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	*/
17	const 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	*/
47	static 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, rxq_idx: i, wait: 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, rxq_idx: 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	*/
73	static 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, n: vsi->alloc_txq,
84	size: sizeof(*vsi->tx_rings), GFP_KERNEL);
85	if (!vsi->tx_rings)
86	return -ENOMEM;
87
88	vsi->rx_rings = devm_kcalloc(dev, n: vsi->alloc_rxq,
89	size: 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, n: (vsi->alloc_txq + num_possible_cpus()),
100	size: 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, n: vsi->alloc_rxq,
106	size: 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, n: vsi->num_q_vectors,
116	size: 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
126	err_zc_qps:
127	devm_kfree(dev, p: vsi->q_vectors);
128	err_vectors:
129	devm_kfree(dev, p: vsi->rxq_map);
130	err_rxq_map:
131	devm_kfree(dev, p: vsi->txq_map);
132	err_txq_map:
133	devm_kfree(dev, p: vsi->rx_rings);
134	err_rings:
135	devm_kfree(dev, p: 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	*/
143	static 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	*/
172	static 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 = vf->num_msix - 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	*/
264	static 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	*/
288	static 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(size: 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(hw: &pf->hw, vsi_handle: vsi->idx, vsi_ctx: ctxt, keep_vsi_alloc: 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(objp: 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	*/
317	static 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(bitmap: vsi->af_xdp_zc_qps);
325	vsi->af_xdp_zc_qps = NULL;
326	/* free the ring and vector containers */
327	devm_kfree(dev, p: vsi->q_vectors);
328	vsi->q_vectors = NULL;
329	devm_kfree(dev, p: vsi->tx_rings);
330	vsi->tx_rings = NULL;
331	devm_kfree(dev, p: vsi->rx_rings);
332	vsi->rx_rings = NULL;
333	devm_kfree(dev, p: vsi->txq_map);
334	vsi->txq_map = NULL;
335	devm_kfree(dev, p: 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	*/
343	static 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(objp: vsi_stat->tx_ring_stats);
373	kfree(objp: vsi_stat->rx_ring_stats);
374	kfree(objp: 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	*/
382	static 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(size: 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(size: 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
434	err_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	*/
446	static 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(lock: &pf->sw_mutex);
471	devm_kfree(dev, p: vsi);
472	}
473
474	void 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	*/
485	static 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(rx_ring: q_vector->rx.rx_ring, FDIR_RX_DESC_CLEAN_BUDGET);
494	ice_clean_ctrl_tx_irq(tx_ring: 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	*/
504	static 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(n: &q_vector->napi);
514
515	return IRQ_HANDLED;
516	}
517
518	static 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(n: &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	*/
540	static 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(size: sizeof(*vsi_stat), GFP_KERNEL);
555	if (!vsi_stat)
556	return -ENOMEM;
557
558	vsi_stat->tx_ring_stats =
559	kcalloc(n: vsi->alloc_txq, size: 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(n: vsi->alloc_rxq, size: 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
574	err_alloc_rx:
575	kfree(objp: vsi_stat->rx_ring_stats);
576	err_alloc_tx:
577	kfree(objp: vsi_stat->tx_ring_stats);
578	kfree(objp: 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	*/
588	static int
589	ice_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	*/
639	static 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, size: sizeof(*vsi), GFP_KERNEL);
657	if (!vsi)
658	goto unlock_pf;
659
660	vsi->back = pf;
661	set_bit(nr: ICE_VSI_DOWN, addr: 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(array: pf->vsi, size: pf->num_alloc_vsi,
669	curr: pf->next_vsi);
670
671	unlock_pf:
672	mutex_unlock(lock: &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	*/
684	static 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	*/
774	static 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(qs_cfg: &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(qs_cfg: &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	*/
819	static 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(nr: vsi->txq_map[i], addr: 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(nr: vsi->rxq_map[i], addr: pf->avail_rxqs);
833	vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
834	}
835
836	mutex_unlock(lock: &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	*/
845	bool 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	*/
856	bool 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	*/
868	static 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(hw: &pf->hw, vsi_handle: 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	*/
886	static 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, p: vsi->rss_hkey_user);
894	devm_kfree(dev, p: 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(hw: &pf->hw, vsi_handle: vsi->idx);
900	}
901
902	/**
903	* ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
904	* @vsi: the VSI being configured
905	*/
906	static 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	*/
960	static 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	ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
983	ctxt->info.outer_vlan_flags =
984	(ICE_AQ_VSI_OUTER_VLAN_TX_MODE_ALL <<
985	ICE_AQ_VSI_OUTER_VLAN_TX_MODE_S) &
986	ICE_AQ_VSI_OUTER_VLAN_TX_MODE_M;
987	ctxt->info.outer_vlan_flags |=
988	(ICE_AQ_VSI_OUTER_TAG_VLAN_8100 <<
989	ICE_AQ_VSI_OUTER_TAG_TYPE_S) &
990	ICE_AQ_VSI_OUTER_TAG_TYPE_M;
991	ctxt->info.outer_vlan_flags |=
992	FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_EMODE_M,
993	ICE_AQ_VSI_OUTER_VLAN_EMODE_NOTHING);
994	}
995	/* Have 1:1 UP mapping for both ingress/egress tables */
996	table |= ICE_UP_TABLE_TRANSLATE(0, 0);
997	table |= ICE_UP_TABLE_TRANSLATE(1, 1);
998	table |= ICE_UP_TABLE_TRANSLATE(2, 2);
999	table |= ICE_UP_TABLE_TRANSLATE(3, 3);
1000	table |= ICE_UP_TABLE_TRANSLATE(4, 4);
1001	table |= ICE_UP_TABLE_TRANSLATE(5, 5);
1002	table |= ICE_UP_TABLE_TRANSLATE(6, 6);
1003	table |= ICE_UP_TABLE_TRANSLATE(7, 7);
1004	ctxt->info.ingress_table = cpu_to_le32(table);
1005	ctxt->info.egress_table = cpu_to_le32(table);
1006	/* Have 1:1 UP mapping for outer to inner UP table */
1007	ctxt->info.outer_up_table = cpu_to_le32(table);
1008	/* No Outer tag support outer_tag_flags remains to zero */
1009	}
1010
1011	/**
1012	* ice_vsi_setup_q_map - Setup a VSI queue map
1013	* @vsi: the VSI being configured
1014	* @ctxt: VSI context structure
1015	*/
1016	static int ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1017	{
1018	u16 offset = 0, qmap = 0, tx_count = 0, rx_count = 0, pow = 0;
1019	u16 num_txq_per_tc, num_rxq_per_tc;
1020	u16 qcount_tx = vsi->alloc_txq;
1021	u16 qcount_rx = vsi->alloc_rxq;
1022	u8 netdev_tc = 0;
1023	int i;
1024
1025	if (!vsi->tc_cfg.numtc) {
1026	/* at least TC0 should be enabled by default */
1027	vsi->tc_cfg.numtc = 1;
1028	vsi->tc_cfg.ena_tc = 1;
1029	}
1030
1031	num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
1032	if (!num_rxq_per_tc)
1033	num_rxq_per_tc = 1;
1034	num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
1035	if (!num_txq_per_tc)
1036	num_txq_per_tc = 1;
1037
1038	/* find the (rounded up) power-of-2 of qcount */
1039	pow = (u16)order_base_2(num_rxq_per_tc);
1040
1041	/* TC mapping is a function of the number of Rx queues assigned to the
1042	* VSI for each traffic class and the offset of these queues.
1043	* The first 10 bits are for queue offset for TC0, next 4 bits for no:of
1044	* queues allocated to TC0. No:of queues is a power-of-2.
1045	*
1046	* If TC is not enabled, the queue offset is set to 0, and allocate one
1047	* queue, this way, traffic for the given TC will be sent to the default
1048	* queue.
1049	*
1050	* Setup number and offset of Rx queues for all TCs for the VSI
1051	*/
1052	ice_for_each_traffic_class(i) {
1053	if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
1054	/* TC is not enabled */
1055	vsi->tc_cfg.tc_info[i].qoffset = 0;
1056	vsi->tc_cfg.tc_info[i].qcount_rx = 1;
1057	vsi->tc_cfg.tc_info[i].qcount_tx = 1;
1058	vsi->tc_cfg.tc_info[i].netdev_tc = 0;
1059	ctxt->info.tc_mapping[i] = 0;
1060	continue;
1061	}
1062
1063	/* TC is enabled */
1064	vsi->tc_cfg.tc_info[i].qoffset = offset;
1065	vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
1066	vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
1067	vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
1068
1069	qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
1070	ICE_AQ_VSI_TC_Q_OFFSET_M) |
1071	((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
1072	ICE_AQ_VSI_TC_Q_NUM_M);
1073	offset += num_rxq_per_tc;
1074	tx_count += num_txq_per_tc;
1075	ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1076	}
1077
1078	/* if offset is non-zero, means it is calculated correctly based on
1079	* enabled TCs for a given VSI otherwise qcount_rx will always
1080	* be correct and non-zero because it is based off - VSI's
1081	* allocated Rx queues which is at least 1 (hence qcount_tx will be
1082	* at least 1)
1083	*/
1084	if (offset)
1085	rx_count = offset;
1086	else
1087	rx_count = num_rxq_per_tc;
1088
1089	if (rx_count > vsi->alloc_rxq) {
1090	dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
1091	rx_count, vsi->alloc_rxq);
1092	return -EINVAL;
1093	}
1094
1095	if (tx_count > vsi->alloc_txq) {
1096	dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
1097	tx_count, vsi->alloc_txq);
1098	return -EINVAL;
1099	}
1100
1101	vsi->num_txq = tx_count;
1102	vsi->num_rxq = rx_count;
1103
1104	if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
1105	dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
1106	/* since there is a chance that num_rxq could have been changed
1107	* in the above for loop, make num_txq equal to num_rxq.
1108	*/
1109	vsi->num_txq = vsi->num_rxq;
1110	}
1111
1112	/* Rx queue mapping */
1113	ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1114	/* q_mapping buffer holds the info for the first queue allocated for
1115	* this VSI in the PF space and also the number of queues associated
1116	* with this VSI.
1117	*/
1118	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
1119	ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
1120
1121	return 0;
1122	}
1123
1124	/**
1125	* ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
1126	* @ctxt: the VSI context being set
1127	* @vsi: the VSI being configured
1128	*/
1129	static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1130	{
1131	u8 dflt_q_group, dflt_q_prio;
1132	u16 dflt_q, report_q, val;
1133
1134	if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
1135	vsi->type != ICE_VSI_VF && vsi->type != ICE_VSI_CHNL)
1136	return;
1137
1138	val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
1139	ctxt->info.valid_sections |= cpu_to_le16(val);
1140	dflt_q = 0;
1141	dflt_q_group = 0;
1142	report_q = 0;
1143	dflt_q_prio = 0;
1144
1145	/* enable flow director filtering/programming */
1146	val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
1147	ctxt->info.fd_options = cpu_to_le16(val);
1148	/* max of allocated flow director filters */
1149	ctxt->info.max_fd_fltr_dedicated =
1150	cpu_to_le16(vsi->num_gfltr);
1151	/* max of shared flow director filters any VSI may program */
1152	ctxt->info.max_fd_fltr_shared =
1153	cpu_to_le16(vsi->num_bfltr);
1154	/* default queue index within the VSI of the default FD */
1155	val = ((dflt_q << ICE_AQ_VSI_FD_DEF_Q_S) &
1156	ICE_AQ_VSI_FD_DEF_Q_M);
1157	/* target queue or queue group to the FD filter */
1158	val |= ((dflt_q_group << ICE_AQ_VSI_FD_DEF_GRP_S) &
1159	ICE_AQ_VSI_FD_DEF_GRP_M);
1160	ctxt->info.fd_def_q = cpu_to_le16(val);
1161	/* queue index on which FD filter completion is reported */
1162	val = ((report_q << ICE_AQ_VSI_FD_REPORT_Q_S) &
1163	ICE_AQ_VSI_FD_REPORT_Q_M);
1164	/* priority of the default qindex action */
1165	val |= ((dflt_q_prio << ICE_AQ_VSI_FD_DEF_PRIORITY_S) &
1166	ICE_AQ_VSI_FD_DEF_PRIORITY_M);
1167	ctxt->info.fd_report_opt = cpu_to_le16(val);
1168	}
1169
1170	/**
1171	* ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
1172	* @ctxt: the VSI context being set
1173	* @vsi: the VSI being configured
1174	*/
1175	static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1176	{
1177	u8 lut_type, hash_type;
1178	struct device *dev;
1179	struct ice_pf *pf;
1180
1181	pf = vsi->back;
1182	dev = ice_pf_to_dev(pf);
1183
1184	switch (vsi->type) {
1185	case ICE_VSI_CHNL:
1186	case ICE_VSI_PF:
1187	/* PF VSI will inherit RSS instance of PF */
1188	lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
1189	hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
1190	break;
1191	case ICE_VSI_VF:
1192	/* VF VSI will gets a small RSS table which is a VSI LUT type */
1193	lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
1194	hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
1195	break;
1196	default:
1197	dev_dbg(dev, "Unsupported VSI type %s\n",
1198	ice_vsi_type_str(vsi->type));
1199	return;
1200	}
1201
1202	ctxt->info.q_opt_rss = ((lut_type << ICE_AQ_VSI_Q_OPT_RSS_LUT_S) &
1203	ICE_AQ_VSI_Q_OPT_RSS_LUT_M) |
1204	(hash_type & ICE_AQ_VSI_Q_OPT_RSS_HASH_M);
1205	}
1206
1207	static void
1208	ice_chnl_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1209	{
1210	struct ice_pf *pf = vsi->back;
1211	u16 qcount, qmap;
1212	u8 offset = 0;
1213	int pow;
1214
1215	qcount = min_t(int, vsi->num_rxq, pf->num_lan_msix);
1216
1217	pow = order_base_2(qcount);
1218	qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
1219	ICE_AQ_VSI_TC_Q_OFFSET_M) |
1220	((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
1221	ICE_AQ_VSI_TC_Q_NUM_M);
1222
1223	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1224	ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1225	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->next_base_q);
1226	ctxt->info.q_mapping[1] = cpu_to_le16(qcount);
1227	}
1228
1229	/**
1230	* ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
1231	* @vsi: VSI to check whether or not VLAN pruning is enabled.
1232	*
1233	* returns true if Rx VLAN pruning is enabled and false otherwise.
1234	*/
1235	static bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
1236	{
1237	return vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1238	}
1239
1240	/**
1241	* ice_vsi_init - Create and initialize a VSI
1242	* @vsi: the VSI being configured
1243	* @vsi_flags: VSI configuration flags
1244	*
1245	* Set ICE_FLAG_VSI_INIT to initialize a new VSI context, clear it to
1246	* reconfigure an existing context.
