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 */
ice_vsi_type_str(enum ice_vsi_type vsi_type)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 */
ice_vsi_ctrl_all_rx_rings(struct ice_vsi * vsi,bool ena)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, i, false);
54
55 ice_flush(&vsi->back->hw);
56
57 ice_for_each_rxq(vsi, i) {
58 ret = ice_vsi_wait_one_rx_ring(vsi, ena, i);
59 if (ret)
60 break;
61 }
62
63 return ret;
64 }
65
66 /**
67 * ice_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the VSI
68 * @vsi: VSI pointer
69 *
70 * On error: returns error code (negative)
71 * On success: returns 0
72 */
ice_vsi_alloc_arrays(struct ice_vsi * vsi)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, vsi->alloc_txq,
84 sizeof(*vsi->tx_rings), GFP_KERNEL);
85 if (!vsi->tx_rings)
86 return -ENOMEM;
87
88 vsi->rx_rings = devm_kcalloc(dev, vsi->alloc_rxq,
89 sizeof(*vsi->rx_rings), GFP_KERNEL);
90 if (!vsi->rx_rings)
91 goto err_rings;
92
93 /* txq_map needs to have enough space to track both Tx (stack) rings
94 * and XDP rings; at this point vsi->num_xdp_txq might not be set,
95 * so use num_possible_cpus() as we want to always provide XDP ring
96 * per CPU, regardless of queue count settings from user that might
97 * have come from ethtool's set_channels() callback;
98 */
99 vsi->txq_map = devm_kcalloc(dev, (vsi->alloc_txq + num_possible_cpus()),
100 sizeof(*vsi->txq_map), GFP_KERNEL);
101
102 if (!vsi->txq_map)
103 goto err_txq_map;
104
105 vsi->rxq_map = devm_kcalloc(dev, vsi->alloc_rxq,
106 sizeof(*vsi->rxq_map), GFP_KERNEL);
107 if (!vsi->rxq_map)
108 goto err_rxq_map;
109
110 /* There is no need to allocate q_vectors for a loopback VSI. */
111 if (vsi->type == ICE_VSI_LB)
112 return 0;
113
114 /* allocate memory for q_vector pointers */
115 vsi->q_vectors = devm_kcalloc(dev, vsi->num_q_vectors,
116 sizeof(*vsi->q_vectors), GFP_KERNEL);
117 if (!vsi->q_vectors)
118 goto err_vectors;
119
120 vsi->af_xdp_zc_qps = bitmap_zalloc(max_t(int, vsi->alloc_txq, vsi->alloc_rxq), GFP_KERNEL);
121 if (!vsi->af_xdp_zc_qps)
122 goto err_zc_qps;
123
124 return 0;
125
126 err_zc_qps:
127 devm_kfree(dev, vsi->q_vectors);
128 err_vectors:
129 devm_kfree(dev, vsi->rxq_map);
130 err_rxq_map:
131 devm_kfree(dev, vsi->txq_map);
132 err_txq_map:
133 devm_kfree(dev, vsi->rx_rings);
134 err_rings:
135 devm_kfree(dev, vsi->tx_rings);
136 return -ENOMEM;
137 }
138
139 /**
140 * ice_vsi_set_num_desc - Set number of descriptors for queues on this VSI
141 * @vsi: the VSI being configured
142 */
ice_vsi_set_num_desc(struct ice_vsi * vsi)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 * @vf: the VF associated with this VSI, if any
170 *
171 * Return 0 on success and a negative value on error
172 */
ice_vsi_set_num_qs(struct ice_vsi * vsi,struct ice_vf * vf)173 static void ice_vsi_set_num_qs(struct ice_vsi *vsi, struct ice_vf *vf)
174 {
175 enum ice_vsi_type vsi_type = vsi->type;
176 struct ice_pf *pf = vsi->back;
177
178 if (WARN_ON(vsi_type == ICE_VSI_VF && !vf))
179 return;
180
181 switch (vsi_type) {
182 case ICE_VSI_PF:
183 if (vsi->req_txq) {
184 vsi->alloc_txq = vsi->req_txq;
185 vsi->num_txq = vsi->req_txq;
186 } else {
187 vsi->alloc_txq = min3(pf->num_lan_msix,
188 ice_get_avail_txq_count(pf),
189 (u16)num_online_cpus());
190 }
191
192 pf->num_lan_tx = vsi->alloc_txq;
193
194 /* only 1 Rx queue unless RSS is enabled */
195 if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
196 vsi->alloc_rxq = 1;
197 } else {
198 if (vsi->req_rxq) {
199 vsi->alloc_rxq = vsi->req_rxq;
200 vsi->num_rxq = vsi->req_rxq;
201 } else {
202 vsi->alloc_rxq = min3(pf->num_lan_msix,
203 ice_get_avail_rxq_count(pf),
204 (u16)num_online_cpus());
205 }
206 }
207
208 pf->num_lan_rx = vsi->alloc_rxq;
209
210 vsi->num_q_vectors = min_t(int, pf->num_lan_msix,
211 max_t(int, vsi->alloc_rxq,
212 vsi->alloc_txq));
213 break;
214 case ICE_VSI_SWITCHDEV_CTRL:
215 /* The number of queues for ctrl VSI is equal to number of VFs.
216 * Each ring is associated to the corresponding VF_PR netdev.
217 */
218 vsi->alloc_txq = ice_get_num_vfs(pf);
219 vsi->alloc_rxq = vsi->alloc_txq;
220 vsi->num_q_vectors = 1;
221 break;
222 case ICE_VSI_VF:
223 if (vf->num_req_qs)
224 vf->num_vf_qs = vf->num_req_qs;
225 vsi->alloc_txq = vf->num_vf_qs;
226 vsi->alloc_rxq = vf->num_vf_qs;
227 /* pf->vfs.num_msix_per includes (VF miscellaneous vector +
228 * data queue interrupts). Since vsi->num_q_vectors is number
229 * of queues vectors, subtract 1 (ICE_NONQ_VECS_VF) from the
230 * original vector count
231 */
232 vsi->num_q_vectors = pf->vfs.num_msix_per - ICE_NONQ_VECS_VF;
233 break;
234 case ICE_VSI_CTRL:
235 vsi->alloc_txq = 1;
236 vsi->alloc_rxq = 1;
237 vsi->num_q_vectors = 1;
238 break;
239 case ICE_VSI_CHNL:
240 vsi->alloc_txq = 0;
241 vsi->alloc_rxq = 0;
242 break;
243 case ICE_VSI_LB:
244 vsi->alloc_txq = 1;
245 vsi->alloc_rxq = 1;
246 break;
247 default:
248 dev_warn(ice_pf_to_dev(pf), "Unknown VSI type %d\n", vsi_type);
249 break;
250 }
251
252 ice_vsi_set_num_desc(vsi);
253 }
254
255 /**
256 * ice_get_free_slot - get the next non-NULL location index in array
257 * @array: array to search
258 * @size: size of the array
259 * @curr: last known occupied index to be used as a search hint
260 *
261 * void * is being used to keep the functionality generic. This lets us use this
262 * function on any array of pointers.
263 */
ice_get_free_slot(void * array,int size,int curr)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 - delete a VSI from the switch
286 * @vsi: pointer to VSI being removed
287 */
ice_vsi_delete(struct ice_vsi * vsi)288 void ice_vsi_delete(struct ice_vsi *vsi)
289 {
290 struct ice_pf *pf = vsi->back;
291 struct ice_vsi_ctx *ctxt;
292 int status;
293
294 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
295 if (!ctxt)
296 return;
297
298 if (vsi->type == ICE_VSI_VF)
299 ctxt->vf_num = vsi->vf->vf_id;
300 ctxt->vsi_num = vsi->vsi_num;
301
302 memcpy(&ctxt->info, &vsi->info, sizeof(ctxt->info));
303
304 status = ice_free_vsi(&pf->hw, vsi->idx, ctxt, false, NULL);
305 if (status)
306 dev_err(ice_pf_to_dev(pf), "Failed to delete VSI %i in FW - error: %d\n",
307 vsi->vsi_num, status);
308
309 kfree(ctxt);
310 }
311
312 /**
313 * ice_vsi_free_arrays - De-allocate queue and vector pointer arrays for the VSI
314 * @vsi: pointer to VSI being cleared
315 */
ice_vsi_free_arrays(struct ice_vsi * vsi)316 static void ice_vsi_free_arrays(struct ice_vsi *vsi)
317 {
318 struct ice_pf *pf = vsi->back;
319 struct device *dev;
320
321 dev = ice_pf_to_dev(pf);
322
323 if (vsi->af_xdp_zc_qps) {
324 bitmap_free(vsi->af_xdp_zc_qps);
325 vsi->af_xdp_zc_qps = NULL;
326 }
327 /* free the ring and vector containers */
328 if (vsi->q_vectors) {
329 devm_kfree(dev, vsi->q_vectors);
330 vsi->q_vectors = NULL;
331 }
332 if (vsi->tx_rings) {
333 devm_kfree(dev, vsi->tx_rings);
334 vsi->tx_rings = NULL;
335 }
336 if (vsi->rx_rings) {
337 devm_kfree(dev, vsi->rx_rings);
338 vsi->rx_rings = NULL;
339 }
340 if (vsi->txq_map) {
341 devm_kfree(dev, vsi->txq_map);
342 vsi->txq_map = NULL;
343 }
344 if (vsi->rxq_map) {
345 devm_kfree(dev, vsi->rxq_map);
346 vsi->rxq_map = NULL;
347 }
348 }
349
350 /**
351 * ice_vsi_clear - clean up and deallocate the provided VSI
352 * @vsi: pointer to VSI being cleared
353 *
354 * This deallocates the VSI's queue resources, removes it from the PF's
355 * VSI array if necessary, and deallocates the VSI
356 *
357 * Returns 0 on success, negative on failure
358 */
ice_vsi_clear(struct ice_vsi * vsi)359 int ice_vsi_clear(struct ice_vsi *vsi)
360 {
361 struct ice_pf *pf = NULL;
362 struct device *dev;
363
364 if (!vsi)
365 return 0;
366
367 if (!vsi->back)
368 return -EINVAL;
369
370 pf = vsi->back;
371 dev = ice_pf_to_dev(pf);
372
373 if (!pf->vsi[vsi->idx] || pf->vsi[vsi->idx] != vsi) {
374 dev_dbg(dev, "vsi does not exist at pf->vsi[%d]\n", vsi->idx);
375 return -EINVAL;
376 }
377
378 mutex_lock(&pf->sw_mutex);
379 /* updates the PF for this cleared VSI */
380
381 pf->vsi[vsi->idx] = NULL;
382 if (vsi->idx < pf->next_vsi && vsi->type != ICE_VSI_CTRL)
383 pf->next_vsi = vsi->idx;
384 if (vsi->idx < pf->next_vsi && vsi->type == ICE_VSI_CTRL && vsi->vf)
385 pf->next_vsi = vsi->idx;
386
387 ice_vsi_free_arrays(vsi);
388 mutex_unlock(&pf->sw_mutex);
389 devm_kfree(dev, vsi);
390
391 return 0;
392 }
393
394 /**
395 * ice_msix_clean_ctrl_vsi - MSIX mode interrupt handler for ctrl VSI
396 * @irq: interrupt number
397 * @data: pointer to a q_vector
398 */
ice_msix_clean_ctrl_vsi(int __always_unused irq,void * data)399 static irqreturn_t ice_msix_clean_ctrl_vsi(int __always_unused irq, void *data)
400 {
401 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
402
403 if (!q_vector->tx.tx_ring)
404 return IRQ_HANDLED;
405
406 #define FDIR_RX_DESC_CLEAN_BUDGET 64
407 ice_clean_rx_irq(q_vector->rx.rx_ring, FDIR_RX_DESC_CLEAN_BUDGET);
408 ice_clean_ctrl_tx_irq(q_vector->tx.tx_ring);
409
410 return IRQ_HANDLED;
411 }
412
413 /**
414 * ice_msix_clean_rings - MSIX mode Interrupt Handler
415 * @irq: interrupt number
416 * @data: pointer to a q_vector
417 */
ice_msix_clean_rings(int __always_unused irq,void * data)418 static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
419 {
420 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
421
422 if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
423 return IRQ_HANDLED;
424
425 q_vector->total_events++;
426
427 napi_schedule(&q_vector->napi);
428
429 return IRQ_HANDLED;
430 }
431
ice_eswitch_msix_clean_rings(int __always_unused irq,void * data)432 static irqreturn_t ice_eswitch_msix_clean_rings(int __always_unused irq, void *data)
433 {
434 struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
435 struct ice_pf *pf = q_vector->vsi->back;
436 struct ice_vf *vf;
437 unsigned int bkt;
438
439 if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
440 return IRQ_HANDLED;
441
442 rcu_read_lock();
443 ice_for_each_vf_rcu(pf, bkt, vf)
444 napi_schedule(&vf->repr->q_vector->napi);
445 rcu_read_unlock();
446
447 return IRQ_HANDLED;
448 }
449
450 /**
451 * ice_vsi_alloc - Allocates the next available struct VSI in the PF
452 * @pf: board private structure
453 * @vsi_type: type of VSI
454 * @ch: ptr to channel
455 * @vf: VF for ICE_VSI_VF and ICE_VSI_CTRL
456 *
457 * The VF pointer is used for ICE_VSI_VF and ICE_VSI_CTRL. For ICE_VSI_CTRL,
458 * it may be NULL in the case there is no association with a VF. For
459 * ICE_VSI_VF the VF pointer *must not* be NULL.
460 *
461 * returns a pointer to a VSI on success, NULL on failure.
462 */
463 static struct ice_vsi *
ice_vsi_alloc(struct ice_pf * pf,enum ice_vsi_type vsi_type,struct ice_channel * ch,struct ice_vf * vf)464 ice_vsi_alloc(struct ice_pf *pf, enum ice_vsi_type vsi_type,
465 struct ice_channel *ch, struct ice_vf *vf)
466 {
467 struct device *dev = ice_pf_to_dev(pf);
468 struct ice_vsi *vsi = NULL;
469
470 if (WARN_ON(vsi_type == ICE_VSI_VF && !vf))
471 return NULL;
472
473 /* Need to protect the allocation of the VSIs at the PF level */
474 mutex_lock(&pf->sw_mutex);
475
476 /* If we have already allocated our maximum number of VSIs,
477 * pf->next_vsi will be ICE_NO_VSI. If not, pf->next_vsi index
478 * is available to be populated
479 */
480 if (pf->next_vsi == ICE_NO_VSI) {
481 dev_dbg(dev, "out of VSI slots!\n");
482 goto unlock_pf;
483 }
484
485 vsi = devm_kzalloc(dev, sizeof(*vsi), GFP_KERNEL);
486 if (!vsi)
487 goto unlock_pf;
488
489 vsi->type = vsi_type;
490 vsi->back = pf;
491 set_bit(ICE_VSI_DOWN, vsi->state);
492
493 if (vsi_type == ICE_VSI_VF)
494 ice_vsi_set_num_qs(vsi, vf);
495 else if (vsi_type != ICE_VSI_CHNL)
496 ice_vsi_set_num_qs(vsi, NULL);
497
498 switch (vsi->type) {
499 case ICE_VSI_SWITCHDEV_CTRL:
500 if (ice_vsi_alloc_arrays(vsi))
501 goto err_rings;
502
503 /* Setup eswitch MSIX irq handler for VSI */
504 vsi->irq_handler = ice_eswitch_msix_clean_rings;
505 break;
506 case ICE_VSI_PF:
507 if (ice_vsi_alloc_arrays(vsi))
508 goto err_rings;
509
510 /* Setup default MSIX irq handler for VSI */
511 vsi->irq_handler = ice_msix_clean_rings;
512 break;
513 case ICE_VSI_CTRL:
514 if (ice_vsi_alloc_arrays(vsi))
515 goto err_rings;
516
517 /* Setup ctrl VSI MSIX irq handler */
518 vsi->irq_handler = ice_msix_clean_ctrl_vsi;
519
520 /* For the PF control VSI this is NULL, for the VF control VSI
521 * this will be the first VF to allocate it.
522 */
523 vsi->vf = vf;
524 break;
525 case ICE_VSI_VF:
526 if (ice_vsi_alloc_arrays(vsi))
527 goto err_rings;
528 vsi->vf = vf;
529 break;
530 case ICE_VSI_CHNL:
531 if (!ch)
532 goto err_rings;
533 vsi->num_rxq = ch->num_rxq;
534 vsi->num_txq = ch->num_txq;
535 vsi->next_base_q = ch->base_q;
536 break;
537 case ICE_VSI_LB:
538 if (ice_vsi_alloc_arrays(vsi))
539 goto err_rings;
540 break;
541 default:
542 dev_warn(dev, "Unknown VSI type %d\n", vsi->type);
543 goto unlock_pf;
544 }
545
546 if (vsi->type == ICE_VSI_CTRL && !vf) {
547 /* Use the last VSI slot as the index for PF control VSI */
548 vsi->idx = pf->num_alloc_vsi - 1;
549 pf->ctrl_vsi_idx = vsi->idx;
550 pf->vsi[vsi->idx] = vsi;
551 } else {
552 /* fill slot and make note of the index */
553 vsi->idx = pf->next_vsi;
554 pf->vsi[pf->next_vsi] = vsi;
555
556 /* prepare pf->next_vsi for next use */
557 pf->next_vsi = ice_get_free_slot(pf->vsi, pf->num_alloc_vsi,
558 pf->next_vsi);
559 }
560
561 if (vsi->type == ICE_VSI_CTRL && vf)
562 vf->ctrl_vsi_idx = vsi->idx;
563 goto unlock_pf;
564
565 err_rings:
566 devm_kfree(dev, vsi);
567 vsi = NULL;
568 unlock_pf:
569 mutex_unlock(&pf->sw_mutex);
570 return vsi;
571 }
572
573 /**
574 * ice_alloc_fd_res - Allocate FD resource for a VSI
575 * @vsi: pointer to the ice_vsi
576 *
577 * This allocates the FD resources
578 *
579 * Returns 0 on success, -EPERM on no-op or -EIO on failure
580 */
ice_alloc_fd_res(struct ice_vsi * vsi)581 static int ice_alloc_fd_res(struct ice_vsi *vsi)
582 {
583 struct ice_pf *pf = vsi->back;
584 u32 g_val, b_val;
585
586 /* Flow Director filters are only allocated/assigned to the PF VSI or
587 * CHNL VSI which passes the traffic. The CTRL VSI is only used to
588 * add/delete filters so resources are not allocated to it
589 */
590 if (!test_bit(ICE_FLAG_FD_ENA, pf->flags))
591 return -EPERM;
592
593 if (!(vsi->type == ICE_VSI_PF || vsi->type == ICE_VSI_VF ||
594 vsi->type == ICE_VSI_CHNL))
595 return -EPERM;
596
597 /* FD filters from guaranteed pool per VSI */
598 g_val = pf->hw.func_caps.fd_fltr_guar;
599 if (!g_val)
600 return -EPERM;
601
602 /* FD filters from best effort pool */
603 b_val = pf->hw.func_caps.fd_fltr_best_effort;
604 if (!b_val)
605 return -EPERM;
606
607 /* PF main VSI gets only 64 FD resources from guaranteed pool
608 * when ADQ is configured.
