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 VFs.
216		 * Each ring is associated to the corresponding VF_PR netdev.
217		 */
218		vsi->alloc_txq = ice_get_num_vfs(pf);
219		vsi->alloc_rxq = vsi->alloc_txq;
220		vsi->num_q_vectors = 1;
221		break;
222	case ICE_VSI_VF:
223		if (vf->num_req_qs)
224			vf->num_vf_qs = vf->num_req_qs;
225		vsi->alloc_txq = vf->num_vf_qs;
226		vsi->alloc_rxq = vf->num_vf_qs;
227		/* pf->vfs.num_msix_per includes (VF miscellaneous vector +
228		 * data queue interrupts). Since vsi->num_q_vectors is number
229		 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
230		 * original vector count
231		 */
232		vsi->num_q_vectors = pf->vfs.num_msix_per - ICE_NONQ_VECS_VF;
233		break;
234	case ICE_VSI_CTRL:
235		vsi->alloc_txq = 1;
236		vsi->alloc_rxq = 1;
237		vsi->num_q_vectors = 1;
238		break;
239	case ICE_VSI_CHNL:
240		vsi->alloc_txq = 0;
241		vsi->alloc_rxq = 0;
242		break;
243	case ICE_VSI_LB:
244		vsi->alloc_txq = 1;
245		vsi->alloc_rxq = 1;
246		break;
247	default:
248		dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi_type);
249		break;
250	}
251
252	ice_vsi_set_num_desc(vsi);
253}
254
255/**
256 * ice_get_free_slot - get the next non-NULL location index in array
257 * @array: array to search
258 * @size: size of the array
259 * @curr: last known occupied index to be used as a search hint
260 *
261 * void * is being used to keep the functionality generic. This lets us use this
262 * function on any array of pointers.
263 */
264static int ice_get_free_slot(void *array, int size, int curr)
265{
266	int **tmp_array = (int **)array;
267	int next;
268
269	if (curr < (size - 1) && !tmp_array[curr + 1]) {
270		next = curr + 1;
271	} else {
272		int i = 0;
273
274		while ((i < size) && (tmp_array[i]))
275			i++;
276		if (i == size)
277			next = ICE_NO_VSI;
278		else
279			next = i;
280	}
281	return next;
282}
283
284/**
285 * ice_vsi_delete_from_hw - delete a VSI from the switch
286 * @vsi: pointer to VSI being removed
287 */
288static void ice_vsi_delete_from_hw(struct ice_vsi *vsi)
289{
290	struct ice_pf *pf = vsi->back;
291	struct ice_vsi_ctx *ctxt;
292	int status;
293
294	ice_fltr_remove_all(vsi);
295	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
296	if (!ctxt)
297		return;
298
299	if (vsi->type == ICE_VSI_VF)
300		ctxt->vf_num = vsi->vf->vf_id;
301	ctxt->vsi_num = vsi->vsi_num;
302
303	memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
304
305	status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
306	if (status)
307		dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %d\n",
308			vsi->vsi_num, status);
309
310	kfree(ctxt);
311}
312
313/**
314 * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
315 * @vsi: pointer to VSI being cleared
316 */
317static void ice_vsi_free_arrays(struct ice_vsi *vsi)
318{
319	struct ice_pf *pf = vsi->back;
320	struct device *dev;
321
322	dev = ice_pf_to_dev(pf);
323
324	bitmap_free(vsi->af_xdp_zc_qps);
325	vsi->af_xdp_zc_qps = NULL;
326	/* free the ring and vector containers */
327	devm_kfree(dev, vsi->q_vectors);
328	vsi->q_vectors = NULL;
329	devm_kfree(dev, vsi->tx_rings);
330	vsi->tx_rings = NULL;
331	devm_kfree(dev, vsi->rx_rings);
332	vsi->rx_rings = NULL;
333	devm_kfree(dev, vsi->txq_map);
334	vsi->txq_map = NULL;
335	devm_kfree(dev, vsi->rxq_map);
336	vsi->rxq_map = NULL;
337}
338
339/**
340 * ice_vsi_free_stats - Free the ring statistics structures
341 * @vsi: VSI pointer
342 */
343static void ice_vsi_free_stats(struct ice_vsi *vsi)
344{
345	struct ice_vsi_stats *vsi_stat;
346	struct ice_pf *pf = vsi->back;
347	int i;
348
349	if (vsi->type == ICE_VSI_CHNL)
350		return;
351	if (!pf->vsi_stats)
352		return;
353
354	vsi_stat = pf->vsi_stats[vsi->idx];
355	if (!vsi_stat)
356		return;
357
358	ice_for_each_alloc_txq(vsi, i) {
359		if (vsi_stat->tx_ring_stats[i]) {
360			kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
361			WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
362		}
363	}
364
365	ice_for_each_alloc_rxq(vsi, i) {
366		if (vsi_stat->rx_ring_stats[i]) {
367			kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
368			WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
369		}
370	}
371
372	kfree(vsi_stat->tx_ring_stats);
373	kfree(vsi_stat->rx_ring_stats);
374	kfree(vsi_stat);
375	pf->vsi_stats[vsi->idx] = NULL;
376}
377
378/**
379 * ice_vsi_alloc_ring_stats - Allocates Tx and Rx ring stats for the VSI
380 * @vsi: VSI which is having stats allocated
381 */
382static int ice_vsi_alloc_ring_stats(struct ice_vsi *vsi)
383{
384	struct ice_ring_stats **tx_ring_stats;
385	struct ice_ring_stats **rx_ring_stats;
386	struct ice_vsi_stats *vsi_stats;
387	struct ice_pf *pf = vsi->back;
388	u16 i;
389
390	vsi_stats = pf->vsi_stats[vsi->idx];
391	tx_ring_stats = vsi_stats->tx_ring_stats;
392	rx_ring_stats = vsi_stats->rx_ring_stats;
393
394	/* Allocate Tx ring stats */
395	ice_for_each_alloc_txq(vsi, i) {
396		struct ice_ring_stats *ring_stats;
397		struct ice_tx_ring *ring;
398
399		ring = vsi->tx_rings[i];
400		ring_stats = tx_ring_stats[i];
401
402		if (!ring_stats) {
403			ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
404			if (!ring_stats)
405				goto err_out;
406
407			WRITE_ONCE(tx_ring_stats[i], ring_stats);
408		}
409
410		ring->ring_stats = ring_stats;
411	}
412
413	/* Allocate Rx ring stats */
414	ice_for_each_alloc_rxq(vsi, i) {
415		struct ice_ring_stats *ring_stats;
416		struct ice_rx_ring *ring;
417
418		ring = vsi->rx_rings[i];
419		ring_stats = rx_ring_stats[i];
420
421		if (!ring_stats) {
422			ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
423			if (!ring_stats)
424				goto err_out;
425
426			WRITE_ONCE(rx_ring_stats[i], ring_stats);
427		}
428
429		ring->ring_stats = ring_stats;
430	}
431
432	return 0;
433
434err_out:
435	ice_vsi_free_stats(vsi);
436	return -ENOMEM;
437}
438
439/**
440 * ice_vsi_free - clean up and deallocate the provided VSI
441 * @vsi: pointer to VSI being cleared
442 *
443 * This deallocates the VSI's queue resources, removes it from the PF's
444 * VSI array if necessary, and deallocates the VSI
445 */
446static void ice_vsi_free(struct ice_vsi *vsi)
447{
448	struct ice_pf *pf = NULL;
449	struct device *dev;
450
451	if (!vsi || !vsi->back)
452		return;
453
454	pf = vsi->back;
455	dev = ice_pf_to_dev(pf);
456
457	if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
458		dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
459		return;
460	}
461
462	mutex_lock(&pf->sw_mutex);
463	/* updates the PF for this cleared VSI */
464
465	pf->vsi[vsi->idx] = NULL;
466	pf->next_vsi = vsi->idx;
467
468	ice_vsi_free_stats(vsi);
469	ice_vsi_free_arrays(vsi);
470	mutex_unlock(&pf->sw_mutex);
471	devm_kfree(dev, vsi);
472}
473
474void ice_vsi_delete(struct ice_vsi *vsi)
475{
476	ice_vsi_delete_from_hw(vsi);
477	ice_vsi_free(vsi);
478}
479
480/**
481 * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
482 * @irq: interrupt number
483 * @data: pointer to a q_vector
484 */
485static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
486{
487	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
488
489	if (!q_vector->tx.tx_ring)
490		return IRQ_HANDLED;
491
492#define FDIR_RX_DESC_CLEAN_BUDGET 64
493	ice_clean_rx_irq(q_vector->rx.rx_ring, FDIR_RX_DESC_CLEAN_BUDGET);
494	ice_clean_ctrl_tx_irq(q_vector->tx.tx_ring);
495
496	return IRQ_HANDLED;
497}
498
499/**
500 * ice_msix_clean_rings - MSIX mode Interrupt Handler
501 * @irq: interrupt number
502 * @data: pointer to a q_vector
503 */
504static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
505{
506	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
507
508	if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
509		return IRQ_HANDLED;
510
511	q_vector->total_events++;
512
513	napi_schedule(&q_vector->napi);
514
515	return IRQ_HANDLED;
516}
517
518static irqreturn_t ice_eswitch_msix_clean_rings(int __always_unused irq, void *data)
519{
520	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
521	struct ice_pf *pf = q_vector->vsi->back;
522	struct ice_vf *vf;
523	unsigned int bkt;
524
525	if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
526		return IRQ_HANDLED;
527
528	rcu_read_lock();
529	ice_for_each_vf_rcu(pf, bkt, vf)
530		napi_schedule(&vf->repr->q_vector->napi);
531	rcu_read_unlock();
532
533	return IRQ_HANDLED;
534}
535
536/**
537 * ice_vsi_alloc_stat_arrays - Allocate statistics arrays
538 * @vsi: VSI pointer
539 */
540static int ice_vsi_alloc_stat_arrays(struct ice_vsi *vsi)
541{
542	struct ice_vsi_stats *vsi_stat;
543	struct ice_pf *pf = vsi->back;
544
545	if (vsi->type == ICE_VSI_CHNL)
546		return 0;
547	if (!pf->vsi_stats)
548		return -ENOENT;
549
550	if (pf->vsi_stats[vsi->idx])
551	/* realloc will happen in rebuild path */
552		return 0;
553
554	vsi_stat = kzalloc(sizeof(*vsi_stat), GFP_KERNEL);
555	if (!vsi_stat)
556		return -ENOMEM;
557
558	vsi_stat->tx_ring_stats =
559		kcalloc(vsi->alloc_txq, sizeof(*vsi_stat->tx_ring_stats),
560			GFP_KERNEL);
561	if (!vsi_stat->tx_ring_stats)
562		goto err_alloc_tx;
563
564	vsi_stat->rx_ring_stats =
565		kcalloc(vsi->alloc_rxq, sizeof(*vsi_stat->rx_ring_stats),
566			GFP_KERNEL);
567	if (!vsi_stat->rx_ring_stats)
568		goto err_alloc_rx;
569
570	pf->vsi_stats[vsi->idx] = vsi_stat;
571
572	return 0;
573
574err_alloc_rx:
575	kfree(vsi_stat->rx_ring_stats);
576err_alloc_tx:
577	kfree(vsi_stat->tx_ring_stats);
578	kfree(vsi_stat);
579	pf->vsi_stats[vsi->idx] = NULL;
580	return -ENOMEM;
581}
582
583/**
584 * ice_vsi_alloc_def - set default values for already allocated VSI
585 * @vsi: ptr to VSI
586 * @ch: ptr to channel
587 */
588static int
589ice_vsi_alloc_def(struct ice_vsi *vsi, struct ice_channel *ch)
590{
591	if (vsi->type != ICE_VSI_CHNL) {
592		ice_vsi_set_num_qs(vsi);
593		if (ice_vsi_alloc_arrays(vsi))
594			return -ENOMEM;
595	}
596
597	switch (vsi->type) {
598	case ICE_VSI_SWITCHDEV_CTRL:
599		/* Setup eswitch MSIX irq handler for VSI */
600		vsi->irq_handler = ice_eswitch_msix_clean_rings;
601		break;
602	case ICE_VSI_PF:
603		/* Setup default MSIX irq handler for VSI */
604		vsi->irq_handler = ice_msix_clean_rings;
605		break;
606	case ICE_VSI_CTRL:
607		/* Setup ctrl VSI MSIX irq handler */
608		vsi->irq_handler = ice_msix_clean_ctrl_vsi;
609		break;
610	case ICE_VSI_CHNL:
611		if (!ch)
612			return -EINVAL;
613
614		vsi->num_rxq = ch->num_rxq;
615		vsi->num_txq = ch->num_txq;
616		vsi->next_base_q = ch->base_q;
617		break;
618	case ICE_VSI_VF:
619	case ICE_VSI_LB:
620		break;
621	default:
622		ice_vsi_free_arrays(vsi);
623		return -EINVAL;
624	}
625
626	return 0;
627}
628
629/**
630 * ice_vsi_alloc - Allocates the next available struct VSI in the PF
631 * @pf: board private structure
632 *
633 * Reserves a VSI index from the PF and allocates an empty VSI structure
634 * without a type. The VSI structure must later be initialized by calling
635 * ice_vsi_cfg().
636 *
637 * returns a pointer to a VSI on success, NULL on failure.
638 */
639static struct ice_vsi *ice_vsi_alloc(struct ice_pf *pf)
640{
641	struct device *dev = ice_pf_to_dev(pf);
642	struct ice_vsi *vsi = NULL;
643
644	/* Need to protect the allocation of the VSIs at the PF level */
645	mutex_lock(&pf->sw_mutex);
646
647	/* If we have already allocated our maximum number of VSIs,
648	 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
649	 * is available to be populated
650	 */
651	if (pf->next_vsi == ICE_NO_VSI) {
652		dev_dbg(dev, "out of VSI slots!\n");
653		goto unlock_pf;
654	}
655
656	vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
657	if (!vsi)
658		goto unlock_pf;
659
660	vsi->back = pf;
661	set_bit(ICE_VSI_DOWN, vsi->state);
662
663	/* fill slot and make note of the index */
664	vsi->idx = pf->next_vsi;
665	pf->vsi[pf->next_vsi] = vsi;
666
667	/* prepare pf->next_vsi for next use */
668	pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
669					 pf->next_vsi);
670
671unlock_pf:
672	mutex_unlock(&pf->sw_mutex);
673	return vsi;
674}
675
676/**
677 * ice_alloc_fd_res - Allocate FD resource for a VSI
678 * @vsi: pointer to the ice_vsi
679 *
680 * This allocates the FD resources
681 *
682 * Returns 0 on success, -EPERM on no-op or -EIO on failure
683 */
684static int ice_alloc_fd_res(struct ice_vsi *vsi)
685{
686	struct ice_pf *pf = vsi->back;
687	u32 g_val, b_val;
688
689	/* Flow Director filters are only allocated/assigned to the PF VSI or
690	 * CHNL VSI which passes the traffic. The CTRL VSI is only used to
691	 * add/delete filters so resources are not allocated to it
692	 */
693	if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
694		return -EPERM;
695
696	if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF ||
697	      vsi->type == ICE_VSI_CHNL))
698		return -EPERM;
699
700	/* FD filters from guaranteed pool per VSI */
701	g_val = pf->hw.func_caps.fd_fltr_guar;
702	if (!g_val)
703		return -EPERM;
704
705	/* FD filters from best effort pool */
706	b_val = pf->hw.func_caps.fd_fltr_best_effort;
707	if (!b_val)
708		return -EPERM;
709
710	/* PF main VSI gets only 64 FD resources from guaranteed pool
711	 * when ADQ is configured.
