xref: /linux/drivers/net/ethernet/intel/ice/ice_sriov.c (revision 9410645520e9b820069761f3450ef6661418e279)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, Intel Corporation. */
3 
4 #include "ice.h"
5 #include "ice_vf_lib_private.h"
6 #include "ice_base.h"
7 #include "ice_lib.h"
8 #include "ice_fltr.h"
9 #include "ice_dcb_lib.h"
10 #include "ice_flow.h"
11 #include "ice_eswitch.h"
12 #include "ice_virtchnl_allowlist.h"
13 #include "ice_flex_pipe.h"
14 #include "ice_vf_vsi_vlan_ops.h"
15 #include "ice_vlan.h"
16 
17 /**
18  * ice_free_vf_entries - Free all VF entries from the hash table
19  * @pf: pointer to the PF structure
20  *
21  * Iterate over the VF hash table, removing and releasing all VF entries.
22  * Called during VF teardown or as cleanup during failed VF initialization.
23  */
ice_free_vf_entries(struct ice_pf * pf)24 static void ice_free_vf_entries(struct ice_pf *pf)
25 {
26 	struct ice_vfs *vfs = &pf->vfs;
27 	struct hlist_node *tmp;
28 	struct ice_vf *vf;
29 	unsigned int bkt;
30 
31 	/* Remove all VFs from the hash table and release their main
32 	 * reference. Once all references to the VF are dropped, ice_put_vf()
33 	 * will call ice_release_vf which will remove the VF memory.
34 	 */
35 	lockdep_assert_held(&vfs->table_lock);
36 
37 	hash_for_each_safe(vfs->table, bkt, tmp, vf, entry) {
38 		hash_del_rcu(&vf->entry);
39 		ice_put_vf(vf);
40 	}
41 }
42 
43 /**
44  * ice_free_vf_res - Free a VF's resources
45  * @vf: pointer to the VF info
46  */
ice_free_vf_res(struct ice_vf * vf)47 static void ice_free_vf_res(struct ice_vf *vf)
48 {
49 	struct ice_pf *pf = vf->pf;
50 	int i, last_vector_idx;
51 
52 	/* First, disable VF's configuration API to prevent OS from
53 	 * accessing the VF's VSI after it's freed or invalidated.
54 	 */
55 	clear_bit(ICE_VF_STATE_INIT, vf->vf_states);
56 	ice_vf_fdir_exit(vf);
57 	/* free VF control VSI */
58 	if (vf->ctrl_vsi_idx != ICE_NO_VSI)
59 		ice_vf_ctrl_vsi_release(vf);
60 
61 	/* free VSI and disconnect it from the parent uplink */
62 	if (vf->lan_vsi_idx != ICE_NO_VSI) {
63 		ice_vf_vsi_release(vf);
64 		vf->num_mac = 0;
65 	}
66 
67 	last_vector_idx = vf->first_vector_idx + vf->num_msix - 1;
68 
69 	/* clear VF MDD event information */
70 	memset(&vf->mdd_tx_events, 0, sizeof(vf->mdd_tx_events));
71 	memset(&vf->mdd_rx_events, 0, sizeof(vf->mdd_rx_events));
72 
73 	/* Disable interrupts so that VF starts in a known state */
74 	for (i = vf->first_vector_idx; i <= last_vector_idx; i++) {
75 		wr32(&pf->hw, GLINT_DYN_CTL(i), GLINT_DYN_CTL_CLEARPBA_M);
76 		ice_flush(&pf->hw);
77 	}
78 	/* reset some of the state variables keeping track of the resources */
79 	clear_bit(ICE_VF_STATE_MC_PROMISC, vf->vf_states);
80 	clear_bit(ICE_VF_STATE_UC_PROMISC, vf->vf_states);
81 }
82 
83 /**
84  * ice_dis_vf_mappings
85  * @vf: pointer to the VF structure
86  */
ice_dis_vf_mappings(struct ice_vf * vf)87 static void ice_dis_vf_mappings(struct ice_vf *vf)
88 {
89 	struct ice_pf *pf = vf->pf;
90 	struct ice_vsi *vsi;
91 	struct device *dev;
92 	int first, last, v;
93 	struct ice_hw *hw;
94 
95 	hw = &pf->hw;
96 	vsi = ice_get_vf_vsi(vf);
97 	if (WARN_ON(!vsi))
98 		return;
99 
100 	dev = ice_pf_to_dev(pf);
101 	wr32(hw, VPINT_ALLOC(vf->vf_id), 0);
102 	wr32(hw, VPINT_ALLOC_PCI(vf->vf_id), 0);
103 
104 	first = vf->first_vector_idx;
105 	last = first + vf->num_msix - 1;
106 	for (v = first; v <= last; v++) {
107 		u32 reg;
108 
109 		reg = FIELD_PREP(GLINT_VECT2FUNC_IS_PF_M, 1) |
110 		      FIELD_PREP(GLINT_VECT2FUNC_PF_NUM_M, hw->pf_id);
111 		wr32(hw, GLINT_VECT2FUNC(v), reg);
112 	}
113 
114 	if (vsi->tx_mapping_mode == ICE_VSI_MAP_CONTIG)
115 		wr32(hw, VPLAN_TX_QBASE(vf->vf_id), 0);
116 	else
117 		dev_err(dev, "Scattered mode for VF Tx queues is not yet implemented\n");
118 
119 	if (vsi->rx_mapping_mode == ICE_VSI_MAP_CONTIG)
120 		wr32(hw, VPLAN_RX_QBASE(vf->vf_id), 0);
121 	else
122 		dev_err(dev, "Scattered mode for VF Rx queues is not yet implemented\n");
123 }
124 
125 /**
126  * ice_sriov_free_msix_res - Reset/free any used MSIX resources
127  * @pf: pointer to the PF structure
128  *
129  * Since no MSIX entries are taken from the pf->irq_tracker then just clear
130  * the pf->sriov_base_vector.
131  *
132  * Returns 0 on success, and -EINVAL on error.
133  */
ice_sriov_free_msix_res(struct ice_pf * pf)134 static int ice_sriov_free_msix_res(struct ice_pf *pf)
135 {
136 	if (!pf)
137 		return -EINVAL;
138 
139 	bitmap_free(pf->sriov_irq_bm);
140 	pf->sriov_irq_size = 0;
141 	pf->sriov_base_vector = 0;
142 
143 	return 0;
144 }
145 
146 /**
147  * ice_free_vfs - Free all VFs
148  * @pf: pointer to the PF structure
149  */
ice_free_vfs(struct ice_pf * pf)150 void ice_free_vfs(struct ice_pf *pf)
151 {
152 	struct device *dev = ice_pf_to_dev(pf);
153 	struct ice_vfs *vfs = &pf->vfs;
154 	struct ice_hw *hw = &pf->hw;
155 	struct ice_vf *vf;
156 	unsigned int bkt;
157 
158 	if (!ice_has_vfs(pf))
159 		return;
160 
161 	while (test_and_set_bit(ICE_VF_DIS, pf->state))
162 		usleep_range(1000, 2000);
163 
164 	/* Disable IOV before freeing resources. This lets any VF drivers
165 	 * running in the host get themselves cleaned up before we yank
166 	 * the carpet out from underneath their feet.
167 	 */
168 	if (!pci_vfs_assigned(pf->pdev))
169 		pci_disable_sriov(pf->pdev);
170 	else
171 		dev_warn(dev, "VFs are assigned - not disabling SR-IOV\n");
172 
173 	mutex_lock(&vfs->table_lock);
174 
175 	ice_for_each_vf(pf, bkt, vf) {
176 		mutex_lock(&vf->cfg_lock);
177 
178 		ice_eswitch_detach_vf(pf, vf);
179 		ice_dis_vf_qs(vf);
180 
181 		if (test_bit(ICE_VF_STATE_INIT, vf->vf_states)) {
182 			/* disable VF qp mappings and set VF disable state */
183 			ice_dis_vf_mappings(vf);
184 			set_bit(ICE_VF_STATE_DIS, vf->vf_states);
185 			ice_free_vf_res(vf);
186 		}
187 
188 		if (!pci_vfs_assigned(pf->pdev)) {
189 			u32 reg_idx, bit_idx;
190 
191 			reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32;
192 			bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32;
193 			wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx));
194 		}
195 
196 		/* clear malicious info since the VF is getting released */
197 		list_del(&vf->mbx_info.list_entry);
198 
199 		mutex_unlock(&vf->cfg_lock);
200 	}
201 
202 	if (ice_sriov_free_msix_res(pf))
203 		dev_err(dev, "Failed to free MSIX resources used by SR-IOV\n");
204 
205 	vfs->num_qps_per = 0;
206 	ice_free_vf_entries(pf);
207 
208 	mutex_unlock(&vfs->table_lock);
209 
210 	clear_bit(ICE_VF_DIS, pf->state);
211 	clear_bit(ICE_FLAG_SRIOV_ENA, pf->flags);
212 }
213 
214 /**
215  * ice_vf_vsi_setup - Set up a VF VSI
216  * @vf: VF to setup VSI for
217  *
218  * Returns pointer to the successfully allocated VSI struct on success,
219  * otherwise returns NULL on failure.
220  */
ice_vf_vsi_setup(struct ice_vf * vf)221 static struct ice_vsi *ice_vf_vsi_setup(struct ice_vf *vf)
222 {
223 	struct ice_vsi_cfg_params params = {};
224 	struct ice_pf *pf = vf->pf;
225 	struct ice_vsi *vsi;
226 
227 	params.type = ICE_VSI_VF;
228 	params.port_info = ice_vf_get_port_info(vf);
229 	params.vf = vf;
230 	params.flags = ICE_VSI_FLAG_INIT;
231 
232 	vsi = ice_vsi_setup(pf, &params);
233 
234 	if (!vsi) {
235 		dev_err(ice_pf_to_dev(pf), "Failed to create VF VSI\n");
236 		ice_vf_invalidate_vsi(vf);
237 		return NULL;
238 	}
239 
240 	vf->lan_vsi_idx = vsi->idx;
241 
242 	return vsi;
243 }
244 
245 
246 /**
247  * ice_ena_vf_msix_mappings - enable VF MSIX mappings in hardware
248  * @vf: VF to enable MSIX mappings for
249  *
250  * Some of the registers need to be indexed/configured using hardware global
251  * device values and other registers need 0-based values, which represent PF
252  * based values.
