1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, Intel Corporation. */
3 
4 #include "ice.h"
5 #include "ice_base.h"
6 #include "ice_flow.h"
7 #include "ice_lib.h"
8 #include "ice_fltr.h"
9 #include "ice_dcb_lib.h"
10 #include "ice_devlink.h"
11 
12 /**
13  * ice_vsi_type_str - maps VSI type enum to string equivalents
14  * @vsi_type: VSI type enum
15  */
16 const char *ice_vsi_type_str(enum ice_vsi_type vsi_type)
17 {
18 	switch (vsi_type) {
19 	case ICE_VSI_PF:
20 		return "ICE_VSI_PF";
21 	case ICE_VSI_VF:
22 		return "ICE_VSI_VF";
23 	case ICE_VSI_CTRL:
24 		return "ICE_VSI_CTRL";
25 	case ICE_VSI_LB:
26 		return "ICE_VSI_LB";
27 	default:
28 		return "unknown";
29 	}
30 }
31 
32 /**
33  * ice_vsi_ctrl_all_rx_rings - Start or stop a VSI's Rx rings
34  * @vsi: the VSI being configured
35  * @ena: start or stop the Rx rings
36  *
37  * First enable/disable all of the Rx rings, flush any remaining writes, and
38  * then verify that they have all been enabled/disabled successfully. This will
39  * let all of the register writes complete when enabling/disabling the Rx rings
40  * before waiting for the change in hardware to complete.
41  */
42 static int ice_vsi_ctrl_all_rx_rings(struct ice_vsi *vsi, bool ena)
43 {
44 	int ret = 0;
45 	u16 i;
46 
47 	for (i = 0; i < vsi->num_rxq; i++)
48 		ice_vsi_ctrl_one_rx_ring(vsi, ena, i, false);
49 
50 	ice_flush(&vsi->back->hw);
51 
52 	for (i = 0; i < vsi->num_rxq; i++) {
53 		ret = ice_vsi_wait_one_rx_ring(vsi, ena, i);
54 		if (ret)
55 			break;
56 	}
57 
58 	return ret;
59 }
60 
61 /**
62  * ice_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the VSI
63  * @vsi: VSI pointer
64  *
65  * On error: returns error code (negative)
66  * On success: returns 0
67  */
68 static int ice_vsi_alloc_arrays(struct ice_vsi *vsi)
69 {
70 	struct ice_pf *pf = vsi->back;
71 	struct device *dev;
72 
73 	dev = ice_pf_to_dev(pf);
74 
75 	/* allocate memory for both Tx and Rx ring pointers */
76 	vsi->tx_rings = devm_kcalloc(dev, vsi->alloc_txq,
77 				     sizeof(*vsi->tx_rings), GFP_KERNEL);
78 	if (!vsi->tx_rings)
79 		return -ENOMEM;
80 
81 	vsi->rx_rings = devm_kcalloc(dev, vsi->alloc_rxq,
82 				     sizeof(*vsi->rx_rings), GFP_KERNEL);
83 	if (!vsi->rx_rings)
84 		goto err_rings;
85 
86 	/* XDP will have vsi->alloc_txq Tx queues as well, so double the size */
87 	vsi->txq_map = devm_kcalloc(dev, (2 * vsi->alloc_txq),
88 				    sizeof(*vsi->txq_map), GFP_KERNEL);
89 
90 	if (!vsi->txq_map)
91 		goto err_txq_map;
92 
93 	vsi->rxq_map = devm_kcalloc(dev, vsi->alloc_rxq,
94 				    sizeof(*vsi->rxq_map), GFP_KERNEL);
95 	if (!vsi->rxq_map)
96 		goto err_rxq_map;
97 
98 	/* There is no need to allocate q_vectors for a loopback VSI. */
99 	if (vsi->type == ICE_VSI_LB)
100 		return 0;
101 
102 	/* allocate memory for q_vector pointers */
103 	vsi->q_vectors = devm_kcalloc(dev, vsi->num_q_vectors,
104 				      sizeof(*vsi->q_vectors), GFP_KERNEL);
105 	if (!vsi->q_vectors)
106 		goto err_vectors;
107 
108 	return 0;
109 
110 err_vectors:
111 	devm_kfree(dev, vsi->rxq_map);
112 err_rxq_map:
113 	devm_kfree(dev, vsi->txq_map);
114 err_txq_map:
115 	devm_kfree(dev, vsi->rx_rings);
116 err_rings:
117 	devm_kfree(dev, vsi->tx_rings);
118 	return -ENOMEM;
119 }
120 
121 /**
122  * ice_vsi_set_num_desc - Set number of descriptors for queues on this VSI
123  * @vsi: the VSI being configured
124  */
125 static void ice_vsi_set_num_desc(struct ice_vsi *vsi)
126 {
127 	switch (vsi->type) {
128 	case ICE_VSI_PF:
129 	case ICE_VSI_CTRL:
130 	case ICE_VSI_LB:
131 		/* a user could change the values of num_[tr]x_desc using
132 		 * ethtool -G so we should keep those values instead of
133 		 * overwriting them with the defaults.
134 		 */
135 		if (!vsi->num_rx_desc)
136 			vsi->num_rx_desc = ICE_DFLT_NUM_RX_DESC;
137 		if (!vsi->num_tx_desc)
138 			vsi->num_tx_desc = ICE_DFLT_NUM_TX_DESC;
139 		break;
140 	default:
141 		dev_dbg(ice_pf_to_dev(vsi->back), "Not setting number of Tx/Rx descriptors for VSI type %d\n",
142 			vsi->type);
143 		break;
144 	}
145 }
146 
147 /**
148  * ice_vsi_set_num_qs - Set number of queues, descriptors and vectors for a VSI
149  * @vsi: the VSI being configured
150  * @vf_id: ID of the VF being configured
151  *
152  * Return 0 on success and a negative value on error
153  */
154 static void ice_vsi_set_num_qs(struct ice_vsi *vsi, u16 vf_id)
155 {
156 	struct ice_pf *pf = vsi->back;
157 	struct ice_vf *vf = NULL;
158 
159 	if (vsi->type == ICE_VSI_VF)
160 		vsi->vf_id = vf_id;
161 	else
162 		vsi->vf_id = ICE_INVAL_VFID;
163 
164 	switch (vsi->type) {
165 	case ICE_VSI_PF:
166 		vsi->alloc_txq = min3(pf->num_lan_msix,
167 				      ice_get_avail_txq_count(pf),
168 				      (u16)num_online_cpus());
169 		if (vsi->req_txq) {
170 			vsi->alloc_txq = vsi->req_txq;
171 			vsi->num_txq = vsi->req_txq;
172 		}
173 
174 		pf->num_lan_tx = vsi->alloc_txq;
175 
176 		/* only 1 Rx queue unless RSS is enabled */
177 		if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
178 			vsi->alloc_rxq = 1;
179 		} else {
180 			vsi->alloc_rxq = min3(pf->num_lan_msix,
181 					      ice_get_avail_rxq_count(pf),
182 					      (u16)num_online_cpus());
183 			if (vsi->req_rxq) {
184 				vsi->alloc_rxq = vsi->req_rxq;
185 				vsi->num_rxq = vsi->req_rxq;
186 			}
187 		}
188 
189 		pf->num_lan_rx = vsi->alloc_rxq;
190 
191 		vsi->num_q_vectors = min_t(int, pf->num_lan_msix,
192 					   max_t(int, vsi->alloc_rxq,
193 						 vsi->alloc_txq));
194 		break;
195 	case ICE_VSI_VF:
196 		vf = &pf->vf[vsi->vf_id];
197 		vsi->alloc_txq = vf->num_vf_qs;
198 		vsi->alloc_rxq = vf->num_vf_qs;
199 		/* pf->num_msix_per_vf includes (VF miscellaneous vector +
200 		 * data queue interrupts). Since vsi->num_q_vectors is number
201 		 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
202 		 * original vector count
203 		 */
204 		vsi->num_q_vectors = pf->num_msix_per_vf - ICE_NONQ_VECS_VF;
205 		break;
206 	case ICE_VSI_CTRL:
207 		vsi->alloc_txq = 1;
208 		vsi->alloc_rxq = 1;
209 		vsi->num_q_vectors = 1;
210 		break;
211 	case ICE_VSI_LB:
212 		vsi->alloc_txq = 1;
213 		vsi->alloc_rxq = 1;
214 		break;
215 	default:
216 		dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi->type);
217 		break;
218 	}
219 
220 	ice_vsi_set_num_desc(vsi);
221 }
222 
223 /**
224  * ice_get_free_slot - get the next non-NULL location index in array
225  * @array: array to search
226  * @size: size of the array
227  * @curr: last known occupied index to be used as a search hint
228  *
229  * void * is being used to keep the functionality generic. This lets us use this
230  * function on any array of pointers.
231  */
232 static int ice_get_free_slot(void *array, int size, int curr)
233 {
234 	int **tmp_array = (int **)array;
235 	int next;
236 
237 	if (curr < (size - 1) && !tmp_array[curr + 1]) {
238 		next = curr + 1;
239 	} else {
240 		int i = 0;
241 
242 		while ((i < size) && (tmp_array[i]))
243 			i++;
244 		if (i == size)
245 			next = ICE_NO_VSI;
246 		else
247 			next = i;
248 	}
249 	return next;
250 }
251 
252 /**
253  * ice_vsi_delete - delete a VSI from the switch
254  * @vsi: pointer to VSI being removed
255  */
256 static void ice_vsi_delete(struct ice_vsi *vsi)
257 {
258 	struct ice_pf *pf = vsi->back;
259 	struct ice_vsi_ctx *ctxt;
260 	enum ice_status status;
261 
262 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
263 	if (!ctxt)
264 		return;
265 
266 	if (vsi->type == ICE_VSI_VF)
267 		ctxt->vf_num = vsi->vf_id;
268 	ctxt->vsi_num = vsi->vsi_num;
269 
270 	memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
271 
272 	status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
273 	if (status)
274 		dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %s\n",
275 			vsi->vsi_num, ice_stat_str(status));
276 
277 	kfree(ctxt);
278 }
279 
280 /**
281  * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
282  * @vsi: pointer to VSI being cleared
283  */
284 static void ice_vsi_free_arrays(struct ice_vsi *vsi)
285 {
286 	struct ice_pf *pf = vsi->back;
287 	struct device *dev;
288 
289 	dev = ice_pf_to_dev(pf);
290 
291 	/* free the ring and vector containers */
292 	if (vsi->q_vectors) {
293 		devm_kfree(dev, vsi->q_vectors);
294 		vsi->q_vectors = NULL;
295 	}
296 	if (vsi->tx_rings) {
297 		devm_kfree(dev, vsi->tx_rings);
298 		vsi->tx_rings = NULL;
299 	}
300 	if (vsi->rx_rings) {
301 		devm_kfree(dev, vsi->rx_rings);
302 		vsi->rx_rings = NULL;
303 	}
304 	if (vsi->txq_map) {
305 		devm_kfree(dev, vsi->txq_map);
306 		vsi->txq_map = NULL;
307 	}
308 	if (vsi->rxq_map) {
309 		devm_kfree(dev, vsi->rxq_map);
310 		vsi->rxq_map = NULL;
311 	}
312 }
313 
314 /**
315  * ice_vsi_clear - clean up and deallocate the provided VSI
316  * @vsi: pointer to VSI being cleared
317  *
318  * This deallocates the VSI's queue resources, removes it from the PF's
319  * VSI array if necessary, and deallocates the VSI
320  *
321  * Returns 0 on success, negative on failure
322  */
323 static int ice_vsi_clear(struct ice_vsi *vsi)
324 {
325 	struct ice_pf *pf = NULL;
326 	struct device *dev;
327 
328 	if (!vsi)
329 		return 0;
330 
331 	if (!vsi->back)
332 		return -EINVAL;
333 
334 	pf = vsi->back;
335 	dev = ice_pf_to_dev(pf);
336 
337 	if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
338 		dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
339 		return -EINVAL;
340 	}
341 
342 	mutex_lock(&pf->sw_mutex);
343 	/* updates the PF for this cleared VSI */
344 
345 	pf->vsi[vsi->idx] = NULL;
346 	if (vsi->idx < pf->next_vsi && vsi->type != ICE_VSI_CTRL)
347 		pf->next_vsi = vsi->idx;
348 	if (vsi->idx < pf->next_vsi && vsi->type == ICE_VSI_CTRL &&
349 	    vsi->vf_id != ICE_INVAL_VFID)
350 		pf->next_vsi = vsi->idx;
351 
352 	ice_vsi_free_arrays(vsi);
353 	mutex_unlock(&pf->sw_mutex);
354 	devm_kfree(dev, vsi);
355 
356 	return 0;
357 }
358 
359 /**
360  * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
361  * @irq: interrupt number
362  * @data: pointer to a q_vector
363  */
364 static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
365 {
366 	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
367 
368 	if (!q_vector->tx.ring)
369 		return IRQ_HANDLED;
370 
371 #define FDIR_RX_DESC_CLEAN_BUDGET 64
372 	ice_clean_rx_irq(q_vector->rx.ring, FDIR_RX_DESC_CLEAN_BUDGET);
373 	ice_clean_ctrl_tx_irq(q_vector->tx.ring);
374 
375 	return IRQ_HANDLED;
376 }
377 
378 /**
379  * ice_msix_clean_rings - MSIX mode Interrupt Handler
380  * @irq: interrupt number
381  * @data: pointer to a q_vector
382  */
383 static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
384 {
385 	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
386 
387 	if (!q_vector->tx.ring && !q_vector->rx.ring)
388 		return IRQ_HANDLED;
389 
390 	q_vector->total_events++;
391 
392 	napi_schedule(&q_vector->napi);
393 
394 	return IRQ_HANDLED;
395 }
396 
397 /**
398  * ice_vsi_alloc - Allocates the next available struct VSI in the PF
399  * @pf: board private structure
400  * @vsi_type: type of VSI
401  * @vf_id: ID of the VF being configured
402  *
403  * returns a pointer to a VSI on success, NULL on failure.
