1 /*******************************************************************************
2  *
3  * Intel Ethernet Controller XL710 Family Linux Driver
4  * Copyright(c) 2013 - 2017 Intel Corporation.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along
16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * The full GNU General Public License is included in this distribution in
19  * the file called "COPYING".
20  *
21  * Contact Information:
22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
24  *
25  ******************************************************************************/
26 
27 #include <linux/etherdevice.h>
28 #include <linux/of_net.h>
29 #include <linux/pci.h>
30 #include <linux/bpf.h>
31 
32 /* Local includes */
33 #include "i40e.h"
34 #include "i40e_diag.h"
35 #include <net/udp_tunnel.h>
36 /* All i40e tracepoints are defined by the include below, which
37  * must be included exactly once across the whole kernel with
38  * CREATE_TRACE_POINTS defined
39  */
40 #define CREATE_TRACE_POINTS
41 #include "i40e_trace.h"
42 
43 const char i40e_driver_name[] = "i40e";
44 static const char i40e_driver_string[] =
45 			"Intel(R) Ethernet Connection XL710 Network Driver";
46 
47 #define DRV_KERN "-k"
48 
49 #define DRV_VERSION_MAJOR 2
50 #define DRV_VERSION_MINOR 1
51 #define DRV_VERSION_BUILD 14
52 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
53 	     __stringify(DRV_VERSION_MINOR) "." \
54 	     __stringify(DRV_VERSION_BUILD)    DRV_KERN
55 const char i40e_driver_version_str[] = DRV_VERSION;
56 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
57 
58 /* a bit of forward declarations */
59 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
60 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired);
61 static int i40e_add_vsi(struct i40e_vsi *vsi);
62 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
63 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
64 static int i40e_setup_misc_vector(struct i40e_pf *pf);
65 static void i40e_determine_queue_usage(struct i40e_pf *pf);
66 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
67 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired);
68 static int i40e_reset(struct i40e_pf *pf);
69 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired);
70 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
71 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
72 
73 /* i40e_pci_tbl - PCI Device ID Table
74  *
75  * Last entry must be all 0s
76  *
77  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
78  *   Class, Class Mask, private data (not used) }
79  */
80 static const struct pci_device_id i40e_pci_tbl[] = {
81 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
82 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
83 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
84 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
85 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
86 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
87 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
88 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
89 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
90 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
91 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
92 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
93 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
94 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
95 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
96 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
97 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
98 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_B), 0},
99 	{PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_SFP28), 0},
100 	/* required last entry */
101 	{0, }
102 };
103 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
104 
105 #define I40E_MAX_VF_COUNT 128
106 static int debug = -1;
107 module_param(debug, uint, 0);
108 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
109 
110 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
111 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
112 MODULE_LICENSE("GPL");
113 MODULE_VERSION(DRV_VERSION);
114 
115 static struct workqueue_struct *i40e_wq;
116 
117 /**
118  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
119  * @hw:   pointer to the HW structure
120  * @mem:  ptr to mem struct to fill out
121  * @size: size of memory requested
122  * @alignment: what to align the allocation to
123  **/
124 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
125 			    u64 size, u32 alignment)
126 {
127 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
128 
129 	mem->size = ALIGN(size, alignment);
130 	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
131 				      &mem->pa, GFP_KERNEL);
132 	if (!mem->va)
133 		return -ENOMEM;
134 
135 	return 0;
136 }
137 
138 /**
139  * i40e_free_dma_mem_d - OS specific memory free for shared code
140  * @hw:   pointer to the HW structure
141  * @mem:  ptr to mem struct to free
142  **/
143 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
144 {
145 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
146 
147 	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
148 	mem->va = NULL;
149 	mem->pa = 0;
150 	mem->size = 0;
151 
152 	return 0;
153 }
154 
155 /**
156  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
157  * @hw:   pointer to the HW structure
158  * @mem:  ptr to mem struct to fill out
159  * @size: size of memory requested
160  **/
161 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
162 			     u32 size)
163 {
164 	mem->size = size;
165 	mem->va = kzalloc(size, GFP_KERNEL);
166 
167 	if (!mem->va)
168 		return -ENOMEM;
169 
170 	return 0;
171 }
172 
173 /**
174  * i40e_free_virt_mem_d - OS specific memory free for shared code
175  * @hw:   pointer to the HW structure
176  * @mem:  ptr to mem struct to free
177  **/
178 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
179 {
180 	/* it's ok to kfree a NULL pointer */
181 	kfree(mem->va);
182 	mem->va = NULL;
183 	mem->size = 0;
184 
185 	return 0;
186 }
187 
188 /**
189  * i40e_get_lump - find a lump of free generic resource
190  * @pf: board private structure
191  * @pile: the pile of resource to search
192  * @needed: the number of items needed
193  * @id: an owner id to stick on the items assigned
194  *
195  * Returns the base item index of the lump, or negative for error
196  *
197  * The search_hint trick and lack of advanced fit-finding only work
198  * because we're highly likely to have all the same size lump requests.
199  * Linear search time and any fragmentation should be minimal.
200  **/
201 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
202 			 u16 needed, u16 id)
203 {
204 	int ret = -ENOMEM;
205 	int i, j;
206 
207 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
208 		dev_info(&pf->pdev->dev,
209 			 "param err: pile=%p needed=%d id=0x%04x\n",
210 			 pile, needed, id);
211 		return -EINVAL;
212 	}
213 
214 	/* start the linear search with an imperfect hint */
215 	i = pile->search_hint;
216 	while (i < pile->num_entries) {
217 		/* skip already allocated entries */
218 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
219 			i++;
220 			continue;
221 		}
222 
223 		/* do we have enough in this lump? */
224 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
225 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
226 				break;
227 		}
228 
229 		if (j == needed) {
230 			/* there was enough, so assign it to the requestor */
231 			for (j = 0; j < needed; j++)
232 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
233 			ret = i;
234 			pile->search_hint = i + j;
235 			break;
236 		}
237 
238 		/* not enough, so skip over it and continue looking */
239 		i += j;
240 	}
241 
242 	return ret;
243 }
244 
245 /**
246  * i40e_put_lump - return a lump of generic resource
247  * @pile: the pile of resource to search
248  * @index: the base item index
249  * @id: the owner id of the items assigned
250  *
251  * Returns the count of items in the lump
252  **/
253 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
254 {
255 	int valid_id = (id | I40E_PILE_VALID_BIT);
256 	int count = 0;
257 	int i;
258 
259 	if (!pile || index >= pile->num_entries)
260 		return -EINVAL;
261 
262 	for (i = index;
263 	     i < pile->num_entries && pile->list[i] == valid_id;
264 	     i++) {
265 		pile->list[i] = 0;
266 		count++;
267 	}
268 
269 	if (count && index < pile->search_hint)
270 		pile->search_hint = index;
271 
272 	return count;
273 }
274 
275 /**
276  * i40e_find_vsi_from_id - searches for the vsi with the given id
277  * @pf - the pf structure to search for the vsi
278  * @id - id of the vsi it is searching for
279  **/
280 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
281 {
282 	int i;
283 
284 	for (i = 0; i < pf->num_alloc_vsi; i++)
285 		if (pf->vsi[i] && (pf->vsi[i]->id == id))
286 			return pf->vsi[i];
287 
288 	return NULL;
289 }
290 
291 /**
292  * i40e_service_event_schedule - Schedule the service task to wake up
293  * @pf: board private structure
294  *
295  * If not already scheduled, this puts the task into the work queue
296  **/
297 void i40e_service_event_schedule(struct i40e_pf *pf)
298 {
299 	if (!test_bit(__I40E_DOWN, pf->state) &&
300 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
301 		queue_work(i40e_wq, &pf->service_task);
302 }
303 
304 /**
305  * i40e_tx_timeout - Respond to a Tx Hang
306  * @netdev: network interface device structure
307  *
308  * If any port has noticed a Tx timeout, it is likely that the whole
309  * device is munged, not just the one netdev port, so go for the full
310  * reset.
311  **/
312 static void i40e_tx_timeout(struct net_device *netdev)
313 {
314 	struct i40e_netdev_priv *np = netdev_priv(netdev);
315 	struct i40e_vsi *vsi = np->vsi;
316 	struct i40e_pf *pf = vsi->back;
317 	struct i40e_ring *tx_ring = NULL;
318 	unsigned int i, hung_queue = 0;
319 	u32 head, val;
320 
321 	pf->tx_timeout_count++;
322 
323 	/* find the stopped queue the same way the stack does */
324 	for (i = 0; i < netdev->num_tx_queues; i++) {
325 		struct netdev_queue *q;
326 		unsigned long trans_start;
327 
328 		q = netdev_get_tx_queue(netdev, i);
329 		trans_start = q->trans_start;
330 		if (netif_xmit_stopped(q) &&
331 		    time_after(jiffies,
332 			       (trans_start + netdev->watchdog_timeo))) {
333 			hung_queue = i;
334 			break;
335 		}
336 	}
337 
338 	if (i == netdev->num_tx_queues) {
339 		netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
340 	} else {
341 		/* now that we have an index, find the tx_ring struct */
342 		for (i = 0; i < vsi->num_queue_pairs; i++) {
343 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
344 				if (hung_queue ==
345 				    vsi->tx_rings[i]->queue_index) {
346 					tx_ring = vsi->tx_rings[i];
347 					break;
348 				}
349 			}
350 		}
351 	}
352 
353 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
354 		pf->tx_timeout_recovery_level = 1;  /* reset after some time */
355 	else if (time_before(jiffies,
356 		      (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
357 		return;   /* don't do any new action before the next timeout */
358 
359 	if (tx_ring) {
360 		head = i40e_get_head(tx_ring);
361 		/* Read interrupt register */
362 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
363 			val = rd32(&pf->hw,
364 			     I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
365 						tx_ring->vsi->base_vector - 1));
366 		else
367 			val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
368 
369 		netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
370 			    vsi->seid, hung_queue, tx_ring->next_to_clean,
371 			    head, tx_ring->next_to_use,
372 			    readl(tx_ring->tail), val);
373 	}
374 
375 	pf->tx_timeout_last_recovery = jiffies;
376 	netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
377 		    pf->tx_timeout_recovery_level, hung_queue);
378 
379 	switch (pf->tx_timeout_recovery_level) {
380 	case 1:
381 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
382 		break;
383 	case 2:
384 		set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
385 		break;
386 	case 3:
387 		set_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
388 		break;
389 	default:
390 		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
391 		break;
392 	}
393 
394 	i40e_service_event_schedule(pf);
395 	pf->tx_timeout_recovery_level++;
396 }
397 
398 /**
399  * i40e_get_vsi_stats_struct - Get System Network Statistics
400  * @vsi: the VSI we care about
401  *
402  * Returns the address of the device statistics structure.
403  * The statistics are actually updated from the service task.
404  **/
405 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
406 {
407 	return &vsi->net_stats;
408 }
409 
410 /**
411  * i40e_get_netdev_stats_struct_tx - populate stats from a Tx ring
412  * @ring: Tx ring to get statistics from
413  * @stats: statistics entry to be updated
414  **/
415 static void i40e_get_netdev_stats_struct_tx(struct i40e_ring *ring,
416 					    struct rtnl_link_stats64 *stats)
417 {
418 	u64 bytes, packets;
419 	unsigned int start;
420 
421 	do {
422 		start = u64_stats_fetch_begin_irq(&ring->syncp);
423 		packets = ring->stats.packets;
424 		bytes   = ring->stats.bytes;
425 	} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
426 
427 	stats->tx_packets += packets;
428 	stats->tx_bytes   += bytes;
429 }
430 
431 /**
432  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
433  * @netdev: network interface device structure
434  *
435  * Returns the address of the device statistics structure.
436  * The statistics are actually updated from the service task.
437  **/
438 static void i40e_get_netdev_stats_struct(struct net_device *netdev,
439 				  struct rtnl_link_stats64 *stats)
440 {
441 	struct i40e_netdev_priv *np = netdev_priv(netdev);
442 	struct i40e_ring *tx_ring, *rx_ring;
443 	struct i40e_vsi *vsi = np->vsi;
444 	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
445 	int i;
446 
447 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
448 		return;
449 
450 	if (!vsi->tx_rings)
451 		return;
452 
453 	rcu_read_lock();
454 	for (i = 0; i < vsi->num_queue_pairs; i++) {
455 		u64 bytes, packets;
456 		unsigned int start;
457 
458 		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
459 		if (!tx_ring)
460 			continue;
461 		i40e_get_netdev_stats_struct_tx(tx_ring, stats);
462 
463 		rx_ring = &tx_ring[1];
464 
465 		do {
466 			start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
467 			packets = rx_ring->stats.packets;
468 			bytes   = rx_ring->stats.bytes;
469 		} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
470 
471 		stats->rx_packets += packets;
472 		stats->rx_bytes   += bytes;
473 
474 		if (i40e_enabled_xdp_vsi(vsi))
475 			i40e_get_netdev_stats_struct_tx(&rx_ring[1], stats);
476 	}
477 	rcu_read_unlock();
478 
479 	/* following stats updated by i40e_watchdog_subtask() */
480 	stats->multicast	= vsi_stats->multicast;
481 	stats->tx_errors	= vsi_stats->tx_errors;
482 	stats->tx_dropped	= vsi_stats->tx_dropped;
483 	stats->rx_errors	= vsi_stats->rx_errors;
484 	stats->rx_dropped	= vsi_stats->rx_dropped;
485 	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
486 	stats->rx_length_errors	= vsi_stats->rx_length_errors;
487 }
488 
489 /**
490  * i40e_vsi_reset_stats - Resets all stats of the given vsi
491  * @vsi: the VSI to have its stats reset
492  **/
493 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
494 {
495 	struct rtnl_link_stats64 *ns;
496 	int i;
497 
498 	if (!vsi)
499 		return;
500 
501 	ns = i40e_get_vsi_stats_struct(vsi);
502 	memset(ns, 0, sizeof(*ns));
503 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
504 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
505 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
506 	if (vsi->rx_rings && vsi->rx_rings[0]) {
507 		for (i = 0; i < vsi->num_queue_pairs; i++) {
508 			memset(&vsi->rx_rings[i]->stats, 0,
509 			       sizeof(vsi->rx_rings[i]->stats));
510 			memset(&vsi->rx_rings[i]->rx_stats, 0,
511 			       sizeof(vsi->rx_rings[i]->rx_stats));
512 			memset(&vsi->tx_rings[i]->stats, 0,
513 			       sizeof(vsi->tx_rings[i]->stats));
514 			memset(&vsi->tx_rings[i]->tx_stats, 0,
515 			       sizeof(vsi->tx_rings[i]->tx_stats));
516 		}
517 	}
518 	vsi->stat_offsets_loaded = false;
519 }
520 
521 /**
522  * i40e_pf_reset_stats - Reset all of the stats for the given PF
523  * @pf: the PF to be reset
524  **/
525 void i40e_pf_reset_stats(struct i40e_pf *pf)
526 {
527 	int i;
528 
529 	memset(&pf->stats, 0, sizeof(pf->stats));
530 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
531 	pf->stat_offsets_loaded = false;
532 
533 	for (i = 0; i < I40E_MAX_VEB; i++) {
534 		if (pf->veb[i]) {
535 			memset(&pf->veb[i]->stats, 0,
536 			       sizeof(pf->veb[i]->stats));
537 			memset(&pf->veb[i]->stats_offsets, 0,
538 			       sizeof(pf->veb[i]->stats_offsets));
539 			pf->veb[i]->stat_offsets_loaded = false;
540 		}
541 	}
542 	pf->hw_csum_rx_error = 0;
543 }
544 
545 /**
546  * i40e_stat_update48 - read and update a 48 bit stat from the chip
547  * @hw: ptr to the hardware info
548  * @hireg: the high 32 bit reg to read
549  * @loreg: the low 32 bit reg to read
550  * @offset_loaded: has the initial offset been loaded yet
551  * @offset: ptr to current offset value
552  * @stat: ptr to the stat
553  *
554  * Since the device stats are not reset at PFReset, they likely will not
555  * be zeroed when the driver starts.  We'll save the first values read
556  * and use them as offsets to be subtracted from the raw values in order
557  * to report stats that count from zero.  In the process, we also manage
558  * the potential roll-over.
559  **/
560 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
561 			       bool offset_loaded, u64 *offset, u64 *stat)
562 {
563 	u64 new_data;
564 
565 	if (hw->device_id == I40E_DEV_ID_QEMU) {
566 		new_data = rd32(hw, loreg);
567 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
568 	} else {
569 		new_data = rd64(hw, loreg);
570 	}
571 	if (!offset_loaded)
572 		*offset = new_data;
573 	if (likely(new_data >= *offset))
574 		*stat = new_data - *offset;
575 	else
576 		*stat = (new_data + BIT_ULL(48)) - *offset;
577 	*stat &= 0xFFFFFFFFFFFFULL;
578 }
579 
580 /**
581  * i40e_stat_update32 - read and update a 32 bit stat from the chip
582  * @hw: ptr to the hardware info
583  * @reg: the hw reg to read
584  * @offset_loaded: has the initial offset been loaded yet
585  * @offset: ptr to current offset value
586  * @stat: ptr to the stat
587  **/
588 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
589 			       bool offset_loaded, u64 *offset, u64 *stat)
590 {
591 	u32 new_data;
592 
593 	new_data = rd32(hw, reg);
594 	if (!offset_loaded)
595 		*offset = new_data;
596 	if (likely(new_data >= *offset))
597 		*stat = (u32)(new_data - *offset);
598 	else
599 		*stat = (u32)((new_data + BIT_ULL(32)) - *offset);
600 }
601 
602 /**
603  * i40e_stat_update_and_clear32 - read and clear hw reg, update a 32 bit stat
604  * @hw: ptr to the hardware info
605  * @reg: the hw reg to read and clear
606  * @stat: ptr to the stat
607  **/
608 static void i40e_stat_update_and_clear32(struct i40e_hw *hw, u32 reg, u64 *stat)
609 {
610 	u32 new_data = rd32(hw, reg);
611 
612 	wr32(hw, reg, 1); /* must write a nonzero value to clear register */
613 	*stat += new_data;
614 }
615 
616 /**
617  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
618  * @vsi: the VSI to be updated
619  **/
620 void i40e_update_eth_stats(struct i40e_vsi *vsi)
621 {
622 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
623 	struct i40e_pf *pf = vsi->back;
624 	struct i40e_hw *hw = &pf->hw;
625 	struct i40e_eth_stats *oes;
626 	struct i40e_eth_stats *es;     /* device's eth stats */
627 
628 	es = &vsi->eth_stats;
629 	oes = &vsi->eth_stats_offsets;
630 
631 	/* Gather up the stats that the hw collects */
632 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
633 			   vsi->stat_offsets_loaded,
634 			   &oes->tx_errors, &es->tx_errors);
635 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
636 			   vsi->stat_offsets_loaded,
637 			   &oes->rx_discards, &es->rx_discards);
638 	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
639 			   vsi->stat_offsets_loaded,
640 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
641 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
642 			   vsi->stat_offsets_loaded,
643 			   &oes->tx_errors, &es->tx_errors);
644 
645 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
646 			   I40E_GLV_GORCL(stat_idx),
647 			   vsi->stat_offsets_loaded,
648 			   &oes->rx_bytes, &es->rx_bytes);
649 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
650 			   I40E_GLV_UPRCL(stat_idx),
651 			   vsi->stat_offsets_loaded,
652 			   &oes->rx_unicast, &es->rx_unicast);
653 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
654 			   I40E_GLV_MPRCL(stat_idx),
655 			   vsi->stat_offsets_loaded,
656 			   &oes->rx_multicast, &es->rx_multicast);
657 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
658 			   I40E_GLV_BPRCL(stat_idx),
659 			   vsi->stat_offsets_loaded,
660 			   &oes->rx_broadcast, &es->rx_broadcast);
661 
662 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
663 			   I40E_GLV_GOTCL(stat_idx),
664 			   vsi->stat_offsets_loaded,
665 			   &oes->tx_bytes, &es->tx_bytes);
666 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
667 			   I40E_GLV_UPTCL(stat_idx),
668 			   vsi->stat_offsets_loaded,
669 			   &oes->tx_unicast, &es->tx_unicast);
670 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
671 			   I40E_GLV_MPTCL(stat_idx),
672 			   vsi->stat_offsets_loaded,
673 			   &oes->tx_multicast, &es->tx_multicast);
674 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
675 			   I40E_GLV_BPTCL(stat_idx),
676 			   vsi->stat_offsets_loaded,
677 			   &oes->tx_broadcast, &es->tx_broadcast);
678 	vsi->stat_offsets_loaded = true;
679 }
680 
681 /**
682  * i40e_update_veb_stats - Update Switch component statistics
683  * @veb: the VEB being updated
684  **/
685 static void i40e_update_veb_stats(struct i40e_veb *veb)
686 {
687 	struct i40e_pf *pf = veb->pf;
688 	struct i40e_hw *hw = &pf->hw;
689 	struct i40e_eth_stats *oes;
690 	struct i40e_eth_stats *es;     /* device's eth stats */
691 	struct i40e_veb_tc_stats *veb_oes;
692 	struct i40e_veb_tc_stats *veb_es;
693 	int i, idx = 0;
694 
695 	idx = veb->stats_idx;
696 	es = &veb->stats;
697 	oes = &veb->stats_offsets;
698 	veb_es = &veb->tc_stats;
699 	veb_oes = &veb->tc_stats_offsets;
700 
701 	/* Gather up the stats that the hw collects */
702 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
703 			   veb->stat_offsets_loaded,
704 			   &oes->tx_discards, &es->tx_discards);
705 	if (hw->revision_id > 0)
706 		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
707 				   veb->stat_offsets_loaded,
708 				   &oes->rx_unknown_protocol,
709 				   &es->rx_unknown_protocol);
710 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
711 			   veb->stat_offsets_loaded,
712 			   &oes->rx_bytes, &es->rx_bytes);
713 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
714 			   veb->stat_offsets_loaded,
715 			   &oes->rx_unicast, &es->rx_unicast);
716 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
717 			   veb->stat_offsets_loaded,
718 			   &oes->rx_multicast, &es->rx_multicast);
719 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
720 			   veb->stat_offsets_loaded,
721 			   &oes->rx_broadcast, &es->rx_broadcast);
722 
723 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
724 			   veb->stat_offsets_loaded,
725 			   &oes->tx_bytes, &es->tx_bytes);
726 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
727 			   veb->stat_offsets_loaded,
728 			   &oes->tx_unicast, &es->tx_unicast);
729 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
730 			   veb->stat_offsets_loaded,
731 			   &oes->tx_multicast, &es->tx_multicast);
732 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
733 			   veb->stat_offsets_loaded,
734 			   &oes->tx_broadcast, &es->tx_broadcast);
735 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
736 		i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
737 				   I40E_GLVEBTC_RPCL(i, idx),
738 				   veb->stat_offsets_loaded,
739 				   &veb_oes->tc_rx_packets[i],
740 				   &veb_es->tc_rx_packets[i]);
741 		i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
742 				   I40E_GLVEBTC_RBCL(i, idx),
743 				   veb->stat_offsets_loaded,
744 				   &veb_oes->tc_rx_bytes[i],
745 				   &veb_es->tc_rx_bytes[i]);
746 		i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
747 				   I40E_GLVEBTC_TPCL(i, idx),
748 				   veb->stat_offsets_loaded,
749 				   &veb_oes->tc_tx_packets[i],
750 				   &veb_es->tc_tx_packets[i]);
751 		i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
752 				   I40E_GLVEBTC_TBCL(i, idx),
753 				   veb->stat_offsets_loaded,
754 				   &veb_oes->tc_tx_bytes[i],
755 				   &veb_es->tc_tx_bytes[i]);
756 	}
757 	veb->stat_offsets_loaded = true;
758 }
759 
760 /**
761  * i40e_update_vsi_stats - Update the vsi statistics counters.
762  * @vsi: the VSI to be updated
763  *
764  * There are a few instances where we store the same stat in a
765  * couple of different structs.  This is partly because we have
766  * the netdev stats that need to be filled out, which is slightly
767  * different from the "eth_stats" defined by the chip and used in
768  * VF communications.  We sort it out here.
769  **/
770 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
771 {
772 	struct i40e_pf *pf = vsi->back;
773 	struct rtnl_link_stats64 *ons;
774 	struct rtnl_link_stats64 *ns;   /* netdev stats */
775 	struct i40e_eth_stats *oes;
776 	struct i40e_eth_stats *es;     /* device's eth stats */
777 	u32 tx_restart, tx_busy;
778 	struct i40e_ring *p;
779 	u32 rx_page, rx_buf;
780 	u64 bytes, packets;
781 	unsigned int start;
782 	u64 tx_linearize;
783 	u64 tx_force_wb;
784 	u64 rx_p, rx_b;
785 	u64 tx_p, tx_b;
786 	u16 q;
787 
788 	if (test_bit(__I40E_VSI_DOWN, vsi->state) ||
789 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
790 		return;
791 
792 	ns = i40e_get_vsi_stats_struct(vsi);
793 	ons = &vsi->net_stats_offsets;
794 	es = &vsi->eth_stats;
795 	oes = &vsi->eth_stats_offsets;
796 
797 	/* Gather up the netdev and vsi stats that the driver collects
798 	 * on the fly during packet processing
799 	 */
800 	rx_b = rx_p = 0;
801 	tx_b = tx_p = 0;
802 	tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
803 	rx_page = 0;
804 	rx_buf = 0;
805 	rcu_read_lock();
806 	for (q = 0; q < vsi->num_queue_pairs; q++) {
807 		/* locate Tx ring */
808 		p = ACCESS_ONCE(vsi->tx_rings[q]);
809 
810 		do {
811 			start = u64_stats_fetch_begin_irq(&p->syncp);
812 			packets = p->stats.packets;
813 			bytes = p->stats.bytes;
814 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
815 		tx_b += bytes;
816 		tx_p += packets;
817 		tx_restart += p->tx_stats.restart_queue;
818 		tx_busy += p->tx_stats.tx_busy;
819 		tx_linearize += p->tx_stats.tx_linearize;
820 		tx_force_wb += p->tx_stats.tx_force_wb;
821 
822 		/* Rx queue is part of the same block as Tx queue */
823 		p = &p[1];
824 		do {
825 			start = u64_stats_fetch_begin_irq(&p->syncp);
826 			packets = p->stats.packets;
827 			bytes = p->stats.bytes;
828 		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
829 		rx_b += bytes;
830 		rx_p += packets;
831 		rx_buf += p->rx_stats.alloc_buff_failed;
832 		rx_page += p->rx_stats.alloc_page_failed;
833 	}
834 	rcu_read_unlock();
835 	vsi->tx_restart = tx_restart;
836 	vsi->tx_busy = tx_busy;
837 	vsi->tx_linearize = tx_linearize;
838 	vsi->tx_force_wb = tx_force_wb;
839 	vsi->rx_page_failed = rx_page;
840 	vsi->rx_buf_failed = rx_buf;
841 
842 	ns->rx_packets = rx_p;
843 	ns->rx_bytes = rx_b;
844 	ns->tx_packets = tx_p;
845 	ns->tx_bytes = tx_b;
846 
847 	/* update netdev stats from eth stats */
848 	i40e_update_eth_stats(vsi);
849 	ons->tx_errors = oes->tx_errors;
850 	ns->tx_errors = es->tx_errors;
851 	ons->multicast = oes->rx_multicast;
852 	ns->multicast = es->rx_multicast;
853 	ons->rx_dropped = oes->rx_discards;
854 	ns->rx_dropped = es->rx_discards;
855 	ons->tx_dropped = oes->tx_discards;
856 	ns->tx_dropped = es->tx_discards;
857 
858 	/* pull in a couple PF stats if this is the main vsi */
859 	if (vsi == pf->vsi[pf->lan_vsi]) {
860 		ns->rx_crc_errors = pf->stats.crc_errors;
861 		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
862 		ns->rx_length_errors = pf->stats.rx_length_errors;
863 	}
864 }
865 
866 /**
867  * i40e_update_pf_stats - Update the PF statistics counters.
868  * @pf: the PF to be updated
869  **/
870 static void i40e_update_pf_stats(struct i40e_pf *pf)
871 {
872 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
873 	struct i40e_hw_port_stats *nsd = &pf->stats;
874 	struct i40e_hw *hw = &pf->hw;
875 	u32 val;
876 	int i;
877 
878 	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
879 			   I40E_GLPRT_GORCL(hw->port),
880 			   pf->stat_offsets_loaded,
881 			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
882 	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
883 			   I40E_GLPRT_GOTCL(hw->port),
884 			   pf->stat_offsets_loaded,
885 			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
886 	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
887 			   pf->stat_offsets_loaded,
888 			   &osd->eth.rx_discards,
889 			   &nsd->eth.rx_discards);
890 	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
891 			   I40E_GLPRT_UPRCL(hw->port),
892 			   pf->stat_offsets_loaded,
893 			   &osd->eth.rx_unicast,
894 			   &nsd->eth.rx_unicast);
895 	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
896 			   I40E_GLPRT_MPRCL(hw->port),
897 			   pf->stat_offsets_loaded,
898 			   &osd->eth.rx_multicast,
899 			   &nsd->eth.rx_multicast);
900 	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
901 			   I40E_GLPRT_BPRCL(hw->port),
902 			   pf->stat_offsets_loaded,
903 			   &osd->eth.rx_broadcast,
904 			   &nsd->eth.rx_broadcast);
905 	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
906 			   I40E_GLPRT_UPTCL(hw->port),
907 			   pf->stat_offsets_loaded,
908 			   &osd->eth.tx_unicast,
909 			   &nsd->eth.tx_unicast);
910 	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
911 			   I40E_GLPRT_MPTCL(hw->port),
912 			   pf->stat_offsets_loaded,
913 			   &osd->eth.tx_multicast,
914 			   &nsd->eth.tx_multicast);
915 	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
916 			   I40E_GLPRT_BPTCL(hw->port),
917 			   pf->stat_offsets_loaded,
918 			   &osd->eth.tx_broadcast,
919 			   &nsd->eth.tx_broadcast);
920 
921 	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
922 			   pf->stat_offsets_loaded,
923 			   &osd->tx_dropped_link_down,
924 			   &nsd->tx_dropped_link_down);
925 
926 	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
927 			   pf->stat_offsets_loaded,
928 			   &osd->crc_errors, &nsd->crc_errors);
929 
930 	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
931 			   pf->stat_offsets_loaded,
932 			   &osd->illegal_bytes, &nsd->illegal_bytes);
933 
934 	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
935 			   pf->stat_offsets_loaded,
936 			   &osd->mac_local_faults,
937 			   &nsd->mac_local_faults);
938 	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
939 			   pf->stat_offsets_loaded,
940 			   &osd->mac_remote_faults,
941 			   &nsd->mac_remote_faults);
942 
943 	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
944 			   pf->stat_offsets_loaded,
945 			   &osd->rx_length_errors,
946 			   &nsd->rx_length_errors);
947 
948 	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
949 			   pf->stat_offsets_loaded,
950 			   &osd->link_xon_rx, &nsd->link_xon_rx);
951 	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
952 			   pf->stat_offsets_loaded,
953 			   &osd->link_xon_tx, &nsd->link_xon_tx);
954 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
955 			   pf->stat_offsets_loaded,
956 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
957 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
958 			   pf->stat_offsets_loaded,
959 			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
960 
961 	for (i = 0; i < 8; i++) {
962 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
963 				   pf->stat_offsets_loaded,
964 				   &osd->priority_xoff_rx[i],
965 				   &nsd->priority_xoff_rx[i]);
966 		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
967 				   pf->stat_offsets_loaded,
968 				   &osd->priority_xon_rx[i],
969 				   &nsd->priority_xon_rx[i]);
970 		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
971 				   pf->stat_offsets_loaded,
972 				   &osd->priority_xon_tx[i],
973 				   &nsd->priority_xon_tx[i]);
974 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
975 				   pf->stat_offsets_loaded,
976 				   &osd->priority_xoff_tx[i],
977 				   &nsd->priority_xoff_tx[i]);
978 		i40e_stat_update32(hw,
979 				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
980 				   pf->stat_offsets_loaded,
981 				   &osd->priority_xon_2_xoff[i],
982 				   &nsd->priority_xon_2_xoff[i]);
983 	}
984 
985 	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
986 			   I40E_GLPRT_PRC64L(hw->port),
987 			   pf->stat_offsets_loaded,
988 			   &osd->rx_size_64, &nsd->rx_size_64);
989 	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
990 			   I40E_GLPRT_PRC127L(hw->port),
991 			   pf->stat_offsets_loaded,
992 			   &osd->rx_size_127, &nsd->rx_size_127);
993 	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
994 			   I40E_GLPRT_PRC255L(hw->port),
995 			   pf->stat_offsets_loaded,
996 			   &osd->rx_size_255, &nsd->rx_size_255);
997 	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
998 			   I40E_GLPRT_PRC511L(hw->port),
999 			   pf->stat_offsets_loaded,
1000 			   &osd->rx_size_511, &nsd->rx_size_511);
1001 	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1002 			   I40E_GLPRT_PRC1023L(hw->port),
1003 			   pf->stat_offsets_loaded,
1004 			   &osd->rx_size_1023, &nsd->rx_size_1023);
1005 	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1006 			   I40E_GLPRT_PRC1522L(hw->port),
1007 			   pf->stat_offsets_loaded,
1008 			   &osd->rx_size_1522, &nsd->rx_size_1522);
1009 	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1010 			   I40E_GLPRT_PRC9522L(hw->port),
1011 			   pf->stat_offsets_loaded,
1012 			   &osd->rx_size_big, &nsd->rx_size_big);
1013 
1014 	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1015 			   I40E_GLPRT_PTC64L(hw->port),
1016 			   pf->stat_offsets_loaded,
1017 			   &osd->tx_size_64, &nsd->tx_size_64);
1018 	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1019 			   I40E_GLPRT_PTC127L(hw->port),
1020 			   pf->stat_offsets_loaded,
1021 			   &osd->tx_size_127, &nsd->tx_size_127);
1022 	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1023 			   I40E_GLPRT_PTC255L(hw->port),
1024 			   pf->stat_offsets_loaded,
1025 			   &osd->tx_size_255, &nsd->tx_size_255);
1026 	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1027 			   I40E_GLPRT_PTC511L(hw->port),
1028 			   pf->stat_offsets_loaded,
1029 			   &osd->tx_size_511, &nsd->tx_size_511);
1030 	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1031 			   I40E_GLPRT_PTC1023L(hw->port),
1032 			   pf->stat_offsets_loaded,
1033 			   &osd->tx_size_1023, &nsd->tx_size_1023);
1034 	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1035 			   I40E_GLPRT_PTC1522L(hw->port),
1036 			   pf->stat_offsets_loaded,
1037 			   &osd->tx_size_1522, &nsd->tx_size_1522);
1038 	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1039 			   I40E_GLPRT_PTC9522L(hw->port),
1040 			   pf->stat_offsets_loaded,
1041 			   &osd->tx_size_big, &nsd->tx_size_big);
1042 
1043 	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1044 			   pf->stat_offsets_loaded,
1045 			   &osd->rx_undersize, &nsd->rx_undersize);
1046 	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1047 			   pf->stat_offsets_loaded,
1048 			   &osd->rx_fragments, &nsd->rx_fragments);
1049 	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1050 			   pf->stat_offsets_loaded,
1051 			   &osd->rx_oversize, &nsd->rx_oversize);
1052 	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1053 			   pf->stat_offsets_loaded,
1054 			   &osd->rx_jabber, &nsd->rx_jabber);
1055 
1056 	/* FDIR stats */
1057 	i40e_stat_update_and_clear32(hw,
1058 			I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(hw->pf_id)),
1059 			&nsd->fd_atr_match);
1060 	i40e_stat_update_and_clear32(hw,
1061 			I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(hw->pf_id)),
1062 			&nsd->fd_sb_match);
1063 	i40e_stat_update_and_clear32(hw,
1064 			I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(hw->pf_id)),
1065 			&nsd->fd_atr_tunnel_match);
1066 
1067 	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1068 	nsd->tx_lpi_status =
1069 		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1070 			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1071 	nsd->rx_lpi_status =
1072 		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1073 			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1074 	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1075 			   pf->stat_offsets_loaded,
1076 			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1077 	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1078 			   pf->stat_offsets_loaded,
1079 			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1080 
1081 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1082 	    !(pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED))
1083 		nsd->fd_sb_status = true;
1084 	else
1085 		nsd->fd_sb_status = false;
1086 
1087 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1088 	    !(pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED))
1089 		nsd->fd_atr_status = true;
1090 	else
1091 		nsd->fd_atr_status = false;
1092 
1093 	pf->stat_offsets_loaded = true;
1094 }
1095 
1096 /**
1097  * i40e_update_stats - Update the various statistics counters.
1098  * @vsi: the VSI to be updated
1099  *
1100  * Update the various stats for this VSI and its related entities.
1101  **/
1102 void i40e_update_stats(struct i40e_vsi *vsi)
1103 {
1104 	struct i40e_pf *pf = vsi->back;
1105 
1106 	if (vsi == pf->vsi[pf->lan_vsi])
1107 		i40e_update_pf_stats(pf);
1108 
1109 	i40e_update_vsi_stats(vsi);
1110 }
1111 
1112 /**
1113  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1114  * @vsi: the VSI to be searched
1115  * @macaddr: the MAC address
1116  * @vlan: the vlan
1117  *
1118  * Returns ptr to the filter object or NULL
1119  **/
1120 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1121 						const u8 *macaddr, s16 vlan)
1122 {
1123 	struct i40e_mac_filter *f;
1124 	u64 key;
1125 
1126 	if (!vsi || !macaddr)
1127 		return NULL;
1128 
1129 	key = i40e_addr_to_hkey(macaddr);
1130 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1131 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1132 		    (vlan == f->vlan))
1133 			return f;
1134 	}
1135 	return NULL;
1136 }
1137 
1138 /**
1139  * i40e_find_mac - Find a mac addr in the macvlan filters list
1140  * @vsi: the VSI to be searched
1141  * @macaddr: the MAC address we are searching for
1142  *
1143  * Returns the first filter with the provided MAC address or NULL if
1144  * MAC address was not found
1145  **/
1146 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1147 {
1148 	struct i40e_mac_filter *f;
1149 	u64 key;
1150 
1151 	if (!vsi || !macaddr)
1152 		return NULL;
1153 
1154 	key = i40e_addr_to_hkey(macaddr);
1155 	hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1156 		if ((ether_addr_equal(macaddr, f->macaddr)))
1157 			return f;
1158 	}
1159 	return NULL;
1160 }
1161 
1162 /**
1163  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1164  * @vsi: the VSI to be searched
1165  *
1166  * Returns true if VSI is in vlan mode or false otherwise
1167  **/
1168 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1169 {
1170 	/* If we have a PVID, always operate in VLAN mode */
1171 	if (vsi->info.pvid)
1172 		return true;
1173 
1174 	/* We need to operate in VLAN mode whenever we have any filters with
1175 	 * a VLAN other than I40E_VLAN_ALL. We could check the table each
1176 	 * time, incurring search cost repeatedly. However, we can notice two
1177 	 * things:
1178 	 *
1179 	 * 1) the only place where we can gain a VLAN filter is in
1180 	 *    i40e_add_filter.
1181 	 *
1182 	 * 2) the only place where filters are actually removed is in
1183 	 *    i40e_sync_filters_subtask.
1184 	 *
1185 	 * Thus, we can simply use a boolean value, has_vlan_filters which we
1186 	 * will set to true when we add a VLAN filter in i40e_add_filter. Then
1187 	 * we have to perform the full search after deleting filters in
1188 	 * i40e_sync_filters_subtask, but we already have to search
1189 	 * filters here and can perform the check at the same time. This
1190 	 * results in avoiding embedding a loop for VLAN mode inside another
1191 	 * loop over all the filters, and should maintain correctness as noted
1192 	 * above.
1193 	 */
1194 	return vsi->has_vlan_filter;
1195 }
1196 
1197 /**
1198  * i40e_correct_mac_vlan_filters - Correct non-VLAN filters if necessary
1199  * @vsi: the VSI to configure
1200  * @tmp_add_list: list of filters ready to be added
1201  * @tmp_del_list: list of filters ready to be deleted
1202  * @vlan_filters: the number of active VLAN filters
1203  *
1204  * Update VLAN=0 and VLAN=-1 (I40E_VLAN_ANY) filters properly so that they
1205  * behave as expected. If we have any active VLAN filters remaining or about
1206  * to be added then we need to update non-VLAN filters to be marked as VLAN=0
1207  * so that they only match against untagged traffic. If we no longer have any
1208  * active VLAN filters, we need to make all non-VLAN filters marked as VLAN=-1
1209  * so that they match against both tagged and untagged traffic. In this way,
1210  * we ensure that we correctly receive the desired traffic. This ensures that
1211  * when we have an active VLAN we will receive only untagged traffic and
1212  * traffic matching active VLANs. If we have no active VLANs then we will
1213  * operate in non-VLAN mode and receive all traffic, tagged or untagged.
1214  *
1215  * Finally, in a similar fashion, this function also corrects filters when
1216  * there is an active PVID assigned to this VSI.
1217  *
1218  * In case of memory allocation failure return -ENOMEM. Otherwise, return 0.
1219  *
1220  * This function is only expected to be called from within
1221  * i40e_sync_vsi_filters.
1222  *
1223  * NOTE: This function expects to be called while under the
1224  * mac_filter_hash_lock
1225  */
1226 static int i40e_correct_mac_vlan_filters(struct i40e_vsi *vsi,
1227 					 struct hlist_head *tmp_add_list,
1228 					 struct hlist_head *tmp_del_list,
1229 					 int vlan_filters)
1230 {
1231 	s16 pvid = le16_to_cpu(vsi->info.pvid);
1232 	struct i40e_mac_filter *f, *add_head;
1233 	struct i40e_new_mac_filter *new;
1234 	struct hlist_node *h;
1235 	int bkt, new_vlan;
1236 
1237 	/* To determine if a particular filter needs to be replaced we
1238 	 * have the three following conditions:
1239 	 *
1240 	 * a) if we have a PVID assigned, then all filters which are
1241 	 *    not marked as VLAN=PVID must be replaced with filters that
1242 	 *    are.
1243 	 * b) otherwise, if we have any active VLANS, all filters
1244 	 *    which are marked as VLAN=-1 must be replaced with
1245 	 *    filters marked as VLAN=0
1246 	 * c) finally, if we do not have any active VLANS, all filters
1247 	 *    which are marked as VLAN=0 must be replaced with filters
1248 	 *    marked as VLAN=-1
1249 	 */
1250 
1251 	/* Update the filters about to be added in place */
1252 	hlist_for_each_entry(new, tmp_add_list, hlist) {
1253 		if (pvid && new->f->vlan != pvid)
1254 			new->f->vlan = pvid;
1255 		else if (vlan_filters && new->f->vlan == I40E_VLAN_ANY)
1256 			new->f->vlan = 0;
1257 		else if (!vlan_filters && new->f->vlan == 0)
1258 			new->f->vlan = I40E_VLAN_ANY;
1259 	}
1260 
1261 	/* Update the remaining active filters */
1262 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1263 		/* Combine the checks for whether a filter needs to be changed
1264 		 * and then determine the new VLAN inside the if block, in
1265 		 * order to avoid duplicating code for adding the new filter
1266 		 * then deleting the old filter.
1267 		 */
1268 		if ((pvid && f->vlan != pvid) ||
1269 		    (vlan_filters && f->vlan == I40E_VLAN_ANY) ||
1270 		    (!vlan_filters && f->vlan == 0)) {
1271 			/* Determine the new vlan we will be adding */
1272 			if (pvid)
1273 				new_vlan = pvid;
1274 			else if (vlan_filters)
1275 				new_vlan = 0;
1276 			else
1277 				new_vlan = I40E_VLAN_ANY;
1278 
1279 			/* Create the new filter */
1280 			add_head = i40e_add_filter(vsi, f->macaddr, new_vlan);
1281 			if (!add_head)
1282 				return -ENOMEM;
1283 
1284 			/* Create a temporary i40e_new_mac_filter */
1285 			new = kzalloc(sizeof(*new), GFP_ATOMIC);
1286 			if (!new)
1287 				return -ENOMEM;
1288 
1289 			new->f = add_head;
1290 			new->state = add_head->state;
1291 
1292 			/* Add the new filter to the tmp list */
1293 			hlist_add_head(&new->hlist, tmp_add_list);
1294 
1295 			/* Put the original filter into the delete list */
1296 			f->state = I40E_FILTER_REMOVE;
1297 			hash_del(&f->hlist);
1298 			hlist_add_head(&f->hlist, tmp_del_list);
1299 		}
1300 	}
1301 
1302 	vsi->has_vlan_filter = !!vlan_filters;
1303 
1304 	return 0;
1305 }
1306 
1307 /**
1308  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1309  * @vsi: the PF Main VSI - inappropriate for any other VSI
1310  * @macaddr: the MAC address
1311  *
1312  * Remove whatever filter the firmware set up so the driver can manage
1313  * its own filtering intelligently.
1314  **/
1315 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1316 {
1317 	struct i40e_aqc_remove_macvlan_element_data element;
1318 	struct i40e_pf *pf = vsi->back;
1319 
1320 	/* Only appropriate for the PF main VSI */
1321 	if (vsi->type != I40E_VSI_MAIN)
1322 		return;
1323 
1324 	memset(&element, 0, sizeof(element));
1325 	ether_addr_copy(element.mac_addr, macaddr);
1326 	element.vlan_tag = 0;
1327 	/* Ignore error returns, some firmware does it this way... */
1328 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1329 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1330 
1331 	memset(&element, 0, sizeof(element));
1332 	ether_addr_copy(element.mac_addr, macaddr);
1333 	element.vlan_tag = 0;
1334 	/* ...and some firmware does it this way. */
1335 	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1336 			I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1337 	i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1338 }
1339 
1340 /**
1341  * i40e_add_filter - Add a mac/vlan filter to the VSI
1342  * @vsi: the VSI to be searched
1343  * @macaddr: the MAC address
1344  * @vlan: the vlan
1345  *
1346  * Returns ptr to the filter object or NULL when no memory available.
1347  *
1348  * NOTE: This function is expected to be called with mac_filter_hash_lock
1349  * being held.
1350  **/
1351 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1352 					const u8 *macaddr, s16 vlan)
1353 {
1354 	struct i40e_mac_filter *f;
1355 	u64 key;
1356 
1357 	if (!vsi || !macaddr)
1358 		return NULL;
1359 
1360 	f = i40e_find_filter(vsi, macaddr, vlan);
1361 	if (!f) {
1362 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1363 		if (!f)
1364 			return NULL;
1365 
1366 		/* Update the boolean indicating if we need to function in
1367 		 * VLAN mode.
1368 		 */
1369 		if (vlan >= 0)
1370 			vsi->has_vlan_filter = true;
1371 
1372 		ether_addr_copy(f->macaddr, macaddr);
1373 		f->vlan = vlan;
1374 		/* If we're in overflow promisc mode, set the state directly
1375 		 * to failed, so we don't bother to try sending the filter
1376 		 * to the hardware.
1377 		 */
1378 		if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state))
1379 			f->state = I40E_FILTER_FAILED;
1380 		else
1381 			f->state = I40E_FILTER_NEW;
1382 		INIT_HLIST_NODE(&f->hlist);
1383 
1384 		key = i40e_addr_to_hkey(macaddr);
1385 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1386 
1387 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1388 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1389 	}
1390 
1391 	/* If we're asked to add a filter that has been marked for removal, it
1392 	 * is safe to simply restore it to active state. __i40e_del_filter
1393 	 * will have simply deleted any filters which were previously marked
1394 	 * NEW or FAILED, so if it is currently marked REMOVE it must have
1395 	 * previously been ACTIVE. Since we haven't yet run the sync filters
1396 	 * task, just restore this filter to the ACTIVE state so that the
1397 	 * sync task leaves it in place
1398 	 */
1399 	if (f->state == I40E_FILTER_REMOVE)
1400 		f->state = I40E_FILTER_ACTIVE;
1401 
1402 	return f;
1403 }
1404 
1405 /**
1406  * __i40e_del_filter - Remove a specific filter from the VSI
1407  * @vsi: VSI to remove from
1408  * @f: the filter to remove from the list
1409  *
1410  * This function should be called instead of i40e_del_filter only if you know
1411  * the exact filter you will remove already, such as via i40e_find_filter or
1412  * i40e_find_mac.
1413  *
1414  * NOTE: This function is expected to be called with mac_filter_hash_lock
1415  * being held.
1416  * ANOTHER NOTE: This function MUST be called from within the context of
1417  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1418  * instead of list_for_each_entry().
1419  **/
1420 void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1421 {
1422 	if (!f)
1423 		return;
1424 
1425 	/* If the filter was never added to firmware then we can just delete it
1426 	 * directly and we don't want to set the status to remove or else an
1427 	 * admin queue command will unnecessarily fire.
1428 	 */
1429 	if ((f->state == I40E_FILTER_FAILED) ||
1430 	    (f->state == I40E_FILTER_NEW)) {
1431 		hash_del(&f->hlist);
1432 		kfree(f);
1433 	} else {
1434 		f->state = I40E_FILTER_REMOVE;
1435 	}
1436 
1437 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1438 	vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1439 }
1440 
1441 /**
1442  * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1443  * @vsi: the VSI to be searched
1444  * @macaddr: the MAC address
1445  * @vlan: the VLAN
1446  *
1447  * NOTE: This function is expected to be called with mac_filter_hash_lock
1448  * being held.
1449  * ANOTHER NOTE: This function MUST be called from within the context of
1450  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1451  * instead of list_for_each_entry().
