xref: /linux/drivers/net/ethernet/intel/ice/ice_main.c (revision 42422993cf28d456778ee9168d73758ec037cd51)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018-2023, Intel Corporation. */
3 
4 /* Intel(R) Ethernet Connection E800 Series Linux Driver */
5 
6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
7 
8 #include <generated/utsrelease.h>
9 #include <linux/crash_dump.h>
10 #include "ice.h"
11 #include "ice_base.h"
12 #include "ice_lib.h"
13 #include "ice_fltr.h"
14 #include "ice_dcb_lib.h"
15 #include "ice_dcb_nl.h"
16 #include "ice_devlink.h"
17 /* Including ice_trace.h with CREATE_TRACE_POINTS defined will generate the
18  * ice tracepoint functions. This must be done exactly once across the
19  * ice driver.
20  */
21 #define CREATE_TRACE_POINTS
22 #include "ice_trace.h"
23 #include "ice_eswitch.h"
24 #include "ice_tc_lib.h"
25 #include "ice_vsi_vlan_ops.h"
26 #include <net/xdp_sock_drv.h>
27 
28 #define DRV_SUMMARY	"Intel(R) Ethernet Connection E800 Series Linux Driver"
29 static const char ice_driver_string[] = DRV_SUMMARY;
30 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
31 
32 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
33 #define ICE_DDP_PKG_PATH	"intel/ice/ddp/"
34 #define ICE_DDP_PKG_FILE	ICE_DDP_PKG_PATH "ice.pkg"
35 
36 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
37 MODULE_DESCRIPTION(DRV_SUMMARY);
38 MODULE_LICENSE("GPL v2");
39 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
40 
41 static int debug = -1;
42 module_param(debug, int, 0644);
43 #ifndef CONFIG_DYNAMIC_DEBUG
44 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
45 #else
46 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
47 #endif /* !CONFIG_DYNAMIC_DEBUG */
48 
49 DEFINE_STATIC_KEY_FALSE(ice_xdp_locking_key);
50 EXPORT_SYMBOL(ice_xdp_locking_key);
51 
52 /**
53  * ice_hw_to_dev - Get device pointer from the hardware structure
54  * @hw: pointer to the device HW structure
55  *
56  * Used to access the device pointer from compilation units which can't easily
57  * include the definition of struct ice_pf without leading to circular header
58  * dependencies.
59  */
60 struct device *ice_hw_to_dev(struct ice_hw *hw)
61 {
62 	struct ice_pf *pf = container_of(hw, struct ice_pf, hw);
63 
64 	return &pf->pdev->dev;
65 }
66 
67 static struct workqueue_struct *ice_wq;
68 struct workqueue_struct *ice_lag_wq;
69 static const struct net_device_ops ice_netdev_safe_mode_ops;
70 static const struct net_device_ops ice_netdev_ops;
71 
72 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
73 
74 static void ice_vsi_release_all(struct ice_pf *pf);
75 
76 static int ice_rebuild_channels(struct ice_pf *pf);
77 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_adv_fltr);
78 
79 static int
80 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
81 		     void *cb_priv, enum tc_setup_type type, void *type_data,
82 		     void *data,
83 		     void (*cleanup)(struct flow_block_cb *block_cb));
84 
85 bool netif_is_ice(const struct net_device *dev)
86 {
87 	return dev && (dev->netdev_ops == &ice_netdev_ops);
88 }
89 
90 /**
91  * ice_get_tx_pending - returns number of Tx descriptors not processed
92  * @ring: the ring of descriptors
93  */
94 static u16 ice_get_tx_pending(struct ice_tx_ring *ring)
95 {
96 	u16 head, tail;
97 
98 	head = ring->next_to_clean;
99 	tail = ring->next_to_use;
100 
101 	if (head != tail)
102 		return (head < tail) ?
103 			tail - head : (tail + ring->count - head);
104 	return 0;
105 }
106 
107 /**
108  * ice_check_for_hang_subtask - check for and recover hung queues
109  * @pf: pointer to PF struct
110  */
111 static void ice_check_for_hang_subtask(struct ice_pf *pf)
112 {
113 	struct ice_vsi *vsi = NULL;
114 	struct ice_hw *hw;
115 	unsigned int i;
116 	int packets;
117 	u32 v;
118 
119 	ice_for_each_vsi(pf, v)
120 		if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
121 			vsi = pf->vsi[v];
122 			break;
123 		}
124 
125 	if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
126 		return;
127 
128 	if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
129 		return;
130 
131 	hw = &vsi->back->hw;
132 
133 	ice_for_each_txq(vsi, i) {
134 		struct ice_tx_ring *tx_ring = vsi->tx_rings[i];
135 		struct ice_ring_stats *ring_stats;
136 
137 		if (!tx_ring)
138 			continue;
139 		if (ice_ring_ch_enabled(tx_ring))
140 			continue;
141 
142 		ring_stats = tx_ring->ring_stats;
143 		if (!ring_stats)
144 			continue;
145 
146 		if (tx_ring->desc) {
147 			/* If packet counter has not changed the queue is
148 			 * likely stalled, so force an interrupt for this
149 			 * queue.
150 			 *
151 			 * prev_pkt would be negative if there was no
152 			 * pending work.
153 			 */
154 			packets = ring_stats->stats.pkts & INT_MAX;
155 			if (ring_stats->tx_stats.prev_pkt == packets) {
156 				/* Trigger sw interrupt to revive the queue */
157 				ice_trigger_sw_intr(hw, tx_ring->q_vector);
158 				continue;
159 			}
160 
161 			/* Memory barrier between read of packet count and call
162 			 * to ice_get_tx_pending()
163 			 */
164 			smp_rmb();
165 			ring_stats->tx_stats.prev_pkt =
166 			    ice_get_tx_pending(tx_ring) ? packets : -1;
167 		}
168 	}
169 }
170 
171 /**
172  * ice_init_mac_fltr - Set initial MAC filters
173  * @pf: board private structure
174  *
175  * Set initial set of MAC filters for PF VSI; configure filters for permanent
176  * address and broadcast address. If an error is encountered, netdevice will be
177  * unregistered.
178  */
179 static int ice_init_mac_fltr(struct ice_pf *pf)
180 {
181 	struct ice_vsi *vsi;
182 	u8 *perm_addr;
183 
184 	vsi = ice_get_main_vsi(pf);
185 	if (!vsi)
186 		return -EINVAL;
187 
188 	perm_addr = vsi->port_info->mac.perm_addr;
189 	return ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
190 }
191 
192 /**
193  * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
194  * @netdev: the net device on which the sync is happening
195  * @addr: MAC address to sync
196  *
197  * This is a callback function which is called by the in kernel device sync
198  * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
199  * populates the tmp_sync_list, which is later used by ice_add_mac to add the
200  * MAC filters from the hardware.
201  */
202 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
203 {
204 	struct ice_netdev_priv *np = netdev_priv(netdev);
205 	struct ice_vsi *vsi = np->vsi;
206 
207 	if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
208 				     ICE_FWD_TO_VSI))
209 		return -EINVAL;
210 
211 	return 0;
212 }
213 
214 /**
215  * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
216  * @netdev: the net device on which the unsync is happening
217  * @addr: MAC address to unsync
218  *
219  * This is a callback function which is called by the in kernel device unsync
220  * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
221  * populates the tmp_unsync_list, which is later used by ice_remove_mac to
222  * delete the MAC filters from the hardware.
223  */
224 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
225 {
226 	struct ice_netdev_priv *np = netdev_priv(netdev);
227 	struct ice_vsi *vsi = np->vsi;
228 
229 	/* Under some circumstances, we might receive a request to delete our
230 	 * own device address from our uc list. Because we store the device
231 	 * address in the VSI's MAC filter list, we need to ignore such
232 	 * requests and not delete our device address from this list.
233 	 */
234 	if (ether_addr_equal(addr, netdev->dev_addr))
235 		return 0;
236 
237 	if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
238 				     ICE_FWD_TO_VSI))
239 		return -EINVAL;
240 
241 	return 0;
242 }
243 
244 /**
245  * ice_vsi_fltr_changed - check if filter state changed
246  * @vsi: VSI to be checked
247  *
248  * returns true if filter state has changed, false otherwise.