1247	*
1248	* This initializes a VSI context depending on the VSI type to be added and
1249	* passes it down to the add_vsi aq command to create a new VSI.
1250	*/
1251	static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
1252	{
1253	struct ice_pf *pf = vsi->back;
1254	struct ice_hw *hw = &pf->hw;
1255	struct ice_vsi_ctx *ctxt;
1256	struct device *dev;
1257	int ret = 0;
1258
1259	dev = ice_pf_to_dev(pf);
1260	ctxt = kzalloc(size: sizeof(*ctxt), GFP_KERNEL);
1261	if (!ctxt)
1262	return -ENOMEM;
1263
1264	switch (vsi->type) {
1265	case ICE_VSI_CTRL:
1266	case ICE_VSI_LB:
1267	case ICE_VSI_PF:
1268	ctxt->flags = ICE_AQ_VSI_TYPE_PF;
1269	break;
1270	case ICE_VSI_SWITCHDEV_CTRL:
1271	case ICE_VSI_CHNL:
1272	ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
1273	break;
1274	case ICE_VSI_VF:
1275	ctxt->flags = ICE_AQ_VSI_TYPE_VF;
1276	/* VF number here is the absolute VF number (0-255) */
1277	ctxt->vf_num = vsi->vf->vf_id + hw->func_caps.vf_base_id;
1278	break;
1279	default:
1280	ret = -ENODEV;
1281	goto out;
1282	}
1283
1284	/* Handle VLAN pruning for channel VSI if main VSI has VLAN
1285	* prune enabled
1286	*/
1287	if (vsi->type == ICE_VSI_CHNL) {
1288	struct ice_vsi *main_vsi;
1289
1290	main_vsi = ice_get_main_vsi(pf);
1291	if (main_vsi && ice_vsi_is_vlan_pruning_ena(vsi: main_vsi))
1292	ctxt->info.sw_flags2 |=
1293	ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1294	else
1295	ctxt->info.sw_flags2 &=
1296	~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1297	}
1298
1299	ice_set_dflt_vsi_ctx(hw, ctxt);
1300	if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
1301	ice_set_fd_vsi_ctx(ctxt, vsi);
1302	/* if the switch is in VEB mode, allow VSI loopback */
1303	if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
1304	ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
1305
1306	/* Set LUT type and HASH type if RSS is enabled */
1307	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1308	vsi->type != ICE_VSI_CTRL) {
1309	ice_set_rss_vsi_ctx(ctxt, vsi);
1310	/* if updating VSI context, make sure to set valid_section:
1311	* to indicate which section of VSI context being updated
1312	*/
1313	if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1314	ctxt->info.valid_sections |=
1315	cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1316	}
1317
1318	ctxt->info.sw_id = vsi->port_info->sw_id;
1319	if (vsi->type == ICE_VSI_CHNL) {
1320	ice_chnl_vsi_setup_q_map(vsi, ctxt);
1321	} else {
1322	ret = ice_vsi_setup_q_map(vsi, ctxt);
1323	if (ret)
1324	goto out;
1325
1326	if (!(vsi_flags & ICE_VSI_FLAG_INIT))
1327	/* means VSI being updated */
1328	/* must to indicate which section of VSI context are
1329	* being modified
1330	*/
1331	ctxt->info.valid_sections |=
1332	cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1333	}
1334
1335	/* Allow control frames out of main VSI */
1336	if (vsi->type == ICE_VSI_PF) {
1337	ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1338	ctxt->info.valid_sections |=
1339	cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1340	}
1341
1342	if (vsi_flags & ICE_VSI_FLAG_INIT) {
1343	ret = ice_add_vsi(hw, vsi_handle: vsi->idx, vsi_ctx: ctxt, NULL);
1344	if (ret) {
1345	dev_err(dev, "Add VSI failed, err %d\n", ret);
1346	ret = -EIO;
1347	goto out;
1348	}
1349	} else {
1350	ret = ice_update_vsi(hw, vsi_handle: vsi->idx, vsi_ctx: ctxt, NULL);
1351	if (ret) {
1352	dev_err(dev, "Update VSI failed, err %d\n", ret);
1353	ret = -EIO;
1354	goto out;
1355	}
1356	}
1357
1358	/* keep context for update VSI operations */
1359	vsi->info = ctxt->info;
1360
1361	/* record VSI number returned */
1362	vsi->vsi_num = ctxt->vsi_num;
1363
1364	out:
1365	kfree(objp: ctxt);
1366	return ret;
1367	}
1368
1369	/**
1370	* ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1371	* @vsi: the VSI having rings deallocated
1372	*/
1373	static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1374	{
1375	int i;
1376
1377	/* Avoid stale references by clearing map from vector to ring */
1378	if (vsi->q_vectors) {
1379	ice_for_each_q_vector(vsi, i) {
1380	struct ice_q_vector *q_vector = vsi->q_vectors[i];
1381
1382	if (q_vector) {
1383	q_vector->tx.tx_ring = NULL;
1384	q_vector->rx.rx_ring = NULL;
1385	}
1386	}
1387	}
1388
1389	if (vsi->tx_rings) {
1390	ice_for_each_alloc_txq(vsi, i) {
1391	if (vsi->tx_rings[i]) {
1392	kfree_rcu(vsi->tx_rings[i], rcu);
1393	WRITE_ONCE(vsi->tx_rings[i], NULL);
1394	}
1395	}
1396	}
1397	if (vsi->rx_rings) {
1398	ice_for_each_alloc_rxq(vsi, i) {
1399	if (vsi->rx_rings[i]) {
1400	kfree_rcu(vsi->rx_rings[i], rcu);
1401	WRITE_ONCE(vsi->rx_rings[i], NULL);
1402	}
1403	}
1404	}
1405	}
1406
1407	/**
1408	* ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1409	* @vsi: VSI which is having rings allocated
1410	*/
1411	static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1412	{
1413	bool dvm_ena = ice_is_dvm_ena(hw: &vsi->back->hw);
1414	struct ice_pf *pf = vsi->back;
1415	struct device *dev;
1416	u16 i;
1417
1418	dev = ice_pf_to_dev(pf);
1419	/* Allocate Tx rings */
1420	ice_for_each_alloc_txq(vsi, i) {
1421	struct ice_tx_ring *ring;
1422
1423	/* allocate with kzalloc(), free with kfree_rcu() */
1424	ring = kzalloc(size: sizeof(*ring), GFP_KERNEL);
1425
1426	if (!ring)
1427	goto err_out;
1428
1429	ring->q_index = i;
1430	ring->reg_idx = vsi->txq_map[i];
1431	ring->vsi = vsi;
1432	ring->tx_tstamps = &pf->ptp.port.tx;
1433	ring->dev = dev;
1434	ring->count = vsi->num_tx_desc;
1435	ring->txq_teid = ICE_INVAL_TEID;
1436	if (dvm_ena)
1437	ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG2;
1438	else
1439	ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG1;
1440	WRITE_ONCE(vsi->tx_rings[i], ring);
1441	}
1442
1443	/* Allocate Rx rings */
1444	ice_for_each_alloc_rxq(vsi, i) {
1445	struct ice_rx_ring *ring;
1446
1447	/* allocate with kzalloc(), free with kfree_rcu() */
1448	ring = kzalloc(size: sizeof(*ring), GFP_KERNEL);
1449	if (!ring)
1450	goto err_out;
1451
1452	ring->q_index = i;
1453	ring->reg_idx = vsi->rxq_map[i];
1454	ring->vsi = vsi;
1455	ring->netdev = vsi->netdev;
1456	ring->dev = dev;
1457	ring->count = vsi->num_rx_desc;
1458	ring->cached_phctime = pf->ptp.cached_phc_time;
1459	WRITE_ONCE(vsi->rx_rings[i], ring);
1460	}
1461
1462	return 0;
1463
1464	err_out:
1465	ice_vsi_clear_rings(vsi);
1466	return -ENOMEM;
1467	}
1468
1469	/**
1470	* ice_vsi_manage_rss_lut - disable/enable RSS
1471	* @vsi: the VSI being changed
1472	* @ena: boolean value indicating if this is an enable or disable request
1473	*
1474	* In the event of disable request for RSS, this function will zero out RSS
1475	* LUT, while in the event of enable request for RSS, it will reconfigure RSS
1476	* LUT.
1477	*/
1478	void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1479	{
1480	u8 *lut;
1481
1482	lut = kzalloc(size: vsi->rss_table_size, GFP_KERNEL);
1483	if (!lut)
1484	return;
1485
1486	if (ena) {
1487	if (vsi->rss_lut_user)
1488	memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1489	else
1490	ice_fill_rss_lut(lut, rss_table_size: vsi->rss_table_size,
1491	rss_size: vsi->rss_size);
1492	}
1493
1494	ice_set_rss_lut(vsi, lut, lut_size: vsi->rss_table_size);
1495	kfree(objp: lut);
1496	}
1497
1498	/**
1499	* ice_vsi_cfg_crc_strip - Configure CRC stripping for a VSI
1500	* @vsi: VSI to be configured
1501	* @disable: set to true to have FCS / CRC in the frame data
1502	*/
1503	void ice_vsi_cfg_crc_strip(struct ice_vsi *vsi, bool disable)
1504	{
1505	int i;
1506
1507	ice_for_each_rxq(vsi, i)
1508	if (disable)
1509	vsi->rx_rings[i]->flags |= ICE_RX_FLAGS_CRC_STRIP_DIS;
1510	else
1511	vsi->rx_rings[i]->flags &= ~ICE_RX_FLAGS_CRC_STRIP_DIS;
1512	}
1513
1514	/**
1515	* ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1516	* @vsi: VSI to be configured
1517	*/
1518	int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1519	{
1520	struct ice_pf *pf = vsi->back;
1521	struct device *dev;
1522	u8 *lut, *key;
1523	int err;
1524
1525	dev = ice_pf_to_dev(pf);
1526	if (vsi->type == ICE_VSI_PF && vsi->ch_rss_size &&
1527	(test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))) {
1528	vsi->rss_size = min_t(u16, vsi->rss_size, vsi->ch_rss_size);
1529	} else {
1530	vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1531
1532	/* If orig_rss_size is valid and it is less than determined
1533	* main VSI's rss_size, update main VSI's rss_size to be
1534	* orig_rss_size so that when tc-qdisc is deleted, main VSI
1535	* RSS table gets programmed to be correct (whatever it was
1536	* to begin with (prior to setup-tc for ADQ config)
1537	*/
1538	if (vsi->orig_rss_size && vsi->rss_size < vsi->orig_rss_size &&
1539	vsi->orig_rss_size <= vsi->num_rxq) {
1540	vsi->rss_size = vsi->orig_rss_size;
1541	/* now orig_rss_size is used, reset it to zero */
1542	vsi->orig_rss_size = 0;
1543	}
1544	}
1545
1546	lut = kzalloc(size: vsi->rss_table_size, GFP_KERNEL);
1547	if (!lut)
1548	return -ENOMEM;
1549
1550	if (vsi->rss_lut_user)
1551	memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1552	else
1553	ice_fill_rss_lut(lut, rss_table_size: vsi->rss_table_size, rss_size: vsi->rss_size);
1554
1555	err = ice_set_rss_lut(vsi, lut, lut_size: vsi->rss_table_size);
1556	if (err) {
1557	dev_err(dev, "set_rss_lut failed, error %d\n", err);
1558	goto ice_vsi_cfg_rss_exit;
1559	}
1560
1561	key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1562	if (!key) {
1563	err = -ENOMEM;
1564	goto ice_vsi_cfg_rss_exit;
1565	}
1566
1567	if (vsi->rss_hkey_user)
1568	memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1569	else
1570	netdev_rss_key_fill(buffer: (void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1571
1572	err = ice_set_rss_key(vsi, seed: key);
1573	if (err)
1574	dev_err(dev, "set_rss_key failed, error %d\n", err);
1575
1576	kfree(objp: key);
1577	ice_vsi_cfg_rss_exit:
1578	kfree(objp: lut);
1579	return err;
1580	}
1581
1582	/**
1583	* ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1584	* @vsi: VSI to be configured
1585	*
1586	* This function will only be called during the VF VSI setup. Upon successful
1587	* completion of package download, this function will configure default RSS
1588	* input sets for VF VSI.
1589	*/
1590	static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1591	{
1592	struct ice_pf *pf = vsi->back;
1593	struct device *dev;
1594	int status;
1595
1596	dev = ice_pf_to_dev(pf);
1597	if (ice_is_safe_mode(pf)) {
1598	dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1599	vsi->vsi_num);
1600	return;
1601	}
1602
1603	status = ice_add_avf_rss_cfg(hw: &pf->hw, vsi_handle: vsi->idx, ICE_DEFAULT_RSS_HENA);
1604	if (status)
1605	dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
1606	vsi->vsi_num, status);
1607	}
1608
1609	/**
1610	* ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1611	* @vsi: VSI to be configured
1612	*
1613	* This function will only be called after successful download package call
1614	* during initialization of PF. Since the downloaded package will erase the
1615	* RSS section, this function will configure RSS input sets for different
1616	* flow types. The last profile added has the highest priority, therefore 2
1617	* tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1618	* (i.e. IPv4 src/dst TCP src/dst port).