1452  **/
1453 void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1454 {
1455 	struct i40e_mac_filter *f;
1456 
1457 	if (!vsi || !macaddr)
1458 		return;
1459 
1460 	f = i40e_find_filter(vsi, macaddr, vlan);
1461 	__i40e_del_filter(vsi, f);
1462 }
1463 
1464 /**
1465  * i40e_add_mac_filter - Add a MAC filter for all active VLANs
1466  * @vsi: the VSI to be searched
1467  * @macaddr: the mac address to be filtered
1468  *
1469  * If we're not in VLAN mode, just add the filter to I40E_VLAN_ANY. Otherwise,
1470  * go through all the macvlan filters and add a macvlan filter for each
1471  * unique vlan that already exists. If a PVID has been assigned, instead only
1472  * add the macaddr to that VLAN.
1473  *
1474  * Returns last filter added on success, else NULL
1475  **/
1476 struct i40e_mac_filter *i40e_add_mac_filter(struct i40e_vsi *vsi,
1477 					    const u8 *macaddr)
1478 {
1479 	struct i40e_mac_filter *f, *add = NULL;
1480 	struct hlist_node *h;
1481 	int bkt;
1482 
1483 	if (vsi->info.pvid)
1484 		return i40e_add_filter(vsi, macaddr,
1485 				       le16_to_cpu(vsi->info.pvid));
1486 
1487 	if (!i40e_is_vsi_in_vlan(vsi))
1488 		return i40e_add_filter(vsi, macaddr, I40E_VLAN_ANY);
1489 
1490 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1491 		if (f->state == I40E_FILTER_REMOVE)
1492 			continue;
1493 		add = i40e_add_filter(vsi, macaddr, f->vlan);
1494 		if (!add)
1495 			return NULL;
1496 	}
1497 
1498 	return add;
1499 }
1500 
1501 /**
1502  * i40e_del_mac_filter - Remove a MAC filter from all VLANs
1503  * @vsi: the VSI to be searched
1504  * @macaddr: the mac address to be removed
1505  *
1506  * Removes a given MAC address from a VSI regardless of what VLAN it has been
1507  * associated with.
1508  *
1509  * Returns 0 for success, or error
1510  **/
1511 int i40e_del_mac_filter(struct i40e_vsi *vsi, const u8 *macaddr)
1512 {
1513 	struct i40e_mac_filter *f;
1514 	struct hlist_node *h;
1515 	bool found = false;
1516 	int bkt;
1517 
1518 	WARN(!spin_is_locked(&vsi->mac_filter_hash_lock),
1519 	     "Missing mac_filter_hash_lock\n");
1520 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1521 		if (ether_addr_equal(macaddr, f->macaddr)) {
1522 			__i40e_del_filter(vsi, f);
1523 			found = true;
1524 		}
1525 	}
1526 
1527 	if (found)
1528 		return 0;
1529 	else
1530 		return -ENOENT;
1531 }
1532 
1533 /**
1534  * i40e_set_mac - NDO callback to set mac address
1535  * @netdev: network interface device structure
1536  * @p: pointer to an address structure
1537  *
1538  * Returns 0 on success, negative on failure
1539  **/
1540 static int i40e_set_mac(struct net_device *netdev, void *p)
1541 {
1542 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1543 	struct i40e_vsi *vsi = np->vsi;
1544 	struct i40e_pf *pf = vsi->back;
1545 	struct i40e_hw *hw = &pf->hw;
1546 	struct sockaddr *addr = p;
1547 
1548 	if (!is_valid_ether_addr(addr->sa_data))
1549 		return -EADDRNOTAVAIL;
1550 
1551 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1552 		netdev_info(netdev, "already using mac address %pM\n",
1553 			    addr->sa_data);
1554 		return 0;
1555 	}
1556 
1557 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
1558 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
1559 		return -EADDRNOTAVAIL;
1560 
1561 	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1562 		netdev_info(netdev, "returning to hw mac address %pM\n",
1563 			    hw->mac.addr);
1564 	else
1565 		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1566 
1567 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1568 	i40e_del_mac_filter(vsi, netdev->dev_addr);
1569 	i40e_add_mac_filter(vsi, addr->sa_data);
1570 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1571 	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1572 	if (vsi->type == I40E_VSI_MAIN) {
1573 		i40e_status ret;
1574 
1575 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1576 						I40E_AQC_WRITE_TYPE_LAA_WOL,
1577 						addr->sa_data, NULL);
1578 		if (ret)
1579 			netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1580 				    i40e_stat_str(hw, ret),
1581 				    i40e_aq_str(hw, hw->aq.asq_last_status));
1582 	}
1583 
1584 	/* schedule our worker thread which will take care of
1585 	 * applying the new filter changes
1586 	 */
1587 	i40e_service_event_schedule(vsi->back);
1588 	return 0;
1589 }
1590 
1591 /**
1592  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1593  * @vsi: the VSI being setup
1594  * @ctxt: VSI context structure
1595  * @enabled_tc: Enabled TCs bitmap
1596  * @is_add: True if called before Add VSI
1597  *
1598  * Setup VSI queue mapping for enabled traffic classes.
1599  **/
1600 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1601 				     struct i40e_vsi_context *ctxt,
1602 				     u8 enabled_tc,
1603 				     bool is_add)
1604 {
1605 	struct i40e_pf *pf = vsi->back;
1606 	u16 sections = 0;
1607 	u8 netdev_tc = 0;
1608 	u16 numtc = 0;
1609 	u16 qcount;
1610 	u8 offset;
1611 	u16 qmap;
1612 	int i;
1613 	u16 num_tc_qps = 0;
1614 
1615 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1616 	offset = 0;
1617 
1618 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1619 		/* Find numtc from enabled TC bitmap */
1620 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1621 			if (enabled_tc & BIT(i)) /* TC is enabled */
1622 				numtc++;
1623 		}
1624 		if (!numtc) {
1625 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1626 			numtc = 1;
1627 		}
1628 	} else {
1629 		/* At least TC0 is enabled in case of non-DCB case */
1630 		numtc = 1;
1631 	}
1632 
1633 	vsi->tc_config.numtc = numtc;
1634 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1635 	/* Number of queues per enabled TC */
1636 	qcount = vsi->alloc_queue_pairs;
1637 
1638 	num_tc_qps = qcount / numtc;
1639 	num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1640 
1641 	/* Setup queue offset/count for all TCs for given VSI */
1642 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1643 		/* See if the given TC is enabled for the given VSI */
1644 		if (vsi->tc_config.enabled_tc & BIT(i)) {
1645 			/* TC is enabled */
1646 			int pow, num_qps;
1647 
1648 			switch (vsi->type) {
1649 			case I40E_VSI_MAIN:
1650 				qcount = min_t(int, pf->alloc_rss_size,
1651 					       num_tc_qps);
1652 				break;
1653 			case I40E_VSI_FDIR:
1654 			case I40E_VSI_SRIOV:
1655 			case I40E_VSI_VMDQ2:
1656 			default:
1657 				qcount = num_tc_qps;
1658 				WARN_ON(i != 0);
1659 				break;
1660 			}
1661 			vsi->tc_config.tc_info[i].qoffset = offset;
1662 			vsi->tc_config.tc_info[i].qcount = qcount;
1663 
1664 			/* find the next higher power-of-2 of num queue pairs */
1665 			num_qps = qcount;
1666 			pow = 0;
1667 			while (num_qps && (BIT_ULL(pow) < qcount)) {
1668 				pow++;
1669 				num_qps >>= 1;
1670 			}
1671 
1672 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1673 			qmap =
1674 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1675 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1676 
1677 			offset += qcount;
1678 		} else {
1679 			/* TC is not enabled so set the offset to
1680 			 * default queue and allocate one queue
1681 			 * for the given TC.
1682 			 */
1683 			vsi->tc_config.tc_info[i].qoffset = 0;
1684 			vsi->tc_config.tc_info[i].qcount = 1;
1685 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1686 
1687 			qmap = 0;
1688 		}
1689 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1690 	}
1691 
1692 	/* Set actual Tx/Rx queue pairs */
1693 	vsi->num_queue_pairs = offset;
1694 	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1695 		if (vsi->req_queue_pairs > 0)
1696 			vsi->num_queue_pairs = vsi->req_queue_pairs;
1697 		else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1698 			vsi->num_queue_pairs = pf->num_lan_msix;
1699 	}
1700 
1701 	/* Scheduler section valid can only be set for ADD VSI */
1702 	if (is_add) {
1703 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1704 
1705 		ctxt->info.up_enable_bits = enabled_tc;
1706 	}
1707 	if (vsi->type == I40E_VSI_SRIOV) {
1708 		ctxt->info.mapping_flags |=
1709 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1710 		for (i = 0; i < vsi->num_queue_pairs; i++)
1711 			ctxt->info.queue_mapping[i] =
1712 					       cpu_to_le16(vsi->base_queue + i);
1713 	} else {
1714 		ctxt->info.mapping_flags |=
1715 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1716 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1717 	}
1718 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1719 }
1720 
1721 /**
1722  * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1723  * @netdev: the netdevice
1724  * @addr: address to add
1725  *
1726  * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1727  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1728  */
1729 static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1730 {
1731 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1732 	struct i40e_vsi *vsi = np->vsi;
1733 
1734 	if (i40e_add_mac_filter(vsi, addr))
1735 		return 0;
1736 	else
1737 		return -ENOMEM;
1738 }
1739 
1740 /**
1741  * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1742  * @netdev: the netdevice
1743  * @addr: address to add
1744  *
1745  * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1746  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1747  */
1748 static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1749 {
1750 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1751 	struct i40e_vsi *vsi = np->vsi;
1752 
1753 	i40e_del_mac_filter(vsi, addr);
1754 
1755 	return 0;
1756 }
1757 
1758 /**
1759  * i40e_set_rx_mode - NDO callback to set the netdev filters
1760  * @netdev: network interface device structure
1761  **/
1762 static void i40e_set_rx_mode(struct net_device *netdev)
1763 {
1764 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1765 	struct i40e_vsi *vsi = np->vsi;
1766 
1767 	spin_lock_bh(&vsi->mac_filter_hash_lock);
1768 
1769 	__dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1770 	__dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1771 
1772 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
1773 
1774 	/* check for other flag changes */
1775 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1776 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1777 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1778 	}
1779 
1780 	/* schedule our worker thread which will take care of
1781 	 * applying the new filter changes
1782 	 */
1783 	i40e_service_event_schedule(vsi->back);
1784 }
1785 
1786 /**
1787  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1788  * @vsi: Pointer to VSI struct
1789  * @from: Pointer to list which contains MAC filter entries - changes to
1790  *        those entries needs to be undone.
1791  *
1792  * MAC filter entries from this list were slated for deletion.
1793  **/
1794 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1795 					 struct hlist_head *from)
1796 {
1797 	struct i40e_mac_filter *f;
1798 	struct hlist_node *h;
1799 
1800 	hlist_for_each_entry_safe(f, h, from, hlist) {
1801 		u64 key = i40e_addr_to_hkey(f->macaddr);
1802 
1803 		/* Move the element back into MAC filter list*/
1804 		hlist_del(&f->hlist);
1805 		hash_add(vsi->mac_filter_hash, &f->hlist, key);
1806 	}
1807 }
1808 
1809 /**
1810  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1811  * @vsi: Pointer to vsi struct
1812  * @from: Pointer to list which contains MAC filter entries - changes to
1813  *        those entries needs to be undone.
1814  *
1815  * MAC filter entries from this list were slated for addition.
1816  **/
1817 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi,
1818 					 struct hlist_head *from)
1819 {
1820 	struct i40e_new_mac_filter *new;
1821 	struct hlist_node *h;
1822 
1823 	hlist_for_each_entry_safe(new, h, from, hlist) {
1824 		/* We can simply free the wrapper structure */
1825 		hlist_del(&new->hlist);
1826 		kfree(new);
1827 	}
1828 }
1829 
1830 /**
1831  * i40e_next_entry - Get the next non-broadcast filter from a list
1832  * @next: pointer to filter in list
1833  *
1834  * Returns the next non-broadcast filter in the list. Required so that we
1835  * ignore broadcast filters within the list, since these are not handled via
1836  * the normal firmware update path.
1837  */
1838 static
1839 struct i40e_new_mac_filter *i40e_next_filter(struct i40e_new_mac_filter *next)
1840 {
1841 	hlist_for_each_entry_continue(next, hlist) {
1842 		if (!is_broadcast_ether_addr(next->f->macaddr))
1843 			return next;
1844 	}
1845 
1846 	return NULL;
1847 }
1848 
1849 /**
1850  * i40e_update_filter_state - Update filter state based on return data
1851  * from firmware
1852  * @count: Number of filters added
1853  * @add_list: return data from fw
1854  * @head: pointer to first filter in current batch
1855  *
1856  * MAC filter entries from list were slated to be added to device. Returns
1857  * number of successful filters. Note that 0 does NOT mean success!
1858  **/
1859 static int
1860 i40e_update_filter_state(int count,
1861 			 struct i40e_aqc_add_macvlan_element_data *add_list,
1862 			 struct i40e_new_mac_filter *add_head)
1863 {
1864 	int retval = 0;
1865 	int i;
1866 
1867 	for (i = 0; i < count; i++) {
1868 		/* Always check status of each filter. We don't need to check
1869 		 * the firmware return status because we pre-set the filter
1870 		 * status to I40E_AQC_MM_ERR_NO_RES when sending the filter
1871 		 * request to the adminq. Thus, if it no longer matches then
1872 		 * we know the filter is active.
1873 		 */
1874 		if (add_list[i].match_method == I40E_AQC_MM_ERR_NO_RES) {
1875 			add_head->state = I40E_FILTER_FAILED;
1876 		} else {
1877 			add_head->state = I40E_FILTER_ACTIVE;
1878 			retval++;
1879 		}
1880 
1881 		add_head = i40e_next_filter(add_head);
1882 		if (!add_head)
1883 			break;
1884 	}
1885 
1886 	return retval;
1887 }
1888 
1889 /**
1890  * i40e_aqc_del_filters - Request firmware to delete a set of filters
1891  * @vsi: ptr to the VSI
1892  * @vsi_name: name to display in messages
1893  * @list: the list of filters to send to firmware
1894  * @num_del: the number of filters to delete
1895  * @retval: Set to -EIO on failure to delete
1896  *
1897  * Send a request to firmware via AdminQ to delete a set of filters. Uses
1898  * *retval instead of a return value so that success does not force ret_val to
1899  * be set to 0. This ensures that a sequence of calls to this function
1900  * preserve the previous value of *retval on successful delete.
1901  */
1902 static
1903 void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
1904 			  struct i40e_aqc_remove_macvlan_element_data *list,
1905 			  int num_del, int *retval)
1906 {
1907 	struct i40e_hw *hw = &vsi->back->hw;
1908 	i40e_status aq_ret;
1909 	int aq_err;
1910 
1911 	aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
1912 	aq_err = hw->aq.asq_last_status;
1913 
1914 	/* Explicitly ignore and do not report when firmware returns ENOENT */
1915 	if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1916 		*retval = -EIO;
1917 		dev_info(&vsi->back->pdev->dev,
1918 			 "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1919 			 vsi_name, i40e_stat_str(hw, aq_ret),
1920 			 i40e_aq_str(hw, aq_err));
1921 	}
1922 }
1923 
1924 /**
1925  * i40e_aqc_add_filters - Request firmware to add a set of filters
1926  * @vsi: ptr to the VSI
1927  * @vsi_name: name to display in messages
1928  * @list: the list of filters to send to firmware
1929  * @add_head: Position in the add hlist
1930  * @num_add: the number of filters to add
1931  * @promisc_change: set to true on exit if promiscuous mode was forced on
1932  *
1933  * Send a request to firmware via AdminQ to add a chunk of filters. Will set
1934  * promisc_changed to true if the firmware has run out of space for more
1935  * filters.
1936  */
1937 static
1938 void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
1939 			  struct i40e_aqc_add_macvlan_element_data *list,
1940 			  struct i40e_new_mac_filter *add_head,
1941 			  int num_add, bool *promisc_changed)
1942 {
1943 	struct i40e_hw *hw = &vsi->back->hw;
1944 	int aq_err, fcnt;
1945 
1946 	i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
1947 	aq_err = hw->aq.asq_last_status;
1948 	fcnt = i40e_update_filter_state(num_add, list, add_head);
1949 
1950 	if (fcnt != num_add) {
1951 		*promisc_changed = true;
1952 		set_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
1953 		dev_warn(&vsi->back->pdev->dev,
1954 			 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
1955 			 i40e_aq_str(hw, aq_err),
1956 			 vsi_name);
1957 	}
1958 }
1959 
1960 /**
1961  * i40e_aqc_broadcast_filter - Set promiscuous broadcast flags
1962  * @vsi: pointer to the VSI
1963  * @f: filter data
1964  *
1965  * This function sets or clears the promiscuous broadcast flags for VLAN
1966  * filters in order to properly receive broadcast frames. Assumes that only
1967  * broadcast filters are passed.
1968  *
1969  * Returns status indicating success or failure;
1970  **/
1971 static i40e_status
1972 i40e_aqc_broadcast_filter(struct i40e_vsi *vsi, const char *vsi_name,
1973 			  struct i40e_mac_filter *f)
1974 {
1975 	bool enable = f->state == I40E_FILTER_NEW;
1976 	struct i40e_hw *hw = &vsi->back->hw;
1977 	i40e_status aq_ret;
1978 
1979 	if (f->vlan == I40E_VLAN_ANY) {
1980 		aq_ret = i40e_aq_set_vsi_broadcast(hw,
1981 						   vsi->seid,
1982 						   enable,
1983 						   NULL);
1984 	} else {
1985 		aq_ret = i40e_aq_set_vsi_bc_promisc_on_vlan(hw,
1986 							    vsi->seid,
1987 							    enable,
1988 							    f->vlan,
1989 							    NULL);
1990 	}
1991 
1992 	if (aq_ret)
1993 		dev_warn(&vsi->back->pdev->dev,
1994 			 "Error %s setting broadcast promiscuous mode on %s\n",
1995 			 i40e_aq_str(hw, hw->aq.asq_last_status),
1996 			 vsi_name);
1997 
1998 	return aq_ret;
1999 }
2000 
2001 /**
2002  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
2003  * @vsi: ptr to the VSI
2004  *
2005  * Push any outstanding VSI filter changes through the AdminQ.
2006  *
2007  * Returns 0 or error value
2008  **/
2009 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
2010 {
2011 	struct hlist_head tmp_add_list, tmp_del_list;
2012 	struct i40e_mac_filter *f;
2013 	struct i40e_new_mac_filter *new, *add_head = NULL;
2014 	struct i40e_hw *hw = &vsi->back->hw;
2015 	unsigned int failed_filters = 0;
2016 	unsigned int vlan_filters = 0;
2017 	bool promisc_changed = false;
2018 	char vsi_name[16] = "PF";
2019 	int filter_list_len = 0;
2020 	i40e_status aq_ret = 0;
2021 	u32 changed_flags = 0;
2022 	struct hlist_node *h;
2023 	struct i40e_pf *pf;
2024 	int num_add = 0;
2025 	int num_del = 0;
2026 	int retval = 0;
2027 	u16 cmd_flags;
2028 	int list_size;
2029 	int bkt;
2030 
2031 	/* empty array typed pointers, kcalloc later */
2032 	struct i40e_aqc_add_macvlan_element_data *add_list;
2033 	struct i40e_aqc_remove_macvlan_element_data *del_list;
2034 
2035 	while (test_and_set_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state))
2036 		usleep_range(1000, 2000);
2037 	pf = vsi->back;
2038 
2039 	if (vsi->netdev) {
2040 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
2041 		vsi->current_netdev_flags = vsi->netdev->flags;
2042 	}
2043 
2044 	INIT_HLIST_HEAD(&tmp_add_list);
2045 	INIT_HLIST_HEAD(&tmp_del_list);
2046 
2047 	if (vsi->type == I40E_VSI_SRIOV)
2048 		snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
2049 	else if (vsi->type != I40E_VSI_MAIN)
2050 		snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
2051 
2052 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
2053 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
2054 
2055 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2056 		/* Create a list of filters to delete. */
2057 		hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2058 			if (f->state == I40E_FILTER_REMOVE) {
2059 				/* Move the element into temporary del_list */
2060 				hash_del(&f->hlist);
2061 				hlist_add_head(&f->hlist, &tmp_del_list);
2062 
2063 				/* Avoid counting removed filters */
2064 				continue;
2065 			}
2066 			if (f->state == I40E_FILTER_NEW) {
2067 				/* Create a temporary i40e_new_mac_filter */
2068 				new = kzalloc(sizeof(*new), GFP_ATOMIC);
2069 				if (!new)
2070 					goto err_no_memory_locked;
2071 
2072 				/* Store pointer to the real filter */
2073 				new->f = f;
2074 				new->state = f->state;
2075 
2076 				/* Add it to the hash list */
2077 				hlist_add_head(&new->hlist, &tmp_add_list);
2078 			}
2079 
2080 			/* Count the number of active (current and new) VLAN
2081 			 * filters we have now. Does not count filters which
2082 			 * are marked for deletion.
2083 			 */
2084 			if (f->vlan > 0)
2085 				vlan_filters++;
2086 		}
2087 
2088 		retval = i40e_correct_mac_vlan_filters(vsi,
2089 						       &tmp_add_list,
2090 						       &tmp_del_list,
2091 						       vlan_filters);
2092 		if (retval)
2093 			goto err_no_memory_locked;
2094 
2095 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2096 	}
2097 
2098 	/* Now process 'del_list' outside the lock */
2099 	if (!hlist_empty(&tmp_del_list)) {
2100 		filter_list_len = hw->aq.asq_buf_size /
2101 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2102 		list_size = filter_list_len *
2103 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
2104 		del_list = kzalloc(list_size, GFP_ATOMIC);
2105 		if (!del_list)
2106 			goto err_no_memory;
2107 
2108 		hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2109 			cmd_flags = 0;
2110 
2111 			/* handle broadcast filters by updating the broadcast
2112 			 * promiscuous flag and release filter list.
2113 			 */
2114 			if (is_broadcast_ether_addr(f->macaddr)) {
2115 				i40e_aqc_broadcast_filter(vsi, vsi_name, f);
2116 
2117 				hlist_del(&f->hlist);
2118 				kfree(f);
2119 				continue;
2120 			}
2121 
2122 			/* add to delete list */
2123 			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2124 			if (f->vlan == I40E_VLAN_ANY) {
2125 				del_list[num_del].vlan_tag = 0;
2126 				cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2127 			} else {
2128 				del_list[num_del].vlan_tag =
2129 					cpu_to_le16((u16)(f->vlan));
2130 			}
2131 
2132 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2133 			del_list[num_del].flags = cmd_flags;
2134 			num_del++;
2135 
2136 			/* flush a full buffer */
2137 			if (num_del == filter_list_len) {
2138 				i40e_aqc_del_filters(vsi, vsi_name, del_list,
2139 						     num_del, &retval);
2140 				memset(del_list, 0, list_size);
2141 				num_del = 0;
2142 			}
2143 			/* Release memory for MAC filter entries which were
2144 			 * synced up with HW.
2145 			 */
2146 			hlist_del(&f->hlist);
2147 			kfree(f);
2148 		}
2149 
2150 		if (num_del) {
2151 			i40e_aqc_del_filters(vsi, vsi_name, del_list,
2152 					     num_del, &retval);
2153 		}
2154 
2155 		kfree(del_list);
2156 		del_list = NULL;
2157 	}
2158 
2159 	if (!hlist_empty(&tmp_add_list)) {
2160 		/* Do all the adds now. */
2161 		filter_list_len = hw->aq.asq_buf_size /
2162 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2163 		list_size = filter_list_len *
2164 			       sizeof(struct i40e_aqc_add_macvlan_element_data);
2165 		add_list = kzalloc(list_size, GFP_ATOMIC);
2166 		if (!add_list)
2167 			goto err_no_memory;
2168 
2169 		num_add = 0;
2170 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2171 			if (test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2172 				     vsi->state)) {
2173 				new->state = I40E_FILTER_FAILED;
2174 				continue;
2175 			}
2176 
2177 			/* handle broadcast filters by updating the broadcast
2178 			 * promiscuous flag instead of adding a MAC filter.
2179 			 */
2180 			if (is_broadcast_ether_addr(new->f->macaddr)) {
2181 				if (i40e_aqc_broadcast_filter(vsi, vsi_name,
2182 							      new->f))
2183 					new->state = I40E_FILTER_FAILED;
2184 				else
2185 					new->state = I40E_FILTER_ACTIVE;
2186 				continue;
2187 			}
2188 
2189 			/* add to add array */
2190 			if (num_add == 0)
2191 				add_head = new;
2192 			cmd_flags = 0;
2193 			ether_addr_copy(add_list[num_add].mac_addr,
2194 					new->f->macaddr);
2195 			if (new->f->vlan == I40E_VLAN_ANY) {
2196 				add_list[num_add].vlan_tag = 0;
2197 				cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2198 			} else {
2199 				add_list[num_add].vlan_tag =
2200 					cpu_to_le16((u16)(new->f->vlan));
2201 			}
2202 			add_list[num_add].queue_number = 0;
2203 			/* set invalid match method for later detection */
2204 			add_list[num_add].match_method = I40E_AQC_MM_ERR_NO_RES;
2205 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2206 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
2207 			num_add++;
2208 
2209 			/* flush a full buffer */
2210 			if (num_add == filter_list_len) {
2211 				i40e_aqc_add_filters(vsi, vsi_name, add_list,
2212 						     add_head, num_add,
2213 						     &promisc_changed);
2214 				memset(add_list, 0, list_size);
2215 				num_add = 0;
2216 			}
2217 		}
2218 		if (num_add) {
2219 			i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2220 					     num_add, &promisc_changed);
2221 		}
2222 		/* Now move all of the filters from the temp add list back to
2223 		 * the VSI's list.
2224 		 */
2225 		spin_lock_bh(&vsi->mac_filter_hash_lock);
2226 		hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2227 			/* Only update the state if we're still NEW */
2228 			if (new->f->state == I40E_FILTER_NEW)
2229 				new->f->state = new->state;
2230 			hlist_del(&new->hlist);
2231 			kfree(new);
2232 		}
2233 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
2234 		kfree(add_list);
2235 		add_list = NULL;
2236 	}
2237 
2238 	/* Determine the number of active and failed filters. */
2239 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2240 	vsi->active_filters = 0;
2241 	hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2242 		if (f->state == I40E_FILTER_ACTIVE)
2243 			vsi->active_filters++;
2244 		else if (f->state == I40E_FILTER_FAILED)
2245 			failed_filters++;
2246 	}
2247 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2248 
2249 	/* If promiscuous mode has changed, we need to calculate a new
2250 	 * threshold for when we are safe to exit
2251 	 */
2252 	if (promisc_changed)
2253 		vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
2254 
2255 	/* Check if we are able to exit overflow promiscuous mode. We can
2256 	 * safely exit if we didn't just enter, we no longer have any failed
2257 	 * filters, and we have reduced filters below the threshold value.
2258 	 */
2259 	if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state) &&
2260 	    !promisc_changed && !failed_filters &&
2261 	    (vsi->active_filters < vsi->promisc_threshold)) {
2262 		dev_info(&pf->pdev->dev,
2263 			 "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2264 			 vsi_name);
2265 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2266 		promisc_changed = true;
2267 		vsi->promisc_threshold = 0;
2268 	}
2269 
2270 	/* if the VF is not trusted do not do promisc */
2271 	if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2272 		clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2273 		goto out;
2274 	}
2275 
2276 	/* check for changes in promiscuous modes */
2277 	if (changed_flags & IFF_ALLMULTI) {
2278 		bool cur_multipromisc;
2279 
2280 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2281 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2282 							       vsi->seid,
2283 							       cur_multipromisc,
2284 							       NULL);
2285 		if (aq_ret) {
2286 			retval = i40e_aq_rc_to_posix(aq_ret,
2287 						     hw->aq.asq_last_status);
2288 			dev_info(&pf->pdev->dev,
2289 				 "set multi promisc failed on %s, err %s aq_err %s\n",
2290 				 vsi_name,
2291 				 i40e_stat_str(hw, aq_ret),
2292 				 i40e_aq_str(hw, hw->aq.asq_last_status));
2293 		}
2294 	}
2295 
2296 	if ((changed_flags & IFF_PROMISC) || promisc_changed) {
2297 		bool cur_promisc;
2298 
2299 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2300 			       test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2301 					vsi->state));
2302 		if ((vsi->type == I40E_VSI_MAIN) &&
2303 		    (pf->lan_veb != I40E_NO_VEB) &&
2304 		    !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2305 			/* set defport ON for Main VSI instead of true promisc
2306 			 * this way we will get all unicast/multicast and VLAN
2307 			 * promisc behavior but will not get VF or VMDq traffic
2308 			 * replicated on the Main VSI.
2309 			 */
2310 			if (pf->cur_promisc != cur_promisc) {
2311 				pf->cur_promisc = cur_promisc;
2312 				if (cur_promisc)
2313 					aq_ret =
2314 					      i40e_aq_set_default_vsi(hw,
2315 								      vsi->seid,
2316 								      NULL);
2317 				else
2318 					aq_ret =
2319 					    i40e_aq_clear_default_vsi(hw,
2320 								      vsi->seid,
2321 								      NULL);
2322 				if (aq_ret) {
2323 					retval = i40e_aq_rc_to_posix(aq_ret,
2324 							hw->aq.asq_last_status);
2325 					dev_info(&pf->pdev->dev,
2326 						 "Set default VSI failed on %s, err %s, aq_err %s\n",
2327 						 vsi_name,
2328 						 i40e_stat_str(hw, aq_ret),
2329 						 i40e_aq_str(hw,
2330 						     hw->aq.asq_last_status));
2331 				}
2332 			}
2333 		} else {
2334 			aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2335 							  hw,
2336 							  vsi->seid,
2337 							  cur_promisc, NULL,
2338 							  true);
2339 			if (aq_ret) {
2340 				retval =
2341 				i40e_aq_rc_to_posix(aq_ret,
2342 						    hw->aq.asq_last_status);
2343 				dev_info(&pf->pdev->dev,
2344 					 "set unicast promisc failed on %s, err %s, aq_err %s\n",
2345 					 vsi_name,
2346 					 i40e_stat_str(hw, aq_ret),
2347 					 i40e_aq_str(hw,
2348 						     hw->aq.asq_last_status));
2349 			}
2350 			aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2351 							  hw,
2352 							  vsi->seid,
2353 							  cur_promisc, NULL);
2354 			if (aq_ret) {
2355 				retval =
2356 				i40e_aq_rc_to_posix(aq_ret,
2357 						    hw->aq.asq_last_status);
2358 				dev_info(&pf->pdev->dev,
2359 					 "set multicast promisc failed on %s, err %s, aq_err %s\n",
2360 					 vsi_name,
2361 					 i40e_stat_str(hw, aq_ret),
2362 					 i40e_aq_str(hw,
2363 						     hw->aq.asq_last_status));
2364 			}
2365 		}
2366 		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2367 						   vsi->seid,
2368 						   cur_promisc, NULL);
2369 		if (aq_ret) {
2370 			retval = i40e_aq_rc_to_posix(aq_ret,
2371 						     pf->hw.aq.asq_last_status);
2372 			dev_info(&pf->pdev->dev,
2373 				 "set brdcast promisc failed, err %s, aq_err %s\n",
2374 					 i40e_stat_str(hw, aq_ret),
2375 					 i40e_aq_str(hw,
2376 						     hw->aq.asq_last_status));
2377 		}
2378 	}
2379 out:
2380 	/* if something went wrong then set the changed flag so we try again */
2381 	if (retval)
2382 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2383 
2384 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2385 	return retval;
2386 
2387 err_no_memory:
2388 	/* Restore elements on the temporary add and delete lists */
2389 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2390 err_no_memory_locked:
2391 	i40e_undo_del_filter_entries(vsi, &tmp_del_list);
2392 	i40e_undo_add_filter_entries(vsi, &tmp_add_list);
2393 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2394 
2395 	vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2396 	clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2397 	return -ENOMEM;
2398 }
2399 
2400 /**
2401  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2402  * @pf: board private structure
2403  **/
2404 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2405 {
2406 	int v;
2407 
2408 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2409 		return;
2410 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2411 
2412 	for (v = 0; v < pf->num_alloc_vsi; v++) {
2413 		if (pf->vsi[v] &&
2414 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2415 			int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2416 
2417 			if (ret) {
2418 				/* come back and try again later */
2419 				pf->flags |= I40E_FLAG_FILTER_SYNC;
2420 				break;
2421 			}
2422 		}
2423 	}
2424 }
2425 
2426 /**
2427  * i40e_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2428  * @vsi: the vsi
2429  **/
2430 static int i40e_max_xdp_frame_size(struct i40e_vsi *vsi)
2431 {
2432 	if (PAGE_SIZE >= 8192 || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
2433 		return I40E_RXBUFFER_2048;
2434 	else
2435 		return I40E_RXBUFFER_3072;
2436 }
2437 
2438 /**
2439  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2440  * @netdev: network interface device structure
2441  * @new_mtu: new value for maximum frame size
2442  *
2443  * Returns 0 on success, negative on failure
2444  **/
2445 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2446 {
2447 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2448 	struct i40e_vsi *vsi = np->vsi;
2449 	struct i40e_pf *pf = vsi->back;
2450 
2451 	if (i40e_enabled_xdp_vsi(vsi)) {
2452 		int frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2453 
2454 		if (frame_size > i40e_max_xdp_frame_size(vsi))
2455 			return -EINVAL;
2456 	}
2457 
2458 	netdev_info(netdev, "changing MTU from %d to %d\n",
2459 		    netdev->mtu, new_mtu);
2460 	netdev->mtu = new_mtu;
2461 	if (netif_running(netdev))
2462 		i40e_vsi_reinit_locked(vsi);
2463 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
2464 		      I40E_FLAG_CLIENT_L2_CHANGE);
2465 	return 0;
2466 }
2467 
2468 /**
2469  * i40e_ioctl - Access the hwtstamp interface
2470  * @netdev: network interface device structure
2471  * @ifr: interface request data
2472  * @cmd: ioctl command
2473  **/
2474 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2475 {
2476 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2477 	struct i40e_pf *pf = np->vsi->back;
2478 
2479 	switch (cmd) {
2480 	case SIOCGHWTSTAMP:
2481 		return i40e_ptp_get_ts_config(pf, ifr);
2482 	case SIOCSHWTSTAMP:
2483 		return i40e_ptp_set_ts_config(pf, ifr);
2484 	default:
2485 		return -EOPNOTSUPP;
2486 	}
2487 }
2488 
2489 /**
2490  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2491  * @vsi: the vsi being adjusted
2492  **/
2493 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2494 {
2495 	struct i40e_vsi_context ctxt;
2496 	i40e_status ret;
2497 
2498 	if ((vsi->info.valid_sections &
2499 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2500 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2501 		return;  /* already enabled */
2502 
2503 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2504 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2505 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2506 
2507 	ctxt.seid = vsi->seid;
2508 	ctxt.info = vsi->info;
2509 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2510 	if (ret) {
2511 		dev_info(&vsi->back->pdev->dev,
2512 			 "update vlan stripping failed, err %s aq_err %s\n",
2513 			 i40e_stat_str(&vsi->back->hw, ret),
2514 			 i40e_aq_str(&vsi->back->hw,
2515 				     vsi->back->hw.aq.asq_last_status));
2516 	}
2517 }
2518 
2519 /**
2520  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2521  * @vsi: the vsi being adjusted
2522  **/
2523 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2524 {
2525 	struct i40e_vsi_context ctxt;
2526 	i40e_status ret;
2527 
2528 	if ((vsi->info.valid_sections &
2529 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2530 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2531 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2532 		return;  /* already disabled */
2533 
2534 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2535 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2536 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2537 
2538 	ctxt.seid = vsi->seid;
2539 	ctxt.info = vsi->info;
2540 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2541 	if (ret) {
2542 		dev_info(&vsi->back->pdev->dev,
2543 			 "update vlan stripping failed, err %s aq_err %s\n",
2544 			 i40e_stat_str(&vsi->back->hw, ret),
2545 			 i40e_aq_str(&vsi->back->hw,
2546 				     vsi->back->hw.aq.asq_last_status));
2547 	}
2548 }
2549 
2550 /**
2551  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2552  * @netdev: network interface to be adjusted
2553  * @features: netdev features to test if VLAN offload is enabled or not
2554  **/
2555 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2556 {
2557 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2558 	struct i40e_vsi *vsi = np->vsi;
2559 
2560 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2561 		i40e_vlan_stripping_enable(vsi);
2562 	else
2563 		i40e_vlan_stripping_disable(vsi);
2564 }
2565 
2566 /**
2567  * i40e_add_vlan_all_mac - Add a MAC/VLAN filter for each existing MAC address
2568  * @vsi: the vsi being configured
2569  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2570  *
2571  * This is a helper function for adding a new MAC/VLAN filter with the
2572  * specified VLAN for each existing MAC address already in the hash table.
2573  * This function does *not* perform any accounting to update filters based on
2574  * VLAN mode.
2575  *
2576  * NOTE: this function expects to be called while under the
2577  * mac_filter_hash_lock
2578  **/
2579 int i40e_add_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2580 {
2581 	struct i40e_mac_filter *f, *add_f;
2582 	struct hlist_node *h;
2583 	int bkt;
2584 
2585 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2586 		if (f->state == I40E_FILTER_REMOVE)
2587 			continue;
2588 		add_f = i40e_add_filter(vsi, f->macaddr, vid);
2589 		if (!add_f) {
2590 			dev_info(&vsi->back->pdev->dev,
2591 				 "Could not add vlan filter %d for %pM\n",
2592 				 vid, f->macaddr);
2593 			return -ENOMEM;
2594 		}
2595 	}
2596 
2597 	return 0;
2598 }
2599 
2600 /**
2601  * i40e_vsi_add_vlan - Add VSI membership for given VLAN
2602  * @vsi: the VSI being configured
2603  * @vid: VLAN id to be added
2604  **/
2605 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, u16 vid)
2606 {
2607 	int err;
2608 
2609 	if (vsi->info.pvid)
2610 		return -EINVAL;
2611 
2612 	/* The network stack will attempt to add VID=0, with the intention to
2613 	 * receive priority tagged packets with a VLAN of 0. Our HW receives
2614 	 * these packets by default when configured to receive untagged
2615 	 * packets, so we don't need to add a filter for this case.
2616 	 * Additionally, HW interprets adding a VID=0 filter as meaning to
2617 	 * receive *only* tagged traffic and stops receiving untagged traffic.
2618 	 * Thus, we do not want to actually add a filter for VID=0
2619 	 */
2620 	if (!vid)
2621 		return 0;
2622 
2623 	/* Locked once because all functions invoked below iterates list*/
2624 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2625 	err = i40e_add_vlan_all_mac(vsi, vid);
2626 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2627 	if (err)
2628 		return err;
2629 
2630 	/* schedule our worker thread which will take care of
2631 	 * applying the new filter changes
2632 	 */
2633 	i40e_service_event_schedule(vsi->back);
2634 	return 0;
2635 }
2636 
2637 /**
2638  * i40e_rm_vlan_all_mac - Remove MAC/VLAN pair for all MAC with the given VLAN
2639  * @vsi: the vsi being configured
2640  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2641  *
2642  * This function should be used to remove all VLAN filters which match the
2643  * given VID. It does not schedule the service event and does not take the
2644  * mac_filter_hash_lock so it may be combined with other operations under
2645  * a single invocation of the mac_filter_hash_lock.
2646  *
2647  * NOTE: this function expects to be called while under the
2648  * mac_filter_hash_lock
2649  */
2650 void i40e_rm_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2651 {
2652 	struct i40e_mac_filter *f;
2653 	struct hlist_node *h;
2654 	int bkt;
2655 
2656 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2657 		if (f->vlan == vid)
2658 			__i40e_del_filter(vsi, f);
2659 	}
2660 }
2661 
2662 /**
2663  * i40e_vsi_kill_vlan - Remove VSI membership for given VLAN
2664  * @vsi: the VSI being configured
2665  * @vid: VLAN id to be removed
2666  **/
2667 void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, u16 vid)
2668 {
2669 	if (!vid || vsi->info.pvid)
2670 		return;
2671 
2672 	spin_lock_bh(&vsi->mac_filter_hash_lock);
2673 	i40e_rm_vlan_all_mac(vsi, vid);
2674 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
2675 
2676 	/* schedule our worker thread which will take care of
2677 	 * applying the new filter changes
2678 	 */
2679 	i40e_service_event_schedule(vsi->back);
2680 }
2681 
2682 /**
2683  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2684  * @netdev: network interface to be adjusted
2685  * @vid: vlan id to be added
2686  *
2687  * net_device_ops implementation for adding vlan ids
2688  **/
2689 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2690 				__always_unused __be16 proto, u16 vid)
2691 {
2692 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2693 	struct i40e_vsi *vsi = np->vsi;
2694 	int ret = 0;
2695 
2696 	if (vid >= VLAN_N_VID)
2697 		return -EINVAL;
2698 
2699 	ret = i40e_vsi_add_vlan(vsi, vid);
2700 	if (!ret)
2701 		set_bit(vid, vsi->active_vlans);
2702 
2703 	return ret;
2704 }
2705 
2706 /**
2707  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2708  * @netdev: network interface to be adjusted
2709  * @vid: vlan id to be removed
2710  *
2711  * net_device_ops implementation for removing vlan ids
2712  **/
2713 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2714 				 __always_unused __be16 proto, u16 vid)
2715 {
2716 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2717 	struct i40e_vsi *vsi = np->vsi;
2718 
2719 	/* return code is ignored as there is nothing a user
2720 	 * can do about failure to remove and a log message was
2721 	 * already printed from the other function
2722 	 */
2723 	i40e_vsi_kill_vlan(vsi, vid);
2724 
2725 	clear_bit(vid, vsi->active_vlans);
2726 
2727 	return 0;
2728 }
2729 
2730 /**
2731  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2732  * @vsi: the vsi being brought back up
2733  **/
2734 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2735 {
2736 	u16 vid;
2737 
2738 	if (!vsi->netdev)
2739 		return;
2740 
2741 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2742 
2743 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2744 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2745 				     vid);
2746 }
2747 
2748 /**
2749  * i40e_vsi_add_pvid - Add pvid for the VSI
2750  * @vsi: the vsi being adjusted
2751  * @vid: the vlan id to set as a PVID
2752  **/
2753 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2754 {
2755 	struct i40e_vsi_context ctxt;
2756 	i40e_status ret;
2757 
2758 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2759 	vsi->info.pvid = cpu_to_le16(vid);
2760 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2761 				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2762 				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2763 
2764 	ctxt.seid = vsi->seid;
2765 	ctxt.info = vsi->info;
2766 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2767 	if (ret) {
2768 		dev_info(&vsi->back->pdev->dev,
2769 			 "add pvid failed, err %s aq_err %s\n",
2770 			 i40e_stat_str(&vsi->back->hw, ret),
2771 			 i40e_aq_str(&vsi->back->hw,
2772 				     vsi->back->hw.aq.asq_last_status));
2773 		return -ENOENT;
2774 	}
2775 
2776 	return 0;
2777 }
2778 
2779 /**
2780  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2781  * @vsi: the vsi being adjusted
2782  *
2783  * Just use the vlan_rx_register() service to put it back to normal
2784  **/
2785 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2786 {
2787 	i40e_vlan_stripping_disable(vsi);
2788 
2789 	vsi->info.pvid = 0;
2790 }
2791 
2792 /**
2793  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2794  * @vsi: ptr to the VSI
2795  *
2796  * If this function returns with an error, then it's possible one or
2797  * more of the rings is populated (while the rest are not).  It is the
2798  * callers duty to clean those orphaned rings.
2799  *
2800  * Return 0 on success, negative on failure
2801  **/
2802 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2803 {
2804 	int i, err = 0;
2805 
2806 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2807 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2808 
2809 	if (!i40e_enabled_xdp_vsi(vsi))
2810 		return err;
2811 
2812 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2813 		err = i40e_setup_tx_descriptors(vsi->xdp_rings[i]);
2814 
2815 	return err;
2816 }
2817 
2818 /**
2819  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2820  * @vsi: ptr to the VSI
2821  *
2822  * Free VSI's transmit software resources
2823  **/
2824 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2825 {
2826 	int i;
2827 
2828 	if (vsi->tx_rings) {
2829 		for (i = 0; i < vsi->num_queue_pairs; i++)
2830 			if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2831 				i40e_free_tx_resources(vsi->tx_rings[i]);
2832 	}
2833 
2834 	if (vsi->xdp_rings) {
2835 		for (i = 0; i < vsi->num_queue_pairs; i++)
2836 			if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2837 				i40e_free_tx_resources(vsi->xdp_rings[i]);
2838 	}
2839 }
2840 
2841 /**
2842  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2843  * @vsi: ptr to the VSI
2844  *
2845  * If this function returns with an error, then it's possible one or
2846  * more of the rings is populated (while the rest are not).  It is the
2847  * callers duty to clean those orphaned rings.
2848  *
2849  * Return 0 on success, negative on failure
2850  **/
2851 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2852 {
2853 	int i, err = 0;
2854 
2855 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2856 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2857 	return err;
2858 }
2859 
2860 /**
2861  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2862  * @vsi: ptr to the VSI
2863  *
2864  * Free all receive software resources
2865  **/
2866 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2867 {
2868 	int i;
2869 
2870 	if (!vsi->rx_rings)
2871 		return;
2872 
2873 	for (i = 0; i < vsi->num_queue_pairs; i++)
2874 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2875 			i40e_free_rx_resources(vsi->rx_rings[i]);
2876 }
2877 
2878 /**
2879  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2880  * @ring: The Tx ring to configure
2881  *
2882  * This enables/disables XPS for a given Tx descriptor ring
2883  * based on the TCs enabled for the VSI that ring belongs to.
2884  **/
2885 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2886 {
2887 	struct i40e_vsi *vsi = ring->vsi;
2888 
2889 	if (!ring->q_vector || !ring->netdev)
2890 		return;
2891 
2892 	if ((vsi->tc_config.numtc <= 1) &&
2893 	    !test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state)) {
2894 		netif_set_xps_queue(ring->netdev,
2895 				    get_cpu_mask(ring->q_vector->v_idx),
2896 				    ring->queue_index);
2897 	}
2898 
2899 	/* schedule our worker thread which will take care of
2900 	 * applying the new filter changes
2901 	 */
2902 	i40e_service_event_schedule(vsi->back);
2903 }
2904 
2905 /**
2906  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2907  * @ring: The Tx ring to configure
2908  *
2909  * Configure the Tx descriptor ring in the HMC context.
2910  **/
2911 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2912 {
2913 	struct i40e_vsi *vsi = ring->vsi;
2914 	u16 pf_q = vsi->base_queue + ring->queue_index;
2915 	struct i40e_hw *hw = &vsi->back->hw;
2916 	struct i40e_hmc_obj_txq tx_ctx;
2917 	i40e_status err = 0;
2918 	u32 qtx_ctl = 0;
2919 
2920 	/* some ATR related tx ring init */
2921 	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2922 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2923 		ring->atr_count = 0;
2924 	} else {
2925 		ring->atr_sample_rate = 0;
2926 	}
2927 
2928 	/* configure XPS */
2929 	i40e_config_xps_tx_ring(ring);
2930 
2931 	/* clear the context structure first */
2932 	memset(&tx_ctx, 0, sizeof(tx_ctx));
2933 
2934 	tx_ctx.new_context = 1;
2935 	tx_ctx.base = (ring->dma / 128);
2936 	tx_ctx.qlen = ring->count;
2937 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2938 					       I40E_FLAG_FD_ATR_ENABLED));
2939 	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2940 	/* FDIR VSI tx ring can still use RS bit and writebacks */
2941 	if (vsi->type != I40E_VSI_FDIR)
2942 		tx_ctx.head_wb_ena = 1;
2943 	tx_ctx.head_wb_addr = ring->dma +
2944 			      (ring->count * sizeof(struct i40e_tx_desc));
2945 
2946 	/* As part of VSI creation/update, FW allocates certain
2947 	 * Tx arbitration queue sets for each TC enabled for
2948 	 * the VSI. The FW returns the handles to these queue
2949 	 * sets as part of the response buffer to Add VSI,
2950 	 * Update VSI, etc. AQ commands. It is expected that
2951 	 * these queue set handles be associated with the Tx
2952 	 * queues by the driver as part of the TX queue context
2953 	 * initialization. This has to be done regardless of
2954 	 * DCB as by default everything is mapped to TC0.
2955 	 */
2956 	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2957 	tx_ctx.rdylist_act = 0;
2958 
2959 	/* clear the context in the HMC */
2960 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2961 	if (err) {
2962 		dev_info(&vsi->back->pdev->dev,
2963 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2964 			 ring->queue_index, pf_q, err);
2965 		return -ENOMEM;
2966 	}
2967 
2968 	/* set the context in the HMC */
2969 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2970 	if (err) {
2971 		dev_info(&vsi->back->pdev->dev,
2972 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2973 			 ring->queue_index, pf_q, err);
2974 		return -ENOMEM;
2975 	}
2976 
2977 	/* Now associate this queue with this PCI function */
2978 	if (vsi->type == I40E_VSI_VMDQ2) {
2979 		qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2980 		qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2981 			   I40E_QTX_CTL_VFVM_INDX_MASK;
2982 	} else {
2983 		qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2984 	}
2985 
2986 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2987 		    I40E_QTX_CTL_PF_INDX_MASK);
2988 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2989 	i40e_flush(hw);
2990 
2991 	/* cache tail off for easier writes later */
2992 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2993 
2994 	return 0;
2995 }
2996 
2997 /**
2998  * i40e_configure_rx_ring - Configure a receive ring context
2999  * @ring: The Rx ring to configure
3000  *
3001  * Configure the Rx descriptor ring in the HMC context.