712	 */
713#define ICE_PF_VSI_GFLTR	64
714
715	/* determine FD filter resources per VSI from shared(best effort) and
716	 * dedicated pool
717	 */
718	if (vsi->type == ICE_VSI_PF) {
719		vsi->num_gfltr = g_val;
720		/* if MQPRIO is configured, main VSI doesn't get all FD
721		 * resources from guaranteed pool. PF VSI gets 64 FD resources
722		 */
723		if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
724			if (g_val < ICE_PF_VSI_GFLTR)
725				return -EPERM;
726			/* allow bare minimum entries for PF VSI */
727			vsi->num_gfltr = ICE_PF_VSI_GFLTR;
728		}
729
730		/* each VSI gets same "best_effort" quota */
731		vsi->num_bfltr = b_val;
732	} else if (vsi->type == ICE_VSI_VF) {
733		vsi->num_gfltr = 0;
734
735		/* each VSI gets same "best_effort" quota */
736		vsi->num_bfltr = b_val;
737	} else {
738		struct ice_vsi *main_vsi;
739		int numtc;
740
741		main_vsi = ice_get_main_vsi(pf);
742		if (!main_vsi)
743			return -EPERM;
744
745		if (!main_vsi->all_numtc)
746			return -EINVAL;
747
748		/* figure out ADQ numtc */
749		numtc = main_vsi->all_numtc - ICE_CHNL_START_TC;
750
751		/* only one TC but still asking resources for channels,
752		 * invalid config
753		 */
754		if (numtc < ICE_CHNL_START_TC)
755			return -EPERM;
756
757		g_val -= ICE_PF_VSI_GFLTR;
758		/* channel VSIs gets equal share from guaranteed pool */
759		vsi->num_gfltr = g_val / numtc;
760
761		/* each VSI gets same "best_effort" quota */
762		vsi->num_bfltr = b_val;
763	}
764
765	return 0;
766}
767
768/**
769 * ice_vsi_get_qs - Assign queues from PF to VSI
770 * @vsi: the VSI to assign queues to
771 *
772 * Returns 0 on success and a negative value on error
773 */
774static int ice_vsi_get_qs(struct ice_vsi *vsi)
775{
776	struct ice_pf *pf = vsi->back;
777	struct ice_qs_cfg tx_qs_cfg = {
778		.qs_mutex = &pf->avail_q_mutex,
779		.pf_map = pf->avail_txqs,
780		.pf_map_size = pf->max_pf_txqs,
781		.q_count = vsi->alloc_txq,
782		.scatter_count = ICE_MAX_SCATTER_TXQS,
783		.vsi_map = vsi->txq_map,
784		.vsi_map_offset = 0,
785		.mapping_mode = ICE_VSI_MAP_CONTIG
786	};
787	struct ice_qs_cfg rx_qs_cfg = {
788		.qs_mutex = &pf->avail_q_mutex,
789		.pf_map = pf->avail_rxqs,
790		.pf_map_size = pf->max_pf_rxqs,
791		.q_count = vsi->alloc_rxq,
792		.scatter_count = ICE_MAX_SCATTER_RXQS,
793		.vsi_map = vsi->rxq_map,
794		.vsi_map_offset = 0,
795		.mapping_mode = ICE_VSI_MAP_CONTIG
796	};
797	int ret;
798
799	if (vsi->type == ICE_VSI_CHNL)
800		return 0;
801
802	ret = __ice_vsi_get_qs(&tx_qs_cfg);
803	if (ret)
804		return ret;
805	vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
806
807	ret = __ice_vsi_get_qs(&rx_qs_cfg);
808	if (ret)
809		return ret;
810	vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
811
812	return 0;
813}
814
815/**
816 * ice_vsi_put_qs - Release queues from VSI to PF
817 * @vsi: the VSI that is going to release queues
818 */
819static void ice_vsi_put_qs(struct ice_vsi *vsi)
820{
821	struct ice_pf *pf = vsi->back;
822	int i;
823
824	mutex_lock(&pf->avail_q_mutex);
825
826	ice_for_each_alloc_txq(vsi, i) {
827		clear_bit(vsi->txq_map[i], pf->avail_txqs);
828		vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
829	}
830
831	ice_for_each_alloc_rxq(vsi, i) {
832		clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
833		vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
834	}
835
836	mutex_unlock(&pf->avail_q_mutex);
837}
838
839/**
840 * ice_is_safe_mode
841 * @pf: pointer to the PF struct
842 *
843 * returns true if driver is in safe mode, false otherwise
844 */
845bool ice_is_safe_mode(struct ice_pf *pf)
846{
847	return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
848}
849
850/**
851 * ice_is_rdma_ena
852 * @pf: pointer to the PF struct
853 *
854 * returns true if RDMA is currently supported, false otherwise
855 */
856bool ice_is_rdma_ena(struct ice_pf *pf)
857{
858	return test_bit(ICE_FLAG_RDMA_ENA, pf->flags);
859}
860
861/**
862 * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
863 * @vsi: the VSI being cleaned up
864 *
865 * This function deletes RSS input set for all flows that were configured
866 * for this VSI
867 */
868static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
869{
870	struct ice_pf *pf = vsi->back;
871	int status;
872
873	if (ice_is_safe_mode(pf))
874		return;
875
876	status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
877	if (status)
878		dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %d\n",
879			vsi->vsi_num, status);
880}
881
882/**
883 * ice_rss_clean - Delete RSS related VSI structures and configuration
884 * @vsi: the VSI being removed
885 */
886static void ice_rss_clean(struct ice_vsi *vsi)
887{
888	struct ice_pf *pf = vsi->back;
889	struct device *dev;
890
891	dev = ice_pf_to_dev(pf);
892
893	devm_kfree(dev, vsi->rss_hkey_user);
894	devm_kfree(dev, vsi->rss_lut_user);
895
896	ice_vsi_clean_rss_flow_fld(vsi);
897	/* remove RSS replay list */
898	if (!ice_is_safe_mode(pf))
899		ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
900}
901
902/**
903 * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
904 * @vsi: the VSI being configured
905 */
906static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
907{
908	struct ice_hw_common_caps *cap;
909	struct ice_pf *pf = vsi->back;
910	u16 max_rss_size;
911
912	if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
913		vsi->rss_size = 1;
914		return;
915	}
916
917	cap = &pf->hw.func_caps.common_cap;
918	max_rss_size = BIT(cap->rss_table_entry_width);
919	switch (vsi->type) {
920	case ICE_VSI_CHNL:
921	case ICE_VSI_PF:
922		/* PF VSI will inherit RSS instance of PF */
923		vsi->rss_table_size = (u16)cap->rss_table_size;
924		if (vsi->type == ICE_VSI_CHNL)
925			vsi->rss_size = min_t(u16, vsi->num_rxq, max_rss_size);
926		else
927			vsi->rss_size = min_t(u16, num_online_cpus(),
928					      max_rss_size);
929		vsi->rss_lut_type = ICE_LUT_PF;
930		break;
931	case ICE_VSI_SWITCHDEV_CTRL:
932		vsi->rss_table_size = ICE_LUT_VSI_SIZE;
933		vsi->rss_size = min_t(u16, num_online_cpus(), max_rss_size);
934		vsi->rss_lut_type = ICE_LUT_VSI;
935		break;
936	case ICE_VSI_VF:
937		/* VF VSI will get a small RSS table.
938		 * For VSI_LUT, LUT size should be set to 64 bytes.
939		 */
940		vsi->rss_table_size = ICE_LUT_VSI_SIZE;
941		vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
942		vsi->rss_lut_type = ICE_LUT_VSI;
943		break;
944	case ICE_VSI_LB:
945		break;
946	default:
947		dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
948			ice_vsi_type_str(vsi->type));
949		break;
950	}
951}
952
953/**
954 * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
955 * @hw: HW structure used to determine the VLAN mode of the device
956 * @ctxt: the VSI context being set
957 *
958 * This initializes a default VSI context for all sections except the Queues.
959 */
960static void ice_set_dflt_vsi_ctx(struct ice_hw *hw, struct ice_vsi_ctx *ctxt)
961{
962	u32 table = 0;
963
964	memset(&ctxt->info, 0, sizeof(ctxt->info));
965	/* VSI's should be allocated from shared pool */
966	ctxt->alloc_from_pool = true;
967	/* Src pruning enabled by default */
968	ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
969	/* Traffic from VSI can be sent to LAN */
970	ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
971	/* allow all untagged/tagged packets by default on Tx */
972	ctxt->info.inner_vlan_flags = ((ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL &
973				  ICE_AQ_VSI_INNER_VLAN_TX_MODE_M) >>
974				 ICE_AQ_VSI_INNER_VLAN_TX_MODE_S);
975	/* SVM - by default bits 3 and 4 in inner_vlan_flags are 0's which
976	 * results in legacy behavior (show VLAN, DEI, and UP) in descriptor.
977	 *
978	 * DVM - leave inner VLAN in packet by default
979	 */
980	if (ice_is_dvm_ena(hw)) {
981		ctxt->info.inner_vlan_flags |=
982			FIELD_PREP(ICE_AQ_VSI_INNER_VLAN_EMODE_M,
983				   ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING);
984		ctxt->info.outer_vlan_flags =
985			(ICE_AQ_VSI_OUTER_VLAN_TX_MODE_ALL <<
986			 ICE_AQ_VSI_OUTER_VLAN_TX_MODE_S) &
987			ICE_AQ_VSI_OUTER_VLAN_TX_MODE_M;
988		ctxt->info.outer_vlan_flags |=
989			(ICE_AQ_VSI_OUTER_TAG_VLAN_8100 <<
990			 ICE_AQ_VSI_OUTER_TAG_TYPE_S) &
991			ICE_AQ_VSI_OUTER_TAG_TYPE_M;
992		ctxt->info.outer_vlan_flags |=
993			FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_EMODE_M,
994				   ICE_AQ_VSI_OUTER_VLAN_EMODE_NOTHING);
995	}
996	/* Have 1:1 UP mapping for both ingress/egress tables */
997	table |= ICE_UP_TABLE_TRANSLATE(0, 0);
998	table |= ICE_UP_TABLE_TRANSLATE(1, 1);
999	table |= ICE_UP_TABLE_TRANSLATE(2, 2);
1000	table |= ICE_UP_TABLE_TRANSLATE(3, 3);
1001	table |= ICE_UP_TABLE_TRANSLATE(4, 4);
1002	table |= ICE_UP_TABLE_TRANSLATE(5, 5);
1003	table |= ICE_UP_TABLE_TRANSLATE(6, 6);
1004	table |= ICE_UP_TABLE_TRANSLATE(7, 7);
1005	ctxt->info.ingress_table = cpu_to_le32(table);
1006	ctxt->info.egress_table = cpu_to_le32(table);
1007	/* Have 1:1 UP mapping for outer to inner UP table */
1008	ctxt->info.outer_up_table = cpu_to_le32(table);
1009	/* No Outer tag support outer_tag_flags remains to zero */
1010}
1011
1012/**
1013 * ice_vsi_setup_q_map - Setup a VSI queue map
1014 * @vsi: the VSI being configured
1015 * @ctxt: VSI context structure
1016 */
1017static int ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1018{
1019	u16 offset = 0, qmap = 0, tx_count = 0, rx_count = 0, pow = 0;
1020	u16 num_txq_per_tc, num_rxq_per_tc;
1021	u16 qcount_tx = vsi->alloc_txq;
1022	u16 qcount_rx = vsi->alloc_rxq;
1023	u8 netdev_tc = 0;
1024	int i;
1025
1026	if (!vsi->tc_cfg.numtc) {
1027		/* at least TC0 should be enabled by default */
1028		vsi->tc_cfg.numtc = 1;
1029		vsi->tc_cfg.ena_tc = 1;
1030	}
1031
1032	num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
1033	if (!num_rxq_per_tc)
1034		num_rxq_per_tc = 1;
1035	num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
1036	if (!num_txq_per_tc)
1037		num_txq_per_tc = 1;
1038
1039	/* find the (rounded up) power-of-2 of qcount */
1040	pow = (u16)order_base_2(num_rxq_per_tc);
1041
1042	/* TC mapping is a function of the number of Rx queues assigned to the
1043	 * VSI for each traffic class and the offset of these queues.
1044	 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
1045	 * queues allocated to TC0. No:of queues is a power-of-2.
1046	 *
1047	 * If TC is not enabled, the queue offset is set to 0, and allocate one
1048	 * queue, this way, traffic for the given TC will be sent to the default
1049	 * queue.
1050	 *
1051	 * Setup number and offset of Rx queues for all TCs for the VSI
1052	 */
1053	ice_for_each_traffic_class(i) {
1054		if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
1055			/* TC is not enabled */
1056			vsi->tc_cfg.tc_info[i].qoffset = 0;
1057			vsi->tc_cfg.tc_info[i].qcount_rx = 1;
1058			vsi->tc_cfg.tc_info[i].qcount_tx = 1;
1059			vsi->tc_cfg.tc_info[i].netdev_tc = 0;
1060			ctxt->info.tc_mapping[i] = 0;
1061			continue;
1062		}
1063
1064		/* TC is enabled */
1065		vsi->tc_cfg.tc_info[i].qoffset = offset;
1066		vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
1067		vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
1068		vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
1069
1070		qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
1071			ICE_AQ_VSI_TC_Q_OFFSET_M) |
1072			((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
1073			 ICE_AQ_VSI_TC_Q_NUM_M);
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 = ((dflt_q << ICE_AQ_VSI_FD_DEF_Q_S) &
1157	       ICE_AQ_VSI_FD_DEF_Q_M);
1158	/* target queue or queue group to the FD filter */
1159	val |= ((dflt_q_group << ICE_AQ_VSI_FD_DEF_GRP_S) &
1160		ICE_AQ_VSI_FD_DEF_GRP_M);
1161	ctxt->info.fd_def_q = cpu_to_le16(val);
1162	/* queue index on which FD filter completion is reported */
1163	val = ((report_q << ICE_AQ_VSI_FD_REPORT_Q_S) &
1164	       ICE_AQ_VSI_FD_REPORT_Q_M);
1165	/* priority of the default qindex action */
1166	val |= ((dflt_q_prio << ICE_AQ_VSI_FD_DEF_PRIORITY_S) &
1167		ICE_AQ_VSI_FD_DEF_PRIORITY_M);
1168	ctxt->info.fd_report_opt = cpu_to_le16(val);
1169}
1170
1171/**
1172 * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
1173 * @ctxt: the VSI context being set
1174 * @vsi: the VSI being configured
1175 */
1176static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1177{
1178	u8 lut_type, hash_type;
1179	struct device *dev;
1180	struct ice_pf *pf;
1181
1182	pf = vsi->back;
1183	dev = ice_pf_to_dev(pf);
1184
1185	switch (vsi->type) {
1186	case ICE_VSI_CHNL:
1187	case ICE_VSI_PF:
1188		/* PF VSI will inherit RSS instance of PF */
1189		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
1190		hash_type = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
1191		break;
1192	case ICE_VSI_VF:
1193		/* VF VSI will gets a small RSS table which is a VSI LUT type */
1194		lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
1195		hash_type = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
1196		break;
1197	default:
1198		dev_dbg(dev, "Unsupported VSI type %s\n",
1199			ice_vsi_type_str(vsi->type));
1200		return;
1201	}
1202
1203	ctxt->info.q_opt_rss = ((lut_type << ICE_AQ_VSI_Q_OPT_RSS_LUT_S) &
1204				ICE_AQ_VSI_Q_OPT_RSS_LUT_M) |
1205				(hash_type & ICE_AQ_VSI_Q_OPT_RSS_HASH_M);
1206}
1207
1208static void
1209ice_chnl_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1210{
1211	struct ice_pf *pf = vsi->back;
1212	u16 qcount, qmap;
1213	u8 offset = 0;
1214	int pow;
1215
1216	qcount = min_t(int, vsi->num_rxq, pf->num_lan_msix);
1217
1218	pow = order_base_2(qcount);
1219	qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
1220		 ICE_AQ_VSI_TC_Q_OFFSET_M) |
1221		 ((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
1222		   ICE_AQ_VSI_TC_Q_NUM_M);
1223
1224	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1225	ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1226	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->next_base_q);
1227	ctxt->info.q_mapping[1] = cpu_to_le16(qcount);
1228}
1229
1230/**
1231 * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
1232 * @vsi: VSI to check whether or not VLAN pruning is enabled.