249  */
250 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
251 {
252 	return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
253 	       test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
254 }
255 
256 /**
257  * ice_set_promisc - Enable promiscuous mode for a given PF
258  * @vsi: the VSI being configured
259  * @promisc_m: mask of promiscuous config bits
260  *
261  */
262 static int ice_set_promisc(struct ice_vsi *vsi, u8 promisc_m)
263 {
264 	int status;
265 
266 	if (vsi->type != ICE_VSI_PF)
267 		return 0;
268 
269 	if (ice_vsi_has_non_zero_vlans(vsi)) {
270 		promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
271 		status = ice_fltr_set_vlan_vsi_promisc(&vsi->back->hw, vsi,
272 						       promisc_m);
273 	} else {
274 		status = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
275 						  promisc_m, 0);
276 	}
277 	if (status && status != -EEXIST)
278 		return status;
279 
280 	netdev_dbg(vsi->netdev, "set promisc filter bits for VSI %i: 0x%x\n",
281 		   vsi->vsi_num, promisc_m);
282 	return 0;
283 }
284 
285 /**
286  * ice_clear_promisc - Disable promiscuous mode for a given PF
287  * @vsi: the VSI being configured
288  * @promisc_m: mask of promiscuous config bits
289  *
290  */
291 static int ice_clear_promisc(struct ice_vsi *vsi, u8 promisc_m)
292 {
293 	int status;
294 
295 	if (vsi->type != ICE_VSI_PF)
296 		return 0;
297 
298 	if (ice_vsi_has_non_zero_vlans(vsi)) {
299 		promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
300 		status = ice_fltr_clear_vlan_vsi_promisc(&vsi->back->hw, vsi,
301 							 promisc_m);
302 	} else {
303 		status = ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
304 						    promisc_m, 0);
305 	}
306 
307 	netdev_dbg(vsi->netdev, "clear promisc filter bits for VSI %i: 0x%x\n",
308 		   vsi->vsi_num, promisc_m);
309 	return status;
310 }
311 
312 /**
313  * ice_vsi_sync_fltr - Update the VSI filter list to the HW
314  * @vsi: ptr to the VSI
315  *
316  * Push any outstanding VSI filter changes through the AdminQ.
317  */
318 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
319 {
320 	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
321 	struct device *dev = ice_pf_to_dev(vsi->back);
322 	struct net_device *netdev = vsi->netdev;
323 	bool promisc_forced_on = false;
324 	struct ice_pf *pf = vsi->back;
325 	struct ice_hw *hw = &pf->hw;
326 	u32 changed_flags = 0;
327 	int err;
328 
329 	if (!vsi->netdev)
330 		return -EINVAL;
331 
332 	while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
333 		usleep_range(1000, 2000);
334 
335 	changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
336 	vsi->current_netdev_flags = vsi->netdev->flags;
337 
338 	INIT_LIST_HEAD(&vsi->tmp_sync_list);
339 	INIT_LIST_HEAD(&vsi->tmp_unsync_list);
340 
341 	if (ice_vsi_fltr_changed(vsi)) {
342 		clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
343 		clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
344 
345 		/* grab the netdev's addr_list_lock */
346 		netif_addr_lock_bh(netdev);
347 		__dev_uc_sync(netdev, ice_add_mac_to_sync_list,
348 			      ice_add_mac_to_unsync_list);
349 		__dev_mc_sync(netdev, ice_add_mac_to_sync_list,
350 			      ice_add_mac_to_unsync_list);
351 		/* our temp lists are populated. release lock */
352 		netif_addr_unlock_bh(netdev);
353 	}
354 
355 	/* Remove MAC addresses in the unsync list */
356 	err = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
357 	ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
358 	if (err) {
359 		netdev_err(netdev, "Failed to delete MAC filters\n");
360 		/* if we failed because of alloc failures, just bail */
361 		if (err == -ENOMEM)
362 			goto out;
363 	}
364 
365 	/* Add MAC addresses in the sync list */
366 	err = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
367 	ice_fltr_free_list(dev, &vsi->tmp_sync_list);
368 	/* If filter is added successfully or already exists, do not go into
369 	 * 'if' condition and report it as error. Instead continue processing
370 	 * rest of the function.
371 	 */
372 	if (err && err != -EEXIST) {
373 		netdev_err(netdev, "Failed to add MAC filters\n");
374 		/* If there is no more space for new umac filters, VSI
375 		 * should go into promiscuous mode. There should be some
376 		 * space reserved for promiscuous filters.
377 		 */
378 		if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
379 		    !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
380 				      vsi->state)) {
381 			promisc_forced_on = true;
382 			netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
383 				    vsi->vsi_num);
384 		} else {
385 			goto out;
386 		}
387 	}
388 	err = 0;
389 	/* check for changes in promiscuous modes */
390 	if (changed_flags & IFF_ALLMULTI) {
391 		if (vsi->current_netdev_flags & IFF_ALLMULTI) {
392 			err = ice_set_promisc(vsi, ICE_MCAST_PROMISC_BITS);
393 			if (err) {
394 				vsi->current_netdev_flags &= ~IFF_ALLMULTI;
395 				goto out_promisc;
396 			}
397 		} else {
398 			/* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
399 			err = ice_clear_promisc(vsi, ICE_MCAST_PROMISC_BITS);
400 			if (err) {
401 				vsi->current_netdev_flags |= IFF_ALLMULTI;
402 				goto out_promisc;
403 			}
404 		}
405 	}
406 
407 	if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
408 	    test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
409 		clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
410 		if (vsi->current_netdev_flags & IFF_PROMISC) {
411 			/* Apply Rx filter rule to get traffic from wire */
412 			if (!ice_is_dflt_vsi_in_use(vsi->port_info)) {
413 				err = ice_set_dflt_vsi(vsi);
414 				if (err && err != -EEXIST) {
415 					netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
416 						   err, vsi->vsi_num);
417 					vsi->current_netdev_flags &=
418 						~IFF_PROMISC;
419 					goto out_promisc;
420 				}
421 				err = 0;
422 				vlan_ops->dis_rx_filtering(vsi);
423 
424 				/* promiscuous mode implies allmulticast so
425 				 * that VSIs that are in promiscuous mode are
426 				 * subscribed to multicast packets coming to
427 				 * the port
428 				 */
429 				err = ice_set_promisc(vsi,
430 						      ICE_MCAST_PROMISC_BITS);
431 				if (err)
432 					goto out_promisc;
433 			}
434 		} else {
435 			/* Clear Rx filter to remove traffic from wire */
436 			if (ice_is_vsi_dflt_vsi(vsi)) {
437 				err = ice_clear_dflt_vsi(vsi);
438 				if (err) {
439 					netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
440 						   err, vsi->vsi_num);
441 					vsi->current_netdev_flags |=
442 						IFF_PROMISC;
443 					goto out_promisc;
444 				}
445 				if (vsi->netdev->features &
446 				    NETIF_F_HW_VLAN_CTAG_FILTER)
447 					vlan_ops->ena_rx_filtering(vsi);
448 			}
449 
450 			/* disable allmulti here, but only if allmulti is not
451 			 * still enabled for the netdev
452 			 */
453 			if (!(vsi->current_netdev_flags & IFF_ALLMULTI)) {
454 				err = ice_clear_promisc(vsi,
455 							ICE_MCAST_PROMISC_BITS);
456 				if (err) {
457 					netdev_err(netdev, "Error %d clearing multicast promiscuous on VSI %i\n",
458 						   err, vsi->vsi_num);
459 				}
460 			}
461 		}
462 	}
463 	goto exit;
464 
465 out_promisc:
466 	set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
467 	goto exit;
468 out:
469 	/* if something went wrong then set the changed flag so we try again */
470 	set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
471 	set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
472 exit:
473 	clear_bit(ICE_CFG_BUSY, vsi->state);
474 	return err;
475 }
476 
477 /**
478  * ice_sync_fltr_subtask - Sync the VSI filter list with HW
479  * @pf: board private structure
480  */
481 static void ice_sync_fltr_subtask(struct ice_pf *pf)
482 {
483 	int v;
484 
485 	if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
486 		return;
487 
488 	clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
489 
490 	ice_for_each_vsi(pf, v)
491 		if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
492 		    ice_vsi_sync_fltr(pf->vsi[v])) {
493 			/* come back and try again later */
494 			set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
495 			break;
496 		}
497 }
498 
499 /**
500  * ice_pf_dis_all_vsi - Pause all VSIs on a PF
501  * @pf: the PF
502  * @locked: is the rtnl_lock already held
503  */
504 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
505 {
506 	int node;
507 	int v;
508 
509 	ice_for_each_vsi(pf, v)
510 		if (pf->vsi[v])
511 			ice_dis_vsi(pf->vsi[v], locked);
512 
513 	for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
514 		pf->pf_agg_node[node].num_vsis = 0;
515 
516 	for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
517 		pf->vf_agg_node[node].num_vsis = 0;
518 }
519 
520 /**
521  * ice_clear_sw_switch_recipes - clear switch recipes
522  * @pf: board private structure
523  *
524  * Mark switch recipes as not created in sw structures. There are cases where
525  * rules (especially advanced rules) need to be restored, either re-read from
526  * hardware or added again. For example after the reset. 'recp_created' flag
527  * prevents from doing that and need to be cleared upfront.
528  */
529 static void ice_clear_sw_switch_recipes(struct ice_pf *pf)
530 {
531 	struct ice_sw_recipe *recp;
532 	u8 i;
533 
534 	recp = pf->hw.switch_info->recp_list;
535 	for (i = 0; i < ICE_MAX_NUM_RECIPES; i++)
536 		recp[i].recp_created = false;
537 }
538 
539 /**
540  * ice_prepare_for_reset - prep for reset
541  * @pf: board private structure
542  * @reset_type: reset type requested
543  *
544  * Inform or close all dependent features in prep for reset.