404  */
405 static struct ice_vsi *
406 ice_vsi_alloc(struct ice_pf *pf, enum ice_vsi_type vsi_type, u16 vf_id)
407 {
408 	struct device *dev = ice_pf_to_dev(pf);
409 	struct ice_vsi *vsi = NULL;
410 
411 	/* Need to protect the allocation of the VSIs at the PF level */
412 	mutex_lock(&pf->sw_mutex);
413 
414 	/* If we have already allocated our maximum number of VSIs,
415 	 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
416 	 * is available to be populated
417 	 */
418 	if (pf->next_vsi == ICE_NO_VSI) {
419 		dev_dbg(dev, "out of VSI slots!\n");
420 		goto unlock_pf;
421 	}
422 
423 	vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
424 	if (!vsi)
425 		goto unlock_pf;
426 
427 	vsi->type = vsi_type;
428 	vsi->back = pf;
429 	set_bit(ICE_VSI_DOWN, vsi->state);
430 
431 	if (vsi_type == ICE_VSI_VF)
432 		ice_vsi_set_num_qs(vsi, vf_id);
433 	else
434 		ice_vsi_set_num_qs(vsi, ICE_INVAL_VFID);
435 
436 	switch (vsi->type) {
437 	case ICE_VSI_PF:
438 		if (ice_vsi_alloc_arrays(vsi))
439 			goto err_rings;
440 
441 		/* Setup default MSIX irq handler for VSI */
442 		vsi->irq_handler = ice_msix_clean_rings;
443 		break;
444 	case ICE_VSI_CTRL:
445 		if (ice_vsi_alloc_arrays(vsi))
446 			goto err_rings;
447 
448 		/* Setup ctrl VSI MSIX irq handler */
449 		vsi->irq_handler = ice_msix_clean_ctrl_vsi;
450 		break;
451 	case ICE_VSI_VF:
452 		if (ice_vsi_alloc_arrays(vsi))
453 			goto err_rings;
454 		break;
455 	case ICE_VSI_LB:
456 		if (ice_vsi_alloc_arrays(vsi))
457 			goto err_rings;
458 		break;
459 	default:
460 		dev_warn(dev, "Unknown VSI type %d\n", vsi->type);
461 		goto unlock_pf;
462 	}
463 
464 	if (vsi->type == ICE_VSI_CTRL && vf_id == ICE_INVAL_VFID) {
465 		/* Use the last VSI slot as the index for PF control VSI */
466 		vsi->idx = pf->num_alloc_vsi - 1;
467 		pf->ctrl_vsi_idx = vsi->idx;
468 		pf->vsi[vsi->idx] = vsi;
469 	} else {
470 		/* fill slot and make note of the index */
471 		vsi->idx = pf->next_vsi;
472 		pf->vsi[pf->next_vsi] = vsi;
473 
474 		/* prepare pf->next_vsi for next use */
475 		pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
476 						 pf->next_vsi);
477 	}
478 
479 	if (vsi->type == ICE_VSI_CTRL && vf_id != ICE_INVAL_VFID)
480 		pf->vf[vf_id].ctrl_vsi_idx = vsi->idx;
481 	goto unlock_pf;
482 
483 err_rings:
484 	devm_kfree(dev, vsi);
485 	vsi = NULL;
486 unlock_pf:
487 	mutex_unlock(&pf->sw_mutex);
488 	return vsi;
489 }
490 
491 /**
492  * ice_alloc_fd_res - Allocate FD resource for a VSI
493  * @vsi: pointer to the ice_vsi
494  *
495  * This allocates the FD resources
496  *
497  * Returns 0 on success, -EPERM on no-op or -EIO on failure
498  */
499 static int ice_alloc_fd_res(struct ice_vsi *vsi)
500 {
501 	struct ice_pf *pf = vsi->back;
502 	u32 g_val, b_val;
503 
504 	/* Flow Director filters are only allocated/assigned to the PF VSI which
505 	 * passes the traffic. The CTRL VSI is only used to add/delete filters
506 	 * so we don't allocate resources to it
507 	 */
508 
509 	/* FD filters from guaranteed pool per VSI */
510 	g_val = pf->hw.func_caps.fd_fltr_guar;
511 	if (!g_val)
512 		return -EPERM;
513 
514 	/* FD filters from best effort pool */
515 	b_val = pf->hw.func_caps.fd_fltr_best_effort;
516 	if (!b_val)
517 		return -EPERM;
518 
519 	if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF))
520 		return -EPERM;
521 
522 	if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
523 		return -EPERM;
524 
525 	vsi->num_gfltr = g_val / pf->num_alloc_vsi;
526 
527 	/* each VSI gets same "best_effort" quota */
528 	vsi->num_bfltr = b_val;
529 
530 	if (vsi->type == ICE_VSI_VF) {
531 		vsi->num_gfltr = 0;
532 
533 		/* each VSI gets same "best_effort" quota */
534 		vsi->num_bfltr = b_val;
535 	}
536 
537 	return 0;
538 }
539 
540 /**
541  * ice_vsi_get_qs - Assign queues from PF to VSI
542  * @vsi: the VSI to assign queues to
543  *
544  * Returns 0 on success and a negative value on error
545  */
546 static int ice_vsi_get_qs(struct ice_vsi *vsi)
547 {
548 	struct ice_pf *pf = vsi->back;
549 	struct ice_qs_cfg tx_qs_cfg = {
550 		.qs_mutex = &pf->avail_q_mutex,
551 		.pf_map = pf->avail_txqs,
552 		.pf_map_size = pf->max_pf_txqs,
553 		.q_count = vsi->alloc_txq,
554 		.scatter_count = ICE_MAX_SCATTER_TXQS,
555 		.vsi_map = vsi->txq_map,
556 		.vsi_map_offset = 0,
557 		.mapping_mode = ICE_VSI_MAP_CONTIG
558 	};
559 	struct ice_qs_cfg rx_qs_cfg = {
560 		.qs_mutex = &pf->avail_q_mutex,
561 		.pf_map = pf->avail_rxqs,
562 		.pf_map_size = pf->max_pf_rxqs,
563 		.q_count = vsi->alloc_rxq,
564 		.scatter_count = ICE_MAX_SCATTER_RXQS,
565 		.vsi_map = vsi->rxq_map,
566 		.vsi_map_offset = 0,
567 		.mapping_mode = ICE_VSI_MAP_CONTIG
568 	};
569 	int ret;
570 
571 	ret = __ice_vsi_get_qs(&tx_qs_cfg);
572 	if (ret)
573 		return ret;
574 	vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
575 
576 	ret = __ice_vsi_get_qs(&rx_qs_cfg);
577 	if (ret)
578 		return ret;
579 	vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
580 
581 	return 0;
582 }
583 
584 /**
585  * ice_vsi_put_qs - Release queues from VSI to PF
586  * @vsi: the VSI that is going to release queues
587  */
588 static void ice_vsi_put_qs(struct ice_vsi *vsi)
589 {
590 	struct ice_pf *pf = vsi->back;
591 	int i;
592 
593 	mutex_lock(&pf->avail_q_mutex);
594 
595 	for (i = 0; i < vsi->alloc_txq; i++) {
596 		clear_bit(vsi->txq_map[i], pf->avail_txqs);
597 		vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
598 	}
599 
600 	for (i = 0; i < vsi->alloc_rxq; i++) {
601 		clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
602 		vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
603 	}
604 
605 	mutex_unlock(&pf->avail_q_mutex);
606 }
607 
608 /**
609  * ice_is_safe_mode
610  * @pf: pointer to the PF struct
611  *
612  * returns true if driver is in safe mode, false otherwise
613  */
614 bool ice_is_safe_mode(struct ice_pf *pf)
615 {
616 	return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
617 }
618 
619 /**
620  * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
621  * @vsi: the VSI being cleaned up
622  *
623  * This function deletes RSS input set for all flows that were configured
624  * for this VSI
625  */
626 static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
627 {
628 	struct ice_pf *pf = vsi->back;
629 	enum ice_status status;
630 
631 	if (ice_is_safe_mode(pf))
632 		return;
633 
634 	status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
635 	if (status)
636 		dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %s\n",
637 			vsi->vsi_num, ice_stat_str(status));
638 }
639 
640 /**
641  * ice_rss_clean - Delete RSS related VSI structures and configuration
642  * @vsi: the VSI being removed
643  */
644 static void ice_rss_clean(struct ice_vsi *vsi)
645 {
646 	struct ice_pf *pf = vsi->back;
647 	struct device *dev;
648 
649 	dev = ice_pf_to_dev(pf);
650 
651 	if (vsi->rss_hkey_user)
652 		devm_kfree(dev, vsi->rss_hkey_user);
653 	if (vsi->rss_lut_user)
654 		devm_kfree(dev, vsi->rss_lut_user);
655 
656 	ice_vsi_clean_rss_flow_fld(vsi);
657 	/* remove RSS replay list */
658 	if (!ice_is_safe_mode(pf))
659 		ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
660 }
661 
662 /**
663  * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
664  * @vsi: the VSI being configured
665  */
666 static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
667 {
668 	struct ice_hw_common_caps *cap;
669 	struct ice_pf *pf = vsi->back;
670 
671 	if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
672 		vsi->rss_size = 1;
673 		return;
674 	}
675 
676 	cap = &pf->hw.func_caps.common_cap;
677 	switch (vsi->type) {
678 	case ICE_VSI_PF:
679 		/* PF VSI will inherit RSS instance of PF */
680 		vsi->rss_table_size = (u16)cap->rss_table_size;
681 		vsi->rss_size = min_t(u16, num_online_cpus(),
682 				      BIT(cap->rss_table_entry_width));
683 		vsi->rss_lut_type = ICE_AQC_GSET_RSS_LUT_TABLE_TYPE_PF;
684 		break;
685 	case ICE_VSI_VF:
686 		/* VF VSI will get a small RSS table.
687 		 * For VSI_LUT, LUT size should be set to 64 bytes.
688 		 */
689 		vsi->rss_table_size = ICE_VSIQF_HLUT_ARRAY_SIZE;
690 		vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
691 		vsi->rss_lut_type = ICE_AQC_GSET_RSS_LUT_TABLE_TYPE_VSI;
692 		break;
693 	case ICE_VSI_LB:
694 		break;
695 	default:
696 		dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
697 			ice_vsi_type_str(vsi->type));
698 		break;
699 	}
700 }
701 
702 /**
703  * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
704  * @ctxt: the VSI context being set
705  *
706  * This initializes a default VSI context for all sections except the Queues.
707  */
708 static void ice_set_dflt_vsi_ctx(struct ice_vsi_ctx *ctxt)
709 {
710 	u32 table = 0;
711 
712 	memset(&ctxt->info, 0, sizeof(ctxt->info));
713 	/* VSI's should be allocated from shared pool */
714 	ctxt->alloc_from_pool = true;
715 	/* Src pruning enabled by default */
716 	ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
717 	/* Traffic from VSI can be sent to LAN */
718 	ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
719 	/* By default bits 3 and 4 in vlan_flags are 0's which results in legacy
720 	 * behavior (show VLAN, DEI, and UP) in descriptor. Also, allow all
721 	 * packets untagged/tagged.
722 	 */
723 	ctxt->info.vlan_flags = ((ICE_AQ_VSI_VLAN_MODE_ALL &
724 				  ICE_AQ_VSI_VLAN_MODE_M) >>
725 				 ICE_AQ_VSI_VLAN_MODE_S);
726 	/* Have 1:1 UP mapping for both ingress/egress tables */
727 	table |= ICE_UP_TABLE_TRANSLATE(0, 0);
728 	table |= ICE_UP_TABLE_TRANSLATE(1, 1);
729 	table |= ICE_UP_TABLE_TRANSLATE(2, 2);
730 	table |= ICE_UP_TABLE_TRANSLATE(3, 3);
731 	table |= ICE_UP_TABLE_TRANSLATE(4, 4);
732 	table |= ICE_UP_TABLE_TRANSLATE(5, 5);
733 	table |= ICE_UP_TABLE_TRANSLATE(6, 6);
734 	table |= ICE_UP_TABLE_TRANSLATE(7, 7);
735 	ctxt->info.ingress_table = cpu_to_le32(table);
736 	ctxt->info.egress_table = cpu_to_le32(table);
737 	/* Have 1:1 UP mapping for outer to inner UP table */
738 	ctxt->info.outer_up_table = cpu_to_le32(table);
739 	/* No Outer tag support outer_tag_flags remains to zero */
740 }
741 
742 /**
743  * ice_vsi_setup_q_map - Setup a VSI queue map
744  * @vsi: the VSI being configured
745  * @ctxt: VSI context structure
746  */
747 static void ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
748 {
749 	u16 offset = 0, qmap = 0, tx_count = 0, pow = 0;
750 	u16 num_txq_per_tc, num_rxq_per_tc;
751 	u16 qcount_tx = vsi->alloc_txq;
752 	u16 qcount_rx = vsi->alloc_rxq;
753 	bool ena_tc0 = false;
754 	u8 netdev_tc = 0;
755 	int i;
756 
757 	/* at least TC0 should be enabled by default */
758 	if (vsi->tc_cfg.numtc) {
759 		if (!(vsi->tc_cfg.ena_tc & BIT(0)))
760 			ena_tc0 = true;
761 	} else {
762 		ena_tc0 = true;
763 	}
764 
765 	if (ena_tc0) {
766 		vsi->tc_cfg.numtc++;
767 		vsi->tc_cfg.ena_tc |= 1;
768 	}
769 
770 	num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
771 	if (!num_rxq_per_tc)
772 		num_rxq_per_tc = 1;
773 	num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
774 	if (!num_txq_per_tc)
775 		num_txq_per_tc = 1;
776 
777 	/* find the (rounded up) power-of-2 of qcount */
778 	pow = (u16)order_base_2(num_rxq_per_tc);
779 
780 	/* TC mapping is a function of the number of Rx queues assigned to the
781 	 * VSI for each traffic class and the offset of these queues.
782 	 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
783 	 * queues allocated to TC0. No:of queues is a power-of-2.
784 	 *
785 	 * If TC is not enabled, the queue offset is set to 0, and allocate one
786 	 * queue, this way, traffic for the given TC will be sent to the default
787 	 * queue.
788 	 *
789 	 * Setup number and offset of Rx queues for all TCs for the VSI
790 	 */
791 	ice_for_each_traffic_class(i) {
792 		if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
793 			/* TC is not enabled */
794 			vsi->tc_cfg.tc_info[i].qoffset = 0;
795 			vsi->tc_cfg.tc_info[i].qcount_rx = 1;
796 			vsi->tc_cfg.tc_info[i].qcount_tx = 1;
797 			vsi->tc_cfg.tc_info[i].netdev_tc = 0;
798 			ctxt->info.tc_mapping[i] = 0;
799 			continue;
800 		}
801 
802 		/* TC is enabled */
803 		vsi->tc_cfg.tc_info[i].qoffset = offset;
804 		vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
805 		vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
806 		vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
807 
808 		qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
809 			ICE_AQ_VSI_TC_Q_OFFSET_M) |
810 			((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
811 			 ICE_AQ_VSI_TC_Q_NUM_M);
812 		offset += num_rxq_per_tc;
813 		tx_count += num_txq_per_tc;
814 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
815 	}
816 
817 	/* if offset is non-zero, means it is calculated correctly based on
818 	 * enabled TCs for a given VSI otherwise qcount_rx will always
819 	 * be correct and non-zero because it is based off - VSI's
820 	 * allocated Rx queues which is at least 1 (hence qcount_tx will be
821 	 * at least 1)
822 	 */
823 	if (offset)
824 		vsi->num_rxq = offset;
825 	else
826 		vsi->num_rxq = num_rxq_per_tc;
827 
828 	vsi->num_txq = tx_count;
829 
830 	if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
831 		dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
832 		/* since there is a chance that num_rxq could have been changed
833 		 * in the above for loop, make num_txq equal to num_rxq.
834 		 */
835 		vsi->num_txq = vsi->num_rxq;
836 	}
837 
838 	/* Rx queue mapping */
839 	ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
840 	/* q_mapping buffer holds the info for the first queue allocated for
841 	 * this VSI in the PF space and also the number of queues associated
842 	 * with this VSI.