1233 *
1234 * returns true if Rx VLAN pruning is enabled and false otherwise.
1235 */
1236static bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
1237{
1238	return vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1239}
1240
1241/**
1242 * ice_vsi_init - Create and initialize a VSI
1243 * @vsi: the VSI being configured
1244 * @vsi_flags: VSI configuration flags
1245 *
1246 * Set ICE_FLAG_VSI_INIT to initialize a new VSI context, clear it to
1247 * reconfigure an existing context.
1248 *
1249 * This initializes a VSI context depending on the VSI type to be added and
1250 * passes it down to the add_vsi aq command to create a new VSI.
1251 */
1252static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
1253{
1254	struct ice_pf *pf = vsi->back;
1255	struct ice_hw *hw = &pf->hw;
1256	struct ice_vsi_ctx *ctxt;
1257	struct device *dev;
1258	int ret = 0;
1259
1260	dev = ice_pf_to_dev(pf);
1261	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1262	if (!ctxt)
1263		return -ENOMEM;
1264
1265	switch (vsi->type) {
1266	case ICE_VSI_CTRL:
1267	case ICE_VSI_LB:
1268	case ICE_VSI_PF:
1269		ctxt->flags = ICE_AQ_VSI_TYPE_PF;
1270		break;
1271	case ICE_VSI_SWITCHDEV_CTRL:
1272	case ICE_VSI_CHNL:
1273		ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
1274		break;
1275	case ICE_VSI_VF:
1276		ctxt->flags = ICE_AQ_VSI_TYPE_VF;
1277		/* VF number here is the absolute VF number (0-255) */
1278		ctxt->vf_num = vsi->vf->vf_id + hw->func_caps.vf_base_id;
1279		break;
1280	default:
1281		ret = -ENODEV;
1282		goto out;
1283	}
1284
1285	/* Handle VLAN pruning for channel VSI if main VSI has VLAN
1286	 * prune enabled
1287	 */
1288	if (vsi->type == ICE_VSI_CHNL) {
1289		struct ice_vsi *main_vsi;
1290
1291		main_vsi = ice_get_main_vsi(pf);
1292		if (main_vsi && ice_vsi_is_vlan_pruning_ena(main_vsi))
1293			ctxt->info.sw_flags2 |=
1294				ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1295		else
1296			ctxt->info.sw_flags2 &=
1297				~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1298	}
1299
1300	ice_set_dflt_vsi_ctx(hw, ctxt);
1301	if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
1302		ice_set_fd_vsi_ctx(ctxt, vsi);
1303	/* if the switch is in VEB mode, allow VSI loopback */
1304	if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
1305		ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
1306
1307	/* Set LUT type and HASH type if RSS is enabled */
1308	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1309	    vsi->type != ICE_VSI_CTRL) {
1310		ice_set_rss_vsi_ctx(ctxt, vsi);
1311		/* if updating VSI context, make sure to set valid_section:
1312		 * to indicate which section of VSI context being updated
1313		 */
1314		if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1315			ctxt->info.valid_sections |=
1316				cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1317	}
1318
1319	ctxt->info.sw_id = vsi->port_info->sw_id;
1320	if (vsi->type == ICE_VSI_CHNL) {
1321		ice_chnl_vsi_setup_q_map(vsi, ctxt);
1322	} else {
1323		ret = ice_vsi_setup_q_map(vsi, ctxt);
1324		if (ret)
1325			goto out;
1326
1327		if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1328			/* means VSI being updated */
1329			/* must to indicate which section of VSI context are
1330			 * being modified
1331			 */
1332			ctxt->info.valid_sections |=
1333				cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1334	}
1335
1336	/* Allow control frames out of main VSI */
1337	if (vsi->type == ICE_VSI_PF) {
1338		ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1339		ctxt->info.valid_sections |=
1340			cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1341	}
1342
1343	if (vsi_flags & ICE_VSI_FLAG_INIT) {
1344		ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1345		if (ret) {
1346			dev_err(dev, "Add VSI failed, err %d\n", ret);
1347			ret = -EIO;
1348			goto out;
1349		}
1350	} else {
1351		ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1352		if (ret) {
1353			dev_err(dev, "Update VSI failed, err %d\n", ret);
1354			ret = -EIO;
1355			goto out;
1356		}
1357	}
1358
1359	/* keep context for update VSI operations */
1360	vsi->info = ctxt->info;
1361
1362	/* record VSI number returned */
1363	vsi->vsi_num = ctxt->vsi_num;
1364
1365out:
1366	kfree(ctxt);
1367	return ret;
1368}
1369
1370/**
1371 * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1372 * @vsi: the VSI having rings deallocated
1373 */
1374static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1375{
1376	int i;
1377
1378	/* Avoid stale references by clearing map from vector to ring */
1379	if (vsi->q_vectors) {
1380		ice_for_each_q_vector(vsi, i) {
1381			struct ice_q_vector *q_vector = vsi->q_vectors[i];
1382
1383			if (q_vector) {
1384				q_vector->tx.tx_ring = NULL;
1385				q_vector->rx.rx_ring = NULL;
1386			}
1387		}
1388	}
1389
1390	if (vsi->tx_rings) {
1391		ice_for_each_alloc_txq(vsi, i) {
1392			if (vsi->tx_rings[i]) {
1393				kfree_rcu(vsi->tx_rings[i], rcu);
1394				WRITE_ONCE(vsi->tx_rings[i], NULL);
1395			}
1396		}
1397	}
1398	if (vsi->rx_rings) {
1399		ice_for_each_alloc_rxq(vsi, i) {
1400			if (vsi->rx_rings[i]) {
1401				kfree_rcu(vsi->rx_rings[i], rcu);
1402				WRITE_ONCE(vsi->rx_rings[i], NULL);
1403			}
1404		}
1405	}
1406}
1407
1408/**
1409 * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1410 * @vsi: VSI which is having rings allocated
1411 */
1412static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1413{
1414	bool dvm_ena = ice_is_dvm_ena(&vsi->back->hw);
1415	struct ice_pf *pf = vsi->back;
1416	struct device *dev;
1417	u16 i;
1418
1419	dev = ice_pf_to_dev(pf);
1420	/* Allocate Tx rings */
1421	ice_for_each_alloc_txq(vsi, i) {
1422		struct ice_tx_ring *ring;
1423
1424		/* allocate with kzalloc(), free with kfree_rcu() */
1425		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1426
1427		if (!ring)
1428			goto err_out;
1429
1430		ring->q_index = i;
1431		ring->reg_idx = vsi->txq_map[i];
1432		ring->vsi = vsi;
1433		ring->tx_tstamps = &pf->ptp.port.tx;
1434		ring->dev = dev;
1435		ring->count = vsi->num_tx_desc;
1436		ring->txq_teid = ICE_INVAL_TEID;
1437		if (dvm_ena)
1438			ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG2;
1439		else
1440			ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG1;
1441		WRITE_ONCE(vsi->tx_rings[i], ring);
1442	}
1443
1444	/* Allocate Rx rings */
1445	ice_for_each_alloc_rxq(vsi, i) {
1446		struct ice_rx_ring *ring;
1447
1448		/* allocate with kzalloc(), free with kfree_rcu() */
1449		ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1450		if (!ring)
1451			goto err_out;
1452
1453		ring->q_index = i;
1454		ring->reg_idx = vsi->rxq_map[i];
1455		ring->vsi = vsi;
1456		ring->netdev = vsi->netdev;
1457		ring->dev = dev;
1458		ring->count = vsi->num_rx_desc;
1459		ring->cached_phctime = pf->ptp.cached_phc_time;
1460		WRITE_ONCE(vsi->rx_rings[i], ring);
1461	}
1462
1463	return 0;
1464
1465err_out:
1466	ice_vsi_clear_rings(vsi);
1467	return -ENOMEM;
1468}
1469
1470/**
1471 * ice_vsi_manage_rss_lut - disable/enable RSS
1472 * @vsi: the VSI being changed
1473 * @ena: boolean value indicating if this is an enable or disable request
1474 *
1475 * In the event of disable request for RSS, this function will zero out RSS
1476 * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1477 * LUT.
1478 */
1479void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1480{
1481	u8 *lut;
1482
1483	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1484	if (!lut)
1485		return;
1486
1487	if (ena) {
1488		if (vsi->rss_lut_user)
1489			memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1490		else
1491			ice_fill_rss_lut(lut, vsi->rss_table_size,
1492					 vsi->rss_size);
1493	}
1494
1495	ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1496	kfree(lut);
1497}
1498
1499/**
1500 * ice_vsi_cfg_crc_strip - Configure CRC stripping for a VSI
1501 * @vsi: VSI to be configured
1502 * @disable: set to true to have FCS / CRC in the frame data
1503 */
1504void ice_vsi_cfg_crc_strip(struct ice_vsi *vsi, bool disable)
1505{
1506	int i;
1507
1508	ice_for_each_rxq(vsi, i)
1509		if (disable)
1510			vsi->rx_rings[i]->flags |= ICE_RX_FLAGS_CRC_STRIP_DIS;
1511		else
1512			vsi->rx_rings[i]->flags &= ~ICE_RX_FLAGS_CRC_STRIP_DIS;
1513}
1514
1515/**
1516 * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1517 * @vsi: VSI to be configured
1518 */
1519int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1520{
1521	struct ice_pf *pf = vsi->back;
1522	struct device *dev;
1523	u8 *lut, *key;
1524	int err;
1525
1526	dev = ice_pf_to_dev(pf);
1527	if (vsi->type == ICE_VSI_PF && vsi->ch_rss_size &&
1528	    (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))) {
1529		vsi->rss_size = min_t(u16, vsi->rss_size, vsi->ch_rss_size);
1530	} else {
1531		vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1532
1533		/* If orig_rss_size is valid and it is less than determined
1534		 * main VSI's rss_size, update main VSI's rss_size to be
1535		 * orig_rss_size so that when tc-qdisc is deleted, main VSI
1536		 * RSS table gets programmed to be correct (whatever it was
1537		 * to begin with (prior to setup-tc for ADQ config)
1538		 */
1539		if (vsi->orig_rss_size && vsi->rss_size < vsi->orig_rss_size &&
1540		    vsi->orig_rss_size <= vsi->num_rxq) {
1541			vsi->rss_size = vsi->orig_rss_size;
1542			/* now orig_rss_size is used, reset it to zero */
1543			vsi->orig_rss_size = 0;
1544		}
1545	}
1546
1547	lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1548	if (!lut)
1549		return -ENOMEM;
1550
1551	if (vsi->rss_lut_user)
1552		memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1553	else
1554		ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1555
1556	err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1557	if (err) {
1558		dev_err(dev, "set_rss_lut failed, error %d\n", err);
1559		goto ice_vsi_cfg_rss_exit;
1560	}
1561
1562	key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1563	if (!key) {
1564		err = -ENOMEM;
1565		goto ice_vsi_cfg_rss_exit;
1566	}
1567
1568	if (vsi->rss_hkey_user)
1569		memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1570	else
1571		netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1572
1573	err = ice_set_rss_key(vsi, key);
1574	if (err)
1575		dev_err(dev, "set_rss_key failed, error %d\n", err);
1576
1577	kfree(key);
1578ice_vsi_cfg_rss_exit:
1579	kfree(lut);
1580	return err;
1581}
1582
1583/**
1584 * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1585 * @vsi: VSI to be configured
1586 *
1587 * This function will only be called during the VF VSI setup. Upon successful
1588 * completion of package download, this function will configure default RSS
1589 * input sets for VF VSI.
1590 */
1591static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1592{
1593	struct ice_pf *pf = vsi->back;
1594	struct device *dev;
1595	int status;
1596
1597	dev = ice_pf_to_dev(pf);
1598	if (ice_is_safe_mode(pf)) {
1599		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1600			vsi->vsi_num);
1601		return;
1602	}
1603
1604	status = ice_add_avf_rss_cfg(&pf->hw, vsi->idx, ICE_DEFAULT_RSS_HENA);
1605	if (status)
1606		dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
1607			vsi->vsi_num, status);
1608}
1609
1610/**
1611 * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1612 * @vsi: VSI to be configured
1613 *
1614 * This function will only be called after successful download package call
1615 * during initialization of PF. Since the downloaded package will erase the
1616 * RSS section, this function will configure RSS input sets for different
1617 * flow types. The last profile added has the highest priority, therefore 2
1618 * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1619 * (i.e. IPv4 src/dst TCP src/dst port).
1620 */
1621static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1622{
1623	u16 vsi_handle = vsi->idx, vsi_num = vsi->vsi_num;
1624	struct ice_pf *pf = vsi->back;
1625	struct ice_hw *hw = &pf->hw;
1626	struct device *dev;
1627	int status;
1628
1629	dev = ice_pf_to_dev(pf);
1630	if (ice_is_safe_mode(pf)) {
1631		dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1632			vsi_num);
1633		return;
1634	}
1635	/* configure RSS for IPv4 with input set IP src/dst */
1636	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1637				 ICE_FLOW_SEG_HDR_IPV4);
1638	if (status)
1639		dev_dbg(dev, "ice_add_rss_cfg failed for ipv4 flow, vsi = %d, error = %d\n",
1640			vsi_num, status);
1641
1642	/* configure RSS for IPv6 with input set IPv6 src/dst */
1643	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1644				 ICE_FLOW_SEG_HDR_IPV6);
1645	if (status)
1646		dev_dbg(dev, "ice_add_rss_cfg failed for ipv6 flow, vsi = %d, error = %d\n",
1647			vsi_num, status);
1648
1649	/* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1650	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV4,
1651				 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4);
1652	if (status)
1653		dev_dbg(dev, "ice_add_rss_cfg failed for tcp4 flow, vsi = %d, error = %d\n",
1654			vsi_num, status);
1655
1656	/* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1657	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV4,
1658				 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4);
1659	if (status)
1660		dev_dbg(dev, "ice_add_rss_cfg failed for udp4 flow, vsi = %d, error = %d\n",
1661			vsi_num, status);
1662
1663	/* configure RSS for sctp4 with input set IP src/dst */
1664	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1665				 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4);
1666	if (status)
1667		dev_dbg(dev, "ice_add_rss_cfg failed for sctp4 flow, vsi = %d, error = %d\n",
1668			vsi_num, status);
1669
1670	/* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1671	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV6,
1672				 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6);
1673	if (status)
1674		dev_dbg(dev, "ice_add_rss_cfg failed for tcp6 flow, vsi = %d, error = %d\n",
1675			vsi_num, status);
1676
1677	/* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1678	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV6,
1679				 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6);
1680	if (status)
1681		dev_dbg(dev, "ice_add_rss_cfg failed for udp6 flow, vsi = %d, error = %d\n",
1682			vsi_num, status);
1683
1684	/* configure RSS for sctp6 with input set IPv6 src/dst */
1685	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1686				 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6);
1687	if (status)
1688		dev_dbg(dev, "ice_add_rss_cfg failed for sctp6 flow, vsi = %d, error = %d\n",
1689			vsi_num, status);
1690
1691	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_ESP_SPI,
1692				 ICE_FLOW_SEG_HDR_ESP);
1693	if (status)
1694		dev_dbg(dev, "ice_add_rss_cfg failed for esp/spi flow, vsi = %d, error = %d\n",
1695			vsi_num, status);
1696}
1697
1698/**
1699 * ice_vsi_cfg_frame_size - setup max frame size and Rx buffer length
1700 * @vsi: VSI
1701 */
1702static void ice_vsi_cfg_frame_size(struct ice_vsi *vsi)
1703{
1704	if (!vsi->netdev || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) {
1705		vsi->max_frame = ICE_MAX_FRAME_LEGACY_RX;
1706		vsi->rx_buf_len = ICE_RXBUF_1664;
1707#if (PAGE_SIZE < 8192)
1708	} else if (!ICE_2K_TOO_SMALL_WITH_PADDING &&
1709		   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
1710		vsi->max_frame = ICE_RXBUF_1536 - NET_IP_ALIGN;
1711		vsi->rx_buf_len = ICE_RXBUF_1536 - NET_IP_ALIGN;
1712#endif
1713	} else {
1714		vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1715		vsi->rx_buf_len = ICE_RXBUF_3072;
1716	}
1717}
1718
1719/**
1720 * ice_pf_state_is_nominal - checks the PF for nominal state
1721 * @pf: pointer to PF to check
1722 *
1723 * Check the PF's state for a collection of bits that would indicate
1724 * the PF is in a state that would inhibit normal operation for
1725 * driver functionality.
