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#include "ice_vsi_vlan_ops.h"
12
13/**
14 * ice_vsi_type_str - maps VSI type enum to string equivalents
15 * @vsi_type: VSI type enum
16 */
17const char *ice_vsi_type_str(enum ice_vsi_type vsi_type)
18{
19	switch (vsi_type) {
20	case ICE_VSI_PF:
21		return "ICE_VSI_PF";
22	case ICE_VSI_VF:
23		return "ICE_VSI_VF";
24	case ICE_VSI_CTRL:
25		return "ICE_VSI_CTRL";
26	case ICE_VSI_CHNL:
27		return "ICE_VSI_CHNL";
28	case ICE_VSI_LB:
29		return "ICE_VSI_LB";
30	case ICE_VSI_SWITCHDEV_CTRL:
31		return "ICE_VSI_SWITCHDEV_CTRL";
32	default:
33		return "unknown";
34	}
35}
36
37/**
38 * ice_vsi_ctrl_all_rx_rings - Start or stop a VSI's Rx rings
39 * @vsi: the VSI being configured
40 * @ena: start or stop the Rx rings
41 *
42 * First enable/disable all of the Rx rings, flush any remaining writes, and
43 * then verify that they have all been enabled/disabled successfully. This will
44 * let all of the register writes complete when enabling/disabling the Rx rings
45 * before waiting for the change in hardware to complete.
46 */
47static int ice_vsi_ctrl_all_rx_rings(struct ice_vsi *vsi, bool ena)
48{
49	int ret = 0;
50	u16 i;
51
52	ice_for_each_rxq(vsi, i)
53		ice_vsi_ctrl_one_rx_ring(vsi, ena, i, false);
54
55	ice_flush(&vsi->back->hw);
56
57	ice_for_each_rxq(vsi, i) {
58		ret = ice_vsi_wait_one_rx_ring(vsi, ena, i);
59		if (ret)
60			break;
61	}
62
63	return ret;
64}
65
66/**
67 * ice_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the VSI
68 * @vsi: VSI pointer
69 *
70 * On error: returns error code (negative)
71 * On success: returns 0
72 */
73static int ice_vsi_alloc_arrays(struct ice_vsi *vsi)
74{
75	struct ice_pf *pf = vsi->back;
76	struct device *dev;
77
78	dev = ice_pf_to_dev(pf);
79	if (vsi->type == ICE_VSI_CHNL)
80		return 0;
81
82	/* allocate memory for both Tx and Rx ring pointers */
83	vsi->tx_rings = devm_kcalloc(dev, vsi->alloc_txq,
84				     sizeof(*vsi->tx_rings), GFP_KERNEL);
85	if (!vsi->tx_rings)
86		return -ENOMEM;
87
88	vsi->rx_rings = devm_kcalloc(dev, vsi->alloc_rxq,
89				     sizeof(*vsi->rx_rings), GFP_KERNEL);
90	if (!vsi->rx_rings)
91		goto err_rings;
92
93	/* txq_map needs to have enough space to track both Tx (stack) rings
94	 * and XDP rings; at this point vsi->num_xdp_txq might not be set,
95	 * so use num_possible_cpus() as we want to always provide XDP ring
96	 * per CPU, regardless of queue count settings from user that might
97	 * have come from ethtool's set_channels() callback;
98	 */
99	vsi->txq_map = devm_kcalloc(dev, (vsi->alloc_txq + num_possible_cpus()),
100				    sizeof(*vsi->txq_map), GFP_KERNEL);
101
102	if (!vsi->txq_map)
103		goto err_txq_map;
104
105	vsi->rxq_map = devm_kcalloc(dev, vsi->alloc_rxq,
106				    sizeof(*vsi->rxq_map), GFP_KERNEL);
107	if (!vsi->rxq_map)
108		goto err_rxq_map;
109
110	/* There is no need to allocate q_vectors for a loopback VSI. */
111	if (vsi->type == ICE_VSI_LB)
112		return 0;
113
114	/* allocate memory for q_vector pointers */
115	vsi->q_vectors = devm_kcalloc(dev, vsi->num_q_vectors,
116				      sizeof(*vsi->q_vectors), GFP_KERNEL);
117	if (!vsi->q_vectors)
118		goto err_vectors;
119
120	vsi->af_xdp_zc_qps = bitmap_zalloc(max_t(int, vsi->alloc_txq, vsi->alloc_rxq), GFP_KERNEL);
121	if (!vsi->af_xdp_zc_qps)
122		goto err_zc_qps;
123
124	return 0;
125
126err_zc_qps:
127	devm_kfree(dev, vsi->q_vectors);
128err_vectors:
129	devm_kfree(dev, vsi->rxq_map);
130err_rxq_map:
131	devm_kfree(dev, vsi->txq_map);
132err_txq_map:
133	devm_kfree(dev, vsi->rx_rings);
134err_rings:
135	devm_kfree(dev, vsi->tx_rings);
136	return -ENOMEM;
137}
138
139/**
140 * ice_vsi_set_num_desc - Set number of descriptors for queues on this VSI
141 * @vsi: the VSI being configured
142 */
143static void ice_vsi_set_num_desc(struct ice_vsi *vsi)
144{
145	switch (vsi->type) {
146	case ICE_VSI_PF:
147	case ICE_VSI_SWITCHDEV_CTRL:
148	case ICE_VSI_CTRL:
149	case ICE_VSI_LB:
150		/* a user could change the values of num_[tr]x_desc using
151		 * ethtool -G so we should keep those values instead of
152		 * overwriting them with the defaults.
153		 */
154		if (!vsi->num_rx_desc)
155			vsi->num_rx_desc = ICE_DFLT_NUM_RX_DESC;
156		if (!vsi->num_tx_desc)
157			vsi->num_tx_desc = ICE_DFLT_NUM_TX_DESC;
158		break;
159	default:
160		dev_dbg(ice_pf_to_dev(vsi->back), "Not setting number of Tx/Rx descriptors for VSI type %d\n",
161			vsi->type);
162		break;
163	}
164}
165
166/**
167 * ice_vsi_set_num_qs - Set number of queues, descriptors and vectors for a VSI
168 * @vsi: the VSI being configured
169 *
170 * Return 0 on success and a negative value on error
171 */
172static void ice_vsi_set_num_qs(struct ice_vsi *vsi)
173{
174	enum ice_vsi_type vsi_type = vsi->type;
175	struct ice_pf *pf = vsi->back;
176	struct ice_vf *vf = vsi->vf;
177
178	if (WARN_ON(vsi_type == ICE_VSI_VF && !vf))
179		return;
180
181	switch (vsi_type) {
182	case ICE_VSI_PF:
183		if (vsi->req_txq) {
184			vsi->alloc_txq = vsi->req_txq;
185			vsi->num_txq = vsi->req_txq;
186		} else {
187			vsi->alloc_txq = min3(pf->num_lan_msix,
188					      ice_get_avail_txq_count(pf),
189					      (u16)num_online_cpus());
190		}
191
192		pf->num_lan_tx = vsi->alloc_txq;
193
194		/* only 1 Rx queue unless RSS is enabled */
195		if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
196			vsi->alloc_rxq = 1;
197		} else {
198			if (vsi->req_rxq) {
199				vsi->alloc_rxq = vsi->req_rxq;
200				vsi->num_rxq = vsi->req_rxq;
201			} else {
202				vsi->alloc_rxq = min3(pf->num_lan_msix,
203						      ice_get_avail_rxq_count(pf),
204						      (u16)num_online_cpus());
205			}
206		}
207
208		pf->num_lan_rx = vsi->alloc_rxq;
209
210		vsi->num_q_vectors = min_t(int, pf->num_lan_msix,
211					   max_t(int, vsi->alloc_rxq,
212						 vsi->alloc_txq));
213		break;
214	case ICE_VSI_SWITCHDEV_CTRL:
215		/* The number of queues for ctrl VSI is equal to number of PRs
216		 * Each ring is associated to the corresponding VF_PR netdev.
217		 * Tx and Rx rings are always equal
218		 */
219		if (vsi->req_txq && vsi->req_rxq) {
220			vsi->alloc_txq = vsi->req_txq;
221			vsi->alloc_rxq = vsi->req_rxq;
222		} else {
223			vsi->alloc_txq = 1;
224			vsi->alloc_rxq = 1;
225		}
226
227		vsi->num_q_vectors = 1;
228		break;
229	case ICE_VSI_VF:
230		if (vf->num_req_qs)
231			vf->num_vf_qs = vf->num_req_qs;
232		vsi->alloc_txq = vf->num_vf_qs;
233		vsi->alloc_rxq = vf->num_vf_qs;
234		/* pf->vfs.num_msix_per includes (VF miscellaneous vector +
235		 * data queue interrupts). Since vsi->num_q_vectors is number
236		 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
237		 * original vector count
238		 */
239		vsi->num_q_vectors = vf->num_msix - ICE_NONQ_VECS_VF;
240		break;
241	case ICE_VSI_CTRL:
242		vsi->alloc_txq = 1;
243		vsi->alloc_rxq = 1;
244		vsi->num_q_vectors = 1;
245		break;
246	case ICE_VSI_CHNL:
247		vsi->alloc_txq = 0;
248		vsi->alloc_rxq = 0;
249		break;
250	case ICE_VSI_LB:
251		vsi->alloc_txq = 1;
252		vsi->alloc_rxq = 1;
253		break;
254	default:
255		dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi_type);
256		break;
257	}
258
259	ice_vsi_set_num_desc(vsi);
260}
261
262/**
263 * ice_get_free_slot - get the next non-NULL location index in array
264 * @array: array to search
265 * @size: size of the array
266 * @curr: last known occupied index to be used as a search hint
267 *
268 * void * is being used to keep the functionality generic. This lets us use this
269 * function on any array of pointers.
270 */
271static int ice_get_free_slot(void *array, int size, int curr)
272{
273	int **tmp_array = (int **)array;
274	int next;
275
276	if (curr < (size - 1) && !tmp_array[curr + 1]) {
277		next = curr + 1;
278	} else {
279		int i = 0;
280
281		while ((i < size) && (tmp_array[i]))
282			i++;
283		if (i == size)
284			next = ICE_NO_VSI;
285		else
286			next = i;
287	}
288	return next;
289}
290
291/**
292 * ice_vsi_delete_from_hw - delete a VSI from the switch
293 * @vsi: pointer to VSI being removed
294 */
295static void ice_vsi_delete_from_hw(struct ice_vsi *vsi)
296{
297	struct ice_pf *pf = vsi->back;
298	struct ice_vsi_ctx *ctxt;
299	int status;
300
301	ice_fltr_remove_all(vsi);
302	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
303	if (!ctxt)
304		return;
305
306	if (vsi->type == ICE_VSI_VF)
307		ctxt->vf_num = vsi->vf->vf_id;
308	ctxt->vsi_num = vsi->vsi_num;
309
310	memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
311
312	status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
313	if (status)
314		dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %d\n",
315			vsi->vsi_num, status);
316
317	kfree(ctxt);
318}
319
320/**
321 * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
322 * @vsi: pointer to VSI being cleared
323 */
324static void ice_vsi_free_arrays(struct ice_vsi *vsi)
325{
326	struct ice_pf *pf = vsi->back;
327	struct device *dev;
328
329	dev = ice_pf_to_dev(pf);
330
331	bitmap_free(vsi->af_xdp_zc_qps);
332	vsi->af_xdp_zc_qps = NULL;
333	/* free the ring and vector containers */
334	devm_kfree(dev, vsi->q_vectors);
335	vsi->q_vectors = NULL;
336	devm_kfree(dev, vsi->tx_rings);
337	vsi->tx_rings = NULL;
338	devm_kfree(dev, vsi->rx_rings);
339	vsi->rx_rings = NULL;
340	devm_kfree(dev, vsi->txq_map);
341	vsi->txq_map = NULL;
342	devm_kfree(dev, vsi->rxq_map);
343	vsi->rxq_map = NULL;
344}
345
346/**
347 * ice_vsi_free_stats - Free the ring statistics structures
348 * @vsi: VSI pointer
349 */
350static void ice_vsi_free_stats(struct ice_vsi *vsi)
351{
352	struct ice_vsi_stats *vsi_stat;
353	struct ice_pf *pf = vsi->back;
354	int i;
355
356	if (vsi->type == ICE_VSI_CHNL)
357		return;
358	if (!pf->vsi_stats)
359		return;
360
361	vsi_stat = pf->vsi_stats[vsi->idx];
362	if (!vsi_stat)
363		return;
364
365	ice_for_each_alloc_txq(vsi, i) {
366		if (vsi_stat->tx_ring_stats[i]) {
367			kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
368			WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
369		}
370	}
371
372	ice_for_each_alloc_rxq(vsi, i) {
373		if (vsi_stat->rx_ring_stats[i]) {
374			kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
375			WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
376		}
377	}
378
379	kfree(vsi_stat->tx_ring_stats);
380	kfree(vsi_stat->rx_ring_stats);
381	kfree(vsi_stat);
382	pf->vsi_stats[vsi->idx] = NULL;
383}
384
385/**
386 * ice_vsi_alloc_ring_stats - Allocates Tx and Rx ring stats for the VSI
387 * @vsi: VSI which is having stats allocated
388 */
389static int ice_vsi_alloc_ring_stats(struct ice_vsi *vsi)
390{
391	struct ice_ring_stats **tx_ring_stats;
392	struct ice_ring_stats **rx_ring_stats;
393	struct ice_vsi_stats *vsi_stats;
394	struct ice_pf *pf = vsi->back;
395	u16 i;
396
397	vsi_stats = pf->vsi_stats[vsi->idx];
398	tx_ring_stats = vsi_stats->tx_ring_stats;
399	rx_ring_stats = vsi_stats->rx_ring_stats;
400
401	/* Allocate Tx ring stats */
402	ice_for_each_alloc_txq(vsi, i) {
403		struct ice_ring_stats *ring_stats;
404		struct ice_tx_ring *ring;
405
406		ring = vsi->tx_rings[i];
407		ring_stats = tx_ring_stats[i];
408
409		if (!ring_stats) {
410			ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
411			if (!ring_stats)
412				goto err_out;
413
414			WRITE_ONCE(tx_ring_stats[i], ring_stats);
415		}
416
417		ring->ring_stats = ring_stats;
418	}
419
420	/* Allocate Rx ring stats */
421	ice_for_each_alloc_rxq(vsi, i) {
422		struct ice_ring_stats *ring_stats;
423		struct ice_rx_ring *ring;
424
425		ring = vsi->rx_rings[i];
426		ring_stats = rx_ring_stats[i];
427
428		if (!ring_stats) {
429			ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
430			if (!ring_stats)
431				goto err_out;
432
433			WRITE_ONCE(rx_ring_stats[i], ring_stats);
434		}
435
436		ring->ring_stats = ring_stats;
437	}
438
439	return 0;
440
441err_out:
442	ice_vsi_free_stats(vsi);
443	return -ENOMEM;
444}
445
446/**
447 * ice_vsi_free - clean up and deallocate the provided VSI
448 * @vsi: pointer to VSI being cleared
449 *
450 * This deallocates the VSI's queue resources, removes it from the PF's
451 * VSI array if necessary, and deallocates the VSI
452 */
453static void ice_vsi_free(struct ice_vsi *vsi)
454{
455	struct ice_pf *pf = NULL;
456	struct device *dev;
457
458	if (!vsi || !vsi->back)
459		return;
460
461	pf = vsi->back;
462	dev = ice_pf_to_dev(pf);
463
464	if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
465		dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
466		return;
467	}
468
469	mutex_lock(&pf->sw_mutex);
470	/* updates the PF for this cleared VSI */
471
472	pf->vsi[vsi->idx] = NULL;
473	pf->next_vsi = vsi->idx;
474
475	ice_vsi_free_stats(vsi);
476	ice_vsi_free_arrays(vsi);
477	mutex_unlock(&pf->sw_mutex);
478	devm_kfree(dev, vsi);
479}
480
481void ice_vsi_delete(struct ice_vsi *vsi)
482{
483	ice_vsi_delete_from_hw(vsi);
484	ice_vsi_free(vsi);
485}
486
487/**
488 * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
489 * @irq: interrupt number
490 * @data: pointer to a q_vector
491 */
492static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
493{
494	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
495
496	if (!q_vector->tx.tx_ring)
497		return IRQ_HANDLED;
498
499#define FDIR_RX_DESC_CLEAN_BUDGET 64
500	ice_clean_rx_irq(q_vector->rx.rx_ring, FDIR_RX_DESC_CLEAN_BUDGET);
501	ice_clean_ctrl_tx_irq(q_vector->tx.tx_ring);
502
503	return IRQ_HANDLED;
504}
505
506/**
507 * ice_msix_clean_rings - MSIX mode Interrupt Handler
508 * @irq: interrupt number
509 * @data: pointer to a q_vector
510 */
511static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
512{
513	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
514
515	if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
516		return IRQ_HANDLED;
517
518	q_vector->total_events++;
519
520	napi_schedule(&q_vector->napi);
521
522	return IRQ_HANDLED;
523}
524
525static irqreturn_t ice_eswitch_msix_clean_rings(int __always_unused irq, void *data)
526{
527	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
528	struct ice_pf *pf = q_vector->vsi->back;
529	struct ice_repr *repr;
530	unsigned long id;
531
532	if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
533		return IRQ_HANDLED;
534
535	xa_for_each(&pf->eswitch.reprs, id, repr)
536		napi_schedule(&repr->q_vector->napi);
537
538	return IRQ_HANDLED;
539}
540
541/**
542 * ice_vsi_alloc_stat_arrays - Allocate statistics arrays
543 * @vsi: VSI pointer
544 */
545static int ice_vsi_alloc_stat_arrays(struct ice_vsi *vsi)
546{
547	struct ice_vsi_stats *vsi_stat;
548	struct ice_pf *pf = vsi->back;
549
550	if (vsi->type == ICE_VSI_CHNL)
551		return 0;
552	if (!pf->vsi_stats)
553		return -ENOENT;
554
555	if (pf->vsi_stats[vsi->idx])
556	/* realloc will happen in rebuild path */
557		return 0;
558
559	vsi_stat = kzalloc(sizeof(*vsi_stat), GFP_KERNEL);
560	if (!vsi_stat)
561		return -ENOMEM;
562
563	vsi_stat->tx_ring_stats =
564		kcalloc(vsi->alloc_txq, sizeof(*vsi_stat->tx_ring_stats),
565			GFP_KERNEL);
566	if (!vsi_stat->tx_ring_stats)
567		goto err_alloc_tx;
568
569	vsi_stat->rx_ring_stats =
570		kcalloc(vsi->alloc_rxq, sizeof(*vsi_stat->rx_ring_stats),
571			GFP_KERNEL);
572	if (!vsi_stat->rx_ring_stats)
573		goto err_alloc_rx;
574
575	pf->vsi_stats[vsi->idx] = vsi_stat;
576
577	return 0;
578
579err_alloc_rx:
580	kfree(vsi_stat->rx_ring_stats);
581err_alloc_tx:
582	kfree(vsi_stat->tx_ring_stats);
583	kfree(vsi_stat);
584	pf->vsi_stats[vsi->idx] = NULL;
585	return -ENOMEM;
586}
587
588/**
589 * ice_vsi_alloc_def - set default values for already allocated VSI
590 * @vsi: ptr to VSI
591 * @ch: ptr to channel
592 */
593static int
594ice_vsi_alloc_def(struct ice_vsi *vsi, struct ice_channel *ch)
595{
596	if (vsi->type != ICE_VSI_CHNL) {
597		ice_vsi_set_num_qs(vsi);
598		if (ice_vsi_alloc_arrays(vsi))
599			return -ENOMEM;
600	}
601
602	switch (vsi->type) {
603	case ICE_VSI_SWITCHDEV_CTRL:
604		/* Setup eswitch MSIX irq handler for VSI */
605		vsi->irq_handler = ice_eswitch_msix_clean_rings;
606		break;
607	case ICE_VSI_PF:
608		/* Setup default MSIX irq handler for VSI */
609		vsi->irq_handler = ice_msix_clean_rings;
610		break;
611	case ICE_VSI_CTRL:
612		/* Setup ctrl VSI MSIX irq handler */
613		vsi->irq_handler = ice_msix_clean_ctrl_vsi;
614		break;
615	case ICE_VSI_CHNL:
616		if (!ch)
617			return -EINVAL;
618
619		vsi->num_rxq = ch->num_rxq;
620		vsi->num_txq = ch->num_txq;
621		vsi->next_base_q = ch->base_q;
622		break;
623	case ICE_VSI_VF:
624	case ICE_VSI_LB:
625		break;
626	default:
627		ice_vsi_free_arrays(vsi);
628		return -EINVAL;
629	}
630
631	return 0;
632}
633
634/**
635 * ice_vsi_alloc - Allocates the next available struct VSI in the PF
636 * @pf: board private structure
637 *
638 * Reserves a VSI index from the PF and allocates an empty VSI structure
639 * without a type. The VSI structure must later be initialized by calling
640 * ice_vsi_cfg().