3002  **/
3003 static int i40e_configure_rx_ring(struct i40e_ring *ring)
3004 {
3005 	struct i40e_vsi *vsi = ring->vsi;
3006 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
3007 	u16 pf_q = vsi->base_queue + ring->queue_index;
3008 	struct i40e_hw *hw = &vsi->back->hw;
3009 	struct i40e_hmc_obj_rxq rx_ctx;
3010 	i40e_status err = 0;
3011 
3012 	ring->state = 0;
3013 
3014 	/* clear the context structure first */
3015 	memset(&rx_ctx, 0, sizeof(rx_ctx));
3016 
3017 	ring->rx_buf_len = vsi->rx_buf_len;
3018 
3019 	rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3020 				    BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3021 
3022 	rx_ctx.base = (ring->dma / 128);
3023 	rx_ctx.qlen = ring->count;
3024 
3025 	/* use 32 byte descriptors */
3026 	rx_ctx.dsize = 1;
3027 
3028 	/* descriptor type is always zero
3029 	 * rx_ctx.dtype = 0;
3030 	 */
3031 	rx_ctx.hsplit_0 = 0;
3032 
3033 	rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3034 	if (hw->revision_id == 0)
3035 		rx_ctx.lrxqthresh = 0;
3036 	else
3037 		rx_ctx.lrxqthresh = 2;
3038 	rx_ctx.crcstrip = 1;
3039 	rx_ctx.l2tsel = 1;
3040 	/* this controls whether VLAN is stripped from inner headers */
3041 	rx_ctx.showiv = 0;
3042 	/* set the prefena field to 1 because the manual says to */
3043 	rx_ctx.prefena = 1;
3044 
3045 	/* clear the context in the HMC */
3046 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3047 	if (err) {
3048 		dev_info(&vsi->back->pdev->dev,
3049 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3050 			 ring->queue_index, pf_q, err);
3051 		return -ENOMEM;
3052 	}
3053 
3054 	/* set the context in the HMC */
3055 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3056 	if (err) {
3057 		dev_info(&vsi->back->pdev->dev,
3058 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3059 			 ring->queue_index, pf_q, err);
3060 		return -ENOMEM;
3061 	}
3062 
3063 	/* configure Rx buffer alignment */
3064 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3065 		clear_ring_build_skb_enabled(ring);
3066 	else
3067 		set_ring_build_skb_enabled(ring);
3068 
3069 	/* cache tail for quicker writes, and clear the reg before use */
3070 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3071 	writel(0, ring->tail);
3072 
3073 	i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3074 
3075 	return 0;
3076 }
3077 
3078 /**
3079  * i40e_vsi_configure_tx - Configure the VSI for Tx
3080  * @vsi: VSI structure describing this set of rings and resources
3081  *
3082  * Configure the Tx VSI for operation.
3083  **/
3084 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3085 {
3086 	int err = 0;
3087 	u16 i;
3088 
3089 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3090 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3091 
3092 	if (!i40e_enabled_xdp_vsi(vsi))
3093 		return err;
3094 
3095 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3096 		err = i40e_configure_tx_ring(vsi->xdp_rings[i]);
3097 
3098 	return err;
3099 }
3100 
3101 /**
3102  * i40e_vsi_configure_rx - Configure the VSI for Rx
3103  * @vsi: the VSI being configured
3104  *
3105  * Configure the Rx VSI for operation.
3106  **/
3107 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3108 {
3109 	int err = 0;
3110 	u16 i;
3111 
3112 	if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3113 		vsi->max_frame = I40E_MAX_RXBUFFER;
3114 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
3115 #if (PAGE_SIZE < 8192)
3116 	} else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3117 		   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3118 		vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3119 		vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3120 #endif
3121 	} else {
3122 		vsi->max_frame = I40E_MAX_RXBUFFER;
3123 		vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3124 						       I40E_RXBUFFER_2048;
3125 	}
3126 
3127 	/* set up individual rings */
3128 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3129 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3130 
3131 	return err;
3132 }
3133 
3134 /**
3135  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3136  * @vsi: ptr to the VSI
3137  **/
3138 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3139 {
3140 	struct i40e_ring *tx_ring, *rx_ring;
3141 	u16 qoffset, qcount;
3142 	int i, n;
3143 
3144 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3145 		/* Reset the TC information */
3146 		for (i = 0; i < vsi->num_queue_pairs; i++) {
3147 			rx_ring = vsi->rx_rings[i];
3148 			tx_ring = vsi->tx_rings[i];
3149 			rx_ring->dcb_tc = 0;
3150 			tx_ring->dcb_tc = 0;
3151 		}
3152 	}
3153 
3154 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3155 		if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3156 			continue;
3157 
3158 		qoffset = vsi->tc_config.tc_info[n].qoffset;
3159 		qcount = vsi->tc_config.tc_info[n].qcount;
3160 		for (i = qoffset; i < (qoffset + qcount); i++) {
3161 			rx_ring = vsi->rx_rings[i];
3162 			tx_ring = vsi->tx_rings[i];
3163 			rx_ring->dcb_tc = n;
3164 			tx_ring->dcb_tc = n;
3165 		}
3166 	}
3167 }
3168 
3169 /**
3170  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3171  * @vsi: ptr to the VSI
3172  **/
3173 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3174 {
3175 	if (vsi->netdev)
3176 		i40e_set_rx_mode(vsi->netdev);
3177 }
3178 
3179 /**
3180  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3181  * @vsi: Pointer to the targeted VSI
3182  *
3183  * This function replays the hlist on the hw where all the SB Flow Director
3184  * filters were saved.
3185  **/
3186 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3187 {
3188 	struct i40e_fdir_filter *filter;
3189 	struct i40e_pf *pf = vsi->back;
3190 	struct hlist_node *node;
3191 
3192 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3193 		return;
3194 
3195 	/* Reset FDir counters as we're replaying all existing filters */
3196 	pf->fd_tcp4_filter_cnt = 0;
3197 	pf->fd_udp4_filter_cnt = 0;
3198 	pf->fd_sctp4_filter_cnt = 0;
3199 	pf->fd_ip4_filter_cnt = 0;
3200 
3201 	hlist_for_each_entry_safe(filter, node,
3202 				  &pf->fdir_filter_list, fdir_node) {
3203 		i40e_add_del_fdir(vsi, filter, true);
3204 	}
3205 }
3206 
3207 /**
3208  * i40e_vsi_configure - Set up the VSI for action
3209  * @vsi: the VSI being configured
3210  **/
3211 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3212 {
3213 	int err;
3214 
3215 	i40e_set_vsi_rx_mode(vsi);
3216 	i40e_restore_vlan(vsi);
3217 	i40e_vsi_config_dcb_rings(vsi);
3218 	err = i40e_vsi_configure_tx(vsi);
3219 	if (!err)
3220 		err = i40e_vsi_configure_rx(vsi);
3221 
3222 	return err;
3223 }
3224 
3225 /**
3226  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3227  * @vsi: the VSI being configured
3228  **/
3229 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3230 {
3231 	bool has_xdp = i40e_enabled_xdp_vsi(vsi);
3232 	struct i40e_pf *pf = vsi->back;
3233 	struct i40e_hw *hw = &pf->hw;
3234 	u16 vector;
3235 	int i, q;
3236 	u32 qp;
3237 
3238 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
3239 	 * and PFINT_LNKLSTn registers, e.g.:
3240 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3241 	 */
3242 	qp = vsi->base_queue;
3243 	vector = vsi->base_vector;
3244 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3245 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3246 
3247 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
3248 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3249 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
3250 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3251 		     q_vector->rx.itr);
3252 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3253 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
3254 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3255 		     q_vector->tx.itr);
3256 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
3257 		     i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3258 
3259 		/* Linked list for the queuepairs assigned to this vector */
3260 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3261 		for (q = 0; q < q_vector->num_ringpairs; q++) {
3262 			u32 nextqp = has_xdp ? qp + vsi->alloc_queue_pairs : qp;
3263 			u32 val;
3264 
3265 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3266 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3267 			      (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3268 			      (nextqp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT) |
3269 			      (I40E_QUEUE_TYPE_TX <<
3270 			       I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3271 
3272 			wr32(hw, I40E_QINT_RQCTL(qp), val);
3273 
3274 			if (has_xdp) {
3275 				val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3276 				      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3277 				      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3278 				      (qp << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3279 				      (I40E_QUEUE_TYPE_TX <<
3280 				       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3281 
3282 				wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3283 			}
3284 
3285 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3286 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3287 			      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3288 			      ((qp + 1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3289 			      (I40E_QUEUE_TYPE_RX <<
3290 			       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3291 
3292 			/* Terminate the linked list */
3293 			if (q == (q_vector->num_ringpairs - 1))
3294 				val |= (I40E_QUEUE_END_OF_LIST <<
3295 					I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3296 
3297 			wr32(hw, I40E_QINT_TQCTL(qp), val);
3298 			qp++;
3299 		}
3300 	}
3301 
3302 	i40e_flush(hw);
3303 }
3304 
3305 /**
3306  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3307  * @hw: ptr to the hardware info
3308  **/
3309 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3310 {
3311 	struct i40e_hw *hw = &pf->hw;
3312 	u32 val;
3313 
3314 	/* clear things first */
3315 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3316 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3317 
3318 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3319 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3320 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
3321 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3322 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3323 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3324 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3325 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3326 
3327 	if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3328 		val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3329 
3330 	if (pf->flags & I40E_FLAG_PTP)
3331 		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3332 
3333 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
3334 
3335 	/* SW_ITR_IDX = 0, but don't change INTENA */
3336 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3337 					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3338 
3339 	/* OTHER_ITR_IDX = 0 */
3340 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3341 }
3342 
3343 /**
3344  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3345  * @vsi: the VSI being configured
3346  **/
3347 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3348 {
3349 	u32 nextqp = i40e_enabled_xdp_vsi(vsi) ? vsi->alloc_queue_pairs : 0;
3350 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3351 	struct i40e_pf *pf = vsi->back;
3352 	struct i40e_hw *hw = &pf->hw;
3353 	u32 val;
3354 
3355 	/* set the ITR configuration */
3356 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
3357 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3358 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
3359 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3360 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3361 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
3362 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3363 
3364 	i40e_enable_misc_int_causes(pf);
3365 
3366 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3367 	wr32(hw, I40E_PFINT_LNKLST0, 0);
3368 
3369 	/* Associate the queue pair to the vector and enable the queue int */
3370 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		       |
3371 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3372 	      (nextqp	   << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3373 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3374 
3375 	wr32(hw, I40E_QINT_RQCTL(0), val);
3376 
3377 	if (i40e_enabled_xdp_vsi(vsi)) {
3378 		val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		     |
3379 		      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)|
3380 		      (I40E_QUEUE_TYPE_TX
3381 		       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3382 
3383 	       wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3384 	}
3385 
3386 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
3387 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3388 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3389 
3390 	wr32(hw, I40E_QINT_TQCTL(0), val);
3391 	i40e_flush(hw);
3392 }
3393 
3394 /**
3395  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3396  * @pf: board private structure
3397  **/
3398 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3399 {
3400 	struct i40e_hw *hw = &pf->hw;
3401 
3402 	wr32(hw, I40E_PFINT_DYN_CTL0,
3403 	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3404 	i40e_flush(hw);
3405 }
3406 
3407 /**
3408  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3409  * @pf: board private structure
3410  * @clearpba: true when all pending interrupt events should be cleared
3411  **/
3412 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3413 {
3414 	struct i40e_hw *hw = &pf->hw;
3415 	u32 val;
3416 
3417 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3418 	      (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3419 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3420 
3421 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
3422 	i40e_flush(hw);
3423 }
3424 
3425 /**
3426  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3427  * @irq: interrupt number
3428  * @data: pointer to a q_vector
3429  **/
3430 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3431 {
3432 	struct i40e_q_vector *q_vector = data;
3433 
3434 	if (!q_vector->tx.ring && !q_vector->rx.ring)
3435 		return IRQ_HANDLED;
3436 
3437 	napi_schedule_irqoff(&q_vector->napi);
3438 
3439 	return IRQ_HANDLED;
3440 }
3441 
3442 /**
3443  * i40e_irq_affinity_notify - Callback for affinity changes
3444  * @notify: context as to what irq was changed
3445  * @mask: the new affinity mask
3446  *
3447  * This is a callback function used by the irq_set_affinity_notifier function
3448  * so that we may register to receive changes to the irq affinity masks.
3449  **/
3450 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3451 				     const cpumask_t *mask)
3452 {
3453 	struct i40e_q_vector *q_vector =
3454 		container_of(notify, struct i40e_q_vector, affinity_notify);
3455 
3456 	cpumask_copy(&q_vector->affinity_mask, mask);
3457 }
3458 
3459 /**
3460  * i40e_irq_affinity_release - Callback for affinity notifier release
3461  * @ref: internal core kernel usage
3462  *
3463  * This is a callback function used by the irq_set_affinity_notifier function
3464  * to inform the current notification subscriber that they will no longer
3465  * receive notifications.
3466  **/
3467 static void i40e_irq_affinity_release(struct kref *ref) {}
3468 
3469 /**
3470  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3471  * @vsi: the VSI being configured
3472  * @basename: name for the vector
3473  *
3474  * Allocates MSI-X vectors and requests interrupts from the kernel.
3475  **/
3476 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3477 {
3478 	int q_vectors = vsi->num_q_vectors;
3479 	struct i40e_pf *pf = vsi->back;
3480 	int base = vsi->base_vector;
3481 	int rx_int_idx = 0;
3482 	int tx_int_idx = 0;
3483 	int vector, err;
3484 	int irq_num;
3485 
3486 	for (vector = 0; vector < q_vectors; vector++) {
3487 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3488 
3489 		irq_num = pf->msix_entries[base + vector].vector;
3490 
3491 		if (q_vector->tx.ring && q_vector->rx.ring) {
3492 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3493 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3494 			tx_int_idx++;
3495 		} else if (q_vector->rx.ring) {
3496 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3497 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3498 		} else if (q_vector->tx.ring) {
3499 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3500 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3501 		} else {
3502 			/* skip this unused q_vector */
3503 			continue;
3504 		}
3505 		err = request_irq(irq_num,
3506 				  vsi->irq_handler,
3507 				  0,
3508 				  q_vector->name,
3509 				  q_vector);
3510 		if (err) {
3511 			dev_info(&pf->pdev->dev,
3512 				 "MSIX request_irq failed, error: %d\n", err);
3513 			goto free_queue_irqs;
3514 		}
3515 
3516 		/* register for affinity change notifications */
3517 		q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3518 		q_vector->affinity_notify.release = i40e_irq_affinity_release;
3519 		irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3520 		/* get_cpu_mask returns a static constant mask with
3521 		 * a permanent lifetime so it's ok to use here.
3522 		 */
3523 		irq_set_affinity_hint(irq_num, get_cpu_mask(q_vector->v_idx));
3524 	}
3525 
3526 	vsi->irqs_ready = true;
3527 	return 0;
3528 
3529 free_queue_irqs:
3530 	while (vector) {
3531 		vector--;
3532 		irq_num = pf->msix_entries[base + vector].vector;
3533 		irq_set_affinity_notifier(irq_num, NULL);
3534 		irq_set_affinity_hint(irq_num, NULL);
3535 		free_irq(irq_num, &vsi->q_vectors[vector]);
3536 	}
3537 	return err;
3538 }
3539 
3540 /**
3541  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3542  * @vsi: the VSI being un-configured
3543  **/
3544 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3545 {
3546 	struct i40e_pf *pf = vsi->back;
3547 	struct i40e_hw *hw = &pf->hw;
3548 	int base = vsi->base_vector;
3549 	int i;
3550 
3551 	/* disable interrupt causation from each queue */
3552 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3553 		u32 val;
3554 
3555 		val = rd32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx));
3556 		val &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3557 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), val);
3558 
3559 		val = rd32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx));
3560 		val &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3561 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), val);
3562 
3563 		if (!i40e_enabled_xdp_vsi(vsi))
3564 			continue;
3565 		wr32(hw, I40E_QINT_TQCTL(vsi->xdp_rings[i]->reg_idx), 0);
3566 	}
3567 
3568 	/* disable each interrupt */
3569 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3570 		for (i = vsi->base_vector;
3571 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3572 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3573 
3574 		i40e_flush(hw);
3575 		for (i = 0; i < vsi->num_q_vectors; i++)
3576 			synchronize_irq(pf->msix_entries[i + base].vector);
3577 	} else {
3578 		/* Legacy and MSI mode - this stops all interrupt handling */
3579 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3580 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3581 		i40e_flush(hw);
3582 		synchronize_irq(pf->pdev->irq);
3583 	}
3584 }
3585 
3586 /**
3587  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3588  * @vsi: the VSI being configured
3589  **/
3590 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3591 {
3592 	struct i40e_pf *pf = vsi->back;
3593 	int i;
3594 
3595 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3596 		for (i = 0; i < vsi->num_q_vectors; i++)
3597 			i40e_irq_dynamic_enable(vsi, i);
3598 	} else {
3599 		i40e_irq_dynamic_enable_icr0(pf, true);
3600 	}
3601 
3602 	i40e_flush(&pf->hw);
3603 	return 0;
3604 }
3605 
3606 /**
3607  * i40e_free_misc_vector - Free the vector that handles non-queue events
3608  * @pf: board private structure
3609  **/
3610 static void i40e_free_misc_vector(struct i40e_pf *pf)
3611 {
3612 	/* Disable ICR 0 */
3613 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3614 	i40e_flush(&pf->hw);
3615 
3616 	if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
3617 		synchronize_irq(pf->msix_entries[0].vector);
3618 		free_irq(pf->msix_entries[0].vector, pf);
3619 		clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
3620 	}
3621 }
3622 
3623 /**
3624  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3625  * @irq: interrupt number
3626  * @data: pointer to a q_vector
3627  *
3628  * This is the handler used for all MSI/Legacy interrupts, and deals
3629  * with both queue and non-queue interrupts.  This is also used in
3630  * MSIX mode to handle the non-queue interrupts.
3631  **/
3632 static irqreturn_t i40e_intr(int irq, void *data)
3633 {
3634 	struct i40e_pf *pf = (struct i40e_pf *)data;
3635 	struct i40e_hw *hw = &pf->hw;
3636 	irqreturn_t ret = IRQ_NONE;
3637 	u32 icr0, icr0_remaining;
3638 	u32 val, ena_mask;
3639 
3640 	icr0 = rd32(hw, I40E_PFINT_ICR0);
3641 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3642 
3643 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3644 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3645 		goto enable_intr;
3646 
3647 	/* if interrupt but no bits showing, must be SWINT */
3648 	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3649 	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3650 		pf->sw_int_count++;
3651 
3652 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3653 	    (icr0 & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3654 		ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3655 		dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3656 		set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
3657 	}
3658 
3659 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3660 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3661 		struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3662 		struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3663 
3664 		/* We do not have a way to disarm Queue causes while leaving
3665 		 * interrupt enabled for all other causes, ideally
3666 		 * interrupt should be disabled while we are in NAPI but
3667 		 * this is not a performance path and napi_schedule()
3668 		 * can deal with rescheduling.
3669 		 */
3670 		if (!test_bit(__I40E_DOWN, pf->state))
3671 			napi_schedule_irqoff(&q_vector->napi);
3672 	}
3673 
3674 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3675 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3676 		set_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
3677 		i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3678 	}
3679 
3680 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3681 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3682 		set_bit(__I40E_MDD_EVENT_PENDING, pf->state);
3683 	}
3684 
3685 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3686 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3687 		set_bit(__I40E_VFLR_EVENT_PENDING, pf->state);
3688 	}
3689 
3690 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3691 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
3692 			set_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
3693 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3694 		val = rd32(hw, I40E_GLGEN_RSTAT);
3695 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3696 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3697 		if (val == I40E_RESET_CORER) {
3698 			pf->corer_count++;
3699 		} else if (val == I40E_RESET_GLOBR) {
3700 			pf->globr_count++;
3701 		} else if (val == I40E_RESET_EMPR) {
3702 			pf->empr_count++;
3703 			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state);
3704 		}
3705 	}
3706 
3707 	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3708 		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3709 		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3710 		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3711 			 rd32(hw, I40E_PFHMC_ERRORINFO),
3712 			 rd32(hw, I40E_PFHMC_ERRORDATA));
3713 	}
3714 
3715 	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3716 		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3717 
3718 		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3719 			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3720 			i40e_ptp_tx_hwtstamp(pf);
3721 		}
3722 	}
3723 
3724 	/* If a critical error is pending we have no choice but to reset the
3725 	 * device.
3726 	 * Report and mask out any remaining unexpected interrupts.
3727 	 */
3728 	icr0_remaining = icr0 & ena_mask;
3729 	if (icr0_remaining) {
3730 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3731 			 icr0_remaining);
3732 		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3733 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3734 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3735 			dev_info(&pf->pdev->dev, "device will be reset\n");
3736 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
3737 			i40e_service_event_schedule(pf);
3738 		}
3739 		ena_mask &= ~icr0_remaining;
3740 	}
3741 	ret = IRQ_HANDLED;
3742 
3743 enable_intr:
3744 	/* re-enable interrupt causes */
3745 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3746 	if (!test_bit(__I40E_DOWN, pf->state)) {
3747 		i40e_service_event_schedule(pf);
3748 		i40e_irq_dynamic_enable_icr0(pf, false);
3749 	}
3750 
3751 	return ret;
3752 }
3753 
3754 /**
3755  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3756  * @tx_ring:  tx ring to clean
3757  * @budget:   how many cleans we're allowed
3758  *
3759  * Returns true if there's any budget left (e.g. the clean is finished)
3760  **/
3761 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3762 {
3763 	struct i40e_vsi *vsi = tx_ring->vsi;
3764 	u16 i = tx_ring->next_to_clean;
3765 	struct i40e_tx_buffer *tx_buf;
3766 	struct i40e_tx_desc *tx_desc;
3767 
3768 	tx_buf = &tx_ring->tx_bi[i];
3769 	tx_desc = I40E_TX_DESC(tx_ring, i);
3770 	i -= tx_ring->count;
3771 
3772 	do {
3773 		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3774 
3775 		/* if next_to_watch is not set then there is no work pending */
3776 		if (!eop_desc)
3777 			break;
3778 
3779 		/* prevent any other reads prior to eop_desc */
3780 		read_barrier_depends();
3781 
3782 		/* if the descriptor isn't done, no work yet to do */
3783 		if (!(eop_desc->cmd_type_offset_bsz &
3784 		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3785 			break;
3786 
3787 		/* clear next_to_watch to prevent false hangs */
3788 		tx_buf->next_to_watch = NULL;
3789 
3790 		tx_desc->buffer_addr = 0;
3791 		tx_desc->cmd_type_offset_bsz = 0;
3792 		/* move past filter desc */
3793 		tx_buf++;
3794 		tx_desc++;
3795 		i++;
3796 		if (unlikely(!i)) {
3797 			i -= tx_ring->count;
3798 			tx_buf = tx_ring->tx_bi;
3799 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3800 		}
3801 		/* unmap skb header data */
3802 		dma_unmap_single(tx_ring->dev,
3803 				 dma_unmap_addr(tx_buf, dma),
3804 				 dma_unmap_len(tx_buf, len),
3805 				 DMA_TO_DEVICE);
3806 		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3807 			kfree(tx_buf->raw_buf);
3808 
3809 		tx_buf->raw_buf = NULL;
3810 		tx_buf->tx_flags = 0;
3811 		tx_buf->next_to_watch = NULL;
3812 		dma_unmap_len_set(tx_buf, len, 0);
3813 		tx_desc->buffer_addr = 0;
3814 		tx_desc->cmd_type_offset_bsz = 0;
3815 
3816 		/* move us past the eop_desc for start of next FD desc */
3817 		tx_buf++;
3818 		tx_desc++;
3819 		i++;
3820 		if (unlikely(!i)) {
3821 			i -= tx_ring->count;
3822 			tx_buf = tx_ring->tx_bi;
3823 			tx_desc = I40E_TX_DESC(tx_ring, 0);
3824 		}
3825 
3826 		/* update budget accounting */
3827 		budget--;
3828 	} while (likely(budget));
3829 
3830 	i += tx_ring->count;
3831 	tx_ring->next_to_clean = i;
3832 
3833 	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3834 		i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3835 
3836 	return budget > 0;
3837 }
3838 
3839 /**
3840  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3841  * @irq: interrupt number
3842  * @data: pointer to a q_vector
3843  **/
3844 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3845 {
3846 	struct i40e_q_vector *q_vector = data;
3847 	struct i40e_vsi *vsi;
3848 
3849 	if (!q_vector->tx.ring)
3850 		return IRQ_HANDLED;
3851 
3852 	vsi = q_vector->tx.ring->vsi;
3853 	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3854 
3855 	return IRQ_HANDLED;
3856 }
3857 
3858 /**
3859  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3860  * @vsi: the VSI being configured
3861  * @v_idx: vector index
3862  * @qp_idx: queue pair index
3863  **/
3864 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3865 {
3866 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3867 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3868 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3869 
3870 	tx_ring->q_vector = q_vector;
3871 	tx_ring->next = q_vector->tx.ring;
3872 	q_vector->tx.ring = tx_ring;
3873 	q_vector->tx.count++;
3874 
3875 	/* Place XDP Tx ring in the same q_vector ring list as regular Tx */
3876 	if (i40e_enabled_xdp_vsi(vsi)) {
3877 		struct i40e_ring *xdp_ring = vsi->xdp_rings[qp_idx];
3878 
3879 		xdp_ring->q_vector = q_vector;
3880 		xdp_ring->next = q_vector->tx.ring;
3881 		q_vector->tx.ring = xdp_ring;
3882 		q_vector->tx.count++;
3883 	}
3884 
3885 	rx_ring->q_vector = q_vector;
3886 	rx_ring->next = q_vector->rx.ring;
3887 	q_vector->rx.ring = rx_ring;
3888 	q_vector->rx.count++;
3889 }
3890 
3891 /**
3892  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3893  * @vsi: the VSI being configured
3894  *
3895  * This function maps descriptor rings to the queue-specific vectors
3896  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3897  * one vector per queue pair, but on a constrained vector budget, we
3898  * group the queue pairs as "efficiently" as possible.
3899  **/
3900 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3901 {
3902 	int qp_remaining = vsi->num_queue_pairs;
3903 	int q_vectors = vsi->num_q_vectors;
3904 	int num_ringpairs;
3905 	int v_start = 0;
3906 	int qp_idx = 0;
3907 
3908 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3909 	 * group them so there are multiple queues per vector.
3910 	 * It is also important to go through all the vectors available to be
3911 	 * sure that if we don't use all the vectors, that the remaining vectors
3912 	 * are cleared. This is especially important when decreasing the
3913 	 * number of queues in use.
3914 	 */
3915 	for (; v_start < q_vectors; v_start++) {
3916 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3917 
3918 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3919 
3920 		q_vector->num_ringpairs = num_ringpairs;
3921 
3922 		q_vector->rx.count = 0;
3923 		q_vector->tx.count = 0;
3924 		q_vector->rx.ring = NULL;
3925 		q_vector->tx.ring = NULL;
3926 
3927 		while (num_ringpairs--) {
3928 			i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3929 			qp_idx++;
3930 			qp_remaining--;
3931 		}
3932 	}
3933 }
3934 
3935 /**
3936  * i40e_vsi_request_irq - Request IRQ from the OS
3937  * @vsi: the VSI being configured
3938  * @basename: name for the vector
3939  **/
3940 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3941 {
3942 	struct i40e_pf *pf = vsi->back;
3943 	int err;
3944 
3945 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3946 		err = i40e_vsi_request_irq_msix(vsi, basename);
3947 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3948 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
3949 				  pf->int_name, pf);
3950 	else
3951 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3952 				  pf->int_name, pf);
3953 
3954 	if (err)
3955 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3956 
3957 	return err;
3958 }
3959 
3960 #ifdef CONFIG_NET_POLL_CONTROLLER
3961 /**
3962  * i40e_netpoll - A Polling 'interrupt' handler
3963  * @netdev: network interface device structure
3964  *
3965  * This is used by netconsole to send skbs without having to re-enable
3966  * interrupts.  It's not called while the normal interrupt routine is executing.
3967  **/
3968 static void i40e_netpoll(struct net_device *netdev)
3969 {
3970 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3971 	struct i40e_vsi *vsi = np->vsi;
3972 	struct i40e_pf *pf = vsi->back;
3973 	int i;
3974 
3975 	/* if interface is down do nothing */
3976 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
3977 		return;
3978 
3979 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3980 		for (i = 0; i < vsi->num_q_vectors; i++)
3981 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3982 	} else {
3983 		i40e_intr(pf->pdev->irq, netdev);
3984 	}
3985 }
3986 #endif
3987 
3988 #define I40E_QTX_ENA_WAIT_COUNT 50
3989 
3990 /**
3991  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3992  * @pf: the PF being configured
3993  * @pf_q: the PF queue
3994  * @enable: enable or disable state of the queue
3995  *
3996  * This routine will wait for the given Tx queue of the PF to reach the
3997  * enabled or disabled state.
3998  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3999  * multiple retries; else will return 0 in case of success.
4000  **/
4001 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4002 {
4003 	int i;
4004 	u32 tx_reg;
4005 
4006 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4007 		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
4008 		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4009 			break;
4010 
4011 		usleep_range(10, 20);
4012 	}
4013 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4014 		return -ETIMEDOUT;
4015 
4016 	return 0;
4017 }
4018 
4019 /**
4020  * i40e_control_tx_q - Start or stop a particular Tx queue
4021  * @pf: the PF structure
4022  * @pf_q: the PF queue to configure
4023  * @enable: start or stop the queue
4024  *
4025  * This function enables or disables a single queue. Note that any delay
4026  * required after the operation is expected to be handled by the caller of
4027  * this function.
4028  **/
4029 static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
4030 {
4031 	struct i40e_hw *hw = &pf->hw;
4032 	u32 tx_reg;
4033 	int i;
4034 
4035 	/* warn the TX unit of coming changes */
4036 	i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
4037 	if (!enable)
4038 		usleep_range(10, 20);
4039 
4040 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4041 		tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
4042 		if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
4043 		    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
4044 			break;
4045 		usleep_range(1000, 2000);
4046 	}
4047 
4048 	/* Skip if the queue is already in the requested state */
4049 	if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4050 		return;
4051 
4052 	/* turn on/off the queue */
4053 	if (enable) {
4054 		wr32(hw, I40E_QTX_HEAD(pf_q), 0);
4055 		tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
4056 	} else {
4057 		tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
4058 	}
4059 
4060 	wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
4061 }
4062 
4063 /**
4064  * i40e_control_wait_tx_q - Start/stop Tx queue and wait for completion
4065  * @seid: VSI SEID
4066  * @pf: the PF structure
4067  * @pf_q: the PF queue to configure
4068  * @is_xdp: true if the queue is used for XDP
4069  * @enable: start or stop the queue
4070  **/
4071 static int i40e_control_wait_tx_q(int seid, struct i40e_pf *pf, int pf_q,
4072 				  bool is_xdp, bool enable)
4073 {
4074 	int ret;
4075 
4076 	i40e_control_tx_q(pf, pf_q, enable);
4077 
4078 	/* wait for the change to finish */
4079 	ret = i40e_pf_txq_wait(pf, pf_q, enable);
4080 	if (ret) {
4081 		dev_info(&pf->pdev->dev,
4082 			 "VSI seid %d %sTx ring %d %sable timeout\n",
4083 			 seid, (is_xdp ? "XDP " : ""), pf_q,
4084 			 (enable ? "en" : "dis"));
4085 	}
4086 
4087 	return ret;
4088 }
4089 
4090 /**
4091  * i40e_vsi_control_tx - Start or stop a VSI's rings
4092  * @vsi: the VSI being configured
4093  * @enable: start or stop the rings
4094  **/
4095 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4096 {
4097 	struct i40e_pf *pf = vsi->back;
4098 	int i, pf_q, ret = 0;
4099 
4100 	pf_q = vsi->base_queue;
4101 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4102 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4103 					     pf_q,
4104 					     false /*is xdp*/, enable);
4105 		if (ret)
4106 			break;
4107 
4108 		if (!i40e_enabled_xdp_vsi(vsi))
4109 			continue;
4110 
4111 		ret = i40e_control_wait_tx_q(vsi->seid, pf,
4112 					     pf_q + vsi->alloc_queue_pairs,
4113 					     true /*is xdp*/, enable);
4114 		if (ret)
4115 			break;
4116 	}
4117 
4118 	return ret;
4119 }
4120 
4121 /**
4122  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4123  * @pf: the PF being configured
4124  * @pf_q: the PF queue
4125  * @enable: enable or disable state of the queue
4126  *
4127  * This routine will wait for the given Rx queue of the PF to reach the
4128  * enabled or disabled state.
4129  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4130  * multiple retries; else will return 0 in case of success.
4131  **/
4132 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4133 {
4134 	int i;
4135 	u32 rx_reg;
4136 
4137 	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4138 		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4139 		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4140 			break;
4141 
4142 		usleep_range(10, 20);
4143 	}
4144 	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4145 		return -ETIMEDOUT;
4146 
4147 	return 0;
4148 }
4149 
4150 /**
4151  * i40e_control_rx_q - Start or stop a particular Rx queue
4152  * @pf: the PF structure
4153  * @pf_q: the PF queue to configure
4154  * @enable: start or stop the queue
4155  *
4156  * This function enables or disables a single queue. Note that any delay
4157  * required after the operation is expected to be handled by the caller of
4158  * this function.
4159  **/
4160 static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4161 {
4162 	struct i40e_hw *hw = &pf->hw;
4163 	u32 rx_reg;
4164 	int i;
4165 
4166 	for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4167 		rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4168 		if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4169 		    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4170 			break;
4171 		usleep_range(1000, 2000);
4172 	}
4173 
4174 	/* Skip if the queue is already in the requested state */
4175 	if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4176 		return;
4177 
4178 	/* turn on/off the queue */
4179 	if (enable)
4180 		rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4181 	else
4182 		rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4183 
4184 	wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4185 }
4186 
4187 /**
4188  * i40e_vsi_control_rx - Start or stop a VSI's rings
4189  * @vsi: the VSI being configured
4190  * @enable: start or stop the rings
4191  **/
4192 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4193 {
4194 	struct i40e_pf *pf = vsi->back;
4195 	int i, pf_q, ret = 0;
4196 
4197 	pf_q = vsi->base_queue;
4198 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4199 		i40e_control_rx_q(pf, pf_q, enable);
4200 
4201 		/* wait for the change to finish */
4202 		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4203 		if (ret) {
4204 			dev_info(&pf->pdev->dev,
4205 				 "VSI seid %d Rx ring %d %sable timeout\n",
4206 				 vsi->seid, pf_q, (enable ? "en" : "dis"));
4207 			break;
4208 		}
4209 	}
4210 
4211 	/* Due to HW errata, on Rx disable only, the register can indicate done
4212 	 * before it really is. Needs 50ms to be sure
4213 	 */
4214 	if (!enable)
4215 		mdelay(50);
4216 
4217 	return ret;
4218 }
4219 
4220 /**
4221  * i40e_vsi_start_rings - Start a VSI's rings
4222  * @vsi: the VSI being configured
4223  **/
4224 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4225 {
4226 	int ret = 0;
4227 
4228 	/* do rx first for enable and last for disable */
4229 	ret = i40e_vsi_control_rx(vsi, true);
4230 	if (ret)
4231 		return ret;
4232 	ret = i40e_vsi_control_tx(vsi, true);
4233 
4234 	return ret;
4235 }
4236 
4237 /**
4238  * i40e_vsi_stop_rings - Stop a VSI's rings
4239  * @vsi: the VSI being configured
4240  **/
4241 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4242 {
4243 	/* When port TX is suspended, don't wait */
4244 	if (test_bit(__I40E_PORT_SUSPENDED, vsi->back->state))
4245 		return i40e_vsi_stop_rings_no_wait(vsi);
4246 
4247 	/* do rx first for enable and last for disable
4248 	 * Ignore return value, we need to shutdown whatever we can
4249 	 */
4250 	i40e_vsi_control_tx(vsi, false);
4251 	i40e_vsi_control_rx(vsi, false);
4252 }
4253 
4254 /**
4255  * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4256  * @vsi: the VSI being shutdown
4257  *
4258  * This function stops all the rings for a VSI but does not delay to verify
4259  * that rings have been disabled. It is expected that the caller is shutting
4260  * down multiple VSIs at once and will delay together for all the VSIs after
4261  * initiating the shutdown. This is particularly useful for shutting down lots
4262  * of VFs together. Otherwise, a large delay can be incurred while configuring
4263  * each VSI in serial.
4264  **/
4265 void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4266 {
4267 	struct i40e_pf *pf = vsi->back;
4268 	int i, pf_q;
4269 
4270 	pf_q = vsi->base_queue;
4271 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4272 		i40e_control_tx_q(pf, pf_q, false);
4273 		i40e_control_rx_q(pf, pf_q, false);
4274 	}
4275 }
4276 
4277 /**
4278  * i40e_vsi_free_irq - Free the irq association with the OS
4279  * @vsi: the VSI being configured
4280  **/
4281 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4282 {
4283 	struct i40e_pf *pf = vsi->back;
4284 	struct i40e_hw *hw = &pf->hw;
4285 	int base = vsi->base_vector;
4286 	u32 val, qp;
4287 	int i;
4288 
4289 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4290 		if (!vsi->q_vectors)
4291 			return;
4292 
4293 		if (!vsi->irqs_ready)
4294 			return;
4295 
4296 		vsi->irqs_ready = false;
4297 		for (i = 0; i < vsi->num_q_vectors; i++) {
4298 			int irq_num;
4299 			u16 vector;
4300 
4301 			vector = i + base;
4302 			irq_num = pf->msix_entries[vector].vector;
4303 
4304 			/* free only the irqs that were actually requested */
4305 			if (!vsi->q_vectors[i] ||
4306 			    !vsi->q_vectors[i]->num_ringpairs)
4307 				continue;
4308 
4309 			/* clear the affinity notifier in the IRQ descriptor */
4310 			irq_set_affinity_notifier(irq_num, NULL);
4311 			/* remove our suggested affinity mask for this IRQ */
4312 			irq_set_affinity_hint(irq_num, NULL);
4313 			synchronize_irq(irq_num);
4314 			free_irq(irq_num, vsi->q_vectors[i]);
4315 
4316 			/* Tear down the interrupt queue link list
4317 			 *
4318 			 * We know that they come in pairs and always
4319 			 * the Rx first, then the Tx.  To clear the
4320 			 * link list, stick the EOL value into the
4321 			 * next_q field of the registers.
4322 			 */
4323 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4324 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4325 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4326 			val |= I40E_QUEUE_END_OF_LIST
4327 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4328 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4329 
4330 			while (qp != I40E_QUEUE_END_OF_LIST) {
4331 				u32 next;
4332 
4333 				val = rd32(hw, I40E_QINT_RQCTL(qp));
4334 
4335 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4336 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4337 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4338 					 I40E_QINT_RQCTL_INTEVENT_MASK);
4339 
4340 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4341 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4342 
4343 				wr32(hw, I40E_QINT_RQCTL(qp), val);
4344 
4345 				val = rd32(hw, I40E_QINT_TQCTL(qp));
4346 
4347 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4348 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4349 
4350 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4351 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4352 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4353 					 I40E_QINT_TQCTL_INTEVENT_MASK);
4354 
4355 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4356 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4357 
4358 				wr32(hw, I40E_QINT_TQCTL(qp), val);
4359 				qp = next;
4360 			}
4361 		}
4362 	} else {
4363 		free_irq(pf->pdev->irq, pf);
4364 
4365 		val = rd32(hw, I40E_PFINT_LNKLST0);
4366 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4367 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4368 		val |= I40E_QUEUE_END_OF_LIST
4369 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4370 		wr32(hw, I40E_PFINT_LNKLST0, val);
4371 
4372 		val = rd32(hw, I40E_QINT_RQCTL(qp));
4373 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4374 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4375 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4376 			 I40E_QINT_RQCTL_INTEVENT_MASK);
4377 
4378 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4379 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4380 
4381 		wr32(hw, I40E_QINT_RQCTL(qp), val);
4382 
4383 		val = rd32(hw, I40E_QINT_TQCTL(qp));
4384 
4385 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4386 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4387 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4388 			 I40E_QINT_TQCTL_INTEVENT_MASK);
4389 
4390 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4391 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4392 
4393 		wr32(hw, I40E_QINT_TQCTL(qp), val);
4394 	}
4395 }
4396 
4397 /**
4398  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4399  * @vsi: the VSI being configured
4400  * @v_idx: Index of vector to be freed
4401  *
4402  * This function frees the memory allocated to the q_vector.  In addition if
4403  * NAPI is enabled it will delete any references to the NAPI struct prior
4404  * to freeing the q_vector.
4405  **/
4406 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4407 {
4408 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4409 	struct i40e_ring *ring;
4410 
4411 	if (!q_vector)
4412 		return;
4413 
4414 	/* disassociate q_vector from rings */
4415 	i40e_for_each_ring(ring, q_vector->tx)
4416 		ring->q_vector = NULL;
4417 
4418 	i40e_for_each_ring(ring, q_vector->rx)
4419 		ring->q_vector = NULL;
4420 
4421 	/* only VSI w/ an associated netdev is set up w/ NAPI */
4422 	if (vsi->netdev)
4423 		netif_napi_del(&q_vector->napi);
4424 
4425 	vsi->q_vectors[v_idx] = NULL;
4426 
4427 	kfree_rcu(q_vector, rcu);
4428 }
4429 
4430 /**
4431  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4432  * @vsi: the VSI being un-configured
4433  *
4434  * This frees the memory allocated to the q_vectors and
4435  * deletes references to the NAPI struct.
4436  **/
4437 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4438 {
4439 	int v_idx;
4440 
4441 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4442 		i40e_free_q_vector(vsi, v_idx);
4443 }
4444 
4445 /**
4446  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4447  * @pf: board private structure
4448  **/
4449 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4450 {
4451 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4452 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4453 		pci_disable_msix(pf->pdev);
4454 		kfree(pf->msix_entries);
4455 		pf->msix_entries = NULL;
4456 		kfree(pf->irq_pile);
4457 		pf->irq_pile = NULL;
4458 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4459 		pci_disable_msi(pf->pdev);
4460 	}
4461 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4462 }
4463 
4464 /**
4465  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4466  * @pf: board private structure
4467  *
4468  * We go through and clear interrupt specific resources and reset the structure
4469  * to pre-load conditions
4470  **/
4471 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4472 {
4473 	int i;
4474 
4475 	i40e_free_misc_vector(pf);
4476 
4477 	i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4478 		      I40E_IWARP_IRQ_PILE_ID);
4479 
4480 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4481 	for (i = 0; i < pf->num_alloc_vsi; i++)
4482 		if (pf->vsi[i])
4483 			i40e_vsi_free_q_vectors(pf->vsi[i]);
4484 	i40e_reset_interrupt_capability(pf);
4485 }
4486 
4487 /**
4488  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4489  * @vsi: the VSI being configured
4490  **/
4491 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4492 {
4493 	int q_idx;
4494 
4495 	if (!vsi->netdev)
4496 		return;
4497 
4498 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4499 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4500 
4501 		if (q_vector->rx.ring || q_vector->tx.ring)
4502 			napi_enable(&q_vector->napi);
4503 	}
4504 }
4505 
4506 /**
4507  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4508  * @vsi: the VSI being configured
4509  **/
4510 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4511 {
4512 	int q_idx;
4513 
4514 	if (!vsi->netdev)
4515 		return;
4516 
4517 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4518 		struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4519 
4520 		if (q_vector->rx.ring || q_vector->tx.ring)
4521 			napi_disable(&q_vector->napi);
4522 	}
4523 }
4524 
4525 /**
4526  * i40e_vsi_close - Shut down a VSI
4527  * @vsi: the vsi to be quelled
4528  **/
4529 static void i40e_vsi_close(struct i40e_vsi *vsi)
4530 {
4531 	struct i40e_pf *pf = vsi->back;
4532 	if (!test_and_set_bit(__I40E_VSI_DOWN, vsi->state))
4533 		i40e_down(vsi);
4534 	i40e_vsi_free_irq(vsi);
4535 	i40e_vsi_free_tx_resources(vsi);
4536 	i40e_vsi_free_rx_resources(vsi);
4537 	vsi->current_netdev_flags = 0;
4538 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
4539 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
4540 		pf->flags |=  I40E_FLAG_CLIENT_RESET;
4541 }
4542 
4543 /**
4544  * i40e_quiesce_vsi - Pause a given VSI
4545  * @vsi: the VSI being paused
4546  **/
4547 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4548 {
4549 	if (test_bit(__I40E_VSI_DOWN, vsi->state))
4550 		return;
4551 
4552 	set_bit(__I40E_VSI_NEEDS_RESTART, vsi->state);
4553 	if (vsi->netdev && netif_running(vsi->netdev))
4554 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4555 	else
4556 		i40e_vsi_close(vsi);
4557 }
4558 
4559 /**
4560  * i40e_unquiesce_vsi - Resume a given VSI
4561  * @vsi: the VSI being resumed
4562  **/
4563 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4564 {
4565 	if (!test_and_clear_bit(__I40E_VSI_NEEDS_RESTART, vsi->state))
4566 		return;
4567 
4568 	if (vsi->netdev && netif_running(vsi->netdev))
4569 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4570 	else
4571 		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4572 }
4573 
4574 /**
4575  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4576  * @pf: the PF
4577  **/
4578 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4579 {
4580 	int v;
4581 
4582 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4583 		if (pf->vsi[v])
4584 			i40e_quiesce_vsi(pf->vsi[v]);
4585 	}
4586 }
4587 
4588 /**
4589  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4590  * @pf: the PF
4591  **/
4592 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4593 {
4594 	int v;
4595 
4596 	for (v = 0; v < pf->num_alloc_vsi; v++) {
4597 		if (pf->vsi[v])
4598 			i40e_unquiesce_vsi(pf->vsi[v]);
4599 	}
4600 }
4601 
4602 /**
4603  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4604  * @vsi: the VSI being configured
4605  *
4606  * Wait until all queues on a given VSI have been disabled.
4607  **/
4608 int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4609 {
4610 	struct i40e_pf *pf = vsi->back;
4611 	int i, pf_q, ret;
4612 
4613 	pf_q = vsi->base_queue;
4614 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4615 		/* Check and wait for the Tx queue */
4616 		ret = i40e_pf_txq_wait(pf, pf_q, false);
4617 		if (ret) {
4618 			dev_info(&pf->pdev->dev,
4619 				 "VSI seid %d Tx ring %d disable timeout\n",
4620 				 vsi->seid, pf_q);
4621 			return ret;
4622 		}
4623 
4624 		if (!i40e_enabled_xdp_vsi(vsi))
4625 			goto wait_rx;
4626 
4627 		/* Check and wait for the XDP Tx queue */
4628 		ret = i40e_pf_txq_wait(pf, pf_q + vsi->alloc_queue_pairs,
4629 				       false);
4630 		if (ret) {
4631 			dev_info(&pf->pdev->dev,
4632 				 "VSI seid %d XDP Tx ring %d disable timeout\n",
4633 				 vsi->seid, pf_q);
4634 			return ret;
4635 		}
4636 wait_rx:
4637 		/* Check and wait for the Rx queue */
4638 		ret = i40e_pf_rxq_wait(pf, pf_q, false);
4639 		if (ret) {
4640 			dev_info(&pf->pdev->dev,
4641 				 "VSI seid %d Rx ring %d disable timeout\n",
4642 				 vsi->seid, pf_q);
4643 			return ret;
4644 		}
4645 	}
4646 
4647 	return 0;
4648 }
4649 
4650 #ifdef CONFIG_I40E_DCB
4651 /**
4652  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4653  * @pf: the PF
4654  *
4655  * This function waits for the queues to be in disabled state for all the
4656  * VSIs that are managed by this PF.
4657  **/
4658 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4659 {
4660 	int v, ret = 0;
4661 
4662 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4663 		if (pf->vsi[v]) {
4664 			ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4665 			if (ret)
4666 				break;
4667 		}
4668 	}
4669 
4670 	return ret;
4671 }
4672 
4673 #endif
4674 
4675 /**
4676  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4677  * @q_idx: TX queue number
4678  * @vsi: Pointer to VSI struct
4679  *
4680  * This function checks specified queue for given VSI. Detects hung condition.
4681  * We proactively detect hung TX queues by checking if interrupts are disabled
4682  * but there are pending descriptors.  If it appears hung, attempt to recover
4683  * by triggering a SW interrupt.
4684  **/
4685 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4686 {
4687 	struct i40e_ring *tx_ring = NULL;
4688 	struct i40e_pf	*pf;
4689 	u32 val, tx_pending;
4690 	int i;
4691 
4692 	pf = vsi->back;
4693 
4694 	/* now that we have an index, find the tx_ring struct */
4695 	for (i = 0; i < vsi->num_queue_pairs; i++) {
4696 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4697 			if (q_idx == vsi->tx_rings[i]->queue_index) {
4698 				tx_ring = vsi->tx_rings[i];
4699 				break;
4700 			}
4701 		}
4702 	}
4703 
4704 	if (!tx_ring)
4705 		return;
4706 
4707 	/* Read interrupt register */
4708 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4709 		val = rd32(&pf->hw,
4710 			   I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4711 					       tx_ring->vsi->base_vector - 1));
4712 	else
4713 		val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4714 
4715 	tx_pending = i40e_get_tx_pending(tx_ring);
4716 
4717 	/* Interrupts are disabled and TX pending is non-zero,
4718 	 * trigger the SW interrupt (don't wait). Worst case
4719 	 * there will be one extra interrupt which may result
4720 	 * into not cleaning any queues because queues are cleaned.