1726 *
1727 * Returns true if PF is in a nominal state, false otherwise
1728 */
1729bool ice_pf_state_is_nominal(struct ice_pf *pf)
1730{
1731	DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1732
1733	if (!pf)
1734		return false;
1735
1736	bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1737	if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1738		return false;
1739
1740	return true;
1741}
1742
1743/**
1744 * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1745 * @vsi: the VSI to be updated
1746 */
1747void ice_update_eth_stats(struct ice_vsi *vsi)
1748{
1749	struct ice_eth_stats *prev_es, *cur_es;
1750	struct ice_hw *hw = &vsi->back->hw;
1751	struct ice_pf *pf = vsi->back;
1752	u16 vsi_num = vsi->vsi_num;    /* HW absolute index of a VSI */
1753
1754	prev_es = &vsi->eth_stats_prev;
1755	cur_es = &vsi->eth_stats;
1756
1757	if (ice_is_reset_in_progress(pf->state))
1758		vsi->stat_offsets_loaded = false;
1759
1760	ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1761			  &prev_es->rx_bytes, &cur_es->rx_bytes);
1762
1763	ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1764			  &prev_es->rx_unicast, &cur_es->rx_unicast);
1765
1766	ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1767			  &prev_es->rx_multicast, &cur_es->rx_multicast);
1768
1769	ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1770			  &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1771
1772	ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1773			  &prev_es->rx_discards, &cur_es->rx_discards);
1774
1775	ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1776			  &prev_es->tx_bytes, &cur_es->tx_bytes);
1777
1778	ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1779			  &prev_es->tx_unicast, &cur_es->tx_unicast);
1780
1781	ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1782			  &prev_es->tx_multicast, &cur_es->tx_multicast);
1783
1784	ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1785			  &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1786
1787	ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1788			  &prev_es->tx_errors, &cur_es->tx_errors);
1789
1790	vsi->stat_offsets_loaded = true;
1791}
1792
1793/**
1794 * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1795 * @hw: HW pointer
1796 * @pf_q: index of the Rx queue in the PF's queue space
1797 * @rxdid: flexible descriptor RXDID
1798 * @prio: priority for the RXDID for this queue
1799 * @ena_ts: true to enable timestamp and false to disable timestamp
1800 */
1801void
1802ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio,
1803			bool ena_ts)
1804{
1805	int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1806
1807	/* clear any previous values */
1808	regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1809		    QRXFLXP_CNTXT_RXDID_PRIO_M |
1810		    QRXFLXP_CNTXT_TS_M);
1811
1812	regval |= (rxdid << QRXFLXP_CNTXT_RXDID_IDX_S) &
1813		QRXFLXP_CNTXT_RXDID_IDX_M;
1814
1815	regval |= (prio << QRXFLXP_CNTXT_RXDID_PRIO_S) &
1816		QRXFLXP_CNTXT_RXDID_PRIO_M;
1817
1818	if (ena_ts)
1819		/* Enable TimeSync on this queue */
1820		regval |= QRXFLXP_CNTXT_TS_M;
1821
1822	wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1823}
1824
1825int ice_vsi_cfg_single_rxq(struct ice_vsi *vsi, u16 q_idx)
1826{
1827	if (q_idx >= vsi->num_rxq)
1828		return -EINVAL;
1829
1830	return ice_vsi_cfg_rxq(vsi->rx_rings[q_idx]);
1831}
1832
1833int ice_vsi_cfg_single_txq(struct ice_vsi *vsi, struct ice_tx_ring **tx_rings, u16 q_idx)
1834{
1835	struct ice_aqc_add_tx_qgrp *qg_buf;
1836	int err;
1837
1838	if (q_idx >= vsi->alloc_txq || !tx_rings || !tx_rings[q_idx])
1839		return -EINVAL;
1840
1841	qg_buf = kzalloc(struct_size(qg_buf, txqs, 1), GFP_KERNEL);
1842	if (!qg_buf)
1843		return -ENOMEM;
1844
1845	qg_buf->num_txqs = 1;
1846
1847	err = ice_vsi_cfg_txq(vsi, tx_rings[q_idx], qg_buf);
1848	kfree(qg_buf);
1849	return err;
1850}
1851
1852/**
1853 * ice_vsi_cfg_rxqs - Configure the VSI for Rx
1854 * @vsi: the VSI being configured
1855 *
1856 * Return 0 on success and a negative value on error
1857 * Configure the Rx VSI for operation.
1858 */
1859int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
1860{
1861	u16 i;
1862
1863	if (vsi->type == ICE_VSI_VF)
1864		goto setup_rings;
1865
1866	ice_vsi_cfg_frame_size(vsi);
1867setup_rings:
1868	/* set up individual rings */
1869	ice_for_each_rxq(vsi, i) {
1870		int err = ice_vsi_cfg_rxq(vsi->rx_rings[i]);
1871
1872		if (err)
1873			return err;
1874	}
1875
1876	return 0;
1877}
1878
1879/**
1880 * ice_vsi_cfg_txqs - Configure the VSI for Tx
1881 * @vsi: the VSI being configured
1882 * @rings: Tx ring array to be configured
1883 * @count: number of Tx ring array elements
1884 *
1885 * Return 0 on success and a negative value on error
1886 * Configure the Tx VSI for operation.
1887 */
1888static int
1889ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_tx_ring **rings, u16 count)
1890{
1891	struct ice_aqc_add_tx_qgrp *qg_buf;
1892	u16 q_idx = 0;
1893	int err = 0;
1894
1895	qg_buf = kzalloc(struct_size(qg_buf, txqs, 1), GFP_KERNEL);
1896	if (!qg_buf)
1897		return -ENOMEM;
1898
1899	qg_buf->num_txqs = 1;
1900
1901	for (q_idx = 0; q_idx < count; q_idx++) {
1902		err = ice_vsi_cfg_txq(vsi, rings[q_idx], qg_buf);
1903		if (err)
1904			goto err_cfg_txqs;
1905	}
1906
1907err_cfg_txqs:
1908	kfree(qg_buf);
1909	return err;
1910}
1911
1912/**
1913 * ice_vsi_cfg_lan_txqs - Configure the VSI for Tx
1914 * @vsi: the VSI being configured
1915 *
1916 * Return 0 on success and a negative value on error
1917 * Configure the Tx VSI for operation.
1918 */
1919int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
1920{
1921	return ice_vsi_cfg_txqs(vsi, vsi->tx_rings, vsi->num_txq);
1922}
1923
1924/**
1925 * ice_vsi_cfg_xdp_txqs - Configure Tx queues dedicated for XDP in given VSI
1926 * @vsi: the VSI being configured
1927 *
1928 * Return 0 on success and a negative value on error
1929 * Configure the Tx queues dedicated for XDP in given VSI for operation.
1930 */
1931int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
1932{
1933	int ret;
1934	int i;
1935
1936	ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings, vsi->num_xdp_txq);
1937	if (ret)
1938		return ret;
1939
1940	ice_for_each_rxq(vsi, i)
1941		ice_tx_xsk_pool(vsi, i);
1942
1943	return 0;
1944}
1945
1946/**
1947 * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1948 * @intrl: interrupt rate limit in usecs
1949 * @gran: interrupt rate limit granularity in usecs
1950 *
1951 * This function converts a decimal interrupt rate limit in usecs to the format
1952 * expected by firmware.
1953 */
1954static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1955{
1956	u32 val = intrl / gran;
1957
1958	if (val)
1959		return val | GLINT_RATE_INTRL_ENA_M;
1960	return 0;
1961}
1962
1963/**
1964 * ice_write_intrl - write throttle rate limit to interrupt specific register
1965 * @q_vector: pointer to interrupt specific structure
1966 * @intrl: throttle rate limit in microseconds to write
1967 */
1968void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1969{
1970	struct ice_hw *hw = &q_vector->vsi->back->hw;
1971
1972	wr32(hw, GLINT_RATE(q_vector->reg_idx),
1973	     ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1974}
1975
1976static struct ice_q_vector *ice_pull_qvec_from_rc(struct ice_ring_container *rc)
1977{
1978	switch (rc->type) {
1979	case ICE_RX_CONTAINER:
1980		if (rc->rx_ring)
1981			return rc->rx_ring->q_vector;
1982		break;
1983	case ICE_TX_CONTAINER:
1984		if (rc->tx_ring)
1985			return rc->tx_ring->q_vector;
1986		break;
1987	default:
1988		break;
1989	}
1990
1991	return NULL;
1992}
1993
1994/**
1995 * __ice_write_itr - write throttle rate to register
1996 * @q_vector: pointer to interrupt data structure
1997 * @rc: pointer to ring container
1998 * @itr: throttle rate in microseconds to write
1999 */
2000static void __ice_write_itr(struct ice_q_vector *q_vector,
2001			    struct ice_ring_container *rc, u16 itr)
2002{
2003	struct ice_hw *hw = &q_vector->vsi->back->hw;
2004
2005	wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
2006	     ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
2007}
2008
2009/**
2010 * ice_write_itr - write throttle rate to queue specific register
2011 * @rc: pointer to ring container
2012 * @itr: throttle rate in microseconds to write
2013 */
2014void ice_write_itr(struct ice_ring_container *rc, u16 itr)
2015{
2016	struct ice_q_vector *q_vector;
2017
2018	q_vector = ice_pull_qvec_from_rc(rc);
2019	if (!q_vector)
2020		return;
2021
2022	__ice_write_itr(q_vector, rc, itr);
2023}
2024
2025/**
2026 * ice_set_q_vector_intrl - set up interrupt rate limiting
2027 * @q_vector: the vector to be configured
2028 *
2029 * Interrupt rate limiting is local to the vector, not per-queue so we must
2030 * detect if either ring container has dynamic moderation enabled to decide
2031 * what to set the interrupt rate limit to via INTRL settings. In the case that
2032 * dynamic moderation is disabled on both, write the value with the cached
2033 * setting to make sure INTRL register matches the user visible value.
2034 */
2035void ice_set_q_vector_intrl(struct ice_q_vector *q_vector)
2036{
2037	if (ITR_IS_DYNAMIC(&q_vector->tx) || ITR_IS_DYNAMIC(&q_vector->rx)) {
2038		/* in the case of dynamic enabled, cap each vector to no more
2039		 * than (4 us) 250,000 ints/sec, which allows low latency
2040		 * but still less than 500,000 interrupts per second, which
2041		 * reduces CPU a bit in the case of the lowest latency
2042		 * setting. The 4 here is a value in microseconds.
2043		 */
2044		ice_write_intrl(q_vector, 4);
2045	} else {
2046		ice_write_intrl(q_vector, q_vector->intrl);
2047	}
2048}
2049
2050/**
2051 * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
2052 * @vsi: the VSI being configured
2053 *
2054 * This configures MSIX mode interrupts for the PF VSI, and should not be used
2055 * for the VF VSI.
2056 */
2057void ice_vsi_cfg_msix(struct ice_vsi *vsi)
2058{
2059	struct ice_pf *pf = vsi->back;
2060	struct ice_hw *hw = &pf->hw;
2061	u16 txq = 0, rxq = 0;
2062	int i, q;
2063
2064	ice_for_each_q_vector(vsi, i) {
2065		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2066		u16 reg_idx = q_vector->reg_idx;
2067
2068		ice_cfg_itr(hw, q_vector);
2069
2070		/* Both Transmit Queue Interrupt Cause Control register
2071		 * and Receive Queue Interrupt Cause control register
2072		 * expects MSIX_INDX field to be the vector index
2073		 * within the function space and not the absolute
2074		 * vector index across PF or across device.
2075		 * For SR-IOV VF VSIs queue vector index always starts
2076		 * with 1 since first vector index(0) is used for OICR
2077		 * in VF space. Since VMDq and other PF VSIs are within
2078		 * the PF function space, use the vector index that is
2079		 * tracked for this PF.
2080		 */
2081		for (q = 0; q < q_vector->num_ring_tx; q++) {
2082			ice_cfg_txq_interrupt(vsi, txq, reg_idx,
2083					      q_vector->tx.itr_idx);
2084			txq++;
2085		}
2086
2087		for (q = 0; q < q_vector->num_ring_rx; q++) {
2088			ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
2089					      q_vector->rx.itr_idx);
2090			rxq++;
2091		}
2092	}
2093}
2094
2095/**
2096 * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
2097 * @vsi: the VSI whose rings are to be enabled
2098 *
2099 * Returns 0 on success and a negative value on error
2100 */
2101int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
2102{
2103	return ice_vsi_ctrl_all_rx_rings(vsi, true);
2104}
2105
2106/**
2107 * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
2108 * @vsi: the VSI whose rings are to be disabled
2109 *
2110 * Returns 0 on success and a negative value on error
2111 */
2112int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
2113{
2114	return ice_vsi_ctrl_all_rx_rings(vsi, false);
2115}
2116
2117/**
2118 * ice_vsi_stop_tx_rings - Disable Tx rings
2119 * @vsi: the VSI being configured
2120 * @rst_src: reset source
2121 * @rel_vmvf_num: Relative ID of VF/VM
2122 * @rings: Tx ring array to be stopped
2123 * @count: number of Tx ring array elements
2124 */
2125static int
2126ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2127		      u16 rel_vmvf_num, struct ice_tx_ring **rings, u16 count)
2128{
2129	u16 q_idx;
2130
2131	if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2132		return -EINVAL;
2133
2134	for (q_idx = 0; q_idx < count; q_idx++) {
2135		struct ice_txq_meta txq_meta = { };
2136		int status;
2137
2138		if (!rings || !rings[q_idx])
2139			return -EINVAL;
2140
2141		ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
2142		status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2143					      rings[q_idx], &txq_meta);
2144
2145		if (status)
2146			return status;
2147	}
2148
2149	return 0;
2150}
2151
2152/**
2153 * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2154 * @vsi: the VSI being configured
2155 * @rst_src: reset source
2156 * @rel_vmvf_num: Relative ID of VF/VM
2157 */
2158int
2159ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2160			  u16 rel_vmvf_num)
2161{
2162	return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings, vsi->num_txq);
2163}
2164
2165/**
2166 * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2167 * @vsi: the VSI being configured
2168 */
2169int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2170{
2171	return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings, vsi->num_xdp_txq);
2172}
2173
2174/**
2175 * ice_vsi_is_rx_queue_active
2176 * @vsi: the VSI being configured
2177 *
2178 * Return true if at least one queue is active.