1619	*/
1620	static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1621	{
1622	u16 vsi_handle = vsi->idx, vsi_num = vsi->vsi_num;
1623	struct ice_pf *pf = vsi->back;
1624	struct ice_hw *hw = &pf->hw;
1625	struct device *dev;
1626	int status;
1627
1628	dev = ice_pf_to_dev(pf);
1629	if (ice_is_safe_mode(pf)) {
1630	dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1631	vsi_num);
1632	return;
1633	}
1634	/* configure RSS for IPv4 with input set IP src/dst */
1635	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1636	addl_hdrs: ICE_FLOW_SEG_HDR_IPV4);
1637	if (status)
1638	dev_dbg(dev, "ice_add_rss_cfg failed for ipv4 flow, vsi = %d, error = %d\n",
1639	vsi_num, status);
1640
1641	/* configure RSS for IPv6 with input set IPv6 src/dst */
1642	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1643	addl_hdrs: ICE_FLOW_SEG_HDR_IPV6);
1644	if (status)
1645	dev_dbg(dev, "ice_add_rss_cfg failed for ipv6 flow, vsi = %d, error = %d\n",
1646	vsi_num, status);
1647
1648	/* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1649	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV4,
1650	addl_hdrs: ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4);
1651	if (status)
1652	dev_dbg(dev, "ice_add_rss_cfg failed for tcp4 flow, vsi = %d, error = %d\n",
1653	vsi_num, status);
1654
1655	/* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1656	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV4,
1657	addl_hdrs: ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4);
1658	if (status)
1659	dev_dbg(dev, "ice_add_rss_cfg failed for udp4 flow, vsi = %d, error = %d\n",
1660	vsi_num, status);
1661
1662	/* configure RSS for sctp4 with input set IP src/dst */
1663	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1664	addl_hdrs: ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4);
1665	if (status)
1666	dev_dbg(dev, "ice_add_rss_cfg failed for sctp4 flow, vsi = %d, error = %d\n",
1667	vsi_num, status);
1668
1669	/* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1670	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV6,
1671	addl_hdrs: ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6);
1672	if (status)
1673	dev_dbg(dev, "ice_add_rss_cfg failed for tcp6 flow, vsi = %d, error = %d\n",
1674	vsi_num, status);
1675
1676	/* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1677	status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV6,
1678	addl_hdrs: ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6);
1679	if (status)
1680	dev_dbg(dev, "ice_add_rss_cfg failed for udp6 flow, vsi = %d, error = %d\n",
1681	vsi_num, status);
1682
1683	/* configure RSS for sctp6 with input set IPv6 src/dst */
1684	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1685	addl_hdrs: ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6);
1686	if (status)
1687	dev_dbg(dev, "ice_add_rss_cfg failed for sctp6 flow, vsi = %d, error = %d\n",
1688	vsi_num, status);
1689
1690	status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_ESP_SPI,
1691	addl_hdrs: ICE_FLOW_SEG_HDR_ESP);
1692	if (status)
1693	dev_dbg(dev, "ice_add_rss_cfg failed for esp/spi flow, vsi = %d, error = %d\n",
1694	vsi_num, status);
1695	}
1696
1697	/**
1698	* ice_vsi_cfg_frame_size - setup max frame size and Rx buffer length
1699	* @vsi: VSI
1700	*/
1701	static void ice_vsi_cfg_frame_size(struct ice_vsi *vsi)
1702	{
1703	if (!vsi->netdev || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) {
1704	vsi->max_frame = ICE_MAX_FRAME_LEGACY_RX;
1705	vsi->rx_buf_len = ICE_RXBUF_1664;
1706	#if (PAGE_SIZE < 8192)
1707	} else if (!ICE_2K_TOO_SMALL_WITH_PADDING &&
1708	(vsi->netdev->mtu <= ETH_DATA_LEN)) {
1709	vsi->max_frame = ICE_RXBUF_1536 - NET_IP_ALIGN;
1710	vsi->rx_buf_len = ICE_RXBUF_1536 - NET_IP_ALIGN;
1711	#endif
1712	} else {
1713	vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1714	vsi->rx_buf_len = ICE_RXBUF_3072;
1715	}
1716	}
1717
1718	/**
1719	* ice_pf_state_is_nominal - checks the PF for nominal state
1720	* @pf: pointer to PF to check
1721	*
1722	* Check the PF's state for a collection of bits that would indicate
1723	* the PF is in a state that would inhibit normal operation for
1724	* driver functionality.
1725	*
1726	* Returns true if PF is in a nominal state, false otherwise
1727	*/
1728	bool ice_pf_state_is_nominal(struct ice_pf *pf)
1729	{
1730	DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1731
1732	if (!pf)
1733	return false;
1734
1735	bitmap_set(map: check_bits, start: 0, nbits: ICE_STATE_NOMINAL_CHECK_BITS);
1736	if (bitmap_intersects(src1: pf->state, src2: check_bits, nbits: ICE_STATE_NBITS))
1737	return false;
1738
1739	return true;
1740	}
1741
1742	/**
1743	* ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1744	* @vsi: the VSI to be updated
1745	*/
1746	void ice_update_eth_stats(struct ice_vsi *vsi)
1747	{
1748	struct ice_eth_stats *prev_es, *cur_es;
1749	struct ice_hw *hw = &vsi->back->hw;
1750	struct ice_pf *pf = vsi->back;
1751	u16 vsi_num = vsi->vsi_num; /* HW absolute index of a VSI */
1752
1753	prev_es = &vsi->eth_stats_prev;
1754	cur_es = &vsi->eth_stats;
1755
1756	if (ice_is_reset_in_progress(state: pf->state))
1757	vsi->stat_offsets_loaded = false;
1758
1759	ice_stat_update40(hw, GLV_GORCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1760	prev_stat: &prev_es->rx_bytes, cur_stat: &cur_es->rx_bytes);
1761
1762	ice_stat_update40(hw, GLV_UPRCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1763	prev_stat: &prev_es->rx_unicast, cur_stat: &cur_es->rx_unicast);
1764
1765	ice_stat_update40(hw, GLV_MPRCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1766	prev_stat: &prev_es->rx_multicast, cur_stat: &cur_es->rx_multicast);
1767
1768	ice_stat_update40(hw, GLV_BPRCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1769	prev_stat: &prev_es->rx_broadcast, cur_stat: &cur_es->rx_broadcast);
1770
1771	ice_stat_update32(hw, GLV_RDPC(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1772	prev_stat: &prev_es->rx_discards, cur_stat: &cur_es->rx_discards);
1773
1774	ice_stat_update40(hw, GLV_GOTCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1775	prev_stat: &prev_es->tx_bytes, cur_stat: &cur_es->tx_bytes);
1776
1777	ice_stat_update40(hw, GLV_UPTCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1778	prev_stat: &prev_es->tx_unicast, cur_stat: &cur_es->tx_unicast);
1779
1780	ice_stat_update40(hw, GLV_MPTCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1781	prev_stat: &prev_es->tx_multicast, cur_stat: &cur_es->tx_multicast);
1782
1783	ice_stat_update40(hw, GLV_BPTCL(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1784	prev_stat: &prev_es->tx_broadcast, cur_stat: &cur_es->tx_broadcast);
1785
1786	ice_stat_update32(hw, GLV_TEPC(vsi_num), prev_stat_loaded: vsi->stat_offsets_loaded,
1787	prev_stat: &prev_es->tx_errors, cur_stat: &cur_es->tx_errors);
1788
1789	vsi->stat_offsets_loaded = true;
1790	}
1791
1792	/**
1793	* ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1794	* @hw: HW pointer
1795	* @pf_q: index of the Rx queue in the PF's queue space
1796	* @rxdid: flexible descriptor RXDID
1797	* @prio: priority for the RXDID for this queue
1798	* @ena_ts: true to enable timestamp and false to disable timestamp
1799	*/
1800	void
1801	ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio,
1802	bool ena_ts)
1803	{
1804	int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1805
1806	/* clear any previous values */
1807	regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1808	QRXFLXP_CNTXT_RXDID_PRIO_M |
1809	QRXFLXP_CNTXT_TS_M);
1810
1811	regval |= (rxdid << QRXFLXP_CNTXT_RXDID_IDX_S) &
1812	QRXFLXP_CNTXT_RXDID_IDX_M;
1813
1814	regval |= (prio << QRXFLXP_CNTXT_RXDID_PRIO_S) &
1815	QRXFLXP_CNTXT_RXDID_PRIO_M;
1816
1817	if (ena_ts)
1818	/* Enable TimeSync on this queue */
1819	regval |= QRXFLXP_CNTXT_TS_M;
1820
1821	wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1822	}
1823
1824	int ice_vsi_cfg_single_rxq(struct ice_vsi *vsi, u16 q_idx)
1825	{
1826	if (q_idx >= vsi->num_rxq)
1827	return -EINVAL;
1828
1829	return ice_vsi_cfg_rxq(ring: vsi->rx_rings[q_idx]);
1830	}
1831
1832	int ice_vsi_cfg_single_txq(struct ice_vsi *vsi, struct ice_tx_ring **tx_rings, u16 q_idx)
1833	{
1834	DEFINE_FLEX(struct ice_aqc_add_tx_qgrp, qg_buf, txqs, 1);
1835
1836	if (q_idx >= vsi->alloc_txq || !tx_rings || !tx_rings[q_idx])
1837	return -EINVAL;
1838
1839	qg_buf->num_txqs = 1;
1840
1841	return ice_vsi_cfg_txq(vsi, ring: tx_rings[q_idx], qg_buf);
1842	}
1843
1844	/**
1845	* ice_vsi_cfg_rxqs - Configure the VSI for Rx
1846	* @vsi: the VSI being configured
1847	*
1848	* Return 0 on success and a negative value on error
1849	* Configure the Rx VSI for operation.
1850	*/
1851	int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
1852	{
1853	u16 i;
1854
1855	if (vsi->type == ICE_VSI_VF)
1856	goto setup_rings;
1857
1858	ice_vsi_cfg_frame_size(vsi);
1859	setup_rings:
1860	/* set up individual rings */
1861	ice_for_each_rxq(vsi, i) {
1862	int err = ice_vsi_cfg_rxq(ring: vsi->rx_rings[i]);
1863
1864	if (err)
1865	return err;
1866	}
1867
1868	return 0;
1869	}
1870
1871	/**
1872	* ice_vsi_cfg_txqs - Configure the VSI for Tx
1873	* @vsi: the VSI being configured
1874	* @rings: Tx ring array to be configured
1875	* @count: number of Tx ring array elements
1876	*
1877	* Return 0 on success and a negative value on error
1878	* Configure the Tx VSI for operation.
1879	*/
1880	static int
1881	ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_tx_ring **rings, u16 count)
1882	{
1883	DEFINE_FLEX(struct ice_aqc_add_tx_qgrp, qg_buf, txqs, 1);
1884	int err = 0;
1885	u16 q_idx;
1886
1887	qg_buf->num_txqs = 1;
1888
1889	for (q_idx = 0; q_idx < count; q_idx++) {
1890	err = ice_vsi_cfg_txq(vsi, ring: rings[q_idx], qg_buf);
1891	if (err)
1892	break;
1893	}
1894
1895	return err;
1896	}
1897
1898	/**
1899	* ice_vsi_cfg_lan_txqs - Configure the VSI for Tx
1900	* @vsi: the VSI being configured
1901	*
1902	* Return 0 on success and a negative value on error
1903	* Configure the Tx VSI for operation.
1904	*/
1905	int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
1906	{
1907	return ice_vsi_cfg_txqs(vsi, rings: vsi->tx_rings, count: vsi->num_txq);
1908	}
1909
1910	/**
1911	* ice_vsi_cfg_xdp_txqs - Configure Tx queues dedicated for XDP in given VSI
1912	* @vsi: the VSI being configured
1913	*
1914	* Return 0 on success and a negative value on error
1915	* Configure the Tx queues dedicated for XDP in given VSI for operation.
1916	*/
1917	int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
1918	{
1919	int ret;
1920	int i;
1921
1922	ret = ice_vsi_cfg_txqs(vsi, rings: vsi->xdp_rings, count: vsi->num_xdp_txq);
1923	if (ret)
1924	return ret;
1925
1926	ice_for_each_rxq(vsi, i)
1927	ice_tx_xsk_pool(vsi, qid: i);
1928
1929	return 0;
1930	}
1931
1932	/**
1933	* ice_intrl_usec_to_reg - convert interrupt rate limit to register value
1934	* @intrl: interrupt rate limit in usecs
1935	* @gran: interrupt rate limit granularity in usecs
1936	*
1937	* This function converts a decimal interrupt rate limit in usecs to the format
1938	* expected by firmware.
1939	*/
1940	static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
1941	{
1942	u32 val = intrl / gran;
1943
1944	if (val)
1945	return val | GLINT_RATE_INTRL_ENA_M;
1946	return 0;
1947	}
1948
1949	/**
1950	* ice_write_intrl - write throttle rate limit to interrupt specific register
1951	* @q_vector: pointer to interrupt specific structure
1952	* @intrl: throttle rate limit in microseconds to write
1953	*/
1954	void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
1955	{
1956	struct ice_hw *hw = &q_vector->vsi->back->hw;
1957
1958	wr32(hw, GLINT_RATE(q_vector->reg_idx),
1959	ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
1960	}
1961
1962	static struct ice_q_vector *ice_pull_qvec_from_rc(struct ice_ring_container *rc)
1963	{
1964	switch (rc->type) {
1965	case ICE_RX_CONTAINER:
1966	if (rc->rx_ring)
1967	return rc->rx_ring->q_vector;
1968	break;
1969	case ICE_TX_CONTAINER:
1970	if (rc->tx_ring)
1971	return rc->tx_ring->q_vector;
1972	break;
1973	default:
1974	break;
1975	}
1976
1977	return NULL;
1978	}
1979
1980	/**
1981	* __ice_write_itr - write throttle rate to register
1982	* @q_vector: pointer to interrupt data structure
1983	* @rc: pointer to ring container
1984	* @itr: throttle rate in microseconds to write
1985	*/
1986	static void __ice_write_itr(struct ice_q_vector *q_vector,
1987	struct ice_ring_container *rc, u16 itr)
1988	{
1989	struct ice_hw *hw = &q_vector->vsi->back->hw;
1990
1991	wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
1992	ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
1993	}
1994
1995	/**
1996	* ice_write_itr - write throttle rate to queue specific register
1997	* @rc: pointer to ring container
1998	* @itr: throttle rate in microseconds to write
1999	*/
2000	void ice_write_itr(struct ice_ring_container *rc, u16 itr)
2001	{
2002	struct ice_q_vector *q_vector;
2003
2004	q_vector = ice_pull_qvec_from_rc(rc);
2005	if (!q_vector)
2006	return;
2007
2008	__ice_write_itr(q_vector, rc, itr);
2009	}
2010
2011	/**
2012	* ice_set_q_vector_intrl - set up interrupt rate limiting
2013	* @q_vector: the vector to be configured
2014	*
2015	* Interrupt rate limiting is local to the vector, not per-queue so we must
2016	* detect if either ring container has dynamic moderation enabled to decide
2017	* what to set the interrupt rate limit to via INTRL settings. In the case that
2018	* dynamic moderation is disabled on both, write the value with the cached
2019	* setting to make sure INTRL register matches the user visible value.
2020	*/
2021	void ice_set_q_vector_intrl(struct ice_q_vector *q_vector)
2022	{
2023	if (ITR_IS_DYNAMIC(&q_vector->tx) || ITR_IS_DYNAMIC(&q_vector->rx)) {
2024	/* in the case of dynamic enabled, cap each vector to no more
2025	* than (4 us) 250,000 ints/sec, which allows low latency
2026	* but still less than 500,000 interrupts per second, which
2027	* reduces CPU a bit in the case of the lowest latency
2028	* setting. The 4 here is a value in microseconds.
2029	*/
2030	ice_write_intrl(q_vector, intrl: 4);
2031	} else {
2032	ice_write_intrl(q_vector, intrl: q_vector->intrl);
2033	}
2034	}
2035
2036	/**
2037	* ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
2038	* @vsi: the VSI being configured
2039	*
2040	* This configures MSIX mode interrupts for the PF VSI, and should not be used
2041	* for the VF VSI.