545  */
546 static void
547 ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
548 {
549 	struct ice_hw *hw = &pf->hw;
550 	struct ice_vsi *vsi;
551 	struct ice_vf *vf;
552 	unsigned int bkt;
553 
554 	dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type);
555 
556 	/* already prepared for reset */
557 	if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
558 		return;
559 
560 	ice_unplug_aux_dev(pf);
561 
562 	/* Notify VFs of impending reset */
563 	if (ice_check_sq_alive(hw, &hw->mailboxq))
564 		ice_vc_notify_reset(pf);
565 
566 	/* Disable VFs until reset is completed */
567 	mutex_lock(&pf->vfs.table_lock);
568 	ice_for_each_vf(pf, bkt, vf)
569 		ice_set_vf_state_dis(vf);
570 	mutex_unlock(&pf->vfs.table_lock);
571 
572 	if (ice_is_eswitch_mode_switchdev(pf)) {
573 		if (reset_type != ICE_RESET_PFR)
574 			ice_clear_sw_switch_recipes(pf);
575 	}
576 
577 	/* release ADQ specific HW and SW resources */
578 	vsi = ice_get_main_vsi(pf);
579 	if (!vsi)
580 		goto skip;
581 
582 	/* to be on safe side, reset orig_rss_size so that normal flow
583 	 * of deciding rss_size can take precedence
584 	 */
585 	vsi->orig_rss_size = 0;
586 
587 	if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
588 		if (reset_type == ICE_RESET_PFR) {
589 			vsi->old_ena_tc = vsi->all_enatc;
590 			vsi->old_numtc = vsi->all_numtc;
591 		} else {
592 			ice_remove_q_channels(vsi, true);
593 
594 			/* for other reset type, do not support channel rebuild
595 			 * hence reset needed info
596 			 */
597 			vsi->old_ena_tc = 0;
598 			vsi->all_enatc = 0;
599 			vsi->old_numtc = 0;
600 			vsi->all_numtc = 0;
601 			vsi->req_txq = 0;
602 			vsi->req_rxq = 0;
603 			clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
604 			memset(&vsi->mqprio_qopt, 0, sizeof(vsi->mqprio_qopt));
605 		}
606 	}
607 skip:
608 
609 	/* clear SW filtering DB */
610 	ice_clear_hw_tbls(hw);
611 	/* disable the VSIs and their queues that are not already DOWN */
612 	ice_pf_dis_all_vsi(pf, false);
613 
614 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
615 		ice_ptp_prepare_for_reset(pf);
616 
617 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
618 		ice_gnss_exit(pf);
619 
620 	if (hw->port_info)
621 		ice_sched_clear_port(hw->port_info);
622 
623 	ice_shutdown_all_ctrlq(hw);
624 
625 	set_bit(ICE_PREPARED_FOR_RESET, pf->state);
626 }
627 
628 /**
629  * ice_do_reset - Initiate one of many types of resets
630  * @pf: board private structure
631  * @reset_type: reset type requested before this function was called.
632  */
633 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
634 {
635 	struct device *dev = ice_pf_to_dev(pf);
636 	struct ice_hw *hw = &pf->hw;
637 
638 	dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
639 
640 	if (pf->lag && pf->lag->bonded && reset_type == ICE_RESET_PFR) {
641 		dev_dbg(dev, "PFR on a bonded interface, promoting to CORER\n");
642 		reset_type = ICE_RESET_CORER;
643 	}
644 
645 	ice_prepare_for_reset(pf, reset_type);
646 
647 	/* trigger the reset */
648 	if (ice_reset(hw, reset_type)) {
649 		dev_err(dev, "reset %d failed\n", reset_type);
650 		set_bit(ICE_RESET_FAILED, pf->state);
651 		clear_bit(ICE_RESET_OICR_RECV, pf->state);
652 		clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
653 		clear_bit(ICE_PFR_REQ, pf->state);
654 		clear_bit(ICE_CORER_REQ, pf->state);
655 		clear_bit(ICE_GLOBR_REQ, pf->state);
656 		wake_up(&pf->reset_wait_queue);
657 		return;
658 	}
659 
660 	/* PFR is a bit of a special case because it doesn't result in an OICR
661 	 * interrupt. So for PFR, rebuild after the reset and clear the reset-
662 	 * associated state bits.
663 	 */
664 	if (reset_type == ICE_RESET_PFR) {
665 		pf->pfr_count++;
666 		ice_rebuild(pf, reset_type);
667 		clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
668 		clear_bit(ICE_PFR_REQ, pf->state);
669 		wake_up(&pf->reset_wait_queue);
670 		ice_reset_all_vfs(pf);
671 	}
672 }
673 
674 /**
675  * ice_reset_subtask - Set up for resetting the device and driver
676  * @pf: board private structure
677  */
678 static void ice_reset_subtask(struct ice_pf *pf)
679 {
680 	enum ice_reset_req reset_type = ICE_RESET_INVAL;
681 
682 	/* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
683 	 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
684 	 * of reset is pending and sets bits in pf->state indicating the reset
685 	 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
686 	 * prepare for pending reset if not already (for PF software-initiated
687 	 * global resets the software should already be prepared for it as
688 	 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
689 	 * by firmware or software on other PFs, that bit is not set so prepare
690 	 * for the reset now), poll for reset done, rebuild and return.
691 	 */
692 	if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
693 		/* Perform the largest reset requested */
694 		if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
695 			reset_type = ICE_RESET_CORER;
696 		if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
697 			reset_type = ICE_RESET_GLOBR;
698 		if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
699 			reset_type = ICE_RESET_EMPR;
700 		/* return if no valid reset type requested */
701 		if (reset_type == ICE_RESET_INVAL)
702 			return;
703 		ice_prepare_for_reset(pf, reset_type);
704 
705 		/* make sure we are ready to rebuild */
706 		if (ice_check_reset(&pf->hw)) {
707 			set_bit(ICE_RESET_FAILED, pf->state);
708 		} else {
709 			/* done with reset. start rebuild */
710 			pf->hw.reset_ongoing = false;
711 			ice_rebuild(pf, reset_type);
712 			/* clear bit to resume normal operations, but
713 			 * ICE_NEEDS_RESTART bit is set in case rebuild failed
714 			 */
715 			clear_bit(ICE_RESET_OICR_RECV, pf->state);
716 			clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
717 			clear_bit(ICE_PFR_REQ, pf->state);
718 			clear_bit(ICE_CORER_REQ, pf->state);
719 			clear_bit(ICE_GLOBR_REQ, pf->state);
720 			wake_up(&pf->reset_wait_queue);
721 			ice_reset_all_vfs(pf);
722 		}
723 
724 		return;
725 	}
726 
727 	/* No pending resets to finish processing. Check for new resets */
728 	if (test_bit(ICE_PFR_REQ, pf->state)) {
729 		reset_type = ICE_RESET_PFR;
730 		if (pf->lag && pf->lag->bonded) {
731 			dev_dbg(ice_pf_to_dev(pf), "PFR on a bonded interface, promoting to CORER\n");
732 			reset_type = ICE_RESET_CORER;
733 		}
734 	}
735 	if (test_bit(ICE_CORER_REQ, pf->state))
736 		reset_type = ICE_RESET_CORER;
737 	if (test_bit(ICE_GLOBR_REQ, pf->state))
738 		reset_type = ICE_RESET_GLOBR;
739 	/* If no valid reset type requested just return */
740 	if (reset_type == ICE_RESET_INVAL)
741 		return;
742 
743 	/* reset if not already down or busy */
744 	if (!test_bit(ICE_DOWN, pf->state) &&
745 	    !test_bit(ICE_CFG_BUSY, pf->state)) {
746 		ice_do_reset(pf, reset_type);
747 	}
748 }
749 
750 /**
751  * ice_print_topo_conflict - print topology conflict message
752  * @vsi: the VSI whose topology status is being checked
753  */
754 static void ice_print_topo_conflict(struct ice_vsi *vsi)
755 {
756 	switch (vsi->port_info->phy.link_info.topo_media_conflict) {
757 	case ICE_AQ_LINK_TOPO_CONFLICT:
758 	case ICE_AQ_LINK_MEDIA_CONFLICT:
759 	case ICE_AQ_LINK_TOPO_UNREACH_PRT:
760 	case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
761 	case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
762 		netdev_info(vsi->netdev, "Potential misconfiguration of the Ethernet port detected. If it was not intended, please use the Intel (R) Ethernet Port Configuration Tool to address the issue.\n");
763 		break;
764 	case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
765 		if (test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, vsi->back->flags))
766 			netdev_warn(vsi->netdev, "An unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules\n");
767 		else
768 			netdev_err(vsi->netdev, "Rx/Tx is disabled on this device because an unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
769 		break;
770 	default:
771 		break;
772 	}
773 }
774 
775 /**
776  * ice_print_link_msg - print link up or down message
777  * @vsi: the VSI whose link status is being queried
778  * @isup: boolean for if the link is now up or down
779  */
780 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
781 {
782 	struct ice_aqc_get_phy_caps_data *caps;
783 	const char *an_advertised;
784 	const char *fec_req;
785 	const char *speed;
786 	const char *fec;
787 	const char *fc;
788 	const char *an;
789 	int status;
790 
791 	if (!vsi)
792 		return;
793 
794 	if (vsi->current_isup == isup)
795 		return;
796 
797 	vsi->current_isup = isup;
798 
799 	if (!isup) {
800 		netdev_info(vsi->netdev, "NIC Link is Down\n");
801 		return;
802 	}
803 
804 	switch (vsi->port_info->phy.link_info.link_speed) {
805 	case ICE_AQ_LINK_SPEED_100GB:
806 		speed = "100 G";
807 		break;
808 	case ICE_AQ_LINK_SPEED_50GB:
809 		speed = "50 G";
810 		break;
811 	case ICE_AQ_LINK_SPEED_40GB:
812 		speed = "40 G";
813 		break;
814 	case ICE_AQ_LINK_SPEED_25GB:
815 		speed = "25 G";
816 		break;
817 	case ICE_AQ_LINK_SPEED_20GB:
818 		speed = "20 G";
819 		break;
820 	case ICE_AQ_LINK_SPEED_10GB:
821 		speed = "10 G";
822 		break;
823 	case ICE_AQ_LINK_SPEED_5GB:
824 		speed = "5 G";
825 		break;
826 	case ICE_AQ_LINK_SPEED_2500MB:
827 		speed = "2.5 G";
828 		break;
829 	case ICE_AQ_LINK_SPEED_1000MB:
830 		speed = "1 G";
831 		break;
832 	case ICE_AQ_LINK_SPEED_100MB:
833 		speed = "100 M";
834 		break;
835 	default:
836 		speed = "Unknown ";
837 		break;
838 	}
839 
840 	switch (vsi->port_info->fc.current_mode) {
841 	case ICE_FC_FULL:
842 		fc = "Rx/Tx";
843 		break;
844 	case ICE_FC_TX_PAUSE:
845 		fc = "Tx";
846 		break;
847 	case ICE_FC_RX_PAUSE:
848 		fc = "Rx";
849 		break;
850 	case ICE_FC_NONE:
851 		fc = "None";
852 		break;
853 	default:
854 		fc = "Unknown";
855 		break;
856 	}
857 
858 	/* Get FEC mode based on negotiated link info */
859 	switch (vsi->port_info->phy.link_info.fec_info) {
860 	case ICE_AQ_LINK_25G_RS_528_FEC_EN:
861 	case ICE_AQ_LINK_25G_RS_544_FEC_EN:
862 		fec = "RS-FEC";
863 		break;
864 	case ICE_AQ_LINK_25G_KR_FEC_EN:
865 		fec = "FC-FEC/BASE-R";
866 		break;
867 	default:
868 		fec = "NONE";
869 		break;
870 	}
871 
872 	/* check if autoneg completed, might be false due to not supported */
873 	if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
874 		an = "True";
875 	else
876 		an = "False";
877 
878 	/* Get FEC mode requested based on PHY caps last SW configuration */
879 	caps = kzalloc(sizeof(*caps), GFP_KERNEL);
880 	if (!caps) {
881 		fec_req = "Unknown";
882 		an_advertised = "Unknown";
883 		goto done;
884 	}
885 
886 	status = ice_aq_get_phy_caps(vsi->port_info, false,
887 				     ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
888 	if (status)
889 		netdev_info(vsi->netdev, "Get phy capability failed.\n");
890 
891 	an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
892 
893 	if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
894 	    caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
895 		fec_req = "RS-FEC";
896 	else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
897 		 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
898 		fec_req = "FC-FEC/BASE-R";
899 	else
900 		fec_req = "NONE";
901 
902 	kfree(caps);
903 
904 done:
905 	netdev_info(vsi->netdev, "NIC Link is up %sbps Full Duplex, Requested FEC: %s, Negotiated FEC: %s, Autoneg Advertised: %s, Autoneg Negotiated: %s, Flow Control: %s\n",
906 		    speed, fec_req, fec, an_advertised, an, fc);
907 	ice_print_topo_conflict(vsi);
908 }
909 
910 /**
911  * ice_vsi_link_event - update the VSI's netdev
912  * @vsi: the VSI on which the link event occurred
913  * @link_up: whether or not the VSI needs to be set up or down
914  */
915 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
916 {
917 	if (!vsi)
918 		return;
919 
920 	if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
921 		return;
922 
923 	if (vsi->type == ICE_VSI_PF) {
924 		if (link_up == netif_carrier_ok(vsi->netdev))
925 			return;
926 
927 		if (link_up) {
928 			netif_carrier_on(vsi->netdev);
929 			netif_tx_wake_all_queues(vsi->netdev);
930 		} else {
931 			netif_carrier_off(vsi->netdev);
932 			netif_tx_stop_all_queues(vsi->netdev);
933 		}
934 	}
935 }
936 
937 /**
938  * ice_set_dflt_mib - send a default config MIB to the FW
939  * @pf: private PF struct
940  *
941  * This function sends a default configuration MIB to the FW.
942  *
943  * If this function errors out at any point, the driver is still able to
944  * function.  The main impact is that LFC may not operate as expected.
945  * Therefore an error state in this function should be treated with a DBG
946  * message and continue on with driver rebuild/reenable.
947  */
948 static void ice_set_dflt_mib(struct ice_pf *pf)
949 {
950 	struct device *dev = ice_pf_to_dev(pf);
951 	u8 mib_type, *buf, *lldpmib = NULL;
952 	u16 len, typelen, offset = 0;
953 	struct ice_lldp_org_tlv *tlv;
954 	struct ice_hw *hw = &pf->hw;
955 	u32 ouisubtype;
956 
957 	mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
958 	lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
959 	if (!lldpmib) {
960 		dev_dbg(dev, "%s Failed to allocate MIB memory\n",
961 			__func__);
962 		return;
963 	}
964 
965 	/* Add ETS CFG TLV */
966 	tlv = (struct ice_lldp_org_tlv *)lldpmib;
967 	typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
968 		   ICE_IEEE_ETS_TLV_LEN);
969 	tlv->typelen = htons(typelen);
970 	ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
971 		      ICE_IEEE_SUBTYPE_ETS_CFG);
972 	tlv->ouisubtype = htonl(ouisubtype);
973 
974 	buf = tlv->tlvinfo;
975 	buf[0] = 0;
976 
977 	/* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
978 	 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
979 	 * Octets 13 - 20 are TSA values - leave as zeros
980 	 */
981 	buf[5] = 0x64;
982 	len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
983 	offset += len + 2;
984 	tlv = (struct ice_lldp_org_tlv *)
985 		((char *)tlv + sizeof(tlv->typelen) + len);
986 
987 	/* Add ETS REC TLV */
988 	buf = tlv->tlvinfo;
989 	tlv->typelen = htons(typelen);
990 
991 	ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
992 		      ICE_IEEE_SUBTYPE_ETS_REC);
993 	tlv->ouisubtype = htonl(ouisubtype);
994 
995 	/* First octet of buf is reserved
996 	 * Octets 1 - 4 map UP to TC - all UPs map to zero
997 	 * Octets 5 - 12 are BW values - set TC 0 to 100%.