4721 	 */
4722 	if (tx_pending && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)))
4723 		i40e_force_wb(vsi, tx_ring->q_vector);
4724 }
4725 
4726 /**
4727  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4728  * @pf:  pointer to PF struct
4729  *
4730  * LAN VSI has netdev and netdev has TX queues. This function is to check
4731  * each of those TX queues if they are hung, trigger recovery by issuing
4732  * SW interrupt.
4733  **/
4734 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4735 {
4736 	struct net_device *netdev;
4737 	struct i40e_vsi *vsi;
4738 	unsigned int i;
4739 
4740 	/* Only for LAN VSI */
4741 	vsi = pf->vsi[pf->lan_vsi];
4742 
4743 	if (!vsi)
4744 		return;
4745 
4746 	/* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4747 	if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
4748 	    test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
4749 		return;
4750 
4751 	/* Make sure type is MAIN VSI */
4752 	if (vsi->type != I40E_VSI_MAIN)
4753 		return;
4754 
4755 	netdev = vsi->netdev;
4756 	if (!netdev)
4757 		return;
4758 
4759 	/* Bail out if netif_carrier is not OK */
4760 	if (!netif_carrier_ok(netdev))
4761 		return;
4762 
4763 	/* Go thru' TX queues for netdev */
4764 	for (i = 0; i < netdev->num_tx_queues; i++) {
4765 		struct netdev_queue *q;
4766 
4767 		q = netdev_get_tx_queue(netdev, i);
4768 		if (q)
4769 			i40e_detect_recover_hung_queue(i, vsi);
4770 	}
4771 }
4772 
4773 /**
4774  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4775  * @pf: pointer to PF
4776  *
4777  * Get TC map for ISCSI PF type that will include iSCSI TC
4778  * and LAN TC.
4779  **/
4780 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4781 {
4782 	struct i40e_dcb_app_priority_table app;
4783 	struct i40e_hw *hw = &pf->hw;
4784 	u8 enabled_tc = 1; /* TC0 is always enabled */
4785 	u8 tc, i;
4786 	/* Get the iSCSI APP TLV */
4787 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4788 
4789 	for (i = 0; i < dcbcfg->numapps; i++) {
4790 		app = dcbcfg->app[i];
4791 		if (app.selector == I40E_APP_SEL_TCPIP &&
4792 		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4793 			tc = dcbcfg->etscfg.prioritytable[app.priority];
4794 			enabled_tc |= BIT(tc);
4795 			break;
4796 		}
4797 	}
4798 
4799 	return enabled_tc;
4800 }
4801 
4802 /**
4803  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4804  * @dcbcfg: the corresponding DCBx configuration structure
4805  *
4806  * Return the number of TCs from given DCBx configuration
4807  **/
4808 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4809 {
4810 	int i, tc_unused = 0;
4811 	u8 num_tc = 0;
4812 	u8 ret = 0;
4813 
4814 	/* Scan the ETS Config Priority Table to find
4815 	 * traffic class enabled for a given priority
4816 	 * and create a bitmask of enabled TCs
4817 	 */
4818 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4819 		num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4820 
4821 	/* Now scan the bitmask to check for
4822 	 * contiguous TCs starting with TC0
4823 	 */
4824 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4825 		if (num_tc & BIT(i)) {
4826 			if (!tc_unused) {
4827 				ret++;
4828 			} else {
4829 				pr_err("Non-contiguous TC - Disabling DCB\n");
4830 				return 1;
4831 			}
4832 		} else {
4833 			tc_unused = 1;
4834 		}
4835 	}
4836 
4837 	/* There is always at least TC0 */
4838 	if (!ret)
4839 		ret = 1;
4840 
4841 	return ret;
4842 }
4843 
4844 /**
4845  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4846  * @dcbcfg: the corresponding DCBx configuration structure
4847  *
4848  * Query the current DCB configuration and return the number of
4849  * traffic classes enabled from the given DCBX config
4850  **/
4851 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4852 {
4853 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4854 	u8 enabled_tc = 1;
4855 	u8 i;
4856 
4857 	for (i = 0; i < num_tc; i++)
4858 		enabled_tc |= BIT(i);
4859 
4860 	return enabled_tc;
4861 }
4862 
4863 /**
4864  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4865  * @pf: PF being queried
4866  *
4867  * Return number of traffic classes enabled for the given PF
4868  **/
4869 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4870 {
4871 	struct i40e_hw *hw = &pf->hw;
4872 	u8 i, enabled_tc = 1;
4873 	u8 num_tc = 0;
4874 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4875 
4876 	/* If DCB is not enabled then always in single TC */
4877 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4878 		return 1;
4879 
4880 	/* SFP mode will be enabled for all TCs on port */
4881 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4882 		return i40e_dcb_get_num_tc(dcbcfg);
4883 
4884 	/* MFP mode return count of enabled TCs for this PF */
4885 	if (pf->hw.func_caps.iscsi)
4886 		enabled_tc =  i40e_get_iscsi_tc_map(pf);
4887 	else
4888 		return 1; /* Only TC0 */
4889 
4890 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4891 		if (enabled_tc & BIT(i))
4892 			num_tc++;
4893 	}
4894 	return num_tc;
4895 }
4896 
4897 /**
4898  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4899  * @pf: PF being queried
4900  *
4901  * Return a bitmap for enabled traffic classes for this PF.
4902  **/
4903 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4904 {
4905 	/* If DCB is not enabled for this PF then just return default TC */
4906 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4907 		return I40E_DEFAULT_TRAFFIC_CLASS;
4908 
4909 	/* SFP mode we want PF to be enabled for all TCs */
4910 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4911 		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4912 
4913 	/* MFP enabled and iSCSI PF type */
4914 	if (pf->hw.func_caps.iscsi)
4915 		return i40e_get_iscsi_tc_map(pf);
4916 	else
4917 		return I40E_DEFAULT_TRAFFIC_CLASS;
4918 }
4919 
4920 /**
4921  * i40e_vsi_get_bw_info - Query VSI BW Information
4922  * @vsi: the VSI being queried
4923  *
4924  * Returns 0 on success, negative value on failure
4925  **/
4926 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4927 {
4928 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4929 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4930 	struct i40e_pf *pf = vsi->back;
4931 	struct i40e_hw *hw = &pf->hw;
4932 	i40e_status ret;
4933 	u32 tc_bw_max;
4934 	int i;
4935 
4936 	/* Get the VSI level BW configuration */
4937 	ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4938 	if (ret) {
4939 		dev_info(&pf->pdev->dev,
4940 			 "couldn't get PF vsi bw config, err %s aq_err %s\n",
4941 			 i40e_stat_str(&pf->hw, ret),
4942 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4943 		return -EINVAL;
4944 	}
4945 
4946 	/* Get the VSI level BW configuration per TC */
4947 	ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4948 					       NULL);
4949 	if (ret) {
4950 		dev_info(&pf->pdev->dev,
4951 			 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4952 			 i40e_stat_str(&pf->hw, ret),
4953 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4954 		return -EINVAL;
4955 	}
4956 
4957 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4958 		dev_info(&pf->pdev->dev,
4959 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4960 			 bw_config.tc_valid_bits,
4961 			 bw_ets_config.tc_valid_bits);
4962 		/* Still continuing */
4963 	}
4964 
4965 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4966 	vsi->bw_max_quanta = bw_config.max_bw;
4967 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4968 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4969 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4970 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4971 		vsi->bw_ets_limit_credits[i] =
4972 					le16_to_cpu(bw_ets_config.credits[i]);
4973 		/* 3 bits out of 4 for each TC */
4974 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4975 	}
4976 
4977 	return 0;
4978 }
4979 
4980 /**
4981  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4982  * @vsi: the VSI being configured
4983  * @enabled_tc: TC bitmap
4984  * @bw_credits: BW shared credits per TC
4985  *
4986  * Returns 0 on success, negative value on failure
4987  **/
4988 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4989 				       u8 *bw_share)
4990 {
4991 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4992 	i40e_status ret;
4993 	int i;
4994 
4995 	bw_data.tc_valid_bits = enabled_tc;
4996 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4997 		bw_data.tc_bw_credits[i] = bw_share[i];
4998 
4999 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
5000 				       NULL);
5001 	if (ret) {
5002 		dev_info(&vsi->back->pdev->dev,
5003 			 "AQ command Config VSI BW allocation per TC failed = %d\n",
5004 			 vsi->back->hw.aq.asq_last_status);
5005 		return -EINVAL;
5006 	}
5007 
5008 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5009 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
5010 
5011 	return 0;
5012 }
5013 
5014 /**
5015  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
5016  * @vsi: the VSI being configured
5017  * @enabled_tc: TC map to be enabled
5018  *
5019  **/
5020 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5021 {
5022 	struct net_device *netdev = vsi->netdev;
5023 	struct i40e_pf *pf = vsi->back;
5024 	struct i40e_hw *hw = &pf->hw;
5025 	u8 netdev_tc = 0;
5026 	int i;
5027 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5028 
5029 	if (!netdev)
5030 		return;
5031 
5032 	if (!enabled_tc) {
5033 		netdev_reset_tc(netdev);
5034 		return;
5035 	}
5036 
5037 	/* Set up actual enabled TCs on the VSI */
5038 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
5039 		return;
5040 
5041 	/* set per TC queues for the VSI */
5042 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5043 		/* Only set TC queues for enabled tcs
5044 		 *
5045 		 * e.g. For a VSI that has TC0 and TC3 enabled the
5046 		 * enabled_tc bitmap would be 0x00001001; the driver
5047 		 * will set the numtc for netdev as 2 that will be
5048 		 * referenced by the netdev layer as TC 0 and 1.
5049 		 */
5050 		if (vsi->tc_config.enabled_tc & BIT(i))
5051 			netdev_set_tc_queue(netdev,
5052 					vsi->tc_config.tc_info[i].netdev_tc,
5053 					vsi->tc_config.tc_info[i].qcount,
5054 					vsi->tc_config.tc_info[i].qoffset);
5055 	}
5056 
5057 	/* Assign UP2TC map for the VSI */
5058 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
5059 		/* Get the actual TC# for the UP */
5060 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
5061 		/* Get the mapped netdev TC# for the UP */
5062 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
5063 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
5064 	}
5065 }
5066 
5067 /**
5068  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
5069  * @vsi: the VSI being configured
5070  * @ctxt: the ctxt buffer returned from AQ VSI update param command
5071  **/
5072 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
5073 				      struct i40e_vsi_context *ctxt)
5074 {
5075 	/* copy just the sections touched not the entire info
5076 	 * since not all sections are valid as returned by
5077 	 * update vsi params
5078 	 */
5079 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
5080 	memcpy(&vsi->info.queue_mapping,
5081 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
5082 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
5083 	       sizeof(vsi->info.tc_mapping));
5084 }
5085 
5086 /**
5087  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
5088  * @vsi: VSI to be configured
5089  * @enabled_tc: TC bitmap
5090  *
5091  * This configures a particular VSI for TCs that are mapped to the
5092  * given TC bitmap. It uses default bandwidth share for TCs across
5093  * VSIs to configure TC for a particular VSI.
5094  *
5095  * NOTE:
5096  * It is expected that the VSI queues have been quisced before calling
5097  * this function.
5098  **/
5099 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5100 {
5101 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5102 	struct i40e_vsi_context ctxt;
5103 	int ret = 0;
5104 	int i;
5105 
5106 	/* Check if enabled_tc is same as existing or new TCs */
5107 	if (vsi->tc_config.enabled_tc == enabled_tc)
5108 		return ret;
5109 
5110 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
5111 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5112 		if (enabled_tc & BIT(i))
5113 			bw_share[i] = 1;
5114 	}
5115 
5116 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5117 	if (ret) {
5118 		dev_info(&vsi->back->pdev->dev,
5119 			 "Failed configuring TC map %d for VSI %d\n",
5120 			 enabled_tc, vsi->seid);
5121 		goto out;
5122 	}
5123 
5124 	/* Update Queue Pairs Mapping for currently enabled UPs */
5125 	ctxt.seid = vsi->seid;
5126 	ctxt.pf_num = vsi->back->hw.pf_id;
5127 	ctxt.vf_num = 0;
5128 	ctxt.uplink_seid = vsi->uplink_seid;
5129 	ctxt.info = vsi->info;
5130 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5131 
5132 	if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5133 		ctxt.info.valid_sections |=
5134 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5135 		ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5136 	}
5137 
5138 	/* Update the VSI after updating the VSI queue-mapping information */
5139 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5140 	if (ret) {
5141 		dev_info(&vsi->back->pdev->dev,
5142 			 "Update vsi tc config failed, err %s aq_err %s\n",
5143 			 i40e_stat_str(&vsi->back->hw, ret),
5144 			 i40e_aq_str(&vsi->back->hw,
5145 				     vsi->back->hw.aq.asq_last_status));
5146 		goto out;
5147 	}
5148 	/* update the local VSI info with updated queue map */
5149 	i40e_vsi_update_queue_map(vsi, &ctxt);
5150 	vsi->info.valid_sections = 0;
5151 
5152 	/* Update current VSI BW information */
5153 	ret = i40e_vsi_get_bw_info(vsi);
5154 	if (ret) {
5155 		dev_info(&vsi->back->pdev->dev,
5156 			 "Failed updating vsi bw info, err %s aq_err %s\n",
5157 			 i40e_stat_str(&vsi->back->hw, ret),
5158 			 i40e_aq_str(&vsi->back->hw,
5159 				     vsi->back->hw.aq.asq_last_status));
5160 		goto out;
5161 	}
5162 
5163 	/* Update the netdev TC setup */
5164 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5165 out:
5166 	return ret;
5167 }
5168 
5169 /**
5170  * i40e_veb_config_tc - Configure TCs for given VEB
5171  * @veb: given VEB
5172  * @enabled_tc: TC bitmap
5173  *
5174  * Configures given TC bitmap for VEB (switching) element
5175  **/
5176 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
5177 {
5178 	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
5179 	struct i40e_pf *pf = veb->pf;
5180 	int ret = 0;
5181 	int i;
5182 
5183 	/* No TCs or already enabled TCs just return */
5184 	if (!enabled_tc || veb->enabled_tc == enabled_tc)
5185 		return ret;
5186 
5187 	bw_data.tc_valid_bits = enabled_tc;
5188 	/* bw_data.absolute_credits is not set (relative) */
5189 
5190 	/* Enable ETS TCs with equal BW Share for now */
5191 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5192 		if (enabled_tc & BIT(i))
5193 			bw_data.tc_bw_share_credits[i] = 1;
5194 	}
5195 
5196 	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
5197 						   &bw_data, NULL);
5198 	if (ret) {
5199 		dev_info(&pf->pdev->dev,
5200 			 "VEB bw config failed, err %s aq_err %s\n",
5201 			 i40e_stat_str(&pf->hw, ret),
5202 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5203 		goto out;
5204 	}
5205 
5206 	/* Update the BW information */
5207 	ret = i40e_veb_get_bw_info(veb);
5208 	if (ret) {
5209 		dev_info(&pf->pdev->dev,
5210 			 "Failed getting veb bw config, err %s aq_err %s\n",
5211 			 i40e_stat_str(&pf->hw, ret),
5212 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5213 	}
5214 
5215 out:
5216 	return ret;
5217 }
5218 
5219 #ifdef CONFIG_I40E_DCB
5220 /**
5221  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
5222  * @pf: PF struct
5223  *
5224  * Reconfigure VEB/VSIs on a given PF; it is assumed that
5225  * the caller would've quiesce all the VSIs before calling
5226  * this function
5227  **/
5228 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
5229 {
5230 	u8 tc_map = 0;
5231 	int ret;
5232 	u8 v;
5233 
5234 	/* Enable the TCs available on PF to all VEBs */
5235 	tc_map = i40e_pf_get_tc_map(pf);
5236 	for (v = 0; v < I40E_MAX_VEB; v++) {
5237 		if (!pf->veb[v])
5238 			continue;
5239 		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
5240 		if (ret) {
5241 			dev_info(&pf->pdev->dev,
5242 				 "Failed configuring TC for VEB seid=%d\n",
5243 				 pf->veb[v]->seid);
5244 			/* Will try to configure as many components */
5245 		}
5246 	}
5247 
5248 	/* Update each VSI */
5249 	for (v = 0; v < pf->num_alloc_vsi; v++) {
5250 		if (!pf->vsi[v])
5251 			continue;
5252 
5253 		/* - Enable all TCs for the LAN VSI
5254 		 * - For all others keep them at TC0 for now
5255 		 */
5256 		if (v == pf->lan_vsi)
5257 			tc_map = i40e_pf_get_tc_map(pf);
5258 		else
5259 			tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
5260 
5261 		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5262 		if (ret) {
5263 			dev_info(&pf->pdev->dev,
5264 				 "Failed configuring TC for VSI seid=%d\n",
5265 				 pf->vsi[v]->seid);
5266 			/* Will try to configure as many components */
5267 		} else {
5268 			/* Re-configure VSI vectors based on updated TC map */
5269 			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5270 			if (pf->vsi[v]->netdev)
5271 				i40e_dcbnl_set_all(pf->vsi[v]);
5272 		}
5273 	}
5274 }
5275 
5276 /**
5277  * i40e_resume_port_tx - Resume port Tx
5278  * @pf: PF struct
5279  *
5280  * Resume a port's Tx and issue a PF reset in case of failure to
5281  * resume.
5282  **/
5283 static int i40e_resume_port_tx(struct i40e_pf *pf)
5284 {
5285 	struct i40e_hw *hw = &pf->hw;
5286 	int ret;
5287 
5288 	ret = i40e_aq_resume_port_tx(hw, NULL);
5289 	if (ret) {
5290 		dev_info(&pf->pdev->dev,
5291 			 "Resume Port Tx failed, err %s aq_err %s\n",
5292 			  i40e_stat_str(&pf->hw, ret),
5293 			  i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5294 		/* Schedule PF reset to recover */
5295 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
5296 		i40e_service_event_schedule(pf);
5297 	}
5298 
5299 	return ret;
5300 }
5301 
5302 /**
5303  * i40e_init_pf_dcb - Initialize DCB configuration
5304  * @pf: PF being configured
5305  *
5306  * Query the current DCB configuration and cache it
5307  * in the hardware structure
5308  **/
5309 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5310 {
5311 	struct i40e_hw *hw = &pf->hw;
5312 	int err = 0;
5313 
5314 	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5315 	if (pf->hw_features & I40E_HW_NO_DCB_SUPPORT)
5316 		goto out;
5317 
5318 	/* Get the initial DCB configuration */
5319 	err = i40e_init_dcb(hw);
5320 	if (!err) {
5321 		/* Device/Function is not DCBX capable */
5322 		if ((!hw->func_caps.dcb) ||
5323 		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5324 			dev_info(&pf->pdev->dev,
5325 				 "DCBX offload is not supported or is disabled for this PF.\n");
5326 		} else {
5327 			/* When status is not DISABLED then DCBX in FW */
5328 			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5329 				       DCB_CAP_DCBX_VER_IEEE;
5330 
5331 			pf->flags |= I40E_FLAG_DCB_CAPABLE;
5332 			/* Enable DCB tagging only when more than one TC
5333 			 * or explicitly disable if only one TC
5334 			 */
5335 			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5336 				pf->flags |= I40E_FLAG_DCB_ENABLED;
5337 			else
5338 				pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5339 			dev_dbg(&pf->pdev->dev,
5340 				"DCBX offload is supported for this PF.\n");
5341 		}
5342 	} else {
5343 		dev_info(&pf->pdev->dev,
5344 			 "Query for DCB configuration failed, err %s aq_err %s\n",
5345 			 i40e_stat_str(&pf->hw, err),
5346 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5347 	}
5348 
5349 out:
5350 	return err;
5351 }
5352 #endif /* CONFIG_I40E_DCB */
5353 #define SPEED_SIZE 14
5354 #define FC_SIZE 8
5355 /**
5356  * i40e_print_link_message - print link up or down
5357  * @vsi: the VSI for which link needs a message
5358  */
5359 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5360 {
5361 	enum i40e_aq_link_speed new_speed;
5362 	struct i40e_pf *pf = vsi->back;
5363 	char *speed = "Unknown";
5364 	char *fc = "Unknown";
5365 	char *fec = "";
5366 	char *req_fec = "";
5367 	char *an = "";
5368 
5369 	new_speed = pf->hw.phy.link_info.link_speed;
5370 
5371 	if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
5372 		return;
5373 	vsi->current_isup = isup;
5374 	vsi->current_speed = new_speed;
5375 	if (!isup) {
5376 		netdev_info(vsi->netdev, "NIC Link is Down\n");
5377 		return;
5378 	}
5379 
5380 	/* Warn user if link speed on NPAR enabled partition is not at
5381 	 * least 10GB
5382 	 */
5383 	if (pf->hw.func_caps.npar_enable &&
5384 	    (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5385 	     pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5386 		netdev_warn(vsi->netdev,
5387 			    "The partition detected link speed that is less than 10Gbps\n");
5388 
5389 	switch (pf->hw.phy.link_info.link_speed) {
5390 	case I40E_LINK_SPEED_40GB:
5391 		speed = "40 G";
5392 		break;
5393 	case I40E_LINK_SPEED_20GB:
5394 		speed = "20 G";
5395 		break;
5396 	case I40E_LINK_SPEED_25GB:
5397 		speed = "25 G";
5398 		break;
5399 	case I40E_LINK_SPEED_10GB:
5400 		speed = "10 G";
5401 		break;
5402 	case I40E_LINK_SPEED_1GB:
5403 		speed = "1000 M";
5404 		break;
5405 	case I40E_LINK_SPEED_100MB:
5406 		speed = "100 M";
5407 		break;
5408 	default:
5409 		break;
5410 	}
5411 
5412 	switch (pf->hw.fc.current_mode) {
5413 	case I40E_FC_FULL:
5414 		fc = "RX/TX";
5415 		break;
5416 	case I40E_FC_TX_PAUSE:
5417 		fc = "TX";
5418 		break;
5419 	case I40E_FC_RX_PAUSE:
5420 		fc = "RX";
5421 		break;
5422 	default:
5423 		fc = "None";
5424 		break;
5425 	}
5426 
5427 	if (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
5428 		req_fec = ", Requested FEC: None";
5429 		fec = ", FEC: None";
5430 		an = ", Autoneg: False";
5431 
5432 		if (pf->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
5433 			an = ", Autoneg: True";
5434 
5435 		if (pf->hw.phy.link_info.fec_info &
5436 		    I40E_AQ_CONFIG_FEC_KR_ENA)
5437 			fec = ", FEC: CL74 FC-FEC/BASE-R";
5438 		else if (pf->hw.phy.link_info.fec_info &
5439 			 I40E_AQ_CONFIG_FEC_RS_ENA)
5440 			fec = ", FEC: CL108 RS-FEC";
5441 
5442 		/* 'CL108 RS-FEC' should be displayed when RS is requested, or
5443 		 * both RS and FC are requested
5444 		 */
5445 		if (vsi->back->hw.phy.link_info.req_fec_info &
5446 		    (I40E_AQ_REQUEST_FEC_KR | I40E_AQ_REQUEST_FEC_RS)) {
5447 			if (vsi->back->hw.phy.link_info.req_fec_info &
5448 			    I40E_AQ_REQUEST_FEC_RS)
5449 				req_fec = ", Requested FEC: CL108 RS-FEC";
5450 			else
5451 				req_fec = ", Requested FEC: CL74 FC-FEC/BASE-R";
5452 		}
5453 	}
5454 
5455 	netdev_info(vsi->netdev, "NIC Link is Up, %sbps Full Duplex%s%s%s, Flow Control: %s\n",
5456 		    speed, req_fec, fec, an, fc);
5457 }
5458 
5459 /**
5460  * i40e_up_complete - Finish the last steps of bringing up a connection
5461  * @vsi: the VSI being configured
5462  **/
5463 static int i40e_up_complete(struct i40e_vsi *vsi)
5464 {
5465 	struct i40e_pf *pf = vsi->back;
5466 	int err;
5467 
5468 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5469 		i40e_vsi_configure_msix(vsi);
5470 	else
5471 		i40e_configure_msi_and_legacy(vsi);
5472 
5473 	/* start rings */
5474 	err = i40e_vsi_start_rings(vsi);
5475 	if (err)
5476 		return err;
5477 
5478 	clear_bit(__I40E_VSI_DOWN, vsi->state);
5479 	i40e_napi_enable_all(vsi);
5480 	i40e_vsi_enable_irq(vsi);
5481 
5482 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5483 	    (vsi->netdev)) {
5484 		i40e_print_link_message(vsi, true);
5485 		netif_tx_start_all_queues(vsi->netdev);
5486 		netif_carrier_on(vsi->netdev);
5487 	}
5488 
5489 	/* replay FDIR SB filters */
5490 	if (vsi->type == I40E_VSI_FDIR) {
5491 		/* reset fd counters */
5492 		pf->fd_add_err = 0;
5493 		pf->fd_atr_cnt = 0;
5494 		i40e_fdir_filter_restore(vsi);
5495 	}
5496 
5497 	/* On the next run of the service_task, notify any clients of the new
5498 	 * opened netdev
5499 	 */
5500 	pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5501 	i40e_service_event_schedule(pf);
5502 
5503 	return 0;
5504 }
5505 
5506 /**
5507  * i40e_vsi_reinit_locked - Reset the VSI
5508  * @vsi: the VSI being configured
5509  *
5510  * Rebuild the ring structs after some configuration
5511  * has changed, e.g. MTU size.
5512  **/
5513 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5514 {
5515 	struct i40e_pf *pf = vsi->back;
5516 
5517 	WARN_ON(in_interrupt());
5518 	while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state))
5519 		usleep_range(1000, 2000);
5520 	i40e_down(vsi);
5521 
5522 	i40e_up(vsi);
5523 	clear_bit(__I40E_CONFIG_BUSY, pf->state);
5524 }
5525 
5526 /**
5527  * i40e_up - Bring the connection back up after being down
5528  * @vsi: the VSI being configured
5529  **/
5530 int i40e_up(struct i40e_vsi *vsi)
5531 {
5532 	int err;
5533 
5534 	err = i40e_vsi_configure(vsi);
5535 	if (!err)
5536 		err = i40e_up_complete(vsi);
5537 
5538 	return err;
5539 }
5540 
5541 /**
5542  * i40e_down - Shutdown the connection processing
5543  * @vsi: the VSI being stopped
5544  **/
5545 void i40e_down(struct i40e_vsi *vsi)
5546 {
5547 	int i;
5548 
5549 	/* It is assumed that the caller of this function
5550 	 * sets the vsi->state __I40E_VSI_DOWN bit.
5551 	 */
5552 	if (vsi->netdev) {
5553 		netif_carrier_off(vsi->netdev);
5554 		netif_tx_disable(vsi->netdev);
5555 	}
5556 	i40e_vsi_disable_irq(vsi);
5557 	i40e_vsi_stop_rings(vsi);
5558 	i40e_napi_disable_all(vsi);
5559 
5560 	for (i = 0; i < vsi->num_queue_pairs; i++) {
5561 		i40e_clean_tx_ring(vsi->tx_rings[i]);
5562 		if (i40e_enabled_xdp_vsi(vsi))
5563 			i40e_clean_tx_ring(vsi->xdp_rings[i]);
5564 		i40e_clean_rx_ring(vsi->rx_rings[i]);
5565 	}
5566 
5567 }
5568 
5569 /**
5570  * i40e_setup_tc - configure multiple traffic classes
5571  * @netdev: net device to configure
5572  * @tc: number of traffic classes to enable
5573  **/
5574 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5575 {
5576 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5577 	struct i40e_vsi *vsi = np->vsi;
5578 	struct i40e_pf *pf = vsi->back;
5579 	u8 enabled_tc = 0;
5580 	int ret = -EINVAL;
5581 	int i;
5582 
5583 	/* Check if DCB enabled to continue */
5584 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5585 		netdev_info(netdev, "DCB is not enabled for adapter\n");
5586 		goto exit;
5587 	}
5588 
5589 	/* Check if MFP enabled */
5590 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5591 		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5592 		goto exit;
5593 	}
5594 
5595 	/* Check whether tc count is within enabled limit */
5596 	if (tc > i40e_pf_get_num_tc(pf)) {
5597 		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5598 		goto exit;
5599 	}
5600 
5601 	/* Generate TC map for number of tc requested */
5602 	for (i = 0; i < tc; i++)
5603 		enabled_tc |= BIT(i);
5604 
5605 	/* Requesting same TC configuration as already enabled */
5606 	if (enabled_tc == vsi->tc_config.enabled_tc)
5607 		return 0;
5608 
5609 	/* Quiesce VSI queues */
5610 	i40e_quiesce_vsi(vsi);
5611 
5612 	/* Configure VSI for enabled TCs */
5613 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
5614 	if (ret) {
5615 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5616 			    vsi->seid);
5617 		goto exit;
5618 	}
5619 
5620 	/* Unquiesce VSI */
5621 	i40e_unquiesce_vsi(vsi);
5622 
5623 exit:
5624 	return ret;
5625 }
5626 
5627 static int __i40e_setup_tc(struct net_device *netdev, enum tc_setup_type type,
5628 			   void *type_data)
5629 {
5630 	struct tc_mqprio_qopt *mqprio = type_data;
5631 
5632 	if (type != TC_SETUP_MQPRIO)
5633 		return -EOPNOTSUPP;
5634 
5635 	mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
5636 
5637 	return i40e_setup_tc(netdev, mqprio->num_tc);
5638 }
5639 
5640 /**
5641  * i40e_open - Called when a network interface is made active
5642  * @netdev: network interface device structure
5643  *
5644  * The open entry point is called when a network interface is made
5645  * active by the system (IFF_UP).  At this point all resources needed
5646  * for transmit and receive operations are allocated, the interrupt
5647  * handler is registered with the OS, the netdev watchdog subtask is
5648  * enabled, and the stack is notified that the interface is ready.
5649  *
5650  * Returns 0 on success, negative value on failure
5651  **/
5652 int i40e_open(struct net_device *netdev)
5653 {
5654 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5655 	struct i40e_vsi *vsi = np->vsi;
5656 	struct i40e_pf *pf = vsi->back;
5657 	int err;
5658 
5659 	/* disallow open during test or if eeprom is broken */
5660 	if (test_bit(__I40E_TESTING, pf->state) ||
5661 	    test_bit(__I40E_BAD_EEPROM, pf->state))
5662 		return -EBUSY;
5663 
5664 	netif_carrier_off(netdev);
5665 
5666 	err = i40e_vsi_open(vsi);
5667 	if (err)
5668 		return err;
5669 
5670 	/* configure global TSO hardware offload settings */
5671 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5672 						       TCP_FLAG_FIN) >> 16);
5673 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5674 						       TCP_FLAG_FIN |
5675 						       TCP_FLAG_CWR) >> 16);
5676 	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5677 
5678 	udp_tunnel_get_rx_info(netdev);
5679 
5680 	return 0;
5681 }
5682 
5683 /**
5684  * i40e_vsi_open -
5685  * @vsi: the VSI to open
5686  *
5687  * Finish initialization of the VSI.
5688  *
5689  * Returns 0 on success, negative value on failure
5690  *
5691  * Note: expects to be called while under rtnl_lock()
5692  **/
5693 int i40e_vsi_open(struct i40e_vsi *vsi)
5694 {
5695 	struct i40e_pf *pf = vsi->back;
5696 	char int_name[I40E_INT_NAME_STR_LEN];
5697 	int err;
5698 
5699 	/* allocate descriptors */
5700 	err = i40e_vsi_setup_tx_resources(vsi);
5701 	if (err)
5702 		goto err_setup_tx;
5703 	err = i40e_vsi_setup_rx_resources(vsi);
5704 	if (err)
5705 		goto err_setup_rx;
5706 
5707 	err = i40e_vsi_configure(vsi);
5708 	if (err)
5709 		goto err_setup_rx;
5710 
5711 	if (vsi->netdev) {
5712 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5713 			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5714 		err = i40e_vsi_request_irq(vsi, int_name);
5715 		if (err)
5716 			goto err_setup_rx;
5717 
5718 		/* Notify the stack of the actual queue counts. */
5719 		err = netif_set_real_num_tx_queues(vsi->netdev,
5720 						   vsi->num_queue_pairs);
5721 		if (err)
5722 			goto err_set_queues;
5723 
5724 		err = netif_set_real_num_rx_queues(vsi->netdev,
5725 						   vsi->num_queue_pairs);
5726 		if (err)
5727 			goto err_set_queues;
5728 
5729 	} else if (vsi->type == I40E_VSI_FDIR) {
5730 		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5731 			 dev_driver_string(&pf->pdev->dev),
5732 			 dev_name(&pf->pdev->dev));
5733 		err = i40e_vsi_request_irq(vsi, int_name);
5734 
5735 	} else {
5736 		err = -EINVAL;
5737 		goto err_setup_rx;
5738 	}
5739 
5740 	err = i40e_up_complete(vsi);
5741 	if (err)
5742 		goto err_up_complete;
5743 
5744 	return 0;
5745 
5746 err_up_complete:
5747 	i40e_down(vsi);
5748 err_set_queues:
5749 	i40e_vsi_free_irq(vsi);
5750 err_setup_rx:
5751 	i40e_vsi_free_rx_resources(vsi);
5752 err_setup_tx:
5753 	i40e_vsi_free_tx_resources(vsi);
5754 	if (vsi == pf->vsi[pf->lan_vsi])
5755 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
5756 
5757 	return err;
5758 }
5759 
5760 /**
5761  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5762  * @pf: Pointer to PF
5763  *
5764  * This function destroys the hlist where all the Flow Director
5765  * filters were saved.
5766  **/
5767 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5768 {
5769 	struct i40e_fdir_filter *filter;
5770 	struct i40e_flex_pit *pit_entry, *tmp;
5771 	struct hlist_node *node2;
5772 
5773 	hlist_for_each_entry_safe(filter, node2,
5774 				  &pf->fdir_filter_list, fdir_node) {
5775 		hlist_del(&filter->fdir_node);
5776 		kfree(filter);
5777 	}
5778 
5779 	list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
5780 		list_del(&pit_entry->list);
5781 		kfree(pit_entry);
5782 	}
5783 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
5784 
5785 	list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
5786 		list_del(&pit_entry->list);
5787 		kfree(pit_entry);
5788 	}
5789 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
5790 
5791 	pf->fdir_pf_active_filters = 0;
5792 	pf->fd_tcp4_filter_cnt = 0;
5793 	pf->fd_udp4_filter_cnt = 0;
5794 	pf->fd_sctp4_filter_cnt = 0;
5795 	pf->fd_ip4_filter_cnt = 0;
5796 
5797 	/* Reprogram the default input set for TCP/IPv4 */
5798 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
5799 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5800 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5801 
5802 	/* Reprogram the default input set for UDP/IPv4 */
5803 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
5804 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5805 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5806 
5807 	/* Reprogram the default input set for SCTP/IPv4 */
5808 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
5809 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5810 				I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5811 
5812 	/* Reprogram the default input set for Other/IPv4 */
5813 	i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
5814 				I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
5815 }
5816 
5817 /**
5818  * i40e_close - Disables a network interface
5819  * @netdev: network interface device structure
5820  *
5821  * The close entry point is called when an interface is de-activated
5822  * by the OS.  The hardware is still under the driver's control, but
5823  * this netdev interface is disabled.
5824  *
5825  * Returns 0, this is not allowed to fail
5826  **/
5827 int i40e_close(struct net_device *netdev)
5828 {
5829 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5830 	struct i40e_vsi *vsi = np->vsi;
5831 
5832 	i40e_vsi_close(vsi);
5833 
5834 	return 0;
5835 }
5836 
5837 /**
5838  * i40e_do_reset - Start a PF or Core Reset sequence
5839  * @pf: board private structure
5840  * @reset_flags: which reset is requested
5841  * @lock_acquired: indicates whether or not the lock has been acquired
5842  * before this function was called.
5843  *
5844  * The essential difference in resets is that the PF Reset
5845  * doesn't clear the packet buffers, doesn't reset the PE
5846  * firmware, and doesn't bother the other PFs on the chip.
5847  **/
5848 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
5849 {
5850 	u32 val;
5851 
5852 	WARN_ON(in_interrupt());
5853 
5854 
5855 	/* do the biggest reset indicated */
5856 	if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5857 
5858 		/* Request a Global Reset
5859 		 *
5860 		 * This will start the chip's countdown to the actual full
5861 		 * chip reset event, and a warning interrupt to be sent
5862 		 * to all PFs, including the requestor.  Our handler
5863 		 * for the warning interrupt will deal with the shutdown
5864 		 * and recovery of the switch setup.
5865 		 */
5866 		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5867 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5868 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5869 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5870 
5871 	} else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5872 
5873 		/* Request a Core Reset
5874 		 *
5875 		 * Same as Global Reset, except does *not* include the MAC/PHY
5876 		 */
5877 		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5878 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5879 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
5880 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5881 		i40e_flush(&pf->hw);
5882 
5883 	} else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5884 
5885 		/* Request a PF Reset
5886 		 *
5887 		 * Resets only the PF-specific registers
5888 		 *
5889 		 * This goes directly to the tear-down and rebuild of
5890 		 * the switch, since we need to do all the recovery as
5891 		 * for the Core Reset.
5892 		 */
5893 		dev_dbg(&pf->pdev->dev, "PFR requested\n");
5894 		i40e_handle_reset_warning(pf, lock_acquired);
5895 
5896 	} else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5897 		int v;
5898 
5899 		/* Find the VSI(s) that requested a re-init */
5900 		dev_info(&pf->pdev->dev,
5901 			 "VSI reinit requested\n");
5902 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5903 			struct i40e_vsi *vsi = pf->vsi[v];
5904 
5905 			if (vsi != NULL &&
5906 			    test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
5907 					       vsi->state))
5908 				i40e_vsi_reinit_locked(pf->vsi[v]);
5909 		}
5910 	} else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5911 		int v;
5912 
5913 		/* Find the VSI(s) that needs to be brought down */
5914 		dev_info(&pf->pdev->dev, "VSI down requested\n");
5915 		for (v = 0; v < pf->num_alloc_vsi; v++) {
5916 			struct i40e_vsi *vsi = pf->vsi[v];
5917 
5918 			if (vsi != NULL &&
5919 			    test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
5920 					       vsi->state)) {
5921 				set_bit(__I40E_VSI_DOWN, vsi->state);
5922 				i40e_down(vsi);
5923 			}
5924 		}
5925 	} else {
5926 		dev_info(&pf->pdev->dev,
5927 			 "bad reset request 0x%08x\n", reset_flags);
5928 	}
5929 }
5930 
5931 #ifdef CONFIG_I40E_DCB
5932 /**
5933  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5934  * @pf: board private structure
5935  * @old_cfg: current DCB config
5936  * @new_cfg: new DCB config
5937  **/
5938 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5939 			    struct i40e_dcbx_config *old_cfg,
5940 			    struct i40e_dcbx_config *new_cfg)
5941 {
5942 	bool need_reconfig = false;
5943 
5944 	/* Check if ETS configuration has changed */
5945 	if (memcmp(&new_cfg->etscfg,
5946 		   &old_cfg->etscfg,
5947 		   sizeof(new_cfg->etscfg))) {
5948 		/* If Priority Table has changed reconfig is needed */
5949 		if (memcmp(&new_cfg->etscfg.prioritytable,
5950 			   &old_cfg->etscfg.prioritytable,
5951 			   sizeof(new_cfg->etscfg.prioritytable))) {
5952 			need_reconfig = true;
5953 			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5954 		}
5955 
5956 		if (memcmp(&new_cfg->etscfg.tcbwtable,
5957 			   &old_cfg->etscfg.tcbwtable,
5958 			   sizeof(new_cfg->etscfg.tcbwtable)))
5959 			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5960 
5961 		if (memcmp(&new_cfg->etscfg.tsatable,
5962 			   &old_cfg->etscfg.tsatable,
5963 			   sizeof(new_cfg->etscfg.tsatable)))
5964 			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5965 	}
5966 
5967 	/* Check if PFC configuration has changed */
5968 	if (memcmp(&new_cfg->pfc,
5969 		   &old_cfg->pfc,
5970 		   sizeof(new_cfg->pfc))) {
5971 		need_reconfig = true;
5972 		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5973 	}
5974 
5975 	/* Check if APP Table has changed */
5976 	if (memcmp(&new_cfg->app,
5977 		   &old_cfg->app,
5978 		   sizeof(new_cfg->app))) {
5979 		need_reconfig = true;
5980 		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5981 	}
5982 
5983 	dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5984 	return need_reconfig;
5985 }
5986 
5987 /**
5988  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5989  * @pf: board private structure
5990  * @e: event info posted on ARQ
5991  **/
5992 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5993 				  struct i40e_arq_event_info *e)
5994 {
5995 	struct i40e_aqc_lldp_get_mib *mib =
5996 		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5997 	struct i40e_hw *hw = &pf->hw;
5998 	struct i40e_dcbx_config tmp_dcbx_cfg;
5999 	bool need_reconfig = false;
6000 	int ret = 0;
6001 	u8 type;
6002 
6003 	/* Not DCB capable or capability disabled */
6004 	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
6005 		return ret;
6006 
6007 	/* Ignore if event is not for Nearest Bridge */
6008 	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
6009 		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
6010 	dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
6011 	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
6012 		return ret;
6013 
6014 	/* Check MIB Type and return if event for Remote MIB update */
6015 	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
6016 	dev_dbg(&pf->pdev->dev,
6017 		"LLDP event mib type %s\n", type ? "remote" : "local");
6018 	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
6019 		/* Update the remote cached instance and return */
6020 		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
6021 				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
6022 				&hw->remote_dcbx_config);
6023 		goto exit;
6024 	}
6025 
6026 	/* Store the old configuration */
6027 	tmp_dcbx_cfg = hw->local_dcbx_config;
6028 
6029 	/* Reset the old DCBx configuration data */
6030 	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
6031 	/* Get updated DCBX data from firmware */
6032 	ret = i40e_get_dcb_config(&pf->hw);
6033 	if (ret) {
6034 		dev_info(&pf->pdev->dev,
6035 			 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
6036 			 i40e_stat_str(&pf->hw, ret),
6037 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6038 		goto exit;
6039 	}
6040 
6041 	/* No change detected in DCBX configs */
6042 	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
6043 		    sizeof(tmp_dcbx_cfg))) {
6044 		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
6045 		goto exit;
6046 	}
6047 
6048 	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
6049 					       &hw->local_dcbx_config);
6050 
6051 	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
6052 
6053 	if (!need_reconfig)
6054 		goto exit;
6055 
6056 	/* Enable DCB tagging only when more than one TC */
6057 	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
6058 		pf->flags |= I40E_FLAG_DCB_ENABLED;
6059 	else
6060 		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6061 
6062 	set_bit(__I40E_PORT_SUSPENDED, pf->state);
6063 	/* Reconfiguration needed quiesce all VSIs */
6064 	i40e_pf_quiesce_all_vsi(pf);
6065 
6066 	/* Changes in configuration update VEB/VSI */
6067 	i40e_dcb_reconfigure(pf);
6068 
6069 	ret = i40e_resume_port_tx(pf);
6070 
6071 	clear_bit(__I40E_PORT_SUSPENDED, pf->state);
6072 	/* In case of error no point in resuming VSIs */
6073 	if (ret)
6074 		goto exit;
6075 
6076 	/* Wait for the PF's queues to be disabled */
6077 	ret = i40e_pf_wait_queues_disabled(pf);
6078 	if (ret) {
6079 		/* Schedule PF reset to recover */
6080 		set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6081 		i40e_service_event_schedule(pf);
6082 	} else {
6083 		i40e_pf_unquiesce_all_vsi(pf);
6084 	pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
6085 		      I40E_FLAG_CLIENT_L2_CHANGE);
6086 	}
6087 
6088 exit:
6089 	return ret;
6090 }
6091 #endif /* CONFIG_I40E_DCB */
6092 
6093 /**
6094  * i40e_do_reset_safe - Protected reset path for userland calls.
6095  * @pf: board private structure
6096  * @reset_flags: which reset is requested
6097  *
6098  **/
6099 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
6100 {
6101 	rtnl_lock();
6102 	i40e_do_reset(pf, reset_flags, true);
6103 	rtnl_unlock();
6104 }
6105 
6106 /**
6107  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
6108  * @pf: board private structure
6109  * @e: event info posted on ARQ
6110  *
6111  * Handler for LAN Queue Overflow Event generated by the firmware for PF
6112  * and VF queues
6113  **/
6114 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
6115 					   struct i40e_arq_event_info *e)
6116 {
6117 	struct i40e_aqc_lan_overflow *data =
6118 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
6119 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
6120 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
6121 	struct i40e_hw *hw = &pf->hw;
6122 	struct i40e_vf *vf;
6123 	u16 vf_id;
6124 
6125 	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
6126 		queue, qtx_ctl);
6127 
6128 	/* Queue belongs to VF, find the VF and issue VF reset */
6129 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
6130 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
6131 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
6132 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
6133 		vf_id -= hw->func_caps.vf_base_id;
6134 		vf = &pf->vf[vf_id];
6135 		i40e_vc_notify_vf_reset(vf);
6136 		/* Allow VF to process pending reset notification */
6137 		msleep(20);
6138 		i40e_reset_vf(vf, false);
6139 	}
6140 }
6141 
6142 /**
6143  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
6144  * @pf: board private structure
6145  **/
6146 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
6147 {
6148 	u32 val, fcnt_prog;
6149 
6150 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6151 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
6152 	return fcnt_prog;
6153 }
6154 
6155 /**
6156  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
6157  * @pf: board private structure
6158  **/
6159 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
6160 {
6161 	u32 val, fcnt_prog;
6162 
6163 	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6164 	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
6165 		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
6166 		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
6167 	return fcnt_prog;
6168 }
6169 
6170 /**
6171  * i40e_get_global_fd_count - Get total FD filters programmed on device
6172  * @pf: board private structure
6173  **/
6174 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
6175 {
6176 	u32 val, fcnt_prog;
6177 
6178 	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
6179 	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
6180 		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
6181 		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
6182 	return fcnt_prog;
6183 }
6184 
6185 /**
6186  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
6187  * @pf: board private structure
6188  **/
6189 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
6190 {
6191 	struct i40e_fdir_filter *filter;
6192 	u32 fcnt_prog, fcnt_avail;
6193 	struct hlist_node *node;
6194 
6195 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
6196 		return;
6197 
6198 	/* Check if we have enough room to re-enable FDir SB capability. */
6199 	fcnt_prog = i40e_get_global_fd_count(pf);
6200 	fcnt_avail = pf->fdir_pf_filter_count;
6201 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
6202 	    (pf->fd_add_err == 0) ||
6203 	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
6204 		if (pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED) {
6205 			pf->flags &= ~I40E_FLAG_FD_SB_AUTO_DISABLED;
6206 			if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
6207 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
6208 				dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
6209 		}
6210 	}
6211 
6212 	/* We should wait for even more space before re-enabling ATR.
6213 	 * Additionally, we cannot enable ATR as long as we still have TCP SB
6214 	 * rules active.
6215 	 */
6216 	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
6217 	    (pf->fd_tcp4_filter_cnt == 0)) {
6218 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
6219 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
6220 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
6221 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
6222 				dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
6223 		}
6224 	}
6225 
6226 	/* if hw had a problem adding a filter, delete it */
6227 	if (pf->fd_inv > 0) {
6228 		hlist_for_each_entry_safe(filter, node,
6229 					  &pf->fdir_filter_list, fdir_node) {
6230 			if (filter->fd_id == pf->fd_inv) {
6231 				hlist_del(&filter->fdir_node);
6232 				kfree(filter);
6233 				pf->fdir_pf_active_filters--;
6234 			}
6235 		}
6236 	}
6237 }
6238 
6239 #define I40E_MIN_FD_FLUSH_INTERVAL 10
6240 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
6241 /**
6242  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
6243  * @pf: board private structure
6244  **/
6245 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
6246 {
6247 	unsigned long min_flush_time;
6248 	int flush_wait_retry = 50;
6249 	bool disable_atr = false;
6250 	int fd_room;
6251 	int reg;
6252 
6253 	if (!time_after(jiffies, pf->fd_flush_timestamp +
6254 				 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
6255 		return;
6256 
6257 	/* If the flush is happening too quick and we have mostly SB rules we
6258 	 * should not re-enable ATR for some time.
6259 	 */
6260 	min_flush_time = pf->fd_flush_timestamp +
6261 			 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
6262 	fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
6263 
6264 	if (!(time_after(jiffies, min_flush_time)) &&
6265 	    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
6266 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
6267 			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
6268 		disable_atr = true;
6269 	}
6270 
6271 	pf->fd_flush_timestamp = jiffies;
6272 	pf->flags |= I40E_FLAG_FD_ATR_AUTO_DISABLED;
6273 	/* flush all filters */
6274 	wr32(&pf->hw, I40E_PFQF_CTL_1,
6275 	     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6276 	i40e_flush(&pf->hw);
6277 	pf->fd_flush_cnt++;
6278 	pf->fd_add_err = 0;
6279 	do {
6280 		/* Check FD flush status every 5-6msec */
6281 		usleep_range(5000, 6000);
6282 		reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6283 		if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6284 			break;
6285 	} while (flush_wait_retry--);
6286 	if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6287 		dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6288 	} else {
6289 		/* replay sideband filters */
6290 		i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6291 		if (!disable_atr && !pf->fd_tcp4_filter_cnt)
6292 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
6293 		clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
6294 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
6295 			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6296 	}
6297 }
6298 
6299 /**
6300  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6301  * @pf: board private structure
6302  **/
6303 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6304 {
6305 	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6306 }
6307 
6308 /* We can see up to 256 filter programming desc in transit if the filters are
6309  * being applied really fast; before we see the first
6310  * filter miss error on Rx queue 0. Accumulating enough error messages before
6311  * reacting will make sure we don't cause flush too often.
