xref: /linux/drivers/net/ethernet/intel/ice/ice_lib.c (revision 511c537bb5647662ff7df7a41180a1721c078720)
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 	vsi->af_xdp_zc_qps = bitmap_zalloc(max_t(int, vsi->alloc_txq, vsi->alloc_rxq), GFP_KERNEL);
109 	if (!vsi->af_xdp_zc_qps)
110 		goto err_zc_qps;
111 
112 	return 0;
113 
114 err_zc_qps:
115 	devm_kfree(dev, vsi->q_vectors);
116 err_vectors:
117 	devm_kfree(dev, vsi->rxq_map);
118 err_rxq_map:
119 	devm_kfree(dev, vsi->txq_map);
120 err_txq_map:
121 	devm_kfree(dev, vsi->rx_rings);
122 err_rings:
123 	devm_kfree(dev, vsi->tx_rings);
124 	return -ENOMEM;
125 }
126 
127 /**
128  * ice_vsi_set_num_desc - Set number of descriptors for queues on this VSI
129  * @vsi: the VSI being configured
130  */
131 static void ice_vsi_set_num_desc(struct ice_vsi *vsi)
132 {
133 	switch (vsi->type) {
134 	case ICE_VSI_PF:
135 	case ICE_VSI_CTRL:
136 	case ICE_VSI_LB:
137 		/* a user could change the values of num_[tr]x_desc using
138 		 * ethtool -G so we should keep those values instead of
139 		 * overwriting them with the defaults.
140 		 */
141 		if (!vsi->num_rx_desc)
142 			vsi->num_rx_desc = ICE_DFLT_NUM_RX_DESC;
143 		if (!vsi->num_tx_desc)
144 			vsi->num_tx_desc = ICE_DFLT_NUM_TX_DESC;
145 		break;
146 	default:
147 		dev_dbg(ice_pf_to_dev(vsi->back), "Not setting number of Tx/Rx descriptors for VSI type %d\n",
148 			vsi->type);
149 		break;
150 	}
151 }
152 
153 /**
154  * ice_vsi_set_num_qs - Set number of queues, descriptors and vectors for a VSI
155  * @vsi: the VSI being configured
156  * @vf_id: ID of the VF being configured
157  *
158  * Return 0 on success and a negative value on error
159  */
160 static void ice_vsi_set_num_qs(struct ice_vsi *vsi, u16 vf_id)
161 {
162 	struct ice_pf *pf = vsi->back;
163 	struct ice_vf *vf = NULL;
164 
165 	if (vsi->type == ICE_VSI_VF)
166 		vsi->vf_id = vf_id;
167 	else
168 		vsi->vf_id = ICE_INVAL_VFID;
169 
170 	switch (vsi->type) {
171 	case ICE_VSI_PF:
172 		if (vsi->req_txq) {
173 			vsi->alloc_txq = vsi->req_txq;
174 			vsi->num_txq = vsi->req_txq;
175 		} else {
176 			vsi->alloc_txq = min3(pf->num_lan_msix,
177 					      ice_get_avail_txq_count(pf),
178 					      (u16)num_online_cpus());
179 		}
180 
181 		pf->num_lan_tx = vsi->alloc_txq;
182 
183 		/* only 1 Rx queue unless RSS is enabled */
184 		if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
185 			vsi->alloc_rxq = 1;
186 		} else {
187 			if (vsi->req_rxq) {
188 				vsi->alloc_rxq = vsi->req_rxq;
189 				vsi->num_rxq = vsi->req_rxq;
190 			} else {
191 				vsi->alloc_rxq = min3(pf->num_lan_msix,
192 						      ice_get_avail_rxq_count(pf),
193 						      (u16)num_online_cpus());
194 			}
195 		}
196 
197 		pf->num_lan_rx = vsi->alloc_rxq;
198 
199 		vsi->num_q_vectors = min_t(int, pf->num_lan_msix,
200 					   max_t(int, vsi->alloc_rxq,
201 						 vsi->alloc_txq));
202 		break;
203 	case ICE_VSI_VF:
204 		vf = &pf->vf[vsi->vf_id];
205 		if (vf->num_req_qs)
206 			vf->num_vf_qs = vf->num_req_qs;
207 		vsi->alloc_txq = vf->num_vf_qs;
208 		vsi->alloc_rxq = vf->num_vf_qs;
209 		/* pf->num_msix_per_vf includes (VF miscellaneous vector +
210 		 * data queue interrupts). Since vsi->num_q_vectors is number
211 		 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
212 		 * original vector count
213 		 */
214 		vsi->num_q_vectors = pf->num_msix_per_vf - ICE_NONQ_VECS_VF;
215 		break;
216 	case ICE_VSI_CTRL:
217 		vsi->alloc_txq = 1;
218 		vsi->alloc_rxq = 1;
219 		vsi->num_q_vectors = 1;
220 		break;
221 	case ICE_VSI_LB:
222 		vsi->alloc_txq = 1;
223 		vsi->alloc_rxq = 1;
224 		break;
225 	default:
226 		dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi->type);
227 		break;
228 	}
229 
230 	ice_vsi_set_num_desc(vsi);
231 }
232 
233 /**
234  * ice_get_free_slot - get the next non-NULL location index in array
235  * @array: array to search
236  * @size: size of the array
237  * @curr: last known occupied index to be used as a search hint
238  *
239  * void * is being used to keep the functionality generic. This lets us use this
240  * function on any array of pointers.
241  */
242 static int ice_get_free_slot(void *array, int size, int curr)
243 {
244 	int **tmp_array = (int **)array;
245 	int next;
246 
247 	if (curr < (size - 1) && !tmp_array[curr + 1]) {
248 		next = curr + 1;
249 	} else {
250 		int i = 0;
251 
252 		while ((i < size) && (tmp_array[i]))
253 			i++;
254 		if (i == size)
255 			next = ICE_NO_VSI;
256 		else
257 			next = i;
258 	}
259 	return next;
260 }
261 
262 /**
263  * ice_vsi_delete - delete a VSI from the switch
264  * @vsi: pointer to VSI being removed
265  */
266 static void ice_vsi_delete(struct ice_vsi *vsi)
267 {
268 	struct ice_pf *pf = vsi->back;
269 	struct ice_vsi_ctx *ctxt;
270 	enum ice_status status;
271 
272 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
273 	if (!ctxt)
274 		return;
275 
276 	if (vsi->type == ICE_VSI_VF)
277 		ctxt->vf_num = vsi->vf_id;
278 	ctxt->vsi_num = vsi->vsi_num;
279 
280 	memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
281 
282 	status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
283 	if (status)
284 		dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %s\n",
285 			vsi->vsi_num, ice_stat_str(status));
286 
287 	kfree(ctxt);
288 }
289 
290 /**
291  * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
292  * @vsi: pointer to VSI being cleared
293  */
294 static void ice_vsi_free_arrays(struct ice_vsi *vsi)
295 {
296 	struct ice_pf *pf = vsi->back;
297 	struct device *dev;
298 
299 	dev = ice_pf_to_dev(pf);
300 
301 	if (vsi->af_xdp_zc_qps) {
302 		bitmap_free(vsi->af_xdp_zc_qps);
303 		vsi->af_xdp_zc_qps = NULL;
304 	}
305 	/* free the ring and vector containers */
306 	if (vsi->q_vectors) {
307 		devm_kfree(dev, vsi->q_vectors);
308 		vsi->q_vectors = NULL;
309 	}
310 	if (vsi->tx_rings) {
311 		devm_kfree(dev, vsi->tx_rings);
312 		vsi->tx_rings = NULL;
313 	}
314 	if (vsi->rx_rings) {
315 		devm_kfree(dev, vsi->rx_rings);
316 		vsi->rx_rings = NULL;
317 	}
318 	if (vsi->txq_map) {
319 		devm_kfree(dev, vsi->txq_map);
320 		vsi->txq_map = NULL;
321 	}
322 	if (vsi->rxq_map) {
323 		devm_kfree(dev, vsi->rxq_map);
324 		vsi->rxq_map = NULL;
325 	}
326 }
327 
328 /**
329  * ice_vsi_clear - clean up and deallocate the provided VSI
330  * @vsi: pointer to VSI being cleared
331  *
332  * This deallocates the VSI's queue resources, removes it from the PF's
333  * VSI array if necessary, and deallocates the VSI
334  *
335  * Returns 0 on success, negative on failure
336  */
337 static int ice_vsi_clear(struct ice_vsi *vsi)
338 {
339 	struct ice_pf *pf = NULL;
340 	struct device *dev;
341 
342 	if (!vsi)
343 		return 0;
344 
345 	if (!vsi->back)
346 		return -EINVAL;
347 
348 	pf = vsi->back;
349 	dev = ice_pf_to_dev(pf);
350 
351 	if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
352 		dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
353 		return -EINVAL;
354 	}
355 
356 	mutex_lock(&pf->sw_mutex);
357 	/* updates the PF for this cleared VSI */
358 
359 	pf->vsi[vsi->idx] = NULL;
360 	if (vsi->idx < pf->next_vsi && vsi->type != ICE_VSI_CTRL)
361 		pf->next_vsi = vsi->idx;
362 	if (vsi->idx < pf->next_vsi && vsi->type == ICE_VSI_CTRL &&
363 	    vsi->vf_id != ICE_INVAL_VFID)
364 		pf->next_vsi = vsi->idx;
365 
366 	ice_vsi_free_arrays(vsi);
367 	mutex_unlock(&pf->sw_mutex);
368 	devm_kfree(dev, vsi);
369 
370 	return 0;
371 }
372 
373 /**
374  * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
375  * @irq: interrupt number
376  * @data: pointer to a q_vector
377  */
378 static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
379 {
380 	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
381 
382 	if (!q_vector->tx.ring)
383 		return IRQ_HANDLED;
384 
385 #define FDIR_RX_DESC_CLEAN_BUDGET 64
386 	ice_clean_rx_irq(q_vector->rx.ring, FDIR_RX_DESC_CLEAN_BUDGET);
387 	ice_clean_ctrl_tx_irq(q_vector->tx.ring);
388 
389 	return IRQ_HANDLED;
390 }
391 
392 /**
393  * ice_msix_clean_rings - MSIX mode Interrupt Handler
394  * @irq: interrupt number
395  * @data: pointer to a q_vector
396  */
397 static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
398 {
399 	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
400 
401 	if (!q_vector->tx.ring && !q_vector->rx.ring)
402 		return IRQ_HANDLED;
403 
404 	q_vector->total_events++;
405 
406 	napi_schedule(&q_vector->napi);
407 
408 	return IRQ_HANDLED;
409 }
410 
411 /**
412  * ice_vsi_alloc - Allocates the next available struct VSI in the PF
413  * @pf: board private structure
414  * @vsi_type: type of VSI
415  * @vf_id: ID of the VF being configured
416  *
417  * returns a pointer to a VSI on success, NULL on failure.