609 */
610 #define ICE_PF_VSI_GFLTR 64
611
612 /* determine FD filter resources per VSI from shared(best effort) and
613 * dedicated pool
614 */
615 if (vsi->type == ICE_VSI_PF) {
616 vsi->num_gfltr = g_val;
617 /* if MQPRIO is configured, main VSI doesn't get all FD
618 * resources from guaranteed pool. PF VSI gets 64 FD resources
619 */
620 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
621 if (g_val < ICE_PF_VSI_GFLTR)
622 return -EPERM;
623 /* allow bare minimum entries for PF VSI */
624 vsi->num_gfltr = ICE_PF_VSI_GFLTR;
625 }
626
627 /* each VSI gets same "best_effort" quota */
628 vsi->num_bfltr = b_val;
629 } else if (vsi->type == ICE_VSI_VF) {
630 vsi->num_gfltr = 0;
631
632 /* each VSI gets same "best_effort" quota */
633 vsi->num_bfltr = b_val;
634 } else {
635 struct ice_vsi *main_vsi;
636 int numtc;
637
638 main_vsi = ice_get_main_vsi(pf);
639 if (!main_vsi)
640 return -EPERM;
641
642 if (!main_vsi->all_numtc)
643 return -EINVAL;
644
645 /* figure out ADQ numtc */
646 numtc = main_vsi->all_numtc - ICE_CHNL_START_TC;
647
648 /* only one TC but still asking resources for channels,
649 * invalid config
650 */
651 if (numtc < ICE_CHNL_START_TC)
652 return -EPERM;
653
654 g_val -= ICE_PF_VSI_GFLTR;
655 /* channel VSIs gets equal share from guaranteed pool */
656 vsi->num_gfltr = g_val / numtc;
657
658 /* each VSI gets same "best_effort" quota */
659 vsi->num_bfltr = b_val;
660 }
661
662 return 0;
663 }
664
665 /**
666 * ice_vsi_get_qs - Assign queues from PF to VSI
667 * @vsi: the VSI to assign queues to
668 *
669 * Returns 0 on success and a negative value on error
670 */
ice_vsi_get_qs(struct ice_vsi * vsi)671 static int ice_vsi_get_qs(struct ice_vsi *vsi)
672 {
673 struct ice_pf *pf = vsi->back;
674 struct ice_qs_cfg tx_qs_cfg = {
675 .qs_mutex = &pf->avail_q_mutex,
676 .pf_map = pf->avail_txqs,
677 .pf_map_size = pf->max_pf_txqs,
678 .q_count = vsi->alloc_txq,
679 .scatter_count = ICE_MAX_SCATTER_TXQS,
680 .vsi_map = vsi->txq_map,
681 .vsi_map_offset = 0,
682 .mapping_mode = ICE_VSI_MAP_CONTIG
683 };
684 struct ice_qs_cfg rx_qs_cfg = {
685 .qs_mutex = &pf->avail_q_mutex,
686 .pf_map = pf->avail_rxqs,
687 .pf_map_size = pf->max_pf_rxqs,
688 .q_count = vsi->alloc_rxq,
689 .scatter_count = ICE_MAX_SCATTER_RXQS,
690 .vsi_map = vsi->rxq_map,
691 .vsi_map_offset = 0,
692 .mapping_mode = ICE_VSI_MAP_CONTIG
693 };
694 int ret;
695
696 if (vsi->type == ICE_VSI_CHNL)
697 return 0;
698
699 ret = __ice_vsi_get_qs(&tx_qs_cfg);
700 if (ret)
701 return ret;
702 vsi->tx_mapping_mode = tx_qs_cfg.mapping_mode;
703
704 ret = __ice_vsi_get_qs(&rx_qs_cfg);
705 if (ret)
706 return ret;
707 vsi->rx_mapping_mode = rx_qs_cfg.mapping_mode;
708
709 return 0;
710 }
711
712 /**
713 * ice_vsi_put_qs - Release queues from VSI to PF
714 * @vsi: the VSI that is going to release queues
715 */
ice_vsi_put_qs(struct ice_vsi * vsi)716 static void ice_vsi_put_qs(struct ice_vsi *vsi)
717 {
718 struct ice_pf *pf = vsi->back;
719 int i;
720
721 mutex_lock(&pf->avail_q_mutex);
722
723 ice_for_each_alloc_txq(vsi, i) {
724 clear_bit(vsi->txq_map[i], pf->avail_txqs);
725 vsi->txq_map[i] = ICE_INVAL_Q_INDEX;
726 }
727
728 ice_for_each_alloc_rxq(vsi, i) {
729 clear_bit(vsi->rxq_map[i], pf->avail_rxqs);
730 vsi->rxq_map[i] = ICE_INVAL_Q_INDEX;
731 }
732
733 mutex_unlock(&pf->avail_q_mutex);
734 }
735
736 /**
737 * ice_is_safe_mode
738 * @pf: pointer to the PF struct
739 *
740 * returns true if driver is in safe mode, false otherwise
741 */
ice_is_safe_mode(struct ice_pf * pf)742 bool ice_is_safe_mode(struct ice_pf *pf)
743 {
744 return !test_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
745 }
746
747 /**
748 * ice_is_rdma_ena
749 * @pf: pointer to the PF struct
750 *
751 * returns true if RDMA is currently supported, false otherwise
752 */
ice_is_rdma_ena(struct ice_pf * pf)753 bool ice_is_rdma_ena(struct ice_pf *pf)
754 {
755 return test_bit(ICE_FLAG_RDMA_ENA, pf->flags);
756 }
757
758 /**
759 * ice_vsi_clean_rss_flow_fld - Delete RSS configuration
760 * @vsi: the VSI being cleaned up
761 *
762 * This function deletes RSS input set for all flows that were configured
763 * for this VSI
764 */
ice_vsi_clean_rss_flow_fld(struct ice_vsi * vsi)765 static void ice_vsi_clean_rss_flow_fld(struct ice_vsi *vsi)
766 {
767 struct ice_pf *pf = vsi->back;
768 int status;
769
770 if (ice_is_safe_mode(pf))
771 return;
772
773 status = ice_rem_vsi_rss_cfg(&pf->hw, vsi->idx);
774 if (status)
775 dev_dbg(ice_pf_to_dev(pf), "ice_rem_vsi_rss_cfg failed for vsi = %d, error = %d\n",
776 vsi->vsi_num, status);
777 }
778
779 /**
780 * ice_rss_clean - Delete RSS related VSI structures and configuration
781 * @vsi: the VSI being removed
782 */
ice_rss_clean(struct ice_vsi * vsi)783 static void ice_rss_clean(struct ice_vsi *vsi)
784 {
785 struct ice_pf *pf = vsi->back;
786 struct device *dev;
787
788 dev = ice_pf_to_dev(pf);
789
790 if (vsi->rss_hkey_user)
791 devm_kfree(dev, vsi->rss_hkey_user);
792 if (vsi->rss_lut_user)
793 devm_kfree(dev, vsi->rss_lut_user);
794
795 ice_vsi_clean_rss_flow_fld(vsi);
796 /* remove RSS replay list */
797 if (!ice_is_safe_mode(pf))
798 ice_rem_vsi_rss_list(&pf->hw, vsi->idx);
799 }
800
801 /**
802 * ice_vsi_set_rss_params - Setup RSS capabilities per VSI type
803 * @vsi: the VSI being configured
804 */
ice_vsi_set_rss_params(struct ice_vsi * vsi)805 static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
806 {
807 struct ice_hw_common_caps *cap;
808 struct ice_pf *pf = vsi->back;
809
810 if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
811 vsi->rss_size = 1;
812 return;
813 }
814
815 cap = &pf->hw.func_caps.common_cap;
816 switch (vsi->type) {
817 case ICE_VSI_CHNL:
818 case ICE_VSI_PF:
819 /* PF VSI will inherit RSS instance of PF */
820 vsi->rss_table_size = (u16)cap->rss_table_size;
821 if (vsi->type == ICE_VSI_CHNL)
822 vsi->rss_size = min_t(u16, vsi->num_rxq,
823 BIT(cap->rss_table_entry_width));
824 else
825 vsi->rss_size = min_t(u16, num_online_cpus(),
826 BIT(cap->rss_table_entry_width));
827 vsi->rss_lut_type = ICE_AQC_GSET_RSS_LUT_TABLE_TYPE_PF;
828 break;
829 case ICE_VSI_SWITCHDEV_CTRL:
830 vsi->rss_table_size = ICE_VSIQF_HLUT_ARRAY_SIZE;
831 vsi->rss_size = min_t(u16, num_online_cpus(),
832 BIT(cap->rss_table_entry_width));
833 vsi->rss_lut_type = ICE_AQC_GSET_RSS_LUT_TABLE_TYPE_VSI;
834 break;
835 case ICE_VSI_VF:
836 /* VF VSI will get a small RSS table.
837 * For VSI_LUT, LUT size should be set to 64 bytes.
838 */
839 vsi->rss_table_size = ICE_VSIQF_HLUT_ARRAY_SIZE;
840 vsi->rss_size = ICE_MAX_RSS_QS_PER_VF;
841 vsi->rss_lut_type = ICE_AQC_GSET_RSS_LUT_TABLE_TYPE_VSI;
842 break;
843 case ICE_VSI_LB:
844 break;
845 default:
846 dev_dbg(ice_pf_to_dev(pf), "Unsupported VSI type %s\n",
847 ice_vsi_type_str(vsi->type));
848 break;
849 }
850 }
851
852 /**
853 * ice_set_dflt_vsi_ctx - Set default VSI context before adding a VSI
854 * @hw: HW structure used to determine the VLAN mode of the device
855 * @ctxt: the VSI context being set
856 *
857 * This initializes a default VSI context for all sections except the Queues.
858 */
ice_set_dflt_vsi_ctx(struct ice_hw * hw,struct ice_vsi_ctx * ctxt)859 static void ice_set_dflt_vsi_ctx(struct ice_hw *hw, struct ice_vsi_ctx *ctxt)
860 {
861 u32 table = 0;
862
863 memset(&ctxt->info, 0, sizeof(ctxt->info));
864 /* VSI's should be allocated from shared pool */
865 ctxt->alloc_from_pool = true;
866 /* Src pruning enabled by default */
867 ctxt->info.sw_flags = ICE_AQ_VSI_SW_FLAG_SRC_PRUNE;
868 /* Traffic from VSI can be sent to LAN */
869 ctxt->info.sw_flags2 = ICE_AQ_VSI_SW_FLAG_LAN_ENA;
870 /* allow all untagged/tagged packets by default on Tx */
871 ctxt->info.inner_vlan_flags = ((ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL &
872 ICE_AQ_VSI_INNER_VLAN_TX_MODE_M) >>
873 ICE_AQ_VSI_INNER_VLAN_TX_MODE_S);
874 /* SVM - by default bits 3 and 4 in inner_vlan_flags are 0's which
875 * results in legacy behavior (show VLAN, DEI, and UP) in descriptor.
876 *
877 * DVM - leave inner VLAN in packet by default
878 */
879 if (ice_is_dvm_ena(hw)) {
880 ctxt->info.inner_vlan_flags |=
881 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
882 ctxt->info.outer_vlan_flags =
883 (ICE_AQ_VSI_OUTER_VLAN_TX_MODE_ALL <<
884 ICE_AQ_VSI_OUTER_VLAN_TX_MODE_S) &
885 ICE_AQ_VSI_OUTER_VLAN_TX_MODE_M;
886 ctxt->info.outer_vlan_flags |=
887 (ICE_AQ_VSI_OUTER_TAG_VLAN_8100 <<
888 ICE_AQ_VSI_OUTER_TAG_TYPE_S) &
889 ICE_AQ_VSI_OUTER_TAG_TYPE_M;
890 ctxt->info.outer_vlan_flags |=
891 FIELD_PREP(ICE_AQ_VSI_OUTER_VLAN_EMODE_M,
892 ICE_AQ_VSI_OUTER_VLAN_EMODE_NOTHING);
893 }
894 /* Have 1:1 UP mapping for both ingress/egress tables */
895 table |= ICE_UP_TABLE_TRANSLATE(0, 0);
896 table |= ICE_UP_TABLE_TRANSLATE(1, 1);
897 table |= ICE_UP_TABLE_TRANSLATE(2, 2);
898 table |= ICE_UP_TABLE_TRANSLATE(3, 3);
899 table |= ICE_UP_TABLE_TRANSLATE(4, 4);
900 table |= ICE_UP_TABLE_TRANSLATE(5, 5);
901 table |= ICE_UP_TABLE_TRANSLATE(6, 6);
902 table |= ICE_UP_TABLE_TRANSLATE(7, 7);
903 ctxt->info.ingress_table = cpu_to_le32(table);
904 ctxt->info.egress_table = cpu_to_le32(table);
905 /* Have 1:1 UP mapping for outer to inner UP table */
906 ctxt->info.outer_up_table = cpu_to_le32(table);
907 /* No Outer tag support outer_tag_flags remains to zero */
908 }
909
910 /**
911 * ice_vsi_setup_q_map - Setup a VSI queue map
912 * @vsi: the VSI being configured
913 * @ctxt: VSI context structure
914 */
ice_vsi_setup_q_map(struct ice_vsi * vsi,struct ice_vsi_ctx * ctxt)915 static int ice_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
916 {
917 u16 offset = 0, qmap = 0, tx_count = 0, rx_count = 0, pow = 0;
918 u16 num_txq_per_tc, num_rxq_per_tc;
919 u16 qcount_tx = vsi->alloc_txq;
920 u16 qcount_rx = vsi->alloc_rxq;
921 u8 netdev_tc = 0;
922 int i;
923
924 if (!vsi->tc_cfg.numtc) {
925 /* at least TC0 should be enabled by default */
926 vsi->tc_cfg.numtc = 1;
927 vsi->tc_cfg.ena_tc = 1;
928 }
929
930 num_rxq_per_tc = min_t(u16, qcount_rx / vsi->tc_cfg.numtc, ICE_MAX_RXQS_PER_TC);
931 if (!num_rxq_per_tc)
932 num_rxq_per_tc = 1;
933 num_txq_per_tc = qcount_tx / vsi->tc_cfg.numtc;
934 if (!num_txq_per_tc)
935 num_txq_per_tc = 1;
936
937 /* find the (rounded up) power-of-2 of qcount */
938 pow = (u16)order_base_2(num_rxq_per_tc);
939
940 /* TC mapping is a function of the number of Rx queues assigned to the
941 * VSI for each traffic class and the offset of these queues.
942 * The first 10 bits are for queue offset for TC0, next 4 bits for no:of
943 * queues allocated to TC0. No:of queues is a power-of-2.
944 *
945 * If TC is not enabled, the queue offset is set to 0, and allocate one
946 * queue, this way, traffic for the given TC will be sent to the default
947 * queue.
948 *
949 * Setup number and offset of Rx queues for all TCs for the VSI
950 */
951 ice_for_each_traffic_class(i) {
952 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
953 /* TC is not enabled */
954 vsi->tc_cfg.tc_info[i].qoffset = 0;
955 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
956 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
957 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
958 ctxt->info.tc_mapping[i] = 0;
959 continue;
960 }
961
962 /* TC is enabled */
963 vsi->tc_cfg.tc_info[i].qoffset = offset;
964 vsi->tc_cfg.tc_info[i].qcount_rx = num_rxq_per_tc;
965 vsi->tc_cfg.tc_info[i].qcount_tx = num_txq_per_tc;
966 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
967
968 qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
969 ICE_AQ_VSI_TC_Q_OFFSET_M) |
970 ((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
971 ICE_AQ_VSI_TC_Q_NUM_M);
972 offset += num_rxq_per_tc;
973 tx_count += num_txq_per_tc;
974 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
975 }
976
977 /* if offset is non-zero, means it is calculated correctly based on
978 * enabled TCs for a given VSI otherwise qcount_rx will always
979 * be correct and non-zero because it is based off - VSI's
980 * allocated Rx queues which is at least 1 (hence qcount_tx will be
981 * at least 1)
982 */
983 if (offset)
984 rx_count = offset;
985 else
986 rx_count = num_rxq_per_tc;
987
988 if (rx_count > vsi->alloc_rxq) {
989 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
990 rx_count, vsi->alloc_rxq);
991 return -EINVAL;
992 }
993
994 if (tx_count > vsi->alloc_txq) {
995 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
996 tx_count, vsi->alloc_txq);
997 return -EINVAL;
998 }
999
1000 vsi->num_txq = tx_count;
1001 vsi->num_rxq = rx_count;
1002
1003 if (vsi->type == ICE_VSI_VF && vsi->num_txq != vsi->num_rxq) {
1004 dev_dbg(ice_pf_to_dev(vsi->back), "VF VSI should have same number of Tx and Rx queues. Hence making them equal\n");
1005 /* since there is a chance that num_rxq could have been changed
1006 * in the above for loop, make num_txq equal to num_rxq.
1007 */
1008 vsi->num_txq = vsi->num_rxq;
1009 }
1010
1011 /* Rx queue mapping */
1012 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1013 /* q_mapping buffer holds the info for the first queue allocated for
1014 * this VSI in the PF space and also the number of queues associated
1015 * with this VSI.
1016 */
1017 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
1018 ctxt->info.q_mapping[1] = cpu_to_le16(vsi->num_rxq);
1019
1020 return 0;
1021 }
1022
1023 /**
1024 * ice_set_fd_vsi_ctx - Set FD VSI context before adding a VSI
1025 * @ctxt: the VSI context being set
1026 * @vsi: the VSI being configured
1027 */
ice_set_fd_vsi_ctx(struct ice_vsi_ctx * ctxt,struct ice_vsi * vsi)1028 static void ice_set_fd_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1029 {
1030 u8 dflt_q_group, dflt_q_prio;
1031 u16 dflt_q, report_q, val;
1032
1033 if (vsi->type != ICE_VSI_PF && vsi->type != ICE_VSI_CTRL &&
1034 vsi->type != ICE_VSI_VF && vsi->type != ICE_VSI_CHNL)
1035 return;
1036
1037 val = ICE_AQ_VSI_PROP_FLOW_DIR_VALID;
1038 ctxt->info.valid_sections |= cpu_to_le16(val);
1039 dflt_q = 0;
1040 dflt_q_group = 0;
1041 report_q = 0;
1042 dflt_q_prio = 0;
1043
1044 /* enable flow director filtering/programming */
1045 val = ICE_AQ_VSI_FD_ENABLE | ICE_AQ_VSI_FD_PROG_ENABLE;
1046 ctxt->info.fd_options = cpu_to_le16(val);
1047 /* max of allocated flow director filters */
1048 ctxt->info.max_fd_fltr_dedicated =
1049 cpu_to_le16(vsi->num_gfltr);
1050 /* max of shared flow director filters any VSI may program */
1051 ctxt->info.max_fd_fltr_shared =
1052 cpu_to_le16(vsi->num_bfltr);
1053 /* default queue index within the VSI of the default FD */
1054 val = ((dflt_q << ICE_AQ_VSI_FD_DEF_Q_S) &
1055 ICE_AQ_VSI_FD_DEF_Q_M);
1056 /* target queue or queue group to the FD filter */
1057 val |= ((dflt_q_group << ICE_AQ_VSI_FD_DEF_GRP_S) &
1058 ICE_AQ_VSI_FD_DEF_GRP_M);
1059 ctxt->info.fd_def_q = cpu_to_le16(val);
1060 /* queue index on which FD filter completion is reported */
1061 val = ((report_q << ICE_AQ_VSI_FD_REPORT_Q_S) &
1062 ICE_AQ_VSI_FD_REPORT_Q_M);
1063 /* priority of the default qindex action */
1064 val |= ((dflt_q_prio << ICE_AQ_VSI_FD_DEF_PRIORITY_S) &
1065 ICE_AQ_VSI_FD_DEF_PRIORITY_M);
1066 ctxt->info.fd_report_opt = cpu_to_le16(val);
1067 }
1068
1069 /**
1070 * ice_set_rss_vsi_ctx - Set RSS VSI context before adding a VSI
1071 * @ctxt: the VSI context being set
1072 * @vsi: the VSI being configured
1073 */
ice_set_rss_vsi_ctx(struct ice_vsi_ctx * ctxt,struct ice_vsi * vsi)1074 static void ice_set_rss_vsi_ctx(struct ice_vsi_ctx *ctxt, struct ice_vsi *vsi)
1075 {
1076 u8 lut_type, hash_type;
1077 struct device *dev;
1078 struct ice_pf *pf;
1079
1080 pf = vsi->back;
1081 dev = ice_pf_to_dev(pf);
1082
1083 switch (vsi->type) {
1084 case ICE_VSI_CHNL:
1085 case ICE_VSI_PF:
1086 /* PF VSI will inherit RSS instance of PF */
1087 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_PF;
1088 hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
1089 break;
1090 case ICE_VSI_VF:
1091 /* VF VSI will gets a small RSS table which is a VSI LUT type */
1092 lut_type = ICE_AQ_VSI_Q_OPT_RSS_LUT_VSI;
1093 hash_type = ICE_AQ_VSI_Q_OPT_RSS_TPLZ;
1094 break;
1095 default:
1096 dev_dbg(dev, "Unsupported VSI type %s\n",
1097 ice_vsi_type_str(vsi->type));
1098 return;
1099 }
1100
1101 ctxt->info.q_opt_rss = ((lut_type << ICE_AQ_VSI_Q_OPT_RSS_LUT_S) &
1102 ICE_AQ_VSI_Q_OPT_RSS_LUT_M) |
1103 ((hash_type << ICE_AQ_VSI_Q_OPT_RSS_HASH_S) &
1104 ICE_AQ_VSI_Q_OPT_RSS_HASH_M);
1105 }
1106
1107 static void
ice_chnl_vsi_setup_q_map(struct ice_vsi * vsi,struct ice_vsi_ctx * ctxt)1108 ice_chnl_vsi_setup_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt)
1109 {
1110 struct ice_pf *pf = vsi->back;
1111 u16 qcount, qmap;
1112 u8 offset = 0;
1113 int pow;
1114
1115 qcount = min_t(int, vsi->num_rxq, pf->num_lan_msix);
1116
1117 pow = order_base_2(qcount);
1118 qmap = ((offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
1119 ICE_AQ_VSI_TC_Q_OFFSET_M) |
1120 ((pow << ICE_AQ_VSI_TC_Q_NUM_S) &
1121 ICE_AQ_VSI_TC_Q_NUM_M);
1122
1123 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1124 ctxt->info.mapping_flags |= cpu_to_le16(ICE_AQ_VSI_Q_MAP_CONTIG);
1125 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->next_base_q);
1126 ctxt->info.q_mapping[1] = cpu_to_le16(qcount);
1127 }
1128
1129 /**
1130 * ice_vsi_init - Create and initialize a VSI
1131 * @vsi: the VSI being configured
1132 * @init_vsi: is this call creating a VSI
1133 *
1134 * This initializes a VSI context depending on the VSI type to be added and
1135 * passes it down to the add_vsi aq command to create a new VSI.