843 	 */
844 	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
845 	ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
846 }
847 
848 /**
849  * ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
850  * @ctxt: the VSI context being set
851  * @vsi: the VSI being configured
852  */
853 static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
854 {
855 	u8 dflt_q_group, dflt_q_prio;
856 	u16 dflt_q, report_q, val;
857 
858 	if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
859 	    vsi->type != ICE_VSI_VF)
860 		return;
861 
862 	val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
863 	ctxt->info.valid_sections |= cpu_to_le16(val);
864 	dflt_q = 0;
865 	dflt_q_group = 0;
866 	report_q = 0;
867 	dflt_q_prio = 0;
868 
869 	/* enable flow director filtering/programming */
870 	val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
871 	ctxt->info.fd_options = cpu_to_le16(val);
872 	/* max of allocated flow director filters */
873 	ctxt->info.max_fd_fltr_dedicated =
874 			cpu_to_le16(vsi->num_gfltr);
875 	/* max of shared flow director filters any VSI may program */
876 	ctxt->info.max_fd_fltr_shared =
877 			cpu_to_le16(vsi->num_bfltr);
878 	/* default queue index within the VSI of the default FD */
879 	val = ((dflt_q << ICE_AQ_VSI_FD_DEF_Q_S) &
880 	       ICE_AQ_VSI_FD_DEF_Q_M);
881 	/* target queue or queue group to the FD filter */
882 	val |= ((dflt_q_group << ICE_AQ_VSI_FD_DEF_GRP_S) &
883 		ICE_AQ_VSI_FD_DEF_GRP_M);
884 	ctxt->info.fd_def_q = cpu_to_le16(val);
885 	/* queue index on which FD filter completion is reported */
886 	val = ((report_q << ICE_AQ_VSI_FD_REPORT_Q_S) &
887 	       ICE_AQ_VSI_FD_REPORT_Q_M);
888 	/* priority of the default qindex action */
889 	val |= ((dflt_q_prio << ICE_AQ_VSI_FD_DEF_PRIORITY_S) &
890 		ICE_AQ_VSI_FD_DEF_PRIORITY_M);
891 	ctxt->info.fd_report_opt = cpu_to_le16(val);
892 }
893 
894 /**
895  * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
896  * @ctxt: the VSI context being set
897  * @vsi: the VSI being configured
898  */
899 static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
900 {
901 	u8 lut_type, hash_type;
902 	struct device *dev;
903 	struct ice_pf *pf;
904 
905 	pf = vsi->back;
906 	dev = ice_pf_to_dev(pf);
907 
908 	switch (vsi->type) {
909 	case ICE_VSI_PF:
910 		/* PF VSI will inherit RSS instance of PF */
911 		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
912 		hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
913 		break;
914 	case ICE_VSI_VF:
915 		/* VF VSI will gets a small RSS table which is a VSI LUT type */
916 		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
917 		hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
918 		break;
919 	default:
920 		dev_dbg(dev, "Unsupported VSI type %s\n",
921 			ice_vsi_type_str(vsi->type));
922 		return;
923 	}
924 
925 	ctxt->info.q_opt_rss = ((lut_type << ICE_AQ_VSI_Q_OPT_RSS_LUT_S) &
926 				ICE_AQ_VSI_Q_OPT_RSS_LUT_M) |
927 				((hash_type << ICE_AQ_VSI_Q_OPT_RSS_HASH_S) &
928 				 ICE_AQ_VSI_Q_OPT_RSS_HASH_M);
929 }
930 
931 /**
932  * ice_vsi_init - Create and initialize a VSI
933  * @vsi: the VSI being configured
934  * @init_vsi: is this call creating a VSI
935  *
936  * This initializes a VSI context depending on the VSI type to be added and
937  * passes it down to the add_vsi aq command to create a new VSI.
938  */
939 static int ice_vsi_init(struct ice_vsi *vsi, bool init_vsi)
940 {
941 	struct ice_pf *pf = vsi->back;
942 	struct ice_hw *hw = &pf->hw;
943 	struct ice_vsi_ctx *ctxt;
944 	struct device *dev;
945 	int ret = 0;
946 
947 	dev = ice_pf_to_dev(pf);
948 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
949 	if (!ctxt)
950 		return -ENOMEM;
951 
952 	switch (vsi->type) {
953 	case ICE_VSI_CTRL:
954 	case ICE_VSI_LB:
955 	case ICE_VSI_PF:
956 		ctxt->flags = ICE_AQ_VSI_TYPE_PF;
957 		break;
958 	case ICE_VSI_VF:
959 		ctxt->flags = ICE_AQ_VSI_TYPE_VF;
960 		/* VF number here is the absolute VF number (0-255) */
961 		ctxt->vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
962 		break;
963 	default:
964 		ret = -ENODEV;
965 		goto out;
966 	}
967 
968 	ice_set_dflt_vsi_ctx(ctxt);
969 	if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
970 		ice_set_fd_vsi_ctx(ctxt, vsi);
971 	/* if the switch is in VEB mode, allow VSI loopback */
972 	if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
973 		ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
974 
975 	/* Set LUT type and HASH type if RSS is enabled */
976 	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
977 	    vsi->type != ICE_VSI_CTRL) {
978 		ice_set_rss_vsi_ctx(ctxt, vsi);
979 		/* if updating VSI context, make sure to set valid_section:
980 		 * to indicate which section of VSI context being updated
981 		 */
982 		if (!init_vsi)
983 			ctxt->info.valid_sections |=
984 				cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
985 	}
986 
987 	ctxt->info.sw_id = vsi->port_info->sw_id;
988 	ice_vsi_setup_q_map(vsi, ctxt);
989 	if (!init_vsi) /* means VSI being updated */
990 		/* must to indicate which section of VSI context are
991 		 * being modified
992 		 */
993 		ctxt->info.valid_sections |=
994 			cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
995 
996 	/* enable/disable MAC and VLAN anti-spoof when spoofchk is on/off
997 	 * respectively
998 	 */
999 	if (vsi->type == ICE_VSI_VF) {
1000 		ctxt->info.valid_sections |=
1001 			cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1002 		if (pf->vf[vsi->vf_id].spoofchk) {
1003 			ctxt->info.sec_flags |=
1004 				ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
1005 				(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
1006 				 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
1007 		} else {
1008 			ctxt->info.sec_flags &=
1009 				~(ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
1010 				  (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
1011 				   ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S));
1012 		}
1013 	}
1014 
1015 	/* Allow control frames out of main VSI */
1016 	if (vsi->type == ICE_VSI_PF) {
1017 		ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1018 		ctxt->info.valid_sections |=
1019 			cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1020 	}
1021 
1022 	if (init_vsi) {
1023 		ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1024 		if (ret) {
1025 			dev_err(dev, "Add VSI failed, err %d\n", ret);
1026 			ret = -EIO;
1027 			goto out;
1028 		}
1029 	} else {
1030 		ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1031 		if (ret) {
1032 			dev_err(dev, "Update VSI failed, err %d\n", ret);
1033 			ret = -EIO;
1034 			goto out;
1035 		}
1036 	}
1037 
1038 	/* keep context for update VSI operations */
1039 	vsi->info = ctxt->info;
1040 
1041 	/* record VSI number returned */
1042 	vsi->vsi_num = ctxt->vsi_num;
1043 
1044 out:
1045 	kfree(ctxt);
1046 	return ret;
1047 }
1048 
1049 /**
1050  * ice_free_res - free a block of resources
1051  * @res: pointer to the resource
1052  * @index: starting index previously returned by ice_get_res
1053  * @id: identifier to track owner
1054  *
1055  * Returns number of resources freed
1056  */
1057 int ice_free_res(struct ice_res_tracker *res, u16 index, u16 id)
1058 {
1059 	int count = 0;
1060 	int i;
1061 
1062 	if (!res || index >= res->end)
1063 		return -EINVAL;
1064 
1065 	id |= ICE_RES_VALID_BIT;
1066 	for (i = index; i < res->end && res->list[i] == id; i++) {
1067 		res->list[i] = 0;
1068 		count++;
1069 	}
1070 
1071 	return count;
1072 }
1073 
1074 /**
1075  * ice_search_res - Search the tracker for a block of resources
1076  * @res: pointer to the resource
1077  * @needed: size of the block needed
1078  * @id: identifier to track owner
1079  *
1080  * Returns the base item index of the block, or -ENOMEM for error
1081  */
1082 static int ice_search_res(struct ice_res_tracker *res, u16 needed, u16 id)
1083 {
1084 	u16 start = 0, end = 0;
1085 
1086 	if (needed > res->end)
1087 		return -ENOMEM;
1088 
1089 	id |= ICE_RES_VALID_BIT;
1090 
1091 	do {
1092 		/* skip already allocated entries */
1093 		if (res->list[end++] & ICE_RES_VALID_BIT) {
1094 			start = end;
1095 			if ((start + needed) > res->end)
1096 				break;
1097 		}
1098 
1099 		if (end == (start + needed)) {
1100 			int i = start;
1101 
1102 			/* there was enough, so assign it to the requestor */
1103 			while (i != end)
1104 				res->list[i++] = id;
1105 
1106 			return start;
1107 		}
1108 	} while (end < res->end);
1109 
1110 	return -ENOMEM;
1111 }
1112 
1113 /**
1114  * ice_get_free_res_count - Get free count from a resource tracker
1115  * @res: Resource tracker instance
1116  */
1117 static u16 ice_get_free_res_count(struct ice_res_tracker *res)
1118 {
1119 	u16 i, count = 0;
1120 
1121 	for (i = 0; i < res->end; i++)
1122 		if (!(res->list[i] & ICE_RES_VALID_BIT))
1123 			count++;
1124 
1125 	return count;
1126 }
1127 
1128 /**
1129  * ice_get_res - get a block of resources
1130  * @pf: board private structure
1131  * @res: pointer to the resource
1132  * @needed: size of the block needed
1133  * @id: identifier to track owner
1134  *
1135  * Returns the base item index of the block, or negative for error
1136  */
1137 int
1138 ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res, u16 needed, u16 id)
1139 {
1140 	if (!res || !pf)
1141 		return -EINVAL;
1142 
1143 	if (!needed || needed > res->num_entries || id >= ICE_RES_VALID_BIT) {
1144 		dev_err(ice_pf_to_dev(pf), "param err: needed=%d, num_entries = %d id=0x%04x\n",
1145 			needed, res->num_entries, id);
1146 		return -EINVAL;
1147 	}
1148 
1149 	return ice_search_res(res, needed, id);
1150 }
1151 
1152 /**
1153  * ice_vsi_setup_vector_base - Set up the base vector for the given VSI
1154  * @vsi: ptr to the VSI
1155  *
1156  * This should only be called after ice_vsi_alloc() which allocates the
1157  * corresponding SW VSI structure and initializes num_queue_pairs for the
1158  * newly allocated VSI.
1159  *
1160  * Returns 0 on success or negative on failure
1161  */
1162 static int ice_vsi_setup_vector_base(struct ice_vsi *vsi)
1163 {
1164 	struct ice_pf *pf = vsi->back;
1165 	struct device *dev;
1166 	u16 num_q_vectors;
1167 	int base;
1168 
1169 	dev = ice_pf_to_dev(pf);
1170 	/* SRIOV doesn't grab irq_tracker entries for each VSI */
1171 	if (vsi->type == ICE_VSI_VF)
1172 		return 0;
1173 
1174 	if (vsi->base_vector) {
1175 		dev_dbg(dev, "VSI %d has non-zero base vector %d\n",
1176 			vsi->vsi_num, vsi->base_vector);
1177 		return -EEXIST;
1178 	}
1179 
1180 	num_q_vectors = vsi->num_q_vectors;
1181 	/* reserve slots from OS requested IRQs */
1182 	if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID) {
1183 		struct ice_vf *vf;
1184 		int i;
1185 
1186 		ice_for_each_vf(pf, i) {
1187 			vf = &pf->vf[i];
1188 			if (i != vsi->vf_id && vf->ctrl_vsi_idx != ICE_NO_VSI) {
1189 				base = pf->vsi[vf->ctrl_vsi_idx]->base_vector;
1190 				break;
1191 			}
1192 		}
1193 		if (i == pf->num_alloc_vfs)
1194 			base = ice_get_res(pf, pf->irq_tracker, num_q_vectors,
1195 					   ICE_RES_VF_CTRL_VEC_ID);
1196 	} else {
1197 		base = ice_get_res(pf, pf->irq_tracker, num_q_vectors,
1198 				   vsi->idx);
1199 	}
1200 
1201 	if (base < 0) {
1202 		dev_err(dev, "%d MSI-X interrupts available. %s %d failed to get %d MSI-X vectors\n",
1203 			ice_get_free_res_count(pf->irq_tracker),
1204 			ice_vsi_type_str(vsi->type), vsi->idx, num_q_vectors);
1205 		return -ENOENT;
1206 	}
1207 	vsi->base_vector = (u16)base;
1208 	pf->num_avail_sw_msix -= num_q_vectors;
1209 
1210 	return 0;
1211 }
1212 
1213 /**
1214  * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1215  * @vsi: the VSI having rings deallocated
1216  */
1217 static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1218 {
1219 	int i;
1220 
1221 	/* Avoid stale references by clearing map from vector to ring */
1222 	if (vsi->q_vectors) {
1223 		ice_for_each_q_vector(vsi, i) {
1224 			struct ice_q_vector *q_vector = vsi->q_vectors[i];
1225 
1226 			if (q_vector) {
1227 				q_vector->tx.ring = NULL;
1228 				q_vector->rx.ring = NULL;
1229 			}
1230 		}
1231 	}
1232 
1233 	if (vsi->tx_rings) {
1234 		for (i = 0; i < vsi->alloc_txq; i++) {
1235 			if (vsi->tx_rings[i]) {
1236 				kfree_rcu(vsi->tx_rings[i], rcu);
1237 				WRITE_ONCE(vsi->tx_rings[i], NULL);
1238 			}
1239 		}
1240 	}
1241 	if (vsi->rx_rings) {
1242 		for (i = 0; i < vsi->alloc_rxq; i++) {
1243 			if (vsi->rx_rings[i]) {
1244 				kfree_rcu(vsi->rx_rings[i], rcu);
1245 				WRITE_ONCE(vsi->rx_rings[i], NULL);
1246 			}
1247 		}
1248 	}
1249 }
1250 
1251 /**
1252  * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1253  * @vsi: VSI which is having rings allocated
1254  */
1255 static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1256 {
1257 	struct ice_pf *pf = vsi->back;
1258 	struct device *dev;
1259 	u16 i;
1260 
1261 	dev = ice_pf_to_dev(pf);
1262 	/* Allocate Tx rings */
1263 	for (i = 0; i < vsi->alloc_txq; i++) {
1264 		struct ice_ring *ring;
1265 
1266 		/* allocate with kzalloc(), free with kfree_rcu() */
1267 		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1268 
1269 		if (!ring)
1270 			goto err_out;
1271 
1272 		ring->q_index = i;
1273 		ring->reg_idx = vsi->txq_map[i];
1274 		ring->ring_active = false;
1275 		ring->vsi = vsi;
1276 		ring->dev = dev;
1277 		ring->count = vsi->num_tx_desc;
1278 		WRITE_ONCE(vsi->tx_rings[i], ring);
1279 	}
1280 
1281 	/* Allocate Rx rings */
1282 	for (i = 0; i < vsi->alloc_rxq; i++) {
1283 		struct ice_ring *ring;
1284 
1285 		/* allocate with kzalloc(), free with kfree_rcu() */
1286 		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1287 		if (!ring)
1288 			goto err_out;
1289 
1290 		ring->q_index = i;
1291 		ring->reg_idx = vsi->rxq_map[i];
1292 		ring->ring_active = false;
1293 		ring->vsi = vsi;
1294 		ring->netdev = vsi->netdev;
1295 		ring->dev = dev;
1296 		ring->count = vsi->num_rx_desc;
1297 		WRITE_ONCE(vsi->rx_rings[i], ring);
1298 	}
1299 
1300 	return 0;
1301 
1302 err_out:
1303 	ice_vsi_clear_rings(vsi);
1304 	return -ENOMEM;
1305 }
1306 
1307 /**
1308  * ice_vsi_manage_rss_lut - disable/enable RSS
1309  * @vsi: the VSI being changed
1310  * @ena: boolean value indicating if this is an enable or disable request
1311  *
1312  * In the event of disable request for RSS, this function will zero out RSS
1313  * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1314  * LUT.