418  */
419 static struct ice_vsi *
420 ice_vsi_alloc(struct ice_pf *pf, enum ice_vsi_type vsi_type, u16 vf_id)
421 {
422 	struct device *dev = ice_pf_to_dev(pf);
423 	struct ice_vsi *vsi = NULL;
424 
425 	/* Need to protect the allocation of the VSIs at the PF level */
426 	mutex_lock(&pf->sw_mutex);
427 
428 	/* If we have already allocated our maximum number of VSIs,
429 	 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
430 	 * is available to be populated
431 	 */
432 	if (pf->next_vsi == ICE_NO_VSI) {
433 		dev_dbg(dev, "out of VSI slots!\n");
434 		goto unlock_pf;
435 	}
436 
437 	vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
438 	if (!vsi)
439 		goto unlock_pf;
440 
441 	vsi->type = vsi_type;
442 	vsi->back = pf;
443 	set_bit(ICE_VSI_DOWN, vsi->state);
444 
445 	if (vsi_type == ICE_VSI_VF)
446 		ice_vsi_set_num_qs(vsi, vf_id);
447 	else
448 		ice_vsi_set_num_qs(vsi, ICE_INVAL_VFID);
449 
450 	switch (vsi->type) {
451 	case ICE_VSI_PF:
452 		if (ice_vsi_alloc_arrays(vsi))
453 			goto err_rings;
454 
455 		/* Setup default MSIX irq handler for VSI */
456 		vsi->irq_handler = ice_msix_clean_rings;
457 		break;
458 	case ICE_VSI_CTRL:
459 		if (ice_vsi_alloc_arrays(vsi))
460 			goto err_rings;
461 
462 		/* Setup ctrl VSI MSIX irq handler */
463 		vsi->irq_handler = ice_msix_clean_ctrl_vsi;
464 		break;
465 	case ICE_VSI_VF:
466 		if (ice_vsi_alloc_arrays(vsi))
467 			goto err_rings;
468 		break;
469 	case ICE_VSI_LB:
470 		if (ice_vsi_alloc_arrays(vsi))
471 			goto err_rings;
472 		break;
473 	default:
474 		dev_warn(dev, "Unknown VSI type %d\n", vsi->type);
475 		goto unlock_pf;
476 	}
477 
478 	if (vsi->type == ICE_VSI_CTRL && vf_id == ICE_INVAL_VFID) {
479 		/* Use the last VSI slot as the index for PF control VSI */
480 		vsi->idx = pf->num_alloc_vsi - 1;
481 		pf->ctrl_vsi_idx = vsi->idx;
482 		pf->vsi[vsi->idx] = vsi;
483 	} else {
484 		/* fill slot and make note of the index */
485 		vsi->idx = pf->next_vsi;
486 		pf->vsi[pf->next_vsi] = vsi;
487 
488 		/* prepare pf->next_vsi for next use */
489 		pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
490 						 pf->next_vsi);
491 	}
492 
493 	if (vsi->type == ICE_VSI_CTRL && vf_id != ICE_INVAL_VFID)
494 		pf->vf[vf_id].ctrl_vsi_idx = vsi->idx;
495 	goto unlock_pf;
496 
497 err_rings:
498 	devm_kfree(dev, vsi);
499 	vsi = NULL;
500 unlock_pf:
501 	mutex_unlock(&pf->sw_mutex);
502 	return vsi;
503 }
504 
505 /**
506  * ice_alloc_fd_res - Allocate FD resource for a VSI
507  * @vsi: pointer to the ice_vsi
508  *
509  * This allocates the FD resources
510  *
511  * Returns 0 on success, -EPERM on no-op or -EIO on failure
512  */
513 static int ice_alloc_fd_res(struct ice_vsi *vsi)
514 {
515 	struct ice_pf *pf = vsi->back;
516 	u32 g_val, b_val;
517 
518 	/* Flow Director filters are only allocated/assigned to the PF VSI which
519 	 * passes the traffic. The CTRL VSI is only used to add/delete filters
520 	 * so we don't allocate resources to it
521 	 */
522 
523 	/* FD filters from guaranteed pool per VSI */
524 	g_val = pf->hw.func_caps.fd_fltr_guar;
525 	if (!g_val)
526 		return -EPERM;
527 
528 	/* FD filters from best effort pool */
529 	b_val = pf->hw.func_caps.fd_fltr_best_effort;
530 	if (!b_val)
531 		return -EPERM;
532 
533 	if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF))
534 		return -EPERM;
535 
536 	if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
537 		return -EPERM;
538 
539 	vsi->num_gfltr = g_val / pf->num_alloc_vsi;
540 
541 	/* each VSI gets same "best_effort" quota */
542 	vsi->num_bfltr = b_val;
543 
544 	if (vsi->type == ICE_VSI_VF) {
545 		vsi->num_gfltr = 0;
546 
547 		/* each VSI gets same "best_effort" quota */
548 		vsi->num_bfltr = b_val;
549 	}
550 
551 	return 0;
552 }
553 
554 /**
555  * ice_vsi_get_qs - Assign queues from PF to VSI
556  * @vsi: the VSI to assign queues to
557  *
558  * Returns 0 on success and a negative value on error
559  */
560 static int ice_vsi_get_qs(struct ice_vsi *vsi)
561 {
562 	struct ice_pf *pf = vsi->back;
563 	struct ice_qs_cfg tx_qs_cfg = {
564 		.qs_mutex = &pf->avail_q_mutex,
565 		.pf_map = pf->avail_txqs,
566 		.pf_map_size = pf->max_pf_txqs,
567 		.q_count = vsi->alloc_txq,
568 		.scatter_count = ICE_MAX_SCATTER_TXQS,
569 		.vsi_map = vsi->txq_map,
570 		.vsi_map_offset = 0,
571 		.mapping_mode = ICE_VSI_MAP_CONTIG
572 	};
573 	struct ice_qs_cfg rx_qs_cfg = {
574 		.qs_mutex = &pf->avail_q_mutex,
575 		.pf_map = pf->avail_rxqs,
576 		.pf_map_size = pf->max_pf_rxqs,
577 		.q_count = vsi->alloc_rxq,
578 		.scatter_count = ICE_MAX_SCATTER_RXQS,
579 		.vsi_map = vsi->rxq_map,
580 		.vsi_map_offset = 0,
581 		.mapping_mode = ICE_VSI_MAP_CONTIG
582 	};
583 	int ret;
584 
585 	ret = __ice_vsi_get_qs(&tx_qs_cfg);
586 	if (ret)
587 		return ret;
588 	vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
589 
590 	ret = __ice_vsi_get_qs(&rx_qs_cfg);
591 	if (ret)
592 		return ret;
593 	vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
594 
595 	return 0;
596 }
597 
598 /**
599  * ice_vsi_put_qs - Release queues from VSI to PF
600  * @vsi: the VSI that is going to release queues
601  */
602 static void ice_vsi_put_qs(struct ice_vsi *vsi)
603 {
604 	struct ice_pf *pf = vsi->back;
605 	int i;
606 
607 	mutex_lock(&pf->avail_q_mutex);
608 
609 	for (i = 0; i < vsi->alloc_txq; i++) {
610 		clear_bit(vsi->txq_map[i], pf->avail_txqs);
611 		vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
612 	}
613 
614 	for (i = 0; i < vsi->alloc_rxq; i++) {
615 		clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
616 		vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
617 	}
618 
619 	mutex_unlock(&pf->avail_q_mutex);
620 }
621 
622 /**
623  * ice_is_safe_mode
624  * @pf: pointer to the PF struct
625  *
626  * returns true if driver is in safe mode, false otherwise
627  */
628 bool ice_is_safe_mode(struct ice_pf *pf)
629 {
630 	return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
631 }
632 
633 /**
634  * ice_is_aux_ena
635  * @pf: pointer to the PF struct
636  *
637  * returns true if AUX devices/drivers are supported, false otherwise
638  */
639 bool ice_is_aux_ena(struct ice_pf *pf)
640 {
641 	return test_bit(ICE_FLAG_AUX_ENA, pf->flags);
642 }
643 
644 /**
645  * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
646  * @vsi: the VSI being cleaned up
647  *
648  * This function deletes RSS input set for all flows that were configured
649  * for this VSI
650  */
651 static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
652 {
653 	struct ice_pf *pf = vsi->back;
654 	enum ice_status status;
655 
656 	if (ice_is_safe_mode(pf))
657 		return;
658 
659 	status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
660 	if (status)
661 		dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %s\n",
662 			vsi->vsi_num, ice_stat_str(status));
663 }
664 
665 /**
666  * ice_rss_clean - Delete RSS related VSI structures and configuration
667  * @vsi: the VSI being removed
668  */
669 static void ice_rss_clean(struct ice_vsi *vsi)
670 {
671 	struct ice_pf *pf = vsi->back;
672 	struct device *dev;
673 
674 	dev = ice_pf_to_dev(pf);
675 
676 	if (vsi->rss_hkey_user)
677 		devm_kfree(dev, vsi->rss_hkey_user);
678 	if (vsi->rss_lut_user)
679 		devm_kfree(dev, vsi->rss_lut_user);
680 
681 	ice_vsi_clean_rss_flow_fld(vsi);
682 	/* remove RSS replay list */
683 	if (!ice_is_safe_mode(pf))
684 		ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
685 }
686 
687 /**
688  * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
689  * @vsi: the VSI being configured
690  */
691 static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
692 {
693 	struct ice_hw_common_caps *cap;
694 	struct ice_pf *pf = vsi->back;
695 
696 	if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
697 		vsi->rss_size = 1;
698 		return;
699 	}
700 
701 	cap = &pf->hw.func_caps.common_cap;
702 	switch (vsi->type) {
703 	case ICE_VSI_PF:
704 		/* PF VSI will inherit RSS instance of PF */
705 		vsi->rss_table_size = (u16)cap->rss_table_size;
706 		vsi->rss_size = min_t(u16, num_online_cpus(),
707 				      BIT(cap->rss_table_entry_width));
708 		vsi->rss_lut_type = ICE_AQC_GSET_RSS_LUT_TABLE_TYPE_PF;
709 		break;
710 	case ICE_VSI_VF:
711 		/* VF VSI will get a small RSS table.
712 		 * For VSI_LUT, LUT size should be set to 64 bytes.
713 		 */
714 		vsi->rss_table_size = ICE_VSIQF_HLUT_ARRAY_SIZE;
715 		vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
716 		vsi->rss_lut_type = ICE_AQC_GSET_RSS_LUT_TABLE_TYPE_VSI;
717 		break;
718 	case ICE_VSI_LB:
719 		break;
720 	default:
721 		dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
722 			ice_vsi_type_str(vsi->type));
723 		break;
724 	}
725 }
726 
727 /**
728  * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
729  * @ctxt: the VSI context being set
730  *
731  * This initializes a default VSI context for all sections except the Queues.
732  */
733 static void ice_set_dflt_vsi_ctx(struct ice_vsi_ctx *ctxt)
734 {
735 	u32 table = 0;
736 
737 	memset(&ctxt->info, 0, sizeof(ctxt->info));
738 	/* VSI's should be allocated from shared pool */
739 	ctxt->alloc_from_pool = true;
740 	/* Src pruning enabled by default */
741 	ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
742 	/* Traffic from VSI can be sent to LAN */
743 	ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
744 	/* By default bits 3 and 4 in vlan_flags are 0's which results in legacy
745 	 * behavior (show VLAN, DEI, and UP) in descriptor. Also, allow all
746 	 * packets untagged/tagged.
747 	 */
748 	ctxt->info.vlan_flags = ((ICE_AQ_VSI_VLAN_MODE_ALL &
749 				  ICE_AQ_VSI_VLAN_MODE_M) >>
750 				 ICE_AQ_VSI_VLAN_MODE_S);
751 	/* Have 1:1 UP mapping for both ingress/egress tables */
752 	table |= ICE_UP_TABLE_TRANSLATE(0, 0);
753 	table |= ICE_UP_TABLE_TRANSLATE(1, 1);
754 	table |= ICE_UP_TABLE_TRANSLATE(2, 2);
755 	table |= ICE_UP_TABLE_TRANSLATE(3, 3);
756 	table |= ICE_UP_TABLE_TRANSLATE(4, 4);
757 	table |= ICE_UP_TABLE_TRANSLATE(5, 5);
758 	table |= ICE_UP_TABLE_TRANSLATE(6, 6);
759 	table |= ICE_UP_TABLE_TRANSLATE(7, 7);
760 	ctxt->info.ingress_table = cpu_to_le32(table);
761 	ctxt->info.egress_table = cpu_to_le32(table);
762 	/* Have 1:1 UP mapping for outer to inner UP table */
763 	ctxt->info.outer_up_table = cpu_to_le32(table);
764 	/* No Outer tag support outer_tag_flags remains to zero */
765 }
766 
767 /**
768  * ice_vsi_setup_q_map - Setup a VSI queue map
769  * @vsi: the VSI being configured
770  * @ctxt: VSI context structure
771  */
772 static void ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
773 {
774 	u16 offset = 0, qmap = 0, tx_count = 0, pow = 0;
775 	u16 num_txq_per_tc, num_rxq_per_tc;
776 	u16 qcount_tx = vsi->alloc_txq;
777 	u16 qcount_rx = vsi->alloc_rxq;
778 	bool ena_tc0 = false;
779 	u8 netdev_tc = 0;
780 	int i;
781 
782 	/* at least TC0 should be enabled by default */
783 	if (vsi->tc_cfg.numtc) {
784 		if (!(vsi->tc_cfg.ena_tc & BIT(0)))
785 			ena_tc0 = true;
786 	} else {
787 		ena_tc0 = true;
788 	}
789 
790 	if (ena_tc0) {
791 		vsi->tc_cfg.numtc++;
792 		vsi->tc_cfg.ena_tc |= 1;
793 	}
794 
795 	num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
796 	if (!num_rxq_per_tc)
797 		num_rxq_per_tc = 1;
798 	num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
799 	if (!num_txq_per_tc)
800 		num_txq_per_tc = 1;
801 
802 	/* find the (rounded up) power-of-2 of qcount */
803 	pow = (u16)order_base_2(num_rxq_per_tc);
804 
805 	/* TC mapping is a function of the number of Rx queues assigned to the
806 	 * VSI for each traffic class and the offset of these queues.
807 	 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
808 	 * queues allocated to TC0. No:of queues is a power-of-2.
809 	 *
810 	 * If TC is not enabled, the queue offset is set to 0, and allocate one
811 	 * queue, this way, traffic for the given TC will be sent to the default
812 	 * queue.
813 	 *
814 	 * Setup number and offset of Rx queues for all TCs for the VSI
815 	 */
816 	ice_for_each_traffic_class(i) {
817 		if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
818 			/* TC is not enabled */
819 			vsi->tc_cfg.tc_info[i].qoffset = 0;
820 			vsi->tc_cfg.tc_info[i].qcount_rx = 1;
821 			vsi->tc_cfg.tc_info[i].qcount_tx = 1;
822 			vsi->tc_cfg.tc_info[i].netdev_tc = 0;
823 			ctxt->info.tc_mapping[i] = 0;
824 			continue;
825 		}
826 
827 		/* TC is enabled */
828 		vsi->tc_cfg.tc_info[i].qoffset = offset;
829 		vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
830 		vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
831 		vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
832 
833 		qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
834 			ICE_AQ_VSI_TC_Q_OFFSET_M) |
835 			((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
836 			 ICE_AQ_VSI_TC_Q_NUM_M);
837 		offset += num_rxq_per_tc;
838 		tx_count += num_txq_per_tc;
839 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
840 	}
841 
842 	/* if offset is non-zero, means it is calculated correctly based on
843 	 * enabled TCs for a given VSI otherwise qcount_rx will always
844 	 * be correct and non-zero because it is based off - VSI's
845 	 * allocated Rx queues which is at least 1 (hence qcount_tx will be
846 	 * at least 1)
847 	 */
848 	if (offset)
849 		vsi->num_rxq = offset;
850 	else
851 		vsi->num_rxq = num_rxq_per_tc;
852 
853 	vsi->num_txq = tx_count;
854 
855 	if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
856 		dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
857 		/* since there is a chance that num_rxq could have been changed
858 		 * in the above for loop, make num_txq equal to num_rxq.
859 		 */
860 		vsi->num_txq = vsi->num_rxq;
861 	}
862 
863 	/* Rx queue mapping */
864 	ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
865 	/* q_mapping buffer holds the info for the first queue allocated for
866 	 * this VSI in the PF space and also the number of queues associated
867 	 * with this VSI.
868 	 */
869 	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
870 	ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
871 }
872 
873 /**
874  * ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
875  * @ctxt: the VSI context being set
876  * @vsi: the VSI being configured
877  */
878 static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
879 {
880 	u8 dflt_q_group, dflt_q_prio;
881 	u16 dflt_q, report_q, val;
882 
883 	if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
884 	    vsi->type != ICE_VSI_VF)
885 		return;
886 
887 	val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
888 	ctxt->info.valid_sections |= cpu_to_le16(val);
889 	dflt_q = 0;
890 	dflt_q_group = 0;
891 	report_q = 0;
892 	dflt_q_prio = 0;
893 
894 	/* enable flow director filtering/programming */
895 	val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
896 	ctxt->info.fd_options = cpu_to_le16(val);
897 	/* max of allocated flow director filters */
898 	ctxt->info.max_fd_fltr_dedicated =
899 			cpu_to_le16(vsi->num_gfltr);
900 	/* max of shared flow director filters any VSI may program */
901 	ctxt->info.max_fd_fltr_shared =
902 			cpu_to_le16(vsi->num_bfltr);
903 	/* default queue index within the VSI of the default FD */
904 	val = ((dflt_q << ICE_AQ_VSI_FD_DEF_Q_S) &
905 	       ICE_AQ_VSI_FD_DEF_Q_M);
906 	/* target queue or queue group to the FD filter */
907 	val |= ((dflt_q_group << ICE_AQ_VSI_FD_DEF_GRP_S) &
908 		ICE_AQ_VSI_FD_DEF_GRP_M);
909 	ctxt->info.fd_def_q = cpu_to_le16(val);
910 	/* queue index on which FD filter completion is reported */
911 	val = ((report_q << ICE_AQ_VSI_FD_REPORT_Q_S) &
912 	       ICE_AQ_VSI_FD_REPORT_Q_M);
913 	/* priority of the default qindex action */
914 	val |= ((dflt_q_prio << ICE_AQ_VSI_FD_DEF_PRIORITY_S) &
915 		ICE_AQ_VSI_FD_DEF_PRIORITY_M);
916 	ctxt->info.fd_report_opt = cpu_to_le16(val);
917 }
918 
919 /**
920  * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
921  * @ctxt: the VSI context being set
922  * @vsi: the VSI being configured
923  */
924 static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
925 {
926 	u8 lut_type, hash_type;
927 	struct device *dev;
928 	struct ice_pf *pf;
929 
930 	pf = vsi->back;
931 	dev = ice_pf_to_dev(pf);
932 
933 	switch (vsi->type) {
934 	case ICE_VSI_PF:
935 		/* PF VSI will inherit RSS instance of PF */
936 		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
937 		hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
938 		break;
939 	case ICE_VSI_VF:
940 		/* VF VSI will gets a small RSS table which is a VSI LUT type */
941 		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
942 		hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
943 		break;
944 	default:
945 		dev_dbg(dev, "Unsupported VSI type %s\n",
946 			ice_vsi_type_str(vsi->type));
947 		return;
948 	}
949 
950 	ctxt->info.q_opt_rss = ((lut_type << ICE_AQ_VSI_Q_OPT_RSS_LUT_S) &
951 				ICE_AQ_VSI_Q_OPT_RSS_LUT_M) |
952 				((hash_type << ICE_AQ_VSI_Q_OPT_RSS_HASH_S) &
953 				 ICE_AQ_VSI_Q_OPT_RSS_HASH_M);
954 }
955 
956 /**
957  * ice_vsi_init - Create and initialize a VSI
958  * @vsi: the VSI being configured
959  * @init_vsi: is this call creating a VSI
960  *
961  * This initializes a VSI context depending on the VSI type to be added and
962  * passes it down to the add_vsi aq command to create a new VSI.