253  */
ice_ena_vf_msix_mappings(struct ice_vf * vf)254 static void ice_ena_vf_msix_mappings(struct ice_vf *vf)
255 {
256 	int device_based_first_msix, device_based_last_msix;
257 	int pf_based_first_msix, pf_based_last_msix, v;
258 	struct ice_pf *pf = vf->pf;
259 	int device_based_vf_id;
260 	struct ice_hw *hw;
261 	u32 reg;
262 
263 	hw = &pf->hw;
264 	pf_based_first_msix = vf->first_vector_idx;
265 	pf_based_last_msix = (pf_based_first_msix + vf->num_msix) - 1;
266 
267 	device_based_first_msix = pf_based_first_msix +
268 		pf->hw.func_caps.common_cap.msix_vector_first_id;
269 	device_based_last_msix =
270 		(device_based_first_msix + vf->num_msix) - 1;
271 	device_based_vf_id = vf->vf_id + hw->func_caps.vf_base_id;
272 
273 	reg = FIELD_PREP(VPINT_ALLOC_FIRST_M, device_based_first_msix) |
274 	      FIELD_PREP(VPINT_ALLOC_LAST_M, device_based_last_msix) |
275 	      VPINT_ALLOC_VALID_M;
276 	wr32(hw, VPINT_ALLOC(vf->vf_id), reg);
277 
278 	reg = FIELD_PREP(VPINT_ALLOC_PCI_FIRST_M, device_based_first_msix) |
279 	      FIELD_PREP(VPINT_ALLOC_PCI_LAST_M, device_based_last_msix) |
280 	      VPINT_ALLOC_PCI_VALID_M;
281 	wr32(hw, VPINT_ALLOC_PCI(vf->vf_id), reg);
282 
283 	/* map the interrupts to its functions */
284 	for (v = pf_based_first_msix; v <= pf_based_last_msix; v++) {
285 		reg = FIELD_PREP(GLINT_VECT2FUNC_VF_NUM_M, device_based_vf_id) |
286 		      FIELD_PREP(GLINT_VECT2FUNC_PF_NUM_M, hw->pf_id);
287 		wr32(hw, GLINT_VECT2FUNC(v), reg);
288 	}
289 
290 	/* Map mailbox interrupt to VF MSI-X vector 0 */
291 	wr32(hw, VPINT_MBX_CTL(device_based_vf_id), VPINT_MBX_CTL_CAUSE_ENA_M);
292 }
293 
294 /**
295  * ice_ena_vf_q_mappings - enable Rx/Tx queue mappings for a VF
296  * @vf: VF to enable the mappings for
297  * @max_txq: max Tx queues allowed on the VF's VSI
298  * @max_rxq: max Rx queues allowed on the VF's VSI
299  */
ice_ena_vf_q_mappings(struct ice_vf * vf,u16 max_txq,u16 max_rxq)300 static void ice_ena_vf_q_mappings(struct ice_vf *vf, u16 max_txq, u16 max_rxq)
301 {
302 	struct device *dev = ice_pf_to_dev(vf->pf);
303 	struct ice_vsi *vsi = ice_get_vf_vsi(vf);
304 	struct ice_hw *hw = &vf->pf->hw;
305 	u32 reg;
306 
307 	if (WARN_ON(!vsi))
308 		return;
309 
310 	/* set regardless of mapping mode */
311 	wr32(hw, VPLAN_TXQ_MAPENA(vf->vf_id), VPLAN_TXQ_MAPENA_TX_ENA_M);
312 
313 	/* VF Tx queues allocation */
314 	if (vsi->tx_mapping_mode == ICE_VSI_MAP_CONTIG) {
315 		/* set the VF PF Tx queue range
316 		 * VFNUMQ value should be set to (number of queues - 1). A value
317 		 * of 0 means 1 queue and a value of 255 means 256 queues
318 		 */
319 		reg = FIELD_PREP(VPLAN_TX_QBASE_VFFIRSTQ_M, vsi->txq_map[0]) |
320 		      FIELD_PREP(VPLAN_TX_QBASE_VFNUMQ_M, max_txq - 1);
321 		wr32(hw, VPLAN_TX_QBASE(vf->vf_id), reg);
322 	} else {
323 		dev_err(dev, "Scattered mode for VF Tx queues is not yet implemented\n");
324 	}
325 
326 	/* set regardless of mapping mode */
327 	wr32(hw, VPLAN_RXQ_MAPENA(vf->vf_id), VPLAN_RXQ_MAPENA_RX_ENA_M);
328 
329 	/* VF Rx queues allocation */
330 	if (vsi->rx_mapping_mode == ICE_VSI_MAP_CONTIG) {
331 		/* set the VF PF Rx queue range
332 		 * VFNUMQ value should be set to (number of queues - 1). A value
333 		 * of 0 means 1 queue and a value of 255 means 256 queues
334 		 */
335 		reg = FIELD_PREP(VPLAN_RX_QBASE_VFFIRSTQ_M, vsi->rxq_map[0]) |
336 		      FIELD_PREP(VPLAN_RX_QBASE_VFNUMQ_M, max_rxq - 1);
337 		wr32(hw, VPLAN_RX_QBASE(vf->vf_id), reg);
338 	} else {
339 		dev_err(dev, "Scattered mode for VF Rx queues is not yet implemented\n");
340 	}
341 }
342 
343 /**
344  * ice_ena_vf_mappings - enable VF MSIX and queue mapping
345  * @vf: pointer to the VF structure
346  */
ice_ena_vf_mappings(struct ice_vf * vf)347 static void ice_ena_vf_mappings(struct ice_vf *vf)
348 {
349 	struct ice_vsi *vsi = ice_get_vf_vsi(vf);
350 
351 	if (WARN_ON(!vsi))
352 		return;
353 
354 	ice_ena_vf_msix_mappings(vf);
355 	ice_ena_vf_q_mappings(vf, vsi->alloc_txq, vsi->alloc_rxq);
356 }
357 
358 /**
359  * ice_calc_vf_reg_idx - Calculate the VF's register index in the PF space
360  * @vf: VF to calculate the register index for
361  * @q_vector: a q_vector associated to the VF
362  */
ice_calc_vf_reg_idx(struct ice_vf * vf,struct ice_q_vector * q_vector)363 void ice_calc_vf_reg_idx(struct ice_vf *vf, struct ice_q_vector *q_vector)
364 {
365 	if (!vf || !q_vector)
366 		return;
367 
368 	/* always add one to account for the OICR being the first MSIX */
369 	q_vector->vf_reg_idx = q_vector->v_idx + ICE_NONQ_VECS_VF;
370 	q_vector->reg_idx = vf->first_vector_idx + q_vector->vf_reg_idx;
371 }
372 
373 /**
374  * ice_sriov_set_msix_res - Set any used MSIX resources
375  * @pf: pointer to PF structure
376  * @num_msix_needed: number of MSIX vectors needed for all SR-IOV VFs
377  *
378  * This function allows SR-IOV resources to be taken from the end of the PF's
379  * allowed HW MSIX vectors so that the irq_tracker will not be affected. We
380  * just set the pf->sriov_base_vector and return success.
381  *
382  * If there are not enough resources available, return an error. This should
383  * always be caught by ice_set_per_vf_res().
384  *
385  * Return 0 on success, and -EINVAL when there are not enough MSIX vectors
386  * in the PF's space available for SR-IOV.
387  */
ice_sriov_set_msix_res(struct ice_pf * pf,u16 num_msix_needed)388 static int ice_sriov_set_msix_res(struct ice_pf *pf, u16 num_msix_needed)
389 {
390 	u16 total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
391 	int vectors_used = ice_get_max_used_msix_vector(pf);
392 	int sriov_base_vector;
393 
394 	sriov_base_vector = total_vectors - num_msix_needed;
395 
396 	/* make sure we only grab irq_tracker entries from the list end and
397 	 * that we have enough available MSIX vectors
398 	 */
399 	if (sriov_base_vector < vectors_used)
400 		return -EINVAL;
401 
402 	pf->sriov_base_vector = sriov_base_vector;
403 
404 	return 0;
405 }
406 
407 /**
408  * ice_set_per_vf_res - check if vectors and queues are available
409  * @pf: pointer to the PF structure
410  * @num_vfs: the number of SR-IOV VFs being configured
411  *
412  * First, determine HW interrupts from common pool. If we allocate fewer VFs, we
413  * get more vectors and can enable more queues per VF. Note that this does not
414  * grab any vectors from the SW pool already allocated. Also note, that all
415  * vector counts include one for each VF's miscellaneous interrupt vector
416  * (i.e. OICR).
417  *
418  * Minimum VFs - 2 vectors, 1 queue pair
419  * Small VFs - 5 vectors, 4 queue pairs
420  * Medium VFs - 17 vectors, 16 queue pairs
421  *
422  * Second, determine number of queue pairs per VF by starting with a pre-defined
423  * maximum each VF supports. If this is not possible, then we adjust based on
424  * queue pairs available on the device.
425  *
426  * Lastly, set queue and MSI-X VF variables tracked by the PF so it can be used
427  * by each VF during VF initialization and reset.
