Mqprio options is not supported when DCB is enabled\n");
6764 			return ret;
6765 		}
6766 		if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
6767 			return ret;
6768 		ret = i40e_validate_mqprio_qopt(vsi, mqprio_qopt);
6769 		if (ret)
6770 			return ret;
6771 		memcpy(&vsi->mqprio_qopt, mqprio_qopt,
6772 		       sizeof(*mqprio_qopt));
6773 		pf->flags |= I40E_FLAG_TC_MQPRIO;
6774 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6775 		break;
6776 	default:
6777 		return -EINVAL;
6778 	}
6779 
6780 config_tc:
6781 	/* Generate TC map for number of tc requested */
6782 	for (i = 0; i < num_tc; i++)
6783 		enabled_tc |= BIT(i);
6784 
6785 	/* Requesting same TC configuration as already enabled */
6786 	if (enabled_tc == vsi->tc_config.enabled_tc &&
6787 	    mode != TC_MQPRIO_MODE_CHANNEL)
6788 		return 0;
6789 
6790 	/* Quiesce VSI queues */
6791 	i40e_quiesce_vsi(vsi);
6792 
6793 	if (!hw && !(pf->flags & I40E_FLAG_TC_MQPRIO))
6794 		i40e_remove_queue_channels(vsi);
6795 
6796 	/* Configure VSI for enabled TCs */
6797 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
6798 	if (ret) {
6799 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
6800 			    vsi->seid);
6801 		need_reset = true;
6802 		goto exit;
6803 	}
6804 
6805 	if (pf->flags & I40E_FLAG_TC_MQPRIO) {
6806 		if (vsi->mqprio_qopt.max_rate[0]) {
6807 			u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
6808 
6809 			do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
6810 			ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
6811 			if (!ret) {
6812 				u64 credits = max_tx_rate;
6813 
6814 				do_div(credits, I40E_BW_CREDIT_DIVISOR);
6815 				dev_dbg(&vsi->back->pdev->dev,
6816 					"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
6817 					max_tx_rate,
6818 					credits,
6819 					vsi->seid);
6820 			} else {
6821 				need_reset = true;
6822 				goto exit;
6823 			}
6824 		}
6825 		ret = i40e_configure_queue_channels(vsi);
6826 		if (ret) {
6827 			netdev_info(netdev,
6828 				    "Failed configuring queue channels\n");
6829 			need_reset = true;
6830 			goto exit;
6831 		}
6832 	}
6833 
6834 exit:
6835 	/* Reset the configuration data to defaults, only TC0 is enabled */
6836 	if (need_reset) {
6837 		i40e_vsi_set_default_tc_config(vsi);
6838 		need_reset = false;
6839 	}
6840 
6841 	/* Unquiesce VSI */
6842 	i40e_unquiesce_vsi(vsi);
6843 	return ret;
6844 }
6845 
6846 /**
6847  * i40e_set_cld_element - sets cloud filter element data
6848  * @filter: cloud filter rule
6849  * @cld: ptr to cloud filter element data
6850  *
6851  * This is helper function to copy data into cloud filter element
6852  **/
6853 static inline void
6854 i40e_set_cld_element(struct i40e_cloud_filter *filter,
6855 		     struct i40e_aqc_cloud_filters_element_data *cld)
6856 {
6857 	int i, j;
6858 	u32 ipa;
6859 
6860 	memset(cld, 0, sizeof(*cld));
6861 	ether_addr_copy(cld->outer_mac, filter->dst_mac);
6862 	ether_addr_copy(cld->inner_mac, filter->src_mac);
6863 
6864 	if (filter->n_proto != ETH_P_IP && filter->n_proto != ETH_P_IPV6)
6865 		return;
6866 
6867 	if (filter->n_proto == ETH_P_IPV6) {
6868 #define IPV6_MAX_INDEX	(ARRAY_SIZE(filter->dst_ipv6) - 1)
6869 		for (i = 0, j = 0; i < ARRAY_SIZE(filter->dst_ipv6);
6870 		     i++, j += 2) {
6871 			ipa = be32_to_cpu(filter->dst_ipv6[IPV6_MAX_INDEX - i]);
6872 			ipa = cpu_to_le32(ipa);
6873 			memcpy(&cld->ipaddr.raw_v6.data[j], &ipa, sizeof(ipa));
6874 		}
6875 	} else {
6876 		ipa = be32_to_cpu(filter->dst_ipv4);
6877 		memcpy(&cld->ipaddr.v4.data, &ipa, sizeof(ipa));
6878 	}
6879 
6880 	cld->inner_vlan = cpu_to_le16(ntohs(filter->vlan_id));
6881 
6882 	/* tenant_id is not supported by FW now, once the support is enabled
6883 	 * fill the cld->tenant_id with cpu_to_le32(filter->tenant_id)
6884 	 */
6885 	if (filter->tenant_id)
6886 		return;
6887 }
6888 
6889 /**
6890  * i40e_add_del_cloud_filter - Add/del cloud filter
6891  * @vsi: pointer to VSI
6892  * @filter: cloud filter rule
6893  * @add: if true, add, if false, delete
6894  *
6895  * Add or delete a cloud filter for a specific flow spec.
6896  * Returns 0 if the filter were successfully added.
6897  **/
6898 static int i40e_add_del_cloud_filter(struct i40e_vsi *vsi,
6899 				     struct i40e_cloud_filter *filter, bool add)
6900 {
6901 	struct i40e_aqc_cloud_filters_element_data cld_filter;
6902 	struct i40e_pf *pf = vsi->back;
6903 	int ret;
6904 	static const u16 flag_table[128] = {
6905 		[I40E_CLOUD_FILTER_FLAGS_OMAC]  =
6906 			I40E_AQC_ADD_CLOUD_FILTER_OMAC,
6907 		[I40E_CLOUD_FILTER_FLAGS_IMAC]  =
6908 			I40E_AQC_ADD_CLOUD_FILTER_IMAC,
6909 		[I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN]  =
6910 			I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN,
6911 		[I40E_CLOUD_FILTER_FLAGS_IMAC_TEN_ID] =
6912 			I40E_AQC_ADD_CLOUD_FILTER_IMAC_TEN_ID,
6913 		[I40E_CLOUD_FILTER_FLAGS_OMAC_TEN_ID_IMAC] =
6914 			I40E_AQC_ADD_CLOUD_FILTER_OMAC_TEN_ID_IMAC,
6915 		[I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN_TEN_ID] =
6916 			I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN_TEN_ID,
6917 		[I40E_CLOUD_FILTER_FLAGS_IIP] =
6918 			I40E_AQC_ADD_CLOUD_FILTER_IIP,
6919 	};
6920 
6921 	if (filter->flags >= ARRAY_SIZE(flag_table))
6922 		return I40E_ERR_CONFIG;
6923 
6924 	/* copy element needed to add cloud filter from filter */
6925 	i40e_set_cld_element(filter, &cld_filter);
6926 
6927 	if (filter->tunnel_type != I40E_CLOUD_TNL_TYPE_NONE)
6928 		cld_filter.flags = cpu_to_le16(filter->tunnel_type <<
6929 					     I40E_AQC_ADD_CLOUD_TNL_TYPE_SHIFT);
6930 
6931 	if (filter->n_proto == ETH_P_IPV6)
6932 		cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
6933 						I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
6934 	else
6935 		cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
6936 						I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
6937 
6938 	if (add)
6939 		ret = i40e_aq_add_cloud_filters(&pf->hw, filter->seid,
6940 						&cld_filter, 1);
6941 	else
6942 		ret = i40e_aq_rem_cloud_filters(&pf->hw, filter->seid,
6943 						&cld_filter, 1);
6944 	if (ret)
6945 		dev_dbg(&pf->pdev->dev,
6946 			"Failed to %s cloud filter using l4 port %u, err %d aq_err %d\n",
6947 			add ? "add" : "delete", filter->dst_port, ret,
6948 			pf->hw.aq.asq_last_status);
6949 	else
6950 		dev_info(&pf->pdev->dev,
6951 			 "%s cloud filter for VSI: %d\n",
6952 			 add ? "Added" : "Deleted", filter->seid);
6953 	return ret;
6954 }
6955 
6956 /**
6957  * i40e_add_del_cloud_filter_big_buf - Add/del cloud filter using big_buf
6958  * @vsi: pointer to VSI
6959  * @filter: cloud filter rule
6960  * @add: if true, add, if false, delete
6961  *
6962  * Add or delete a cloud filter for a specific flow spec using big buffer.
6963  * Returns 0 if the filter were successfully added.
6964  **/
6965 static int i40e_add_del_cloud_filter_big_buf(struct i40e_vsi *vsi,
6966 					     struct i40e_cloud_filter *filter,
6967 					     bool add)
6968 {
6969 	struct i40e_aqc_cloud_filters_element_bb cld_filter;
6970 	struct i40e_pf *pf = vsi->back;
6971 	int ret;
6972 
6973 	/* Both (src/dst) valid mac_addr are not supported */
6974 	if ((is_valid_ether_addr(filter->dst_mac) &&
6975 	     is_valid_ether_addr(filter->src_mac)) ||
6976 	    (is_multicast_ether_addr(filter->dst_mac) &&
6977 	     is_multicast_ether_addr(filter->src_mac)))
6978 		return -EINVAL;
6979 
6980 	/* Make sure port is specified, otherwise bail out, for channel
6981 	 * specific cloud filter needs 'L4 port' to be non-zero
6982 	 */
6983 	if (!filter->dst_port)
6984 		return -EINVAL;
6985 
6986 	/* adding filter using src_port/src_ip is not supported at this stage */
6987 	if (filter->src_port || filter->src_ipv4 ||
6988 	    !ipv6_addr_any(&filter->ip.v6.src_ip6))
6989 		return -EINVAL;
6990 
6991 	/* copy element needed to add cloud filter from filter */
6992 	i40e_set_cld_element(filter, &cld_filter.element);
6993 
6994 	if (is_valid_ether_addr(filter->dst_mac) ||
6995 	    is_valid_ether_addr(filter->src_mac) ||
6996 	    is_multicast_ether_addr(filter->dst_mac) ||
6997 	    is_multicast_ether_addr(filter->src_mac)) {
6998 		/* MAC + IP : unsupported mode */
6999 		if (filter->dst_ipv4)
7000 			return -EINVAL;
7001 
7002 		/* since we validated that L4 port must be valid before
7003 		 * we get here, start with respective "flags" value
7004 		 * and update if vlan is present or not
7005 		 */
7006 		cld_filter.element.flags =
7007 			cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_PORT);
7008 
7009 		if (filter->vlan_id) {
7010 			cld_filter.element.flags =
7011 			cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_VLAN_PORT);
7012 		}
7013 
7014 	} else if (filter->dst_ipv4 ||
7015 		   !ipv6_addr_any(&filter->ip.v6.dst_ip6)) {
7016 		cld_filter.element.flags =
7017 				cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_IP_PORT);
7018 		if (filter->n_proto == ETH_P_IPV6)
7019 			cld_filter.element.flags |=
7020 				cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
7021 		else
7022 			cld_filter.element.flags |=
7023 				cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
7024 	} else {
7025 		dev_err(&pf->pdev->dev,
7026 			"either mac or ip has to be valid for cloud filter\n");
7027 		return -EINVAL;
7028 	}
7029 
7030 	/* Now copy L4 port in Byte 6..7 in general fields */
7031 	cld_filter.general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X16_WORD0] =
7032 						be16_to_cpu(filter->dst_port);
7033 
7034 	if (add) {
7035 		/* Validate current device switch mode, change if necessary */
7036 		ret = i40e_validate_and_set_switch_mode(vsi);
7037 		if (ret) {
7038 			dev_err(&pf->pdev->dev,
7039 				"failed to set switch mode, ret %d\n",
7040 				ret);
7041 			return ret;
7042 		}
7043 
7044 		ret = i40e_aq_add_cloud_filters_bb(&pf->hw, filter->seid,
7045 						   &cld_filter, 1);
7046 	} else {
7047 		ret = i40e_aq_rem_cloud_filters_bb(&pf->hw, filter->seid,
7048 						   &cld_filter, 1);
7049 	}
7050 
7051 	if (ret)
7052 		dev_dbg(&pf->pdev->dev,
7053 			"Failed to %s cloud filter(big buffer) err %d aq_err %d\n",
7054 			add ? "add" : "delete", ret, pf->hw.aq.asq_last_status);
7055 	else
7056 		dev_info(&pf->pdev->dev,
7057 			 "%s cloud filter for VSI: %d, L4 port: %d\n",
7058 			 add ? "add" : "delete", filter->seid,
7059 			 ntohs(filter->dst_port));
7060 	return ret;
7061 }
7062 
7063 /**
7064  * i40e_parse_cls_flower - Parse tc flower filters provided by kernel
7065  * @vsi: Pointer to VSI
7066  * @cls_flower: Pointer to struct tc_cls_flower_offload
7067  * @filter: Pointer to cloud filter structure
7068  *
7069  **/
7070 static int i40e_parse_cls_flower(struct i40e_vsi *vsi,
7071 				 struct tc_cls_flower_offload *f,
7072 				 struct i40e_cloud_filter *filter)
7073 {
7074 	u16 n_proto_mask = 0, n_proto_key = 0, addr_type = 0;
7075 	struct i40e_pf *pf = vsi->back;
7076 	u8 field_flags = 0;
7077 
7078 	if (f->dissector->used_keys &
7079 	    ~(BIT(FLOW_DISSECTOR_KEY_CONTROL) |
7080 	      BIT(FLOW_DISSECTOR_KEY_BASIC) |
7081 	      BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
7082 	      BIT(FLOW_DISSECTOR_KEY_VLAN) |
7083 	      BIT(FLOW_DISSECTOR_KEY_IPV4_ADDRS) |
7084 	      BIT(FLOW_DISSECTOR_KEY_IPV6_ADDRS) |
7085 	      BIT(FLOW_DISSECTOR_KEY_PORTS) |
7086 	      BIT(FLOW_DISSECTOR_KEY_ENC_KEYID))) {
7087 		dev_err(&pf->pdev->dev, "Unsupported key used: 0x%x\n",
7088 			f->dissector->used_keys);
7089 		return -EOPNOTSUPP;
7090 	}
7091 
7092 	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
7093 		struct flow_dissector_key_keyid *key =
7094 			skb_flow_dissector_target(f->dissector,
7095 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
7096 						  f->key);
7097 
7098 		struct flow_dissector_key_keyid *mask =
7099 			skb_flow_dissector_target(f->dissector,
7100 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
7101 						  f->mask);
7102 
7103 		if (mask->keyid != 0)
7104 			field_flags |= I40E_CLOUD_FIELD_TEN_ID;
7105 
7106 		filter->tenant_id = be32_to_cpu(key->keyid);
7107 	}
7108 
7109 	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_BASIC)) {
7110 		struct flow_dissector_key_basic *key =
7111 			skb_flow_dissector_target(f->dissector,
7112 						  FLOW_DISSECTOR_KEY_BASIC,
7113 						  f->key);
7114 
7115 		struct flow_dissector_key_basic *mask =
7116 			skb_flow_dissector_target(f->dissector,
7117 						  FLOW_DISSECTOR_KEY_BASIC,
7118 						  f->mask);
7119 
7120 		n_proto_key = ntohs(key->n_proto);
7121 		n_proto_mask = ntohs(mask->n_proto);
7122 
7123 		if (n_proto_key == ETH_P_ALL) {
7124 			n_proto_key = 0;
7125 			n_proto_mask = 0;
7126 		}
7127 		filter->n_proto = n_proto_key & n_proto_mask;
7128 		filter->ip_proto = key->ip_proto;
7129 	}
7130 
7131 	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
7132 		struct flow_dissector_key_eth_addrs *key =
7133 			skb_flow_dissector_target(f->dissector,
7134 						  FLOW_DISSECTOR_KEY_ETH_ADDRS,
7135 						  f->key);
7136 
7137 		struct flow_dissector_key_eth_addrs *mask =
7138 			skb_flow_dissector_target(f->dissector,
7139 						  FLOW_DISSECTOR_KEY_ETH_ADDRS,
7140 						  f->mask);
7141 
7142 		/* use is_broadcast and is_zero to check for all 0xf or 0 */
7143 		if (!is_zero_ether_addr(mask->dst)) {
7144 			if (is_broadcast_ether_addr(mask->dst)) {
7145 				field_flags |= I40E_CLOUD_FIELD_OMAC;
7146 			} else {
7147 				dev_err(&pf->pdev->dev, "Bad ether dest mask %pM\n",
7148 					mask->dst);
7149 				return I40E_ERR_CONFIG;
7150 			}
7151 		}
7152 
7153 		if (!is_zero_ether_addr(mask->src)) {
7154 			if (is_broadcast_ether_addr(mask->src)) {
7155 				field_flags |= I40E_CLOUD_FIELD_IMAC;
7156 			} else {
7157 				dev_err(&pf->pdev->dev, "Bad ether src mask %pM\n",
7158 					mask->src);
7159 				return I40E_ERR_CONFIG;
7160 			}
7161 		}
7162 		ether_addr_copy(filter->dst_mac, key->dst);
7163 		ether_addr_copy(filter->src_mac, key->src);
7164 	}
7165 
7166 	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_VLAN)) {
7167 		struct flow_dissector_key_vlan *key =
7168 			skb_flow_dissector_target(f->dissector,
7169 						  FLOW_DISSECTOR_KEY_VLAN,
7170 						  f->key);
7171 		struct flow_dissector_key_vlan *mask =
7172 			skb_flow_dissector_target(f->dissector,
7173 						  FLOW_DISSECTOR_KEY_VLAN,
7174 						  f->mask);
7175 
7176 		if (mask->vlan_id) {
7177 			if (mask->vlan_id == VLAN_VID_MASK) {
7178 				field_flags |= I40E_CLOUD_FIELD_IVLAN;
7179 
7180 			} else {
7181 				dev_err(&pf->pdev->dev, "Bad vlan mask 0x%04x\n",
7182 					mask->vlan_id);
7183 				return I40E_ERR_CONFIG;
7184 			}
7185 		}
7186 
7187 		filter->vlan_id = cpu_to_be16(key->vlan_id);
7188 	}
7189 
7190 	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_CONTROL)) {
7191 		struct flow_dissector_key_control *key =
7192 			skb_flow_dissector_target(f->dissector,
7193 						  FLOW_DISSECTOR_KEY_CONTROL,
7194 						  f->key);
7195 
7196 		addr_type = key->addr_type;
7197 	}
7198 
7199 	if (addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
7200 		struct flow_dissector_key_ipv4_addrs *key =
7201 			skb_flow_dissector_target(f->dissector,
7202 						  FLOW_DISSECTOR_KEY_IPV4_ADDRS,
7203 						  f->key);
7204 		struct flow_dissector_key_ipv4_addrs *mask =
7205 			skb_flow_dissector_target(f->dissector,
7206 						  FLOW_DISSECTOR_KEY_IPV4_ADDRS,
7207 						  f->mask);
7208 
7209 		if (mask->dst) {
7210 			if (mask->dst == cpu_to_be32(0xffffffff)) {
7211 				field_flags |= I40E_CLOUD_FIELD_IIP;
7212 			} else {
7213 				mask->dst = be32_to_cpu(mask->dst);
7214 				dev_err(&pf->pdev->dev, "Bad ip dst mask %pI4\n",
7215 					&mask->dst);
7216 				return I40E_ERR_CONFIG;
7217 			}
7218 		}
7219 
7220 		if (mask->src) {
7221 			if (mask->src == cpu_to_be32(0xffffffff)) {
7222 				field_flags |= I40E_CLOUD_FIELD_IIP;
7223 			} else {
7224 				mask->src = be32_to_cpu(mask->src);
7225 				dev_err(&pf->pdev->dev, "Bad ip src mask %pI4\n",
7226 					&mask->src);
7227 				return I40E_ERR_CONFIG;
7228 			}
7229 		}
7230 
7231 		if (field_flags & I40E_CLOUD_FIELD_TEN_ID) {
7232 			dev_err(&pf->pdev->dev, "Tenant id not allowed for ip filter\n");
7233 			return I40E_ERR_CONFIG;
7234 		}
7235 		filter->dst_ipv4 = key->dst;
7236 		filter->src_ipv4 = key->src;
7237 	}
7238 
7239 	if (addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
7240 		struct flow_dissector_key_ipv6_addrs *key =
7241 			skb_flow_dissector_target(f->dissector,
7242 						  FLOW_DISSECTOR_KEY_IPV6_ADDRS,
7243 						  f->key);
7244 		struct flow_dissector_key_ipv6_addrs *mask =
7245 			skb_flow_dissector_target(f->dissector,
7246 						  FLOW_DISSECTOR_KEY_IPV6_ADDRS,
7247 						  f->mask);
7248 
7249 		/* src and dest IPV6 address should not be LOOPBACK
7250 		 * (0:0:0:0:0:0:0:1), which can be represented as ::1
7251 		 */
7252 		if (ipv6_addr_loopback(&key->dst) ||
7253 		    ipv6_addr_loopback(&key->src)) {
7254 			dev_err(&pf->pdev->dev,
7255 				"Bad ipv6, addr is LOOPBACK\n");
7256 			return I40E_ERR_CONFIG;
7257 		}
7258 		if (!ipv6_addr_any(&mask->dst) || !ipv6_addr_any(&mask->src))
7259 			field_flags |= I40E_CLOUD_FIELD_IIP;
7260 
7261 		memcpy(&filter->src_ipv6, &key->src.s6_addr32,
7262 		       sizeof(filter->src_ipv6));
7263 		memcpy(&filter->dst_ipv6, &key->dst.s6_addr32,
7264 		       sizeof(filter->dst_ipv6));
7265 	}
7266 
7267 	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_PORTS)) {
7268 		struct flow_dissector_key_ports *key =
7269 			skb_flow_dissector_target(f->dissector,
7270 						  FLOW_DISSECTOR_KEY_PORTS,
7271 						  f->key);
7272 		struct flow_dissector_key_ports *mask =
7273 			skb_flow_dissector_target(f->dissector,
7274 						  FLOW_DISSECTOR_KEY_PORTS,
7275 						  f->mask);
7276 
7277 		if (mask->src) {
7278 			if (mask->src == cpu_to_be16(0xffff)) {
7279 				field_flags |= I40E_CLOUD_FIELD_IIP;
7280 			} else {
7281 				dev_err(&pf->pdev->dev, "Bad src port mask 0x%04x\n",
7282 					be16_to_cpu(mask->src));
7283 				return I40E_ERR_CONFIG;
7284 			}
7285 		}
7286 
7287 		if (mask->dst) {
7288 			if (mask->dst == cpu_to_be16(0xffff)) {
7289 				field_flags |= I40E_CLOUD_FIELD_IIP;
7290 			} else {
7291 				dev_err(&pf->pdev->dev, "Bad dst port mask 0x%04x\n",
7292 					be16_to_cpu(mask->dst));
7293 				return I40E_ERR_CONFIG;
7294 			}
7295 		}
7296 
7297 		filter->dst_port = key->dst;
7298 		filter->src_port = key->src;
7299 
7300 		switch (filter->ip_proto) {
7301 		case IPPROTO_TCP:
7302 		case IPPROTO_UDP:
7303 			break;
7304 		default:
7305 			dev_err(&pf->pdev->dev,
7306 				"Only UDP and TCP transport are supported\n");
7307 			return -EINVAL;
7308 		}
7309 	}
7310 	filter->flags = field_flags;
7311 	return 0;
7312 }
7313 
7314 /**
7315  * i40e_handle_tclass: Forward to a traffic class on the device
7316  * @vsi: Pointer to VSI
7317  * @tc: traffic class index on the device
7318  * @filter: Pointer to cloud filter structure
7319  *
7320  **/
7321 static int i40e_handle_tclass(struct i40e_vsi *vsi, u32 tc,
7322 			      struct i40e_cloud_filter *filter)
7323 {
7324 	struct i40e_channel *ch, *ch_tmp;
7325 
7326 	/* direct to a traffic class on the same device */
7327 	if (tc == 0) {
7328 		filter->seid = vsi->seid;
7329 		return 0;
7330 	} else if (vsi->tc_config.enabled_tc & BIT(tc)) {
7331 		if (!filter->dst_port) {
7332 			dev_err(&vsi->back->pdev->dev,
7333 				"Specify destination port to direct to traffic class that is not default\n");
7334 			return -EINVAL;
7335 		}
7336 		if (list_empty(&vsi->ch_list))
7337 			return -EINVAL;
7338 		list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list,
7339 					 list) {
7340 			if (ch->seid == vsi->tc_seid_map[tc])
7341 				filter->seid = ch->seid;
7342 		}
7343 		return 0;
7344 	}
7345 	dev_err(&vsi->back->pdev->dev, "TC is not enabled\n");
7346 	return -EINVAL;
7347 }
7348 
7349 /**
7350  * i40e_configure_clsflower - Configure tc flower filters
7351  * @vsi: Pointer to VSI
7352  * @cls_flower: Pointer to struct tc_cls_flower_offload
7353  *
7354  **/
7355 static int i40e_configure_clsflower(struct i40e_vsi *vsi,
7356 				    struct tc_cls_flower_offload *cls_flower)
7357 {
7358 	int tc = tc_classid_to_hwtc(vsi->netdev, cls_flower->classid);
7359 	struct i40e_cloud_filter *filter = NULL;
7360 	struct i40e_pf *pf = vsi->back;
7361 	int err = 0;
7362 
7363 	if (tc < 0) {
7364 		dev_err(&vsi->back->pdev->dev, "Invalid traffic class\n");
7365 		return -EINVAL;
7366 	}
7367 
7368 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state) ||
7369 	    test_bit(__I40E_RESET_INTR_RECEIVED, pf->state))
7370 		return -EBUSY;
7371 
7372 	if (pf->fdir_pf_active_filters ||
7373 	    (!hlist_empty(&pf->fdir_filter_list))) {
7374 		dev_err(&vsi->back->pdev->dev,
7375 			"Flow Director Sideband filters exists, turn ntuple off to configure cloud filters\n");
7376 		return -EINVAL;
7377 	}
7378 
7379 	if (vsi->back->flags & I40E_FLAG_FD_SB_ENABLED) {
7380 		dev_err(&vsi->back->pdev->dev,
7381 			"Disable Flow Director Sideband, configuring Cloud filters via tc-flower\n");
7382 		vsi->back->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7383 		vsi->back->flags |= I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
7384 	}
7385 
7386 	filter = kzalloc(sizeof(*filter), GFP_KERNEL);
7387 	if (!filter)
7388 		return -ENOMEM;
7389 
7390 	filter->cookie = cls_flower->cookie;
7391 
7392 	err = i40e_parse_cls_flower(vsi, cls_flower, filter);
7393 	if (err < 0)
7394 		goto err;
7395 
7396 	err = i40e_handle_tclass(vsi, tc, filter);
7397 	if (err < 0)
7398 		goto err;
7399 
7400 	/* Add cloud filter */
7401 	if (filter->dst_port)
7402 		err = i40e_add_del_cloud_filter_big_buf(vsi, filter, true);
7403 	else
7404 		err = i40e_add_del_cloud_filter(vsi, filter, true);
7405 
7406 	if (err) {
7407 		dev_err(&pf->pdev->dev,
7408 			"Failed to add cloud filter, err %s\n",
7409 			i40e_stat_str(&pf->hw, err));
7410 		err = i40e_aq_rc_to_posix(err, pf->hw.aq.asq_last_status);
7411 		goto err;
7412 	}
7413 
7414 	/* add filter to the ordered list */
7415 	INIT_HLIST_NODE(&filter->cloud_node);
7416 
7417 	hlist_add_head(&filter->cloud_node, &pf->cloud_filter_list);
7418 
7419 	pf->num_cloud_filters++;
7420 
7421 	return err;
7422 err:
7423 	kfree(filter);
7424 	return err;
7425 }
7426 
7427 /**
7428  * i40e_find_cloud_filter - Find the could filter in the list
7429  * @vsi: Pointer to VSI
7430  * @cookie: filter specific cookie
7431  *
7432  **/
7433 static struct i40e_cloud_filter *i40e_find_cloud_filter(struct i40e_vsi *vsi,
7434 							unsigned long *cookie)
7435 {
7436 	struct i40e_cloud_filter *filter = NULL;
7437 	struct hlist_node *node2;
7438 
7439 	hlist_for_each_entry_safe(filter, node2,
7440 				  &vsi->back->cloud_filter_list, cloud_node)
7441 		if (!memcmp(cookie, &filter->cookie, sizeof(filter->cookie)))
7442 			return filter;
7443 	return NULL;
7444 }
7445 
7446 /**
7447  * i40e_delete_clsflower - Remove tc flower filters
7448  * @vsi: Pointer to VSI
7449  * @cls_flower: Pointer to struct tc_cls_flower_offload
7450  *
7451  **/
7452 static int i40e_delete_clsflower(struct i40e_vsi *vsi,
7453 				 struct tc_cls_flower_offload *cls_flower)
7454 {
7455 	struct i40e_cloud_filter *filter = NULL;
7456 	struct i40e_pf *pf = vsi->back;
7457 	int err = 0;
7458 
7459 	filter = i40e_find_cloud_filter(vsi, &cls_flower->cookie);
7460 
7461 	if (!filter)
7462 		return -EINVAL;
7463 
7464 	hash_del(&filter->cloud_node);
7465 
7466 	if (filter->dst_port)
7467 		err = i40e_add_del_cloud_filter_big_buf(vsi, filter, false);
7468 	else
7469 		err = i40e_add_del_cloud_filter(vsi, filter, false);
7470 
7471 	kfree(filter);
7472 	if (err) {
7473 		dev_err(&pf->pdev->dev,
7474 			"Failed to delete cloud filter, err %s\n",
7475 			i40e_stat_str(&pf->hw, err));
7476 		return i40e_aq_rc_to_posix(err, pf->hw.aq.asq_last_status);
7477 	}
7478 
7479 	pf->num_cloud_filters--;
7480 	if (!pf->num_cloud_filters)
7481 		if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
7482 		    !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
7483 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
7484 			pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
7485 			pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
7486 		}
7487 	return 0;
7488 }
7489 
7490 /**
7491  * i40e_setup_tc_cls_flower - flower classifier offloads
7492  * @netdev: net device to configure
7493  * @type_data: offload data
7494  **/
7495 static int i40e_setup_tc_cls_flower(struct i40e_netdev_priv *np,
7496 				    struct tc_cls_flower_offload *cls_flower)
7497 {
7498 	struct i40e_vsi *vsi = np->vsi;
7499 
7500 	if (cls_flower->common.chain_index)
7501 		return -EOPNOTSUPP;
7502 
7503 	switch (cls_flower->command) {
7504 	case TC_CLSFLOWER_REPLACE:
7505 		return i40e_configure_clsflower(vsi, cls_flower);
7506 	case TC_CLSFLOWER_DESTROY:
7507 		return i40e_delete_clsflower(vsi, cls_flower);
7508 	case TC_CLSFLOWER_STATS:
7509 		return -EOPNOTSUPP;
7510 	default:
7511 		return -EINVAL;
7512 	}
7513 }
7514 
7515 static int i40e_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
7516 				  void *cb_priv)
7517 {
7518 	struct i40e_netdev_priv *np = cb_priv;
7519 
7520 	switch (type) {
7521 	case TC_SETUP_CLSFLOWER:
7522 		return i40e_setup_tc_cls_flower(np, type_data);
7523 
7524 	default:
7525 		return -EOPNOTSUPP;
7526 	}
7527 }
7528 
7529 static int i40e_setup_tc_block(struct net_device *dev,
7530 			       struct tc_block_offload *f)
7531 {
7532 	struct i40e_netdev_priv *np = netdev_priv(dev);
7533 
7534 	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
7535 		return -EOPNOTSUPP;
7536 
7537 	switch (f->command) {
7538 	case TC_BLOCK_BIND:
7539 		return tcf_block_cb_register(f->block, i40e_setup_tc_block_cb,
7540 					     np, np);
7541 	case TC_BLOCK_UNBIND:
7542 		tcf_block_cb_unregister(f->block, i40e_setup_tc_block_cb, np);
7543 		return 0;
7544 	default:
7545 		return -EOPNOTSUPP;
7546 	}
7547 }
7548 
7549 static int __i40e_setup_tc(struct net_device *netdev, enum tc_setup_type type,
7550 			   void *type_data)
7551 {
7552 	switch (type) {
7553 	case TC_SETUP_QDISC_MQPRIO:
7554 		return i40e_setup_tc(netdev, type_data);
7555 	case TC_SETUP_BLOCK:
7556 		return i40e_setup_tc_block(netdev, type_data);
7557 	default:
7558 		return -EOPNOTSUPP;
7559 	}
7560 }
7561 
7562 /**
7563  * i40e_open - Called when a network interface is made active
7564  * @netdev: network interface device structure
7565  *
7566  * The open entry point is called when a network interface is made
7567  * active by the system (IFF_UP).  At this point all resources needed
7568  * for transmit and receive operations are allocated, the interrupt
7569  * handler is registered with the OS, the netdev watchdog subtask is
7570  * enabled, and the stack is notified that the interface is ready.