641 *
642 * returns a pointer to a VSI on success, NULL on failure.
643 */
644static struct ice_vsi *ice_vsi_alloc(struct ice_pf *pf)
645{
646	struct device *dev = ice_pf_to_dev(pf);
647	struct ice_vsi *vsi = NULL;
648
649	/* Need to protect the allocation of the VSIs at the PF level */
650	mutex_lock(&pf->sw_mutex);
651
652	/* If we have already allocated our maximum number of VSIs,
653	 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
654	 * is available to be populated
655	 */
656	if (pf->next_vsi == ICE_NO_VSI) {
657		dev_dbg(dev, "out of VSI slots!\n");
658		goto unlock_pf;
659	}
660
661	vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
662	if (!vsi)
663		goto unlock_pf;
664
665	vsi->back = pf;
666	set_bit(ICE_VSI_DOWN, vsi->state);
667
668	/* fill slot and make note of the index */
669	vsi->idx = pf->next_vsi;
670	pf->vsi[pf->next_vsi] = vsi;
671
672	/* prepare pf->next_vsi for next use */
673	pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
674					 pf->next_vsi);
675
676unlock_pf:
677	mutex_unlock(&pf->sw_mutex);
678	return vsi;
679}
680
681/**
682 * ice_alloc_fd_res - Allocate FD resource for a VSI
683 * @vsi: pointer to the ice_vsi
684 *
685 * This allocates the FD resources
686 *
687 * Returns 0 on success, -EPERM on no-op or -EIO on failure
688 */
689static int ice_alloc_fd_res(struct ice_vsi *vsi)
690{
691	struct ice_pf *pf = vsi->back;
692	u32 g_val, b_val;
693
694	/* Flow Director filters are only allocated/assigned to the PF VSI or
695	 * CHNL VSI which passes the traffic. The CTRL VSI is only used to
696	 * add/delete filters so resources are not allocated to it
697	 */
698	if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
699		return -EPERM;
700
701	if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF ||
702	      vsi->type == ICE_VSI_CHNL))
703		return -EPERM;
704
705	/* FD filters from guaranteed pool per VSI */
706	g_val = pf->hw.func_caps.fd_fltr_guar;
707	if (!g_val)
708		return -EPERM;
709
710	/* FD filters from best effort pool */
711	b_val = pf->hw.func_caps.fd_fltr_best_effort;
712	if (!b_val)
713		return -EPERM;
714
715	/* PF main VSI gets only 64 FD resources from guaranteed pool
716	 * when ADQ is configured.
717	 */
718#define ICE_PF_VSI_GFLTR	64
719
720	/* determine FD filter resources per VSI from shared(best effort) and
721	 * dedicated pool
722	 */
723	if (vsi->type == ICE_VSI_PF) {
724		vsi->num_gfltr = g_val;
725		/* if MQPRIO is configured, main VSI doesn't get all FD
726		 * resources from guaranteed pool. PF VSI gets 64 FD resources
727		 */
728		if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
729			if (g_val < ICE_PF_VSI_GFLTR)
730				return -EPERM;
731			/* allow bare minimum entries for PF VSI */
732			vsi->num_gfltr = ICE_PF_VSI_GFLTR;
733		}
734
735		/* each VSI gets same "best_effort" quota */
736		vsi->num_bfltr = b_val;
737	} else if (vsi->type == ICE_VSI_VF) {
738		vsi->num_gfltr = 0;
739
740		/* each VSI gets same "best_effort" quota */
741		vsi->num_bfltr = b_val;
742	} else {
743		struct ice_vsi *main_vsi;
744		int numtc;
745
746		main_vsi = ice_get_main_vsi(pf);
747		if (!main_vsi)
748			return -EPERM;
749
750		if (!main_vsi->all_numtc)
751			return -EINVAL;
752
753		/* figure out ADQ numtc */
754		numtc = main_vsi->all_numtc - ICE_CHNL_START_TC;
755
756		/* only one TC but still asking resources for channels,
757		 * invalid config
758		 */
759		if (numtc < ICE_CHNL_START_TC)
760			return -EPERM;
761
762		g_val -= ICE_PF_VSI_GFLTR;
763		/* channel VSIs gets equal share from guaranteed pool */
764		vsi->num_gfltr = g_val / numtc;
765
766		/* each VSI gets same "best_effort" quota */
767		vsi->num_bfltr = b_val;
768	}
769
770	return 0;
771}
772
773/**
774 * ice_vsi_get_qs - Assign queues from PF to VSI
775 * @vsi: the VSI to assign queues to
776 *
777 * Returns 0 on success and a negative value on error
778 */
779static int ice_vsi_get_qs(struct ice_vsi *vsi)
780{
781	struct ice_pf *pf = vsi->back;
782	struct ice_qs_cfg tx_qs_cfg = {
783		.qs_mutex = &pf->avail_q_mutex,
784		.pf_map = pf->avail_txqs,
785		.pf_map_size = pf->max_pf_txqs,
786		.q_count = vsi->alloc_txq,
787		.scatter_count = ICE_MAX_SCATTER_TXQS,
788		.vsi_map = vsi->txq_map,
789		.vsi_map_offset = 0,
790		.mapping_mode = ICE_VSI_MAP_CONTIG
791	};
792	struct ice_qs_cfg rx_qs_cfg = {
793		.qs_mutex = &pf->avail_q_mutex,
794		.pf_map = pf->avail_rxqs,
795		.pf_map_size = pf->max_pf_rxqs,
796		.q_count = vsi->alloc_rxq,
797		.scatter_count = ICE_MAX_SCATTER_RXQS,
798		.vsi_map = vsi->rxq_map,
799		.vsi_map_offset = 0,
800		.mapping_mode = ICE_VSI_MAP_CONTIG
801	};
802	int ret;
803
804	if (vsi->type == ICE_VSI_CHNL)
805		return 0;
806
807	ret = __ice_vsi_get_qs(&tx_qs_cfg);
808	if (ret)
809		return ret;
810	vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
811
812	ret = __ice_vsi_get_qs(&rx_qs_cfg);
813	if (ret)
814		return ret;
815	vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
816
817	return 0;
818}
819
820/**
821 * ice_vsi_put_qs - Release queues from VSI to PF
822 * @vsi: the VSI that is going to release queues
823 */
824static void ice_vsi_put_qs(struct ice_vsi *vsi)
825{
826	struct ice_pf *pf = vsi->back;
827	int i;
828
829	mutex_lock(&pf->avail_q_mutex);
830
831	ice_for_each_alloc_txq(vsi, i) {
832		clear_bit(vsi->txq_map[i], pf->avail_txqs);
833		vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
834	}
835
836	ice_for_each_alloc_rxq(vsi, i) {
837		clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
838		vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
839	}
840
841	mutex_unlock(&pf->avail_q_mutex);
842}
843
844/**
845 * ice_is_safe_mode
846 * @pf: pointer to the PF struct
847 *
848 * returns true if driver is in safe mode, false otherwise
849 */
850bool ice_is_safe_mode(struct ice_pf *pf)
851{
852	return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
853}
854
855/**
856 * ice_is_rdma_ena
857 * @pf: pointer to the PF struct
858 *
859 * returns true if RDMA is currently supported, false otherwise
860 */
861bool ice_is_rdma_ena(struct ice_pf *pf)
862{
863	return test_bit(ICE_FLAG_RDMA_ENA, pf->flags);
864}
865
866/**
867 * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
868 * @vsi: the VSI being cleaned up
869 *
870 * This function deletes RSS input set for all flows that were configured
871 * for this VSI
872 */
873static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
874{
875	struct ice_pf *pf = vsi->back;
876	int status;
877
878	if (ice_is_safe_mode(pf))
879		return;
880
881	status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
882	if (status)
883		dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %d\n",
884			vsi->vsi_num, status);
885}
886
887/**
888 * ice_rss_clean - Delete RSS related VSI structures and configuration
889 * @vsi: the VSI being removed
890 */
891static void ice_rss_clean(struct ice_vsi *vsi)
892{
893	struct ice_pf *pf = vsi->back;
894	struct device *dev;
895
896	dev = ice_pf_to_dev(pf);
897
898	devm_kfree(dev, vsi->rss_hkey_user);
899	devm_kfree(dev, vsi->rss_lut_user);
900
901	ice_vsi_clean_rss_flow_fld(vsi);
902	/* remove RSS replay list */
903	if (!ice_is_safe_mode(pf))
904		ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
905}
906
907/**
908 * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
909 * @vsi: the VSI being configured
910 */
911static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
912{
913	struct ice_hw_common_caps *cap;
914	struct ice_pf *pf = vsi->back;
915	u16 max_rss_size;
916
917	if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
918		vsi->rss_size = 1;
919		return;
920	}
921
922	cap = &pf->hw.func_caps.common_cap;
923	max_rss_size = BIT(cap->rss_table_entry_width);
924	switch (vsi->type) {
925	case ICE_VSI_CHNL:
926	case ICE_VSI_PF:
927		/* PF VSI will inherit RSS instance of PF */
928		vsi->rss_table_size = (u16)cap->rss_table_size;
929		if (vsi->type == ICE_VSI_CHNL)
930			vsi->rss_size = min_t(u16, vsi->num_rxq, max_rss_size);
931		else
932			vsi->rss_size = min_t(u16, num_online_cpus(),
933					      max_rss_size);
934		vsi->rss_lut_type = ICE_LUT_PF;
935		break;
936	case ICE_VSI_SWITCHDEV_CTRL:
937		vsi->rss_table_size = ICE_LUT_VSI_SIZE;
938		vsi->rss_size = min_t(u16, num_online_cpus(), max_rss_size);
939		vsi->rss_lut_type = ICE_LUT_VSI;
940		break;
941	case ICE_VSI_VF:
942		/* VF VSI will get a small RSS table.
943		 * For VSI_LUT, LUT size should be set to 64 bytes.
944		 */
945		vsi->rss_table_size = ICE_LUT_VSI_SIZE;
946		vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
947		vsi->rss_lut_type = ICE_LUT_VSI;
948		break;
949	case ICE_VSI_LB:
950		break;
951	default:
952		dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
953			ice_vsi_type_str(vsi->type));
954		break;
955	}
956}
957
958/**
959 * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
960 * @hw: HW structure used to determine the VLAN mode of the device
961 * @ctxt: the VSI context being set
962 *
963 * This initializes a default VSI context for all sections except the Queues.
964 */
965static void ice_set_dflt_vsi_ctx(struct ice_hw *hw, struct ice_vsi_ctx *ctxt)
966{
967	u32 table = 0;
968
969	memset(&ctxt->info, 0, sizeof(ctxt->info));
970	/* VSI's should be allocated from shared pool */
971	ctxt->alloc_from_pool = true;
972	/* Src pruning enabled by default */
973	ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
974	/* Traffic from VSI can be sent to LAN */
975	ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
976	/* allow all untagged/tagged packets by default on Tx */
977	ctxt->info.inner_vlan_flags = FIELD_PREP(ICE_AQ_VSI_INNER_VLAN_TX_MODE_M,
978						 ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL);
979	/* SVM - by default bits 3 and 4 in inner_vlan_flags are 0's which
980	 * results in legacy behavior (show VLAN, DEI, and UP) in descriptor.
981	 *
982	 * DVM - leave inner VLAN in packet by default
983	 */
984	if (ice_is_dvm_ena(hw)) {
985		ctxt->info.inner_vlan_flags |=
986			FIELD_PREP(ICE_AQ_VSI_INNER_VLAN_EMODE_M,
987				   ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING);
988		ctxt->info.outer_vlan_flags =
989			FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_TX_MODE_M,
990				   ICE_AQ_VSI_OUTER_VLAN_TX_MODE_ALL);
991		ctxt->info.outer_vlan_flags |=
992			FIELD_PREP(ICE_AQ_VSI_OUTER_TAG_TYPE_M,
993				   ICE_AQ_VSI_OUTER_TAG_VLAN_8100);
994		ctxt->info.outer_vlan_flags |=
995			FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_EMODE_M,
996				   ICE_AQ_VSI_OUTER_VLAN_EMODE_NOTHING);
997	}
998	/* Have 1:1 UP mapping for both ingress/egress tables */
999	table |= ICE_UP_TABLE_TRANSLATE(0, 0);
1000	table |= ICE_UP_TABLE_TRANSLATE(1, 1);
1001	table |= ICE_UP_TABLE_TRANSLATE(2, 2);
1002	table |= ICE_UP_TABLE_TRANSLATE(3, 3);
1003	table |= ICE_UP_TABLE_TRANSLATE(4, 4);
1004	table |= ICE_UP_TABLE_TRANSLATE(5, 5);
1005	table |= ICE_UP_TABLE_TRANSLATE(6, 6);
1006	table |= ICE_UP_TABLE_TRANSLATE(7, 7);
1007	ctxt->info.ingress_table = cpu_to_le32(table);
1008	ctxt->info.egress_table = cpu_to_le32(table);
1009	/* Have 1:1 UP mapping for outer to inner UP table */
1010	ctxt->info.outer_up_table = cpu_to_le32(table);
1011	/* No Outer tag support outer_tag_flags remains to zero */
1012}
1013
1014/**
1015 * ice_vsi_setup_q_map - Setup a VSI queue map
1016 * @vsi: the VSI being configured
1017 * @ctxt: VSI context structure
1018 */
1019static int ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1020{
1021	u16 offset = 0, qmap = 0, tx_count = 0, rx_count = 0, pow = 0;
1022	u16 num_txq_per_tc, num_rxq_per_tc;
1023	u16 qcount_tx = vsi->alloc_txq;
1024	u16 qcount_rx = vsi->alloc_rxq;
1025	u8 netdev_tc = 0;
1026	int i;
1027
1028	if (!vsi->tc_cfg.numtc) {
1029		/* at least TC0 should be enabled by default */
1030		vsi->tc_cfg.numtc = 1;
1031		vsi->tc_cfg.ena_tc = 1;
1032	}
1033
1034	num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
1035	if (!num_rxq_per_tc)
1036		num_rxq_per_tc = 1;
1037	num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
1038	if (!num_txq_per_tc)
1039		num_txq_per_tc = 1;
1040
1041	/* find the (rounded up) power-of-2 of qcount */
1042	pow = (u16)order_base_2(num_rxq_per_tc);
1043
1044	/* TC mapping is a function of the number of Rx queues assigned to the
1045	 * VSI for each traffic class and the offset of these queues.