2042	*/
2043	void ice_vsi_cfg_msix(struct ice_vsi *vsi)
2044	{
2045	struct ice_pf *pf = vsi->back;
2046	struct ice_hw *hw = &pf->hw;
2047	u16 txq = 0, rxq = 0;
2048	int i, q;
2049
2050	ice_for_each_q_vector(vsi, i) {
2051	struct ice_q_vector *q_vector = vsi->q_vectors[i];
2052	u16 reg_idx = q_vector->reg_idx;
2053
2054	ice_cfg_itr(hw, q_vector);
2055
2056	/* Both Transmit Queue Interrupt Cause Control register
2057	* and Receive Queue Interrupt Cause control register
2058	* expects MSIX_INDX field to be the vector index
2059	* within the function space and not the absolute
2060	* vector index across PF or across device.
2061	* For SR-IOV VF VSIs queue vector index always starts
2062	* with 1 since first vector index(0) is used for OICR
2063	* in VF space. Since VMDq and other PF VSIs are within
2064	* the PF function space, use the vector index that is
2065	* tracked for this PF.
2066	*/
2067	for (q = 0; q < q_vector->num_ring_tx; q++) {
2068	ice_cfg_txq_interrupt(vsi, txq, msix_idx: reg_idx,
2069	itr_idx: q_vector->tx.itr_idx);
2070	txq++;
2071	}
2072
2073	for (q = 0; q < q_vector->num_ring_rx; q++) {
2074	ice_cfg_rxq_interrupt(vsi, rxq, msix_idx: reg_idx,
2075	itr_idx: q_vector->rx.itr_idx);
2076	rxq++;
2077	}
2078	}
2079	}
2080
2081	/**
2082	* ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
2083	* @vsi: the VSI whose rings are to be enabled
2084	*
2085	* Returns 0 on success and a negative value on error
2086	*/
2087	int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
2088	{
2089	return ice_vsi_ctrl_all_rx_rings(vsi, ena: true);
2090	}
2091
2092	/**
2093	* ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
2094	* @vsi: the VSI whose rings are to be disabled
2095	*
2096	* Returns 0 on success and a negative value on error
2097	*/
2098	int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
2099	{
2100	return ice_vsi_ctrl_all_rx_rings(vsi, ena: false);
2101	}
2102
2103	/**
2104	* ice_vsi_stop_tx_rings - Disable Tx rings
2105	* @vsi: the VSI being configured
2106	* @rst_src: reset source
2107	* @rel_vmvf_num: Relative ID of VF/VM
2108	* @rings: Tx ring array to be stopped
2109	* @count: number of Tx ring array elements
2110	*/
2111	static int
2112	ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2113	u16 rel_vmvf_num, struct ice_tx_ring **rings, u16 count)
2114	{
2115	u16 q_idx;
2116
2117	if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2118	return -EINVAL;
2119
2120	for (q_idx = 0; q_idx < count; q_idx++) {
2121	struct ice_txq_meta txq_meta = { };
2122	int status;
2123
2124	if (!rings || !rings[q_idx])
2125	return -EINVAL;
2126
2127	ice_fill_txq_meta(vsi, ring: rings[q_idx], txq_meta: &txq_meta);
2128	status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2129	ring: rings[q_idx], txq_meta: &txq_meta);
2130
2131	if (status)
2132	return status;
2133	}
2134
2135	return 0;
2136	}
2137
2138	/**
2139	* ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2140	* @vsi: the VSI being configured
2141	* @rst_src: reset source
2142	* @rel_vmvf_num: Relative ID of VF/VM
2143	*/
2144	int
2145	ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2146	u16 rel_vmvf_num)
2147	{
2148	return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, rings: vsi->tx_rings, count: vsi->num_txq);
2149	}
2150
2151	/**
2152	* ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2153	* @vsi: the VSI being configured
2154	*/
2155	int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2156	{
2157	return ice_vsi_stop_tx_rings(vsi, rst_src: ICE_NO_RESET, rel_vmvf_num: 0, rings: vsi->xdp_rings, count: vsi->num_xdp_txq);
2158	}
2159
2160	/**
2161	* ice_vsi_is_rx_queue_active
2162	* @vsi: the VSI being configured
2163	*
2164	* Return true if at least one queue is active.
2165	*/
2166	bool ice_vsi_is_rx_queue_active(struct ice_vsi *vsi)
2167	{
2168	struct ice_pf *pf = vsi->back;
2169	struct ice_hw *hw = &pf->hw;
2170	int i;
2171
2172	ice_for_each_rxq(vsi, i) {
2173	u32 rx_reg;
2174	int pf_q;
2175
2176	pf_q = vsi->rxq_map[i];
2177	rx_reg = rd32(hw, QRX_CTRL(pf_q));
2178	if (rx_reg & QRX_CTRL_QENA_STAT_M)
2179	return true;
2180	}
2181
2182	return false;
2183	}
2184
2185	static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2186	{
2187	if (!test_bit(ICE_FLAG_DCB_ENA, vsi->back->flags)) {
2188	vsi->tc_cfg.ena_tc = ICE_DFLT_TRAFFIC_CLASS;
2189	vsi->tc_cfg.numtc = 1;
2190	return;
2191	}
2192
2193	/* set VSI TC information based on DCB config */
2194	ice_vsi_set_dcb_tc_cfg(vsi);
2195	}
2196
2197	/**
2198	* ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2199	* @vsi: the VSI being configured
2200	* @tx: bool to determine Tx or Rx rule
2201	* @create: bool to determine create or remove Rule
2202	*/
2203	void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2204	{
2205	int (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2206	enum ice_sw_fwd_act_type act);
2207	struct ice_pf *pf = vsi->back;
2208	struct device *dev;
2209	int status;
2210
2211	dev = ice_pf_to_dev(pf);
2212	eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2213
2214	if (tx) {
2215	status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2216	ICE_DROP_PACKET);
2217	} else {
2218	if (ice_fw_supports_lldp_fltr_ctrl(hw: &pf->hw)) {
2219	status = ice_lldp_fltr_add_remove(hw: &pf->hw, vsi_num: vsi->vsi_num,
2220	add: create);
2221	} else {
2222	status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2223	ICE_FWD_TO_VSI);
2224	}
2225	}
2226
2227	if (status)
2228	dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %d\n",
2229	create ? "adding" : "removing", tx ? "TX" : "RX",
2230	vsi->vsi_num, status);
2231	}
2232
2233	/**
2234	* ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2235	* @vsi: pointer to the VSI
2236	*
2237	* This function will allocate new scheduler aggregator now if needed and will
2238	* move specified VSI into it.
2239	*/
2240	static void ice_set_agg_vsi(struct ice_vsi *vsi)
2241	{
2242	struct device *dev = ice_pf_to_dev(vsi->back);
2243	struct ice_agg_node *agg_node_iter = NULL;
2244	u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2245	struct ice_agg_node *agg_node = NULL;
2246	int node_offset, max_agg_nodes = 0;
2247	struct ice_port_info *port_info;
2248	struct ice_pf *pf = vsi->back;
2249	u32 agg_node_id_start = 0;
2250	int status;
2251
2252	/* create (as needed) scheduler aggregator node and move VSI into
2253	* corresponding aggregator node
2254	* - PF aggregator node to contains VSIs of type _PF and _CTRL
2255	* - VF aggregator nodes will contain VF VSI
2256	*/
2257	port_info = pf->hw.port_info;
2258	if (!port_info)
2259	return;
2260
2261	switch (vsi->type) {
2262	case ICE_VSI_CTRL:
2263	case ICE_VSI_CHNL:
2264	case ICE_VSI_LB:
2265	case ICE_VSI_PF:
2266	case ICE_VSI_SWITCHDEV_CTRL:
2267	max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2268	agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2269	agg_node_iter = &pf->pf_agg_node[0];
2270	break;
2271	case ICE_VSI_VF:
2272	/* user can create 'n' VFs on a given PF, but since max children
2273	* per aggregator node can be only 64. Following code handles
2274	* aggregator(s) for VF VSIs, either selects a agg_node which
2275	* was already created provided num_vsis < 64, otherwise
2276	* select next available node, which will be created
2277	*/
2278	max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2279	agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2280	agg_node_iter = &pf->vf_agg_node[0];
2281	break;
2282	default:
2283	/* other VSI type, handle later if needed */
2284	dev_dbg(dev, "unexpected VSI type %s\n",
2285	ice_vsi_type_str(vsi->type));
2286	return;
2287	}
2288
2289	/* find the appropriate aggregator node */
2290	for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2291	/* see if we can find space in previously created
2292	* node if num_vsis < 64, otherwise skip
2293	*/
2294	if (agg_node_iter->num_vsis &&
2295	agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2296	agg_node_iter++;
2297	continue;
2298	}
2299
2300	if (agg_node_iter->valid &&
2301	agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2302	agg_id = agg_node_iter->agg_id;
2303	agg_node = agg_node_iter;
2304	break;
2305	}
2306
2307	/* find unclaimed agg_id */
2308	if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2309	agg_id = node_offset + agg_node_id_start;
2310	agg_node = agg_node_iter;
2311	break;
2312	}
2313	/* move to next agg_node */
2314	agg_node_iter++;
2315	}
2316
2317	if (!agg_node)
2318	return;
2319
2320	/* if selected aggregator node was not created, create it */
2321	if (!agg_node->valid) {
2322	status = ice_cfg_agg(pi: port_info, agg_id, agg_type: ICE_AGG_TYPE_AGG,
2323	tc_bitmap: (u8)vsi->tc_cfg.ena_tc);
2324	if (status) {
2325	dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2326	agg_id);
2327	return;
2328	}
2329	/* aggregator node is created, store the needed info */
2330	agg_node->valid = true;
2331	agg_node->agg_id = agg_id;
2332	}
2333
2334	/* move VSI to corresponding aggregator node */
2335	status = ice_move_vsi_to_agg(pi: port_info, agg_id, vsi_handle: vsi->idx,
2336	tc_bitmap: (u8)vsi->tc_cfg.ena_tc);
2337	if (status) {
2338	dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2339	vsi->idx, agg_id);
2340	return;
2341	}
2342
2343	/* keep active children count for aggregator node */
2344	agg_node->num_vsis++;
2345
2346	/* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2347	* to aggregator node
2348	*/
2349	vsi->agg_node = agg_node;
2350	dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2351	vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2352	vsi->agg_node->num_vsis);
2353	}
2354
2355	static int ice_vsi_cfg_tc_lan(struct ice_pf *pf, struct ice_vsi *vsi)
2356	{
2357	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2358	struct device *dev = ice_pf_to_dev(pf);
2359	int ret, i;
2360
2361	/* configure VSI nodes based on number of queues and TC's */
2362	ice_for_each_traffic_class(i) {
2363	if (!(vsi->tc_cfg.ena_tc & BIT(i)))
2364	continue;
2365
2366	if (vsi->type == ICE_VSI_CHNL) {
2367	if (!vsi->alloc_txq && vsi->num_txq)
2368	max_txqs[i] = vsi->num_txq;
2369	else
2370	max_txqs[i] = pf->num_lan_tx;
2371	} else {
2372	max_txqs[i] = vsi->alloc_txq;
2373	}
2374	}
2375
2376	dev_dbg(dev, "vsi->tc_cfg.ena_tc = %d\n", vsi->tc_cfg.ena_tc);
2377	ret = ice_cfg_vsi_lan(pi: vsi->port_info, vsi_handle: vsi->idx, tc_bitmap: vsi->tc_cfg.ena_tc,
2378	max_lanqs: max_txqs);
2379	if (ret) {
2380	dev_err(dev, "VSI %d failed lan queue config, error %d\n",
2381	vsi->vsi_num, ret);
2382	return ret;
2383	}
2384
2385	return 0;
2386	}
2387
2388	/**
2389	* ice_vsi_cfg_def - configure default VSI based on the type
2390	* @vsi: pointer to VSI
2391	* @params: the parameters to configure this VSI with
2392	*/
2393	static int
2394	ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2395	{
2396	struct device *dev = ice_pf_to_dev(vsi->back);
2397	struct ice_pf *pf = vsi->back;
2398	int ret;
2399
2400	vsi->vsw = pf->first_sw;
2401
2402	ret = ice_vsi_alloc_def(vsi, ch: params->ch);
2403	if (ret)
2404	return ret;
2405
2406	/* allocate memory for Tx/Rx ring stat pointers */
2407	ret = ice_vsi_alloc_stat_arrays(vsi);
2408	if (ret)
2409	goto unroll_vsi_alloc;
2410
2411	ice_alloc_fd_res(vsi);
2412
2413	ret = ice_vsi_get_qs(vsi);
2414	if (ret) {
2415	dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2416	vsi->idx);
2417	goto unroll_vsi_alloc_stat;
2418	}
2419
2420	/* set RSS capabilities */
2421	ice_vsi_set_rss_params(vsi);
2422
2423	/* set TC configuration */
2424	ice_vsi_set_tc_cfg(vsi);
2425
2426	/* create the VSI */
2427	ret = ice_vsi_init(vsi, vsi_flags: params->flags);
2428	if (ret)
2429	goto unroll_get_qs;
2430
2431	ice_vsi_init_vlan_ops(vsi);
2432
2433	switch (vsi->type) {
2434	case ICE_VSI_CTRL:
2435	case ICE_VSI_SWITCHDEV_CTRL:
2436	case ICE_VSI_PF:
2437	ret = ice_vsi_alloc_q_vectors(vsi);
2438	if (ret)
2439	goto unroll_vsi_init;
2440
2441	ret = ice_vsi_alloc_rings(vsi);
2442	if (ret)
2443	goto unroll_vector_base;
2444
2445	ret = ice_vsi_alloc_ring_stats(vsi);
2446	if (ret)
2447	goto unroll_vector_base;
2448
2449	ice_vsi_map_rings_to_vectors(vsi);
2450	vsi->stat_offsets_loaded = false;
2451
2452	if (ice_is_xdp_ena_vsi(vsi)) {
2453	ret = ice_vsi_determine_xdp_res(vsi);
2454	if (ret)
2455	goto unroll_vector_base;
2456	ret = ice_prepare_xdp_rings(vsi, prog: vsi->xdp_prog);
2457	if (ret)
2458	goto unroll_vector_base;
2459	}
2460
2461	/* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2462	if (vsi->type != ICE_VSI_CTRL)
2463	/* Do not exit if configuring RSS had an issue, at
2464	* least receive traffic on first queue. Hence no
2465	* need to capture return value
2466	*/
2467	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2468	ice_vsi_cfg_rss_lut_key(vsi);
2469	ice_vsi_set_rss_flow_fld(vsi);
2470	}
2471	ice_init_arfs(vsi);
2472	break;
2473	case ICE_VSI_CHNL:
2474	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2475	ice_vsi_cfg_rss_lut_key(vsi);