7571  *
7572  * Returns 0 on success, negative value on failure
7573  **/
7574 int i40e_open(struct net_device *netdev)
7575 {
7576 	struct i40e_netdev_priv *np = netdev_priv(netdev);
7577 	struct i40e_vsi *vsi = np->vsi;
7578 	struct i40e_pf *pf = vsi->back;
7579 	int err;
7580 
7581 	/* disallow open during test or if eeprom is broken */
7582 	if (test_bit(__I40E_TESTING, pf->state) ||
7583 	    test_bit(__I40E_BAD_EEPROM, pf->state))
7584 		return -EBUSY;
7585 
7586 	netif_carrier_off(netdev);
7587 
7588 	err = i40e_vsi_open(vsi);
7589 	if (err)
7590 		return err;
7591 
7592 	/* configure global TSO hardware offload settings */
7593 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
7594 						       TCP_FLAG_FIN) >> 16);
7595 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
7596 						       TCP_FLAG_FIN |
7597 						       TCP_FLAG_CWR) >> 16);
7598 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
7599 
7600 	udp_tunnel_get_rx_info(netdev);
7601 
7602 	return 0;
7603 }
7604 
7605 /**
7606  * i40e_vsi_open -
7607  * @vsi: the VSI to open
7608  *
7609  * Finish initialization of the VSI.
7610  *
7611  * Returns 0 on success, negative value on failure
7612  *
7613  * Note: expects to be called while under rtnl_lock()
7614  **/
7615 int i40e_vsi_open(struct i40e_vsi *vsi)
7616 {
7617 	struct i40e_pf *pf = vsi->back;
7618 	char int_name[I40E_INT_NAME_STR_LEN];
7619 	int err;
7620 
7621 	/* allocate descriptors */
7622 	err = i40e_vsi_setup_tx_resources(vsi);
7623 	if (err)
7624 		goto err_setup_tx;
7625 	err = i40e_vsi_setup_rx_resources(vsi);
7626 	if (err)
7627 		goto err_setup_rx;
7628 
7629 	err = i40e_vsi_configure(vsi);
7630 	if (err)
7631 		goto err_setup_rx;
7632 
7633 	if (vsi->netdev) {
7634 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
7635 			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
7636 		err = i40e_vsi_request_irq(vsi, int_name);
7637 		if (err)
7638 			goto err_setup_rx;
7639 
7640 		/* Notify the stack of the actual queue counts. */
7641 		err = netif_set_real_num_tx_queues(vsi->netdev,
7642 						   vsi->num_queue_pairs);
7643 		if (err)
7644 			goto err_set_queues;
7645 
7646 		err = netif_set_real_num_rx_queues(vsi->netdev,
7647 						   vsi->num_queue_pairs);
7648 		if (err)
7649 			goto err_set_queues;
7650 
7651 	} else if (vsi->type == I40E_VSI_FDIR) {
7652 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
7653 			 dev_driver_string(&pf->pdev->dev),
7654 			 dev_name(&pf->pdev->dev));
7655 		err = i40e_vsi_request_irq(vsi, int_name);
7656 
7657 	} else {
7658 		err = -EINVAL;
7659 		goto err_setup_rx;
7660 	}
7661 
7662 	err = i40e_up_complete(vsi);
7663 	if (err)
7664 		goto err_up_complete;
7665 
7666 	return 0;
7667 
7668 err_up_complete:
7669 	i40e_down(vsi);
7670 err_set_queues:
7671 	i40e_vsi_free_irq(vsi);
7672 err_setup_rx:
7673 	i40e_vsi_free_rx_resources(vsi);
7674 err_setup_tx:
7675 	i40e_vsi_free_tx_resources(vsi);
7676 	if (vsi == pf->vsi[pf->lan_vsi])
7677 		i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
7678 
7679 	return err;
7680 }
7681 
7682 /**
7683  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
7684  * @pf: Pointer to PF
7685  *
7686  * This function destroys the hlist where all the Flow Director
7687  * filters were saved.
7688  **/
7689 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
7690 {
7691 	struct i40e_fdir_filter *filter;
7692 	struct i40e_flex_pit *pit_entry, *tmp;
7693 	struct hlist_node *node2;
7694 
7695 	hlist_for_each_entry_safe(filter, node2,
7696 				  &pf->fdir_filter_list, fdir_node) {
7697 		hlist_del(&filter->fdir_node);
7698 		kfree(filter);
7699 	}
7700 
7701 	list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
7702 		list_del(&pit_entry->list);
7703 		kfree(pit_entry);
7704 	}
7705 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
7706 
7707 	list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
7708 		list_del(&pit_entry->list);
7709 		kfree(pit_entry);
7710 	}
7711 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
7712 
7713 	pf->fdir_pf_active_filters = 0;
7714 	pf->fd_tcp4_filter_cnt = 0;
7715 	pf->fd_udp4_filter_cnt = 0;
7716 	pf->fd_sctp4_filter_cnt = 0;
7717 	pf->fd_ip4_filter_cnt = 0;
7718 
7719 	/* Reprogram the default input set for TCP/IPv4 */
7720 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
7721 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
7722 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
7723 
7724 	/* Reprogram the default input set for UDP/IPv4 */
7725 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
7726 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
7727 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
7728 
7729 	/* Reprogram the default input set for SCTP/IPv4 */
7730 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
7731 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
7732 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
7733 
7734 	/* Reprogram the default input set for Other/IPv4 */
7735 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
7736 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
7737 }
7738 
7739 /**
7740  * i40e_cloud_filter_exit - Cleans up the cloud filters
7741  * @pf: Pointer to PF
7742  *
7743  * This function destroys the hlist where all the cloud filters
7744  * were saved.
7745  **/
7746 static void i40e_cloud_filter_exit(struct i40e_pf *pf)
7747 {
7748 	struct i40e_cloud_filter *cfilter;
7749 	struct hlist_node *node;
7750 
7751 	hlist_for_each_entry_safe(cfilter, node,
7752 				  &pf->cloud_filter_list, cloud_node) {
7753 		hlist_del(&cfilter->cloud_node);
7754 		kfree(cfilter);
7755 	}
7756 	pf->num_cloud_filters = 0;
7757 
7758 	if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
7759 	    !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
7760 		pf->flags |= I40E_FLAG_FD_SB_ENABLED;
7761 		pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
7762 		pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
7763 	}
7764 }
7765 
7766 /**
7767  * i40e_close - Disables a network interface
7768  * @netdev: network interface device structure
7769  *
7770  * The close entry point is called when an interface is de-activated
7771  * by the OS.  The hardware is still under the driver's control, but
7772  * this netdev interface is disabled.
7773  *
7774  * Returns 0, this is not allowed to fail
7775  **/
7776 int i40e_close(struct net_device *netdev)
7777 {
7778 	struct i40e_netdev_priv *np = netdev_priv(netdev);
7779 	struct i40e_vsi *vsi = np->vsi;
7780 
7781 	i40e_vsi_close(vsi);
7782 
7783 	return 0;
7784 }
7785 
7786 /**
7787  * i40e_do_reset - Start a PF or Core Reset sequence
7788  * @pf: board private structure
7789  * @reset_flags: which reset is requested
7790  * @lock_acquired: indicates whether or not the lock has been acquired
7791  * before this function was called.
7792  *
7793  * The essential difference in resets is that the PF Reset
7794  * doesn't clear the packet buffers, doesn't reset the PE
7795  * firmware, and doesn't bother the other PFs on the chip.
7796  **/
7797 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
7798 {
7799 	u32 val;
7800 
7801 	WARN_ON(in_interrupt());
7802 
7803 
7804 	/* do the biggest reset indicated */
7805 	if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
7806 
7807 		/* Request a Global Reset
7808 		 *
7809 		 * This will start the chip's countdown to the actual full
7810 		 * chip reset event, and a warning interrupt to be sent
7811 		 * to all PFs, including the requestor.  Our handler
7812 		 * for the warning interrupt will deal with the shutdown
7813 		 * and recovery of the switch setup.
7814 		 */
7815 		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
7816 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
7817 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
7818 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
7819 
7820 	} else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
7821 
7822 		/* Request a Core Reset
7823 		 *
7824 		 * Same as Global Reset, except does *not* include the MAC/PHY
7825 		 */
7826 		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
7827 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
7828 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
7829 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
7830 		i40e_flush(&pf->hw);
7831 
7832 	} else if (reset_flags & I40E_PF_RESET_FLAG) {
7833 
7834 		/* Request a PF Reset
7835 		 *
7836 		 * Resets only the PF-specific registers
7837 		 *
7838 		 * This goes directly to the tear-down and rebuild of
7839 		 * the switch, since we need to do all the recovery as
7840 		 * for the Core Reset.
7841 		 */
7842 		dev_dbg(&pf->pdev->dev, "PFR requested\n");
7843 		i40e_handle_reset_warning(pf, lock_acquired);
7844 
7845 	} else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
7846 		int v;
7847 
7848 		/* Find the VSI(s) that requested a re-init */
7849 		dev_info(&pf->pdev->dev,
7850 			 "VSI reinit requested\n");
7851 		for (v = 0; v < pf->num_alloc_vsi; v++) {
7852 			struct i40e_vsi *vsi = pf->vsi[v];
7853 
7854 			if (vsi != NULL &&
7855 			    test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
7856 					       vsi->state))
7857 				i40e_vsi_reinit_locked(pf->vsi[v]);
7858 		}
7859 	} else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
7860 		int v;
7861 
7862 		/* Find the VSI(s) that needs to be brought down */
7863 		dev_info(&pf->pdev->dev, "VSI down requested\n");
7864 		for (v = 0; v < pf->num_alloc_vsi; v++) {
7865 			struct i40e_vsi *vsi = pf->vsi[v];
7866 
7867 			if (vsi != NULL &&
7868 			    test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
7869 					       vsi->state)) {
7870 				set_bit(__I40E_VSI_DOWN, vsi->state);
7871 				i40e_down(vsi);
7872 			}
7873 		}
7874 	} else {
7875 		dev_info(&pf->pdev->dev,
7876 			 "bad reset request 0x%08x\n", reset_flags);
7877 	}
7878 }
7879 
7880 #ifdef CONFIG_I40E_DCB
7881 /**
7882  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
7883  * @pf: board private structure
7884  * @old_cfg: current DCB config
7885  * @new_cfg: new DCB config
7886  **/
7887 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
7888 			    struct i40e_dcbx_config *old_cfg,
7889 			    struct i40e_dcbx_config *new_cfg)
7890 {
7891 	bool need_reconfig = false;
7892 
7893 	/* Check if ETS configuration has changed */
7894 	if (memcmp(&new_cfg->etscfg,
7895 		   &old_cfg->etscfg,
7896 		   sizeof(new_cfg->etscfg))) {
7897 		/* If Priority Table has changed reconfig is needed */
7898 		if (memcmp(&new_cfg->etscfg.prioritytable,
7899 			   &old_cfg->etscfg.prioritytable,
7900 			   sizeof(new_cfg->etscfg.prioritytable))) {
7901 			need_reconfig = true;
7902 			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
7903 		}
7904 
7905 		if (memcmp(&new_cfg->etscfg.tcbwtable,
7906 			   &old_cfg->etscfg.tcbwtable,
7907 			   sizeof(new_cfg->etscfg.tcbwtable)))
7908 			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
7909 
7910 		if (memcmp(&new_cfg->etscfg.tsatable,
7911 			   &old_cfg->etscfg.tsatable,
7912 			   sizeof(new_cfg->etscfg.tsatable)))
7913 			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
7914 	}
7915 
7916 	/* Check if PFC configuration has changed */
7917 	if (memcmp(&new_cfg->pfc,
7918 		   &old_cfg->pfc,
7919 		   sizeof(new_cfg->pfc))) {
7920 		need_reconfig = true;
7921 		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
7922 	}
7923 
7924 	/* Check if APP Table has changed */
7925 	if (memcmp(&new_cfg->app,
7926 		   &old_cfg->app,
7927 		   sizeof(new_cfg->app))) {
7928 		need_reconfig = true;
7929 		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
7930 	}
7931 
7932 	dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
7933 	return need_reconfig;
7934 }
7935 
7936 /**
7937  * i40e_handle_lldp_event - Handle LLDP Change MIB event
7938  * @pf: board private structure
7939  * @e: event info posted on ARQ
7940  **/
7941 static int i40e_handle_lldp_event(struct i40e_pf *pf,
7942 				  struct i40e_arq_event_info *e)
7943 {
7944 	struct i40e_aqc_lldp_get_mib *mib =
7945 		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
7946 	struct i40e_hw *hw = &pf->hw;
7947 	struct i40e_dcbx_config tmp_dcbx_cfg;
7948 	bool need_reconfig = false;
7949 	int ret = 0;
7950 	u8 type;
7951 
7952 	/* Not DCB capable or capability disabled */
7953 	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
7954 		return ret;
7955 
7956 	/* Ignore if event is not for Nearest Bridge */
7957 	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
7958 		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
7959 	dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
7960 	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
7961 		return ret;
7962 
7963 	/* Check MIB Type and return if event for Remote MIB update */
7964 	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
7965 	dev_dbg(&pf->pdev->dev,
7966 		"LLDP event mib type %s\n", type ? "remote" : "local");
7967 	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
7968 		/* Update the remote cached instance and return */
7969 		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
7970 				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
7971 				&hw->remote_dcbx_config);
7972 		goto exit;
7973 	}
7974 
7975 	/* Store the old configuration */
7976 	tmp_dcbx_cfg = hw->local_dcbx_config;
7977 
7978 	/* Reset the old DCBx configuration data */
7979 	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
7980 	/* Get updated DCBX data from firmware */
7981 	ret = i40e_get_dcb_config(&pf->hw);
7982 	if (ret) {
7983 		dev_info(&pf->pdev->dev,
7984 			 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
7985 			 i40e_stat_str(&pf->hw, ret),
7986 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7987 		goto exit;
7988 	}
7989 
7990 	/* No change detected in DCBX configs */
7991 	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
7992 		    sizeof(tmp_dcbx_cfg))) {
7993 		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
7994 		goto exit;
7995 	}
7996 
7997 	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
7998 					       &hw->local_dcbx_config);
7999 
8000 	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
8001 
8002 	if (!need_reconfig)
8003 		goto exit;
8004 
8005 	/* Enable DCB tagging only when more than one TC */
8006 	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
8007 		pf->flags |= I40E_FLAG_DCB_ENABLED;
8008 	else
8009 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
8010 
8011 	set_bit(__I40E_PORT_SUSPENDED, pf->state);
8012 	/* Reconfiguration needed quiesce all VSIs */
8013 	i40e_pf_quiesce_all_vsi(pf);
8014 
8015 	/* Changes in configuration update VEB/VSI */
8016 	i40e_dcb_reconfigure(pf);
8017 
8018 	ret = i40e_resume_port_tx(pf);
8019 
8020 	clear_bit(__I40E_PORT_SUSPENDED, pf->state);
8021 	/* In case of error no point in resuming VSIs */
8022 	if (ret)
8023 		goto exit;
8024 
8025 	/* Wait for the PF's queues to be disabled */
8026 	ret = i40e_pf_wait_queues_disabled(pf);
8027 	if (ret) {
8028 		/* Schedule PF reset to recover */
8029 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
8030 		i40e_service_event_schedule(pf);
8031 	} else {
8032 		i40e_pf_unquiesce_all_vsi(pf);
8033 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
8034 		      I40E_FLAG_CLIENT_L2_CHANGE);
8035 	}
8036 
8037 exit:
8038 	return ret;
8039 }
8040 #endif /* CONFIG_I40E_DCB */
8041 
8042 /**
8043  * i40e_do_reset_safe - Protected reset path for userland calls.
8044  * @pf: board private structure
8045  * @reset_flags: which reset is requested
8046  *
8047  **/
8048 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
8049 {
8050 	rtnl_lock();
8051 	i40e_do_reset(pf, reset_flags, true);
8052 	rtnl_unlock();
8053 }
8054 
8055 /**
8056  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
8057  * @pf: board private structure
8058  * @e: event info posted on ARQ
8059  *
8060  * Handler for LAN Queue Overflow Event generated by the firmware for PF
8061  * and VF queues
8062  **/
8063 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
8064 					   struct i40e_arq_event_info *e)
8065 {
8066 	struct i40e_aqc_lan_overflow *data =
8067 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
8068 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
8069 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
8070 	struct i40e_hw *hw = &pf->hw;
8071 	struct i40e_vf *vf;
8072 	u16 vf_id;
8073 
8074 	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
8075 		queue, qtx_ctl);
8076 
8077 	/* Queue belongs to VF, find the VF and issue VF reset */
8078 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
8079 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
8080 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
8081 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
8082 		vf_id -= hw->func_caps.vf_base_id;
8083 		vf = &pf->vf[vf_id];
8084 		i40e_vc_notify_vf_reset(vf);
8085 		/* Allow VF to process pending reset notification */
8086 		msleep(20);
8087 		i40e_reset_vf(vf, false);
8088 	}
8089 }
8090 
8091 /**
8092  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
8093  * @pf: board private structure
8094  **/
8095 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
8096 {
8097 	u32 val, fcnt_prog;
8098 
8099 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8100 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
8101 	return fcnt_prog;
8102 }
8103 
8104 /**
8105  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
8106  * @pf: board private structure
8107  **/
8108 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
8109 {
8110 	u32 val, fcnt_prog;
8111 
8112 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8113 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
8114 		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
8115 		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
8116 	return fcnt_prog;
8117 }
8118 
8119 /**
8120  * i40e_get_global_fd_count - Get total FD filters programmed on device
8121  * @pf: board private structure
8122  **/
8123 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
8124 {
8125 	u32 val, fcnt_prog;
8126 
8127 	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
8128 	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
8129 		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
8130 		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
8131 	return fcnt_prog;
8132 }
8133 
8134 /**
8135  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
8136  * @pf: board private structure
8137  **/
8138 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
8139 {
8140 	struct i40e_fdir_filter *filter;
8141 	u32 fcnt_prog, fcnt_avail;
8142 	struct hlist_node *node;
8143 
8144 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8145 		return;
8146 
8147 	/* Check if we have enough room to re-enable FDir SB capability. */
8148 	fcnt_prog = i40e_get_global_fd_count(pf);
8149 	fcnt_avail = pf->fdir_pf_filter_count;
8150 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
8151 	    (pf->fd_add_err == 0) ||
8152 	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
8153 		if (pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED) {
8154 			pf->flags &= ~I40E_FLAG_FD_SB_AUTO_DISABLED;
8155 			if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
8156 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
8157 				dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
8158 		}
8159 	}
8160 
8161 	/* We should wait for even more space before re-enabling ATR.
8162 	 * Additionally, we cannot enable ATR as long as we still have TCP SB
8163 	 * rules active.
8164 	 */
8165 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
8166 	    (pf->fd_tcp4_filter_cnt == 0)) {
8167 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
8168 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
8169 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8170 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
8171 				dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
8172 		}
8173 	}
8174 
8175 	/* if hw had a problem adding a filter, delete it */
8176 	if (pf->fd_inv > 0) {
8177 		hlist_for_each_entry_safe(filter, node,
8178 					  &pf->fdir_filter_list, fdir_node) {
8179 			if (filter->fd_id == pf->fd_inv) {
8180 				hlist_del(&filter->fdir_node);
8181 				kfree(filter);
8182 				pf->fdir_pf_active_filters--;
8183 				pf->fd_inv = 0;
8184 			}
8185 		}
8186 	}
8187 }
8188 
8189 #define I40E_MIN_FD_FLUSH_INTERVAL 10
8190 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
8191 /**
8192  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
8193  * @pf: board private structure
8194  **/
8195 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
8196 {
8197 	unsigned long min_flush_time;
8198 	int flush_wait_retry = 50;
8199 	bool disable_atr = false;
8200 	int fd_room;
8201 	int reg;
8202 
8203 	if (!time_after(jiffies, pf->fd_flush_timestamp +
8204 				 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
8205 		return;
8206 
8207 	/* If the flush is happening too quick and we have mostly SB rules we
8208 	 * should not re-enable ATR for some time.
8209 	 */
8210 	min_flush_time = pf->fd_flush_timestamp +
8211 			 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
8212 	fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
8213 
8214 	if (!(time_after(jiffies, min_flush_time)) &&
8215 	    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
8216 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
8217 			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
8218 		disable_atr = true;
8219 	}
8220 
8221 	pf->fd_flush_timestamp = jiffies;
8222 	pf->flags |= I40E_FLAG_FD_ATR_AUTO_DISABLED;
8223 	/* flush all filters */
8224 	wr32(&pf->hw, I40E_PFQF_CTL_1,
8225 	     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
8226 	i40e_flush(&pf->hw);
8227 	pf->fd_flush_cnt++;
8228 	pf->fd_add_err = 0;
8229 	do {
8230 		/* Check FD flush status every 5-6msec */
8231 		usleep_range(5000, 6000);
8232 		reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
8233 		if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
8234 			break;
8235 	} while (flush_wait_retry--);
8236 	if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
8237 		dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
8238 	} else {
8239 		/* replay sideband filters */
8240 		i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
8241 		if (!disable_atr && !pf->fd_tcp4_filter_cnt)
8242 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
8243 		clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
8244 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
8245 			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
8246 	}
8247 }
8248 
8249 /**
8250  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
8251  * @pf: board private structure
8252  **/
8253 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
8254 {
8255 	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
8256 }
8257 
8258 /* We can see up to 256 filter programming desc in transit if the filters are
8259  * being applied really fast; before we see the first
8260  * filter miss error on Rx queue 0. Accumulating enough error messages before
8261  * reacting will make sure we don't cause flush too often.
8262  */
8263 #define I40E_MAX_FD_PROGRAM_ERROR 256
8264 
8265 /**
8266  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
8267  * @pf: board private structure
8268  **/
8269 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
8270 {
8271 
8272 	/* if interface is down do nothing */
8273 	if (test_bit(__I40E_DOWN, pf->state))
8274 		return;
8275 
8276 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8277 		i40e_fdir_flush_and_replay(pf);
8278 
8279 	i40e_fdir_check_and_reenable(pf);
8280 
8281 }
8282 
8283 /**
8284  * i40e_vsi_link_event - notify VSI of a link event
8285  * @vsi: vsi to be notified
8286  * @link_up: link up or down
8287  **/
8288 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
8289 {
8290 	if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
8291 		return;
8292 
8293 	switch (vsi->type) {
8294 	case I40E_VSI_MAIN:
8295 		if (!vsi->netdev || !vsi->netdev_registered)
8296 			break;
8297 
8298 		if (link_up) {
8299 			netif_carrier_on(vsi->netdev);
8300 			netif_tx_wake_all_queues(vsi->netdev);
8301 		} else {
8302 			netif_carrier_off(vsi->netdev);
8303 			netif_tx_stop_all_queues(vsi->netdev);
8304 		}
8305 		break;
8306 
8307 	case I40E_VSI_SRIOV:
8308 	case I40E_VSI_VMDQ2:
8309 	case I40E_VSI_CTRL:
8310 	case I40E_VSI_IWARP:
8311 	case I40E_VSI_MIRROR:
8312 	default:
8313 		/* there is no notification for other VSIs */
8314 		break;
8315 	}
8316 }
8317 
8318 /**
8319  * i40e_veb_link_event - notify elements on the veb of a link event
8320  * @veb: veb to be notified
8321  * @link_up: link up or down
8322  **/
8323 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
8324 {
8325 	struct i40e_pf *pf;
8326 	int i;
8327 
8328 	if (!veb || !veb->pf)
8329 		return;
8330 	pf = veb->pf;
8331 
8332 	/* depth first... */
8333 	for (i = 0; i < I40E_MAX_VEB; i++)
8334 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
8335 			i40e_veb_link_event(pf->veb[i], link_up);
8336 
8337 	/* ... now the local VSIs */
8338 	for (i = 0; i < pf->num_alloc_vsi; i++)
8339 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
8340 			i40e_vsi_link_event(pf->vsi[i], link_up);
8341 }
8342 
8343 /**
8344  * i40e_link_event - Update netif_carrier status
8345  * @pf: board private structure
8346  **/
8347 static void i40e_link_event(struct i40e_pf *pf)
8348 {
8349 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8350 	u8 new_link_speed, old_link_speed;
8351 	i40e_status status;
8352 	bool new_link, old_link;
8353 
8354 	/* save off old link status information */
8355 	pf->hw.phy.link_info_old = pf->hw.phy.link_info;
8356 
8357 	/* set this to force the get_link_status call to refresh state */
8358 	pf->hw.phy.get_link_info = true;
8359 
8360 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
8361 
8362 	status = i40e_get_link_status(&pf->hw, &new_link);
8363 
8364 	/* On success, disable temp link polling */
8365 	if (status == I40E_SUCCESS) {
8366 		if (pf->flags & I40E_FLAG_TEMP_LINK_POLLING)
8367 			pf->flags &= ~I40E_FLAG_TEMP_LINK_POLLING;
8368 	} else {
8369 		/* Enable link polling temporarily until i40e_get_link_status
8370 		 * returns I40E_SUCCESS
8371 		 */
8372 		pf->flags |= I40E_FLAG_TEMP_LINK_POLLING;
8373 		dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
8374 			status);
8375 		return;
8376 	}
8377 
8378 	old_link_speed = pf->hw.phy.link_info_old.link_speed;
8379 	new_link_speed = pf->hw.phy.link_info.link_speed;
8380 
8381 	if (new_link == old_link &&
8382 	    new_link_speed == old_link_speed &&
8383 	    (test_bit(__I40E_VSI_DOWN, vsi->state) ||
8384 	     new_link == netif_carrier_ok(vsi->netdev)))
8385 		return;
8386 
8387 	i40e_print_link_message(vsi, new_link);
8388 
8389 	/* Notify the base of the switch tree connected to
8390 	 * the link.  Floating VEBs are not notified.
8391 	 */
8392 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
8393 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
8394 	else
8395 		i40e_vsi_link_event(vsi, new_link);
8396 
8397 	if (pf->vf)
8398 		i40e_vc_notify_link_state(pf);
8399 
8400 	if (pf->flags & I40E_FLAG_PTP)
8401 		i40e_ptp_set_increment(pf);
8402 }
8403 
8404 /**
8405  * i40e_watchdog_subtask - periodic checks not using event driven response
8406  * @pf: board private structure
8407  **/
8408 static void i40e_watchdog_subtask(struct i40e_pf *pf)
8409 {
8410 	int i;
8411 
8412 	/* if interface is down do nothing */
8413 	if (test_bit(__I40E_DOWN, pf->state) ||
8414 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
8415 		return;
8416 
8417 	/* make sure we don't do these things too often */
8418 	if (time_before(jiffies, (pf->service_timer_previous +
8419 				  pf->service_timer_period)))
8420 		return;
8421 	pf->service_timer_previous = jiffies;
8422 
8423 	if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
8424 	    (pf->flags & I40E_FLAG_TEMP_LINK_POLLING))
8425 		i40e_link_event(pf);
8426 
8427 	/* Update the stats for active netdevs so the network stack
8428 	 * can look at updated numbers whenever it cares to
8429 	 */
8430 	for (i = 0; i < pf->num_alloc_vsi; i++)
8431 		if (pf->vsi[i] && pf->vsi[i]->netdev)
8432 			i40e_update_stats(pf->vsi[i]);
8433 
8434 	if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
8435 		/* Update the stats for the active switching components */
8436 		for (i = 0; i < I40E_MAX_VEB; i++)
8437 			if (pf->veb[i])
8438 				i40e_update_veb_stats(pf->veb[i]);
8439 	}
8440 
8441 	i40e_ptp_rx_hang(pf);
8442 	i40e_ptp_tx_hang(pf);
8443 }
8444 
8445 /**
8446  * i40e_reset_subtask - Set up for resetting the device and driver
8447  * @pf: board private structure
8448  **/
8449 static void i40e_reset_subtask(struct i40e_pf *pf)
8450 {
8451 	u32 reset_flags = 0;
8452 
8453 	if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
8454 		reset_flags |= BIT(__I40E_REINIT_REQUESTED);
8455 		clear_bit(__I40E_REINIT_REQUESTED, pf->state);
8456 	}
8457 	if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
8458 		reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
8459 		clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
8460 	}
8461 	if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
8462 		reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
8463 		clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
8464 	}
8465 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
8466 		reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
8467 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
8468 	}
8469 	if (test_bit(__I40E_DOWN_REQUESTED, pf->state)) {
8470 		reset_flags |= BIT(__I40E_DOWN_REQUESTED);
8471 		clear_bit(__I40E_DOWN_REQUESTED, pf->state);
8472 	}
8473 
8474 	/* If there's a recovery already waiting, it takes
8475 	 * precedence before starting a new reset sequence.
8476 	 */
8477 	if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
8478 		i40e_prep_for_reset(pf, false);
8479 		i40e_reset(pf);
8480 		i40e_rebuild(pf, false, false);
8481 	}
8482 
8483 	/* If we're already down or resetting, just bail */
8484 	if (reset_flags &&
8485 	    !test_bit(__I40E_DOWN, pf->state) &&
8486 	    !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
8487 		i40e_do_reset(pf, reset_flags, false);
8488 	}
8489 }
8490 
8491 /**
8492  * i40e_handle_link_event - Handle link event
8493  * @pf: board private structure
8494  * @e: event info posted on ARQ
8495  **/
8496 static void i40e_handle_link_event(struct i40e_pf *pf,
8497 				   struct i40e_arq_event_info *e)
8498 {
8499 	struct i40e_aqc_get_link_status *status =
8500 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
8501 
8502 	/* Do a new status request to re-enable LSE reporting
8503 	 * and load new status information into the hw struct
8504 	 * This completely ignores any state information
8505 	 * in the ARQ event info, instead choosing to always
8506 	 * issue the AQ update link status command.
8507 	 */
8508 	i40e_link_event(pf);
8509 
8510 	/* Check if module meets thermal requirements */
8511 	if (status->phy_type == I40E_PHY_TYPE_NOT_SUPPORTED_HIGH_TEMP) {
8512 		dev_err(&pf->pdev->dev,
8513 			"Rx/Tx is disabled on this device because the module does not meet thermal requirements.\n");
8514 		dev_err(&pf->pdev->dev,
8515 			"Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
8516 	} else {
8517 		/* check for unqualified module, if link is down, suppress
8518 		 * the message if link was forced to be down.