1046	 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
1047	 * queues allocated to TC0. No:of queues is a power-of-2.
1048	 *
1049	 * If TC is not enabled, the queue offset is set to 0, and allocate one
1050	 * queue, this way, traffic for the given TC will be sent to the default
1051	 * queue.
1052	 *
1053	 * Setup number and offset of Rx queues for all TCs for the VSI
1054	 */
1055	ice_for_each_traffic_class(i) {
1056		if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
1057			/* TC is not enabled */
1058			vsi->tc_cfg.tc_info[i].qoffset = 0;
1059			vsi->tc_cfg.tc_info[i].qcount_rx = 1;
1060			vsi->tc_cfg.tc_info[i].qcount_tx = 1;
1061			vsi->tc_cfg.tc_info[i].netdev_tc = 0;
1062			ctxt->info.tc_mapping[i] = 0;
1063			continue;
1064		}
1065
1066		/* TC is enabled */
1067		vsi->tc_cfg.tc_info[i].qoffset = offset;
1068		vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
1069		vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
1070		vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
1071
1072		qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, offset);
1073		qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
1074		offset += num_rxq_per_tc;
1075		tx_count += num_txq_per_tc;
1076		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1077	}
1078
1079	/* if offset is non-zero, means it is calculated correctly based on
1080	 * enabled TCs for a given VSI otherwise qcount_rx will always
1081	 * be correct and non-zero because it is based off - VSI's
1082	 * allocated Rx queues which is at least 1 (hence qcount_tx will be
1083	 * at least 1)
1084	 */
1085	if (offset)
1086		rx_count = offset;
1087	else
1088		rx_count = num_rxq_per_tc;
1089
1090	if (rx_count > vsi->alloc_rxq) {
1091		dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
1092			rx_count, vsi->alloc_rxq);
1093		return -EINVAL;
1094	}
1095
1096	if (tx_count > vsi->alloc_txq) {
1097		dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
1098			tx_count, vsi->alloc_txq);
1099		return -EINVAL;
1100	}
1101
1102	vsi->num_txq = tx_count;
1103	vsi->num_rxq = rx_count;
1104
1105	if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
1106		dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
1107		/* since there is a chance that num_rxq could have been changed
1108		 * in the above for loop, make num_txq equal to num_rxq.
1109		 */
1110		vsi->num_txq = vsi->num_rxq;
1111	}
1112
1113	/* Rx queue mapping */
1114	ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1115	/* q_mapping buffer holds the info for the first queue allocated for
1116	 * this VSI in the PF space and also the number of queues associated
1117	 * with this VSI.
1118	 */
1119	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
1120	ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
1121
1122	return 0;
1123}
1124
1125/**
1126 * ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
1127 * @ctxt: the VSI context being set
1128 * @vsi: the VSI being configured
1129 */
1130static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1131{
1132	u8 dflt_q_group, dflt_q_prio;
1133	u16 dflt_q, report_q, val;
1134
1135	if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
1136	    vsi->type != ICE_VSI_VF && vsi->type != ICE_VSI_CHNL)
1137		return;
1138
1139	val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
1140	ctxt->info.valid_sections |= cpu_to_le16(val);
1141	dflt_q = 0;
1142	dflt_q_group = 0;
1143	report_q = 0;
1144	dflt_q_prio = 0;
1145
1146	/* enable flow director filtering/programming */
1147	val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
1148	ctxt->info.fd_options = cpu_to_le16(val);
1149	/* max of allocated flow director filters */
1150	ctxt->info.max_fd_fltr_dedicated =
1151			cpu_to_le16(vsi->num_gfltr);
1152	/* max of shared flow director filters any VSI may program */
1153	ctxt->info.max_fd_fltr_shared =
1154			cpu_to_le16(vsi->num_bfltr);
1155	/* default queue index within the VSI of the default FD */
1156	val = FIELD_PREP(ICE_AQ_VSI_FD_DEF_Q_M, dflt_q);
1157	/* target queue or queue group to the FD filter */
1158	val |= FIELD_PREP(ICE_AQ_VSI_FD_DEF_GRP_M, dflt_q_group);
1159	ctxt->info.fd_def_q = cpu_to_le16(val);
1160	/* queue index on which FD filter completion is reported */
1161	val = FIELD_PREP(ICE_AQ_VSI_FD_REPORT_Q_M, report_q);
1162	/* priority of the default qindex action */
1163	val |= FIELD_PREP(ICE_AQ_VSI_FD_DEF_PRIORITY_M, dflt_q_prio);
1164	ctxt->info.fd_report_opt = cpu_to_le16(val);
1165}
1166
1167/**
1168 * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
1169 * @ctxt: the VSI context being set
1170 * @vsi: the VSI being configured
1171 */
1172static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1173{
1174	u8 lut_type, hash_type;
1175	struct device *dev;
1176	struct ice_pf *pf;
1177
1178	pf = vsi->back;
1179	dev = ice_pf_to_dev(pf);
1180
1181	switch (vsi->type) {
1182	case ICE_VSI_CHNL:
1183	case ICE_VSI_PF:
1184		/* PF VSI will inherit RSS instance of PF */
1185		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
1186		break;
1187	case ICE_VSI_VF:
1188		/* VF VSI will gets a small RSS table which is a VSI LUT type */
1189		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
1190		break;
1191	default:
1192		dev_dbg(dev, "Unsupported VSI type %s\n",
1193			ice_vsi_type_str(vsi->type));
1194		return;
1195	}
1196
1197	hash_type = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
1198	vsi->rss_hfunc = hash_type;
1199
1200	ctxt->info.q_opt_rss =
1201		FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_LUT_M, lut_type) |
1202		FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_HASH_M, hash_type);
1203}
1204
1205static void
1206ice_chnl_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1207{
1208	struct ice_pf *pf = vsi->back;
1209	u16 qcount, qmap;
1210	u8 offset = 0;
1211	int pow;
1212
1213	qcount = min_t(int, vsi->num_rxq, pf->num_lan_msix);
1214
1215	pow = order_base_2(qcount);
1216	qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, offset);
1217	qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
1218
1219	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1220	ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1221	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->next_base_q);
1222	ctxt->info.q_mapping[1] = cpu_to_le16(qcount);
1223}
1224
1225/**
1226 * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
1227 * @vsi: VSI to check whether or not VLAN pruning is enabled.
1228 *
1229 * returns true if Rx VLAN pruning is enabled and false otherwise.
1230 */
1231static bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
1232{
1233	return vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1234}
1235
1236/**
1237 * ice_vsi_init - Create and initialize a VSI
1238 * @vsi: the VSI being configured
1239 * @vsi_flags: VSI configuration flags
1240 *
1241 * Set ICE_FLAG_VSI_INIT to initialize a new VSI context, clear it to
1242 * reconfigure an existing context.
1243 *
1244 * This initializes a VSI context depending on the VSI type to be added and
1245 * passes it down to the add_vsi aq command to create a new VSI.
1246 */
1247static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
1248{
1249	struct ice_pf *pf = vsi->back;
1250	struct ice_hw *hw = &pf->hw;
1251	struct ice_vsi_ctx *ctxt;
1252	struct device *dev;
1253	int ret = 0;
1254
1255	dev = ice_pf_to_dev(pf);
1256	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1257	if (!ctxt)
1258		return -ENOMEM;
1259
1260	switch (vsi->type) {
1261	case ICE_VSI_CTRL:
1262	case ICE_VSI_LB:
1263	case ICE_VSI_PF:
1264		ctxt->flags = ICE_AQ_VSI_TYPE_PF;
1265		break;
1266	case ICE_VSI_SWITCHDEV_CTRL:
1267	case ICE_VSI_CHNL:
1268		ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
1269		break;
1270	case ICE_VSI_VF:
1271		ctxt->flags = ICE_AQ_VSI_TYPE_VF;
1272		/* VF number here is the absolute VF number (0-255) */
1273		ctxt->vf_num = vsi->vf->vf_id + hw->func_caps.vf_base_id;
1274		break;
1275	default:
1276		ret = -ENODEV;
1277		goto out;
1278	}
1279
1280	/* Handle VLAN pruning for channel VSI if main VSI has VLAN
1281	 * prune enabled
1282	 */
1283	if (vsi->type == ICE_VSI_CHNL) {
1284		struct ice_vsi *main_vsi;
1285
1286		main_vsi = ice_get_main_vsi(pf);
1287		if (main_vsi && ice_vsi_is_vlan_pruning_ena(main_vsi))
1288			ctxt->info.sw_flags2 |=
1289				ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1290		else
1291			ctxt->info.sw_flags2 &=
1292				~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1293	}
1294
1295	ice_set_dflt_vsi_ctx(hw, ctxt);
1296	if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
1297		ice_set_fd_vsi_ctx(ctxt, vsi);
1298	/* if the switch is in VEB mode, allow VSI loopback */
1299	if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
1300		ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
1301
1302	/* Set LUT type and HASH type if RSS is enabled */
1303	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1304	    vsi->type != ICE_VSI_CTRL) {
1305		ice_set_rss_vsi_ctx(ctxt, vsi);
1306		/* if updating VSI context, make sure to set valid_section:
1307		 * to indicate which section of VSI context being updated
1308		 */
1309		if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1310			ctxt->info.valid_sections |=
1311				cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1312	}
1313
1314	ctxt->info.sw_id = vsi->port_info->sw_id;
1315	if (vsi->type == ICE_VSI_CHNL) {
1316		ice_chnl_vsi_setup_q_map(vsi, ctxt);
1317	} else {
1318		ret = ice_vsi_setup_q_map(vsi, ctxt);
1319		if (ret)
1320			goto out;
1321
1322		if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1323			/* means VSI being updated */
1324			/* must to indicate which section of VSI context are
1325			 * being modified
1326			 */
1327			ctxt->info.valid_sections |=
1328				cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1329	}
1330
1331	/* Allow control frames out of main VSI */
1332	if (vsi->type == ICE_VSI_PF) {
1333		ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1334		ctxt->info.valid_sections |=
1335			cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1336	}
1337
1338	if (vsi_flags & ICE_VSI_FLAG_INIT) {
1339		ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1340		if (ret) {
1341			dev_err(dev, "Add VSI failed, err %d\n", ret);
1342			ret = -EIO;
1343			goto out;
1344		}
1345	} else {
1346		ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1347		if (ret) {
1348			dev_err(dev, "Update VSI failed, err %d\n", ret);
1349			ret = -EIO;
1350			goto out;
1351		}
1352	}
1353
1354	/* keep context for update VSI operations */
1355	vsi->info = ctxt->info;
1356
1357	/* record VSI number returned */
1358	vsi->vsi_num = ctxt->vsi_num;
1359
1360out:
1361	kfree(ctxt);
1362	return ret;
1363}
1364
1365/**
1366 * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1367 * @vsi: the VSI having rings deallocated
1368 */
1369static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1370{
1371	int i;
1372
1373	/* Avoid stale references by clearing map from vector to ring */
1374	if (vsi->q_vectors) {
1375		ice_for_each_q_vector(vsi, i) {
1376			struct ice_q_vector *q_vector = vsi->q_vectors[i];
1377
1378			if (q_vector) {
1379				q_vector->tx.tx_ring = NULL;
1380				q_vector->rx.rx_ring = NULL;
1381			}
1382		}
1383	}
1384
1385	if (vsi->tx_rings) {
1386		ice_for_each_alloc_txq(vsi, i) {
1387			if (vsi->tx_rings[i]) {
1388				kfree_rcu(vsi->tx_rings[i], rcu);
1389				WRITE_ONCE(vsi->tx_rings[i], NULL);
1390			}
1391		}
1392	}
1393	if (vsi->rx_rings) {
1394		ice_for_each_alloc_rxq(vsi, i) {
1395			if (vsi->rx_rings[i]) {
1396				kfree_rcu(vsi->rx_rings[i], rcu);
1397				WRITE_ONCE(vsi->rx_rings[i], NULL);
1398			}
1399		}
1400	}
1401}
1402
1403/**
1404 * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1405 * @vsi: VSI which is having rings allocated
1406 */
1407static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1408{
1409	bool dvm_ena = ice_is_dvm_ena(&vsi->back->hw);
1410	struct ice_pf *pf = vsi->back;
1411	struct device *dev;
1412	u16 i;
1413
1414	dev = ice_pf_to_dev(pf);
1415	/* Allocate Tx rings */
1416	ice_for_each_alloc_txq(vsi, i) {
1417		struct ice_tx_ring *ring;
1418
1419		/* allocate with kzalloc(), free with kfree_rcu() */
1420		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1421
1422		if (!ring)
1423			goto err_out;
1424
1425		ring->q_index = i;
1426		ring->reg_idx = vsi->txq_map[i];
1427		ring->vsi = vsi;
1428		ring->tx_tstamps = &pf->ptp.port.tx;
1429		ring->dev = dev;
1430		ring->count = vsi->num_tx_desc;
1431		ring->txq_teid = ICE_INVAL_TEID;
1432		if (dvm_ena)
1433			ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG2;
1434		else
1435			ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG1;
1436		WRITE_ONCE(vsi->tx_rings[i], ring);
1437	}
1438
1439	/* Allocate Rx rings */
1440	ice_for_each_alloc_rxq(vsi, i) {
1441		struct ice_rx_ring *ring;
1442
1443		/* allocate with kzalloc(), free with kfree_rcu() */
1444		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1445		if (!ring)
1446			goto err_out;
1447
1448		ring->q_index = i;
1449		ring->reg_idx = vsi->rxq_map[i];
1450		ring->vsi = vsi;
1451		ring->netdev = vsi->netdev;
1452		ring->dev = dev;
1453		ring->count = vsi->num_rx_desc;
1454		ring->cached_phctime = pf->ptp.cached_phc_time;
1455		WRITE_ONCE(vsi->rx_rings[i], ring);
1456	}
1457
1458	return 0;
1459
1460err_out:
1461	ice_vsi_clear_rings(vsi);
1462	return -ENOMEM;
1463}
1464
1465/**
1466 * ice_vsi_manage_rss_lut - disable/enable RSS
1467 * @vsi: the VSI being changed
1468 * @ena: boolean value indicating if this is an enable or disable request
1469 *
1470 * In the event of disable request for RSS, this function will zero out RSS
1471 * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1472 * LUT.
1473 */
1474void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1475{
1476	u8 *lut;
1477
1478	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1479	if (!lut)
1480		return;
1481
1482	if (ena) {
1483		if (vsi->rss_lut_user)
1484			memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1485		else
1486			ice_fill_rss_lut(lut, vsi->rss_table_size,
1487					 vsi->rss_size);
1488	}
1489
1490	ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1491	kfree(lut);
1492}
1493
1494/**
1495 * ice_vsi_cfg_crc_strip - Configure CRC stripping for a VSI
1496 * @vsi: VSI to be configured
1497 * @disable: set to true to have FCS / CRC in the frame data
1498 */
1499void ice_vsi_cfg_crc_strip(struct ice_vsi *vsi, bool disable)
1500{
1501	int i;
1502
1503	ice_for_each_rxq(vsi, i)
1504		if (disable)
1505			vsi->rx_rings[i]->flags |= ICE_RX_FLAGS_CRC_STRIP_DIS;
1506		else
1507			vsi->rx_rings[i]->flags &= ~ICE_RX_FLAGS_CRC_STRIP_DIS;
1508}
1509
1510/**
1511 * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1512 * @vsi: VSI to be configured
1513 */
1514int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1515{
1516	struct ice_pf *pf = vsi->back;
1517	struct device *dev;
1518	u8 *lut, *key;
1519	int err;
1520
1521	dev = ice_pf_to_dev(pf);
1522	if (vsi->type == ICE_VSI_PF && vsi->ch_rss_size &&
1523	    (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))) {
1524		vsi->rss_size = min_t(u16, vsi->rss_size, vsi->ch_rss_size);
1525	} else {
1526		vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1527
1528		/* If orig_rss_size is valid and it is less than determined
1529		 * main VSI's rss_size, update main VSI's rss_size to be
1530		 * orig_rss_size so that when tc-qdisc is deleted, main VSI
1531		 * RSS table gets programmed to be correct (whatever it was
1532		 * to begin with (prior to setup-tc for ADQ config)
1533		 */
1534		if (vsi->orig_rss_size && vsi->rss_size < vsi->orig_rss_size &&
1535		    vsi->orig_rss_size <= vsi->num_rxq) {
1536			vsi->rss_size = vsi->orig_rss_size;
1537			/* now orig_rss_size is used, reset it to zero */
1538			vsi->orig_rss_size = 0;
1539		}
1540	}
1541
1542	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1543	if (!lut)
1544		return -ENOMEM;
1545
1546	if (vsi->rss_lut_user)
1547		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1548	else
1549		ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1550
1551	err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1552	if (err) {
1553		dev_err(dev, "set_rss_lut failed, error %d\n", err);
1554		goto ice_vsi_cfg_rss_exit;
1555	}
1556
1557	key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1558	if (!key) {
1559		err = -ENOMEM;
1560		goto ice_vsi_cfg_rss_exit;
1561	}
1562
1563	if (vsi->rss_hkey_user)
1564		memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1565	else
1566		netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1567
1568	err = ice_set_rss_key(vsi, key);
1569	if (err)
1570		dev_err(dev, "set_rss_key failed, error %d\n", err);
1571
1572	kfree(key);
1573ice_vsi_cfg_rss_exit:
1574	kfree(lut);
1575	return err;
1576}
1577
1578/**
1579 * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1580 * @vsi: VSI to be configured
1581 *
1582 * This function will only be called during the VF VSI setup. Upon successful
1583 * completion of package download, this function will configure default RSS
1584 * input sets for VF VSI.