2179 */
2180bool ice_vsi_is_rx_queue_active(struct ice_vsi *vsi)
2181{
2182	struct ice_pf *pf = vsi->back;
2183	struct ice_hw *hw = &pf->hw;
2184	int i;
2185
2186	ice_for_each_rxq(vsi, i) {
2187		u32 rx_reg;
2188		int pf_q;
2189
2190		pf_q = vsi->rxq_map[i];
2191		rx_reg = rd32(hw, QRX_CTRL(pf_q));
2192		if (rx_reg & QRX_CTRL_QENA_STAT_M)
2193			return true;
2194	}
2195
2196	return false;
2197}
2198
2199static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2200{
2201	if (!test_bit(ICE_FLAG_DCB_ENA, vsi->back->flags)) {
2202		vsi->tc_cfg.ena_tc = ICE_DFLT_TRAFFIC_CLASS;
2203		vsi->tc_cfg.numtc = 1;
2204		return;
2205	}
2206
2207	/* set VSI TC information based on DCB config */
2208	ice_vsi_set_dcb_tc_cfg(vsi);
2209}
2210
2211/**
2212 * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2213 * @vsi: the VSI being configured
2214 * @tx: bool to determine Tx or Rx rule
2215 * @create: bool to determine create or remove Rule
2216 */
2217void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2218{
2219	int (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2220			enum ice_sw_fwd_act_type act);
2221	struct ice_pf *pf = vsi->back;
2222	struct device *dev;
2223	int status;
2224
2225	dev = ice_pf_to_dev(pf);
2226	eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2227
2228	if (tx) {
2229		status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2230				  ICE_DROP_PACKET);
2231	} else {
2232		if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2233			status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2234							  create);
2235		} else {
2236			status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2237					  ICE_FWD_TO_VSI);
2238		}
2239	}
2240
2241	if (status)
2242		dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %d\n",
2243			create ? "adding" : "removing", tx ? "TX" : "RX",
2244			vsi->vsi_num, status);
2245}
2246
2247/**
2248 * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2249 * @vsi: pointer to the VSI
2250 *
2251 * This function will allocate new scheduler aggregator now if needed and will
2252 * move specified VSI into it.
2253 */
2254static void ice_set_agg_vsi(struct ice_vsi *vsi)
2255{
2256	struct device *dev = ice_pf_to_dev(vsi->back);
2257	struct ice_agg_node *agg_node_iter = NULL;
2258	u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2259	struct ice_agg_node *agg_node = NULL;
2260	int node_offset, max_agg_nodes = 0;
2261	struct ice_port_info *port_info;
2262	struct ice_pf *pf = vsi->back;
2263	u32 agg_node_id_start = 0;
2264	int status;
2265
2266	/* create (as needed) scheduler aggregator node and move VSI into
2267	 * corresponding aggregator node
2268	 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2269	 * - VF aggregator nodes will contain VF VSI
2270	 */
2271	port_info = pf->hw.port_info;
2272	if (!port_info)
2273		return;
2274
2275	switch (vsi->type) {
2276	case ICE_VSI_CTRL:
2277	case ICE_VSI_CHNL:
2278	case ICE_VSI_LB:
2279	case ICE_VSI_PF:
2280	case ICE_VSI_SWITCHDEV_CTRL:
2281		max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2282		agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2283		agg_node_iter = &pf->pf_agg_node[0];
2284		break;
2285	case ICE_VSI_VF:
2286		/* user can create 'n' VFs on a given PF, but since max children
2287		 * per aggregator node can be only 64. Following code handles
2288		 * aggregator(s) for VF VSIs, either selects a agg_node which
2289		 * was already created provided num_vsis < 64, otherwise
2290		 * select next available node, which will be created
2291		 */
2292		max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2293		agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2294		agg_node_iter = &pf->vf_agg_node[0];
2295		break;
2296	default:
2297		/* other VSI type, handle later if needed */
2298		dev_dbg(dev, "unexpected VSI type %s\n",
2299			ice_vsi_type_str(vsi->type));
2300		return;
2301	}
2302
2303	/* find the appropriate aggregator node */
2304	for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2305		/* see if we can find space in previously created
2306		 * node if num_vsis < 64, otherwise skip
2307		 */
2308		if (agg_node_iter->num_vsis &&
2309		    agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2310			agg_node_iter++;
2311			continue;
2312		}
2313
2314		if (agg_node_iter->valid &&
2315		    agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2316			agg_id = agg_node_iter->agg_id;
2317			agg_node = agg_node_iter;
2318			break;
2319		}
2320
2321		/* find unclaimed agg_id */
2322		if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2323			agg_id = node_offset + agg_node_id_start;
2324			agg_node = agg_node_iter;
2325			break;
2326		}
2327		/* move to next agg_node */
2328		agg_node_iter++;
2329	}
2330
2331	if (!agg_node)
2332		return;
2333
2334	/* if selected aggregator node was not created, create it */
2335	if (!agg_node->valid) {
2336		status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2337				     (u8)vsi->tc_cfg.ena_tc);
2338		if (status) {
2339			dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2340				agg_id);
2341			return;
2342		}
2343		/* aggregator node is created, store the needed info */
2344		agg_node->valid = true;
2345		agg_node->agg_id = agg_id;
2346	}
2347
2348	/* move VSI to corresponding aggregator node */
2349	status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2350				     (u8)vsi->tc_cfg.ena_tc);
2351	if (status) {
2352		dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2353			vsi->idx, agg_id);
2354		return;
2355	}
2356
2357	/* keep active children count for aggregator node */
2358	agg_node->num_vsis++;
2359
2360	/* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2361	 * to aggregator node
2362	 */
2363	vsi->agg_node = agg_node;
2364	dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2365		vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2366		vsi->agg_node->num_vsis);
2367}
2368
2369static int ice_vsi_cfg_tc_lan(struct ice_pf *pf, struct ice_vsi *vsi)
2370{
2371	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2372	struct device *dev = ice_pf_to_dev(pf);
2373	int ret, i;
2374
2375	/* configure VSI nodes based on number of queues and TC's */
2376	ice_for_each_traffic_class(i) {
2377		if (!(vsi->tc_cfg.ena_tc & BIT(i)))
2378			continue;
2379
2380		if (vsi->type == ICE_VSI_CHNL) {
2381			if (!vsi->alloc_txq && vsi->num_txq)
2382				max_txqs[i] = vsi->num_txq;
2383			else
2384				max_txqs[i] = pf->num_lan_tx;
2385		} else {
2386			max_txqs[i] = vsi->alloc_txq;
2387		}
2388
2389		if (vsi->type == ICE_VSI_PF)
2390			max_txqs[i] += vsi->num_xdp_txq;
2391	}
2392
2393	dev_dbg(dev, "vsi->tc_cfg.ena_tc = %d\n", vsi->tc_cfg.ena_tc);
2394	ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2395			      max_txqs);
2396	if (ret) {
2397		dev_err(dev, "VSI %d failed lan queue config, error %d\n",
2398			vsi->vsi_num, ret);
2399		return ret;
2400	}
2401
2402	return 0;
2403}
2404
2405/**
2406 * ice_vsi_cfg_def - configure default VSI based on the type
2407 * @vsi: pointer to VSI
2408 * @params: the parameters to configure this VSI with
2409 */
2410static int
2411ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2412{
2413	struct device *dev = ice_pf_to_dev(vsi->back);
2414	struct ice_pf *pf = vsi->back;
2415	int ret;
2416
2417	vsi->vsw = pf->first_sw;
2418
2419	ret = ice_vsi_alloc_def(vsi, params->ch);
2420	if (ret)
2421		return ret;
2422
2423	/* allocate memory for Tx/Rx ring stat pointers */
2424	ret = ice_vsi_alloc_stat_arrays(vsi);
2425	if (ret)
2426		goto unroll_vsi_alloc;
2427
2428	ice_alloc_fd_res(vsi);
2429
2430	ret = ice_vsi_get_qs(vsi);
2431	if (ret) {
2432		dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2433			vsi->idx);
2434		goto unroll_vsi_alloc_stat;
2435	}
2436
2437	/* set RSS capabilities */
2438	ice_vsi_set_rss_params(vsi);
2439
2440	/* set TC configuration */
2441	ice_vsi_set_tc_cfg(vsi);
2442
2443	/* create the VSI */
2444	ret = ice_vsi_init(vsi, params->flags);
2445	if (ret)
2446		goto unroll_get_qs;
2447
2448	ice_vsi_init_vlan_ops(vsi);
2449
2450	switch (vsi->type) {
2451	case ICE_VSI_CTRL:
2452	case ICE_VSI_SWITCHDEV_CTRL:
2453	case ICE_VSI_PF:
2454		ret = ice_vsi_alloc_q_vectors(vsi);
2455		if (ret)
2456			goto unroll_vsi_init;
2457
2458		ret = ice_vsi_alloc_rings(vsi);
2459		if (ret)
2460			goto unroll_vector_base;
2461
2462		ret = ice_vsi_alloc_ring_stats(vsi);
2463		if (ret)
2464			goto unroll_vector_base;
2465
2466		ice_vsi_map_rings_to_vectors(vsi);
2467		vsi->stat_offsets_loaded = false;
2468
2469		if (ice_is_xdp_ena_vsi(vsi)) {
2470			ret = ice_vsi_determine_xdp_res(vsi);
2471			if (ret)
2472				goto unroll_vector_base;
2473			ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog);
2474			if (ret)
2475				goto unroll_vector_base;
2476		}
2477
2478		/* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2479		if (vsi->type != ICE_VSI_CTRL)
2480			/* Do not exit if configuring RSS had an issue, at
2481			 * least receive traffic on first queue. Hence no
2482			 * need to capture return value
2483			 */
2484			if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2485				ice_vsi_cfg_rss_lut_key(vsi);
2486				ice_vsi_set_rss_flow_fld(vsi);
2487			}
2488		ice_init_arfs(vsi);
2489		break;
2490	case ICE_VSI_CHNL:
2491		if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2492			ice_vsi_cfg_rss_lut_key(vsi);
2493			ice_vsi_set_rss_flow_fld(vsi);
2494		}
2495		break;
2496	case ICE_VSI_VF:
2497		/* VF driver will take care of creating netdev for this type and
2498		 * map queues to vectors through Virtchnl, PF driver only
2499		 * creates a VSI and corresponding structures for bookkeeping
2500		 * purpose
2501		 */
2502		ret = ice_vsi_alloc_q_vectors(vsi);
2503		if (ret)
2504			goto unroll_vsi_init;
2505
2506		ret = ice_vsi_alloc_rings(vsi);
2507		if (ret)
2508			goto unroll_alloc_q_vector;
2509
2510		ret = ice_vsi_alloc_ring_stats(vsi);
2511		if (ret)
2512			goto unroll_vector_base;
2513
2514		vsi->stat_offsets_loaded = false;
2515
2516		/* Do not exit if configuring RSS had an issue, at least
2517		 * receive traffic on first queue. Hence no need to capture
2518		 * return value
2519		 */
2520		if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2521			ice_vsi_cfg_rss_lut_key(vsi);
2522			ice_vsi_set_vf_rss_flow_fld(vsi);
2523		}
2524		break;
2525	case ICE_VSI_LB:
2526		ret = ice_vsi_alloc_rings(vsi);
2527		if (ret)
2528			goto unroll_vsi_init;
2529
2530		ret = ice_vsi_alloc_ring_stats(vsi);
2531		if (ret)
2532			goto unroll_vector_base;
2533
2534		break;
2535	default:
2536		/* clean up the resources and exit */
2537		ret = -EINVAL;
2538		goto unroll_vsi_init;
2539	}
2540
2541	return 0;
2542
2543unroll_vector_base:
2544	/* reclaim SW interrupts back to the common pool */
2545unroll_alloc_q_vector:
2546	ice_vsi_free_q_vectors(vsi);
2547unroll_vsi_init:
2548	ice_vsi_delete_from_hw(vsi);
2549unroll_get_qs:
2550	ice_vsi_put_qs(vsi);
2551unroll_vsi_alloc_stat:
2552	ice_vsi_free_stats(vsi);
2553unroll_vsi_alloc:
2554	ice_vsi_free_arrays(vsi);
2555	return ret;
2556}
2557
2558/**
2559 * ice_vsi_cfg - configure a previously allocated VSI
2560 * @vsi: pointer to VSI
2561 * @params: parameters used to configure this VSI
2562 */
2563int ice_vsi_cfg(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2564{
2565	struct ice_pf *pf = vsi->back;
2566	int ret;
2567
2568	if (WARN_ON(params->type == ICE_VSI_VF && !params->vf))
2569		return -EINVAL;
2570
2571	vsi->type = params->type;
2572	vsi->port_info = params->pi;
2573
2574	/* For VSIs which don't have a connected VF, this will be NULL */
2575	vsi->vf = params->vf;
2576
2577	ret = ice_vsi_cfg_def(vsi, params);
2578	if (ret)
2579		return ret;
2580
2581	ret = ice_vsi_cfg_tc_lan(vsi->back, vsi);
2582	if (ret)
2583		ice_vsi_decfg(vsi);
2584
2585	if (vsi->type == ICE_VSI_CTRL) {
2586		if (vsi->vf) {
2587			WARN_ON(vsi->vf->ctrl_vsi_idx != ICE_NO_VSI);
2588			vsi->vf->ctrl_vsi_idx = vsi->idx;
2589		} else {
2590			WARN_ON(pf->ctrl_vsi_idx != ICE_NO_VSI);
2591			pf->ctrl_vsi_idx = vsi->idx;
2592		}
2593	}
2594
2595	return ret;
2596}
2597
2598/**
2599 * ice_vsi_decfg - remove all VSI configuration
2600 * @vsi: pointer to VSI
2601 */
2602void ice_vsi_decfg(struct ice_vsi *vsi)
2603{
2604	struct ice_pf *pf = vsi->back;
2605	int err;
2606
2607	/* The Rx rule will only exist to remove if the LLDP FW
2608	 * engine is currently stopped
2609	 */
2610	if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF &&
2611	    !test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2612		ice_cfg_sw_lldp(vsi, false, false);
2613
2614	ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
2615	err = ice_rm_vsi_rdma_cfg(vsi->port_info, vsi->idx);
2616	if (err)
2617		dev_err(ice_pf_to_dev(pf), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
2618			vsi->vsi_num, err);
2619
2620	if (ice_is_xdp_ena_vsi(vsi))
2621		/* return value check can be skipped here, it always returns
2622		 * 0 if reset is in progress
2623		 */
2624		ice_destroy_xdp_rings(vsi);
2625
2626	ice_vsi_clear_rings(vsi);
2627	ice_vsi_free_q_vectors(vsi);
2628	ice_vsi_put_qs(vsi);
2629	ice_vsi_free_arrays(vsi);
2630
2631	/* SR-IOV determines needed MSIX resources all at once instead of per
2632	 * VSI since when VFs are spawned we know how many VFs there are and how
2633	 * many interrupts each VF needs. SR-IOV MSIX resources are also
2634	 * cleared in the same manner.
2635	 */
2636
2637	if (vsi->type == ICE_VSI_VF &&
2638	    vsi->agg_node && vsi->agg_node->valid)
2639		vsi->agg_node->num_vsis--;
2640}
2641
2642/**
2643 * ice_vsi_setup - Set up a VSI by a given type
2644 * @pf: board private structure
2645 * @params: parameters to use when creating the VSI
2646 *
2647 * This allocates the sw VSI structure and its queue resources.
2648 *
2649 * Returns pointer to the successfully allocated and configured VSI sw struct on
2650 * success, NULL on failure.
2651 */
2652struct ice_vsi *
2653ice_vsi_setup(struct ice_pf *pf, struct ice_vsi_cfg_params *params)
2654{
2655	struct device *dev = ice_pf_to_dev(pf);
2656	struct ice_vsi *vsi;
2657	int ret;
2658
2659	/* ice_vsi_setup can only initialize a new VSI, and we must have
2660	 * a port_info structure for it.
2661	 */
2662	if (WARN_ON(!(params->flags & ICE_VSI_FLAG_INIT)) ||
2663	    WARN_ON(!params->pi))
2664		return NULL;
2665
2666	vsi = ice_vsi_alloc(pf);
2667	if (!vsi) {
2668		dev_err(dev, "could not allocate VSI\n");
2669		return NULL;
2670	}
2671
2672	ret = ice_vsi_cfg(vsi, params);
2673	if (ret)
2674		goto err_vsi_cfg;
2675
2676	/* Add switch rule to drop all Tx Flow Control Frames, of look up
2677	 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2678	 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2679	 * The rule is added once for PF VSI in order to create appropriate
2680	 * recipe, since VSI/VSI list is ignored with drop action...
2681	 * Also add rules to handle LLDP Tx packets.  Tx LLDP packets need to
2682	 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2683	 * settings in the HW.