8519 		 */
8520 		if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
8521 		    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
8522 		    (!(status->link_info & I40E_AQ_LINK_UP)) &&
8523 		    (!(pf->flags & I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED))) {
8524 			dev_err(&pf->pdev->dev,
8525 				"Rx/Tx is disabled on this device because an unsupported SFP module type was detected.\n");
8526 			dev_err(&pf->pdev->dev,
8527 				"Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
8528 		}
8529 	}
8530 }
8531 
8532 /**
8533  * i40e_clean_adminq_subtask - Clean the AdminQ rings
8534  * @pf: board private structure
8535  **/
8536 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
8537 {
8538 	struct i40e_arq_event_info event;
8539 	struct i40e_hw *hw = &pf->hw;
8540 	u16 pending, i = 0;
8541 	i40e_status ret;
8542 	u16 opcode;
8543 	u32 oldval;
8544 	u32 val;
8545 
8546 	/* Do not run clean AQ when PF reset fails */
8547 	if (test_bit(__I40E_RESET_FAILED, pf->state))
8548 		return;
8549 
8550 	/* check for error indications */
8551 	val = rd32(&pf->hw, pf->hw.aq.arq.len);
8552 	oldval = val;
8553 	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
8554 		if (hw->debug_mask & I40E_DEBUG_AQ)
8555 			dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
8556 		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
8557 	}
8558 	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
8559 		if (hw->debug_mask & I40E_DEBUG_AQ)
8560 			dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
8561 		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
8562 		pf->arq_overflows++;
8563 	}
8564 	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
8565 		if (hw->debug_mask & I40E_DEBUG_AQ)
8566 			dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
8567 		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
8568 	}
8569 	if (oldval != val)
8570 		wr32(&pf->hw, pf->hw.aq.arq.len, val);
8571 
8572 	val = rd32(&pf->hw, pf->hw.aq.asq.len);
8573 	oldval = val;
8574 	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
8575 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
8576 			dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
8577 		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
8578 	}
8579 	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
8580 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
8581 			dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
8582 		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
8583 	}
8584 	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
8585 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
8586 			dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
8587 		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
8588 	}
8589 	if (oldval != val)
8590 		wr32(&pf->hw, pf->hw.aq.asq.len, val);
8591 
8592 	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
8593 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
8594 	if (!event.msg_buf)
8595 		return;
8596 
8597 	do {
8598 		ret = i40e_clean_arq_element(hw, &event, &pending);
8599 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
8600 			break;
8601 		else if (ret) {
8602 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
8603 			break;
8604 		}
8605 
8606 		opcode = le16_to_cpu(event.desc.opcode);
8607 		switch (opcode) {
8608 
8609 		case i40e_aqc_opc_get_link_status:
8610 			i40e_handle_link_event(pf, &event);
8611 			break;
8612 		case i40e_aqc_opc_send_msg_to_pf:
8613 			ret = i40e_vc_process_vf_msg(pf,
8614 					le16_to_cpu(event.desc.retval),
8615 					le32_to_cpu(event.desc.cookie_high),
8616 					le32_to_cpu(event.desc.cookie_low),
8617 					event.msg_buf,
8618 					event.msg_len);
8619 			break;
8620 		case i40e_aqc_opc_lldp_update_mib:
8621 			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
8622 #ifdef CONFIG_I40E_DCB
8623 			rtnl_lock();
8624 			ret = i40e_handle_lldp_event(pf, &event);
8625 			rtnl_unlock();
8626 #endif /* CONFIG_I40E_DCB */
8627 			break;
8628 		case i40e_aqc_opc_event_lan_overflow:
8629 			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
8630 			i40e_handle_lan_overflow_event(pf, &event);
8631 			break;
8632 		case i40e_aqc_opc_send_msg_to_peer:
8633 			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
8634 			break;
8635 		case i40e_aqc_opc_nvm_erase:
8636 		case i40e_aqc_opc_nvm_update:
8637 		case i40e_aqc_opc_oem_post_update:
8638 			i40e_debug(&pf->hw, I40E_DEBUG_NVM,
8639 				   "ARQ NVM operation 0x%04x completed\n",
8640 				   opcode);
8641 			break;
8642 		default:
8643 			dev_info(&pf->pdev->dev,
8644 				 "ARQ: Unknown event 0x%04x ignored\n",
8645 				 opcode);
8646 			break;
8647 		}
8648 	} while (i++ < pf->adminq_work_limit);
8649 
8650 	if (i < pf->adminq_work_limit)
8651 		clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
8652 
8653 	/* re-enable Admin queue interrupt cause */
8654 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
8655 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
8656 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
8657 	i40e_flush(hw);
8658 
8659 	kfree(event.msg_buf);
8660 }
8661 
8662 /**
8663  * i40e_verify_eeprom - make sure eeprom is good to use
8664  * @pf: board private structure
8665  **/
8666 static void i40e_verify_eeprom(struct i40e_pf *pf)
8667 {
8668 	int err;
8669 
8670 	err = i40e_diag_eeprom_test(&pf->hw);
8671 	if (err) {
8672 		/* retry in case of garbage read */
8673 		err = i40e_diag_eeprom_test(&pf->hw);
8674 		if (err) {
8675 			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
8676 				 err);
8677 			set_bit(__I40E_BAD_EEPROM, pf->state);
8678 		}
8679 	}
8680 
8681 	if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
8682 		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
8683 		clear_bit(__I40E_BAD_EEPROM, pf->state);
8684 	}
8685 }
8686 
8687 /**
8688  * i40e_enable_pf_switch_lb
8689  * @pf: pointer to the PF structure
8690  *
8691  * enable switch loop back or die - no point in a return value
8692  **/
8693 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
8694 {
8695 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8696 	struct i40e_vsi_context ctxt;
8697 	int ret;
8698 
8699 	ctxt.seid = pf->main_vsi_seid;
8700 	ctxt.pf_num = pf->hw.pf_id;
8701 	ctxt.vf_num = 0;
8702 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
8703 	if (ret) {
8704 		dev_info(&pf->pdev->dev,
8705 			 "couldn't get PF vsi config, err %s aq_err %s\n",
8706 			 i40e_stat_str(&pf->hw, ret),
8707 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8708 		return;
8709 	}
8710 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8711 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8712 	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8713 
8714 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
8715 	if (ret) {
8716 		dev_info(&pf->pdev->dev,
8717 			 "update vsi switch failed, err %s aq_err %s\n",
8718 			 i40e_stat_str(&pf->hw, ret),
8719 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8720 	}
8721 }
8722 
8723 /**
8724  * i40e_disable_pf_switch_lb
8725  * @pf: pointer to the PF structure
8726  *
8727  * disable switch loop back or die - no point in a return value
8728  **/
8729 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
8730 {
8731 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8732 	struct i40e_vsi_context ctxt;
8733 	int ret;
8734 
8735 	ctxt.seid = pf->main_vsi_seid;
8736 	ctxt.pf_num = pf->hw.pf_id;
8737 	ctxt.vf_num = 0;
8738 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
8739 	if (ret) {
8740 		dev_info(&pf->pdev->dev,
8741 			 "couldn't get PF vsi config, err %s aq_err %s\n",
8742 			 i40e_stat_str(&pf->hw, ret),
8743 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8744 		return;
8745 	}
8746 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8747 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8748 	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8749 
8750 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
8751 	if (ret) {
8752 		dev_info(&pf->pdev->dev,
8753 			 "update vsi switch failed, err %s aq_err %s\n",
8754 			 i40e_stat_str(&pf->hw, ret),
8755 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8756 	}
8757 }
8758 
8759 /**
8760  * i40e_config_bridge_mode - Configure the HW bridge mode
8761  * @veb: pointer to the bridge instance
8762  *
8763  * Configure the loop back mode for the LAN VSI that is downlink to the
8764  * specified HW bridge instance. It is expected this function is called
8765  * when a new HW bridge is instantiated.
8766  **/
8767 static void i40e_config_bridge_mode(struct i40e_veb *veb)
8768 {
8769 	struct i40e_pf *pf = veb->pf;
8770 
8771 	if (pf->hw.debug_mask & I40E_DEBUG_LAN)
8772 		dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
8773 			 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
8774 	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
8775 		i40e_disable_pf_switch_lb(pf);
8776 	else
8777 		i40e_enable_pf_switch_lb(pf);
8778 }
8779 
8780 /**
8781  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
8782  * @veb: pointer to the VEB instance
8783  *
8784  * This is a recursive function that first builds the attached VSIs then
8785  * recurses in to build the next layer of VEB.  We track the connections
8786  * through our own index numbers because the seid's from the HW could
8787  * change across the reset.
8788  **/
8789 static int i40e_reconstitute_veb(struct i40e_veb *veb)
8790 {
8791 	struct i40e_vsi *ctl_vsi = NULL;
8792 	struct i40e_pf *pf = veb->pf;
8793 	int v, veb_idx;
8794 	int ret;
8795 
8796 	/* build VSI that owns this VEB, temporarily attached to base VEB */
8797 	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
8798 		if (pf->vsi[v] &&
8799 		    pf->vsi[v]->veb_idx == veb->idx &&
8800 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
8801 			ctl_vsi = pf->vsi[v];
8802 			break;
8803 		}
8804 	}
8805 	if (!ctl_vsi) {
8806 		dev_info(&pf->pdev->dev,
8807 			 "missing owner VSI for veb_idx %d\n", veb->idx);
8808 		ret = -ENOENT;
8809 		goto end_reconstitute;
8810 	}
8811 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
8812 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
8813 	ret = i40e_add_vsi(ctl_vsi);
8814 	if (ret) {
8815 		dev_info(&pf->pdev->dev,
8816 			 "rebuild of veb_idx %d owner VSI failed: %d\n",
8817 			 veb->idx, ret);
8818 		goto end_reconstitute;
8819 	}
8820 	i40e_vsi_reset_stats(ctl_vsi);
8821 
8822 	/* create the VEB in the switch and move the VSI onto the VEB */
8823 	ret = i40e_add_veb(veb, ctl_vsi);
8824 	if (ret)
8825 		goto end_reconstitute;
8826 
8827 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
8828 		veb->bridge_mode = BRIDGE_MODE_VEB;
8829 	else
8830 		veb->bridge_mode = BRIDGE_MODE_VEPA;
8831 	i40e_config_bridge_mode(veb);
8832 
8833 	/* create the remaining VSIs attached to this VEB */
8834 	for (v = 0; v < pf->num_alloc_vsi; v++) {
8835 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
8836 			continue;
8837 
8838 		if (pf->vsi[v]->veb_idx == veb->idx) {
8839 			struct i40e_vsi *vsi = pf->vsi[v];
8840 
8841 			vsi->uplink_seid = veb->seid;
8842 			ret = i40e_add_vsi(vsi);
8843 			if (ret) {
8844 				dev_info(&pf->pdev->dev,
8845 					 "rebuild of vsi_idx %d failed: %d\n",
8846 					 v, ret);
8847 				goto end_reconstitute;
8848 			}
8849 			i40e_vsi_reset_stats(vsi);
8850 		}
8851 	}
8852 
8853 	/* create any VEBs attached to this VEB - RECURSION */
8854 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
8855 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
8856 			pf->veb[veb_idx]->uplink_seid = veb->seid;
8857 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
8858 			if (ret)
8859 				break;
8860 		}
8861 	}
8862 
8863 end_reconstitute:
8864 	return ret;
8865 }
8866 
8867 /**
8868  * i40e_get_capabilities - get info about the HW
8869  * @pf: the PF struct
8870  **/
8871 static int i40e_get_capabilities(struct i40e_pf *pf,
8872 				 enum i40e_admin_queue_opc list_type)
8873 {
8874 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
8875 	u16 data_size;
8876 	int buf_len;
8877 	int err;
8878 
8879 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
8880 	do {
8881 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
8882 		if (!cap_buf)
8883 			return -ENOMEM;
8884 
8885 		/* this loads the data into the hw struct for us */
8886 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
8887 						    &data_size, list_type,
8888 						    NULL);
8889 		/* data loaded, buffer no longer needed */
8890 		kfree(cap_buf);
8891 
8892 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
8893 			/* retry with a larger buffer */
8894 			buf_len = data_size;
8895 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
8896 			dev_info(&pf->pdev->dev,
8897 				 "capability discovery failed, err %s aq_err %s\n",
8898 				 i40e_stat_str(&pf->hw, err),
8899 				 i40e_aq_str(&pf->hw,
8900 					     pf->hw.aq.asq_last_status));
8901 			return -ENODEV;
8902 		}
8903 	} while (err);
8904 
8905 	if (pf->hw.debug_mask & I40E_DEBUG_USER) {
8906 		if (list_type == i40e_aqc_opc_list_func_capabilities) {
8907 			dev_info(&pf->pdev->dev,
8908 				 "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",
8909 				 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
8910 				 pf->hw.func_caps.num_msix_vectors,
8911 				 pf->hw.func_caps.num_msix_vectors_vf,
8912 				 pf->hw.func_caps.fd_filters_guaranteed,
8913 				 pf->hw.func_caps.fd_filters_best_effort,
8914 				 pf->hw.func_caps.num_tx_qp,
8915 				 pf->hw.func_caps.num_vsis);
8916 		} else if (list_type == i40e_aqc_opc_list_dev_capabilities) {
8917 			dev_info(&pf->pdev->dev,
8918 				 "switch_mode=0x%04x, function_valid=0x%08x\n",
8919 				 pf->hw.dev_caps.switch_mode,
8920 				 pf->hw.dev_caps.valid_functions);
8921 			dev_info(&pf->pdev->dev,
8922 				 "SR-IOV=%d, num_vfs for all function=%u\n",
8923 				 pf->hw.dev_caps.sr_iov_1_1,
8924 				 pf->hw.dev_caps.num_vfs);
8925 			dev_info(&pf->pdev->dev,
8926 				 "num_vsis=%u, num_rx:%u, num_tx=%u\n",
8927 				 pf->hw.dev_caps.num_vsis,
8928 				 pf->hw.dev_caps.num_rx_qp,
8929 				 pf->hw.dev_caps.num_tx_qp);
8930 		}
8931 	}
8932 	if (list_type == i40e_aqc_opc_list_func_capabilities) {
8933 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
8934 		       + pf->hw.func_caps.num_vfs)
8935 		if (pf->hw.revision_id == 0 &&
8936 		    pf->hw.func_caps.num_vsis < DEF_NUM_VSI) {
8937 			dev_info(&pf->pdev->dev,
8938 				 "got num_vsis %d, setting num_vsis to %d\n",
8939 				 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
8940 			pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
8941 		}
8942 	}
8943 	return 0;
8944 }
8945 
8946 static int i40e_vsi_clear(struct i40e_vsi *vsi);
8947 
8948 /**
8949  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
8950  * @pf: board private structure
8951  **/
8952 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
8953 {
8954 	struct i40e_vsi *vsi;
8955 
8956 	/* quick workaround for an NVM issue that leaves a critical register
8957 	 * uninitialized
8958 	 */
8959 	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
8960 		static const u32 hkey[] = {
8961 			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
8962 			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
8963 			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
8964 			0x95b3a76d};
8965 		int i;
8966 
8967 		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
8968 			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
8969 	}
8970 
8971 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8972 		return;
8973 
8974 	/* find existing VSI and see if it needs configuring */
8975 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
8976 
8977 	/* create a new VSI if none exists */
8978 	if (!vsi) {
8979 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
8980 				     pf->vsi[pf->lan_vsi]->seid, 0);
8981 		if (!vsi) {
8982 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
8983 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8984 			pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
8985 			return;
8986 		}
8987 	}
8988 
8989 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
8990 }
8991 
8992 /**
8993  * i40e_fdir_teardown - release the Flow Director resources
8994  * @pf: board private structure
8995  **/
8996 static void i40e_fdir_teardown(struct i40e_pf *pf)
8997 {
8998 	struct i40e_vsi *vsi;
8999 
9000 	i40e_fdir_filter_exit(pf);
9001 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
9002 	if (vsi)
9003 		i40e_vsi_release(vsi);
9004 }
9005 
9006 /**
9007  * i40e_rebuild_cloud_filters - Rebuilds cloud filters for VSIs
9008  * @vsi: PF main vsi
9009  * @seid: seid of main or channel VSIs
9010  *
9011  * Rebuilds cloud filters associated with main VSI and channel VSIs if they
9012  * existed before reset
9013  **/
9014 static int i40e_rebuild_cloud_filters(struct i40e_vsi *vsi, u16 seid)
9015 {
9016 	struct i40e_cloud_filter *cfilter;
9017 	struct i40e_pf *pf = vsi->back;
9018 	struct hlist_node *node;
9019 	i40e_status ret;
9020 
9021 	/* Add cloud filters back if they exist */
9022 	hlist_for_each_entry_safe(cfilter, node, &pf->cloud_filter_list,
9023 				  cloud_node) {
9024 		if (cfilter->seid != seid)
9025 			continue;
9026 
9027 		if (cfilter->dst_port)
9028 			ret = i40e_add_del_cloud_filter_big_buf(vsi, cfilter,
9029 								true);
9030 		else
9031 			ret = i40e_add_del_cloud_filter(vsi, cfilter, true);
9032 
9033 		if (ret) {
9034 			dev_dbg(&pf->pdev->dev,
9035 				"Failed to rebuild cloud filter, err %s aq_err %s\n",
9036 				i40e_stat_str(&pf->hw, ret),
9037 				i40e_aq_str(&pf->hw,
9038 					    pf->hw.aq.asq_last_status));
9039 			return ret;
9040 		}
9041 	}
9042 	return 0;
9043 }
9044 
9045 /**
9046  * i40e_rebuild_channels - Rebuilds channel VSIs if they existed before reset
9047  * @vsi: PF main vsi
9048  *
9049  * Rebuilds channel VSIs if they existed before reset
9050  **/
9051 static int i40e_rebuild_channels(struct i40e_vsi *vsi)
9052 {
9053 	struct i40e_channel *ch, *ch_tmp;
9054 	i40e_status ret;
9055 
9056 	if (list_empty(&vsi->ch_list))
9057 		return 0;
9058 
9059 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
9060 		if (!ch->initialized)
9061 			break;
9062 		/* Proceed with creation of channel (VMDq2) VSI */
9063 		ret = i40e_add_channel(vsi->back, vsi->uplink_seid, ch);
9064 		if (ret) {
9065 			dev_info(&vsi->back->pdev->dev,
9066 				 "failed to rebuild channels using uplink_seid %u\n",
9067 				 vsi->uplink_seid);
9068 			return ret;
9069 		}
9070 		if (ch->max_tx_rate) {
9071 			u64 credits = ch->max_tx_rate;
9072 
9073 			if (i40e_set_bw_limit(vsi, ch->seid,
9074 					      ch->max_tx_rate))
9075 				return -EINVAL;
9076 
9077 			do_div(credits, I40E_BW_CREDIT_DIVISOR);
9078 			dev_dbg(&vsi->back->pdev->dev,
9079 				"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
9080 				ch->max_tx_rate,
9081 				credits,
9082 				ch->seid);
9083 		}
9084 		ret = i40e_rebuild_cloud_filters(vsi, ch->seid);
9085 		if (ret) {
9086 			dev_dbg(&vsi->back->pdev->dev,
9087 				"Failed to rebuild cloud filters for channel VSI %u\n",
9088 				ch->seid);
9089 			return ret;
9090 		}
9091 	}
9092 	return 0;
9093 }
9094 
9095 /**
9096  * i40e_prep_for_reset - prep for the core to reset
9097  * @pf: board private structure
9098  * @lock_acquired: indicates whether or not the lock has been acquired
9099  * before this function was called.
9100  *
9101  * Close up the VFs and other things in prep for PF Reset.
9102   **/
9103 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
9104 {
9105 	struct i40e_hw *hw = &pf->hw;
9106 	i40e_status ret = 0;
9107 	u32 v;
9108 
9109 	clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
9110 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
9111 		return;
9112 	if (i40e_check_asq_alive(&pf->hw))
9113 		i40e_vc_notify_reset(pf);
9114 
9115 	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
9116 
9117 	/* quiesce the VSIs and their queues that are not already DOWN */
9118 	/* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
9119 	if (!lock_acquired)
9120 		rtnl_lock();
9121 	i40e_pf_quiesce_all_vsi(pf);
9122 	if (!lock_acquired)
9123 		rtnl_unlock();
9124 
9125 	for (v = 0; v < pf->num_alloc_vsi; v++) {
9126 		if (pf->vsi[v])
9127 			pf->vsi[v]->seid = 0;
9128 	}
9129 
9130 	i40e_shutdown_adminq(&pf->hw);
9131 
9132 	/* call shutdown HMC */
9133 	if (hw->hmc.hmc_obj) {
9134 		ret = i40e_shutdown_lan_hmc(hw);
9135 		if (ret)
9136 			dev_warn(&pf->pdev->dev,
9137 				 "shutdown_lan_hmc failed: %d\n", ret);
9138 	}
9139 }
9140 
9141 /**
9142  * i40e_send_version - update firmware with driver version
9143  * @pf: PF struct
9144  */
9145 static void i40e_send_version(struct i40e_pf *pf)
9146 {
9147 	struct i40e_driver_version dv;
9148 
9149 	dv.major_version = DRV_VERSION_MAJOR;
9150 	dv.minor_version = DRV_VERSION_MINOR;
9151 	dv.build_version = DRV_VERSION_BUILD;
9152 	dv.subbuild_version = 0;
9153 	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
9154 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
9155 }
9156 
9157 /**
9158  * i40e_get_oem_version - get OEM specific version information
9159  * @hw: pointer to the hardware structure
9160  **/
9161 static void i40e_get_oem_version(struct i40e_hw *hw)
9162 {
9163 	u16 block_offset = 0xffff;
9164 	u16 block_length = 0;
9165 	u16 capabilities = 0;
9166 	u16 gen_snap = 0;
9167 	u16 release = 0;
9168 
9169 #define I40E_SR_NVM_OEM_VERSION_PTR		0x1B
9170 #define I40E_NVM_OEM_LENGTH_OFFSET		0x00
9171 #define I40E_NVM_OEM_CAPABILITIES_OFFSET	0x01
9172 #define I40E_NVM_OEM_GEN_OFFSET			0x02
9173 #define I40E_NVM_OEM_RELEASE_OFFSET		0x03
9174 #define I40E_NVM_OEM_CAPABILITIES_MASK		0x000F
9175 #define I40E_NVM_OEM_LENGTH			3
9176 
9177 	/* Check if pointer to OEM version block is valid. */
9178 	i40e_read_nvm_word(hw, I40E_SR_NVM_OEM_VERSION_PTR, &block_offset);
9179 	if (block_offset == 0xffff)
9180 		return;
9181 
9182 	/* Check if OEM version block has correct length. */
9183 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_LENGTH_OFFSET,
9184 			   &block_length);
9185 	if (block_length < I40E_NVM_OEM_LENGTH)
9186 		return;
9187 
9188 	/* Check if OEM version format is as expected. */
9189 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_CAPABILITIES_OFFSET,
9190 			   &capabilities);
9191 	if ((capabilities & I40E_NVM_OEM_CAPABILITIES_MASK) != 0)
9192 		return;
9193 
9194 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_GEN_OFFSET,
9195 			   &gen_snap);
9196 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_RELEASE_OFFSET,
9197 			   &release);
9198 	hw->nvm.oem_ver = (gen_snap << I40E_OEM_SNAP_SHIFT) | release;
9199 	hw->nvm.eetrack = I40E_OEM_EETRACK_ID;
9200 }
9201 
9202 /**
9203  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
9204  * @pf: board private structure
9205  **/
9206 static int i40e_reset(struct i40e_pf *pf)
9207 {
9208 	struct i40e_hw *hw = &pf->hw;
9209 	i40e_status ret;
9210 
9211 	ret = i40e_pf_reset(hw);
9212 	if (ret) {
9213 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
9214 		set_bit(__I40E_RESET_FAILED, pf->state);
9215 		clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
9216 	} else {
9217 		pf->pfr_count++;
9218 	}
9219 	return ret;
9220 }
9221 
9222 /**
9223  * i40e_rebuild - rebuild using a saved config
9224  * @pf: board private structure
9225  * @reinit: if the Main VSI needs to re-initialized.
9226  * @lock_acquired: indicates whether or not the lock has been acquired
9227  * before this function was called.
9228  **/
9229 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
9230 {
9231 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9232 	struct i40e_hw *hw = &pf->hw;
9233 	u8 set_fc_aq_fail = 0;
9234 	i40e_status ret;
9235 	u32 val;
9236 	int v;
9237 
9238 	if (test_bit(__I40E_DOWN, pf->state))
9239 		goto clear_recovery;
9240 	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
9241 
9242 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
9243 	ret = i40e_init_adminq(&pf->hw);
9244 	if (ret) {
9245 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
9246 			 i40e_stat_str(&pf->hw, ret),
9247 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9248 		goto clear_recovery;
9249 	}
9250 	i40e_get_oem_version(&pf->hw);
9251 
9252 	/* re-verify the eeprom if we just had an EMP reset */
9253 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
9254 		i40e_verify_eeprom(pf);
9255 
9256 	i40e_clear_pxe_mode(hw);
9257 	ret = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
9258 	if (ret)
9259 		goto end_core_reset;
9260 
9261 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
9262 				hw->func_caps.num_rx_qp, 0, 0);
9263 	if (ret) {
9264 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
9265 		goto end_core_reset;
9266 	}
9267 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
9268 	if (ret) {
9269 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
9270 		goto end_core_reset;
9271 	}
9272 
9273 #ifdef CONFIG_I40E_DCB
9274 	ret = i40e_init_pf_dcb(pf);
9275 	if (ret) {
9276 		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
9277 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
9278 		/* Continue without DCB enabled */
9279 	}
9280 #endif /* CONFIG_I40E_DCB */
9281 	/* do basic switch setup */
9282 	if (!lock_acquired)
9283 		rtnl_lock();
9284 	ret = i40e_setup_pf_switch(pf, reinit);
9285 	if (ret)
9286 		goto end_unlock;
9287 
9288 	/* The driver only wants link up/down and module qualification
9289 	 * reports from firmware.  Note the negative logic.
9290 	 */
9291 	ret = i40e_aq_set_phy_int_mask(&pf->hw,
9292 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
9293 					 I40E_AQ_EVENT_MEDIA_NA |
9294 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
9295 	if (ret)
9296 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
9297 			 i40e_stat_str(&pf->hw, ret),
9298 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9299 
9300 	/* make sure our flow control settings are restored */
9301 	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
9302 	if (ret)
9303 		dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
9304 			i40e_stat_str(&pf->hw, ret),
9305 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9306 
9307 	/* Rebuild the VSIs and VEBs that existed before reset.
9308 	 * They are still in our local switch element arrays, so only
9309 	 * need to rebuild the switch model in the HW.
9310 	 *
9311 	 * If there were VEBs but the reconstitution failed, we'll try
9312 	 * try to recover minimal use by getting the basic PF VSI working.
9313 	 */
9314 	if (vsi->uplink_seid != pf->mac_seid) {
9315 		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
9316 		/* find the one VEB connected to the MAC, and find orphans */
9317 		for (v = 0; v < I40E_MAX_VEB; v++) {
9318 			if (!pf->veb[v])
9319 				continue;
9320 
9321 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
9322 			    pf->veb[v]->uplink_seid == 0) {
9323 				ret = i40e_reconstitute_veb(pf->veb[v]);
9324 
9325 				if (!ret)
9326 					continue;
9327 
9328 				/* If Main VEB failed, we're in deep doodoo,
9329 				 * so give up rebuilding the switch and set up
9330 				 * for minimal rebuild of PF VSI.
9331 				 * If orphan failed, we'll report the error
9332 				 * but try to keep going.
9333 				 */
9334 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
9335 					dev_info(&pf->pdev->dev,
9336 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
9337 						 ret);
9338 					vsi->uplink_seid = pf->mac_seid;
9339 					break;
9340 				} else if (pf->veb[v]->uplink_seid == 0) {
9341 					dev_info(&pf->pdev->dev,
9342 						 "rebuild of orphan VEB failed: %d\n",
9343 						 ret);
9344 				}
9345 			}
9346 		}
9347 	}
9348 
9349 	if (vsi->uplink_seid == pf->mac_seid) {
9350 		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
9351 		/* no VEB, so rebuild only the Main VSI */
9352 		ret = i40e_add_vsi(vsi);
9353 		if (ret) {
9354 			dev_info(&pf->pdev->dev,
9355 				 "rebuild of Main VSI failed: %d\n", ret);
9356 			goto end_unlock;
9357 		}
9358 	}
9359 
9360 	if (vsi->mqprio_qopt.max_rate[0]) {
9361 		u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
9362 		u64 credits = 0;
9363 
9364 		do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
9365 		ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
9366 		if (ret)
9367 			goto end_unlock;
9368 
9369 		credits = max_tx_rate;
9370 		do_div(credits, I40E_BW_CREDIT_DIVISOR);
9371 		dev_dbg(&vsi->back->pdev->dev,
9372 			"Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
9373 			max_tx_rate,
9374 			credits,
9375 			vsi->seid);
9376 	}
9377 
9378 	ret = i40e_rebuild_cloud_filters(vsi, vsi->seid);
9379 	if (ret)
9380 		goto end_unlock;
9381 
9382 	/* PF Main VSI is rebuild by now, go ahead and rebuild channel VSIs
9383 	 * for this main VSI if they exist
9384 	 */
9385 	ret = i40e_rebuild_channels(vsi);
9386 	if (ret)
9387 		goto end_unlock;
9388 
9389 	/* Reconfigure hardware for allowing smaller MSS in the case
9390 	 * of TSO, so that we avoid the MDD being fired and causing
9391 	 * a reset in the case of small MSS+TSO.
9392 	 */
9393 #define I40E_REG_MSS          0x000E64DC
9394 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
9395 #define I40E_64BYTE_MSS       0x400000
9396 	val = rd32(hw, I40E_REG_MSS);
9397 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
9398 		val &= ~I40E_REG_MSS_MIN_MASK;
9399 		val |= I40E_64BYTE_MSS;
9400 		wr32(hw, I40E_REG_MSS, val);
9401 	}
9402 
9403 	if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
9404 		msleep(75);
9405 		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
9406 		if (ret)
9407 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
9408 				 i40e_stat_str(&pf->hw, ret),
9409 				 i40e_aq_str(&pf->hw,
9410 					     pf->hw.aq.asq_last_status));
9411 	}
9412 	/* reinit the misc interrupt */
9413 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
9414 		ret = i40e_setup_misc_vector(pf);
9415 
9416 	/* Add a filter to drop all Flow control frames from any VSI from being
9417 	 * transmitted. By doing so we stop a malicious VF from sending out
9418 	 * PAUSE or PFC frames and potentially controlling traffic for other
9419 	 * PF/VF VSIs.
9420 	 * The FW can still send Flow control frames if enabled.
9421 	 */
9422 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
9423 						       pf->main_vsi_seid);
9424 
9425 	/* restart the VSIs that were rebuilt and running before the reset */
9426 	i40e_pf_unquiesce_all_vsi(pf);
9427 
9428 	/* Release the RTNL lock before we start resetting VFs */
9429 	if (!lock_acquired)
9430 		rtnl_unlock();
9431 
9432 	i40e_reset_all_vfs(pf, true);
9433 
9434 	/* tell the firmware that we're starting */
9435 	i40e_send_version(pf);
9436 
9437 	/* We've already released the lock, so don't do it again */
9438 	goto end_core_reset;
9439 
9440 end_unlock:
9441 	if (!lock_acquired)
9442 		rtnl_unlock();
9443 end_core_reset:
9444 	clear_bit(__I40E_RESET_FAILED, pf->state);
9445 clear_recovery:
9446 	clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
9447 }
9448 
9449 /**
9450  * i40e_reset_and_rebuild - reset and rebuild using a saved config
9451  * @pf: board private structure
9452  * @reinit: if the Main VSI needs to re-initialized.
9453  * @lock_acquired: indicates whether or not the lock has been acquired
9454  * before this function was called.
9455  **/
9456 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
9457 				   bool lock_acquired)
9458 {
9459 	int ret;
9460 	/* Now we wait for GRST to settle out.
9461 	 * We don't have to delete the VEBs or VSIs from the hw switch
9462 	 * because the reset will make them disappear.
9463 	 */
9464 	ret = i40e_reset(pf);
9465 	if (!ret)
9466 		i40e_rebuild(pf, reinit, lock_acquired);
9467 }
9468 
9469 /**
9470  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
9471  * @pf: board private structure
9472  *
9473  * Close up the VFs and other things in prep for a Core Reset,
9474  * then get ready to rebuild the world.
9475  * @lock_acquired: indicates whether or not the lock has been acquired
9476  * before this function was called.