1315  */
1316 void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1317 {
1318 	u8 *lut;
1319 
1320 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1321 	if (!lut)
1322 		return;
1323 
1324 	if (ena) {
1325 		if (vsi->rss_lut_user)
1326 			memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1327 		else
1328 			ice_fill_rss_lut(lut, vsi->rss_table_size,
1329 					 vsi->rss_size);
1330 	}
1331 
1332 	ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1333 	kfree(lut);
1334 }
1335 
1336 /**
1337  * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1338  * @vsi: VSI to be configured
1339  */
1340 static int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1341 {
1342 	struct ice_pf *pf = vsi->back;
1343 	struct device *dev;
1344 	u8 *lut, *key;
1345 	int err;
1346 
1347 	dev = ice_pf_to_dev(pf);
1348 	vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1349 
1350 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1351 	if (!lut)
1352 		return -ENOMEM;
1353 
1354 	if (vsi->rss_lut_user)
1355 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1356 	else
1357 		ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1358 
1359 	err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1360 	if (err) {
1361 		dev_err(dev, "set_rss_lut failed, error %d\n", err);
1362 		goto ice_vsi_cfg_rss_exit;
1363 	}
1364 
1365 	key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1366 	if (!key) {
1367 		err = -ENOMEM;
1368 		goto ice_vsi_cfg_rss_exit;
1369 	}
1370 
1371 	if (vsi->rss_hkey_user)
1372 		memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1373 	else
1374 		netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1375 
1376 	err = ice_set_rss_key(vsi, key);
1377 	if (err)
1378 		dev_err(dev, "set_rss_key failed, error %d\n", err);
1379 
1380 	kfree(key);
1381 ice_vsi_cfg_rss_exit:
1382 	kfree(lut);
1383 	return err;
1384 }
1385 
1386 /**
1387  * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1388  * @vsi: VSI to be configured
1389  *
1390  * This function will only be called during the VF VSI setup. Upon successful
1391  * completion of package download, this function will configure default RSS
1392  * input sets for VF VSI.
1393  */
1394 static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1395 {
1396 	struct ice_pf *pf = vsi->back;
1397 	enum ice_status status;
1398 	struct device *dev;
1399 
1400 	dev = ice_pf_to_dev(pf);
1401 	if (ice_is_safe_mode(pf)) {
1402 		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1403 			vsi->vsi_num);
1404 		return;
1405 	}
1406 
1407 	status = ice_add_avf_rss_cfg(&pf->hw, vsi->idx, ICE_DEFAULT_RSS_HENA);
1408 	if (status)
1409 		dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %s\n",
1410 			vsi->vsi_num, ice_stat_str(status));
1411 }
1412 
1413 /**
1414  * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1415  * @vsi: VSI to be configured
1416  *
1417  * This function will only be called after successful download package call
1418  * during initialization of PF. Since the downloaded package will erase the
1419  * RSS section, this function will configure RSS input sets for different
1420  * flow types. The last profile added has the highest priority, therefore 2
1421  * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1422  * (i.e. IPv4 src/dst TCP src/dst port).
1423  */
1424 static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1425 {
1426 	u16 vsi_handle = vsi->idx, vsi_num = vsi->vsi_num;
1427 	struct ice_pf *pf = vsi->back;
1428 	struct ice_hw *hw = &pf->hw;
1429 	enum ice_status status;
1430 	struct device *dev;
1431 
1432 	dev = ice_pf_to_dev(pf);
1433 	if (ice_is_safe_mode(pf)) {
1434 		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1435 			vsi_num);
1436 		return;
1437 	}
1438 	/* configure RSS for IPv4 with input set IP src/dst */
1439 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1440 				 ICE_FLOW_SEG_HDR_IPV4);
1441 	if (status)
1442 		dev_dbg(dev, "ice_add_rss_cfg failed for ipv4 flow, vsi = %d, error = %s\n",
1443 			vsi_num, ice_stat_str(status));
1444 
1445 	/* configure RSS for IPv6 with input set IPv6 src/dst */
1446 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1447 				 ICE_FLOW_SEG_HDR_IPV6);
1448 	if (status)
1449 		dev_dbg(dev, "ice_add_rss_cfg failed for ipv6 flow, vsi = %d, error = %s\n",
1450 			vsi_num, ice_stat_str(status));
1451 
1452 	/* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1453 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV4,
1454 				 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4);
1455 	if (status)
1456 		dev_dbg(dev, "ice_add_rss_cfg failed for tcp4 flow, vsi = %d, error = %s\n",
1457 			vsi_num, ice_stat_str(status));
1458 
1459 	/* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1460 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV4,
1461 				 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4);
1462 	if (status)
1463 		dev_dbg(dev, "ice_add_rss_cfg failed for udp4 flow, vsi = %d, error = %s\n",
1464 			vsi_num, ice_stat_str(status));
1465 
1466 	/* configure RSS for sctp4 with input set IP src/dst */
1467 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1468 				 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4);
1469 	if (status)
1470 		dev_dbg(dev, "ice_add_rss_cfg failed for sctp4 flow, vsi = %d, error = %s\n",
1471 			vsi_num, ice_stat_str(status));
1472 
1473 	/* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1474 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV6,
1475 				 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6);
1476 	if (status)
1477 		dev_dbg(dev, "ice_add_rss_cfg failed for tcp6 flow, vsi = %d, error = %s\n",
1478 			vsi_num, ice_stat_str(status));
1479 
1480 	/* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1481 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV6,
1482 				 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6);
1483 	if (status)
1484 		dev_dbg(dev, "ice_add_rss_cfg failed for udp6 flow, vsi = %d, error = %s\n",
1485 			vsi_num, ice_stat_str(status));
1486 
1487 	/* configure RSS for sctp6 with input set IPv6 src/dst */
1488 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1489 				 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6);
1490 	if (status)
1491 		dev_dbg(dev, "ice_add_rss_cfg failed for sctp6 flow, vsi = %d, error = %s\n",
1492 			vsi_num, ice_stat_str(status));
1493 }
1494 
1495 /**
1496  * ice_pf_state_is_nominal - checks the PF for nominal state
1497  * @pf: pointer to PF to check
1498  *
1499  * Check the PF's state for a collection of bits that would indicate
1500  * the PF is in a state that would inhibit normal operation for
1501  * driver functionality.
1502  *
1503  * Returns true if PF is in a nominal state, false otherwise
1504  */
1505 bool ice_pf_state_is_nominal(struct ice_pf *pf)
1506 {
1507 	DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1508 
1509 	if (!pf)
1510 		return false;
1511 
1512 	bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1513 	if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1514 		return false;
1515 
1516 	return true;
1517 }
1518 
1519 /**
1520  * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1521  * @vsi: the VSI to be updated
1522  */
1523 void ice_update_eth_stats(struct ice_vsi *vsi)
1524 {
1525 	struct ice_eth_stats *prev_es, *cur_es;
1526 	struct ice_hw *hw = &vsi->back->hw;
1527 	u16 vsi_num = vsi->vsi_num;    /* HW absolute index of a VSI */
1528 
1529 	prev_es = &vsi->eth_stats_prev;
1530 	cur_es = &vsi->eth_stats;
1531 
1532 	ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1533 			  &prev_es->rx_bytes, &cur_es->rx_bytes);
1534 
1535 	ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1536 			  &prev_es->rx_unicast, &cur_es->rx_unicast);
1537 
1538 	ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1539 			  &prev_es->rx_multicast, &cur_es->rx_multicast);
1540 
1541 	ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1542 			  &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1543 
1544 	ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1545 			  &prev_es->rx_discards, &cur_es->rx_discards);
1546 
1547 	ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1548 			  &prev_es->tx_bytes, &cur_es->tx_bytes);
1549 
1550 	ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1551 			  &prev_es->tx_unicast, &cur_es->tx_unicast);
1552 
1553 	ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1554 			  &prev_es->tx_multicast, &cur_es->tx_multicast);
1555 
1556 	ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1557 			  &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1558 
1559 	ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1560 			  &prev_es->tx_errors, &cur_es->tx_errors);
1561 
1562 	vsi->stat_offsets_loaded = true;
1563 }
1564 
1565 /**
1566  * ice_vsi_add_vlan - Add VSI membership for given VLAN
1567  * @vsi: the VSI being configured
1568  * @vid: VLAN ID to be added
1569  * @action: filter action to be performed on match
1570  */
1571 int
1572 ice_vsi_add_vlan(struct ice_vsi *vsi, u16 vid, enum ice_sw_fwd_act_type action)
1573 {
1574 	struct ice_pf *pf = vsi->back;
1575 	struct device *dev;
1576 	int err = 0;
1577 
1578 	dev = ice_pf_to_dev(pf);
1579 
1580 	if (!ice_fltr_add_vlan(vsi, vid, action)) {
1581 		vsi->num_vlan++;
1582 	} else {
1583 		err = -ENODEV;
1584 		dev_err(dev, "Failure Adding VLAN %d on VSI %i\n", vid,
1585 			vsi->vsi_num);
1586 	}
1587 
1588 	return err;
1589 }
1590 
1591 /**
1592  * ice_vsi_kill_vlan - Remove VSI membership for a given VLAN
1593  * @vsi: the VSI being configured
1594  * @vid: VLAN ID to be removed
1595  *
1596  * Returns 0 on success and negative on failure
1597  */
1598 int ice_vsi_kill_vlan(struct ice_vsi *vsi, u16 vid)
1599 {
1600 	struct ice_pf *pf = vsi->back;
1601 	enum ice_status status;
1602 	struct device *dev;
1603 	int err = 0;
1604 
1605 	dev = ice_pf_to_dev(pf);
1606 
1607 	status = ice_fltr_remove_vlan(vsi, vid, ICE_FWD_TO_VSI);
1608 	if (!status) {
1609 		vsi->num_vlan--;
1610 	} else if (status == ICE_ERR_DOES_NOT_EXIST) {
1611 		dev_dbg(dev, "Failed to remove VLAN %d on VSI %i, it does not exist, status: %s\n",
1612 			vid, vsi->vsi_num, ice_stat_str(status));
1613 	} else {
1614 		dev_err(dev, "Error removing VLAN %d on vsi %i error: %s\n",
1615 			vid, vsi->vsi_num, ice_stat_str(status));
1616 		err = -EIO;
1617 	}
1618 
1619 	return err;
1620 }
1621 
1622 /**
1623  * ice_vsi_cfg_frame_size - setup max frame size and Rx buffer length
1624  * @vsi: VSI
1625  */
1626 void ice_vsi_cfg_frame_size(struct ice_vsi *vsi)
1627 {
1628 	if (!vsi->netdev || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) {
1629 		vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1630 		vsi->rx_buf_len = ICE_RXBUF_2048;
1631 #if (PAGE_SIZE < 8192)
1632 	} else if (!ICE_2K_TOO_SMALL_WITH_PADDING &&
1633 		   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
1634 		vsi->max_frame = ICE_RXBUF_1536 - NET_IP_ALIGN;
1635 		vsi->rx_buf_len = ICE_RXBUF_1536 - NET_IP_ALIGN;
1636 #endif
1637 	} else {
1638 		vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1639 #if (PAGE_SIZE < 8192)
1640 		vsi->rx_buf_len = ICE_RXBUF_3072;
1641 #else
1642 		vsi->rx_buf_len = ICE_RXBUF_2048;
1643 #endif
1644 	}
1645 }
1646 
1647 /**
1648  * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1649  * @hw: HW pointer
1650  * @pf_q: index of the Rx queue in the PF's queue space
1651  * @rxdid: flexible descriptor RXDID
1652  * @prio: priority for the RXDID for this queue
1653  */
1654 void
1655 ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio)
1656 {
1657 	int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1658 
1659 	/* clear any previous values */
1660 	regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1661 		    QRXFLXP_CNTXT_RXDID_PRIO_M |
1662 		    QRXFLXP_CNTXT_TS_M);
1663 
1664 	regval |= (rxdid << QRXFLXP_CNTXT_RXDID_IDX_S) &
1665 		QRXFLXP_CNTXT_RXDID_IDX_M;
1666 
1667 	regval |= (prio << QRXFLXP_CNTXT_RXDID_PRIO_S) &
1668 		QRXFLXP_CNTXT_RXDID_PRIO_M;
1669 
1670 	wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1671 }
1672 
1673 /**
1674  * ice_vsi_cfg_rxqs - Configure the VSI for Rx
1675  * @vsi: the VSI being configured
1676  *
1677  * Return 0 on success and a negative value on error
1678  * Configure the Rx VSI for operation.
1679  */
1680 int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
1681 {
1682 	u16 i;
1683 
1684 	if (vsi->type == ICE_VSI_VF)
1685 		goto setup_rings;
1686 
1687 	ice_vsi_cfg_frame_size(vsi);
1688 setup_rings:
1689 	/* set up individual rings */
1690 	for (i = 0; i < vsi->num_rxq; i++) {
1691 		int err;
1692 
1693 		err = ice_setup_rx_ctx(vsi->rx_rings[i]);
1694 		if (err) {
1695 			dev_err(ice_pf_to_dev(vsi->back), "ice_setup_rx_ctx failed for RxQ %d, err %d\n",
1696 				i, err);
1697 			return err;
1698 		}
1699 	}
1700 
1701 	return 0;
1702 }
1703 
1704 /**
1705  * ice_vsi_cfg_txqs - Configure the VSI for Tx
1706  * @vsi: the VSI being configured
1707  * @rings: Tx ring array to be configured
1708  *
1709  * Return 0 on success and a negative value on error
1710  * Configure the Tx VSI for operation.
1711  */
1712 static int
1713 ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_ring **rings)
1714 {
1715 	struct ice_aqc_add_tx_qgrp *qg_buf;
1716 	u16 q_idx = 0;
1717 	int err = 0;
1718 
1719 	qg_buf = kzalloc(struct_size(qg_buf, txqs, 1), GFP_KERNEL);
1720 	if (!qg_buf)
1721 		return -ENOMEM;
1722 
1723 	qg_buf->num_txqs = 1;
1724 
1725 	for (q_idx = 0; q_idx < vsi->num_txq; q_idx++) {
1726 		err = ice_vsi_cfg_txq(vsi, rings[q_idx], qg_buf);
1727 		if (err)
1728 			goto err_cfg_txqs;
1729 	}
1730 
1731 err_cfg_txqs:
1732 	kfree(qg_buf);
1733 	return err;
1734 }
1735 
1736 /**
1737  * ice_vsi_cfg_lan_txqs - Configure the VSI for Tx
1738  * @vsi: the VSI being configured
1739  *
1740  * Return 0 on success and a negative value on error
1741  * Configure the Tx VSI for operation.