1136 */
ice_vsi_init(struct ice_vsi * vsi,bool init_vsi)1137 static int ice_vsi_init(struct ice_vsi *vsi, bool init_vsi)
1138 {
1139 struct ice_pf *pf = vsi->back;
1140 struct ice_hw *hw = &pf->hw;
1141 struct ice_vsi_ctx *ctxt;
1142 struct device *dev;
1143 int ret = 0;
1144
1145 dev = ice_pf_to_dev(pf);
1146 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1147 if (!ctxt)
1148 return -ENOMEM;
1149
1150 switch (vsi->type) {
1151 case ICE_VSI_CTRL:
1152 case ICE_VSI_LB:
1153 case ICE_VSI_PF:
1154 ctxt->flags = ICE_AQ_VSI_TYPE_PF;
1155 break;
1156 case ICE_VSI_SWITCHDEV_CTRL:
1157 case ICE_VSI_CHNL:
1158 ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
1159 break;
1160 case ICE_VSI_VF:
1161 ctxt->flags = ICE_AQ_VSI_TYPE_VF;
1162 /* VF number here is the absolute VF number (0-255) */
1163 ctxt->vf_num = vsi->vf->vf_id + hw->func_caps.vf_base_id;
1164 break;
1165 default:
1166 ret = -ENODEV;
1167 goto out;
1168 }
1169
1170 /* Handle VLAN pruning for channel VSI if main VSI has VLAN
1171 * prune enabled
1172 */
1173 if (vsi->type == ICE_VSI_CHNL) {
1174 struct ice_vsi *main_vsi;
1175
1176 main_vsi = ice_get_main_vsi(pf);
1177 if (main_vsi && ice_vsi_is_vlan_pruning_ena(main_vsi))
1178 ctxt->info.sw_flags2 |=
1179 ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1180 else
1181 ctxt->info.sw_flags2 &=
1182 ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
1183 }
1184
1185 ice_set_dflt_vsi_ctx(hw, ctxt);
1186 if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
1187 ice_set_fd_vsi_ctx(ctxt, vsi);
1188 /* if the switch is in VEB mode, allow VSI loopback */
1189 if (vsi->vsw->bridge_mode == BRIDGE_MODE_VEB)
1190 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
1191
1192 /* Set LUT type and HASH type if RSS is enabled */
1193 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags) &&
1194 vsi->type != ICE_VSI_CTRL) {
1195 ice_set_rss_vsi_ctx(ctxt, vsi);
1196 /* if updating VSI context, make sure to set valid_section:
1197 * to indicate which section of VSI context being updated
1198 */
1199 if (!init_vsi)
1200 ctxt->info.valid_sections |=
1201 cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
1202 }
1203
1204 ctxt->info.sw_id = vsi->port_info->sw_id;
1205 if (vsi->type == ICE_VSI_CHNL) {
1206 ice_chnl_vsi_setup_q_map(vsi, ctxt);
1207 } else {
1208 ret = ice_vsi_setup_q_map(vsi, ctxt);
1209 if (ret)
1210 goto out;
1211
1212 if (!init_vsi) /* means VSI being updated */
1213 /* must to indicate which section of VSI context are
1214 * being modified
1215 */
1216 ctxt->info.valid_sections |=
1217 cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
1218 }
1219
1220 /* Allow control frames out of main VSI */
1221 if (vsi->type == ICE_VSI_PF) {
1222 ctxt->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
1223 ctxt->info.valid_sections |=
1224 cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
1225 }
1226
1227 if (init_vsi) {
1228 ret = ice_add_vsi(hw, vsi->idx, ctxt, NULL);
1229 if (ret) {
1230 dev_err(dev, "Add VSI failed, err %d\n", ret);
1231 ret = -EIO;
1232 goto out;
1233 }
1234 } else {
1235 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
1236 if (ret) {
1237 dev_err(dev, "Update VSI failed, err %d\n", ret);
1238 ret = -EIO;
1239 goto out;
1240 }
1241 }
1242
1243 /* keep context for update VSI operations */
1244 vsi->info = ctxt->info;
1245
1246 /* record VSI number returned */
1247 vsi->vsi_num = ctxt->vsi_num;
1248
1249 out:
1250 kfree(ctxt);
1251 return ret;
1252 }
1253
1254 /**
1255 * ice_free_res - free a block of resources
1256 * @res: pointer to the resource
1257 * @index: starting index previously returned by ice_get_res
1258 * @id: identifier to track owner
1259 *
1260 * Returns number of resources freed
1261 */
ice_free_res(struct ice_res_tracker * res,u16 index,u16 id)1262 int ice_free_res(struct ice_res_tracker *res, u16 index, u16 id)
1263 {
1264 int count = 0;
1265 int i;
1266
1267 if (!res || index >= res->end)
1268 return -EINVAL;
1269
1270 id |= ICE_RES_VALID_BIT;
1271 for (i = index; i < res->end && res->list[i] == id; i++) {
1272 res->list[i] = 0;
1273 count++;
1274 }
1275
1276 return count;
1277 }
1278
1279 /**
1280 * ice_search_res - Search the tracker for a block of resources
1281 * @res: pointer to the resource
1282 * @needed: size of the block needed
1283 * @id: identifier to track owner
1284 *
1285 * Returns the base item index of the block, or -ENOMEM for error
1286 */
ice_search_res(struct ice_res_tracker * res,u16 needed,u16 id)1287 static int ice_search_res(struct ice_res_tracker *res, u16 needed, u16 id)
1288 {
1289 u16 start = 0, end = 0;
1290
1291 if (needed > res->end)
1292 return -ENOMEM;
1293
1294 id |= ICE_RES_VALID_BIT;
1295
1296 do {
1297 /* skip already allocated entries */
1298 if (res->list[end++] & ICE_RES_VALID_BIT) {
1299 start = end;
1300 if ((start + needed) > res->end)
1301 break;
1302 }
1303
1304 if (end == (start + needed)) {
1305 int i = start;
1306
1307 /* there was enough, so assign it to the requestor */
1308 while (i != end)
1309 res->list[i++] = id;
1310
1311 return start;
1312 }
1313 } while (end < res->end);
1314
1315 return -ENOMEM;
1316 }
1317
1318 /**
1319 * ice_get_free_res_count - Get free count from a resource tracker
1320 * @res: Resource tracker instance
1321 */
ice_get_free_res_count(struct ice_res_tracker * res)1322 static u16 ice_get_free_res_count(struct ice_res_tracker *res)
1323 {
1324 u16 i, count = 0;
1325
1326 for (i = 0; i < res->end; i++)
1327 if (!(res->list[i] & ICE_RES_VALID_BIT))
1328 count++;
1329
1330 return count;
1331 }
1332
1333 /**
1334 * ice_get_res - get a block of resources
1335 * @pf: board private structure
1336 * @res: pointer to the resource
1337 * @needed: size of the block needed
1338 * @id: identifier to track owner
1339 *
1340 * Returns the base item index of the block, or negative for error
1341 */
1342 int
ice_get_res(struct ice_pf * pf,struct ice_res_tracker * res,u16 needed,u16 id)1343 ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res, u16 needed, u16 id)
1344 {
1345 if (!res || !pf)
1346 return -EINVAL;
1347
1348 if (!needed || needed > res->num_entries || id >= ICE_RES_VALID_BIT) {
1349 dev_err(ice_pf_to_dev(pf), "param err: needed=%d, num_entries = %d id=0x%04x\n",
1350 needed, res->num_entries, id);
1351 return -EINVAL;
1352 }
1353
1354 return ice_search_res(res, needed, id);
1355 }
1356
1357 /**
1358 * ice_get_vf_ctrl_res - Get VF control VSI resource
1359 * @pf: pointer to the PF structure
1360 * @vsi: the VSI to allocate a resource for
1361 *
1362 * Look up whether another VF has already allocated the control VSI resource.
1363 * If so, re-use this resource so that we share it among all VFs.
1364 *
1365 * Otherwise, allocate the resource and return it.
1366 */
ice_get_vf_ctrl_res(struct ice_pf * pf,struct ice_vsi * vsi)1367 static int ice_get_vf_ctrl_res(struct ice_pf *pf, struct ice_vsi *vsi)
1368 {
1369 struct ice_vf *vf;
1370 unsigned int bkt;
1371 int base;
1372
1373 rcu_read_lock();
1374 ice_for_each_vf_rcu(pf, bkt, vf) {
1375 if (vf != vsi->vf && vf->ctrl_vsi_idx != ICE_NO_VSI) {
1376 base = pf->vsi[vf->ctrl_vsi_idx]->base_vector;
1377 rcu_read_unlock();
1378 return base;
1379 }
1380 }
1381 rcu_read_unlock();
1382
1383 return ice_get_res(pf, pf->irq_tracker, vsi->num_q_vectors,
1384 ICE_RES_VF_CTRL_VEC_ID);
1385 }
1386
1387 /**
1388 * ice_vsi_setup_vector_base - Set up the base vector for the given VSI
1389 * @vsi: ptr to the VSI
1390 *
1391 * This should only be called after ice_vsi_alloc() which allocates the
1392 * corresponding SW VSI structure and initializes num_queue_pairs for the
1393 * newly allocated VSI.
1394 *
1395 * Returns 0 on success or negative on failure
1396 */
ice_vsi_setup_vector_base(struct ice_vsi * vsi)1397 static int ice_vsi_setup_vector_base(struct ice_vsi *vsi)
1398 {
1399 struct ice_pf *pf = vsi->back;
1400 struct device *dev;
1401 u16 num_q_vectors;
1402 int base;
1403
1404 dev = ice_pf_to_dev(pf);
1405 /* SRIOV doesn't grab irq_tracker entries for each VSI */
1406 if (vsi->type == ICE_VSI_VF)
1407 return 0;
1408 if (vsi->type == ICE_VSI_CHNL)
1409 return 0;
1410
1411 if (vsi->base_vector) {
1412 dev_dbg(dev, "VSI %d has non-zero base vector %d\n",
1413 vsi->vsi_num, vsi->base_vector);
1414 return -EEXIST;
1415 }
1416
1417 num_q_vectors = vsi->num_q_vectors;
1418 /* reserve slots from OS requested IRQs */
1419 if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
1420 base = ice_get_vf_ctrl_res(pf, vsi);
1421 } else {
1422 base = ice_get_res(pf, pf->irq_tracker, num_q_vectors,
1423 vsi->idx);
1424 }
1425
1426 if (base < 0) {
1427 dev_err(dev, "%d MSI-X interrupts available. %s %d failed to get %d MSI-X vectors\n",
1428 ice_get_free_res_count(pf->irq_tracker),
1429 ice_vsi_type_str(vsi->type), vsi->idx, num_q_vectors);
1430 return -ENOENT;
1431 }
1432 vsi->base_vector = (u16)base;
1433 pf->num_avail_sw_msix -= num_q_vectors;
1434
1435 return 0;
1436 }
1437
1438 /**
1439 * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
1440 * @vsi: the VSI having rings deallocated
1441 */
ice_vsi_clear_rings(struct ice_vsi * vsi)1442 static void ice_vsi_clear_rings(struct ice_vsi *vsi)
1443 {
1444 int i;
1445
1446 /* Avoid stale references by clearing map from vector to ring */
1447 if (vsi->q_vectors) {
1448 ice_for_each_q_vector(vsi, i) {
1449 struct ice_q_vector *q_vector = vsi->q_vectors[i];
1450
1451 if (q_vector) {
1452 q_vector->tx.tx_ring = NULL;
1453 q_vector->rx.rx_ring = NULL;
1454 }
1455 }
1456 }
1457
1458 if (vsi->tx_rings) {
1459 ice_for_each_alloc_txq(vsi, i) {
1460 if (vsi->tx_rings[i]) {
1461 kfree_rcu(vsi->tx_rings[i], rcu);
1462 WRITE_ONCE(vsi->tx_rings[i], NULL);
1463 }
1464 }
1465 }
1466 if (vsi->rx_rings) {
1467 ice_for_each_alloc_rxq(vsi, i) {
1468 if (vsi->rx_rings[i]) {
1469 kfree_rcu(vsi->rx_rings[i], rcu);
1470 WRITE_ONCE(vsi->rx_rings[i], NULL);
1471 }
1472 }
1473 }
1474 }
1475
1476 /**
1477 * ice_vsi_alloc_rings - Allocates Tx and Rx rings for the VSI
1478 * @vsi: VSI which is having rings allocated
1479 */
ice_vsi_alloc_rings(struct ice_vsi * vsi)1480 static int ice_vsi_alloc_rings(struct ice_vsi *vsi)
1481 {
1482 bool dvm_ena = ice_is_dvm_ena(&vsi->back->hw);
1483 struct ice_pf *pf = vsi->back;
1484 struct device *dev;
1485 u16 i;
1486
1487 dev = ice_pf_to_dev(pf);
1488 /* Allocate Tx rings */
1489 ice_for_each_alloc_txq(vsi, i) {
1490 struct ice_tx_ring *ring;
1491
1492 /* allocate with kzalloc(), free with kfree_rcu() */
1493 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1494
1495 if (!ring)
1496 goto err_out;
1497
1498 ring->q_index = i;
1499 ring->reg_idx = vsi->txq_map[i];
1500 ring->vsi = vsi;
1501 ring->tx_tstamps = &pf->ptp.port.tx;
1502 ring->dev = dev;
1503 ring->count = vsi->num_tx_desc;
1504 ring->txq_teid = ICE_INVAL_TEID;
1505 if (dvm_ena)
1506 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG2;
1507 else
1508 ring->flags |= ICE_TX_FLAGS_RING_VLAN_L2TAG1;
1509 WRITE_ONCE(vsi->tx_rings[i], ring);
1510 }
1511
1512 /* Allocate Rx rings */
1513 ice_for_each_alloc_rxq(vsi, i) {
1514 struct ice_rx_ring *ring;
1515
1516 /* allocate with kzalloc(), free with kfree_rcu() */
1517 ring = kzalloc(sizeof(*ring), GFP_KERNEL);
1518 if (!ring)
1519 goto err_out;
1520
1521 ring->q_index = i;
1522 ring->reg_idx = vsi->rxq_map[i];
1523 ring->vsi = vsi;
1524 ring->netdev = vsi->netdev;
1525 ring->dev = dev;
1526 ring->count = vsi->num_rx_desc;
1527 ring->cached_phctime = pf->ptp.cached_phc_time;
1528 WRITE_ONCE(vsi->rx_rings[i], ring);
1529 }
1530
1531 return 0;
1532
1533 err_out:
1534 ice_vsi_clear_rings(vsi);
1535 return -ENOMEM;
1536 }
1537
1538 /**
1539 * ice_vsi_manage_rss_lut - disable/enable RSS
1540 * @vsi: the VSI being changed
1541 * @ena: boolean value indicating if this is an enable or disable request
1542 *
1543 * In the event of disable request for RSS, this function will zero out RSS
1544 * LUT, while in the event of enable request for RSS, it will reconfigure RSS
1545 * LUT.
1546 */
ice_vsi_manage_rss_lut(struct ice_vsi * vsi,bool ena)1547 void ice_vsi_manage_rss_lut(struct ice_vsi *vsi, bool ena)
1548 {
1549 u8 *lut;
1550
1551 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1552 if (!lut)
1553 return;
1554
1555 if (ena) {
1556 if (vsi->rss_lut_user)
1557 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1558 else
1559 ice_fill_rss_lut(lut, vsi->rss_table_size,
1560 vsi->rss_size);
1561 }
1562
1563 ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1564 kfree(lut);
1565 }
1566
1567 /**
1568 * ice_vsi_cfg_crc_strip - Configure CRC stripping for a VSI
1569 * @vsi: VSI to be configured
1570 * @disable: set to true to have FCS / CRC in the frame data
1571 */
ice_vsi_cfg_crc_strip(struct ice_vsi * vsi,bool disable)1572 void ice_vsi_cfg_crc_strip(struct ice_vsi *vsi, bool disable)
1573 {
1574 int i;
1575
1576 ice_for_each_rxq(vsi, i)
1577 if (disable)
1578 vsi->rx_rings[i]->flags |= ICE_RX_FLAGS_CRC_STRIP_DIS;
1579 else
1580 vsi->rx_rings[i]->flags &= ~ICE_RX_FLAGS_CRC_STRIP_DIS;
1581 }
1582
1583 /**
1584 * ice_vsi_cfg_rss_lut_key - Configure RSS params for a VSI
1585 * @vsi: VSI to be configured
1586 */
ice_vsi_cfg_rss_lut_key(struct ice_vsi * vsi)1587 int ice_vsi_cfg_rss_lut_key(struct ice_vsi *vsi)
1588 {
1589 struct ice_pf *pf = vsi->back;
1590 struct device *dev;
1591 u8 *lut, *key;
1592 int err;
1593
1594 dev = ice_pf_to_dev(pf);
1595 if (vsi->type == ICE_VSI_PF && vsi->ch_rss_size &&
1596 (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))) {
1597 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->ch_rss_size);
1598 } else {
1599 vsi->rss_size = min_t(u16, vsi->rss_size, vsi->num_rxq);
1600
1601 /* If orig_rss_size is valid and it is less than determined
1602 * main VSI's rss_size, update main VSI's rss_size to be
1603 * orig_rss_size so that when tc-qdisc is deleted, main VSI
1604 * RSS table gets programmed to be correct (whatever it was
1605 * to begin with (prior to setup-tc for ADQ config)
1606 */
1607 if (vsi->orig_rss_size && vsi->rss_size < vsi->orig_rss_size &&
1608 vsi->orig_rss_size <= vsi->num_rxq) {
1609 vsi->rss_size = vsi->orig_rss_size;
1610 /* now orig_rss_size is used, reset it to zero */
1611 vsi->orig_rss_size = 0;
1612 }
1613 }
1614
1615 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1616 if (!lut)
1617 return -ENOMEM;
1618
1619 if (vsi->rss_lut_user)
1620 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1621 else
1622 ice_fill_rss_lut(lut, vsi->rss_table_size, vsi->rss_size);
1623
1624 err = ice_set_rss_lut(vsi, lut, vsi->rss_table_size);
1625 if (err) {
1626 dev_err(dev, "set_rss_lut failed, error %d\n", err);
1627 goto ice_vsi_cfg_rss_exit;
1628 }
1629
1630 key = kzalloc(ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE, GFP_KERNEL);
1631 if (!key) {
1632 err = -ENOMEM;
1633 goto ice_vsi_cfg_rss_exit;
1634 }
1635
1636 if (vsi->rss_hkey_user)
1637 memcpy(key, vsi->rss_hkey_user, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1638 else
1639 netdev_rss_key_fill((void *)key, ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE);
1640
1641 err = ice_set_rss_key(vsi, key);
1642 if (err)
1643 dev_err(dev, "set_rss_key failed, error %d\n", err);
1644
1645 kfree(key);
1646 ice_vsi_cfg_rss_exit:
1647 kfree(lut);
1648 return err;
1649 }
1650
1651 /**
1652 * ice_vsi_set_vf_rss_flow_fld - Sets VF VSI RSS input set for different flows
1653 * @vsi: VSI to be configured
1654 *
1655 * This function will only be called during the VF VSI setup. Upon successful
1656 * completion of package download, this function will configure default RSS
1657 * input sets for VF VSI.
1658 */
ice_vsi_set_vf_rss_flow_fld(struct ice_vsi * vsi)1659 static void ice_vsi_set_vf_rss_flow_fld(struct ice_vsi *vsi)
1660 {
1661 struct ice_pf *pf = vsi->back;
1662 struct device *dev;
1663 int status;
1664
1665 dev = ice_pf_to_dev(pf);
1666 if (ice_is_safe_mode(pf)) {
1667 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1668 vsi->vsi_num);
1669 return;
1670 }
1671
1672 status = ice_add_avf_rss_cfg(&pf->hw, vsi->idx, ICE_DEFAULT_RSS_HENA);
1673 if (status)
1674 dev_dbg(dev, "ice_add_avf_rss_cfg failed for vsi = %d, error = %d\n",
1675 vsi->vsi_num, status);
1676 }
1677
1678 /**
1679 * ice_vsi_set_rss_flow_fld - Sets RSS input set for different flows
1680 * @vsi: VSI to be configured
1681 *
1682 * This function will only be called after successful download package call
1683 * during initialization of PF. Since the downloaded package will erase the
1684 * RSS section, this function will configure RSS input sets for different
1685 * flow types. The last profile added has the highest priority, therefore 2
1686 * tuple profiles (i.e. IPv4 src/dst) are added before 4 tuple profiles
1687 * (i.e. IPv4 src/dst TCP src/dst port).