963  */
964 static int ice_vsi_init(struct ice_vsi *vsi, bool init_vsi)
965 {
966 	struct ice_pf *pf = vsi->back;
967 	struct ice_hw *hw = &pf->hw;
968 	struct ice_vsi_ctx *ctxt;
969 	struct device *dev;
970 	int ret = 0;
971 
972 	dev = ice_pf_to_dev(pf);
973 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
974 	if (!ctxt)
975 		return -ENOMEM;
976 
977 	switch (vsi->type) {
978 	case ICE_VSI_CTRL:
979 	case ICE_VSI_LB:
980 	case ICE_VSI_PF:
981 		ctxt->flags = ICE_AQ_VSI_TYPE_PF;
982 		break;
983 	case ICE_VSI_VF:
984 		ctxt->flags = ICE_AQ_VSI_TYPE_VF;
985 		/* VF number here is the absolute VF number (0-255) */
986 		ctxt->vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
987 		break;
988 	default:
989 		ret = -ENODEV;
990 		goto out;
991 	}
992 
993 	ice_set_dflt_vsi_ctx(ctxt);
994 	if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
995 		ice_set_fd_vsi_ctx(ctxt, vsi);
996 	/* if the switch is in VEB mode, allow VSI loopback */
997 	if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
998 		ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
999 
1000 	/* Set LUT type and HASH type if RSS is enabled */
1001 	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1002 	    vsi->type != ICE_VSI_CTRL) {
1003 		ice_set_rss_vsi_ctx(ctxt, vsi);
1004 		/* if updating VSI context, make sure to set valid_section:
1005 		 * to indicate which section of VSI context being updated
1006 		 */
1007 		if (!init_vsi)
1008 			ctxt->info.valid_sections |=
1009 				cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1010 	}
1011 
1012 	ctxt->info.sw_id = vsi->port_info->sw_id;
1013 	ice_vsi_setup_q_map(vsi, ctxt);
1014 	if (!init_vsi) /* means VSI being updated */
1015 		/* must to indicate which section of VSI context are
1016 		 * being modified
1017 		 */
1018 		ctxt->info.valid_sections |=
1019 			cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1020 
1021 	/* enable/disable MAC and VLAN anti-spoof when spoofchk is on/off
1022 	 * respectively
1023 	 */
1024 	if (vsi->type == ICE_VSI_VF) {
1025 		ctxt->info.valid_sections |=
1026 			cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1027 		if (pf->vf[vsi->vf_id].spoofchk) {
1028 			ctxt->info.sec_flags |=
1029 				ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
1030 				(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
1031 				 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
1032 		} else {
1033 			ctxt->info.sec_flags &=
1034 				~(ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
1035 				  (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
1036 				   ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S));
1037 		}
1038 	}
1039 
1040 	/* Allow control frames out of main VSI */
1041 	if (vsi->type == ICE_VSI_PF) {
1042 		ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1043 		ctxt->info.valid_sections |=
1044 			cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1045 	}
1046 
1047 	if (init_vsi) {
1048 		ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1049 		if (ret) {
1050 			dev_err(dev, "Add VSI failed, err %d\n", ret);
1051 			ret = -EIO;
1052 			goto out;
1053 		}
1054 	} else {
1055 		ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1056 		if (ret) {
1057 			dev_err(dev, "Update VSI failed, err %d\n", ret);
1058 			ret = -EIO;
1059 			goto out;
1060 		}
1061 	}
1062 
1063 	/* keep context for update VSI operations */
1064 	vsi->info = ctxt->info;
1065 
1066 	/* record VSI number returned */
1067 	vsi->vsi_num = ctxt->vsi_num;
1068 
1069 out:
1070 	kfree(ctxt);
1071 	return ret;
1072 }
1073 
1074 /**
1075  * ice_free_res - free a block of resources
1076  * @res: pointer to the resource
1077  * @index: starting index previously returned by ice_get_res
1078  * @id: identifier to track owner
1079  *
1080  * Returns number of resources freed
1081  */
1082 int ice_free_res(struct ice_res_tracker *res, u16 index, u16 id)
1083 {
1084 	int count = 0;
1085 	int i;
1086 
1087 	if (!res || index >= res->end)
1088 		return -EINVAL;
1089 
1090 	id |= ICE_RES_VALID_BIT;
1091 	for (i = index; i < res->end && res->list[i] == id; i++) {
1092 		res->list[i] = 0;
1093 		count++;
1094 	}
1095 
1096 	return count;
1097 }
1098 
1099 /**
1100  * ice_search_res - Search the tracker for a block of resources
1101  * @res: pointer to the resource
1102  * @needed: size of the block needed
1103  * @id: identifier to track owner
1104  *
1105  * Returns the base item index of the block, or -ENOMEM for error
1106  */
1107 static int ice_search_res(struct ice_res_tracker *res, u16 needed, u16 id)
1108 {
1109 	u16 start = 0, end = 0;
1110 
1111 	if (needed > res->end)
1112 		return -ENOMEM;
1113 
1114 	id |= ICE_RES_VALID_BIT;
1115 
1116 	do {
1117 		/* skip already allocated entries */
1118 		if (res->list[end++] & ICE_RES_VALID_BIT) {
1119 			start = end;
1120 			if ((start + needed) > res->end)
1121 				break;
1122 		}
1123 
1124 		if (end == (start + needed)) {
1125 			int i = start;
1126 
1127 			/* there was enough, so assign it to the requestor */
1128 			while (i != end)
1129 				res->list[i++] = id;
1130 
1131 			return start;
1132 		}
1133 	} while (end < res->end);
1134 
1135 	return -ENOMEM;
1136 }
1137 
1138 /**
1139  * ice_get_free_res_count - Get free count from a resource tracker
1140  * @res: Resource tracker instance
1141  */
1142 static u16 ice_get_free_res_count(struct ice_res_tracker *res)
1143 {
1144 	u16 i, count = 0;
1145 
1146 	for (i = 0; i < res->end; i++)
1147 		if (!(res->list[i] & ICE_RES_VALID_BIT))
1148 			count++;
1149 
1150 	return count;
1151 }
1152 
1153 /**
1154  * ice_get_res - get a block of resources
1155  * @pf: board private structure
1156  * @res: pointer to the resource
1157  * @needed: size of the block needed
1158  * @id: identifier to track owner
1159  *
1160  * Returns the base item index of the block, or negative for error
1161  */
1162 int
1163 ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res, u16 needed, u16 id)
1164 {
1165 	if (!res || !pf)
1166 		return -EINVAL;
1167 
1168 	if (!needed || needed > res->num_entries || id >= ICE_RES_VALID_BIT) {
1169 		dev_err(ice_pf_to_dev(pf), "param err: needed=%d, num_entries = %d id=0x%04x\n",
1170 			needed, res->num_entries, id);
1171 		return -EINVAL;
1172 	}
1173 
1174 	return ice_search_res(res, needed, id);
1175 }
1176 
1177 /**
1178  * ice_vsi_setup_vector_base - Set up the base vector for the given VSI
1179  * @vsi: ptr to the VSI
1180  *
1181  * This should only be called after ice_vsi_alloc() which allocates the
1182  * corresponding SW VSI structure and initializes num_queue_pairs for the
1183  * newly allocated VSI.
1184  *
1185  * Returns 0 on success or negative on failure
1186  */
1187 static int ice_vsi_setup_vector_base(struct ice_vsi *vsi)
1188 {
1189 	struct ice_pf *pf = vsi->back;
1190 	struct device *dev;
1191 	u16 num_q_vectors;
1192 	int base;
1193 
1194 	dev = ice_pf_to_dev(pf);
1195 	/* SRIOV doesn't grab irq_tracker entries for each VSI */
1196 	if (vsi->type == ICE_VSI_VF)
1197 		return 0;
1198 
1199 	if (vsi->base_vector) {
1200 		dev_dbg(dev, "VSI %d has non-zero base vector %d\n",
1201 			vsi->vsi_num, vsi->base_vector);
1202 		return -EEXIST;
1203 	}
1204 
1205 	num_q_vectors = vsi->num_q_vectors;
1206 	/* reserve slots from OS requested IRQs */
1207 	if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID) {
1208 		struct ice_vf *vf;
1209 		int i;
1210 
1211 		ice_for_each_vf(pf, i) {
1212 			vf = &pf->vf[i];
1213 			if (i != vsi->vf_id && vf->ctrl_vsi_idx != ICE_NO_VSI) {
1214 				base = pf->vsi[vf->ctrl_vsi_idx]->base_vector;
1215 				break;
1216 			}
1217 		}
1218 		if (i == pf->num_alloc_vfs)
1219 			base = ice_get_res(pf, pf->irq_tracker, num_q_vectors,
1220 					   ICE_RES_VF_CTRL_VEC_ID);
1221 	} else {
1222 		base = ice_get_res(pf, pf->irq_tracker, num_q_vectors,
1223 				   vsi->idx);
1224 	}
1225 
1226 	if (base < 0) {
1227 		dev_err(dev, "%d MSI-X interrupts available. %s %d failed to get %d MSI-X vectors\n",
1228 			ice_get_free_res_count(pf->irq_tracker),
1229 			ice_vsi_type_str(vsi->type), vsi->idx, num_q_vectors);
1230 		return -ENOENT;
1231 	}
1232 	vsi->base_vector = (u16)base;
1233 	pf->num_avail_sw_msix -= num_q_vectors;
1234 
1235 	return 0;
1236 }
1237 
1238 /**
1239  * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1240  * @vsi: the VSI having rings deallocated
1241  */
1242 static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1243 {
1244 	int i;
1245 
1246 	/* Avoid stale references by clearing map from vector to ring */
1247 	if (vsi->q_vectors) {
1248 		ice_for_each_q_vector(vsi, i) {
1249 			struct ice_q_vector *q_vector = vsi->q_vectors[i];
1250 
1251 			if (q_vector) {
1252 				q_vector->tx.ring = NULL;
1253 				q_vector->rx.ring = NULL;
1254 			}
1255 		}
1256 	}
1257 
1258 	if (vsi->tx_rings) {
1259 		for (i = 0; i < vsi->alloc_txq; i++) {
1260 			if (vsi->tx_rings[i]) {
1261 				kfree_rcu(vsi->tx_rings[i], rcu);
1262 				WRITE_ONCE(vsi->tx_rings[i], NULL);
1263 			}
1264 		}
1265 	}
1266 	if (vsi->rx_rings) {
1267 		for (i = 0; i < vsi->alloc_rxq; i++) {
1268 			if (vsi->rx_rings[i]) {
1269 				kfree_rcu(vsi->rx_rings[i], rcu);
1270 				WRITE_ONCE(vsi->rx_rings[i], NULL);
1271 			}
1272 		}
1273 	}
1274 }
1275 
1276 /**
1277  * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1278  * @vsi: VSI which is having rings allocated
1279  */
1280 static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1281 {
1282 	struct ice_pf *pf = vsi->back;
1283 	struct device *dev;
1284 	u16 i;
1285 
1286 	dev = ice_pf_to_dev(pf);
1287 	/* Allocate Tx rings */
1288 	for (i = 0; i < vsi->alloc_txq; i++) {
1289 		struct ice_ring *ring;
1290 
1291 		/* allocate with kzalloc(), free with kfree_rcu() */
1292 		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1293 
1294 		if (!ring)
1295 			goto err_out;
1296 
1297 		ring->q_index = i;
1298 		ring->reg_idx = vsi->txq_map[i];
1299 		ring->ring_active = false;
1300 		ring->vsi = vsi;
1301 		ring->dev = dev;
1302 		ring->count = vsi->num_tx_desc;
1303 		WRITE_ONCE(vsi->tx_rings[i], ring);
1304 	}
1305 
1306 	/* Allocate Rx rings */
1307 	for (i = 0; i < vsi->alloc_rxq; i++) {
1308 		struct ice_ring *ring;
1309 
1310 		/* allocate with kzalloc(), free with kfree_rcu() */
1311 		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1312 		if (!ring)
1313 			goto err_out;
1314 
1315 		ring->q_index = i;
1316 		ring->reg_idx = vsi->rxq_map[i];
1317 		ring->ring_active = false;
1318 		ring->vsi = vsi;
1319 		ring->netdev = vsi->netdev;
1320 		ring->dev = dev;
1321 		ring->count = vsi->num_rx_desc;
1322 		WRITE_ONCE(vsi->rx_rings[i], ring);
1323 	}
1324 
1325 	return 0;
1326 
1327 err_out:
1328 	ice_vsi_clear_rings(vsi);
1329 	return -ENOMEM;
1330 }
1331 
1332 /**
1333  * ice_vsi_manage_rss_lut - disable/enable RSS
1334  * @vsi: the VSI being changed
1335  * @ena: boolean value indicating if this is an enable or disable request
1336  *
1337  * In the event of disable request for RSS, this function will zero out RSS
1338  * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1339  * LUT.
1340  */
1341 void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1342 {
1343 	u8 *lut;
1344 
1345 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1346 	if (!lut)
1347 		return;
1348 
1349 	if (ena) {
1350 		if (vsi->rss_lut_user)
1351 			memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1352 		else
1353 			ice_fill_rss_lut(lut, vsi->rss_table_size,
1354 					 vsi->rss_size);
1355 	}
1356 
1357 	ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1358 	kfree(lut);
1359 }
1360 
1361 /**
1362  * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1363  * @vsi: VSI to be configured
1364  */
1365 static int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1366 {
1367 	struct ice_pf *pf = vsi->back;
1368 	struct device *dev;
1369 	u8 *lut, *key;
1370 	int err;
1371 
1372 	dev = ice_pf_to_dev(pf);
1373 	vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1374 
1375 	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1376 	if (!lut)
1377 		return -ENOMEM;
1378 
1379 	if (vsi->rss_lut_user)
1380 		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1381 	else
1382 		ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1383 
1384 	err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1385 	if (err) {
1386 		dev_err(dev, "set_rss_lut failed, error %d\n", err);
1387 		goto ice_vsi_cfg_rss_exit;
1388 	}
1389 
1390 	key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1391 	if (!key) {
1392 		err = -ENOMEM;
1393 		goto ice_vsi_cfg_rss_exit;
1394 	}
1395 
1396 	if (vsi->rss_hkey_user)
1397 		memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1398 	else
1399 		netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1400 
1401 	err = ice_set_rss_key(vsi, key);
1402 	if (err)
1403 		dev_err(dev, "set_rss_key failed, error %d\n", err);
1404 
1405 	kfree(key);
1406 ice_vsi_cfg_rss_exit:
1407 	kfree(lut);
1408 	return err;
1409 }
1410 
1411 /**
1412  * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1413  * @vsi: VSI to be configured
1414  *
1415  * This function will only be called during the VF VSI setup. Upon successful
1416  * completion of package download, this function will configure default RSS
1417  * input sets for VF VSI.