2476	ice_vsi_set_rss_flow_fld(vsi);
2477	}
2478	break;
2479	case ICE_VSI_VF:
2480	/* VF driver will take care of creating netdev for this type and
2481	* map queues to vectors through Virtchnl, PF driver only
2482	* creates a VSI and corresponding structures for bookkeeping
2483	* purpose
2484	*/
2485	ret = ice_vsi_alloc_q_vectors(vsi);
2486	if (ret)
2487	goto unroll_vsi_init;
2488
2489	ret = ice_vsi_alloc_rings(vsi);
2490	if (ret)
2491	goto unroll_alloc_q_vector;
2492
2493	ret = ice_vsi_alloc_ring_stats(vsi);
2494	if (ret)
2495	goto unroll_vector_base;
2496
2497	vsi->stat_offsets_loaded = false;
2498
2499	/* Do not exit if configuring RSS had an issue, at least
2500	* receive traffic on first queue. Hence no need to capture
2501	* return value
2502	*/
2503	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2504	ice_vsi_cfg_rss_lut_key(vsi);
2505	ice_vsi_set_vf_rss_flow_fld(vsi);
2506	}
2507	break;
2508	case ICE_VSI_LB:
2509	ret = ice_vsi_alloc_rings(vsi);
2510	if (ret)
2511	goto unroll_vsi_init;
2512
2513	ret = ice_vsi_alloc_ring_stats(vsi);
2514	if (ret)
2515	goto unroll_vector_base;
2516
2517	break;
2518	default:
2519	/* clean up the resources and exit */
2520	ret = -EINVAL;
2521	goto unroll_vsi_init;
2522	}
2523
2524	return 0;
2525
2526	unroll_vector_base:
2527	/* reclaim SW interrupts back to the common pool */
2528	unroll_alloc_q_vector:
2529	ice_vsi_free_q_vectors(vsi);
2530	unroll_vsi_init:
2531	ice_vsi_delete_from_hw(vsi);
2532	unroll_get_qs:
2533	ice_vsi_put_qs(vsi);
2534	unroll_vsi_alloc_stat:
2535	ice_vsi_free_stats(vsi);
2536	unroll_vsi_alloc:
2537	ice_vsi_free_arrays(vsi);
2538	return ret;
2539	}
2540
2541	/**
2542	* ice_vsi_cfg - configure a previously allocated VSI
2543	* @vsi: pointer to VSI
2544	* @params: parameters used to configure this VSI
2545	*/
2546	int ice_vsi_cfg(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
2547	{
2548	struct ice_pf *pf = vsi->back;
2549	int ret;
2550
2551	if (WARN_ON(params->type == ICE_VSI_VF && !params->vf))
2552	return -EINVAL;
2553
2554	vsi->type = params->type;
2555	vsi->port_info = params->pi;
2556
2557	/* For VSIs which don't have a connected VF, this will be NULL */
2558	vsi->vf = params->vf;
2559
2560	ret = ice_vsi_cfg_def(vsi, params);
2561	if (ret)
2562	return ret;
2563
2564	ret = ice_vsi_cfg_tc_lan(pf: vsi->back, vsi);
2565	if (ret)
2566	ice_vsi_decfg(vsi);
2567
2568	if (vsi->type == ICE_VSI_CTRL) {
2569	if (vsi->vf) {
2570	WARN_ON(vsi->vf->ctrl_vsi_idx != ICE_NO_VSI);
2571	vsi->vf->ctrl_vsi_idx = vsi->idx;
2572	} else {
2573	WARN_ON(pf->ctrl_vsi_idx != ICE_NO_VSI);
2574	pf->ctrl_vsi_idx = vsi->idx;
2575	}
2576	}
2577
2578	return ret;
2579	}
2580
2581	/**
2582	* ice_vsi_decfg - remove all VSI configuration
2583	* @vsi: pointer to VSI
2584	*/
2585	void ice_vsi_decfg(struct ice_vsi *vsi)
2586	{
2587	struct ice_pf *pf = vsi->back;
2588	int err;
2589
2590	/* The Rx rule will only exist to remove if the LLDP FW
2591	* engine is currently stopped
2592	*/
2593	if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF &&
2594	!test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
2595	ice_cfg_sw_lldp(vsi, tx: false, create: false);
2596
2597	ice_rm_vsi_lan_cfg(pi: vsi->port_info, vsi_handle: vsi->idx);
2598	err = ice_rm_vsi_rdma_cfg(pi: vsi->port_info, vsi_handle: vsi->idx);
2599	if (err)
2600	dev_err(ice_pf_to_dev(pf), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
2601	vsi->vsi_num, err);
2602
2603	if (ice_is_xdp_ena_vsi(vsi))
2604	/* return value check can be skipped here, it always returns
2605	* 0 if reset is in progress
2606	*/
2607	ice_destroy_xdp_rings(vsi);
2608
2609	ice_vsi_clear_rings(vsi);
2610	ice_vsi_free_q_vectors(vsi);
2611	ice_vsi_put_qs(vsi);
2612	ice_vsi_free_arrays(vsi);
2613
2614	/* SR-IOV determines needed MSIX resources all at once instead of per
2615	* VSI since when VFs are spawned we know how many VFs there are and how
2616	* many interrupts each VF needs. SR-IOV MSIX resources are also
2617	* cleared in the same manner.
2618	*/
2619
2620	if (vsi->type == ICE_VSI_VF &&
2621	vsi->agg_node && vsi->agg_node->valid)
2622	vsi->agg_node->num_vsis--;
2623	if (vsi->agg_node) {
2624	vsi->agg_node->valid = false;
2625	vsi->agg_node->agg_id = 0;
2626	}
2627	}
2628
2629	/**
2630	* ice_vsi_setup - Set up a VSI by a given type
2631	* @pf: board private structure
2632	* @params: parameters to use when creating the VSI
2633	*
2634	* This allocates the sw VSI structure and its queue resources.
2635	*
2636	* Returns pointer to the successfully allocated and configured VSI sw struct on
2637	* success, NULL on failure.
2638	*/
2639	struct ice_vsi *
2640	ice_vsi_setup(struct ice_pf *pf, struct ice_vsi_cfg_params *params)
2641	{
2642	struct device *dev = ice_pf_to_dev(pf);
2643	struct ice_vsi *vsi;
2644	int ret;
2645
2646	/* ice_vsi_setup can only initialize a new VSI, and we must have
2647	* a port_info structure for it.
2648	*/
2649	if (WARN_ON(!(params->flags & ICE_VSI_FLAG_INIT)) ||
2650	WARN_ON(!params->pi))
2651	return NULL;
2652
2653	vsi = ice_vsi_alloc(pf);
2654	if (!vsi) {
2655	dev_err(dev, "could not allocate VSI\n");
2656	return NULL;
2657	}
2658
2659	ret = ice_vsi_cfg(vsi, params);
2660	if (ret)
2661	goto err_vsi_cfg;
2662
2663	/* Add switch rule to drop all Tx Flow Control Frames, of look up
2664	* type ETHERTYPE from VSIs, and restrict malicious VF from sending
2665	* out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2666	* The rule is added once for PF VSI in order to create appropriate
2667	* recipe, since VSI/VSI list is ignored with drop action...
2668	* Also add rules to handle LLDP Tx packets. Tx LLDP packets need to
2669	* be dropped so that VFs cannot send LLDP packets to reconfig DCB
2670	* settings in the HW.
2671	*/
2672	if (!ice_is_safe_mode(pf) && vsi->type == ICE_VSI_PF) {
2673	ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2674	action: ICE_DROP_PACKET);
2675	ice_cfg_sw_lldp(vsi, tx: true, create: true);
2676	}
2677
2678	if (!vsi->agg_node)
2679	ice_set_agg_vsi(vsi);
2680
2681	return vsi;
2682
2683	err_vsi_cfg:
2684	ice_vsi_free(vsi);
2685
2686	return NULL;
2687	}
2688
2689	/**
2690	* ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2691	* @vsi: the VSI being cleaned up
2692	*/
2693	static void ice_vsi_release_msix(struct ice_vsi *vsi)
2694	{
2695	struct ice_pf *pf = vsi->back;
2696	struct ice_hw *hw = &pf->hw;
2697	u32 txq = 0;
2698	u32 rxq = 0;
2699	int i, q;
2700
2701	ice_for_each_q_vector(vsi, i) {
2702	struct ice_q_vector *q_vector = vsi->q_vectors[i];
2703
2704	ice_write_intrl(q_vector, intrl: 0);
2705	for (q = 0; q < q_vector->num_ring_tx; q++) {
2706	ice_write_itr(rc: &q_vector->tx, itr: 0);
2707	wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2708	if (ice_is_xdp_ena_vsi(vsi)) {
2709	u32 xdp_txq = txq + vsi->num_xdp_txq;
2710
2711	wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2712	}
2713	txq++;
2714	}
2715
2716	for (q = 0; q < q_vector->num_ring_rx; q++) {
2717	ice_write_itr(rc: &q_vector->rx, itr: 0);
2718	wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2719	rxq++;
2720	}
2721	}
2722
2723	ice_flush(hw);
2724	}
2725
2726	/**
2727	* ice_vsi_free_irq - Free the IRQ association with the OS
2728	* @vsi: the VSI being configured
2729	*/
2730	void ice_vsi_free_irq(struct ice_vsi *vsi)
2731	{
2732	struct ice_pf *pf = vsi->back;
2733	int i;
2734
2735	if (!vsi->q_vectors || !vsi->irqs_ready)
2736	return;
2737
2738	ice_vsi_release_msix(vsi);
2739	if (vsi->type == ICE_VSI_VF)
2740	return;
2741
2742	vsi->irqs_ready = false;
2743	ice_free_cpu_rx_rmap(vsi);
2744
2745	ice_for_each_q_vector(vsi, i) {
2746	int irq_num;
2747
2748	irq_num = vsi->q_vectors[i]->irq.virq;
2749
2750	/* free only the irqs that were actually requested */
2751	if (!vsi->q_vectors[i] ||
2752	!(vsi->q_vectors[i]->num_ring_tx ||
2753	vsi->q_vectors[i]->num_ring_rx))
2754	continue;
2755
2756	/* clear the affinity notifier in the IRQ descriptor */
2757	if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2758	irq_set_affinity_notifier(irq: irq_num, NULL);
2759
2760	/* clear the affinity_mask in the IRQ descriptor */
2761	irq_set_affinity_hint(irq: irq_num, NULL);
2762	synchronize_irq(irq: irq_num);
2763	devm_free_irq(ice_pf_to_dev(pf), irq: irq_num, dev_id: vsi->q_vectors[i]);
2764	}
2765	}
2766
2767	/**
2768	* ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2769	* @vsi: the VSI having resources freed
2770	*/
2771	void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2772	{
2773	int i;
2774
2775	if (!vsi->tx_rings)
2776	return;
2777
2778	ice_for_each_txq(vsi, i)
2779	if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2780	ice_free_tx_ring(tx_ring: vsi->tx_rings[i]);
2781	}
2782
2783	/**
2784	* ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2785	* @vsi: the VSI having resources freed
2786	*/
2787	void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2788	{
2789	int i;
2790
2791	if (!vsi->rx_rings)
2792	return;
2793
2794	ice_for_each_rxq(vsi, i)
2795	if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2796	ice_free_rx_ring(rx_ring: vsi->rx_rings[i]);
2797	}
2798
2799	/**
2800	* ice_vsi_close - Shut down a VSI
2801	* @vsi: the VSI being shut down
2802	*/
2803	void ice_vsi_close(struct ice_vsi *vsi)
2804	{
2805	if (!test_and_set_bit(nr: ICE_VSI_DOWN, addr: vsi->state))
2806	ice_down(vsi);
2807
2808	ice_vsi_free_irq(vsi);
2809	ice_vsi_free_tx_rings(vsi);
2810	ice_vsi_free_rx_rings(vsi);
2811	}
2812
2813	/**
2814	* ice_ena_vsi - resume a VSI
2815	* @vsi: the VSI being resume
2816	* @locked: is the rtnl_lock already held
2817	*/
2818	int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2819	{
2820	int err = 0;
2821
2822	if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2823	return 0;
2824
2825	clear_bit(nr: ICE_VSI_NEEDS_RESTART, addr: vsi->state);
2826
2827	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2828	if (netif_running(dev: vsi->netdev)) {
2829	if (!locked)
2830	rtnl_lock();
2831
2832	err = ice_open_internal(netdev: vsi->netdev);
2833
2834	if (!locked)
2835	rtnl_unlock();
2836	}
2837	} else if (vsi->type == ICE_VSI_CTRL) {
2838	err = ice_vsi_open_ctrl(vsi);
2839	}
2840
2841	return err;
2842	}
2843
2844	/**
2845	* ice_dis_vsi - pause a VSI
2846	* @vsi: the VSI being paused
2847	* @locked: is the rtnl_lock already held
2848	*/
2849	void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2850	{
2851	if (test_bit(ICE_VSI_DOWN, vsi->state))
2852	return;
2853
2854	set_bit(nr: ICE_VSI_NEEDS_RESTART, addr: vsi->state);
2855
2856	if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2857	if (netif_running(dev: vsi->netdev)) {
2858	if (!locked)
2859	rtnl_lock();
2860
2861	ice_vsi_close(vsi);
2862
2863	if (!locked)
2864	rtnl_unlock();
2865	} else {
2866	ice_vsi_close(vsi);
2867	}
2868	} else if (vsi->type == ICE_VSI_CTRL ||
2869	vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
2870	ice_vsi_close(vsi);
2871	}
2872	}
2873
2874	/**
2875	* ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
2876	* @vsi: the VSI being un-configured
2877	*/
2878	void ice_vsi_dis_irq(struct ice_vsi *vsi)
2879	{
2880	struct ice_pf *pf = vsi->back;
2881	struct ice_hw *hw = &pf->hw;
2882	u32 val;
2883	int i;
2884
2885	/* disable interrupt causation from each queue */
2886	if (vsi->tx_rings) {
2887	ice_for_each_txq(vsi, i) {
2888	if (vsi->tx_rings[i]) {
2889	u16 reg;
2890
2891	reg = vsi->tx_rings[i]->reg_idx;
2892	val = rd32(hw, QINT_TQCTL(reg));
2893	val &= ~QINT_TQCTL_CAUSE_ENA_M;
2894	wr32(hw, QINT_TQCTL(reg), val);
2895	}
2896	}
2897	}
2898
2899	if (vsi->rx_rings) {
2900	ice_for_each_rxq(vsi, i) {
2901	if (vsi->rx_rings[i]) {
2902	u16 reg;
2903
2904	reg = vsi->rx_rings[i]->reg_idx;
2905	val = rd32(hw, QINT_RQCTL(reg));
2906	val &= ~QINT_RQCTL_CAUSE_ENA_M;
2907	wr32(hw, QINT_RQCTL(reg), val);
2908	}
2909	}
2910	}
2911
2912	/* disable each interrupt */
2913	ice_for_each_q_vector(vsi, i) {
2914	if (!vsi->q_vectors[i])
2915	continue;
2916	wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
2917	}
2918
2919	ice_flush(hw);
2920
2921	/* don't call synchronize_irq() for VF's from the host */
2922	if (vsi->type == ICE_VSI_VF)
2923	return;
2924
2925	ice_for_each_q_vector(vsi, i)
2926	synchronize_irq(irq: vsi->q_vectors[i]->irq.virq);
2927	}
2928
2929	/**
2930	* ice_vsi_release - Delete a VSI and free its resources
2931	* @vsi: the VSI being removed
2932	*
2933	* Returns 0 on success or < 0 on error
2934	*/
2935	int ice_vsi_release(struct ice_vsi *vsi)
2936	{
2937	struct ice_pf *pf;
2938
2939	if (!vsi->back)
2940	return -ENODEV;
2941	pf = vsi->back;
2942
2943	if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2944	ice_rss_clean(vsi);
2945
2946	ice_vsi_close(vsi);
2947	ice_vsi_decfg(vsi);
2948
2949	/* retain SW VSI data structure since it is needed to unregister and
2950	* free VSI netdev when PF is not in reset recovery pending state,\
2951	* for ex: during rmmod.