2684	 */
2685	if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF) {
2686		ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2687				 ICE_DROP_PACKET);
2688		ice_cfg_sw_lldp(vsi, true, true);
2689	}
2690
2691	if (!vsi->agg_node)
2692		ice_set_agg_vsi(vsi);
2693
2694	return vsi;
2695
2696err_vsi_cfg:
2697	ice_vsi_free(vsi);
2698
2699	return NULL;
2700}
2701
2702/**
2703 * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2704 * @vsi: the VSI being cleaned up
2705 */
2706static void ice_vsi_release_msix(struct ice_vsi *vsi)
2707{
2708	struct ice_pf *pf = vsi->back;
2709	struct ice_hw *hw = &pf->hw;
2710	u32 txq = 0;
2711	u32 rxq = 0;
2712	int i, q;
2713
2714	ice_for_each_q_vector(vsi, i) {
2715		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2716
2717		ice_write_intrl(q_vector, 0);
2718		for (q = 0; q < q_vector->num_ring_tx; q++) {
2719			ice_write_itr(&q_vector->tx, 0);
2720			wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2721			if (ice_is_xdp_ena_vsi(vsi)) {
2722				u32 xdp_txq = txq + vsi->num_xdp_txq;
2723
2724				wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2725			}
2726			txq++;
2727		}
2728
2729		for (q = 0; q < q_vector->num_ring_rx; q++) {
2730			ice_write_itr(&q_vector->rx, 0);
2731			wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2732			rxq++;
2733		}
2734	}
2735
2736	ice_flush(hw);
2737}
2738
2739/**
2740 * ice_vsi_free_irq - Free the IRQ association with the OS
2741 * @vsi: the VSI being configured
2742 */
2743void ice_vsi_free_irq(struct ice_vsi *vsi)
2744{
2745	struct ice_pf *pf = vsi->back;
2746	int i;
2747
2748	if (!vsi->q_vectors || !vsi->irqs_ready)
2749		return;
2750
2751	ice_vsi_release_msix(vsi);
2752	if (vsi->type == ICE_VSI_VF)
2753		return;
2754
2755	vsi->irqs_ready = false;
2756	ice_free_cpu_rx_rmap(vsi);
2757
2758	ice_for_each_q_vector(vsi, i) {
2759		int irq_num;
2760
2761		irq_num = vsi->q_vectors[i]->irq.virq;
2762
2763		/* free only the irqs that were actually requested */
2764		if (!vsi->q_vectors[i] ||
2765		    !(vsi->q_vectors[i]->num_ring_tx ||
2766		      vsi->q_vectors[i]->num_ring_rx))
2767			continue;
2768
2769		/* clear the affinity notifier in the IRQ descriptor */
2770		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2771			irq_set_affinity_notifier(irq_num, NULL);
2772
2773		/* clear the affinity_mask in the IRQ descriptor */
2774		irq_set_affinity_hint(irq_num, NULL);
2775		synchronize_irq(irq_num);
2776		devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2777	}
2778}
2779
2780/**
2781 * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2782 * @vsi: the VSI having resources freed
2783 */
2784void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2785{
2786	int i;
2787
2788	if (!vsi->tx_rings)
2789		return;
2790
2791	ice_for_each_txq(vsi, i)
2792		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2793			ice_free_tx_ring(vsi->tx_rings[i]);
2794}
2795
2796/**
2797 * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2798 * @vsi: the VSI having resources freed
2799 */
2800void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2801{
2802	int i;
2803
2804	if (!vsi->rx_rings)
2805		return;
2806
2807	ice_for_each_rxq(vsi, i)
2808		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2809			ice_free_rx_ring(vsi->rx_rings[i]);
2810}
2811
2812/**
2813 * ice_vsi_close - Shut down a VSI
2814 * @vsi: the VSI being shut down
2815 */
2816void ice_vsi_close(struct ice_vsi *vsi)
2817{
2818	if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2819		ice_down(vsi);
2820
2821	ice_vsi_free_irq(vsi);
2822	ice_vsi_free_tx_rings(vsi);
2823	ice_vsi_free_rx_rings(vsi);
2824}
2825
2826/**
2827 * ice_ena_vsi - resume a VSI
2828 * @vsi: the VSI being resume
2829 * @locked: is the rtnl_lock already held
2830 */
2831int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2832{
2833	int err = 0;
2834
2835	if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2836		return 0;
2837
2838	clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2839
2840	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2841		if (netif_running(vsi->netdev)) {
2842			if (!locked)
2843				rtnl_lock();
2844
2845			err = ice_open_internal(vsi->netdev);
2846
2847			if (!locked)
2848				rtnl_unlock();
2849		}
2850	} else if (vsi->type == ICE_VSI_CTRL) {
2851		err = ice_vsi_open_ctrl(vsi);
2852	}
2853
2854	return err;
2855}
2856
2857/**
2858 * ice_dis_vsi - pause a VSI
2859 * @vsi: the VSI being paused
2860 * @locked: is the rtnl_lock already held
2861 */
2862void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2863{
2864	if (test_bit(ICE_VSI_DOWN, vsi->state))
2865		return;
2866
2867	set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2868
2869	if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2870		if (netif_running(vsi->netdev)) {
2871			if (!locked)
2872				rtnl_lock();
2873
2874			ice_vsi_close(vsi);
2875
2876			if (!locked)
2877				rtnl_unlock();
2878		} else {
2879			ice_vsi_close(vsi);
2880		}
2881	} else if (vsi->type == ICE_VSI_CTRL ||
2882		   vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
2883		ice_vsi_close(vsi);
2884	}
2885}
2886
2887/**
2888 * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
2889 * @vsi: the VSI being un-configured
2890 */
2891void ice_vsi_dis_irq(struct ice_vsi *vsi)
2892{
2893	struct ice_pf *pf = vsi->back;
2894	struct ice_hw *hw = &pf->hw;
2895	u32 val;
2896	int i;
2897
2898	/* disable interrupt causation from each queue */
2899	if (vsi->tx_rings) {
2900		ice_for_each_txq(vsi, i) {
2901			if (vsi->tx_rings[i]) {
2902				u16 reg;
2903
2904				reg = vsi->tx_rings[i]->reg_idx;
2905				val = rd32(hw, QINT_TQCTL(reg));
2906				val &= ~QINT_TQCTL_CAUSE_ENA_M;
2907				wr32(hw, QINT_TQCTL(reg), val);
2908			}
2909		}
2910	}
2911
2912	if (vsi->rx_rings) {
2913		ice_for_each_rxq(vsi, i) {
2914			if (vsi->rx_rings[i]) {
2915				u16 reg;
2916
2917				reg = vsi->rx_rings[i]->reg_idx;
2918				val = rd32(hw, QINT_RQCTL(reg));
2919				val &= ~QINT_RQCTL_CAUSE_ENA_M;
2920				wr32(hw, QINT_RQCTL(reg), val);
2921			}
2922		}
2923	}
2924
2925	/* disable each interrupt */
2926	ice_for_each_q_vector(vsi, i) {
2927		if (!vsi->q_vectors[i])
2928			continue;
2929		wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
2930	}
2931
2932	ice_flush(hw);
2933
2934	/* don't call synchronize_irq() for VF's from the host */
2935	if (vsi->type == ICE_VSI_VF)
2936		return;
2937
2938	ice_for_each_q_vector(vsi, i)
2939		synchronize_irq(vsi->q_vectors[i]->irq.virq);
2940}
2941
2942/**
2943 * ice_vsi_release - Delete a VSI and free its resources
2944 * @vsi: the VSI being removed
2945 *
2946 * Returns 0 on success or < 0 on error
2947 */
2948int ice_vsi_release(struct ice_vsi *vsi)
2949{
2950	struct ice_pf *pf;
2951
2952	if (!vsi->back)
2953		return -ENODEV;
2954	pf = vsi->back;
2955
2956	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2957		ice_rss_clean(vsi);
2958
2959	ice_vsi_close(vsi);
2960	ice_vsi_decfg(vsi);
2961
2962	/* retain SW VSI data structure since it is needed to unregister and
2963	 * free VSI netdev when PF is not in reset recovery pending state,\
2964	 * for ex: during rmmod.
2965	 */
2966	if (!ice_is_reset_in_progress(pf->state))
2967		ice_vsi_delete(vsi);
2968
2969	return 0;
2970}
2971
2972/**
2973 * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
2974 * @vsi: VSI connected with q_vectors
2975 * @coalesce: array of struct with stored coalesce
2976 *
2977 * Returns array size.
2978 */
2979static int
2980ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
2981			     struct ice_coalesce_stored *coalesce)
2982{
2983	int i;
2984
2985	ice_for_each_q_vector(vsi, i) {
2986		struct ice_q_vector *q_vector = vsi->q_vectors[i];
2987
2988		coalesce[i].itr_tx = q_vector->tx.itr_settings;
2989		coalesce[i].itr_rx = q_vector->rx.itr_settings;
2990		coalesce[i].intrl = q_vector->intrl;
2991
2992		if (i < vsi->num_txq)
2993			coalesce[i].tx_valid = true;
2994		if (i < vsi->num_rxq)
2995			coalesce[i].rx_valid = true;
2996	}
2997
2998	return vsi->num_q_vectors;
2999}
3000
3001/**
3002 * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
3003 * @vsi: VSI connected with q_vectors
3004 * @coalesce: pointer to array of struct with stored coalesce
3005 * @size: size of coalesce array
3006 *
3007 * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
3008 * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
3009 * to default value.
3010 */
3011static void
3012ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
3013			     struct ice_coalesce_stored *coalesce, int size)
3014{
3015	struct ice_ring_container *rc;
3016	int i;
3017
3018	if ((size && !coalesce) || !vsi)
3019		return;
3020
3021	/* There are a couple of cases that have to be handled here:
3022	 *   1. The case where the number of queue vectors stays the same, but
3023	 *      the number of Tx or Rx rings changes (the first for loop)
3024	 *   2. The case where the number of queue vectors increased (the
3025	 *      second for loop)
3026	 */
3027	for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
3028		/* There are 2 cases to handle here and they are the same for
3029		 * both Tx and Rx:
3030		 *   if the entry was valid previously (coalesce[i].[tr]x_valid
3031		 *   and the loop variable is less than the number of rings
3032		 *   allocated, then write the previous values
3033		 *
3034		 *   if the entry was not valid previously, but the number of
3035		 *   rings is less than are allocated (this means the number of
3036		 *   rings increased from previously), then write out the
3037		 *   values in the first element
3038		 *
3039		 *   Also, always write the ITR, even if in ITR_IS_DYNAMIC
3040		 *   as there is no harm because the dynamic algorithm
3041		 *   will just overwrite.
3042		 */
3043		if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
3044			rc = &vsi->q_vectors[i]->rx;
3045			rc->itr_settings = coalesce[i].itr_rx;
3046			ice_write_itr(rc, rc->itr_setting);
3047		} else if (i < vsi->alloc_rxq) {
3048			rc = &vsi->q_vectors[i]->rx;
3049			rc->itr_settings = coalesce[0].itr_rx;
3050			ice_write_itr(rc, rc->itr_setting);
3051		}
3052
3053		if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
3054			rc = &vsi->q_vectors[i]->tx;
3055			rc->itr_settings = coalesce[i].itr_tx;
3056			ice_write_itr(rc, rc->itr_setting);
3057		} else if (i < vsi->alloc_txq) {
3058			rc = &vsi->q_vectors[i]->tx;
3059			rc->itr_settings = coalesce[0].itr_tx;
3060			ice_write_itr(rc, rc->itr_setting);
3061		}
3062
3063		vsi->q_vectors[i]->intrl = coalesce[i].intrl;
3064		ice_set_q_vector_intrl(vsi->q_vectors[i]);
3065	}
3066
3067	/* the number of queue vectors increased so write whatever is in
3068	 * the first element
3069	 */
3070	for (; i < vsi->num_q_vectors; i++) {
3071		/* transmit */
3072		rc = &vsi->q_vectors[i]->tx;
3073		rc->itr_settings = coalesce[0].itr_tx;
3074		ice_write_itr(rc, rc->itr_setting);
3075
3076		/* receive */
3077		rc = &vsi->q_vectors[i]->rx;
3078		rc->itr_settings = coalesce[0].itr_rx;
3079		ice_write_itr(rc, rc->itr_setting);
3080
3081		vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3082		ice_set_q_vector_intrl(vsi->q_vectors[i]);
3083	}
3084}
3085
3086/**
3087 * ice_vsi_realloc_stat_arrays - Frees unused stat structures
3088 * @vsi: VSI pointer
3089 * @prev_txq: Number of Tx rings before ring reallocation
3090 * @prev_rxq: Number of Rx rings before ring reallocation
3091 */
3092static void
3093ice_vsi_realloc_stat_arrays(struct ice_vsi *vsi, int prev_txq, int prev_rxq)
3094{
3095	struct ice_vsi_stats *vsi_stat;
3096	struct ice_pf *pf = vsi->back;
3097	int i;
3098
3099	if (!prev_txq || !prev_rxq)
3100		return;
3101	if (vsi->type == ICE_VSI_CHNL)
3102		return;
3103
3104	vsi_stat = pf->vsi_stats[vsi->idx];
3105
3106	if (vsi->num_txq < prev_txq) {
3107		for (i = vsi->num_txq; i < prev_txq; i++) {
3108			if (vsi_stat->tx_ring_stats[i]) {
3109				kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
3110				WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
3111			}
3112		}
3113	}
3114
3115	if (vsi->num_rxq < prev_rxq) {
3116		for (i = vsi->num_rxq; i < prev_rxq; i++) {
3117			if (vsi_stat->rx_ring_stats[i]) {
3118				kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
3119				WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
3120			}
3121		}
3122	}
3123}
3124
3125/**
3126 * ice_vsi_rebuild - Rebuild VSI after reset
3127 * @vsi: VSI to be rebuild
3128 * @vsi_flags: flags used for VSI rebuild flow
3129 *
3130 * Set vsi_flags to ICE_VSI_FLAG_INIT to initialize a new VSI, or
3131 * ICE_VSI_FLAG_NO_INIT to rebuild an existing VSI in hardware.
3132 *
3133 * Returns 0 on success and negative value on failure
3134 */
3135int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags)
3136{
3137	struct ice_vsi_cfg_params params = {};
3138	struct ice_coalesce_stored *coalesce;
3139	int ret, prev_txq, prev_rxq;
3140	int prev_num_q_vectors = 0;
3141	struct ice_pf *pf;
3142
3143	if (!vsi)
3144		return -EINVAL;
3145
3146	params = ice_vsi_to_params(vsi);
3147	params.flags = vsi_flags;
3148
3149	pf = vsi->back;
3150	if (WARN_ON(vsi->type == ICE_VSI_VF && !vsi->vf))
3151		return -EINVAL;
3152
3153	coalesce = kcalloc(vsi->num_q_vectors,
3154			   sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3155	if (!coalesce)
3156		return -ENOMEM;
3157
3158	prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3159
3160	prev_txq = vsi->num_txq;
3161	prev_rxq = vsi->num_rxq;
3162
3163	ice_vsi_decfg(vsi);
3164	ret = ice_vsi_cfg_def(vsi, &params);
3165	if (ret)
3166		goto err_vsi_cfg;
3167
3168	ret = ice_vsi_cfg_tc_lan(pf, vsi);
3169	if (ret) {
3170		if (vsi_flags & ICE_VSI_FLAG_INIT) {
3171			ret = -EIO;
3172			goto err_vsi_cfg_tc_lan;
3173		}
3174
3175		kfree(coalesce);
3176		return ice_schedule_reset(pf, ICE_RESET_PFR);
3177	}
3178
3179	ice_vsi_realloc_stat_arrays(vsi, prev_txq, prev_rxq);
3180
3181	ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3182	kfree(coalesce);
3183
3184	return 0;
3185
3186err_vsi_cfg_tc_lan:
3187	ice_vsi_decfg(vsi);
3188err_vsi_cfg:
3189	kfree(coalesce);
3190	return ret;
3191}
3192
3193/**
3194 * ice_is_reset_in_progress - check for a reset in progress
3195 * @state: PF state field
3196 */
3197bool ice_is_reset_in_progress(unsigned long *state)
3198{
3199	return test_bit(ICE_RESET_OICR_RECV, state) ||
3200	       test_bit(ICE_PFR_REQ, state) ||
3201	       test_bit(ICE_CORER_REQ, state) ||
3202	       test_bit(ICE_GLOBR_REQ, state);
3203}
3204
3205/**
3206 * ice_wait_for_reset - Wait for driver to finish reset and rebuild
3207 * @pf: pointer to the PF structure
3208 * @timeout: length of time to wait, in jiffies
3209 *
3210 * Wait (sleep) for a short time until the driver finishes cleaning up from
3211 * a device reset. The caller must be able to sleep. Use this to delay
3212 * operations that could fail while the driver is cleaning up after a device
3213 * reset.