1418  */
1419 static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1420 {
1421 	struct ice_pf *pf = vsi->back;
1422 	enum ice_status status;
1423 	struct device *dev;
1424 
1425 	dev = ice_pf_to_dev(pf);
1426 	if (ice_is_safe_mode(pf)) {
1427 		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1428 			vsi->vsi_num);
1429 		return;
1430 	}
1431 
1432 	status = ice_add_avf_rss_cfg(&pf->hw, vsi->idx, ICE_DEFAULT_RSS_HENA);
1433 	if (status)
1434 		dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %s\n",
1435 			vsi->vsi_num, ice_stat_str(status));
1436 }
1437 
1438 /**
1439  * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1440  * @vsi: VSI to be configured
1441  *
1442  * This function will only be called after successful download package call
1443  * during initialization of PF. Since the downloaded package will erase the
1444  * RSS section, this function will configure RSS input sets for different
1445  * flow types. The last profile added has the highest priority, therefore 2
1446  * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1447  * (i.e. IPv4 src/dst TCP src/dst port).
1448  */
1449 static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1450 {
1451 	u16 vsi_handle = vsi->idx, vsi_num = vsi->vsi_num;
1452 	struct ice_pf *pf = vsi->back;
1453 	struct ice_hw *hw = &pf->hw;
1454 	enum ice_status status;
1455 	struct device *dev;
1456 
1457 	dev = ice_pf_to_dev(pf);
1458 	if (ice_is_safe_mode(pf)) {
1459 		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1460 			vsi_num);
1461 		return;
1462 	}
1463 	/* configure RSS for IPv4 with input set IP src/dst */
1464 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1465 				 ICE_FLOW_SEG_HDR_IPV4);
1466 	if (status)
1467 		dev_dbg(dev, "ice_add_rss_cfg failed for ipv4 flow, vsi = %d, error = %s\n",
1468 			vsi_num, ice_stat_str(status));
1469 
1470 	/* configure RSS for IPv6 with input set IPv6 src/dst */
1471 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1472 				 ICE_FLOW_SEG_HDR_IPV6);
1473 	if (status)
1474 		dev_dbg(dev, "ice_add_rss_cfg failed for ipv6 flow, vsi = %d, error = %s\n",
1475 			vsi_num, ice_stat_str(status));
1476 
1477 	/* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1478 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV4,
1479 				 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4);
1480 	if (status)
1481 		dev_dbg(dev, "ice_add_rss_cfg failed for tcp4 flow, vsi = %d, error = %s\n",
1482 			vsi_num, ice_stat_str(status));
1483 
1484 	/* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1485 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV4,
1486 				 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4);
1487 	if (status)
1488 		dev_dbg(dev, "ice_add_rss_cfg failed for udp4 flow, vsi = %d, error = %s\n",
1489 			vsi_num, ice_stat_str(status));
1490 
1491 	/* configure RSS for sctp4 with input set IP src/dst */
1492 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1493 				 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4);
1494 	if (status)
1495 		dev_dbg(dev, "ice_add_rss_cfg failed for sctp4 flow, vsi = %d, error = %s\n",
1496 			vsi_num, ice_stat_str(status));
1497 
1498 	/* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1499 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV6,
1500 				 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6);
1501 	if (status)
1502 		dev_dbg(dev, "ice_add_rss_cfg failed for tcp6 flow, vsi = %d, error = %s\n",
1503 			vsi_num, ice_stat_str(status));
1504 
1505 	/* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1506 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV6,
1507 				 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6);
1508 	if (status)
1509 		dev_dbg(dev, "ice_add_rss_cfg failed for udp6 flow, vsi = %d, error = %s\n",
1510 			vsi_num, ice_stat_str(status));
1511 
1512 	/* configure RSS for sctp6 with input set IPv6 src/dst */
1513 	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1514 				 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6);
1515 	if (status)
1516 		dev_dbg(dev, "ice_add_rss_cfg failed for sctp6 flow, vsi = %d, error = %s\n",
1517 			vsi_num, ice_stat_str(status));
1518 }
1519 
1520 /**
1521  * ice_pf_state_is_nominal - checks the PF for nominal state
1522  * @pf: pointer to PF to check
1523  *
1524  * Check the PF's state for a collection of bits that would indicate
1525  * the PF is in a state that would inhibit normal operation for
1526  * driver functionality.
1527  *
1528  * Returns true if PF is in a nominal state, false otherwise
1529  */
1530 bool ice_pf_state_is_nominal(struct ice_pf *pf)
1531 {
1532 	DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1533 
1534 	if (!pf)
1535 		return false;
1536 
1537 	bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1538 	if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1539 		return false;
1540 
1541 	return true;
1542 }
1543 
1544 /**
1545  * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1546  * @vsi: the VSI to be updated
1547  */
1548 void ice_update_eth_stats(struct ice_vsi *vsi)
1549 {
1550 	struct ice_eth_stats *prev_es, *cur_es;
1551 	struct ice_hw *hw = &vsi->back->hw;
1552 	u16 vsi_num = vsi->vsi_num;    /* HW absolute index of a VSI */
1553 
1554 	prev_es = &vsi->eth_stats_prev;
1555 	cur_es = &vsi->eth_stats;
1556 
1557 	ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1558 			  &prev_es->rx_bytes, &cur_es->rx_bytes);
1559 
1560 	ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1561 			  &prev_es->rx_unicast, &cur_es->rx_unicast);
1562 
1563 	ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1564 			  &prev_es->rx_multicast, &cur_es->rx_multicast);
1565 
1566 	ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1567 			  &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1568 
1569 	ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1570 			  &prev_es->rx_discards, &cur_es->rx_discards);
1571 
1572 	ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1573 			  &prev_es->tx_bytes, &cur_es->tx_bytes);
1574 
1575 	ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1576 			  &prev_es->tx_unicast, &cur_es->tx_unicast);
1577 
1578 	ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1579 			  &prev_es->tx_multicast, &cur_es->tx_multicast);
1580 
1581 	ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1582 			  &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1583 
1584 	ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1585 			  &prev_es->tx_errors, &cur_es->tx_errors);
1586 
1587 	vsi->stat_offsets_loaded = true;
1588 }
1589 
1590 /**
1591  * ice_vsi_add_vlan - Add VSI membership for given VLAN
1592  * @vsi: the VSI being configured
1593  * @vid: VLAN ID to be added
1594  * @action: filter action to be performed on match
1595  */
1596 int
1597 ice_vsi_add_vlan(struct ice_vsi *vsi, u16 vid, enum ice_sw_fwd_act_type action)
1598 {
1599 	struct ice_pf *pf = vsi->back;
1600 	struct device *dev;
1601 	int err = 0;
1602 
1603 	dev = ice_pf_to_dev(pf);
1604 
1605 	if (!ice_fltr_add_vlan(vsi, vid, action)) {
1606 		vsi->num_vlan++;
1607 	} else {
1608 		err = -ENODEV;
1609 		dev_err(dev, "Failure Adding VLAN %d on VSI %i\n", vid,
1610 			vsi->vsi_num);
1611 	}
1612 
1613 	return err;
1614 }
1615 
1616 /**
1617  * ice_vsi_kill_vlan - Remove VSI membership for a given VLAN
1618  * @vsi: the VSI being configured
1619  * @vid: VLAN ID to be removed
1620  *
1621  * Returns 0 on success and negative on failure
1622  */
1623 int ice_vsi_kill_vlan(struct ice_vsi *vsi, u16 vid)
1624 {
1625 	struct ice_pf *pf = vsi->back;
1626 	enum ice_status status;
1627 	struct device *dev;
1628 	int err = 0;
1629 
1630 	dev = ice_pf_to_dev(pf);
1631 
1632 	status = ice_fltr_remove_vlan(vsi, vid, ICE_FWD_TO_VSI);
1633 	if (!status) {
1634 		vsi->num_vlan--;
1635 	} else if (status == ICE_ERR_DOES_NOT_EXIST) {
1636 		dev_dbg(dev, "Failed to remove VLAN %d on VSI %i, it does not exist, status: %s\n",
1637 			vid, vsi->vsi_num, ice_stat_str(status));
1638 	} else {
1639 		dev_err(dev, "Error removing VLAN %d on vsi %i error: %s\n",
1640 			vid, vsi->vsi_num, ice_stat_str(status));
1641 		err = -EIO;
1642 	}
1643 
1644 	return err;
1645 }
1646 
1647 /**
1648  * ice_vsi_cfg_frame_size - setup max frame size and Rx buffer length
1649  * @vsi: VSI
1650  */
1651 void ice_vsi_cfg_frame_size(struct ice_vsi *vsi)
1652 {
1653 	if (!vsi->netdev || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) {
1654 		vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1655 		vsi->rx_buf_len = ICE_RXBUF_2048;
1656 #if (PAGE_SIZE < 8192)
1657 	} else if (!ICE_2K_TOO_SMALL_WITH_PADDING &&
1658 		   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
1659 		vsi->max_frame = ICE_RXBUF_1536 - NET_IP_ALIGN;
1660 		vsi->rx_buf_len = ICE_RXBUF_1536 - NET_IP_ALIGN;
1661 #endif
1662 	} else {
1663 		vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1664 #if (PAGE_SIZE < 8192)
1665 		vsi->rx_buf_len = ICE_RXBUF_3072;
1666 #else
1667 		vsi->rx_buf_len = ICE_RXBUF_2048;
1668 #endif
1669 	}
1670 }
1671 
1672 /**
1673  * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1674  * @hw: HW pointer
1675  * @pf_q: index of the Rx queue in the PF's queue space
1676  * @rxdid: flexible descriptor RXDID
1677  * @prio: priority for the RXDID for this queue
1678  */
1679 void
1680 ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio)
1681 {
1682 	int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1683 
1684 	/* clear any previous values */
1685 	regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1686 		    QRXFLXP_CNTXT_RXDID_PRIO_M |
1687 		    QRXFLXP_CNTXT_TS_M);
1688 
1689 	regval |= (rxdid << QRXFLXP_CNTXT_RXDID_IDX_S) &
1690 		QRXFLXP_CNTXT_RXDID_IDX_M;
1691 
1692 	regval |= (prio << QRXFLXP_CNTXT_RXDID_PRIO_S) &
1693 		QRXFLXP_CNTXT_RXDID_PRIO_M;
1694 
1695 	wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1696 }
1697 
1698 int ice_vsi_cfg_single_rxq(struct ice_vsi *vsi, u16 q_idx)
1699 {
1700 	if (q_idx >= vsi->num_rxq)
1701 		return -EINVAL;
1702 
1703 	return ice_vsi_cfg_rxq(vsi->rx_rings[q_idx]);
1704 }
1705 
1706 int ice_vsi_cfg_single_txq(struct ice_vsi *vsi, struct ice_ring **tx_rings, u16 q_idx)
1707 {
1708 	struct ice_aqc_add_tx_qgrp *qg_buf;
1709 	int err;
1710 
1711 	if (q_idx >= vsi->alloc_txq || !tx_rings || !tx_rings[q_idx])
1712 		return -EINVAL;
1713 
1714 	qg_buf = kzalloc(struct_size(qg_buf, txqs, 1), GFP_KERNEL);
1715 	if (!qg_buf)
1716 		return -ENOMEM;
1717 
1718 	qg_buf->num_txqs = 1;
1719 
1720 	err = ice_vsi_cfg_txq(vsi, tx_rings[q_idx], qg_buf);
1721 	kfree(qg_buf);
1722 	return err;
1723 }
1724 
1725 /**
1726  * ice_vsi_cfg_rxqs - Configure the VSI for Rx
1727  * @vsi: the VSI being configured
1728  *
1729  * Return 0 on success and a negative value on error
1730  * Configure the Rx VSI for operation.
1731  */
1732 int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
1733 {
1734 	u16 i;
1735 
1736 	if (vsi->type == ICE_VSI_VF)
1737 		goto setup_rings;
1738 
1739 	ice_vsi_cfg_frame_size(vsi);
1740 setup_rings:
1741 	/* set up individual rings */
1742 	ice_for_each_rxq(vsi, i) {
1743 		int err = ice_vsi_cfg_rxq(vsi->rx_rings[i]);
1744 
1745 		if (err)
1746 			return err;
1747 	}
1748 
1749 	return 0;
1750 }
1751 
1752 /**
1753  * ice_vsi_cfg_txqs - Configure the VSI for Tx
1754  * @vsi: the VSI being configured
1755  * @rings: Tx ring array to be configured
1756  *
1757  * Return 0 on success and a negative value on error
1758  * Configure the Tx VSI for operation.
1759  */
1760 static int
1761 ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_ring **rings)
1762 {
1763 	struct ice_aqc_add_tx_qgrp *qg_buf;
1764 	u16 q_idx = 0;
1765 	int err = 0;
1766 
1767 	qg_buf = kzalloc(struct_size(qg_buf, txqs, 1), GFP_KERNEL);
1768 	if (!qg_buf)
1769 		return -ENOMEM;
1770 
1771 	qg_buf->num_txqs = 1;
1772 
1773 	for (q_idx = 0; q_idx < vsi->num_txq; q_idx++) {
1774 		err = ice_vsi_cfg_txq(vsi, rings[q_idx], qg_buf);
1775 		if (err)
1776 			goto err_cfg_txqs;
1777 	}
1778 
1779 err_cfg_txqs:
1780 	kfree(qg_buf);
1781 	return err;
1782 }
1783 
1784 /**
1785  * ice_vsi_cfg_lan_txqs - Configure the VSI for Tx
1786  * @vsi: the VSI being configured
1787  *
1788  * Return 0 on success and a negative value on error
1789  * Configure the Tx VSI for operation.
1790  */
1791 int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
1792 {
1793 	return ice_vsi_cfg_txqs(vsi, vsi->tx_rings);
1794 }
1795 
1796 /**
1797  * ice_vsi_cfg_xdp_txqs - Configure Tx queues dedicated for XDP in given VSI
1798  * @vsi: the VSI being configured
1799  *
1800  * Return 0 on success and a negative value on error
1801  * Configure the Tx queues dedicated for XDP in given VSI for operation.
1802  */
1803 int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
1804 {
1805 	int ret;
1806 	int i;
1807 
1808 	ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings);
1809 	if (ret)
1810 		return ret;
1811 
1812 	for (i = 0; i < vsi->num_xdp_txq; i++)
1813 		vsi->xdp_rings[i]->xsk_pool = ice_xsk_pool(vsi->xdp_rings[i]);
1814 
1815 	return ret;
1816 }
1817 
1818 /**
1819  * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1820  * @intrl: interrupt rate limit in usecs
1821  * @gran: interrupt rate limit granularity in usecs
1822  *
1823  * This function converts a decimal interrupt rate limit in usecs to the format
1824  * expected by firmware.