1688 */
ice_vsi_set_rss_flow_fld(struct ice_vsi * vsi)1689 static void ice_vsi_set_rss_flow_fld(struct ice_vsi *vsi)
1690 {
1691 u16 vsi_handle = vsi->idx, vsi_num = vsi->vsi_num;
1692 struct ice_pf *pf = vsi->back;
1693 struct ice_hw *hw = &pf->hw;
1694 struct device *dev;
1695 int status;
1696
1697 dev = ice_pf_to_dev(pf);
1698 if (ice_is_safe_mode(pf)) {
1699 dev_dbg(dev, "Advanced RSS disabled. Package download failed, vsi num = %d\n",
1700 vsi_num);
1701 return;
1702 }
1703 /* configure RSS for IPv4 with input set IP src/dst */
1704 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1705 ICE_FLOW_SEG_HDR_IPV4);
1706 if (status)
1707 dev_dbg(dev, "ice_add_rss_cfg failed for ipv4 flow, vsi = %d, error = %d\n",
1708 vsi_num, status);
1709
1710 /* configure RSS for IPv6 with input set IPv6 src/dst */
1711 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1712 ICE_FLOW_SEG_HDR_IPV6);
1713 if (status)
1714 dev_dbg(dev, "ice_add_rss_cfg failed for ipv6 flow, vsi = %d, error = %d\n",
1715 vsi_num, status);
1716
1717 /* configure RSS for tcp4 with input set IP src/dst, TCP src/dst */
1718 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV4,
1719 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV4);
1720 if (status)
1721 dev_dbg(dev, "ice_add_rss_cfg failed for tcp4 flow, vsi = %d, error = %d\n",
1722 vsi_num, status);
1723
1724 /* configure RSS for udp4 with input set IP src/dst, UDP src/dst */
1725 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV4,
1726 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV4);
1727 if (status)
1728 dev_dbg(dev, "ice_add_rss_cfg failed for udp4 flow, vsi = %d, error = %d\n",
1729 vsi_num, status);
1730
1731 /* configure RSS for sctp4 with input set IP src/dst */
1732 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV4,
1733 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV4);
1734 if (status)
1735 dev_dbg(dev, "ice_add_rss_cfg failed for sctp4 flow, vsi = %d, error = %d\n",
1736 vsi_num, status);
1737
1738 /* configure RSS for tcp6 with input set IPv6 src/dst, TCP src/dst */
1739 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_TCP_IPV6,
1740 ICE_FLOW_SEG_HDR_TCP | ICE_FLOW_SEG_HDR_IPV6);
1741 if (status)
1742 dev_dbg(dev, "ice_add_rss_cfg failed for tcp6 flow, vsi = %d, error = %d\n",
1743 vsi_num, status);
1744
1745 /* configure RSS for udp6 with input set IPv6 src/dst, UDP src/dst */
1746 status = ice_add_rss_cfg(hw, vsi_handle, ICE_HASH_UDP_IPV6,
1747 ICE_FLOW_SEG_HDR_UDP | ICE_FLOW_SEG_HDR_IPV6);
1748 if (status)
1749 dev_dbg(dev, "ice_add_rss_cfg failed for udp6 flow, vsi = %d, error = %d\n",
1750 vsi_num, status);
1751
1752 /* configure RSS for sctp6 with input set IPv6 src/dst */
1753 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_IPV6,
1754 ICE_FLOW_SEG_HDR_SCTP | ICE_FLOW_SEG_HDR_IPV6);
1755 if (status)
1756 dev_dbg(dev, "ice_add_rss_cfg failed for sctp6 flow, vsi = %d, error = %d\n",
1757 vsi_num, status);
1758
1759 status = ice_add_rss_cfg(hw, vsi_handle, ICE_FLOW_HASH_ESP_SPI,
1760 ICE_FLOW_SEG_HDR_ESP);
1761 if (status)
1762 dev_dbg(dev, "ice_add_rss_cfg failed for esp/spi flow, vsi = %d, error = %d\n",
1763 vsi_num, status);
1764 }
1765
1766 /**
1767 * ice_pf_state_is_nominal - checks the PF for nominal state
1768 * @pf: pointer to PF to check
1769 *
1770 * Check the PF's state for a collection of bits that would indicate
1771 * the PF is in a state that would inhibit normal operation for
1772 * driver functionality.
1773 *
1774 * Returns true if PF is in a nominal state, false otherwise
1775 */
ice_pf_state_is_nominal(struct ice_pf * pf)1776 bool ice_pf_state_is_nominal(struct ice_pf *pf)
1777 {
1778 DECLARE_BITMAP(check_bits, ICE_STATE_NBITS) = { 0 };
1779
1780 if (!pf)
1781 return false;
1782
1783 bitmap_set(check_bits, 0, ICE_STATE_NOMINAL_CHECK_BITS);
1784 if (bitmap_intersects(pf->state, check_bits, ICE_STATE_NBITS))
1785 return false;
1786
1787 return true;
1788 }
1789
1790 /**
1791 * ice_update_eth_stats - Update VSI-specific ethernet statistics counters
1792 * @vsi: the VSI to be updated
1793 */
ice_update_eth_stats(struct ice_vsi * vsi)1794 void ice_update_eth_stats(struct ice_vsi *vsi)
1795 {
1796 struct ice_eth_stats *prev_es, *cur_es;
1797 struct ice_hw *hw = &vsi->back->hw;
1798 u16 vsi_num = vsi->vsi_num; /* HW absolute index of a VSI */
1799
1800 prev_es = &vsi->eth_stats_prev;
1801 cur_es = &vsi->eth_stats;
1802
1803 ice_stat_update40(hw, GLV_GORCL(vsi_num), vsi->stat_offsets_loaded,
1804 &prev_es->rx_bytes, &cur_es->rx_bytes);
1805
1806 ice_stat_update40(hw, GLV_UPRCL(vsi_num), vsi->stat_offsets_loaded,
1807 &prev_es->rx_unicast, &cur_es->rx_unicast);
1808
1809 ice_stat_update40(hw, GLV_MPRCL(vsi_num), vsi->stat_offsets_loaded,
1810 &prev_es->rx_multicast, &cur_es->rx_multicast);
1811
1812 ice_stat_update40(hw, GLV_BPRCL(vsi_num), vsi->stat_offsets_loaded,
1813 &prev_es->rx_broadcast, &cur_es->rx_broadcast);
1814
1815 ice_stat_update32(hw, GLV_RDPC(vsi_num), vsi->stat_offsets_loaded,
1816 &prev_es->rx_discards, &cur_es->rx_discards);
1817
1818 ice_stat_update40(hw, GLV_GOTCL(vsi_num), vsi->stat_offsets_loaded,
1819 &prev_es->tx_bytes, &cur_es->tx_bytes);
1820
1821 ice_stat_update40(hw, GLV_UPTCL(vsi_num), vsi->stat_offsets_loaded,
1822 &prev_es->tx_unicast, &cur_es->tx_unicast);
1823
1824 ice_stat_update40(hw, GLV_MPTCL(vsi_num), vsi->stat_offsets_loaded,
1825 &prev_es->tx_multicast, &cur_es->tx_multicast);
1826
1827 ice_stat_update40(hw, GLV_BPTCL(vsi_num), vsi->stat_offsets_loaded,
1828 &prev_es->tx_broadcast, &cur_es->tx_broadcast);
1829
1830 ice_stat_update32(hw, GLV_TEPC(vsi_num), vsi->stat_offsets_loaded,
1831 &prev_es->tx_errors, &cur_es->tx_errors);
1832
1833 vsi->stat_offsets_loaded = true;
1834 }
1835
1836 /**
1837 * ice_vsi_cfg_frame_size - setup max frame size and Rx buffer length
1838 * @vsi: VSI
1839 */
ice_vsi_cfg_frame_size(struct ice_vsi * vsi)1840 void ice_vsi_cfg_frame_size(struct ice_vsi *vsi)
1841 {
1842 if (!vsi->netdev || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags)) {
1843 vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1844 vsi->rx_buf_len = ICE_RXBUF_2048;
1845 #if (PAGE_SIZE < 8192)
1846 } else if (!ICE_2K_TOO_SMALL_WITH_PADDING &&
1847 (vsi->netdev->mtu <= ETH_DATA_LEN)) {
1848 vsi->max_frame = ICE_RXBUF_1536 - NET_IP_ALIGN;
1849 vsi->rx_buf_len = ICE_RXBUF_1536 - NET_IP_ALIGN;
1850 #endif
1851 } else {
1852 vsi->max_frame = ICE_AQ_SET_MAC_FRAME_SIZE_MAX;
1853 #if (PAGE_SIZE < 8192)
1854 vsi->rx_buf_len = ICE_RXBUF_3072;
1855 #else
1856 vsi->rx_buf_len = ICE_RXBUF_2048;
1857 #endif
1858 }
1859 }
1860
1861 /**
1862 * ice_write_qrxflxp_cntxt - write/configure QRXFLXP_CNTXT register
1863 * @hw: HW pointer
1864 * @pf_q: index of the Rx queue in the PF's queue space
1865 * @rxdid: flexible descriptor RXDID
1866 * @prio: priority for the RXDID for this queue
1867 * @ena_ts: true to enable timestamp and false to disable timestamp
1868 */
1869 void
ice_write_qrxflxp_cntxt(struct ice_hw * hw,u16 pf_q,u32 rxdid,u32 prio,bool ena_ts)1870 ice_write_qrxflxp_cntxt(struct ice_hw *hw, u16 pf_q, u32 rxdid, u32 prio,
1871 bool ena_ts)
1872 {
1873 int regval = rd32(hw, QRXFLXP_CNTXT(pf_q));
1874
1875 /* clear any previous values */
1876 regval &= ~(QRXFLXP_CNTXT_RXDID_IDX_M |
1877 QRXFLXP_CNTXT_RXDID_PRIO_M |
1878 QRXFLXP_CNTXT_TS_M);
1879
1880 regval |= (rxdid << QRXFLXP_CNTXT_RXDID_IDX_S) &
1881 QRXFLXP_CNTXT_RXDID_IDX_M;
1882
1883 regval |= (prio << QRXFLXP_CNTXT_RXDID_PRIO_S) &
1884 QRXFLXP_CNTXT_RXDID_PRIO_M;
1885
1886 if (ena_ts)
1887 /* Enable TimeSync on this queue */
1888 regval |= QRXFLXP_CNTXT_TS_M;
1889
1890 wr32(hw, QRXFLXP_CNTXT(pf_q), regval);
1891 }
1892
ice_vsi_cfg_single_rxq(struct ice_vsi * vsi,u16 q_idx)1893 int ice_vsi_cfg_single_rxq(struct ice_vsi *vsi, u16 q_idx)
1894 {
1895 if (q_idx >= vsi->num_rxq)
1896 return -EINVAL;
1897
1898 return ice_vsi_cfg_rxq(vsi->rx_rings[q_idx]);
1899 }
1900
ice_vsi_cfg_single_txq(struct ice_vsi * vsi,struct ice_tx_ring ** tx_rings,u16 q_idx)1901 int ice_vsi_cfg_single_txq(struct ice_vsi *vsi, struct ice_tx_ring **tx_rings, u16 q_idx)
1902 {
1903 struct ice_aqc_add_tx_qgrp *qg_buf;
1904 int err;
1905
1906 if (q_idx >= vsi->alloc_txq || !tx_rings || !tx_rings[q_idx])
1907 return -EINVAL;
1908
1909 qg_buf = kzalloc(struct_size(qg_buf, txqs, 1), GFP_KERNEL);
1910 if (!qg_buf)
1911 return -ENOMEM;
1912
1913 qg_buf->num_txqs = 1;
1914
1915 err = ice_vsi_cfg_txq(vsi, tx_rings[q_idx], qg_buf);
1916 kfree(qg_buf);
1917 return err;
1918 }
1919
1920 /**
1921 * ice_vsi_cfg_rxqs - Configure the VSI for Rx
1922 * @vsi: the VSI being configured
1923 *
1924 * Return 0 on success and a negative value on error
1925 * Configure the Rx VSI for operation.
1926 */
ice_vsi_cfg_rxqs(struct ice_vsi * vsi)1927 int ice_vsi_cfg_rxqs(struct ice_vsi *vsi)
1928 {
1929 u16 i;
1930
1931 if (vsi->type == ICE_VSI_VF)
1932 goto setup_rings;
1933
1934 ice_vsi_cfg_frame_size(vsi);
1935 setup_rings:
1936 /* set up individual rings */
1937 ice_for_each_rxq(vsi, i) {
1938 int err = ice_vsi_cfg_rxq(vsi->rx_rings[i]);
1939
1940 if (err)
1941 return err;
1942 }
1943
1944 return 0;
1945 }
1946
1947 /**
1948 * ice_vsi_cfg_txqs - Configure the VSI for Tx
1949 * @vsi: the VSI being configured
1950 * @rings: Tx ring array to be configured
1951 * @count: number of Tx ring array elements
1952 *
1953 * Return 0 on success and a negative value on error
1954 * Configure the Tx VSI for operation.
1955 */
1956 static int
ice_vsi_cfg_txqs(struct ice_vsi * vsi,struct ice_tx_ring ** rings,u16 count)1957 ice_vsi_cfg_txqs(struct ice_vsi *vsi, struct ice_tx_ring **rings, u16 count)
1958 {
1959 struct ice_aqc_add_tx_qgrp *qg_buf;
1960 u16 q_idx = 0;
1961 int err = 0;
1962
1963 qg_buf = kzalloc(struct_size(qg_buf, txqs, 1), GFP_KERNEL);
1964 if (!qg_buf)
1965 return -ENOMEM;
1966
1967 qg_buf->num_txqs = 1;
1968
1969 for (q_idx = 0; q_idx < count; q_idx++) {
1970 err = ice_vsi_cfg_txq(vsi, rings[q_idx], qg_buf);
1971 if (err)
1972 goto err_cfg_txqs;
1973 }
1974
1975 err_cfg_txqs:
1976 kfree(qg_buf);
1977 return err;
1978 }
1979
1980 /**
1981 * ice_vsi_cfg_lan_txqs - Configure the VSI for Tx
1982 * @vsi: the VSI being configured
1983 *
1984 * Return 0 on success and a negative value on error
1985 * Configure the Tx VSI for operation.
1986 */
ice_vsi_cfg_lan_txqs(struct ice_vsi * vsi)1987 int ice_vsi_cfg_lan_txqs(struct ice_vsi *vsi)
1988 {
1989 return ice_vsi_cfg_txqs(vsi, vsi->tx_rings, vsi->num_txq);
1990 }
1991
1992 /**
1993 * ice_vsi_cfg_xdp_txqs - Configure Tx queues dedicated for XDP in given VSI
1994 * @vsi: the VSI being configured
1995 *
1996 * Return 0 on success and a negative value on error
1997 * Configure the Tx queues dedicated for XDP in given VSI for operation.
1998 */
ice_vsi_cfg_xdp_txqs(struct ice_vsi * vsi)1999 int ice_vsi_cfg_xdp_txqs(struct ice_vsi *vsi)
2000 {
2001 int ret;
2002 int i;
2003
2004 ret = ice_vsi_cfg_txqs(vsi, vsi->xdp_rings, vsi->num_xdp_txq);
2005 if (ret)
2006 return ret;
2007
2008 ice_for_each_rxq(vsi, i)
2009 ice_tx_xsk_pool(vsi, i);
2010
2011 return ret;
2012 }
2013
2014 /**
2015 * ice_intrl_usec_to_reg - convert interrupt rate limit to register value
2016 * @intrl: interrupt rate limit in usecs
2017 * @gran: interrupt rate limit granularity in usecs
2018 *
2019 * This function converts a decimal interrupt rate limit in usecs to the format
2020 * expected by firmware.
2021 */
ice_intrl_usec_to_reg(u8 intrl,u8 gran)2022 static u32 ice_intrl_usec_to_reg(u8 intrl, u8 gran)
2023 {
2024 u32 val = intrl / gran;
2025
2026 if (val)
2027 return val | GLINT_RATE_INTRL_ENA_M;
2028 return 0;
2029 }
2030
2031 /**
2032 * ice_write_intrl - write throttle rate limit to interrupt specific register
2033 * @q_vector: pointer to interrupt specific structure
2034 * @intrl: throttle rate limit in microseconds to write
2035 */
ice_write_intrl(struct ice_q_vector * q_vector,u8 intrl)2036 void ice_write_intrl(struct ice_q_vector *q_vector, u8 intrl)
2037 {
2038 struct ice_hw *hw = &q_vector->vsi->back->hw;
2039
2040 wr32(hw, GLINT_RATE(q_vector->reg_idx),
2041 ice_intrl_usec_to_reg(intrl, ICE_INTRL_GRAN_ABOVE_25));
2042 }
2043
ice_pull_qvec_from_rc(struct ice_ring_container * rc)2044 static struct ice_q_vector *ice_pull_qvec_from_rc(struct ice_ring_container *rc)
2045 {
2046 switch (rc->type) {
2047 case ICE_RX_CONTAINER:
2048 if (rc->rx_ring)
2049 return rc->rx_ring->q_vector;
2050 break;
2051 case ICE_TX_CONTAINER:
2052 if (rc->tx_ring)
2053 return rc->tx_ring->q_vector;
2054 break;
2055 default:
2056 break;
2057 }
2058
2059 return NULL;
2060 }
2061
2062 /**
2063 * __ice_write_itr - write throttle rate to register
2064 * @q_vector: pointer to interrupt data structure
2065 * @rc: pointer to ring container
2066 * @itr: throttle rate in microseconds to write
2067 */
__ice_write_itr(struct ice_q_vector * q_vector,struct ice_ring_container * rc,u16 itr)2068 static void __ice_write_itr(struct ice_q_vector *q_vector,
2069 struct ice_ring_container *rc, u16 itr)
2070 {
2071 struct ice_hw *hw = &q_vector->vsi->back->hw;
2072
2073 wr32(hw, GLINT_ITR(rc->itr_idx, q_vector->reg_idx),
2074 ITR_REG_ALIGN(itr) >> ICE_ITR_GRAN_S);
2075 }
2076
2077 /**
2078 * ice_write_itr - write throttle rate to queue specific register
2079 * @rc: pointer to ring container
2080 * @itr: throttle rate in microseconds to write
2081 */
ice_write_itr(struct ice_ring_container * rc,u16 itr)2082 void ice_write_itr(struct ice_ring_container *rc, u16 itr)
2083 {
2084 struct ice_q_vector *q_vector;
2085
2086 q_vector = ice_pull_qvec_from_rc(rc);
2087 if (!q_vector)
2088 return;
2089
2090 __ice_write_itr(q_vector, rc, itr);
2091 }
2092
2093 /**
2094 * ice_set_q_vector_intrl - set up interrupt rate limiting
2095 * @q_vector: the vector to be configured
2096 *
2097 * Interrupt rate limiting is local to the vector, not per-queue so we must
2098 * detect if either ring container has dynamic moderation enabled to decide
2099 * what to set the interrupt rate limit to via INTRL settings. In the case that
2100 * dynamic moderation is disabled on both, write the value with the cached
2101 * setting to make sure INTRL register matches the user visible value.
2102 */
ice_set_q_vector_intrl(struct ice_q_vector * q_vector)2103 void ice_set_q_vector_intrl(struct ice_q_vector *q_vector)
2104 {
2105 if (ITR_IS_DYNAMIC(&q_vector->tx) || ITR_IS_DYNAMIC(&q_vector->rx)) {
2106 /* in the case of dynamic enabled, cap each vector to no more
2107 * than (4 us) 250,000 ints/sec, which allows low latency
2108 * but still less than 500,000 interrupts per second, which
2109 * reduces CPU a bit in the case of the lowest latency
2110 * setting. The 4 here is a value in microseconds.
2111 */
2112 ice_write_intrl(q_vector, 4);
2113 } else {
2114 ice_write_intrl(q_vector, q_vector->intrl);
2115 }
2116 }
2117
2118 /**
2119 * ice_vsi_cfg_msix - MSIX mode Interrupt Config in the HW
2120 * @vsi: the VSI being configured
2121 *
2122 * This configures MSIX mode interrupts for the PF VSI, and should not be used
2123 * for the VF VSI.
2124 */
ice_vsi_cfg_msix(struct ice_vsi * vsi)2125 void ice_vsi_cfg_msix(struct ice_vsi *vsi)
2126 {
2127 struct ice_pf *pf = vsi->back;
2128 struct ice_hw *hw = &pf->hw;
2129 u16 txq = 0, rxq = 0;
2130 int i, q;
2131
2132 ice_for_each_q_vector(vsi, i) {
2133 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2134 u16 reg_idx = q_vector->reg_idx;
2135
2136 ice_cfg_itr(hw, q_vector);
2137
2138 /* Both Transmit Queue Interrupt Cause Control register
2139 * and Receive Queue Interrupt Cause control register
2140 * expects MSIX_INDX field to be the vector index
2141 * within the function space and not the absolute
2142 * vector index across PF or across device.
2143 * For SR-IOV VF VSIs queue vector index always starts
2144 * with 1 since first vector index(0) is used for OICR
2145 * in VF space. Since VMDq and other PF VSIs are within
2146 * the PF function space, use the vector index that is
2147 * tracked for this PF.