9477  **/
9478 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
9479 {
9480 	i40e_prep_for_reset(pf, lock_acquired);
9481 	i40e_reset_and_rebuild(pf, false, lock_acquired);
9482 }
9483 
9484 /**
9485  * i40e_handle_mdd_event
9486  * @pf: pointer to the PF structure
9487  *
9488  * Called from the MDD irq handler to identify possibly malicious vfs
9489  **/
9490 static void i40e_handle_mdd_event(struct i40e_pf *pf)
9491 {
9492 	struct i40e_hw *hw = &pf->hw;
9493 	bool mdd_detected = false;
9494 	bool pf_mdd_detected = false;
9495 	struct i40e_vf *vf;
9496 	u32 reg;
9497 	int i;
9498 
9499 	if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
9500 		return;
9501 
9502 	/* find what triggered the MDD event */
9503 	reg = rd32(hw, I40E_GL_MDET_TX);
9504 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
9505 		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
9506 				I40E_GL_MDET_TX_PF_NUM_SHIFT;
9507 		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
9508 				I40E_GL_MDET_TX_VF_NUM_SHIFT;
9509 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
9510 				I40E_GL_MDET_TX_EVENT_SHIFT;
9511 		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
9512 				I40E_GL_MDET_TX_QUEUE_SHIFT) -
9513 				pf->hw.func_caps.base_queue;
9514 		if (netif_msg_tx_err(pf))
9515 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
9516 				 event, queue, pf_num, vf_num);
9517 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
9518 		mdd_detected = true;
9519 	}
9520 	reg = rd32(hw, I40E_GL_MDET_RX);
9521 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
9522 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
9523 				I40E_GL_MDET_RX_FUNCTION_SHIFT;
9524 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
9525 				I40E_GL_MDET_RX_EVENT_SHIFT;
9526 		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
9527 				I40E_GL_MDET_RX_QUEUE_SHIFT) -
9528 				pf->hw.func_caps.base_queue;
9529 		if (netif_msg_rx_err(pf))
9530 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
9531 				 event, queue, func);
9532 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
9533 		mdd_detected = true;
9534 	}
9535 
9536 	if (mdd_detected) {
9537 		reg = rd32(hw, I40E_PF_MDET_TX);
9538 		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
9539 			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
9540 			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
9541 			pf_mdd_detected = true;
9542 		}
9543 		reg = rd32(hw, I40E_PF_MDET_RX);
9544 		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
9545 			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
9546 			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
9547 			pf_mdd_detected = true;
9548 		}
9549 		/* Queue belongs to the PF, initiate a reset */
9550 		if (pf_mdd_detected) {
9551 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
9552 			i40e_service_event_schedule(pf);
9553 		}
9554 	}
9555 
9556 	/* see if one of the VFs needs its hand slapped */
9557 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
9558 		vf = &(pf->vf[i]);
9559 		reg = rd32(hw, I40E_VP_MDET_TX(i));
9560 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
9561 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
9562 			vf->num_mdd_events++;
9563 			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
9564 				 i);
9565 		}
9566 
9567 		reg = rd32(hw, I40E_VP_MDET_RX(i));
9568 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
9569 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
9570 			vf->num_mdd_events++;
9571 			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
9572 				 i);
9573 		}
9574 
9575 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
9576 			dev_info(&pf->pdev->dev,
9577 				 "Too many MDD events on VF %d, disabled\n", i);
9578 			dev_info(&pf->pdev->dev,
9579 				 "Use PF Control I/F to re-enable the VF\n");
9580 			set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
9581 		}
9582 	}
9583 
9584 	/* re-enable mdd interrupt cause */
9585 	clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
9586 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
9587 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
9588 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
9589 	i40e_flush(hw);
9590 }
9591 
9592 static const char *i40e_tunnel_name(struct i40e_udp_port_config *port)
9593 {
9594 	switch (port->type) {
9595 	case UDP_TUNNEL_TYPE_VXLAN:
9596 		return "vxlan";
9597 	case UDP_TUNNEL_TYPE_GENEVE:
9598 		return "geneve";
9599 	default:
9600 		return "unknown";
9601 	}
9602 }
9603 
9604 /**
9605  * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
9606  * @pf: board private structure
9607  **/
9608 static void i40e_sync_udp_filters(struct i40e_pf *pf)
9609 {
9610 	int i;
9611 
9612 	/* loop through and set pending bit for all active UDP filters */
9613 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
9614 		if (pf->udp_ports[i].port)
9615 			pf->pending_udp_bitmap |= BIT_ULL(i);
9616 	}
9617 
9618 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9619 }
9620 
9621 /**
9622  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
9623  * @pf: board private structure
9624  **/
9625 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
9626 {
9627 	struct i40e_hw *hw = &pf->hw;
9628 	i40e_status ret;
9629 	u16 port;
9630 	int i;
9631 
9632 	if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
9633 		return;
9634 
9635 	pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
9636 
9637 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
9638 		if (pf->pending_udp_bitmap & BIT_ULL(i)) {
9639 			pf->pending_udp_bitmap &= ~BIT_ULL(i);
9640 			port = pf->udp_ports[i].port;
9641 			if (port)
9642 				ret = i40e_aq_add_udp_tunnel(hw, port,
9643 							pf->udp_ports[i].type,
9644 							NULL, NULL);
9645 			else
9646 				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
9647 
9648 			if (ret) {
9649 				dev_info(&pf->pdev->dev,
9650 					 "%s %s port %d, index %d failed, err %s aq_err %s\n",
9651 					 i40e_tunnel_name(&pf->udp_ports[i]),
9652 					 port ? "add" : "delete",
9653 					 port, i,
9654 					 i40e_stat_str(&pf->hw, ret),
9655 					 i40e_aq_str(&pf->hw,
9656 						     pf->hw.aq.asq_last_status));
9657 				pf->udp_ports[i].port = 0;
9658 			}
9659 		}
9660 	}
9661 }
9662 
9663 /**
9664  * i40e_service_task - Run the driver's async subtasks
9665  * @work: pointer to work_struct containing our data
9666  **/
9667 static void i40e_service_task(struct work_struct *work)
9668 {
9669 	struct i40e_pf *pf = container_of(work,
9670 					  struct i40e_pf,
9671 					  service_task);
9672 	unsigned long start_time = jiffies;
9673 
9674 	/* don't bother with service tasks if a reset is in progress */
9675 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
9676 		return;
9677 
9678 	if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
9679 		return;
9680 
9681 	i40e_detect_recover_hung(pf);
9682 	i40e_sync_filters_subtask(pf);
9683 	i40e_reset_subtask(pf);
9684 	i40e_handle_mdd_event(pf);
9685 	i40e_vc_process_vflr_event(pf);
9686 	i40e_watchdog_subtask(pf);
9687 	i40e_fdir_reinit_subtask(pf);
9688 	if (pf->flags & I40E_FLAG_CLIENT_RESET) {
9689 		/* Client subtask will reopen next time through. */
9690 		i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], true);
9691 		pf->flags &= ~I40E_FLAG_CLIENT_RESET;
9692 	} else {
9693 		i40e_client_subtask(pf);
9694 		if (pf->flags & I40E_FLAG_CLIENT_L2_CHANGE) {
9695 			i40e_notify_client_of_l2_param_changes(
9696 							pf->vsi[pf->lan_vsi]);
9697 			pf->flags &= ~I40E_FLAG_CLIENT_L2_CHANGE;
9698 		}
9699 	}
9700 	i40e_sync_filters_subtask(pf);
9701 	i40e_sync_udp_filters_subtask(pf);
9702 	i40e_clean_adminq_subtask(pf);
9703 
9704 	/* flush memory to make sure state is correct before next watchdog */
9705 	smp_mb__before_atomic();
9706 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
9707 
9708 	/* If the tasks have taken longer than one timer cycle or there
9709 	 * is more work to be done, reschedule the service task now
9710 	 * rather than wait for the timer to tick again.
9711 	 */
9712 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
9713 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)		 ||
9714 	    test_bit(__I40E_MDD_EVENT_PENDING, pf->state)		 ||
9715 	    test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
9716 		i40e_service_event_schedule(pf);
9717 }
9718 
9719 /**
9720  * i40e_service_timer - timer callback
9721  * @data: pointer to PF struct
9722  **/
9723 static void i40e_service_timer(struct timer_list *t)
9724 {
9725 	struct i40e_pf *pf = from_timer(pf, t, service_timer);
9726 
9727 	mod_timer(&pf->service_timer,
9728 		  round_jiffies(jiffies + pf->service_timer_period));
9729 	i40e_service_event_schedule(pf);
9730 }
9731 
9732 /**
9733  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
9734  * @vsi: the VSI being configured
9735  **/
9736 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
9737 {
9738 	struct i40e_pf *pf = vsi->back;
9739 
9740 	switch (vsi->type) {
9741 	case I40E_VSI_MAIN:
9742 		vsi->alloc_queue_pairs = pf->num_lan_qps;
9743 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
9744 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
9745 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
9746 			vsi->num_q_vectors = pf->num_lan_msix;
9747 		else
9748 			vsi->num_q_vectors = 1;
9749 
9750 		break;
9751 
9752 	case I40E_VSI_FDIR:
9753 		vsi->alloc_queue_pairs = 1;
9754 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
9755 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
9756 		vsi->num_q_vectors = pf->num_fdsb_msix;
9757 		break;
9758 
9759 	case I40E_VSI_VMDQ2:
9760 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
9761 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
9762 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
9763 		vsi->num_q_vectors = pf->num_vmdq_msix;
9764 		break;
9765 
9766 	case I40E_VSI_SRIOV:
9767 		vsi->alloc_queue_pairs = pf->num_vf_qps;
9768 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
9769 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
9770 		break;
9771 
9772 	default:
9773 		WARN_ON(1);
9774 		return -ENODATA;
9775 	}
9776 
9777 	return 0;
9778 }
9779 
9780 /**
9781  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
9782  * @vsi: VSI pointer
9783  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
9784  *
9785  * On error: returns error code (negative)
9786  * On success: returns 0
9787  **/
9788 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
9789 {
9790 	struct i40e_ring **next_rings;
9791 	int size;
9792 	int ret = 0;
9793 
9794 	/* allocate memory for both Tx, XDP Tx and Rx ring pointers */
9795 	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs *
9796 	       (i40e_enabled_xdp_vsi(vsi) ? 3 : 2);
9797 	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
9798 	if (!vsi->tx_rings)
9799 		return -ENOMEM;
9800 	next_rings = vsi->tx_rings + vsi->alloc_queue_pairs;
9801 	if (i40e_enabled_xdp_vsi(vsi)) {
9802 		vsi->xdp_rings = next_rings;
9803 		next_rings += vsi->alloc_queue_pairs;
9804 	}
9805 	vsi->rx_rings = next_rings;
9806 
9807 	if (alloc_qvectors) {
9808 		/* allocate memory for q_vector pointers */
9809 		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
9810 		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
9811 		if (!vsi->q_vectors) {
9812 			ret = -ENOMEM;
9813 			goto err_vectors;
9814 		}
9815 	}
9816 	return ret;
9817 
9818 err_vectors:
9819 	kfree(vsi->tx_rings);
9820 	return ret;
9821 }
9822 
9823 /**
9824  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
9825  * @pf: board private structure
9826  * @type: type of VSI
9827  *
9828  * On error: returns error code (negative)
9829  * On success: returns vsi index in PF (positive)
9830  **/
9831 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
9832 {
9833 	int ret = -ENODEV;
9834 	struct i40e_vsi *vsi;
9835 	int vsi_idx;
9836 	int i;
9837 
9838 	/* Need to protect the allocation of the VSIs at the PF level */
9839 	mutex_lock(&pf->switch_mutex);
9840 
9841 	/* VSI list may be fragmented if VSI creation/destruction has
9842 	 * been happening.  We can afford to do a quick scan to look
9843 	 * for any free VSIs in the list.
9844 	 *
9845 	 * find next empty vsi slot, looping back around if necessary
9846 	 */
9847 	i = pf->next_vsi;
9848 	while (i < pf->num_alloc_vsi && pf->vsi[i])
9849 		i++;
9850 	if (i >= pf->num_alloc_vsi) {
9851 		i = 0;
9852 		while (i < pf->next_vsi && pf->vsi[i])
9853 			i++;
9854 	}
9855 
9856 	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
9857 		vsi_idx = i;             /* Found one! */
9858 	} else {
9859 		ret = -ENODEV;
9860 		goto unlock_pf;  /* out of VSI slots! */
9861 	}
9862 	pf->next_vsi = ++i;
9863 
9864 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
9865 	if (!vsi) {
9866 		ret = -ENOMEM;
9867 		goto unlock_pf;
9868 	}
9869 	vsi->type = type;
9870 	vsi->back = pf;
9871 	set_bit(__I40E_VSI_DOWN, vsi->state);
9872 	vsi->flags = 0;
9873 	vsi->idx = vsi_idx;
9874 	vsi->int_rate_limit = 0;
9875 	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
9876 				pf->rss_table_size : 64;
9877 	vsi->netdev_registered = false;
9878 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
9879 	hash_init(vsi->mac_filter_hash);
9880 	vsi->irqs_ready = false;
9881 
9882 	ret = i40e_set_num_rings_in_vsi(vsi);
9883 	if (ret)
9884 		goto err_rings;
9885 
9886 	ret = i40e_vsi_alloc_arrays(vsi, true);
9887 	if (ret)
9888 		goto err_rings;
9889 
9890 	/* Setup default MSIX irq handler for VSI */
9891 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
9892 
9893 	/* Initialize VSI lock */
9894 	spin_lock_init(&vsi->mac_filter_hash_lock);
9895 	pf->vsi[vsi_idx] = vsi;
9896 	ret = vsi_idx;
9897 	goto unlock_pf;
9898 
9899 err_rings:
9900 	pf->next_vsi = i - 1;
9901 	kfree(vsi);
9902 unlock_pf:
9903 	mutex_unlock(&pf->switch_mutex);
9904 	return ret;
9905 }
9906 
9907 /**
9908  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
9909  * @type: VSI pointer
9910  * @free_qvectors: a bool to specify if q_vectors need to be freed.
9911  *
9912  * On error: returns error code (negative)
9913  * On success: returns 0
9914  **/
9915 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
9916 {
9917 	/* free the ring and vector containers */
9918 	if (free_qvectors) {
9919 		kfree(vsi->q_vectors);
9920 		vsi->q_vectors = NULL;
9921 	}
9922 	kfree(vsi->tx_rings);
9923 	vsi->tx_rings = NULL;
9924 	vsi->rx_rings = NULL;
9925 	vsi->xdp_rings = NULL;
9926 }
9927 
9928 /**
9929  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
9930  * and lookup table
9931  * @vsi: Pointer to VSI structure
9932  */
9933 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
9934 {
9935 	if (!vsi)
9936 		return;
9937 
9938 	kfree(vsi->rss_hkey_user);
9939 	vsi->rss_hkey_user = NULL;
9940 
9941 	kfree(vsi->rss_lut_user);
9942 	vsi->rss_lut_user = NULL;
9943 }
9944 
9945 /**
9946  * i40e_vsi_clear - Deallocate the VSI provided
9947  * @vsi: the VSI being un-configured
9948  **/
9949 static int i40e_vsi_clear(struct i40e_vsi *vsi)
9950 {
9951 	struct i40e_pf *pf;
9952 
9953 	if (!vsi)
9954 		return 0;
9955 
9956 	if (!vsi->back)
9957 		goto free_vsi;
9958 	pf = vsi->back;
9959 
9960 	mutex_lock(&pf->switch_mutex);
9961 	if (!pf->vsi[vsi->idx]) {
9962 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
9963 			vsi->idx, vsi->idx, vsi, vsi->type);
9964 		goto unlock_vsi;
9965 	}
9966 
9967 	if (pf->vsi[vsi->idx] != vsi) {
9968 		dev_err(&pf->pdev->dev,
9969 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
9970 			pf->vsi[vsi->idx]->idx,
9971 			pf->vsi[vsi->idx],
9972 			pf->vsi[vsi->idx]->type,
9973 			vsi->idx, vsi, vsi->type);
9974 		goto unlock_vsi;
9975 	}
9976 
9977 	/* updates the PF for this cleared vsi */
9978 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9979 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
9980 
9981 	i40e_vsi_free_arrays(vsi, true);
9982 	i40e_clear_rss_config_user(vsi);
9983 
9984 	pf->vsi[vsi->idx] = NULL;
9985 	if (vsi->idx < pf->next_vsi)
9986 		pf->next_vsi = vsi->idx;
9987 
9988 unlock_vsi:
9989 	mutex_unlock(&pf->switch_mutex);
9990 free_vsi:
9991 	kfree(vsi);
9992 
9993 	return 0;
9994 }
9995 
9996 /**
9997  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
9998  * @vsi: the VSI being cleaned
9999  **/
10000 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
10001 {
10002 	int i;
10003 
10004 	if (vsi->tx_rings && vsi->tx_rings[0]) {
10005 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10006 			kfree_rcu(vsi->tx_rings[i], rcu);
10007 			vsi->tx_rings[i] = NULL;
10008 			vsi->rx_rings[i] = NULL;
10009 			if (vsi->xdp_rings)
10010 				vsi->xdp_rings[i] = NULL;
10011 		}
10012 	}
10013 }
10014 
10015 /**
10016  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
10017  * @vsi: the VSI being configured
10018  **/
10019 static int i40e_alloc_rings(struct i40e_vsi *vsi)
10020 {
10021 	int i, qpv = i40e_enabled_xdp_vsi(vsi) ? 3 : 2;
10022 	struct i40e_pf *pf = vsi->back;
10023 	struct i40e_ring *ring;
10024 
10025 	/* Set basic values in the rings to be used later during open() */
10026 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10027 		/* allocate space for both Tx and Rx in one shot */
10028 		ring = kcalloc(qpv, sizeof(struct i40e_ring), GFP_KERNEL);
10029 		if (!ring)
10030 			goto err_out;
10031 
10032 		ring->queue_index = i;
10033 		ring->reg_idx = vsi->base_queue + i;
10034 		ring->ring_active = false;
10035 		ring->vsi = vsi;
10036 		ring->netdev = vsi->netdev;
10037 		ring->dev = &pf->pdev->dev;
10038 		ring->count = vsi->num_desc;
10039 		ring->size = 0;
10040 		ring->dcb_tc = 0;
10041 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10042 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10043 		ring->tx_itr_setting = pf->tx_itr_default;
10044 		vsi->tx_rings[i] = ring++;
10045 
10046 		if (!i40e_enabled_xdp_vsi(vsi))
10047 			goto setup_rx;
10048 
10049 		ring->queue_index = vsi->alloc_queue_pairs + i;
10050 		ring->reg_idx = vsi->base_queue + ring->queue_index;
10051 		ring->ring_active = false;
10052 		ring->vsi = vsi;
10053 		ring->netdev = NULL;
10054 		ring->dev = &pf->pdev->dev;
10055 		ring->count = vsi->num_desc;
10056 		ring->size = 0;
10057 		ring->dcb_tc = 0;
10058 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10059 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10060 		set_ring_xdp(ring);
10061 		ring->tx_itr_setting = pf->tx_itr_default;
10062 		vsi->xdp_rings[i] = ring++;
10063 
10064 setup_rx:
10065 		ring->queue_index = i;
10066 		ring->reg_idx = vsi->base_queue + i;
10067 		ring->ring_active = false;
10068 		ring->vsi = vsi;
10069 		ring->netdev = vsi->netdev;
10070 		ring->dev = &pf->pdev->dev;
10071 		ring->count = vsi->num_desc;
10072 		ring->size = 0;
10073 		ring->dcb_tc = 0;
10074 		ring->rx_itr_setting = pf->rx_itr_default;
10075 		vsi->rx_rings[i] = ring;
10076 	}
10077 
10078 	return 0;
10079 
10080 err_out:
10081 	i40e_vsi_clear_rings(vsi);
10082 	return -ENOMEM;
10083 }
10084 
10085 /**
10086  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
10087  * @pf: board private structure
10088  * @vectors: the number of MSI-X vectors to request
10089  *
10090  * Returns the number of vectors reserved, or error
10091  **/
10092 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
10093 {
10094 	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
10095 					I40E_MIN_MSIX, vectors);
10096 	if (vectors < 0) {
10097 		dev_info(&pf->pdev->dev,
10098 			 "MSI-X vector reservation failed: %d\n", vectors);
10099 		vectors = 0;
10100 	}
10101 
10102 	return vectors;
10103 }
10104 
10105 /**
10106  * i40e_init_msix - Setup the MSIX capability
10107  * @pf: board private structure
10108  *
10109  * Work with the OS to set up the MSIX vectors needed.
10110  *
10111  * Returns the number of vectors reserved or negative on failure
10112  **/
10113 static int i40e_init_msix(struct i40e_pf *pf)
10114 {
10115 	struct i40e_hw *hw = &pf->hw;
10116 	int cpus, extra_vectors;
10117 	int vectors_left;
10118 	int v_budget, i;
10119 	int v_actual;
10120 	int iwarp_requested = 0;
10121 
10122 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
10123 		return -ENODEV;
10124 
10125 	/* The number of vectors we'll request will be comprised of:
10126 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
10127 	 *   - The number of LAN queue pairs
10128 	 *	- Queues being used for RSS.
10129 	 *		We don't need as many as max_rss_size vectors.
10130 	 *		use rss_size instead in the calculation since that
10131 	 *		is governed by number of cpus in the system.
10132 	 *	- assumes symmetric Tx/Rx pairing
10133 	 *   - The number of VMDq pairs
10134 	 *   - The CPU count within the NUMA node if iWARP is enabled
10135 	 * Once we count this up, try the request.
10136 	 *
10137 	 * If we can't get what we want, we'll simplify to nearly nothing
10138 	 * and try again.  If that still fails, we punt.
10139 	 */
10140 	vectors_left = hw->func_caps.num_msix_vectors;
10141 	v_budget = 0;
10142 
10143 	/* reserve one vector for miscellaneous handler */
10144 	if (vectors_left) {
10145 		v_budget++;
10146 		vectors_left--;
10147 	}
10148 
10149 	/* reserve some vectors for the main PF traffic queues. Initially we
10150 	 * only reserve at most 50% of the available vectors, in the case that
10151 	 * the number of online CPUs is large. This ensures that we can enable
10152 	 * extra features as well. Once we've enabled the other features, we
10153 	 * will use any remaining vectors to reach as close as we can to the
10154 	 * number of online CPUs.
10155 	 */
10156 	cpus = num_online_cpus();
10157 	pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
10158 	vectors_left -= pf->num_lan_msix;
10159 
10160 	/* reserve one vector for sideband flow director */
10161 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10162 		if (vectors_left) {
10163 			pf->num_fdsb_msix = 1;
10164 			v_budget++;
10165 			vectors_left--;
10166 		} else {
10167 			pf->num_fdsb_msix = 0;
10168 		}
10169 	}
10170 
10171 	/* can we reserve enough for iWARP? */
10172 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
10173 		iwarp_requested = pf->num_iwarp_msix;
10174 
10175 		if (!vectors_left)
10176 			pf->num_iwarp_msix = 0;
10177 		else if (vectors_left < pf->num_iwarp_msix)
10178 			pf->num_iwarp_msix = 1;
10179 		v_budget += pf->num_iwarp_msix;
10180 		vectors_left -= pf->num_iwarp_msix;
10181 	}
10182 
10183 	/* any vectors left over go for VMDq support */
10184 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
10185 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
10186 		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
10187 
10188 		if (!vectors_left) {
10189 			pf->num_vmdq_msix = 0;
10190 			pf->num_vmdq_qps = 0;
10191 		} else {
10192 			/* if we're short on vectors for what's desired, we limit
10193 			 * the queues per vmdq.  If this is still more than are
10194 			 * available, the user will need to change the number of
10195 			 * queues/vectors used by the PF later with the ethtool
10196 			 * channels command
10197 			 */
10198 			if (vmdq_vecs < vmdq_vecs_wanted)
10199 				pf->num_vmdq_qps = 1;
10200 			pf->num_vmdq_msix = pf->num_vmdq_qps;
10201 
10202 			v_budget += vmdq_vecs;
10203 			vectors_left -= vmdq_vecs;
10204 		}
10205 	}
10206 
10207 	/* On systems with a large number of SMP cores, we previously limited
10208 	 * the number of vectors for num_lan_msix to be at most 50% of the
10209 	 * available vectors, to allow for other features. Now, we add back
10210 	 * the remaining vectors. However, we ensure that the total
10211 	 * num_lan_msix will not exceed num_online_cpus(). To do this, we
10212 	 * calculate the number of vectors we can add without going over the
10213 	 * cap of CPUs. For systems with a small number of CPUs this will be
10214 	 * zero.
10215 	 */
10216 	extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
10217 	pf->num_lan_msix += extra_vectors;
10218 	vectors_left -= extra_vectors;
10219 
10220 	WARN(vectors_left < 0,
10221 	     "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
10222 
10223 	v_budget += pf->num_lan_msix;
10224 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
10225 				   GFP_KERNEL);
10226 	if (!pf->msix_entries)
10227 		return -ENOMEM;
10228 
10229 	for (i = 0; i < v_budget; i++)
10230 		pf->msix_entries[i].entry = i;
10231 	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
10232 
10233 	if (v_actual < I40E_MIN_MSIX) {
10234 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
10235 		kfree(pf->msix_entries);
10236 		pf->msix_entries = NULL;
10237 		pci_disable_msix(pf->pdev);
10238 		return -ENODEV;
10239 
10240 	} else if (v_actual == I40E_MIN_MSIX) {
10241 		/* Adjust for minimal MSIX use */
10242 		pf->num_vmdq_vsis = 0;
10243 		pf->num_vmdq_qps = 0;
10244 		pf->num_lan_qps = 1;
10245 		pf->num_lan_msix = 1;
10246 
10247 	} else if (v_actual != v_budget) {
10248 		/* If we have limited resources, we will start with no vectors
10249 		 * for the special features and then allocate vectors to some
10250 		 * of these features based on the policy and at the end disable
10251 		 * the features that did not get any vectors.
10252 		 */
10253 		int vec;
10254 
10255 		dev_info(&pf->pdev->dev,
10256 			 "MSI-X vector limit reached with %d, wanted %d, attempting to redistribute vectors\n",
10257 			 v_actual, v_budget);
10258 		/* reserve the misc vector */
10259 		vec = v_actual - 1;
10260 
10261 		/* Scale vector usage down */
10262 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
10263 		pf->num_vmdq_vsis = 1;
10264 		pf->num_vmdq_qps = 1;
10265 
10266 		/* partition out the remaining vectors */
10267 		switch (vec) {
10268 		case 2:
10269 			pf->num_lan_msix = 1;
10270 			break;
10271 		case 3:
10272 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
10273 				pf->num_lan_msix = 1;
10274 				pf->num_iwarp_msix = 1;
10275 			} else {
10276 				pf->num_lan_msix = 2;
10277 			}
10278 			break;
10279 		default:
10280 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
10281 				pf->num_iwarp_msix = min_t(int, (vec / 3),
10282 						 iwarp_requested);
10283 				pf->num_vmdq_vsis = min_t(int, (vec / 3),
10284 						  I40E_DEFAULT_NUM_VMDQ_VSI);
10285 			} else {
10286 				pf->num_vmdq_vsis = min_t(int, (vec / 2),
10287 						  I40E_DEFAULT_NUM_VMDQ_VSI);
10288 			}
10289 			if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10290 				pf->num_fdsb_msix = 1;
10291 				vec--;
10292 			}
10293 			pf->num_lan_msix = min_t(int,
10294 			       (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
10295 							      pf->num_lan_msix);
10296 			pf->num_lan_qps = pf->num_lan_msix;
10297 			break;
10298 		}
10299 	}
10300 
10301 	if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
10302 	    (pf->num_fdsb_msix == 0)) {
10303 		dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
10304 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10305 		pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
10306 	}
10307 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10308 	    (pf->num_vmdq_msix == 0)) {
10309 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
10310 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
10311 	}
10312 
10313 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
10314 	    (pf->num_iwarp_msix == 0)) {
10315 		dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
10316 		pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
10317 	}
10318 	i40e_debug(&pf->hw, I40E_DEBUG_INIT,
10319 		   "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
10320 		   pf->num_lan_msix,
10321 		   pf->num_vmdq_msix * pf->num_vmdq_vsis,
10322 		   pf->num_fdsb_msix,
10323 		   pf->num_iwarp_msix);
10324 
10325 	return v_actual;
10326 }
10327 
10328 /**
10329  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
10330  * @vsi: the VSI being configured
10331  * @v_idx: index of the vector in the vsi struct
10332  * @cpu: cpu to be used on affinity_mask
10333  *
10334  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
10335  **/
10336 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
10337 {
10338 	struct i40e_q_vector *q_vector;
10339 
10340 	/* allocate q_vector */
10341 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
10342 	if (!q_vector)
10343 		return -ENOMEM;
10344 
10345 	q_vector->vsi = vsi;
10346 	q_vector->v_idx = v_idx;
10347 	cpumask_copy(&q_vector->affinity_mask, cpu_possible_mask);
10348 
10349 	if (vsi->netdev)
10350 		netif_napi_add(vsi->netdev, &q_vector->napi,
10351 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
10352 
10353 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
10354 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
10355 
10356 	/* tie q_vector and vsi together */
10357 	vsi->q_vectors[v_idx] = q_vector;
10358 
10359 	return 0;
10360 }
10361 
10362 /**
10363  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
10364  * @vsi: the VSI being configured
10365  *
10366  * We allocate one q_vector per queue interrupt.  If allocation fails we
10367  * return -ENOMEM.
10368  **/
10369 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
10370 {
10371 	struct i40e_pf *pf = vsi->back;
10372 	int err, v_idx, num_q_vectors, current_cpu;
10373 
10374 	/* if not MSIX, give the one vector only to the LAN VSI */
10375 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
10376 		num_q_vectors = vsi->num_q_vectors;
10377 	else if (vsi == pf->vsi[pf->lan_vsi])
10378 		num_q_vectors = 1;
10379 	else
10380 		return -EINVAL;
10381 
10382 	current_cpu = cpumask_first(cpu_online_mask);
10383 
10384 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
10385 		err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
10386 		if (err)
10387 			goto err_out;
10388 		current_cpu = cpumask_next(current_cpu, cpu_online_mask);
10389 		if (unlikely(current_cpu >= nr_cpu_ids))
10390 			current_cpu = cpumask_first(cpu_online_mask);
10391 	}
10392 
10393 	return 0;
10394 
10395 err_out:
10396 	while (v_idx--)
10397 		i40e_free_q_vector(vsi, v_idx);
10398 
10399 	return err;
10400 }
10401 
10402 /**
10403  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
10404  * @pf: board private structure to initialize
10405  **/
10406 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
10407 {
10408 	int vectors = 0;
10409 	ssize_t size;
10410 
10411 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
10412 		vectors = i40e_init_msix(pf);
10413 		if (vectors < 0) {
10414 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
10415 				       I40E_FLAG_IWARP_ENABLED	|
10416 				       I40E_FLAG_RSS_ENABLED	|
10417 				       I40E_FLAG_DCB_CAPABLE	|
10418 				       I40E_FLAG_DCB_ENABLED	|
10419 				       I40E_FLAG_SRIOV_ENABLED	|
10420 				       I40E_FLAG_FD_SB_ENABLED	|
10421 				       I40E_FLAG_FD_ATR_ENABLED	|
10422 				       I40E_FLAG_VMDQ_ENABLED);
10423 			pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
10424 
10425 			/* rework the queue expectations without MSIX */
10426 			i40e_determine_queue_usage(pf);
10427 		}
10428 	}
10429 
10430 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10431 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
10432 		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
10433 		vectors = pci_enable_msi(pf->pdev);
10434 		if (vectors < 0) {
10435 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
10436 				 vectors);
10437 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
10438 		}
10439 		vectors = 1;  /* one MSI or Legacy vector */
10440 	}
10441 
10442 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
10443 		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
10444 
10445 	/* set up vector assignment tracking */
10446 	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
10447 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
10448 	if (!pf->irq_pile) {
10449 		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
10450 		return -ENOMEM;
10451 	}
10452 	pf->irq_pile->num_entries = vectors;
10453 	pf->irq_pile->search_hint = 0;
10454 
10455 	/* track first vector for misc interrupts, ignore return */
10456 	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
10457 
10458 	return 0;
10459 }
10460 
10461 /**
10462  * i40e_restore_interrupt_scheme - Restore the interrupt scheme
10463  * @pf: private board data structure
10464  *
10465  * Restore the interrupt scheme that was cleared when we suspended the
10466  * device. This should be called during resume to re-allocate the q_vectors
10467  * and reacquire IRQs.
10468  */
10469 static int i40e_restore_interrupt_scheme(struct i40e_pf *pf)
10470 {
10471 	int err, i;
10472 
10473 	/* We cleared the MSI and MSI-X flags when disabling the old interrupt
10474 	 * scheme. We need to re-enabled them here in order to attempt to
10475 	 * re-acquire the MSI or MSI-X vectors
10476 	 */
10477 	pf->flags |= (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
10478 
10479 	err = i40e_init_interrupt_scheme(pf);
10480 	if (err)
10481 		return err;
10482 
10483 	/* Now that we've re-acquired IRQs, we need to remap the vectors and
10484 	 * rings together again.
10485 	 */
10486 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10487 		if (pf->vsi[i]) {
10488 			err = i40e_vsi_alloc_q_vectors(pf->vsi[i]);
10489 			if (err)
10490 				goto err_unwind;
10491 			i40e_vsi_map_rings_to_vectors(pf->vsi[i]);
10492 		}
10493 	}
10494 
10495 	err = i40e_setup_misc_vector(pf);
10496 	if (err)
10497 		goto err_unwind;
10498 
10499 	return 0;
10500 
10501 err_unwind:
10502 	while (i--) {
10503 		if (pf->vsi[i])
10504 			i40e_vsi_free_q_vectors(pf->vsi[i]);
10505 	}
10506 
10507 	return err;
10508 }
10509 
10510 /**
10511  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
10512  * @pf: board private structure
10513  *
10514  * This sets up the handler for MSIX 0, which is used to manage the
10515  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
10516  * when in MSI or Legacy interrupt mode.