998 	 * Octets 13 - 20 are TSA value - leave as zeros
999 	 */
1000 	buf[5] = 0x64;
1001 	offset += len + 2;
1002 	tlv = (struct ice_lldp_org_tlv *)
1003 		((char *)tlv + sizeof(tlv->typelen) + len);
1004 
1005 	/* Add PFC CFG TLV */
1006 	typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
1007 		   ICE_IEEE_PFC_TLV_LEN);
1008 	tlv->typelen = htons(typelen);
1009 
1010 	ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
1011 		      ICE_IEEE_SUBTYPE_PFC_CFG);
1012 	tlv->ouisubtype = htonl(ouisubtype);
1013 
1014 	/* Octet 1 left as all zeros - PFC disabled */
1015 	buf[0] = 0x08;
1016 	len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
1017 	offset += len + 2;
1018 
1019 	if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
1020 		dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
1021 
1022 	kfree(lldpmib);
1023 }
1024 
1025 /**
1026  * ice_check_phy_fw_load - check if PHY FW load failed
1027  * @pf: pointer to PF struct
1028  * @link_cfg_err: bitmap from the link info structure
1029  *
1030  * check if external PHY FW load failed and print an error message if it did
1031  */
1032 static void ice_check_phy_fw_load(struct ice_pf *pf, u8 link_cfg_err)
1033 {
1034 	if (!(link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE)) {
1035 		clear_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1036 		return;
1037 	}
1038 
1039 	if (test_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags))
1040 		return;
1041 
1042 	if (link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE) {
1043 		dev_err(ice_pf_to_dev(pf), "Device failed to load the FW for the external PHY. Please download and install the latest NVM for your device and try again\n");
1044 		set_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1045 	}
1046 }
1047 
1048 /**
1049  * ice_check_module_power
1050  * @pf: pointer to PF struct
1051  * @link_cfg_err: bitmap from the link info structure
1052  *
1053  * check module power level returned by a previous call to aq_get_link_info
1054  * and print error messages if module power level is not supported
1055  */
1056 static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
1057 {
1058 	/* if module power level is supported, clear the flag */
1059 	if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
1060 			      ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
1061 		clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1062 		return;
1063 	}
1064 
1065 	/* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
1066 	 * above block didn't clear this bit, there's nothing to do
1067 	 */
1068 	if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
1069 		return;
1070 
1071 	if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
1072 		dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
1073 		set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1074 	} else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
1075 		dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
1076 		set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1077 	}
1078 }
1079 
1080 /**
1081  * ice_check_link_cfg_err - check if link configuration failed
1082  * @pf: pointer to the PF struct
1083  * @link_cfg_err: bitmap from the link info structure
1084  *
1085  * print if any link configuration failure happens due to the value in the
1086  * link_cfg_err parameter in the link info structure
1087  */
1088 static void ice_check_link_cfg_err(struct ice_pf *pf, u8 link_cfg_err)
1089 {
1090 	ice_check_module_power(pf, link_cfg_err);
1091 	ice_check_phy_fw_load(pf, link_cfg_err);
1092 }
1093 
1094 /**
1095  * ice_link_event - process the link event
1096  * @pf: PF that the link event is associated with
1097  * @pi: port_info for the port that the link event is associated with
1098  * @link_up: true if the physical link is up and false if it is down
1099  * @link_speed: current link speed received from the link event
1100  *
1101  * Returns 0 on success and negative on failure
1102  */
1103 static int
1104 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
1105 	       u16 link_speed)
1106 {
1107 	struct device *dev = ice_pf_to_dev(pf);
1108 	struct ice_phy_info *phy_info;
1109 	struct ice_vsi *vsi;
1110 	u16 old_link_speed;
1111 	bool old_link;
1112 	int status;
1113 
1114 	phy_info = &pi->phy;
1115 	phy_info->link_info_old = phy_info->link_info;
1116 
1117 	old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
1118 	old_link_speed = phy_info->link_info_old.link_speed;
1119 
1120 	/* update the link info structures and re-enable link events,
1121 	 * don't bail on failure due to other book keeping needed
1122 	 */
1123 	status = ice_update_link_info(pi);
1124 	if (status)
1125 		dev_dbg(dev, "Failed to update link status on port %d, err %d aq_err %s\n",
1126 			pi->lport, status,
1127 			ice_aq_str(pi->hw->adminq.sq_last_status));
1128 
1129 	ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
1130 
1131 	/* Check if the link state is up after updating link info, and treat
1132 	 * this event as an UP event since the link is actually UP now.
1133 	 */
1134 	if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
1135 		link_up = true;
1136 
1137 	vsi = ice_get_main_vsi(pf);
1138 	if (!vsi || !vsi->port_info)
1139 		return -EINVAL;
1140 
1141 	/* turn off PHY if media was removed */
1142 	if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
1143 	    !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
1144 		set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
1145 		ice_set_link(vsi, false);
1146 	}
1147 
1148 	/* if the old link up/down and speed is the same as the new */
1149 	if (link_up == old_link && link_speed == old_link_speed)
1150 		return 0;
1151 
1152 	ice_ptp_link_change(pf, pf->hw.pf_id, link_up);
1153 
1154 	if (ice_is_dcb_active(pf)) {
1155 		if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
1156 			ice_dcb_rebuild(pf);
1157 	} else {
1158 		if (link_up)
1159 			ice_set_dflt_mib(pf);
1160 	}
1161 	ice_vsi_link_event(vsi, link_up);
1162 	ice_print_link_msg(vsi, link_up);
1163 
1164 	ice_vc_notify_link_state(pf);
1165 
1166 	return 0;
1167 }
1168 
1169 /**
1170  * ice_watchdog_subtask - periodic tasks not using event driven scheduling
1171  * @pf: board private structure
1172  */
1173 static void ice_watchdog_subtask(struct ice_pf *pf)
1174 {
1175 	int i;
1176 
1177 	/* if interface is down do nothing */
1178 	if (test_bit(ICE_DOWN, pf->state) ||
1179 	    test_bit(ICE_CFG_BUSY, pf->state))
1180 		return;
1181 
1182 	/* make sure we don't do these things too often */
1183 	if (time_before(jiffies,
1184 			pf->serv_tmr_prev + pf->serv_tmr_period))
1185 		return;
1186 
1187 	pf->serv_tmr_prev = jiffies;
1188 
1189 	/* Update the stats for active netdevs so the network stack
1190 	 * can look at updated numbers whenever it cares to
1191 	 */
1192 	ice_update_pf_stats(pf);
1193 	ice_for_each_vsi(pf, i)
1194 		if (pf->vsi[i] && pf->vsi[i]->netdev)
1195 			ice_update_vsi_stats(pf->vsi[i]);
1196 }
1197 
1198 /**
1199  * ice_init_link_events - enable/initialize link events
1200  * @pi: pointer to the port_info instance
1201  *
1202  * Returns -EIO on failure, 0 on success
1203  */
1204 static int ice_init_link_events(struct ice_port_info *pi)
1205 {
1206 	u16 mask;
1207 
1208 	mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1209 		       ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL |
1210 		       ICE_AQ_LINK_EVENT_PHY_FW_LOAD_FAIL));
1211 
1212 	if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1213 		dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1214 			pi->lport);
1215 		return -EIO;
1216 	}
1217 
1218 	if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1219 		dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1220 			pi->lport);
1221 		return -EIO;
1222 	}
1223 
1224 	return 0;
1225 }
1226 
1227 /**
1228  * ice_handle_link_event - handle link event via ARQ
1229  * @pf: PF that the link event is associated with
1230  * @event: event structure containing link status info
1231  */
1232 static int
1233 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1234 {
1235 	struct ice_aqc_get_link_status_data *link_data;
1236 	struct ice_port_info *port_info;
1237 	int status;
1238 
1239 	link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1240 	port_info = pf->hw.port_info;
1241 	if (!port_info)
1242 		return -EINVAL;
1243 
1244 	status = ice_link_event(pf, port_info,
1245 				!!(link_data->link_info & ICE_AQ_LINK_UP),
1246 				le16_to_cpu(link_data->link_speed));
1247 	if (status)
1248 		dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1249 			status);
1250 
1251 	return status;
1252 }
1253 
1254 /**
1255  * ice_aq_prep_for_event - Prepare to wait for an AdminQ event from firmware
1256  * @pf: pointer to the PF private structure
1257  * @task: intermediate helper storage and identifier for waiting
1258  * @opcode: the opcode to wait for
1259  *
1260  * Prepares to wait for a specific AdminQ completion event on the ARQ for
1261  * a given PF. Actual wait would be done by a call to ice_aq_wait_for_event().
1262  *
1263  * Calls are separated to allow caller registering for event before sending
1264  * the command, which mitigates a race between registering and FW responding.
1265  *
1266  * To obtain only the descriptor contents, pass an task->event with null
1267  * msg_buf. If the complete data buffer is desired, allocate the
1268  * task->event.msg_buf with enough space ahead of time.
1269  */
1270 void ice_aq_prep_for_event(struct ice_pf *pf, struct ice_aq_task *task,
1271 			   u16 opcode)
1272 {
1273 	INIT_HLIST_NODE(&task->entry);
1274 	task->opcode = opcode;
1275 	task->state = ICE_AQ_TASK_WAITING;
1276 
1277 	spin_lock_bh(&pf->aq_wait_lock);
1278 	hlist_add_head(&task->entry, &pf->aq_wait_list);
1279 	spin_unlock_bh(&pf->aq_wait_lock);
1280 }
1281 
1282 /**
1283  * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1284  * @pf: pointer to the PF private structure
1285  * @task: ptr prepared by ice_aq_prep_for_event()
1286  * @timeout: how long to wait, in jiffies
1287  *
1288  * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1289  * current thread will be put to sleep until the specified event occurs or
1290  * until the given timeout is reached.
1291  *
1292  * Returns: zero on success, or a negative error code on failure.