6312  */
6313 #define I40E_MAX_FD_PROGRAM_ERROR 256
6314 
6315 /**
6316  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6317  * @pf: board private structure
6318  **/
6319 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6320 {
6321 
6322 	/* if interface is down do nothing */
6323 	if (test_bit(__I40E_DOWN, pf->state))
6324 		return;
6325 
6326 	if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
6327 		i40e_fdir_flush_and_replay(pf);
6328 
6329 	i40e_fdir_check_and_reenable(pf);
6330 
6331 }
6332 
6333 /**
6334  * i40e_vsi_link_event - notify VSI of a link event
6335  * @vsi: vsi to be notified
6336  * @link_up: link up or down
6337  **/
6338 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6339 {
6340 	if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
6341 		return;
6342 
6343 	switch (vsi->type) {
6344 	case I40E_VSI_MAIN:
6345 		if (!vsi->netdev || !vsi->netdev_registered)
6346 			break;
6347 
6348 		if (link_up) {
6349 			netif_carrier_on(vsi->netdev);
6350 			netif_tx_wake_all_queues(vsi->netdev);
6351 		} else {
6352 			netif_carrier_off(vsi->netdev);
6353 			netif_tx_stop_all_queues(vsi->netdev);
6354 		}
6355 		break;
6356 
6357 	case I40E_VSI_SRIOV:
6358 	case I40E_VSI_VMDQ2:
6359 	case I40E_VSI_CTRL:
6360 	case I40E_VSI_IWARP:
6361 	case I40E_VSI_MIRROR:
6362 	default:
6363 		/* there is no notification for other VSIs */
6364 		break;
6365 	}
6366 }
6367 
6368 /**
6369  * i40e_veb_link_event - notify elements on the veb of a link event
6370  * @veb: veb to be notified
6371  * @link_up: link up or down
6372  **/
6373 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6374 {
6375 	struct i40e_pf *pf;
6376 	int i;
6377 
6378 	if (!veb || !veb->pf)
6379 		return;
6380 	pf = veb->pf;
6381 
6382 	/* depth first... */
6383 	for (i = 0; i < I40E_MAX_VEB; i++)
6384 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6385 			i40e_veb_link_event(pf->veb[i], link_up);
6386 
6387 	/* ... now the local VSIs */
6388 	for (i = 0; i < pf->num_alloc_vsi; i++)
6389 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6390 			i40e_vsi_link_event(pf->vsi[i], link_up);
6391 }
6392 
6393 /**
6394  * i40e_link_event - Update netif_carrier status
6395  * @pf: board private structure
6396  **/
6397 static void i40e_link_event(struct i40e_pf *pf)
6398 {
6399 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6400 	u8 new_link_speed, old_link_speed;
6401 	i40e_status status;
6402 	bool new_link, old_link;
6403 
6404 	/* save off old link status information */
6405 	pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6406 
6407 	/* set this to force the get_link_status call to refresh state */
6408 	pf->hw.phy.get_link_info = true;
6409 
6410 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6411 
6412 	status = i40e_get_link_status(&pf->hw, &new_link);
6413 
6414 	/* On success, disable temp link polling */
6415 	if (status == I40E_SUCCESS) {
6416 		if (pf->flags & I40E_FLAG_TEMP_LINK_POLLING)
6417 			pf->flags &= ~I40E_FLAG_TEMP_LINK_POLLING;
6418 	} else {
6419 		/* Enable link polling temporarily until i40e_get_link_status
6420 		 * returns I40E_SUCCESS
6421 		 */
6422 		pf->flags |= I40E_FLAG_TEMP_LINK_POLLING;
6423 		dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6424 			status);
6425 		return;
6426 	}
6427 
6428 	old_link_speed = pf->hw.phy.link_info_old.link_speed;
6429 	new_link_speed = pf->hw.phy.link_info.link_speed;
6430 
6431 	if (new_link == old_link &&
6432 	    new_link_speed == old_link_speed &&
6433 	    (test_bit(__I40E_VSI_DOWN, vsi->state) ||
6434 	     new_link == netif_carrier_ok(vsi->netdev)))
6435 		return;
6436 
6437 	i40e_print_link_message(vsi, new_link);
6438 
6439 	/* Notify the base of the switch tree connected to
6440 	 * the link.  Floating VEBs are not notified.
6441 	 */
6442 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6443 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6444 	else
6445 		i40e_vsi_link_event(vsi, new_link);
6446 
6447 	if (pf->vf)
6448 		i40e_vc_notify_link_state(pf);
6449 
6450 	if (pf->flags & I40E_FLAG_PTP)
6451 		i40e_ptp_set_increment(pf);
6452 }
6453 
6454 /**
6455  * i40e_watchdog_subtask - periodic checks not using event driven response
6456  * @pf: board private structure
6457  **/
6458 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6459 {
6460 	int i;
6461 
6462 	/* if interface is down do nothing */
6463 	if (test_bit(__I40E_DOWN, pf->state) ||
6464 	    test_bit(__I40E_CONFIG_BUSY, pf->state))
6465 		return;
6466 
6467 	/* make sure we don't do these things too often */
6468 	if (time_before(jiffies, (pf->service_timer_previous +
6469 				  pf->service_timer_period)))
6470 		return;
6471 	pf->service_timer_previous = jiffies;
6472 
6473 	if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
6474 	    (pf->flags & I40E_FLAG_TEMP_LINK_POLLING))
6475 		i40e_link_event(pf);
6476 
6477 	/* Update the stats for active netdevs so the network stack
6478 	 * can look at updated numbers whenever it cares to
6479 	 */
6480 	for (i = 0; i < pf->num_alloc_vsi; i++)
6481 		if (pf->vsi[i] && pf->vsi[i]->netdev)
6482 			i40e_update_stats(pf->vsi[i]);
6483 
6484 	if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6485 		/* Update the stats for the active switching components */
6486 		for (i = 0; i < I40E_MAX_VEB; i++)
6487 			if (pf->veb[i])
6488 				i40e_update_veb_stats(pf->veb[i]);
6489 	}
6490 
6491 	i40e_ptp_rx_hang(pf);
6492 	i40e_ptp_tx_hang(pf);
6493 }
6494 
6495 /**
6496  * i40e_reset_subtask - Set up for resetting the device and driver
6497  * @pf: board private structure
6498  **/
6499 static void i40e_reset_subtask(struct i40e_pf *pf)
6500 {
6501 	u32 reset_flags = 0;
6502 
6503 	if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
6504 		reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6505 		clear_bit(__I40E_REINIT_REQUESTED, pf->state);
6506 	}
6507 	if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
6508 		reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6509 		clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6510 	}
6511 	if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
6512 		reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6513 		clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
6514 	}
6515 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
6516 		reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6517 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
6518 	}
6519 	if (test_bit(__I40E_DOWN_REQUESTED, pf->state)) {
6520 		reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6521 		clear_bit(__I40E_DOWN_REQUESTED, pf->state);
6522 	}
6523 
6524 	/* If there's a recovery already waiting, it takes
6525 	 * precedence before starting a new reset sequence.
6526 	 */
6527 	if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
6528 		i40e_prep_for_reset(pf, false);
6529 		i40e_reset(pf);
6530 		i40e_rebuild(pf, false, false);
6531 	}
6532 
6533 	/* If we're already down or resetting, just bail */
6534 	if (reset_flags &&
6535 	    !test_bit(__I40E_DOWN, pf->state) &&
6536 	    !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
6537 		i40e_do_reset(pf, reset_flags, false);
6538 	}
6539 }
6540 
6541 /**
6542  * i40e_handle_link_event - Handle link event
6543  * @pf: board private structure
6544  * @e: event info posted on ARQ
6545  **/
6546 static void i40e_handle_link_event(struct i40e_pf *pf,
6547 				   struct i40e_arq_event_info *e)
6548 {
6549 	struct i40e_aqc_get_link_status *status =
6550 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6551 
6552 	/* Do a new status request to re-enable LSE reporting
6553 	 * and load new status information into the hw struct
6554 	 * This completely ignores any state information
6555 	 * in the ARQ event info, instead choosing to always
6556 	 * issue the AQ update link status command.
6557 	 */
6558 	i40e_link_event(pf);
6559 
6560 	/* check for unqualified module, if link is down */
6561 	if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6562 	    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6563 	    (!(status->link_info & I40E_AQ_LINK_UP)))
6564 		dev_err(&pf->pdev->dev,
6565 			"The driver failed to link because an unqualified module was detected.\n");
6566 }
6567 
6568 /**
6569  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6570  * @pf: board private structure
6571  **/
6572 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6573 {
6574 	struct i40e_arq_event_info event;
6575 	struct i40e_hw *hw = &pf->hw;
6576 	u16 pending, i = 0;
6577 	i40e_status ret;
6578 	u16 opcode;
6579 	u32 oldval;
6580 	u32 val;
6581 
6582 	/* Do not run clean AQ when PF reset fails */
6583 	if (test_bit(__I40E_RESET_FAILED, pf->state))
6584 		return;
6585 
6586 	/* check for error indications */
6587 	val = rd32(&pf->hw, pf->hw.aq.arq.len);
6588 	oldval = val;
6589 	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6590 		if (hw->debug_mask & I40E_DEBUG_AQ)
6591 			dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6592 		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6593 	}
6594 	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6595 		if (hw->debug_mask & I40E_DEBUG_AQ)
6596 			dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6597 		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6598 		pf->arq_overflows++;
6599 	}
6600 	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6601 		if (hw->debug_mask & I40E_DEBUG_AQ)
6602 			dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6603 		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6604 	}
6605 	if (oldval != val)
6606 		wr32(&pf->hw, pf->hw.aq.arq.len, val);
6607 
6608 	val = rd32(&pf->hw, pf->hw.aq.asq.len);
6609 	oldval = val;
6610 	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6611 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6612 			dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6613 		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6614 	}
6615 	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6616 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6617 			dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6618 		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6619 	}
6620 	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6621 		if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6622 			dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6623 		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6624 	}
6625 	if (oldval != val)
6626 		wr32(&pf->hw, pf->hw.aq.asq.len, val);
6627 
6628 	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6629 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6630 	if (!event.msg_buf)
6631 		return;
6632 
6633 	do {
6634 		ret = i40e_clean_arq_element(hw, &event, &pending);
6635 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6636 			break;
6637 		else if (ret) {
6638 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6639 			break;
6640 		}
6641 
6642 		opcode = le16_to_cpu(event.desc.opcode);
6643 		switch (opcode) {
6644 
6645 		case i40e_aqc_opc_get_link_status:
6646 			i40e_handle_link_event(pf, &event);
6647 			break;
6648 		case i40e_aqc_opc_send_msg_to_pf:
6649 			ret = i40e_vc_process_vf_msg(pf,
6650 					le16_to_cpu(event.desc.retval),
6651 					le32_to_cpu(event.desc.cookie_high),
6652 					le32_to_cpu(event.desc.cookie_low),
6653 					event.msg_buf,
6654 					event.msg_len);
6655 			break;
6656 		case i40e_aqc_opc_lldp_update_mib:
6657 			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6658 #ifdef CONFIG_I40E_DCB
6659 			rtnl_lock();
6660 			ret = i40e_handle_lldp_event(pf, &event);
6661 			rtnl_unlock();
6662 #endif /* CONFIG_I40E_DCB */
6663 			break;
6664 		case i40e_aqc_opc_event_lan_overflow:
6665 			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6666 			i40e_handle_lan_overflow_event(pf, &event);
6667 			break;
6668 		case i40e_aqc_opc_send_msg_to_peer:
6669 			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6670 			break;
6671 		case i40e_aqc_opc_nvm_erase:
6672 		case i40e_aqc_opc_nvm_update:
6673 		case i40e_aqc_opc_oem_post_update:
6674 			i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6675 				   "ARQ NVM operation 0x%04x completed\n",
6676 				   opcode);
6677 			break;
6678 		default:
6679 			dev_info(&pf->pdev->dev,
6680 				 "ARQ: Unknown event 0x%04x ignored\n",
6681 				 opcode);
6682 			break;
6683 		}
6684 	} while (i++ < pf->adminq_work_limit);
6685 
6686 	if (i < pf->adminq_work_limit)
6687 		clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
6688 
6689 	/* re-enable Admin queue interrupt cause */
6690 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
6691 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6692 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
6693 	i40e_flush(hw);
6694 
6695 	kfree(event.msg_buf);
6696 }
6697 
6698 /**
6699  * i40e_verify_eeprom - make sure eeprom is good to use
6700  * @pf: board private structure
6701  **/
6702 static void i40e_verify_eeprom(struct i40e_pf *pf)
6703 {
6704 	int err;
6705 
6706 	err = i40e_diag_eeprom_test(&pf->hw);
6707 	if (err) {
6708 		/* retry in case of garbage read */
6709 		err = i40e_diag_eeprom_test(&pf->hw);
6710 		if (err) {
6711 			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6712 				 err);
6713 			set_bit(__I40E_BAD_EEPROM, pf->state);
6714 		}
6715 	}
6716 
6717 	if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
6718 		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6719 		clear_bit(__I40E_BAD_EEPROM, pf->state);
6720 	}
6721 }
6722 
6723 /**
6724  * i40e_enable_pf_switch_lb
6725  * @pf: pointer to the PF structure
6726  *
6727  * enable switch loop back or die - no point in a return value
6728  **/
6729 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6730 {
6731 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6732 	struct i40e_vsi_context ctxt;
6733 	int ret;
6734 
6735 	ctxt.seid = pf->main_vsi_seid;
6736 	ctxt.pf_num = pf->hw.pf_id;
6737 	ctxt.vf_num = 0;
6738 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6739 	if (ret) {
6740 		dev_info(&pf->pdev->dev,
6741 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6742 			 i40e_stat_str(&pf->hw, ret),
6743 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6744 		return;
6745 	}
6746 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6747 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6748 	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6749 
6750 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6751 	if (ret) {
6752 		dev_info(&pf->pdev->dev,
6753 			 "update vsi switch failed, err %s aq_err %s\n",
6754 			 i40e_stat_str(&pf->hw, ret),
6755 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6756 	}
6757 }
6758 
6759 /**
6760  * i40e_disable_pf_switch_lb
6761  * @pf: pointer to the PF structure
6762  *
6763  * disable switch loop back or die - no point in a return value
6764  **/
6765 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6766 {
6767 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6768 	struct i40e_vsi_context ctxt;
6769 	int ret;
6770 
6771 	ctxt.seid = pf->main_vsi_seid;
6772 	ctxt.pf_num = pf->hw.pf_id;
6773 	ctxt.vf_num = 0;
6774 	ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6775 	if (ret) {
6776 		dev_info(&pf->pdev->dev,
6777 			 "couldn't get PF vsi config, err %s aq_err %s\n",
6778 			 i40e_stat_str(&pf->hw, ret),
6779 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6780 		return;
6781 	}
6782 	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6783 	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6784 	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6785 
6786 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6787 	if (ret) {
6788 		dev_info(&pf->pdev->dev,
6789 			 "update vsi switch failed, err %s aq_err %s\n",
6790 			 i40e_stat_str(&pf->hw, ret),
6791 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6792 	}
6793 }
6794 
6795 /**
6796  * i40e_config_bridge_mode - Configure the HW bridge mode
6797  * @veb: pointer to the bridge instance
6798  *
6799  * Configure the loop back mode for the LAN VSI that is downlink to the
6800  * specified HW bridge instance. It is expected this function is called
6801  * when a new HW bridge is instantiated.
6802  **/
6803 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6804 {
6805 	struct i40e_pf *pf = veb->pf;
6806 
6807 	if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6808 		dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6809 			 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6810 	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6811 		i40e_disable_pf_switch_lb(pf);
6812 	else
6813 		i40e_enable_pf_switch_lb(pf);
6814 }
6815 
6816 /**
6817  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6818  * @veb: pointer to the VEB instance
6819  *
6820  * This is a recursive function that first builds the attached VSIs then
6821  * recurses in to build the next layer of VEB.  We track the connections
6822  * through our own index numbers because the seid's from the HW could
6823  * change across the reset.
6824  **/
6825 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6826 {
6827 	struct i40e_vsi *ctl_vsi = NULL;
6828 	struct i40e_pf *pf = veb->pf;
6829 	int v, veb_idx;
6830 	int ret;
6831 
6832 	/* build VSI that owns this VEB, temporarily attached to base VEB */
6833 	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6834 		if (pf->vsi[v] &&
6835 		    pf->vsi[v]->veb_idx == veb->idx &&
6836 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6837 			ctl_vsi = pf->vsi[v];
6838 			break;
6839 		}
6840 	}
6841 	if (!ctl_vsi) {
6842 		dev_info(&pf->pdev->dev,
6843 			 "missing owner VSI for veb_idx %d\n", veb->idx);
6844 		ret = -ENOENT;
6845 		goto end_reconstitute;
6846 	}
6847 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
6848 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6849 	ret = i40e_add_vsi(ctl_vsi);
6850 	if (ret) {
6851 		dev_info(&pf->pdev->dev,
6852 			 "rebuild of veb_idx %d owner VSI failed: %d\n",
6853 			 veb->idx, ret);
6854 		goto end_reconstitute;
6855 	}
6856 	i40e_vsi_reset_stats(ctl_vsi);
6857 
6858 	/* create the VEB in the switch and move the VSI onto the VEB */
6859 	ret = i40e_add_veb(veb, ctl_vsi);
6860 	if (ret)
6861 		goto end_reconstitute;
6862 
6863 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6864 		veb->bridge_mode = BRIDGE_MODE_VEB;
6865 	else
6866 		veb->bridge_mode = BRIDGE_MODE_VEPA;
6867 	i40e_config_bridge_mode(veb);
6868 
6869 	/* create the remaining VSIs attached to this VEB */
6870 	for (v = 0; v < pf->num_alloc_vsi; v++) {
6871 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6872 			continue;
6873 
6874 		if (pf->vsi[v]->veb_idx == veb->idx) {
6875 			struct i40e_vsi *vsi = pf->vsi[v];
6876 
6877 			vsi->uplink_seid = veb->seid;
6878 			ret = i40e_add_vsi(vsi);
6879 			if (ret) {
6880 				dev_info(&pf->pdev->dev,
6881 					 "rebuild of vsi_idx %d failed: %d\n",
6882 					 v, ret);
6883 				goto end_reconstitute;
6884 			}
6885 			i40e_vsi_reset_stats(vsi);
6886 		}
6887 	}
6888 
6889 	/* create any VEBs attached to this VEB - RECURSION */
6890 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6891 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6892 			pf->veb[veb_idx]->uplink_seid = veb->seid;
6893 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6894 			if (ret)
6895 				break;
6896 		}
6897 	}
6898 
6899 end_reconstitute:
6900 	return ret;
6901 }
6902 
6903 /**
6904  * i40e_get_capabilities - get info about the HW
6905  * @pf: the PF struct
6906  **/
6907 static int i40e_get_capabilities(struct i40e_pf *pf)
6908 {
6909 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6910 	u16 data_size;
6911 	int buf_len;
6912 	int err;
6913 
6914 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6915 	do {
6916 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
6917 		if (!cap_buf)
6918 			return -ENOMEM;
6919 
6920 		/* this loads the data into the hw struct for us */
6921 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6922 					    &data_size,
6923 					    i40e_aqc_opc_list_func_capabilities,
6924 					    NULL);
6925 		/* data loaded, buffer no longer needed */
6926 		kfree(cap_buf);
6927 
6928 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6929 			/* retry with a larger buffer */
6930 			buf_len = data_size;
6931 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6932 			dev_info(&pf->pdev->dev,
6933 				 "capability discovery failed, err %s aq_err %s\n",
6934 				 i40e_stat_str(&pf->hw, err),
6935 				 i40e_aq_str(&pf->hw,
6936 					     pf->hw.aq.asq_last_status));
6937 			return -ENODEV;
6938 		}
6939 	} while (err);
6940 
6941 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
6942 		dev_info(&pf->pdev->dev,
6943 			 "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",
6944 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6945 			 pf->hw.func_caps.num_msix_vectors,
6946 			 pf->hw.func_caps.num_msix_vectors_vf,
6947 			 pf->hw.func_caps.fd_filters_guaranteed,
6948 			 pf->hw.func_caps.fd_filters_best_effort,
6949 			 pf->hw.func_caps.num_tx_qp,
6950 			 pf->hw.func_caps.num_vsis);
6951 
6952 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6953 		       + pf->hw.func_caps.num_vfs)
6954 	if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6955 		dev_info(&pf->pdev->dev,
6956 			 "got num_vsis %d, setting num_vsis to %d\n",
6957 			 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6958 		pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6959 	}
6960 
6961 	return 0;
6962 }
6963 
6964 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6965 
6966 /**
6967  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6968  * @pf: board private structure
6969  **/
6970 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6971 {
6972 	struct i40e_vsi *vsi;
6973 
6974 	/* quick workaround for an NVM issue that leaves a critical register
6975 	 * uninitialized
6976 	 */
6977 	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6978 		static const u32 hkey[] = {
6979 			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6980 			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6981 			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6982 			0x95b3a76d};
6983 		int i;
6984 
6985 		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6986 			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6987 	}
6988 
6989 	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6990 		return;
6991 
6992 	/* find existing VSI and see if it needs configuring */
6993 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
6994 
6995 	/* create a new VSI if none exists */
6996 	if (!vsi) {
6997 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6998 				     pf->vsi[pf->lan_vsi]->seid, 0);
6999 		if (!vsi) {
7000 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
7001 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7002 			return;
7003 		}
7004 	}
7005 
7006 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
7007 }
7008 
7009 /**
7010  * i40e_fdir_teardown - release the Flow Director resources
7011  * @pf: board private structure
7012  **/
7013 static void i40e_fdir_teardown(struct i40e_pf *pf)
7014 {
7015 	struct i40e_vsi *vsi;
7016 
7017 	i40e_fdir_filter_exit(pf);
7018 	vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
7019 	if (vsi)
7020 		i40e_vsi_release(vsi);
7021 }
7022 
7023 /**
7024  * i40e_prep_for_reset - prep for the core to reset
7025  * @pf: board private structure
7026  * @lock_acquired: indicates whether or not the lock has been acquired
7027  * before this function was called.
7028  *
7029  * Close up the VFs and other things in prep for PF Reset.
7030   **/
7031 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
7032 {
7033 	struct i40e_hw *hw = &pf->hw;
7034 	i40e_status ret = 0;
7035 	u32 v;
7036 
7037 	clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
7038 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
7039 		return;
7040 	if (i40e_check_asq_alive(&pf->hw))
7041 		i40e_vc_notify_reset(pf);
7042 
7043 	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
7044 
7045 	/* quiesce the VSIs and their queues that are not already DOWN */
7046 	/* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
7047 	if (!lock_acquired)
7048 		rtnl_lock();
7049 	i40e_pf_quiesce_all_vsi(pf);
7050 	if (!lock_acquired)
7051 		rtnl_unlock();
7052 
7053 	for (v = 0; v < pf->num_alloc_vsi; v++) {
7054 		if (pf->vsi[v])
7055 			pf->vsi[v]->seid = 0;
7056 	}
7057 
7058 	i40e_shutdown_adminq(&pf->hw);
7059 
7060 	/* call shutdown HMC */
7061 	if (hw->hmc.hmc_obj) {
7062 		ret = i40e_shutdown_lan_hmc(hw);
7063 		if (ret)
7064 			dev_warn(&pf->pdev->dev,
7065 				 "shutdown_lan_hmc failed: %d\n", ret);
7066 	}
7067 }
7068 
7069 /**
7070  * i40e_send_version - update firmware with driver version
7071  * @pf: PF struct
7072  */
7073 static void i40e_send_version(struct i40e_pf *pf)
7074 {
7075 	struct i40e_driver_version dv;
7076 
7077 	dv.major_version = DRV_VERSION_MAJOR;
7078 	dv.minor_version = DRV_VERSION_MINOR;
7079 	dv.build_version = DRV_VERSION_BUILD;
7080 	dv.subbuild_version = 0;
7081 	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
7082 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
7083 }
7084 
7085 /**
7086  * i40e_get_oem_version - get OEM specific version information
7087  * @hw: pointer to the hardware structure
7088  **/
7089 static void i40e_get_oem_version(struct i40e_hw *hw)
7090 {
7091 	u16 block_offset = 0xffff;
7092 	u16 block_length = 0;
7093 	u16 capabilities = 0;
7094 	u16 gen_snap = 0;
7095 	u16 release = 0;
7096 
7097 #define I40E_SR_NVM_OEM_VERSION_PTR		0x1B
7098 #define I40E_NVM_OEM_LENGTH_OFFSET		0x00
7099 #define I40E_NVM_OEM_CAPABILITIES_OFFSET	0x01
7100 #define I40E_NVM_OEM_GEN_OFFSET			0x02
7101 #define I40E_NVM_OEM_RELEASE_OFFSET		0x03
7102 #define I40E_NVM_OEM_CAPABILITIES_MASK		0x000F
7103 #define I40E_NVM_OEM_LENGTH			3
7104 
7105 	/* Check if pointer to OEM version block is valid. */
7106 	i40e_read_nvm_word(hw, I40E_SR_NVM_OEM_VERSION_PTR, &block_offset);
7107 	if (block_offset == 0xffff)
7108 		return;
7109 
7110 	/* Check if OEM version block has correct length. */
7111 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_LENGTH_OFFSET,
7112 			   &block_length);
7113 	if (block_length < I40E_NVM_OEM_LENGTH)
7114 		return;
7115 
7116 	/* Check if OEM version format is as expected. */
7117 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_CAPABILITIES_OFFSET,
7118 			   &capabilities);
7119 	if ((capabilities & I40E_NVM_OEM_CAPABILITIES_MASK) != 0)
7120 		return;
7121 
7122 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_GEN_OFFSET,
7123 			   &gen_snap);
7124 	i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_RELEASE_OFFSET,
7125 			   &release);
7126 	hw->nvm.oem_ver = (gen_snap << I40E_OEM_SNAP_SHIFT) | release;
7127 	hw->nvm.eetrack = I40E_OEM_EETRACK_ID;
7128 }
7129 
7130 /**
7131  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
7132  * @pf: board private structure
7133  **/
7134 static int i40e_reset(struct i40e_pf *pf)
7135 {
7136 	struct i40e_hw *hw = &pf->hw;
7137 	i40e_status ret;
7138 
7139 	ret = i40e_pf_reset(hw);
7140 	if (ret) {
7141 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
7142 		set_bit(__I40E_RESET_FAILED, pf->state);
7143 		clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
7144 	} else {
7145 		pf->pfr_count++;
7146 	}
7147 	return ret;
7148 }
7149 
7150 /**
7151  * i40e_rebuild - rebuild using a saved config
7152  * @pf: board private structure
7153  * @reinit: if the Main VSI needs to re-initialized.
7154  * @lock_acquired: indicates whether or not the lock has been acquired
7155  * before this function was called.
7156  **/
7157 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
7158 {
7159 	struct i40e_hw *hw = &pf->hw;
7160 	u8 set_fc_aq_fail = 0;
7161 	i40e_status ret;
7162 	u32 val;
7163 	int v;
7164 
7165 	if (test_bit(__I40E_DOWN, pf->state))
7166 		goto clear_recovery;
7167 	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
7168 
7169 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
7170 	ret = i40e_init_adminq(&pf->hw);
7171 	if (ret) {
7172 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
7173 			 i40e_stat_str(&pf->hw, ret),
7174 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7175 		goto clear_recovery;
7176 	}
7177 	i40e_get_oem_version(&pf->hw);
7178 
7179 	/* re-verify the eeprom if we just had an EMP reset */
7180 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
7181 		i40e_verify_eeprom(pf);
7182 
7183 	i40e_clear_pxe_mode(hw);
7184 	ret = i40e_get_capabilities(pf);
7185 	if (ret)
7186 		goto end_core_reset;
7187 
7188 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
7189 				hw->func_caps.num_rx_qp, 0, 0);
7190 	if (ret) {
7191 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
7192 		goto end_core_reset;
7193 	}
7194 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
7195 	if (ret) {
7196 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
7197 		goto end_core_reset;
7198 	}
7199 
7200 #ifdef CONFIG_I40E_DCB
7201 	ret = i40e_init_pf_dcb(pf);
7202 	if (ret) {
7203 		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
7204 		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
7205 		/* Continue without DCB enabled */
7206 	}
7207 #endif /* CONFIG_I40E_DCB */
7208 	/* do basic switch setup */
7209 	if (!lock_acquired)
7210 		rtnl_lock();
7211 	ret = i40e_setup_pf_switch(pf, reinit);
7212 	if (ret)
7213 		goto end_unlock;
7214 
7215 	/* The driver only wants link up/down and module qualification
7216 	 * reports from firmware.  Note the negative logic.
7217 	 */
7218 	ret = i40e_aq_set_phy_int_mask(&pf->hw,
7219 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
7220 					 I40E_AQ_EVENT_MEDIA_NA |
7221 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
7222 	if (ret)
7223 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
7224 			 i40e_stat_str(&pf->hw, ret),
7225 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7226 
7227 	/* make sure our flow control settings are restored */
7228 	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
7229 	if (ret)
7230 		dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
7231 			i40e_stat_str(&pf->hw, ret),
7232 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7233 
7234 	/* Rebuild the VSIs and VEBs that existed before reset.
7235 	 * They are still in our local switch element arrays, so only
7236 	 * need to rebuild the switch model in the HW.
7237 	 *
7238 	 * If there were VEBs but the reconstitution failed, we'll try
7239 	 * try to recover minimal use by getting the basic PF VSI working.
7240 	 */
7241 	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
7242 		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
7243 		/* find the one VEB connected to the MAC, and find orphans */
7244 		for (v = 0; v < I40E_MAX_VEB; v++) {
7245 			if (!pf->veb[v])
7246 				continue;
7247 
7248 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
7249 			    pf->veb[v]->uplink_seid == 0) {
7250 				ret = i40e_reconstitute_veb(pf->veb[v]);
7251 
7252 				if (!ret)
7253 					continue;
7254 
7255 				/* If Main VEB failed, we're in deep doodoo,
7256 				 * so give up rebuilding the switch and set up
7257 				 * for minimal rebuild of PF VSI.
7258 				 * If orphan failed, we'll report the error
7259 				 * but try to keep going.
7260 				 */
7261 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
7262 					dev_info(&pf->pdev->dev,
7263 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
7264 						 ret);
7265 					pf->vsi[pf->lan_vsi]->uplink_seid
7266 								= pf->mac_seid;
7267 					break;
7268 				} else if (pf->veb[v]->uplink_seid == 0) {
7269 					dev_info(&pf->pdev->dev,
7270 						 "rebuild of orphan VEB failed: %d\n",
7271 						 ret);
7272 				}
7273 			}
7274 		}
7275 	}
7276 
7277 	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
7278 		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
7279 		/* no VEB, so rebuild only the Main VSI */
7280 		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
7281 		if (ret) {
7282 			dev_info(&pf->pdev->dev,
7283 				 "rebuild of Main VSI failed: %d\n", ret);
7284 			goto end_unlock;
7285 		}
7286 	}
7287 
7288 	/* Reconfigure hardware for allowing smaller MSS in the case
7289 	 * of TSO, so that we avoid the MDD being fired and causing
7290 	 * a reset in the case of small MSS+TSO.
7291 	 */
7292 #define I40E_REG_MSS          0x000E64DC
7293 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
7294 #define I40E_64BYTE_MSS       0x400000
7295 	val = rd32(hw, I40E_REG_MSS);
7296 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
7297 		val &= ~I40E_REG_MSS_MIN_MASK;
7298 		val |= I40E_64BYTE_MSS;
7299 		wr32(hw, I40E_REG_MSS, val);
7300 	}
7301 
7302 	if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
7303 		msleep(75);
7304 		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
7305 		if (ret)
7306 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
7307 				 i40e_stat_str(&pf->hw, ret),
7308 				 i40e_aq_str(&pf->hw,
7309 					     pf->hw.aq.asq_last_status));
7310 	}
7311 	/* reinit the misc interrupt */
7312 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7313 		ret = i40e_setup_misc_vector(pf);
7314 
7315 	/* Add a filter to drop all Flow control frames from any VSI from being
7316 	 * transmitted. By doing so we stop a malicious VF from sending out
7317 	 * PAUSE or PFC frames and potentially controlling traffic for other
7318 	 * PF/VF VSIs.
7319 	 * The FW can still send Flow control frames if enabled.
7320 	 */
7321 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
7322 						       pf->main_vsi_seid);
7323 
7324 	/* restart the VSIs that were rebuilt and running before the reset */
7325 	i40e_pf_unquiesce_all_vsi(pf);
7326 
7327 	/* Release the RTNL lock before we start resetting VFs */
7328 	if (!lock_acquired)
7329 		rtnl_unlock();
7330 
7331 	i40e_reset_all_vfs(pf, true);
7332 
7333 	/* tell the firmware that we're starting */
7334 	i40e_send_version(pf);
7335 
7336 	/* We've already released the lock, so don't do it again */
7337 	goto end_core_reset;
7338 
7339 end_unlock:
7340 	if (!lock_acquired)
7341 		rtnl_unlock();
7342 end_core_reset:
7343 	clear_bit(__I40E_RESET_FAILED, pf->state);
7344 clear_recovery:
7345 	clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
7346 }
7347 
7348 /**
7349  * i40e_reset_and_rebuild - reset and rebuild using a saved config
7350  * @pf: board private structure
7351  * @reinit: if the Main VSI needs to re-initialized.
7352  * @lock_acquired: indicates whether or not the lock has been acquired
7353  * before this function was called.
7354  **/
7355 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
7356 				   bool lock_acquired)
7357 {
7358 	int ret;
7359 	/* Now we wait for GRST to settle out.
7360 	 * We don't have to delete the VEBs or VSIs from the hw switch
7361 	 * because the reset will make them disappear.
7362 	 */
7363 	ret = i40e_reset(pf);
7364 	if (!ret)
7365 		i40e_rebuild(pf, reinit, lock_acquired);
7366 }
7367 
7368 /**
7369  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7370  * @pf: board private structure
7371  *
7372  * Close up the VFs and other things in prep for a Core Reset,
7373  * then get ready to rebuild the world.
7374  * @lock_acquired: indicates whether or not the lock has been acquired
7375  * before this function was called.
7376  **/
7377 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
7378 {
7379 	i40e_prep_for_reset(pf, lock_acquired);
7380 	i40e_reset_and_rebuild(pf, false, lock_acquired);
7381 }
7382 
7383 /**
7384  * i40e_handle_mdd_event
7385  * @pf: pointer to the PF structure
7386  *
7387  * Called from the MDD irq handler to identify possibly malicious vfs
7388  **/
7389 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7390 {
7391 	struct i40e_hw *hw = &pf->hw;
7392 	bool mdd_detected = false;
7393 	bool pf_mdd_detected = false;
7394 	struct i40e_vf *vf;
7395 	u32 reg;
7396 	int i;
7397 
7398 	if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
7399 		return;
7400 
7401 	/* find what triggered the MDD event */
7402 	reg = rd32(hw, I40E_GL_MDET_TX);
7403 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7404 		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7405 				I40E_GL_MDET_TX_PF_NUM_SHIFT;
7406 		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7407 				I40E_GL_MDET_TX_VF_NUM_SHIFT;
7408 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7409 				I40E_GL_MDET_TX_EVENT_SHIFT;
7410 		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7411 				I40E_GL_MDET_TX_QUEUE_SHIFT) -
7412 				pf->hw.func_caps.base_queue;
7413 		if (netif_msg_tx_err(pf))
7414 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7415 				 event, queue, pf_num, vf_num);
7416 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7417 		mdd_detected = true;
7418 	}
7419 	reg = rd32(hw, I40E_GL_MDET_RX);
7420 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7421 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7422 				I40E_GL_MDET_RX_FUNCTION_SHIFT;
7423 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7424 				I40E_GL_MDET_RX_EVENT_SHIFT;
7425 		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7426 				I40E_GL_MDET_RX_QUEUE_SHIFT) -
7427 				pf->hw.func_caps.base_queue;
7428 		if (netif_msg_rx_err(pf))
7429 			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7430 				 event, queue, func);
7431 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7432 		mdd_detected = true;
7433 	}
7434 
7435 	if (mdd_detected) {
7436 		reg = rd32(hw, I40E_PF_MDET_TX);
7437 		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7438 			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7439 			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7440 			pf_mdd_detected = true;
7441 		}
7442 		reg = rd32(hw, I40E_PF_MDET_RX);
7443 		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7444 			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7445 			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7446 			pf_mdd_detected = true;
7447 		}
7448 		/* Queue belongs to the PF, initiate a reset */
7449 		if (pf_mdd_detected) {
7450 			set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
7451 			i40e_service_event_schedule(pf);
7452 		}
7453 	}
7454 
7455 	/* see if one of the VFs needs its hand slapped */
7456 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7457 		vf = &(pf->vf[i]);
7458 		reg = rd32(hw, I40E_VP_MDET_TX(i));
7459 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7460 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7461 			vf->num_mdd_events++;
7462 			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7463 				 i);
7464 		}
7465 
7466 		reg = rd32(hw, I40E_VP_MDET_RX(i));
7467 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7468 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7469 			vf->num_mdd_events++;
7470 			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7471 				 i);
7472 		}
7473 
7474 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7475 			dev_info(&pf->pdev->dev,
7476 				 "Too many MDD events on VF %d, disabled\n", i);
7477 			dev_info(&pf->pdev->dev,
7478 				 "Use PF Control I/F to re-enable the VF\n");
7479 			set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
7480 		}
7481 	}
7482 
7483 	/* re-enable mdd interrupt cause */
7484 	clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
7485 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7486 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7487 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7488 	i40e_flush(hw);
7489 }
7490 
7491 static const char *i40e_tunnel_name(struct i40e_udp_port_config *port)
7492 {
7493 	switch (port->type) {
7494 	case UDP_TUNNEL_TYPE_VXLAN:
7495 		return "vxlan";
7496 	case UDP_TUNNEL_TYPE_GENEVE:
7497 		return "geneve";
7498 	default:
7499 		return "unknown";
7500 	}
7501 }
7502 
7503 /**
7504  * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
7505  * @pf: board private structure
7506  **/
7507 static void i40e_sync_udp_filters(struct i40e_pf *pf)
7508 {
7509 	int i;
7510 
7511 	/* loop through and set pending bit for all active UDP filters */
7512 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7513 		if (pf->udp_ports[i].port)
7514 			pf->pending_udp_bitmap |= BIT_ULL(i);
7515 	}
7516 
7517 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
7518 }
7519 
7520 /**
7521  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7522  * @pf: board private structure
7523  **/
7524 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7525 {
7526 	struct i40e_hw *hw = &pf->hw;
7527 	i40e_status ret;
7528 	u16 port;
7529 	int i;
7530 
7531 	if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7532 		return;
7533 
7534 	pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7535 
7536 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7537 		if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7538 			pf->pending_udp_bitmap &= ~BIT_ULL(i);
7539 			port = pf->udp_ports[i].port;
7540 			if (port)
7541 				ret = i40e_aq_add_udp_tunnel(hw, port,
7542 							pf->udp_ports[i].type,
7543 							NULL, NULL);
7544 			else
7545 				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7546 
7547 			if (ret) {
7548 				dev_info(&pf->pdev->dev,
7549 					 "%s %s port %d, index %d failed, err %s aq_err %s\n",
7550 					 i40e_tunnel_name(&pf->udp_ports[i]),
7551 					 port ? "add" : "delete",
7552 					 port, i,
7553 					 i40e_stat_str(&pf->hw, ret),
7554 					 i40e_aq_str(&pf->hw,
7555 						     pf->hw.aq.asq_last_status));
7556 				pf->udp_ports[i].port = 0;
7557 			}
7558 		}
7559 	}
7560 }
7561 
7562 /**
7563  * i40e_service_task - Run the driver's async subtasks
7564  * @work: pointer to work_struct containing our data
7565  **/
7566 static void i40e_service_task(struct work_struct *work)
7567 {
7568 	struct i40e_pf *pf = container_of(work,
7569 					  struct i40e_pf,
7570 					  service_task);
7571 	unsigned long start_time = jiffies;
7572 
7573 	/* don't bother with service tasks if a reset is in progress */
7574 	if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
7575 		return;
7576 
7577 	if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
7578 		return;
7579 
7580 	i40e_detect_recover_hung(pf);
7581 	i40e_sync_filters_subtask(pf);
7582 	i40e_reset_subtask(pf);
7583 	i40e_handle_mdd_event(pf);
7584 	i40e_vc_process_vflr_event(pf);
7585 	i40e_watchdog_subtask(pf);
7586 	i40e_fdir_reinit_subtask(pf);
7587 	if (pf->flags & I40E_FLAG_CLIENT_RESET) {
7588 		/* Client subtask will reopen next time through. */
7589 		i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], true);
7590 		pf->flags &= ~I40E_FLAG_CLIENT_RESET;
7591 	} else {
7592 		i40e_client_subtask(pf);
7593 		if (pf->flags & I40E_FLAG_CLIENT_L2_CHANGE) {
7594 			i40e_notify_client_of_l2_param_changes(
7595 							pf->vsi[pf->lan_vsi]);
7596 			pf->flags &= ~I40E_FLAG_CLIENT_L2_CHANGE;
7597 		}
7598 	}
7599 	i40e_sync_filters_subtask(pf);
7600 	i40e_sync_udp_filters_subtask(pf);
7601 	i40e_clean_adminq_subtask(pf);
7602 
7603 	/* flush memory to make sure state is correct before next watchdog */
7604 	smp_mb__before_atomic();
7605 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
7606 
7607 	/* If the tasks have taken longer than one timer cycle or there
7608 	 * is more work to be done, reschedule the service task now
7609 	 * rather than wait for the timer to tick again.
7610 	 */
7611 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7612 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)		 ||
7613 	    test_bit(__I40E_MDD_EVENT_PENDING, pf->state)		 ||
7614 	    test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
7615 		i40e_service_event_schedule(pf);
7616 }
7617 
7618 /**
7619  * i40e_service_timer - timer callback
7620  * @data: pointer to PF struct
7621  **/
7622 static void i40e_service_timer(unsigned long data)
7623 {
7624 	struct i40e_pf *pf = (struct i40e_pf *)data;
7625 
7626 	mod_timer(&pf->service_timer,
7627 		  round_jiffies(jiffies + pf->service_timer_period));
7628 	i40e_service_event_schedule(pf);
7629 }
7630 
7631 /**
7632  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7633  * @vsi: the VSI being configured
7634  **/
7635 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7636 {
7637 	struct i40e_pf *pf = vsi->back;
7638 
7639 	switch (vsi->type) {
7640 	case I40E_VSI_MAIN:
7641 		vsi->alloc_queue_pairs = pf->num_lan_qps;
7642 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7643 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7644 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7645 			vsi->num_q_vectors = pf->num_lan_msix;
7646 		else
7647 			vsi->num_q_vectors = 1;
7648 
7649 		break;
7650 
7651 	case I40E_VSI_FDIR:
7652 		vsi->alloc_queue_pairs = 1;
7653 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7654 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7655 		vsi->num_q_vectors = pf->num_fdsb_msix;
7656 		break;
7657 
7658 	case I40E_VSI_VMDQ2:
7659 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7660 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7661 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7662 		vsi->num_q_vectors = pf->num_vmdq_msix;
7663 		break;
7664 
7665 	case I40E_VSI_SRIOV:
7666 		vsi->alloc_queue_pairs = pf->num_vf_qps;
7667 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7668 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
7669 		break;
7670 
7671 	default:
7672 		WARN_ON(1);
7673 		return -ENODATA;
7674 	}
7675 
7676 	return 0;
7677 }
7678 
7679 /**
7680  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7681  * @type: VSI pointer
7682  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7683  *
7684  * On error: returns error code (negative)
7685  * On success: returns 0
7686  **/
7687 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7688 {
7689 	struct i40e_ring **next_rings;
7690 	int size;
7691 	int ret = 0;
7692 
7693 	/* allocate memory for both Tx, XDP Tx and Rx ring pointers */
7694 	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs *
7695 	       (i40e_enabled_xdp_vsi(vsi) ? 3 : 2);
7696 	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7697 	if (!vsi->tx_rings)
7698 		return -ENOMEM;
7699 	next_rings = vsi->tx_rings + vsi->alloc_queue_pairs;
7700 	if (i40e_enabled_xdp_vsi(vsi)) {
7701 		vsi->xdp_rings = next_rings;
7702 		next_rings += vsi->alloc_queue_pairs;
7703 	}
7704 	vsi->rx_rings = next_rings;
7705 
7706 	if (alloc_qvectors) {
7707 		/* allocate memory for q_vector pointers */
7708 		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7709 		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7710 		if (!vsi->q_vectors) {
7711 			ret = -ENOMEM;
7712 			goto err_vectors;
7713 		}
7714 	}
7715 	return ret;
7716 
7717 err_vectors:
7718 	kfree(vsi->tx_rings);
7719 	return ret;
7720 }
7721 
7722 /**
7723  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7724  * @pf: board private structure
7725  * @type: type of VSI
7726  *
7727  * On error: returns error code (negative)
7728  * On success: returns vsi index in PF (positive)
7729  **/
7730 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7731 {
7732 	int ret = -ENODEV;
7733 	struct i40e_vsi *vsi;
7734 	int vsi_idx;
7735 	int i;
7736 
7737 	/* Need to protect the allocation of the VSIs at the PF level */
7738 	mutex_lock(&pf->switch_mutex);
7739 
7740 	/* VSI list may be fragmented if VSI creation/destruction has
7741 	 * been happening.  We can afford to do a quick scan to look
7742 	 * for any free VSIs in the list.