1825  */
1826 static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1827 {
1828 	u32 val = intrl / gran;
1829 
1830 	if (val)
1831 		return val | GLINT_RATE_INTRL_ENA_M;
1832 	return 0;
1833 }
1834 
1835 /**
1836  * ice_write_intrl - write throttle rate limit to interrupt specific register
1837  * @q_vector: pointer to interrupt specific structure
1838  * @intrl: throttle rate limit in microseconds to write
1839  */
1840 void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1841 {
1842 	struct ice_hw *hw = &q_vector->vsi->back->hw;
1843 
1844 	wr32(hw, GLINT_RATE(q_vector->reg_idx),
1845 	     ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1846 }
1847 
1848 /**
1849  * __ice_write_itr - write throttle rate to register
1850  * @q_vector: pointer to interrupt data structure
1851  * @rc: pointer to ring container
1852  * @itr: throttle rate in microseconds to write
1853  */
1854 static void __ice_write_itr(struct ice_q_vector *q_vector,
1855 			    struct ice_ring_container *rc, u16 itr)
1856 {
1857 	struct ice_hw *hw = &q_vector->vsi->back->hw;
1858 
1859 	wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
1860 	     ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
1861 }
1862 
1863 /**
1864  * ice_write_itr - write throttle rate to queue specific register
1865  * @rc: pointer to ring container
1866  * @itr: throttle rate in microseconds to write
1867  */
1868 void ice_write_itr(struct ice_ring_container *rc, u16 itr)
1869 {
1870 	struct ice_q_vector *q_vector;
1871 
1872 	if (!rc->ring)
1873 		return;
1874 
1875 	q_vector = rc->ring->q_vector;
1876 
1877 	__ice_write_itr(q_vector, rc, itr);
1878 }
1879 
1880 /**
1881  * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
1882  * @vsi: the VSI being configured
1883  *
1884  * This configures MSIX mode interrupts for the PF VSI, and should not be used
1885  * for the VF VSI.
1886  */
1887 void ice_vsi_cfg_msix(struct ice_vsi *vsi)
1888 {
1889 	struct ice_pf *pf = vsi->back;
1890 	struct ice_hw *hw = &pf->hw;
1891 	u16 txq = 0, rxq = 0;
1892 	int i, q;
1893 
1894 	for (i = 0; i < vsi->num_q_vectors; i++) {
1895 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
1896 		u16 reg_idx = q_vector->reg_idx;
1897 
1898 		ice_cfg_itr(hw, q_vector);
1899 
1900 		/* Both Transmit Queue Interrupt Cause Control register
1901 		 * and Receive Queue Interrupt Cause control register
1902 		 * expects MSIX_INDX field to be the vector index
1903 		 * within the function space and not the absolute
1904 		 * vector index across PF or across device.
1905 		 * For SR-IOV VF VSIs queue vector index always starts
1906 		 * with 1 since first vector index(0) is used for OICR
1907 		 * in VF space. Since VMDq and other PF VSIs are within
1908 		 * the PF function space, use the vector index that is
1909 		 * tracked for this PF.
1910 		 */
1911 		for (q = 0; q < q_vector->num_ring_tx; q++) {
1912 			ice_cfg_txq_interrupt(vsi, txq, reg_idx,
1913 					      q_vector->tx.itr_idx);
1914 			txq++;
1915 		}
1916 
1917 		for (q = 0; q < q_vector->num_ring_rx; q++) {
1918 			ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
1919 					      q_vector->rx.itr_idx);
1920 			rxq++;
1921 		}
1922 	}
1923 }
1924 
1925 /**
1926  * ice_vsi_manage_vlan_insertion - Manage VLAN insertion for the VSI for Tx
1927  * @vsi: the VSI being changed
1928  */
1929 int ice_vsi_manage_vlan_insertion(struct ice_vsi *vsi)
1930 {
1931 	struct ice_hw *hw = &vsi->back->hw;
1932 	struct ice_vsi_ctx *ctxt;
1933 	enum ice_status status;
1934 	int ret = 0;
1935 
1936 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1937 	if (!ctxt)
1938 		return -ENOMEM;
1939 
1940 	/* Here we are configuring the VSI to let the driver add VLAN tags by
1941 	 * setting vlan_flags to ICE_AQ_VSI_VLAN_MODE_ALL. The actual VLAN tag
1942 	 * insertion happens in the Tx hot path, in ice_tx_map.
1943 	 */
1944 	ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL;
1945 
1946 	/* Preserve existing VLAN strip setting */
1947 	ctxt->info.vlan_flags |= (vsi->info.vlan_flags &
1948 				  ICE_AQ_VSI_VLAN_EMOD_M);
1949 
1950 	ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID);
1951 
1952 	status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1953 	if (status) {
1954 		dev_err(ice_pf_to_dev(vsi->back), "update VSI for VLAN insert failed, err %s aq_err %s\n",
1955 			ice_stat_str(status),
1956 			ice_aq_str(hw->adminq.sq_last_status));
1957 		ret = -EIO;
1958 		goto out;
1959 	}
1960 
1961 	vsi->info.vlan_flags = ctxt->info.vlan_flags;
1962 out:
1963 	kfree(ctxt);
1964 	return ret;
1965 }
1966 
1967 /**
1968  * ice_vsi_manage_vlan_stripping - Manage VLAN stripping for the VSI for Rx
1969  * @vsi: the VSI being changed
1970  * @ena: boolean value indicating if this is a enable or disable request
1971  */
1972 int ice_vsi_manage_vlan_stripping(struct ice_vsi *vsi, bool ena)
1973 {
1974 	struct ice_hw *hw = &vsi->back->hw;
1975 	struct ice_vsi_ctx *ctxt;
1976 	enum ice_status status;
1977 	int ret = 0;
1978 
1979 	/* do not allow modifying VLAN stripping when a port VLAN is configured
1980 	 * on this VSI
1981 	 */
1982 	if (vsi->info.pvid)
1983 		return 0;
1984 
1985 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1986 	if (!ctxt)
1987 		return -ENOMEM;
1988 
1989 	/* Here we are configuring what the VSI should do with the VLAN tag in
1990 	 * the Rx packet. We can either leave the tag in the packet or put it in
1991 	 * the Rx descriptor.
1992 	 */
1993 	if (ena)
1994 		/* Strip VLAN tag from Rx packet and put it in the desc */
1995 		ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_EMOD_STR_BOTH;
1996 	else
1997 		/* Disable stripping. Leave tag in packet */
1998 		ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_EMOD_NOTHING;
1999 
2000 	/* Allow all packets untagged/tagged */
2001 	ctxt->info.vlan_flags |= ICE_AQ_VSI_VLAN_MODE_ALL;
2002 
2003 	ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID);
2004 
2005 	status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
2006 	if (status) {
2007 		dev_err(ice_pf_to_dev(vsi->back), "update VSI for VLAN strip failed, ena = %d err %s aq_err %s\n",
2008 			ena, ice_stat_str(status),
2009 			ice_aq_str(hw->adminq.sq_last_status));
2010 		ret = -EIO;
2011 		goto out;
2012 	}
2013 
2014 	vsi->info.vlan_flags = ctxt->info.vlan_flags;
2015 out:
2016 	kfree(ctxt);
2017 	return ret;
2018 }
2019 
2020 /**
2021  * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
2022  * @vsi: the VSI whose rings are to be enabled
2023  *
2024  * Returns 0 on success and a negative value on error
2025  */
2026 int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
2027 {
2028 	return ice_vsi_ctrl_all_rx_rings(vsi, true);
2029 }
2030 
2031 /**
2032  * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
2033  * @vsi: the VSI whose rings are to be disabled
2034  *
2035  * Returns 0 on success and a negative value on error
2036  */
2037 int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
2038 {
2039 	return ice_vsi_ctrl_all_rx_rings(vsi, false);
2040 }
2041 
2042 /**
2043  * ice_vsi_stop_tx_rings - Disable Tx rings
2044  * @vsi: the VSI being configured
2045  * @rst_src: reset source
2046  * @rel_vmvf_num: Relative ID of VF/VM
2047  * @rings: Tx ring array to be stopped
2048  */
2049 static int
2050 ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2051 		      u16 rel_vmvf_num, struct ice_ring **rings)
2052 {
2053 	u16 q_idx;
2054 
2055 	if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2056 		return -EINVAL;
2057 
2058 	for (q_idx = 0; q_idx < vsi->num_txq; q_idx++) {
2059 		struct ice_txq_meta txq_meta = { };
2060 		int status;
2061 
2062 		if (!rings || !rings[q_idx])
2063 			return -EINVAL;
2064 
2065 		ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
2066 		status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2067 					      rings[q_idx], &txq_meta);
2068 
2069 		if (status)
2070 			return status;
2071 	}
2072 
2073 	return 0;
2074 }
2075 
2076 /**
2077  * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2078  * @vsi: the VSI being configured
2079  * @rst_src: reset source
2080  * @rel_vmvf_num: Relative ID of VF/VM
2081  */
2082 int
2083 ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2084 			  u16 rel_vmvf_num)
2085 {
2086 	return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings);
2087 }
2088 
2089 /**
2090  * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2091  * @vsi: the VSI being configured
2092  */
2093 int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2094 {
2095 	return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings);
2096 }
2097 
2098 /**
2099  * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
2100  * @vsi: VSI to check whether or not VLAN pruning is enabled.
2101  *
2102  * returns true if Rx VLAN pruning is enabled and false otherwise.
2103  */
2104 bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
2105 {
2106 	if (!vsi)
2107 		return false;
2108 
2109 	return (vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA);
2110 }
2111 
2112 /**
2113  * ice_cfg_vlan_pruning - enable or disable VLAN pruning on the VSI
2114  * @vsi: VSI to enable or disable VLAN pruning on
2115  * @ena: set to true to enable VLAN pruning and false to disable it
2116  * @vlan_promisc: enable valid security flags if not in VLAN promiscuous mode
2117  *
2118  * returns 0 if VSI is updated, negative otherwise
2119  */
2120 int ice_cfg_vlan_pruning(struct ice_vsi *vsi, bool ena, bool vlan_promisc)
2121 {
2122 	struct ice_vsi_ctx *ctxt;
2123 	struct ice_pf *pf;
2124 	int status;
2125 
2126 	if (!vsi)
2127 		return -EINVAL;
2128 
2129 	/* Don't enable VLAN pruning if the netdev is currently in promiscuous
2130 	 * mode. VLAN pruning will be enabled when the interface exits
2131 	 * promiscuous mode if any VLAN filters are active.
2132 	 */
2133 	if (vsi->netdev && vsi->netdev->flags & IFF_PROMISC && ena)
2134 		return 0;
2135 
2136 	pf = vsi->back;
2137 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
2138 	if (!ctxt)
2139 		return -ENOMEM;
2140 
2141 	ctxt->info = vsi->info;
2142 
2143 	if (ena)
2144 		ctxt->info.sw_flags2 |= ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
2145 	else
2146 		ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
2147 
2148 	if (!vlan_promisc)
2149 		ctxt->info.valid_sections =
2150 			cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
2151 
2152 	status = ice_update_vsi(&pf->hw, vsi->idx, ctxt, NULL);
2153 	if (status) {
2154 		netdev_err(vsi->netdev, "%sabling VLAN pruning on VSI handle: %d, VSI HW ID: %d failed, err = %s, aq_err = %s\n",
2155 			   ena ? "En" : "Dis", vsi->idx, vsi->vsi_num,
2156 			   ice_stat_str(status),
2157 			   ice_aq_str(pf->hw.adminq.sq_last_status));
2158 		goto err_out;
2159 	}
2160 
2161 	vsi->info.sw_flags2 = ctxt->info.sw_flags2;
2162 
2163 	kfree(ctxt);
2164 	return 0;
2165 
2166 err_out:
2167 	kfree(ctxt);
2168 	return -EIO;
2169 }
2170 
2171 static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2172 {
2173 	struct ice_dcbx_cfg *cfg = &vsi->port_info->qos_cfg.local_dcbx_cfg;
2174 
2175 	vsi->tc_cfg.ena_tc = ice_dcb_get_ena_tc(cfg);
2176 	vsi->tc_cfg.numtc = ice_dcb_get_num_tc(cfg);
2177 }
2178 
2179 /**
2180  * ice_vsi_set_q_vectors_reg_idx - set the HW register index for all q_vectors
2181  * @vsi: VSI to set the q_vectors register index on
2182  */
2183 static int
2184 ice_vsi_set_q_vectors_reg_idx(struct ice_vsi *vsi)
2185 {
2186 	u16 i;
2187 
2188 	if (!vsi || !vsi->q_vectors)
2189 		return -EINVAL;
2190 
2191 	ice_for_each_q_vector(vsi, i) {
2192 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2193 
2194 		if (!q_vector) {
2195 			dev_err(ice_pf_to_dev(vsi->back), "Failed to set reg_idx on q_vector %d VSI %d\n",
2196 				i, vsi->vsi_num);
2197 			goto clear_reg_idx;
2198 		}
2199 
2200 		if (vsi->type == ICE_VSI_VF) {
2201 			struct ice_vf *vf = &vsi->back->vf[vsi->vf_id];
2202 
2203 			q_vector->reg_idx = ice_calc_vf_reg_idx(vf, q_vector);
2204 		} else {
2205 			q_vector->reg_idx =
2206 				q_vector->v_idx + vsi->base_vector;
2207 		}
2208 	}
2209 
2210 	return 0;
2211 
2212 clear_reg_idx:
2213 	ice_for_each_q_vector(vsi, i) {
2214 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2215 
2216 		if (q_vector)
2217 			q_vector->reg_idx = 0;
2218 	}
2219 
2220 	return -EINVAL;
2221 }
2222 
2223 /**
2224  * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2225  * @vsi: the VSI being configured
2226  * @tx: bool to determine Tx or Rx rule
2227  * @create: bool to determine create or remove Rule
2228  */
2229 void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2230 {
2231 	enum ice_status (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2232 				    enum ice_sw_fwd_act_type act);
2233 	struct ice_pf *pf = vsi->back;
2234 	enum ice_status status;
2235 	struct device *dev;
2236 
2237 	dev = ice_pf_to_dev(pf);
2238 	eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2239 
2240 	if (tx) {
2241 		status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2242 				  ICE_DROP_PACKET);
2243 	} else {
2244 		if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2245 			status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2246 							  create);
2247 		} else {
2248 			status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2249 					  ICE_FWD_TO_VSI);
2250 		}
2251 	}
2252 
2253 	if (status)
2254 		dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %s\n",
2255 			create ? "adding" : "removing", tx ? "TX" : "RX",
2256 			vsi->vsi_num, ice_stat_str(status));
2257 }
2258 
2259 /**
2260  * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2261  * @vsi: pointer to the VSI
2262  *
2263  * This function will allocate new scheduler aggregator now if needed and will
2264  * move specified VSI into it.