10517  **/
10518 static int i40e_setup_misc_vector(struct i40e_pf *pf)
10519 {
10520 	struct i40e_hw *hw = &pf->hw;
10521 	int err = 0;
10522 
10523 	/* Only request the IRQ once, the first time through. */
10524 	if (!test_and_set_bit(__I40E_MISC_IRQ_REQUESTED, pf->state)) {
10525 		err = request_irq(pf->msix_entries[0].vector,
10526 				  i40e_intr, 0, pf->int_name, pf);
10527 		if (err) {
10528 			clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
10529 			dev_info(&pf->pdev->dev,
10530 				 "request_irq for %s failed: %d\n",
10531 				 pf->int_name, err);
10532 			return -EFAULT;
10533 		}
10534 	}
10535 
10536 	i40e_enable_misc_int_causes(pf);
10537 
10538 	/* associate no queues to the misc vector */
10539 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
10540 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
10541 
10542 	i40e_flush(hw);
10543 
10544 	i40e_irq_dynamic_enable_icr0(pf);
10545 
10546 	return err;
10547 }
10548 
10549 /**
10550  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
10551  * @vsi: Pointer to vsi structure
10552  * @seed: Buffter to store the hash keys
10553  * @lut: Buffer to store the lookup table entries
10554  * @lut_size: Size of buffer to store the lookup table entries
10555  *
10556  * Return 0 on success, negative on failure
10557  */
10558 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
10559 			   u8 *lut, u16 lut_size)
10560 {
10561 	struct i40e_pf *pf = vsi->back;
10562 	struct i40e_hw *hw = &pf->hw;
10563 	int ret = 0;
10564 
10565 	if (seed) {
10566 		ret = i40e_aq_get_rss_key(hw, vsi->id,
10567 			(struct i40e_aqc_get_set_rss_key_data *)seed);
10568 		if (ret) {
10569 			dev_info(&pf->pdev->dev,
10570 				 "Cannot get RSS key, err %s aq_err %s\n",
10571 				 i40e_stat_str(&pf->hw, ret),
10572 				 i40e_aq_str(&pf->hw,
10573 					     pf->hw.aq.asq_last_status));
10574 			return ret;
10575 		}
10576 	}
10577 
10578 	if (lut) {
10579 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
10580 
10581 		ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
10582 		if (ret) {
10583 			dev_info(&pf->pdev->dev,
10584 				 "Cannot get RSS lut, err %s aq_err %s\n",
10585 				 i40e_stat_str(&pf->hw, ret),
10586 				 i40e_aq_str(&pf->hw,
10587 					     pf->hw.aq.asq_last_status));
10588 			return ret;
10589 		}
10590 	}
10591 
10592 	return ret;
10593 }
10594 
10595 /**
10596  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
10597  * @vsi: Pointer to vsi structure
10598  * @seed: RSS hash seed
10599  * @lut: Lookup table
10600  * @lut_size: Lookup table size
10601  *
10602  * Returns 0 on success, negative on failure
10603  **/
10604 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
10605 			       const u8 *lut, u16 lut_size)
10606 {
10607 	struct i40e_pf *pf = vsi->back;
10608 	struct i40e_hw *hw = &pf->hw;
10609 	u16 vf_id = vsi->vf_id;
10610 	u8 i;
10611 
10612 	/* Fill out hash function seed */
10613 	if (seed) {
10614 		u32 *seed_dw = (u32 *)seed;
10615 
10616 		if (vsi->type == I40E_VSI_MAIN) {
10617 			for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
10618 				wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
10619 		} else if (vsi->type == I40E_VSI_SRIOV) {
10620 			for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
10621 				wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
10622 		} else {
10623 			dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
10624 		}
10625 	}
10626 
10627 	if (lut) {
10628 		u32 *lut_dw = (u32 *)lut;
10629 
10630 		if (vsi->type == I40E_VSI_MAIN) {
10631 			if (lut_size != I40E_HLUT_ARRAY_SIZE)
10632 				return -EINVAL;
10633 			for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
10634 				wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
10635 		} else if (vsi->type == I40E_VSI_SRIOV) {
10636 			if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
10637 				return -EINVAL;
10638 			for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
10639 				wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
10640 		} else {
10641 			dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
10642 		}
10643 	}
10644 	i40e_flush(hw);
10645 
10646 	return 0;
10647 }
10648 
10649 /**
10650  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
10651  * @vsi: Pointer to VSI structure
10652  * @seed: Buffer to store the keys
10653  * @lut: Buffer to store the lookup table entries
10654  * @lut_size: Size of buffer to store the lookup table entries
10655  *
10656  * Returns 0 on success, negative on failure
10657  */
10658 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
10659 			    u8 *lut, u16 lut_size)
10660 {
10661 	struct i40e_pf *pf = vsi->back;
10662 	struct i40e_hw *hw = &pf->hw;
10663 	u16 i;
10664 
10665 	if (seed) {
10666 		u32 *seed_dw = (u32 *)seed;
10667 
10668 		for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
10669 			seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
10670 	}
10671 	if (lut) {
10672 		u32 *lut_dw = (u32 *)lut;
10673 
10674 		if (lut_size != I40E_HLUT_ARRAY_SIZE)
10675 			return -EINVAL;
10676 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
10677 			lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
10678 	}
10679 
10680 	return 0;
10681 }
10682 
10683 /**
10684  * i40e_config_rss - Configure RSS keys and lut
10685  * @vsi: Pointer to VSI structure
10686  * @seed: RSS hash seed
10687  * @lut: Lookup table
10688  * @lut_size: Lookup table size
10689  *
10690  * Returns 0 on success, negative on failure
10691  */
10692 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
10693 {
10694 	struct i40e_pf *pf = vsi->back;
10695 
10696 	if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
10697 		return i40e_config_rss_aq(vsi, seed, lut, lut_size);
10698 	else
10699 		return i40e_config_rss_reg(vsi, seed, lut, lut_size);
10700 }
10701 
10702 /**
10703  * i40e_get_rss - Get RSS keys and lut
10704  * @vsi: Pointer to VSI structure
10705  * @seed: Buffer to store the keys
10706  * @lut: Buffer to store the lookup table entries
10707  * lut_size: Size of buffer to store the lookup table entries
10708  *
10709  * Returns 0 on success, negative on failure
10710  */
10711 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
10712 {
10713 	struct i40e_pf *pf = vsi->back;
10714 
10715 	if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
10716 		return i40e_get_rss_aq(vsi, seed, lut, lut_size);
10717 	else
10718 		return i40e_get_rss_reg(vsi, seed, lut, lut_size);
10719 }
10720 
10721 /**
10722  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
10723  * @pf: Pointer to board private structure
10724  * @lut: Lookup table
10725  * @rss_table_size: Lookup table size
10726  * @rss_size: Range of queue number for hashing
10727  */
10728 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
10729 		       u16 rss_table_size, u16 rss_size)
10730 {
10731 	u16 i;
10732 
10733 	for (i = 0; i < rss_table_size; i++)
10734 		lut[i] = i % rss_size;
10735 }
10736 
10737 /**
10738  * i40e_pf_config_rss - Prepare for RSS if used
10739  * @pf: board private structure
10740  **/
10741 static int i40e_pf_config_rss(struct i40e_pf *pf)
10742 {
10743 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
10744 	u8 seed[I40E_HKEY_ARRAY_SIZE];
10745 	u8 *lut;
10746 	struct i40e_hw *hw = &pf->hw;
10747 	u32 reg_val;
10748 	u64 hena;
10749 	int ret;
10750 
10751 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
10752 	hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
10753 		((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
10754 	hena |= i40e_pf_get_default_rss_hena(pf);
10755 
10756 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
10757 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
10758 
10759 	/* Determine the RSS table size based on the hardware capabilities */
10760 	reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
10761 	reg_val = (pf->rss_table_size == 512) ?
10762 			(reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
10763 			(reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
10764 	i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
10765 
10766 	/* Determine the RSS size of the VSI */
10767 	if (!vsi->rss_size) {
10768 		u16 qcount;
10769 
10770 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
10771 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
10772 	}
10773 	if (!vsi->rss_size)
10774 		return -EINVAL;
10775 
10776 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
10777 	if (!lut)
10778 		return -ENOMEM;
10779 
10780 	/* Use user configured lut if there is one, otherwise use default */
10781 	if (vsi->rss_lut_user)
10782 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
10783 	else
10784 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
10785 
10786 	/* Use user configured hash key if there is one, otherwise
10787 	 * use default.
10788 	 */
10789 	if (vsi->rss_hkey_user)
10790 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
10791 	else
10792 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
10793 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
10794 	kfree(lut);
10795 
10796 	return ret;
10797 }
10798 
10799 /**
10800  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
10801  * @pf: board private structure
10802  * @queue_count: the requested queue count for rss.
10803  *
10804  * returns 0 if rss is not enabled, if enabled returns the final rss queue
10805  * count which may be different from the requested queue count.
10806  * Note: expects to be called while under rtnl_lock()
10807  **/
10808 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
10809 {
10810 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
10811 	int new_rss_size;
10812 
10813 	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
10814 		return 0;
10815 
10816 	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
10817 
10818 	if (queue_count != vsi->num_queue_pairs) {
10819 		u16 qcount;
10820 
10821 		vsi->req_queue_pairs = queue_count;
10822 		i40e_prep_for_reset(pf, true);
10823 
10824 		pf->alloc_rss_size = new_rss_size;
10825 
10826 		i40e_reset_and_rebuild(pf, true, true);
10827 
10828 		/* Discard the user configured hash keys and lut, if less
10829 		 * queues are enabled.
10830 		 */
10831 		if (queue_count < vsi->rss_size) {
10832 			i40e_clear_rss_config_user(vsi);
10833 			dev_dbg(&pf->pdev->dev,
10834 				"discard user configured hash keys and lut\n");
10835 		}
10836 
10837 		/* Reset vsi->rss_size, as number of enabled queues changed */
10838 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
10839 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
10840 
10841 		i40e_pf_config_rss(pf);
10842 	}
10843 	dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
10844 		 vsi->req_queue_pairs, pf->rss_size_max);
10845 	return pf->alloc_rss_size;
10846 }
10847 
10848 /**
10849  * i40e_get_partition_bw_setting - Retrieve BW settings for this PF partition
10850  * @pf: board private structure
10851  **/
10852 i40e_status i40e_get_partition_bw_setting(struct i40e_pf *pf)
10853 {
10854 	i40e_status status;
10855 	bool min_valid, max_valid;
10856 	u32 max_bw, min_bw;
10857 
10858 	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
10859 					   &min_valid, &max_valid);
10860 
10861 	if (!status) {
10862 		if (min_valid)
10863 			pf->min_bw = min_bw;
10864 		if (max_valid)
10865 			pf->max_bw = max_bw;
10866 	}
10867 
10868 	return status;
10869 }
10870 
10871 /**
10872  * i40e_set_partition_bw_setting - Set BW settings for this PF partition
10873  * @pf: board private structure
10874  **/
10875 i40e_status i40e_set_partition_bw_setting(struct i40e_pf *pf)
10876 {
10877 	struct i40e_aqc_configure_partition_bw_data bw_data;
10878 	i40e_status status;
10879 
10880 	/* Set the valid bit for this PF */
10881 	bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
10882 	bw_data.max_bw[pf->hw.pf_id] = pf->max_bw & I40E_ALT_BW_VALUE_MASK;
10883 	bw_data.min_bw[pf->hw.pf_id] = pf->min_bw & I40E_ALT_BW_VALUE_MASK;
10884 
10885 	/* Set the new bandwidths */
10886 	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
10887 
10888 	return status;
10889 }
10890 
10891 /**
10892  * i40e_commit_partition_bw_setting - Commit BW settings for this PF partition
10893  * @pf: board private structure
10894  **/
10895 i40e_status i40e_commit_partition_bw_setting(struct i40e_pf *pf)
10896 {
10897 	/* Commit temporary BW setting to permanent NVM image */
10898 	enum i40e_admin_queue_err last_aq_status;
10899 	i40e_status ret;
10900 	u16 nvm_word;
10901 
10902 	if (pf->hw.partition_id != 1) {
10903 		dev_info(&pf->pdev->dev,
10904 			 "Commit BW only works on partition 1! This is partition %d",
10905 			 pf->hw.partition_id);
10906 		ret = I40E_NOT_SUPPORTED;
10907 		goto bw_commit_out;
10908 	}
10909 
10910 	/* Acquire NVM for read access */
10911 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
10912 	last_aq_status = pf->hw.aq.asq_last_status;
10913 	if (ret) {
10914 		dev_info(&pf->pdev->dev,
10915 			 "Cannot acquire NVM for read access, err %s aq_err %s\n",
10916 			 i40e_stat_str(&pf->hw, ret),
10917 			 i40e_aq_str(&pf->hw, last_aq_status));
10918 		goto bw_commit_out;
10919 	}
10920 
10921 	/* Read word 0x10 of NVM - SW compatibility word 1 */
10922 	ret = i40e_aq_read_nvm(&pf->hw,
10923 			       I40E_SR_NVM_CONTROL_WORD,
10924 			       0x10, sizeof(nvm_word), &nvm_word,
10925 			       false, NULL);
10926 	/* Save off last admin queue command status before releasing
10927 	 * the NVM
10928 	 */
10929 	last_aq_status = pf->hw.aq.asq_last_status;
10930 	i40e_release_nvm(&pf->hw);
10931 	if (ret) {
10932 		dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
10933 			 i40e_stat_str(&pf->hw, ret),
10934 			 i40e_aq_str(&pf->hw, last_aq_status));
10935 		goto bw_commit_out;
10936 	}
10937 
10938 	/* Wait a bit for NVM release to complete */
10939 	msleep(50);
10940 
10941 	/* Acquire NVM for write access */
10942 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
10943 	last_aq_status = pf->hw.aq.asq_last_status;
10944 	if (ret) {
10945 		dev_info(&pf->pdev->dev,
10946 			 "Cannot acquire NVM for write access, err %s aq_err %s\n",
10947 			 i40e_stat_str(&pf->hw, ret),
10948 			 i40e_aq_str(&pf->hw, last_aq_status));
10949 		goto bw_commit_out;
10950 	}
10951 	/* Write it back out unchanged to initiate update NVM,
10952 	 * which will force a write of the shadow (alt) RAM to
10953 	 * the NVM - thus storing the bandwidth values permanently.
10954 	 */
10955 	ret = i40e_aq_update_nvm(&pf->hw,
10956 				 I40E_SR_NVM_CONTROL_WORD,
10957 				 0x10, sizeof(nvm_word),
10958 				 &nvm_word, true, NULL);
10959 	/* Save off last admin queue command status before releasing
10960 	 * the NVM
10961 	 */
10962 	last_aq_status = pf->hw.aq.asq_last_status;
10963 	i40e_release_nvm(&pf->hw);
10964 	if (ret)
10965 		dev_info(&pf->pdev->dev,
10966 			 "BW settings NOT SAVED, err %s aq_err %s\n",
10967 			 i40e_stat_str(&pf->hw, ret),
10968 			 i40e_aq_str(&pf->hw, last_aq_status));
10969 bw_commit_out:
10970 
10971 	return ret;
10972 }
10973 
10974 /**
10975  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
10976  * @pf: board private structure to initialize
10977  *
10978  * i40e_sw_init initializes the Adapter private data structure.
10979  * Fields are initialized based on PCI device information and
10980  * OS network device settings (MTU size).
10981  **/
10982 static int i40e_sw_init(struct i40e_pf *pf)
10983 {
10984 	int err = 0;
10985 	int size;
10986 
10987 	/* Set default capability flags */
10988 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
10989 		    I40E_FLAG_MSI_ENABLED     |
10990 		    I40E_FLAG_MSIX_ENABLED;
10991 
10992 	/* Set default ITR */
10993 	pf->rx_itr_default = I40E_ITR_RX_DEF;
10994 	pf->tx_itr_default = I40E_ITR_TX_DEF;
10995 
10996 	/* Depending on PF configurations, it is possible that the RSS
10997 	 * maximum might end up larger than the available queues
10998 	 */
10999 	pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
11000 	pf->alloc_rss_size = 1;
11001 	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
11002 	pf->rss_size_max = min_t(int, pf->rss_size_max,
11003 				 pf->hw.func_caps.num_tx_qp);
11004 	if (pf->hw.func_caps.rss) {
11005 		pf->flags |= I40E_FLAG_RSS_ENABLED;
11006 		pf->alloc_rss_size = min_t(int, pf->rss_size_max,
11007 					   num_online_cpus());
11008 	}
11009 
11010 	/* MFP mode enabled */
11011 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
11012 		pf->flags |= I40E_FLAG_MFP_ENABLED;
11013 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
11014 		if (i40e_get_partition_bw_setting(pf)) {
11015 			dev_warn(&pf->pdev->dev,
11016 				 "Could not get partition bw settings\n");
11017 		} else {
11018 			dev_info(&pf->pdev->dev,
11019 				 "Partition BW Min = %8.8x, Max = %8.8x\n",
11020 				 pf->min_bw, pf->max_bw);
11021 
11022 			/* nudge the Tx scheduler */
11023 			i40e_set_partition_bw_setting(pf);
11024 		}
11025 	}
11026 
11027 	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
11028 	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
11029 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
11030 		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
11031 		if (pf->flags & I40E_FLAG_MFP_ENABLED &&
11032 		    pf->hw.num_partitions > 1)
11033 			dev_info(&pf->pdev->dev,
11034 				 "Flow Director Sideband mode Disabled in MFP mode\n");
11035 		else
11036 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
11037 		pf->fdir_pf_filter_count =
11038 				 pf->hw.func_caps.fd_filters_guaranteed;
11039 		pf->hw.fdir_shared_filter_count =
11040 				 pf->hw.func_caps.fd_filters_best_effort;
11041 	}
11042 
11043 	if (pf->hw.mac.type == I40E_MAC_X722) {
11044 		pf->hw_features |= (I40E_HW_RSS_AQ_CAPABLE |
11045 				    I40E_HW_128_QP_RSS_CAPABLE |
11046 				    I40E_HW_ATR_EVICT_CAPABLE |
11047 				    I40E_HW_WB_ON_ITR_CAPABLE |
11048 				    I40E_HW_MULTIPLE_TCP_UDP_RSS_PCTYPE |
11049 				    I40E_HW_NO_PCI_LINK_CHECK |
11050 				    I40E_HW_USE_SET_LLDP_MIB |
11051 				    I40E_HW_GENEVE_OFFLOAD_CAPABLE |
11052 				    I40E_HW_PTP_L4_CAPABLE |
11053 				    I40E_HW_WOL_MC_MAGIC_PKT_WAKE |
11054 				    I40E_HW_OUTER_UDP_CSUM_CAPABLE);
11055 
11056 #define I40E_FDEVICT_PCTYPE_DEFAULT 0xc03
11057 		if (rd32(&pf->hw, I40E_GLQF_FDEVICTENA(1)) !=
11058 		    I40E_FDEVICT_PCTYPE_DEFAULT) {
11059 			dev_warn(&pf->pdev->dev,
11060 				 "FD EVICT PCTYPES are not right, disable FD HW EVICT\n");
11061 			pf->hw_features &= ~I40E_HW_ATR_EVICT_CAPABLE;
11062 		}
11063 	} else if ((pf->hw.aq.api_maj_ver > 1) ||
11064 		   ((pf->hw.aq.api_maj_ver == 1) &&
11065 		    (pf->hw.aq.api_min_ver > 4))) {
11066 		/* Supported in FW API version higher than 1.4 */
11067 		pf->hw_features |= I40E_HW_GENEVE_OFFLOAD_CAPABLE;
11068 	}
11069 
11070 	/* Enable HW ATR eviction if possible */
11071 	if (pf->hw_features & I40E_HW_ATR_EVICT_CAPABLE)
11072 		pf->flags |= I40E_FLAG_HW_ATR_EVICT_ENABLED;
11073 
11074 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11075 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
11076 	    (pf->hw.aq.fw_maj_ver < 4))) {
11077 		pf->hw_features |= I40E_HW_RESTART_AUTONEG;
11078 		/* No DCB support  for FW < v4.33 */
11079 		pf->hw_features |= I40E_HW_NO_DCB_SUPPORT;
11080 	}
11081 
11082 	/* Disable FW LLDP if FW < v4.3 */
11083 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11084 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
11085 	    (pf->hw.aq.fw_maj_ver < 4)))
11086 		pf->hw_features |= I40E_HW_STOP_FW_LLDP;
11087 
11088 	/* Use the FW Set LLDP MIB API if FW > v4.40 */
11089 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11090 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
11091 	    (pf->hw.aq.fw_maj_ver >= 5)))
11092 		pf->hw_features |= I40E_HW_USE_SET_LLDP_MIB;
11093 
11094 	/* Enable PTP L4 if FW > v6.0 */
11095 	if (pf->hw.mac.type == I40E_MAC_XL710 &&
11096 	    pf->hw.aq.fw_maj_ver >= 6)
11097 		pf->hw_features |= I40E_HW_PTP_L4_CAPABLE;
11098 
11099 	if (pf->hw.func_caps.vmdq) {
11100 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
11101 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
11102 		pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
11103 	}
11104 
11105 	if (pf->hw.func_caps.iwarp) {
11106 		pf->flags |= I40E_FLAG_IWARP_ENABLED;
11107 		/* IWARP needs one extra vector for CQP just like MISC.*/
11108 		pf->num_iwarp_msix = (int)num_online_cpus() + 1;
11109 	}
11110 
11111 #ifdef CONFIG_PCI_IOV
11112 	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
11113 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
11114 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
11115 		pf->num_req_vfs = min_t(int,
11116 					pf->hw.func_caps.num_vfs,
11117 					I40E_MAX_VF_COUNT);
11118 	}
11119 #endif /* CONFIG_PCI_IOV */
11120 	pf->eeprom_version = 0xDEAD;
11121 	pf->lan_veb = I40E_NO_VEB;
11122 	pf->lan_vsi = I40E_NO_VSI;
11123 
11124 	/* By default FW has this off for performance reasons */
11125 	pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
11126 
11127 	/* set up queue assignment tracking */
11128 	size = sizeof(struct i40e_lump_tracking)
11129 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
11130 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
11131 	if (!pf->qp_pile) {
11132 		err = -ENOMEM;
11133 		goto sw_init_done;
11134 	}
11135 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
11136 	pf->qp_pile->search_hint = 0;
11137 
11138 	pf->tx_timeout_recovery_level = 1;
11139 
11140 	mutex_init(&pf->switch_mutex);
11141 
11142 sw_init_done:
11143 	return err;
11144 }
11145 
11146 /**
11147  * i40e_set_ntuple - set the ntuple feature flag and take action
11148  * @pf: board private structure to initialize
11149  * @features: the feature set that the stack is suggesting
11150  *
11151  * returns a bool to indicate if reset needs to happen
11152  **/
11153 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
11154 {
11155 	bool need_reset = false;
11156 
11157 	/* Check if Flow Director n-tuple support was enabled or disabled.  If
11158 	 * the state changed, we need to reset.
11159 	 */
11160 	if (features & NETIF_F_NTUPLE) {
11161 		/* Enable filters and mark for reset */
11162 		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
11163 			need_reset = true;
11164 		/* enable FD_SB only if there is MSI-X vector and no cloud
11165 		 * filters exist
11166 		 */
11167 		if (pf->num_fdsb_msix > 0 && !pf->num_cloud_filters) {
11168 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
11169 			pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
11170 		}
11171 	} else {
11172 		/* turn off filters, mark for reset and clear SW filter list */
11173 		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11174 			need_reset = true;
11175 			i40e_fdir_filter_exit(pf);
11176 		}
11177 		pf->flags &= ~(I40E_FLAG_FD_SB_ENABLED |
11178 			       I40E_FLAG_FD_SB_AUTO_DISABLED);
11179 		pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
11180 
11181 		/* reset fd counters */
11182 		pf->fd_add_err = 0;
11183 		pf->fd_atr_cnt = 0;
11184 		/* if ATR was auto disabled it can be re-enabled. */
11185 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
11186 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
11187 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
11188 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
11189 				dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
11190 		}
11191 	}
11192 	return need_reset;
11193 }
11194 
11195 /**
11196  * i40e_clear_rss_lut - clear the rx hash lookup table
11197  * @vsi: the VSI being configured
11198  **/
11199 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
11200 {
11201 	struct i40e_pf *pf = vsi->back;
11202 	struct i40e_hw *hw = &pf->hw;
11203 	u16 vf_id = vsi->vf_id;
11204 	u8 i;
11205 
11206 	if (vsi->type == I40E_VSI_MAIN) {
11207 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
11208 			wr32(hw, I40E_PFQF_HLUT(i), 0);
11209 	} else if (vsi->type == I40E_VSI_SRIOV) {
11210 		for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
11211 			i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
11212 	} else {
11213 		dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
11214 	}
11215 }
11216 
11217 /**
11218  * i40e_set_features - set the netdev feature flags
11219  * @netdev: ptr to the netdev being adjusted
11220  * @features: the feature set that the stack is suggesting
11221  * Note: expects to be called while under rtnl_lock()
11222  **/
11223 static int i40e_set_features(struct net_device *netdev,
11224 			     netdev_features_t features)
11225 {
11226 	struct i40e_netdev_priv *np = netdev_priv(netdev);
11227 	struct i40e_vsi *vsi = np->vsi;
11228 	struct i40e_pf *pf = vsi->back;
11229 	bool need_reset;
11230 
11231 	if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
11232 		i40e_pf_config_rss(pf);
11233 	else if (!(features & NETIF_F_RXHASH) &&
11234 		 netdev->features & NETIF_F_RXHASH)
11235 		i40e_clear_rss_lut(vsi);
11236 
11237 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
11238 		i40e_vlan_stripping_enable(vsi);
11239 	else
11240 		i40e_vlan_stripping_disable(vsi);
11241 
11242 	if (!(features & NETIF_F_HW_TC) && pf->num_cloud_filters) {
11243 		dev_err(&pf->pdev->dev,
11244 			"Offloaded tc filters active, can't turn hw_tc_offload off");
11245 		return -EINVAL;
11246 	}
11247 
11248 	need_reset = i40e_set_ntuple(pf, features);
11249 
11250 	if (need_reset)
11251 		i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
11252 
11253 	return 0;
11254 }
11255 
11256 /**
11257  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
11258  * @pf: board private structure
11259  * @port: The UDP port to look up
11260  *
11261  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
11262  **/
11263 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
11264 {
11265 	u8 i;
11266 
11267 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
11268 		if (pf->udp_ports[i].port == port)
11269 			return i;
11270 	}
11271 
11272 	return i;
11273 }
11274 
11275 /**
11276  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
11277  * @netdev: This physical port's netdev
11278  * @ti: Tunnel endpoint information
11279  **/
11280 static void i40e_udp_tunnel_add(struct net_device *netdev,
11281 				struct udp_tunnel_info *ti)
11282 {
11283 	struct i40e_netdev_priv *np = netdev_priv(netdev);
11284 	struct i40e_vsi *vsi = np->vsi;
11285 	struct i40e_pf *pf = vsi->back;
11286 	u16 port = ntohs(ti->port);
11287 	u8 next_idx;
11288 	u8 idx;
11289 
11290 	idx = i40e_get_udp_port_idx(pf, port);
11291 
11292 	/* Check if port already exists */
11293 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
11294 		netdev_info(netdev, "port %d already offloaded\n", port);
11295 		return;
11296 	}
11297 
11298 	/* Now check if there is space to add the new port */
11299 	next_idx = i40e_get_udp_port_idx(pf, 0);
11300 
11301 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
11302 		netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
11303 			    port);
11304 		return;
11305 	}
11306 
11307 	switch (ti->type) {
11308 	case UDP_TUNNEL_TYPE_VXLAN:
11309 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
11310 		break;
11311 	case UDP_TUNNEL_TYPE_GENEVE:
11312 		if (!(pf->hw_features & I40E_HW_GENEVE_OFFLOAD_CAPABLE))
11313 			return;
11314 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
11315 		break;
11316 	default:
11317 		return;
11318 	}
11319 
11320 	/* New port: add it and mark its index in the bitmap */
11321 	pf->udp_ports[next_idx].port = port;
11322 	pf->pending_udp_bitmap |= BIT_ULL(next_idx);
11323 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
11324 }
11325 
11326 /**
11327  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
11328  * @netdev: This physical port's netdev
11329  * @ti: Tunnel endpoint information
11330  **/
11331 static void i40e_udp_tunnel_del(struct net_device *netdev,
11332 				struct udp_tunnel_info *ti)
11333 {
11334 	struct i40e_netdev_priv *np = netdev_priv(netdev);
11335 	struct i40e_vsi *vsi = np->vsi;
11336 	struct i40e_pf *pf = vsi->back;
11337 	u16 port = ntohs(ti->port);
11338 	u8 idx;
11339 
11340 	idx = i40e_get_udp_port_idx(pf, port);
11341 
11342 	/* Check if port already exists */
11343 	if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
11344 		goto not_found;
11345 
11346 	switch (ti->type) {
11347 	case UDP_TUNNEL_TYPE_VXLAN:
11348 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
11349 			goto not_found;
11350 		break;
11351 	case UDP_TUNNEL_TYPE_GENEVE:
11352 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
11353 			goto not_found;
11354 		break;
11355 	default:
11356 		goto not_found;
11357 	}
11358 
11359 	/* if port exists, set it to 0 (mark for deletion)
11360 	 * and make it pending
11361 	 */
11362 	pf->udp_ports[idx].port = 0;
11363 	pf->pending_udp_bitmap |= BIT_ULL(idx);
11364 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
11365 
11366 	return;
11367 not_found:
11368 	netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
11369 		    port);
11370 }
11371 
11372 static int i40e_get_phys_port_id(struct net_device *netdev,
11373 				 struct netdev_phys_item_id *ppid)
11374 {
11375 	struct i40e_netdev_priv *np = netdev_priv(netdev);
11376 	struct i40e_pf *pf = np->vsi->back;
11377 	struct i40e_hw *hw = &pf->hw;
11378 
11379 	if (!(pf->hw_features & I40E_HW_PORT_ID_VALID))
11380 		return -EOPNOTSUPP;
11381 
11382 	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
11383 	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
11384 
11385 	return 0;
11386 }
11387 
11388 /**
11389  * i40e_ndo_fdb_add - add an entry to the hardware database
11390  * @ndm: the input from the stack
11391  * @tb: pointer to array of nladdr (unused)
11392  * @dev: the net device pointer
11393  * @addr: the MAC address entry being added
11394  * @flags: instructions from stack about fdb operation
11395  */
11396 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
11397 			    struct net_device *dev,
11398 			    const unsigned char *addr, u16 vid,
11399 			    u16 flags)
11400 {
11401 	struct i40e_netdev_priv *np = netdev_priv(dev);
11402 	struct i40e_pf *pf = np->vsi->back;
11403 	int err = 0;
11404 
11405 	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
11406 		return -EOPNOTSUPP;
11407 
11408 	if (vid) {
11409 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
11410 		return -EINVAL;
11411 	}
11412 
11413 	/* Hardware does not support aging addresses so if a
11414 	 * ndm_state is given only allow permanent addresses
11415 	 */
11416 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
11417 		netdev_info(dev, "FDB only supports static addresses\n");
11418 		return -EINVAL;
11419 	}
11420 
11421 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
11422 		err = dev_uc_add_excl(dev, addr);
11423 	else if (is_multicast_ether_addr(addr))
11424 		err = dev_mc_add_excl(dev, addr);
11425 	else
11426 		err = -EINVAL;
11427 
11428 	/* Only return duplicate errors if NLM_F_EXCL is set */
11429 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
11430 		err = 0;
11431 
11432 	return err;
11433 }
11434 
11435 /**
11436  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
11437  * @dev: the netdev being configured
11438  * @nlh: RTNL message
11439  *
11440  * Inserts a new hardware bridge if not already created and
11441  * enables the bridging mode requested (VEB or VEPA). If the
11442  * hardware bridge has already been inserted and the request
11443  * is to change the mode then that requires a PF reset to
11444  * allow rebuild of the components with required hardware
11445  * bridge mode enabled.
11446  *
11447  * Note: expects to be called while under rtnl_lock()
11448  **/
11449 static int i40e_ndo_bridge_setlink(struct net_device *dev,
11450 				   struct nlmsghdr *nlh,
11451 				   u16 flags)
11452 {
11453 	struct i40e_netdev_priv *np = netdev_priv(dev);
11454 	struct i40e_vsi *vsi = np->vsi;
11455 	struct i40e_pf *pf = vsi->back;
11456 	struct i40e_veb *veb = NULL;
11457 	struct nlattr *attr, *br_spec;
11458 	int i, rem;
11459 
11460 	/* Only for PF VSI for now */
11461 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
11462 		return -EOPNOTSUPP;
11463 
11464 	/* Find the HW bridge for PF VSI */
11465 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
11466 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
11467 			veb = pf->veb[i];
11468 	}
11469 
11470 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
11471 
11472 	nla_for_each_nested(attr, br_spec, rem) {
11473 		__u16 mode;
11474 
11475 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
11476 			continue;
11477 
11478 		mode = nla_get_u16(attr);
11479 		if ((mode != BRIDGE_MODE_VEPA) &&
11480 		    (mode != BRIDGE_MODE_VEB))
11481 			return -EINVAL;
11482 
11483 		/* Insert a new HW bridge */
11484 		if (!veb) {
11485 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
11486 					     vsi->tc_config.enabled_tc);
11487 			if (veb) {
11488 				veb->bridge_mode = mode;
11489 				i40e_config_bridge_mode(veb);
11490 			} else {
11491 				/* No Bridge HW offload available */
11492 				return -ENOENT;
11493 			}
11494 			break;
11495 		} else if (mode != veb->bridge_mode) {
11496 			/* Existing HW bridge but different mode needs reset */
11497 			veb->bridge_mode = mode;
11498 			/* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
11499 			if (mode == BRIDGE_MODE_VEB)
11500 				pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11501 			else
11502 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
11503 			i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
11504 			break;
11505 		}
11506 	}
11507 
11508 	return 0;
11509 }
11510 
11511 /**
11512  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
11513  * @skb: skb buff
11514  * @pid: process id
11515  * @seq: RTNL message seq #
11516  * @dev: the netdev being configured
11517  * @filter_mask: unused
11518  * @nlflags: netlink flags passed in
11519  *
11520  * Return the mode in which the hardware bridge is operating in
11521  * i.e VEB or VEPA.