1293  */
1294 int ice_aq_wait_for_event(struct ice_pf *pf, struct ice_aq_task *task,
1295 			  unsigned long timeout)
1296 {
1297 	enum ice_aq_task_state *state = &task->state;
1298 	struct device *dev = ice_pf_to_dev(pf);
1299 	unsigned long start = jiffies;
1300 	long ret;
1301 	int err;
1302 
1303 	ret = wait_event_interruptible_timeout(pf->aq_wait_queue,
1304 					       *state != ICE_AQ_TASK_WAITING,
1305 					       timeout);
1306 	switch (*state) {
1307 	case ICE_AQ_TASK_NOT_PREPARED:
1308 		WARN(1, "call to %s without ice_aq_prep_for_event()", __func__);
1309 		err = -EINVAL;
1310 		break;
1311 	case ICE_AQ_TASK_WAITING:
1312 		err = ret < 0 ? ret : -ETIMEDOUT;
1313 		break;
1314 	case ICE_AQ_TASK_CANCELED:
1315 		err = ret < 0 ? ret : -ECANCELED;
1316 		break;
1317 	case ICE_AQ_TASK_COMPLETE:
1318 		err = ret < 0 ? ret : 0;
1319 		break;
1320 	default:
1321 		WARN(1, "Unexpected AdminQ wait task state %u", *state);
1322 		err = -EINVAL;
1323 		break;
1324 	}
1325 
1326 	dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1327 		jiffies_to_msecs(jiffies - start),
1328 		jiffies_to_msecs(timeout),
1329 		task->opcode);
1330 
1331 	spin_lock_bh(&pf->aq_wait_lock);
1332 	hlist_del(&task->entry);
1333 	spin_unlock_bh(&pf->aq_wait_lock);
1334 
1335 	return err;
1336 }
1337 
1338 /**
1339  * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1340  * @pf: pointer to the PF private structure
1341  * @opcode: the opcode of the event
1342  * @event: the event to check
1343  *
1344  * Loops over the current list of pending threads waiting for an AdminQ event.
1345  * For each matching task, copy the contents of the event into the task
1346  * structure and wake up the thread.
1347  *
1348  * If multiple threads wait for the same opcode, they will all be woken up.
1349  *
1350  * Note that event->msg_buf will only be duplicated if the event has a buffer
1351  * with enough space already allocated. Otherwise, only the descriptor and
1352  * message length will be copied.
1353  *
1354  * Returns: true if an event was found, false otherwise
1355  */
1356 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1357 				struct ice_rq_event_info *event)
1358 {
1359 	struct ice_rq_event_info *task_ev;
1360 	struct ice_aq_task *task;
1361 	bool found = false;
1362 
1363 	spin_lock_bh(&pf->aq_wait_lock);
1364 	hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1365 		if (task->state != ICE_AQ_TASK_WAITING)
1366 			continue;
1367 		if (task->opcode != opcode)
1368 			continue;
1369 
1370 		task_ev = &task->event;
1371 		memcpy(&task_ev->desc, &event->desc, sizeof(event->desc));
1372 		task_ev->msg_len = event->msg_len;
1373 
1374 		/* Only copy the data buffer if a destination was set */
1375 		if (task_ev->msg_buf && task_ev->buf_len >= event->buf_len) {
1376 			memcpy(task_ev->msg_buf, event->msg_buf,
1377 			       event->buf_len);
1378 			task_ev->buf_len = event->buf_len;
1379 		}
1380 
1381 		task->state = ICE_AQ_TASK_COMPLETE;
1382 		found = true;
1383 	}
1384 	spin_unlock_bh(&pf->aq_wait_lock);
1385 
1386 	if (found)
1387 		wake_up(&pf->aq_wait_queue);
1388 }
1389 
1390 /**
1391  * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1392  * @pf: the PF private structure
1393  *
1394  * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1395  * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1396  */
1397 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1398 {
1399 	struct ice_aq_task *task;
1400 
1401 	spin_lock_bh(&pf->aq_wait_lock);
1402 	hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1403 		task->state = ICE_AQ_TASK_CANCELED;
1404 	spin_unlock_bh(&pf->aq_wait_lock);
1405 
1406 	wake_up(&pf->aq_wait_queue);
1407 }
1408 
1409 #define ICE_MBX_OVERFLOW_WATERMARK 64
1410 
1411 /**
1412  * __ice_clean_ctrlq - helper function to clean controlq rings
1413  * @pf: ptr to struct ice_pf
1414  * @q_type: specific Control queue type
1415  */
1416 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1417 {
1418 	struct device *dev = ice_pf_to_dev(pf);
1419 	struct ice_rq_event_info event;
1420 	struct ice_hw *hw = &pf->hw;
1421 	struct ice_ctl_q_info *cq;
1422 	u16 pending, i = 0;
1423 	const char *qtype;
1424 	u32 oldval, val;
1425 
1426 	/* Do not clean control queue if/when PF reset fails */
1427 	if (test_bit(ICE_RESET_FAILED, pf->state))
1428 		return 0;
1429 
1430 	switch (q_type) {
1431 	case ICE_CTL_Q_ADMIN:
1432 		cq = &hw->adminq;
1433 		qtype = "Admin";
1434 		break;
1435 	case ICE_CTL_Q_SB:
1436 		cq = &hw->sbq;
1437 		qtype = "Sideband";
1438 		break;
1439 	case ICE_CTL_Q_MAILBOX:
1440 		cq = &hw->mailboxq;
1441 		qtype = "Mailbox";
1442 		/* we are going to try to detect a malicious VF, so set the
1443 		 * state to begin detection
1444 		 */
1445 		hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1446 		break;
1447 	default:
1448 		dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1449 		return 0;
1450 	}
1451 
1452 	/* check for error indications - PF_xx_AxQLEN register layout for
1453 	 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1454 	 */
1455 	val = rd32(hw, cq->rq.len);
1456 	if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1457 		   PF_FW_ARQLEN_ARQCRIT_M)) {
1458 		oldval = val;
1459 		if (val & PF_FW_ARQLEN_ARQVFE_M)
1460 			dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1461 				qtype);
1462 		if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1463 			dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1464 				qtype);
1465 		}
1466 		if (val & PF_FW_ARQLEN_ARQCRIT_M)
1467 			dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1468 				qtype);
1469 		val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1470 			 PF_FW_ARQLEN_ARQCRIT_M);
1471 		if (oldval != val)
1472 			wr32(hw, cq->rq.len, val);
1473 	}
1474 
1475 	val = rd32(hw, cq->sq.len);
1476 	if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1477 		   PF_FW_ATQLEN_ATQCRIT_M)) {
1478 		oldval = val;
1479 		if (val & PF_FW_ATQLEN_ATQVFE_M)
1480 			dev_dbg(dev, "%s Send Queue VF Error detected\n",
1481 				qtype);
1482 		if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1483 			dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1484 				qtype);
1485 		}
1486 		if (val & PF_FW_ATQLEN_ATQCRIT_M)
1487 			dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1488 				qtype);
1489 		val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1490 			 PF_FW_ATQLEN_ATQCRIT_M);
1491 		if (oldval != val)
1492 			wr32(hw, cq->sq.len, val);
1493 	}
1494 
1495 	event.buf_len = cq->rq_buf_size;
1496 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1497 	if (!event.msg_buf)
1498 		return 0;
1499 
1500 	do {
1501 		struct ice_mbx_data data = {};
1502 		u16 opcode;
1503 		int ret;
1504 
1505 		ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1506 		if (ret == -EALREADY)
1507 			break;
1508 		if (ret) {
1509 			dev_err(dev, "%s Receive Queue event error %d\n", qtype,
1510 				ret);
1511 			break;
1512 		}
1513 
1514 		opcode = le16_to_cpu(event.desc.opcode);
1515 
1516 		/* Notify any thread that might be waiting for this event */
1517 		ice_aq_check_events(pf, opcode, &event);
1518 
1519 		switch (opcode) {
1520 		case ice_aqc_opc_get_link_status:
1521 			if (ice_handle_link_event(pf, &event))
1522 				dev_err(dev, "Could not handle link event\n");
1523 			break;
1524 		case ice_aqc_opc_event_lan_overflow:
1525 			ice_vf_lan_overflow_event(pf, &event);
1526 			break;
1527 		case ice_mbx_opc_send_msg_to_pf:
1528 			data.num_msg_proc = i;
1529 			data.num_pending_arq = pending;
1530 			data.max_num_msgs_mbx = hw->mailboxq.num_rq_entries;
1531 			data.async_watermark_val = ICE_MBX_OVERFLOW_WATERMARK;
1532 
1533 			ice_vc_process_vf_msg(pf, &event, &data);
1534 			break;
1535 		case ice_aqc_opc_fw_logging:
1536 			ice_output_fw_log(hw, &event.desc, event.msg_buf);
1537 			break;
1538 		case ice_aqc_opc_lldp_set_mib_change:
1539 			ice_dcb_process_lldp_set_mib_change(pf, &event);
1540 			break;
1541 		default:
1542 			dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1543 				qtype, opcode);
1544 			break;
1545 		}
1546 	} while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1547 
1548 	kfree(event.msg_buf);
1549 
1550 	return pending && (i == ICE_DFLT_IRQ_WORK);
1551 }
1552 
1553 /**
1554  * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1555  * @hw: pointer to hardware info
1556  * @cq: control queue information
1557  *
1558  * returns true if there are pending messages in a queue, false if there aren't
1559  */
1560 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1561 {
1562 	u16 ntu;
1563 
1564 	ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1565 	return cq->rq.next_to_clean != ntu;
1566 }
1567 
1568 /**
1569  * ice_clean_adminq_subtask - clean the AdminQ rings
1570  * @pf: board private structure
1571  */
1572 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1573 {
1574 	struct ice_hw *hw = &pf->hw;
1575 
1576 	if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1577 		return;
1578 
1579 	if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1580 		return;
1581 
1582 	clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1583 
1584 	/* There might be a situation where new messages arrive to a control
1585 	 * queue between processing the last message and clearing the
1586 	 * EVENT_PENDING bit. So before exiting, check queue head again (using
1587 	 * ice_ctrlq_pending) and process new messages if any.