2148 */
2149 for (q = 0; q < q_vector->num_ring_tx; q++) {
2150 ice_cfg_txq_interrupt(vsi, txq, reg_idx,
2151 q_vector->tx.itr_idx);
2152 txq++;
2153 }
2154
2155 for (q = 0; q < q_vector->num_ring_rx; q++) {
2156 ice_cfg_rxq_interrupt(vsi, rxq, reg_idx,
2157 q_vector->rx.itr_idx);
2158 rxq++;
2159 }
2160 }
2161 }
2162
2163 /**
2164 * ice_vsi_start_all_rx_rings - start/enable all of a VSI's Rx rings
2165 * @vsi: the VSI whose rings are to be enabled
2166 *
2167 * Returns 0 on success and a negative value on error
2168 */
ice_vsi_start_all_rx_rings(struct ice_vsi * vsi)2169 int ice_vsi_start_all_rx_rings(struct ice_vsi *vsi)
2170 {
2171 return ice_vsi_ctrl_all_rx_rings(vsi, true);
2172 }
2173
2174 /**
2175 * ice_vsi_stop_all_rx_rings - stop/disable all of a VSI's Rx rings
2176 * @vsi: the VSI whose rings are to be disabled
2177 *
2178 * Returns 0 on success and a negative value on error
2179 */
ice_vsi_stop_all_rx_rings(struct ice_vsi * vsi)2180 int ice_vsi_stop_all_rx_rings(struct ice_vsi *vsi)
2181 {
2182 return ice_vsi_ctrl_all_rx_rings(vsi, false);
2183 }
2184
2185 /**
2186 * ice_vsi_stop_tx_rings - Disable Tx rings
2187 * @vsi: the VSI being configured
2188 * @rst_src: reset source
2189 * @rel_vmvf_num: Relative ID of VF/VM
2190 * @rings: Tx ring array to be stopped
2191 * @count: number of Tx ring array elements
2192 */
2193 static int
ice_vsi_stop_tx_rings(struct ice_vsi * vsi,enum ice_disq_rst_src rst_src,u16 rel_vmvf_num,struct ice_tx_ring ** rings,u16 count)2194 ice_vsi_stop_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2195 u16 rel_vmvf_num, struct ice_tx_ring **rings, u16 count)
2196 {
2197 u16 q_idx;
2198
2199 if (vsi->num_txq > ICE_LAN_TXQ_MAX_QDIS)
2200 return -EINVAL;
2201
2202 for (q_idx = 0; q_idx < count; q_idx++) {
2203 struct ice_txq_meta txq_meta = { };
2204 int status;
2205
2206 if (!rings || !rings[q_idx])
2207 return -EINVAL;
2208
2209 ice_fill_txq_meta(vsi, rings[q_idx], &txq_meta);
2210 status = ice_vsi_stop_tx_ring(vsi, rst_src, rel_vmvf_num,
2211 rings[q_idx], &txq_meta);
2212
2213 if (status)
2214 return status;
2215 }
2216
2217 return 0;
2218 }
2219
2220 /**
2221 * ice_vsi_stop_lan_tx_rings - Disable LAN Tx rings
2222 * @vsi: the VSI being configured
2223 * @rst_src: reset source
2224 * @rel_vmvf_num: Relative ID of VF/VM
2225 */
2226 int
ice_vsi_stop_lan_tx_rings(struct ice_vsi * vsi,enum ice_disq_rst_src rst_src,u16 rel_vmvf_num)2227 ice_vsi_stop_lan_tx_rings(struct ice_vsi *vsi, enum ice_disq_rst_src rst_src,
2228 u16 rel_vmvf_num)
2229 {
2230 return ice_vsi_stop_tx_rings(vsi, rst_src, rel_vmvf_num, vsi->tx_rings, vsi->num_txq);
2231 }
2232
2233 /**
2234 * ice_vsi_stop_xdp_tx_rings - Disable XDP Tx rings
2235 * @vsi: the VSI being configured
2236 */
ice_vsi_stop_xdp_tx_rings(struct ice_vsi * vsi)2237 int ice_vsi_stop_xdp_tx_rings(struct ice_vsi *vsi)
2238 {
2239 return ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0, vsi->xdp_rings, vsi->num_xdp_txq);
2240 }
2241
2242 /**
2243 * ice_vsi_is_rx_queue_active
2244 * @vsi: the VSI being configured
2245 *
2246 * Return true if at least one queue is active.
2247 */
ice_vsi_is_rx_queue_active(struct ice_vsi * vsi)2248 bool ice_vsi_is_rx_queue_active(struct ice_vsi *vsi)
2249 {
2250 struct ice_pf *pf = vsi->back;
2251 struct ice_hw *hw = &pf->hw;
2252 int i;
2253
2254 ice_for_each_rxq(vsi, i) {
2255 u32 rx_reg;
2256 int pf_q;
2257
2258 pf_q = vsi->rxq_map[i];
2259 rx_reg = rd32(hw, QRX_CTRL(pf_q));
2260 if (rx_reg & QRX_CTRL_QENA_STAT_M)
2261 return true;
2262 }
2263
2264 return false;
2265 }
2266
2267 /**
2268 * ice_vsi_is_vlan_pruning_ena - check if VLAN pruning is enabled or not
2269 * @vsi: VSI to check whether or not VLAN pruning is enabled.
2270 *
2271 * returns true if Rx VLAN pruning is enabled and false otherwise.
2272 */
ice_vsi_is_vlan_pruning_ena(struct ice_vsi * vsi)2273 bool ice_vsi_is_vlan_pruning_ena(struct ice_vsi *vsi)
2274 {
2275 if (!vsi)
2276 return false;
2277
2278 return (vsi->info.sw_flags2 & ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA);
2279 }
2280
ice_vsi_set_tc_cfg(struct ice_vsi * vsi)2281 static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
2282 {
2283 if (!test_bit(ICE_FLAG_DCB_ENA, vsi->back->flags)) {
2284 vsi->tc_cfg.ena_tc = ICE_DFLT_TRAFFIC_CLASS;
2285 vsi->tc_cfg.numtc = 1;
2286 return;
2287 }
2288
2289 /* set VSI TC information based on DCB config */
2290 ice_vsi_set_dcb_tc_cfg(vsi);
2291 }
2292
2293 /**
2294 * ice_vsi_set_q_vectors_reg_idx - set the HW register index for all q_vectors
2295 * @vsi: VSI to set the q_vectors register index on
2296 */
2297 static int
ice_vsi_set_q_vectors_reg_idx(struct ice_vsi * vsi)2298 ice_vsi_set_q_vectors_reg_idx(struct ice_vsi *vsi)
2299 {
2300 u16 i;
2301
2302 if (!vsi || !vsi->q_vectors)
2303 return -EINVAL;
2304
2305 ice_for_each_q_vector(vsi, i) {
2306 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2307
2308 if (!q_vector) {
2309 dev_err(ice_pf_to_dev(vsi->back), "Failed to set reg_idx on q_vector %d VSI %d\n",
2310 i, vsi->vsi_num);
2311 goto clear_reg_idx;
2312 }
2313
2314 if (vsi->type == ICE_VSI_VF) {
2315 struct ice_vf *vf = vsi->vf;
2316
2317 q_vector->reg_idx = ice_calc_vf_reg_idx(vf, q_vector);
2318 } else {
2319 q_vector->reg_idx =
2320 q_vector->v_idx + vsi->base_vector;
2321 }
2322 }
2323
2324 return 0;
2325
2326 clear_reg_idx:
2327 ice_for_each_q_vector(vsi, i) {
2328 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2329
2330 if (q_vector)
2331 q_vector->reg_idx = 0;
2332 }
2333
2334 return -EINVAL;
2335 }
2336
2337 /**
2338 * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
2339 * @vsi: the VSI being configured
2340 * @tx: bool to determine Tx or Rx rule
2341 * @create: bool to determine create or remove Rule
2342 */
ice_cfg_sw_lldp(struct ice_vsi * vsi,bool tx,bool create)2343 void ice_cfg_sw_lldp(struct ice_vsi *vsi, bool tx, bool create)
2344 {
2345 int (*eth_fltr)(struct ice_vsi *v, u16 type, u16 flag,
2346 enum ice_sw_fwd_act_type act);
2347 struct ice_pf *pf = vsi->back;
2348 struct device *dev;
2349 int status;
2350
2351 dev = ice_pf_to_dev(pf);
2352 eth_fltr = create ? ice_fltr_add_eth : ice_fltr_remove_eth;
2353
2354 if (tx) {
2355 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_TX,
2356 ICE_DROP_PACKET);
2357 } else {
2358 if (ice_fw_supports_lldp_fltr_ctrl(&pf->hw)) {
2359 status = ice_lldp_fltr_add_remove(&pf->hw, vsi->vsi_num,
2360 create);
2361 } else {
2362 status = eth_fltr(vsi, ETH_P_LLDP, ICE_FLTR_RX,
2363 ICE_FWD_TO_VSI);
2364 }
2365 }
2366
2367 if (status)
2368 dev_dbg(dev, "Fail %s %s LLDP rule on VSI %i error: %d\n",
2369 create ? "adding" : "removing", tx ? "TX" : "RX",
2370 vsi->vsi_num, status);
2371 }
2372
2373 /**
2374 * ice_set_agg_vsi - sets up scheduler aggregator node and move VSI into it
2375 * @vsi: pointer to the VSI
2376 *
2377 * This function will allocate new scheduler aggregator now if needed and will
2378 * move specified VSI into it.
2379 */
ice_set_agg_vsi(struct ice_vsi * vsi)2380 static void ice_set_agg_vsi(struct ice_vsi *vsi)
2381 {
2382 struct device *dev = ice_pf_to_dev(vsi->back);
2383 struct ice_agg_node *agg_node_iter = NULL;
2384 u32 agg_id = ICE_INVALID_AGG_NODE_ID;
2385 struct ice_agg_node *agg_node = NULL;
2386 int node_offset, max_agg_nodes = 0;
2387 struct ice_port_info *port_info;
2388 struct ice_pf *pf = vsi->back;
2389 u32 agg_node_id_start = 0;
2390 int status;
2391
2392 /* create (as needed) scheduler aggregator node and move VSI into
2393 * corresponding aggregator node
2394 * - PF aggregator node to contains VSIs of type _PF and _CTRL
2395 * - VF aggregator nodes will contain VF VSI
2396 */
2397 port_info = pf->hw.port_info;
2398 if (!port_info)
2399 return;
2400
2401 switch (vsi->type) {
2402 case ICE_VSI_CTRL:
2403 case ICE_VSI_CHNL:
2404 case ICE_VSI_LB:
2405 case ICE_VSI_PF:
2406 case ICE_VSI_SWITCHDEV_CTRL:
2407 max_agg_nodes = ICE_MAX_PF_AGG_NODES;
2408 agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
2409 agg_node_iter = &pf->pf_agg_node[0];
2410 break;
2411 case ICE_VSI_VF:
2412 /* user can create 'n' VFs on a given PF, but since max children
2413 * per aggregator node can be only 64. Following code handles
2414 * aggregator(s) for VF VSIs, either selects a agg_node which
2415 * was already created provided num_vsis < 64, otherwise
2416 * select next available node, which will be created
2417 */
2418 max_agg_nodes = ICE_MAX_VF_AGG_NODES;
2419 agg_node_id_start = ICE_VF_AGG_NODE_ID_START;
2420 agg_node_iter = &pf->vf_agg_node[0];
2421 break;
2422 default:
2423 /* other VSI type, handle later if needed */
2424 dev_dbg(dev, "unexpected VSI type %s\n",
2425 ice_vsi_type_str(vsi->type));
2426 return;
2427 }
2428
2429 /* find the appropriate aggregator node */
2430 for (node_offset = 0; node_offset < max_agg_nodes; node_offset++) {
2431 /* see if we can find space in previously created
2432 * node if num_vsis < 64, otherwise skip
2433 */
2434 if (agg_node_iter->num_vsis &&
2435 agg_node_iter->num_vsis == ICE_MAX_VSIS_IN_AGG_NODE) {
2436 agg_node_iter++;
2437 continue;
2438 }
2439
2440 if (agg_node_iter->valid &&
2441 agg_node_iter->agg_id != ICE_INVALID_AGG_NODE_ID) {
2442 agg_id = agg_node_iter->agg_id;
2443 agg_node = agg_node_iter;
2444 break;
2445 }
2446
2447 /* find unclaimed agg_id */
2448 if (agg_node_iter->agg_id == ICE_INVALID_AGG_NODE_ID) {
2449 agg_id = node_offset + agg_node_id_start;
2450 agg_node = agg_node_iter;
2451 break;
2452 }
2453 /* move to next agg_node */
2454 agg_node_iter++;
2455 }
2456
2457 if (!agg_node)
2458 return;
2459
2460 /* if selected aggregator node was not created, create it */
2461 if (!agg_node->valid) {
2462 status = ice_cfg_agg(port_info, agg_id, ICE_AGG_TYPE_AGG,
2463 (u8)vsi->tc_cfg.ena_tc);
2464 if (status) {
2465 dev_err(dev, "unable to create aggregator node with agg_id %u\n",
2466 agg_id);
2467 return;
2468 }
2469 /* aggregator node is created, store the needed info */
2470 agg_node->valid = true;
2471 agg_node->agg_id = agg_id;
2472 }
2473
2474 /* move VSI to corresponding aggregator node */
2475 status = ice_move_vsi_to_agg(port_info, agg_id, vsi->idx,
2476 (u8)vsi->tc_cfg.ena_tc);
2477 if (status) {
2478 dev_err(dev, "unable to move VSI idx %u into aggregator %u node",
2479 vsi->idx, agg_id);
2480 return;
2481 }
2482
2483 /* keep active children count for aggregator node */
2484 agg_node->num_vsis++;
2485
2486 /* cache the 'agg_id' in VSI, so that after reset - VSI will be moved
2487 * to aggregator node
2488 */
2489 vsi->agg_node = agg_node;
2490 dev_dbg(dev, "successfully moved VSI idx %u tc_bitmap 0x%x) into aggregator node %d which has num_vsis %u\n",
2491 vsi->idx, vsi->tc_cfg.ena_tc, vsi->agg_node->agg_id,
2492 vsi->agg_node->num_vsis);
2493 }
2494
2495 /**
2496 * ice_vsi_setup - Set up a VSI by a given type
2497 * @pf: board private structure
2498 * @pi: pointer to the port_info instance
2499 * @vsi_type: VSI type
2500 * @vf: pointer to VF to which this VSI connects. This field is used primarily
2501 * for the ICE_VSI_VF type. Other VSI types should pass NULL.
2502 * @ch: ptr to channel
2503 *
2504 * This allocates the sw VSI structure and its queue resources.
2505 *
2506 * Returns pointer to the successfully allocated and configured VSI sw struct on
2507 * success, NULL on failure.
2508 */
2509 struct ice_vsi *
ice_vsi_setup(struct ice_pf * pf,struct ice_port_info * pi,enum ice_vsi_type vsi_type,struct ice_vf * vf,struct ice_channel * ch)2510 ice_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
2511 enum ice_vsi_type vsi_type, struct ice_vf *vf,
2512 struct ice_channel *ch)
2513 {
2514 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2515 struct device *dev = ice_pf_to_dev(pf);
2516 struct ice_vsi *vsi;
2517 int ret, i;
2518
2519 if (vsi_type == ICE_VSI_CHNL)
2520 vsi = ice_vsi_alloc(pf, vsi_type, ch, NULL);
2521 else if (vsi_type == ICE_VSI_VF || vsi_type == ICE_VSI_CTRL)
2522 vsi = ice_vsi_alloc(pf, vsi_type, NULL, vf);
2523 else
2524 vsi = ice_vsi_alloc(pf, vsi_type, NULL, NULL);
2525
2526 if (!vsi) {
2527 dev_err(dev, "could not allocate VSI\n");
2528 return NULL;
2529 }
2530
2531 vsi->port_info = pi;
2532 vsi->vsw = pf->first_sw;
2533 if (vsi->type == ICE_VSI_PF)
2534 vsi->ethtype = ETH_P_PAUSE;
2535
2536 ice_alloc_fd_res(vsi);
2537
2538 if (vsi_type != ICE_VSI_CHNL) {
2539 if (ice_vsi_get_qs(vsi)) {
2540 dev_err(dev, "Failed to allocate queues. vsi->idx = %d\n",
2541 vsi->idx);
2542 goto unroll_vsi_alloc;
2543 }
2544 }
2545
2546 /* set RSS capabilities */
2547 ice_vsi_set_rss_params(vsi);
2548
2549 /* set TC configuration */
2550 ice_vsi_set_tc_cfg(vsi);
2551
2552 /* create the VSI */
2553 ret = ice_vsi_init(vsi, true);
2554 if (ret)
2555 goto unroll_get_qs;
2556
2557 ice_vsi_init_vlan_ops(vsi);
2558
2559 switch (vsi->type) {
2560 case ICE_VSI_CTRL:
2561 case ICE_VSI_SWITCHDEV_CTRL:
2562 case ICE_VSI_PF:
2563 ret = ice_vsi_alloc_q_vectors(vsi);
2564 if (ret)
2565 goto unroll_vsi_init;
2566
2567 ret = ice_vsi_setup_vector_base(vsi);
2568 if (ret)
2569 goto unroll_alloc_q_vector;
2570
2571 ret = ice_vsi_set_q_vectors_reg_idx(vsi);
2572 if (ret)
2573 goto unroll_vector_base;
2574
2575 ret = ice_vsi_alloc_rings(vsi);
2576 if (ret)
2577 goto unroll_vector_base;
2578
2579 ice_vsi_map_rings_to_vectors(vsi);
2580
2581 /* ICE_VSI_CTRL does not need RSS so skip RSS processing */
2582 if (vsi->type != ICE_VSI_CTRL)
2583 /* Do not exit if configuring RSS had an issue, at
2584 * least receive traffic on first queue. Hence no
2585 * need to capture return value
2586 */
2587 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2588 ice_vsi_cfg_rss_lut_key(vsi);
2589 ice_vsi_set_rss_flow_fld(vsi);
2590 }
2591 ice_init_arfs(vsi);
2592 break;
2593 case ICE_VSI_CHNL:
2594 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2595 ice_vsi_cfg_rss_lut_key(vsi);
2596 ice_vsi_set_rss_flow_fld(vsi);
2597 }
2598 break;
2599 case ICE_VSI_VF:
2600 /* VF driver will take care of creating netdev for this type and
2601 * map queues to vectors through Virtchnl, PF driver only
2602 * creates a VSI and corresponding structures for bookkeeping
2603 * purpose
2604 */
2605 ret = ice_vsi_alloc_q_vectors(vsi);
2606 if (ret)
2607 goto unroll_vsi_init;
2608
2609 ret = ice_vsi_alloc_rings(vsi);
2610 if (ret)
2611 goto unroll_alloc_q_vector;
2612
2613 ret = ice_vsi_set_q_vectors_reg_idx(vsi);
2614 if (ret)
2615 goto unroll_vector_base;
2616
2617 /* Do not exit if configuring RSS had an issue, at least
2618 * receive traffic on first queue. Hence no need to capture
2619 * return value
2620 */
2621 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
2622 ice_vsi_cfg_rss_lut_key(vsi);
2623 ice_vsi_set_vf_rss_flow_fld(vsi);
2624 }
2625 break;
2626 case ICE_VSI_LB:
2627 ret = ice_vsi_alloc_rings(vsi);
2628 if (ret)
2629 goto unroll_vsi_init;
2630 break;
2631 default:
2632 /* clean up the resources and exit */
2633 goto unroll_vsi_init;
2634 }
2635
2636 /* configure VSI nodes based on number of queues and TC's */
2637 ice_for_each_traffic_class(i) {
2638 if (!(vsi->tc_cfg.ena_tc & BIT(i)))
2639 continue;
2640
2641 if (vsi->type == ICE_VSI_CHNL) {
2642 if (!vsi->alloc_txq && vsi->num_txq)
2643 max_txqs[i] = vsi->num_txq;
2644 else
2645 max_txqs[i] = pf->num_lan_tx;
2646 } else {
2647 max_txqs[i] = vsi->alloc_txq;
2648 }
2649 }
2650
2651 dev_dbg(dev, "vsi->tc_cfg.ena_tc = %d\n", vsi->tc_cfg.ena_tc);
2652 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2653 max_txqs);
2654 if (ret) {
2655 dev_err(dev, "VSI %d failed lan queue config, error %d\n",
2656 vsi->vsi_num, ret);
2657 goto unroll_clear_rings;
2658 }
2659
2660 /* Add switch rule to drop all Tx Flow Control Frames, of look up
2661 * type ETHERTYPE from VSIs, and restrict malicious VF from sending
2662 * out PAUSE or PFC frames. If enabled, FW can still send FC frames.
2663 * The rule is added once for PF VSI in order to create appropriate
2664 * recipe, since VSI/VSI list is ignored with drop action...
2665 * Also add rules to handle LLDP Tx packets. Tx LLDP packets need to
2666 * be dropped so that VFs cannot send LLDP packets to reconfig DCB
2667 * settings in the HW.