428  */
ice_set_per_vf_res(struct ice_pf * pf,u16 num_vfs)429 static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
430 {
431 	int vectors_used = ice_get_max_used_msix_vector(pf);
432 	u16 num_msix_per_vf, num_txq, num_rxq, avail_qs;
433 	int msix_avail_per_vf, msix_avail_for_sriov;
434 	struct device *dev = ice_pf_to_dev(pf);
435 	int err;
436 
437 	lockdep_assert_held(&pf->vfs.table_lock);
438 
439 	if (!num_vfs)
440 		return -EINVAL;
441 
442 	/* determine MSI-X resources per VF */
443 	msix_avail_for_sriov = pf->hw.func_caps.common_cap.num_msix_vectors -
444 		vectors_used;
445 	msix_avail_per_vf = msix_avail_for_sriov / num_vfs;
446 	if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MED) {
447 		num_msix_per_vf = ICE_NUM_VF_MSIX_MED;
448 	} else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_SMALL) {
449 		num_msix_per_vf = ICE_NUM_VF_MSIX_SMALL;
450 	} else if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MULTIQ_MIN) {
451 		num_msix_per_vf = ICE_NUM_VF_MSIX_MULTIQ_MIN;
452 	} else if (msix_avail_per_vf >= ICE_MIN_INTR_PER_VF) {
453 		num_msix_per_vf = ICE_MIN_INTR_PER_VF;
454 	} else {
455 		dev_err(dev, "Only %d MSI-X interrupts available for SR-IOV. Not enough to support minimum of %d MSI-X interrupts per VF for %d VFs\n",
456 			msix_avail_for_sriov, ICE_MIN_INTR_PER_VF,
457 			num_vfs);
458 		return -ENOSPC;
459 	}
460 
461 	num_txq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF,
462 			ICE_MAX_RSS_QS_PER_VF);
463 	avail_qs = ice_get_avail_txq_count(pf) / num_vfs;
464 	if (!avail_qs)
465 		num_txq = 0;
466 	else if (num_txq > avail_qs)
467 		num_txq = rounddown_pow_of_two(avail_qs);
468 
469 	num_rxq = min_t(u16, num_msix_per_vf - ICE_NONQ_VECS_VF,
470 			ICE_MAX_RSS_QS_PER_VF);
471 	avail_qs = ice_get_avail_rxq_count(pf) / num_vfs;
472 	if (!avail_qs)
473 		num_rxq = 0;
474 	else if (num_rxq > avail_qs)
475 		num_rxq = rounddown_pow_of_two(avail_qs);
476 
477 	if (num_txq < ICE_MIN_QS_PER_VF || num_rxq < ICE_MIN_QS_PER_VF) {
478 		dev_err(dev, "Not enough queues to support minimum of %d queue pairs per VF for %d VFs\n",
479 			ICE_MIN_QS_PER_VF, num_vfs);
480 		return -ENOSPC;
481 	}
482 
483 	err = ice_sriov_set_msix_res(pf, num_msix_per_vf * num_vfs);
484 	if (err) {
485 		dev_err(dev, "Unable to set MSI-X resources for %d VFs, err %d\n",
486 			num_vfs, err);
487 		return err;
488 	}
489 
490 	/* only allow equal Tx/Rx queue count (i.e. queue pairs) */
491 	pf->vfs.num_qps_per = min_t(int, num_txq, num_rxq);
492 	pf->vfs.num_msix_per = num_msix_per_vf;
493 	dev_info(dev, "Enabling %d VFs with %d vectors and %d queues per VF\n",
494 		 num_vfs, pf->vfs.num_msix_per, pf->vfs.num_qps_per);
495 
496 	return 0;
497 }
498 
499 /**
500  * ice_sriov_get_irqs - get irqs for SR-IOV usacase
501  * @pf: pointer to PF structure
502  * @needed: number of irqs to get
503  *
504  * This returns the first MSI-X vector index in PF space that is used by this
505  * VF. This index is used when accessing PF relative registers such as
506  * GLINT_VECT2FUNC and GLINT_DYN_CTL.
507  * This will always be the OICR index in the AVF driver so any functionality
508  * using vf->first_vector_idx for queue configuration_id: id of VF which will
509  * use this irqs
510  *
511  * Only SRIOV specific vectors are tracked in sriov_irq_bm. SRIOV vectors are
512  * allocated from the end of global irq index. First bit in sriov_irq_bm means
513  * last irq index etc. It simplifies extension of SRIOV vectors.
514  * They will be always located from sriov_base_vector to the last irq
515  * index. While increasing/decreasing sriov_base_vector can be moved.
516  */
ice_sriov_get_irqs(struct ice_pf * pf,u16 needed)517 static int ice_sriov_get_irqs(struct ice_pf *pf, u16 needed)
518 {
519 	int res = bitmap_find_next_zero_area(pf->sriov_irq_bm,
520 					     pf->sriov_irq_size, 0, needed, 0);
521 	/* conversion from number in bitmap to global irq index */
522 	int index = pf->sriov_irq_size - res - needed;
523 
524 	if (res >= pf->sriov_irq_size || index < pf->sriov_base_vector)
525 		return -ENOENT;
526 
527 	bitmap_set(pf->sriov_irq_bm, res, needed);
528 	return index;
529 }
530 
531 /**
532  * ice_sriov_free_irqs - free irqs used by the VF
533  * @pf: pointer to PF structure
534  * @vf: pointer to VF structure
535  */
ice_sriov_free_irqs(struct ice_pf * pf,struct ice_vf * vf)536 static void ice_sriov_free_irqs(struct ice_pf *pf, struct ice_vf *vf)
537 {
538 	/* Move back from first vector index to first index in bitmap */
539 	int bm_i = pf->sriov_irq_size - vf->first_vector_idx - vf->num_msix;
540 
541 	bitmap_clear(pf->sriov_irq_bm, bm_i, vf->num_msix);
542 	vf->first_vector_idx = 0;
543 }
544 
545 /**
546  * ice_init_vf_vsi_res - initialize/setup VF VSI resources
547  * @vf: VF to initialize/setup the VSI for
548  *
549  * This function creates a VSI for the VF, adds a VLAN 0 filter, and sets up the
550  * VF VSI's broadcast filter and is only used during initial VF creation.
551  */
ice_init_vf_vsi_res(struct ice_vf * vf)552 static int ice_init_vf_vsi_res(struct ice_vf *vf)
553 {
554 	struct ice_pf *pf = vf->pf;
555 	struct ice_vsi *vsi;
556 	int err;
557 
558 	vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
559 	if (vf->first_vector_idx < 0)
560 		return -ENOMEM;
561 
562 	vsi = ice_vf_vsi_setup(vf);
563 	if (!vsi)
564 		return -ENOMEM;
565 
566 	err = ice_vf_init_host_cfg(vf, vsi);
567 	if (err)
568 		goto release_vsi;
569 
570 	return 0;
571 
572 release_vsi:
573 	ice_vf_vsi_release(vf);
574 	return err;
575 }
576 
577 /**
578  * ice_start_vfs - start VFs so they are ready to be used by SR-IOV
579  * @pf: PF the VFs are associated with
580  */
ice_start_vfs(struct ice_pf * pf)581 static int ice_start_vfs(struct ice_pf *pf)
582 {
583 	struct ice_hw *hw = &pf->hw;
584 	unsigned int bkt, it_cnt;
585 	struct ice_vf *vf;
586 	int retval;
587 
588 	lockdep_assert_held(&pf->vfs.table_lock);
589 
590 	it_cnt = 0;
591 	ice_for_each_vf(pf, bkt, vf) {
592 		vf->vf_ops->clear_reset_trigger(vf);
593 
594 		retval = ice_init_vf_vsi_res(vf);
595 		if (retval) {
596 			dev_err(ice_pf_to_dev(pf), "Failed to initialize VSI resources for VF %d, error %d\n",
597 				vf->vf_id, retval);
598 			goto teardown;
599 		}
600 
601 		retval = ice_eswitch_attach_vf(pf, vf);
602 		if (retval) {
603 			dev_err(ice_pf_to_dev(pf), "Failed to attach VF %d to eswitch, error %d",
604 				vf->vf_id, retval);
605 			ice_vf_vsi_release(vf);
606 			goto teardown;
607 		}
608 
609 		set_bit(ICE_VF_STATE_INIT, vf->vf_states);
610 		ice_ena_vf_mappings(vf);
611 		wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE);
612 		it_cnt++;
613 	}
614 
615 	ice_flush(hw);
616 	return 0;
617 
618 teardown:
619 	ice_for_each_vf(pf, bkt, vf) {
620 		if (it_cnt == 0)
621 			break;
622 
623 		ice_dis_vf_mappings(vf);
624 		ice_vf_vsi_release(vf);
625 		it_cnt--;
626 	}
627 
628 	return retval;
629 }
630 
631 /**
632  * ice_sriov_free_vf - Free VF memory after all references are dropped
633  * @vf: pointer to VF to free
634  *
635  * Called by ice_put_vf through ice_release_vf once the last reference to a VF
636  * structure has been dropped.
637  */
ice_sriov_free_vf(struct ice_vf * vf)638 static void ice_sriov_free_vf(struct ice_vf *vf)
639 {
640 	mutex_destroy(&vf->cfg_lock);
641 
642 	kfree_rcu(vf, rcu);
643 }
644 
645 /**
646  * ice_sriov_clear_reset_state - clears VF Reset status register
647  * @vf: the vf to configure
648  */
ice_sriov_clear_reset_state(struct ice_vf * vf)649 static void ice_sriov_clear_reset_state(struct ice_vf *vf)
650 {
651 	struct ice_hw *hw = &vf->pf->hw;
652 
653 	/* Clear the reset status register so that VF immediately sees that
654 	 * the device is resetting, even if hardware hasn't yet gotten around
655 	 * to clearing VFGEN_RSTAT for us.
656 	 */
657 	wr32(hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_INPROGRESS);
658 }
659 
660 /**
661  * ice_sriov_clear_mbx_register - clears SRIOV VF's mailbox registers
662  * @vf: the vf to configure
663  */
ice_sriov_clear_mbx_register(struct ice_vf * vf)664 static void ice_sriov_clear_mbx_register(struct ice_vf *vf)
665 {
666 	struct ice_pf *pf = vf->pf;
667 
668 	wr32(&pf->hw, VF_MBX_ARQLEN(vf->vf_id), 0);
669 	wr32(&pf->hw, VF_MBX_ATQLEN(vf->vf_id), 0);
670 }
671 
672 /**
673  * ice_sriov_trigger_reset_register - trigger VF reset for SRIOV VF
674  * @vf: pointer to VF structure
675  * @is_vflr: true if reset occurred due to VFLR
676  *
677  * Trigger and cleanup after a VF reset for a SR-IOV VF.
678  */
ice_sriov_trigger_reset_register(struct ice_vf * vf,bool is_vflr)679 static void ice_sriov_trigger_reset_register(struct ice_vf *vf, bool is_vflr)
680 {
681 	struct ice_pf *pf = vf->pf;
682 	u32 reg, reg_idx, bit_idx;
683 	unsigned int vf_abs_id, i;
684 	struct device *dev;
685 	struct ice_hw *hw;
686 
687 	dev = ice_pf_to_dev(pf);
688 	hw = &pf->hw;
689 	vf_abs_id = vf->vf_id + hw->func_caps.vf_base_id;
690 
691 	/* In the case of a VFLR, HW has already reset the VF and we just need
692 	 * to clean up. Otherwise we must first trigger the reset using the
693 	 * VFRTRIG register.