1742  */
1743 int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
1744 {
1745 	return ice_vsi_cfg_txqs(vsi, vsi->tx_rings);
1746 }
1747 
1748 /**
1749  * ice_vsi_cfg_xdp_txqs - Configure Tx queues dedicated for XDP in given VSI
1750  * @vsi: the VSI being configured
1751  *
1752  * Return 0 on success and a negative value on error
1753  * Configure the Tx queues dedicated for XDP in given VSI for operation.
1754  */
1755 int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
1756 {
1757 	int ret;
1758 	int i;
1759 
1760 	ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings);
1761 	if (ret)
1762 		return ret;
1763 
1764 	for (i = 0; i < vsi->num_xdp_txq; i++)
1765 		vsi->xdp_rings[i]->xsk_pool = ice_xsk_pool(vsi->xdp_rings[i]);
1766 
1767 	return ret;
1768 }
1769 
1770 /**
1771  * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1772  * @intrl: interrupt rate limit in usecs
1773  * @gran: interrupt rate limit granularity in usecs
1774  *
1775  * This function converts a decimal interrupt rate limit in usecs to the format
1776  * expected by firmware.
1777  */
1778 static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1779 {
1780 	u32 val = intrl / gran;
1781 
1782 	if (val)
1783 		return val | GLINT_RATE_INTRL_ENA_M;
1784 	return 0;
1785 }
1786 
1787 /**
1788  * ice_write_intrl - write throttle rate limit to interrupt specific register
1789  * @q_vector: pointer to interrupt specific structure
1790  * @intrl: throttle rate limit in microseconds to write
1791  */
1792 void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1793 {
1794 	struct ice_hw *hw = &q_vector->vsi->back->hw;
1795 
1796 	wr32(hw, GLINT_RATE(q_vector->reg_idx),
1797 	     ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1798 }
1799 
1800 /**
1801  * __ice_write_itr - write throttle rate to register
1802  * @q_vector: pointer to interrupt data structure
1803  * @rc: pointer to ring container
1804  * @itr: throttle rate in microseconds to write
1805  */
1806 static void __ice_write_itr(struct ice_q_vector *q_vector,
1807 			    struct ice_ring_container *rc, u16 itr)
1808 {
1809 	struct ice_hw *hw = &q_vector->vsi->back->hw;
1810 
1811 	wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
1812 	     ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
1813 }
1814 
1815 /**
1816  * ice_write_itr - write throttle rate to queue specific register
1817  * @rc: pointer to ring container
1818  * @itr: throttle rate in microseconds to write
1819  */
1820 void ice_write_itr(struct ice_ring_container *rc, u16 itr)
1821 {
1822 	struct ice_q_vector *q_vector;
1823 
1824 	if (!rc->ring)
1825 		return;
1826 
1827 	q_vector = rc->ring->q_vector;
1828 
1829 	__ice_write_itr(q_vector, rc, itr);
1830 }
1831 
1832 /**
1833  * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
1834  * @vsi: the VSI being configured
1835  *
1836  * This configures MSIX mode interrupts for the PF VSI, and should not be used
1837  * for the VF VSI.
1838  */
1839 void ice_vsi_cfg_msix(struct ice_vsi *vsi)
1840 {
1841 	struct ice_pf *pf = vsi->back;
1842 	struct ice_hw *hw = &pf->hw;
1843 	u16 txq = 0, rxq = 0;
1844 	int i, q;
1845 
1846 	for (i = 0; i < vsi->num_q_vectors; i++) {
1847 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
1848 		u16 reg_idx = q_vector->reg_idx;
1849 
1850 		ice_cfg_itr(hw, q_vector);
1851 
1852 		/* Both Transmit Queue Interrupt Cause Control register
1853 		 * and Receive Queue Interrupt Cause control register
1854 		 * expects MSIX_INDX field to be the vector index
1855 		 * within the function space and not the absolute
1856 		 * vector index across PF or across device.
1857 		 * For SR-IOV VF VSIs queue vector index always starts
1858 		 * with 1 since first vector index(0) is used for OICR
1859 		 * in VF space. Since VMDq and other PF VSIs are within
1860 		 * the PF function space, use the vector index that is
1861 		 * tracked for this PF.
1862 		 */
1863 		for (q = 0; q < q_vector->num_ring_tx; q++) {
1864 			ice_cfg_txq_interrupt(vsi, txq, reg_idx,
1865 					      q_vector->tx.itr_idx);
1866 			txq++;
1867 		}
1868 
1869 		for (q = 0; q < q_vector->num_ring_rx; q++) {
1870 			ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
1871 					      q_vector->rx.itr_idx);
1872 			rxq++;
1873 		}
1874 	}
1875 }
1876 
1877 /**
1878  * ice_vsi_manage_vlan_insertion - Manage VLAN insertion for the VSI for Tx
1879  * @vsi: the VSI being changed
1880  */
1881 int ice_vsi_manage_vlan_insertion(struct ice_vsi *vsi)
1882 {
1883 	struct ice_hw *hw = &vsi->back->hw;
1884 	struct ice_vsi_ctx *ctxt;
1885 	enum ice_status status;
1886 	int ret = 0;
1887 
1888 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1889 	if (!ctxt)
1890 		return -ENOMEM;
1891 
1892 	/* Here we are configuring the VSI to let the driver add VLAN tags by
1893 	 * setting vlan_flags to ICE_AQ_VSI_VLAN_MODE_ALL. The actual VLAN tag
1894 	 * insertion happens in the Tx hot path, in ice_tx_map.
1895 	 */
1896 	ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL;
1897 
1898 	/* Preserve existing VLAN strip setting */
1899 	ctxt->info.vlan_flags |= (vsi->info.vlan_flags &
1900 				  ICE_AQ_VSI_VLAN_EMOD_M);
1901 
1902 	ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID);
1903 
1904 	status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1905 	if (status) {
1906 		dev_err(ice_pf_to_dev(vsi->back), "update VSI for VLAN insert failed, err %s aq_err %s\n",
1907 			ice_stat_str(status),
1908 			ice_aq_str(hw->adminq.sq_last_status));
1909 		ret = -EIO;
1910 		goto out;
1911 	}
1912 
1913 	vsi->info.vlan_flags = ctxt->info.vlan_flags;
1914 out:
1915 	kfree(ctxt);
1916 	return ret;
1917 }
1918 
1919 /**
1920  * ice_vsi_manage_vlan_stripping - Manage VLAN stripping for the VSI for Rx
1921  * @vsi: the VSI being changed
1922  * @ena: boolean value indicating if this is a enable or disable request
1923  */
1924 int ice_vsi_manage_vlan_stripping(struct ice_vsi *vsi, bool ena)
1925 {
1926 	struct ice_hw *hw = &vsi->back->hw;
1927 	struct ice_vsi_ctx *ctxt;
1928 	enum ice_status status;
1929 	int ret = 0;
1930 
1931 	/* do not allow modifying VLAN stripping when a port VLAN is configured
1932 	 * on this VSI
1933 	 */
1934 	if (vsi->info.pvid)
1935 		return 0;
1936 
1937 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1938 	if (!ctxt)
1939 		return -ENOMEM;
1940 
1941 	/* Here we are configuring what the VSI should do with the VLAN tag in
1942 	 * the Rx packet. We can either leave the tag in the packet or put it in
1943 	 * the Rx descriptor.
1944 	 */
1945 	if (ena)
1946 		/* Strip VLAN tag from Rx packet and put it in the desc */
1947 		ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_EMOD_STR_BOTH;
1948 	else
1949 		/* Disable stripping. Leave tag in packet */
1950 		ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_EMOD_NOTHING;
1951 
1952 	/* Allow all packets untagged/tagged */
1953 	ctxt->info.vlan_flags |= ICE_AQ_VSI_VLAN_MODE_ALL;
1954 
1955 	ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID);
1956 
1957 	status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1958 	if (status) {
1959 		dev_err(ice_pf_to_dev(vsi->back), "update VSI for VLAN strip failed, ena = %d err %s aq_err %s\n",
1960 			ena, ice_stat_str(status),
1961 			ice_aq_str(hw->adminq.sq_last_status));
1962 		ret = -EIO;
1963 		goto out;
1964 	}
1965 
1966 	vsi->info.vlan_flags = ctxt->info.vlan_flags;
1967 out:
1968 	kfree(ctxt);
1969 	return ret;
1970 }
1971 
1972 /**
1973  * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
1974  * @vsi: the VSI whose rings are to be enabled
1975  *
1976  * Returns 0 on success and a negative value on error
1977  */
1978 int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
1979 {
1980 	return ice_vsi_ctrl_all_rx_rings(vsi, true);
1981 }
1982 
1983 /**
1984  * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
1985  * @vsi: the VSI whose rings are to be disabled
1986  *
1987  * Returns 0 on success and a negative value on error
1988  */
1989 int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
1990 {
1991 	return ice_vsi_ctrl_all_rx_rings(vsi, false);
1992 }
1993 
1994 /**
1995  * ice_vsi_stop_tx_rings - Disable Tx rings
1996  * @vsi: the VSI being configured
1997  * @rst_src: reset source
1998  * @rel_vmvf_num: Relative ID of VF/VM
1999  * @rings: Tx ring array to be stopped
2000  */
2001 static int
2002 ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2003 		      u16 rel_vmvf_num, struct ice_ring **rings)
2004 {
2005 	u16 q_idx;
2006 
2007 	if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2008 		return -EINVAL;
2009 
2010 	for (q_idx = 0; q_idx < vsi->num_txq; q_idx++) {
2011 		struct ice_txq_meta txq_meta = { };
2012 		int status;
2013 
2014 		if (!rings || !rings[q_idx])
2015 			return -EINVAL;
2016 
2017 		ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
2018 		status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2019 					      rings[q_idx], &txq_meta);
2020 
2021 		if (status)
2022 			return status;
2023 	}
2024 
2025 	return 0;
2026 }
2027 
2028 /**
2029  * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2030  * @vsi: the VSI being configured
2031  * @rst_src: reset source
2032  * @rel_vmvf_num: Relative ID of VF/VM
2033  */
2034 int
2035 ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2036 			  u16 rel_vmvf_num)
2037 {
2038 	return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings);
2039 }
2040 
2041 /**
2042  * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2043  * @vsi: the VSI being configured
2044  */
2045 int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2046 {
2047 	return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings);
2048 }
2049 
2050 /**
2051  * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
2052  * @vsi: VSI to check whether or not VLAN pruning is enabled.
2053  *
2054  * returns true if Rx VLAN pruning is enabled and false otherwise.
2055  */
2056 bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
2057 {
2058 	if (!vsi)
2059 		return false;
2060 
2061 	return (vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA);
2062 }
2063 
2064 /**
2065  * ice_cfg_vlan_pruning - enable or disable VLAN pruning on the VSI
2066  * @vsi: VSI to enable or disable VLAN pruning on
2067  * @ena: set to true to enable VLAN pruning and false to disable it
2068  * @vlan_promisc: enable valid security flags if not in VLAN promiscuous mode
2069  *
2070  * returns 0 if VSI is updated, negative otherwise
2071  */
2072 int ice_cfg_vlan_pruning(struct ice_vsi *vsi, bool ena, bool vlan_promisc)
2073 {
2074 	struct ice_vsi_ctx *ctxt;
2075 	struct ice_pf *pf;
2076 	int status;
2077 
2078 	if (!vsi)
2079 		return -EINVAL;
2080 
2081 	/* Don't enable VLAN pruning if the netdev is currently in promiscuous
2082 	 * mode. VLAN pruning will be enabled when the interface exits
2083 	 * promiscuous mode if any VLAN filters are active.
2084 	 */
2085 	if (vsi->netdev && vsi->netdev->flags & IFF_PROMISC && ena)
2086 		return 0;
2087 
2088 	pf = vsi->back;
2089 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
2090 	if (!ctxt)
2091 		return -ENOMEM;
2092 
2093 	ctxt->info = vsi->info;
2094 
2095 	if (ena)
2096 		ctxt->info.sw_flags2 |= ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
2097 	else
2098 		ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
2099 
2100 	if (!vlan_promisc)
2101 		ctxt->info.valid_sections =
2102 			cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
2103 
2104 	status = ice_update_vsi(&pf->hw, vsi->idx, ctxt, NULL);
2105 	if (status) {
2106 		netdev_err(vsi->netdev, "%sabling VLAN pruning on VSI handle: %d, VSI HW ID: %d failed, err = %s, aq_err = %s\n",
2107 			   ena ? "En" : "Dis", vsi->idx, vsi->vsi_num,
2108 			   ice_stat_str(status),
2109 			   ice_aq_str(pf->hw.adminq.sq_last_status));
2110 		goto err_out;
2111 	}
2112 
2113 	vsi->info.sw_flags2 = ctxt->info.sw_flags2;
2114 
2115 	kfree(ctxt);
2116 	return 0;
2117 
2118 err_out:
2119 	kfree(ctxt);
2120 	return -EIO;
2121 }
2122 
2123 static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2124 {
2125 	struct ice_dcbx_cfg *cfg = &vsi->port_info->qos_cfg.local_dcbx_cfg;
2126 
2127 	vsi->tc_cfg.ena_tc = ice_dcb_get_ena_tc(cfg);
2128 	vsi->tc_cfg.numtc = ice_dcb_get_num_tc(cfg);
2129 }
2130 
2131 /**
2132  * ice_vsi_set_q_vectors_reg_idx - set the HW register index for all q_vectors
2133  * @vsi: VSI to set the q_vectors register index on
2134  */
2135 static int
2136 ice_vsi_set_q_vectors_reg_idx(struct ice_vsi *vsi)
2137 {
2138 	u16 i;
2139 
2140 	if (!vsi || !vsi->q_vectors)
2141 		return -EINVAL;
2142 
2143 	ice_for_each_q_vector(vsi, i) {
2144 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2145 
2146 		if (!q_vector) {
2147 			dev_err(ice_pf_to_dev(vsi->back), "Failed to set reg_idx on q_vector %d VSI %d\n",
2148 				i, vsi->vsi_num);
2149 			goto clear_reg_idx;
2150 		}
2151 
2152 		if (vsi->type == ICE_VSI_VF) {
2153 			struct ice_vf *vf = &vsi->back->vf[vsi->vf_id];
2154 
2155 			q_vector->reg_idx = ice_calc_vf_reg_idx(vf, q_vector);
2156 		} else {
2157 			q_vector->reg_idx =
2158 				q_vector->v_idx + vsi->base_vector;
2159 		}
2160 	}
2161 
2162 	return 0;
2163 
2164 clear_reg_idx:
2165 	ice_for_each_q_vector(vsi, i) {
2166 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2167 
2168 		if (q_vector)
2169 			q_vector->reg_idx = 0;
2170 	}
2171 
2172 	return -EINVAL;
2173 }
2174 
2175 /**
2176  * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2177  * @vsi: the VSI being configured
2178  * @tx: bool to determine Tx or Rx rule
2179  * @create: bool to determine create or remove Rule
2180  */
2181 void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2182 {
2183 	enum ice_status (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2184 				    enum ice_sw_fwd_act_type act);
2185 	struct ice_pf *pf = vsi->back;
2186 	enum ice_status status;
2187 	struct device *dev;
2188 
2189 	dev = ice_pf_to_dev(pf);
2190 	eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2191 
2192 	if (tx) {
2193 		status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2194 				  ICE_DROP_PACKET);
2195 	} else {
2196 		if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2197 			status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2198 							  create);
2199 		} else {
2200 			status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2201 					  ICE_FWD_TO_VSI);
2202 		}
2203 	}
2204 
2205 	if (status)
2206 		dev_err(dev, "Fail %s %s LLDP rule on VSI %i error: %s\n",
2207 			create ? "adding" : "removing", tx ? "TX" : "RX",
2208 			vsi->vsi_num, ice_stat_str(status));
2209 }
2210 
2211 /**
2212  * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2213  * @vsi: pointer to the VSI
2214  *
2215  * This function will allocate new scheduler aggregator now if needed and will
2216  * move specified VSI into it.