11522  **/
11523 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
11524 				   struct net_device *dev,
11525 				   u32 __always_unused filter_mask,
11526 				   int nlflags)
11527 {
11528 	struct i40e_netdev_priv *np = netdev_priv(dev);
11529 	struct i40e_vsi *vsi = np->vsi;
11530 	struct i40e_pf *pf = vsi->back;
11531 	struct i40e_veb *veb = NULL;
11532 	int i;
11533 
11534 	/* Only for PF VSI for now */
11535 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
11536 		return -EOPNOTSUPP;
11537 
11538 	/* Find the HW bridge for the PF VSI */
11539 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
11540 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
11541 			veb = pf->veb[i];
11542 	}
11543 
11544 	if (!veb)
11545 		return 0;
11546 
11547 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
11548 				       0, 0, nlflags, filter_mask, NULL);
11549 }
11550 
11551 /**
11552  * i40e_features_check - Validate encapsulated packet conforms to limits
11553  * @skb: skb buff
11554  * @dev: This physical port's netdev
11555  * @features: Offload features that the stack believes apply
11556  **/
11557 static netdev_features_t i40e_features_check(struct sk_buff *skb,
11558 					     struct net_device *dev,
11559 					     netdev_features_t features)
11560 {
11561 	size_t len;
11562 
11563 	/* No point in doing any of this if neither checksum nor GSO are
11564 	 * being requested for this frame.  We can rule out both by just
11565 	 * checking for CHECKSUM_PARTIAL
11566 	 */
11567 	if (skb->ip_summed != CHECKSUM_PARTIAL)
11568 		return features;
11569 
11570 	/* We cannot support GSO if the MSS is going to be less than
11571 	 * 64 bytes.  If it is then we need to drop support for GSO.
11572 	 */
11573 	if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
11574 		features &= ~NETIF_F_GSO_MASK;
11575 
11576 	/* MACLEN can support at most 63 words */
11577 	len = skb_network_header(skb) - skb->data;
11578 	if (len & ~(63 * 2))
11579 		goto out_err;
11580 
11581 	/* IPLEN and EIPLEN can support at most 127 dwords */
11582 	len = skb_transport_header(skb) - skb_network_header(skb);
11583 	if (len & ~(127 * 4))
11584 		goto out_err;
11585 
11586 	if (skb->encapsulation) {
11587 		/* L4TUNLEN can support 127 words */
11588 		len = skb_inner_network_header(skb) - skb_transport_header(skb);
11589 		if (len & ~(127 * 2))
11590 			goto out_err;
11591 
11592 		/* IPLEN can support at most 127 dwords */
11593 		len = skb_inner_transport_header(skb) -
11594 		      skb_inner_network_header(skb);
11595 		if (len & ~(127 * 4))
11596 			goto out_err;
11597 	}
11598 
11599 	/* No need to validate L4LEN as TCP is the only protocol with a
11600 	 * a flexible value and we support all possible values supported
11601 	 * by TCP, which is at most 15 dwords
11602 	 */
11603 
11604 	return features;
11605 out_err:
11606 	return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
11607 }
11608 
11609 /**
11610  * i40e_xdp_setup - add/remove an XDP program
11611  * @vsi: VSI to changed
11612  * @prog: XDP program
11613  **/
11614 static int i40e_xdp_setup(struct i40e_vsi *vsi,
11615 			  struct bpf_prog *prog)
11616 {
11617 	int frame_size = vsi->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
11618 	struct i40e_pf *pf = vsi->back;
11619 	struct bpf_prog *old_prog;
11620 	bool need_reset;
11621 	int i;
11622 
11623 	/* Don't allow frames that span over multiple buffers */
11624 	if (frame_size > vsi->rx_buf_len)
11625 		return -EINVAL;
11626 
11627 	if (!i40e_enabled_xdp_vsi(vsi) && !prog)
11628 		return 0;
11629 
11630 	/* When turning XDP on->off/off->on we reset and rebuild the rings. */
11631 	need_reset = (i40e_enabled_xdp_vsi(vsi) != !!prog);
11632 
11633 	if (need_reset)
11634 		i40e_prep_for_reset(pf, true);
11635 
11636 	old_prog = xchg(&vsi->xdp_prog, prog);
11637 
11638 	if (need_reset)
11639 		i40e_reset_and_rebuild(pf, true, true);
11640 
11641 	for (i = 0; i < vsi->num_queue_pairs; i++)
11642 		WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
11643 
11644 	if (old_prog)
11645 		bpf_prog_put(old_prog);
11646 
11647 	return 0;
11648 }
11649 
11650 /**
11651  * i40e_xdp - implements ndo_bpf for i40e
11652  * @dev: netdevice
11653  * @xdp: XDP command
11654  **/
11655 static int i40e_xdp(struct net_device *dev,
11656 		    struct netdev_bpf *xdp)
11657 {
11658 	struct i40e_netdev_priv *np = netdev_priv(dev);
11659 	struct i40e_vsi *vsi = np->vsi;
11660 
11661 	if (vsi->type != I40E_VSI_MAIN)
11662 		return -EINVAL;
11663 
11664 	switch (xdp->command) {
11665 	case XDP_SETUP_PROG:
11666 		return i40e_xdp_setup(vsi, xdp->prog);
11667 	case XDP_QUERY_PROG:
11668 		xdp->prog_attached = i40e_enabled_xdp_vsi(vsi);
11669 		xdp->prog_id = vsi->xdp_prog ? vsi->xdp_prog->aux->id : 0;
11670 		return 0;
11671 	default:
11672 		return -EINVAL;
11673 	}
11674 }
11675 
11676 static const struct net_device_ops i40e_netdev_ops = {
11677 	.ndo_open		= i40e_open,
11678 	.ndo_stop		= i40e_close,
11679 	.ndo_start_xmit		= i40e_lan_xmit_frame,
11680 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
11681 	.ndo_set_rx_mode	= i40e_set_rx_mode,
11682 	.ndo_validate_addr	= eth_validate_addr,
11683 	.ndo_set_mac_address	= i40e_set_mac,
11684 	.ndo_change_mtu		= i40e_change_mtu,
11685 	.ndo_do_ioctl		= i40e_ioctl,
11686 	.ndo_tx_timeout		= i40e_tx_timeout,
11687 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
11688 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
11689 #ifdef CONFIG_NET_POLL_CONTROLLER
11690 	.ndo_poll_controller	= i40e_netpoll,
11691 #endif
11692 	.ndo_setup_tc		= __i40e_setup_tc,
11693 	.ndo_set_features	= i40e_set_features,
11694 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
11695 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
11696 	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
11697 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
11698 	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
11699 	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
11700 	.ndo_set_vf_trust	= i40e_ndo_set_vf_trust,
11701 	.ndo_udp_tunnel_add	= i40e_udp_tunnel_add,
11702 	.ndo_udp_tunnel_del	= i40e_udp_tunnel_del,
11703 	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
11704 	.ndo_fdb_add		= i40e_ndo_fdb_add,
11705 	.ndo_features_check	= i40e_features_check,
11706 	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
11707 	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
11708 	.ndo_bpf		= i40e_xdp,
11709 };
11710 
11711 /**
11712  * i40e_config_netdev - Setup the netdev flags
11713  * @vsi: the VSI being configured
11714  *
11715  * Returns 0 on success, negative value on failure
11716  **/
11717 static int i40e_config_netdev(struct i40e_vsi *vsi)
11718 {
11719 	struct i40e_pf *pf = vsi->back;
11720 	struct i40e_hw *hw = &pf->hw;
11721 	struct i40e_netdev_priv *np;
11722 	struct net_device *netdev;
11723 	u8 broadcast[ETH_ALEN];
11724 	u8 mac_addr[ETH_ALEN];
11725 	int etherdev_size;
11726 	netdev_features_t hw_enc_features;
11727 	netdev_features_t hw_features;
11728 
11729 	etherdev_size = sizeof(struct i40e_netdev_priv);
11730 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
11731 	if (!netdev)
11732 		return -ENOMEM;
11733 
11734 	vsi->netdev = netdev;
11735 	np = netdev_priv(netdev);
11736 	np->vsi = vsi;
11737 
11738 	hw_enc_features = NETIF_F_SG			|
11739 			  NETIF_F_IP_CSUM		|
11740 			  NETIF_F_IPV6_CSUM		|
11741 			  NETIF_F_HIGHDMA		|
11742 			  NETIF_F_SOFT_FEATURES		|
11743 			  NETIF_F_TSO			|
11744 			  NETIF_F_TSO_ECN		|
11745 			  NETIF_F_TSO6			|
11746 			  NETIF_F_GSO_GRE		|
11747 			  NETIF_F_GSO_GRE_CSUM		|
11748 			  NETIF_F_GSO_PARTIAL		|
11749 			  NETIF_F_GSO_UDP_TUNNEL	|
11750 			  NETIF_F_GSO_UDP_TUNNEL_CSUM	|
11751 			  NETIF_F_SCTP_CRC		|
11752 			  NETIF_F_RXHASH		|
11753 			  NETIF_F_RXCSUM		|
11754 			  0;
11755 
11756 	if (!(pf->hw_features & I40E_HW_OUTER_UDP_CSUM_CAPABLE))
11757 		netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
11758 
11759 	netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
11760 
11761 	netdev->hw_enc_features |= hw_enc_features;
11762 
11763 	/* record features VLANs can make use of */
11764 	netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
11765 
11766 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
11767 		netdev->hw_features |= NETIF_F_NTUPLE | NETIF_F_HW_TC;
11768 
11769 	hw_features = hw_enc_features		|
11770 		      NETIF_F_HW_VLAN_CTAG_TX	|
11771 		      NETIF_F_HW_VLAN_CTAG_RX;
11772 
11773 	netdev->hw_features |= hw_features;
11774 
11775 	netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
11776 	netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
11777 
11778 	if (vsi->type == I40E_VSI_MAIN) {
11779 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
11780 		ether_addr_copy(mac_addr, hw->mac.perm_addr);
11781 		/* The following steps are necessary for two reasons. First,
11782 		 * some older NVM configurations load a default MAC-VLAN
11783 		 * filter that will accept any tagged packet, and we want to
11784 		 * replace this with a normal filter. Additionally, it is
11785 		 * possible our MAC address was provided by the platform using
11786 		 * Open Firmware or similar.
11787 		 *
11788 		 * Thus, we need to remove the default filter and install one
11789 		 * specific to the MAC address.
11790 		 */
11791 		i40e_rm_default_mac_filter(vsi, mac_addr);
11792 		spin_lock_bh(&vsi->mac_filter_hash_lock);
11793 		i40e_add_mac_filter(vsi, mac_addr);
11794 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
11795 	} else {
11796 		/* Relate the VSI_VMDQ name to the VSI_MAIN name. Note that we
11797 		 * are still limited by IFNAMSIZ, but we're adding 'v%d\0' to
11798 		 * the end, which is 4 bytes long, so force truncation of the
11799 		 * original name by IFNAMSIZ - 4
11800 		 */
11801 		snprintf(netdev->name, IFNAMSIZ, "%.*sv%%d",
11802 			 IFNAMSIZ - 4,
11803 			 pf->vsi[pf->lan_vsi]->netdev->name);
11804 		random_ether_addr(mac_addr);
11805 
11806 		spin_lock_bh(&vsi->mac_filter_hash_lock);
11807 		i40e_add_mac_filter(vsi, mac_addr);
11808 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
11809 	}
11810 
11811 	/* Add the broadcast filter so that we initially will receive
11812 	 * broadcast packets. Note that when a new VLAN is first added the
11813 	 * driver will convert all filters marked I40E_VLAN_ANY into VLAN
11814 	 * specific filters as part of transitioning into "vlan" operation.
11815 	 * When more VLANs are added, the driver will copy each existing MAC
11816 	 * filter and add it for the new VLAN.
11817 	 *
11818 	 * Broadcast filters are handled specially by
11819 	 * i40e_sync_filters_subtask, as the driver must to set the broadcast
11820 	 * promiscuous bit instead of adding this directly as a MAC/VLAN
11821 	 * filter. The subtask will update the correct broadcast promiscuous
11822 	 * bits as VLANs become active or inactive.
11823 	 */
11824 	eth_broadcast_addr(broadcast);
11825 	spin_lock_bh(&vsi->mac_filter_hash_lock);
11826 	i40e_add_mac_filter(vsi, broadcast);
11827 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
11828 
11829 	ether_addr_copy(netdev->dev_addr, mac_addr);
11830 	ether_addr_copy(netdev->perm_addr, mac_addr);
11831 
11832 	netdev->priv_flags |= IFF_UNICAST_FLT;
11833 	netdev->priv_flags |= IFF_SUPP_NOFCS;
11834 	/* Setup netdev TC information */
11835 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
11836 
11837 	netdev->netdev_ops = &i40e_netdev_ops;
11838 	netdev->watchdog_timeo = 5 * HZ;
11839 	i40e_set_ethtool_ops(netdev);
11840 
11841 	/* MTU range: 68 - 9706 */
11842 	netdev->min_mtu = ETH_MIN_MTU;
11843 	netdev->max_mtu = I40E_MAX_RXBUFFER - I40E_PACKET_HDR_PAD;
11844 
11845 	return 0;
11846 }
11847 
11848 /**
11849  * i40e_vsi_delete - Delete a VSI from the switch
11850  * @vsi: the VSI being removed
11851  *
11852  * Returns 0 on success, negative value on failure
11853  **/
11854 static void i40e_vsi_delete(struct i40e_vsi *vsi)
11855 {
11856 	/* remove default VSI is not allowed */
11857 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
11858 		return;
11859 
11860 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
11861 }
11862 
11863 /**
11864  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
11865  * @vsi: the VSI being queried
11866  *
11867  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
11868  **/
11869 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
11870 {
11871 	struct i40e_veb *veb;
11872 	struct i40e_pf *pf = vsi->back;
11873 
11874 	/* Uplink is not a bridge so default to VEB */
11875 	if (vsi->veb_idx == I40E_NO_VEB)
11876 		return 1;
11877 
11878 	veb = pf->veb[vsi->veb_idx];
11879 	if (!veb) {
11880 		dev_info(&pf->pdev->dev,
11881 			 "There is no veb associated with the bridge\n");
11882 		return -ENOENT;
11883 	}
11884 
11885 	/* Uplink is a bridge in VEPA mode */
11886 	if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
11887 		return 0;
11888 	} else {
11889 		/* Uplink is a bridge in VEB mode */
11890 		return 1;
11891 	}
11892 
11893 	/* VEPA is now default bridge, so return 0 */
11894 	return 0;
11895 }
11896 
11897 /**
11898  * i40e_add_vsi - Add a VSI to the switch
11899  * @vsi: the VSI being configured
11900  *
11901  * This initializes a VSI context depending on the VSI type to be added and
11902  * passes it down to the add_vsi aq command.
11903  **/
11904 static int i40e_add_vsi(struct i40e_vsi *vsi)
11905 {
11906 	int ret = -ENODEV;
11907 	struct i40e_pf *pf = vsi->back;
11908 	struct i40e_hw *hw = &pf->hw;
11909 	struct i40e_vsi_context ctxt;
11910 	struct i40e_mac_filter *f;
11911 	struct hlist_node *h;
11912 	int bkt;
11913 
11914 	u8 enabled_tc = 0x1; /* TC0 enabled */
11915 	int f_count = 0;
11916 
11917 	memset(&ctxt, 0, sizeof(ctxt));
11918 	switch (vsi->type) {
11919 	case I40E_VSI_MAIN:
11920 		/* The PF's main VSI is already setup as part of the
11921 		 * device initialization, so we'll not bother with
11922 		 * the add_vsi call, but we will retrieve the current
11923 		 * VSI context.
11924 		 */
11925 		ctxt.seid = pf->main_vsi_seid;
11926 		ctxt.pf_num = pf->hw.pf_id;
11927 		ctxt.vf_num = 0;
11928 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
11929 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
11930 		if (ret) {
11931 			dev_info(&pf->pdev->dev,
11932 				 "couldn't get PF vsi config, err %s aq_err %s\n",
11933 				 i40e_stat_str(&pf->hw, ret),
11934 				 i40e_aq_str(&pf->hw,
11935 					     pf->hw.aq.asq_last_status));
11936 			return -ENOENT;
11937 		}
11938 		vsi->info = ctxt.info;
11939 		vsi->info.valid_sections = 0;
11940 
11941 		vsi->seid = ctxt.seid;
11942 		vsi->id = ctxt.vsi_number;
11943 
11944 		enabled_tc = i40e_pf_get_tc_map(pf);
11945 
11946 		/* Source pruning is enabled by default, so the flag is
11947 		 * negative logic - if it's set, we need to fiddle with
11948 		 * the VSI to disable source pruning.
11949 		 */
11950 		if (pf->flags & I40E_FLAG_SOURCE_PRUNING_DISABLED) {
11951 			memset(&ctxt, 0, sizeof(ctxt));
11952 			ctxt.seid = pf->main_vsi_seid;
11953 			ctxt.pf_num = pf->hw.pf_id;
11954 			ctxt.vf_num = 0;
11955 			ctxt.info.valid_sections |=
11956 				     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
11957 			ctxt.info.switch_id =
11958 				   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
11959 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
11960 			if (ret) {
11961 				dev_info(&pf->pdev->dev,
11962 					 "update vsi failed, err %s aq_err %s\n",
11963 					 i40e_stat_str(&pf->hw, ret),
11964 					 i40e_aq_str(&pf->hw,
11965 						     pf->hw.aq.asq_last_status));
11966 				ret = -ENOENT;
11967 				goto err;
11968 			}
11969 		}
11970 
11971 		/* MFP mode setup queue map and update VSI */
11972 		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
11973 		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
11974 			memset(&ctxt, 0, sizeof(ctxt));
11975 			ctxt.seid = pf->main_vsi_seid;
11976 			ctxt.pf_num = pf->hw.pf_id;
11977 			ctxt.vf_num = 0;
11978 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
11979 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
11980 			if (ret) {
11981 				dev_info(&pf->pdev->dev,
11982 					 "update vsi failed, err %s aq_err %s\n",
11983 					 i40e_stat_str(&pf->hw, ret),
11984 					 i40e_aq_str(&pf->hw,
11985 						    pf->hw.aq.asq_last_status));
11986 				ret = -ENOENT;
11987 				goto err;
11988 			}
11989 			/* update the local VSI info queue map */
11990 			i40e_vsi_update_queue_map(vsi, &ctxt);
11991 			vsi->info.valid_sections = 0;
11992 		} else {
11993 			/* Default/Main VSI is only enabled for TC0
11994 			 * reconfigure it to enable all TCs that are
11995 			 * available on the port in SFP mode.
11996 			 * For MFP case the iSCSI PF would use this
11997 			 * flow to enable LAN+iSCSI TC.
11998 			 */
11999 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
12000 			if (ret) {
12001 				/* Single TC condition is not fatal,
12002 				 * message and continue
12003 				 */
12004 				dev_info(&pf->pdev->dev,
12005 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
12006 					 enabled_tc,
12007 					 i40e_stat_str(&pf->hw, ret),
12008 					 i40e_aq_str(&pf->hw,
12009 						    pf->hw.aq.asq_last_status));
12010 			}
12011 		}
12012 		break;
12013 
12014 	case I40E_VSI_FDIR:
12015 		ctxt.pf_num = hw->pf_id;
12016 		ctxt.vf_num = 0;
12017 		ctxt.uplink_seid = vsi->uplink_seid;
12018 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
12019 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
12020 		if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
12021 		    (i40e_is_vsi_uplink_mode_veb(vsi))) {
12022 			ctxt.info.valid_sections |=
12023 			     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
12024 			ctxt.info.switch_id =
12025 			   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
12026 		}
12027 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
12028 		break;
12029 
12030 	case I40E_VSI_VMDQ2:
12031 		ctxt.pf_num = hw->pf_id;
12032 		ctxt.vf_num = 0;
12033 		ctxt.uplink_seid = vsi->uplink_seid;
12034 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
12035 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
12036 
12037 		/* This VSI is connected to VEB so the switch_id
12038 		 * should be set to zero by default.
12039 		 */
12040 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
12041 			ctxt.info.valid_sections |=
12042 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
12043 			ctxt.info.switch_id =
12044 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
12045 		}
12046 
12047 		/* Setup the VSI tx/rx queue map for TC0 only for now */
12048 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
12049 		break;
12050 
12051 	case I40E_VSI_SRIOV:
12052 		ctxt.pf_num = hw->pf_id;
12053 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
12054 		ctxt.uplink_seid = vsi->uplink_seid;
12055 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
12056 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
12057 
12058 		/* This VSI is connected to VEB so the switch_id
12059 		 * should be set to zero by default.
12060 		 */
12061 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
12062 			ctxt.info.valid_sections |=
12063 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
12064 			ctxt.info.switch_id =
12065 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
12066 		}
12067 
12068 		if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
12069 			ctxt.info.valid_sections |=
12070 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
12071 			ctxt.info.queueing_opt_flags |=
12072 				(I40E_AQ_VSI_QUE_OPT_TCP_ENA |
12073 				 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
12074 		}
12075 
12076 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
12077 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
12078 		if (pf->vf[vsi->vf_id].spoofchk) {
12079 			ctxt.info.valid_sections |=
12080 				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
12081 			ctxt.info.sec_flags |=
12082 				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
12083 				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
12084 		}
12085 		/* Setup the VSI tx/rx queue map for TC0 only for now */
12086 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
12087 		break;
12088 
12089 	case I40E_VSI_IWARP:
12090 		/* send down message to iWARP */
12091 		break;
12092 
12093 	default:
12094 		return -ENODEV;
12095 	}
12096 
12097 	if (vsi->type != I40E_VSI_MAIN) {
12098 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
12099 		if (ret) {
12100 			dev_info(&vsi->back->pdev->dev,
12101 				 "add vsi failed, err %s aq_err %s\n",
12102 				 i40e_stat_str(&pf->hw, ret),
12103 				 i40e_aq_str(&pf->hw,
12104 					     pf->hw.aq.asq_last_status));
12105 			ret = -ENOENT;
12106 			goto err;
12107 		}
12108 		vsi->info = ctxt.info;
12109 		vsi->info.valid_sections = 0;
12110 		vsi->seid = ctxt.seid;
12111 		vsi->id = ctxt.vsi_number;
12112 	}
12113 
12114 	vsi->active_filters = 0;
12115 	clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
12116 	spin_lock_bh(&vsi->mac_filter_hash_lock);
12117 	/* If macvlan filters already exist, force them to get loaded */
12118 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
12119 		f->state = I40E_FILTER_NEW;
12120 		f_count++;
12121 	}
12122 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
12123 
12124 	if (f_count) {
12125 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
12126 		pf->flags |= I40E_FLAG_FILTER_SYNC;
12127 	}
12128 
12129 	/* Update VSI BW information */
12130 	ret = i40e_vsi_get_bw_info(vsi);
12131 	if (ret) {
12132 		dev_info(&pf->pdev->dev,
12133 			 "couldn't get vsi bw info, err %s aq_err %s\n",
12134 			 i40e_stat_str(&pf->hw, ret),
12135 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12136 		/* VSI is already added so not tearing that up */
12137 		ret = 0;
12138 	}
12139 
12140 err:
12141 	return ret;
12142 }
12143 
12144 /**
12145  * i40e_vsi_release - Delete a VSI and free its resources
12146  * @vsi: the VSI being removed
12147  *
12148  * Returns 0 on success or < 0 on error
12149  **/
12150 int i40e_vsi_release(struct i40e_vsi *vsi)
12151 {
12152 	struct i40e_mac_filter *f;
12153 	struct hlist_node *h;
12154 	struct i40e_veb *veb = NULL;
12155 	struct i40e_pf *pf;
12156 	u16 uplink_seid;
12157 	int i, n, bkt;
12158 
12159 	pf = vsi->back;
12160 
12161 	/* release of a VEB-owner or last VSI is not allowed */
12162 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
12163 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
12164 			 vsi->seid, vsi->uplink_seid);
12165 		return -ENODEV;
12166 	}
12167 	if (vsi == pf->vsi[pf->lan_vsi] &&
12168 	    !test_bit(__I40E_DOWN, pf->state)) {
12169 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
12170 		return -ENODEV;
12171 	}
12172 
12173 	uplink_seid = vsi->uplink_seid;
12174 	if (vsi->type != I40E_VSI_SRIOV) {
12175 		if (vsi->netdev_registered) {
12176 			vsi->netdev_registered = false;
12177 			if (vsi->netdev) {
12178 				/* results in a call to i40e_close() */
12179 				unregister_netdev(vsi->netdev);
12180 			}
12181 		} else {
12182 			i40e_vsi_close(vsi);
12183 		}
12184 		i40e_vsi_disable_irq(vsi);
12185 	}
12186 
12187 	spin_lock_bh(&vsi->mac_filter_hash_lock);
12188 
12189 	/* clear the sync flag on all filters */
12190 	if (vsi->netdev) {
12191 		__dev_uc_unsync(vsi->netdev, NULL);
12192 		__dev_mc_unsync(vsi->netdev, NULL);
12193 	}
12194 
12195 	/* make sure any remaining filters are marked for deletion */
12196 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
12197 		__i40e_del_filter(vsi, f);
12198 
12199 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
12200 
12201 	i40e_sync_vsi_filters(vsi);
12202 
12203 	i40e_vsi_delete(vsi);
12204 	i40e_vsi_free_q_vectors(vsi);
12205 	if (vsi->netdev) {
12206 		free_netdev(vsi->netdev);
12207 		vsi->netdev = NULL;
12208 	}
12209 	i40e_vsi_clear_rings(vsi);
12210 	i40e_vsi_clear(vsi);
12211 
12212 	/* If this was the last thing on the VEB, except for the
12213 	 * controlling VSI, remove the VEB, which puts the controlling
12214 	 * VSI onto the next level down in the switch.
12215 	 *
12216 	 * Well, okay, there's one more exception here: don't remove
12217 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
12218 	 * from up the network stack.
12219 	 */
12220 	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
12221 		if (pf->vsi[i] &&
12222 		    pf->vsi[i]->uplink_seid == uplink_seid &&
12223 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
12224 			n++;      /* count the VSIs */
12225 		}
12226 	}
12227 	for (i = 0; i < I40E_MAX_VEB; i++) {
12228 		if (!pf->veb[i])
12229 			continue;
12230 		if (pf->veb[i]->uplink_seid == uplink_seid)
12231 			n++;     /* count the VEBs */
12232 		if (pf->veb[i]->seid == uplink_seid)
12233 			veb = pf->veb[i];
12234 	}
12235 	if (n == 0 && veb && veb->uplink_seid != 0)
12236 		i40e_veb_release(veb);
12237 
12238 	return 0;
12239 }
12240 
12241 /**
12242  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
12243  * @vsi: ptr to the VSI
12244  *
12245  * This should only be called after i40e_vsi_mem_alloc() which allocates the
12246  * corresponding SW VSI structure and initializes num_queue_pairs for the
12247  * newly allocated VSI.
12248  *
12249  * Returns 0 on success or negative on failure
12250  **/
12251 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
12252 {
12253 	int ret = -ENOENT;
12254 	struct i40e_pf *pf = vsi->back;
12255 
12256 	if (vsi->q_vectors[0]) {
12257 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
12258 			 vsi->seid);
12259 		return -EEXIST;
12260 	}
12261 
12262 	if (vsi->base_vector) {
12263 		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
12264 			 vsi->seid, vsi->base_vector);
12265 		return -EEXIST;
12266 	}
12267 
12268 	ret = i40e_vsi_alloc_q_vectors(vsi);
12269 	if (ret) {
12270 		dev_info(&pf->pdev->dev,
12271 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
12272 			 vsi->num_q_vectors, vsi->seid, ret);
12273 		vsi->num_q_vectors = 0;
12274 		goto vector_setup_out;
12275 	}
12276 
12277 	/* In Legacy mode, we do not have to get any other vector since we
12278 	 * piggyback on the misc/ICR0 for queue interrupts.
12279 	*/
12280 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
12281 		return ret;
12282 	if (vsi->num_q_vectors)
12283 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
12284 						 vsi->num_q_vectors, vsi->idx);
12285 	if (vsi->base_vector < 0) {
12286 		dev_info(&pf->pdev->dev,
12287 			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
12288 			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
12289 		i40e_vsi_free_q_vectors(vsi);
12290 		ret = -ENOENT;
12291 		goto vector_setup_out;
12292 	}
12293 
12294 vector_setup_out:
12295 	return ret;
12296 }
12297 
12298 /**
12299  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
12300  * @vsi: pointer to the vsi.
12301  *
12302  * This re-allocates a vsi's queue resources.
12303  *
12304  * Returns pointer to the successfully allocated and configured VSI sw struct
12305  * on success, otherwise returns NULL on failure.
12306  **/
12307 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
12308 {
12309 	u16 alloc_queue_pairs;
12310 	struct i40e_pf *pf;
12311 	u8 enabled_tc;
12312 	int ret;
12313 
12314 	if (!vsi)
12315 		return NULL;
12316 
12317 	pf = vsi->back;
12318 
12319 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
12320 	i40e_vsi_clear_rings(vsi);
12321 
12322 	i40e_vsi_free_arrays(vsi, false);
12323 	i40e_set_num_rings_in_vsi(vsi);
12324 	ret = i40e_vsi_alloc_arrays(vsi, false);
12325 	if (ret)
12326 		goto err_vsi;
12327 
12328 	alloc_queue_pairs = vsi->alloc_queue_pairs *
12329 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
12330 
12331 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
12332 	if (ret < 0) {
12333 		dev_info(&pf->pdev->dev,
12334 			 "failed to get tracking for %d queues for VSI %d err %d\n",
12335 			 alloc_queue_pairs, vsi->seid, ret);
12336 		goto err_vsi;
12337 	}
12338 	vsi->base_queue = ret;
12339 
12340 	/* Update the FW view of the VSI. Force a reset of TC and queue
12341 	 * layout configurations.
12342 	 */
12343 	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
12344 	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
12345 	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
12346 	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
12347 	if (vsi->type == I40E_VSI_MAIN)
12348 		i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
12349 
12350 	/* assign it some queues */
12351 	ret = i40e_alloc_rings(vsi);
12352 	if (ret)
12353 		goto err_rings;
12354 
12355 	/* map all of the rings to the q_vectors */
12356 	i40e_vsi_map_rings_to_vectors(vsi);
12357 	return vsi;
12358 
12359 err_rings:
12360 	i40e_vsi_free_q_vectors(vsi);
12361 	if (vsi->netdev_registered) {
12362 		vsi->netdev_registered = false;
12363 		unregister_netdev(vsi->netdev);
12364 		free_netdev(vsi->netdev);
12365 		vsi->netdev = NULL;
12366 	}
12367 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
12368 err_vsi:
12369 	i40e_vsi_clear(vsi);
12370 	return NULL;
12371 }
12372 
12373 /**
12374  * i40e_vsi_setup - Set up a VSI by a given type
12375  * @pf: board private structure
12376  * @type: VSI type
12377  * @uplink_seid: the switch element to link to
12378  * @param1: usage depends upon VSI type. For VF types, indicates VF id
12379  *
12380  * This allocates the sw VSI structure and its queue resources, then add a VSI
12381  * to the identified VEB.
12382  *
12383  * Returns pointer to the successfully allocated and configure VSI sw struct on
12384  * success, otherwise returns NULL on failure.