694 	 */
695 	if (!is_vflr) {
696 		reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id));
697 		reg |= VPGEN_VFRTRIG_VFSWR_M;
698 		wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg);
699 	}
700 
701 	/* clear the VFLR bit in GLGEN_VFLRSTAT */
702 	reg_idx = (vf_abs_id) / 32;
703 	bit_idx = (vf_abs_id) % 32;
704 	wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx));
705 	ice_flush(hw);
706 
707 	wr32(hw, PF_PCI_CIAA,
708 	     VF_DEVICE_STATUS | (vf_abs_id << PF_PCI_CIAA_VF_NUM_S));
709 	for (i = 0; i < ICE_PCI_CIAD_WAIT_COUNT; i++) {
710 		reg = rd32(hw, PF_PCI_CIAD);
711 		/* no transactions pending so stop polling */
712 		if ((reg & VF_TRANS_PENDING_M) == 0)
713 			break;
714 
715 		dev_err(dev, "VF %u PCI transactions stuck\n", vf->vf_id);
716 		udelay(ICE_PCI_CIAD_WAIT_DELAY_US);
717 	}
718 }
719 
720 /**
721  * ice_sriov_poll_reset_status - poll SRIOV VF reset status
722  * @vf: pointer to VF structure
723  *
724  * Returns true when reset is successful, else returns false
725  */
ice_sriov_poll_reset_status(struct ice_vf * vf)726 static bool ice_sriov_poll_reset_status(struct ice_vf *vf)
727 {
728 	struct ice_pf *pf = vf->pf;
729 	unsigned int i;
730 	u32 reg;
731 
732 	for (i = 0; i < 10; i++) {
733 		/* VF reset requires driver to first reset the VF and then
734 		 * poll the status register to make sure that the reset
735 		 * completed successfully.
736 		 */
737 		reg = rd32(&pf->hw, VPGEN_VFRSTAT(vf->vf_id));
738 		if (reg & VPGEN_VFRSTAT_VFRD_M)
739 			return true;
740 
741 		/* only sleep if the reset is not done */
742 		usleep_range(10, 20);
743 	}
744 	return false;
745 }
746 
747 /**
748  * ice_sriov_clear_reset_trigger - enable VF to access hardware
749  * @vf: VF to enabled hardware access for
750  */
ice_sriov_clear_reset_trigger(struct ice_vf * vf)751 static void ice_sriov_clear_reset_trigger(struct ice_vf *vf)
752 {
753 	struct ice_hw *hw = &vf->pf->hw;
754 	u32 reg;
755 
756 	reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_id));
757 	reg &= ~VPGEN_VFRTRIG_VFSWR_M;
758 	wr32(hw, VPGEN_VFRTRIG(vf->vf_id), reg);
759 	ice_flush(hw);
760 }
761 
762 /**
763  * ice_sriov_post_vsi_rebuild - tasks to do after the VF's VSI have been rebuilt
764  * @vf: VF to perform tasks on
765  */
ice_sriov_post_vsi_rebuild(struct ice_vf * vf)766 static void ice_sriov_post_vsi_rebuild(struct ice_vf *vf)
767 {
768 	ice_ena_vf_mappings(vf);
769 	wr32(&vf->pf->hw, VFGEN_RSTAT(vf->vf_id), VIRTCHNL_VFR_VFACTIVE);
770 }
771 
772 static const struct ice_vf_ops ice_sriov_vf_ops = {
773 	.reset_type = ICE_VF_RESET,
774 	.free = ice_sriov_free_vf,
775 	.clear_reset_state = ice_sriov_clear_reset_state,
776 	.clear_mbx_register = ice_sriov_clear_mbx_register,
777 	.trigger_reset_register = ice_sriov_trigger_reset_register,
778 	.poll_reset_status = ice_sriov_poll_reset_status,
779 	.clear_reset_trigger = ice_sriov_clear_reset_trigger,
780 	.irq_close = NULL,
781 	.post_vsi_rebuild = ice_sriov_post_vsi_rebuild,
782 };
783 
784 /**
785  * ice_create_vf_entries - Allocate and insert VF entries
786  * @pf: pointer to the PF structure
787  * @num_vfs: the number of VFs to allocate
788  *
789  * Allocate new VF entries and insert them into the hash table. Set some
790  * basic default fields for initializing the new VFs.
791  *
792  * After this function exits, the hash table will have num_vfs entries
793  * inserted.
794  *
795  * Returns 0 on success or an integer error code on failure.
796  */
ice_create_vf_entries(struct ice_pf * pf,u16 num_vfs)797 static int ice_create_vf_entries(struct ice_pf *pf, u16 num_vfs)
798 {
799 	struct pci_dev *pdev = pf->pdev;
800 	struct ice_vfs *vfs = &pf->vfs;
801 	struct pci_dev *vfdev = NULL;
802 	struct ice_vf *vf;
803 	u16 vf_pdev_id;
804 	int err, pos;
805 
806 	lockdep_assert_held(&vfs->table_lock);
807 
808 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
809 	pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_pdev_id);
810 
811 	for (u16 vf_id = 0; vf_id < num_vfs; vf_id++) {
812 		vf = kzalloc(sizeof(*vf), GFP_KERNEL);
813 		if (!vf) {
814 			err = -ENOMEM;
815 			goto err_free_entries;
816 		}
817 		kref_init(&vf->refcnt);
818 
819 		vf->pf = pf;
820 		vf->vf_id = vf_id;
821 
822 		/* set sriov vf ops for VFs created during SRIOV flow */
823 		vf->vf_ops = &ice_sriov_vf_ops;
824 
825 		ice_initialize_vf_entry(vf);
826 
827 		do {
828 			vfdev = pci_get_device(pdev->vendor, vf_pdev_id, vfdev);
829 		} while (vfdev && vfdev->physfn != pdev);
830 		vf->vfdev = vfdev;
831 		vf->vf_sw_id = pf->first_sw;
832 
833 		pci_dev_get(vfdev);
834 
835 		hash_add_rcu(vfs->table, &vf->entry, vf_id);
836 	}
837 
838 	/* Decrement of refcount done by pci_get_device() inside the loop does
839 	 * not touch the last iteration's vfdev, so it has to be done manually
840 	 * to balance pci_dev_get() added within the loop.
841 	 */
842 	pci_dev_put(vfdev);
843 
844 	return 0;
845 
846 err_free_entries:
847 	ice_free_vf_entries(pf);
848 	return err;
849 }
850 
851 /**
852  * ice_ena_vfs - enable VFs so they are ready to be used
853  * @pf: pointer to the PF structure
854  * @num_vfs: number of VFs to enable
855  */
ice_ena_vfs(struct ice_pf * pf,u16 num_vfs)856 static int ice_ena_vfs(struct ice_pf *pf, u16 num_vfs)
857 {
858 	int total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
859 	struct device *dev = ice_pf_to_dev(pf);
860 	struct ice_hw *hw = &pf->hw;
861 	int ret;
862 
863 	pf->sriov_irq_bm = bitmap_zalloc(total_vectors, GFP_KERNEL);
864 	if (!pf->sriov_irq_bm)
865 		return -ENOMEM;
866 	pf->sriov_irq_size = total_vectors;
867 
868 	/* Disable global interrupt 0 so we don't try to handle the VFLR. */
869 	wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
870 	     ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S);
871 	set_bit(ICE_OICR_INTR_DIS, pf->state);
872 	ice_flush(hw);
873 
874 	ret = pci_enable_sriov(pf->pdev, num_vfs);
875 	if (ret)
876 		goto err_unroll_intr;
877 
878 	mutex_lock(&pf->vfs.table_lock);
879 
880 	ret = ice_set_per_vf_res(pf, num_vfs);
881 	if (ret) {
882 		dev_err(dev, "Not enough resources for %d VFs, err %d. Try with fewer number of VFs\n",
883 			num_vfs, ret);
884 		goto err_unroll_sriov;
885 	}
886 
887 	ret = ice_create_vf_entries(pf, num_vfs);
888 	if (ret) {
889 		dev_err(dev, "Failed to allocate VF entries for %d VFs\n",
890 			num_vfs);
891 		goto err_unroll_sriov;
892 	}
893 
894 	ret = ice_start_vfs(pf);
895 	if (ret) {
896 		dev_err(dev, "Failed to start %d VFs, err %d\n", num_vfs, ret);
897 		ret = -EAGAIN;
898 		goto err_unroll_vf_entries;
899 	}
900 
901 	clear_bit(ICE_VF_DIS, pf->state);
902 
903 	/* rearm global interrupts */
904 	if (test_and_clear_bit(ICE_OICR_INTR_DIS, pf->state))
905 		ice_irq_dynamic_ena(hw, NULL, NULL);
906 
907 	mutex_unlock(&pf->vfs.table_lock);
908 
909 	return 0;
910 
911 err_unroll_vf_entries:
912 	ice_free_vf_entries(pf);
913 err_unroll_sriov:
914 	mutex_unlock(&pf->vfs.table_lock);
915 	pci_disable_sriov(pf->pdev);
916 err_unroll_intr:
917 	/* rearm interrupts here */
918 	ice_irq_dynamic_ena(hw, NULL, NULL);
919 	clear_bit(ICE_OICR_INTR_DIS, pf->state);
920 	bitmap_free(pf->sriov_irq_bm);
921 	return ret;
922 }
923 
924 /**
925  * ice_pci_sriov_ena - Enable or change number of VFs
926  * @pf: pointer to the PF structure
927  * @num_vfs: number of VFs to allocate
928  *
929  * Returns 0 on success and negative on failure
930  */
ice_pci_sriov_ena(struct ice_pf * pf,int num_vfs)931 static int ice_pci_sriov_ena(struct ice_pf *pf, int num_vfs)
932 {
933 	struct device *dev = ice_pf_to_dev(pf);
934 	int err;
935 
936 	if (!num_vfs) {
937 		ice_free_vfs(pf);
938 		return 0;
939 	}
940 
941 	if (num_vfs > pf->vfs.num_supported) {
942 		dev_err(dev, "Can't enable %d VFs, max VFs supported is %d\n",
943 			num_vfs, pf->vfs.num_supported);
944 		return -EOPNOTSUPP;
945 	}
946 
947 	dev_info(dev, "Enabling %d VFs\n", num_vfs);
948 	err = ice_ena_vfs(pf, num_vfs);
949 	if (err) {
950 		dev_err(dev, "Failed to enable SR-IOV: %d\n", err);
951 		return err;
952 	}
953 
954 	set_bit(ICE_FLAG_SRIOV_ENA, pf->flags);
955 	return 0;
956 }
957 
958 /**
959  * ice_check_sriov_allowed - check if SR-IOV is allowed based on various checks
960  * @pf: PF to enabled SR-IOV on
961  */
ice_check_sriov_allowed(struct ice_pf * pf)962 static int ice_check_sriov_allowed(struct ice_pf *pf)
963 {
964 	struct device *dev = ice_pf_to_dev(pf);
965 
966 	if (!test_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags)) {