2265  */
2266 static void ice_set_agg_vsi(struct ice_vsi *vsi)
2267 {
2268 	struct device *dev = ice_pf_to_dev(vsi->back);
2269 	struct ice_agg_node *agg_node_iter = NULL;
2270 	u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2271 	struct ice_agg_node *agg_node = NULL;
2272 	int node_offset, max_agg_nodes = 0;
2273 	struct ice_port_info *port_info;
2274 	struct ice_pf *pf = vsi->back;
2275 	u32 agg_node_id_start = 0;
2276 	enum ice_status status;
2277 
2278 	/* create (as needed) scheduler aggregator node and move VSI into
2279 	 * corresponding aggregator node
2280 	 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2281 	 * - VF aggregator nodes will contain VF VSI
2282 	 */
2283 	port_info = pf->hw.port_info;
2284 	if (!port_info)
2285 		return;
2286 
2287 	switch (vsi->type) {
2288 	case ICE_VSI_CTRL:
2289 	case ICE_VSI_LB:
2290 	case ICE_VSI_PF:
2291 		max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2292 		agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2293 		agg_node_iter = &pf->pf_agg_node[0];
2294 		break;
2295 	case ICE_VSI_VF:
2296 		/* user can create 'n' VFs on a given PF, but since max children
2297 		 * per aggregator node can be only 64. Following code handles
2298 		 * aggregator(s) for VF VSIs, either selects a agg_node which
2299 		 * was already created provided num_vsis < 64, otherwise
2300 		 * select next available node, which will be created
2301 		 */
2302 		max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2303 		agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2304 		agg_node_iter = &pf->vf_agg_node[0];
2305 		break;
2306 	default:
2307 		/* other VSI type, handle later if needed */
2308 		dev_dbg(dev, "unexpected VSI type %s\n",
2309 			ice_vsi_type_str(vsi->type));
2310 		return;
2311 	}
2312 
2313 	/* find the appropriate aggregator node */
2314 	for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2315 		/* see if we can find space in previously created
2316 		 * node if num_vsis < 64, otherwise skip
2317 		 */
2318 		if (agg_node_iter->num_vsis &&
2319 		    agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2320 			agg_node_iter++;
2321 			continue;
2322 		}
2323 
2324 		if (agg_node_iter->valid &&
2325 		    agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2326 			agg_id = agg_node_iter->agg_id;
2327 			agg_node = agg_node_iter;
2328 			break;
2329 		}
2330 
2331 		/* find unclaimed agg_id */
2332 		if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2333 			agg_id = node_offset + agg_node_id_start;
2334 			agg_node = agg_node_iter;
2335 			break;
2336 		}
2337 		/* move to next agg_node */
2338 		agg_node_iter++;
2339 	}
2340 
2341 	if (!agg_node)
2342 		return;
2343 
2344 	/* if selected aggregator node was not created, create it */
2345 	if (!agg_node->valid) {
2346 		status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2347 				     (u8)vsi->tc_cfg.ena_tc);
2348 		if (status) {
2349 			dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2350 				agg_id);
2351 			return;
2352 		}
2353 		/* aggregator node is created, store the neeeded info */
2354 		agg_node->valid = true;
2355 		agg_node->agg_id = agg_id;
2356 	}
2357 
2358 	/* move VSI to corresponding aggregator node */
2359 	status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2360 				     (u8)vsi->tc_cfg.ena_tc);
2361 	if (status) {
2362 		dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2363 			vsi->idx, agg_id);
2364 		return;
2365 	}
2366 
2367 	/* keep active children count for aggregator node */
2368 	agg_node->num_vsis++;
2369 
2370 	/* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2371 	 * to aggregator node
2372 	 */
2373 	vsi->agg_node = agg_node;
2374 	dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2375 		vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2376 		vsi->agg_node->num_vsis);
2377 }
2378 
2379 /**
2380  * ice_vsi_setup - Set up a VSI by a given type
2381  * @pf: board private structure
2382  * @pi: pointer to the port_info instance
2383  * @vsi_type: VSI type
2384  * @vf_id: defines VF ID to which this VSI connects. This field is meant to be
2385  *         used only for ICE_VSI_VF VSI type. For other VSI types, should
2386  *         fill-in ICE_INVAL_VFID as input.
2387  *
2388  * This allocates the sw VSI structure and its queue resources.
2389  *
2390  * Returns pointer to the successfully allocated and configured VSI sw struct on
2391  * success, NULL on failure.
2392  */
2393 struct ice_vsi *
2394 ice_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
2395 	      enum ice_vsi_type vsi_type, u16 vf_id)
2396 {
2397 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2398 	struct device *dev = ice_pf_to_dev(pf);
2399 	enum ice_status status;
2400 	struct ice_vsi *vsi;
2401 	int ret, i;
2402 
2403 	if (vsi_type == ICE_VSI_VF || vsi_type == ICE_VSI_CTRL)
2404 		vsi = ice_vsi_alloc(pf, vsi_type, vf_id);
2405 	else
2406 		vsi = ice_vsi_alloc(pf, vsi_type, ICE_INVAL_VFID);
2407 
2408 	if (!vsi) {
2409 		dev_err(dev, "could not allocate VSI\n");
2410 		return NULL;
2411 	}
2412 
2413 	vsi->port_info = pi;
2414 	vsi->vsw = pf->first_sw;
2415 	if (vsi->type == ICE_VSI_PF)
2416 		vsi->ethtype = ETH_P_PAUSE;
2417 
2418 	if (vsi->type == ICE_VSI_VF || vsi->type == ICE_VSI_CTRL)
2419 		vsi->vf_id = vf_id;
2420 
2421 	ice_alloc_fd_res(vsi);
2422 
2423 	if (ice_vsi_get_qs(vsi)) {
2424 		dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2425 			vsi->idx);
2426 		goto unroll_vsi_alloc;
2427 	}
2428 
2429 	/* set RSS capabilities */
2430 	ice_vsi_set_rss_params(vsi);
2431 
2432 	/* set TC configuration */
2433 	ice_vsi_set_tc_cfg(vsi);
2434 
2435 	/* create the VSI */
2436 	ret = ice_vsi_init(vsi, true);
2437 	if (ret)
2438 		goto unroll_get_qs;
2439 
2440 	switch (vsi->type) {
2441 	case ICE_VSI_CTRL:
2442 	case ICE_VSI_PF:
2443 		ret = ice_vsi_alloc_q_vectors(vsi);
2444 		if (ret)
2445 			goto unroll_vsi_init;
2446 
2447 		ret = ice_vsi_setup_vector_base(vsi);
2448 		if (ret)
2449 			goto unroll_alloc_q_vector;
2450 
2451 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
2452 		if (ret)
2453 			goto unroll_vector_base;
2454 
2455 		ret = ice_vsi_alloc_rings(vsi);
2456 		if (ret)
2457 			goto unroll_vector_base;
2458 
2459 		/* Always add VLAN ID 0 switch rule by default. This is needed
2460 		 * in order to allow all untagged and 0 tagged priority traffic
2461 		 * if Rx VLAN pruning is enabled. Also there are cases where we
2462 		 * don't get the call to add VLAN 0 via ice_vlan_rx_add_vid()
2463 		 * so this handles those cases (i.e. adding the PF to a bridge
2464 		 * without the 8021q module loaded).
2465 		 */
2466 		ret = ice_vsi_add_vlan(vsi, 0, ICE_FWD_TO_VSI);
2467 		if (ret)
2468 			goto unroll_clear_rings;
2469 
2470 		ice_vsi_map_rings_to_vectors(vsi);
2471 
2472 		/* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2473 		if (vsi->type != ICE_VSI_CTRL)
2474 			/* Do not exit if configuring RSS had an issue, at
2475 			 * least receive traffic on first queue. Hence no
2476 			 * need to capture return value
2477 			 */
2478 			if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2479 				ice_vsi_cfg_rss_lut_key(vsi);
2480 				ice_vsi_set_rss_flow_fld(vsi);
2481 			}
2482 		ice_init_arfs(vsi);
2483 		break;
2484 	case ICE_VSI_VF:
2485 		/* VF driver will take care of creating netdev for this type and
2486 		 * map queues to vectors through Virtchnl, PF driver only
2487 		 * creates a VSI and corresponding structures for bookkeeping
2488 		 * purpose
2489 		 */
2490 		ret = ice_vsi_alloc_q_vectors(vsi);
2491 		if (ret)
2492 			goto unroll_vsi_init;
2493 
2494 		ret = ice_vsi_alloc_rings(vsi);
2495 		if (ret)
2496 			goto unroll_alloc_q_vector;
2497 
2498 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
2499 		if (ret)
2500 			goto unroll_vector_base;
2501 
2502 		/* Do not exit if configuring RSS had an issue, at least
2503 		 * receive traffic on first queue. Hence no need to capture
2504 		 * return value
2505 		 */
2506 		if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2507 			ice_vsi_cfg_rss_lut_key(vsi);
2508 			ice_vsi_set_vf_rss_flow_fld(vsi);
2509 		}
2510 		break;
2511 	case ICE_VSI_LB:
2512 		ret = ice_vsi_alloc_rings(vsi);
2513 		if (ret)
2514 			goto unroll_vsi_init;
2515 		break;
2516 	default:
2517 		/* clean up the resources and exit */
2518 		goto unroll_vsi_init;
2519 	}
2520 
2521 	/* configure VSI nodes based on number of queues and TC's */
2522 	for (i = 0; i < vsi->tc_cfg.numtc; i++)
2523 		max_txqs[i] = vsi->alloc_txq;
2524 
2525 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2526 				 max_txqs);
2527 	if (status) {
2528 		dev_err(dev, "VSI %d failed lan queue config, error %s\n",
2529 			vsi->vsi_num, ice_stat_str(status));
2530 		goto unroll_clear_rings;
2531 	}
2532 
2533 	/* Add switch rule to drop all Tx Flow Control Frames, of look up
2534 	 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2535 	 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2536 	 * The rule is added once for PF VSI in order to create appropriate
2537 	 * recipe, since VSI/VSI list is ignored with drop action...
2538 	 * Also add rules to handle LLDP Tx packets.  Tx LLDP packets need to
2539 	 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2540 	 * settings in the HW.
2541 	 */
2542 	if (!ice_is_safe_mode(pf))
2543 		if (vsi->type == ICE_VSI_PF) {
2544 			ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2545 					 ICE_DROP_PACKET);
2546 			ice_cfg_sw_lldp(vsi, true, true);
2547 		}
2548 
2549 	if (!vsi->agg_node)
2550 		ice_set_agg_vsi(vsi);
2551 	return vsi;
2552 
2553 unroll_clear_rings:
2554 	ice_vsi_clear_rings(vsi);
2555 unroll_vector_base:
2556 	/* reclaim SW interrupts back to the common pool */
2557 	ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
2558 	pf->num_avail_sw_msix += vsi->num_q_vectors;
2559 unroll_alloc_q_vector:
2560 	ice_vsi_free_q_vectors(vsi);
2561 unroll_vsi_init:
2562 	ice_vsi_delete(vsi);
2563 unroll_get_qs:
2564 	ice_vsi_put_qs(vsi);
2565 unroll_vsi_alloc:
2566 	if (vsi_type == ICE_VSI_VF)
2567 		ice_enable_lag(pf->lag);
2568 	ice_vsi_clear(vsi);
2569 
2570 	return NULL;
2571 }
2572 
2573 /**
2574  * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2575  * @vsi: the VSI being cleaned up
2576  */
2577 static void ice_vsi_release_msix(struct ice_vsi *vsi)
2578 {
2579 	struct ice_pf *pf = vsi->back;
2580 	struct ice_hw *hw = &pf->hw;
2581 	u32 txq = 0;
2582 	u32 rxq = 0;
2583 	int i, q;
2584 
2585 	for (i = 0; i < vsi->num_q_vectors; i++) {
2586 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2587 
2588 		ice_write_intrl(q_vector, 0);
2589 		for (q = 0; q < q_vector->num_ring_tx; q++) {
2590 			ice_write_itr(&q_vector->tx, 0);
2591 			wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2592 			if (ice_is_xdp_ena_vsi(vsi)) {
2593 				u32 xdp_txq = txq + vsi->num_xdp_txq;
2594 
2595 				wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2596 			}
2597 			txq++;
2598 		}
2599 
2600 		for (q = 0; q < q_vector->num_ring_rx; q++) {
2601 			ice_write_itr(&q_vector->rx, 0);
2602 			wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2603 			rxq++;
2604 		}
2605 	}
2606 
2607 	ice_flush(hw);
2608 }
2609 
2610 /**
2611  * ice_vsi_free_irq - Free the IRQ association with the OS
2612  * @vsi: the VSI being configured
2613  */
2614 void ice_vsi_free_irq(struct ice_vsi *vsi)
2615 {
2616 	struct ice_pf *pf = vsi->back;
2617 	int base = vsi->base_vector;
2618 	int i;
2619 
2620 	if (!vsi->q_vectors || !vsi->irqs_ready)
2621 		return;
2622 
2623 	ice_vsi_release_msix(vsi);
2624 	if (vsi->type == ICE_VSI_VF)
2625 		return;
2626 
2627 	vsi->irqs_ready = false;
2628 	ice_for_each_q_vector(vsi, i) {
2629 		u16 vector = i + base;
2630 		int irq_num;
2631 
2632 		irq_num = pf->msix_entries[vector].vector;
2633 
2634 		/* free only the irqs that were actually requested */
2635 		if (!vsi->q_vectors[i] ||
2636 		    !(vsi->q_vectors[i]->num_ring_tx ||
2637 		      vsi->q_vectors[i]->num_ring_rx))
2638 			continue;
2639 
2640 		/* clear the affinity notifier in the IRQ descriptor */
2641 		irq_set_affinity_notifier(irq_num, NULL);
2642 
2643 		/* clear the affinity_mask in the IRQ descriptor */
2644 		irq_set_affinity_hint(irq_num, NULL);
2645 		synchronize_irq(irq_num);
2646 		devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2647 	}
2648 }
2649 
2650 /**
2651  * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2652  * @vsi: the VSI having resources freed
2653  */
2654 void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2655 {
2656 	int i;
2657 
2658 	if (!vsi->tx_rings)
2659 		return;
2660 
2661 	ice_for_each_txq(vsi, i)
2662 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2663 			ice_free_tx_ring(vsi->tx_rings[i]);
2664 }
2665 
2666 /**
2667  * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2668  * @vsi: the VSI having resources freed
2669  */
2670 void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2671 {
2672 	int i;
2673 
2674 	if (!vsi->rx_rings)
2675 		return;
2676 
2677 	ice_for_each_rxq(vsi, i)