2217  */
2218 static void ice_set_agg_vsi(struct ice_vsi *vsi)
2219 {
2220 	struct device *dev = ice_pf_to_dev(vsi->back);
2221 	struct ice_agg_node *agg_node_iter = NULL;
2222 	u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2223 	struct ice_agg_node *agg_node = NULL;
2224 	int node_offset, max_agg_nodes = 0;
2225 	struct ice_port_info *port_info;
2226 	struct ice_pf *pf = vsi->back;
2227 	u32 agg_node_id_start = 0;
2228 	enum ice_status status;
2229 
2230 	/* create (as needed) scheduler aggregator node and move VSI into
2231 	 * corresponding aggregator node
2232 	 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2233 	 * - VF aggregator nodes will contain VF VSI
2234 	 */
2235 	port_info = pf->hw.port_info;
2236 	if (!port_info)
2237 		return;
2238 
2239 	switch (vsi->type) {
2240 	case ICE_VSI_CTRL:
2241 	case ICE_VSI_LB:
2242 	case ICE_VSI_PF:
2243 		max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2244 		agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2245 		agg_node_iter = &pf->pf_agg_node[0];
2246 		break;
2247 	case ICE_VSI_VF:
2248 		/* user can create 'n' VFs on a given PF, but since max children
2249 		 * per aggregator node can be only 64. Following code handles
2250 		 * aggregator(s) for VF VSIs, either selects a agg_node which
2251 		 * was already created provided num_vsis < 64, otherwise
2252 		 * select next available node, which will be created
2253 		 */
2254 		max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2255 		agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2256 		agg_node_iter = &pf->vf_agg_node[0];
2257 		break;
2258 	default:
2259 		/* other VSI type, handle later if needed */
2260 		dev_dbg(dev, "unexpected VSI type %s\n",
2261 			ice_vsi_type_str(vsi->type));
2262 		return;
2263 	}
2264 
2265 	/* find the appropriate aggregator node */
2266 	for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2267 		/* see if we can find space in previously created
2268 		 * node if num_vsis < 64, otherwise skip
2269 		 */
2270 		if (agg_node_iter->num_vsis &&
2271 		    agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2272 			agg_node_iter++;
2273 			continue;
2274 		}
2275 
2276 		if (agg_node_iter->valid &&
2277 		    agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2278 			agg_id = agg_node_iter->agg_id;
2279 			agg_node = agg_node_iter;
2280 			break;
2281 		}
2282 
2283 		/* find unclaimed agg_id */
2284 		if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2285 			agg_id = node_offset + agg_node_id_start;
2286 			agg_node = agg_node_iter;
2287 			break;
2288 		}
2289 		/* move to next agg_node */
2290 		agg_node_iter++;
2291 	}
2292 
2293 	if (!agg_node)
2294 		return;
2295 
2296 	/* if selected aggregator node was not created, create it */
2297 	if (!agg_node->valid) {
2298 		status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2299 				     (u8)vsi->tc_cfg.ena_tc);
2300 		if (status) {
2301 			dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2302 				agg_id);
2303 			return;
2304 		}
2305 		/* aggregator node is created, store the neeeded info */
2306 		agg_node->valid = true;
2307 		agg_node->agg_id = agg_id;
2308 	}
2309 
2310 	/* move VSI to corresponding aggregator node */
2311 	status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2312 				     (u8)vsi->tc_cfg.ena_tc);
2313 	if (status) {
2314 		dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2315 			vsi->idx, agg_id);
2316 		return;
2317 	}
2318 
2319 	/* keep active children count for aggregator node */
2320 	agg_node->num_vsis++;
2321 
2322 	/* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2323 	 * to aggregator node
2324 	 */
2325 	vsi->agg_node = agg_node;
2326 	dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2327 		vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2328 		vsi->agg_node->num_vsis);
2329 }
2330 
2331 /**
2332  * ice_vsi_setup - Set up a VSI by a given type
2333  * @pf: board private structure
2334  * @pi: pointer to the port_info instance
2335  * @vsi_type: VSI type
2336  * @vf_id: defines VF ID to which this VSI connects. This field is meant to be
2337  *         used only for ICE_VSI_VF VSI type. For other VSI types, should
2338  *         fill-in ICE_INVAL_VFID as input.
2339  *
2340  * This allocates the sw VSI structure and its queue resources.
2341  *
2342  * Returns pointer to the successfully allocated and configured VSI sw struct on
2343  * success, NULL on failure.
2344  */
2345 struct ice_vsi *
2346 ice_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
2347 	      enum ice_vsi_type vsi_type, u16 vf_id)
2348 {
2349 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2350 	struct device *dev = ice_pf_to_dev(pf);
2351 	enum ice_status status;
2352 	struct ice_vsi *vsi;
2353 	int ret, i;
2354 
2355 	if (vsi_type == ICE_VSI_VF || vsi_type == ICE_VSI_CTRL)
2356 		vsi = ice_vsi_alloc(pf, vsi_type, vf_id);
2357 	else
2358 		vsi = ice_vsi_alloc(pf, vsi_type, ICE_INVAL_VFID);
2359 
2360 	if (!vsi) {
2361 		dev_err(dev, "could not allocate VSI\n");
2362 		return NULL;
2363 	}
2364 
2365 	vsi->port_info = pi;
2366 	vsi->vsw = pf->first_sw;
2367 	if (vsi->type == ICE_VSI_PF)
2368 		vsi->ethtype = ETH_P_PAUSE;
2369 
2370 	if (vsi->type == ICE_VSI_VF || vsi->type == ICE_VSI_CTRL)
2371 		vsi->vf_id = vf_id;
2372 
2373 	ice_alloc_fd_res(vsi);
2374 
2375 	if (ice_vsi_get_qs(vsi)) {
2376 		dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2377 			vsi->idx);
2378 		goto unroll_vsi_alloc;
2379 	}
2380 
2381 	/* set RSS capabilities */
2382 	ice_vsi_set_rss_params(vsi);
2383 
2384 	/* set TC configuration */
2385 	ice_vsi_set_tc_cfg(vsi);
2386 
2387 	/* create the VSI */
2388 	ret = ice_vsi_init(vsi, true);
2389 	if (ret)
2390 		goto unroll_get_qs;
2391 
2392 	switch (vsi->type) {
2393 	case ICE_VSI_CTRL:
2394 	case ICE_VSI_PF:
2395 		ret = ice_vsi_alloc_q_vectors(vsi);
2396 		if (ret)
2397 			goto unroll_vsi_init;
2398 
2399 		ret = ice_vsi_setup_vector_base(vsi);
2400 		if (ret)
2401 			goto unroll_alloc_q_vector;
2402 
2403 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
2404 		if (ret)
2405 			goto unroll_vector_base;
2406 
2407 		ret = ice_vsi_alloc_rings(vsi);
2408 		if (ret)
2409 			goto unroll_vector_base;
2410 
2411 		/* Always add VLAN ID 0 switch rule by default. This is needed
2412 		 * in order to allow all untagged and 0 tagged priority traffic
2413 		 * if Rx VLAN pruning is enabled. Also there are cases where we
2414 		 * don't get the call to add VLAN 0 via ice_vlan_rx_add_vid()
2415 		 * so this handles those cases (i.e. adding the PF to a bridge
2416 		 * without the 8021q module loaded).
2417 		 */
2418 		ret = ice_vsi_add_vlan(vsi, 0, ICE_FWD_TO_VSI);
2419 		if (ret)
2420 			goto unroll_clear_rings;
2421 
2422 		ice_vsi_map_rings_to_vectors(vsi);
2423 
2424 		/* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2425 		if (vsi->type != ICE_VSI_CTRL)
2426 			/* Do not exit if configuring RSS had an issue, at
2427 			 * least receive traffic on first queue. Hence no
2428 			 * need to capture return value
2429 			 */
2430 			if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2431 				ice_vsi_cfg_rss_lut_key(vsi);
2432 				ice_vsi_set_rss_flow_fld(vsi);
2433 			}
2434 		ice_init_arfs(vsi);
2435 		break;
2436 	case ICE_VSI_VF:
2437 		/* VF driver will take care of creating netdev for this type and
2438 		 * map queues to vectors through Virtchnl, PF driver only
2439 		 * creates a VSI and corresponding structures for bookkeeping
2440 		 * purpose
2441 		 */
2442 		ret = ice_vsi_alloc_q_vectors(vsi);
2443 		if (ret)
2444 			goto unroll_vsi_init;
2445 
2446 		ret = ice_vsi_alloc_rings(vsi);
2447 		if (ret)
2448 			goto unroll_alloc_q_vector;
2449 
2450 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
2451 		if (ret)
2452 			goto unroll_vector_base;
2453 
2454 		/* Do not exit if configuring RSS had an issue, at least
2455 		 * receive traffic on first queue. Hence no need to capture
2456 		 * return value
2457 		 */
2458 		if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2459 			ice_vsi_cfg_rss_lut_key(vsi);
2460 			ice_vsi_set_vf_rss_flow_fld(vsi);
2461 		}
2462 		break;
2463 	case ICE_VSI_LB:
2464 		ret = ice_vsi_alloc_rings(vsi);
2465 		if (ret)
2466 			goto unroll_vsi_init;
2467 		break;
2468 	default:
2469 		/* clean up the resources and exit */
2470 		goto unroll_vsi_init;
2471 	}
2472 
2473 	/* configure VSI nodes based on number of queues and TC's */
2474 	for (i = 0; i < vsi->tc_cfg.numtc; i++)
2475 		max_txqs[i] = vsi->alloc_txq;
2476 
2477 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2478 				 max_txqs);
2479 	if (status) {
2480 		dev_err(dev, "VSI %d failed lan queue config, error %s\n",
2481 			vsi->vsi_num, ice_stat_str(status));
2482 		goto unroll_clear_rings;
2483 	}
2484 
2485 	/* Add switch rule to drop all Tx Flow Control Frames, of look up
2486 	 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2487 	 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2488 	 * The rule is added once for PF VSI in order to create appropriate
2489 	 * recipe, since VSI/VSI list is ignored with drop action...
2490 	 * Also add rules to handle LLDP Tx packets.  Tx LLDP packets need to
2491 	 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2492 	 * settings in the HW.
2493 	 */
2494 	if (!ice_is_safe_mode(pf))
2495 		if (vsi->type == ICE_VSI_PF) {
2496 			ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2497 					 ICE_DROP_PACKET);
2498 			ice_cfg_sw_lldp(vsi, true, true);
2499 		}
2500 
2501 	if (!vsi->agg_node)
2502 		ice_set_agg_vsi(vsi);
2503 	return vsi;
2504 
2505 unroll_clear_rings:
2506 	ice_vsi_clear_rings(vsi);
2507 unroll_vector_base:
2508 	/* reclaim SW interrupts back to the common pool */
2509 	ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
2510 	pf->num_avail_sw_msix += vsi->num_q_vectors;
2511 unroll_alloc_q_vector:
2512 	ice_vsi_free_q_vectors(vsi);
2513 unroll_vsi_init:
2514 	ice_vsi_delete(vsi);
2515 unroll_get_qs:
2516 	ice_vsi_put_qs(vsi);
2517 unroll_vsi_alloc:
2518 	if (vsi_type == ICE_VSI_VF)
2519 		ice_enable_lag(pf->lag);
2520 	ice_vsi_clear(vsi);
2521 
2522 	return NULL;
2523 }
2524 
2525 /**
2526  * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2527  * @vsi: the VSI being cleaned up
2528  */
2529 static void ice_vsi_release_msix(struct ice_vsi *vsi)
2530 {
2531 	struct ice_pf *pf = vsi->back;
2532 	struct ice_hw *hw = &pf->hw;
2533 	u32 txq = 0;
2534 	u32 rxq = 0;
2535 	int i, q;
2536 
2537 	for (i = 0; i < vsi->num_q_vectors; i++) {
2538 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2539 
2540 		ice_write_intrl(q_vector, 0);
2541 		for (q = 0; q < q_vector->num_ring_tx; q++) {
2542 			ice_write_itr(&q_vector->tx, 0);
2543 			wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2544 			if (ice_is_xdp_ena_vsi(vsi)) {
2545 				u32 xdp_txq = txq + vsi->num_xdp_txq;
2546 
2547 				wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2548 			}
2549 			txq++;
2550 		}
2551 
2552 		for (q = 0; q < q_vector->num_ring_rx; q++) {
2553 			ice_write_itr(&q_vector->rx, 0);
2554 			wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2555 			rxq++;
2556 		}
2557 	}
2558 
2559 	ice_flush(hw);
2560 }
2561 
2562 /**
2563  * ice_vsi_free_irq - Free the IRQ association with the OS
2564  * @vsi: the VSI being configured
2565  */
2566 void ice_vsi_free_irq(struct ice_vsi *vsi)
2567 {
2568 	struct ice_pf *pf = vsi->back;
2569 	int base = vsi->base_vector;
2570 	int i;
2571 
2572 	if (!vsi->q_vectors || !vsi->irqs_ready)
2573 		return;
2574 
2575 	ice_vsi_release_msix(vsi);
2576 	if (vsi->type == ICE_VSI_VF)
2577 		return;
2578 
2579 	vsi->irqs_ready = false;
2580 	ice_for_each_q_vector(vsi, i) {
2581 		u16 vector = i + base;
2582 		int irq_num;
2583 
2584 		irq_num = pf->msix_entries[vector].vector;
2585 
2586 		/* free only the irqs that were actually requested */
2587 		if (!vsi->q_vectors[i] ||
2588 		    !(vsi->q_vectors[i]->num_ring_tx ||
2589 		      vsi->q_vectors[i]->num_ring_rx))
2590 			continue;
2591 
2592 		/* clear the affinity notifier in the IRQ descriptor */
2593 		irq_set_affinity_notifier(irq_num, NULL);
2594 
2595 		/* clear the affinity_mask in the IRQ descriptor */
2596 		irq_set_affinity_hint(irq_num, NULL);
2597 		synchronize_irq(irq_num);
2598 		devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2599 	}
2600 }
2601 
2602 /**
2603  * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2604  * @vsi: the VSI having resources freed
2605  */
2606 void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2607 {
2608 	int i;
2609 
2610 	if (!vsi->tx_rings)
2611 		return;
2612 
2613 	ice_for_each_txq(vsi, i)
2614 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2615 			ice_free_tx_ring(vsi->tx_rings[i]);
2616 }
2617 
2618 /**
2619  * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2620  * @vsi: the VSI having resources freed
2621  */
2622 void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2623 {
2624 	int i;
2625 
2626 	if (!vsi->rx_rings)
2627 		return;
2628 
2629 	ice_for_each_rxq(vsi, i)