1585 */
1586static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1587{
1588	struct ice_pf *pf = vsi->back;
1589	struct device *dev;
1590	int status;
1591
1592	dev = ice_pf_to_dev(pf);
1593	if (ice_is_safe_mode(pf)) {
1594		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1595			vsi->vsi_num);
1596		return;
1597	}
1598
1599	status = ice_add_avf_rss_cfg(&pf->hw, vsi, ICE_DEFAULT_RSS_HENA);
1600	if (status)
1601		dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
1602			vsi->vsi_num, status);
1603}
1604
1605static const struct ice_rss_hash_cfg default_rss_cfgs[] = {
1606	/* configure RSS for IPv4 with input set IP src/dst */
1607	{ICE_FLOW_SEG_HDR_IPV4, ICE_FLOW_HASH_IPV4, ICE_RSS_ANY_HEADERS, false},
1608	/* configure RSS for IPv6 with input set IPv6 src/dst */
1609	{ICE_FLOW_SEG_HDR_IPV6, ICE_FLOW_HASH_IPV6, ICE_RSS_ANY_HEADERS, false},
1610	/* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1611	{ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4,
1612				ICE_HASH_TCP_IPV4,  ICE_RSS_ANY_HEADERS, false},
1613	/* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1614	{ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4,
1615				ICE_HASH_UDP_IPV4,  ICE_RSS_ANY_HEADERS, false},
1616	/* configure RSS for sctp4 with input set IP src/dst - only support
1617	 * RSS on SCTPv4 on outer headers (non-tunneled)
1618	 */
1619	{ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4,
1620		ICE_HASH_SCTP_IPV4, ICE_RSS_OUTER_HEADERS, false},
1621	/* configure RSS for gtpc4 with input set IPv4 src/dst */
1622	{ICE_FLOW_SEG_HDR_GTPC | ICE_FLOW_SEG_HDR_IPV4,
1623		ICE_FLOW_HASH_IPV4, ICE_RSS_OUTER_HEADERS, false},
1624	/* configure RSS for gtpc4t with input set IPv4 src/dst */
1625	{ICE_FLOW_SEG_HDR_GTPC_TEID | ICE_FLOW_SEG_HDR_IPV4,
1626		ICE_FLOW_HASH_GTP_C_IPV4_TEID, ICE_RSS_OUTER_HEADERS, false},
1627	/* configure RSS for gtpu4 with input set IPv4 src/dst */
1628	{ICE_FLOW_SEG_HDR_GTPU_IP | ICE_FLOW_SEG_HDR_IPV4,
1629		ICE_FLOW_HASH_GTP_U_IPV4_TEID, ICE_RSS_OUTER_HEADERS, false},
1630	/* configure RSS for gtpu4e with input set IPv4 src/dst */
1631	{ICE_FLOW_SEG_HDR_GTPU_EH | ICE_FLOW_SEG_HDR_IPV4,
1632		ICE_FLOW_HASH_GTP_U_IPV4_EH, ICE_RSS_OUTER_HEADERS, false},
1633	/* configure RSS for gtpu4u with input set IPv4 src/dst */
1634	{ ICE_FLOW_SEG_HDR_GTPU_UP | ICE_FLOW_SEG_HDR_IPV4,
1635		ICE_FLOW_HASH_GTP_U_IPV4_UP, ICE_RSS_OUTER_HEADERS, false},
1636	/* configure RSS for gtpu4d with input set IPv4 src/dst */
1637	{ICE_FLOW_SEG_HDR_GTPU_DWN | ICE_FLOW_SEG_HDR_IPV4,
1638		ICE_FLOW_HASH_GTP_U_IPV4_DWN, ICE_RSS_OUTER_HEADERS, false},
1639
1640	/* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1641	{ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6,
1642				ICE_HASH_TCP_IPV6,  ICE_RSS_ANY_HEADERS, false},
1643	/* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1644	{ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6,
1645				ICE_HASH_UDP_IPV6,  ICE_RSS_ANY_HEADERS, false},
1646	/* configure RSS for sctp6 with input set IPv6 src/dst - only support
1647	 * RSS on SCTPv6 on outer headers (non-tunneled)
1648	 */
1649	{ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6,
1650		ICE_HASH_SCTP_IPV6, ICE_RSS_OUTER_HEADERS, false},
1651	/* configure RSS for IPSEC ESP SPI with input set MAC_IPV4_SPI */
1652	{ICE_FLOW_SEG_HDR_ESP,
1653		ICE_FLOW_HASH_ESP_SPI, ICE_RSS_OUTER_HEADERS, false},
1654	/* configure RSS for gtpc6 with input set IPv6 src/dst */
1655	{ICE_FLOW_SEG_HDR_GTPC | ICE_FLOW_SEG_HDR_IPV6,
1656		ICE_FLOW_HASH_IPV6, ICE_RSS_OUTER_HEADERS, false},
1657	/* configure RSS for gtpc6t with input set IPv6 src/dst */
1658	{ICE_FLOW_SEG_HDR_GTPC_TEID | ICE_FLOW_SEG_HDR_IPV6,
1659		ICE_FLOW_HASH_GTP_C_IPV6_TEID, ICE_RSS_OUTER_HEADERS, false},
1660	/* configure RSS for gtpu6 with input set IPv6 src/dst */
1661	{ICE_FLOW_SEG_HDR_GTPU_IP | ICE_FLOW_SEG_HDR_IPV6,
1662		ICE_FLOW_HASH_GTP_U_IPV6_TEID, ICE_RSS_OUTER_HEADERS, false},
1663	/* configure RSS for gtpu6e with input set IPv6 src/dst */
1664	{ICE_FLOW_SEG_HDR_GTPU_EH | ICE_FLOW_SEG_HDR_IPV6,
1665		ICE_FLOW_HASH_GTP_U_IPV6_EH, ICE_RSS_OUTER_HEADERS, false},
1666	/* configure RSS for gtpu6u with input set IPv6 src/dst */
1667	{ ICE_FLOW_SEG_HDR_GTPU_UP | ICE_FLOW_SEG_HDR_IPV6,
1668		ICE_FLOW_HASH_GTP_U_IPV6_UP, ICE_RSS_OUTER_HEADERS, false},
1669	/* configure RSS for gtpu6d with input set IPv6 src/dst */
1670	{ICE_FLOW_SEG_HDR_GTPU_DWN | ICE_FLOW_SEG_HDR_IPV6,
1671		ICE_FLOW_HASH_GTP_U_IPV6_DWN, ICE_RSS_OUTER_HEADERS, false},
1672};
1673
1674/**
1675 * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1676 * @vsi: VSI to be configured
1677 *
1678 * This function will only be called after successful download package call
1679 * during initialization of PF. Since the downloaded package will erase the
1680 * RSS section, this function will configure RSS input sets for different
1681 * flow types. The last profile added has the highest priority, therefore 2
1682 * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1683 * (i.e. IPv4 src/dst TCP src/dst port).
1684 */
1685static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1686{
1687	u16 vsi_num = vsi->vsi_num;
1688	struct ice_pf *pf = vsi->back;
1689	struct ice_hw *hw = &pf->hw;
1690	struct device *dev;
1691	int status;
1692	u32 i;
1693
1694	dev = ice_pf_to_dev(pf);
1695	if (ice_is_safe_mode(pf)) {
1696		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1697			vsi_num);
1698		return;
1699	}
1700	for (i = 0; i < ARRAY_SIZE(default_rss_cfgs); i++) {
1701		const struct ice_rss_hash_cfg *cfg = &default_rss_cfgs[i];
1702
1703		status = ice_add_rss_cfg(hw, vsi, cfg);
1704		if (status)
1705			dev_dbg(dev, "ice_add_rss_cfg failed, addl_hdrs = %x, hash_flds = %llx, hdr_type = %d, symm = %d\n",
1706				cfg->addl_hdrs, cfg->hash_flds,
1707				cfg->hdr_type, cfg->symm);
1708	}
1709}
1710
1711/**
1712 * ice_pf_state_is_nominal - checks the PF for nominal state
1713 * @pf: pointer to PF to check
1714 *
1715 * Check the PF's state for a collection of bits that would indicate
1716 * the PF is in a state that would inhibit normal operation for
1717 * driver functionality.
1718 *
1719 * Returns true if PF is in a nominal state, false otherwise
1720 */
1721bool ice_pf_state_is_nominal(struct ice_pf *pf)
1722{
1723	DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1724
1725	if (!pf)
1726		return false;
1727
1728	bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1729	if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1730		return false;
1731
1732	return true;
1733}
1734
1735/**
1736 * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1737 * @vsi: the VSI to be updated
1738 */
1739void ice_update_eth_stats(struct ice_vsi *vsi)
1740{
1741	struct ice_eth_stats *prev_es, *cur_es;
1742	struct ice_hw *hw = &vsi->back->hw;
1743	struct ice_pf *pf = vsi->back;
1744	u16 vsi_num = vsi->vsi_num;    /* HW absolute index of a VSI */
1745
1746	prev_es = &vsi->eth_stats_prev;
1747	cur_es = &vsi->eth_stats;
1748
1749	if (ice_is_reset_in_progress(pf->state))
1750		vsi->stat_offsets_loaded = false;
1751
1752	ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1753			  &prev_es->rx_bytes, &cur_es->rx_bytes);
1754
1755	ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1756			  &prev_es->rx_unicast, &cur_es->rx_unicast);
1757
1758	ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1759			  &prev_es->rx_multicast, &cur_es->rx_multicast);
1760
1761	ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1762			  &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1763
1764	ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1765			  &prev_es->rx_discards, &cur_es->rx_discards);
1766
1767	ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1768			  &prev_es->tx_bytes, &cur_es->tx_bytes);
1769
1770	ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1771			  &prev_es->tx_unicast, &cur_es->tx_unicast);
1772
1773	ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1774			  &prev_es->tx_multicast, &cur_es->tx_multicast);
1775
1776	ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1777			  &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1778
1779	ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1780			  &prev_es->tx_errors, &cur_es->tx_errors);
1781
1782	vsi->stat_offsets_loaded = true;
1783}
1784
1785/**
1786 * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1787 * @hw: HW pointer
1788 * @pf_q: index of the Rx queue in the PF's queue space
1789 * @rxdid: flexible descriptor RXDID
1790 * @prio: priority for the RXDID for this queue
1791 * @ena_ts: true to enable timestamp and false to disable timestamp
1792 */
1793void
1794ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio,
1795			bool ena_ts)
1796{
1797	int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1798
1799	/* clear any previous values */
1800	regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1801		    QRXFLXP_CNTXT_RXDID_PRIO_M |
1802		    QRXFLXP_CNTXT_TS_M);
1803
1804	regval |= FIELD_PREP(QRXFLXP_CNTXT_RXDID_IDX_M, rxdid);
1805	regval |= FIELD_PREP(QRXFLXP_CNTXT_RXDID_PRIO_M, prio);
1806
1807	if (ena_ts)
1808		/* Enable TimeSync on this queue */
1809		regval |= QRXFLXP_CNTXT_TS_M;
1810
1811	wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1812}
1813
1814/**
1815 * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1816 * @intrl: interrupt rate limit in usecs
1817 * @gran: interrupt rate limit granularity in usecs
1818 *
1819 * This function converts a decimal interrupt rate limit in usecs to the format
1820 * expected by firmware.
1821 */
1822static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1823{
1824	u32 val = intrl / gran;
1825
1826	if (val)
1827		return val | GLINT_RATE_INTRL_ENA_M;
1828	return 0;
1829}
1830
1831/**
1832 * ice_write_intrl - write throttle rate limit to interrupt specific register
1833 * @q_vector: pointer to interrupt specific structure
1834 * @intrl: throttle rate limit in microseconds to write
1835 */
1836void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1837{
1838	struct ice_hw *hw = &q_vector->vsi->back->hw;
1839
1840	wr32(hw, GLINT_RATE(q_vector->reg_idx),
1841	     ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1842}
1843
1844static struct ice_q_vector *ice_pull_qvec_from_rc(struct ice_ring_container *rc)
1845{
1846	switch (rc->type) {
1847	case ICE_RX_CONTAINER:
1848		if (rc->rx_ring)
1849			return rc->rx_ring->q_vector;
1850		break;
1851	case ICE_TX_CONTAINER:
1852		if (rc->tx_ring)
1853			return rc->tx_ring->q_vector;
1854		break;
1855	default:
1856		break;
1857	}
1858
1859	return NULL;
1860}
1861
1862/**
1863 * __ice_write_itr - write throttle rate to register
1864 * @q_vector: pointer to interrupt data structure
1865 * @rc: pointer to ring container
1866 * @itr: throttle rate in microseconds to write
1867 */
1868static void __ice_write_itr(struct ice_q_vector *q_vector,
1869			    struct ice_ring_container *rc, u16 itr)
1870{
1871	struct ice_hw *hw = &q_vector->vsi->back->hw;
1872
1873	wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
1874	     ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
1875}
1876
1877/**
1878 * ice_write_itr - write throttle rate to queue specific register
1879 * @rc: pointer to ring container
1880 * @itr: throttle rate in microseconds to write
1881 */
1882void ice_write_itr(struct ice_ring_container *rc, u16 itr)
1883{
1884	struct ice_q_vector *q_vector;
1885
1886	q_vector = ice_pull_qvec_from_rc(rc);
1887	if (!q_vector)
1888		return;
1889
1890	__ice_write_itr(q_vector, rc, itr);
1891}
1892
1893/**
1894 * ice_set_q_vector_intrl - set up interrupt rate limiting
1895 * @q_vector: the vector to be configured
1896 *
1897 * Interrupt rate limiting is local to the vector, not per-queue so we must
1898 * detect if either ring container has dynamic moderation enabled to decide
1899 * what to set the interrupt rate limit to via INTRL settings. In the case that
1900 * dynamic moderation is disabled on both, write the value with the cached
1901 * setting to make sure INTRL register matches the user visible value.
1902 */
1903void ice_set_q_vector_intrl(struct ice_q_vector *q_vector)
1904{
1905	if (ITR_IS_DYNAMIC(&q_vector->tx) || ITR_IS_DYNAMIC(&q_vector->rx)) {
1906		/* in the case of dynamic enabled, cap each vector to no more
1907		 * than (4 us) 250,000 ints/sec, which allows low latency
1908		 * but still less than 500,000 interrupts per second, which
1909		 * reduces CPU a bit in the case of the lowest latency
1910		 * setting. The 4 here is a value in microseconds.
1911		 */
1912		ice_write_intrl(q_vector, 4);
1913	} else {
1914		ice_write_intrl(q_vector, q_vector->intrl);
1915	}
1916}
1917
1918/**
1919 * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
1920 * @vsi: the VSI being configured
1921 *
1922 * This configures MSIX mode interrupts for the PF VSI, and should not be used
1923 * for the VF VSI.
1924 */
1925void ice_vsi_cfg_msix(struct ice_vsi *vsi)
1926{
1927	struct ice_pf *pf = vsi->back;
1928	struct ice_hw *hw = &pf->hw;
1929	u16 txq = 0, rxq = 0;
1930	int i, q;
1931
1932	ice_for_each_q_vector(vsi, i) {
1933		struct ice_q_vector *q_vector = vsi->q_vectors[i];
1934		u16 reg_idx = q_vector->reg_idx;
1935
1936		ice_cfg_itr(hw, q_vector);
1937
1938		/* Both Transmit Queue Interrupt Cause Control register
1939		 * and Receive Queue Interrupt Cause control register
1940		 * expects MSIX_INDX field to be the vector index
1941		 * within the function space and not the absolute
1942		 * vector index across PF or across device.
1943		 * For SR-IOV VF VSIs queue vector index always starts
1944		 * with 1 since first vector index(0) is used for OICR
1945		 * in VF space. Since VMDq and other PF VSIs are within
1946		 * the PF function space, use the vector index that is
1947		 * tracked for this PF.
1948		 */
1949		for (q = 0; q < q_vector->num_ring_tx; q++) {
1950			ice_cfg_txq_interrupt(vsi, txq, reg_idx,
1951					      q_vector->tx.itr_idx);
1952			txq++;
1953		}
1954
1955		for (q = 0; q < q_vector->num_ring_rx; q++) {
1956			ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
1957					      q_vector->rx.itr_idx);
1958			rxq++;
1959		}
1960	}
1961}
1962
1963/**
1964 * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
1965 * @vsi: the VSI whose rings are to be enabled
1966 *
1967 * Returns 0 on success and a negative value on error
1968 */
1969int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
1970{
1971	return ice_vsi_ctrl_all_rx_rings(vsi, true);
1972}
1973
1974/**
1975 * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
1976 * @vsi: the VSI whose rings are to be disabled
1977 *
1978 * Returns 0 on success and a negative value on error
1979 */
1980int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
1981{
1982	return ice_vsi_ctrl_all_rx_rings(vsi, false);
1983}
1984
1985/**
1986 * ice_vsi_stop_tx_rings - Disable Tx rings
1987 * @vsi: the VSI being configured
1988 * @rst_src: reset source
1989 * @rel_vmvf_num: Relative ID of VF/VM
1990 * @rings: Tx ring array to be stopped
1991 * @count: number of Tx ring array elements
1992 */
1993static int
1994ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
1995		      u16 rel_vmvf_num, struct ice_tx_ring **rings, u16 count)
1996{
1997	u16 q_idx;
1998
1999	if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2000		return -EINVAL;
2001
2002	for (q_idx = 0; q_idx < count; q_idx++) {
2003		struct ice_txq_meta txq_meta = { };
2004		int status;
2005
2006		if (!rings || !rings[q_idx])
2007			return -EINVAL;
2008
2009		ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
2010		status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2011					      rings[q_idx], &txq_meta);
2012
2013		if (status)
2014			return status;
2015	}
2016
2017	return 0;
2018}
2019
2020/**
2021 * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2022 * @vsi: the VSI being configured
2023 * @rst_src: reset source
2024 * @rel_vmvf_num: Relative ID of VF/VM
2025 */
2026int
2027ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2028			  u16 rel_vmvf_num)
2029{
2030	return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings, vsi->num_txq);
2031}
2032
2033/**
2034 * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2035 * @vsi: the VSI being configured
2036 */
2037int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2038{
2039	return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings, vsi->num_xdp_txq);
2040}
2041
2042/**
2043 * ice_vsi_is_rx_queue_active
2044 * @vsi: the VSI being configured
2045 *
2046 * Return true if at least one queue is active.