2678 		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2679 			ice_free_rx_ring(vsi->rx_rings[i]);
2680 }
2681 
2682 /**
2683  * ice_vsi_close - Shut down a VSI
2684  * @vsi: the VSI being shut down
2685  */
2686 void ice_vsi_close(struct ice_vsi *vsi)
2687 {
2688 	if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2689 		ice_down(vsi);
2690 
2691 	ice_vsi_free_irq(vsi);
2692 	ice_vsi_free_tx_rings(vsi);
2693 	ice_vsi_free_rx_rings(vsi);
2694 }
2695 
2696 /**
2697  * ice_ena_vsi - resume a VSI
2698  * @vsi: the VSI being resume
2699  * @locked: is the rtnl_lock already held
2700  */
2701 int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2702 {
2703 	int err = 0;
2704 
2705 	if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2706 		return 0;
2707 
2708 	clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2709 
2710 	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2711 		if (netif_running(vsi->netdev)) {
2712 			if (!locked)
2713 				rtnl_lock();
2714 
2715 			err = ice_open_internal(vsi->netdev);
2716 
2717 			if (!locked)
2718 				rtnl_unlock();
2719 		}
2720 	} else if (vsi->type == ICE_VSI_CTRL) {
2721 		err = ice_vsi_open_ctrl(vsi);
2722 	}
2723 
2724 	return err;
2725 }
2726 
2727 /**
2728  * ice_dis_vsi - pause a VSI
2729  * @vsi: the VSI being paused
2730  * @locked: is the rtnl_lock already held
2731  */
2732 void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2733 {
2734 	if (test_bit(ICE_VSI_DOWN, vsi->state))
2735 		return;
2736 
2737 	set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2738 
2739 	if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2740 		if (netif_running(vsi->netdev)) {
2741 			if (!locked)
2742 				rtnl_lock();
2743 
2744 			ice_vsi_close(vsi);
2745 
2746 			if (!locked)
2747 				rtnl_unlock();
2748 		} else {
2749 			ice_vsi_close(vsi);
2750 		}
2751 	} else if (vsi->type == ICE_VSI_CTRL) {
2752 		ice_vsi_close(vsi);
2753 	}
2754 }
2755 
2756 /**
2757  * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
2758  * @vsi: the VSI being un-configured
2759  */
2760 void ice_vsi_dis_irq(struct ice_vsi *vsi)
2761 {
2762 	int base = vsi->base_vector;
2763 	struct ice_pf *pf = vsi->back;
2764 	struct ice_hw *hw = &pf->hw;
2765 	u32 val;
2766 	int i;
2767 
2768 	/* disable interrupt causation from each queue */
2769 	if (vsi->tx_rings) {
2770 		ice_for_each_txq(vsi, i) {
2771 			if (vsi->tx_rings[i]) {
2772 				u16 reg;
2773 
2774 				reg = vsi->tx_rings[i]->reg_idx;
2775 				val = rd32(hw, QINT_TQCTL(reg));
2776 				val &= ~QINT_TQCTL_CAUSE_ENA_M;
2777 				wr32(hw, QINT_TQCTL(reg), val);
2778 			}
2779 		}
2780 	}
2781 
2782 	if (vsi->rx_rings) {
2783 		ice_for_each_rxq(vsi, i) {
2784 			if (vsi->rx_rings[i]) {
2785 				u16 reg;
2786 
2787 				reg = vsi->rx_rings[i]->reg_idx;
2788 				val = rd32(hw, QINT_RQCTL(reg));
2789 				val &= ~QINT_RQCTL_CAUSE_ENA_M;
2790 				wr32(hw, QINT_RQCTL(reg), val);
2791 			}
2792 		}
2793 	}
2794 
2795 	/* disable each interrupt */
2796 	ice_for_each_q_vector(vsi, i) {
2797 		if (!vsi->q_vectors[i])
2798 			continue;
2799 		wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
2800 	}
2801 
2802 	ice_flush(hw);
2803 
2804 	/* don't call synchronize_irq() for VF's from the host */
2805 	if (vsi->type == ICE_VSI_VF)
2806 		return;
2807 
2808 	ice_for_each_q_vector(vsi, i)
2809 		synchronize_irq(pf->msix_entries[i + base].vector);
2810 }
2811 
2812 /**
2813  * ice_napi_del - Remove NAPI handler for the VSI
2814  * @vsi: VSI for which NAPI handler is to be removed
2815  */
2816 void ice_napi_del(struct ice_vsi *vsi)
2817 {
2818 	int v_idx;
2819 
2820 	if (!vsi->netdev)
2821 		return;
2822 
2823 	ice_for_each_q_vector(vsi, v_idx)
2824 		netif_napi_del(&vsi->q_vectors[v_idx]->napi);
2825 }
2826 
2827 /**
2828  * ice_vsi_release - Delete a VSI and free its resources
2829  * @vsi: the VSI being removed
2830  *
2831  * Returns 0 on success or < 0 on error
2832  */
2833 int ice_vsi_release(struct ice_vsi *vsi)
2834 {
2835 	struct ice_pf *pf;
2836 
2837 	if (!vsi->back)
2838 		return -ENODEV;
2839 	pf = vsi->back;
2840 
2841 	/* do not unregister while driver is in the reset recovery pending
2842 	 * state. Since reset/rebuild happens through PF service task workqueue,
2843 	 * it's not a good idea to unregister netdev that is associated to the
2844 	 * PF that is running the work queue items currently. This is done to
2845 	 * avoid check_flush_dependency() warning on this wq
2846 	 */
2847 	if (vsi->netdev && !ice_is_reset_in_progress(pf->state) &&
2848 	    (test_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state))) {
2849 		unregister_netdev(vsi->netdev);
2850 		clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
2851 	}
2852 
2853 	ice_devlink_destroy_port(vsi);
2854 
2855 	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2856 		ice_rss_clean(vsi);
2857 
2858 	/* Disable VSI and free resources */
2859 	if (vsi->type != ICE_VSI_LB)
2860 		ice_vsi_dis_irq(vsi);
2861 	ice_vsi_close(vsi);
2862 
2863 	/* SR-IOV determines needed MSIX resources all at once instead of per
2864 	 * VSI since when VFs are spawned we know how many VFs there are and how
2865 	 * many interrupts each VF needs. SR-IOV MSIX resources are also
2866 	 * cleared in the same manner.
2867 	 */
2868 	if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID) {
2869 		struct ice_vf *vf;
2870 		int i;
2871 
2872 		ice_for_each_vf(pf, i) {
2873 			vf = &pf->vf[i];
2874 			if (i != vsi->vf_id && vf->ctrl_vsi_idx != ICE_NO_VSI)
2875 				break;
2876 		}
2877 		if (i == pf->num_alloc_vfs) {
2878 			/* No other VFs left that have control VSI, reclaim SW
2879 			 * interrupts back to the common pool
2880 			 */
2881 			ice_free_res(pf->irq_tracker, vsi->base_vector,
2882 				     ICE_RES_VF_CTRL_VEC_ID);
2883 			pf->num_avail_sw_msix += vsi->num_q_vectors;
2884 		}
2885 	} else if (vsi->type != ICE_VSI_VF) {
2886 		/* reclaim SW interrupts back to the common pool */
2887 		ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
2888 		pf->num_avail_sw_msix += vsi->num_q_vectors;
2889 	}
2890 
2891 	if (!ice_is_safe_mode(pf)) {
2892 		if (vsi->type == ICE_VSI_PF) {
2893 			ice_fltr_remove_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2894 					    ICE_DROP_PACKET);
2895 			ice_cfg_sw_lldp(vsi, true, false);
2896 			/* The Rx rule will only exist to remove if the LLDP FW
2897 			 * engine is currently stopped
2898 			 */
2899 			if (!test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2900 				ice_cfg_sw_lldp(vsi, false, false);
2901 		}
2902 	}
2903 
2904 	ice_fltr_remove_all(vsi);
2905 	ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
2906 	ice_vsi_delete(vsi);
2907 	ice_vsi_free_q_vectors(vsi);
2908 
2909 	if (vsi->netdev) {
2910 		if (test_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state)) {
2911 			unregister_netdev(vsi->netdev);
2912 			clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
2913 		}
2914 		if (test_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state)) {
2915 			free_netdev(vsi->netdev);
2916 			vsi->netdev = NULL;
2917 			clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
2918 		}
2919 	}
2920 
2921 	if (vsi->type == ICE_VSI_VF &&
2922 	    vsi->agg_node && vsi->agg_node->valid)
2923 		vsi->agg_node->num_vsis--;
2924 	ice_vsi_clear_rings(vsi);
2925 
2926 	ice_vsi_put_qs(vsi);
2927 
2928 	/* retain SW VSI data structure since it is needed to unregister and
2929 	 * free VSI netdev when PF is not in reset recovery pending state,\
2930 	 * for ex: during rmmod.
2931 	 */
2932 	if (!ice_is_reset_in_progress(pf->state))
2933 		ice_vsi_clear(vsi);
2934 
2935 	return 0;
2936 }
2937 
2938 /**
2939  * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
2940  * @vsi: VSI connected with q_vectors
2941  * @coalesce: array of struct with stored coalesce
2942  *
2943  * Returns array size.
2944  */
2945 static int
2946 ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
2947 			     struct ice_coalesce_stored *coalesce)
2948 {
2949 	int i;
2950 
2951 	ice_for_each_q_vector(vsi, i) {
2952 		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2953 
2954 		coalesce[i].itr_tx = q_vector->tx.itr_setting;
2955 		coalesce[i].itr_rx = q_vector->rx.itr_setting;
2956 		coalesce[i].intrl = q_vector->intrl;
2957 
2958 		if (i < vsi->num_txq)
2959 			coalesce[i].tx_valid = true;
2960 		if (i < vsi->num_rxq)
2961 			coalesce[i].rx_valid = true;
2962 	}
2963 
2964 	return vsi->num_q_vectors;
2965 }
2966 
2967 /**
2968  * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
2969  * @vsi: VSI connected with q_vectors
2970  * @coalesce: pointer to array of struct with stored coalesce
2971  * @size: size of coalesce array
2972  *
2973  * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
2974  * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
2975  * to default value.
2976  */
2977 static void
2978 ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
2979 			     struct ice_coalesce_stored *coalesce, int size)
2980 {
2981 	struct ice_ring_container *rc;
2982 	int i;
2983 
2984 	if ((size && !coalesce) || !vsi)
2985 		return;
2986 
2987 	/* There are a couple of cases that have to be handled here:
2988 	 *   1. The case where the number of queue vectors stays the same, but
2989 	 *      the number of Tx or Rx rings changes (the first for loop)
2990 	 *   2. The case where the number of queue vectors increased (the
2991 	 *      second for loop)
2992 	 */
2993 	for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
2994 		/* There are 2 cases to handle here and they are the same for
2995 		 * both Tx and Rx:
2996 		 *   if the entry was valid previously (coalesce[i].[tr]x_valid
2997 		 *   and the loop variable is less than the number of rings
2998 		 *   allocated, then write the previous values
2999 		 *
3000 		 *   if the entry was not valid previously, but the number of
3001 		 *   rings is less than are allocated (this means the number of
3002 		 *   rings increased from previously), then write out the
3003 		 *   values in the first element
3004 		 *
3005 		 *   Also, always write the ITR, even if in ITR_IS_DYNAMIC
3006 		 *   as there is no harm because the dynamic algorithm
3007 		 *   will just overwrite.
3008 		 */
3009 		if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
3010 			rc = &vsi->q_vectors[i]->rx;
3011 			rc->itr_setting = coalesce[i].itr_rx;
3012 			ice_write_itr(rc, rc->itr_setting);
3013 		} else if (i < vsi->alloc_rxq) {
3014 			rc = &vsi->q_vectors[i]->rx;
3015 			rc->itr_setting = coalesce[0].itr_rx;
3016 			ice_write_itr(rc, rc->itr_setting);
3017 		}
3018 
3019 		if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
3020 			rc = &vsi->q_vectors[i]->tx;
3021 			rc->itr_setting = coalesce[i].itr_tx;
3022 			ice_write_itr(rc, rc->itr_setting);
3023 		} else if (i < vsi->alloc_txq) {
3024 			rc = &vsi->q_vectors[i]->tx;
3025 			rc->itr_setting = coalesce[0].itr_tx;
3026 			ice_write_itr(rc, rc->itr_setting);
3027 		}
3028 
3029 		vsi->q_vectors[i]->intrl = coalesce[i].intrl;
3030 		ice_write_intrl(vsi->q_vectors[i], coalesce[i].intrl);
3031 	}
3032 
3033 	/* the number of queue vectors increased so write whatever is in
3034 	 * the first element
3035 	 */
3036 	for (; i < vsi->num_q_vectors; i++) {
3037 		/* transmit */
3038 		rc = &vsi->q_vectors[i]->tx;
3039 		rc->itr_setting = coalesce[0].itr_tx;
3040 		ice_write_itr(rc, rc->itr_setting);
3041 
3042 		/* receive */
3043 		rc = &vsi->q_vectors[i]->rx;
3044 		rc->itr_setting = coalesce[0].itr_rx;
3045 		ice_write_itr(rc, rc->itr_setting);
3046 
3047 		vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3048 		ice_write_intrl(vsi->q_vectors[i], coalesce[0].intrl);
3049 	}
3050 }
3051 
3052 /**
3053  * ice_vsi_rebuild - Rebuild VSI after reset
3054  * @vsi: VSI to be rebuild
3055  * @init_vsi: is this an initialization or a reconfigure of the VSI
3056  *
3057  * Returns 0 on success and negative value on failure
3058  */
3059 int ice_vsi_rebuild(struct ice_vsi *vsi, bool init_vsi)
3060 {
3061 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3062 	struct ice_coalesce_stored *coalesce;
3063 	int prev_num_q_vectors = 0;
3064 	struct ice_vf *vf = NULL;
3065 	enum ice_vsi_type vtype;
3066 	enum ice_status status;
3067 	struct ice_pf *pf;
3068 	int ret, i;
3069 
3070 	if (!vsi)
3071 		return -EINVAL;
3072 
3073 	pf = vsi->back;
3074 	vtype = vsi->type;
3075 	if (vtype == ICE_VSI_VF)
3076 		vf = &pf->vf[vsi->vf_id];
3077 
3078 	coalesce = kcalloc(vsi->num_q_vectors,
3079 			   sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3080 	if (!coalesce)
3081 		return -ENOMEM;
3082 
3083 	prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3084 
3085 	ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
3086 	ice_vsi_free_q_vectors(vsi);
3087 
3088 	/* SR-IOV determines needed MSIX resources all at once instead of per
3089 	 * VSI since when VFs are spawned we know how many VFs there are and how
3090 	 * many interrupts each VF needs. SR-IOV MSIX resources are also
3091 	 * cleared in the same manner.