967 		dev_err(dev, "This device is not capable of SR-IOV\n");
968 		return -EOPNOTSUPP;
969 	}
970 
971 	if (ice_is_safe_mode(pf)) {
972 		dev_err(dev, "SR-IOV cannot be configured - Device is in Safe Mode\n");
973 		return -EOPNOTSUPP;
974 	}
975 
976 	if (!ice_pf_state_is_nominal(pf)) {
977 		dev_err(dev, "Cannot enable SR-IOV, device not ready\n");
978 		return -EBUSY;
979 	}
980 
981 	return 0;
982 }
983 
984 /**
985  * ice_sriov_get_vf_total_msix - return number of MSI-X used by VFs
986  * @pdev: pointer to pci_dev struct
987  *
988  * The function is called via sysfs ops
989  */
ice_sriov_get_vf_total_msix(struct pci_dev * pdev)990 u32 ice_sriov_get_vf_total_msix(struct pci_dev *pdev)
991 {
992 	struct ice_pf *pf = pci_get_drvdata(pdev);
993 
994 	return pf->sriov_irq_size - ice_get_max_used_msix_vector(pf);
995 }
996 
ice_sriov_move_base_vector(struct ice_pf * pf,int move)997 static int ice_sriov_move_base_vector(struct ice_pf *pf, int move)
998 {
999 	if (pf->sriov_base_vector - move < ice_get_max_used_msix_vector(pf))
1000 		return -ENOMEM;
1001 
1002 	pf->sriov_base_vector -= move;
1003 	return 0;
1004 }
1005 
ice_sriov_remap_vectors(struct ice_pf * pf,u16 restricted_id)1006 static void ice_sriov_remap_vectors(struct ice_pf *pf, u16 restricted_id)
1007 {
1008 	u16 vf_ids[ICE_MAX_SRIOV_VFS];
1009 	struct ice_vf *tmp_vf;
1010 	int to_remap = 0, bkt;
1011 
1012 	/* For better irqs usage try to remap irqs of VFs
1013 	 * that aren't running yet
1014 	 */
1015 	ice_for_each_vf(pf, bkt, tmp_vf) {
1016 		/* skip VF which is changing the number of MSI-X */
1017 		if (restricted_id == tmp_vf->vf_id ||
1018 		    test_bit(ICE_VF_STATE_ACTIVE, tmp_vf->vf_states))
1019 			continue;
1020 
1021 		ice_dis_vf_mappings(tmp_vf);
1022 		ice_sriov_free_irqs(pf, tmp_vf);
1023 
1024 		vf_ids[to_remap] = tmp_vf->vf_id;
1025 		to_remap += 1;
1026 	}
1027 
1028 	for (int i = 0; i < to_remap; i++) {
1029 		tmp_vf = ice_get_vf_by_id(pf, vf_ids[i]);
1030 		if (!tmp_vf)
1031 			continue;
1032 
1033 		tmp_vf->first_vector_idx =
1034 			ice_sriov_get_irqs(pf, tmp_vf->num_msix);
1035 		/* there is no need to rebuild VSI as we are only changing the
1036 		 * vector indexes not amount of MSI-X or queues
1037 		 */
1038 		ice_ena_vf_mappings(tmp_vf);
1039 		ice_put_vf(tmp_vf);
1040 	}
1041 }
1042 
1043 /**
1044  * ice_sriov_set_msix_vec_count
1045  * @vf_dev: pointer to pci_dev struct of VF device
1046  * @msix_vec_count: new value for MSI-X amount on this VF
1047  *
1048  * Set requested MSI-X, queues and registers for @vf_dev.
1049  *
1050  * First do some sanity checks like if there are any VFs, if the new value
1051  * is correct etc. Then disable old mapping (MSI-X and queues registers), change
1052  * MSI-X and queues, rebuild VSI and enable new mapping.
1053  *
1054  * If it is possible (driver not binded to VF) try to remap also other VFs to
1055  * linearize irqs register usage.
1056  */
ice_sriov_set_msix_vec_count(struct pci_dev * vf_dev,int msix_vec_count)1057 int ice_sriov_set_msix_vec_count(struct pci_dev *vf_dev, int msix_vec_count)
1058 {
1059 	struct pci_dev *pdev = pci_physfn(vf_dev);
1060 	struct ice_pf *pf = pci_get_drvdata(pdev);
1061 	u16 prev_msix, prev_queues, queues;
1062 	bool needs_rebuild = false;
1063 	struct ice_vsi *vsi;
1064 	struct ice_vf *vf;
1065 	int id;
1066 
1067 	if (!ice_get_num_vfs(pf))
1068 		return -ENOENT;
1069 
1070 	if (!msix_vec_count)
1071 		return 0;
1072 
1073 	queues = msix_vec_count;
1074 	/* add 1 MSI-X for OICR */
1075 	msix_vec_count += 1;
1076 
1077 	if (queues > min(ice_get_avail_txq_count(pf),
1078 			 ice_get_avail_rxq_count(pf)))
1079 		return -EINVAL;
1080 
1081 	if (msix_vec_count < ICE_MIN_INTR_PER_VF)
1082 		return -EINVAL;
1083 
1084 	/* Transition of PCI VF function number to function_id */
1085 	for (id = 0; id < pci_num_vf(pdev); id++) {
1086 		if (vf_dev->devfn == pci_iov_virtfn_devfn(pdev, id))
1087 			break;
1088 	}
1089 
1090 	if (id == pci_num_vf(pdev))
1091 		return -ENOENT;
1092 
1093 	vf = ice_get_vf_by_id(pf, id);
1094 
1095 	if (!vf)
1096 		return -ENOENT;
1097 
1098 	vsi = ice_get_vf_vsi(vf);
1099 	if (!vsi)
1100 		return -ENOENT;
1101 
1102 	prev_msix = vf->num_msix;
1103 	prev_queues = vf->num_vf_qs;
1104 
1105 	if (ice_sriov_move_base_vector(pf, msix_vec_count - prev_msix)) {
1106 		ice_put_vf(vf);
1107 		return -ENOSPC;
1108 	}
1109 
1110 	ice_dis_vf_mappings(vf);
1111 	ice_sriov_free_irqs(pf, vf);
1112 
1113 	/* Remap all VFs beside the one is now configured */
1114 	ice_sriov_remap_vectors(pf, vf->vf_id);
1115 
1116 	vf->num_msix = msix_vec_count;
1117 	vf->num_vf_qs = queues;
1118 	vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
1119 	if (vf->first_vector_idx < 0)
1120 		goto unroll;
1121 
1122 	if (ice_vf_reconfig_vsi(vf) || ice_vf_init_host_cfg(vf, vsi)) {
1123 		/* Try to rebuild with previous values */
1124 		needs_rebuild = true;
1125 		goto unroll;
1126 	}
1127 
1128 	dev_info(ice_pf_to_dev(pf),
1129 		 "Changing VF %d resources to %d vectors and %d queues\n",
1130 		 vf->vf_id, vf->num_msix, vf->num_vf_qs);
1131 
1132 	ice_ena_vf_mappings(vf);
1133 	ice_put_vf(vf);
1134 
1135 	return 0;
1136 
1137 unroll:
1138 	dev_info(ice_pf_to_dev(pf),
1139 		 "Can't set %d vectors on VF %d, falling back to %d\n",
1140 		 vf->num_msix, vf->vf_id, prev_msix);
1141 
1142 	vf->num_msix = prev_msix;
1143 	vf->num_vf_qs = prev_queues;
1144 	vf->first_vector_idx = ice_sriov_get_irqs(pf, vf->num_msix);
1145 	if (vf->first_vector_idx < 0)
1146 		return -EINVAL;
1147 
1148 	if (needs_rebuild) {
1149 		ice_vf_reconfig_vsi(vf);
1150 		ice_vf_init_host_cfg(vf, vsi);
1151 	}
1152 
1153 	ice_ena_vf_mappings(vf);
1154 	ice_put_vf(vf);
1155 
1156 	return -EINVAL;
1157 }
1158 
1159 /**
1160  * ice_sriov_configure - Enable or change number of VFs via sysfs
1161  * @pdev: pointer to a pci_dev structure
1162  * @num_vfs: number of VFs to allocate or 0 to free VFs
1163  *
1164  * This function is called when the user updates the number of VFs in sysfs. On
1165  * success return whatever num_vfs was set to by the caller. Return negative on
1166  * failure.
1167  */
ice_sriov_configure(struct pci_dev * pdev,int num_vfs)1168 int ice_sriov_configure(struct pci_dev *pdev, int num_vfs)
1169 {
1170 	struct ice_pf *pf = pci_get_drvdata(pdev);
1171 	struct device *dev = ice_pf_to_dev(pf);
1172 	int err;
1173 
1174 	err = ice_check_sriov_allowed(pf);
1175 	if (err)
1176 		return err;
1177 
1178 	if (!num_vfs) {
1179 		if (!pci_vfs_assigned(pdev)) {
1180 			ice_free_vfs(pf);
1181 			return 0;
1182 		}
1183 
1184 		dev_err(dev, "can't free VFs because some are assigned to VMs.\n");
1185 		return -EBUSY;
1186 	}
1187 
1188 	err = ice_pci_sriov_ena(pf, num_vfs);
1189 	if (err)
1190 		return err;
1191 
1192 	return num_vfs;
1193 }
1194 
1195 /**
1196  * ice_process_vflr_event - Free VF resources via IRQ calls
1197  * @pf: pointer to the PF structure
1198  *
1199  * called from the VFLR IRQ handler to
1200  * free up VF resources and state variables
1201  */
ice_process_vflr_event(struct ice_pf * pf)1202 void ice_process_vflr_event(struct ice_pf *pf)
1203 {
1204 	struct ice_hw *hw = &pf->hw;
1205 	struct ice_vf *vf;
1206 	unsigned int bkt;
1207 	u32 reg;
1208 
1209 	if (!test_and_clear_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
1210 	    !ice_has_vfs(pf))
1211 		return;
1212 
1213 	mutex_lock(&pf->vfs.table_lock);
1214 	ice_for_each_vf(pf, bkt, vf) {
1215 		u32 reg_idx, bit_idx;
1216 
1217 		reg_idx = (hw->func_caps.vf_base_id + vf->vf_id) / 32;
1218 		bit_idx = (hw->func_caps.vf_base_id + vf->vf_id) % 32;
1219 		/* read GLGEN_VFLRSTAT register to find out the flr VFs */
1220 		reg = rd32(hw, GLGEN_VFLRSTAT(reg_idx));
1221 		if (reg & BIT(bit_idx))
1222 			/* GLGEN_VFLRSTAT bit will be cleared in ice_reset_vf */
1223 			ice_reset_vf(vf, ICE_VF_RESET_VFLR | ICE_VF_RESET_LOCK);
1224 	}
1225 	mutex_unlock(&pf->vfs.table_lock);
1226 }
1227 
1228 /**
1229  * ice_get_vf_from_pfq - get the VF who owns the PF space queue passed in
1230  * @pf: PF used to index all VFs
1231  * @pfq: queue index relative to the PF's function space
1232  *
1233  * If no VF is found who owns the pfq then return NULL, otherwise return a
1234  * pointer to the VF who owns the pfq
1235  *
1236  * If this function returns non-NULL, it acquires a reference count of the VF
1237  * structure. The caller is responsible for calling ice_put_vf() to drop this
1238  * reference.