2952	*/
2953	if (!ice_is_reset_in_progress(state: pf->state))
2954	ice_vsi_delete(vsi);
2955
2956	return 0;
2957	}
2958
2959	/**
2960	* ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
2961	* @vsi: VSI connected with q_vectors
2962	* @coalesce: array of struct with stored coalesce
2963	*
2964	* Returns array size.
2965	*/
2966	static int
2967	ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
2968	struct ice_coalesce_stored *coalesce)
2969	{
2970	int i;
2971
2972	ice_for_each_q_vector(vsi, i) {
2973	struct ice_q_vector *q_vector = vsi->q_vectors[i];
2974
2975	coalesce[i].itr_tx = q_vector->tx.itr_settings;
2976	coalesce[i].itr_rx = q_vector->rx.itr_settings;
2977	coalesce[i].intrl = q_vector->intrl;
2978
2979	if (i < vsi->num_txq)
2980	coalesce[i].tx_valid = true;
2981	if (i < vsi->num_rxq)
2982	coalesce[i].rx_valid = true;
2983	}
2984
2985	return vsi->num_q_vectors;
2986	}
2987
2988	/**
2989	* ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
2990	* @vsi: VSI connected with q_vectors
2991	* @coalesce: pointer to array of struct with stored coalesce
2992	* @size: size of coalesce array
2993	*
2994	* Before this function, ice_vsi_rebuild_get_coalesce should be called to save
2995	* ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
2996	* to default value.
2997	*/
2998	static void
2999	ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
3000	struct ice_coalesce_stored *coalesce, int size)
3001	{
3002	struct ice_ring_container *rc;
3003	int i;
3004
3005	if ((size && !coalesce) || !vsi)
3006	return;
3007
3008	/* There are a couple of cases that have to be handled here:
3009	* 1. The case where the number of queue vectors stays the same, but
3010	* the number of Tx or Rx rings changes (the first for loop)
3011	* 2. The case where the number of queue vectors increased (the
3012	* second for loop)
3013	*/
3014	for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
3015	/* There are 2 cases to handle here and they are the same for
3016	* both Tx and Rx:
3017	* if the entry was valid previously (coalesce[i].[tr]x_valid
3018	* and the loop variable is less than the number of rings
3019	* allocated, then write the previous values
3020	*
3021	* if the entry was not valid previously, but the number of
3022	* rings is less than are allocated (this means the number of
3023	* rings increased from previously), then write out the
3024	* values in the first element
3025	*
3026	* Also, always write the ITR, even if in ITR_IS_DYNAMIC
3027	* as there is no harm because the dynamic algorithm
3028	* will just overwrite.
3029	*/
3030	if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
3031	rc = &vsi->q_vectors[i]->rx;
3032	rc->itr_settings = coalesce[i].itr_rx;
3033	ice_write_itr(rc, itr: rc->itr_setting);
3034	} else if (i < vsi->alloc_rxq) {
3035	rc = &vsi->q_vectors[i]->rx;
3036	rc->itr_settings = coalesce[0].itr_rx;
3037	ice_write_itr(rc, itr: rc->itr_setting);
3038	}
3039
3040	if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
3041	rc = &vsi->q_vectors[i]->tx;
3042	rc->itr_settings = coalesce[i].itr_tx;
3043	ice_write_itr(rc, itr: rc->itr_setting);
3044	} else if (i < vsi->alloc_txq) {
3045	rc = &vsi->q_vectors[i]->tx;
3046	rc->itr_settings = coalesce[0].itr_tx;
3047	ice_write_itr(rc, itr: rc->itr_setting);
3048	}
3049
3050	vsi->q_vectors[i]->intrl = coalesce[i].intrl;
3051	ice_set_q_vector_intrl(q_vector: vsi->q_vectors[i]);
3052	}
3053
3054	/* the number of queue vectors increased so write whatever is in
3055	* the first element
3056	*/
3057	for (; i < vsi->num_q_vectors; i++) {
3058	/* transmit */
3059	rc = &vsi->q_vectors[i]->tx;
3060	rc->itr_settings = coalesce[0].itr_tx;
3061	ice_write_itr(rc, itr: rc->itr_setting);
3062
3063	/* receive */
3064	rc = &vsi->q_vectors[i]->rx;
3065	rc->itr_settings = coalesce[0].itr_rx;
3066	ice_write_itr(rc, itr: rc->itr_setting);
3067
3068	vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3069	ice_set_q_vector_intrl(q_vector: vsi->q_vectors[i]);
3070	}
3071	}
3072
3073	/**
3074	* ice_vsi_realloc_stat_arrays - Frees unused stat structures
3075	* @vsi: VSI pointer
3076	* @prev_txq: Number of Tx rings before ring reallocation
3077	* @prev_rxq: Number of Rx rings before ring reallocation
3078	*/
3079	static void
3080	ice_vsi_realloc_stat_arrays(struct ice_vsi *vsi, int prev_txq, int prev_rxq)
3081	{
3082	struct ice_vsi_stats *vsi_stat;
3083	struct ice_pf *pf = vsi->back;
3084	int i;
3085
3086	if (!prev_txq || !prev_rxq)
3087	return;
3088	if (vsi->type == ICE_VSI_CHNL)
3089	return;
3090
3091	vsi_stat = pf->vsi_stats[vsi->idx];
3092
3093	if (vsi->num_txq < prev_txq) {
3094	for (i = vsi->num_txq; i < prev_txq; i++) {
3095	if (vsi_stat->tx_ring_stats[i]) {
3096	kfree_rcu(vsi_stat->tx_ring_stats[i], rcu);
3097	WRITE_ONCE(vsi_stat->tx_ring_stats[i], NULL);
3098	}
3099	}
3100	}
3101
3102	if (vsi->num_rxq < prev_rxq) {
3103	for (i = vsi->num_rxq; i < prev_rxq; i++) {
3104	if (vsi_stat->rx_ring_stats[i]) {
3105	kfree_rcu(vsi_stat->rx_ring_stats[i], rcu);
3106	WRITE_ONCE(vsi_stat->rx_ring_stats[i], NULL);
3107	}
3108	}
3109	}
3110	}
3111
3112	/**
3113	* ice_vsi_rebuild - Rebuild VSI after reset
3114	* @vsi: VSI to be rebuild
3115	* @vsi_flags: flags used for VSI rebuild flow
3116	*
3117	* Set vsi_flags to ICE_VSI_FLAG_INIT to initialize a new VSI, or
3118	* ICE_VSI_FLAG_NO_INIT to rebuild an existing VSI in hardware.
3119	*
3120	* Returns 0 on success and negative value on failure
3121	*/
3122	int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags)
3123	{
3124	struct ice_vsi_cfg_params params = {};
3125	struct ice_coalesce_stored *coalesce;
3126	int ret, prev_txq, prev_rxq;
3127	int prev_num_q_vectors = 0;
3128	struct ice_pf *pf;
3129
3130	if (!vsi)
3131	return -EINVAL;
3132
3133	params = ice_vsi_to_params(vsi);
3134	params.flags = vsi_flags;
3135
3136	pf = vsi->back;
3137	if (WARN_ON(vsi->type == ICE_VSI_VF && !vsi->vf))
3138	return -EINVAL;
3139
3140	coalesce = kcalloc(n: vsi->num_q_vectors,
3141	size: sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3142	if (!coalesce)
3143	return -ENOMEM;
3144
3145	prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3146
3147	prev_txq = vsi->num_txq;
3148	prev_rxq = vsi->num_rxq;
3149
3150	ice_vsi_decfg(vsi);
3151	ret = ice_vsi_cfg_def(vsi, params: &params);
3152	if (ret)
3153	goto err_vsi_cfg;
3154
3155	ret = ice_vsi_cfg_tc_lan(pf, vsi);
3156	if (ret) {
3157	if (vsi_flags & ICE_VSI_FLAG_INIT) {
3158	ret = -EIO;
3159	goto err_vsi_cfg_tc_lan;
3160	}
3161
3162	kfree(objp: coalesce);
3163	return ice_schedule_reset(pf, reset: ICE_RESET_PFR);
3164	}
3165
3166	ice_vsi_realloc_stat_arrays(vsi, prev_txq, prev_rxq);
3167
3168	ice_vsi_rebuild_set_coalesce(vsi, coalesce, size: prev_num_q_vectors);
3169	kfree(objp: coalesce);
3170
3171	return 0;
3172
3173	err_vsi_cfg_tc_lan:
3174	ice_vsi_decfg(vsi);
3175	err_vsi_cfg:
3176	kfree(objp: coalesce);
3177	return ret;
3178	}
3179
3180	/**
3181	* ice_is_reset_in_progress - check for a reset in progress
3182	* @state: PF state field
3183	*/
3184	bool ice_is_reset_in_progress(unsigned long *state)
3185	{
3186	return test_bit(ICE_RESET_OICR_RECV, state) ||
3187	test_bit(ICE_PFR_REQ, state) ||
3188	test_bit(ICE_CORER_REQ, state) ||
3189	test_bit(ICE_GLOBR_REQ, state);
3190	}
3191
3192	/**
3193	* ice_wait_for_reset - Wait for driver to finish reset and rebuild
3194	* @pf: pointer to the PF structure
3195	* @timeout: length of time to wait, in jiffies
3196	*
3197	* Wait (sleep) for a short time until the driver finishes cleaning up from
3198	* a device reset. The caller must be able to sleep. Use this to delay
3199	* operations that could fail while the driver is cleaning up after a device
3200	* reset.
3201	*
3202	* Returns 0 on success, -EBUSY if the reset is not finished within the
3203	* timeout, and -ERESTARTSYS if the thread was interrupted.
3204	*/
3205	int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3206	{
3207	long ret;
3208
3209	ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3210	!ice_is_reset_in_progress(pf->state),
3211	timeout);
3212	if (ret < 0)
3213	return ret;
3214	else if (!ret)
3215	return -EBUSY;
3216	else
3217	return 0;
3218	}
3219
3220	/**
3221	* ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3222	* @vsi: VSI being configured
3223	* @ctx: the context buffer returned from AQ VSI update command
3224	*/
3225	static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3226	{
3227	vsi->info.mapping_flags = ctx->info.mapping_flags;
3228	memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3229	sizeof(vsi->info.q_mapping));
3230	memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3231	sizeof(vsi->info.tc_mapping));
3232	}
3233
3234	/**
3235	* ice_vsi_cfg_netdev_tc - Setup the netdev TC configuration
3236	* @vsi: the VSI being configured
3237	* @ena_tc: TC map to be enabled
3238	*/
3239	void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc)
3240	{
3241	struct net_device *netdev = vsi->netdev;
3242	struct ice_pf *pf = vsi->back;
3243	int numtc = vsi->tc_cfg.numtc;
3244	struct ice_dcbx_cfg *dcbcfg;
3245	u8 netdev_tc;
3246	int i;
3247
3248	if (!netdev)
3249	return;
3250
3251	/* CHNL VSI doesn't have it's own netdev, hence, no netdev_tc */
3252	if (vsi->type == ICE_VSI_CHNL)
3253	return;
3254
3255	if (!ena_tc) {
3256	netdev_reset_tc(dev: netdev);
3257	return;
3258	}
3259
3260	if (vsi->type == ICE_VSI_PF && ice_is_adq_active(pf))
3261	numtc = vsi->all_numtc;
3262
3263	if (netdev_set_num_tc(dev: netdev, num_tc: numtc))
3264	return;
3265
3266	dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg;
3267
3268	ice_for_each_traffic_class(i)
3269	if (vsi->tc_cfg.ena_tc & BIT(i))
3270	netdev_set_tc_queue(dev: netdev,
3271	tc: vsi->tc_cfg.tc_info[i].netdev_tc,
3272	count: vsi->tc_cfg.tc_info[i].qcount_tx,
3273	offset: vsi->tc_cfg.tc_info[i].qoffset);
3274	/* setup TC queue map for CHNL TCs */
3275	ice_for_each_chnl_tc(i) {
3276	if (!(vsi->all_enatc & BIT(i)))
3277	break;
3278	if (!vsi->mqprio_qopt.qopt.count[i])
3279	break;
3280	netdev_set_tc_queue(dev: netdev, tc: i,
3281	count: vsi->mqprio_qopt.qopt.count[i],
3282	offset: vsi->mqprio_qopt.qopt.offset[i]);
3283	}
3284
3285	if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3286	return;
3287
3288	for (i = 0; i < ICE_MAX_USER_PRIORITY; i++) {
3289	u8 ets_tc = dcbcfg->etscfg.prio_table[i];
3290
3291	/* Get the mapped netdev TC# for the UP */
3292	netdev_tc = vsi->tc_cfg.tc_info[ets_tc].netdev_tc;
3293	netdev_set_prio_tc_map(dev: netdev, prio: i, tc: netdev_tc);
3294	}
3295	}
3296
3297	/**
3298	* ice_vsi_setup_q_map_mqprio - Prepares mqprio based tc_config
3299	* @vsi: the VSI being configured,
3300	* @ctxt: VSI context structure
3301	* @ena_tc: number of traffic classes to enable
3302	*
3303	* Prepares VSI tc_config to have queue configurations based on MQPRIO options.