3214 *
3215 * Returns 0 on success, -EBUSY if the reset is not finished within the
3216 * timeout, and -ERESTARTSYS if the thread was interrupted.
3217 */
3218int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3219{
3220	long ret;
3221
3222	ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3223					       !ice_is_reset_in_progress(pf->state),
3224					       timeout);
3225	if (ret < 0)
3226		return ret;
3227	else if (!ret)
3228		return -EBUSY;
3229	else
3230		return 0;
3231}
3232
3233/**
3234 * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3235 * @vsi: VSI being configured
3236 * @ctx: the context buffer returned from AQ VSI update command
3237 */
3238static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3239{
3240	vsi->info.mapping_flags = ctx->info.mapping_flags;
3241	memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3242	       sizeof(vsi->info.q_mapping));
3243	memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3244	       sizeof(vsi->info.tc_mapping));
3245}
3246
3247/**
3248 * ice_vsi_cfg_netdev_tc - Setup the netdev TC configuration
3249 * @vsi: the VSI being configured
3250 * @ena_tc: TC map to be enabled
3251 */
3252void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc)
3253{
3254	struct net_device *netdev = vsi->netdev;
3255	struct ice_pf *pf = vsi->back;
3256	int numtc = vsi->tc_cfg.numtc;
3257	struct ice_dcbx_cfg *dcbcfg;
3258	u8 netdev_tc;
3259	int i;
3260
3261	if (!netdev)
3262		return;
3263
3264	/* CHNL VSI doesn't have it's own netdev, hence, no netdev_tc */
3265	if (vsi->type == ICE_VSI_CHNL)
3266		return;
3267
3268	if (!ena_tc) {
3269		netdev_reset_tc(netdev);
3270		return;
3271	}
3272
3273	if (vsi->type == ICE_VSI_PF && ice_is_adq_active(pf))
3274		numtc = vsi->all_numtc;
3275
3276	if (netdev_set_num_tc(netdev, numtc))
3277		return;
3278
3279	dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg;
3280
3281	ice_for_each_traffic_class(i)
3282		if (vsi->tc_cfg.ena_tc & BIT(i))
3283			netdev_set_tc_queue(netdev,
3284					    vsi->tc_cfg.tc_info[i].netdev_tc,
3285					    vsi->tc_cfg.tc_info[i].qcount_tx,
3286					    vsi->tc_cfg.tc_info[i].qoffset);
3287	/* setup TC queue map for CHNL TCs */
3288	ice_for_each_chnl_tc(i) {
3289		if (!(vsi->all_enatc & BIT(i)))
3290			break;
3291		if (!vsi->mqprio_qopt.qopt.count[i])
3292			break;
3293		netdev_set_tc_queue(netdev, i,
3294				    vsi->mqprio_qopt.qopt.count[i],
3295				    vsi->mqprio_qopt.qopt.offset[i]);
3296	}
3297
3298	if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3299		return;
3300
3301	for (i = 0; i < ICE_MAX_USER_PRIORITY; i++) {
3302		u8 ets_tc = dcbcfg->etscfg.prio_table[i];
3303
3304		/* Get the mapped netdev TC# for the UP */
3305		netdev_tc = vsi->tc_cfg.tc_info[ets_tc].netdev_tc;
3306		netdev_set_prio_tc_map(netdev, i, netdev_tc);
3307	}
3308}
3309
3310/**
3311 * ice_vsi_setup_q_map_mqprio - Prepares mqprio based tc_config
3312 * @vsi: the VSI being configured,
3313 * @ctxt: VSI context structure
3314 * @ena_tc: number of traffic classes to enable
3315 *
3316 * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
3317 */
3318static int
3319ice_vsi_setup_q_map_mqprio(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt,
3320			   u8 ena_tc)
3321{
3322	u16 pow, offset = 0, qcount_tx = 0, qcount_rx = 0, qmap;
3323	u16 tc0_offset = vsi->mqprio_qopt.qopt.offset[0];
3324	int tc0_qcount = vsi->mqprio_qopt.qopt.count[0];
3325	u16 new_txq, new_rxq;
3326	u8 netdev_tc = 0;
3327	int i;
3328
3329	vsi->tc_cfg.ena_tc = ena_tc ? ena_tc : 1;
3330
3331	pow = order_base_2(tc0_qcount);
3332	qmap = ((tc0_offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
3333		ICE_AQ_VSI_TC_Q_OFFSET_M) |
3334		((pow << ICE_AQ_VSI_TC_Q_NUM_S) & ICE_AQ_VSI_TC_Q_NUM_M);
3335
3336	ice_for_each_traffic_class(i) {
3337		if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
3338			/* TC is not enabled */
3339			vsi->tc_cfg.tc_info[i].qoffset = 0;
3340			vsi->tc_cfg.tc_info[i].qcount_rx = 1;
3341			vsi->tc_cfg.tc_info[i].qcount_tx = 1;
3342			vsi->tc_cfg.tc_info[i].netdev_tc = 0;
3343			ctxt->info.tc_mapping[i] = 0;
3344			continue;
3345		}
3346
3347		offset = vsi->mqprio_qopt.qopt.offset[i];
3348		qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3349		qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3350		vsi->tc_cfg.tc_info[i].qoffset = offset;
3351		vsi->tc_cfg.tc_info[i].qcount_rx = qcount_rx;
3352		vsi->tc_cfg.tc_info[i].qcount_tx = qcount_tx;
3353		vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
3354	}
3355
3356	if (vsi->all_numtc && vsi->all_numtc != vsi->tc_cfg.numtc) {
3357		ice_for_each_chnl_tc(i) {
3358			if (!(vsi->all_enatc & BIT(i)))
3359				continue;
3360			offset = vsi->mqprio_qopt.qopt.offset[i];
3361			qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3362			qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3363		}
3364	}
3365
3366	new_txq = offset + qcount_tx;
3367	if (new_txq > vsi->alloc_txq) {
3368		dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
3369			new_txq, vsi->alloc_txq);
3370		return -EINVAL;
3371	}
3372
3373	new_rxq = offset + qcount_rx;
3374	if (new_rxq > vsi->alloc_rxq) {
3375		dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
3376			new_rxq, vsi->alloc_rxq);
3377		return -EINVAL;
3378	}
3379
3380	/* Set actual Tx/Rx queue pairs */
3381	vsi->num_txq = new_txq;
3382	vsi->num_rxq = new_rxq;
3383
3384	/* Setup queue TC[0].qmap for given VSI context */
3385	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
3386	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
3387	ctxt->info.q_mapping[1] = cpu_to_le16(tc0_qcount);
3388
3389	/* Find queue count available for channel VSIs and starting offset
3390	 * for channel VSIs
3391	 */
3392	if (tc0_qcount && tc0_qcount < vsi->num_rxq) {
3393		vsi->cnt_q_avail = vsi->num_rxq - tc0_qcount;
3394		vsi->next_base_q = tc0_qcount;
3395	}
3396	dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_txq = %d\n",  vsi->num_txq);
3397	dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_rxq = %d\n",  vsi->num_rxq);
3398	dev_dbg(ice_pf_to_dev(vsi->back), "all_numtc %u, all_enatc: 0x%04x, tc_cfg.numtc %u\n",
3399		vsi->all_numtc, vsi->all_enatc, vsi->tc_cfg.numtc);
3400
3401	return 0;
3402}
3403
3404/**
3405 * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3406 * @vsi: VSI to be configured
3407 * @ena_tc: TC bitmap
3408 *
3409 * VSI queues expected to be quiesced before calling this function
3410 */
3411int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3412{
3413	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3414	struct ice_pf *pf = vsi->back;
3415	struct ice_tc_cfg old_tc_cfg;
3416	struct ice_vsi_ctx *ctx;
3417	struct device *dev;
3418	int i, ret = 0;
3419	u8 num_tc = 0;
3420
3421	dev = ice_pf_to_dev(pf);
3422	if (vsi->tc_cfg.ena_tc == ena_tc &&
3423	    vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
3424		return 0;
3425
3426	ice_for_each_traffic_class(i) {
3427		/* build bitmap of enabled TCs */
3428		if (ena_tc & BIT(i))
3429			num_tc++;
3430		/* populate max_txqs per TC */
3431		max_txqs[i] = vsi->alloc_txq;
3432		/* Update max_txqs if it is CHNL VSI, because alloc_t[r]xq are
3433		 * zero for CHNL VSI, hence use num_txq instead as max_txqs
3434		 */
3435		if (vsi->type == ICE_VSI_CHNL &&
3436		    test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3437			max_txqs[i] = vsi->num_txq;
3438	}
3439
3440	memcpy(&old_tc_cfg, &vsi->tc_cfg, sizeof(old_tc_cfg));
3441	vsi->tc_cfg.ena_tc = ena_tc;
3442	vsi->tc_cfg.numtc = num_tc;
3443
3444	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3445	if (!ctx)
3446		return -ENOMEM;
3447
3448	ctx->vf_num = 0;
3449	ctx->info = vsi->info;
3450
3451	if (vsi->type == ICE_VSI_PF &&
3452	    test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3453		ret = ice_vsi_setup_q_map_mqprio(vsi, ctx, ena_tc);
3454	else
3455		ret = ice_vsi_setup_q_map(vsi, ctx);
3456
3457	if (ret) {
3458		memcpy(&vsi->tc_cfg, &old_tc_cfg, sizeof(vsi->tc_cfg));
3459		goto out;
3460	}
3461
3462	/* must to indicate which section of VSI context are being modified */
3463	ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3464	ret = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3465	if (ret) {
3466		dev_info(dev, "Failed VSI Update\n");
3467		goto out;
3468	}
3469
3470	if (vsi->type == ICE_VSI_PF &&
3471	    test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3472		ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, 1, max_txqs);
3473	else
3474		ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx,
3475				      vsi->tc_cfg.ena_tc, max_txqs);
3476
3477	if (ret) {
3478		dev_err(dev, "VSI %d failed TC config, error %d\n",
3479			vsi->vsi_num, ret);
3480		goto out;
3481	}
3482	ice_vsi_update_q_map(vsi, ctx);
3483	vsi->info.valid_sections = 0;
3484
3485	ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3486out:
3487	kfree(ctx);
3488	return ret;
3489}
3490
3491/**
3492 * ice_update_ring_stats - Update ring statistics
3493 * @stats: stats to be updated
3494 * @pkts: number of processed packets
3495 * @bytes: number of processed bytes
3496 *
3497 * This function assumes that caller has acquired a u64_stats_sync lock.
3498 */
3499static void ice_update_ring_stats(struct ice_q_stats *stats, u64 pkts, u64 bytes)
3500{
3501	stats->bytes += bytes;
3502	stats->pkts += pkts;
3503}
3504
3505/**
3506 * ice_update_tx_ring_stats - Update Tx ring specific counters
3507 * @tx_ring: ring to update
3508 * @pkts: number of processed packets
3509 * @bytes: number of processed bytes
3510 */
3511void ice_update_tx_ring_stats(struct ice_tx_ring *tx_ring, u64 pkts, u64 bytes)
3512{
3513	u64_stats_update_begin(&tx_ring->ring_stats->syncp);
3514	ice_update_ring_stats(&tx_ring->ring_stats->stats, pkts, bytes);
3515	u64_stats_update_end(&tx_ring->ring_stats->syncp);
3516}
3517
3518/**
3519 * ice_update_rx_ring_stats - Update Rx ring specific counters
3520 * @rx_ring: ring to update
3521 * @pkts: number of processed packets
3522 * @bytes: number of processed bytes
3523 */
3524void ice_update_rx_ring_stats(struct ice_rx_ring *rx_ring, u64 pkts, u64 bytes)
3525{
3526	u64_stats_update_begin(&rx_ring->ring_stats->syncp);
3527	ice_update_ring_stats(&rx_ring->ring_stats->stats, pkts, bytes);
3528	u64_stats_update_end(&rx_ring->ring_stats->syncp);
3529}
3530
3531/**
3532 * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3533 * @pi: port info of the switch with default VSI
3534 *
3535 * Return true if the there is a single VSI in default forwarding VSI list
3536 */
3537bool ice_is_dflt_vsi_in_use(struct ice_port_info *pi)
3538{
3539	bool exists = false;
3540
3541	ice_check_if_dflt_vsi(pi, 0, &exists);
3542	return exists;
3543}
3544
3545/**
3546 * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3547 * @vsi: VSI to compare against default forwarding VSI
3548 *
3549 * If this VSI passed in is the default forwarding VSI then return true, else
3550 * return false
3551 */
3552bool ice_is_vsi_dflt_vsi(struct ice_vsi *vsi)
3553{
3554	return ice_check_if_dflt_vsi(vsi->port_info, vsi->idx, NULL);
3555}
3556
3557/**
3558 * ice_set_dflt_vsi - set the default forwarding VSI
3559 * @vsi: VSI getting set as the default forwarding VSI on the switch
3560 *
3561 * If the VSI passed in is already the default VSI and it's enabled just return
3562 * success.
3563 *
3564 * Otherwise try to set the VSI passed in as the switch's default VSI and
3565 * return the result.
3566 */
3567int ice_set_dflt_vsi(struct ice_vsi *vsi)
3568{
3569	struct device *dev;
3570	int status;
3571
3572	if (!vsi)
3573		return -EINVAL;
3574
3575	dev = ice_pf_to_dev(vsi->back);
3576
3577	if (ice_lag_is_switchdev_running(vsi->back)) {
3578		dev_dbg(dev, "VSI %d passed is a part of LAG containing interfaces in switchdev mode, nothing to do\n",
3579			vsi->vsi_num);
3580		return 0;
3581	}
3582
3583	/* the VSI passed in is already the default VSI */
3584	if (ice_is_vsi_dflt_vsi(vsi)) {
3585		dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3586			vsi->vsi_num);
3587		return 0;
3588	}
3589
3590	status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, true, ICE_FLTR_RX);
3591	if (status) {
3592		dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %d\n",
3593			vsi->vsi_num, status);
3594		return status;
3595	}
3596
3597	return 0;
3598}
3599
3600/**
3601 * ice_clear_dflt_vsi - clear the default forwarding VSI
3602 * @vsi: VSI to remove from filter list
3603 *
3604 * If the switch has no default VSI or it's not enabled then return error.
3605 *
3606 * Otherwise try to clear the default VSI and return the result.
3607 */
3608int ice_clear_dflt_vsi(struct ice_vsi *vsi)
3609{
3610	struct device *dev;
3611	int status;
3612
3613	if (!vsi)
3614		return -EINVAL;
3615
3616	dev = ice_pf_to_dev(vsi->back);
3617
3618	/* there is no default VSI configured */
3619	if (!ice_is_dflt_vsi_in_use(vsi->port_info))
3620		return -ENODEV;
3621
3622	status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, false,
3623				  ICE_FLTR_RX);
3624	if (status) {
3625		dev_err(dev, "Failed to clear the default forwarding VSI %d, error %d\n",
3626			vsi->vsi_num, status);
3627		return -EIO;
3628	}
3629
3630	return 0;
3631}
3632
3633/**
3634 * ice_get_link_speed_mbps - get link speed in Mbps
3635 * @vsi: the VSI whose link speed is being queried
3636 *
3637 * Return current VSI link speed and 0 if the speed is unknown.