1239  */
ice_get_vf_from_pfq(struct ice_pf * pf,u16 pfq)1240 static struct ice_vf *ice_get_vf_from_pfq(struct ice_pf *pf, u16 pfq)
1241 {
1242 	struct ice_vf *vf;
1243 	unsigned int bkt;
1244 
1245 	rcu_read_lock();
1246 	ice_for_each_vf_rcu(pf, bkt, vf) {
1247 		struct ice_vsi *vsi;
1248 		u16 rxq_idx;
1249 
1250 		vsi = ice_get_vf_vsi(vf);
1251 		if (!vsi)
1252 			continue;
1253 
1254 		ice_for_each_rxq(vsi, rxq_idx)
1255 			if (vsi->rxq_map[rxq_idx] == pfq) {
1256 				struct ice_vf *found;
1257 
1258 				if (kref_get_unless_zero(&vf->refcnt))
1259 					found = vf;
1260 				else
1261 					found = NULL;
1262 				rcu_read_unlock();
1263 				return found;
1264 			}
1265 	}
1266 	rcu_read_unlock();
1267 
1268 	return NULL;
1269 }
1270 
1271 /**
1272  * ice_globalq_to_pfq - convert from global queue index to PF space queue index
1273  * @pf: PF used for conversion
1274  * @globalq: global queue index used to convert to PF space queue index
1275  */
ice_globalq_to_pfq(struct ice_pf * pf,u32 globalq)1276 static u32 ice_globalq_to_pfq(struct ice_pf *pf, u32 globalq)
1277 {
1278 	return globalq - pf->hw.func_caps.common_cap.rxq_first_id;
1279 }
1280 
1281 /**
1282  * ice_vf_lan_overflow_event - handle LAN overflow event for a VF
1283  * @pf: PF that the LAN overflow event happened on
1284  * @event: structure holding the event information for the LAN overflow event
1285  *
1286  * Determine if the LAN overflow event was caused by a VF queue. If it was not
1287  * caused by a VF, do nothing. If a VF caused this LAN overflow event trigger a
1288  * reset on the offending VF.
1289  */
1290 void
ice_vf_lan_overflow_event(struct ice_pf * pf,struct ice_rq_event_info * event)1291 ice_vf_lan_overflow_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1292 {
1293 	u32 gldcb_rtctq, queue;
1294 	struct ice_vf *vf;
1295 
1296 	gldcb_rtctq = le32_to_cpu(event->desc.params.lan_overflow.prtdcb_ruptq);
1297 	dev_dbg(ice_pf_to_dev(pf), "GLDCB_RTCTQ: 0x%08x\n", gldcb_rtctq);
1298 
1299 	/* event returns device global Rx queue number */
1300 	queue = FIELD_GET(GLDCB_RTCTQ_RXQNUM_M, gldcb_rtctq);
1301 
1302 	vf = ice_get_vf_from_pfq(pf, ice_globalq_to_pfq(pf, queue));
1303 	if (!vf)
1304 		return;
1305 
1306 	ice_reset_vf(vf, ICE_VF_RESET_NOTIFY | ICE_VF_RESET_LOCK);
1307 	ice_put_vf(vf);
1308 }
1309 
1310 /**
1311  * ice_set_vf_spoofchk
1312  * @netdev: network interface device structure
1313  * @vf_id: VF identifier
1314  * @ena: flag to enable or disable feature
1315  *
1316  * Enable or disable VF spoof checking
1317  */
ice_set_vf_spoofchk(struct net_device * netdev,int vf_id,bool ena)1318 int ice_set_vf_spoofchk(struct net_device *netdev, int vf_id, bool ena)
1319 {
1320 	struct ice_netdev_priv *np = netdev_priv(netdev);
1321 	struct ice_pf *pf = np->vsi->back;
1322 	struct ice_vsi *vf_vsi;
1323 	struct device *dev;
1324 	struct ice_vf *vf;
1325 	int ret;
1326 
1327 	dev = ice_pf_to_dev(pf);
1328 
1329 	vf = ice_get_vf_by_id(pf, vf_id);
1330 	if (!vf)
1331 		return -EINVAL;
1332 
1333 	ret = ice_check_vf_ready_for_cfg(vf);
1334 	if (ret)
1335 		goto out_put_vf;
1336 
1337 	vf_vsi = ice_get_vf_vsi(vf);
1338 	if (!vf_vsi) {
1339 		netdev_err(netdev, "VSI %d for VF %d is null\n",
1340 			   vf->lan_vsi_idx, vf->vf_id);
1341 		ret = -EINVAL;
1342 		goto out_put_vf;
1343 	}
1344 
1345 	if (vf_vsi->type != ICE_VSI_VF) {
1346 		netdev_err(netdev, "Type %d of VSI %d for VF %d is no ICE_VSI_VF\n",
1347 			   vf_vsi->type, vf_vsi->vsi_num, vf->vf_id);
1348 		ret = -ENODEV;
1349 		goto out_put_vf;
1350 	}
1351 
1352 	if (ena == vf->spoofchk) {
1353 		dev_dbg(dev, "VF spoofchk already %s\n", ena ? "ON" : "OFF");
1354 		ret = 0;
1355 		goto out_put_vf;
1356 	}
1357 
1358 	ret = ice_vsi_apply_spoofchk(vf_vsi, ena);
1359 	if (ret)
1360 		dev_err(dev, "Failed to set spoofchk %s for VF %d VSI %d\n error %d\n",
1361 			ena ? "ON" : "OFF", vf->vf_id, vf_vsi->vsi_num, ret);
1362 	else
1363 		vf->spoofchk = ena;
1364 
1365 out_put_vf:
1366 	ice_put_vf(vf);
1367 	return ret;
1368 }
1369 
1370 /**
1371  * ice_get_vf_cfg
1372  * @netdev: network interface device structure
1373  * @vf_id: VF identifier
1374  * @ivi: VF configuration structure
1375  *
1376  * return VF configuration
1377  */
1378 int
ice_get_vf_cfg(struct net_device * netdev,int vf_id,struct ifla_vf_info * ivi)1379 ice_get_vf_cfg(struct net_device *netdev, int vf_id, struct ifla_vf_info *ivi)
1380 {
1381 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
1382 	struct ice_vf *vf;
1383 	int ret;
1384 
1385 	vf = ice_get_vf_by_id(pf, vf_id);
1386 	if (!vf)
1387 		return -EINVAL;
1388 
1389 	ret = ice_check_vf_ready_for_cfg(vf);
1390 	if (ret)
1391 		goto out_put_vf;
1392 
1393 	ivi->vf = vf_id;
1394 	ether_addr_copy(ivi->mac, vf->hw_lan_addr);
1395 
1396 	/* VF configuration for VLAN and applicable QoS */
1397 	ivi->vlan = ice_vf_get_port_vlan_id(vf);
1398 	ivi->qos = ice_vf_get_port_vlan_prio(vf);
1399 	if (ice_vf_is_port_vlan_ena(vf))
1400 		ivi->vlan_proto = cpu_to_be16(ice_vf_get_port_vlan_tpid(vf));
1401 
1402 	ivi->trusted = vf->trusted;
1403 	ivi->spoofchk = vf->spoofchk;
1404 	if (!vf->link_forced)
1405 		ivi->linkstate = IFLA_VF_LINK_STATE_AUTO;
1406 	else if (vf->link_up)
1407 		ivi->linkstate = IFLA_VF_LINK_STATE_ENABLE;
1408 	else
1409 		ivi->linkstate = IFLA_VF_LINK_STATE_DISABLE;
1410 	ivi->max_tx_rate = vf->max_tx_rate;
1411 	ivi->min_tx_rate = vf->min_tx_rate;
1412 
1413 out_put_vf:
1414 	ice_put_vf(vf);
1415 	return ret;
1416 }
1417 
1418 /**
1419  * __ice_set_vf_mac - program VF MAC address
1420  * @pf: PF to be configure
1421  * @vf_id: VF identifier
1422  * @mac: MAC address
1423  *
1424  * program VF MAC address
1425  * Return: zero on success or an error code on failure
1426  */
__ice_set_vf_mac(struct ice_pf * pf,u16 vf_id,const u8 * mac)1427 int __ice_set_vf_mac(struct ice_pf *pf, u16 vf_id, const u8 *mac)
1428 {
1429 	struct device *dev;
1430 	struct ice_vf *vf;
1431 	int ret;
1432 
1433 	dev = ice_pf_to_dev(pf);
1434 	if (is_multicast_ether_addr(mac)) {
1435 		dev_err(dev, "%pM not a valid unicast address\n", mac);
1436 		return -EINVAL;
1437 	}
1438 
1439 	vf = ice_get_vf_by_id(pf, vf_id);
1440 	if (!vf)
1441 		return -EINVAL;
1442 
1443 	/* nothing left to do, unicast MAC already set */
1444 	if (ether_addr_equal(vf->dev_lan_addr, mac) &&
1445 	    ether_addr_equal(vf->hw_lan_addr, mac)) {
1446 		ret = 0;
1447 		goto out_put_vf;
1448 	}
1449 
1450 	ret = ice_check_vf_ready_for_cfg(vf);
1451 	if (ret)
1452 		goto out_put_vf;
1453 
1454 	mutex_lock(&vf->cfg_lock);
1455 
1456 	/* VF is notified of its new MAC via the PF's response to the
1457 	 * VIRTCHNL_OP_GET_VF_RESOURCES message after the VF has been reset
1458 	 */
1459 	ether_addr_copy(vf->dev_lan_addr, mac);
1460 	ether_addr_copy(vf->hw_lan_addr, mac);
1461 	if (is_zero_ether_addr(mac)) {
1462 		/* VF will send VIRTCHNL_OP_ADD_ETH_ADDR message with its MAC */
1463 		vf->pf_set_mac = false;
1464 		dev_info(dev, "Removing MAC on VF %d. VF driver will be reinitialized\n",
1465 			 vf->vf_id);
1466 	} else {
1467 		/* PF will add MAC rule for the VF */
1468 		vf->pf_set_mac = true;
1469 		dev_info(dev, "Setting MAC %pM on VF %d. VF driver will be reinitialized\n",
1470 			 mac, vf_id);
1471 	}
1472 
1473 	ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1474 	mutex_unlock(&vf->cfg_lock);
1475 
1476 out_put_vf:
1477 	ice_put_vf(vf);
1478 	return ret;
1479 }
1480 
1481 /**
1482  * ice_set_vf_mac - .ndo_set_vf_mac handler
1483  * @netdev: network interface device structure
1484  * @vf_id: VF identifier
1485  * @mac: MAC address
1486  *
1487  * program VF MAC address
1488  * Return: zero on success or an error code on failure
1489  */
ice_set_vf_mac(struct net_device * netdev,int vf_id,u8 * mac)1490 int ice_set_vf_mac(struct net_device *netdev, int vf_id, u8 *mac)
1491 {
1492 	return __ice_set_vf_mac(ice_netdev_to_pf(netdev), vf_id, mac);
1493 }
1494 
1495 /**
1496  * ice_set_vf_trust
1497  * @netdev: network interface device structure
1498  * @vf_id: VF identifier
1499  * @trusted: Boolean value to enable/disable trusted VF
1500  *
1501  * Enable or disable a given VF as trusted
1502  */