2668 */
2669 if (!ice_is_safe_mode(pf))
2670 if (vsi->type == ICE_VSI_PF) {
2671 ice_fltr_add_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
2672 ICE_DROP_PACKET);
2673 ice_cfg_sw_lldp(vsi, true, true);
2674 }
2675
2676 if (!vsi->agg_node)
2677 ice_set_agg_vsi(vsi);
2678 return vsi;
2679
2680 unroll_clear_rings:
2681 ice_vsi_clear_rings(vsi);
2682 unroll_vector_base:
2683 /* reclaim SW interrupts back to the common pool */
2684 ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
2685 pf->num_avail_sw_msix += vsi->num_q_vectors;
2686 unroll_alloc_q_vector:
2687 ice_vsi_free_q_vectors(vsi);
2688 unroll_vsi_init:
2689 ice_vsi_delete(vsi);
2690 unroll_get_qs:
2691 ice_vsi_put_qs(vsi);
2692 unroll_vsi_alloc:
2693 if (vsi_type == ICE_VSI_VF)
2694 ice_enable_lag(pf->lag);
2695 ice_vsi_clear(vsi);
2696
2697 return NULL;
2698 }
2699
2700 /**
2701 * ice_vsi_release_msix - Clear the queue to Interrupt mapping in HW
2702 * @vsi: the VSI being cleaned up
2703 */
ice_vsi_release_msix(struct ice_vsi * vsi)2704 static void ice_vsi_release_msix(struct ice_vsi *vsi)
2705 {
2706 struct ice_pf *pf = vsi->back;
2707 struct ice_hw *hw = &pf->hw;
2708 u32 txq = 0;
2709 u32 rxq = 0;
2710 int i, q;
2711
2712 ice_for_each_q_vector(vsi, i) {
2713 struct ice_q_vector *q_vector = vsi->q_vectors[i];
2714
2715 ice_write_intrl(q_vector, 0);
2716 for (q = 0; q < q_vector->num_ring_tx; q++) {
2717 ice_write_itr(&q_vector->tx, 0);
2718 wr32(hw, QINT_TQCTL(vsi->txq_map[txq]), 0);
2719 if (ice_is_xdp_ena_vsi(vsi)) {
2720 u32 xdp_txq = txq + vsi->num_xdp_txq;
2721
2722 wr32(hw, QINT_TQCTL(vsi->txq_map[xdp_txq]), 0);
2723 }
2724 txq++;
2725 }
2726
2727 for (q = 0; q < q_vector->num_ring_rx; q++) {
2728 ice_write_itr(&q_vector->rx, 0);
2729 wr32(hw, QINT_RQCTL(vsi->rxq_map[rxq]), 0);
2730 rxq++;
2731 }
2732 }
2733
2734 ice_flush(hw);
2735 }
2736
2737 /**
2738 * ice_vsi_free_irq - Free the IRQ association with the OS
2739 * @vsi: the VSI being configured
2740 */
ice_vsi_free_irq(struct ice_vsi * vsi)2741 void ice_vsi_free_irq(struct ice_vsi *vsi)
2742 {
2743 struct ice_pf *pf = vsi->back;
2744 int base = vsi->base_vector;
2745 int i;
2746
2747 if (!vsi->q_vectors || !vsi->irqs_ready)
2748 return;
2749
2750 ice_vsi_release_msix(vsi);
2751 if (vsi->type == ICE_VSI_VF)
2752 return;
2753
2754 vsi->irqs_ready = false;
2755 ice_free_cpu_rx_rmap(vsi);
2756
2757 ice_for_each_q_vector(vsi, i) {
2758 u16 vector = i + base;
2759 int irq_num;
2760
2761 irq_num = pf->msix_entries[vector].vector;
2762
2763 /* free only the irqs that were actually requested */
2764 if (!vsi->q_vectors[i] ||
2765 !(vsi->q_vectors[i]->num_ring_tx ||
2766 vsi->q_vectors[i]->num_ring_rx))
2767 continue;
2768
2769 /* clear the affinity notifier in the IRQ descriptor */
2770 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2771 irq_set_affinity_notifier(irq_num, NULL);
2772
2773 /* clear the affinity_mask in the IRQ descriptor */
2774 irq_set_affinity_hint(irq_num, NULL);
2775 synchronize_irq(irq_num);
2776 devm_free_irq(ice_pf_to_dev(pf), irq_num, vsi->q_vectors[i]);
2777 }
2778 }
2779
2780 /**
2781 * ice_vsi_free_tx_rings - Free Tx resources for VSI queues
2782 * @vsi: the VSI having resources freed
2783 */
ice_vsi_free_tx_rings(struct ice_vsi * vsi)2784 void ice_vsi_free_tx_rings(struct ice_vsi *vsi)
2785 {
2786 int i;
2787
2788 if (!vsi->tx_rings)
2789 return;
2790
2791 ice_for_each_txq(vsi, i)
2792 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2793 ice_free_tx_ring(vsi->tx_rings[i]);
2794 }
2795
2796 /**
2797 * ice_vsi_free_rx_rings - Free Rx resources for VSI queues
2798 * @vsi: the VSI having resources freed
2799 */
ice_vsi_free_rx_rings(struct ice_vsi * vsi)2800 void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
2801 {
2802 int i;
2803
2804 if (!vsi->rx_rings)
2805 return;
2806
2807 ice_for_each_rxq(vsi, i)
2808 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2809 ice_free_rx_ring(vsi->rx_rings[i]);
2810 }
2811
2812 /**
2813 * ice_vsi_close - Shut down a VSI
2814 * @vsi: the VSI being shut down
2815 */
ice_vsi_close(struct ice_vsi * vsi)2816 void ice_vsi_close(struct ice_vsi *vsi)
2817 {
2818 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state))
2819 ice_down(vsi);
2820
2821 ice_vsi_free_irq(vsi);
2822 ice_vsi_free_tx_rings(vsi);
2823 ice_vsi_free_rx_rings(vsi);
2824 }
2825
2826 /**
2827 * ice_ena_vsi - resume a VSI
2828 * @vsi: the VSI being resume
2829 * @locked: is the rtnl_lock already held
2830 */
ice_ena_vsi(struct ice_vsi * vsi,bool locked)2831 int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
2832 {
2833 int err = 0;
2834
2835 if (!test_bit(ICE_VSI_NEEDS_RESTART, vsi->state))
2836 return 0;
2837
2838 clear_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2839
2840 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
2841 if (netif_running(vsi->netdev)) {
2842 if (!locked)
2843 rtnl_lock();
2844
2845 err = ice_open_internal(vsi->netdev);
2846
2847 if (!locked)
2848 rtnl_unlock();
2849 }
2850 } else if (vsi->type == ICE_VSI_CTRL) {
2851 err = ice_vsi_open_ctrl(vsi);
2852 }
2853
2854 return err;
2855 }
2856
2857 /**
2858 * ice_dis_vsi - pause a VSI
2859 * @vsi: the VSI being paused
2860 * @locked: is the rtnl_lock already held
2861 */
ice_dis_vsi(struct ice_vsi * vsi,bool locked)2862 void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
2863 {
2864 if (test_bit(ICE_VSI_DOWN, vsi->state))
2865 return;
2866
2867 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
2868
2869 if (vsi->type == ICE_VSI_PF && vsi->netdev) {
2870 if (netif_running(vsi->netdev)) {
2871 if (!locked)
2872 rtnl_lock();
2873
2874 ice_vsi_close(vsi);
2875
2876 if (!locked)
2877 rtnl_unlock();
2878 } else {
2879 ice_vsi_close(vsi);
2880 }
2881 } else if (vsi->type == ICE_VSI_CTRL ||
2882 vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
2883 ice_vsi_close(vsi);
2884 }
2885 }
2886
2887 /**
2888 * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
2889 * @vsi: the VSI being un-configured
2890 */
ice_vsi_dis_irq(struct ice_vsi * vsi)2891 void ice_vsi_dis_irq(struct ice_vsi *vsi)
2892 {
2893 int base = vsi->base_vector;
2894 struct ice_pf *pf = vsi->back;
2895 struct ice_hw *hw = &pf->hw;
2896 u32 val;
2897 int i;
2898
2899 /* disable interrupt causation from each queue */
2900 if (vsi->tx_rings) {
2901 ice_for_each_txq(vsi, i) {
2902 if (vsi->tx_rings[i]) {
2903 u16 reg;
2904
2905 reg = vsi->tx_rings[i]->reg_idx;
2906 val = rd32(hw, QINT_TQCTL(reg));
2907 val &= ~QINT_TQCTL_CAUSE_ENA_M;
2908 wr32(hw, QINT_TQCTL(reg), val);
2909 }
2910 }
2911 }
2912
2913 if (vsi->rx_rings) {
2914 ice_for_each_rxq(vsi, i) {
2915 if (vsi->rx_rings[i]) {
2916 u16 reg;
2917
2918 reg = vsi->rx_rings[i]->reg_idx;
2919 val = rd32(hw, QINT_RQCTL(reg));
2920 val &= ~QINT_RQCTL_CAUSE_ENA_M;
2921 wr32(hw, QINT_RQCTL(reg), val);
2922 }
2923 }
2924 }
2925
2926 /* disable each interrupt */
2927 ice_for_each_q_vector(vsi, i) {
2928 if (!vsi->q_vectors[i])
2929 continue;
2930 wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
2931 }
2932
2933 ice_flush(hw);
2934
2935 /* don't call synchronize_irq() for VF's from the host */
2936 if (vsi->type == ICE_VSI_VF)
2937 return;
2938
2939 ice_for_each_q_vector(vsi, i)
2940 synchronize_irq(pf->msix_entries[i + base].vector);
2941 }
2942
2943 /**
2944 * ice_napi_del - Remove NAPI handler for the VSI
2945 * @vsi: VSI for which NAPI handler is to be removed
2946 */
ice_napi_del(struct ice_vsi * vsi)2947 void ice_napi_del(struct ice_vsi *vsi)
2948 {
2949 int v_idx;
2950
2951 if (!vsi->netdev)
2952 return;
2953
2954 ice_for_each_q_vector(vsi, v_idx)
2955 netif_napi_del(&vsi->q_vectors[v_idx]->napi);
2956 }
2957
2958 /**
2959 * ice_free_vf_ctrl_res - Free the VF control VSI resource
2960 * @pf: pointer to PF structure
2961 * @vsi: the VSI to free resources for
2962 *
2963 * Check if the VF control VSI resource is still in use. If no VF is using it
2964 * any more, release the VSI resource. Otherwise, leave it to be cleaned up
2965 * once no other VF uses it.
2966 */
ice_free_vf_ctrl_res(struct ice_pf * pf,struct ice_vsi * vsi)2967 static void ice_free_vf_ctrl_res(struct ice_pf *pf, struct ice_vsi *vsi)
2968 {
2969 struct ice_vf *vf;
2970 unsigned int bkt;
2971
2972 rcu_read_lock();
2973 ice_for_each_vf_rcu(pf, bkt, vf) {
2974 if (vf != vsi->vf && vf->ctrl_vsi_idx != ICE_NO_VSI) {
2975 rcu_read_unlock();
2976 return;
2977 }
2978 }
2979 rcu_read_unlock();
2980
2981 /* No other VFs left that have control VSI. It is now safe to reclaim
2982 * SW interrupts back to the common pool.
2983 */
2984 ice_free_res(pf->irq_tracker, vsi->base_vector,
2985 ICE_RES_VF_CTRL_VEC_ID);
2986 pf->num_avail_sw_msix += vsi->num_q_vectors;
2987 }
2988
2989 /**
2990 * ice_vsi_release - Delete a VSI and free its resources
2991 * @vsi: the VSI being removed
2992 *
2993 * Returns 0 on success or < 0 on error
2994 */
ice_vsi_release(struct ice_vsi * vsi)2995 int ice_vsi_release(struct ice_vsi *vsi)
2996 {
2997 struct ice_pf *pf;
2998 int err;
2999
3000 if (!vsi->back)
3001 return -ENODEV;
3002 pf = vsi->back;
3003
3004 /* do not unregister while driver is in the reset recovery pending
3005 * state. Since reset/rebuild happens through PF service task workqueue,
3006 * it's not a good idea to unregister netdev that is associated to the
3007 * PF that is running the work queue items currently. This is done to
3008 * avoid check_flush_dependency() warning on this wq
3009 */
3010 if (vsi->netdev && !ice_is_reset_in_progress(pf->state) &&
3011 (test_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state))) {
3012 unregister_netdev(vsi->netdev);
3013 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
3014 }
3015
3016 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
3017 ice_rss_clean(vsi);
3018
3019 /* Disable VSI and free resources */
3020 if (vsi->type != ICE_VSI_LB)
3021 ice_vsi_dis_irq(vsi);
3022 ice_vsi_close(vsi);
3023
3024 /* SR-IOV determines needed MSIX resources all at once instead of per
3025 * VSI since when VFs are spawned we know how many VFs there are and how
3026 * many interrupts each VF needs. SR-IOV MSIX resources are also
3027 * cleared in the same manner.
3028 */
3029 if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
3030 ice_free_vf_ctrl_res(pf, vsi);
3031 } else if (vsi->type != ICE_VSI_VF) {
3032 /* reclaim SW interrupts back to the common pool */
3033 ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
3034 pf->num_avail_sw_msix += vsi->num_q_vectors;
3035 }
3036
3037 if (!ice_is_safe_mode(pf)) {
3038 if (vsi->type == ICE_VSI_PF) {
3039 ice_fltr_remove_eth(vsi, ETH_P_PAUSE, ICE_FLTR_TX,
3040 ICE_DROP_PACKET);
3041 ice_cfg_sw_lldp(vsi, true, false);
3042 /* The Rx rule will only exist to remove if the LLDP FW
3043 * engine is currently stopped
3044 */
3045 if (!test_bit(ICE_FLAG_FW_LLDP_AGENT, pf->flags))
3046 ice_cfg_sw_lldp(vsi, false, false);
3047 }
3048 }
3049
3050 if (ice_is_vsi_dflt_vsi(vsi))
3051 ice_clear_dflt_vsi(vsi);
3052 ice_fltr_remove_all(vsi);
3053 ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
3054 err = ice_rm_vsi_rdma_cfg(vsi->port_info, vsi->idx);
3055 if (err)
3056 dev_err(ice_pf_to_dev(vsi->back), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
3057 vsi->vsi_num, err);
3058 ice_vsi_delete(vsi);
3059 ice_vsi_free_q_vectors(vsi);
3060
3061 if (vsi->netdev) {
3062 if (test_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state)) {
3063 unregister_netdev(vsi->netdev);
3064 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
3065 }
3066 if (test_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state)) {
3067 free_netdev(vsi->netdev);
3068 vsi->netdev = NULL;
3069 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3070 }
3071 }
3072
3073 if (vsi->type == ICE_VSI_PF)
3074 ice_devlink_destroy_pf_port(pf);
3075
3076 if (vsi->type == ICE_VSI_VF &&
3077 vsi->agg_node && vsi->agg_node->valid)
3078 vsi->agg_node->num_vsis--;
3079 ice_vsi_clear_rings(vsi);
3080
3081 ice_vsi_put_qs(vsi);
3082
3083 /* retain SW VSI data structure since it is needed to unregister and
3084 * free VSI netdev when PF is not in reset recovery pending state,\
3085 * for ex: during rmmod.
3086 */
3087 if (!ice_is_reset_in_progress(pf->state))
3088 ice_vsi_clear(vsi);
3089
3090 return 0;
3091 }
3092
3093 /**
3094 * ice_vsi_rebuild_get_coalesce - get coalesce from all q_vectors
3095 * @vsi: VSI connected with q_vectors
3096 * @coalesce: array of struct with stored coalesce
3097 *
3098 * Returns array size.
3099 */
3100 static int
ice_vsi_rebuild_get_coalesce(struct ice_vsi * vsi,struct ice_coalesce_stored * coalesce)3101 ice_vsi_rebuild_get_coalesce(struct ice_vsi *vsi,
3102 struct ice_coalesce_stored *coalesce)
3103 {
3104 int i;
3105
3106 ice_for_each_q_vector(vsi, i) {
3107 struct ice_q_vector *q_vector = vsi->q_vectors[i];
3108
3109 coalesce[i].itr_tx = q_vector->tx.itr_settings;
3110 coalesce[i].itr_rx = q_vector->rx.itr_settings;
3111 coalesce[i].intrl = q_vector->intrl;
3112
3113 if (i < vsi->num_txq)
3114 coalesce[i].tx_valid = true;
3115 if (i < vsi->num_rxq)
3116 coalesce[i].rx_valid = true;
3117 }
3118
3119 return vsi->num_q_vectors;
3120 }
3121
3122 /**
3123 * ice_vsi_rebuild_set_coalesce - set coalesce from earlier saved arrays
3124 * @vsi: VSI connected with q_vectors
3125 * @coalesce: pointer to array of struct with stored coalesce
3126 * @size: size of coalesce array
3127 *
3128 * Before this function, ice_vsi_rebuild_get_coalesce should be called to save
3129 * ITR params in arrays. If size is 0 or coalesce wasn't stored set coalesce
3130 * to default value.
3131 */
3132 static void
ice_vsi_rebuild_set_coalesce(struct ice_vsi * vsi,struct ice_coalesce_stored * coalesce,int size)3133 ice_vsi_rebuild_set_coalesce(struct ice_vsi *vsi,
3134 struct ice_coalesce_stored *coalesce, int size)
3135 {
3136 struct ice_ring_container *rc;
3137 int i;
3138
3139 if ((size && !coalesce) || !vsi)
3140 return;
3141
3142 /* There are a couple of cases that have to be handled here:
3143 * 1. The case where the number of queue vectors stays the same, but
3144 * the number of Tx or Rx rings changes (the first for loop)
3145 * 2. The case where the number of queue vectors increased (the
3146 * second for loop)
3147 */
3148 for (i = 0; i < size && i < vsi->num_q_vectors; i++) {
3149 /* There are 2 cases to handle here and they are the same for
3150 * both Tx and Rx:
3151 * if the entry was valid previously (coalesce[i].[tr]x_valid
3152 * and the loop variable is less than the number of rings
3153 * allocated, then write the previous values
3154 *
3155 * if the entry was not valid previously, but the number of
3156 * rings is less than are allocated (this means the number of
3157 * rings increased from previously), then write out the
3158 * values in the first element
3159 *
3160 * Also, always write the ITR, even if in ITR_IS_DYNAMIC
3161 * as there is no harm because the dynamic algorithm
3162 * will just overwrite.
3163 */
3164 if (i < vsi->alloc_rxq && coalesce[i].rx_valid) {
3165 rc = &vsi->q_vectors[i]->rx;
3166 rc->itr_settings = coalesce[i].itr_rx;
3167 ice_write_itr(rc, rc->itr_setting);
3168 } else if (i < vsi->alloc_rxq) {
3169 rc = &vsi->q_vectors[i]->rx;
3170 rc->itr_settings = coalesce[0].itr_rx;
3171 ice_write_itr(rc, rc->itr_setting);
3172 }
3173
3174 if (i < vsi->alloc_txq && coalesce[i].tx_valid) {
3175 rc = &vsi->q_vectors[i]->tx;
3176 rc->itr_settings = coalesce[i].itr_tx;
3177 ice_write_itr(rc, rc->itr_setting);
3178 } else if (i < vsi->alloc_txq) {
3179 rc = &vsi->q_vectors[i]->tx;
3180 rc->itr_settings = coalesce[0].itr_tx;
3181 ice_write_itr(rc, rc->itr_setting);
3182 }
3183
3184 vsi->q_vectors[i]->intrl = coalesce[i].intrl;
3185 ice_set_q_vector_intrl(vsi->q_vectors[i]);
3186 }
3187
3188 /* the number of queue vectors increased so write whatever is in
3189 * the first element
3190 */
3191 for (; i < vsi->num_q_vectors; i++) {
3192 /* transmit */
3193 rc = &vsi->q_vectors[i]->tx;
3194 rc->itr_settings = coalesce[0].itr_tx;
3195 ice_write_itr(rc, rc->itr_setting);
3196
3197 /* receive */
3198 rc = &vsi->q_vectors[i]->rx;
3199 rc->itr_settings = coalesce[0].itr_rx;
3200 ice_write_itr(rc, rc->itr_setting);
3201
3202 vsi->q_vectors[i]->intrl = coalesce[0].intrl;
3203 ice_set_q_vector_intrl(vsi->q_vectors[i]);
3204 }
3205 }
3206
3207 /**
3208 * ice_vsi_rebuild - Rebuild VSI after reset
3209 * @vsi: VSI to be rebuild
3210 * @init_vsi: is this an initialization or a reconfigure of the VSI
3211 *
3212 * Returns 0 on success and negative value on failure
3213 */
ice_vsi_rebuild(struct ice_vsi * vsi,bool init_vsi)3214 int ice_vsi_rebuild(struct ice_vsi *vsi, bool init_vsi)
3215 {
3216 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3217 struct ice_coalesce_stored *coalesce;
3218 int prev_num_q_vectors = 0;
3219 enum ice_vsi_type vtype;
3220 struct ice_pf *pf;
3221 int ret, i;
3222
3223 if (!vsi)
3224 return -EINVAL;
3225
3226 pf = vsi->back;
3227 vtype = vsi->type;
3228 if (WARN_ON(vtype == ICE_VSI_VF && !vsi->vf))
3229 return -EINVAL;
3230
3231 ice_vsi_init_vlan_ops(vsi);
3232
3233 coalesce = kcalloc(vsi->num_q_vectors,
3234 sizeof(struct ice_coalesce_stored), GFP_KERNEL);
3235 if (!coalesce)
3236 return -ENOMEM;
3237
3238 prev_num_q_vectors = ice_vsi_rebuild_get_coalesce(vsi, coalesce);
3239
3240 ice_rm_vsi_lan_cfg(vsi->port_info, vsi->idx);
3241 ret = ice_rm_vsi_rdma_cfg(vsi->port_info, vsi->idx);
3242 if (ret)
3243 dev_err(ice_pf_to_dev(vsi->back), "Failed to remove RDMA scheduler config for VSI %u, err %d\n",
3244 vsi->vsi_num, ret);
3245 ice_vsi_free_q_vectors(vsi);
3246
3247 /* SR-IOV determines needed MSIX resources all at once instead of per
3248 * VSI since when VFs are spawned we know how many VFs there are and how
3249 * many interrupts each VF needs. SR-IOV MSIX resources are also
3250 * cleared in the same manner.