2047 */
2048bool ice_vsi_is_rx_queue_active(struct ice_vsi *vsi)
2049{
2050	struct ice_pf *pf = vsi->back;
2051	struct ice_hw *hw = &pf->hw;
2052	int i;
2053
2054	ice_for_each_rxq(vsi, i) {
2055		u32 rx_reg;
2056		int pf_q;
2057
2058		pf_q = vsi->rxq_map[i];
2059		rx_reg = rd32(hw, QRX_CTRL(pf_q));
2060		if (rx_reg & QRX_CTRL_QENA_STAT_M)
2061			return true;
2062	}
2063
2064	return false;
2065}
2066
2067static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2068{
2069	if (!test_bit(ICE_FLAG_DCB_ENA, vsi->back->flags)) {
2070		vsi->tc_cfg.ena_tc = ICE_DFLT_TRAFFIC_CLASS;
2071		vsi->tc_cfg.numtc = 1;
2072		return;
2073	}
2074
2075	/* set VSI TC information based on DCB config */
2076	ice_vsi_set_dcb_tc_cfg(vsi);
2077}
2078
2079/**
2080 * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2081 * @vsi: the VSI being configured
2082 * @tx: bool to determine Tx or Rx rule
2083 * @create: bool to determine create or remove Rule
2084 */
2085void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2086{
2087	int (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2088			enum ice_sw_fwd_act_type act);
2089	struct ice_pf *pf = vsi->back;
2090	struct device *dev;
2091	int status;
2092
2093	dev = ice_pf_to_dev(pf);
2094	eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2095
2096	if (tx) {
2097		status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2098				  ICE_DROP_PACKET);
2099	} else {
2100		if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2101			status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2102							  create);
2103		} else {
2104			status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2105					  ICE_FWD_TO_VSI);
2106		}
2107	}
2108
2109	if (status)
2110		dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %d\n",
2111			create ? "adding" : "removing", tx ? "TX" : "RX",
2112			vsi->vsi_num, status);
2113}
2114
2115/**
2116 * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2117 * @vsi: pointer to the VSI
2118 *
2119 * This function will allocate new scheduler aggregator now if needed and will
2120 * move specified VSI into it.
2121 */
2122static void ice_set_agg_vsi(struct ice_vsi *vsi)
2123{
2124	struct device *dev = ice_pf_to_dev(vsi->back);
2125	struct ice_agg_node *agg_node_iter = NULL;
2126	u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2127	struct ice_agg_node *agg_node = NULL;
2128	int node_offset, max_agg_nodes = 0;
2129	struct ice_port_info *port_info;
2130	struct ice_pf *pf = vsi->back;
2131	u32 agg_node_id_start = 0;
2132	int status;
2133
2134	/* create (as needed) scheduler aggregator node and move VSI into
2135	 * corresponding aggregator node
2136	 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2137	 * - VF aggregator nodes will contain VF VSI
2138	 */
2139	port_info = pf->hw.port_info;
2140	if (!port_info)
2141		return;
2142
2143	switch (vsi->type) {
2144	case ICE_VSI_CTRL:
2145	case ICE_VSI_CHNL:
2146	case ICE_VSI_LB:
2147	case ICE_VSI_PF:
2148	case ICE_VSI_SWITCHDEV_CTRL:
2149		max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2150		agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2151		agg_node_iter = &pf->pf_agg_node[0];
2152		break;
2153	case ICE_VSI_VF:
2154		/* user can create 'n' VFs on a given PF, but since max children
2155		 * per aggregator node can be only 64. Following code handles
2156		 * aggregator(s) for VF VSIs, either selects a agg_node which
2157		 * was already created provided num_vsis < 64, otherwise
2158		 * select next available node, which will be created
2159		 */
2160		max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2161		agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2162		agg_node_iter = &pf->vf_agg_node[0];
2163		break;
2164	default:
2165		/* other VSI type, handle later if needed */
2166		dev_dbg(dev, "unexpected VSI type %s\n",
2167			ice_vsi_type_str(vsi->type));
2168		return;
2169	}
2170
2171	/* find the appropriate aggregator node */
2172	for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2173		/* see if we can find space in previously created
2174		 * node if num_vsis < 64, otherwise skip
2175		 */
2176		if (agg_node_iter->num_vsis &&
2177		    agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2178			agg_node_iter++;
2179			continue;
2180		}
2181
2182		if (agg_node_iter->valid &&
2183		    agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2184			agg_id = agg_node_iter->agg_id;
2185			agg_node = agg_node_iter;
2186			break;
2187		}
2188
2189		/* find unclaimed agg_id */
2190		if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2191			agg_id = node_offset + agg_node_id_start;
2192			agg_node = agg_node_iter;
2193			break;
2194		}
2195		/* move to next agg_node */
2196		agg_node_iter++;
2197	}
2198
2199	if (!agg_node)
2200		return;
2201
2202	/* if selected aggregator node was not created, create it */
2203	if (!agg_node->valid) {
2204		status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2205				     (u8)vsi->tc_cfg.ena_tc);
2206		if (status) {
2207			dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2208				agg_id);
2209			return;
2210		}
2211		/* aggregator node is created, store the needed info */
2212		agg_node->valid = true;
2213		agg_node->agg_id = agg_id;
2214	}
2215
2216	/* move VSI to corresponding aggregator node */
2217	status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2218				     (u8)vsi->tc_cfg.ena_tc);
2219	if (status) {
2220		dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2221			vsi->idx, agg_id);
2222		return;
2223	}
2224
2225	/* keep active children count for aggregator node */
2226	agg_node->num_vsis++;
2227
2228	/* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2229	 * to aggregator node
2230	 */
2231	vsi->agg_node = agg_node;
2232	dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2233		vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2234		vsi->agg_node->num_vsis);
2235}
2236
2237static int ice_vsi_cfg_tc_lan(struct ice_pf *pf, struct ice_vsi *vsi)
2238{
2239	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2240	struct device *dev = ice_pf_to_dev(pf);
2241	int ret, i;
2242
2243	/* configure VSI nodes based on number of queues and TC's */
2244	ice_for_each_traffic_class(i) {
2245		if (!(vsi->tc_cfg.ena_tc & BIT(i)))
2246			continue;
2247
2248		if (vsi->type == ICE_VSI_CHNL) {
2249			if (!vsi->alloc_txq && vsi->num_txq)
2250				max_txqs[i] = vsi->num_txq;
2251			else
2252				max_txqs[i] = pf->num_lan_tx;
2253		} else {
2254			max_txqs[i] = vsi->alloc_txq;
2255		}
2256
2257		if (vsi->type == ICE_VSI_PF)
2258			max_txqs[i] += vsi->num_xdp_txq;
2259	}
2260
2261	dev_dbg(dev, "vsi->tc_cfg.ena_tc = %d\n", vsi->tc_cfg.ena_tc);
2262	ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2263			      max_txqs);
2264	if (ret) {
2265		dev_err(dev, "VSI %d failed lan queue config, error %d\n",
2266			vsi->vsi_num, ret);
2267		return ret;
2268	}
2269
2270	return 0;
2271}
2272
2273/**
2274 * ice_vsi_cfg_def - configure default VSI based on the type
2275 * @vsi: pointer to VSI
2276 * @params: the parameters to configure this VSI with
2277 */
2278static int
2279ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2280{
2281	struct device *dev = ice_pf_to_dev(vsi->back);
2282	struct ice_pf *pf = vsi->back;
2283	int ret;
2284
2285	vsi->vsw = pf->first_sw;
2286
2287	ret = ice_vsi_alloc_def(vsi, params->ch);
2288	if (ret)
2289		return ret;
2290
2291	/* allocate memory for Tx/Rx ring stat pointers */
2292	ret = ice_vsi_alloc_stat_arrays(vsi);
2293	if (ret)
2294		goto unroll_vsi_alloc;
2295
2296	ice_alloc_fd_res(vsi);
2297
2298	ret = ice_vsi_get_qs(vsi);
2299	if (ret) {
2300		dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2301			vsi->idx);
2302		goto unroll_vsi_alloc_stat;
2303	}
2304
2305	/* set RSS capabilities */
2306	ice_vsi_set_rss_params(vsi);
2307
2308	/* set TC configuration */
2309	ice_vsi_set_tc_cfg(vsi);
2310
2311	/* create the VSI */
2312	ret = ice_vsi_init(vsi, params->flags);
2313	if (ret)
2314		goto unroll_get_qs;
2315
2316	ice_vsi_init_vlan_ops(vsi);
2317
2318	switch (vsi->type) {
2319	case ICE_VSI_CTRL:
2320	case ICE_VSI_SWITCHDEV_CTRL:
2321	case ICE_VSI_PF:
2322		ret = ice_vsi_alloc_q_vectors(vsi);
2323		if (ret)
2324			goto unroll_vsi_init;
2325
2326		ret = ice_vsi_alloc_rings(vsi);
2327		if (ret)
2328			goto unroll_vector_base;
2329
2330		ret = ice_vsi_alloc_ring_stats(vsi);
2331		if (ret)
2332			goto unroll_vector_base;
2333
2334		ice_vsi_map_rings_to_vectors(vsi);
2335
2336		/* Associate q_vector rings to napi */
2337		ice_vsi_set_napi_queues(vsi);
2338
2339		vsi->stat_offsets_loaded = false;
2340
2341		if (ice_is_xdp_ena_vsi(vsi)) {
2342			ret = ice_vsi_determine_xdp_res(vsi);
2343			if (ret)
2344				goto unroll_vector_base;
2345			ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog);
2346			if (ret)
2347				goto unroll_vector_base;
2348		}
2349
2350		/* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2351		if (vsi->type != ICE_VSI_CTRL)
2352			/* Do not exit if configuring RSS had an issue, at
2353			 * least receive traffic on first queue. Hence no
2354			 * need to capture return value
2355			 */
2356			if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2357				ice_vsi_cfg_rss_lut_key(vsi);
2358				ice_vsi_set_rss_flow_fld(vsi);
2359			}
2360		ice_init_arfs(vsi);
2361		break;
2362	case ICE_VSI_CHNL:
2363		if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2364			ice_vsi_cfg_rss_lut_key(vsi);
2365			ice_vsi_set_rss_flow_fld(vsi);
2366		}
2367		break;
2368	case ICE_VSI_VF:
2369		/* VF driver will take care of creating netdev for this type and
2370		 * map queues to vectors through Virtchnl, PF driver only
2371		 * creates a VSI and corresponding structures for bookkeeping
2372		 * purpose
2373		 */
2374		ret = ice_vsi_alloc_q_vectors(vsi);
2375		if (ret)
2376			goto unroll_vsi_init;
2377
2378		ret = ice_vsi_alloc_rings(vsi);
2379		if (ret)
2380			goto unroll_alloc_q_vector;
2381
2382		ret = ice_vsi_alloc_ring_stats(vsi);
2383		if (ret)
2384			goto unroll_vector_base;
2385
2386		vsi->stat_offsets_loaded = false;
2387
2388		/* Do not exit if configuring RSS had an issue, at least
2389		 * receive traffic on first queue. Hence no need to capture
2390		 * return value
2391		 */
2392		if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2393			ice_vsi_cfg_rss_lut_key(vsi);
2394			ice_vsi_set_vf_rss_flow_fld(vsi);
2395		}
2396		break;
2397	case ICE_VSI_LB:
2398		ret = ice_vsi_alloc_rings(vsi);
2399		if (ret)
2400			goto unroll_vsi_init;
2401
2402		ret = ice_vsi_alloc_ring_stats(vsi);
2403		if (ret)
2404			goto unroll_vector_base;
2405
2406		break;
2407	default:
2408		/* clean up the resources and exit */
2409		ret = -EINVAL;
2410		goto unroll_vsi_init;
2411	}
2412
2413	return 0;
2414
2415unroll_vector_base:
2416	/* reclaim SW interrupts back to the common pool */
2417unroll_alloc_q_vector:
2418	ice_vsi_free_q_vectors(vsi);
2419unroll_vsi_init:
2420	ice_vsi_delete_from_hw(vsi);
2421unroll_get_qs:
2422	ice_vsi_put_qs(vsi);
2423unroll_vsi_alloc_stat:
2424	ice_vsi_free_stats(vsi);
2425unroll_vsi_alloc:
2426	ice_vsi_free_arrays(vsi);
2427	return ret;
2428}
2429
2430/**
2431 * ice_vsi_cfg - configure a previously allocated VSI
2432 * @vsi: pointer to VSI
2433 * @params: parameters used to configure this VSI
2434 */
2435int ice_vsi_cfg(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2436{
2437	struct ice_pf *pf = vsi->back;
2438	int ret;
2439
2440	if (WARN_ON(params->type == ICE_VSI_VF && !params->vf))
2441		return -EINVAL;
2442
2443	vsi->type = params->type;
2444	vsi->port_info = params->pi;
2445
2446	/* For VSIs which don't have a connected VF, this will be NULL */
2447	vsi->vf = params->vf;
2448
2449	ret = ice_vsi_cfg_def(vsi, params);
2450	if (ret)
2451		return ret;
2452
2453	ret = ice_vsi_cfg_tc_lan(vsi->back, vsi);
2454	if (ret)
2455		ice_vsi_decfg(vsi);
2456
2457	if (vsi->type == ICE_VSI_CTRL) {
2458		if (vsi->vf) {
2459			WARN_ON(vsi->vf->ctrl_vsi_idx != ICE_NO_VSI);
2460			vsi->vf->ctrl_vsi_idx = vsi->idx;
2461		} else {
2462			WARN_ON(pf->ctrl_vsi_idx != ICE_NO_VSI);
2463			pf->ctrl_vsi_idx = vsi->idx;
2464		}
2465	}
2466
2467	return ret;
2468}
2469
2470/**
2471 * ice_vsi_decfg - remove all VSI configuration
2472 * @vsi: pointer to VSI
2473 */
2474void ice_vsi_decfg(struct ice_vsi *vsi)
2475{
2476	struct ice_pf *pf = vsi->back;
2477	int err;
2478
2479	/* The Rx rule will only exist to remove if the LLDP FW
2480	 * engine is currently stopped
2481	 */
2482	if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF &&
2483	    !test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2484		ice_cfg_sw_lldp(vsi, false, false);
2485
2486	ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
2487	err = ice_rm_vsi_rdma_cfg(vsi->port_info, vsi->idx);
2488	if (err)
2489		dev_err(ice_pf_to_dev(pf), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
2490			vsi->vsi_num, err);
2491
2492	if (ice_is_xdp_ena_vsi(vsi))
2493		/* return value check can be skipped here, it always returns
2494		 * 0 if reset is in progress
2495		 */
2496		ice_destroy_xdp_rings(vsi);
2497
2498	ice_vsi_clear_rings(vsi);
2499	ice_vsi_free_q_vectors(vsi);
2500	ice_vsi_put_qs(vsi);
2501	ice_vsi_free_arrays(vsi);
2502
2503	/* SR-IOV determines needed MSIX resources all at once instead of per
2504	 * VSI since when VFs are spawned we know how many VFs there are and how
2505	 * many interrupts each VF needs. SR-IOV MSIX resources are also
2506	 * cleared in the same manner.
2507	 */
2508
2509	if (vsi->type == ICE_VSI_VF &&
2510	    vsi->agg_node && vsi->agg_node->valid)
2511		vsi->agg_node->num_vsis--;
2512}
2513
2514/**
2515 * ice_vsi_setup - Set up a VSI by a given type
2516 * @pf: board private structure
2517 * @params: parameters to use when creating the VSI
2518 *
2519 * This allocates the sw VSI structure and its queue resources.
2520 *
2521 * Returns pointer to the successfully allocated and configured VSI sw struct on
2522 * success, NULL on failure.
2523 */
2524struct ice_vsi *
2525ice_vsi_setup(struct ice_pf *pf, struct ice_vsi_cfg_params *params)
2526{
2527	struct device *dev = ice_pf_to_dev(pf);
2528	struct ice_vsi *vsi;
2529	int ret;
2530
2531	/* ice_vsi_setup can only initialize a new VSI, and we must have
2532	 * a port_info structure for it.