ice_set_vf_trust(struct net_device * netdev,int vf_id,bool trusted)1503 int ice_set_vf_trust(struct net_device *netdev, int vf_id, bool trusted)
1504 {
1505 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
1506 	struct ice_vf *vf;
1507 	int ret;
1508 
1509 	vf = ice_get_vf_by_id(pf, vf_id);
1510 	if (!vf)
1511 		return -EINVAL;
1512 
1513 	if (ice_is_eswitch_mode_switchdev(pf)) {
1514 		dev_info(ice_pf_to_dev(pf), "Trusted VF is forbidden in switchdev mode\n");
1515 		return -EOPNOTSUPP;
1516 	}
1517 
1518 	ret = ice_check_vf_ready_for_cfg(vf);
1519 	if (ret)
1520 		goto out_put_vf;
1521 
1522 	/* Check if already trusted */
1523 	if (trusted == vf->trusted) {
1524 		ret = 0;
1525 		goto out_put_vf;
1526 	}
1527 
1528 	mutex_lock(&vf->cfg_lock);
1529 
1530 	vf->trusted = trusted;
1531 	ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1532 	dev_info(ice_pf_to_dev(pf), "VF %u is now %strusted\n",
1533 		 vf_id, trusted ? "" : "un");
1534 
1535 	mutex_unlock(&vf->cfg_lock);
1536 
1537 out_put_vf:
1538 	ice_put_vf(vf);
1539 	return ret;
1540 }
1541 
1542 /**
1543  * ice_set_vf_link_state
1544  * @netdev: network interface device structure
1545  * @vf_id: VF identifier
1546  * @link_state: required link state
1547  *
1548  * Set VF's link state, irrespective of physical link state status
1549  */
ice_set_vf_link_state(struct net_device * netdev,int vf_id,int link_state)1550 int ice_set_vf_link_state(struct net_device *netdev, int vf_id, int link_state)
1551 {
1552 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
1553 	struct ice_vf *vf;
1554 	int ret;
1555 
1556 	vf = ice_get_vf_by_id(pf, vf_id);
1557 	if (!vf)
1558 		return -EINVAL;
1559 
1560 	ret = ice_check_vf_ready_for_cfg(vf);
1561 	if (ret)
1562 		goto out_put_vf;
1563 
1564 	switch (link_state) {
1565 	case IFLA_VF_LINK_STATE_AUTO:
1566 		vf->link_forced = false;
1567 		break;
1568 	case IFLA_VF_LINK_STATE_ENABLE:
1569 		vf->link_forced = true;
1570 		vf->link_up = true;
1571 		break;
1572 	case IFLA_VF_LINK_STATE_DISABLE:
1573 		vf->link_forced = true;
1574 		vf->link_up = false;
1575 		break;
1576 	default:
1577 		ret = -EINVAL;
1578 		goto out_put_vf;
1579 	}
1580 
1581 	ice_vc_notify_vf_link_state(vf);
1582 
1583 out_put_vf:
1584 	ice_put_vf(vf);
1585 	return ret;
1586 }
1587 
1588 /**
1589  * ice_calc_all_vfs_min_tx_rate - calculate cumulative min Tx rate on all VFs
1590  * @pf: PF associated with VFs
1591  */
ice_calc_all_vfs_min_tx_rate(struct ice_pf * pf)1592 static int ice_calc_all_vfs_min_tx_rate(struct ice_pf *pf)
1593 {
1594 	struct ice_vf *vf;
1595 	unsigned int bkt;
1596 	int rate = 0;
1597 
1598 	rcu_read_lock();
1599 	ice_for_each_vf_rcu(pf, bkt, vf)
1600 		rate += vf->min_tx_rate;
1601 	rcu_read_unlock();
1602 
1603 	return rate;
1604 }
1605 
1606 /**
1607  * ice_min_tx_rate_oversubscribed - check if min Tx rate causes oversubscription
1608  * @vf: VF trying to configure min_tx_rate
1609  * @min_tx_rate: min Tx rate in Mbps
1610  *
1611  * Check if the min_tx_rate being passed in will cause oversubscription of total
1612  * min_tx_rate based on the current link speed and all other VFs configured
1613  * min_tx_rate
1614  *
1615  * Return true if the passed min_tx_rate would cause oversubscription, else
1616  * return false
1617  */
1618 static bool
ice_min_tx_rate_oversubscribed(struct ice_vf * vf,int min_tx_rate)1619 ice_min_tx_rate_oversubscribed(struct ice_vf *vf, int min_tx_rate)
1620 {
1621 	struct ice_vsi *vsi = ice_get_vf_vsi(vf);
1622 	int all_vfs_min_tx_rate;
1623 	int link_speed_mbps;
1624 
1625 	if (WARN_ON(!vsi))
1626 		return false;
1627 
1628 	link_speed_mbps = ice_get_link_speed_mbps(vsi);
1629 	all_vfs_min_tx_rate = ice_calc_all_vfs_min_tx_rate(vf->pf);
1630 
1631 	/* this VF's previous rate is being overwritten */
1632 	all_vfs_min_tx_rate -= vf->min_tx_rate;
1633 
1634 	if (all_vfs_min_tx_rate + min_tx_rate > link_speed_mbps) {
1635 		dev_err(ice_pf_to_dev(vf->pf), "min_tx_rate of %d Mbps on VF %u would cause oversubscription of %d Mbps based on the current link speed %d Mbps\n",
1636 			min_tx_rate, vf->vf_id,
1637 			all_vfs_min_tx_rate + min_tx_rate - link_speed_mbps,
1638 			link_speed_mbps);
1639 		return true;
1640 	}
1641 
1642 	return false;
1643 }
1644 
1645 /**
1646  * ice_set_vf_bw - set min/max VF bandwidth
1647  * @netdev: network interface device structure
1648  * @vf_id: VF identifier
1649  * @min_tx_rate: Minimum Tx rate in Mbps
1650  * @max_tx_rate: Maximum Tx rate in Mbps
1651  */
1652 int
ice_set_vf_bw(struct net_device * netdev,int vf_id,int min_tx_rate,int max_tx_rate)1653 ice_set_vf_bw(struct net_device *netdev, int vf_id, int min_tx_rate,
1654 	      int max_tx_rate)
1655 {
1656 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
1657 	struct ice_vsi *vsi;
1658 	struct device *dev;
1659 	struct ice_vf *vf;
1660 	int ret;
1661 
1662 	dev = ice_pf_to_dev(pf);
1663 
1664 	vf = ice_get_vf_by_id(pf, vf_id);
1665 	if (!vf)
1666 		return -EINVAL;
1667 
1668 	ret = ice_check_vf_ready_for_cfg(vf);
1669 	if (ret)
1670 		goto out_put_vf;
1671 
1672 	vsi = ice_get_vf_vsi(vf);
1673 	if (!vsi) {
1674 		ret = -EINVAL;
1675 		goto out_put_vf;
1676 	}
1677 
1678 	if (min_tx_rate && ice_is_dcb_active(pf)) {
1679 		dev_err(dev, "DCB on PF is currently enabled. VF min Tx rate limiting not allowed on this PF.\n");
1680 		ret = -EOPNOTSUPP;
1681 		goto out_put_vf;
1682 	}
1683 
1684 	if (ice_min_tx_rate_oversubscribed(vf, min_tx_rate)) {
1685 		ret = -EINVAL;
1686 		goto out_put_vf;
1687 	}
1688 
1689 	if (vf->min_tx_rate != (unsigned int)min_tx_rate) {
1690 		ret = ice_set_min_bw_limit(vsi, (u64)min_tx_rate * 1000);
1691 		if (ret) {
1692 			dev_err(dev, "Unable to set min-tx-rate for VF %d\n",
1693 				vf->vf_id);
1694 			goto out_put_vf;
1695 		}
1696 
1697 		vf->min_tx_rate = min_tx_rate;
1698 	}
1699 
1700 	if (vf->max_tx_rate != (unsigned int)max_tx_rate) {
1701 		ret = ice_set_max_bw_limit(vsi, (u64)max_tx_rate * 1000);
1702 		if (ret) {
1703 			dev_err(dev, "Unable to set max-tx-rate for VF %d\n",
1704 				vf->vf_id);
1705 			goto out_put_vf;
1706 		}
1707 
1708 		vf->max_tx_rate = max_tx_rate;
1709 	}
1710 
1711 out_put_vf:
1712 	ice_put_vf(vf);
1713 	return ret;
1714 }
1715 
1716 /**
1717  * ice_get_vf_stats - populate some stats for the VF
1718  * @netdev: the netdev of the PF
1719  * @vf_id: the host OS identifier (0-255)
1720  * @vf_stats: pointer to the OS memory to be initialized
1721  */
ice_get_vf_stats(struct net_device * netdev,int vf_id,struct ifla_vf_stats * vf_stats)1722 int ice_get_vf_stats(struct net_device *netdev, int vf_id,
1723 		     struct ifla_vf_stats *vf_stats)
1724 {
1725 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
1726 	struct ice_eth_stats *stats;
1727 	struct ice_vsi *vsi;
1728 	struct ice_vf *vf;
1729 	int ret;
1730 
1731 	vf = ice_get_vf_by_id(pf, vf_id);
1732 	if (!vf)
1733 		return -EINVAL;
1734 
1735 	ret = ice_check_vf_ready_for_cfg(vf);
1736 	if (ret)
1737 		goto out_put_vf;
1738 
1739 	vsi = ice_get_vf_vsi(vf);
1740 	if (!vsi) {
1741 		ret = -EINVAL;
1742 		goto out_put_vf;
1743 	}
1744 
1745 	ice_update_eth_stats(vsi);
1746 	stats = &vsi->eth_stats;
1747 
1748 	memset(vf_stats, 0, sizeof(*vf_stats));
1749 
1750 	vf_stats->rx_packets = stats->rx_unicast + stats->rx_broadcast +
1751 		stats->rx_multicast;
1752 	vf_stats->tx_packets = stats->tx_unicast + stats->tx_broadcast +
1753 		stats->tx_multicast;
1754 	vf_stats->rx_bytes   = stats->rx_bytes;
1755 	vf_stats->tx_bytes   = stats->tx_bytes;
1756 	vf_stats->broadcast  = stats->rx_broadcast;
1757 	vf_stats->multicast  = stats->rx_multicast;
1758 	vf_stats->rx_dropped = stats->rx_discards;
1759 	vf_stats->tx_dropped = stats->tx_discards;
1760 
1761 out_put_vf:
1762 	ice_put_vf(vf);
1763 	return ret;
1764 }
1765 
1766 /**
1767  * ice_is_supported_port_vlan_proto - make sure the vlan_proto is supported
1768  * @hw: hardware structure used to check the VLAN mode
1769  * @vlan_proto: VLAN TPID being checked
1770  *
1771  * If the device is configured in Double VLAN Mode (DVM), then both ETH_P_8021Q
1772  * and ETH_P_8021AD are supported. If the device is configured in Single VLAN
1773  * Mode (SVM), then only ETH_P_8021Q is supported.