3251 */
3252 if (vtype != ICE_VSI_VF) {
3253 /* reclaim SW interrupts back to the common pool */
3254 ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
3255 pf->num_avail_sw_msix += vsi->num_q_vectors;
3256 vsi->base_vector = 0;
3257 }
3258
3259 if (ice_is_xdp_ena_vsi(vsi))
3260 /* return value check can be skipped here, it always returns
3261 * 0 if reset is in progress
3262 */
3263 ice_destroy_xdp_rings(vsi);
3264 ice_vsi_put_qs(vsi);
3265 ice_vsi_clear_rings(vsi);
3266 ice_vsi_free_arrays(vsi);
3267 if (vtype == ICE_VSI_VF)
3268 ice_vsi_set_num_qs(vsi, vsi->vf);
3269 else
3270 ice_vsi_set_num_qs(vsi, NULL);
3271
3272 ret = ice_vsi_alloc_arrays(vsi);
3273 if (ret < 0)
3274 goto err_vsi;
3275
3276 ice_vsi_get_qs(vsi);
3277
3278 ice_alloc_fd_res(vsi);
3279 ice_vsi_set_tc_cfg(vsi);
3280
3281 /* Initialize VSI struct elements and create VSI in FW */
3282 ret = ice_vsi_init(vsi, init_vsi);
3283 if (ret < 0)
3284 goto err_vsi;
3285
3286 switch (vtype) {
3287 case ICE_VSI_CTRL:
3288 case ICE_VSI_SWITCHDEV_CTRL:
3289 case ICE_VSI_PF:
3290 ret = ice_vsi_alloc_q_vectors(vsi);
3291 if (ret)
3292 goto err_rings;
3293
3294 ret = ice_vsi_setup_vector_base(vsi);
3295 if (ret)
3296 goto err_vectors;
3297
3298 ret = ice_vsi_set_q_vectors_reg_idx(vsi);
3299 if (ret)
3300 goto err_vectors;
3301
3302 ret = ice_vsi_alloc_rings(vsi);
3303 if (ret)
3304 goto err_vectors;
3305
3306 ice_vsi_map_rings_to_vectors(vsi);
3307 if (ice_is_xdp_ena_vsi(vsi)) {
3308 ret = ice_vsi_determine_xdp_res(vsi);
3309 if (ret)
3310 goto err_vectors;
3311 ret = ice_prepare_xdp_rings(vsi, vsi->xdp_prog);
3312 if (ret)
3313 goto err_vectors;
3314 }
3315 /* ICE_VSI_CTRL does not need RSS so skip RSS processing */
3316 if (vtype != ICE_VSI_CTRL)
3317 /* Do not exit if configuring RSS had an issue, at
3318 * least receive traffic on first queue. Hence no
3319 * need to capture return value
3320 */
3321 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags))
3322 ice_vsi_cfg_rss_lut_key(vsi);
3323
3324 /* disable or enable CRC stripping */
3325 if (vsi->netdev)
3326 ice_vsi_cfg_crc_strip(vsi, !!(vsi->netdev->features &
3327 NETIF_F_RXFCS));
3328
3329 break;
3330 case ICE_VSI_VF:
3331 ret = ice_vsi_alloc_q_vectors(vsi);
3332 if (ret)
3333 goto err_rings;
3334
3335 ret = ice_vsi_set_q_vectors_reg_idx(vsi);
3336 if (ret)
3337 goto err_vectors;
3338
3339 ret = ice_vsi_alloc_rings(vsi);
3340 if (ret)
3341 goto err_vectors;
3342
3343 break;
3344 case ICE_VSI_CHNL:
3345 if (test_bit(ICE_FLAG_RSS_ENA, pf->flags)) {
3346 ice_vsi_cfg_rss_lut_key(vsi);
3347 ice_vsi_set_rss_flow_fld(vsi);
3348 }
3349 break;
3350 default:
3351 break;
3352 }
3353
3354 /* configure VSI nodes based on number of queues and TC's */
3355 for (i = 0; i < vsi->tc_cfg.numtc; i++) {
3356 /* configure VSI nodes based on number of queues and TC's.
3357 * ADQ creates VSIs for each TC/Channel but doesn't
3358 * allocate queues instead it reconfigures the PF queues
3359 * as per the TC command. So max_txqs should point to the
3360 * PF Tx queues.
3361 */
3362 if (vtype == ICE_VSI_CHNL)
3363 max_txqs[i] = pf->num_lan_tx;
3364 else
3365 max_txqs[i] = vsi->alloc_txq;
3366
3367 if (ice_is_xdp_ena_vsi(vsi))
3368 max_txqs[i] += vsi->num_xdp_txq;
3369 }
3370
3371 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3372 /* If MQPRIO is set, means channel code path, hence for main
3373 * VSI's, use TC as 1
3374 */
3375 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, 1, max_txqs);
3376 else
3377 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx,
3378 vsi->tc_cfg.ena_tc, max_txqs);
3379
3380 if (ret) {
3381 dev_err(ice_pf_to_dev(pf), "VSI %d failed lan queue config, error %d\n",
3382 vsi->vsi_num, ret);
3383 if (init_vsi) {
3384 ret = -EIO;
3385 goto err_vectors;
3386 } else {
3387 return ice_schedule_reset(pf, ICE_RESET_PFR);
3388 }
3389 }
3390 ice_vsi_rebuild_set_coalesce(vsi, coalesce, prev_num_q_vectors);
3391 kfree(coalesce);
3392
3393 return 0;
3394
3395 err_vectors:
3396 ice_vsi_free_q_vectors(vsi);
3397 err_rings:
3398 if (vsi->netdev) {
3399 vsi->current_netdev_flags = 0;
3400 unregister_netdev(vsi->netdev);
3401 free_netdev(vsi->netdev);
3402 vsi->netdev = NULL;
3403 }
3404 err_vsi:
3405 ice_vsi_clear(vsi);
3406 set_bit(ICE_RESET_FAILED, pf->state);
3407 kfree(coalesce);
3408 return ret;
3409 }
3410
3411 /**
3412 * ice_is_reset_in_progress - check for a reset in progress
3413 * @state: PF state field
3414 */
ice_is_reset_in_progress(unsigned long * state)3415 bool ice_is_reset_in_progress(unsigned long *state)
3416 {
3417 return test_bit(ICE_RESET_OICR_RECV, state) ||
3418 test_bit(ICE_PFR_REQ, state) ||
3419 test_bit(ICE_CORER_REQ, state) ||
3420 test_bit(ICE_GLOBR_REQ, state);
3421 }
3422
3423 /**
3424 * ice_wait_for_reset - Wait for driver to finish reset and rebuild
3425 * @pf: pointer to the PF structure
3426 * @timeout: length of time to wait, in jiffies
3427 *
3428 * Wait (sleep) for a short time until the driver finishes cleaning up from
3429 * a device reset. The caller must be able to sleep. Use this to delay
3430 * operations that could fail while the driver is cleaning up after a device
3431 * reset.
3432 *
3433 * Returns 0 on success, -EBUSY if the reset is not finished within the
3434 * timeout, and -ERESTARTSYS if the thread was interrupted.
3435 */
ice_wait_for_reset(struct ice_pf * pf,unsigned long timeout)3436 int ice_wait_for_reset(struct ice_pf *pf, unsigned long timeout)
3437 {
3438 long ret;
3439
3440 ret = wait_event_interruptible_timeout(pf->reset_wait_queue,
3441 !ice_is_reset_in_progress(pf->state),
3442 timeout);
3443 if (ret < 0)
3444 return ret;
3445 else if (!ret)
3446 return -EBUSY;
3447 else
3448 return 0;
3449 }
3450
3451 /**
3452 * ice_vsi_update_q_map - update our copy of the VSI info with new queue map
3453 * @vsi: VSI being configured
3454 * @ctx: the context buffer returned from AQ VSI update command
3455 */
ice_vsi_update_q_map(struct ice_vsi * vsi,struct ice_vsi_ctx * ctx)3456 static void ice_vsi_update_q_map(struct ice_vsi *vsi, struct ice_vsi_ctx *ctx)
3457 {
3458 vsi->info.mapping_flags = ctx->info.mapping_flags;
3459 memcpy(&vsi->info.q_mapping, &ctx->info.q_mapping,
3460 sizeof(vsi->info.q_mapping));
3461 memcpy(&vsi->info.tc_mapping, ctx->info.tc_mapping,
3462 sizeof(vsi->info.tc_mapping));
3463 }
3464
3465 /**
3466 * ice_vsi_cfg_netdev_tc - Setup the netdev TC configuration
3467 * @vsi: the VSI being configured
3468 * @ena_tc: TC map to be enabled
3469 */
ice_vsi_cfg_netdev_tc(struct ice_vsi * vsi,u8 ena_tc)3470 void ice_vsi_cfg_netdev_tc(struct ice_vsi *vsi, u8 ena_tc)
3471 {
3472 struct net_device *netdev = vsi->netdev;
3473 struct ice_pf *pf = vsi->back;
3474 int numtc = vsi->tc_cfg.numtc;
3475 struct ice_dcbx_cfg *dcbcfg;
3476 u8 netdev_tc;
3477 int i;
3478
3479 if (!netdev)
3480 return;
3481
3482 /* CHNL VSI doesn't have it's own netdev, hence, no netdev_tc */
3483 if (vsi->type == ICE_VSI_CHNL)
3484 return;
3485
3486 if (!ena_tc) {
3487 netdev_reset_tc(netdev);
3488 return;
3489 }
3490
3491 if (vsi->type == ICE_VSI_PF && ice_is_adq_active(pf))
3492 numtc = vsi->all_numtc;
3493
3494 if (netdev_set_num_tc(netdev, numtc))
3495 return;
3496
3497 dcbcfg = &pf->hw.port_info->qos_cfg.local_dcbx_cfg;
3498
3499 ice_for_each_traffic_class(i)
3500 if (vsi->tc_cfg.ena_tc & BIT(i))
3501 netdev_set_tc_queue(netdev,
3502 vsi->tc_cfg.tc_info[i].netdev_tc,
3503 vsi->tc_cfg.tc_info[i].qcount_tx,
3504 vsi->tc_cfg.tc_info[i].qoffset);
3505 /* setup TC queue map for CHNL TCs */
3506 ice_for_each_chnl_tc(i) {
3507 if (!(vsi->all_enatc & BIT(i)))
3508 break;
3509 if (!vsi->mqprio_qopt.qopt.count[i])
3510 break;
3511 netdev_set_tc_queue(netdev, i,
3512 vsi->mqprio_qopt.qopt.count[i],
3513 vsi->mqprio_qopt.qopt.offset[i]);
3514 }
3515
3516 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3517 return;
3518
3519 for (i = 0; i < ICE_MAX_USER_PRIORITY; i++) {
3520 u8 ets_tc = dcbcfg->etscfg.prio_table[i];
3521
3522 /* Get the mapped netdev TC# for the UP */
3523 netdev_tc = vsi->tc_cfg.tc_info[ets_tc].netdev_tc;
3524 netdev_set_prio_tc_map(netdev, i, netdev_tc);
3525 }
3526 }
3527
3528 /**
3529 * ice_vsi_setup_q_map_mqprio - Prepares mqprio based tc_config
3530 * @vsi: the VSI being configured,
3531 * @ctxt: VSI context structure
3532 * @ena_tc: number of traffic classes to enable
3533 *
3534 * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
3535 */
3536 static int
ice_vsi_setup_q_map_mqprio(struct ice_vsi * vsi,struct ice_vsi_ctx * ctxt,u8 ena_tc)3537 ice_vsi_setup_q_map_mqprio(struct ice_vsi *vsi, struct ice_vsi_ctx *ctxt,
3538 u8 ena_tc)
3539 {
3540 u16 pow, offset = 0, qcount_tx = 0, qcount_rx = 0, qmap;
3541 u16 tc0_offset = vsi->mqprio_qopt.qopt.offset[0];
3542 int tc0_qcount = vsi->mqprio_qopt.qopt.count[0];
3543 u16 new_txq, new_rxq;
3544 u8 netdev_tc = 0;
3545 int i;
3546
3547 vsi->tc_cfg.ena_tc = ena_tc ? ena_tc : 1;
3548
3549 pow = order_base_2(tc0_qcount);
3550 qmap = ((tc0_offset << ICE_AQ_VSI_TC_Q_OFFSET_S) &
3551 ICE_AQ_VSI_TC_Q_OFFSET_M) |
3552 ((pow << ICE_AQ_VSI_TC_Q_NUM_S) & ICE_AQ_VSI_TC_Q_NUM_M);
3553
3554 ice_for_each_traffic_class(i) {
3555 if (!(vsi->tc_cfg.ena_tc & BIT(i))) {
3556 /* TC is not enabled */
3557 vsi->tc_cfg.tc_info[i].qoffset = 0;
3558 vsi->tc_cfg.tc_info[i].qcount_rx = 1;
3559 vsi->tc_cfg.tc_info[i].qcount_tx = 1;
3560 vsi->tc_cfg.tc_info[i].netdev_tc = 0;
3561 ctxt->info.tc_mapping[i] = 0;
3562 continue;
3563 }
3564
3565 offset = vsi->mqprio_qopt.qopt.offset[i];
3566 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3567 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3568 vsi->tc_cfg.tc_info[i].qoffset = offset;
3569 vsi->tc_cfg.tc_info[i].qcount_rx = qcount_rx;
3570 vsi->tc_cfg.tc_info[i].qcount_tx = qcount_tx;
3571 vsi->tc_cfg.tc_info[i].netdev_tc = netdev_tc++;
3572 }
3573
3574 if (vsi->all_numtc && vsi->all_numtc != vsi->tc_cfg.numtc) {
3575 ice_for_each_chnl_tc(i) {
3576 if (!(vsi->all_enatc & BIT(i)))
3577 continue;
3578 offset = vsi->mqprio_qopt.qopt.offset[i];
3579 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
3580 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
3581 }
3582 }
3583
3584 new_txq = offset + qcount_tx;
3585 if (new_txq > vsi->alloc_txq) {
3586 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Tx queues (%u), than were allocated (%u)!\n",
3587 new_txq, vsi->alloc_txq);
3588 return -EINVAL;
3589 }
3590
3591 new_rxq = offset + qcount_rx;
3592 if (new_rxq > vsi->alloc_rxq) {
3593 dev_err(ice_pf_to_dev(vsi->back), "Trying to use more Rx queues (%u), than were allocated (%u)!\n",
3594 new_rxq, vsi->alloc_rxq);
3595 return -EINVAL;
3596 }
3597
3598 /* Set actual Tx/Rx queue pairs */
3599 vsi->num_txq = new_txq;
3600 vsi->num_rxq = new_rxq;
3601
3602 /* Setup queue TC[0].qmap for given VSI context */
3603 ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
3604 ctxt->info.q_mapping[0] = cpu_to_le16(vsi->rxq_map[0]);
3605 ctxt->info.q_mapping[1] = cpu_to_le16(tc0_qcount);
3606
3607 /* Find queue count available for channel VSIs and starting offset
3608 * for channel VSIs
3609 */
3610 if (tc0_qcount && tc0_qcount < vsi->num_rxq) {
3611 vsi->cnt_q_avail = vsi->num_rxq - tc0_qcount;
3612 vsi->next_base_q = tc0_qcount;
3613 }
3614 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_txq = %d\n", vsi->num_txq);
3615 dev_dbg(ice_pf_to_dev(vsi->back), "vsi->num_rxq = %d\n", vsi->num_rxq);
3616 dev_dbg(ice_pf_to_dev(vsi->back), "all_numtc %u, all_enatc: 0x%04x, tc_cfg.numtc %u\n",
3617 vsi->all_numtc, vsi->all_enatc, vsi->tc_cfg.numtc);
3618
3619 return 0;
3620 }
3621
3622 /**
3623 * ice_vsi_cfg_tc - Configure VSI Tx Sched for given TC map
3624 * @vsi: VSI to be configured
3625 * @ena_tc: TC bitmap
3626 *
3627 * VSI queues expected to be quiesced before calling this function
3628 */
ice_vsi_cfg_tc(struct ice_vsi * vsi,u8 ena_tc)3629 int ice_vsi_cfg_tc(struct ice_vsi *vsi, u8 ena_tc)
3630 {
3631 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
3632 struct ice_pf *pf = vsi->back;
3633 struct ice_tc_cfg old_tc_cfg;
3634 struct ice_vsi_ctx *ctx;
3635 struct device *dev;
3636 int i, ret = 0;
3637 u8 num_tc = 0;
3638
3639 dev = ice_pf_to_dev(pf);
3640 if (vsi->tc_cfg.ena_tc == ena_tc &&
3641 vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
3642 return ret;
3643
3644 ice_for_each_traffic_class(i) {
3645 /* build bitmap of enabled TCs */
3646 if (ena_tc & BIT(i))
3647 num_tc++;
3648 /* populate max_txqs per TC */
3649 max_txqs[i] = vsi->alloc_txq;
3650 /* Update max_txqs if it is CHNL VSI, because alloc_t[r]xq are
3651 * zero for CHNL VSI, hence use num_txq instead as max_txqs
3652 */
3653 if (vsi->type == ICE_VSI_CHNL &&
3654 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3655 max_txqs[i] = vsi->num_txq;
3656 }
3657
3658 memcpy(&old_tc_cfg, &vsi->tc_cfg, sizeof(old_tc_cfg));
3659 vsi->tc_cfg.ena_tc = ena_tc;
3660 vsi->tc_cfg.numtc = num_tc;
3661
3662 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
3663 if (!ctx)
3664 return -ENOMEM;
3665
3666 ctx->vf_num = 0;
3667 ctx->info = vsi->info;
3668
3669 if (vsi->type == ICE_VSI_PF &&
3670 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3671 ret = ice_vsi_setup_q_map_mqprio(vsi, ctx, ena_tc);
3672 else
3673 ret = ice_vsi_setup_q_map(vsi, ctx);
3674
3675 if (ret) {
3676 memcpy(&vsi->tc_cfg, &old_tc_cfg, sizeof(vsi->tc_cfg));
3677 goto out;
3678 }
3679
3680 /* must to indicate which section of VSI context are being modified */
3681 ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_RXQ_MAP_VALID);
3682 ret = ice_update_vsi(&pf->hw, vsi->idx, ctx, NULL);
3683 if (ret) {
3684 dev_info(dev, "Failed VSI Update\n");
3685 goto out;
3686 }
3687
3688 if (vsi->type == ICE_VSI_PF &&
3689 test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
3690 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, 1, max_txqs);
3691 else
3692 ret = ice_cfg_vsi_lan(vsi->port_info, vsi->idx,
3693 vsi->tc_cfg.ena_tc, max_txqs);
3694
3695 if (ret) {
3696 dev_err(dev, "VSI %d failed TC config, error %d\n",
3697 vsi->vsi_num, ret);
3698 goto out;
3699 }
3700 ice_vsi_update_q_map(vsi, ctx);
3701 vsi->info.valid_sections = 0;
3702
3703 ice_vsi_cfg_netdev_tc(vsi, ena_tc);
3704 out:
3705 kfree(ctx);
3706 return ret;
3707 }
3708
3709 /**
3710 * ice_update_ring_stats - Update ring statistics
3711 * @stats: stats to be updated
3712 * @pkts: number of processed packets
3713 * @bytes: number of processed bytes
3714 *
3715 * This function assumes that caller has acquired a u64_stats_sync lock.
3716 */
ice_update_ring_stats(struct ice_q_stats * stats,u64 pkts,u64 bytes)3717 static void ice_update_ring_stats(struct ice_q_stats *stats, u64 pkts, u64 bytes)
3718 {
3719 stats->bytes += bytes;
3720 stats->pkts += pkts;
3721 }
3722
3723 /**
3724 * ice_update_tx_ring_stats - Update Tx ring specific counters
3725 * @tx_ring: ring to update
3726 * @pkts: number of processed packets
3727 * @bytes: number of processed bytes
3728 */
ice_update_tx_ring_stats(struct ice_tx_ring * tx_ring,u64 pkts,u64 bytes)3729 void ice_update_tx_ring_stats(struct ice_tx_ring *tx_ring, u64 pkts, u64 bytes)
3730 {
3731 u64_stats_update_begin(&tx_ring->syncp);
3732 ice_update_ring_stats(&tx_ring->stats, pkts, bytes);
3733 u64_stats_update_end(&tx_ring->syncp);
3734 }
3735
3736 /**
3737 * ice_update_rx_ring_stats - Update Rx ring specific counters
3738 * @rx_ring: ring to update
3739 * @pkts: number of processed packets
3740 * @bytes: number of processed bytes
3741 */
ice_update_rx_ring_stats(struct ice_rx_ring * rx_ring,u64 pkts,u64 bytes)3742 void ice_update_rx_ring_stats(struct ice_rx_ring *rx_ring, u64 pkts, u64 bytes)
3743 {
3744 u64_stats_update_begin(&rx_ring->syncp);
3745 ice_update_ring_stats(&rx_ring->stats, pkts, bytes);
3746 u64_stats_update_end(&rx_ring->syncp);
3747 }
3748
3749 /**
3750 * ice_is_dflt_vsi_in_use - check if the default forwarding VSI is being used
3751 * @pi: port info of the switch with default VSI
3752 *
3753 * Return true if the there is a single VSI in default forwarding VSI list
3754 */
ice_is_dflt_vsi_in_use(struct ice_port_info * pi)3755 bool ice_is_dflt_vsi_in_use(struct ice_port_info *pi)
3756 {
3757 bool exists = false;
3758
3759 ice_check_if_dflt_vsi(pi, 0, &exists);
3760 return exists;
3761 }
3762
3763 /**
3764 * ice_is_vsi_dflt_vsi - check if the VSI passed in is the default VSI
3765 * @vsi: VSI to compare against default forwarding VSI
3766 *
3767 * If this VSI passed in is the default forwarding VSI then return true, else
3768 * return false
3769 */
ice_is_vsi_dflt_vsi(struct ice_vsi * vsi)3770 bool ice_is_vsi_dflt_vsi(struct ice_vsi *vsi)
3771 {
3772 return ice_check_if_dflt_vsi(vsi->port_info, vsi->idx, NULL);
3773 }
3774
3775 /**
3776 * ice_set_dflt_vsi - set the default forwarding VSI
3777 * @vsi: VSI getting set as the default forwarding VSI on the switch
3778 *
3779 * If the VSI passed in is already the default VSI and it's enabled just return
3780 * success.