2630 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2631 			ice_free_rx_ring(vsi->rx_rings[i]);
2632 }
2633 
2634 /**
2635  * ice_vsi_close - Shut down a VSI
2636  * @vsi: the VSI being shut down
2637  */
2638 void ice_vsi_close(struct ice_vsi *vsi)
2639 {
2640 	if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2641 		ice_down(vsi);
2642 
2643 	ice_vsi_free_irq(vsi);
2644 	ice_vsi_free_tx_rings(vsi);
2645 	ice_vsi_free_rx_rings(vsi);
2646 }
2647 
2648 /**
2649  * ice_ena_vsi - resume a VSI
2650  * @vsi: the VSI being resume
2651  * @locked: is the rtnl_lock already held
2652  */
2653 int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2654 {
2655 	int err = 0;
2656 
2657 	if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2658 		return 0;
2659 
2660 	clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2661 
2662 	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2663 		if (netif_running(vsi->netdev)) {
2664 			if (!locked)
2665 				rtnl_lock();
2666 
2667 			err = ice_open_internal(vsi->netdev);
2668 
2669 			if (!locked)
2670 				rtnl_unlock();
2671 		}
2672 	} else if (vsi->type == ICE_VSI_CTRL) {
2673 		err = ice_vsi_open_ctrl(vsi);
2674 	}
2675 
2676 	return err;
2677 }
2678 
2679 /**
2680  * ice_dis_vsi - pause a VSI
2681  * @vsi: the VSI being paused
2682  * @locked: is the rtnl_lock already held
2683  */
2684 void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2685 {
2686 	if (test_bit(ICE_VSI_DOWN, vsi->state))
2687 		return;
2688 
2689 	set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2690 
2691 	if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2692 		if (netif_running(vsi->netdev)) {
2693 			if (!locked)
2694 				rtnl_lock();
2695 
2696 			ice_vsi_close(vsi);
2697 
2698 			if (!locked)
2699 				rtnl_unlock();
2700 		} else {
2701 			ice_vsi_close(vsi);
2702 		}
2703 	} else if (vsi->type == ICE_VSI_CTRL) {
2704 		ice_vsi_close(vsi);
2705 	}
2706 }
2707 
2708 /**
2709  * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
2710  * @vsi: the VSI being un-configured
2711  */
2712 void ice_vsi_dis_irq(struct ice_vsi *vsi)
2713 {
2714 	int base = vsi->base_vector;
2715 	struct ice_pf *pf = vsi->back;
2716 	struct ice_hw *hw = &pf->hw;
2717 	u32 val;
2718 	int i;
2719 
2720 	/* disable interrupt causation from each queue */
2721 	if (vsi->tx_rings) {
2722 		ice_for_each_txq(vsi, i) {
2723 			if (vsi->tx_rings[i]) {
2724 				u16 reg;
2725 
2726 				reg = vsi->tx_rings[i]->reg_idx;
2727 				val = rd32(hw, QINT_TQCTL(reg));
2728 				val &= ~QINT_TQCTL_CAUSE_ENA_M;
2729 				wr32(hw, QINT_TQCTL(reg), val);
2730 			}
2731 		}
2732 	}
2733 
2734 	if (vsi->rx_rings) {
2735 		ice_for_each_rxq(vsi, i) {
2736 			if (vsi->rx_rings[i]) {
2737 				u16 reg;
2738 
2739 				reg = vsi->rx_rings[i]->reg_idx;
2740 				val = rd32(hw, QINT_RQCTL(reg));
2741 				val &= ~QINT_RQCTL_CAUSE_ENA_M;
2742 				wr32(hw, QINT_RQCTL(reg), val);
2743 			}
2744 		}
2745 	}
2746 
2747 	/* disable each interrupt */
2748 	ice_for_each_q_vector(vsi, i) {
2749 		if (!vsi->q_vectors[i])
2750 			continue;
2751 		wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
2752 	}
2753 
2754 	ice_flush(hw);
2755 
2756 	/* don't call synchronize_irq() for VF's from the host */
2757 	if (vsi->type == ICE_VSI_VF)
2758 		return;
2759 
2760 	ice_for_each_q_vector(vsi, i)
2761 		synchronize_irq(pf->msix_entries[i + base].vector);
2762 }
2763 
2764 /**
2765  * ice_napi_del - Remove NAPI handler for the VSI
2766  * @vsi: VSI for which NAPI handler is to be removed
2767  */
2768 void ice_napi_del(struct ice_vsi *vsi)
2769 {
2770 	int v_idx;
2771 
2772 	if (!vsi->netdev)
2773 		return;
2774 
2775 	ice_for_each_q_vector(vsi, v_idx)
2776 		netif_napi_del(&vsi->q_vectors[v_idx]->napi);
2777 }
2778 
2779 /**
2780  * ice_vsi_release - Delete a VSI and free its resources
2781  * @vsi: the VSI being removed
2782  *
2783  * Returns 0 on success or < 0 on error
2784  */
2785 int ice_vsi_release(struct ice_vsi *vsi)
2786 {
2787 	struct ice_pf *pf;
2788 
2789 	if (!vsi->back)
2790 		return -ENODEV;
2791 	pf = vsi->back;
2792 
2793 	/* do not unregister while driver is in the reset recovery pending
2794 	 * state. Since reset/rebuild happens through PF service task workqueue,
2795 	 * it's not a good idea to unregister netdev that is associated to the
2796 	 * PF that is running the work queue items currently. This is done to
2797 	 * avoid check_flush_dependency() warning on this wq
2798 	 */
2799 	if (vsi->netdev && !ice_is_reset_in_progress(pf->state) &&
2800 	    (test_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state))) {
2801 		unregister_netdev(vsi->netdev);
2802 		clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
2803 	}
2804 
2805 	ice_devlink_destroy_port(vsi);
2806 
2807 	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2808 		ice_rss_clean(vsi);
2809 
2810 	/* Disable VSI and free resources */
2811 	if (vsi->type != ICE_VSI_LB)
2812 		ice_vsi_dis_irq(vsi);
2813 	ice_vsi_close(vsi);
2814 
2815 	/* SR-IOV determines needed MSIX resources all at once instead of per
2816 	 * VSI since when VFs are spawned we know how many VFs there are and how
2817 	 * many interrupts each VF needs. SR-IOV MSIX resources are also
2818 	 * cleared in the same manner.
2819 	 */
2820 	if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID) {
2821 		struct ice_vf *vf;
2822 		int i;
2823 
2824 		ice_for_each_vf(pf, i) {
2825 			vf = &pf->vf[i];
2826 			if (i != vsi->vf_id && vf->ctrl_vsi_idx != ICE_NO_VSI)
2827 				break;
2828 		}
2829 		if (i == pf->num_alloc_vfs) {
2830 			/* No other VFs left that have control VSI, reclaim SW
2831 			 * interrupts back to the common pool
2832 			 */
2833 			ice_free_res(pf->irq_tracker, vsi->base_vector,
2834 				     ICE_RES_VF_CTRL_VEC_ID);
2835 			pf->num_avail_sw_msix += vsi->num_q_vectors;
2836 		}
2837 	} else if (vsi->type != ICE_VSI_VF) {
2838 		/* reclaim SW interrupts back to the common pool */
2839 		ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
2840 		pf->num_avail_sw_msix += vsi->num_q_vectors;
2841 	}
2842 
2843 	if (!ice_is_safe_mode(pf)) {
2844 		if (vsi->type == ICE_VSI_PF) {
2845 			ice_fltr_remove_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2846 					    ICE_DROP_PACKET);
2847 			ice_cfg_sw_lldp(vsi, true, false);
2848 			/* The Rx rule will only exist to remove if the LLDP FW
2849 			 * engine is currently stopped
2850 			 */
2851 			if (!test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2852 				ice_cfg_sw_lldp(vsi, false, false);
2853 		}
2854 	}
2855 
2856 	ice_fltr_remove_all(vsi);
2857 	ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
2858 	ice_vsi_delete(vsi);
2859 	ice_vsi_free_q_vectors(vsi);
2860 
2861 	if (vsi->netdev) {
2862 		if (test_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state)) {
2863 			unregister_netdev(vsi->netdev);
2864 			clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
2865 		}
2866 		if (test_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state)) {
2867 			free_netdev(vsi->netdev);
2868 			vsi->netdev = NULL;
2869 			clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
2870 		}
2871 	}
2872 
2873 	if (vsi->type == ICE_VSI_VF &&
2874 	    vsi->agg_node && vsi->agg_node->valid)
2875 		vsi->agg_node->num_vsis--;
2876 	ice_vsi_clear_rings(vsi);
2877 
2878 	ice_vsi_put_qs(vsi);
2879 
2880 	/* retain SW VSI data structure since it is needed to unregister and
2881 	 * free VSI netdev when PF is not in reset recovery pending state,\
2882 	 * for ex: during rmmod.
2883 	 */
2884 	if (!ice_is_reset_in_progress(pf->state))
2885 		ice_vsi_clear(vsi);
2886 
2887 	return 0;
2888 }
2889 
2890 /**
2891  * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
2892  * @vsi: VSI connected with q_vectors
2893  * @coalesce: array of struct with stored coalesce
2894  *
2895  * Returns array size.
2896  */
2897 static int
2898 ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
2899 			     struct ice_coalesce_stored *coalesce)
2900 {
2901 	int i;
2902 
2903 	ice_for_each_q_vector(vsi, i) {
2904 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2905 
2906 		coalesce[i].itr_tx = q_vector->tx.itr_setting;
2907 		coalesce[i].itr_rx = q_vector->rx.itr_setting;
2908 		coalesce[i].intrl = q_vector->intrl;
2909 
2910 		if (i < vsi->num_txq)
2911 			coalesce[i].tx_valid = true;
2912 		if (i < vsi->num_rxq)
2913 			coalesce[i].rx_valid = true;
2914 	}
2915 
2916 	return vsi->num_q_vectors;
2917 }
2918 
2919 /**
2920  * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
2921  * @vsi: VSI connected with q_vectors
2922  * @coalesce: pointer to array of struct with stored coalesce
2923  * @size: size of coalesce array
2924  *
2925  * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
2926  * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
2927  * to default value.
2928  */
2929 static void
2930 ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
2931 			     struct ice_coalesce_stored *coalesce, int size)
2932 {
2933 	struct ice_ring_container *rc;
2934 	int i;
2935 
2936 	if ((size && !coalesce) || !vsi)
2937 		return;
2938 
2939 	/* There are a couple of cases that have to be handled here:
2940 	 *   1. The case where the number of queue vectors stays the same, but
2941 	 *      the number of Tx or Rx rings changes (the first for loop)
2942 	 *   2. The case where the number of queue vectors increased (the
2943 	 *      second for loop)
2944 	 */
2945 	for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
2946 		/* There are 2 cases to handle here and they are the same for
2947 		 * both Tx and Rx:
2948 		 *   if the entry was valid previously (coalesce[i].[tr]x_valid
2949 		 *   and the loop variable is less than the number of rings
2950 		 *   allocated, then write the previous values
2951 		 *
2952 		 *   if the entry was not valid previously, but the number of
2953 		 *   rings is less than are allocated (this means the number of
2954 		 *   rings increased from previously), then write out the
2955 		 *   values in the first element
2956 		 *
2957 		 *   Also, always write the ITR, even if in ITR_IS_DYNAMIC
2958 		 *   as there is no harm because the dynamic algorithm
2959 		 *   will just overwrite.
2960 		 */
2961 		if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
2962 			rc = &vsi->q_vectors[i]->rx;
2963 			rc->itr_setting = coalesce[i].itr_rx;
2964 			ice_write_itr(rc, rc->itr_setting);
2965 		} else if (i < vsi->alloc_rxq) {
2966 			rc = &vsi->q_vectors[i]->rx;
2967 			rc->itr_setting = coalesce[0].itr_rx;
2968 			ice_write_itr(rc, rc->itr_setting);
2969 		}
2970 
2971 		if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
2972 			rc = &vsi->q_vectors[i]->tx;
2973 			rc->itr_setting = coalesce[i].itr_tx;
2974 			ice_write_itr(rc, rc->itr_setting);
2975 		} else if (i < vsi->alloc_txq) {
2976 			rc = &vsi->q_vectors[i]->tx;
2977 			rc->itr_setting = coalesce[0].itr_tx;
2978 			ice_write_itr(rc, rc->itr_setting);
2979 		}
2980 
2981 		vsi->q_vectors[i]->intrl = coalesce[i].intrl;
2982 		ice_write_intrl(vsi->q_vectors[i], coalesce[i].intrl);
2983 	}
2984 
2985 	/* the number of queue vectors increased so write whatever is in
2986 	 * the first element
2987 	 */
2988 	for (; i < vsi->num_q_vectors; i++) {
2989 		/* transmit */
2990 		rc = &vsi->q_vectors[i]->tx;
2991 		rc->itr_setting = coalesce[0].itr_tx;
2992 		ice_write_itr(rc, rc->itr_setting);
2993 
2994 		/* receive */
2995 		rc = &vsi->q_vectors[i]->rx;
2996 		rc->itr_setting = coalesce[0].itr_rx;
2997 		ice_write_itr(rc, rc->itr_setting);
2998 
2999 		vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3000 		ice_write_intrl(vsi->q_vectors[i], coalesce[0].intrl);
3001 	}
3002 }
3003 
3004 /**
3005  * ice_vsi_rebuild - Rebuild VSI after reset
3006  * @vsi: VSI to be rebuild
3007  * @init_vsi: is this an initialization or a reconfigure of the VSI
3008  *
3009  * Returns 0 on success and negative value on failure
3010  */
3011 int ice_vsi_rebuild(struct ice_vsi *vsi, bool init_vsi)
3012 {
3013 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3014 	struct ice_coalesce_stored *coalesce;
3015 	int prev_num_q_vectors = 0;
3016 	struct ice_vf *vf = NULL;
3017 	enum ice_vsi_type vtype;
3018 	enum ice_status status;
3019 	struct ice_pf *pf;
3020 	int ret, i;
3021 
3022 	if (!vsi)
3023 		return -EINVAL;
3024 
3025 	pf = vsi->back;
3026 	vtype = vsi->type;
3027 	if (vtype == ICE_VSI_VF)
3028 		vf = &pf->vf[vsi->vf_id];
3029 
3030 	coalesce = kcalloc(vsi->num_q_vectors,
3031 			   sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3032 	if (!coalesce)
3033 		return -ENOMEM;
3034 
3035 	prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3036 
3037 	ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
3038 	ice_vsi_free_q_vectors(vsi);
3039 
3040 	/* SR-IOV determines needed MSIX resources all at once instead of per
3041 	 * VSI since when VFs are spawned we know how many VFs there are and how
3042 	 * many interrupts each VF needs. SR-IOV MSIX resources are also
3043 	 * cleared in the same manner.