2533	 */
2534	if (WARN_ON(!(params->flags & ICE_VSI_FLAG_INIT)) ||
2535	    WARN_ON(!params->pi))
2536		return NULL;
2537
2538	vsi = ice_vsi_alloc(pf);
2539	if (!vsi) {
2540		dev_err(dev, "could not allocate VSI\n");
2541		return NULL;
2542	}
2543
2544	ret = ice_vsi_cfg(vsi, params);
2545	if (ret)
2546		goto err_vsi_cfg;
2547
2548	/* Add switch rule to drop all Tx Flow Control Frames, of look up
2549	 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2550	 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2551	 * The rule is added once for PF VSI in order to create appropriate
2552	 * recipe, since VSI/VSI list is ignored with drop action...
2553	 * Also add rules to handle LLDP Tx packets.  Tx LLDP packets need to
2554	 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2555	 * settings in the HW.
2556	 */
2557	if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF) {
2558		ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2559				 ICE_DROP_PACKET);
2560		ice_cfg_sw_lldp(vsi, true, true);
2561	}
2562
2563	if (!vsi->agg_node)
2564		ice_set_agg_vsi(vsi);
2565
2566	return vsi;
2567
2568err_vsi_cfg:
2569	ice_vsi_free(vsi);
2570
2571	return NULL;
2572}
2573
2574/**
2575 * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2576 * @vsi: the VSI being cleaned up
2577 */
2578static void ice_vsi_release_msix(struct ice_vsi *vsi)
2579{
2580	struct ice_pf *pf = vsi->back;
2581	struct ice_hw *hw = &pf->hw;
2582	u32 txq = 0;
2583	u32 rxq = 0;
2584	int i, q;
2585
2586	ice_for_each_q_vector(vsi, i) {
2587		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2588
2589		ice_write_intrl(q_vector, 0);
2590		for (q = 0; q < q_vector->num_ring_tx; q++) {
2591			ice_write_itr(&q_vector->tx, 0);
2592			wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2593			if (ice_is_xdp_ena_vsi(vsi)) {
2594				u32 xdp_txq = txq + vsi->num_xdp_txq;
2595
2596				wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2597			}
2598			txq++;
2599		}
2600
2601		for (q = 0; q < q_vector->num_ring_rx; q++) {
2602			ice_write_itr(&q_vector->rx, 0);
2603			wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2604			rxq++;
2605		}
2606	}
2607
2608	ice_flush(hw);
2609}
2610
2611/**
2612 * ice_vsi_free_irq - Free the IRQ association with the OS
2613 * @vsi: the VSI being configured
2614 */
2615void ice_vsi_free_irq(struct ice_vsi *vsi)
2616{
2617	struct ice_pf *pf = vsi->back;
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_free_cpu_rx_rmap(vsi);
2629
2630	ice_for_each_q_vector(vsi, i) {
2631		int irq_num;
2632
2633		irq_num = vsi->q_vectors[i]->irq.virq;
2634
2635		/* free only the irqs that were actually requested */
2636		if (!vsi->q_vectors[i] ||
2637		    !(vsi->q_vectors[i]->num_ring_tx ||
2638		      vsi->q_vectors[i]->num_ring_rx))
2639			continue;
2640
2641		/* clear the affinity notifier in the IRQ descriptor */
2642		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2643			irq_set_affinity_notifier(irq_num, NULL);
2644
2645		/* clear the affinity_mask in the IRQ descriptor */
2646		irq_set_affinity_hint(irq_num, NULL);
2647		synchronize_irq(irq_num);
2648		devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2649	}
2650}
2651
2652/**
2653 * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2654 * @vsi: the VSI having resources freed
2655 */
2656void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2657{
2658	int i;
2659
2660	if (!vsi->tx_rings)
2661		return;
2662
2663	ice_for_each_txq(vsi, i)
2664		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2665			ice_free_tx_ring(vsi->tx_rings[i]);
2666}
2667
2668/**
2669 * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2670 * @vsi: the VSI having resources freed
2671 */
2672void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2673{
2674	int i;
2675
2676	if (!vsi->rx_rings)
2677		return;
2678
2679	ice_for_each_rxq(vsi, i)
2680		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2681			ice_free_rx_ring(vsi->rx_rings[i]);
2682}
2683
2684/**
2685 * ice_vsi_close - Shut down a VSI
2686 * @vsi: the VSI being shut down
2687 */
2688void ice_vsi_close(struct ice_vsi *vsi)
2689{
2690	if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2691		ice_down(vsi);
2692
2693	ice_vsi_free_irq(vsi);
2694	ice_vsi_free_tx_rings(vsi);
2695	ice_vsi_free_rx_rings(vsi);
2696}
2697
2698/**
2699 * ice_ena_vsi - resume a VSI
2700 * @vsi: the VSI being resume
2701 * @locked: is the rtnl_lock already held
2702 */
2703int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2704{
2705	int err = 0;
2706
2707	if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2708		return 0;
2709
2710	clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2711
2712	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2713		if (netif_running(vsi->netdev)) {
2714			if (!locked)
2715				rtnl_lock();
2716
2717			err = ice_open_internal(vsi->netdev);
2718
2719			if (!locked)
2720				rtnl_unlock();
2721		}
2722	} else if (vsi->type == ICE_VSI_CTRL) {
2723		err = ice_vsi_open_ctrl(vsi);
2724	}
2725
2726	return err;
2727}
2728
2729/**
2730 * ice_dis_vsi - pause a VSI
2731 * @vsi: the VSI being paused
2732 * @locked: is the rtnl_lock already held
2733 */
2734void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2735{
2736	if (test_bit(ICE_VSI_DOWN, vsi->state))
2737		return;
2738
2739	set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2740
2741	if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2742		if (netif_running(vsi->netdev)) {
2743			if (!locked)
2744				rtnl_lock();
2745
2746			ice_vsi_close(vsi);
2747
2748			if (!locked)
2749				rtnl_unlock();
2750		} else {
2751			ice_vsi_close(vsi);
2752		}
2753	} else if (vsi->type == ICE_VSI_CTRL ||
2754		   vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
2755		ice_vsi_close(vsi);
2756	}
2757}
2758
2759/**
2760 * __ice_queue_set_napi - Set the napi instance for the queue
2761 * @dev: device to which NAPI and queue belong
2762 * @queue_index: Index of queue
2763 * @type: queue type as RX or TX
2764 * @napi: NAPI context
2765 * @locked: is the rtnl_lock already held
2766 *
2767 * Set the napi instance for the queue. Caller indicates the lock status.
2768 */
2769static void
2770__ice_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2771		     enum netdev_queue_type type, struct napi_struct *napi,
2772		     bool locked)
2773{
2774	if (!locked)
2775		rtnl_lock();
2776	netif_queue_set_napi(dev, queue_index, type, napi);
2777	if (!locked)
2778		rtnl_unlock();
2779}
2780
2781/**
2782 * ice_queue_set_napi - Set the napi instance for the queue
2783 * @vsi: VSI being configured
2784 * @queue_index: Index of queue
2785 * @type: queue type as RX or TX
2786 * @napi: NAPI context
2787 *
2788 * Set the napi instance for the queue. The rtnl lock state is derived from the
2789 * execution path.
2790 */
2791void
2792ice_queue_set_napi(struct ice_vsi *vsi, unsigned int queue_index,
2793		   enum netdev_queue_type type, struct napi_struct *napi)
2794{
2795	struct ice_pf *pf = vsi->back;
2796
2797	if (!vsi->netdev)
2798		return;
2799
2800	if (current_work() == &pf->serv_task ||
2801	    test_bit(ICE_PREPARED_FOR_RESET, pf->state) ||
2802	    test_bit(ICE_DOWN, pf->state) ||
2803	    test_bit(ICE_SUSPENDED, pf->state))
2804		__ice_queue_set_napi(vsi->netdev, queue_index, type, napi,
2805				     false);
2806	else
2807		__ice_queue_set_napi(vsi->netdev, queue_index, type, napi,
2808				     true);
2809}
2810
2811/**
2812 * __ice_q_vector_set_napi_queues - Map queue[s] associated with the napi
2813 * @q_vector: q_vector pointer
2814 * @locked: is the rtnl_lock already held
2815 *
2816 * Associate the q_vector napi with all the queue[s] on the vector.
2817 * Caller indicates the lock status.
2818 */
2819void __ice_q_vector_set_napi_queues(struct ice_q_vector *q_vector, bool locked)
2820{
2821	struct ice_rx_ring *rx_ring;
2822	struct ice_tx_ring *tx_ring;
2823
2824	ice_for_each_rx_ring(rx_ring, q_vector->rx)
2825		__ice_queue_set_napi(q_vector->vsi->netdev, rx_ring->q_index,
2826				     NETDEV_QUEUE_TYPE_RX, &q_vector->napi,
2827				     locked);
2828
2829	ice_for_each_tx_ring(tx_ring, q_vector->tx)
2830		__ice_queue_set_napi(q_vector->vsi->netdev, tx_ring->q_index,
2831				     NETDEV_QUEUE_TYPE_TX, &q_vector->napi,
2832				     locked);
2833	/* Also set the interrupt number for the NAPI */
2834	netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq);
2835}
2836
2837/**
2838 * ice_q_vector_set_napi_queues - Map queue[s] associated with the napi
2839 * @q_vector: q_vector pointer
2840 *
2841 * Associate the q_vector napi with all the queue[s] on the vector
2842 */
2843void ice_q_vector_set_napi_queues(struct ice_q_vector *q_vector)
2844{
2845	struct ice_rx_ring *rx_ring;
2846	struct ice_tx_ring *tx_ring;
2847
2848	ice_for_each_rx_ring(rx_ring, q_vector->rx)
2849		ice_queue_set_napi(q_vector->vsi, rx_ring->q_index,
2850				   NETDEV_QUEUE_TYPE_RX, &q_vector->napi);
2851
2852	ice_for_each_tx_ring(tx_ring, q_vector->tx)
2853		ice_queue_set_napi(q_vector->vsi, tx_ring->q_index,
2854				   NETDEV_QUEUE_TYPE_TX, &q_vector->napi);
2855	/* Also set the interrupt number for the NAPI */
2856	netif_napi_set_irq(&q_vector->napi, q_vector->irq.virq);
2857}
2858
2859/**
2860 * ice_vsi_set_napi_queues
2861 * @vsi: VSI pointer
2862 *
2863 * Associate queue[s] with napi for all vectors
2864 */
2865void ice_vsi_set_napi_queues(struct ice_vsi *vsi)
2866{
2867	int i;
2868
2869	if (!vsi->netdev)
2870		return;
2871
2872	ice_for_each_q_vector(vsi, i)
2873		ice_q_vector_set_napi_queues(vsi->q_vectors[i]);
2874}
2875
2876/**
2877 * ice_vsi_release - Delete a VSI and free its resources
2878 * @vsi: the VSI being removed
2879 *
2880 * Returns 0 on success or < 0 on error
2881 */
2882int ice_vsi_release(struct ice_vsi *vsi)
2883{
2884	struct ice_pf *pf;
2885
2886	if (!vsi->back)
2887		return -ENODEV;
2888	pf = vsi->back;
2889
2890	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2891		ice_rss_clean(vsi);
2892
2893	ice_vsi_close(vsi);
2894	ice_vsi_decfg(vsi);
2895
2896	/* retain SW VSI data structure since it is needed to unregister and
2897	 * free VSI netdev when PF is not in reset recovery pending state,\
2898	 * for ex: during rmmod.
2899	 */
2900	if (!ice_is_reset_in_progress(pf->state))
2901		ice_vsi_delete(vsi);
2902
2903	return 0;
2904}
2905
2906/**
2907 * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
2908 * @vsi: VSI connected with q_vectors
2909 * @coalesce: array of struct with stored coalesce
2910 *
2911 * Returns array size.
2912 */
2913static int
2914ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
2915			     struct ice_coalesce_stored *coalesce)
2916{
2917	int i;
2918
2919	ice_for_each_q_vector(vsi, i) {
2920		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2921
2922		coalesce[i].itr_tx = q_vector->tx.itr_settings;
2923		coalesce[i].itr_rx = q_vector->rx.itr_settings;
2924		coalesce[i].intrl = q_vector->intrl;
2925
2926		if (i < vsi->num_txq)
2927			coalesce[i].tx_valid = true;
2928		if (i < vsi->num_rxq)
2929			coalesce[i].rx_valid = true;
2930	}
2931
2932	return vsi->num_q_vectors;
2933}
2934
2935/**
2936 * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
2937 * @vsi: VSI connected with q_vectors
2938 * @coalesce: pointer to array of struct with stored coalesce
2939 * @size: size of coalesce array
2940 *
2941 * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
2942 * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
2943 * to default value.
2944 */
2945static void
2946ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
2947			     struct ice_coalesce_stored *coalesce, int size)
2948{
2949	struct ice_ring_container *rc;
2950	int i;
2951
2952	if ((size && !coalesce) || !vsi)
2953		return;
2954
2955	/* There are a couple of cases that have to be handled here:
2956	 *   1. The case where the number of queue vectors stays the same, but
2957	 *      the number of Tx or Rx rings changes (the first for loop)
2958	 *   2. The case where the number of queue vectors increased (the
2959	 *      second for loop)
2960	 */
2961	for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
2962		/* There are 2 cases to handle here and they are the same for
2963		 * both Tx and Rx:
2964		 *   if the entry was valid previously (coalesce[i].[tr]x_valid
2965		 *   and the loop variable is less than the number of rings
2966		 *   allocated, then write the previous values
2967		 *
2968		 *   if the entry was not valid previously, but the number of
2969		 *   rings is less than are allocated (this means the number of
2970		 *   rings increased from previously), then write out the
2971		 *   values in the first element
2972		 *
2973		 *   Also, always write the ITR, even if in ITR_IS_DYNAMIC
2974		 *   as there is no harm because the dynamic algorithm
2975		 *   will just overwrite.
2976		 */
2977		if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
2978			rc = &vsi->q_vectors[i]->rx;
2979			rc->itr_settings = coalesce[i].itr_rx;
2980			ice_write_itr(rc, rc->itr_setting);
2981		} else if (i < vsi->alloc_rxq) {
2982			rc = &vsi->q_vectors[i]->rx;
2983			rc->itr_settings = coalesce[0].itr_rx;
2984			ice_write_itr(rc, rc->itr_setting);
2985		}
2986
2987		if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
2988			rc = &vsi->q_vectors[i]->tx;
2989			rc->itr_settings = coalesce[i].itr_tx;
2990			ice_write_itr(rc, rc->itr_setting);
2991		} else if (i < vsi->alloc_txq) {
2992			rc = &vsi->q_vectors[i]->tx;
2993			rc->itr_settings = coalesce[0].itr_tx;
2994			ice_write_itr(rc, rc->itr_setting);
2995		}
2996
2997		vsi->q_vectors[i]->intrl = coalesce[i].intrl;
2998		ice_set_q_vector_intrl(vsi->q_vectors[i]);
2999	}
3000
3001	/* the number of queue vectors increased so write whatever is in
3002	 * the first element
3003	 */
3004	for (; i < vsi->num_q_vectors; i++) {
3005		/* transmit */
3006		rc = &vsi->q_vectors[i]->tx;
3007		rc->itr_settings = coalesce[0].itr_tx;
3008		ice_write_itr(rc, rc->itr_setting);
3009
3010		/* receive */
3011		rc = &vsi->q_vectors[i]->rx;
3012		rc->itr_settings = coalesce[0].itr_rx;
3013		ice_write_itr(rc, rc->itr_setting);
3014
3015		vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3016		ice_set_q_vector_intrl(vsi->q_vectors[i]);
3017	}
3018}
3019
3020/**
3021 * ice_vsi_realloc_stat_arrays - Frees unused stat structures or alloc new ones
3022 * @vsi: VSI pointer
3023 */
3024static int
3025ice_vsi_realloc_stat_arrays(struct ice_vsi *vsi)
3026{
3027	u16 req_txq = vsi->req_txq ? vsi->req_txq : vsi->alloc_txq;
3028	u16 req_rxq = vsi->req_rxq ? vsi->req_rxq : vsi->alloc_rxq;
3029	struct ice_ring_stats **tx_ring_stats;
3030	struct ice_ring_stats **rx_ring_stats;
3031	struct ice_vsi_stats *vsi_stat;
3032	struct ice_pf *pf = vsi->back;
3033	u16 prev_txq = vsi->alloc_txq;
3034	u16 prev_rxq = vsi->alloc_rxq;
3035	int i;
3036
3037	vsi_stat = pf->vsi_stats[vsi->idx];
3038
3039	if (req_txq < prev_txq) {
3040		for (i = req_txq; i < prev_txq; i++) {
3041			if (vsi_stat->tx_ring_stats[i]) {
3042				kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
3043				WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
3044			}
3045		}
3046	}
3047
3048	tx_ring_stats = vsi_stat->tx_ring_stats;
3049	vsi_stat->tx_ring_stats =
3050		krealloc_array(vsi_stat->tx_ring_stats, req_txq,
3051			       sizeof(*vsi_stat->tx_ring_stats),
3052			       GFP_KERNEL | __GFP_ZERO);
3053	if (!vsi_stat->tx_ring_stats) {
3054		vsi_stat->tx_ring_stats = tx_ring_stats;
3055		return -ENOMEM;
3056	}
3057
3058	if (req_rxq < prev_rxq) {
3059		for (i = req_rxq; i < prev_rxq; i++) {
3060			if (vsi_stat->rx_ring_stats[i]) {
3061				kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
3062				WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
3063			}
3064		}
3065	}
3066
3067	rx_ring_stats = vsi_stat->rx_ring_stats;
3068	vsi_stat->rx_ring_stats =
3069		krealloc_array(vsi_stat->rx_ring_stats, req_rxq,
3070			       sizeof(*vsi_stat->rx_ring_stats),
3071			       GFP_KERNEL | __GFP_ZERO);
3072	if (!vsi_stat->rx_ring_stats) {
3073		vsi_stat->rx_ring_stats = rx_ring_stats;
3074		return -ENOMEM;
3075	}
3076
3077	return 0;
3078}
3079
3080/**
3081 * ice_vsi_rebuild - Rebuild VSI after reset
3082 * @vsi: VSI to be rebuild
3083 * @vsi_flags: flags used for VSI rebuild flow
3084 *
3085 * Set vsi_flags to ICE_VSI_FLAG_INIT to initialize a new VSI, or
3086 * ICE_VSI_FLAG_NO_INIT to rebuild an existing VSI in hardware.