1774  */
1775 static bool
ice_is_supported_port_vlan_proto(struct ice_hw * hw,u16 vlan_proto)1776 ice_is_supported_port_vlan_proto(struct ice_hw *hw, u16 vlan_proto)
1777 {
1778 	bool is_supported = false;
1779 
1780 	switch (vlan_proto) {
1781 	case ETH_P_8021Q:
1782 		is_supported = true;
1783 		break;
1784 	case ETH_P_8021AD:
1785 		if (ice_is_dvm_ena(hw))
1786 			is_supported = true;
1787 		break;
1788 	}
1789 
1790 	return is_supported;
1791 }
1792 
1793 /**
1794  * ice_set_vf_port_vlan
1795  * @netdev: network interface device structure
1796  * @vf_id: VF identifier
1797  * @vlan_id: VLAN ID being set
1798  * @qos: priority setting
1799  * @vlan_proto: VLAN protocol
1800  *
1801  * program VF Port VLAN ID and/or QoS
1802  */
1803 int
ice_set_vf_port_vlan(struct net_device * netdev,int vf_id,u16 vlan_id,u8 qos,__be16 vlan_proto)1804 ice_set_vf_port_vlan(struct net_device *netdev, int vf_id, u16 vlan_id, u8 qos,
1805 		     __be16 vlan_proto)
1806 {
1807 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
1808 	u16 local_vlan_proto = ntohs(vlan_proto);
1809 	struct device *dev;
1810 	struct ice_vf *vf;
1811 	int ret;
1812 
1813 	dev = ice_pf_to_dev(pf);
1814 
1815 	if (vlan_id >= VLAN_N_VID || qos > 7) {
1816 		dev_err(dev, "Invalid Port VLAN parameters for VF %d, ID %d, QoS %d\n",
1817 			vf_id, vlan_id, qos);
1818 		return -EINVAL;
1819 	}
1820 
1821 	if (!ice_is_supported_port_vlan_proto(&pf->hw, local_vlan_proto)) {
1822 		dev_err(dev, "VF VLAN protocol 0x%04x is not supported\n",
1823 			local_vlan_proto);
1824 		return -EPROTONOSUPPORT;
1825 	}
1826 
1827 	vf = ice_get_vf_by_id(pf, vf_id);
1828 	if (!vf)
1829 		return -EINVAL;
1830 
1831 	ret = ice_check_vf_ready_for_cfg(vf);
1832 	if (ret)
1833 		goto out_put_vf;
1834 
1835 	if (ice_vf_get_port_vlan_prio(vf) == qos &&
1836 	    ice_vf_get_port_vlan_tpid(vf) == local_vlan_proto &&
1837 	    ice_vf_get_port_vlan_id(vf) == vlan_id) {
1838 		/* duplicate request, so just return success */
1839 		dev_dbg(dev, "Duplicate port VLAN %u, QoS %u, TPID 0x%04x request\n",
1840 			vlan_id, qos, local_vlan_proto);
1841 		ret = 0;
1842 		goto out_put_vf;
1843 	}
1844 
1845 	mutex_lock(&vf->cfg_lock);
1846 
1847 	vf->port_vlan_info = ICE_VLAN(local_vlan_proto, vlan_id, qos);
1848 	if (ice_vf_is_port_vlan_ena(vf))
1849 		dev_info(dev, "Setting VLAN %u, QoS %u, TPID 0x%04x on VF %d\n",
1850 			 vlan_id, qos, local_vlan_proto, vf_id);
1851 	else
1852 		dev_info(dev, "Clearing port VLAN on VF %d\n", vf_id);
1853 
1854 	ice_reset_vf(vf, ICE_VF_RESET_NOTIFY);
1855 	mutex_unlock(&vf->cfg_lock);
1856 
1857 out_put_vf:
1858 	ice_put_vf(vf);
1859 	return ret;
1860 }
1861 
1862 /**
1863  * ice_print_vf_rx_mdd_event - print VF Rx malicious driver detect event
1864  * @vf: pointer to the VF structure
1865  */
ice_print_vf_rx_mdd_event(struct ice_vf * vf)1866 void ice_print_vf_rx_mdd_event(struct ice_vf *vf)
1867 {
1868 	struct ice_pf *pf = vf->pf;
1869 	struct device *dev;
1870 
1871 	dev = ice_pf_to_dev(pf);
1872 
1873 	dev_info(dev, "%d Rx Malicious Driver Detection events detected on PF %d VF %d MAC %pM. mdd-auto-reset-vfs=%s\n",
1874 		 vf->mdd_rx_events.count, pf->hw.pf_id, vf->vf_id,
1875 		 vf->dev_lan_addr,
1876 		 test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)
1877 			  ? "on" : "off");
1878 }
1879 
1880 /**
1881  * ice_print_vf_tx_mdd_event - print VF Tx malicious driver detect event
1882  * @vf: pointer to the VF structure
1883  */
ice_print_vf_tx_mdd_event(struct ice_vf * vf)1884 void ice_print_vf_tx_mdd_event(struct ice_vf *vf)
1885 {
1886 	struct ice_pf *pf = vf->pf;
1887 	struct device *dev;
1888 
1889 	dev = ice_pf_to_dev(pf);
1890 
1891 	dev_info(dev, "%d Tx Malicious Driver Detection events detected on PF %d VF %d MAC %pM. mdd-auto-reset-vfs=%s\n",
1892 		 vf->mdd_tx_events.count, pf->hw.pf_id, vf->vf_id,
1893 		 vf->dev_lan_addr,
1894 		 test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)
1895 			  ? "on" : "off");
1896 }
1897 
1898 /**
1899  * ice_print_vfs_mdd_events - print VFs malicious driver detect event
1900  * @pf: pointer to the PF structure
1901  *
1902  * Called from ice_handle_mdd_event to rate limit and print VFs MDD events.
1903  */
ice_print_vfs_mdd_events(struct ice_pf * pf)1904 void ice_print_vfs_mdd_events(struct ice_pf *pf)
1905 {
1906 	struct ice_vf *vf;
1907 	unsigned int bkt;
1908 
1909 	/* check that there are pending MDD events to print */
1910 	if (!test_and_clear_bit(ICE_MDD_VF_PRINT_PENDING, pf->state))
1911 		return;
1912 
1913 	/* VF MDD event logs are rate limited to one second intervals */
1914 	if (time_is_after_jiffies(pf->vfs.last_printed_mdd_jiffies + HZ * 1))
1915 		return;
1916 
1917 	pf->vfs.last_printed_mdd_jiffies = jiffies;
1918 
1919 	mutex_lock(&pf->vfs.table_lock);
1920 	ice_for_each_vf(pf, bkt, vf) {
1921 		/* only print Rx MDD event message if there are new events */
1922 		if (vf->mdd_rx_events.count != vf->mdd_rx_events.last_printed) {
1923 			vf->mdd_rx_events.last_printed =
1924 							vf->mdd_rx_events.count;
1925 			ice_print_vf_rx_mdd_event(vf);
1926 		}
1927 
1928 		/* only print Tx MDD event message if there are new events */
1929 		if (vf->mdd_tx_events.count != vf->mdd_tx_events.last_printed) {
1930 			vf->mdd_tx_events.last_printed =
1931 							vf->mdd_tx_events.count;
1932 			ice_print_vf_tx_mdd_event(vf);
1933 		}
1934 	}
1935 	mutex_unlock(&pf->vfs.table_lock);
1936 }
1937 
1938 /**
1939  * ice_restore_all_vfs_msi_state - restore VF MSI state after PF FLR
1940  * @pf: pointer to the PF structure
1941  *
1942  * Called when recovering from a PF FLR to restore interrupt capability to
1943  * the VFs.
1944  */
ice_restore_all_vfs_msi_state(struct ice_pf * pf)1945 void ice_restore_all_vfs_msi_state(struct ice_pf *pf)
1946 {
1947 	struct ice_vf *vf;
1948 	u32 bkt;
1949 
1950 	ice_for_each_vf(pf, bkt, vf)
1951 		pci_restore_msi_state(vf->vfdev);
1952 }
1953