3638 */
3639int ice_get_link_speed_mbps(struct ice_vsi *vsi)
3640{
3641	unsigned int link_speed;
3642
3643	link_speed = vsi->port_info->phy.link_info.link_speed;
3644
3645	return (int)ice_get_link_speed(fls(link_speed) - 1);
3646}
3647
3648/**
3649 * ice_get_link_speed_kbps - get link speed in Kbps
3650 * @vsi: the VSI whose link speed is being queried
3651 *
3652 * Return current VSI link speed and 0 if the speed is unknown.
3653 */
3654int ice_get_link_speed_kbps(struct ice_vsi *vsi)
3655{
3656	int speed_mbps;
3657
3658	speed_mbps = ice_get_link_speed_mbps(vsi);
3659
3660	return speed_mbps * 1000;
3661}
3662
3663/**
3664 * ice_set_min_bw_limit - setup minimum BW limit for Tx based on min_tx_rate
3665 * @vsi: VSI to be configured
3666 * @min_tx_rate: min Tx rate in Kbps to be configured as BW limit
3667 *
3668 * If the min_tx_rate is specified as 0 that means to clear the minimum BW limit
3669 * profile, otherwise a non-zero value will force a minimum BW limit for the VSI
3670 * on TC 0.
3671 */
3672int ice_set_min_bw_limit(struct ice_vsi *vsi, u64 min_tx_rate)
3673{
3674	struct ice_pf *pf = vsi->back;
3675	struct device *dev;
3676	int status;
3677	int speed;
3678
3679	dev = ice_pf_to_dev(pf);
3680	if (!vsi->port_info) {
3681		dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3682			vsi->idx, vsi->type);
3683		return -EINVAL;
3684	}
3685
3686	speed = ice_get_link_speed_kbps(vsi);
3687	if (min_tx_rate > (u64)speed) {
3688		dev_err(dev, "invalid min Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3689			min_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3690			speed);
3691		return -EINVAL;
3692	}
3693
3694	/* Configure min BW for VSI limit */
3695	if (min_tx_rate) {
3696		status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3697						   ICE_MIN_BW, min_tx_rate);
3698		if (status) {
3699			dev_err(dev, "failed to set min Tx rate(%llu Kbps) for %s %d\n",
3700				min_tx_rate, ice_vsi_type_str(vsi->type),
3701				vsi->idx);
3702			return status;
3703		}
3704
3705		dev_dbg(dev, "set min Tx rate(%llu Kbps) for %s\n",
3706			min_tx_rate, ice_vsi_type_str(vsi->type));
3707	} else {
3708		status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3709							vsi->idx, 0,
3710							ICE_MIN_BW);
3711		if (status) {
3712			dev_err(dev, "failed to clear min Tx rate configuration for %s %d\n",
3713				ice_vsi_type_str(vsi->type), vsi->idx);
3714			return status;
3715		}
3716
3717		dev_dbg(dev, "cleared min Tx rate configuration for %s %d\n",
3718			ice_vsi_type_str(vsi->type), vsi->idx);
3719	}
3720
3721	return 0;
3722}
3723
3724/**
3725 * ice_set_max_bw_limit - setup maximum BW limit for Tx based on max_tx_rate
3726 * @vsi: VSI to be configured
3727 * @max_tx_rate: max Tx rate in Kbps to be configured as BW limit
3728 *
3729 * If the max_tx_rate is specified as 0 that means to clear the maximum BW limit
3730 * profile, otherwise a non-zero value will force a maximum BW limit for the VSI
3731 * on TC 0.
3732 */
3733int ice_set_max_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate)
3734{
3735	struct ice_pf *pf = vsi->back;
3736	struct device *dev;
3737	int status;
3738	int speed;
3739
3740	dev = ice_pf_to_dev(pf);
3741	if (!vsi->port_info) {
3742		dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3743			vsi->idx, vsi->type);
3744		return -EINVAL;
3745	}
3746
3747	speed = ice_get_link_speed_kbps(vsi);
3748	if (max_tx_rate > (u64)speed) {
3749		dev_err(dev, "invalid max Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3750			max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3751			speed);
3752		return -EINVAL;
3753	}
3754
3755	/* Configure max BW for VSI limit */
3756	if (max_tx_rate) {
3757		status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3758						   ICE_MAX_BW, max_tx_rate);
3759		if (status) {
3760			dev_err(dev, "failed setting max Tx rate(%llu Kbps) for %s %d\n",
3761				max_tx_rate, ice_vsi_type_str(vsi->type),
3762				vsi->idx);
3763			return status;
3764		}
3765
3766		dev_dbg(dev, "set max Tx rate(%llu Kbps) for %s %d\n",
3767			max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx);
3768	} else {
3769		status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3770							vsi->idx, 0,
3771							ICE_MAX_BW);
3772		if (status) {
3773			dev_err(dev, "failed clearing max Tx rate configuration for %s %d\n",
3774				ice_vsi_type_str(vsi->type), vsi->idx);
3775			return status;
3776		}
3777
3778		dev_dbg(dev, "cleared max Tx rate configuration for %s %d\n",
3779			ice_vsi_type_str(vsi->type), vsi->idx);
3780	}
3781
3782	return 0;
3783}
3784
3785/**
3786 * ice_set_link - turn on/off physical link
3787 * @vsi: VSI to modify physical link on
3788 * @ena: turn on/off physical link
3789 */
3790int ice_set_link(struct ice_vsi *vsi, bool ena)
3791{
3792	struct device *dev = ice_pf_to_dev(vsi->back);
3793	struct ice_port_info *pi = vsi->port_info;
3794	struct ice_hw *hw = pi->hw;
3795	int status;
3796
3797	if (vsi->type != ICE_VSI_PF)
3798		return -EINVAL;
3799
3800	status = ice_aq_set_link_restart_an(pi, ena, NULL);
3801
3802	/* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3803	 * this is not a fatal error, so print a warning message and return
3804	 * a success code. Return an error if FW returns an error code other
3805	 * than ICE_AQ_RC_EMODE
3806	 */
3807	if (status == -EIO) {
3808		if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3809			dev_dbg(dev, "can't set link to %s, err %d aq_err %s. not fatal, continuing\n",
3810				(ena ? "ON" : "OFF"), status,
3811				ice_aq_str(hw->adminq.sq_last_status));
3812	} else if (status) {
3813		dev_err(dev, "can't set link to %s, err %d aq_err %s\n",
3814			(ena ? "ON" : "OFF"), status,
3815			ice_aq_str(hw->adminq.sq_last_status));
3816		return status;
3817	}
3818
3819	return 0;
3820}
3821
3822/**
3823 * ice_vsi_add_vlan_zero - add VLAN 0 filter(s) for this VSI
3824 * @vsi: VSI used to add VLAN filters
3825 *
3826 * In Single VLAN Mode (SVM), single VLAN filters via ICE_SW_LKUP_VLAN are based
3827 * on the inner VLAN ID, so the VLAN TPID (i.e. 0x8100 or 0x888a8) doesn't
3828 * matter. In Double VLAN Mode (DVM), outer/single VLAN filters via
3829 * ICE_SW_LKUP_VLAN are based on the outer/single VLAN ID + VLAN TPID.
3830 *
3831 * For both modes add a VLAN 0 + no VLAN TPID filter to handle untagged traffic
3832 * when VLAN pruning is enabled. Also, this handles VLAN 0 priority tagged
3833 * traffic in SVM, since the VLAN TPID isn't part of filtering.
3834 *
3835 * If DVM is enabled then an explicit VLAN 0 + VLAN TPID filter needs to be
3836 * added to allow VLAN 0 priority tagged traffic in DVM, since the VLAN TPID is
3837 * part of filtering.
3838 */
3839int ice_vsi_add_vlan_zero(struct ice_vsi *vsi)
3840{
3841	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3842	struct ice_vlan vlan;
3843	int err;
3844
3845	vlan = ICE_VLAN(0, 0, 0);
3846	err = vlan_ops->add_vlan(vsi, &vlan);
3847	if (err && err != -EEXIST)
3848		return err;
3849
3850	/* in SVM both VLAN 0 filters are identical */
3851	if (!ice_is_dvm_ena(&vsi->back->hw))
3852		return 0;
3853
3854	vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3855	err = vlan_ops->add_vlan(vsi, &vlan);
3856	if (err && err != -EEXIST)
3857		return err;
3858
3859	return 0;
3860}
3861
3862/**
3863 * ice_vsi_del_vlan_zero - delete VLAN 0 filter(s) for this VSI
3864 * @vsi: VSI used to add VLAN filters
3865 *
3866 * Delete the VLAN 0 filters in the same manner that they were added in
3867 * ice_vsi_add_vlan_zero.
3868 */
3869int ice_vsi_del_vlan_zero(struct ice_vsi *vsi)
3870{
3871	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3872	struct ice_vlan vlan;
3873	int err;
3874
3875	vlan = ICE_VLAN(0, 0, 0);
3876	err = vlan_ops->del_vlan(vsi, &vlan);
3877	if (err && err != -EEXIST)
3878		return err;
3879
3880	/* in SVM both VLAN 0 filters are identical */
3881	if (!ice_is_dvm_ena(&vsi->back->hw))
3882		return 0;
3883
3884	vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3885	err = vlan_ops->del_vlan(vsi, &vlan);
3886	if (err && err != -EEXIST)
3887		return err;
3888
3889	/* when deleting the last VLAN filter, make sure to disable the VLAN
3890	 * promisc mode so the filter isn't left by accident
3891	 */
3892	return ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3893				    ICE_MCAST_VLAN_PROMISC_BITS, 0);
3894}
3895
3896/**
3897 * ice_vsi_num_zero_vlans - get number of VLAN 0 filters based on VLAN mode
3898 * @vsi: VSI used to get the VLAN mode
3899 *
3900 * If DVM is enabled then 2 VLAN 0 filters are added, else if SVM is enabled
3901 * then 1 VLAN 0 filter is added. See ice_vsi_add_vlan_zero for more details.
3902 */
3903static u16 ice_vsi_num_zero_vlans(struct ice_vsi *vsi)
3904{
3905#define ICE_DVM_NUM_ZERO_VLAN_FLTRS	2
3906#define ICE_SVM_NUM_ZERO_VLAN_FLTRS	1
3907	/* no VLAN 0 filter is created when a port VLAN is active */
3908	if (vsi->type == ICE_VSI_VF) {
3909		if (WARN_ON(!vsi->vf))
3910			return 0;
3911
3912		if (ice_vf_is_port_vlan_ena(vsi->vf))
3913			return 0;
3914	}
3915
3916	if (ice_is_dvm_ena(&vsi->back->hw))
3917		return ICE_DVM_NUM_ZERO_VLAN_FLTRS;
3918	else
3919		return ICE_SVM_NUM_ZERO_VLAN_FLTRS;
3920}
3921
3922/**
3923 * ice_vsi_has_non_zero_vlans - check if VSI has any non-zero VLANs
3924 * @vsi: VSI used to determine if any non-zero VLANs have been added
3925 */
3926bool ice_vsi_has_non_zero_vlans(struct ice_vsi *vsi)
3927{
3928	return (vsi->num_vlan > ice_vsi_num_zero_vlans(vsi));
3929}
3930
3931/**
3932 * ice_vsi_num_non_zero_vlans - get the number of non-zero VLANs for this VSI
3933 * @vsi: VSI used to get the number of non-zero VLANs added
3934 */
3935u16 ice_vsi_num_non_zero_vlans(struct ice_vsi *vsi)
3936{
3937	return (vsi->num_vlan - ice_vsi_num_zero_vlans(vsi));
3938}
3939
3940/**
3941 * ice_is_feature_supported
3942 * @pf: pointer to the struct ice_pf instance
3943 * @f: feature enum to be checked
3944 *
3945 * returns true if feature is supported, false otherwise
3946 */
3947bool ice_is_feature_supported(struct ice_pf *pf, enum ice_feature f)
3948{
3949	if (f < 0 || f >= ICE_F_MAX)
3950		return false;
3951
3952	return test_bit(f, pf->features);
3953}
3954
3955/**
3956 * ice_set_feature_support
3957 * @pf: pointer to the struct ice_pf instance
3958 * @f: feature enum to set
3959 */
3960void ice_set_feature_support(struct ice_pf *pf, enum ice_feature f)
3961{
3962	if (f < 0 || f >= ICE_F_MAX)
3963		return;
3964
3965	set_bit(f, pf->features);
3966}
3967
3968/**
3969 * ice_clear_feature_support
3970 * @pf: pointer to the struct ice_pf instance
3971 * @f: feature enum to clear
3972 */
3973void ice_clear_feature_support(struct ice_pf *pf, enum ice_feature f)
3974{
3975	if (f < 0 || f >= ICE_F_MAX)
3976		return;
3977
3978	clear_bit(f, pf->features);
3979}
3980
3981/**
3982 * ice_init_feature_support
3983 * @pf: pointer to the struct ice_pf instance
3984 *
3985 * called during init to setup supported feature
3986 */
3987void ice_init_feature_support(struct ice_pf *pf)
3988{
3989	switch (pf->hw.device_id) {
3990	case ICE_DEV_ID_E810C_BACKPLANE:
3991	case ICE_DEV_ID_E810C_QSFP:
3992	case ICE_DEV_ID_E810C_SFP:
3993		ice_set_feature_support(pf, ICE_F_DSCP);
3994		ice_set_feature_support(pf, ICE_F_PTP_EXTTS);
3995		if (ice_is_e810t(&pf->hw)) {
3996			ice_set_feature_support(pf, ICE_F_SMA_CTRL);
3997			if (ice_gnss_is_gps_present(&pf->hw))
3998				ice_set_feature_support(pf, ICE_F_GNSS);
3999		}
4000		break;
4001	default:
4002		break;
4003	}
4004}
4005
4006/**
4007 * ice_vsi_update_security - update security block in VSI
4008 * @vsi: pointer to VSI structure
4009 * @fill: function pointer to fill ctx
4010 */
4011int
4012ice_vsi_update_security(struct ice_vsi *vsi, void (*fill)(struct ice_vsi_ctx *))
4013{
4014	struct ice_vsi_ctx ctx = { 0 };
4015
4016	ctx.info = vsi->info;
4017	ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
4018	fill(&ctx);
4019
4020	if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4021		return -ENODEV;
4022
4023	vsi->info = ctx.info;
4024	return 0;
4025}
4026
4027/**
4028 * ice_vsi_ctx_set_antispoof - set antispoof function in VSI ctx
4029 * @ctx: pointer to VSI ctx structure
4030 */
4031void ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx *ctx)
4032{
4033	ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
4034			       (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4035				ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4036}
4037
4038/**
4039 * ice_vsi_ctx_clear_antispoof - clear antispoof function in VSI ctx
4040 * @ctx: pointer to VSI ctx structure
4041 */
4042void ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx *ctx)
4043{
4044	ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF &
4045			       ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4046				 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4047}
4048
4049/**
4050 * ice_vsi_ctx_set_allow_override - allow destination override on VSI
4051 * @ctx: pointer to VSI ctx structure
4052 */
4053void ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx *ctx)
4054{
4055	ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4056}
4057
4058/**
4059 * ice_vsi_ctx_clear_allow_override - turn off destination override on VSI
4060 * @ctx: pointer to VSI ctx structure
4061 */
4062void ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx *ctx)
4063{
4064	ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4065}
4066
4067/**
4068 * ice_vsi_update_local_lb - update sw block in VSI with local loopback bit
4069 * @vsi: pointer to VSI structure
4070 * @set: set or unset the bit
4071 */
4072int
4073ice_vsi_update_local_lb(struct ice_vsi *vsi, bool set)
4074{
4075	struct ice_vsi_ctx ctx = {
4076		.info	= vsi->info,
4077	};
4078
4079	ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
4080	if (set)
4081		ctx.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4082	else
4083		ctx.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4084
4085	if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4086		return -ENODEV;
4087
4088	vsi->info = ctx.info;
4089	return 0;
4090}
4091