3087 *
3088 * Returns 0 on success and negative value on failure
3089 */
3090int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags)
3091{
3092	struct ice_vsi_cfg_params params = {};
3093	struct ice_coalesce_stored *coalesce;
3094	int prev_num_q_vectors;
3095	struct ice_pf *pf;
3096	int ret;
3097
3098	if (!vsi)
3099		return -EINVAL;
3100
3101	params = ice_vsi_to_params(vsi);
3102	params.flags = vsi_flags;
3103
3104	pf = vsi->back;
3105	if (WARN_ON(vsi->type == ICE_VSI_VF && !vsi->vf))
3106		return -EINVAL;
3107
3108	ret = ice_vsi_realloc_stat_arrays(vsi);
3109	if (ret)
3110		goto err_vsi_cfg;
3111
3112	ice_vsi_decfg(vsi);
3113	ret = ice_vsi_cfg_def(vsi, &params);
3114	if (ret)
3115		goto err_vsi_cfg;
3116
3117	coalesce = kcalloc(vsi->num_q_vectors,
3118			   sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3119	if (!coalesce)
3120		return -ENOMEM;
3121
3122	prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3123
3124	ret = ice_vsi_cfg_tc_lan(pf, vsi);
3125	if (ret) {
3126		if (vsi_flags & ICE_VSI_FLAG_INIT) {
3127			ret = -EIO;
3128			goto err_vsi_cfg_tc_lan;
3129		}
3130
3131		kfree(coalesce);
3132		return ice_schedule_reset(pf, ICE_RESET_PFR);
3133	}
3134
3135	ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3136	kfree(coalesce);
3137
3138	return 0;
3139
3140err_vsi_cfg_tc_lan:
3141	ice_vsi_decfg(vsi);
3142	kfree(coalesce);
3143err_vsi_cfg:
3144	return ret;
3145}
3146
3147/**
3148 * ice_is_reset_in_progress - check for a reset in progress
3149 * @state: PF state field
3150 */
3151bool ice_is_reset_in_progress(unsigned long *state)
3152{
3153	return test_bit(ICE_RESET_OICR_RECV, state) ||
3154	       test_bit(ICE_PFR_REQ, state) ||
3155	       test_bit(ICE_CORER_REQ, state) ||
3156	       test_bit(ICE_GLOBR_REQ, state);
3157}
3158
3159/**
3160 * ice_wait_for_reset - Wait for driver to finish reset and rebuild
3161 * @pf: pointer to the PF structure
3162 * @timeout: length of time to wait, in jiffies
3163 *
3164 * Wait (sleep) for a short time until the driver finishes cleaning up from
3165 * a device reset. The caller must be able to sleep. Use this to delay
3166 * operations that could fail while the driver is cleaning up after a device
3167 * reset.
3168 *
3169 * Returns 0 on success, -EBUSY if the reset is not finished within the
3170 * timeout, and -ERESTARTSYS if the thread was interrupted.
3171 */
3172int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3173{
3174	long ret;
3175
3176	ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3177					       !ice_is_reset_in_progress(pf->state),
3178					       timeout);
3179	if (ret < 0)
3180		return ret;
3181	else if (!ret)
3182		return -EBUSY;
3183	else
3184		return 0;
3185}
3186
3187/**
3188 * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3189 * @vsi: VSI being configured
3190 * @ctx: the context buffer returned from AQ VSI update command
3191 */
3192static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3193{
3194	vsi->info.mapping_flags = ctx->info.mapping_flags;
3195	memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3196	       sizeof(vsi->info.q_mapping));
3197	memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3198	       sizeof(vsi->info.tc_mapping));
3199}
3200
3201/**
3202 * ice_vsi_cfg_netdev_tc - Setup the netdev TC configuration
3203 * @vsi: the VSI being configured
3204 * @ena_tc: TC map to be enabled
3205 */
3206void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc)
3207{
3208	struct net_device *netdev = vsi->netdev;
3209	struct ice_pf *pf = vsi->back;
3210	int numtc = vsi->tc_cfg.numtc;
3211	struct ice_dcbx_cfg *dcbcfg;
3212	u8 netdev_tc;
3213	int i;
3214
3215	if (!netdev)
3216		return;
3217
3218	/* CHNL VSI doesn't have it's own netdev, hence, no netdev_tc */
3219	if (vsi->type == ICE_VSI_CHNL)
3220		return;
3221
3222	if (!ena_tc) {
3223		netdev_reset_tc(netdev);
3224		return;
3225	}
3226
3227	if (vsi->type == ICE_VSI_PF && ice_is_adq_active(pf))
3228		numtc = vsi->all_numtc;
3229
3230	if (netdev_set_num_tc(netdev, numtc))
3231		return;
3232
3233	dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg;
3234
3235	ice_for_each_traffic_class(i)
3236		if (vsi->tc_cfg.ena_tc & BIT(i))
3237			netdev_set_tc_queue(netdev,
3238					    vsi->tc_cfg.tc_info[i].netdev_tc,
3239					    vsi->tc_cfg.tc_info[i].qcount_tx,
3240					    vsi->tc_cfg.tc_info[i].qoffset);
3241	/* setup TC queue map for CHNL TCs */
3242	ice_for_each_chnl_tc(i) {
3243		if (!(vsi->all_enatc & BIT(i)))
3244			break;
3245		if (!vsi->mqprio_qopt.qopt.count[i])
3246			break;
3247		netdev_set_tc_queue(netdev, i,
3248				    vsi->mqprio_qopt.qopt.count[i],
3249				    vsi->mqprio_qopt.qopt.offset[i]);
3250	}
3251
3252	if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3253		return;
3254
3255	for (i = 0; i < ICE_MAX_USER_PRIORITY; i++) {
3256		u8 ets_tc = dcbcfg->etscfg.prio_table[i];
3257
3258		/* Get the mapped netdev TC# for the UP */
3259		netdev_tc = vsi->tc_cfg.tc_info[ets_tc].netdev_tc;
3260		netdev_set_prio_tc_map(netdev, i, netdev_tc);
3261	}
3262}
3263
3264/**
3265 * ice_vsi_setup_q_map_mqprio - Prepares mqprio based tc_config
3266 * @vsi: the VSI being configured,
3267 * @ctxt: VSI context structure
3268 * @ena_tc: number of traffic classes to enable
3269 *
3270 * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
3271 */
3272static int
3273ice_vsi_setup_q_map_mqprio(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt,
3274			   u8 ena_tc)
3275{
3276	u16 pow, offset = 0, qcount_tx = 0, qcount_rx = 0, qmap;
3277	u16 tc0_offset = vsi->mqprio_qopt.qopt.offset[0];
3278	int tc0_qcount = vsi->mqprio_qopt.qopt.count[0];
3279	u16 new_txq, new_rxq;
3280	u8 netdev_tc = 0;
3281	int i;
3282
3283	vsi->tc_cfg.ena_tc = ena_tc ? ena_tc : 1;
3284
3285	pow = order_base_2(tc0_qcount);
3286	qmap = FIELD_PREP(ICE_AQ_VSI_TC_Q_OFFSET_M, tc0_offset);
3287	qmap |= FIELD_PREP(ICE_AQ_VSI_TC_Q_NUM_M, pow);
3288
3289	ice_for_each_traffic_class(i) {
3290		if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
3291			/* TC is not enabled */
3292			vsi->tc_cfg.tc_info[i].qoffset = 0;
3293			vsi->tc_cfg.tc_info[i].qcount_rx = 1;
3294			vsi->tc_cfg.tc_info[i].qcount_tx = 1;
3295			vsi->tc_cfg.tc_info[i].netdev_tc = 0;
3296			ctxt->info.tc_mapping[i] = 0;
3297			continue;
3298		}
3299
3300		offset = vsi->mqprio_qopt.qopt.offset[i];
3301		qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3302		qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3303		vsi->tc_cfg.tc_info[i].qoffset = offset;
3304		vsi->tc_cfg.tc_info[i].qcount_rx = qcount_rx;
3305		vsi->tc_cfg.tc_info[i].qcount_tx = qcount_tx;
3306		vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
3307	}
3308
3309	if (vsi->all_numtc && vsi->all_numtc != vsi->tc_cfg.numtc) {
3310		ice_for_each_chnl_tc(i) {
3311			if (!(vsi->all_enatc & BIT(i)))
3312				continue;
3313			offset = vsi->mqprio_qopt.qopt.offset[i];
3314			qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3315			qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3316		}
3317	}
3318
3319	new_txq = offset + qcount_tx;
3320	if (new_txq > vsi->alloc_txq) {
3321		dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
3322			new_txq, vsi->alloc_txq);
3323		return -EINVAL;
3324	}
3325
3326	new_rxq = offset + qcount_rx;
3327	if (new_rxq > vsi->alloc_rxq) {
3328		dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
3329			new_rxq, vsi->alloc_rxq);
3330		return -EINVAL;
3331	}
3332
3333	/* Set actual Tx/Rx queue pairs */
3334	vsi->num_txq = new_txq;
3335	vsi->num_rxq = new_rxq;
3336
3337	/* Setup queue TC[0].qmap for given VSI context */
3338	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
3339	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
3340	ctxt->info.q_mapping[1] = cpu_to_le16(tc0_qcount);
3341
3342	/* Find queue count available for channel VSIs and starting offset
3343	 * for channel VSIs
3344	 */
3345	if (tc0_qcount && tc0_qcount < vsi->num_rxq) {
3346		vsi->cnt_q_avail = vsi->num_rxq - tc0_qcount;
3347		vsi->next_base_q = tc0_qcount;
3348	}
3349	dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_txq = %d\n",  vsi->num_txq);
3350	dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_rxq = %d\n",  vsi->num_rxq);
3351	dev_dbg(ice_pf_to_dev(vsi->back), "all_numtc %u, all_enatc: 0x%04x, tc_cfg.numtc %u\n",
3352		vsi->all_numtc, vsi->all_enatc, vsi->tc_cfg.numtc);
3353
3354	return 0;
3355}
3356
3357/**
3358 * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3359 * @vsi: VSI to be configured
3360 * @ena_tc: TC bitmap
3361 *
3362 * VSI queues expected to be quiesced before calling this function
3363 */
3364int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3365{
3366	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3367	struct ice_pf *pf = vsi->back;
3368	struct ice_tc_cfg old_tc_cfg;
3369	struct ice_vsi_ctx *ctx;
3370	struct device *dev;
3371	int i, ret = 0;
3372	u8 num_tc = 0;
3373
3374	dev = ice_pf_to_dev(pf);
3375	if (vsi->tc_cfg.ena_tc == ena_tc &&
3376	    vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
3377		return 0;
3378
3379	ice_for_each_traffic_class(i) {
3380		/* build bitmap of enabled TCs */
3381		if (ena_tc & BIT(i))
3382			num_tc++;
3383		/* populate max_txqs per TC */
3384		max_txqs[i] = vsi->alloc_txq;
3385		/* Update max_txqs if it is CHNL VSI, because alloc_t[r]xq are
3386		 * zero for CHNL VSI, hence use num_txq instead as max_txqs
3387		 */
3388		if (vsi->type == ICE_VSI_CHNL &&
3389		    test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3390			max_txqs[i] = vsi->num_txq;
3391	}
3392
3393	memcpy(&old_tc_cfg, &vsi->tc_cfg, sizeof(old_tc_cfg));
3394	vsi->tc_cfg.ena_tc = ena_tc;
3395	vsi->tc_cfg.numtc = num_tc;
3396
3397	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3398	if (!ctx)
3399		return -ENOMEM;
3400
3401	ctx->vf_num = 0;
3402	ctx->info = vsi->info;
3403
3404	if (vsi->type == ICE_VSI_PF &&
3405	    test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3406		ret = ice_vsi_setup_q_map_mqprio(vsi, ctx, ena_tc);
3407	else
3408		ret = ice_vsi_setup_q_map(vsi, ctx);
3409
3410	if (ret) {
3411		memcpy(&vsi->tc_cfg, &old_tc_cfg, sizeof(vsi->tc_cfg));
3412		goto out;
3413	}
3414
3415	/* must to indicate which section of VSI context are being modified */
3416	ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3417	ret = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3418	if (ret) {
3419		dev_info(dev, "Failed VSI Update\n");
3420		goto out;
3421	}
3422
3423	if (vsi->type == ICE_VSI_PF &&
3424	    test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3425		ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, 1, max_txqs);
3426	else
3427		ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx,
3428				      vsi->tc_cfg.ena_tc, max_txqs);
3429
3430	if (ret) {
3431		dev_err(dev, "VSI %d failed TC config, error %d\n",
3432			vsi->vsi_num, ret);
3433		goto out;
3434	}
3435	ice_vsi_update_q_map(vsi, ctx);
3436	vsi->info.valid_sections = 0;
3437
3438	ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3439out:
3440	kfree(ctx);
3441	return ret;
3442}
3443
3444/**
3445 * ice_update_ring_stats - Update ring statistics
3446 * @stats: stats to be updated
3447 * @pkts: number of processed packets
3448 * @bytes: number of processed bytes
3449 *
3450 * This function assumes that caller has acquired a u64_stats_sync lock.
3451 */
3452static void ice_update_ring_stats(struct ice_q_stats *stats, u64 pkts, u64 bytes)
3453{
3454	stats->bytes += bytes;
3455	stats->pkts += pkts;
3456}
3457
3458/**
3459 * ice_update_tx_ring_stats - Update Tx ring specific counters
3460 * @tx_ring: ring to update
3461 * @pkts: number of processed packets
3462 * @bytes: number of processed bytes
3463 */
3464void ice_update_tx_ring_stats(struct ice_tx_ring *tx_ring, u64 pkts, u64 bytes)
3465{
3466	u64_stats_update_begin(&tx_ring->ring_stats->syncp);
3467	ice_update_ring_stats(&tx_ring->ring_stats->stats, pkts, bytes);
3468	u64_stats_update_end(&tx_ring->ring_stats->syncp);
3469}
3470
3471/**
3472 * ice_update_rx_ring_stats - Update Rx ring specific counters
3473 * @rx_ring: ring to update
3474 * @pkts: number of processed packets
3475 * @bytes: number of processed bytes
3476 */
3477void ice_update_rx_ring_stats(struct ice_rx_ring *rx_ring, u64 pkts, u64 bytes)
3478{
3479	u64_stats_update_begin(&rx_ring->ring_stats->syncp);
3480	ice_update_ring_stats(&rx_ring->ring_stats->stats, pkts, bytes);
3481	u64_stats_update_end(&rx_ring->ring_stats->syncp);
3482}
3483
3484/**
3485 * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3486 * @pi: port info of the switch with default VSI
3487 *
3488 * Return true if the there is a single VSI in default forwarding VSI list
3489 */
3490bool ice_is_dflt_vsi_in_use(struct ice_port_info *pi)
3491{
3492	bool exists = false;
3493
3494	ice_check_if_dflt_vsi(pi, 0, &exists);
3495	return exists;
3496}
3497
3498/**
3499 * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3500 * @vsi: VSI to compare against default forwarding VSI
3501 *
3502 * If this VSI passed in is the default forwarding VSI then return true, else
3503 * return false
3504 */
3505bool ice_is_vsi_dflt_vsi(struct ice_vsi *vsi)
3506{
3507	return ice_check_if_dflt_vsi(vsi->port_info, vsi->idx, NULL);
3508}
3509
3510/**
3511 * ice_set_dflt_vsi - set the default forwarding VSI
3512 * @vsi: VSI getting set as the default forwarding VSI on the switch
3513 *
3514 * If the VSI passed in is already the default VSI and it's enabled just return
3515 * success.
3516 *
3517 * Otherwise try to set the VSI passed in as the switch's default VSI and
3518 * return the result.
3519 */
3520int ice_set_dflt_vsi(struct ice_vsi *vsi)
3521{
3522	struct device *dev;
3523	int status;
3524
3525	if (!vsi)
3526		return -EINVAL;
3527
3528	dev = ice_pf_to_dev(vsi->back);
3529
3530	if (ice_lag_is_switchdev_running(vsi->back)) {
3531		dev_dbg(dev, "VSI %d passed is a part of LAG containing interfaces in switchdev mode, nothing to do\n",
3532			vsi->vsi_num);
3533		return 0;
3534	}
3535
3536	/* the VSI passed in is already the default VSI */
3537	if (ice_is_vsi_dflt_vsi(vsi)) {
3538		dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3539			vsi->vsi_num);
3540		return 0;
3541	}
3542
3543	status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, true, ICE_FLTR_RX);
3544	if (status) {
3545		dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %d\n",
3546			vsi->vsi_num, status);
3547		return status;
3548	}
3549
3550	return 0;
3551}
3552
3553/**
3554 * ice_clear_dflt_vsi - clear the default forwarding VSI
3555 * @vsi: VSI to remove from filter list
3556 *
3557 * If the switch has no default VSI or it's not enabled then return error.