12385  **/
12386 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
12387 				u16 uplink_seid, u32 param1)
12388 {
12389 	struct i40e_vsi *vsi = NULL;
12390 	struct i40e_veb *veb = NULL;
12391 	u16 alloc_queue_pairs;
12392 	int ret, i;
12393 	int v_idx;
12394 
12395 	/* The requested uplink_seid must be either
12396 	 *     - the PF's port seid
12397 	 *              no VEB is needed because this is the PF
12398 	 *              or this is a Flow Director special case VSI
12399 	 *     - seid of an existing VEB
12400 	 *     - seid of a VSI that owns an existing VEB
12401 	 *     - seid of a VSI that doesn't own a VEB
12402 	 *              a new VEB is created and the VSI becomes the owner
12403 	 *     - seid of the PF VSI, which is what creates the first VEB
12404 	 *              this is a special case of the previous
12405 	 *
12406 	 * Find which uplink_seid we were given and create a new VEB if needed
12407 	 */
12408 	for (i = 0; i < I40E_MAX_VEB; i++) {
12409 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
12410 			veb = pf->veb[i];
12411 			break;
12412 		}
12413 	}
12414 
12415 	if (!veb && uplink_seid != pf->mac_seid) {
12416 
12417 		for (i = 0; i < pf->num_alloc_vsi; i++) {
12418 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
12419 				vsi = pf->vsi[i];
12420 				break;
12421 			}
12422 		}
12423 		if (!vsi) {
12424 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
12425 				 uplink_seid);
12426 			return NULL;
12427 		}
12428 
12429 		if (vsi->uplink_seid == pf->mac_seid)
12430 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
12431 					     vsi->tc_config.enabled_tc);
12432 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
12433 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
12434 					     vsi->tc_config.enabled_tc);
12435 		if (veb) {
12436 			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
12437 				dev_info(&vsi->back->pdev->dev,
12438 					 "New VSI creation error, uplink seid of LAN VSI expected.\n");
12439 				return NULL;
12440 			}
12441 			/* We come up by default in VEPA mode if SRIOV is not
12442 			 * already enabled, in which case we can't force VEPA
12443 			 * mode.
12444 			 */
12445 			if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
12446 				veb->bridge_mode = BRIDGE_MODE_VEPA;
12447 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
12448 			}
12449 			i40e_config_bridge_mode(veb);
12450 		}
12451 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
12452 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
12453 				veb = pf->veb[i];
12454 		}
12455 		if (!veb) {
12456 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
12457 			return NULL;
12458 		}
12459 
12460 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
12461 		uplink_seid = veb->seid;
12462 	}
12463 
12464 	/* get vsi sw struct */
12465 	v_idx = i40e_vsi_mem_alloc(pf, type);
12466 	if (v_idx < 0)
12467 		goto err_alloc;
12468 	vsi = pf->vsi[v_idx];
12469 	if (!vsi)
12470 		goto err_alloc;
12471 	vsi->type = type;
12472 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
12473 
12474 	if (type == I40E_VSI_MAIN)
12475 		pf->lan_vsi = v_idx;
12476 	else if (type == I40E_VSI_SRIOV)
12477 		vsi->vf_id = param1;
12478 	/* assign it some queues */
12479 	alloc_queue_pairs = vsi->alloc_queue_pairs *
12480 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
12481 
12482 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
12483 	if (ret < 0) {
12484 		dev_info(&pf->pdev->dev,
12485 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
12486 			 alloc_queue_pairs, vsi->seid, ret);
12487 		goto err_vsi;
12488 	}
12489 	vsi->base_queue = ret;
12490 
12491 	/* get a VSI from the hardware */
12492 	vsi->uplink_seid = uplink_seid;
12493 	ret = i40e_add_vsi(vsi);
12494 	if (ret)
12495 		goto err_vsi;
12496 
12497 	switch (vsi->type) {
12498 	/* setup the netdev if needed */
12499 	case I40E_VSI_MAIN:
12500 	case I40E_VSI_VMDQ2:
12501 		ret = i40e_config_netdev(vsi);
12502 		if (ret)
12503 			goto err_netdev;
12504 		ret = register_netdev(vsi->netdev);
12505 		if (ret)
12506 			goto err_netdev;
12507 		vsi->netdev_registered = true;
12508 		netif_carrier_off(vsi->netdev);
12509 #ifdef CONFIG_I40E_DCB
12510 		/* Setup DCB netlink interface */
12511 		i40e_dcbnl_setup(vsi);
12512 #endif /* CONFIG_I40E_DCB */
12513 		/* fall through */
12514 
12515 	case I40E_VSI_FDIR:
12516 		/* set up vectors and rings if needed */
12517 		ret = i40e_vsi_setup_vectors(vsi);
12518 		if (ret)
12519 			goto err_msix;
12520 
12521 		ret = i40e_alloc_rings(vsi);
12522 		if (ret)
12523 			goto err_rings;
12524 
12525 		/* map all of the rings to the q_vectors */
12526 		i40e_vsi_map_rings_to_vectors(vsi);
12527 
12528 		i40e_vsi_reset_stats(vsi);
12529 		break;
12530 
12531 	default:
12532 		/* no netdev or rings for the other VSI types */
12533 		break;
12534 	}
12535 
12536 	if ((pf->hw_features & I40E_HW_RSS_AQ_CAPABLE) &&
12537 	    (vsi->type == I40E_VSI_VMDQ2)) {
12538 		ret = i40e_vsi_config_rss(vsi);
12539 	}
12540 	return vsi;
12541 
12542 err_rings:
12543 	i40e_vsi_free_q_vectors(vsi);
12544 err_msix:
12545 	if (vsi->netdev_registered) {
12546 		vsi->netdev_registered = false;
12547 		unregister_netdev(vsi->netdev);
12548 		free_netdev(vsi->netdev);
12549 		vsi->netdev = NULL;
12550 	}
12551 err_netdev:
12552 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
12553 err_vsi:
12554 	i40e_vsi_clear(vsi);
12555 err_alloc:
12556 	return NULL;
12557 }
12558 
12559 /**
12560  * i40e_veb_get_bw_info - Query VEB BW information
12561  * @veb: the veb to query
12562  *
12563  * Query the Tx scheduler BW configuration data for given VEB
12564  **/
12565 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
12566 {
12567 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
12568 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
12569 	struct i40e_pf *pf = veb->pf;
12570 	struct i40e_hw *hw = &pf->hw;
12571 	u32 tc_bw_max;
12572 	int ret = 0;
12573 	int i;
12574 
12575 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
12576 						  &bw_data, NULL);
12577 	if (ret) {
12578 		dev_info(&pf->pdev->dev,
12579 			 "query veb bw config failed, err %s aq_err %s\n",
12580 			 i40e_stat_str(&pf->hw, ret),
12581 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
12582 		goto out;
12583 	}
12584 
12585 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
12586 						   &ets_data, NULL);
12587 	if (ret) {
12588 		dev_info(&pf->pdev->dev,
12589 			 "query veb bw ets config failed, err %s aq_err %s\n",
12590 			 i40e_stat_str(&pf->hw, ret),
12591 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
12592 		goto out;
12593 	}
12594 
12595 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
12596 	veb->bw_max_quanta = ets_data.tc_bw_max;
12597 	veb->is_abs_credits = bw_data.absolute_credits_enable;
12598 	veb->enabled_tc = ets_data.tc_valid_bits;
12599 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
12600 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
12601 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
12602 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
12603 		veb->bw_tc_limit_credits[i] =
12604 					le16_to_cpu(bw_data.tc_bw_limits[i]);
12605 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
12606 	}
12607 
12608 out:
12609 	return ret;
12610 }
12611 
12612 /**
12613  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
12614  * @pf: board private structure
12615  *
12616  * On error: returns error code (negative)
12617  * On success: returns vsi index in PF (positive)
12618  **/
12619 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
12620 {
12621 	int ret = -ENOENT;
12622 	struct i40e_veb *veb;
12623 	int i;
12624 
12625 	/* Need to protect the allocation of switch elements at the PF level */
12626 	mutex_lock(&pf->switch_mutex);
12627 
12628 	/* VEB list may be fragmented if VEB creation/destruction has
12629 	 * been happening.  We can afford to do a quick scan to look
12630 	 * for any free slots in the list.
12631 	 *
12632 	 * find next empty veb slot, looping back around if necessary
12633 	 */
12634 	i = 0;
12635 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
12636 		i++;
12637 	if (i >= I40E_MAX_VEB) {
12638 		ret = -ENOMEM;
12639 		goto err_alloc_veb;  /* out of VEB slots! */
12640 	}
12641 
12642 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
12643 	if (!veb) {
12644 		ret = -ENOMEM;
12645 		goto err_alloc_veb;
12646 	}
12647 	veb->pf = pf;
12648 	veb->idx = i;
12649 	veb->enabled_tc = 1;
12650 
12651 	pf->veb[i] = veb;
12652 	ret = i;
12653 err_alloc_veb:
12654 	mutex_unlock(&pf->switch_mutex);
12655 	return ret;
12656 }
12657 
12658 /**
12659  * i40e_switch_branch_release - Delete a branch of the switch tree
12660  * @branch: where to start deleting
12661  *
12662  * This uses recursion to find the tips of the branch to be
12663  * removed, deleting until we get back to and can delete this VEB.
12664  **/
12665 static void i40e_switch_branch_release(struct i40e_veb *branch)
12666 {
12667 	struct i40e_pf *pf = branch->pf;
12668 	u16 branch_seid = branch->seid;
12669 	u16 veb_idx = branch->idx;
12670 	int i;
12671 
12672 	/* release any VEBs on this VEB - RECURSION */
12673 	for (i = 0; i < I40E_MAX_VEB; i++) {
12674 		if (!pf->veb[i])
12675 			continue;
12676 		if (pf->veb[i]->uplink_seid == branch->seid)
12677 			i40e_switch_branch_release(pf->veb[i]);
12678 	}
12679 
12680 	/* Release the VSIs on this VEB, but not the owner VSI.
12681 	 *
12682 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
12683 	 *       the VEB itself, so don't use (*branch) after this loop.
12684 	 */
12685 	for (i = 0; i < pf->num_alloc_vsi; i++) {
12686 		if (!pf->vsi[i])
12687 			continue;
12688 		if (pf->vsi[i]->uplink_seid == branch_seid &&
12689 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
12690 			i40e_vsi_release(pf->vsi[i]);
12691 		}
12692 	}
12693 
12694 	/* There's one corner case where the VEB might not have been
12695 	 * removed, so double check it here and remove it if needed.
12696 	 * This case happens if the veb was created from the debugfs
12697 	 * commands and no VSIs were added to it.
12698 	 */
12699 	if (pf->veb[veb_idx])
12700 		i40e_veb_release(pf->veb[veb_idx]);
12701 }
12702 
12703 /**
12704  * i40e_veb_clear - remove veb struct
12705  * @veb: the veb to remove
12706  **/
12707 static void i40e_veb_clear(struct i40e_veb *veb)
12708 {
12709 	if (!veb)
12710 		return;
12711 
12712 	if (veb->pf) {
12713 		struct i40e_pf *pf = veb->pf;
12714 
12715 		mutex_lock(&pf->switch_mutex);
12716 		if (pf->veb[veb->idx] == veb)
12717 			pf->veb[veb->idx] = NULL;
12718 		mutex_unlock(&pf->switch_mutex);
12719 	}
12720 
12721 	kfree(veb);
12722 }
12723 
12724 /**
12725  * i40e_veb_release - Delete a VEB and free its resources
12726  * @veb: the VEB being removed
12727  **/
12728 void i40e_veb_release(struct i40e_veb *veb)
12729 {
12730 	struct i40e_vsi *vsi = NULL;
12731 	struct i40e_pf *pf;
12732 	int i, n = 0;
12733 
12734 	pf = veb->pf;
12735 
12736 	/* find the remaining VSI and check for extras */
12737 	for (i = 0; i < pf->num_alloc_vsi; i++) {
12738 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
12739 			n++;
12740 			vsi = pf->vsi[i];
12741 		}
12742 	}
12743 	if (n != 1) {
12744 		dev_info(&pf->pdev->dev,
12745 			 "can't remove VEB %d with %d VSIs left\n",
12746 			 veb->seid, n);
12747 		return;
12748 	}
12749 
12750 	/* move the remaining VSI to uplink veb */
12751 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
12752 	if (veb->uplink_seid) {
12753 		vsi->uplink_seid = veb->uplink_seid;
12754 		if (veb->uplink_seid == pf->mac_seid)
12755 			vsi->veb_idx = I40E_NO_VEB;
12756 		else
12757 			vsi->veb_idx = veb->veb_idx;
12758 	} else {
12759 		/* floating VEB */
12760 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
12761 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
12762 	}
12763 
12764 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
12765 	i40e_veb_clear(veb);
12766 }
12767 
12768 /**
12769  * i40e_add_veb - create the VEB in the switch
12770  * @veb: the VEB to be instantiated
12771  * @vsi: the controlling VSI
12772  **/
12773 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
12774 {
12775 	struct i40e_pf *pf = veb->pf;
12776 	bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
12777 	int ret;
12778 
12779 	ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
12780 			      veb->enabled_tc, false,
12781 			      &veb->seid, enable_stats, NULL);
12782 
12783 	/* get a VEB from the hardware */
12784 	if (ret) {
12785 		dev_info(&pf->pdev->dev,
12786 			 "couldn't add VEB, err %s aq_err %s\n",
12787 			 i40e_stat_str(&pf->hw, ret),
12788 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12789 		return -EPERM;
12790 	}
12791 
12792 	/* get statistics counter */
12793 	ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
12794 					 &veb->stats_idx, NULL, NULL, NULL);
12795 	if (ret) {
12796 		dev_info(&pf->pdev->dev,
12797 			 "couldn't get VEB statistics idx, err %s aq_err %s\n",
12798 			 i40e_stat_str(&pf->hw, ret),
12799 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12800 		return -EPERM;
12801 	}
12802 	ret = i40e_veb_get_bw_info(veb);
12803 	if (ret) {
12804 		dev_info(&pf->pdev->dev,
12805 			 "couldn't get VEB bw info, err %s aq_err %s\n",
12806 			 i40e_stat_str(&pf->hw, ret),
12807 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
12808 		i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
12809 		return -ENOENT;
12810 	}
12811 
12812 	vsi->uplink_seid = veb->seid;
12813 	vsi->veb_idx = veb->idx;
12814 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
12815 
12816 	return 0;
12817 }
12818 
12819 /**
12820  * i40e_veb_setup - Set up a VEB
12821  * @pf: board private structure
12822  * @flags: VEB setup flags
12823  * @uplink_seid: the switch element to link to
12824  * @vsi_seid: the initial VSI seid
12825  * @enabled_tc: Enabled TC bit-map
12826  *
12827  * This allocates the sw VEB structure and links it into the switch
12828  * It is possible and legal for this to be a duplicate of an already
12829  * existing VEB.  It is also possible for both uplink and vsi seids
12830  * to be zero, in order to create a floating VEB.
12831  *
12832  * Returns pointer to the successfully allocated VEB sw struct on
12833  * success, otherwise returns NULL on failure.
12834  **/
12835 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
12836 				u16 uplink_seid, u16 vsi_seid,
12837 				u8 enabled_tc)
12838 {
12839 	struct i40e_veb *veb, *uplink_veb = NULL;
12840 	int vsi_idx, veb_idx;
12841 	int ret;
12842 
12843 	/* if one seid is 0, the other must be 0 to create a floating relay */
12844 	if ((uplink_seid == 0 || vsi_seid == 0) &&
12845 	    (uplink_seid + vsi_seid != 0)) {
12846 		dev_info(&pf->pdev->dev,
12847 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
12848 			 uplink_seid, vsi_seid);
12849 		return NULL;
12850 	}
12851 
12852 	/* make sure there is such a vsi and uplink */
12853 	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
12854 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
12855 			break;
12856 	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
12857 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
12858 			 vsi_seid);
12859 		return NULL;
12860 	}
12861 
12862 	if (uplink_seid && uplink_seid != pf->mac_seid) {
12863 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
12864 			if (pf->veb[veb_idx] &&
12865 			    pf->veb[veb_idx]->seid == uplink_seid) {
12866 				uplink_veb = pf->veb[veb_idx];
12867 				break;
12868 			}
12869 		}
12870 		if (!uplink_veb) {
12871 			dev_info(&pf->pdev->dev,
12872 				 "uplink seid %d not found\n", uplink_seid);
12873 			return NULL;
12874 		}
12875 	}
12876 
12877 	/* get veb sw struct */
12878 	veb_idx = i40e_veb_mem_alloc(pf);
12879 	if (veb_idx < 0)
12880 		goto err_alloc;
12881 	veb = pf->veb[veb_idx];
12882 	veb->flags = flags;
12883 	veb->uplink_seid = uplink_seid;
12884 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
12885 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
12886 
12887 	/* create the VEB in the switch */
12888 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
12889 	if (ret)
12890 		goto err_veb;
12891 	if (vsi_idx == pf->lan_vsi)
12892 		pf->lan_veb = veb->idx;
12893 
12894 	return veb;
12895 
12896 err_veb:
12897 	i40e_veb_clear(veb);
12898 err_alloc:
12899 	return NULL;
12900 }
12901 
12902 /**
12903  * i40e_setup_pf_switch_element - set PF vars based on switch type
12904  * @pf: board private structure
12905  * @ele: element we are building info from
12906  * @num_reported: total number of elements
12907  * @printconfig: should we print the contents
12908  *
12909  * helper function to assist in extracting a few useful SEID values.
12910  **/
12911 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
12912 				struct i40e_aqc_switch_config_element_resp *ele,
12913 				u16 num_reported, bool printconfig)
12914 {
12915 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
12916 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
12917 	u8 element_type = ele->element_type;
12918 	u16 seid = le16_to_cpu(ele->seid);
12919 
12920 	if (printconfig)
12921 		dev_info(&pf->pdev->dev,
12922 			 "type=%d seid=%d uplink=%d downlink=%d\n",
12923 			 element_type, seid, uplink_seid, downlink_seid);
12924 
12925 	switch (element_type) {
12926 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
12927 		pf->mac_seid = seid;
12928 		break;
12929 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
12930 		/* Main VEB? */
12931 		if (uplink_seid != pf->mac_seid)
12932 			break;
12933 		if (pf->lan_veb == I40E_NO_VEB) {
12934 			int v;
12935 
12936 			/* find existing or else empty VEB */
12937 			for (v = 0; v < I40E_MAX_VEB; v++) {
12938 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
12939 					pf->lan_veb = v;
12940 					break;
12941 				}
12942 			}
12943 			if (pf->lan_veb == I40E_NO_VEB) {
12944 				v = i40e_veb_mem_alloc(pf);
12945 				if (v < 0)
12946 					break;
12947 				pf->lan_veb = v;
12948 			}
12949 		}
12950 
12951 		pf->veb[pf->lan_veb]->seid = seid;
12952 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
12953 		pf->veb[pf->lan_veb]->pf = pf;
12954 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
12955 		break;
12956 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
12957 		if (num_reported != 1)
12958 			break;
12959 		/* This is immediately after a reset so we can assume this is
12960 		 * the PF's VSI
12961 		 */
12962 		pf->mac_seid = uplink_seid;
12963 		pf->pf_seid = downlink_seid;
12964 		pf->main_vsi_seid = seid;
12965 		if (printconfig)
12966 			dev_info(&pf->pdev->dev,
12967 				 "pf_seid=%d main_vsi_seid=%d\n",
12968 				 pf->pf_seid, pf->main_vsi_seid);
12969 		break;
12970 	case I40E_SWITCH_ELEMENT_TYPE_PF:
12971 	case I40E_SWITCH_ELEMENT_TYPE_VF:
12972 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
12973 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
12974 	case I40E_SWITCH_ELEMENT_TYPE_PE:
12975 	case I40E_SWITCH_ELEMENT_TYPE_PA:
12976 		/* ignore these for now */
12977 		break;
12978 	default:
12979 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
12980 			 element_type, seid);
12981 		break;
12982 	}
12983 }
12984 
12985 /**
12986  * i40e_fetch_switch_configuration - Get switch config from firmware
12987  * @pf: board private structure
12988  * @printconfig: should we print the contents
12989  *
12990  * Get the current switch configuration from the device and
12991  * extract a few useful SEID values.
12992  **/
12993 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
12994 {
12995 	struct i40e_aqc_get_switch_config_resp *sw_config;
12996 	u16 next_seid = 0;
12997 	int ret = 0;
12998 	u8 *aq_buf;
12999 	int i;
13000 
13001 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
13002 	if (!aq_buf)
13003 		return -ENOMEM;
13004 
13005 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
13006 	do {
13007 		u16 num_reported, num_total;
13008 
13009 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
13010 						I40E_AQ_LARGE_BUF,
13011 						&next_seid, NULL);
13012 		if (ret) {
13013 			dev_info(&pf->pdev->dev,
13014 				 "get switch config failed err %s aq_err %s\n",
13015 				 i40e_stat_str(&pf->hw, ret),
13016 				 i40e_aq_str(&pf->hw,
13017 					     pf->hw.aq.asq_last_status));
13018 			kfree(aq_buf);
13019 			return -ENOENT;
13020 		}
13021 
13022 		num_reported = le16_to_cpu(sw_config->header.num_reported);
13023 		num_total = le16_to_cpu(sw_config->header.num_total);
13024 
13025 		if (printconfig)
13026 			dev_info(&pf->pdev->dev,
13027 				 "header: %d reported %d total\n",
13028 				 num_reported, num_total);
13029 
13030 		for (i = 0; i < num_reported; i++) {
13031 			struct i40e_aqc_switch_config_element_resp *ele =
13032 				&sw_config->element[i];
13033 
13034 			i40e_setup_pf_switch_element(pf, ele, num_reported,
13035 						     printconfig);
13036 		}
13037 	} while (next_seid != 0);
13038 
13039 	kfree(aq_buf);
13040 	return ret;
13041 }
13042 
13043 /**
13044  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
13045  * @pf: board private structure
13046  * @reinit: if the Main VSI needs to re-initialized.
13047  *
13048  * Returns 0 on success, negative value on failure
13049  **/
13050 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
13051 {
13052 	u16 flags = 0;
13053 	int ret;
13054 
13055 	/* find out what's out there already */
13056 	ret = i40e_fetch_switch_configuration(pf, false);
13057 	if (ret) {
13058 		dev_info(&pf->pdev->dev,
13059 			 "couldn't fetch switch config, err %s aq_err %s\n",
13060 			 i40e_stat_str(&pf->hw, ret),
13061 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13062 		return ret;
13063 	}
13064 	i40e_pf_reset_stats(pf);
13065 
13066 	/* set the switch config bit for the whole device to
13067 	 * support limited promisc or true promisc
13068 	 * when user requests promisc. The default is limited
13069 	 * promisc.
13070 	*/
13071 
13072 	if ((pf->hw.pf_id == 0) &&
13073 	    !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT)) {
13074 		flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
13075 		pf->last_sw_conf_flags = flags;
13076 	}
13077 
13078 	if (pf->hw.pf_id == 0) {
13079 		u16 valid_flags;
13080 
13081 		valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
13082 		ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags, 0,
13083 						NULL);
13084 		if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
13085 			dev_info(&pf->pdev->dev,
13086 				 "couldn't set switch config bits, err %s aq_err %s\n",
13087 				 i40e_stat_str(&pf->hw, ret),
13088 				 i40e_aq_str(&pf->hw,
13089 					     pf->hw.aq.asq_last_status));
13090 			/* not a fatal problem, just keep going */
13091 		}
13092 		pf->last_sw_conf_valid_flags = valid_flags;
13093 	}
13094 
13095 	/* first time setup */
13096 	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
13097 		struct i40e_vsi *vsi = NULL;
13098 		u16 uplink_seid;
13099 
13100 		/* Set up the PF VSI associated with the PF's main VSI
13101 		 * that is already in the HW switch
13102 		 */
13103 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
13104 			uplink_seid = pf->veb[pf->lan_veb]->seid;
13105 		else
13106 			uplink_seid = pf->mac_seid;
13107 		if (pf->lan_vsi == I40E_NO_VSI)
13108 			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
13109 		else if (reinit)
13110 			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
13111 		if (!vsi) {
13112 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
13113 			i40e_cloud_filter_exit(pf);
13114 			i40e_fdir_teardown(pf);
13115 			return -EAGAIN;
13116 		}
13117 	} else {
13118 		/* force a reset of TC and queue layout configurations */
13119 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
13120 
13121 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
13122 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
13123 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
13124 	}
13125 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
13126 
13127 	i40e_fdir_sb_setup(pf);
13128 
13129 	/* Setup static PF queue filter control settings */
13130 	ret = i40e_setup_pf_filter_control(pf);
13131 	if (ret) {
13132 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
13133 			 ret);
13134 		/* Failure here should not stop continuing other steps */
13135 	}
13136 
13137 	/* enable RSS in the HW, even for only one queue, as the stack can use
13138 	 * the hash
13139 	 */
13140 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
13141 		i40e_pf_config_rss(pf);
13142 
13143 	/* fill in link information and enable LSE reporting */
13144 	i40e_link_event(pf);
13145 
13146 	/* Initialize user-specific link properties */
13147 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
13148 				  I40E_AQ_AN_COMPLETED) ? true : false);
13149 
13150 	i40e_ptp_init(pf);
13151 
13152 	/* repopulate tunnel port filters */
13153 	i40e_sync_udp_filters(pf);
13154 
13155 	return ret;
13156 }
13157 
13158 /**
13159  * i40e_determine_queue_usage - Work out queue distribution
13160  * @pf: board private structure
13161  **/
13162 static void i40e_determine_queue_usage(struct i40e_pf *pf)
13163 {
13164 	int queues_left;
13165 	int q_max;
13166 
13167 	pf->num_lan_qps = 0;
13168 
13169 	/* Find the max queues to be put into basic use.  We'll always be
13170 	 * using TC0, whether or not DCB is running, and TC0 will get the
13171 	 * big RSS set.
13172 	 */
13173 	queues_left = pf->hw.func_caps.num_tx_qp;
13174 
13175 	if ((queues_left == 1) ||
13176 	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
13177 		/* one qp for PF, no queues for anything else */
13178 		queues_left = 0;
13179 		pf->alloc_rss_size = pf->num_lan_qps = 1;
13180 
13181 		/* make sure all the fancies are disabled */
13182 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
13183 			       I40E_FLAG_IWARP_ENABLED	|
13184 			       I40E_FLAG_FD_SB_ENABLED	|
13185 			       I40E_FLAG_FD_ATR_ENABLED	|
13186 			       I40E_FLAG_DCB_CAPABLE	|
13187 			       I40E_FLAG_DCB_ENABLED	|
13188 			       I40E_FLAG_SRIOV_ENABLED	|
13189 			       I40E_FLAG_VMDQ_ENABLED);
13190 		pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
13191 	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
13192 				  I40E_FLAG_FD_SB_ENABLED |
13193 				  I40E_FLAG_FD_ATR_ENABLED |
13194 				  I40E_FLAG_DCB_CAPABLE))) {
13195 		/* one qp for PF */
13196 		pf->alloc_rss_size = pf->num_lan_qps = 1;
13197 		queues_left -= pf->num_lan_qps;
13198 
13199 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
13200 			       I40E_FLAG_IWARP_ENABLED	|
13201 			       I40E_FLAG_FD_SB_ENABLED	|
13202 			       I40E_FLAG_FD_ATR_ENABLED	|
13203 			       I40E_FLAG_DCB_ENABLED	|
13204 			       I40E_FLAG_VMDQ_ENABLED);
13205 		pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
13206 	} else {
13207 		/* Not enough queues for all TCs */
13208 		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
13209 		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
13210 			pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
13211 					I40E_FLAG_DCB_ENABLED);
13212 			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
13213 		}
13214 
13215 		/* limit lan qps to the smaller of qps, cpus or msix */
13216 		q_max = max_t(int, pf->rss_size_max, num_online_cpus());
13217 		q_max = min_t(int, q_max, pf->hw.func_caps.num_tx_qp);
13218 		q_max = min_t(int, q_max, pf->hw.func_caps.num_msix_vectors);
13219 		pf->num_lan_qps = q_max;
13220 
13221 		queues_left -= pf->num_lan_qps;
13222 	}
13223 
13224 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
13225 		if (queues_left > 1) {
13226 			queues_left -= 1; /* save 1 queue for FD */
13227 		} else {
13228 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
13229 			pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
13230 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
13231 		}
13232 	}
13233 
13234 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
13235 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
13236 		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
13237 					(queues_left / pf->num_vf_qps));
13238 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
13239 	}
13240 
13241 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
13242 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
13243 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
13244 					  (queues_left / pf->num_vmdq_qps));
13245 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
13246 	}
13247 
13248 	pf->queues_left = queues_left;
13249 	dev_dbg(&pf->pdev->dev,
13250 		"qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
13251 		pf->hw.func_caps.num_tx_qp,
13252 		!!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
13253 		pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
13254 		pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
13255 		queues_left);
13256 }
13257 
13258 /**
13259  * i40e_setup_pf_filter_control - Setup PF static filter control
13260  * @pf: PF to be setup
13261  *
13262  * i40e_setup_pf_filter_control sets up a PF's initial filter control
13263  * settings. If PE/FCoE are enabled then it will also set the per PF
13264  * based filter sizes required for them. It also enables Flow director,
13265  * ethertype and macvlan type filter settings for the pf.
13266  *
13267  * Returns 0 on success, negative on failure
13268  **/
13269 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
13270 {
13271 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
13272 
13273 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
13274 
13275 	/* Flow Director is enabled */
13276 	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
13277 		settings->enable_fdir = true;
13278 
13279 	/* Ethtype and MACVLAN filters enabled for PF */
13280 	settings->enable_ethtype = true;
13281 	settings->enable_macvlan = true;
13282 
13283 	if (i40e_set_filter_control(&pf->hw, settings))
13284 		return -ENOENT;
13285 
13286 	return 0;
13287 }
13288 
13289 #define INFO_STRING_LEN 255
13290 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
13291 static void i40e_print_features(struct i40e_pf *pf)
13292 {
13293 	struct i40e_hw *hw = &pf->hw;
13294 	char *buf;
13295 	int i;
13296 
13297 	buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
13298 	if (!buf)
13299 		return;
13300 
13301 	i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
13302 #ifdef CONFIG_PCI_IOV
13303 	i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
13304 #endif
13305 	i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
13306 		      pf->hw.func_caps.num_vsis,
13307 		      pf->vsi[pf->lan_vsi]->num_queue_pairs);
13308 	if (pf->flags & I40E_FLAG_RSS_ENABLED)
13309 		i += snprintf(&buf[i], REMAIN(i), " RSS");
13310 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
13311 		i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
13312 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
13313 		i += snprintf(&buf[i], REMAIN(i), " FD_SB");
13314 		i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
13315 	}
13316 	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
13317 		i += snprintf(&buf[i], REMAIN(i), " DCB");
13318 	i += snprintf(&buf[i], REMAIN(i), " VxLAN");
13319 	i += snprintf(&buf[i], REMAIN(i), " Geneve");
13320 	if (pf->flags & I40E_FLAG_PTP)
13321 		i += snprintf(&buf[i], REMAIN(i), " PTP");
13322 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
13323 		i += snprintf(&buf[i], REMAIN(i), " VEB");
13324 	else
13325 		i += snprintf(&buf[i], REMAIN(i), " VEPA");
13326 
13327 	dev_info(&pf->pdev->dev, "%s\n", buf);
13328 	kfree(buf);
13329 	WARN_ON(i > INFO_STRING_LEN);
13330 }
13331 
13332 /**
13333  * i40e_get_platform_mac_addr - get platform-specific MAC address
13334  * @pdev: PCI device information struct
13335  * @pf: board private structure
13336  *
13337  * Look up the MAC address for the device. First we'll try
13338  * eth_platform_get_mac_address, which will check Open Firmware, or arch
13339  * specific fallback. Otherwise, we'll default to the stored value in
13340  * firmware.
13341  **/
13342 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
13343 {
13344 	if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
13345 		i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
13346 }
13347 
13348 /**
13349  * i40e_probe - Device initialization routine
13350  * @pdev: PCI device information struct
13351  * @ent: entry in i40e_pci_tbl
13352  *
13353  * i40e_probe initializes a PF identified by a pci_dev structure.
13354  * The OS initialization, configuring of the PF private structure,
13355  * and a hardware reset occur.
13356  *
13357  * Returns 0 on success, negative on failure
13358  **/
13359 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
13360 {
13361 	struct i40e_aq_get_phy_abilities_resp abilities;
13362 	struct i40e_pf *pf;
13363 	struct i40e_hw *hw;
13364 	static u16 pfs_found;
13365 	u16 wol_nvm_bits;
13366 	u16 link_status;
13367 	int err;
13368 	u32 val;
13369 	u32 i;
13370 	u8 set_fc_aq_fail;
13371 
13372 	err = pci_enable_device_mem(pdev);
13373 	if (err)
13374 		return err;
13375 
13376 	/* set up for high or low dma */
13377 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
13378 	if (err) {
13379 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
13380 		if (err) {
13381 			dev_err(&pdev->dev,
13382 				"DMA configuration failed: 0x%x\n", err);
13383 			goto err_dma;
13384 		}
13385 	}
13386 
13387 	/* set up pci connections */
13388 	err = pci_request_mem_regions(pdev, i40e_driver_name);
13389 	if (err) {
13390 		dev_info(&pdev->dev,
13391 			 "pci_request_selected_regions failed %d\n", err);
13392 		goto err_pci_reg;
13393 	}
13394 
13395 	pci_enable_pcie_error_reporting(pdev);
13396 	pci_set_master(pdev);
13397 
13398 	/* Now that we have a PCI connection, we need to do the
13399 	 * low level device setup.  This is primarily setting up
13400 	 * the Admin Queue structures and then querying for the
13401 	 * device's current profile information.