7743 	 *
7744 	 * find next empty vsi slot, looping back around if necessary
7745 	 */
7746 	i = pf->next_vsi;
7747 	while (i < pf->num_alloc_vsi && pf->vsi[i])
7748 		i++;
7749 	if (i >= pf->num_alloc_vsi) {
7750 		i = 0;
7751 		while (i < pf->next_vsi && pf->vsi[i])
7752 			i++;
7753 	}
7754 
7755 	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7756 		vsi_idx = i;             /* Found one! */
7757 	} else {
7758 		ret = -ENODEV;
7759 		goto unlock_pf;  /* out of VSI slots! */
7760 	}
7761 	pf->next_vsi = ++i;
7762 
7763 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7764 	if (!vsi) {
7765 		ret = -ENOMEM;
7766 		goto unlock_pf;
7767 	}
7768 	vsi->type = type;
7769 	vsi->back = pf;
7770 	set_bit(__I40E_VSI_DOWN, vsi->state);
7771 	vsi->flags = 0;
7772 	vsi->idx = vsi_idx;
7773 	vsi->int_rate_limit = 0;
7774 	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7775 				pf->rss_table_size : 64;
7776 	vsi->netdev_registered = false;
7777 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7778 	hash_init(vsi->mac_filter_hash);
7779 	vsi->irqs_ready = false;
7780 
7781 	ret = i40e_set_num_rings_in_vsi(vsi);
7782 	if (ret)
7783 		goto err_rings;
7784 
7785 	ret = i40e_vsi_alloc_arrays(vsi, true);
7786 	if (ret)
7787 		goto err_rings;
7788 
7789 	/* Setup default MSIX irq handler for VSI */
7790 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7791 
7792 	/* Initialize VSI lock */
7793 	spin_lock_init(&vsi->mac_filter_hash_lock);
7794 	pf->vsi[vsi_idx] = vsi;
7795 	ret = vsi_idx;
7796 	goto unlock_pf;
7797 
7798 err_rings:
7799 	pf->next_vsi = i - 1;
7800 	kfree(vsi);
7801 unlock_pf:
7802 	mutex_unlock(&pf->switch_mutex);
7803 	return ret;
7804 }
7805 
7806 /**
7807  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7808  * @type: VSI pointer
7809  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7810  *
7811  * On error: returns error code (negative)
7812  * On success: returns 0
7813  **/
7814 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7815 {
7816 	/* free the ring and vector containers */
7817 	if (free_qvectors) {
7818 		kfree(vsi->q_vectors);
7819 		vsi->q_vectors = NULL;
7820 	}
7821 	kfree(vsi->tx_rings);
7822 	vsi->tx_rings = NULL;
7823 	vsi->rx_rings = NULL;
7824 	vsi->xdp_rings = NULL;
7825 }
7826 
7827 /**
7828  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7829  * and lookup table
7830  * @vsi: Pointer to VSI structure
7831  */
7832 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7833 {
7834 	if (!vsi)
7835 		return;
7836 
7837 	kfree(vsi->rss_hkey_user);
7838 	vsi->rss_hkey_user = NULL;
7839 
7840 	kfree(vsi->rss_lut_user);
7841 	vsi->rss_lut_user = NULL;
7842 }
7843 
7844 /**
7845  * i40e_vsi_clear - Deallocate the VSI provided
7846  * @vsi: the VSI being un-configured
7847  **/
7848 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7849 {
7850 	struct i40e_pf *pf;
7851 
7852 	if (!vsi)
7853 		return 0;
7854 
7855 	if (!vsi->back)
7856 		goto free_vsi;
7857 	pf = vsi->back;
7858 
7859 	mutex_lock(&pf->switch_mutex);
7860 	if (!pf->vsi[vsi->idx]) {
7861 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7862 			vsi->idx, vsi->idx, vsi, vsi->type);
7863 		goto unlock_vsi;
7864 	}
7865 
7866 	if (pf->vsi[vsi->idx] != vsi) {
7867 		dev_err(&pf->pdev->dev,
7868 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7869 			pf->vsi[vsi->idx]->idx,
7870 			pf->vsi[vsi->idx],
7871 			pf->vsi[vsi->idx]->type,
7872 			vsi->idx, vsi, vsi->type);
7873 		goto unlock_vsi;
7874 	}
7875 
7876 	/* updates the PF for this cleared vsi */
7877 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7878 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7879 
7880 	i40e_vsi_free_arrays(vsi, true);
7881 	i40e_clear_rss_config_user(vsi);
7882 
7883 	pf->vsi[vsi->idx] = NULL;
7884 	if (vsi->idx < pf->next_vsi)
7885 		pf->next_vsi = vsi->idx;
7886 
7887 unlock_vsi:
7888 	mutex_unlock(&pf->switch_mutex);
7889 free_vsi:
7890 	kfree(vsi);
7891 
7892 	return 0;
7893 }
7894 
7895 /**
7896  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7897  * @vsi: the VSI being cleaned
7898  **/
7899 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7900 {
7901 	int i;
7902 
7903 	if (vsi->tx_rings && vsi->tx_rings[0]) {
7904 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7905 			kfree_rcu(vsi->tx_rings[i], rcu);
7906 			vsi->tx_rings[i] = NULL;
7907 			vsi->rx_rings[i] = NULL;
7908 			if (vsi->xdp_rings)
7909 				vsi->xdp_rings[i] = NULL;
7910 		}
7911 	}
7912 }
7913 
7914 /**
7915  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7916  * @vsi: the VSI being configured
7917  **/
7918 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7919 {
7920 	int i, qpv = i40e_enabled_xdp_vsi(vsi) ? 3 : 2;
7921 	struct i40e_pf *pf = vsi->back;
7922 	struct i40e_ring *ring;
7923 
7924 	/* Set basic values in the rings to be used later during open() */
7925 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7926 		/* allocate space for both Tx and Rx in one shot */
7927 		ring = kcalloc(qpv, sizeof(struct i40e_ring), GFP_KERNEL);
7928 		if (!ring)
7929 			goto err_out;
7930 
7931 		ring->queue_index = i;
7932 		ring->reg_idx = vsi->base_queue + i;
7933 		ring->ring_active = false;
7934 		ring->vsi = vsi;
7935 		ring->netdev = vsi->netdev;
7936 		ring->dev = &pf->pdev->dev;
7937 		ring->count = vsi->num_desc;
7938 		ring->size = 0;
7939 		ring->dcb_tc = 0;
7940 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
7941 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7942 		ring->tx_itr_setting = pf->tx_itr_default;
7943 		vsi->tx_rings[i] = ring++;
7944 
7945 		if (!i40e_enabled_xdp_vsi(vsi))
7946 			goto setup_rx;
7947 
7948 		ring->queue_index = vsi->alloc_queue_pairs + i;
7949 		ring->reg_idx = vsi->base_queue + ring->queue_index;
7950 		ring->ring_active = false;
7951 		ring->vsi = vsi;
7952 		ring->netdev = NULL;
7953 		ring->dev = &pf->pdev->dev;
7954 		ring->count = vsi->num_desc;
7955 		ring->size = 0;
7956 		ring->dcb_tc = 0;
7957 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
7958 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7959 		set_ring_xdp(ring);
7960 		ring->tx_itr_setting = pf->tx_itr_default;
7961 		vsi->xdp_rings[i] = ring++;
7962 
7963 setup_rx:
7964 		ring->queue_index = i;
7965 		ring->reg_idx = vsi->base_queue + i;
7966 		ring->ring_active = false;
7967 		ring->vsi = vsi;
7968 		ring->netdev = vsi->netdev;
7969 		ring->dev = &pf->pdev->dev;
7970 		ring->count = vsi->num_desc;
7971 		ring->size = 0;
7972 		ring->dcb_tc = 0;
7973 		ring->rx_itr_setting = pf->rx_itr_default;
7974 		vsi->rx_rings[i] = ring;
7975 	}
7976 
7977 	return 0;
7978 
7979 err_out:
7980 	i40e_vsi_clear_rings(vsi);
7981 	return -ENOMEM;
7982 }
7983 
7984 /**
7985  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7986  * @pf: board private structure
7987  * @vectors: the number of MSI-X vectors to request
7988  *
7989  * Returns the number of vectors reserved, or error
7990  **/
7991 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7992 {
7993 	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7994 					I40E_MIN_MSIX, vectors);
7995 	if (vectors < 0) {
7996 		dev_info(&pf->pdev->dev,
7997 			 "MSI-X vector reservation failed: %d\n", vectors);
7998 		vectors = 0;
7999 	}
8000 
8001 	return vectors;
8002 }
8003 
8004 /**
8005  * i40e_init_msix - Setup the MSIX capability
8006  * @pf: board private structure
8007  *
8008  * Work with the OS to set up the MSIX vectors needed.
8009  *
8010  * Returns the number of vectors reserved or negative on failure
8011  **/
8012 static int i40e_init_msix(struct i40e_pf *pf)
8013 {
8014 	struct i40e_hw *hw = &pf->hw;
8015 	int cpus, extra_vectors;
8016 	int vectors_left;
8017 	int v_budget, i;
8018 	int v_actual;
8019 	int iwarp_requested = 0;
8020 
8021 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
8022 		return -ENODEV;
8023 
8024 	/* The number of vectors we'll request will be comprised of:
8025 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
8026 	 *   - The number of LAN queue pairs
8027 	 *	- Queues being used for RSS.
8028 	 *		We don't need as many as max_rss_size vectors.
8029 	 *		use rss_size instead in the calculation since that
8030 	 *		is governed by number of cpus in the system.
8031 	 *	- assumes symmetric Tx/Rx pairing
8032 	 *   - The number of VMDq pairs
8033 	 *   - The CPU count within the NUMA node if iWARP is enabled
8034 	 * Once we count this up, try the request.
8035 	 *
8036 	 * If we can't get what we want, we'll simplify to nearly nothing
8037 	 * and try again.  If that still fails, we punt.
8038 	 */
8039 	vectors_left = hw->func_caps.num_msix_vectors;
8040 	v_budget = 0;
8041 
8042 	/* reserve one vector for miscellaneous handler */
8043 	if (vectors_left) {
8044 		v_budget++;
8045 		vectors_left--;
8046 	}
8047 
8048 	/* reserve some vectors for the main PF traffic queues. Initially we
8049 	 * only reserve at most 50% of the available vectors, in the case that
8050 	 * the number of online CPUs is large. This ensures that we can enable
8051 	 * extra features as well. Once we've enabled the other features, we
8052 	 * will use any remaining vectors to reach as close as we can to the
8053 	 * number of online CPUs.
8054 	 */
8055 	cpus = num_online_cpus();
8056 	pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
8057 	vectors_left -= pf->num_lan_msix;
8058 
8059 	/* reserve one vector for sideband flow director */
8060 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8061 		if (vectors_left) {
8062 			pf->num_fdsb_msix = 1;
8063 			v_budget++;
8064 			vectors_left--;
8065 		} else {
8066 			pf->num_fdsb_msix = 0;
8067 		}
8068 	}
8069 
8070 	/* can we reserve enough for iWARP? */
8071 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
8072 		iwarp_requested = pf->num_iwarp_msix;
8073 
8074 		if (!vectors_left)
8075 			pf->num_iwarp_msix = 0;
8076 		else if (vectors_left < pf->num_iwarp_msix)
8077 			pf->num_iwarp_msix = 1;
8078 		v_budget += pf->num_iwarp_msix;
8079 		vectors_left -= pf->num_iwarp_msix;
8080 	}
8081 
8082 	/* any vectors left over go for VMDq support */
8083 	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
8084 		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
8085 		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
8086 
8087 		if (!vectors_left) {
8088 			pf->num_vmdq_msix = 0;
8089 			pf->num_vmdq_qps = 0;
8090 		} else {
8091 			/* if we're short on vectors for what's desired, we limit
8092 			 * the queues per vmdq.  If this is still more than are
8093 			 * available, the user will need to change the number of
8094 			 * queues/vectors used by the PF later with the ethtool
8095 			 * channels command
8096 			 */
8097 			if (vmdq_vecs < vmdq_vecs_wanted)
8098 				pf->num_vmdq_qps = 1;
8099 			pf->num_vmdq_msix = pf->num_vmdq_qps;
8100 
8101 			v_budget += vmdq_vecs;
8102 			vectors_left -= vmdq_vecs;
8103 		}
8104 	}
8105 
8106 	/* On systems with a large number of SMP cores, we previously limited
8107 	 * the number of vectors for num_lan_msix to be at most 50% of the
8108 	 * available vectors, to allow for other features. Now, we add back
8109 	 * the remaining vectors. However, we ensure that the total
8110 	 * num_lan_msix will not exceed num_online_cpus(). To do this, we
8111 	 * calculate the number of vectors we can add without going over the
8112 	 * cap of CPUs. For systems with a small number of CPUs this will be
8113 	 * zero.
8114 	 */
8115 	extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
8116 	pf->num_lan_msix += extra_vectors;
8117 	vectors_left -= extra_vectors;
8118 
8119 	WARN(vectors_left < 0,
8120 	     "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
8121 
8122 	v_budget += pf->num_lan_msix;
8123 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
8124 				   GFP_KERNEL);
8125 	if (!pf->msix_entries)
8126 		return -ENOMEM;
8127 
8128 	for (i = 0; i < v_budget; i++)
8129 		pf->msix_entries[i].entry = i;
8130 	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
8131 
8132 	if (v_actual < I40E_MIN_MSIX) {
8133 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
8134 		kfree(pf->msix_entries);
8135 		pf->msix_entries = NULL;
8136 		pci_disable_msix(pf->pdev);
8137 		return -ENODEV;
8138 
8139 	} else if (v_actual == I40E_MIN_MSIX) {
8140 		/* Adjust for minimal MSIX use */
8141 		pf->num_vmdq_vsis = 0;
8142 		pf->num_vmdq_qps = 0;
8143 		pf->num_lan_qps = 1;
8144 		pf->num_lan_msix = 1;
8145 
8146 	} else if (!vectors_left) {
8147 		/* If we have limited resources, we will start with no vectors
8148 		 * for the special features and then allocate vectors to some
8149 		 * of these features based on the policy and at the end disable
8150 		 * the features that did not get any vectors.
8151 		 */
8152 		int vec;
8153 
8154 		dev_info(&pf->pdev->dev,
8155 			 "MSI-X vector limit reached, attempting to redistribute vectors\n");
8156 		/* reserve the misc vector */
8157 		vec = v_actual - 1;
8158 
8159 		/* Scale vector usage down */
8160 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
8161 		pf->num_vmdq_vsis = 1;
8162 		pf->num_vmdq_qps = 1;
8163 
8164 		/* partition out the remaining vectors */
8165 		switch (vec) {
8166 		case 2:
8167 			pf->num_lan_msix = 1;
8168 			break;
8169 		case 3:
8170 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
8171 				pf->num_lan_msix = 1;
8172 				pf->num_iwarp_msix = 1;
8173 			} else {
8174 				pf->num_lan_msix = 2;
8175 			}
8176 			break;
8177 		default:
8178 			if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
8179 				pf->num_iwarp_msix = min_t(int, (vec / 3),
8180 						 iwarp_requested);
8181 				pf->num_vmdq_vsis = min_t(int, (vec / 3),
8182 						  I40E_DEFAULT_NUM_VMDQ_VSI);
8183 			} else {
8184 				pf->num_vmdq_vsis = min_t(int, (vec / 2),
8185 						  I40E_DEFAULT_NUM_VMDQ_VSI);
8186 			}
8187 			if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8188 				pf->num_fdsb_msix = 1;
8189 				vec--;
8190 			}
8191 			pf->num_lan_msix = min_t(int,
8192 			       (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
8193 							      pf->num_lan_msix);
8194 			pf->num_lan_qps = pf->num_lan_msix;
8195 			break;
8196 		}
8197 	}
8198 
8199 	if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
8200 	    (pf->num_fdsb_msix == 0)) {
8201 		dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
8202 		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8203 	}
8204 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
8205 	    (pf->num_vmdq_msix == 0)) {
8206 		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
8207 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
8208 	}
8209 
8210 	if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
8211 	    (pf->num_iwarp_msix == 0)) {
8212 		dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
8213 		pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
8214 	}
8215 	i40e_debug(&pf->hw, I40E_DEBUG_INIT,
8216 		   "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
8217 		   pf->num_lan_msix,
8218 		   pf->num_vmdq_msix * pf->num_vmdq_vsis,
8219 		   pf->num_fdsb_msix,
8220 		   pf->num_iwarp_msix);
8221 
8222 	return v_actual;
8223 }
8224 
8225 /**
8226  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
8227  * @vsi: the VSI being configured
8228  * @v_idx: index of the vector in the vsi struct
8229  * @cpu: cpu to be used on affinity_mask
8230  *
8231  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
8232  **/
8233 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
8234 {
8235 	struct i40e_q_vector *q_vector;
8236 
8237 	/* allocate q_vector */
8238 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
8239 	if (!q_vector)
8240 		return -ENOMEM;
8241 
8242 	q_vector->vsi = vsi;
8243 	q_vector->v_idx = v_idx;
8244 	cpumask_copy(&q_vector->affinity_mask, cpu_possible_mask);
8245 
8246 	if (vsi->netdev)
8247 		netif_napi_add(vsi->netdev, &q_vector->napi,
8248 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
8249 
8250 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
8251 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
8252 
8253 	/* tie q_vector and vsi together */
8254 	vsi->q_vectors[v_idx] = q_vector;
8255 
8256 	return 0;
8257 }
8258 
8259 /**
8260  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
8261  * @vsi: the VSI being configured
8262  *
8263  * We allocate one q_vector per queue interrupt.  If allocation fails we
8264  * return -ENOMEM.
8265  **/
8266 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
8267 {
8268 	struct i40e_pf *pf = vsi->back;
8269 	int err, v_idx, num_q_vectors, current_cpu;
8270 
8271 	/* if not MSIX, give the one vector only to the LAN VSI */
8272 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
8273 		num_q_vectors = vsi->num_q_vectors;
8274 	else if (vsi == pf->vsi[pf->lan_vsi])
8275 		num_q_vectors = 1;
8276 	else
8277 		return -EINVAL;
8278 
8279 	current_cpu = cpumask_first(cpu_online_mask);
8280 
8281 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
8282 		err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
8283 		if (err)
8284 			goto err_out;
8285 		current_cpu = cpumask_next(current_cpu, cpu_online_mask);
8286 		if (unlikely(current_cpu >= nr_cpu_ids))
8287 			current_cpu = cpumask_first(cpu_online_mask);
8288 	}
8289 
8290 	return 0;
8291 
8292 err_out:
8293 	while (v_idx--)
8294 		i40e_free_q_vector(vsi, v_idx);
8295 
8296 	return err;
8297 }
8298 
8299 /**
8300  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
8301  * @pf: board private structure to initialize
8302  **/
8303 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
8304 {
8305 	int vectors = 0;
8306 	ssize_t size;
8307 
8308 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
8309 		vectors = i40e_init_msix(pf);
8310 		if (vectors < 0) {
8311 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
8312 				       I40E_FLAG_IWARP_ENABLED	|
8313 				       I40E_FLAG_RSS_ENABLED	|
8314 				       I40E_FLAG_DCB_CAPABLE	|
8315 				       I40E_FLAG_DCB_ENABLED	|
8316 				       I40E_FLAG_SRIOV_ENABLED	|
8317 				       I40E_FLAG_FD_SB_ENABLED	|
8318 				       I40E_FLAG_FD_ATR_ENABLED	|
8319 				       I40E_FLAG_VMDQ_ENABLED);
8320 
8321 			/* rework the queue expectations without MSIX */
8322 			i40e_determine_queue_usage(pf);
8323 		}
8324 	}
8325 
8326 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
8327 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
8328 		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
8329 		vectors = pci_enable_msi(pf->pdev);
8330 		if (vectors < 0) {
8331 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
8332 				 vectors);
8333 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
8334 		}
8335 		vectors = 1;  /* one MSI or Legacy vector */
8336 	}
8337 
8338 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
8339 		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
8340 
8341 	/* set up vector assignment tracking */
8342 	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
8343 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
8344 	if (!pf->irq_pile) {
8345 		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
8346 		return -ENOMEM;
8347 	}
8348 	pf->irq_pile->num_entries = vectors;
8349 	pf->irq_pile->search_hint = 0;
8350 
8351 	/* track first vector for misc interrupts, ignore return */
8352 	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
8353 
8354 	return 0;
8355 }
8356 
8357 #ifdef CONFIG_PM
8358 /**
8359  * i40e_restore_interrupt_scheme - Restore the interrupt scheme
8360  * @pf: private board data structure
8361  *
8362  * Restore the interrupt scheme that was cleared when we suspended the
8363  * device. This should be called during resume to re-allocate the q_vectors
8364  * and reacquire IRQs.
8365  */
8366 static int i40e_restore_interrupt_scheme(struct i40e_pf *pf)
8367 {
8368 	int err, i;
8369 
8370 	/* We cleared the MSI and MSI-X flags when disabling the old interrupt
8371 	 * scheme. We need to re-enabled them here in order to attempt to
8372 	 * re-acquire the MSI or MSI-X vectors
8373 	 */
8374 	pf->flags |= (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
8375 
8376 	err = i40e_init_interrupt_scheme(pf);
8377 	if (err)
8378 		return err;
8379 
8380 	/* Now that we've re-acquired IRQs, we need to remap the vectors and
8381 	 * rings together again.
8382 	 */
8383 	for (i = 0; i < pf->num_alloc_vsi; i++) {
8384 		if (pf->vsi[i]) {
8385 			err = i40e_vsi_alloc_q_vectors(pf->vsi[i]);
8386 			if (err)
8387 				goto err_unwind;
8388 			i40e_vsi_map_rings_to_vectors(pf->vsi[i]);
8389 		}
8390 	}
8391 
8392 	err = i40e_setup_misc_vector(pf);
8393 	if (err)
8394 		goto err_unwind;
8395 
8396 	return 0;
8397 
8398 err_unwind:
8399 	while (i--) {
8400 		if (pf->vsi[i])
8401 			i40e_vsi_free_q_vectors(pf->vsi[i]);
8402 	}
8403 
8404 	return err;
8405 }
8406 #endif /* CONFIG_PM */
8407 
8408 /**
8409  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
8410  * @pf: board private structure
8411  *
8412  * This sets up the handler for MSIX 0, which is used to manage the
8413  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
8414  * when in MSI or Legacy interrupt mode.
8415  **/
8416 static int i40e_setup_misc_vector(struct i40e_pf *pf)
8417 {
8418 	struct i40e_hw *hw = &pf->hw;
8419 	int err = 0;
8420 
8421 	/* Only request the IRQ once, the first time through. */
8422 	if (!test_and_set_bit(__I40E_MISC_IRQ_REQUESTED, pf->state)) {
8423 		err = request_irq(pf->msix_entries[0].vector,
8424 				  i40e_intr, 0, pf->int_name, pf);
8425 		if (err) {
8426 			clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
8427 			dev_info(&pf->pdev->dev,
8428 				 "request_irq for %s failed: %d\n",
8429 				 pf->int_name, err);
8430 			return -EFAULT;
8431 		}
8432 	}
8433 
8434 	i40e_enable_misc_int_causes(pf);
8435 
8436 	/* associate no queues to the misc vector */
8437 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
8438 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
8439 
8440 	i40e_flush(hw);
8441 
8442 	i40e_irq_dynamic_enable_icr0(pf, true);
8443 
8444 	return err;
8445 }
8446 
8447 /**
8448  * i40e_config_rss_aq - Prepare for RSS using AQ commands
8449  * @vsi: vsi structure
8450  * @seed: RSS hash seed
8451  **/
8452 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8453 			      u8 *lut, u16 lut_size)
8454 {
8455 	struct i40e_pf *pf = vsi->back;
8456 	struct i40e_hw *hw = &pf->hw;
8457 	int ret = 0;
8458 
8459 	if (seed) {
8460 		struct i40e_aqc_get_set_rss_key_data *seed_dw =
8461 			(struct i40e_aqc_get_set_rss_key_data *)seed;
8462 		ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
8463 		if (ret) {
8464 			dev_info(&pf->pdev->dev,
8465 				 "Cannot set RSS key, err %s aq_err %s\n",
8466 				 i40e_stat_str(hw, ret),
8467 				 i40e_aq_str(hw, hw->aq.asq_last_status));
8468 			return ret;
8469 		}
8470 	}
8471 	if (lut) {
8472 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8473 
8474 		ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8475 		if (ret) {
8476 			dev_info(&pf->pdev->dev,
8477 				 "Cannot set RSS lut, err %s aq_err %s\n",
8478 				 i40e_stat_str(hw, ret),
8479 				 i40e_aq_str(hw, hw->aq.asq_last_status));
8480 			return ret;
8481 		}
8482 	}
8483 	return ret;
8484 }
8485 
8486 /**
8487  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8488  * @vsi: Pointer to vsi structure
8489  * @seed: Buffter to store the hash keys
8490  * @lut: Buffer to store the lookup table entries
8491  * @lut_size: Size of buffer to store the lookup table entries
8492  *
8493  * Return 0 on success, negative on failure
8494  */
8495 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8496 			   u8 *lut, u16 lut_size)
8497 {
8498 	struct i40e_pf *pf = vsi->back;
8499 	struct i40e_hw *hw = &pf->hw;
8500 	int ret = 0;
8501 
8502 	if (seed) {
8503 		ret = i40e_aq_get_rss_key(hw, vsi->id,
8504 			(struct i40e_aqc_get_set_rss_key_data *)seed);
8505 		if (ret) {
8506 			dev_info(&pf->pdev->dev,
8507 				 "Cannot get RSS key, err %s aq_err %s\n",
8508 				 i40e_stat_str(&pf->hw, ret),
8509 				 i40e_aq_str(&pf->hw,
8510 					     pf->hw.aq.asq_last_status));
8511 			return ret;
8512 		}
8513 	}
8514 
8515 	if (lut) {
8516 		bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8517 
8518 		ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8519 		if (ret) {
8520 			dev_info(&pf->pdev->dev,
8521 				 "Cannot get RSS lut, err %s aq_err %s\n",
8522 				 i40e_stat_str(&pf->hw, ret),
8523 				 i40e_aq_str(&pf->hw,
8524 					     pf->hw.aq.asq_last_status));
8525 			return ret;
8526 		}
8527 	}
8528 
8529 	return ret;
8530 }
8531 
8532 /**
8533  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8534  * @vsi: VSI structure
8535  **/
8536 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8537 {
8538 	u8 seed[I40E_HKEY_ARRAY_SIZE];
8539 	struct i40e_pf *pf = vsi->back;
8540 	u8 *lut;
8541 	int ret;
8542 
8543 	if (!(pf->hw_features & I40E_HW_RSS_AQ_CAPABLE))
8544 		return 0;
8545 
8546 	if (!vsi->rss_size)
8547 		vsi->rss_size = min_t(int, pf->alloc_rss_size,
8548 				      vsi->num_queue_pairs);
8549 	if (!vsi->rss_size)
8550 		return -EINVAL;
8551 
8552 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8553 	if (!lut)
8554 		return -ENOMEM;
8555 	/* Use the user configured hash keys and lookup table if there is one,
8556 	 * otherwise use default
8557 	 */
8558 	if (vsi->rss_lut_user)
8559 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8560 	else
8561 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8562 	if (vsi->rss_hkey_user)
8563 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8564 	else
8565 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8566 	ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8567 	kfree(lut);
8568 
8569 	return ret;
8570 }
8571 
8572 /**
8573  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8574  * @vsi: Pointer to vsi structure
8575  * @seed: RSS hash seed
8576  * @lut: Lookup table
8577  * @lut_size: Lookup table size
8578  *
8579  * Returns 0 on success, negative on failure
8580  **/
8581 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8582 			       const u8 *lut, u16 lut_size)
8583 {
8584 	struct i40e_pf *pf = vsi->back;
8585 	struct i40e_hw *hw = &pf->hw;
8586 	u16 vf_id = vsi->vf_id;
8587 	u8 i;
8588 
8589 	/* Fill out hash function seed */
8590 	if (seed) {
8591 		u32 *seed_dw = (u32 *)seed;
8592 
8593 		if (vsi->type == I40E_VSI_MAIN) {
8594 			for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8595 				wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
8596 		} else if (vsi->type == I40E_VSI_SRIOV) {
8597 			for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8598 				wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
8599 		} else {
8600 			dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8601 		}
8602 	}
8603 
8604 	if (lut) {
8605 		u32 *lut_dw = (u32 *)lut;
8606 
8607 		if (vsi->type == I40E_VSI_MAIN) {
8608 			if (lut_size != I40E_HLUT_ARRAY_SIZE)
8609 				return -EINVAL;
8610 			for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8611 				wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8612 		} else if (vsi->type == I40E_VSI_SRIOV) {
8613 			if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8614 				return -EINVAL;
8615 			for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8616 				wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
8617 		} else {
8618 			dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8619 		}
8620 	}
8621 	i40e_flush(hw);
8622 
8623 	return 0;
8624 }
8625 
8626 /**
8627  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8628  * @vsi: Pointer to VSI structure
8629  * @seed: Buffer to store the keys
8630  * @lut: Buffer to store the lookup table entries
8631  * @lut_size: Size of buffer to store the lookup table entries
8632  *
8633  * Returns 0 on success, negative on failure
8634  */
8635 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8636 			    u8 *lut, u16 lut_size)
8637 {
8638 	struct i40e_pf *pf = vsi->back;
8639 	struct i40e_hw *hw = &pf->hw;
8640 	u16 i;
8641 
8642 	if (seed) {
8643 		u32 *seed_dw = (u32 *)seed;
8644 
8645 		for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8646 			seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8647 	}
8648 	if (lut) {
8649 		u32 *lut_dw = (u32 *)lut;
8650 
8651 		if (lut_size != I40E_HLUT_ARRAY_SIZE)
8652 			return -EINVAL;
8653 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8654 			lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8655 	}
8656 
8657 	return 0;
8658 }
8659 
8660 /**
8661  * i40e_config_rss - Configure RSS keys and lut
8662  * @vsi: Pointer to VSI structure
8663  * @seed: RSS hash seed
8664  * @lut: Lookup table
8665  * @lut_size: Lookup table size
8666  *
8667  * Returns 0 on success, negative on failure
8668  */
8669 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8670 {
8671 	struct i40e_pf *pf = vsi->back;
8672 
8673 	if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
8674 		return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8675 	else
8676 		return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8677 }
8678 
8679 /**
8680  * i40e_get_rss - Get RSS keys and lut
8681  * @vsi: Pointer to VSI structure
8682  * @seed: Buffer to store the keys
8683  * @lut: Buffer to store the lookup table entries
8684  * lut_size: Size of buffer to store the lookup table entries
8685  *
8686  * Returns 0 on success, negative on failure
8687  */
8688 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8689 {
8690 	struct i40e_pf *pf = vsi->back;
8691 
8692 	if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
8693 		return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8694 	else
8695 		return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8696 }
8697 
8698 /**
8699  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8700  * @pf: Pointer to board private structure
8701  * @lut: Lookup table
8702  * @rss_table_size: Lookup table size
8703  * @rss_size: Range of queue number for hashing
8704  */
8705 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8706 		       u16 rss_table_size, u16 rss_size)
8707 {
8708 	u16 i;
8709 
8710 	for (i = 0; i < rss_table_size; i++)
8711 		lut[i] = i % rss_size;
8712 }
8713 
8714 /**
8715  * i40e_pf_config_rss - Prepare for RSS if used
8716  * @pf: board private structure
8717  **/
8718 static int i40e_pf_config_rss(struct i40e_pf *pf)
8719 {
8720 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8721 	u8 seed[I40E_HKEY_ARRAY_SIZE];
8722 	u8 *lut;
8723 	struct i40e_hw *hw = &pf->hw;
8724 	u32 reg_val;
8725 	u64 hena;
8726 	int ret;
8727 
8728 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8729 	hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8730 		((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8731 	hena |= i40e_pf_get_default_rss_hena(pf);
8732 
8733 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8734 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8735 
8736 	/* Determine the RSS table size based on the hardware capabilities */
8737 	reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8738 	reg_val = (pf->rss_table_size == 512) ?
8739 			(reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8740 			(reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8741 	i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8742 
8743 	/* Determine the RSS size of the VSI */
8744 	if (!vsi->rss_size) {
8745 		u16 qcount;
8746 
8747 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8748 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8749 	}
8750 	if (!vsi->rss_size)
8751 		return -EINVAL;
8752 
8753 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8754 	if (!lut)
8755 		return -ENOMEM;
8756 
8757 	/* Use user configured lut if there is one, otherwise use default */
8758 	if (vsi->rss_lut_user)
8759 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8760 	else
8761 		i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8762 
8763 	/* Use user configured hash key if there is one, otherwise
8764 	 * use default.
8765 	 */
8766 	if (vsi->rss_hkey_user)
8767 		memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8768 	else
8769 		netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8770 	ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8771 	kfree(lut);
8772 
8773 	return ret;
8774 }
8775 
8776 /**
8777  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8778  * @pf: board private structure
8779  * @queue_count: the requested queue count for rss.
8780  *
8781  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8782  * count which may be different from the requested queue count.
8783  * Note: expects to be called while under rtnl_lock()
8784  **/
8785 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8786 {
8787 	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8788 	int new_rss_size;
8789 
8790 	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8791 		return 0;
8792 
8793 	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8794 
8795 	if (queue_count != vsi->num_queue_pairs) {
8796 		u16 qcount;
8797 
8798 		vsi->req_queue_pairs = queue_count;
8799 		i40e_prep_for_reset(pf, true);
8800 
8801 		pf->alloc_rss_size = new_rss_size;
8802 
8803 		i40e_reset_and_rebuild(pf, true, true);
8804 
8805 		/* Discard the user configured hash keys and lut, if less
8806 		 * queues are enabled.
8807 		 */
8808 		if (queue_count < vsi->rss_size) {
8809 			i40e_clear_rss_config_user(vsi);
8810 			dev_dbg(&pf->pdev->dev,
8811 				"discard user configured hash keys and lut\n");
8812 		}
8813 
8814 		/* Reset vsi->rss_size, as number of enabled queues changed */
8815 		qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8816 		vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8817 
8818 		i40e_pf_config_rss(pf);
8819 	}
8820 	dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
8821 		 vsi->req_queue_pairs, pf->rss_size_max);
8822 	return pf->alloc_rss_size;
8823 }
8824 
8825 /**
8826  * i40e_get_partition_bw_setting - Retrieve BW settings for this PF partition
8827  * @pf: board private structure
8828  **/
8829 i40e_status i40e_get_partition_bw_setting(struct i40e_pf *pf)
8830 {
8831 	i40e_status status;
8832 	bool min_valid, max_valid;
8833 	u32 max_bw, min_bw;
8834 
8835 	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8836 					   &min_valid, &max_valid);
8837 
8838 	if (!status) {
8839 		if (min_valid)
8840 			pf->min_bw = min_bw;
8841 		if (max_valid)
8842 			pf->max_bw = max_bw;
8843 	}
8844 
8845 	return status;
8846 }
8847 
8848 /**
8849  * i40e_set_partition_bw_setting - Set BW settings for this PF partition
8850  * @pf: board private structure
8851  **/
8852 i40e_status i40e_set_partition_bw_setting(struct i40e_pf *pf)
8853 {
8854 	struct i40e_aqc_configure_partition_bw_data bw_data;
8855 	i40e_status status;
8856 
8857 	/* Set the valid bit for this PF */
8858 	bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8859 	bw_data.max_bw[pf->hw.pf_id] = pf->max_bw & I40E_ALT_BW_VALUE_MASK;
8860 	bw_data.min_bw[pf->hw.pf_id] = pf->min_bw & I40E_ALT_BW_VALUE_MASK;
8861 
8862 	/* Set the new bandwidths */
8863 	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8864 
8865 	return status;
8866 }
8867 
8868 /**
8869  * i40e_commit_partition_bw_setting - Commit BW settings for this PF partition
8870  * @pf: board private structure
8871  **/
8872 i40e_status i40e_commit_partition_bw_setting(struct i40e_pf *pf)
8873 {
8874 	/* Commit temporary BW setting to permanent NVM image */
8875 	enum i40e_admin_queue_err last_aq_status;
8876 	i40e_status ret;
8877 	u16 nvm_word;
8878 
8879 	if (pf->hw.partition_id != 1) {
8880 		dev_info(&pf->pdev->dev,
8881 			 "Commit BW only works on partition 1! This is partition %d",
8882 			 pf->hw.partition_id);
8883 		ret = I40E_NOT_SUPPORTED;
8884 		goto bw_commit_out;
8885 	}
8886 
8887 	/* Acquire NVM for read access */
8888 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8889 	last_aq_status = pf->hw.aq.asq_last_status;
8890 	if (ret) {
8891 		dev_info(&pf->pdev->dev,
8892 			 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8893 			 i40e_stat_str(&pf->hw, ret),
8894 			 i40e_aq_str(&pf->hw, last_aq_status));
8895 		goto bw_commit_out;
8896 	}
8897 
8898 	/* Read word 0x10 of NVM - SW compatibility word 1 */
8899 	ret = i40e_aq_read_nvm(&pf->hw,
8900 			       I40E_SR_NVM_CONTROL_WORD,
8901 			       0x10, sizeof(nvm_word), &nvm_word,
8902 			       false, NULL);
8903 	/* Save off last admin queue command status before releasing
8904 	 * the NVM
8905 	 */
8906 	last_aq_status = pf->hw.aq.asq_last_status;
8907 	i40e_release_nvm(&pf->hw);
8908 	if (ret) {
8909 		dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8910 			 i40e_stat_str(&pf->hw, ret),
8911 			 i40e_aq_str(&pf->hw, last_aq_status));
8912 		goto bw_commit_out;
8913 	}
8914 
8915 	/* Wait a bit for NVM release to complete */
8916 	msleep(50);
8917 
8918 	/* Acquire NVM for write access */
8919 	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8920 	last_aq_status = pf->hw.aq.asq_last_status;
8921 	if (ret) {
8922 		dev_info(&pf->pdev->dev,
8923 			 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8924 			 i40e_stat_str(&pf->hw, ret),
8925 			 i40e_aq_str(&pf->hw, last_aq_status));
8926 		goto bw_commit_out;
8927 	}
8928 	/* Write it back out unchanged to initiate update NVM,
8929 	 * which will force a write of the shadow (alt) RAM to
8930 	 * the NVM - thus storing the bandwidth values permanently.
8931 	 */
8932 	ret = i40e_aq_update_nvm(&pf->hw,
8933 				 I40E_SR_NVM_CONTROL_WORD,
8934 				 0x10, sizeof(nvm_word),
8935 				 &nvm_word, true, NULL);
8936 	/* Save off last admin queue command status before releasing
8937 	 * the NVM
8938 	 */
8939 	last_aq_status = pf->hw.aq.asq_last_status;
8940 	i40e_release_nvm(&pf->hw);
8941 	if (ret)
8942 		dev_info(&pf->pdev->dev,
8943 			 "BW settings NOT SAVED, err %s aq_err %s\n",
8944 			 i40e_stat_str(&pf->hw, ret),
8945 			 i40e_aq_str(&pf->hw, last_aq_status));
8946 bw_commit_out:
8947 
8948 	return ret;
8949 }
8950 
8951 /**
8952  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8953  * @pf: board private structure to initialize
8954  *
8955  * i40e_sw_init initializes the Adapter private data structure.
8956  * Fields are initialized based on PCI device information and
8957  * OS network device settings (MTU size).
8958  **/
8959 static int i40e_sw_init(struct i40e_pf *pf)
8960 {
8961 	int err = 0;
8962 	int size;
8963 
8964 	/* Set default capability flags */
8965 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8966 		    I40E_FLAG_MSI_ENABLED     |
8967 		    I40E_FLAG_MSIX_ENABLED;
8968 
8969 	/* Set default ITR */
8970 	pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8971 	pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8972 
8973 	/* Depending on PF configurations, it is possible that the RSS
8974 	 * maximum might end up larger than the available queues
8975 	 */
8976 	pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8977 	pf->alloc_rss_size = 1;
8978 	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8979 	pf->rss_size_max = min_t(int, pf->rss_size_max,
8980 				 pf->hw.func_caps.num_tx_qp);
8981 	if (pf->hw.func_caps.rss) {
8982 		pf->flags |= I40E_FLAG_RSS_ENABLED;
8983 		pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8984 					   num_online_cpus());
8985 	}
8986 
8987 	/* MFP mode enabled */
8988 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8989 		pf->flags |= I40E_FLAG_MFP_ENABLED;
8990 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8991 		if (i40e_get_partition_bw_setting(pf)) {
8992 			dev_warn(&pf->pdev->dev,
8993 				 "Could not get partition bw settings\n");
8994 		} else {
8995 			dev_info(&pf->pdev->dev,
8996 				 "Partition BW Min = %8.8x, Max = %8.8x\n",
8997 				 pf->min_bw, pf->max_bw);
8998 
8999 			/* nudge the Tx scheduler */
9000 			i40e_set_partition_bw_setting(pf);
9001 		}
9002 	}
9003 
9004 	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
9005 	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
9006 		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
9007 		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
9008 		if (pf->flags & I40E_FLAG_MFP_ENABLED &&
9009 		    pf->hw.num_partitions > 1)
9010 			dev_info(&pf->pdev->dev,
9011 				 "Flow Director Sideband mode Disabled in MFP mode\n");
9012 		else
9013 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
9014 		pf->fdir_pf_filter_count =
9015 				 pf->hw.func_caps.fd_filters_guaranteed;
9016 		pf->hw.fdir_shared_filter_count =
9017 				 pf->hw.func_caps.fd_filters_best_effort;
9018 	}
9019 
9020 	if (pf->hw.mac.type == I40E_MAC_X722) {
9021 		pf->hw_features |= (I40E_HW_RSS_AQ_CAPABLE |
9022 				    I40E_HW_128_QP_RSS_CAPABLE |
9023 				    I40E_HW_ATR_EVICT_CAPABLE |
9024 				    I40E_HW_WB_ON_ITR_CAPABLE |
9025 				    I40E_HW_MULTIPLE_TCP_UDP_RSS_PCTYPE |
9026 				    I40E_HW_NO_PCI_LINK_CHECK |
9027 				    I40E_HW_USE_SET_LLDP_MIB |
9028 				    I40E_HW_GENEVE_OFFLOAD_CAPABLE |
9029 				    I40E_HW_PTP_L4_CAPABLE |
9030 				    I40E_HW_WOL_MC_MAGIC_PKT_WAKE |
9031 				    I40E_HW_OUTER_UDP_CSUM_CAPABLE);
9032 
9033 #define I40E_FDEVICT_PCTYPE_DEFAULT 0xc03
9034 		if (rd32(&pf->hw, I40E_GLQF_FDEVICTENA(1)) !=
9035 		    I40E_FDEVICT_PCTYPE_DEFAULT) {
9036 			dev_warn(&pf->pdev->dev,
9037 				 "FD EVICT PCTYPES are not right, disable FD HW EVICT\n");
9038 			pf->hw_features &= ~I40E_HW_ATR_EVICT_CAPABLE;
9039 		}
9040 	} else if ((pf->hw.aq.api_maj_ver > 1) ||
9041 		   ((pf->hw.aq.api_maj_ver == 1) &&
9042 		    (pf->hw.aq.api_min_ver > 4))) {
9043 		/* Supported in FW API version higher than 1.4 */
9044 		pf->hw_features |= I40E_HW_GENEVE_OFFLOAD_CAPABLE;
9045 	}
9046 
9047 	/* Enable HW ATR eviction if possible */
9048 	if (pf->hw_features & I40E_HW_ATR_EVICT_CAPABLE)
9049 		pf->flags |= I40E_FLAG_HW_ATR_EVICT_ENABLED;
9050 
9051 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
9052 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
9053 	    (pf->hw.aq.fw_maj_ver < 4))) {
9054 		pf->hw_features |= I40E_HW_RESTART_AUTONEG;
9055 		/* No DCB support  for FW < v4.33 */
9056 		pf->hw_features |= I40E_HW_NO_DCB_SUPPORT;
9057 	}
9058 
9059 	/* Disable FW LLDP if FW < v4.3 */
9060 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
9061 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
9062 	    (pf->hw.aq.fw_maj_ver < 4)))
9063 		pf->hw_features |= I40E_HW_STOP_FW_LLDP;
9064 
9065 	/* Use the FW Set LLDP MIB API if FW > v4.40 */
9066 	if ((pf->hw.mac.type == I40E_MAC_XL710) &&
9067 	    (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
9068 	    (pf->hw.aq.fw_maj_ver >= 5)))
9069 		pf->hw_features |= I40E_HW_USE_SET_LLDP_MIB;
9070 
9071 	if (pf->hw.func_caps.vmdq) {
9072 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
9073 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
9074 		pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
9075 	}
9076 
9077 	if (pf->hw.func_caps.iwarp) {
9078 		pf->flags |= I40E_FLAG_IWARP_ENABLED;
9079 		/* IWARP needs one extra vector for CQP just like MISC.*/
9080 		pf->num_iwarp_msix = (int)num_online_cpus() + 1;
9081 	}
9082 
9083 #ifdef CONFIG_PCI_IOV
9084 	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
9085 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
9086 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
9087 		pf->num_req_vfs = min_t(int,
9088 					pf->hw.func_caps.num_vfs,
9089 					I40E_MAX_VF_COUNT);
9090 	}
9091 #endif /* CONFIG_PCI_IOV */
9092 	pf->eeprom_version = 0xDEAD;
9093 	pf->lan_veb = I40E_NO_VEB;
9094 	pf->lan_vsi = I40E_NO_VSI;
9095 
9096 	/* By default FW has this off for performance reasons */
9097 	pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
9098 
9099 	/* set up queue assignment tracking */
9100 	size = sizeof(struct i40e_lump_tracking)
9101 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
9102 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
9103 	if (!pf->qp_pile) {
9104 		err = -ENOMEM;
9105 		goto sw_init_done;
9106 	}
9107 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
9108 	pf->qp_pile->search_hint = 0;
9109 
9110 	pf->tx_timeout_recovery_level = 1;
9111 
9112 	mutex_init(&pf->switch_mutex);
9113 
9114 sw_init_done:
9115 	return err;
9116 }
9117 
9118 /**
9119  * i40e_set_ntuple - set the ntuple feature flag and take action
9120  * @pf: board private structure to initialize
9121  * @features: the feature set that the stack is suggesting
9122  *
9123  * returns a bool to indicate if reset needs to happen
9124  **/
9125 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
9126 {
9127 	bool need_reset = false;
9128 
9129 	/* Check if Flow Director n-tuple support was enabled or disabled.  If
9130 	 * the state changed, we need to reset.
9131 	 */
9132 	if (features & NETIF_F_NTUPLE) {
9133 		/* Enable filters and mark for reset */
9134 		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
9135 			need_reset = true;
9136 		/* enable FD_SB only if there is MSI-X vector */
9137 		if (pf->num_fdsb_msix > 0)
9138 			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
9139 	} else {
9140 		/* turn off filters, mark for reset and clear SW filter list */
9141 		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9142 			need_reset = true;
9143 			i40e_fdir_filter_exit(pf);
9144 		}
9145 		pf->flags &= ~(I40E_FLAG_FD_SB_ENABLED |
9146 			       I40E_FLAG_FD_SB_AUTO_DISABLED);
9147 		/* reset fd counters */
9148 		pf->fd_add_err = 0;
9149 		pf->fd_atr_cnt = 0;
9150 		/* if ATR was auto disabled it can be re-enabled. */
9151 		if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
9152 			pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
9153 			if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
9154 			    (I40E_DEBUG_FD & pf->hw.debug_mask))
9155 				dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
9156 		}
9157 	}
9158 	return need_reset;
9159 }
9160 
9161 /**
9162  * i40e_clear_rss_lut - clear the rx hash lookup table
9163  * @vsi: the VSI being configured
9164  **/
9165 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
9166 {
9167 	struct i40e_pf *pf = vsi->back;
9168 	struct i40e_hw *hw = &pf->hw;
9169 	u16 vf_id = vsi->vf_id;
9170 	u8 i;
9171 
9172 	if (vsi->type == I40E_VSI_MAIN) {
9173 		for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
9174 			wr32(hw, I40E_PFQF_HLUT(i), 0);
9175 	} else if (vsi->type == I40E_VSI_SRIOV) {
9176 		for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
9177 			i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
9178 	} else {
9179 		dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
9180 	}
9181 }
9182 
9183 /**
9184  * i40e_set_features - set the netdev feature flags
9185  * @netdev: ptr to the netdev being adjusted
9186  * @features: the feature set that the stack is suggesting
9187  * Note: expects to be called while under rtnl_lock()
9188  **/
9189 static int i40e_set_features(struct net_device *netdev,
9190 			     netdev_features_t features)
9191 {
9192 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9193 	struct i40e_vsi *vsi = np->vsi;
9194 	struct i40e_pf *pf = vsi->back;
9195 	bool need_reset;
9196 
9197 	if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
9198 		i40e_pf_config_rss(pf);
9199 	else if (!(features & NETIF_F_RXHASH) &&
9200 		 netdev->features & NETIF_F_RXHASH)
9201 		i40e_clear_rss_lut(vsi);
9202 
9203 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
9204 		i40e_vlan_stripping_enable(vsi);
9205 	else
9206 		i40e_vlan_stripping_disable(vsi);
9207 
9208 	need_reset = i40e_set_ntuple(pf, features);
9209 
9210 	if (need_reset)
9211 		i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
9212 
9213 	return 0;
9214 }
9215 
9216 /**
9217  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
9218  * @pf: board private structure
9219  * @port: The UDP port to look up
9220  *
9221  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
9222  **/
9223 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
9224 {
9225 	u8 i;
9226 
9227 	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
9228 		if (pf->udp_ports[i].port == port)
9229 			return i;
9230 	}
9231 
9232 	return i;
9233 }
9234 
9235 /**
9236  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
9237  * @netdev: This physical port's netdev
9238  * @ti: Tunnel endpoint information
9239  **/
9240 static void i40e_udp_tunnel_add(struct net_device *netdev,
9241 				struct udp_tunnel_info *ti)
9242 {
9243 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9244 	struct i40e_vsi *vsi = np->vsi;
9245 	struct i40e_pf *pf = vsi->back;
9246 	u16 port = ntohs(ti->port);
9247 	u8 next_idx;
9248 	u8 idx;
9249 
9250 	idx = i40e_get_udp_port_idx(pf, port);
9251 
9252 	/* Check if port already exists */
9253 	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
9254 		netdev_info(netdev, "port %d already offloaded\n", port);
9255 		return;
9256 	}
9257 
9258 	/* Now check if there is space to add the new port */
9259 	next_idx = i40e_get_udp_port_idx(pf, 0);
9260 
9261 	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
9262 		netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
9263 			    port);
9264 		return;
9265 	}
9266 
9267 	switch (ti->type) {
9268 	case UDP_TUNNEL_TYPE_VXLAN:
9269 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
9270 		break;
9271 	case UDP_TUNNEL_TYPE_GENEVE:
9272 		if (!(pf->hw_features & I40E_HW_GENEVE_OFFLOAD_CAPABLE))
9273 			return;
9274 		pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
9275 		break;
9276 	default:
9277 		return;
9278 	}
9279 
9280 	/* New port: add it and mark its index in the bitmap */
9281 	pf->udp_ports[next_idx].port = port;
9282 	pf->pending_udp_bitmap |= BIT_ULL(next_idx);
9283 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9284 }
9285 
9286 /**
9287  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
9288  * @netdev: This physical port's netdev
9289  * @ti: Tunnel endpoint information
9290  **/
9291 static void i40e_udp_tunnel_del(struct net_device *netdev,
9292 				struct udp_tunnel_info *ti)
9293 {
9294 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9295 	struct i40e_vsi *vsi = np->vsi;
9296 	struct i40e_pf *pf = vsi->back;
9297 	u16 port = ntohs(ti->port);
9298 	u8 idx;
9299 
9300 	idx = i40e_get_udp_port_idx(pf, port);
9301 
9302 	/* Check if port already exists */
9303 	if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
9304 		goto not_found;
9305 
9306 	switch (ti->type) {
9307 	case UDP_TUNNEL_TYPE_VXLAN:
9308 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
9309 			goto not_found;
9310 		break;
9311 	case UDP_TUNNEL_TYPE_GENEVE:
9312 		if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
9313 			goto not_found;
9314 		break;
9315 	default:
9316 		goto not_found;
9317 	}
9318 
9319 	/* if port exists, set it to 0 (mark for deletion)
9320 	 * and make it pending
9321 	 */
9322 	pf->udp_ports[idx].port = 0;
9323 	pf->pending_udp_bitmap |= BIT_ULL(idx);
9324 	pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9325 
9326 	return;
9327 not_found:
9328 	netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
9329 		    port);
9330 }
9331 
9332 static int i40e_get_phys_port_id(struct net_device *netdev,
9333 				 struct netdev_phys_item_id *ppid)
9334 {
9335 	struct i40e_netdev_priv *np = netdev_priv(netdev);
9336 	struct i40e_pf *pf = np->vsi->back;
9337 	struct i40e_hw *hw = &pf->hw;
9338 
9339 	if (!(pf->hw_features & I40E_HW_PORT_ID_VALID))
9340 		return -EOPNOTSUPP;
9341 
9342 	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
9343 	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
9344 
9345 	return 0;
9346 }
9347 
9348 /**
9349  * i40e_ndo_fdb_add - add an entry to the hardware database
9350  * @ndm: the input from the stack
9351  * @tb: pointer to array of nladdr (unused)
9352  * @dev: the net device pointer
9353  * @addr: the MAC address entry being added
9354  * @flags: instructions from stack about fdb operation
9355  */
9356 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9357 			    struct net_device *dev,
9358 			    const unsigned char *addr, u16 vid,
9359 			    u16 flags)
9360 {
9361 	struct i40e_netdev_priv *np = netdev_priv(dev);
9362 	struct i40e_pf *pf = np->vsi->back;
9363 	int err = 0;
9364 
9365 	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
9366 		return -EOPNOTSUPP;
9367 
9368 	if (vid) {
9369 		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
9370 		return -EINVAL;
9371 	}
9372 
9373 	/* Hardware does not support aging addresses so if a
9374 	 * ndm_state is given only allow permanent addresses
9375 	 */
9376 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
9377 		netdev_info(dev, "FDB only supports static addresses\n");
9378 		return -EINVAL;
9379 	}
9380 
9381 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
9382 		err = dev_uc_add_excl(dev, addr);
9383 	else if (is_multicast_ether_addr(addr))
9384 		err = dev_mc_add_excl(dev, addr);
9385 	else
9386 		err = -EINVAL;
9387 
9388 	/* Only return duplicate errors if NLM_F_EXCL is set */
9389 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
9390 		err = 0;
9391 
9392 	return err;
9393 }
9394 
9395 /**
9396  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
9397  * @dev: the netdev being configured
9398  * @nlh: RTNL message
9399  *
9400  * Inserts a new hardware bridge if not already created and
9401  * enables the bridging mode requested (VEB or VEPA). If the
9402  * hardware bridge has already been inserted and the request
9403  * is to change the mode then that requires a PF reset to
9404  * allow rebuild of the components with required hardware
9405  * bridge mode enabled.