3044 	 */
3045 	if (vtype != ICE_VSI_VF) {
3046 		/* reclaim SW interrupts back to the common pool */
3047 		ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
3048 		pf->num_avail_sw_msix += vsi->num_q_vectors;
3049 		vsi->base_vector = 0;
3050 	}
3051 
3052 	if (ice_is_xdp_ena_vsi(vsi))
3053 		/* return value check can be skipped here, it always returns
3054 		 * 0 if reset is in progress
3055 		 */
3056 		ice_destroy_xdp_rings(vsi);
3057 	ice_vsi_put_qs(vsi);
3058 	ice_vsi_clear_rings(vsi);
3059 	ice_vsi_free_arrays(vsi);
3060 	if (vtype == ICE_VSI_VF)
3061 		ice_vsi_set_num_qs(vsi, vf->vf_id);
3062 	else
3063 		ice_vsi_set_num_qs(vsi, ICE_INVAL_VFID);
3064 
3065 	ret = ice_vsi_alloc_arrays(vsi);
3066 	if (ret < 0)
3067 		goto err_vsi;
3068 
3069 	ice_vsi_get_qs(vsi);
3070 
3071 	ice_alloc_fd_res(vsi);
3072 	ice_vsi_set_tc_cfg(vsi);
3073 
3074 	/* Initialize VSI struct elements and create VSI in FW */
3075 	ret = ice_vsi_init(vsi, init_vsi);
3076 	if (ret < 0)
3077 		goto err_vsi;
3078 
3079 	switch (vtype) {
3080 	case ICE_VSI_CTRL:
3081 	case ICE_VSI_PF:
3082 		ret = ice_vsi_alloc_q_vectors(vsi);
3083 		if (ret)
3084 			goto err_rings;
3085 
3086 		ret = ice_vsi_setup_vector_base(vsi);
3087 		if (ret)
3088 			goto err_vectors;
3089 
3090 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
3091 		if (ret)
3092 			goto err_vectors;
3093 
3094 		ret = ice_vsi_alloc_rings(vsi);
3095 		if (ret)
3096 			goto err_vectors;
3097 
3098 		ice_vsi_map_rings_to_vectors(vsi);
3099 		if (ice_is_xdp_ena_vsi(vsi)) {
3100 			vsi->num_xdp_txq = vsi->alloc_rxq;
3101 			ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog);
3102 			if (ret)
3103 				goto err_vectors;
3104 		}
3105 		/* ICE_VSI_CTRL does not need RSS so skip RSS processing */
3106 		if (vtype != ICE_VSI_CTRL)
3107 			/* Do not exit if configuring RSS had an issue, at
3108 			 * least receive traffic on first queue. Hence no
3109 			 * need to capture return value
3110 			 */
3111 			if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
3112 				ice_vsi_cfg_rss_lut_key(vsi);
3113 		break;
3114 	case ICE_VSI_VF:
3115 		ret = ice_vsi_alloc_q_vectors(vsi);
3116 		if (ret)
3117 			goto err_rings;
3118 
3119 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
3120 		if (ret)
3121 			goto err_vectors;
3122 
3123 		ret = ice_vsi_alloc_rings(vsi);
3124 		if (ret)
3125 			goto err_vectors;
3126 
3127 		break;
3128 	default:
3129 		break;
3130 	}
3131 
3132 	/* configure VSI nodes based on number of queues and TC's */
3133 	for (i = 0; i < vsi->tc_cfg.numtc; i++) {
3134 		max_txqs[i] = vsi->alloc_txq;
3135 
3136 		if (ice_is_xdp_ena_vsi(vsi))
3137 			max_txqs[i] += vsi->num_xdp_txq;
3138 	}
3139 
3140 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
3141 				 max_txqs);
3142 	if (status) {
3143 		dev_err(ice_pf_to_dev(pf), "VSI %d failed lan queue config, error %s\n",
3144 			vsi->vsi_num, ice_stat_str(status));
3145 		if (init_vsi) {
3146 			ret = -EIO;
3147 			goto err_vectors;
3148 		} else {
3149 			return ice_schedule_reset(pf, ICE_RESET_PFR);
3150 		}
3151 	}
3152 	ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3153 	kfree(coalesce);
3154 
3155 	return 0;
3156 
3157 err_vectors:
3158 	ice_vsi_free_q_vectors(vsi);
3159 err_rings:
3160 	if (vsi->netdev) {
3161 		vsi->current_netdev_flags = 0;
3162 		unregister_netdev(vsi->netdev);
3163 		free_netdev(vsi->netdev);
3164 		vsi->netdev = NULL;
3165 	}
3166 err_vsi:
3167 	ice_vsi_clear(vsi);
3168 	set_bit(ICE_RESET_FAILED, pf->state);
3169 	kfree(coalesce);
3170 	return ret;
3171 }
3172 
3173 /**
3174  * ice_is_reset_in_progress - check for a reset in progress
3175  * @state: PF state field
3176  */
3177 bool ice_is_reset_in_progress(unsigned long *state)
3178 {
3179 	return test_bit(ICE_RESET_OICR_RECV, state) ||
3180 	       test_bit(ICE_PFR_REQ, state) ||
3181 	       test_bit(ICE_CORER_REQ, state) ||
3182 	       test_bit(ICE_GLOBR_REQ, state);
3183 }
3184 
3185 #ifdef CONFIG_DCB
3186 /**
3187  * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3188  * @vsi: VSI being configured
3189  * @ctx: the context buffer returned from AQ VSI update command
3190  */
3191 static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3192 {
3193 	vsi->info.mapping_flags = ctx->info.mapping_flags;
3194 	memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3195 	       sizeof(vsi->info.q_mapping));
3196 	memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3197 	       sizeof(vsi->info.tc_mapping));
3198 }
3199 
3200 /**
3201  * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3202  * @vsi: VSI to be configured
3203  * @ena_tc: TC bitmap
3204  *
3205  * VSI queues expected to be quiesced before calling this function
3206  */
3207 int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3208 {
3209 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3210 	struct ice_pf *pf = vsi->back;
3211 	struct ice_vsi_ctx *ctx;
3212 	enum ice_status status;
3213 	struct device *dev;
3214 	int i, ret = 0;
3215 	u8 num_tc = 0;
3216 
3217 	dev = ice_pf_to_dev(pf);
3218 
3219 	ice_for_each_traffic_class(i) {
3220 		/* build bitmap of enabled TCs */
3221 		if (ena_tc & BIT(i))
3222 			num_tc++;
3223 		/* populate max_txqs per TC */
3224 		max_txqs[i] = vsi->alloc_txq;
3225 	}
3226 
3227 	vsi->tc_cfg.ena_tc = ena_tc;
3228 	vsi->tc_cfg.numtc = num_tc;
3229 
3230 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3231 	if (!ctx)
3232 		return -ENOMEM;
3233 
3234 	ctx->vf_num = 0;
3235 	ctx->info = vsi->info;
3236 
3237 	ice_vsi_setup_q_map(vsi, ctx);
3238 
3239 	/* must to indicate which section of VSI context are being modified */
3240 	ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3241 	status = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3242 	if (status) {
3243 		dev_info(dev, "Failed VSI Update\n");
3244 		ret = -EIO;
3245 		goto out;
3246 	}
3247 
3248 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
3249 				 max_txqs);
3250 
3251 	if (status) {
3252 		dev_err(dev, "VSI %d failed TC config, error %s\n",
3253 			vsi->vsi_num, ice_stat_str(status));
3254 		ret = -EIO;
3255 		goto out;
3256 	}
3257 	ice_vsi_update_q_map(vsi, ctx);
3258 	vsi->info.valid_sections = 0;
3259 
3260 	ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3261 out:
3262 	kfree(ctx);
3263 	return ret;
3264 }
3265 #endif /* CONFIG_DCB */
3266 
3267 /**
3268  * ice_update_ring_stats - Update ring statistics
3269  * @ring: ring to update
3270  * @pkts: number of processed packets
3271  * @bytes: number of processed bytes
3272  *
3273  * This function assumes that caller has acquired a u64_stats_sync lock.
3274  */
3275 static void ice_update_ring_stats(struct ice_ring *ring, u64 pkts, u64 bytes)
3276 {
3277 	ring->stats.bytes += bytes;
3278 	ring->stats.pkts += pkts;
3279 }
3280 
3281 /**
3282  * ice_update_tx_ring_stats - Update Tx ring specific counters
3283  * @tx_ring: ring to update
3284  * @pkts: number of processed packets
3285  * @bytes: number of processed bytes
3286  */
3287 void ice_update_tx_ring_stats(struct ice_ring *tx_ring, u64 pkts, u64 bytes)
3288 {
3289 	u64_stats_update_begin(&tx_ring->syncp);
3290 	ice_update_ring_stats(tx_ring, pkts, bytes);
3291 	u64_stats_update_end(&tx_ring->syncp);
3292 }
3293 
3294 /**
3295  * ice_update_rx_ring_stats - Update Rx ring specific counters
3296  * @rx_ring: ring to update
3297  * @pkts: number of processed packets
3298  * @bytes: number of processed bytes
3299  */
3300 void ice_update_rx_ring_stats(struct ice_ring *rx_ring, u64 pkts, u64 bytes)
3301 {
3302 	u64_stats_update_begin(&rx_ring->syncp);
3303 	ice_update_ring_stats(rx_ring, pkts, bytes);
3304 	u64_stats_update_end(&rx_ring->syncp);
3305 }
3306 
3307 /**
3308  * ice_status_to_errno - convert from enum ice_status to Linux errno
3309  * @err: ice_status value to convert
3310  */
3311 int ice_status_to_errno(enum ice_status err)
3312 {
3313 	switch (err) {
3314 	case ICE_SUCCESS:
3315 		return 0;
3316 	case ICE_ERR_DOES_NOT_EXIST:
3317 		return -ENOENT;
3318 	case ICE_ERR_OUT_OF_RANGE:
3319 		return -ENOTTY;
3320 	case ICE_ERR_PARAM:
3321 		return -EINVAL;
3322 	case ICE_ERR_NO_MEMORY:
3323 		return -ENOMEM;
3324 	case ICE_ERR_MAX_LIMIT:
3325 		return -EAGAIN;
3326 	default:
3327 		return -EINVAL;
3328 	}
3329 }
3330 
3331 /**
3332  * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3333  * @sw: switch to check if its default forwarding VSI is free
3334  *
3335  * Return true if the default forwarding VSI is already being used, else returns
3336  * false signalling that it's available to use.
3337  */
3338 bool ice_is_dflt_vsi_in_use(struct ice_sw *sw)
3339 {
3340 	return (sw->dflt_vsi && sw->dflt_vsi_ena);
3341 }
3342 
3343 /**
3344  * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3345  * @sw: switch for the default forwarding VSI to compare against
3346  * @vsi: VSI to compare against default forwarding VSI
3347  *
3348  * If this VSI passed in is the default forwarding VSI then return true, else
3349  * return false
3350  */
3351 bool ice_is_vsi_dflt_vsi(struct ice_sw *sw, struct ice_vsi *vsi)
3352 {
3353 	return (sw->dflt_vsi == vsi && sw->dflt_vsi_ena);
3354 }
3355 
3356 /**
3357  * ice_set_dflt_vsi - set the default forwarding VSI
3358  * @sw: switch used to assign the default forwarding VSI
3359  * @vsi: VSI getting set as the default forwarding VSI on the switch
3360  *
3361  * If the VSI passed in is already the default VSI and it's enabled just return
3362  * success.
3363  *
3364  * If there is already a default VSI on the switch and it's enabled then return
3365  * -EEXIST since there can only be one default VSI per switch.
3366  *
3367  *  Otherwise try to set the VSI passed in as the switch's default VSI and
3368  *  return the result.
3369  */
3370 int ice_set_dflt_vsi(struct ice_sw *sw, struct ice_vsi *vsi)
3371 {
3372 	enum ice_status status;
3373 	struct device *dev;
3374 
3375 	if (!sw || !vsi)
3376 		return -EINVAL;
3377 
3378 	dev = ice_pf_to_dev(vsi->back);
3379 
3380 	/* the VSI passed in is already the default VSI */
3381 	if (ice_is_vsi_dflt_vsi(sw, vsi)) {
3382 		dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3383 			vsi->vsi_num);
3384 		return 0;
3385 	}
3386 
3387 	/* another VSI is already the default VSI for this switch */
3388 	if (ice_is_dflt_vsi_in_use(sw)) {
3389 		dev_err(dev, "Default forwarding VSI %d already in use, disable it and try again\n",
3390 			sw->dflt_vsi->vsi_num);
3391 		return -EEXIST;
3392 	}
3393 
3394 	status = ice_cfg_dflt_vsi(&vsi->back->hw, vsi->idx, true, ICE_FLTR_RX);
3395 	if (status) {
3396 		dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %s\n",
3397 			vsi->vsi_num, ice_stat_str(status));
3398 		return -EIO;
3399 	}
3400 
3401 	sw->dflt_vsi = vsi;
3402 	sw->dflt_vsi_ena = true;
3403 
3404 	return 0;
3405 }
3406 
3407 /**
3408  * ice_clear_dflt_vsi - clear the default forwarding VSI
3409  * @sw: switch used to clear the default VSI
3410  *
3411  * If the switch has no default VSI or it's not enabled then return error.
3412  *
3413  * Otherwise try to clear the default VSI and return the result.
3414  */
3415 int ice_clear_dflt_vsi(struct ice_sw *sw)
3416 {
3417 	struct ice_vsi *dflt_vsi;
3418 	enum ice_status status;
3419 	struct device *dev;
3420 
3421 	if (!sw)
3422 		return -EINVAL;
3423 
3424 	dev = ice_pf_to_dev(sw->pf);
3425 
3426 	dflt_vsi = sw->dflt_vsi;
3427 
3428 	/* there is no default VSI configured */
3429 	if (!ice_is_dflt_vsi_in_use(sw))
3430 		return -ENODEV;
3431 
3432 	status = ice_cfg_dflt_vsi(&dflt_vsi->back->hw, dflt_vsi->idx, false,
3433 				  ICE_FLTR_RX);
3434 	if (status) {
3435 		dev_err(dev, "Failed to clear the default forwarding VSI %d, error %s\n",
3436 			dflt_vsi->vsi_num, ice_stat_str(status));
3437 		return -EIO;
3438 	}
3439 
3440 	sw->dflt_vsi = NULL;
3441 	sw->dflt_vsi_ena = false;
3442 
3443 	return 0;
3444 }
3445 
3446 /**
3447  * ice_set_link - turn on/off physical link
3448  * @vsi: VSI to modify physical link on
3449  * @ena: turn on/off physical link
3450  */
3451 int ice_set_link(struct ice_vsi *vsi, bool ena)
3452 {
3453 	struct device *dev = ice_pf_to_dev(vsi->back);
3454 	struct ice_port_info *pi = vsi->port_info;
3455 	struct ice_hw *hw = pi->hw;
3456 	enum ice_status status;
3457 
3458 	if (vsi->type != ICE_VSI_PF)
3459 		return -EINVAL;
3460 
3461 	status = ice_aq_set_link_restart_an(pi, ena, NULL);
3462 
3463 	/* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3464 	 * this is not a fatal error, so print a warning message and return
3465 	 * a success code. Return an error if FW returns an error code other
3466 	 * than ICE_AQ_RC_EMODE
3467 	 */
3468 	if (status == ICE_ERR_AQ_ERROR) {
3469 		if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3470 			dev_warn(dev, "can't set link to %s, err %s aq_err %s. not fatal, continuing\n",
3471 				 (ena ? "ON" : "OFF"), ice_stat_str(status),
3472 				 ice_aq_str(hw->adminq.sq_last_status));
3473 	} else if (status) {
3474 		dev_err(dev, "can't set link to %s, err %s aq_err %s\n",
3475 			(ena ? "ON" : "OFF"), ice_stat_str(status),
3476 			ice_aq_str(hw->adminq.sq_last_status));
3477 		return -EIO;
3478 	}
3479 
3480 	return 0;
3481 }
3482