3781 *
3782 * Otherwise try to set the VSI passed in as the switch's default VSI and
3783 * return the result.
3784 */
ice_set_dflt_vsi(struct ice_vsi * vsi)3785 int ice_set_dflt_vsi(struct ice_vsi *vsi)
3786 {
3787 struct device *dev;
3788 int status;
3789
3790 if (!vsi)
3791 return -EINVAL;
3792
3793 dev = ice_pf_to_dev(vsi->back);
3794
3795 /* the VSI passed in is already the default VSI */
3796 if (ice_is_vsi_dflt_vsi(vsi)) {
3797 dev_dbg(dev, "VSI %d passed in is already the default forwarding VSI, nothing to do\n",
3798 vsi->vsi_num);
3799 return 0;
3800 }
3801
3802 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, true, ICE_FLTR_RX);
3803 if (status) {
3804 dev_err(dev, "Failed to set VSI %d as the default forwarding VSI, error %d\n",
3805 vsi->vsi_num, status);
3806 return status;
3807 }
3808
3809 return 0;
3810 }
3811
3812 /**
3813 * ice_clear_dflt_vsi - clear the default forwarding VSI
3814 * @vsi: VSI to remove from filter list
3815 *
3816 * If the switch has no default VSI or it's not enabled then return error.
3817 *
3818 * Otherwise try to clear the default VSI and return the result.
3819 */
ice_clear_dflt_vsi(struct ice_vsi * vsi)3820 int ice_clear_dflt_vsi(struct ice_vsi *vsi)
3821 {
3822 struct device *dev;
3823 int status;
3824
3825 if (!vsi)
3826 return -EINVAL;
3827
3828 dev = ice_pf_to_dev(vsi->back);
3829
3830 /* there is no default VSI configured */
3831 if (!ice_is_dflt_vsi_in_use(vsi->port_info))
3832 return -ENODEV;
3833
3834 status = ice_cfg_dflt_vsi(vsi->port_info, vsi->idx, false,
3835 ICE_FLTR_RX);
3836 if (status) {
3837 dev_err(dev, "Failed to clear the default forwarding VSI %d, error %d\n",
3838 vsi->vsi_num, status);
3839 return -EIO;
3840 }
3841
3842 return 0;
3843 }
3844
3845 /**
3846 * ice_get_link_speed_mbps - get link speed in Mbps
3847 * @vsi: the VSI whose link speed is being queried
3848 *
3849 * Return current VSI link speed and 0 if the speed is unknown.
3850 */
ice_get_link_speed_mbps(struct ice_vsi * vsi)3851 int ice_get_link_speed_mbps(struct ice_vsi *vsi)
3852 {
3853 switch (vsi->port_info->phy.link_info.link_speed) {
3854 case ICE_AQ_LINK_SPEED_100GB:
3855 return SPEED_100000;
3856 case ICE_AQ_LINK_SPEED_50GB:
3857 return SPEED_50000;
3858 case ICE_AQ_LINK_SPEED_40GB:
3859 return SPEED_40000;
3860 case ICE_AQ_LINK_SPEED_25GB:
3861 return SPEED_25000;
3862 case ICE_AQ_LINK_SPEED_20GB:
3863 return SPEED_20000;
3864 case ICE_AQ_LINK_SPEED_10GB:
3865 return SPEED_10000;
3866 case ICE_AQ_LINK_SPEED_5GB:
3867 return SPEED_5000;
3868 case ICE_AQ_LINK_SPEED_2500MB:
3869 return SPEED_2500;
3870 case ICE_AQ_LINK_SPEED_1000MB:
3871 return SPEED_1000;
3872 case ICE_AQ_LINK_SPEED_100MB:
3873 return SPEED_100;
3874 case ICE_AQ_LINK_SPEED_10MB:
3875 return SPEED_10;
3876 case ICE_AQ_LINK_SPEED_UNKNOWN:
3877 default:
3878 return 0;
3879 }
3880 }
3881
3882 /**
3883 * ice_get_link_speed_kbps - get link speed in Kbps
3884 * @vsi: the VSI whose link speed is being queried
3885 *
3886 * Return current VSI link speed and 0 if the speed is unknown.
3887 */
ice_get_link_speed_kbps(struct ice_vsi * vsi)3888 int ice_get_link_speed_kbps(struct ice_vsi *vsi)
3889 {
3890 int speed_mbps;
3891
3892 speed_mbps = ice_get_link_speed_mbps(vsi);
3893
3894 return speed_mbps * 1000;
3895 }
3896
3897 /**
3898 * ice_set_min_bw_limit - setup minimum BW limit for Tx based on min_tx_rate
3899 * @vsi: VSI to be configured
3900 * @min_tx_rate: min Tx rate in Kbps to be configured as BW limit
3901 *
3902 * If the min_tx_rate is specified as 0 that means to clear the minimum BW limit
3903 * profile, otherwise a non-zero value will force a minimum BW limit for the VSI
3904 * on TC 0.
3905 */
ice_set_min_bw_limit(struct ice_vsi * vsi,u64 min_tx_rate)3906 int ice_set_min_bw_limit(struct ice_vsi *vsi, u64 min_tx_rate)
3907 {
3908 struct ice_pf *pf = vsi->back;
3909 struct device *dev;
3910 int status;
3911 int speed;
3912
3913 dev = ice_pf_to_dev(pf);
3914 if (!vsi->port_info) {
3915 dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3916 vsi->idx, vsi->type);
3917 return -EINVAL;
3918 }
3919
3920 speed = ice_get_link_speed_kbps(vsi);
3921 if (min_tx_rate > (u64)speed) {
3922 dev_err(dev, "invalid min Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3923 min_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3924 speed);
3925 return -EINVAL;
3926 }
3927
3928 /* Configure min BW for VSI limit */
3929 if (min_tx_rate) {
3930 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3931 ICE_MIN_BW, min_tx_rate);
3932 if (status) {
3933 dev_err(dev, "failed to set min Tx rate(%llu Kbps) for %s %d\n",
3934 min_tx_rate, ice_vsi_type_str(vsi->type),
3935 vsi->idx);
3936 return status;
3937 }
3938
3939 dev_dbg(dev, "set min Tx rate(%llu Kbps) for %s\n",
3940 min_tx_rate, ice_vsi_type_str(vsi->type));
3941 } else {
3942 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
3943 vsi->idx, 0,
3944 ICE_MIN_BW);
3945 if (status) {
3946 dev_err(dev, "failed to clear min Tx rate configuration for %s %d\n",
3947 ice_vsi_type_str(vsi->type), vsi->idx);
3948 return status;
3949 }
3950
3951 dev_dbg(dev, "cleared min Tx rate configuration for %s %d\n",
3952 ice_vsi_type_str(vsi->type), vsi->idx);
3953 }
3954
3955 return 0;
3956 }
3957
3958 /**
3959 * ice_set_max_bw_limit - setup maximum BW limit for Tx based on max_tx_rate
3960 * @vsi: VSI to be configured
3961 * @max_tx_rate: max Tx rate in Kbps to be configured as BW limit
3962 *
3963 * If the max_tx_rate is specified as 0 that means to clear the maximum BW limit
3964 * profile, otherwise a non-zero value will force a maximum BW limit for the VSI
3965 * on TC 0.
3966 */
ice_set_max_bw_limit(struct ice_vsi * vsi,u64 max_tx_rate)3967 int ice_set_max_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate)
3968 {
3969 struct ice_pf *pf = vsi->back;
3970 struct device *dev;
3971 int status;
3972 int speed;
3973
3974 dev = ice_pf_to_dev(pf);
3975 if (!vsi->port_info) {
3976 dev_dbg(dev, "VSI %d, type %u specified doesn't have valid port_info\n",
3977 vsi->idx, vsi->type);
3978 return -EINVAL;
3979 }
3980
3981 speed = ice_get_link_speed_kbps(vsi);
3982 if (max_tx_rate > (u64)speed) {
3983 dev_err(dev, "invalid max Tx rate %llu Kbps specified for %s %d is greater than current link speed %u Kbps\n",
3984 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx,
3985 speed);
3986 return -EINVAL;
3987 }
3988
3989 /* Configure max BW for VSI limit */
3990 if (max_tx_rate) {
3991 status = ice_cfg_vsi_bw_lmt_per_tc(vsi->port_info, vsi->idx, 0,
3992 ICE_MAX_BW, max_tx_rate);
3993 if (status) {
3994 dev_err(dev, "failed setting max Tx rate(%llu Kbps) for %s %d\n",
3995 max_tx_rate, ice_vsi_type_str(vsi->type),
3996 vsi->idx);
3997 return status;
3998 }
3999
4000 dev_dbg(dev, "set max Tx rate(%llu Kbps) for %s %d\n",
4001 max_tx_rate, ice_vsi_type_str(vsi->type), vsi->idx);
4002 } else {
4003 status = ice_cfg_vsi_bw_dflt_lmt_per_tc(vsi->port_info,
4004 vsi->idx, 0,
4005 ICE_MAX_BW);
4006 if (status) {
4007 dev_err(dev, "failed clearing max Tx rate configuration for %s %d\n",
4008 ice_vsi_type_str(vsi->type), vsi->idx);
4009 return status;
4010 }
4011
4012 dev_dbg(dev, "cleared max Tx rate configuration for %s %d\n",
4013 ice_vsi_type_str(vsi->type), vsi->idx);
4014 }
4015
4016 return 0;
4017 }
4018
4019 /**
4020 * ice_set_link - turn on/off physical link
4021 * @vsi: VSI to modify physical link on
4022 * @ena: turn on/off physical link
4023 */
ice_set_link(struct ice_vsi * vsi,bool ena)4024 int ice_set_link(struct ice_vsi *vsi, bool ena)
4025 {
4026 struct device *dev = ice_pf_to_dev(vsi->back);
4027 struct ice_port_info *pi = vsi->port_info;
4028 struct ice_hw *hw = pi->hw;
4029 int status;
4030
4031 if (vsi->type != ICE_VSI_PF)
4032 return -EINVAL;
4033
4034 status = ice_aq_set_link_restart_an(pi, ena, NULL);
4035
4036 /* if link is owned by manageability, FW will return ICE_AQ_RC_EMODE.
4037 * this is not a fatal error, so print a warning message and return
4038 * a success code. Return an error if FW returns an error code other
4039 * than ICE_AQ_RC_EMODE
4040 */
4041 if (status == -EIO) {
4042 if (hw->adminq.sq_last_status == ICE_AQ_RC_EMODE)
4043 dev_dbg(dev, "can't set link to %s, err %d aq_err %s. not fatal, continuing\n",
4044 (ena ? "ON" : "OFF"), status,
4045 ice_aq_str(hw->adminq.sq_last_status));
4046 } else if (status) {
4047 dev_err(dev, "can't set link to %s, err %d aq_err %s\n",
4048 (ena ? "ON" : "OFF"), status,
4049 ice_aq_str(hw->adminq.sq_last_status));
4050 return status;
4051 }
4052
4053 return 0;
4054 }
4055
4056 /**
4057 * ice_vsi_add_vlan_zero - add VLAN 0 filter(s) for this VSI
4058 * @vsi: VSI used to add VLAN filters
4059 *
4060 * In Single VLAN Mode (SVM), single VLAN filters via ICE_SW_LKUP_VLAN are based
4061 * on the inner VLAN ID, so the VLAN TPID (i.e. 0x8100 or 0x888a8) doesn't
4062 * matter. In Double VLAN Mode (DVM), outer/single VLAN filters via
4063 * ICE_SW_LKUP_VLAN are based on the outer/single VLAN ID + VLAN TPID.
4064 *
4065 * For both modes add a VLAN 0 + no VLAN TPID filter to handle untagged traffic
4066 * when VLAN pruning is enabled. Also, this handles VLAN 0 priority tagged
4067 * traffic in SVM, since the VLAN TPID isn't part of filtering.
4068 *
4069 * If DVM is enabled then an explicit VLAN 0 + VLAN TPID filter needs to be
4070 * added to allow VLAN 0 priority tagged traffic in DVM, since the VLAN TPID is
4071 * part of filtering.
4072 */
ice_vsi_add_vlan_zero(struct ice_vsi * vsi)4073 int ice_vsi_add_vlan_zero(struct ice_vsi *vsi)
4074 {
4075 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
4076 struct ice_vlan vlan;
4077 int err;
4078
4079 vlan = ICE_VLAN(0, 0, 0);
4080 err = vlan_ops->add_vlan(vsi, &vlan);
4081 if (err && err != -EEXIST)
4082 return err;
4083
4084 /* in SVM both VLAN 0 filters are identical */
4085 if (!ice_is_dvm_ena(&vsi->back->hw))
4086 return 0;
4087
4088 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
4089 err = vlan_ops->add_vlan(vsi, &vlan);
4090 if (err && err != -EEXIST)
4091 return err;
4092
4093 return 0;
4094 }
4095
4096 /**
4097 * ice_vsi_del_vlan_zero - delete VLAN 0 filter(s) for this VSI
4098 * @vsi: VSI used to add VLAN filters
4099 *
4100 * Delete the VLAN 0 filters in the same manner that they were added in
4101 * ice_vsi_add_vlan_zero.
4102 */
ice_vsi_del_vlan_zero(struct ice_vsi * vsi)4103 int ice_vsi_del_vlan_zero(struct ice_vsi *vsi)
4104 {
4105 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
4106 struct ice_vlan vlan;
4107 int err;
4108
4109 vlan = ICE_VLAN(0, 0, 0);
4110 err = vlan_ops->del_vlan(vsi, &vlan);
4111 if (err && err != -EEXIST)
4112 return err;
4113
4114 /* in SVM both VLAN 0 filters are identical */
4115 if (!ice_is_dvm_ena(&vsi->back->hw))
4116 return 0;
4117
4118 vlan = ICE_VLAN(ETH_P_8021Q, 0, 0);
4119 err = vlan_ops->del_vlan(vsi, &vlan);
4120 if (err && err != -EEXIST)
4121 return err;
4122
4123 /* when deleting the last VLAN filter, make sure to disable the VLAN
4124 * promisc mode so the filter isn't left by accident
4125 */
4126 return ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
4127 ICE_MCAST_VLAN_PROMISC_BITS, 0);
4128 }
4129
4130 /**
4131 * ice_vsi_num_zero_vlans - get number of VLAN 0 filters based on VLAN mode
4132 * @vsi: VSI used to get the VLAN mode
4133 *
4134 * If DVM is enabled then 2 VLAN 0 filters are added, else if SVM is enabled
4135 * then 1 VLAN 0 filter is added. See ice_vsi_add_vlan_zero for more details.
4136 */
ice_vsi_num_zero_vlans(struct ice_vsi * vsi)4137 static u16 ice_vsi_num_zero_vlans(struct ice_vsi *vsi)
4138 {
4139 #define ICE_DVM_NUM_ZERO_VLAN_FLTRS 2
4140 #define ICE_SVM_NUM_ZERO_VLAN_FLTRS 1
4141 /* no VLAN 0 filter is created when a port VLAN is active */
4142 if (vsi->type == ICE_VSI_VF) {
4143 if (WARN_ON(!vsi->vf))
4144 return 0;
4145
4146 if (ice_vf_is_port_vlan_ena(vsi->vf))
4147 return 0;
4148 }
4149
4150 if (ice_is_dvm_ena(&vsi->back->hw))
4151 return ICE_DVM_NUM_ZERO_VLAN_FLTRS;
4152 else
4153 return ICE_SVM_NUM_ZERO_VLAN_FLTRS;
4154 }
4155
4156 /**
4157 * ice_vsi_has_non_zero_vlans - check if VSI has any non-zero VLANs
4158 * @vsi: VSI used to determine if any non-zero VLANs have been added
4159 */
ice_vsi_has_non_zero_vlans(struct ice_vsi * vsi)4160 bool ice_vsi_has_non_zero_vlans(struct ice_vsi *vsi)
4161 {
4162 return (vsi->num_vlan > ice_vsi_num_zero_vlans(vsi));
4163 }
4164
4165 /**
4166 * ice_vsi_num_non_zero_vlans - get the number of non-zero VLANs for this VSI
4167 * @vsi: VSI used to get the number of non-zero VLANs added
4168 */
ice_vsi_num_non_zero_vlans(struct ice_vsi * vsi)4169 u16 ice_vsi_num_non_zero_vlans(struct ice_vsi *vsi)
4170 {
4171 return (vsi->num_vlan - ice_vsi_num_zero_vlans(vsi));
4172 }
4173
4174 /**
4175 * ice_is_feature_supported
4176 * @pf: pointer to the struct ice_pf instance
4177 * @f: feature enum to be checked
4178 *
4179 * returns true if feature is supported, false otherwise
4180 */
ice_is_feature_supported(struct ice_pf * pf,enum ice_feature f)4181 bool ice_is_feature_supported(struct ice_pf *pf, enum ice_feature f)
4182 {
4183 if (f < 0 || f >= ICE_F_MAX)
4184 return false;
4185
4186 return test_bit(f, pf->features);
4187 }
4188
4189 /**
4190 * ice_set_feature_support
4191 * @pf: pointer to the struct ice_pf instance
4192 * @f: feature enum to set
4193 */
ice_set_feature_support(struct ice_pf * pf,enum ice_feature f)4194 static void ice_set_feature_support(struct ice_pf *pf, enum ice_feature f)
4195 {
4196 if (f < 0 || f >= ICE_F_MAX)
4197 return;
4198
4199 set_bit(f, pf->features);
4200 }
4201
4202 /**
4203 * ice_clear_feature_support
4204 * @pf: pointer to the struct ice_pf instance
4205 * @f: feature enum to clear
4206 */
ice_clear_feature_support(struct ice_pf * pf,enum ice_feature f)4207 void ice_clear_feature_support(struct ice_pf *pf, enum ice_feature f)
4208 {
4209 if (f < 0 || f >= ICE_F_MAX)
4210 return;
4211
4212 clear_bit(f, pf->features);
4213 }
4214
4215 /**
4216 * ice_init_feature_support
4217 * @pf: pointer to the struct ice_pf instance
4218 *
4219 * called during init to setup supported feature
4220 */
ice_init_feature_support(struct ice_pf * pf)4221 void ice_init_feature_support(struct ice_pf *pf)
4222 {
4223 switch (pf->hw.device_id) {
4224 case ICE_DEV_ID_E810C_BACKPLANE:
4225 case ICE_DEV_ID_E810C_QSFP:
4226 case ICE_DEV_ID_E810C_SFP:
4227 ice_set_feature_support(pf, ICE_F_DSCP);
4228 ice_set_feature_support(pf, ICE_F_PTP_EXTTS);
4229 if (ice_is_e810t(&pf->hw)) {
4230 ice_set_feature_support(pf, ICE_F_SMA_CTRL);
4231 if (ice_gnss_is_gps_present(&pf->hw))
4232 ice_set_feature_support(pf, ICE_F_GNSS);
4233 }
4234 break;
4235 default:
4236 break;
4237 }
4238 }
4239
4240 /**
4241 * ice_vsi_update_security - update security block in VSI
4242 * @vsi: pointer to VSI structure
4243 * @fill: function pointer to fill ctx
4244 */
4245 int
ice_vsi_update_security(struct ice_vsi * vsi,void (* fill)(struct ice_vsi_ctx *))4246 ice_vsi_update_security(struct ice_vsi *vsi, void (*fill)(struct ice_vsi_ctx *))
4247 {
4248 struct ice_vsi_ctx ctx = { 0 };
4249
4250 ctx.info = vsi->info;
4251 ctx.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SECURITY_VALID);
4252 fill(&ctx);
4253
4254 if (ice_update_vsi(&vsi->back->hw, vsi->idx, &ctx, NULL))
4255 return -ENODEV;
4256
4257 vsi->info = ctx.info;
4258 return 0;
4259 }
4260
4261 /**
4262 * ice_vsi_ctx_set_antispoof - set antispoof function in VSI ctx
4263 * @ctx: pointer to VSI ctx structure
4264 */
ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx * ctx)4265 void ice_vsi_ctx_set_antispoof(struct ice_vsi_ctx *ctx)
4266 {
4267 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF |
4268 (ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4269 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4270 }
4271
4272 /**
4273 * ice_vsi_ctx_clear_antispoof - clear antispoof function in VSI ctx
4274 * @ctx: pointer to VSI ctx structure
4275 */
ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx * ctx)4276 void ice_vsi_ctx_clear_antispoof(struct ice_vsi_ctx *ctx)
4277 {
4278 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ENA_MAC_ANTI_SPOOF &
4279 ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4280 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4281 }
4282
4283 /**
4284 * ice_vsi_ctx_set_allow_override - allow destination override on VSI
4285 * @ctx: pointer to VSI ctx structure
4286 */
ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx * ctx)4287 void ice_vsi_ctx_set_allow_override(struct ice_vsi_ctx *ctx)
4288 {
4289 ctx->info.sec_flags |= ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4290 }
4291
4292 /**
4293 * ice_vsi_ctx_clear_allow_override - turn off destination override on VSI
4294 * @ctx: pointer to VSI ctx structure
4295 */
ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx * ctx)4296 void ice_vsi_ctx_clear_allow_override(struct ice_vsi_ctx *ctx)
4297 {
4298 ctx->info.sec_flags &= ~ICE_AQ_VSI_SEC_FLAG_ALLOW_DEST_OVRD;
4299 }
4300