9406  *
9407  * Note: expects to be called while under rtnl_lock()
9408  **/
9409 static int i40e_ndo_bridge_setlink(struct net_device *dev,
9410 				   struct nlmsghdr *nlh,
9411 				   u16 flags)
9412 {
9413 	struct i40e_netdev_priv *np = netdev_priv(dev);
9414 	struct i40e_vsi *vsi = np->vsi;
9415 	struct i40e_pf *pf = vsi->back;
9416 	struct i40e_veb *veb = NULL;
9417 	struct nlattr *attr, *br_spec;
9418 	int i, rem;
9419 
9420 	/* Only for PF VSI for now */
9421 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9422 		return -EOPNOTSUPP;
9423 
9424 	/* Find the HW bridge for PF VSI */
9425 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9426 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9427 			veb = pf->veb[i];
9428 	}
9429 
9430 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9431 
9432 	nla_for_each_nested(attr, br_spec, rem) {
9433 		__u16 mode;
9434 
9435 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
9436 			continue;
9437 
9438 		mode = nla_get_u16(attr);
9439 		if ((mode != BRIDGE_MODE_VEPA) &&
9440 		    (mode != BRIDGE_MODE_VEB))
9441 			return -EINVAL;
9442 
9443 		/* Insert a new HW bridge */
9444 		if (!veb) {
9445 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9446 					     vsi->tc_config.enabled_tc);
9447 			if (veb) {
9448 				veb->bridge_mode = mode;
9449 				i40e_config_bridge_mode(veb);
9450 			} else {
9451 				/* No Bridge HW offload available */
9452 				return -ENOENT;
9453 			}
9454 			break;
9455 		} else if (mode != veb->bridge_mode) {
9456 			/* Existing HW bridge but different mode needs reset */
9457 			veb->bridge_mode = mode;
9458 			/* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
9459 			if (mode == BRIDGE_MODE_VEB)
9460 				pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
9461 			else
9462 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9463 			i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED),
9464 				      true);
9465 			break;
9466 		}
9467 	}
9468 
9469 	return 0;
9470 }
9471 
9472 /**
9473  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
9474  * @skb: skb buff
9475  * @pid: process id
9476  * @seq: RTNL message seq #
9477  * @dev: the netdev being configured
9478  * @filter_mask: unused
9479  * @nlflags: netlink flags passed in
9480  *
9481  * Return the mode in which the hardware bridge is operating in
9482  * i.e VEB or VEPA.
9483  **/
9484 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9485 				   struct net_device *dev,
9486 				   u32 __always_unused filter_mask,
9487 				   int nlflags)
9488 {
9489 	struct i40e_netdev_priv *np = netdev_priv(dev);
9490 	struct i40e_vsi *vsi = np->vsi;
9491 	struct i40e_pf *pf = vsi->back;
9492 	struct i40e_veb *veb = NULL;
9493 	int i;
9494 
9495 	/* Only for PF VSI for now */
9496 	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9497 		return -EOPNOTSUPP;
9498 
9499 	/* Find the HW bridge for the PF VSI */
9500 	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9501 		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9502 			veb = pf->veb[i];
9503 	}
9504 
9505 	if (!veb)
9506 		return 0;
9507 
9508 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9509 				       0, 0, nlflags, filter_mask, NULL);
9510 }
9511 
9512 /**
9513  * i40e_features_check - Validate encapsulated packet conforms to limits
9514  * @skb: skb buff
9515  * @dev: This physical port's netdev
9516  * @features: Offload features that the stack believes apply
9517  **/
9518 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9519 					     struct net_device *dev,
9520 					     netdev_features_t features)
9521 {
9522 	size_t len;
9523 
9524 	/* No point in doing any of this if neither checksum nor GSO are
9525 	 * being requested for this frame.  We can rule out both by just
9526 	 * checking for CHECKSUM_PARTIAL
9527 	 */
9528 	if (skb->ip_summed != CHECKSUM_PARTIAL)
9529 		return features;
9530 
9531 	/* We cannot support GSO if the MSS is going to be less than
9532 	 * 64 bytes.  If it is then we need to drop support for GSO.
9533 	 */
9534 	if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
9535 		features &= ~NETIF_F_GSO_MASK;
9536 
9537 	/* MACLEN can support at most 63 words */
9538 	len = skb_network_header(skb) - skb->data;
9539 	if (len & ~(63 * 2))
9540 		goto out_err;
9541 
9542 	/* IPLEN and EIPLEN can support at most 127 dwords */
9543 	len = skb_transport_header(skb) - skb_network_header(skb);
9544 	if (len & ~(127 * 4))
9545 		goto out_err;
9546 
9547 	if (skb->encapsulation) {
9548 		/* L4TUNLEN can support 127 words */
9549 		len = skb_inner_network_header(skb) - skb_transport_header(skb);
9550 		if (len & ~(127 * 2))
9551 			goto out_err;
9552 
9553 		/* IPLEN can support at most 127 dwords */
9554 		len = skb_inner_transport_header(skb) -
9555 		      skb_inner_network_header(skb);
9556 		if (len & ~(127 * 4))
9557 			goto out_err;
9558 	}
9559 
9560 	/* No need to validate L4LEN as TCP is the only protocol with a
9561 	 * a flexible value and we support all possible values supported
9562 	 * by TCP, which is at most 15 dwords
9563 	 */
9564 
9565 	return features;
9566 out_err:
9567 	return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9568 }
9569 
9570 /**
9571  * i40e_xdp_setup - add/remove an XDP program
9572  * @vsi: VSI to changed
9573  * @prog: XDP program
9574  **/
9575 static int i40e_xdp_setup(struct i40e_vsi *vsi,
9576 			  struct bpf_prog *prog)
9577 {
9578 	int frame_size = vsi->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
9579 	struct i40e_pf *pf = vsi->back;
9580 	struct bpf_prog *old_prog;
9581 	bool need_reset;
9582 	int i;
9583 
9584 	/* Don't allow frames that span over multiple buffers */
9585 	if (frame_size > vsi->rx_buf_len)
9586 		return -EINVAL;
9587 
9588 	if (!i40e_enabled_xdp_vsi(vsi) && !prog)
9589 		return 0;
9590 
9591 	/* When turning XDP on->off/off->on we reset and rebuild the rings. */
9592 	need_reset = (i40e_enabled_xdp_vsi(vsi) != !!prog);
9593 
9594 	if (need_reset)
9595 		i40e_prep_for_reset(pf, true);
9596 
9597 	old_prog = xchg(&vsi->xdp_prog, prog);
9598 
9599 	if (need_reset)
9600 		i40e_reset_and_rebuild(pf, true, true);
9601 
9602 	for (i = 0; i < vsi->num_queue_pairs; i++)
9603 		WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
9604 
9605 	if (old_prog)
9606 		bpf_prog_put(old_prog);
9607 
9608 	return 0;
9609 }
9610 
9611 /**
9612  * i40e_xdp - implements ndo_xdp for i40e
9613  * @dev: netdevice
9614  * @xdp: XDP command
9615  **/
9616 static int i40e_xdp(struct net_device *dev,
9617 		    struct netdev_xdp *xdp)
9618 {
9619 	struct i40e_netdev_priv *np = netdev_priv(dev);
9620 	struct i40e_vsi *vsi = np->vsi;
9621 
9622 	if (vsi->type != I40E_VSI_MAIN)
9623 		return -EINVAL;
9624 
9625 	switch (xdp->command) {
9626 	case XDP_SETUP_PROG:
9627 		return i40e_xdp_setup(vsi, xdp->prog);
9628 	case XDP_QUERY_PROG:
9629 		xdp->prog_attached = i40e_enabled_xdp_vsi(vsi);
9630 		xdp->prog_id = vsi->xdp_prog ? vsi->xdp_prog->aux->id : 0;
9631 		return 0;
9632 	default:
9633 		return -EINVAL;
9634 	}
9635 }
9636 
9637 static const struct net_device_ops i40e_netdev_ops = {
9638 	.ndo_open		= i40e_open,
9639 	.ndo_stop		= i40e_close,
9640 	.ndo_start_xmit		= i40e_lan_xmit_frame,
9641 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
9642 	.ndo_set_rx_mode	= i40e_set_rx_mode,
9643 	.ndo_validate_addr	= eth_validate_addr,
9644 	.ndo_set_mac_address	= i40e_set_mac,
9645 	.ndo_change_mtu		= i40e_change_mtu,
9646 	.ndo_do_ioctl		= i40e_ioctl,
9647 	.ndo_tx_timeout		= i40e_tx_timeout,
9648 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
9649 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
9650 #ifdef CONFIG_NET_POLL_CONTROLLER
9651 	.ndo_poll_controller	= i40e_netpoll,
9652 #endif
9653 	.ndo_setup_tc		= __i40e_setup_tc,
9654 	.ndo_set_features	= i40e_set_features,
9655 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
9656 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
9657 	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
9658 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
9659 	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
9660 	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
9661 	.ndo_set_vf_trust	= i40e_ndo_set_vf_trust,
9662 	.ndo_udp_tunnel_add	= i40e_udp_tunnel_add,
9663 	.ndo_udp_tunnel_del	= i40e_udp_tunnel_del,
9664 	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
9665 	.ndo_fdb_add		= i40e_ndo_fdb_add,
9666 	.ndo_features_check	= i40e_features_check,
9667 	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
9668 	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
9669 	.ndo_xdp		= i40e_xdp,
9670 };
9671 
9672 /**
9673  * i40e_config_netdev - Setup the netdev flags
9674  * @vsi: the VSI being configured
9675  *
9676  * Returns 0 on success, negative value on failure
9677  **/
9678 static int i40e_config_netdev(struct i40e_vsi *vsi)
9679 {
9680 	struct i40e_pf *pf = vsi->back;
9681 	struct i40e_hw *hw = &pf->hw;
9682 	struct i40e_netdev_priv *np;
9683 	struct net_device *netdev;
9684 	u8 broadcast[ETH_ALEN];
9685 	u8 mac_addr[ETH_ALEN];
9686 	int etherdev_size;
9687 	netdev_features_t hw_enc_features;
9688 	netdev_features_t hw_features;
9689 
9690 	etherdev_size = sizeof(struct i40e_netdev_priv);
9691 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9692 	if (!netdev)
9693 		return -ENOMEM;
9694 
9695 	vsi->netdev = netdev;
9696 	np = netdev_priv(netdev);
9697 	np->vsi = vsi;
9698 
9699 	hw_enc_features = NETIF_F_SG			|
9700 			  NETIF_F_IP_CSUM		|
9701 			  NETIF_F_IPV6_CSUM		|
9702 			  NETIF_F_HIGHDMA		|
9703 			  NETIF_F_SOFT_FEATURES		|
9704 			  NETIF_F_TSO			|
9705 			  NETIF_F_TSO_ECN		|
9706 			  NETIF_F_TSO6			|
9707 			  NETIF_F_GSO_GRE		|
9708 			  NETIF_F_GSO_GRE_CSUM		|
9709 			  NETIF_F_GSO_PARTIAL		|
9710 			  NETIF_F_GSO_UDP_TUNNEL	|
9711 			  NETIF_F_GSO_UDP_TUNNEL_CSUM	|
9712 			  NETIF_F_SCTP_CRC		|
9713 			  NETIF_F_RXHASH		|
9714 			  NETIF_F_RXCSUM		|
9715 			  0;
9716 
9717 	if (!(pf->hw_features & I40E_HW_OUTER_UDP_CSUM_CAPABLE))
9718 		netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9719 
9720 	netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9721 
9722 	netdev->hw_enc_features |= hw_enc_features;
9723 
9724 	/* record features VLANs can make use of */
9725 	netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
9726 
9727 	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9728 		netdev->hw_features |= NETIF_F_NTUPLE;
9729 	hw_features = hw_enc_features		|
9730 		      NETIF_F_HW_VLAN_CTAG_TX	|
9731 		      NETIF_F_HW_VLAN_CTAG_RX;
9732 
9733 	netdev->hw_features |= hw_features;
9734 
9735 	netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9736 	netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9737 
9738 	if (vsi->type == I40E_VSI_MAIN) {
9739 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9740 		ether_addr_copy(mac_addr, hw->mac.perm_addr);
9741 		/* The following steps are necessary for two reasons. First,
9742 		 * some older NVM configurations load a default MAC-VLAN
9743 		 * filter that will accept any tagged packet, and we want to
9744 		 * replace this with a normal filter. Additionally, it is
9745 		 * possible our MAC address was provided by the platform using
9746 		 * Open Firmware or similar.
9747 		 *
9748 		 * Thus, we need to remove the default filter and install one
9749 		 * specific to the MAC address.
9750 		 */
9751 		i40e_rm_default_mac_filter(vsi, mac_addr);
9752 		spin_lock_bh(&vsi->mac_filter_hash_lock);
9753 		i40e_add_mac_filter(vsi, mac_addr);
9754 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
9755 	} else {
9756 		/* Relate the VSI_VMDQ name to the VSI_MAIN name. Note that we
9757 		 * are still limited by IFNAMSIZ, but we're adding 'v%d\0' to
9758 		 * the end, which is 4 bytes long, so force truncation of the
9759 		 * original name by IFNAMSIZ - 4
9760 		 */
9761 		snprintf(netdev->name, IFNAMSIZ, "%.*sv%%d",
9762 			 IFNAMSIZ - 4,
9763 			 pf->vsi[pf->lan_vsi]->netdev->name);
9764 		random_ether_addr(mac_addr);
9765 
9766 		spin_lock_bh(&vsi->mac_filter_hash_lock);
9767 		i40e_add_mac_filter(vsi, mac_addr);
9768 		spin_unlock_bh(&vsi->mac_filter_hash_lock);
9769 	}
9770 
9771 	/* Add the broadcast filter so that we initially will receive
9772 	 * broadcast packets. Note that when a new VLAN is first added the
9773 	 * driver will convert all filters marked I40E_VLAN_ANY into VLAN
9774 	 * specific filters as part of transitioning into "vlan" operation.
9775 	 * When more VLANs are added, the driver will copy each existing MAC
9776 	 * filter and add it for the new VLAN.
9777 	 *
9778 	 * Broadcast filters are handled specially by
9779 	 * i40e_sync_filters_subtask, as the driver must to set the broadcast
9780 	 * promiscuous bit instead of adding this directly as a MAC/VLAN
9781 	 * filter. The subtask will update the correct broadcast promiscuous
9782 	 * bits as VLANs become active or inactive.
9783 	 */
9784 	eth_broadcast_addr(broadcast);
9785 	spin_lock_bh(&vsi->mac_filter_hash_lock);
9786 	i40e_add_mac_filter(vsi, broadcast);
9787 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
9788 
9789 	ether_addr_copy(netdev->dev_addr, mac_addr);
9790 	ether_addr_copy(netdev->perm_addr, mac_addr);
9791 
9792 	netdev->priv_flags |= IFF_UNICAST_FLT;
9793 	netdev->priv_flags |= IFF_SUPP_NOFCS;
9794 	/* Setup netdev TC information */
9795 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9796 
9797 	netdev->netdev_ops = &i40e_netdev_ops;
9798 	netdev->watchdog_timeo = 5 * HZ;
9799 	i40e_set_ethtool_ops(netdev);
9800 
9801 	/* MTU range: 68 - 9706 */
9802 	netdev->min_mtu = ETH_MIN_MTU;
9803 	netdev->max_mtu = I40E_MAX_RXBUFFER - I40E_PACKET_HDR_PAD;
9804 
9805 	return 0;
9806 }
9807 
9808 /**
9809  * i40e_vsi_delete - Delete a VSI from the switch
9810  * @vsi: the VSI being removed
9811  *
9812  * Returns 0 on success, negative value on failure
9813  **/
9814 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9815 {
9816 	/* remove default VSI is not allowed */
9817 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9818 		return;
9819 
9820 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9821 }
9822 
9823 /**
9824  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9825  * @vsi: the VSI being queried
9826  *
9827  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9828  **/
9829 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9830 {
9831 	struct i40e_veb *veb;
9832 	struct i40e_pf *pf = vsi->back;
9833 
9834 	/* Uplink is not a bridge so default to VEB */
9835 	if (vsi->veb_idx == I40E_NO_VEB)
9836 		return 1;
9837 
9838 	veb = pf->veb[vsi->veb_idx];
9839 	if (!veb) {
9840 		dev_info(&pf->pdev->dev,
9841 			 "There is no veb associated with the bridge\n");
9842 		return -ENOENT;
9843 	}
9844 
9845 	/* Uplink is a bridge in VEPA mode */
9846 	if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9847 		return 0;
9848 	} else {
9849 		/* Uplink is a bridge in VEB mode */
9850 		return 1;
9851 	}
9852 
9853 	/* VEPA is now default bridge, so return 0 */
9854 	return 0;
9855 }
9856 
9857 /**
9858  * i40e_add_vsi - Add a VSI to the switch
9859  * @vsi: the VSI being configured
9860  *
9861  * This initializes a VSI context depending on the VSI type to be added and
9862  * passes it down to the add_vsi aq command.
9863  **/
9864 static int i40e_add_vsi(struct i40e_vsi *vsi)
9865 {
9866 	int ret = -ENODEV;
9867 	struct i40e_pf *pf = vsi->back;
9868 	struct i40e_hw *hw = &pf->hw;
9869 	struct i40e_vsi_context ctxt;
9870 	struct i40e_mac_filter *f;
9871 	struct hlist_node *h;
9872 	int bkt;
9873 
9874 	u8 enabled_tc = 0x1; /* TC0 enabled */
9875 	int f_count = 0;
9876 
9877 	memset(&ctxt, 0, sizeof(ctxt));
9878 	switch (vsi->type) {
9879 	case I40E_VSI_MAIN:
9880 		/* The PF's main VSI is already setup as part of the
9881 		 * device initialization, so we'll not bother with
9882 		 * the add_vsi call, but we will retrieve the current
9883 		 * VSI context.
9884 		 */
9885 		ctxt.seid = pf->main_vsi_seid;
9886 		ctxt.pf_num = pf->hw.pf_id;
9887 		ctxt.vf_num = 0;
9888 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9889 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9890 		if (ret) {
9891 			dev_info(&pf->pdev->dev,
9892 				 "couldn't get PF vsi config, err %s aq_err %s\n",
9893 				 i40e_stat_str(&pf->hw, ret),
9894 				 i40e_aq_str(&pf->hw,
9895 					     pf->hw.aq.asq_last_status));
9896 			return -ENOENT;
9897 		}
9898 		vsi->info = ctxt.info;
9899 		vsi->info.valid_sections = 0;
9900 
9901 		vsi->seid = ctxt.seid;
9902 		vsi->id = ctxt.vsi_number;
9903 
9904 		enabled_tc = i40e_pf_get_tc_map(pf);
9905 
9906 		/* MFP mode setup queue map and update VSI */
9907 		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9908 		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9909 			memset(&ctxt, 0, sizeof(ctxt));
9910 			ctxt.seid = pf->main_vsi_seid;
9911 			ctxt.pf_num = pf->hw.pf_id;
9912 			ctxt.vf_num = 0;
9913 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9914 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9915 			if (ret) {
9916 				dev_info(&pf->pdev->dev,
9917 					 "update vsi failed, err %s aq_err %s\n",
9918 					 i40e_stat_str(&pf->hw, ret),
9919 					 i40e_aq_str(&pf->hw,
9920 						    pf->hw.aq.asq_last_status));
9921 				ret = -ENOENT;
9922 				goto err;
9923 			}
9924 			/* update the local VSI info queue map */
9925 			i40e_vsi_update_queue_map(vsi, &ctxt);
9926 			vsi->info.valid_sections = 0;
9927 		} else {
9928 			/* Default/Main VSI is only enabled for TC0
9929 			 * reconfigure it to enable all TCs that are
9930 			 * available on the port in SFP mode.
9931 			 * For MFP case the iSCSI PF would use this
9932 			 * flow to enable LAN+iSCSI TC.
9933 			 */
9934 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
9935 			if (ret) {
9936 				/* Single TC condition is not fatal,
9937 				 * message and continue
9938 				 */
9939 				dev_info(&pf->pdev->dev,
9940 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9941 					 enabled_tc,
9942 					 i40e_stat_str(&pf->hw, ret),
9943 					 i40e_aq_str(&pf->hw,
9944 						    pf->hw.aq.asq_last_status));
9945 			}
9946 		}
9947 		break;
9948 
9949 	case I40E_VSI_FDIR:
9950 		ctxt.pf_num = hw->pf_id;
9951 		ctxt.vf_num = 0;
9952 		ctxt.uplink_seid = vsi->uplink_seid;
9953 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9954 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9955 		if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9956 		    (i40e_is_vsi_uplink_mode_veb(vsi))) {
9957 			ctxt.info.valid_sections |=
9958 			     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9959 			ctxt.info.switch_id =
9960 			   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9961 		}
9962 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9963 		break;
9964 
9965 	case I40E_VSI_VMDQ2:
9966 		ctxt.pf_num = hw->pf_id;
9967 		ctxt.vf_num = 0;
9968 		ctxt.uplink_seid = vsi->uplink_seid;
9969 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9970 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9971 
9972 		/* This VSI is connected to VEB so the switch_id
9973 		 * should be set to zero by default.
9974 		 */
9975 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9976 			ctxt.info.valid_sections |=
9977 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9978 			ctxt.info.switch_id =
9979 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9980 		}
9981 
9982 		/* Setup the VSI tx/rx queue map for TC0 only for now */
9983 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9984 		break;
9985 
9986 	case I40E_VSI_SRIOV:
9987 		ctxt.pf_num = hw->pf_id;
9988 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9989 		ctxt.uplink_seid = vsi->uplink_seid;
9990 		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9991 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9992 
9993 		/* This VSI is connected to VEB so the switch_id
9994 		 * should be set to zero by default.
9995 		 */
9996 		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9997 			ctxt.info.valid_sections |=
9998 				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9999 			ctxt.info.switch_id =
10000 				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
10001 		}
10002 
10003 		if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
10004 			ctxt.info.valid_sections |=
10005 				cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
10006 			ctxt.info.queueing_opt_flags |=
10007 				(I40E_AQ_VSI_QUE_OPT_TCP_ENA |
10008 				 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
10009 		}
10010 
10011 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
10012 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
10013 		if (pf->vf[vsi->vf_id].spoofchk) {
10014 			ctxt.info.valid_sections |=
10015 				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
10016 			ctxt.info.sec_flags |=
10017 				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
10018 				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
10019 		}
10020 		/* Setup the VSI tx/rx queue map for TC0 only for now */
10021 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
10022 		break;
10023 
10024 	case I40E_VSI_IWARP:
10025 		/* send down message to iWARP */
10026 		break;
10027 
10028 	default:
10029 		return -ENODEV;
10030 	}
10031 
10032 	if (vsi->type != I40E_VSI_MAIN) {
10033 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
10034 		if (ret) {
10035 			dev_info(&vsi->back->pdev->dev,
10036 				 "add vsi failed, err %s aq_err %s\n",
10037 				 i40e_stat_str(&pf->hw, ret),
10038 				 i40e_aq_str(&pf->hw,
10039 					     pf->hw.aq.asq_last_status));
10040 			ret = -ENOENT;
10041 			goto err;
10042 		}
10043 		vsi->info = ctxt.info;
10044 		vsi->info.valid_sections = 0;
10045 		vsi->seid = ctxt.seid;
10046 		vsi->id = ctxt.vsi_number;
10047 	}
10048 
10049 	vsi->active_filters = 0;
10050 	clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
10051 	spin_lock_bh(&vsi->mac_filter_hash_lock);
10052 	/* If macvlan filters already exist, force them to get loaded */
10053 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
10054 		f->state = I40E_FILTER_NEW;
10055 		f_count++;
10056 	}
10057 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
10058 
10059 	if (f_count) {
10060 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
10061 		pf->flags |= I40E_FLAG_FILTER_SYNC;
10062 	}
10063 
10064 	/* Update VSI BW information */
10065 	ret = i40e_vsi_get_bw_info(vsi);
10066 	if (ret) {
10067 		dev_info(&pf->pdev->dev,
10068 			 "couldn't get vsi bw info, err %s aq_err %s\n",
10069 			 i40e_stat_str(&pf->hw, ret),
10070 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10071 		/* VSI is already added so not tearing that up */
10072 		ret = 0;
10073 	}
10074 
10075 err:
10076 	return ret;
10077 }
10078 
10079 /**
10080  * i40e_vsi_release - Delete a VSI and free its resources
10081  * @vsi: the VSI being removed
10082  *
10083  * Returns 0 on success or < 0 on error
10084  **/
10085 int i40e_vsi_release(struct i40e_vsi *vsi)
10086 {
10087 	struct i40e_mac_filter *f;
10088 	struct hlist_node *h;
10089 	struct i40e_veb *veb = NULL;
10090 	struct i40e_pf *pf;
10091 	u16 uplink_seid;
10092 	int i, n, bkt;
10093 
10094 	pf = vsi->back;
10095 
10096 	/* release of a VEB-owner or last VSI is not allowed */
10097 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
10098 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
10099 			 vsi->seid, vsi->uplink_seid);
10100 		return -ENODEV;
10101 	}
10102 	if (vsi == pf->vsi[pf->lan_vsi] &&
10103 	    !test_bit(__I40E_DOWN, pf->state)) {
10104 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
10105 		return -ENODEV;
10106 	}
10107 
10108 	uplink_seid = vsi->uplink_seid;
10109 	if (vsi->type != I40E_VSI_SRIOV) {
10110 		if (vsi->netdev_registered) {
10111 			vsi->netdev_registered = false;
10112 			if (vsi->netdev) {
10113 				/* results in a call to i40e_close() */
10114 				unregister_netdev(vsi->netdev);
10115 			}
10116 		} else {
10117 			i40e_vsi_close(vsi);
10118 		}
10119 		i40e_vsi_disable_irq(vsi);
10120 	}
10121 
10122 	spin_lock_bh(&vsi->mac_filter_hash_lock);
10123 
10124 	/* clear the sync flag on all filters */
10125 	if (vsi->netdev) {
10126 		__dev_uc_unsync(vsi->netdev, NULL);
10127 		__dev_mc_unsync(vsi->netdev, NULL);
10128 	}
10129 
10130 	/* make sure any remaining filters are marked for deletion */
10131 	hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
10132 		__i40e_del_filter(vsi, f);
10133 
10134 	spin_unlock_bh(&vsi->mac_filter_hash_lock);
10135 
10136 	i40e_sync_vsi_filters(vsi);
10137 
10138 	i40e_vsi_delete(vsi);
10139 	i40e_vsi_free_q_vectors(vsi);
10140 	if (vsi->netdev) {
10141 		free_netdev(vsi->netdev);
10142 		vsi->netdev = NULL;
10143 	}
10144 	i40e_vsi_clear_rings(vsi);
10145 	i40e_vsi_clear(vsi);
10146 
10147 	/* If this was the last thing on the VEB, except for the
10148 	 * controlling VSI, remove the VEB, which puts the controlling
10149 	 * VSI onto the next level down in the switch.
10150 	 *
10151 	 * Well, okay, there's one more exception here: don't remove
10152 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
10153 	 * from up the network stack.
10154 	 */
10155 	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
10156 		if (pf->vsi[i] &&
10157 		    pf->vsi[i]->uplink_seid == uplink_seid &&
10158 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10159 			n++;      /* count the VSIs */
10160 		}
10161 	}
10162 	for (i = 0; i < I40E_MAX_VEB; i++) {
10163 		if (!pf->veb[i])
10164 			continue;
10165 		if (pf->veb[i]->uplink_seid == uplink_seid)
10166 			n++;     /* count the VEBs */
10167 		if (pf->veb[i]->seid == uplink_seid)
10168 			veb = pf->veb[i];
10169 	}
10170 	if (n == 0 && veb && veb->uplink_seid != 0)
10171 		i40e_veb_release(veb);
10172 
10173 	return 0;
10174 }
10175 
10176 /**
10177  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
10178  * @vsi: ptr to the VSI
10179  *
10180  * This should only be called after i40e_vsi_mem_alloc() which allocates the
10181  * corresponding SW VSI structure and initializes num_queue_pairs for the
10182  * newly allocated VSI.
10183  *
10184  * Returns 0 on success or negative on failure
10185  **/
10186 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
10187 {
10188 	int ret = -ENOENT;
10189 	struct i40e_pf *pf = vsi->back;
10190 
10191 	if (vsi->q_vectors[0]) {
10192 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
10193 			 vsi->seid);
10194 		return -EEXIST;
10195 	}
10196 
10197 	if (vsi->base_vector) {
10198 		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
10199 			 vsi->seid, vsi->base_vector);
10200 		return -EEXIST;
10201 	}
10202 
10203 	ret = i40e_vsi_alloc_q_vectors(vsi);
10204 	if (ret) {
10205 		dev_info(&pf->pdev->dev,
10206 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
10207 			 vsi->num_q_vectors, vsi->seid, ret);
10208 		vsi->num_q_vectors = 0;
10209 		goto vector_setup_out;
10210 	}
10211 
10212 	/* In Legacy mode, we do not have to get any other vector since we
10213 	 * piggyback on the misc/ICR0 for queue interrupts.
10214 	*/
10215 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
10216 		return ret;
10217 	if (vsi->num_q_vectors)
10218 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
10219 						 vsi->num_q_vectors, vsi->idx);
10220 	if (vsi->base_vector < 0) {
10221 		dev_info(&pf->pdev->dev,
10222 			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
10223 			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
10224 		i40e_vsi_free_q_vectors(vsi);
10225 		ret = -ENOENT;
10226 		goto vector_setup_out;
10227 	}
10228 
10229 vector_setup_out:
10230 	return ret;
10231 }
10232 
10233 /**
10234  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
10235  * @vsi: pointer to the vsi.
10236  *
10237  * This re-allocates a vsi's queue resources.
10238  *
10239  * Returns pointer to the successfully allocated and configured VSI sw struct
10240  * on success, otherwise returns NULL on failure.
10241  **/
10242 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
10243 {
10244 	u16 alloc_queue_pairs;
10245 	struct i40e_pf *pf;
10246 	u8 enabled_tc;
10247 	int ret;
10248 
10249 	if (!vsi)
10250 		return NULL;
10251 
10252 	pf = vsi->back;
10253 
10254 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
10255 	i40e_vsi_clear_rings(vsi);
10256 
10257 	i40e_vsi_free_arrays(vsi, false);
10258 	i40e_set_num_rings_in_vsi(vsi);
10259 	ret = i40e_vsi_alloc_arrays(vsi, false);
10260 	if (ret)
10261 		goto err_vsi;
10262 
10263 	alloc_queue_pairs = vsi->alloc_queue_pairs *
10264 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
10265 
10266 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
10267 	if (ret < 0) {
10268 		dev_info(&pf->pdev->dev,
10269 			 "failed to get tracking for %d queues for VSI %d err %d\n",
10270 			 alloc_queue_pairs, vsi->seid, ret);
10271 		goto err_vsi;
10272 	}
10273 	vsi->base_queue = ret;
10274 
10275 	/* Update the FW view of the VSI. Force a reset of TC and queue
10276 	 * layout configurations.
10277 	 */
10278 	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10279 	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10280 	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10281 	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10282 	if (vsi->type == I40E_VSI_MAIN)
10283 		i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
10284 
10285 	/* assign it some queues */
10286 	ret = i40e_alloc_rings(vsi);
10287 	if (ret)
10288 		goto err_rings;
10289 
10290 	/* map all of the rings to the q_vectors */
10291 	i40e_vsi_map_rings_to_vectors(vsi);
10292 	return vsi;
10293 
10294 err_rings:
10295 	i40e_vsi_free_q_vectors(vsi);
10296 	if (vsi->netdev_registered) {
10297 		vsi->netdev_registered = false;
10298 		unregister_netdev(vsi->netdev);
10299 		free_netdev(vsi->netdev);
10300 		vsi->netdev = NULL;
10301 	}
10302 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
10303 err_vsi:
10304 	i40e_vsi_clear(vsi);
10305 	return NULL;
10306 }
10307 
10308 /**
10309  * i40e_vsi_setup - Set up a VSI by a given type
10310  * @pf: board private structure
10311  * @type: VSI type
10312  * @uplink_seid: the switch element to link to
10313  * @param1: usage depends upon VSI type. For VF types, indicates VF id
10314  *
10315  * This allocates the sw VSI structure and its queue resources, then add a VSI
10316  * to the identified VEB.
10317  *
10318  * Returns pointer to the successfully allocated and configure VSI sw struct on
10319  * success, otherwise returns NULL on failure.
10320  **/
10321 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
10322 				u16 uplink_seid, u32 param1)
10323 {
10324 	struct i40e_vsi *vsi = NULL;
10325 	struct i40e_veb *veb = NULL;
10326 	u16 alloc_queue_pairs;
10327 	int ret, i;
10328 	int v_idx;
10329 
10330 	/* The requested uplink_seid must be either
10331 	 *     - the PF's port seid
10332 	 *              no VEB is needed because this is the PF
10333 	 *              or this is a Flow Director special case VSI
10334 	 *     - seid of an existing VEB
10335 	 *     - seid of a VSI that owns an existing VEB
10336 	 *     - seid of a VSI that doesn't own a VEB
10337 	 *              a new VEB is created and the VSI becomes the owner
10338 	 *     - seid of the PF VSI, which is what creates the first VEB
10339 	 *              this is a special case of the previous
10340 	 *
10341 	 * Find which uplink_seid we were given and create a new VEB if needed
10342 	 */
10343 	for (i = 0; i < I40E_MAX_VEB; i++) {
10344 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
10345 			veb = pf->veb[i];
10346 			break;
10347 		}
10348 	}
10349 
10350 	if (!veb && uplink_seid != pf->mac_seid) {
10351 
10352 		for (i = 0; i < pf->num_alloc_vsi; i++) {
10353 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
10354 				vsi = pf->vsi[i];
10355 				break;
10356 			}
10357 		}
10358 		if (!vsi) {
10359 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
10360 				 uplink_seid);
10361 			return NULL;
10362 		}
10363 
10364 		if (vsi->uplink_seid == pf->mac_seid)
10365 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
10366 					     vsi->tc_config.enabled_tc);
10367 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
10368 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
10369 					     vsi->tc_config.enabled_tc);
10370 		if (veb) {
10371 			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
10372 				dev_info(&vsi->back->pdev->dev,
10373 					 "New VSI creation error, uplink seid of LAN VSI expected.\n");
10374 				return NULL;
10375 			}
10376 			/* We come up by default in VEPA mode if SRIOV is not
10377 			 * already enabled, in which case we can't force VEPA
10378 			 * mode.
10379 			 */
10380 			if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
10381 				veb->bridge_mode = BRIDGE_MODE_VEPA;
10382 				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
10383 			}
10384 			i40e_config_bridge_mode(veb);
10385 		}
10386 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
10387 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
10388 				veb = pf->veb[i];
10389 		}
10390 		if (!veb) {
10391 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
10392 			return NULL;
10393 		}
10394 
10395 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10396 		uplink_seid = veb->seid;
10397 	}
10398 
10399 	/* get vsi sw struct */
10400 	v_idx = i40e_vsi_mem_alloc(pf, type);
10401 	if (v_idx < 0)
10402 		goto err_alloc;
10403 	vsi = pf->vsi[v_idx];
10404 	if (!vsi)
10405 		goto err_alloc;
10406 	vsi->type = type;
10407 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
10408 
10409 	if (type == I40E_VSI_MAIN)
10410 		pf->lan_vsi = v_idx;
10411 	else if (type == I40E_VSI_SRIOV)
10412 		vsi->vf_id = param1;
10413 	/* assign it some queues */
10414 	alloc_queue_pairs = vsi->alloc_queue_pairs *
10415 			    (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
10416 
10417 	ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
10418 	if (ret < 0) {
10419 		dev_info(&pf->pdev->dev,
10420 			 "failed to get tracking for %d queues for VSI %d err=%d\n",
10421 			 alloc_queue_pairs, vsi->seid, ret);
10422 		goto err_vsi;
10423 	}
10424 	vsi->base_queue = ret;
10425 
10426 	/* get a VSI from the hardware */
10427 	vsi->uplink_seid = uplink_seid;
10428 	ret = i40e_add_vsi(vsi);
10429 	if (ret)
10430 		goto err_vsi;
10431 
10432 	switch (vsi->type) {
10433 	/* setup the netdev if needed */
10434 	case I40E_VSI_MAIN:
10435 	case I40E_VSI_VMDQ2:
10436 		ret = i40e_config_netdev(vsi);
10437 		if (ret)
10438 			goto err_netdev;
10439 		ret = register_netdev(vsi->netdev);
10440 		if (ret)
10441 			goto err_netdev;
10442 		vsi->netdev_registered = true;
10443 		netif_carrier_off(vsi->netdev);
10444 #ifdef CONFIG_I40E_DCB
10445 		/* Setup DCB netlink interface */
10446 		i40e_dcbnl_setup(vsi);
10447 #endif /* CONFIG_I40E_DCB */
10448 		/* fall through */
10449 
10450 	case I40E_VSI_FDIR:
10451 		/* set up vectors and rings if needed */
10452 		ret = i40e_vsi_setup_vectors(vsi);
10453 		if (ret)
10454 			goto err_msix;
10455 
10456 		ret = i40e_alloc_rings(vsi);
10457 		if (ret)
10458 			goto err_rings;
10459 
10460 		/* map all of the rings to the q_vectors */
10461 		i40e_vsi_map_rings_to_vectors(vsi);
10462 
10463 		i40e_vsi_reset_stats(vsi);
10464 		break;
10465 
10466 	default:
10467 		/* no netdev or rings for the other VSI types */
10468 		break;
10469 	}
10470 
10471 	if ((pf->hw_features & I40E_HW_RSS_AQ_CAPABLE) &&
10472 	    (vsi->type == I40E_VSI_VMDQ2)) {
10473 		ret = i40e_vsi_config_rss(vsi);
10474 	}
10475 	return vsi;
10476 
10477 err_rings:
10478 	i40e_vsi_free_q_vectors(vsi);
10479 err_msix:
10480 	if (vsi->netdev_registered) {
10481 		vsi->netdev_registered = false;
10482 		unregister_netdev(vsi->netdev);
10483 		free_netdev(vsi->netdev);
10484 		vsi->netdev = NULL;
10485 	}
10486 err_netdev:
10487 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
10488 err_vsi:
10489 	i40e_vsi_clear(vsi);
10490 err_alloc:
10491 	return NULL;
10492 }
10493 
10494 /**
10495  * i40e_veb_get_bw_info - Query VEB BW information
10496  * @veb: the veb to query
10497  *
10498  * Query the Tx scheduler BW configuration data for given VEB
10499  **/
10500 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
10501 {
10502 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
10503 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
10504 	struct i40e_pf *pf = veb->pf;
10505 	struct i40e_hw *hw = &pf->hw;
10506 	u32 tc_bw_max;
10507 	int ret = 0;
10508 	int i;
10509 
10510 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
10511 						  &bw_data, NULL);
10512 	if (ret) {
10513 		dev_info(&pf->pdev->dev,
10514 			 "query veb bw config failed, err %s aq_err %s\n",
10515 			 i40e_stat_str(&pf->hw, ret),
10516 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10517 		goto out;
10518 	}
10519 
10520 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
10521 						   &ets_data, NULL);
10522 	if (ret) {
10523 		dev_info(&pf->pdev->dev,
10524 			 "query veb bw ets config failed, err %s aq_err %s\n",
10525 			 i40e_stat_str(&pf->hw, ret),
10526 			 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10527 		goto out;
10528 	}
10529 
10530 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
10531 	veb->bw_max_quanta = ets_data.tc_bw_max;
10532 	veb->is_abs_credits = bw_data.absolute_credits_enable;
10533 	veb->enabled_tc = ets_data.tc_valid_bits;
10534 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
10535 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
10536 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
10537 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
10538 		veb->bw_tc_limit_credits[i] =
10539 					le16_to_cpu(bw_data.tc_bw_limits[i]);
10540 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
10541 	}
10542 
10543 out:
10544 	return ret;
10545 }
10546 
10547 /**
10548  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
10549  * @pf: board private structure
10550  *
10551  * On error: returns error code (negative)
10552  * On success: returns vsi index in PF (positive)
10553  **/
10554 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
10555 {
10556 	int ret = -ENOENT;
10557 	struct i40e_veb *veb;
10558 	int i;
10559 
10560 	/* Need to protect the allocation of switch elements at the PF level */
10561 	mutex_lock(&pf->switch_mutex);
10562 
10563 	/* VEB list may be fragmented if VEB creation/destruction has
10564 	 * been happening.  We can afford to do a quick scan to look
10565 	 * for any free slots in the list.
10566 	 *
10567 	 * find next empty veb slot, looping back around if necessary
10568 	 */
10569 	i = 0;
10570 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
10571 		i++;
10572 	if (i >= I40E_MAX_VEB) {
10573 		ret = -ENOMEM;
10574 		goto err_alloc_veb;  /* out of VEB slots! */
10575 	}
10576 
10577 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
10578 	if (!veb) {
10579 		ret = -ENOMEM;
10580 		goto err_alloc_veb;
10581 	}
10582 	veb->pf = pf;
10583 	veb->idx = i;
10584 	veb->enabled_tc = 1;
10585 
10586 	pf->veb[i] = veb;
10587 	ret = i;
10588 err_alloc_veb:
10589 	mutex_unlock(&pf->switch_mutex);
10590 	return ret;
10591 }
10592 
10593 /**
10594  * i40e_switch_branch_release - Delete a branch of the switch tree
10595  * @branch: where to start deleting
10596  *
10597  * This uses recursion to find the tips of the branch to be
10598  * removed, deleting until we get back to and can delete this VEB.
10599  **/
10600 static void i40e_switch_branch_release(struct i40e_veb *branch)
10601 {
10602 	struct i40e_pf *pf = branch->pf;
10603 	u16 branch_seid = branch->seid;
10604 	u16 veb_idx = branch->idx;
10605 	int i;
10606 
10607 	/* release any VEBs on this VEB - RECURSION */
10608 	for (i = 0; i < I40E_MAX_VEB; i++) {
10609 		if (!pf->veb[i])
10610 			continue;
10611 		if (pf->veb[i]->uplink_seid == branch->seid)
10612 			i40e_switch_branch_release(pf->veb[i]);
10613 	}
10614 
10615 	/* Release the VSIs on this VEB, but not the owner VSI.
10616 	 *
10617 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10618 	 *       the VEB itself, so don't use (*branch) after this loop.
10619 	 */
10620 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10621 		if (!pf->vsi[i])
10622 			continue;
10623 		if (pf->vsi[i]->uplink_seid == branch_seid &&
10624 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10625 			i40e_vsi_release(pf->vsi[i]);
10626 		}
10627 	}
10628 
10629 	/* There's one corner case where the VEB might not have been
10630 	 * removed, so double check it here and remove it if needed.
10631 	 * This case happens if the veb was created from the debugfs
10632 	 * commands and no VSIs were added to it.
10633 	 */
10634 	if (pf->veb[veb_idx])
10635 		i40e_veb_release(pf->veb[veb_idx]);
10636 }
10637 
10638 /**
10639  * i40e_veb_clear - remove veb struct
10640  * @veb: the veb to remove
10641  **/
10642 static void i40e_veb_clear(struct i40e_veb *veb)
10643 {
10644 	if (!veb)
10645 		return;
10646 
10647 	if (veb->pf) {
10648 		struct i40e_pf *pf = veb->pf;
10649 
10650 		mutex_lock(&pf->switch_mutex);
10651 		if (pf->veb[veb->idx] == veb)
10652 			pf->veb[veb->idx] = NULL;
10653 		mutex_unlock(&pf->switch_mutex);
10654 	}
10655 
10656 	kfree(veb);
10657 }
10658 
10659 /**
10660  * i40e_veb_release - Delete a VEB and free its resources
10661  * @veb: the VEB being removed
10662  **/
10663 void i40e_veb_release(struct i40e_veb *veb)
10664 {
10665 	struct i40e_vsi *vsi = NULL;
10666 	struct i40e_pf *pf;
10667 	int i, n = 0;
10668 
10669 	pf = veb->pf;
10670 
10671 	/* find the remaining VSI and check for extras */
10672 	for (i = 0; i < pf->num_alloc_vsi; i++) {
10673 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10674 			n++;
10675 			vsi = pf->vsi[i];
10676 		}
10677 	}
10678 	if (n != 1) {
10679 		dev_info(&pf->pdev->dev,
10680 			 "can't remove VEB %d with %d VSIs left\n",
10681 			 veb->seid, n);
10682 		return;
10683 	}
10684 
10685 	/* move the remaining VSI to uplink veb */
10686 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10687 	if (veb->uplink_seid) {
10688 		vsi->uplink_seid = veb->uplink_seid;
10689 		if (veb->uplink_seid == pf->mac_seid)
10690 			vsi->veb_idx = I40E_NO_VEB;
10691 		else
10692 			vsi->veb_idx = veb->veb_idx;
10693 	} else {
10694 		/* floating VEB */
10695 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10696 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10697 	}
10698 
10699 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10700 	i40e_veb_clear(veb);
10701 }
10702 
10703 /**
10704  * i40e_add_veb - create the VEB in the switch
10705  * @veb: the VEB to be instantiated
10706  * @vsi: the controlling VSI
10707  **/
10708 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10709 {
10710 	struct i40e_pf *pf = veb->pf;
10711 	bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10712 	int ret;
10713 
10714 	ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10715 			      veb->enabled_tc, false,
10716 			      &veb->seid, enable_stats, NULL);
10717 
10718 	/* get a VEB from the hardware */
10719 	if (ret) {
10720 		dev_info(&pf->pdev->dev,
10721 			 "couldn't add VEB, err %s aq_err %s\n",
10722 			 i40e_stat_str(&pf->hw, ret),
10723 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10724 		return -EPERM;
10725 	}
10726 
10727 	/* get statistics counter */
10728 	ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10729 					 &veb->stats_idx, NULL, NULL, NULL);
10730 	if (ret) {
10731 		dev_info(&pf->pdev->dev,
10732 			 "couldn't get VEB statistics idx, err %s aq_err %s\n",
10733 			 i40e_stat_str(&pf->hw, ret),
10734 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10735 		return -EPERM;
10736 	}
10737 	ret = i40e_veb_get_bw_info(veb);
10738 	if (ret) {
10739 		dev_info(&pf->pdev->dev,
10740 			 "couldn't get VEB bw info, err %s aq_err %s\n",
10741 			 i40e_stat_str(&pf->hw, ret),
10742 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10743 		i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10744 		return -ENOENT;
10745 	}
10746 
10747 	vsi->uplink_seid = veb->seid;
10748 	vsi->veb_idx = veb->idx;
10749 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10750 
10751 	return 0;
10752 }
10753 
10754 /**
10755  * i40e_veb_setup - Set up a VEB
10756  * @pf: board private structure
10757  * @flags: VEB setup flags
10758  * @uplink_seid: the switch element to link to
10759  * @vsi_seid: the initial VSI seid
10760  * @enabled_tc: Enabled TC bit-map
10761  *
10762  * This allocates the sw VEB structure and links it into the switch
10763  * It is possible and legal for this to be a duplicate of an already
10764  * existing VEB.  It is also possible for both uplink and vsi seids
10765  * to be zero, in order to create a floating VEB.
10766  *
10767  * Returns pointer to the successfully allocated VEB sw struct on
10768  * success, otherwise returns NULL on failure.