3304	*/
3305	static int
3306	ice_vsi_setup_q_map_mqprio(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt,
3307	u8 ena_tc)
3308	{
3309	u16 pow, offset = 0, qcount_tx = 0, qcount_rx = 0, qmap;
3310	u16 tc0_offset = vsi->mqprio_qopt.qopt.offset[0];
3311	int tc0_qcount = vsi->mqprio_qopt.qopt.count[0];
3312	u16 new_txq, new_rxq;
3313	u8 netdev_tc = 0;
3314	int i;
3315
3316	vsi->tc_cfg.ena_tc = ena_tc ? ena_tc : 1;
3317
3318	pow = order_base_2(tc0_qcount);
3319	qmap = ((tc0_offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
3320	ICE_AQ_VSI_TC_Q_OFFSET_M) |
3321	((pow << ICE_AQ_VSI_TC_Q_NUM_S) & ICE_AQ_VSI_TC_Q_NUM_M);
3322
3323	ice_for_each_traffic_class(i) {
3324	if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
3325	/* TC is not enabled */
3326	vsi->tc_cfg.tc_info[i].qoffset = 0;
3327	vsi->tc_cfg.tc_info[i].qcount_rx = 1;
3328	vsi->tc_cfg.tc_info[i].qcount_tx = 1;
3329	vsi->tc_cfg.tc_info[i].netdev_tc = 0;
3330	ctxt->info.tc_mapping[i] = 0;
3331	continue;
3332	}
3333
3334	offset = vsi->mqprio_qopt.qopt.offset[i];
3335	qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3336	qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3337	vsi->tc_cfg.tc_info[i].qoffset = offset;
3338	vsi->tc_cfg.tc_info[i].qcount_rx = qcount_rx;
3339	vsi->tc_cfg.tc_info[i].qcount_tx = qcount_tx;
3340	vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
3341	}
3342
3343	if (vsi->all_numtc && vsi->all_numtc != vsi->tc_cfg.numtc) {
3344	ice_for_each_chnl_tc(i) {
3345	if (!(vsi->all_enatc & BIT(i)))
3346	continue;
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	}
3351	}
3352
3353	new_txq = offset + qcount_tx;
3354	if (new_txq > vsi->alloc_txq) {
3355	dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
3356	new_txq, vsi->alloc_txq);
3357	return -EINVAL;
3358	}
3359
3360	new_rxq = offset + qcount_rx;
3361	if (new_rxq > vsi->alloc_rxq) {
3362	dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
3363	new_rxq, vsi->alloc_rxq);
3364	return -EINVAL;
3365	}
3366
3367	/* Set actual Tx/Rx queue pairs */
3368	vsi->num_txq = new_txq;
3369	vsi->num_rxq = new_rxq;
3370
3371	/* Setup queue TC[0].qmap for given VSI context */
3372	ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
3373	ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
3374	ctxt->info.q_mapping[1] = cpu_to_le16(tc0_qcount);
3375
3376	/* Find queue count available for channel VSIs and starting offset
3377	* for channel VSIs
3378	*/
3379	if (tc0_qcount && tc0_qcount < vsi->num_rxq) {
3380	vsi->cnt_q_avail = vsi->num_rxq - tc0_qcount;
3381	vsi->next_base_q = tc0_qcount;
3382	}
3383	dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_txq = %d\n", vsi->num_txq);
3384	dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_rxq = %d\n", vsi->num_rxq);
3385	dev_dbg(ice_pf_to_dev(vsi->back), "all_numtc %u, all_enatc: 0x%04x, tc_cfg.numtc %u\n",
3386	vsi->all_numtc, vsi->all_enatc, vsi->tc_cfg.numtc);
3387
3388	return 0;
3389	}
3390
3391	/**
3392	* ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3393	* @vsi: VSI to be configured
3394	* @ena_tc: TC bitmap
3395	*
3396	* VSI queues expected to be quiesced before calling this function
3397	*/
3398	int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3399	{
3400	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3401	struct ice_pf *pf = vsi->back;
3402	struct ice_tc_cfg old_tc_cfg;
3403	struct ice_vsi_ctx *ctx;
3404	struct device *dev;
3405	int i, ret = 0;
3406	u8 num_tc = 0;
3407
3408	dev = ice_pf_to_dev(pf);
3409	if (vsi->tc_cfg.ena_tc == ena_tc &&
3410	vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
3411	return 0;
3412
3413	ice_for_each_traffic_class(i) {
3414	/* build bitmap of enabled TCs */
3415	if (ena_tc & BIT(i))
3416	num_tc++;
3417	/* populate max_txqs per TC */
3418	max_txqs[i] = vsi->alloc_txq;
3419	/* Update max_txqs if it is CHNL VSI, because alloc_t[r]xq are
3420	* zero for CHNL VSI, hence use num_txq instead as max_txqs
3421	*/
3422	if (vsi->type == ICE_VSI_CHNL &&
3423	test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3424	max_txqs[i] = vsi->num_txq;
3425	}
3426
3427	memcpy(&old_tc_cfg, &vsi->tc_cfg, sizeof(old_tc_cfg));
3428	vsi->tc_cfg.ena_tc = ena_tc;
3429	vsi->tc_cfg.numtc = num_tc;
3430
3431	ctx = kzalloc(size: sizeof(*ctx), GFP_KERNEL);
3432	if (!ctx)
3433	return -ENOMEM;
3434
3435	ctx->vf_num = 0;
3436	ctx->info = vsi->info;
3437
3438	if (vsi->type == ICE_VSI_PF &&
3439	test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3440	ret = ice_vsi_setup_q_map_mqprio(vsi, ctxt: ctx, ena_tc);
3441	else
3442	ret = ice_vsi_setup_q_map(vsi, ctxt: ctx);
3443
3444	if (ret) {
3445	memcpy(&vsi->tc_cfg, &old_tc_cfg, sizeof(vsi->tc_cfg));
3446	goto out;
3447	}
3448
3449	/* must to indicate which section of VSI context are being modified */
3450	ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3451	ret = ice_update_vsi(hw: &pf->hw, vsi_handle: vsi->idx, vsi_ctx: ctx, NULL);
3452	if (ret) {
3453	dev_info(dev, "Failed VSI Update\n");
3454	goto out;
3455	}
3456
3457	if (vsi->type == ICE_VSI_PF &&
3458	test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3459	ret = ice_cfg_vsi_lan(pi: vsi->port_info, vsi_handle: vsi->idx, tc_bitmap: 1, max_lanqs: max_txqs);
3460	else
3461	ret = ice_cfg_vsi_lan(pi: vsi->port_info, vsi_handle: vsi->idx,
3462	tc_bitmap: vsi->tc_cfg.ena_tc, max_lanqs: max_txqs);
3463
3464	if (ret) {
3465	dev_err(dev, "VSI %d failed TC config, error %d\n",
3466	vsi->vsi_num, ret);
3467	goto out;
3468	}
3469	ice_vsi_update_q_map(vsi, ctx);
3470	vsi->info.valid_sections = 0;
3471
3472	ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3473	out:
3474	kfree(objp: ctx);
3475	return ret;
3476	}
3477
3478	/**
3479	* ice_update_ring_stats - Update ring statistics
3480	* @stats: stats to be updated
3481	* @pkts: number of processed packets
3482	* @bytes: number of processed bytes
3483	*
3484	* This function assumes that caller has acquired a u64_stats_sync lock.
3485	*/
3486	static void ice_update_ring_stats(struct ice_q_stats *stats, u64 pkts, u64 bytes)
3487	{
3488	stats->bytes += bytes;
3489	stats->pkts += pkts;
3490	}
3491
3492	/**
3493	* ice_update_tx_ring_stats - Update Tx ring specific counters
3494	* @tx_ring: ring to update
3495	* @pkts: number of processed packets
3496	* @bytes: number of processed bytes
3497	*/
3498	void ice_update_tx_ring_stats(struct ice_tx_ring *tx_ring, u64 pkts, u64 bytes)
3499	{
3500	u64_stats_update_begin(syncp: &tx_ring->ring_stats->syncp);
3501	ice_update_ring_stats(stats: &tx_ring->ring_stats->stats, pkts, bytes);
3502	u64_stats_update_end(syncp: &tx_ring->ring_stats->syncp);
3503	}
3504
3505	/**
3506	* ice_update_rx_ring_stats - Update Rx ring specific counters
3507	* @rx_ring: ring to update
3508	* @pkts: number of processed packets
3509	* @bytes: number of processed bytes
3510	*/
3511	void ice_update_rx_ring_stats(struct ice_rx_ring *rx_ring, u64 pkts, u64 bytes)
3512	{
3513	u64_stats_update_begin(syncp: &rx_ring->ring_stats->syncp);
3514	ice_update_ring_stats(stats: &rx_ring->ring_stats->stats, pkts, bytes);
3515	u64_stats_update_end(syncp: &rx_ring->ring_stats->syncp);
3516	}
3517
3518	/**
3519	* ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3520	* @pi: port info of the switch with default VSI
3521	*
3522	* Return true if the there is a single VSI in default forwarding VSI list
3523	*/
3524	bool ice_is_dflt_vsi_in_use(struct ice_port_info *pi)
3525	{
3526	bool exists = false;
3527
3528	ice_check_if_dflt_vsi(pi, vsi_handle: 0, rule_exists: &exists);
3529	return exists;
3530	}
3531
3532	/**
3533	* ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3534	* @vsi: VSI to compare against default forwarding VSI
3535	*
3536	* If this VSI passed in is the default forwarding VSI then return true, else
3537	* return false
3538	*/
3539	bool ice_is_vsi_dflt_vsi(struct ice_vsi *vsi)
3540	{
3541	return ice_check_if_dflt_vsi(pi: vsi->port_info, vsi_handle: vsi->idx, NULL);
3542	}
3543
3544	/**
3545	* ice_set_dflt_vsi - set the default forwarding VSI
3546	* @vsi: VSI getting set as the default forwarding VSI on the switch
3547	*
3548	* If the VSI passed in is already the default VSI and it's enabled just return
3549	* success.
3550	*
3551	* Otherwise try to set the VSI passed in as the switch's default VSI and
3552	* return the result.
3553	*/
3554	int ice_set_dflt_vsi(struct ice_vsi *vsi)
3555	{
3556	struct device *dev;
3557	int status;
3558
3559	if (!vsi)
3560	return -EINVAL;
3561
3562	dev = ice_pf_to_dev(vsi->back);
3563
3564	if (ice_lag_is_switchdev_running(pf: vsi->back)) {
3565	dev_dbg(dev, "VSI %d passed is a part of LAG containing interfaces in switchdev mode, nothing to do\n",
3566	vsi->vsi_num);
3567	return 0;
3568	}
3569
3570	/* the VSI passed in is already the default VSI */
3571	if (ice_is_vsi_dflt_vsi(vsi)) {
3572	dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3573	vsi->vsi_num);
3574	return 0;
3575	}
3576
3577	status = ice_cfg_dflt_vsi(pi: vsi->port_info, vsi_handle: vsi->idx, set: true, ICE_FLTR_RX);
3578	if (status) {
3579	dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %d\n",
3580	vsi->vsi_num, status);
3581	return status;
3582	}
3583
3584	return 0;
3585	}
3586
3587	/**
3588	* ice_clear_dflt_vsi - clear the default forwarding VSI
3589	* @vsi: VSI to remove from filter list
3590	*
3591	* If the switch has no default VSI or it's not enabled then return error.
3592	*
3593	* Otherwise try to clear the default VSI and return the result.
3594	*/
3595	int ice_clear_dflt_vsi(struct ice_vsi *vsi)
3596	{
3597	struct device *dev;
3598	int status;
3599
3600	if (!vsi)
3601	return -EINVAL;
3602
3603	dev = ice_pf_to_dev(vsi->back);
3604
3605	/* there is no default VSI configured */
3606	if (!ice_is_dflt_vsi_in_use(pi: vsi->port_info))
3607	return -ENODEV;
3608
3609	status = ice_cfg_dflt_vsi(pi: vsi->port_info, vsi_handle: vsi->idx, set: false,
3610	ICE_FLTR_RX);
3611	if (status) {
3612	dev_err(dev, "Failed to clear the default forwarding VSI %d, error %d\n",
3613	vsi->vsi_num, status);
3614	return -EIO;
3615	}
3616
3617	return 0;
3618	}
3619
3620	/**
3621	* ice_get_link_speed_mbps - get link speed in Mbps
3622	* @vsi: the VSI whose link speed is being queried
3623	*
3624	* Return current VSI link speed and 0 if the speed is unknown.
3625	*/
3626	int ice_get_link_speed_mbps(struct ice_vsi *vsi)
3627	{
3628	unsigned int link_speed;
3629
3630	link_speed = vsi->port_info->phy.link_info.link_speed;
3631
3632	return (int)ice_get_link_speed(index: fls(x: link_speed) - 1);
3633	}
3634
3635	/**
3636	* ice_get_link_speed_kbps - get link speed in Kbps
3637	* @vsi: the VSI whose link speed is being queried
3638	*
3639	* Return current VSI link speed and 0 if the speed is unknown.
3640	*/
3641	int ice_get_link_speed_kbps(struct ice_vsi *vsi)
3642	{
3643	int speed_mbps;
3644
3645	speed_mbps = ice_get_link_speed_mbps(vsi);
3646
3647	return speed_mbps * 1000;
3648	}
3649
3650	/**
3651	* ice_set_min_bw_limit - setup minimum BW limit for Tx based on min_tx_rate
3652	* @vsi: VSI to be configured
3653	* @min_tx_rate: min Tx rate in Kbps to be configured as BW limit
3654	*
3655	* If the min_tx_rate is specified as 0 that means to clear the minimum BW limit
3656	* profile, otherwise a non-zero value will force a minimum BW limit for the VSI
3657	* on TC 0.
3658	*/
3659	int ice_set_min_bw_limit(struct ice_vsi *vsi, u64 min_tx_rate)
3660	{
3661	struct ice_pf *pf = vsi->back;
3662	struct device *dev;
3663	int status;
3664	int speed;
3665
3666	dev = ice_pf_to_dev(pf);
3667	if (!vsi->port_info) {
3668	dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3669	vsi->idx, vsi->type);
3670	return -EINVAL;
3671	}
3672
3673	speed = ice_get_link_speed_kbps(vsi);
3674	if (min_tx_rate > (u64)speed) {
3675	dev_err(dev, "invalid min Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3676	min_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3677	speed);
3678	return -EINVAL;
3679	}
3680
3681	/* Configure min BW for VSI limit */
3682	if (min_tx_rate) {
3683	status = ice_cfg_vsi_bw_lmt_per_tc(pi: vsi->port_info, vsi_handle: vsi->idx, tc: 0,
3684	rl_type: ICE_MIN_BW, bw: min_tx_rate);
3685	if (status) {
3686	dev_err(dev, "failed to set min Tx rate(%llu Kbps) for %s %d\n",
3687	min_tx_rate, ice_vsi_type_str(vsi->type),
3688	vsi->idx);
3689	return status;
3690	}
3691
3692	dev_dbg(dev, "set min Tx rate(%llu Kbps) for %s\n",
3693	min_tx_rate, ice_vsi_type_str(vsi->type));
3694	} else {
3695	status = ice_cfg_vsi_bw_dflt_lmt_per_tc(pi: vsi->port_info,
3696	vsi_handle: vsi->idx, tc: 0,
3697	rl_type: ICE_MIN_BW);
3698	if (status) {
3699	dev_err(dev, "failed to clear min Tx rate configuration for %s %d\n",
3700	ice_vsi_type_str(vsi->type), vsi->idx);
3701	return status;
3702	}
3703
3704	dev_dbg(dev, "cleared min Tx rate configuration for %s %d\n",
3705	ice_vsi_type_str(vsi->type), vsi->idx);
3706	}
3707
3708	return 0;
3709	}
3710
3711	/**
3712	* ice_set_max_bw_limit - setup maximum BW limit for Tx based on max_tx_rate
3713	* @vsi: VSI to be configured
3714	* @max_tx_rate: max Tx rate in Kbps to be configured as BW limit
3715	*
3716	* If the max_tx_rate is specified as 0 that means to clear the maximum BW limit
3717	* profile, otherwise a non-zero value will force a maximum BW limit for the VSI
3718	* on TC 0.