3558 *
3559 * Otherwise try to clear the default VSI and return the result.
3560 */
3561int ice_clear_dflt_vsi(struct ice_vsi *vsi)
3562{
3563	struct device *dev;
3564	int status;
3565
3566	if (!vsi)
3567		return -EINVAL;
3568
3569	dev = ice_pf_to_dev(vsi->back);
3570
3571	/* there is no default VSI configured */
3572	if (!ice_is_dflt_vsi_in_use(vsi->port_info))
3573		return -ENODEV;
3574
3575	status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, false,
3576				  ICE_FLTR_RX);
3577	if (status) {
3578		dev_err(dev, "Failed to clear the default forwarding VSI %d, error %d\n",
3579			vsi->vsi_num, status);
3580		return -EIO;
3581	}
3582
3583	return 0;
3584}
3585
3586/**
3587 * ice_get_link_speed_mbps - get link speed in Mbps
3588 * @vsi: the VSI whose link speed is being queried
3589 *
3590 * Return current VSI link speed and 0 if the speed is unknown.
3591 */
3592int ice_get_link_speed_mbps(struct ice_vsi *vsi)
3593{
3594	unsigned int link_speed;
3595
3596	link_speed = vsi->port_info->phy.link_info.link_speed;
3597
3598	return (int)ice_get_link_speed(fls(link_speed) - 1);
3599}
3600
3601/**
3602 * ice_get_link_speed_kbps - get link speed in Kbps
3603 * @vsi: the VSI whose link speed is being queried
3604 *
3605 * Return current VSI link speed and 0 if the speed is unknown.
3606 */
3607int ice_get_link_speed_kbps(struct ice_vsi *vsi)
3608{
3609	int speed_mbps;
3610
3611	speed_mbps = ice_get_link_speed_mbps(vsi);
3612
3613	return speed_mbps * 1000;
3614}
3615
3616/**
3617 * ice_set_min_bw_limit - setup minimum BW limit for Tx based on min_tx_rate
3618 * @vsi: VSI to be configured
3619 * @min_tx_rate: min Tx rate in Kbps to be configured as BW limit
3620 *
3621 * If the min_tx_rate is specified as 0 that means to clear the minimum BW limit
3622 * profile, otherwise a non-zero value will force a minimum BW limit for the VSI
3623 * on TC 0.
3624 */
3625int ice_set_min_bw_limit(struct ice_vsi *vsi, u64 min_tx_rate)
3626{
3627	struct ice_pf *pf = vsi->back;
3628	struct device *dev;
3629	int status;
3630	int speed;
3631
3632	dev = ice_pf_to_dev(pf);
3633	if (!vsi->port_info) {
3634		dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3635			vsi->idx, vsi->type);
3636		return -EINVAL;
3637	}
3638
3639	speed = ice_get_link_speed_kbps(vsi);
3640	if (min_tx_rate > (u64)speed) {
3641		dev_err(dev, "invalid min Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3642			min_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3643			speed);
3644		return -EINVAL;
3645	}
3646
3647	/* Configure min BW for VSI limit */
3648	if (min_tx_rate) {
3649		status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3650						   ICE_MIN_BW, min_tx_rate);
3651		if (status) {
3652			dev_err(dev, "failed to set min Tx rate(%llu Kbps) for %s %d\n",
3653				min_tx_rate, ice_vsi_type_str(vsi->type),
3654				vsi->idx);
3655			return status;
3656		}
3657
3658		dev_dbg(dev, "set min Tx rate(%llu Kbps) for %s\n",
3659			min_tx_rate, ice_vsi_type_str(vsi->type));
3660	} else {
3661		status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3662							vsi->idx, 0,
3663							ICE_MIN_BW);
3664		if (status) {
3665			dev_err(dev, "failed to clear min Tx rate configuration for %s %d\n",
3666				ice_vsi_type_str(vsi->type), vsi->idx);
3667			return status;
3668		}
3669
3670		dev_dbg(dev, "cleared min Tx rate configuration for %s %d\n",
3671			ice_vsi_type_str(vsi->type), vsi->idx);
3672	}
3673
3674	return 0;
3675}
3676
3677/**
3678 * ice_set_max_bw_limit - setup maximum BW limit for Tx based on max_tx_rate
3679 * @vsi: VSI to be configured
3680 * @max_tx_rate: max Tx rate in Kbps to be configured as BW limit
3681 *
3682 * If the max_tx_rate is specified as 0 that means to clear the maximum BW limit
3683 * profile, otherwise a non-zero value will force a maximum BW limit for the VSI
3684 * on TC 0.
3685 */
3686int ice_set_max_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate)
3687{
3688	struct ice_pf *pf = vsi->back;
3689	struct device *dev;
3690	int status;
3691	int speed;
3692
3693	dev = ice_pf_to_dev(pf);
3694	if (!vsi->port_info) {
3695		dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3696			vsi->idx, vsi->type);
3697		return -EINVAL;
3698	}
3699
3700	speed = ice_get_link_speed_kbps(vsi);
3701	if (max_tx_rate > (u64)speed) {
3702		dev_err(dev, "invalid max Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3703			max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3704			speed);
3705		return -EINVAL;
3706	}
3707
3708	/* Configure max BW for VSI limit */
3709	if (max_tx_rate) {
3710		status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3711						   ICE_MAX_BW, max_tx_rate);
3712		if (status) {
3713			dev_err(dev, "failed setting max Tx rate(%llu Kbps) for %s %d\n",
3714				max_tx_rate, ice_vsi_type_str(vsi->type),
3715				vsi->idx);
3716			return status;
3717		}
3718
3719		dev_dbg(dev, "set max Tx rate(%llu Kbps) for %s %d\n",
3720			max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx);
3721	} else {
3722		status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3723							vsi->idx, 0,
3724							ICE_MAX_BW);
3725		if (status) {
3726			dev_err(dev, "failed clearing max Tx rate configuration for %s %d\n",
3727				ice_vsi_type_str(vsi->type), vsi->idx);
3728			return status;
3729		}
3730
3731		dev_dbg(dev, "cleared max Tx rate configuration for %s %d\n",
3732			ice_vsi_type_str(vsi->type), vsi->idx);
3733	}
3734
3735	return 0;
3736}
3737
3738/**
3739 * ice_set_link - turn on/off physical link
3740 * @vsi: VSI to modify physical link on
3741 * @ena: turn on/off physical link
3742 */
3743int ice_set_link(struct ice_vsi *vsi, bool ena)
3744{
3745	struct device *dev = ice_pf_to_dev(vsi->back);
3746	struct ice_port_info *pi = vsi->port_info;
3747	struct ice_hw *hw = pi->hw;
3748	int status;
3749
3750	if (vsi->type != ICE_VSI_PF)
3751		return -EINVAL;
3752
3753	status = ice_aq_set_link_restart_an(pi, ena, NULL);
3754
3755	/* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3756	 * this is not a fatal error, so print a warning message and return
3757	 * a success code. Return an error if FW returns an error code other
3758	 * than ICE_AQ_RC_EMODE
3759	 */
3760	if (status == -EIO) {
3761		if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3762			dev_dbg(dev, "can't set link to %s, err %d aq_err %s. not fatal, continuing\n",
3763				(ena ? "ON" : "OFF"), status,
3764				ice_aq_str(hw->adminq.sq_last_status));
3765	} else if (status) {
3766		dev_err(dev, "can't set link to %s, err %d aq_err %s\n",
3767			(ena ? "ON" : "OFF"), status,
3768			ice_aq_str(hw->adminq.sq_last_status));
3769		return status;
3770	}
3771
3772	return 0;
3773}
3774
3775/**
3776 * ice_vsi_add_vlan_zero - add VLAN 0 filter(s) for this VSI
3777 * @vsi: VSI used to add VLAN filters
3778 *
3779 * In Single VLAN Mode (SVM), single VLAN filters via ICE_SW_LKUP_VLAN are based
3780 * on the inner VLAN ID, so the VLAN TPID (i.e. 0x8100 or 0x888a8) doesn't
3781 * matter. In Double VLAN Mode (DVM), outer/single VLAN filters via
3782 * ICE_SW_LKUP_VLAN are based on the outer/single VLAN ID + VLAN TPID.
3783 *
3784 * For both modes add a VLAN 0 + no VLAN TPID filter to handle untagged traffic
3785 * when VLAN pruning is enabled. Also, this handles VLAN 0 priority tagged
3786 * traffic in SVM, since the VLAN TPID isn't part of filtering.
3787 *
3788 * If DVM is enabled then an explicit VLAN 0 + VLAN TPID filter needs to be
3789 * added to allow VLAN 0 priority tagged traffic in DVM, since the VLAN TPID is
3790 * part of filtering.
3791 */
3792int ice_vsi_add_vlan_zero(struct ice_vsi *vsi)
3793{
3794	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3795	struct ice_vlan vlan;
3796	int err;
3797
3798	vlan = ICE_VLAN(0, 0, 0);
3799	err = vlan_ops->add_vlan(vsi, &vlan);
3800	if (err && err != -EEXIST)
3801		return err;
3802
3803	/* in SVM both VLAN 0 filters are identical */
3804	if (!ice_is_dvm_ena(&vsi->back->hw))
3805		return 0;
3806
3807	vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3808	err = vlan_ops->add_vlan(vsi, &vlan);
3809	if (err && err != -EEXIST)
3810		return err;
3811
3812	return 0;
3813}
3814
3815/**
3816 * ice_vsi_del_vlan_zero - delete VLAN 0 filter(s) for this VSI
3817 * @vsi: VSI used to add VLAN filters
3818 *
3819 * Delete the VLAN 0 filters in the same manner that they were added in
3820 * ice_vsi_add_vlan_zero.
3821 */
3822int ice_vsi_del_vlan_zero(struct ice_vsi *vsi)
3823{
3824	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3825	struct ice_vlan vlan;
3826	int err;
3827
3828	vlan = ICE_VLAN(0, 0, 0);
3829	err = vlan_ops->del_vlan(vsi, &vlan);
3830	if (err && err != -EEXIST)
3831		return err;
3832
3833	/* in SVM both VLAN 0 filters are identical */
3834	if (!ice_is_dvm_ena(&vsi->back->hw))
3835		return 0;
3836
3837	vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3838	err = vlan_ops->del_vlan(vsi, &vlan);
3839	if (err && err != -EEXIST)
3840		return err;
3841
3842	/* when deleting the last VLAN filter, make sure to disable the VLAN
3843	 * promisc mode so the filter isn't left by accident
3844	 */
3845	return ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3846				    ICE_MCAST_VLAN_PROMISC_BITS, 0);
3847}
3848
3849/**
3850 * ice_vsi_num_zero_vlans - get number of VLAN 0 filters based on VLAN mode
3851 * @vsi: VSI used to get the VLAN mode
3852 *
3853 * If DVM is enabled then 2 VLAN 0 filters are added, else if SVM is enabled
3854 * then 1 VLAN 0 filter is added. See ice_vsi_add_vlan_zero for more details.
3855 */
3856static u16 ice_vsi_num_zero_vlans(struct ice_vsi *vsi)
3857{
3858#define ICE_DVM_NUM_ZERO_VLAN_FLTRS	2
3859#define ICE_SVM_NUM_ZERO_VLAN_FLTRS	1
3860	/* no VLAN 0 filter is created when a port VLAN is active */
3861	if (vsi->type == ICE_VSI_VF) {
3862		if (WARN_ON(!vsi->vf))
3863			return 0;
3864
3865		if (ice_vf_is_port_vlan_ena(vsi->vf))
3866			return 0;
3867	}
3868
3869	if (ice_is_dvm_ena(&vsi->back->hw))
3870		return ICE_DVM_NUM_ZERO_VLAN_FLTRS;
3871	else
3872		return ICE_SVM_NUM_ZERO_VLAN_FLTRS;
3873}
3874
3875/**
3876 * ice_vsi_has_non_zero_vlans - check if VSI has any non-zero VLANs
3877 * @vsi: VSI used to determine if any non-zero VLANs have been added
3878 */
3879bool ice_vsi_has_non_zero_vlans(struct ice_vsi *vsi)
3880{
3881	return (vsi->num_vlan > ice_vsi_num_zero_vlans(vsi));
3882}
3883
3884/**
3885 * ice_vsi_num_non_zero_vlans - get the number of non-zero VLANs for this VSI
3886 * @vsi: VSI used to get the number of non-zero VLANs added
3887 */
3888u16 ice_vsi_num_non_zero_vlans(struct ice_vsi *vsi)
3889{
3890	return (vsi->num_vlan - ice_vsi_num_zero_vlans(vsi));
3891}
3892
3893/**
3894 * ice_is_feature_supported
3895 * @pf: pointer to the struct ice_pf instance
3896 * @f: feature enum to be checked
3897 *
3898 * returns true if feature is supported, false otherwise
3899 */
3900bool ice_is_feature_supported(struct ice_pf *pf, enum ice_feature f)
3901{
3902	if (f < 0 || f >= ICE_F_MAX)
3903		return false;
3904
3905	return test_bit(f, pf->features);
3906}
3907
3908/**
3909 * ice_set_feature_support
3910 * @pf: pointer to the struct ice_pf instance
3911 * @f: feature enum to set
3912 */
3913void ice_set_feature_support(struct ice_pf *pf, enum ice_feature f)
3914{
3915	if (f < 0 || f >= ICE_F_MAX)
3916		return;
3917
3918	set_bit(f, pf->features);
3919}
3920
3921/**
3922 * ice_clear_feature_support
3923 * @pf: pointer to the struct ice_pf instance
3924 * @f: feature enum to clear
3925 */
3926void ice_clear_feature_support(struct ice_pf *pf, enum ice_feature f)
3927{
3928	if (f < 0 || f >= ICE_F_MAX)
3929		return;
3930
3931	clear_bit(f, pf->features);
3932}
3933
3934/**
3935 * ice_init_feature_support
3936 * @pf: pointer to the struct ice_pf instance
3937 *
3938 * called during init to setup supported feature
3939 */
3940void ice_init_feature_support(struct ice_pf *pf)
3941{
3942	switch (pf->hw.device_id) {
3943	case ICE_DEV_ID_E810C_BACKPLANE:
3944	case ICE_DEV_ID_E810C_QSFP:
3945	case ICE_DEV_ID_E810C_SFP:
3946	case ICE_DEV_ID_E810_XXV_BACKPLANE:
3947	case ICE_DEV_ID_E810_XXV_QSFP:
3948	case ICE_DEV_ID_E810_XXV_SFP:
3949		ice_set_feature_support(pf, ICE_F_DSCP);
3950		if (ice_is_phy_rclk_in_netlist(&pf->hw))
3951			ice_set_feature_support(pf, ICE_F_PHY_RCLK);
3952		/* If we don't own the timer - don't enable other caps */
3953		if (!ice_pf_src_tmr_owned(pf))
3954			break;
3955		if (ice_is_cgu_in_netlist(&pf->hw))
3956			ice_set_feature_support(pf, ICE_F_CGU);
3957		if (ice_is_clock_mux_in_netlist(&pf->hw))
3958			ice_set_feature_support(pf, ICE_F_SMA_CTRL);
3959		if (ice_gnss_is_gps_present(&pf->hw))
3960			ice_set_feature_support(pf, ICE_F_GNSS);
3961		break;
3962	default:
3963		break;
3964	}
3965}
3966
3967/**
3968 * ice_vsi_update_security - update security block in VSI
3969 * @vsi: pointer to VSI structure
3970 * @fill: function pointer to fill ctx
3971 */
3972int
3973ice_vsi_update_security(struct ice_vsi *vsi, void (*fill)(struct ice_vsi_ctx *))
3974{
3975	struct ice_vsi_ctx ctx = { 0 };
3976
3977	ctx.info = vsi->info;
3978	ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
3979	fill(&ctx);
3980
3981	if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
3982		return -ENODEV;
3983
3984	vsi->info = ctx.info;
3985	return 0;
3986}
3987
3988/**
3989 * ice_vsi_ctx_set_antispoof - set antispoof function in VSI ctx
3990 * @ctx: pointer to VSI ctx structure
3991 */
3992void ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx *ctx)
3993{
3994	ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
3995			       (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3996				ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3997}
3998
3999/**
4000 * ice_vsi_ctx_clear_antispoof - clear antispoof function in VSI ctx
4001 * @ctx: pointer to VSI ctx structure
4002 */
4003void ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx *ctx)
4004{
4005	ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF &
4006			       ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4007				 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4008}
4009
4010/**
4011 * ice_vsi_ctx_set_allow_override - allow destination override on VSI
4012 * @ctx: pointer to VSI ctx structure
4013 */
4014void ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx *ctx)
4015{
4016	ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4017}
4018
4019/**
4020 * ice_vsi_ctx_clear_allow_override - turn off destination override on VSI
4021 * @ctx: pointer to VSI ctx structure
4022 */
4023void ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx *ctx)
4024{
4025	ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4026}
4027
4028/**
4029 * ice_vsi_update_local_lb - update sw block in VSI with local loopback bit
4030 * @vsi: pointer to VSI structure
4031 * @set: set or unset the bit
4032 */
4033int
4034ice_vsi_update_local_lb(struct ice_vsi *vsi, bool set)
4035{
4036	struct ice_vsi_ctx ctx = {
4037		.info	= vsi->info,
4038	};
4039
4040	ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
4041	if (set)
4042		ctx.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4043	else
4044		ctx.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4045
4046	if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4047		return -ENODEV;
4048
4049	vsi->info = ctx.info;
4050	return 0;
4051}
4052