3092 	 */
3093 	if (vtype != ICE_VSI_VF) {
3094 		/* reclaim SW interrupts back to the common pool */
3095 		ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
3096 		pf->num_avail_sw_msix += vsi->num_q_vectors;
3097 		vsi->base_vector = 0;
3098 	}
3099 
3100 	if (ice_is_xdp_ena_vsi(vsi))
3101 		/* return value check can be skipped here, it always returns
3102 		 * 0 if reset is in progress
3103 		 */
3104 		ice_destroy_xdp_rings(vsi);
3105 	ice_vsi_put_qs(vsi);
3106 	ice_vsi_clear_rings(vsi);
3107 	ice_vsi_free_arrays(vsi);
3108 	if (vtype == ICE_VSI_VF)
3109 		ice_vsi_set_num_qs(vsi, vf->vf_id);
3110 	else
3111 		ice_vsi_set_num_qs(vsi, ICE_INVAL_VFID);
3112 
3113 	ret = ice_vsi_alloc_arrays(vsi);
3114 	if (ret < 0)
3115 		goto err_vsi;
3116 
3117 	ice_vsi_get_qs(vsi);
3118 
3119 	ice_alloc_fd_res(vsi);
3120 	ice_vsi_set_tc_cfg(vsi);
3121 
3122 	/* Initialize VSI struct elements and create VSI in FW */
3123 	ret = ice_vsi_init(vsi, init_vsi);
3124 	if (ret < 0)
3125 		goto err_vsi;
3126 
3127 	switch (vtype) {
3128 	case ICE_VSI_CTRL:
3129 	case ICE_VSI_PF:
3130 		ret = ice_vsi_alloc_q_vectors(vsi);
3131 		if (ret)
3132 			goto err_rings;
3133 
3134 		ret = ice_vsi_setup_vector_base(vsi);
3135 		if (ret)
3136 			goto err_vectors;
3137 
3138 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
3139 		if (ret)
3140 			goto err_vectors;
3141 
3142 		ret = ice_vsi_alloc_rings(vsi);
3143 		if (ret)
3144 			goto err_vectors;
3145 
3146 		ice_vsi_map_rings_to_vectors(vsi);
3147 		if (ice_is_xdp_ena_vsi(vsi)) {
3148 			vsi->num_xdp_txq = vsi->alloc_rxq;
3149 			ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog);
3150 			if (ret)
3151 				goto err_vectors;
3152 		}
3153 		/* ICE_VSI_CTRL does not need RSS so skip RSS processing */
3154 		if (vtype != ICE_VSI_CTRL)
3155 			/* Do not exit if configuring RSS had an issue, at
3156 			 * least receive traffic on first queue. Hence no
3157 			 * need to capture return value
3158 			 */
3159 			if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
3160 				ice_vsi_cfg_rss_lut_key(vsi);
3161 		break;
3162 	case ICE_VSI_VF:
3163 		ret = ice_vsi_alloc_q_vectors(vsi);
3164 		if (ret)
3165 			goto err_rings;
3166 
3167 		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
3168 		if (ret)
3169 			goto err_vectors;
3170 
3171 		ret = ice_vsi_alloc_rings(vsi);
3172 		if (ret)
3173 			goto err_vectors;
3174 
3175 		break;
3176 	default:
3177 		break;
3178 	}
3179 
3180 	/* configure VSI nodes based on number of queues and TC's */
3181 	for (i = 0; i < vsi->tc_cfg.numtc; i++) {
3182 		max_txqs[i] = vsi->alloc_txq;
3183 
3184 		if (ice_is_xdp_ena_vsi(vsi))
3185 			max_txqs[i] += vsi->num_xdp_txq;
3186 	}
3187 
3188 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
3189 				 max_txqs);
3190 	if (status) {
3191 		dev_err(ice_pf_to_dev(pf), "VSI %d failed lan queue config, error %s\n",
3192 			vsi->vsi_num, ice_stat_str(status));
3193 		if (init_vsi) {
3194 			ret = -EIO;
3195 			goto err_vectors;
3196 		} else {
3197 			return ice_schedule_reset(pf, ICE_RESET_PFR);
3198 		}
3199 	}
3200 	ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3201 	kfree(coalesce);
3202 
3203 	return 0;
3204 
3205 err_vectors:
3206 	ice_vsi_free_q_vectors(vsi);
3207 err_rings:
3208 	if (vsi->netdev) {
3209 		vsi->current_netdev_flags = 0;
3210 		unregister_netdev(vsi->netdev);
3211 		free_netdev(vsi->netdev);
3212 		vsi->netdev = NULL;
3213 	}
3214 err_vsi:
3215 	ice_vsi_clear(vsi);
3216 	set_bit(ICE_RESET_FAILED, pf->state);
3217 	kfree(coalesce);
3218 	return ret;
3219 }
3220 
3221 /**
3222  * ice_is_reset_in_progress - check for a reset in progress
3223  * @state: PF state field
3224  */
3225 bool ice_is_reset_in_progress(unsigned long *state)
3226 {
3227 	return test_bit(ICE_RESET_OICR_RECV, state) ||
3228 	       test_bit(ICE_PFR_REQ, state) ||
3229 	       test_bit(ICE_CORER_REQ, state) ||
3230 	       test_bit(ICE_GLOBR_REQ, state);
3231 }
3232 
3233 /**
3234  * ice_wait_for_reset - Wait for driver to finish reset and rebuild
3235  * @pf: pointer to the PF structure
3236  * @timeout: length of time to wait, in jiffies
3237  *
3238  * Wait (sleep) for a short time until the driver finishes cleaning up from
3239  * a device reset. The caller must be able to sleep. Use this to delay
3240  * operations that could fail while the driver is cleaning up after a device
3241  * reset.
3242  *
3243  * Returns 0 on success, -EBUSY if the reset is not finished within the
3244  * timeout, and -ERESTARTSYS if the thread was interrupted.
3245  */
3246 int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3247 {
3248 	long ret;
3249 
3250 	ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3251 					       !ice_is_reset_in_progress(pf->state),
3252 					       timeout);
3253 	if (ret < 0)
3254 		return ret;
3255 	else if (!ret)
3256 		return -EBUSY;
3257 	else
3258 		return 0;
3259 }
3260 
3261 #ifdef CONFIG_DCB
3262 /**
3263  * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3264  * @vsi: VSI being configured
3265  * @ctx: the context buffer returned from AQ VSI update command
3266  */
3267 static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3268 {
3269 	vsi->info.mapping_flags = ctx->info.mapping_flags;
3270 	memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3271 	       sizeof(vsi->info.q_mapping));
3272 	memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3273 	       sizeof(vsi->info.tc_mapping));
3274 }
3275 
3276 /**
3277  * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3278  * @vsi: VSI to be configured
3279  * @ena_tc: TC bitmap
3280  *
3281  * VSI queues expected to be quiesced before calling this function
3282  */
3283 int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3284 {
3285 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3286 	struct ice_pf *pf = vsi->back;
3287 	struct ice_vsi_ctx *ctx;
3288 	enum ice_status status;
3289 	struct device *dev;
3290 	int i, ret = 0;
3291 	u8 num_tc = 0;
3292 
3293 	dev = ice_pf_to_dev(pf);
3294 
3295 	ice_for_each_traffic_class(i) {
3296 		/* build bitmap of enabled TCs */
3297 		if (ena_tc & BIT(i))
3298 			num_tc++;
3299 		/* populate max_txqs per TC */
3300 		max_txqs[i] = vsi->alloc_txq;
3301 	}
3302 
3303 	vsi->tc_cfg.ena_tc = ena_tc;
3304 	vsi->tc_cfg.numtc = num_tc;
3305 
3306 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3307 	if (!ctx)
3308 		return -ENOMEM;
3309 
3310 	ctx->vf_num = 0;
3311 	ctx->info = vsi->info;
3312 
3313 	ice_vsi_setup_q_map(vsi, ctx);
3314 
3315 	/* must to indicate which section of VSI context are being modified */
3316 	ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3317 	status = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3318 	if (status) {
3319 		dev_info(dev, "Failed VSI Update\n");
3320 		ret = -EIO;
3321 		goto out;
3322 	}
3323 
3324 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
3325 				 max_txqs);
3326 
3327 	if (status) {
3328 		dev_err(dev, "VSI %d failed TC config, error %s\n",
3329 			vsi->vsi_num, ice_stat_str(status));
3330 		ret = -EIO;
3331 		goto out;
3332 	}
3333 	ice_vsi_update_q_map(vsi, ctx);
3334 	vsi->info.valid_sections = 0;
3335 
3336 	ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3337 out:
3338 	kfree(ctx);
3339 	return ret;
3340 }
3341 #endif /* CONFIG_DCB */
3342 
3343 /**
3344  * ice_update_ring_stats - Update ring statistics
3345  * @ring: ring to update
3346  * @pkts: number of processed packets
3347  * @bytes: number of processed bytes
3348  *
3349  * This function assumes that caller has acquired a u64_stats_sync lock.
3350  */
3351 static void ice_update_ring_stats(struct ice_ring *ring, u64 pkts, u64 bytes)
3352 {
3353 	ring->stats.bytes += bytes;
3354 	ring->stats.pkts += pkts;
3355 }
3356 
3357 /**
3358  * ice_update_tx_ring_stats - Update Tx ring specific counters
3359  * @tx_ring: ring to update
3360  * @pkts: number of processed packets
3361  * @bytes: number of processed bytes
3362  */
3363 void ice_update_tx_ring_stats(struct ice_ring *tx_ring, u64 pkts, u64 bytes)
3364 {
3365 	u64_stats_update_begin(&tx_ring->syncp);
3366 	ice_update_ring_stats(tx_ring, pkts, bytes);
3367 	u64_stats_update_end(&tx_ring->syncp);
3368 }
3369 
3370 /**
3371  * ice_update_rx_ring_stats - Update Rx ring specific counters
3372  * @rx_ring: ring to update
3373  * @pkts: number of processed packets
3374  * @bytes: number of processed bytes
3375  */
3376 void ice_update_rx_ring_stats(struct ice_ring *rx_ring, u64 pkts, u64 bytes)
3377 {
3378 	u64_stats_update_begin(&rx_ring->syncp);
3379 	ice_update_ring_stats(rx_ring, pkts, bytes);
3380 	u64_stats_update_end(&rx_ring->syncp);
3381 }
3382 
3383 /**
3384  * ice_status_to_errno - convert from enum ice_status to Linux errno
3385  * @err: ice_status value to convert
3386  */
3387 int ice_status_to_errno(enum ice_status err)
3388 {
3389 	switch (err) {
3390 	case ICE_SUCCESS:
3391 		return 0;
3392 	case ICE_ERR_DOES_NOT_EXIST:
3393 		return -ENOENT;
3394 	case ICE_ERR_OUT_OF_RANGE:
3395 		return -ENOTTY;
3396 	case ICE_ERR_PARAM:
3397 		return -EINVAL;
3398 	case ICE_ERR_NO_MEMORY:
3399 		return -ENOMEM;
3400 	case ICE_ERR_MAX_LIMIT:
3401 		return -EAGAIN;
3402 	default:
3403 		return -EINVAL;
3404 	}
3405 }
3406 
3407 /**
3408  * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3409  * @sw: switch to check if its default forwarding VSI is free
3410  *
3411  * Return true if the default forwarding VSI is already being used, else returns
3412  * false signalling that it's available to use.
3413  */
3414 bool ice_is_dflt_vsi_in_use(struct ice_sw *sw)
3415 {
3416 	return (sw->dflt_vsi && sw->dflt_vsi_ena);
3417 }
3418 
3419 /**
3420  * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3421  * @sw: switch for the default forwarding VSI to compare against
3422  * @vsi: VSI to compare against default forwarding VSI
3423  *
3424  * If this VSI passed in is the default forwarding VSI then return true, else
3425  * return false
3426  */
3427 bool ice_is_vsi_dflt_vsi(struct ice_sw *sw, struct ice_vsi *vsi)
3428 {
3429 	return (sw->dflt_vsi == vsi && sw->dflt_vsi_ena);
3430 }
3431 
3432 /**
3433  * ice_set_dflt_vsi - set the default forwarding VSI
3434  * @sw: switch used to assign the default forwarding VSI
3435  * @vsi: VSI getting set as the default forwarding VSI on the switch
3436  *
3437  * If the VSI passed in is already the default VSI and it's enabled just return
3438  * success.
3439  *
3440  * If there is already a default VSI on the switch and it's enabled then return
3441  * -EEXIST since there can only be one default VSI per switch.
3442  *
3443  *  Otherwise try to set the VSI passed in as the switch's default VSI and
3444  *  return the result.
3445  */
3446 int ice_set_dflt_vsi(struct ice_sw *sw, struct ice_vsi *vsi)
3447 {
3448 	enum ice_status status;
3449 	struct device *dev;
3450 
3451 	if (!sw || !vsi)
3452 		return -EINVAL;
3453 
3454 	dev = ice_pf_to_dev(vsi->back);
3455 
3456 	/* the VSI passed in is already the default VSI */
3457 	if (ice_is_vsi_dflt_vsi(sw, vsi)) {
3458 		dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3459 			vsi->vsi_num);
3460 		return 0;
3461 	}
3462 
3463 	/* another VSI is already the default VSI for this switch */
3464 	if (ice_is_dflt_vsi_in_use(sw)) {
3465 		dev_err(dev, "Default forwarding VSI %d already in use, disable it and try again\n",
3466 			sw->dflt_vsi->vsi_num);
3467 		return -EEXIST;
3468 	}
3469 
3470 	status = ice_cfg_dflt_vsi(&vsi->back->hw, vsi->idx, true, ICE_FLTR_RX);
3471 	if (status) {
3472 		dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %s\n",
3473 			vsi->vsi_num, ice_stat_str(status));
3474 		return -EIO;
3475 	}
3476 
3477 	sw->dflt_vsi = vsi;
3478 	sw->dflt_vsi_ena = true;
3479 
3480 	return 0;
3481 }
3482 
3483 /**
3484  * ice_clear_dflt_vsi - clear the default forwarding VSI
3485  * @sw: switch used to clear the default VSI
3486  *
3487  * If the switch has no default VSI or it's not enabled then return error.
3488  *
3489  * Otherwise try to clear the default VSI and return the result.
3490  */
3491 int ice_clear_dflt_vsi(struct ice_sw *sw)
3492 {
3493 	struct ice_vsi *dflt_vsi;
3494 	enum ice_status status;
3495 	struct device *dev;
3496 
3497 	if (!sw)
3498 		return -EINVAL;
3499 
3500 	dev = ice_pf_to_dev(sw->pf);
3501 
3502 	dflt_vsi = sw->dflt_vsi;
3503 
3504 	/* there is no default VSI configured */
3505 	if (!ice_is_dflt_vsi_in_use(sw))
3506 		return -ENODEV;
3507 
3508 	status = ice_cfg_dflt_vsi(&dflt_vsi->back->hw, dflt_vsi->idx, false,
3509 				  ICE_FLTR_RX);
3510 	if (status) {
3511 		dev_err(dev, "Failed to clear the default forwarding VSI %d, error %s\n",
3512 			dflt_vsi->vsi_num, ice_stat_str(status));
3513 		return -EIO;
3514 	}
3515 
3516 	sw->dflt_vsi = NULL;
3517 	sw->dflt_vsi_ena = false;
3518 
3519 	return 0;
3520 }
3521 
3522 /**
3523  * ice_set_link - turn on/off physical link
3524  * @vsi: VSI to modify physical link on
3525  * @ena: turn on/off physical link
3526  */
3527 int ice_set_link(struct ice_vsi *vsi, bool ena)
3528 {
3529 	struct device *dev = ice_pf_to_dev(vsi->back);
3530 	struct ice_port_info *pi = vsi->port_info;
3531 	struct ice_hw *hw = pi->hw;
3532 	enum ice_status status;
3533 
3534 	if (vsi->type != ICE_VSI_PF)
3535 		return -EINVAL;
3536 
3537 	status = ice_aq_set_link_restart_an(pi, ena, NULL);
3538 
3539 	/* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3540 	 * this is not a fatal error, so print a warning message and return
3541 	 * a success code. Return an error if FW returns an error code other
3542 	 * than ICE_AQ_RC_EMODE
3543 	 */
3544 	if (status == ICE_ERR_AQ_ERROR) {
3545 		if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3546 			dev_warn(dev, "can't set link to %s, err %s aq_err %s. not fatal, continuing\n",
3547 				 (ena ? "ON" : "OFF"), ice_stat_str(status),
3548 				 ice_aq_str(hw->adminq.sq_last_status));
3549 	} else if (status) {
3550 		dev_err(dev, "can't set link to %s, err %s aq_err %s\n",
3551 			(ena ? "ON" : "OFF"), ice_stat_str(status),
3552 			ice_aq_str(hw->adminq.sq_last_status));
3553 		return -EIO;
3554 	}
3555 
3556 	return 0;
3557 }
3558