3719	*/
3720	int ice_set_max_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate)
3721	{
3722	struct ice_pf *pf = vsi->back;
3723	struct device *dev;
3724	int status;
3725	int speed;
3726
3727	dev = ice_pf_to_dev(pf);
3728	if (!vsi->port_info) {
3729	dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3730	vsi->idx, vsi->type);
3731	return -EINVAL;
3732	}
3733
3734	speed = ice_get_link_speed_kbps(vsi);
3735	if (max_tx_rate > (u64)speed) {
3736	dev_err(dev, "invalid max Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3737	max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3738	speed);
3739	return -EINVAL;
3740	}
3741
3742	/* Configure max BW for VSI limit */
3743	if (max_tx_rate) {
3744	status = ice_cfg_vsi_bw_lmt_per_tc(pi: vsi->port_info, vsi_handle: vsi->idx, tc: 0,
3745	rl_type: ICE_MAX_BW, bw: max_tx_rate);
3746	if (status) {
3747	dev_err(dev, "failed setting max Tx rate(%llu Kbps) for %s %d\n",
3748	max_tx_rate, ice_vsi_type_str(vsi->type),
3749	vsi->idx);
3750	return status;
3751	}
3752
3753	dev_dbg(dev, "set max Tx rate(%llu Kbps) for %s %d\n",
3754	max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx);
3755	} else {
3756	status = ice_cfg_vsi_bw_dflt_lmt_per_tc(pi: vsi->port_info,
3757	vsi_handle: vsi->idx, tc: 0,
3758	rl_type: ICE_MAX_BW);
3759	if (status) {
3760	dev_err(dev, "failed clearing max Tx rate configuration for %s %d\n",
3761	ice_vsi_type_str(vsi->type), vsi->idx);
3762	return status;
3763	}
3764
3765	dev_dbg(dev, "cleared max Tx rate configuration for %s %d\n",
3766	ice_vsi_type_str(vsi->type), vsi->idx);
3767	}
3768
3769	return 0;
3770	}
3771
3772	/**
3773	* ice_set_link - turn on/off physical link
3774	* @vsi: VSI to modify physical link on
3775	* @ena: turn on/off physical link
3776	*/
3777	int ice_set_link(struct ice_vsi *vsi, bool ena)
3778	{
3779	struct device *dev = ice_pf_to_dev(vsi->back);
3780	struct ice_port_info *pi = vsi->port_info;
3781	struct ice_hw *hw = pi->hw;
3782	int status;
3783
3784	if (vsi->type != ICE_VSI_PF)
3785	return -EINVAL;
3786
3787	status = ice_aq_set_link_restart_an(pi, ena_link: ena, NULL);
3788
3789	/* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
3790	* this is not a fatal error, so print a warning message and return
3791	* a success code. Return an error if FW returns an error code other
3792	* than ICE_AQ_RC_EMODE
3793	*/
3794	if (status == -EIO) {
3795	if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
3796	dev_dbg(dev, "can't set link to %s, err %d aq_err %s. not fatal, continuing\n",
3797	(ena ? "ON" : "OFF"), status,
3798	ice_aq_str(hw->adminq.sq_last_status));
3799	} else if (status) {
3800	dev_err(dev, "can't set link to %s, err %d aq_err %s\n",
3801	(ena ? "ON" : "OFF"), status,
3802	ice_aq_str(hw->adminq.sq_last_status));
3803	return status;
3804	}
3805
3806	return 0;
3807	}
3808
3809	/**
3810	* ice_vsi_add_vlan_zero - add VLAN 0 filter(s) for this VSI
3811	* @vsi: VSI used to add VLAN filters
3812	*
3813	* In Single VLAN Mode (SVM), single VLAN filters via ICE_SW_LKUP_VLAN are based
3814	* on the inner VLAN ID, so the VLAN TPID (i.e. 0x8100 or 0x888a8) doesn't
3815	* matter. In Double VLAN Mode (DVM), outer/single VLAN filters via
3816	* ICE_SW_LKUP_VLAN are based on the outer/single VLAN ID + VLAN TPID.
3817	*
3818	* For both modes add a VLAN 0 + no VLAN TPID filter to handle untagged traffic
3819	* when VLAN pruning is enabled. Also, this handles VLAN 0 priority tagged
3820	* traffic in SVM, since the VLAN TPID isn't part of filtering.
3821	*
3822	* If DVM is enabled then an explicit VLAN 0 + VLAN TPID filter needs to be
3823	* added to allow VLAN 0 priority tagged traffic in DVM, since the VLAN TPID is
3824	* part of filtering.
3825	*/
3826	int ice_vsi_add_vlan_zero(struct ice_vsi *vsi)
3827	{
3828	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3829	struct ice_vlan vlan;
3830	int err;
3831
3832	vlan = ICE_VLAN(0, 0, 0);
3833	err = vlan_ops->add_vlan(vsi, &vlan);
3834	if (err && err != -EEXIST)
3835	return err;
3836
3837	/* in SVM both VLAN 0 filters are identical */
3838	if (!ice_is_dvm_ena(hw: &vsi->back->hw))
3839	return 0;
3840
3841	vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3842	err = vlan_ops->add_vlan(vsi, &vlan);
3843	if (err && err != -EEXIST)
3844	return err;
3845
3846	return 0;
3847	}
3848
3849	/**
3850	* ice_vsi_del_vlan_zero - delete VLAN 0 filter(s) for this VSI
3851	* @vsi: VSI used to add VLAN filters
3852	*
3853	* Delete the VLAN 0 filters in the same manner that they were added in
3854	* ice_vsi_add_vlan_zero.
3855	*/
3856	int ice_vsi_del_vlan_zero(struct ice_vsi *vsi)
3857	{
3858	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3859	struct ice_vlan vlan;
3860	int err;
3861
3862	vlan = ICE_VLAN(0, 0, 0);
3863	err = vlan_ops->del_vlan(vsi, &vlan);
3864	if (err && err != -EEXIST)
3865	return err;
3866
3867	/* in SVM both VLAN 0 filters are identical */
3868	if (!ice_is_dvm_ena(hw: &vsi->back->hw))
3869	return 0;
3870
3871	vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
3872	err = vlan_ops->del_vlan(vsi, &vlan);
3873	if (err && err != -EEXIST)
3874	return err;
3875
3876	/* when deleting the last VLAN filter, make sure to disable the VLAN
3877	* promisc mode so the filter isn't left by accident
3878	*/
3879	return ice_clear_vsi_promisc(hw: &vsi->back->hw, vsi_handle: vsi->idx,
3880	ICE_MCAST_VLAN_PROMISC_BITS, vid: 0);
3881	}
3882
3883	/**
3884	* ice_vsi_num_zero_vlans - get number of VLAN 0 filters based on VLAN mode
3885	* @vsi: VSI used to get the VLAN mode
3886	*
3887	* If DVM is enabled then 2 VLAN 0 filters are added, else if SVM is enabled
3888	* then 1 VLAN 0 filter is added. See ice_vsi_add_vlan_zero for more details.
3889	*/
3890	static u16 ice_vsi_num_zero_vlans(struct ice_vsi *vsi)
3891	{
3892	#define ICE_DVM_NUM_ZERO_VLAN_FLTRS 2
3893	#define ICE_SVM_NUM_ZERO_VLAN_FLTRS 1
3894	/* no VLAN 0 filter is created when a port VLAN is active */
3895	if (vsi->type == ICE_VSI_VF) {
3896	if (WARN_ON(!vsi->vf))
3897	return 0;
3898
3899	if (ice_vf_is_port_vlan_ena(vf: vsi->vf))
3900	return 0;
3901	}
3902
3903	if (ice_is_dvm_ena(hw: &vsi->back->hw))
3904	return ICE_DVM_NUM_ZERO_VLAN_FLTRS;
3905	else
3906	return ICE_SVM_NUM_ZERO_VLAN_FLTRS;
3907	}
3908
3909	/**
3910	* ice_vsi_has_non_zero_vlans - check if VSI has any non-zero VLANs
3911	* @vsi: VSI used to determine if any non-zero VLANs have been added
3912	*/
3913	bool ice_vsi_has_non_zero_vlans(struct ice_vsi *vsi)
3914	{
3915	return (vsi->num_vlan > ice_vsi_num_zero_vlans(vsi));
3916	}
3917
3918	/**
3919	* ice_vsi_num_non_zero_vlans - get the number of non-zero VLANs for this VSI
3920	* @vsi: VSI used to get the number of non-zero VLANs added
3921	*/
3922	u16 ice_vsi_num_non_zero_vlans(struct ice_vsi *vsi)
3923	{
3924	return (vsi->num_vlan - ice_vsi_num_zero_vlans(vsi));
3925	}
3926
3927	/**
3928	* ice_is_feature_supported
3929	* @pf: pointer to the struct ice_pf instance
3930	* @f: feature enum to be checked
3931	*
3932	* returns true if feature is supported, false otherwise
3933	*/
3934	bool ice_is_feature_supported(struct ice_pf *pf, enum ice_feature f)
3935	{
3936	if (f < 0 || f >= ICE_F_MAX)
3937	return false;
3938
3939	return test_bit(f, pf->features);
3940	}
3941
3942	/**
3943	* ice_set_feature_support
3944	* @pf: pointer to the struct ice_pf instance
3945	* @f: feature enum to set
3946	*/
3947	void ice_set_feature_support(struct ice_pf *pf, enum ice_feature f)
3948	{
3949	if (f < 0 || f >= ICE_F_MAX)
3950	return;
3951
3952	set_bit(nr: f, addr: pf->features);
3953	}
3954
3955	/**
3956	* ice_clear_feature_support
3957	* @pf: pointer to the struct ice_pf instance
3958	* @f: feature enum to clear
3959	*/
3960	void ice_clear_feature_support(struct ice_pf *pf, enum ice_feature f)
3961	{
3962	if (f < 0 || f >= ICE_F_MAX)
3963	return;
3964
3965	clear_bit(nr: f, addr: pf->features);
3966	}
3967
3968	/**
3969	* ice_init_feature_support
3970	* @pf: pointer to the struct ice_pf instance
3971	*
3972	* called during init to setup supported feature
3973	*/
3974	void ice_init_feature_support(struct ice_pf *pf)
3975	{
3976	switch (pf->hw.device_id) {
3977	case ICE_DEV_ID_E810C_BACKPLANE:
3978	case ICE_DEV_ID_E810C_QSFP:
3979	case ICE_DEV_ID_E810C_SFP:
3980	case ICE_DEV_ID_E810_XXV_BACKPLANE:
3981	case ICE_DEV_ID_E810_XXV_QSFP:
3982	case ICE_DEV_ID_E810_XXV_SFP:
3983	ice_set_feature_support(pf, f: ICE_F_DSCP);
3984	if (ice_is_phy_rclk_in_netlist(hw: &pf->hw))
3985	ice_set_feature_support(pf, f: ICE_F_PHY_RCLK);
3986	/* If we don't own the timer - don't enable other caps */
3987	if (!ice_pf_src_tmr_owned(pf))
3988	break;
3989	if (ice_is_cgu_in_netlist(hw: &pf->hw))
3990	ice_set_feature_support(pf, f: ICE_F_CGU);
3991	if (ice_is_clock_mux_in_netlist(hw: &pf->hw))
3992	ice_set_feature_support(pf, f: ICE_F_SMA_CTRL);
3993	if (ice_gnss_is_gps_present(hw: &pf->hw))
3994	ice_set_feature_support(pf, f: ICE_F_GNSS);
3995	break;
3996	default:
3997	break;
3998	}
3999	}
4000
4001	/**
4002	* ice_vsi_update_security - update security block in VSI
4003	* @vsi: pointer to VSI structure
4004	* @fill: function pointer to fill ctx
4005	*/
4006	int
4007	ice_vsi_update_security(struct ice_vsi *vsi, void (*fill)(struct ice_vsi_ctx *))
4008	{
4009	struct ice_vsi_ctx ctx = { 0 };
4010
4011	ctx.info = vsi->info;
4012	ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
4013	fill(&ctx);
4014
4015	if (ice_update_vsi(hw: &vsi->back->hw, vsi_handle: vsi->idx, vsi_ctx: &ctx, NULL))
4016	return -ENODEV;
4017
4018	vsi->info = ctx.info;
4019	return 0;
4020	}
4021
4022	/**
4023	* ice_vsi_ctx_set_antispoof - set antispoof function in VSI ctx
4024	* @ctx: pointer to VSI ctx structure
4025	*/
4026	void ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx *ctx)
4027	{
4028	ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
4029	(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4030	ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4031	}
4032
4033	/**
4034	* ice_vsi_ctx_clear_antispoof - clear antispoof function in VSI ctx
4035	* @ctx: pointer to VSI ctx structure
4036	*/
4037	void ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx *ctx)
4038	{
4039	ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF &
4040	~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4041	ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4042	}
4043
4044	/**
4045	* ice_vsi_ctx_set_allow_override - allow destination override on VSI
4046	* @ctx: pointer to VSI ctx structure
4047	*/
4048	void ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx *ctx)
4049	{
4050	ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4051	}
4052
4053	/**
4054	* ice_vsi_ctx_clear_allow_override - turn off destination override on VSI
4055	* @ctx: pointer to VSI ctx structure
4056	*/
4057	void ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx *ctx)
4058	{
4059	ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4060	}
4061
4062	/**
4063	* ice_vsi_update_local_lb - update sw block in VSI with local loopback bit
4064	* @vsi: pointer to VSI structure
4065	* @set: set or unset the bit
4066	*/
4067	int
4068	ice_vsi_update_local_lb(struct ice_vsi *vsi, bool set)
4069	{
4070	struct ice_vsi_ctx ctx = {
4071	.info = vsi->info,
4072	};
4073
4074	ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
4075	if (set)
4076	ctx.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4077	else
4078	ctx.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_LOCAL_LB;
4079
4080	if (ice_update_vsi(hw: &vsi->back->hw, vsi_handle: vsi->idx, vsi_ctx: &ctx, NULL))
4081	return -ENODEV;
4082
4083	vsi->info = ctx.info;
4084	return 0;
4085	}
4086