13402 	 */
13403 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
13404 	if (!pf) {
13405 		err = -ENOMEM;
13406 		goto err_pf_alloc;
13407 	}
13408 	pf->next_vsi = 0;
13409 	pf->pdev = pdev;
13410 	set_bit(__I40E_DOWN, pf->state);
13411 
13412 	hw = &pf->hw;
13413 	hw->back = pf;
13414 
13415 	pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
13416 				I40E_MAX_CSR_SPACE);
13417 
13418 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
13419 	if (!hw->hw_addr) {
13420 		err = -EIO;
13421 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
13422 			 (unsigned int)pci_resource_start(pdev, 0),
13423 			 pf->ioremap_len, err);
13424 		goto err_ioremap;
13425 	}
13426 	hw->vendor_id = pdev->vendor;
13427 	hw->device_id = pdev->device;
13428 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
13429 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
13430 	hw->subsystem_device_id = pdev->subsystem_device;
13431 	hw->bus.device = PCI_SLOT(pdev->devfn);
13432 	hw->bus.func = PCI_FUNC(pdev->devfn);
13433 	hw->bus.bus_id = pdev->bus->number;
13434 	pf->instance = pfs_found;
13435 
13436 	/* Select something other than the 802.1ad ethertype for the
13437 	 * switch to use internally and drop on ingress.
13438 	 */
13439 	hw->switch_tag = 0xffff;
13440 	hw->first_tag = ETH_P_8021AD;
13441 	hw->second_tag = ETH_P_8021Q;
13442 
13443 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
13444 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
13445 
13446 	/* set up the locks for the AQ, do this only once in probe
13447 	 * and destroy them only once in remove
13448 	 */
13449 	mutex_init(&hw->aq.asq_mutex);
13450 	mutex_init(&hw->aq.arq_mutex);
13451 
13452 	pf->msg_enable = netif_msg_init(debug,
13453 					NETIF_MSG_DRV |
13454 					NETIF_MSG_PROBE |
13455 					NETIF_MSG_LINK);
13456 	if (debug < -1)
13457 		pf->hw.debug_mask = debug;
13458 
13459 	/* do a special CORER for clearing PXE mode once at init */
13460 	if (hw->revision_id == 0 &&
13461 	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
13462 		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
13463 		i40e_flush(hw);
13464 		msleep(200);
13465 		pf->corer_count++;
13466 
13467 		i40e_clear_pxe_mode(hw);
13468 	}
13469 
13470 	/* Reset here to make sure all is clean and to define PF 'n' */
13471 	i40e_clear_hw(hw);
13472 	err = i40e_pf_reset(hw);
13473 	if (err) {
13474 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
13475 		goto err_pf_reset;
13476 	}
13477 	pf->pfr_count++;
13478 
13479 	hw->aq.num_arq_entries = I40E_AQ_LEN;
13480 	hw->aq.num_asq_entries = I40E_AQ_LEN;
13481 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
13482 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
13483 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
13484 
13485 	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
13486 		 "%s-%s:misc",
13487 		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
13488 
13489 	err = i40e_init_shared_code(hw);
13490 	if (err) {
13491 		dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
13492 			 err);
13493 		goto err_pf_reset;
13494 	}
13495 
13496 	/* set up a default setting for link flow control */
13497 	pf->hw.fc.requested_mode = I40E_FC_NONE;
13498 
13499 	err = i40e_init_adminq(hw);
13500 	if (err) {
13501 		if (err == I40E_ERR_FIRMWARE_API_VERSION)
13502 			dev_info(&pdev->dev,
13503 				 "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");
13504 		else
13505 			dev_info(&pdev->dev,
13506 				 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
13507 
13508 		goto err_pf_reset;
13509 	}
13510 	i40e_get_oem_version(hw);
13511 
13512 	/* provide nvm, fw, api versions */
13513 	dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
13514 		 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
13515 		 hw->aq.api_maj_ver, hw->aq.api_min_ver,
13516 		 i40e_nvm_version_str(hw));
13517 
13518 	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
13519 	    hw->aq.api_min_ver > I40E_FW_MINOR_VERSION(hw))
13520 		dev_info(&pdev->dev,
13521 			 "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");
13522 	else if (hw->aq.api_maj_ver == 1 && hw->aq.api_min_ver < 4)
13523 		dev_info(&pdev->dev,
13524 			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
13525 
13526 	i40e_verify_eeprom(pf);
13527 
13528 	/* Rev 0 hardware was never productized */
13529 	if (hw->revision_id < 1)
13530 		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");
13531 
13532 	i40e_clear_pxe_mode(hw);
13533 	err = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
13534 	if (err)
13535 		goto err_adminq_setup;
13536 
13537 	err = i40e_sw_init(pf);
13538 	if (err) {
13539 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
13540 		goto err_sw_init;
13541 	}
13542 
13543 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
13544 				hw->func_caps.num_rx_qp, 0, 0);
13545 	if (err) {
13546 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
13547 		goto err_init_lan_hmc;
13548 	}
13549 
13550 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
13551 	if (err) {
13552 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
13553 		err = -ENOENT;
13554 		goto err_configure_lan_hmc;
13555 	}
13556 
13557 	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
13558 	 * Ignore error return codes because if it was already disabled via
13559 	 * hardware settings this will fail
13560 	 */
13561 	if (pf->hw_features & I40E_HW_STOP_FW_LLDP) {
13562 		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
13563 		i40e_aq_stop_lldp(hw, true, NULL);
13564 	}
13565 
13566 	/* allow a platform config to override the HW addr */
13567 	i40e_get_platform_mac_addr(pdev, pf);
13568 
13569 	if (!is_valid_ether_addr(hw->mac.addr)) {
13570 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
13571 		err = -EIO;
13572 		goto err_mac_addr;
13573 	}
13574 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
13575 	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
13576 	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
13577 	if (is_valid_ether_addr(hw->mac.port_addr))
13578 		pf->hw_features |= I40E_HW_PORT_ID_VALID;
13579 
13580 	pci_set_drvdata(pdev, pf);
13581 	pci_save_state(pdev);
13582 #ifdef CONFIG_I40E_DCB
13583 	err = i40e_init_pf_dcb(pf);
13584 	if (err) {
13585 		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
13586 		pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
13587 		/* Continue without DCB enabled */
13588 	}
13589 #endif /* CONFIG_I40E_DCB */
13590 
13591 	/* set up periodic task facility */
13592 	timer_setup(&pf->service_timer, i40e_service_timer, 0);
13593 	pf->service_timer_period = HZ;
13594 
13595 	INIT_WORK(&pf->service_task, i40e_service_task);
13596 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
13597 
13598 	/* NVM bit on means WoL disabled for the port */
13599 	i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
13600 	if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
13601 		pf->wol_en = false;
13602 	else
13603 		pf->wol_en = true;
13604 	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
13605 
13606 	/* set up the main switch operations */
13607 	i40e_determine_queue_usage(pf);
13608 	err = i40e_init_interrupt_scheme(pf);
13609 	if (err)
13610 		goto err_switch_setup;
13611 
13612 	/* The number of VSIs reported by the FW is the minimum guaranteed
13613 	 * to us; HW supports far more and we share the remaining pool with
13614 	 * the other PFs. We allocate space for more than the guarantee with
13615 	 * the understanding that we might not get them all later.
13616 	 */
13617 	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
13618 		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
13619 	else
13620 		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
13621 
13622 	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
13623 	pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
13624 			  GFP_KERNEL);
13625 	if (!pf->vsi) {
13626 		err = -ENOMEM;
13627 		goto err_switch_setup;
13628 	}
13629 
13630 #ifdef CONFIG_PCI_IOV
13631 	/* prep for VF support */
13632 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
13633 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
13634 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
13635 		if (pci_num_vf(pdev))
13636 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
13637 	}
13638 #endif
13639 	err = i40e_setup_pf_switch(pf, false);
13640 	if (err) {
13641 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
13642 		goto err_vsis;
13643 	}
13644 	INIT_LIST_HEAD(&pf->vsi[pf->lan_vsi]->ch_list);
13645 
13646 	/* Make sure flow control is set according to current settings */
13647 	err = i40e_set_fc(hw, &set_fc_aq_fail, true);
13648 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
13649 		dev_dbg(&pf->pdev->dev,
13650 			"Set fc with err %s aq_err %s on get_phy_cap\n",
13651 			i40e_stat_str(hw, err),
13652 			i40e_aq_str(hw, hw->aq.asq_last_status));
13653 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
13654 		dev_dbg(&pf->pdev->dev,
13655 			"Set fc with err %s aq_err %s on set_phy_config\n",
13656 			i40e_stat_str(hw, err),
13657 			i40e_aq_str(hw, hw->aq.asq_last_status));
13658 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
13659 		dev_dbg(&pf->pdev->dev,
13660 			"Set fc with err %s aq_err %s on get_link_info\n",
13661 			i40e_stat_str(hw, err),
13662 			i40e_aq_str(hw, hw->aq.asq_last_status));
13663 
13664 	/* if FDIR VSI was set up, start it now */
13665 	for (i = 0; i < pf->num_alloc_vsi; i++) {
13666 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
13667 			i40e_vsi_open(pf->vsi[i]);
13668 			break;
13669 		}
13670 	}
13671 
13672 	/* The driver only wants link up/down and module qualification
13673 	 * reports from firmware.  Note the negative logic.
13674 	 */
13675 	err = i40e_aq_set_phy_int_mask(&pf->hw,
13676 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
13677 					 I40E_AQ_EVENT_MEDIA_NA |
13678 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
13679 	if (err)
13680 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
13681 			 i40e_stat_str(&pf->hw, err),
13682 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13683 
13684 	/* Reconfigure hardware for allowing smaller MSS in the case
13685 	 * of TSO, so that we avoid the MDD being fired and causing
13686 	 * a reset in the case of small MSS+TSO.
13687 	 */
13688 	val = rd32(hw, I40E_REG_MSS);
13689 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
13690 		val &= ~I40E_REG_MSS_MIN_MASK;
13691 		val |= I40E_64BYTE_MSS;
13692 		wr32(hw, I40E_REG_MSS, val);
13693 	}
13694 
13695 	if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
13696 		msleep(75);
13697 		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
13698 		if (err)
13699 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
13700 				 i40e_stat_str(&pf->hw, err),
13701 				 i40e_aq_str(&pf->hw,
13702 					     pf->hw.aq.asq_last_status));
13703 	}
13704 	/* The main driver is (mostly) up and happy. We need to set this state
13705 	 * before setting up the misc vector or we get a race and the vector
13706 	 * ends up disabled forever.
13707 	 */
13708 	clear_bit(__I40E_DOWN, pf->state);
13709 
13710 	/* In case of MSIX we are going to setup the misc vector right here
13711 	 * to handle admin queue events etc. In case of legacy and MSI
13712 	 * the misc functionality and queue processing is combined in
13713 	 * the same vector and that gets setup at open.
13714 	 */
13715 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
13716 		err = i40e_setup_misc_vector(pf);
13717 		if (err) {
13718 			dev_info(&pdev->dev,
13719 				 "setup of misc vector failed: %d\n", err);
13720 			goto err_vsis;
13721 		}
13722 	}
13723 
13724 #ifdef CONFIG_PCI_IOV
13725 	/* prep for VF support */
13726 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
13727 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
13728 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
13729 		/* disable link interrupts for VFs */
13730 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
13731 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
13732 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
13733 		i40e_flush(hw);
13734 
13735 		if (pci_num_vf(pdev)) {
13736 			dev_info(&pdev->dev,
13737 				 "Active VFs found, allocating resources.\n");
13738 			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
13739 			if (err)
13740 				dev_info(&pdev->dev,
13741 					 "Error %d allocating resources for existing VFs\n",
13742 					 err);
13743 		}
13744 	}
13745 #endif /* CONFIG_PCI_IOV */
13746 
13747 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
13748 		pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
13749 						      pf->num_iwarp_msix,
13750 						      I40E_IWARP_IRQ_PILE_ID);
13751 		if (pf->iwarp_base_vector < 0) {
13752 			dev_info(&pdev->dev,
13753 				 "failed to get tracking for %d vectors for IWARP err=%d\n",
13754 				 pf->num_iwarp_msix, pf->iwarp_base_vector);
13755 			pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
13756 		}
13757 	}
13758 
13759 	i40e_dbg_pf_init(pf);
13760 
13761 	/* tell the firmware that we're starting */
13762 	i40e_send_version(pf);
13763 
13764 	/* since everything's happy, start the service_task timer */
13765 	mod_timer(&pf->service_timer,
13766 		  round_jiffies(jiffies + pf->service_timer_period));
13767 
13768 	/* add this PF to client device list and launch a client service task */
13769 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
13770 		err = i40e_lan_add_device(pf);
13771 		if (err)
13772 			dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
13773 				 err);
13774 	}
13775 
13776 #define PCI_SPEED_SIZE 8
13777 #define PCI_WIDTH_SIZE 8
13778 	/* Devices on the IOSF bus do not have this information
13779 	 * and will report PCI Gen 1 x 1 by default so don't bother
13780 	 * checking them.
13781 	 */
13782 	if (!(pf->hw_features & I40E_HW_NO_PCI_LINK_CHECK)) {
13783 		char speed[PCI_SPEED_SIZE] = "Unknown";
13784 		char width[PCI_WIDTH_SIZE] = "Unknown";
13785 
13786 		/* Get the negotiated link width and speed from PCI config
13787 		 * space
13788 		 */
13789 		pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
13790 					  &link_status);
13791 
13792 		i40e_set_pci_config_data(hw, link_status);
13793 
13794 		switch (hw->bus.speed) {
13795 		case i40e_bus_speed_8000:
13796 			strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
13797 		case i40e_bus_speed_5000:
13798 			strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
13799 		case i40e_bus_speed_2500:
13800 			strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
13801 		default:
13802 			break;
13803 		}
13804 		switch (hw->bus.width) {
13805 		case i40e_bus_width_pcie_x8:
13806 			strncpy(width, "8", PCI_WIDTH_SIZE); break;
13807 		case i40e_bus_width_pcie_x4:
13808 			strncpy(width, "4", PCI_WIDTH_SIZE); break;
13809 		case i40e_bus_width_pcie_x2:
13810 			strncpy(width, "2", PCI_WIDTH_SIZE); break;
13811 		case i40e_bus_width_pcie_x1:
13812 			strncpy(width, "1", PCI_WIDTH_SIZE); break;
13813 		default:
13814 			break;
13815 		}
13816 
13817 		dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
13818 			 speed, width);
13819 
13820 		if (hw->bus.width < i40e_bus_width_pcie_x8 ||
13821 		    hw->bus.speed < i40e_bus_speed_8000) {
13822 			dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
13823 			dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
13824 		}
13825 	}
13826 
13827 	/* get the requested speeds from the fw */
13828 	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
13829 	if (err)
13830 		dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
13831 			i40e_stat_str(&pf->hw, err),
13832 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13833 	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
13834 
13835 	/* get the supported phy types from the fw */
13836 	err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
13837 	if (err)
13838 		dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
13839 			i40e_stat_str(&pf->hw, err),
13840 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13841 
13842 	/* Add a filter to drop all Flow control frames from any VSI from being
13843 	 * transmitted. By doing so we stop a malicious VF from sending out
13844 	 * PAUSE or PFC frames and potentially controlling traffic for other
13845 	 * PF/VF VSIs.
13846 	 * The FW can still send Flow control frames if enabled.
13847 	 */
13848 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
13849 						       pf->main_vsi_seid);
13850 
13851 	if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
13852 		(pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
13853 		pf->hw_features |= I40E_HW_PHY_CONTROLS_LEDS;
13854 	if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
13855 		pf->hw_features |= I40E_HW_HAVE_CRT_RETIMER;
13856 	/* print a string summarizing features */
13857 	i40e_print_features(pf);
13858 
13859 	return 0;
13860 
13861 	/* Unwind what we've done if something failed in the setup */
13862 err_vsis:
13863 	set_bit(__I40E_DOWN, pf->state);
13864 	i40e_clear_interrupt_scheme(pf);
13865 	kfree(pf->vsi);
13866 err_switch_setup:
13867 	i40e_reset_interrupt_capability(pf);
13868 	del_timer_sync(&pf->service_timer);
13869 err_mac_addr:
13870 err_configure_lan_hmc:
13871 	(void)i40e_shutdown_lan_hmc(hw);
13872 err_init_lan_hmc:
13873 	kfree(pf->qp_pile);
13874 err_sw_init:
13875 err_adminq_setup:
13876 err_pf_reset:
13877 	iounmap(hw->hw_addr);
13878 err_ioremap:
13879 	kfree(pf);
13880 err_pf_alloc:
13881 	pci_disable_pcie_error_reporting(pdev);
13882 	pci_release_mem_regions(pdev);
13883 err_pci_reg:
13884 err_dma:
13885 	pci_disable_device(pdev);
13886 	return err;
13887 }
13888 
13889 /**
13890  * i40e_remove - Device removal routine
13891  * @pdev: PCI device information struct
13892  *
13893  * i40e_remove is called by the PCI subsystem to alert the driver
13894  * that is should release a PCI device.  This could be caused by a
13895  * Hot-Plug event, or because the driver is going to be removed from
13896  * memory.
13897  **/
13898 static void i40e_remove(struct pci_dev *pdev)
13899 {
13900 	struct i40e_pf *pf = pci_get_drvdata(pdev);
13901 	struct i40e_hw *hw = &pf->hw;
13902 	i40e_status ret_code;
13903 	int i;
13904 
13905 	i40e_dbg_pf_exit(pf);
13906 
13907 	i40e_ptp_stop(pf);
13908 
13909 	/* Disable RSS in hw */
13910 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
13911 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
13912 
13913 	/* no more scheduling of any task */
13914 	set_bit(__I40E_SUSPENDED, pf->state);
13915 	set_bit(__I40E_DOWN, pf->state);
13916 	if (pf->service_timer.function)
13917 		del_timer_sync(&pf->service_timer);
13918 	if (pf->service_task.func)
13919 		cancel_work_sync(&pf->service_task);
13920 
13921 	/* Client close must be called explicitly here because the timer
13922 	 * has been stopped.
13923 	 */
13924 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
13925 
13926 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
13927 		i40e_free_vfs(pf);
13928 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
13929 	}
13930 
13931 	i40e_fdir_teardown(pf);
13932 
13933 	/* If there is a switch structure or any orphans, remove them.
13934 	 * This will leave only the PF's VSI remaining.
13935 	 */
13936 	for (i = 0; i < I40E_MAX_VEB; i++) {
13937 		if (!pf->veb[i])
13938 			continue;
13939 
13940 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
13941 		    pf->veb[i]->uplink_seid == 0)
13942 			i40e_switch_branch_release(pf->veb[i]);
13943 	}
13944 
13945 	/* Now we can shutdown the PF's VSI, just before we kill
13946 	 * adminq and hmc.
13947 	 */
13948 	if (pf->vsi[pf->lan_vsi])
13949 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
13950 
13951 	i40e_cloud_filter_exit(pf);
13952 
13953 	/* remove attached clients */
13954 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
13955 		ret_code = i40e_lan_del_device(pf);
13956 		if (ret_code)
13957 			dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
13958 				 ret_code);
13959 	}
13960 
13961 	/* shutdown and destroy the HMC */
13962 	if (hw->hmc.hmc_obj) {
13963 		ret_code = i40e_shutdown_lan_hmc(hw);
13964 		if (ret_code)
13965 			dev_warn(&pdev->dev,
13966 				 "Failed to destroy the HMC resources: %d\n",
13967 				 ret_code);
13968 	}
13969 
13970 	/* shutdown the adminq */
13971 	i40e_shutdown_adminq(hw);
13972 
13973 	/* destroy the locks only once, here */
13974 	mutex_destroy(&hw->aq.arq_mutex);
13975 	mutex_destroy(&hw->aq.asq_mutex);
13976 
13977 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
13978 	i40e_clear_interrupt_scheme(pf);
13979 	for (i = 0; i < pf->num_alloc_vsi; i++) {
13980 		if (pf->vsi[i]) {
13981 			i40e_vsi_clear_rings(pf->vsi[i]);
13982 			i40e_vsi_clear(pf->vsi[i]);
13983 			pf->vsi[i] = NULL;
13984 		}
13985 	}
13986 
13987 	for (i = 0; i < I40E_MAX_VEB; i++) {
13988 		kfree(pf->veb[i]);
13989 		pf->veb[i] = NULL;
13990 	}
13991 
13992 	kfree(pf->qp_pile);
13993 	kfree(pf->vsi);
13994 
13995 	iounmap(hw->hw_addr);
13996 	kfree(pf);
13997 	pci_release_mem_regions(pdev);
13998 
13999 	pci_disable_pcie_error_reporting(pdev);
14000 	pci_disable_device(pdev);
14001 }
14002 
14003 /**
14004  * i40e_pci_error_detected - warning that something funky happened in PCI land
14005  * @pdev: PCI device information struct
14006  *
14007  * Called to warn that something happened and the error handling steps
14008  * are in progress.  Allows the driver to quiesce things, be ready for
14009  * remediation.
14010  **/
14011 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
14012 						enum pci_channel_state error)
14013 {
14014 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14015 
14016 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
14017 
14018 	if (!pf) {
14019 		dev_info(&pdev->dev,
14020 			 "Cannot recover - error happened during device probe\n");
14021 		return PCI_ERS_RESULT_DISCONNECT;
14022 	}
14023 
14024 	/* shutdown all operations */
14025 	if (!test_bit(__I40E_SUSPENDED, pf->state))
14026 		i40e_prep_for_reset(pf, false);
14027 
14028 	/* Request a slot reset */
14029 	return PCI_ERS_RESULT_NEED_RESET;
14030 }
14031 
14032 /**
14033  * i40e_pci_error_slot_reset - a PCI slot reset just happened
14034  * @pdev: PCI device information struct
14035  *
14036  * Called to find if the driver can work with the device now that
14037  * the pci slot has been reset.  If a basic connection seems good
14038  * (registers are readable and have sane content) then return a
14039  * happy little PCI_ERS_RESULT_xxx.
14040  **/
14041 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
14042 {
14043 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14044 	pci_ers_result_t result;
14045 	int err;
14046 	u32 reg;
14047 
14048 	dev_dbg(&pdev->dev, "%s\n", __func__);
14049 	if (pci_enable_device_mem(pdev)) {
14050 		dev_info(&pdev->dev,
14051 			 "Cannot re-enable PCI device after reset.\n");
14052 		result = PCI_ERS_RESULT_DISCONNECT;
14053 	} else {
14054 		pci_set_master(pdev);
14055 		pci_restore_state(pdev);
14056 		pci_save_state(pdev);
14057 		pci_wake_from_d3(pdev, false);
14058 
14059 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
14060 		if (reg == 0)
14061 			result = PCI_ERS_RESULT_RECOVERED;
14062 		else
14063 			result = PCI_ERS_RESULT_DISCONNECT;
14064 	}
14065 
14066 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
14067 	if (err) {
14068 		dev_info(&pdev->dev,
14069 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
14070 			 err);
14071 		/* non-fatal, continue */
14072 	}
14073 
14074 	return result;
14075 }
14076 
14077 /**
14078  * i40e_pci_error_reset_prepare - prepare device driver for pci reset
14079  * @pdev: PCI device information struct
14080  */
14081 static void i40e_pci_error_reset_prepare(struct pci_dev *pdev)
14082 {
14083 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14084 
14085 	i40e_prep_for_reset(pf, false);
14086 }
14087 
14088 /**
14089  * i40e_pci_error_reset_done - pci reset done, device driver reset can begin
14090  * @pdev: PCI device information struct
14091  */
14092 static void i40e_pci_error_reset_done(struct pci_dev *pdev)
14093 {
14094 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14095 
14096 	i40e_reset_and_rebuild(pf, false, false);
14097 }
14098 
14099 /**
14100  * i40e_pci_error_resume - restart operations after PCI error recovery
14101  * @pdev: PCI device information struct
14102  *
14103  * Called to allow the driver to bring things back up after PCI error
14104  * and/or reset recovery has finished.
14105  **/
14106 static void i40e_pci_error_resume(struct pci_dev *pdev)
14107 {
14108 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14109 
14110 	dev_dbg(&pdev->dev, "%s\n", __func__);
14111 	if (test_bit(__I40E_SUSPENDED, pf->state))
14112 		return;
14113 
14114 	i40e_handle_reset_warning(pf, false);
14115 }
14116 
14117 /**
14118  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
14119  * using the mac_address_write admin q function
14120  * @pf: pointer to i40e_pf struct
14121  **/
14122 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
14123 {
14124 	struct i40e_hw *hw = &pf->hw;
14125 	i40e_status ret;
14126 	u8 mac_addr[6];
14127 	u16 flags = 0;
14128 
14129 	/* Get current MAC address in case it's an LAA */
14130 	if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
14131 		ether_addr_copy(mac_addr,
14132 				pf->vsi[pf->lan_vsi]->netdev->dev_addr);
14133 	} else {
14134 		dev_err(&pf->pdev->dev,
14135 			"Failed to retrieve MAC address; using default\n");
14136 		ether_addr_copy(mac_addr, hw->mac.addr);
14137 	}
14138 
14139 	/* The FW expects the mac address write cmd to first be called with
14140 	 * one of these flags before calling it again with the multicast
14141 	 * enable flags.
14142 	 */
14143 	flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
14144 
14145 	if (hw->func_caps.flex10_enable && hw->partition_id != 1)
14146 		flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
14147 
14148 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
14149 	if (ret) {
14150 		dev_err(&pf->pdev->dev,
14151 			"Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
14152 		return;
14153 	}
14154 
14155 	flags = I40E_AQC_MC_MAG_EN
14156 			| I40E_AQC_WOL_PRESERVE_ON_PFR
14157 			| I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
14158 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
14159 	if (ret)
14160 		dev_err(&pf->pdev->dev,
14161 			"Failed to enable Multicast Magic Packet wake up\n");
14162 }
14163 
14164 /**
14165  * i40e_shutdown - PCI callback for shutting down
14166  * @pdev: PCI device information struct
14167  **/
14168 static void i40e_shutdown(struct pci_dev *pdev)
14169 {
14170 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14171 	struct i40e_hw *hw = &pf->hw;
14172 
14173 	set_bit(__I40E_SUSPENDED, pf->state);
14174 	set_bit(__I40E_DOWN, pf->state);
14175 	rtnl_lock();
14176 	i40e_prep_for_reset(pf, true);
14177 	rtnl_unlock();
14178 
14179 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
14180 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
14181 
14182 	del_timer_sync(&pf->service_timer);
14183 	cancel_work_sync(&pf->service_task);
14184 	i40e_cloud_filter_exit(pf);
14185 	i40e_fdir_teardown(pf);
14186 
14187 	/* Client close must be called explicitly here because the timer
14188 	 * has been stopped.
14189 	 */
14190 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
14191 
14192 	if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
14193 		i40e_enable_mc_magic_wake(pf);
14194 
14195 	i40e_prep_for_reset(pf, false);
14196 
14197 	wr32(hw, I40E_PFPM_APM,
14198 	     (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
14199 	wr32(hw, I40E_PFPM_WUFC,
14200 	     (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
14201 
14202 	i40e_clear_interrupt_scheme(pf);
14203 
14204 	if (system_state == SYSTEM_POWER_OFF) {
14205 		pci_wake_from_d3(pdev, pf->wol_en);
14206 		pci_set_power_state(pdev, PCI_D3hot);
14207 	}
14208 }
14209 
14210 /**
14211  * i40e_suspend - PM callback for moving to D3
14212  * @dev: generic device information structure
14213  **/
14214 static int __maybe_unused i40e_suspend(struct device *dev)
14215 {
14216 	struct pci_dev *pdev = to_pci_dev(dev);
14217 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14218 	struct i40e_hw *hw = &pf->hw;
14219 
14220 	/* If we're already suspended, then there is nothing to do */
14221 	if (test_and_set_bit(__I40E_SUSPENDED, pf->state))
14222 		return 0;
14223 
14224 	set_bit(__I40E_DOWN, pf->state);
14225 
14226 	/* Ensure service task will not be running */
14227 	del_timer_sync(&pf->service_timer);
14228 	cancel_work_sync(&pf->service_task);
14229 
14230 	if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
14231 		i40e_enable_mc_magic_wake(pf);
14232 
14233 	i40e_prep_for_reset(pf, false);
14234 
14235 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
14236 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
14237 
14238 	/* Clear the interrupt scheme and release our IRQs so that the system
14239 	 * can safely hibernate even when there are a large number of CPUs.
14240 	 * Otherwise hibernation might fail when mapping all the vectors back
14241 	 * to CPU0.
14242 	 */
14243 	i40e_clear_interrupt_scheme(pf);
14244 
14245 	return 0;
14246 }
14247 
14248 /**
14249  * i40e_resume - PM callback for waking up from D3
14250  * @dev: generic device information structure
14251  **/
14252 static int __maybe_unused i40e_resume(struct device *dev)
14253 {
14254 	struct pci_dev *pdev = to_pci_dev(dev);
14255 	struct i40e_pf *pf = pci_get_drvdata(pdev);
14256 	int err;
14257 
14258 	/* If we're not suspended, then there is nothing to do */
14259 	if (!test_bit(__I40E_SUSPENDED, pf->state))
14260 		return 0;
14261 
14262 	/* We cleared the interrupt scheme when we suspended, so we need to
14263 	 * restore it now to resume device functionality.
14264 	 */
14265 	err = i40e_restore_interrupt_scheme(pf);
14266 	if (err) {
14267 		dev_err(&pdev->dev, "Cannot restore interrupt scheme: %d\n",
14268 			err);
14269 	}
14270 
14271 	clear_bit(__I40E_DOWN, pf->state);
14272 	i40e_reset_and_rebuild(pf, false, false);
14273 
14274 	/* Clear suspended state last after everything is recovered */
14275 	clear_bit(__I40E_SUSPENDED, pf->state);
14276 
14277 	/* Restart the service task */
14278 	mod_timer(&pf->service_timer,
14279 		  round_jiffies(jiffies + pf->service_timer_period));
14280 
14281 	return 0;
14282 }
14283 
14284 static const struct pci_error_handlers i40e_err_handler = {
14285 	.error_detected = i40e_pci_error_detected,
14286 	.slot_reset = i40e_pci_error_slot_reset,
14287 	.reset_prepare = i40e_pci_error_reset_prepare,
14288 	.reset_done = i40e_pci_error_reset_done,
14289 	.resume = i40e_pci_error_resume,
14290 };
14291 
14292 static SIMPLE_DEV_PM_OPS(i40e_pm_ops, i40e_suspend, i40e_resume);
14293 
14294 static struct pci_driver i40e_driver = {
14295 	.name     = i40e_driver_name,
14296 	.id_table = i40e_pci_tbl,
14297 	.probe    = i40e_probe,
14298 	.remove   = i40e_remove,
14299 	.driver   = {
14300 		.pm = &i40e_pm_ops,
14301 	},
14302 	.shutdown = i40e_shutdown,
14303 	.err_handler = &i40e_err_handler,
14304 	.sriov_configure = i40e_pci_sriov_configure,
14305 };
14306 
14307 /**
14308  * i40e_init_module - Driver registration routine
14309  *
14310  * i40e_init_module is the first routine called when the driver is
14311  * loaded. All it does is register with the PCI subsystem.
14312  **/
14313 static int __init i40e_init_module(void)
14314 {
14315 	pr_info("%s: %s - version %s\n", i40e_driver_name,
14316 		i40e_driver_string, i40e_driver_version_str);
14317 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
14318 
14319 	/* There is no need to throttle the number of active tasks because
14320 	 * each device limits its own task using a state bit for scheduling
14321 	 * the service task, and the device tasks do not interfere with each
14322 	 * other, so we don't set a max task limit. We must set WQ_MEM_RECLAIM
14323 	 * since we need to be able to guarantee forward progress even under
14324 	 * memory pressure.
14325 	 */
14326 	i40e_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, i40e_driver_name);
14327 	if (!i40e_wq) {
14328 		pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
14329 		return -ENOMEM;
14330 	}
14331 
14332 	i40e_dbg_init();
14333 	return pci_register_driver(&i40e_driver);
14334 }
14335 module_init(i40e_init_module);
14336 
14337 /**
14338  * i40e_exit_module - Driver exit cleanup routine
14339  *
14340  * i40e_exit_module is called just before the driver is removed
14341  * from memory.
14342  **/
14343 static void __exit i40e_exit_module(void)
14344 {
14345 	pci_unregister_driver(&i40e_driver);
14346 	destroy_workqueue(i40e_wq);
14347 	i40e_dbg_exit();
14348 }
14349 module_exit(i40e_exit_module);
14350