1588 	 */
1589 	if (ice_ctrlq_pending(hw, &hw->adminq))
1590 		__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1591 
1592 	ice_flush(hw);
1593 }
1594 
1595 /**
1596  * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1597  * @pf: board private structure
1598  */
1599 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1600 {
1601 	struct ice_hw *hw = &pf->hw;
1602 
1603 	if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1604 		return;
1605 
1606 	if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1607 		return;
1608 
1609 	clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1610 
1611 	if (ice_ctrlq_pending(hw, &hw->mailboxq))
1612 		__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1613 
1614 	ice_flush(hw);
1615 }
1616 
1617 /**
1618  * ice_clean_sbq_subtask - clean the Sideband Queue rings
1619  * @pf: board private structure
1620  */
1621 static void ice_clean_sbq_subtask(struct ice_pf *pf)
1622 {
1623 	struct ice_hw *hw = &pf->hw;
1624 
1625 	/* Nothing to do here if sideband queue is not supported */
1626 	if (!ice_is_sbq_supported(hw)) {
1627 		clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1628 		return;
1629 	}
1630 
1631 	if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1632 		return;
1633 
1634 	if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1635 		return;
1636 
1637 	clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1638 
1639 	if (ice_ctrlq_pending(hw, &hw->sbq))
1640 		__ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1641 
1642 	ice_flush(hw);
1643 }
1644 
1645 /**
1646  * ice_service_task_schedule - schedule the service task to wake up
1647  * @pf: board private structure
1648  *
1649  * If not already scheduled, this puts the task into the work queue.
1650  */
1651 void ice_service_task_schedule(struct ice_pf *pf)
1652 {
1653 	if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1654 	    !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1655 	    !test_bit(ICE_NEEDS_RESTART, pf->state))
1656 		queue_work(ice_wq, &pf->serv_task);
1657 }
1658 
1659 /**
1660  * ice_service_task_complete - finish up the service task
1661  * @pf: board private structure
1662  */
1663 static void ice_service_task_complete(struct ice_pf *pf)
1664 {
1665 	WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1666 
1667 	/* force memory (pf->state) to sync before next service task */
1668 	smp_mb__before_atomic();
1669 	clear_bit(ICE_SERVICE_SCHED, pf->state);
1670 }
1671 
1672 /**
1673  * ice_service_task_stop - stop service task and cancel works
1674  * @pf: board private structure
1675  *
1676  * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1677  * 1 otherwise.
1678  */
1679 static int ice_service_task_stop(struct ice_pf *pf)
1680 {
1681 	int ret;
1682 
1683 	ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1684 
1685 	if (pf->serv_tmr.function)
1686 		del_timer_sync(&pf->serv_tmr);
1687 	if (pf->serv_task.func)
1688 		cancel_work_sync(&pf->serv_task);
1689 
1690 	clear_bit(ICE_SERVICE_SCHED, pf->state);
1691 	return ret;
1692 }
1693 
1694 /**
1695  * ice_service_task_restart - restart service task and schedule works
1696  * @pf: board private structure
1697  *
1698  * This function is needed for suspend and resume works (e.g WoL scenario)
1699  */
1700 static void ice_service_task_restart(struct ice_pf *pf)
1701 {
1702 	clear_bit(ICE_SERVICE_DIS, pf->state);
1703 	ice_service_task_schedule(pf);
1704 }
1705 
1706 /**
1707  * ice_service_timer - timer callback to schedule service task
1708  * @t: pointer to timer_list
1709  */
1710 static void ice_service_timer(struct timer_list *t)
1711 {
1712 	struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1713 
1714 	mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1715 	ice_service_task_schedule(pf);
1716 }
1717 
1718 /**
1719  * ice_handle_mdd_event - handle malicious driver detect event
1720  * @pf: pointer to the PF structure
1721  *
1722  * Called from service task. OICR interrupt handler indicates MDD event.
1723  * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1724  * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1725  * disable the queue, the PF can be configured to reset the VF using ethtool
1726  * private flag mdd-auto-reset-vf.
1727  */
1728 static void ice_handle_mdd_event(struct ice_pf *pf)
1729 {
1730 	struct device *dev = ice_pf_to_dev(pf);
1731 	struct ice_hw *hw = &pf->hw;
1732 	struct ice_vf *vf;
1733 	unsigned int bkt;
1734 	u32 reg;
1735 
1736 	if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1737 		/* Since the VF MDD event logging is rate limited, check if
1738 		 * there are pending MDD events.
1739 		 */
1740 		ice_print_vfs_mdd_events(pf);
1741 		return;
1742 	}
1743 
1744 	/* find what triggered an MDD event */
1745 	reg = rd32(hw, GL_MDET_TX_PQM);
1746 	if (reg & GL_MDET_TX_PQM_VALID_M) {
1747 		u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1748 				GL_MDET_TX_PQM_PF_NUM_S;
1749 		u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1750 				GL_MDET_TX_PQM_VF_NUM_S;
1751 		u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1752 				GL_MDET_TX_PQM_MAL_TYPE_S;
1753 		u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1754 				GL_MDET_TX_PQM_QNUM_S);
1755 
1756 		if (netif_msg_tx_err(pf))
1757 			dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1758 				 event, queue, pf_num, vf_num);
1759 		wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1760 	}
1761 
1762 	reg = rd32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw));
1763 	if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1764 		u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1765 				GL_MDET_TX_TCLAN_PF_NUM_S;
1766 		u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1767 				GL_MDET_TX_TCLAN_VF_NUM_S;
1768 		u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1769 				GL_MDET_TX_TCLAN_MAL_TYPE_S;
1770 		u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1771 				GL_MDET_TX_TCLAN_QNUM_S);
1772 
1773 		if (netif_msg_tx_err(pf))
1774 			dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1775 				 event, queue, pf_num, vf_num);
1776 		wr32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw), U32_MAX);
1777 	}
1778 
1779 	reg = rd32(hw, GL_MDET_RX);
1780 	if (reg & GL_MDET_RX_VALID_M) {
1781 		u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1782 				GL_MDET_RX_PF_NUM_S;
1783 		u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1784 				GL_MDET_RX_VF_NUM_S;
1785 		u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1786 				GL_MDET_RX_MAL_TYPE_S;
1787 		u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1788 				GL_MDET_RX_QNUM_S);
1789 
1790 		if (netif_msg_rx_err(pf))
1791 			dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1792 				 event, queue, pf_num, vf_num);
1793 		wr32(hw, GL_MDET_RX, 0xffffffff);
1794 	}
1795 
1796 	/* check to see if this PF caused an MDD event */
1797 	reg = rd32(hw, PF_MDET_TX_PQM);
1798 	if (reg & PF_MDET_TX_PQM_VALID_M) {
1799 		wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1800 		if (netif_msg_tx_err(pf))
1801 			dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1802 	}
1803 
1804 	reg = rd32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw));
1805 	if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1806 		wr32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw), 0xffff);
1807 		if (netif_msg_tx_err(pf))
1808 			dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1809 	}
1810 
1811 	reg = rd32(hw, PF_MDET_RX);
1812 	if (reg & PF_MDET_RX_VALID_M) {
1813 		wr32(hw, PF_MDET_RX, 0xFFFF);
1814 		if (netif_msg_rx_err(pf))
1815 			dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1816 	}
1817 
1818 	/* Check to see if one of the VFs caused an MDD event, and then
1819 	 * increment counters and set print pending
1820 	 */
1821 	mutex_lock(&pf->vfs.table_lock);
1822 	ice_for_each_vf(pf, bkt, vf) {
1823 		reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id));
1824 		if (reg & VP_MDET_TX_PQM_VALID_M) {
1825 			wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF);
1826 			vf->mdd_tx_events.count++;
1827 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1828 			if (netif_msg_tx_err(pf))
1829 				dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1830 					 vf->vf_id);
1831 		}
1832 
1833 		reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id));
1834 		if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1835 			wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF);
1836 			vf->mdd_tx_events.count++;
1837 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1838 			if (netif_msg_tx_err(pf))
1839 				dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1840 					 vf->vf_id);
1841 		}
1842 
1843 		reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id));
1844 		if (reg & VP_MDET_TX_TDPU_VALID_M) {
1845 			wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF);
1846 			vf->mdd_tx_events.count++;
1847 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1848 			if (netif_msg_tx_err(pf))
1849 				dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1850 					 vf->vf_id);
1851 		}
1852 
1853 		reg = rd32(hw, VP_MDET_RX(vf->vf_id));
1854 		if (reg & VP_MDET_RX_VALID_M) {
1855 			wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF);
1856 			vf->mdd_rx_events.count++;
1857 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1858 			if (netif_msg_rx_err(pf))
1859 				dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1860 					 vf->vf_id);
1861 
1862 			/* Since the queue is disabled on VF Rx MDD events, the
1863 			 * PF can be configured to reset the VF through ethtool
1864 			 * private flag mdd-auto-reset-vf.