10769  **/
10770 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10771 				u16 uplink_seid, u16 vsi_seid,
10772 				u8 enabled_tc)
10773 {
10774 	struct i40e_veb *veb, *uplink_veb = NULL;
10775 	int vsi_idx, veb_idx;
10776 	int ret;
10777 
10778 	/* if one seid is 0, the other must be 0 to create a floating relay */
10779 	if ((uplink_seid == 0 || vsi_seid == 0) &&
10780 	    (uplink_seid + vsi_seid != 0)) {
10781 		dev_info(&pf->pdev->dev,
10782 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
10783 			 uplink_seid, vsi_seid);
10784 		return NULL;
10785 	}
10786 
10787 	/* make sure there is such a vsi and uplink */
10788 	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10789 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10790 			break;
10791 	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10792 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10793 			 vsi_seid);
10794 		return NULL;
10795 	}
10796 
10797 	if (uplink_seid && uplink_seid != pf->mac_seid) {
10798 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10799 			if (pf->veb[veb_idx] &&
10800 			    pf->veb[veb_idx]->seid == uplink_seid) {
10801 				uplink_veb = pf->veb[veb_idx];
10802 				break;
10803 			}
10804 		}
10805 		if (!uplink_veb) {
10806 			dev_info(&pf->pdev->dev,
10807 				 "uplink seid %d not found\n", uplink_seid);
10808 			return NULL;
10809 		}
10810 	}
10811 
10812 	/* get veb sw struct */
10813 	veb_idx = i40e_veb_mem_alloc(pf);
10814 	if (veb_idx < 0)
10815 		goto err_alloc;
10816 	veb = pf->veb[veb_idx];
10817 	veb->flags = flags;
10818 	veb->uplink_seid = uplink_seid;
10819 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10820 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10821 
10822 	/* create the VEB in the switch */
10823 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10824 	if (ret)
10825 		goto err_veb;
10826 	if (vsi_idx == pf->lan_vsi)
10827 		pf->lan_veb = veb->idx;
10828 
10829 	return veb;
10830 
10831 err_veb:
10832 	i40e_veb_clear(veb);
10833 err_alloc:
10834 	return NULL;
10835 }
10836 
10837 /**
10838  * i40e_setup_pf_switch_element - set PF vars based on switch type
10839  * @pf: board private structure
10840  * @ele: element we are building info from
10841  * @num_reported: total number of elements
10842  * @printconfig: should we print the contents
10843  *
10844  * helper function to assist in extracting a few useful SEID values.
10845  **/
10846 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10847 				struct i40e_aqc_switch_config_element_resp *ele,
10848 				u16 num_reported, bool printconfig)
10849 {
10850 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10851 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10852 	u8 element_type = ele->element_type;
10853 	u16 seid = le16_to_cpu(ele->seid);
10854 
10855 	if (printconfig)
10856 		dev_info(&pf->pdev->dev,
10857 			 "type=%d seid=%d uplink=%d downlink=%d\n",
10858 			 element_type, seid, uplink_seid, downlink_seid);
10859 
10860 	switch (element_type) {
10861 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
10862 		pf->mac_seid = seid;
10863 		break;
10864 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
10865 		/* Main VEB? */
10866 		if (uplink_seid != pf->mac_seid)
10867 			break;
10868 		if (pf->lan_veb == I40E_NO_VEB) {
10869 			int v;
10870 
10871 			/* find existing or else empty VEB */
10872 			for (v = 0; v < I40E_MAX_VEB; v++) {
10873 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10874 					pf->lan_veb = v;
10875 					break;
10876 				}
10877 			}
10878 			if (pf->lan_veb == I40E_NO_VEB) {
10879 				v = i40e_veb_mem_alloc(pf);
10880 				if (v < 0)
10881 					break;
10882 				pf->lan_veb = v;
10883 			}
10884 		}
10885 
10886 		pf->veb[pf->lan_veb]->seid = seid;
10887 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10888 		pf->veb[pf->lan_veb]->pf = pf;
10889 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10890 		break;
10891 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
10892 		if (num_reported != 1)
10893 			break;
10894 		/* This is immediately after a reset so we can assume this is
10895 		 * the PF's VSI
10896 		 */
10897 		pf->mac_seid = uplink_seid;
10898 		pf->pf_seid = downlink_seid;
10899 		pf->main_vsi_seid = seid;
10900 		if (printconfig)
10901 			dev_info(&pf->pdev->dev,
10902 				 "pf_seid=%d main_vsi_seid=%d\n",
10903 				 pf->pf_seid, pf->main_vsi_seid);
10904 		break;
10905 	case I40E_SWITCH_ELEMENT_TYPE_PF:
10906 	case I40E_SWITCH_ELEMENT_TYPE_VF:
10907 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
10908 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
10909 	case I40E_SWITCH_ELEMENT_TYPE_PE:
10910 	case I40E_SWITCH_ELEMENT_TYPE_PA:
10911 		/* ignore these for now */
10912 		break;
10913 	default:
10914 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10915 			 element_type, seid);
10916 		break;
10917 	}
10918 }
10919 
10920 /**
10921  * i40e_fetch_switch_configuration - Get switch config from firmware
10922  * @pf: board private structure
10923  * @printconfig: should we print the contents
10924  *
10925  * Get the current switch configuration from the device and
10926  * extract a few useful SEID values.
10927  **/
10928 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10929 {
10930 	struct i40e_aqc_get_switch_config_resp *sw_config;
10931 	u16 next_seid = 0;
10932 	int ret = 0;
10933 	u8 *aq_buf;
10934 	int i;
10935 
10936 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10937 	if (!aq_buf)
10938 		return -ENOMEM;
10939 
10940 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10941 	do {
10942 		u16 num_reported, num_total;
10943 
10944 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10945 						I40E_AQ_LARGE_BUF,
10946 						&next_seid, NULL);
10947 		if (ret) {
10948 			dev_info(&pf->pdev->dev,
10949 				 "get switch config failed err %s aq_err %s\n",
10950 				 i40e_stat_str(&pf->hw, ret),
10951 				 i40e_aq_str(&pf->hw,
10952 					     pf->hw.aq.asq_last_status));
10953 			kfree(aq_buf);
10954 			return -ENOENT;
10955 		}
10956 
10957 		num_reported = le16_to_cpu(sw_config->header.num_reported);
10958 		num_total = le16_to_cpu(sw_config->header.num_total);
10959 
10960 		if (printconfig)
10961 			dev_info(&pf->pdev->dev,
10962 				 "header: %d reported %d total\n",
10963 				 num_reported, num_total);
10964 
10965 		for (i = 0; i < num_reported; i++) {
10966 			struct i40e_aqc_switch_config_element_resp *ele =
10967 				&sw_config->element[i];
10968 
10969 			i40e_setup_pf_switch_element(pf, ele, num_reported,
10970 						     printconfig);
10971 		}
10972 	} while (next_seid != 0);
10973 
10974 	kfree(aq_buf);
10975 	return ret;
10976 }
10977 
10978 /**
10979  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10980  * @pf: board private structure
10981  * @reinit: if the Main VSI needs to re-initialized.
10982  *
10983  * Returns 0 on success, negative value on failure
10984  **/
10985 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10986 {
10987 	u16 flags = 0;
10988 	int ret;
10989 
10990 	/* find out what's out there already */
10991 	ret = i40e_fetch_switch_configuration(pf, false);
10992 	if (ret) {
10993 		dev_info(&pf->pdev->dev,
10994 			 "couldn't fetch switch config, err %s aq_err %s\n",
10995 			 i40e_stat_str(&pf->hw, ret),
10996 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10997 		return ret;
10998 	}
10999 	i40e_pf_reset_stats(pf);
11000 
11001 	/* set the switch config bit for the whole device to
11002 	 * support limited promisc or true promisc
11003 	 * when user requests promisc. The default is limited
11004 	 * promisc.
11005 	*/
11006 
11007 	if ((pf->hw.pf_id == 0) &&
11008 	    !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
11009 		flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
11010 
11011 	if (pf->hw.pf_id == 0) {
11012 		u16 valid_flags;
11013 
11014 		valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
11015 		ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
11016 						NULL);
11017 		if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
11018 			dev_info(&pf->pdev->dev,
11019 				 "couldn't set switch config bits, err %s aq_err %s\n",
11020 				 i40e_stat_str(&pf->hw, ret),
11021 				 i40e_aq_str(&pf->hw,
11022 					     pf->hw.aq.asq_last_status));
11023 			/* not a fatal problem, just keep going */
11024 		}
11025 	}
11026 
11027 	/* first time setup */
11028 	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
11029 		struct i40e_vsi *vsi = NULL;
11030 		u16 uplink_seid;
11031 
11032 		/* Set up the PF VSI associated with the PF's main VSI
11033 		 * that is already in the HW switch
11034 		 */
11035 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
11036 			uplink_seid = pf->veb[pf->lan_veb]->seid;
11037 		else
11038 			uplink_seid = pf->mac_seid;
11039 		if (pf->lan_vsi == I40E_NO_VSI)
11040 			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
11041 		else if (reinit)
11042 			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
11043 		if (!vsi) {
11044 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
11045 			i40e_fdir_teardown(pf);
11046 			return -EAGAIN;
11047 		}
11048 	} else {
11049 		/* force a reset of TC and queue layout configurations */
11050 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
11051 
11052 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
11053 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
11054 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
11055 	}
11056 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
11057 
11058 	i40e_fdir_sb_setup(pf);
11059 
11060 	/* Setup static PF queue filter control settings */
11061 	ret = i40e_setup_pf_filter_control(pf);
11062 	if (ret) {
11063 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
11064 			 ret);
11065 		/* Failure here should not stop continuing other steps */
11066 	}
11067 
11068 	/* enable RSS in the HW, even for only one queue, as the stack can use
11069 	 * the hash
11070 	 */
11071 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
11072 		i40e_pf_config_rss(pf);
11073 
11074 	/* fill in link information and enable LSE reporting */
11075 	i40e_link_event(pf);
11076 
11077 	/* Initialize user-specific link properties */
11078 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
11079 				  I40E_AQ_AN_COMPLETED) ? true : false);
11080 
11081 	i40e_ptp_init(pf);
11082 
11083 	/* repopulate tunnel port filters */
11084 	i40e_sync_udp_filters(pf);
11085 
11086 	return ret;
11087 }
11088 
11089 /**
11090  * i40e_determine_queue_usage - Work out queue distribution
11091  * @pf: board private structure
11092  **/
11093 static void i40e_determine_queue_usage(struct i40e_pf *pf)
11094 {
11095 	int queues_left;
11096 	int q_max;
11097 
11098 	pf->num_lan_qps = 0;
11099 
11100 	/* Find the max queues to be put into basic use.  We'll always be
11101 	 * using TC0, whether or not DCB is running, and TC0 will get the
11102 	 * big RSS set.
11103 	 */
11104 	queues_left = pf->hw.func_caps.num_tx_qp;
11105 
11106 	if ((queues_left == 1) ||
11107 	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
11108 		/* one qp for PF, no queues for anything else */
11109 		queues_left = 0;
11110 		pf->alloc_rss_size = pf->num_lan_qps = 1;
11111 
11112 		/* make sure all the fancies are disabled */
11113 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
11114 			       I40E_FLAG_IWARP_ENABLED	|
11115 			       I40E_FLAG_FD_SB_ENABLED	|
11116 			       I40E_FLAG_FD_ATR_ENABLED	|
11117 			       I40E_FLAG_DCB_CAPABLE	|
11118 			       I40E_FLAG_DCB_ENABLED	|
11119 			       I40E_FLAG_SRIOV_ENABLED	|
11120 			       I40E_FLAG_VMDQ_ENABLED);
11121 	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
11122 				  I40E_FLAG_FD_SB_ENABLED |
11123 				  I40E_FLAG_FD_ATR_ENABLED |
11124 				  I40E_FLAG_DCB_CAPABLE))) {
11125 		/* one qp for PF */
11126 		pf->alloc_rss_size = pf->num_lan_qps = 1;
11127 		queues_left -= pf->num_lan_qps;
11128 
11129 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
11130 			       I40E_FLAG_IWARP_ENABLED	|
11131 			       I40E_FLAG_FD_SB_ENABLED	|
11132 			       I40E_FLAG_FD_ATR_ENABLED	|
11133 			       I40E_FLAG_DCB_ENABLED	|
11134 			       I40E_FLAG_VMDQ_ENABLED);
11135 	} else {
11136 		/* Not enough queues for all TCs */
11137 		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
11138 		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
11139 			pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
11140 					I40E_FLAG_DCB_ENABLED);
11141 			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
11142 		}
11143 
11144 		/* limit lan qps to the smaller of qps, cpus or msix */
11145 		q_max = max_t(int, pf->rss_size_max, num_online_cpus());
11146 		q_max = min_t(int, q_max, pf->hw.func_caps.num_tx_qp);
11147 		q_max = min_t(int, q_max, pf->hw.func_caps.num_msix_vectors);
11148 		pf->num_lan_qps = q_max;
11149 
11150 		queues_left -= pf->num_lan_qps;
11151 	}
11152 
11153 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11154 		if (queues_left > 1) {
11155 			queues_left -= 1; /* save 1 queue for FD */
11156 		} else {
11157 			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
11158 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
11159 		}
11160 	}
11161 
11162 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11163 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
11164 		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
11165 					(queues_left / pf->num_vf_qps));
11166 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
11167 	}
11168 
11169 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
11170 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
11171 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
11172 					  (queues_left / pf->num_vmdq_qps));
11173 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
11174 	}
11175 
11176 	pf->queues_left = queues_left;
11177 	dev_dbg(&pf->pdev->dev,
11178 		"qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
11179 		pf->hw.func_caps.num_tx_qp,
11180 		!!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
11181 		pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
11182 		pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
11183 		queues_left);
11184 }
11185 
11186 /**
11187  * i40e_setup_pf_filter_control - Setup PF static filter control
11188  * @pf: PF to be setup
11189  *
11190  * i40e_setup_pf_filter_control sets up a PF's initial filter control
11191  * settings. If PE/FCoE are enabled then it will also set the per PF
11192  * based filter sizes required for them. It also enables Flow director,
11193  * ethertype and macvlan type filter settings for the pf.
11194  *
11195  * Returns 0 on success, negative on failure
11196  **/
11197 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
11198 {
11199 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
11200 
11201 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
11202 
11203 	/* Flow Director is enabled */
11204 	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
11205 		settings->enable_fdir = true;
11206 
11207 	/* Ethtype and MACVLAN filters enabled for PF */
11208 	settings->enable_ethtype = true;
11209 	settings->enable_macvlan = true;
11210 
11211 	if (i40e_set_filter_control(&pf->hw, settings))
11212 		return -ENOENT;
11213 
11214 	return 0;
11215 }
11216 
11217 #define INFO_STRING_LEN 255
11218 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
11219 static void i40e_print_features(struct i40e_pf *pf)
11220 {
11221 	struct i40e_hw *hw = &pf->hw;
11222 	char *buf;
11223 	int i;
11224 
11225 	buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
11226 	if (!buf)
11227 		return;
11228 
11229 	i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
11230 #ifdef CONFIG_PCI_IOV
11231 	i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
11232 #endif
11233 	i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
11234 		      pf->hw.func_caps.num_vsis,
11235 		      pf->vsi[pf->lan_vsi]->num_queue_pairs);
11236 	if (pf->flags & I40E_FLAG_RSS_ENABLED)
11237 		i += snprintf(&buf[i], REMAIN(i), " RSS");
11238 	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
11239 		i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
11240 	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11241 		i += snprintf(&buf[i], REMAIN(i), " FD_SB");
11242 		i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
11243 	}
11244 	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
11245 		i += snprintf(&buf[i], REMAIN(i), " DCB");
11246 	i += snprintf(&buf[i], REMAIN(i), " VxLAN");
11247 	i += snprintf(&buf[i], REMAIN(i), " Geneve");
11248 	if (pf->flags & I40E_FLAG_PTP)
11249 		i += snprintf(&buf[i], REMAIN(i), " PTP");
11250 	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
11251 		i += snprintf(&buf[i], REMAIN(i), " VEB");
11252 	else
11253 		i += snprintf(&buf[i], REMAIN(i), " VEPA");
11254 
11255 	dev_info(&pf->pdev->dev, "%s\n", buf);
11256 	kfree(buf);
11257 	WARN_ON(i > INFO_STRING_LEN);
11258 }
11259 
11260 /**
11261  * i40e_get_platform_mac_addr - get platform-specific MAC address
11262  * @pdev: PCI device information struct
11263  * @pf: board private structure
11264  *
11265  * Look up the MAC address for the device. First we'll try
11266  * eth_platform_get_mac_address, which will check Open Firmware, or arch
11267  * specific fallback. Otherwise, we'll default to the stored value in
11268  * firmware.
11269  **/
11270 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
11271 {
11272 	if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
11273 		i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
11274 }
11275 
11276 /**
11277  * i40e_probe - Device initialization routine
11278  * @pdev: PCI device information struct
11279  * @ent: entry in i40e_pci_tbl
11280  *
11281  * i40e_probe initializes a PF identified by a pci_dev structure.
11282  * The OS initialization, configuring of the PF private structure,
11283  * and a hardware reset occur.
11284  *
11285  * Returns 0 on success, negative on failure
11286  **/
11287 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
11288 {
11289 	struct i40e_aq_get_phy_abilities_resp abilities;
11290 	struct i40e_pf *pf;
11291 	struct i40e_hw *hw;
11292 	static u16 pfs_found;
11293 	u16 wol_nvm_bits;
11294 	u16 link_status;
11295 	int err;
11296 	u32 val;
11297 	u32 i;
11298 	u8 set_fc_aq_fail;
11299 
11300 	err = pci_enable_device_mem(pdev);
11301 	if (err)
11302 		return err;
11303 
11304 	/* set up for high or low dma */
11305 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
11306 	if (err) {
11307 		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
11308 		if (err) {
11309 			dev_err(&pdev->dev,
11310 				"DMA configuration failed: 0x%x\n", err);
11311 			goto err_dma;
11312 		}
11313 	}
11314 
11315 	/* set up pci connections */
11316 	err = pci_request_mem_regions(pdev, i40e_driver_name);
11317 	if (err) {
11318 		dev_info(&pdev->dev,
11319 			 "pci_request_selected_regions failed %d\n", err);
11320 		goto err_pci_reg;
11321 	}
11322 
11323 	pci_enable_pcie_error_reporting(pdev);
11324 	pci_set_master(pdev);
11325 
11326 	/* Now that we have a PCI connection, we need to do the
11327 	 * low level device setup.  This is primarily setting up
11328 	 * the Admin Queue structures and then querying for the
11329 	 * device's current profile information.
11330 	 */
11331 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
11332 	if (!pf) {
11333 		err = -ENOMEM;
11334 		goto err_pf_alloc;
11335 	}
11336 	pf->next_vsi = 0;
11337 	pf->pdev = pdev;
11338 	set_bit(__I40E_DOWN, pf->state);
11339 
11340 	hw = &pf->hw;
11341 	hw->back = pf;
11342 
11343 	pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
11344 				I40E_MAX_CSR_SPACE);
11345 
11346 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
11347 	if (!hw->hw_addr) {
11348 		err = -EIO;
11349 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
11350 			 (unsigned int)pci_resource_start(pdev, 0),
11351 			 pf->ioremap_len, err);
11352 		goto err_ioremap;
11353 	}
11354 	hw->vendor_id = pdev->vendor;
11355 	hw->device_id = pdev->device;
11356 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
11357 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
11358 	hw->subsystem_device_id = pdev->subsystem_device;
11359 	hw->bus.device = PCI_SLOT(pdev->devfn);
11360 	hw->bus.func = PCI_FUNC(pdev->devfn);
11361 	hw->bus.bus_id = pdev->bus->number;
11362 	pf->instance = pfs_found;
11363 
11364 	/* Select something other than the 802.1ad ethertype for the
11365 	 * switch to use internally and drop on ingress.
11366 	 */
11367 	hw->switch_tag = 0xffff;
11368 	hw->first_tag = ETH_P_8021AD;
11369 	hw->second_tag = ETH_P_8021Q;
11370 
11371 	INIT_LIST_HEAD(&pf->l3_flex_pit_list);
11372 	INIT_LIST_HEAD(&pf->l4_flex_pit_list);
11373 
11374 	/* set up the locks for the AQ, do this only once in probe
11375 	 * and destroy them only once in remove
11376 	 */
11377 	mutex_init(&hw->aq.asq_mutex);
11378 	mutex_init(&hw->aq.arq_mutex);
11379 
11380 	pf->msg_enable = netif_msg_init(debug,
11381 					NETIF_MSG_DRV |
11382 					NETIF_MSG_PROBE |
11383 					NETIF_MSG_LINK);
11384 	if (debug < -1)
11385 		pf->hw.debug_mask = debug;
11386 
11387 	/* do a special CORER for clearing PXE mode once at init */
11388 	if (hw->revision_id == 0 &&
11389 	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
11390 		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
11391 		i40e_flush(hw);
11392 		msleep(200);
11393 		pf->corer_count++;
11394 
11395 		i40e_clear_pxe_mode(hw);
11396 	}
11397 
11398 	/* Reset here to make sure all is clean and to define PF 'n' */
11399 	i40e_clear_hw(hw);
11400 	err = i40e_pf_reset(hw);
11401 	if (err) {
11402 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
11403 		goto err_pf_reset;
11404 	}
11405 	pf->pfr_count++;
11406 
11407 	hw->aq.num_arq_entries = I40E_AQ_LEN;
11408 	hw->aq.num_asq_entries = I40E_AQ_LEN;
11409 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11410 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11411 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
11412 
11413 	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
11414 		 "%s-%s:misc",
11415 		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
11416 
11417 	err = i40e_init_shared_code(hw);
11418 	if (err) {
11419 		dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
11420 			 err);
11421 		goto err_pf_reset;
11422 	}
11423 
11424 	/* set up a default setting for link flow control */
11425 	pf->hw.fc.requested_mode = I40E_FC_NONE;
11426 
11427 	err = i40e_init_adminq(hw);
11428 	if (err) {
11429 		if (err == I40E_ERR_FIRMWARE_API_VERSION)
11430 			dev_info(&pdev->dev,
11431 				 "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");
11432 		else
11433 			dev_info(&pdev->dev,
11434 				 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
11435 
11436 		goto err_pf_reset;
11437 	}
11438 	i40e_get_oem_version(hw);
11439 
11440 	/* provide nvm, fw, api versions */
11441 	dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
11442 		 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
11443 		 hw->aq.api_maj_ver, hw->aq.api_min_ver,
11444 		 i40e_nvm_version_str(hw));
11445 
11446 	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
11447 	    hw->aq.api_min_ver > I40E_FW_MINOR_VERSION(hw))
11448 		dev_info(&pdev->dev,
11449 			 "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");
11450 	else if (hw->aq.api_maj_ver == 1 && hw->aq.api_min_ver < 4)
11451 		dev_info(&pdev->dev,
11452 			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
11453 
11454 	i40e_verify_eeprom(pf);
11455 
11456 	/* Rev 0 hardware was never productized */
11457 	if (hw->revision_id < 1)
11458 		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");
11459 
11460 	i40e_clear_pxe_mode(hw);
11461 	err = i40e_get_capabilities(pf);
11462 	if (err)
11463 		goto err_adminq_setup;
11464 
11465 	err = i40e_sw_init(pf);
11466 	if (err) {
11467 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
11468 		goto err_sw_init;
11469 	}
11470 
11471 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
11472 				hw->func_caps.num_rx_qp, 0, 0);
11473 	if (err) {
11474 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
11475 		goto err_init_lan_hmc;
11476 	}
11477 
11478 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
11479 	if (err) {
11480 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
11481 		err = -ENOENT;
11482 		goto err_configure_lan_hmc;
11483 	}
11484 
11485 	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
11486 	 * Ignore error return codes because if it was already disabled via
11487 	 * hardware settings this will fail
11488 	 */
11489 	if (pf->hw_features & I40E_HW_STOP_FW_LLDP) {
11490 		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
11491 		i40e_aq_stop_lldp(hw, true, NULL);
11492 	}
11493 
11494 	/* allow a platform config to override the HW addr */
11495 	i40e_get_platform_mac_addr(pdev, pf);
11496 
11497 	if (!is_valid_ether_addr(hw->mac.addr)) {
11498 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
11499 		err = -EIO;
11500 		goto err_mac_addr;
11501 	}
11502 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
11503 	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
11504 	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
11505 	if (is_valid_ether_addr(hw->mac.port_addr))
11506 		pf->hw_features |= I40E_HW_PORT_ID_VALID;
11507 
11508 	pci_set_drvdata(pdev, pf);
11509 	pci_save_state(pdev);
11510 #ifdef CONFIG_I40E_DCB
11511 	err = i40e_init_pf_dcb(pf);
11512 	if (err) {
11513 		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
11514 		pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
11515 		/* Continue without DCB enabled */
11516 	}
11517 #endif /* CONFIG_I40E_DCB */
11518 
11519 	/* set up periodic task facility */
11520 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
11521 	pf->service_timer_period = HZ;
11522 
11523 	INIT_WORK(&pf->service_task, i40e_service_task);
11524 	clear_bit(__I40E_SERVICE_SCHED, pf->state);
11525 
11526 	/* NVM bit on means WoL disabled for the port */
11527 	i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
11528 	if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
11529 		pf->wol_en = false;
11530 	else
11531 		pf->wol_en = true;
11532 	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
11533 
11534 	/* set up the main switch operations */
11535 	i40e_determine_queue_usage(pf);
11536 	err = i40e_init_interrupt_scheme(pf);
11537 	if (err)
11538 		goto err_switch_setup;
11539 
11540 	/* The number of VSIs reported by the FW is the minimum guaranteed
11541 	 * to us; HW supports far more and we share the remaining pool with
11542 	 * the other PFs. We allocate space for more than the guarantee with
11543 	 * the understanding that we might not get them all later.
11544 	 */
11545 	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
11546 		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
11547 	else
11548 		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
11549 
11550 	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11551 	pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11552 			  GFP_KERNEL);
11553 	if (!pf->vsi) {
11554 		err = -ENOMEM;
11555 		goto err_switch_setup;
11556 	}
11557 
11558 #ifdef CONFIG_PCI_IOV
11559 	/* prep for VF support */
11560 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11561 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11562 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
11563 		if (pci_num_vf(pdev))
11564 			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11565 	}
11566 #endif
11567 	err = i40e_setup_pf_switch(pf, false);
11568 	if (err) {
11569 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11570 		goto err_vsis;
11571 	}
11572 
11573 	/* Make sure flow control is set according to current settings */
11574 	err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11575 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11576 		dev_dbg(&pf->pdev->dev,
11577 			"Set fc with err %s aq_err %s on get_phy_cap\n",
11578 			i40e_stat_str(hw, err),
11579 			i40e_aq_str(hw, hw->aq.asq_last_status));
11580 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11581 		dev_dbg(&pf->pdev->dev,
11582 			"Set fc with err %s aq_err %s on set_phy_config\n",
11583 			i40e_stat_str(hw, err),
11584 			i40e_aq_str(hw, hw->aq.asq_last_status));
11585 	if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11586 		dev_dbg(&pf->pdev->dev,
11587 			"Set fc with err %s aq_err %s on get_link_info\n",
11588 			i40e_stat_str(hw, err),
11589 			i40e_aq_str(hw, hw->aq.asq_last_status));
11590 
11591 	/* if FDIR VSI was set up, start it now */
11592 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11593 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11594 			i40e_vsi_open(pf->vsi[i]);
11595 			break;
11596 		}
11597 	}
11598 
11599 	/* The driver only wants link up/down and module qualification
11600 	 * reports from firmware.  Note the negative logic.
11601 	 */
11602 	err = i40e_aq_set_phy_int_mask(&pf->hw,
11603 				       ~(I40E_AQ_EVENT_LINK_UPDOWN |
11604 					 I40E_AQ_EVENT_MEDIA_NA |
11605 					 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11606 	if (err)
11607 		dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11608 			 i40e_stat_str(&pf->hw, err),
11609 			 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11610 
11611 	/* Reconfigure hardware for allowing smaller MSS in the case
11612 	 * of TSO, so that we avoid the MDD being fired and causing
11613 	 * a reset in the case of small MSS+TSO.
11614 	 */
11615 	val = rd32(hw, I40E_REG_MSS);
11616 	if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11617 		val &= ~I40E_REG_MSS_MIN_MASK;
11618 		val |= I40E_64BYTE_MSS;
11619 		wr32(hw, I40E_REG_MSS, val);
11620 	}
11621 
11622 	if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
11623 		msleep(75);
11624 		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11625 		if (err)
11626 			dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11627 				 i40e_stat_str(&pf->hw, err),
11628 				 i40e_aq_str(&pf->hw,
11629 					     pf->hw.aq.asq_last_status));
11630 	}
11631 	/* The main driver is (mostly) up and happy. We need to set this state
11632 	 * before setting up the misc vector or we get a race and the vector
11633 	 * ends up disabled forever.
11634 	 */
11635 	clear_bit(__I40E_DOWN, pf->state);
11636 
11637 	/* In case of MSIX we are going to setup the misc vector right here
11638 	 * to handle admin queue events etc. In case of legacy and MSI
11639 	 * the misc functionality and queue processing is combined in
11640 	 * the same vector and that gets setup at open.
11641 	 */
11642 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11643 		err = i40e_setup_misc_vector(pf);
11644 		if (err) {
11645 			dev_info(&pdev->dev,
11646 				 "setup of misc vector failed: %d\n", err);
11647 			goto err_vsis;
11648 		}
11649 	}
11650 
11651 #ifdef CONFIG_PCI_IOV
11652 	/* prep for VF support */
11653 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11654 	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11655 	    !test_bit(__I40E_BAD_EEPROM, pf->state)) {
11656 		/* disable link interrupts for VFs */
11657 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11658 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11659 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11660 		i40e_flush(hw);
11661 
11662 		if (pci_num_vf(pdev)) {
11663 			dev_info(&pdev->dev,
11664 				 "Active VFs found, allocating resources.\n");
11665 			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11666 			if (err)
11667 				dev_info(&pdev->dev,
11668 					 "Error %d allocating resources for existing VFs\n",
11669 					 err);
11670 		}
11671 	}
11672 #endif /* CONFIG_PCI_IOV */
11673 
11674 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11675 		pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11676 						      pf->num_iwarp_msix,
11677 						      I40E_IWARP_IRQ_PILE_ID);
11678 		if (pf->iwarp_base_vector < 0) {
11679 			dev_info(&pdev->dev,
11680 				 "failed to get tracking for %d vectors for IWARP err=%d\n",
11681 				 pf->num_iwarp_msix, pf->iwarp_base_vector);
11682 			pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11683 		}
11684 	}
11685 
11686 	i40e_dbg_pf_init(pf);
11687 
11688 	/* tell the firmware that we're starting */
11689 	i40e_send_version(pf);
11690 
11691 	/* since everything's happy, start the service_task timer */
11692 	mod_timer(&pf->service_timer,
11693 		  round_jiffies(jiffies + pf->service_timer_period));
11694 
11695 	/* add this PF to client device list and launch a client service task */
11696 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11697 		err = i40e_lan_add_device(pf);
11698 		if (err)
11699 			dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11700 				 err);
11701 	}
11702 
11703 #define PCI_SPEED_SIZE 8
11704 #define PCI_WIDTH_SIZE 8
11705 	/* Devices on the IOSF bus do not have this information
11706 	 * and will report PCI Gen 1 x 1 by default so don't bother
11707 	 * checking them.
11708 	 */
11709 	if (!(pf->hw_features & I40E_HW_NO_PCI_LINK_CHECK)) {
11710 		char speed[PCI_SPEED_SIZE] = "Unknown";
11711 		char width[PCI_WIDTH_SIZE] = "Unknown";
11712 
11713 		/* Get the negotiated link width and speed from PCI config
11714 		 * space
11715 		 */
11716 		pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11717 					  &link_status);
11718 
11719 		i40e_set_pci_config_data(hw, link_status);
11720 
11721 		switch (hw->bus.speed) {
11722 		case i40e_bus_speed_8000:
11723 			strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11724 		case i40e_bus_speed_5000:
11725 			strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11726 		case i40e_bus_speed_2500:
11727 			strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11728 		default:
11729 			break;
11730 		}
11731 		switch (hw->bus.width) {
11732 		case i40e_bus_width_pcie_x8:
11733 			strncpy(width, "8", PCI_WIDTH_SIZE); break;
11734 		case i40e_bus_width_pcie_x4:
11735 			strncpy(width, "4", PCI_WIDTH_SIZE); break;
11736 		case i40e_bus_width_pcie_x2:
11737 			strncpy(width, "2", PCI_WIDTH_SIZE); break;
11738 		case i40e_bus_width_pcie_x1:
11739 			strncpy(width, "1", PCI_WIDTH_SIZE); break;
11740 		default:
11741 			break;
11742 		}
11743 
11744 		dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11745 			 speed, width);
11746 
11747 		if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11748 		    hw->bus.speed < i40e_bus_speed_8000) {
11749 			dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11750 			dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11751 		}
11752 	}
11753 
11754 	/* get the requested speeds from the fw */
11755 	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11756 	if (err)
11757 		dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11758 			i40e_stat_str(&pf->hw, err),
11759 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11760 	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11761 
11762 	/* get the supported phy types from the fw */
11763 	err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11764 	if (err)
11765 		dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11766 			i40e_stat_str(&pf->hw, err),
11767 			i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11768 
11769 	/* Add a filter to drop all Flow control frames from any VSI from being
11770 	 * transmitted. By doing so we stop a malicious VF from sending out
11771 	 * PAUSE or PFC frames and potentially controlling traffic for other
11772 	 * PF/VF VSIs.
11773 	 * The FW can still send Flow control frames if enabled.
11774 	 */
11775 	i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11776 						       pf->main_vsi_seid);
11777 
11778 	if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11779 		(pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11780 		pf->hw_features |= I40E_HW_PHY_CONTROLS_LEDS;
11781 	if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
11782 		pf->hw_features |= I40E_HW_HAVE_CRT_RETIMER;
11783 	/* print a string summarizing features */
11784 	i40e_print_features(pf);
11785 
11786 	return 0;
11787 
11788 	/* Unwind what we've done if something failed in the setup */
11789 err_vsis:
11790 	set_bit(__I40E_DOWN, pf->state);
11791 	i40e_clear_interrupt_scheme(pf);
11792 	kfree(pf->vsi);
11793 err_switch_setup:
11794 	i40e_reset_interrupt_capability(pf);
11795 	del_timer_sync(&pf->service_timer);
11796 err_mac_addr:
11797 err_configure_lan_hmc:
11798 	(void)i40e_shutdown_lan_hmc(hw);
11799 err_init_lan_hmc:
11800 	kfree(pf->qp_pile);
11801 err_sw_init:
11802 err_adminq_setup:
11803 err_pf_reset:
11804 	iounmap(hw->hw_addr);
11805 err_ioremap:
11806 	kfree(pf);
11807 err_pf_alloc:
11808 	pci_disable_pcie_error_reporting(pdev);
11809 	pci_release_mem_regions(pdev);
11810 err_pci_reg:
11811 err_dma:
11812 	pci_disable_device(pdev);
11813 	return err;
11814 }
11815 
11816 /**
11817  * i40e_remove - Device removal routine
11818  * @pdev: PCI device information struct
11819  *
11820  * i40e_remove is called by the PCI subsystem to alert the driver
11821  * that is should release a PCI device.  This could be caused by a
11822  * Hot-Plug event, or because the driver is going to be removed from
11823  * memory.
11824  **/
11825 static void i40e_remove(struct pci_dev *pdev)
11826 {
11827 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11828 	struct i40e_hw *hw = &pf->hw;
11829 	i40e_status ret_code;
11830 	int i;
11831 
11832 	i40e_dbg_pf_exit(pf);
11833 
11834 	i40e_ptp_stop(pf);
11835 
11836 	/* Disable RSS in hw */
11837 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11838 	i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11839 
11840 	/* no more scheduling of any task */
11841 	set_bit(__I40E_SUSPENDED, pf->state);
11842 	set_bit(__I40E_DOWN, pf->state);
11843 	if (pf->service_timer.data)
11844 		del_timer_sync(&pf->service_timer);
11845 	if (pf->service_task.func)
11846 		cancel_work_sync(&pf->service_task);
11847 
11848 	/* Client close must be called explicitly here because the timer
11849 	 * has been stopped.
11850 	 */
11851 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
11852 
11853 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11854 		i40e_free_vfs(pf);
11855 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11856 	}
11857 
11858 	i40e_fdir_teardown(pf);
11859 
11860 	/* If there is a switch structure or any orphans, remove them.
11861 	 * This will leave only the PF's VSI remaining.
11862 	 */
11863 	for (i = 0; i < I40E_MAX_VEB; i++) {
11864 		if (!pf->veb[i])
11865 			continue;
11866 
11867 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11868 		    pf->veb[i]->uplink_seid == 0)
11869 			i40e_switch_branch_release(pf->veb[i]);
11870 	}
11871 
11872 	/* Now we can shutdown the PF's VSI, just before we kill
11873 	 * adminq and hmc.
11874 	 */
11875 	if (pf->vsi[pf->lan_vsi])
11876 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11877 
11878 	/* remove attached clients */
11879 	if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11880 		ret_code = i40e_lan_del_device(pf);
11881 		if (ret_code)
11882 			dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11883 				 ret_code);
11884 	}
11885 
11886 	/* shutdown and destroy the HMC */
11887 	if (hw->hmc.hmc_obj) {
11888 		ret_code = i40e_shutdown_lan_hmc(hw);
11889 		if (ret_code)
11890 			dev_warn(&pdev->dev,
11891 				 "Failed to destroy the HMC resources: %d\n",
11892 				 ret_code);
11893 	}
11894 
11895 	/* shutdown the adminq */
11896 	i40e_shutdown_adminq(hw);
11897 
11898 	/* destroy the locks only once, here */
11899 	mutex_destroy(&hw->aq.arq_mutex);
11900 	mutex_destroy(&hw->aq.asq_mutex);
11901 
11902 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11903 	i40e_clear_interrupt_scheme(pf);
11904 	for (i = 0; i < pf->num_alloc_vsi; i++) {
11905 		if (pf->vsi[i]) {
11906 			i40e_vsi_clear_rings(pf->vsi[i]);
11907 			i40e_vsi_clear(pf->vsi[i]);
11908 			pf->vsi[i] = NULL;
11909 		}
11910 	}
11911 
11912 	for (i = 0; i < I40E_MAX_VEB; i++) {
11913 		kfree(pf->veb[i]);
11914 		pf->veb[i] = NULL;
11915 	}
11916 
11917 	kfree(pf->qp_pile);
11918 	kfree(pf->vsi);
11919 
11920 	iounmap(hw->hw_addr);
11921 	kfree(pf);
11922 	pci_release_mem_regions(pdev);
11923 
11924 	pci_disable_pcie_error_reporting(pdev);
11925 	pci_disable_device(pdev);
11926 }
11927 
11928 /**
11929  * i40e_pci_error_detected - warning that something funky happened in PCI land
11930  * @pdev: PCI device information struct
11931  *
11932  * Called to warn that something happened and the error handling steps
11933  * are in progress.  Allows the driver to quiesce things, be ready for
11934  * remediation.
11935  **/
11936 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11937 						enum pci_channel_state error)
11938 {
11939 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11940 
11941 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11942 
11943 	if (!pf) {
11944 		dev_info(&pdev->dev,
11945 			 "Cannot recover - error happened during device probe\n");
11946 		return PCI_ERS_RESULT_DISCONNECT;
11947 	}
11948 
11949 	/* shutdown all operations */
11950 	if (!test_bit(__I40E_SUSPENDED, pf->state))
11951 		i40e_prep_for_reset(pf, false);
11952 
11953 	/* Request a slot reset */
11954 	return PCI_ERS_RESULT_NEED_RESET;
11955 }
11956 
11957 /**
11958  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11959  * @pdev: PCI device information struct
11960  *
11961  * Called to find if the driver can work with the device now that
11962  * the pci slot has been reset.  If a basic connection seems good
11963  * (registers are readable and have sane content) then return a
11964  * happy little PCI_ERS_RESULT_xxx.
11965  **/
11966 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11967 {
11968 	struct i40e_pf *pf = pci_get_drvdata(pdev);
11969 	pci_ers_result_t result;
11970 	int err;
11971 	u32 reg;
11972 
11973 	dev_dbg(&pdev->dev, "%s\n", __func__);
11974 	if (pci_enable_device_mem(pdev)) {
11975 		dev_info(&pdev->dev,
11976 			 "Cannot re-enable PCI device after reset.\n");
11977 		result = PCI_ERS_RESULT_DISCONNECT;
11978 	} else {
11979 		pci_set_master(pdev);
11980 		pci_restore_state(pdev);
11981 		pci_save_state(pdev);
11982 		pci_wake_from_d3(pdev, false);
11983 
11984 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11985 		if (reg == 0)
11986 			result = PCI_ERS_RESULT_RECOVERED;
11987 		else
11988 			result = PCI_ERS_RESULT_DISCONNECT;
11989 	}
11990 
11991 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
11992 	if (err) {
11993 		dev_info(&pdev->dev,
11994 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11995 			 err);
11996 		/* non-fatal, continue */
11997 	}
11998 
11999 	return result;
12000 }
12001 
12002 /**
12003  * i40e_pci_error_resume - restart operations after PCI error recovery
12004  * @pdev: PCI device information struct
12005  *
12006  * Called to allow the driver to bring things back up after PCI error
12007  * and/or reset recovery has finished.
12008  **/
12009 static void i40e_pci_error_resume(struct pci_dev *pdev)
12010 {
12011 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12012 
12013 	dev_dbg(&pdev->dev, "%s\n", __func__);
12014 	if (test_bit(__I40E_SUSPENDED, pf->state))
12015 		return;
12016 
12017 	i40e_handle_reset_warning(pf, false);
12018 }
12019 
12020 /**
12021  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
12022  * using the mac_address_write admin q function
12023  * @pf: pointer to i40e_pf struct
12024  **/
12025 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
12026 {
12027 	struct i40e_hw *hw = &pf->hw;
12028 	i40e_status ret;
12029 	u8 mac_addr[6];
12030 	u16 flags = 0;
12031 
12032 	/* Get current MAC address in case it's an LAA */
12033 	if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
12034 		ether_addr_copy(mac_addr,
12035 				pf->vsi[pf->lan_vsi]->netdev->dev_addr);
12036 	} else {
12037 		dev_err(&pf->pdev->dev,
12038 			"Failed to retrieve MAC address; using default\n");
12039 		ether_addr_copy(mac_addr, hw->mac.addr);
12040 	}
12041 
12042 	/* The FW expects the mac address write cmd to first be called with
12043 	 * one of these flags before calling it again with the multicast
12044 	 * enable flags.
12045 	 */
12046 	flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
12047 
12048 	if (hw->func_caps.flex10_enable && hw->partition_id != 1)
12049 		flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
12050 
12051 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
12052 	if (ret) {
12053 		dev_err(&pf->pdev->dev,
12054 			"Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
12055 		return;
12056 	}
12057 
12058 	flags = I40E_AQC_MC_MAG_EN
12059 			| I40E_AQC_WOL_PRESERVE_ON_PFR
12060 			| I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
12061 	ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
12062 	if (ret)
12063 		dev_err(&pf->pdev->dev,
12064 			"Failed to enable Multicast Magic Packet wake up\n");
12065 }
12066 
12067 /**
12068  * i40e_shutdown - PCI callback for shutting down
12069  * @pdev: PCI device information struct
12070  **/
12071 static void i40e_shutdown(struct pci_dev *pdev)
12072 {
12073 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12074 	struct i40e_hw *hw = &pf->hw;
12075 
12076 	set_bit(__I40E_SUSPENDED, pf->state);
12077 	set_bit(__I40E_DOWN, pf->state);
12078 	rtnl_lock();
12079 	i40e_prep_for_reset(pf, true);
12080 	rtnl_unlock();
12081 
12082 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
12083 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
12084 
12085 	del_timer_sync(&pf->service_timer);
12086 	cancel_work_sync(&pf->service_task);
12087 	i40e_fdir_teardown(pf);
12088 
12089 	/* Client close must be called explicitly here because the timer
12090 	 * has been stopped.
12091 	 */
12092 	i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
12093 
12094 	if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
12095 		i40e_enable_mc_magic_wake(pf);
12096 
12097 	i40e_prep_for_reset(pf, false);
12098 
12099 	wr32(hw, I40E_PFPM_APM,
12100 	     (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
12101 	wr32(hw, I40E_PFPM_WUFC,
12102 	     (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
12103 
12104 	i40e_clear_interrupt_scheme(pf);
12105 
12106 	if (system_state == SYSTEM_POWER_OFF) {
12107 		pci_wake_from_d3(pdev, pf->wol_en);
12108 		pci_set_power_state(pdev, PCI_D3hot);
12109 	}
12110 }
12111 
12112 #ifdef CONFIG_PM
12113 /**
12114  * i40e_suspend - PM callback for moving to D3
12115  * @dev: generic device information structure
12116  **/
12117 static int i40e_suspend(struct device *dev)
12118 {
12119 	struct pci_dev *pdev = to_pci_dev(dev);
12120 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12121 	struct i40e_hw *hw = &pf->hw;
12122 
12123 	/* If we're already suspended, then there is nothing to do */
12124 	if (test_and_set_bit(__I40E_SUSPENDED, pf->state))
12125 		return 0;
12126 
12127 	set_bit(__I40E_DOWN, pf->state);
12128 
12129 	/* Ensure service task will not be running */
12130 	del_timer_sync(&pf->service_timer);
12131 	cancel_work_sync(&pf->service_task);
12132 
12133 	if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
12134 		i40e_enable_mc_magic_wake(pf);
12135 
12136 	i40e_prep_for_reset(pf, false);
12137 
12138 	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
12139 	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
12140 
12141 	/* Clear the interrupt scheme and release our IRQs so that the system
12142 	 * can safely hibernate even when there are a large number of CPUs.
12143 	 * Otherwise hibernation might fail when mapping all the vectors back
12144 	 * to CPU0.
12145 	 */
12146 	i40e_clear_interrupt_scheme(pf);
12147 
12148 	return 0;
12149 }
12150 
12151 /**
12152  * i40e_resume - PM callback for waking up from D3
12153  * @dev: generic device information structure
12154  **/
12155 static int i40e_resume(struct device *dev)
12156 {
12157 	struct pci_dev *pdev = to_pci_dev(dev);
12158 	struct i40e_pf *pf = pci_get_drvdata(pdev);
12159 	int err;
12160 
12161 	/* If we're not suspended, then there is nothing to do */
12162 	if (!test_bit(__I40E_SUSPENDED, pf->state))
12163 		return 0;
12164 
12165 	/* We cleared the interrupt scheme when we suspended, so we need to
12166 	 * restore it now to resume device functionality.
12167 	 */
12168 	err = i40e_restore_interrupt_scheme(pf);
12169 	if (err) {
12170 		dev_err(&pdev->dev, "Cannot restore interrupt scheme: %d\n",
12171 			err);
12172 	}
12173 
12174 	clear_bit(__I40E_DOWN, pf->state);
12175 	i40e_reset_and_rebuild(pf, false, false);
12176 
12177 	/* Clear suspended state last after everything is recovered */
12178 	clear_bit(__I40E_SUSPENDED, pf->state);
12179 
12180 	/* Restart the service task */
12181 	mod_timer(&pf->service_timer,
12182 		  round_jiffies(jiffies + pf->service_timer_period));
12183 
12184 	return 0;
12185 }
12186 
12187 #endif /* CONFIG_PM */
12188 
12189 static const struct pci_error_handlers i40e_err_handler = {
12190 	.error_detected = i40e_pci_error_detected,
12191 	.slot_reset = i40e_pci_error_slot_reset,
12192 	.resume = i40e_pci_error_resume,
12193 };
12194 
12195 static SIMPLE_DEV_PM_OPS(i40e_pm_ops, i40e_suspend, i40e_resume);
12196 
12197 static struct pci_driver i40e_driver = {
12198 	.name     = i40e_driver_name,
12199 	.id_table = i40e_pci_tbl,
12200 	.probe    = i40e_probe,
12201 	.remove   = i40e_remove,
12202 #ifdef CONFIG_PM
12203 	.driver   = {
12204 		.pm = &i40e_pm_ops,
12205 	},
12206 #endif /* CONFIG_PM */
12207 	.shutdown = i40e_shutdown,
12208 	.err_handler = &i40e_err_handler,
12209 	.sriov_configure = i40e_pci_sriov_configure,
12210 };
12211 
12212 /**
12213  * i40e_init_module - Driver registration routine
12214  *
12215  * i40e_init_module is the first routine called when the driver is
12216  * loaded. All it does is register with the PCI subsystem.
12217  **/
12218 static int __init i40e_init_module(void)
12219 {
12220 	pr_info("%s: %s - version %s\n", i40e_driver_name,
12221 		i40e_driver_string, i40e_driver_version_str);
12222 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
12223 
12224 	/* There is no need to throttle the number of active tasks because
12225 	 * each device limits its own task using a state bit for scheduling
12226 	 * the service task, and the device tasks do not interfere with each
12227 	 * other, so we don't set a max task limit. We must set WQ_MEM_RECLAIM
12228 	 * since we need to be able to guarantee forward progress even under
12229 	 * memory pressure.
12230 	 */
12231 	i40e_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, i40e_driver_name);
12232 	if (!i40e_wq) {
12233 		pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
12234 		return -ENOMEM;
12235 	}
12236 
12237 	i40e_dbg_init();
12238 	return pci_register_driver(&i40e_driver);
12239 }
12240 module_init(i40e_init_module);
12241 
12242 /**
12243  * i40e_exit_module - Driver exit cleanup routine
12244  *
12245  * i40e_exit_module is called just before the driver is removed
12246  * from memory.
12247  **/
12248 static void __exit i40e_exit_module(void)
12249 {
12250 	pci_unregister_driver(&i40e_driver);
12251 	destroy_workqueue(i40e_wq);
12252 	i40e_dbg_exit();
12253 }
12254 module_exit(i40e_exit_module);
12255