1865 			 */
1866 			if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1867 				/* VF MDD event counters will be cleared by
1868 				 * reset, so print the event prior to reset.
1869 				 */
1870 				ice_print_vf_rx_mdd_event(vf);
1871 				ice_reset_vf(vf, ICE_VF_RESET_LOCK);
1872 			}
1873 		}
1874 	}
1875 	mutex_unlock(&pf->vfs.table_lock);
1876 
1877 	ice_print_vfs_mdd_events(pf);
1878 }
1879 
1880 /**
1881  * ice_force_phys_link_state - Force the physical link state
1882  * @vsi: VSI to force the physical link state to up/down
1883  * @link_up: true/false indicates to set the physical link to up/down
1884  *
1885  * Force the physical link state by getting the current PHY capabilities from
1886  * hardware and setting the PHY config based on the determined capabilities. If
1887  * link changes a link event will be triggered because both the Enable Automatic
1888  * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1889  *
1890  * Returns 0 on success, negative on failure
1891  */
1892 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1893 {
1894 	struct ice_aqc_get_phy_caps_data *pcaps;
1895 	struct ice_aqc_set_phy_cfg_data *cfg;
1896 	struct ice_port_info *pi;
1897 	struct device *dev;
1898 	int retcode;
1899 
1900 	if (!vsi || !vsi->port_info || !vsi->back)
1901 		return -EINVAL;
1902 	if (vsi->type != ICE_VSI_PF)
1903 		return 0;
1904 
1905 	dev = ice_pf_to_dev(vsi->back);
1906 
1907 	pi = vsi->port_info;
1908 
1909 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1910 	if (!pcaps)
1911 		return -ENOMEM;
1912 
1913 	retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1914 				      NULL);
1915 	if (retcode) {
1916 		dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1917 			vsi->vsi_num, retcode);
1918 		retcode = -EIO;
1919 		goto out;
1920 	}
1921 
1922 	/* No change in link */
1923 	if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1924 	    link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1925 		goto out;
1926 
1927 	/* Use the current user PHY configuration. The current user PHY
1928 	 * configuration is initialized during probe from PHY capabilities
1929 	 * software mode, and updated on set PHY configuration.
1930 	 */
1931 	cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1932 	if (!cfg) {
1933 		retcode = -ENOMEM;
1934 		goto out;
1935 	}
1936 
1937 	cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1938 	if (link_up)
1939 		cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1940 	else
1941 		cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1942 
1943 	retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1944 	if (retcode) {
1945 		dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1946 			vsi->vsi_num, retcode);
1947 		retcode = -EIO;
1948 	}
1949 
1950 	kfree(cfg);
1951 out:
1952 	kfree(pcaps);
1953 	return retcode;
1954 }
1955 
1956 /**
1957  * ice_init_nvm_phy_type - Initialize the NVM PHY type
1958  * @pi: port info structure
1959  *
1960  * Initialize nvm_phy_type_[low|high] for link lenient mode support
1961  */
1962 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1963 {
1964 	struct ice_aqc_get_phy_caps_data *pcaps;
1965 	struct ice_pf *pf = pi->hw->back;
1966 	int err;
1967 
1968 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1969 	if (!pcaps)
1970 		return -ENOMEM;
1971 
1972 	err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA,
1973 				  pcaps, NULL);
1974 
1975 	if (err) {
1976 		dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1977 		goto out;
1978 	}
1979 
1980 	pf->nvm_phy_type_hi = pcaps->phy_type_high;
1981 	pf->nvm_phy_type_lo = pcaps->phy_type_low;
1982 
1983 out:
1984 	kfree(pcaps);
1985 	return err;
1986 }
1987 
1988 /**
1989  * ice_init_link_dflt_override - Initialize link default override
1990  * @pi: port info structure
1991  *
1992  * Initialize link default override and PHY total port shutdown during probe
1993  */
1994 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1995 {
1996 	struct ice_link_default_override_tlv *ldo;
1997 	struct ice_pf *pf = pi->hw->back;
1998 
1999 	ldo = &pf->link_dflt_override;
2000 	if (ice_get_link_default_override(ldo, pi))
2001 		return;
2002 
2003 	if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
2004 		return;
2005 
2006 	/* Enable Total Port Shutdown (override/replace link-down-on-close
2007 	 * ethtool private flag) for ports with Port Disable bit set.
2008 	 */
2009 	set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
2010 	set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
2011 }
2012 
2013 /**
2014  * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
2015  * @pi: port info structure
2016  *
2017  * If default override is enabled, initialize the user PHY cfg speed and FEC
2018  * settings using the default override mask from the NVM.
2019  *
2020  * The PHY should only be configured with the default override settings the
2021  * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
2022  * is used to indicate that the user PHY cfg default override is initialized
2023  * and the PHY has not been configured with the default override settings. The
2024  * state is set here, and cleared in ice_configure_phy the first time the PHY is
2025  * configured.
2026  *
2027  * This function should be called only if the FW doesn't support default
2028  * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
2029  */
2030 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
2031 {
2032 	struct ice_link_default_override_tlv *ldo;
2033 	struct ice_aqc_set_phy_cfg_data *cfg;
2034 	struct ice_phy_info *phy = &pi->phy;
2035 	struct ice_pf *pf = pi->hw->back;
2036 
2037 	ldo = &pf->link_dflt_override;
2038 
2039 	/* If link default override is enabled, use to mask NVM PHY capabilities
2040 	 * for speed and FEC default configuration.
2041 	 */
2042 	cfg = &phy->curr_user_phy_cfg;
2043 
2044 	if (ldo->phy_type_low || ldo->phy_type_high) {
2045 		cfg->phy_type_low = pf->nvm_phy_type_lo &
2046 				    cpu_to_le64(ldo->phy_type_low);
2047 		cfg->phy_type_high = pf->nvm_phy_type_hi &
2048 				     cpu_to_le64(ldo->phy_type_high);
2049 	}
2050 	cfg->link_fec_opt = ldo->fec_options;
2051 	phy->curr_user_fec_req = ICE_FEC_AUTO;
2052 
2053 	set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
2054 }
2055 
2056 /**
2057  * ice_init_phy_user_cfg - Initialize the PHY user configuration
2058  * @pi: port info structure
2059  *
2060  * Initialize the current user PHY configuration, speed, FEC, and FC requested
2061  * mode to default. The PHY defaults are from get PHY capabilities topology
2062  * with media so call when media is first available. An error is returned if
2063  * called when media is not available. The PHY initialization completed state is
2064  * set here.
2065  *
2066  * These configurations are used when setting PHY
2067  * configuration. The user PHY configuration is updated on set PHY
2068  * configuration. Returns 0 on success, negative on failure
2069  */
2070 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
2071 {
2072 	struct ice_aqc_get_phy_caps_data *pcaps;
2073 	struct ice_phy_info *phy = &pi->phy;
2074 	struct ice_pf *pf = pi->hw->back;
2075 	int err;
2076 
2077 	if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2078 		return -EIO;
2079 
2080 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2081 	if (!pcaps)
2082 		return -ENOMEM;
2083 
2084 	if (ice_fw_supports_report_dflt_cfg(pi->hw))
2085 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2086 					  pcaps, NULL);
2087 	else
2088 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2089 					  pcaps, NULL);
2090 	if (err) {
2091 		dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
2092 		goto err_out;
2093 	}
2094 
2095 	ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
2096 
2097 	/* check if lenient mode is supported and enabled */
2098 	if (ice_fw_supports_link_override(pi->hw) &&
2099 	    !(pcaps->module_compliance_enforcement &
2100 	      ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
2101 		set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
2102 
2103 		/* if the FW supports default PHY configuration mode, then the driver
2104 		 * does not have to apply link override settings. If not,
2105 		 * initialize user PHY configuration with link override values
2106 		 */
2107 		if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
2108 		    (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
2109 			ice_init_phy_cfg_dflt_override(pi);
2110 			goto out;
2111 		}
2112 	}
2113 
2114 	/* if link default override is not enabled, set user flow control and
2115 	 * FEC settings based on what get_phy_caps returned
2116 	 */
2117 	phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
2118 						      pcaps->link_fec_options);
2119 	phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
2120 
2121 out:
2122 	phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
2123 	set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
2124 err_out:
2125 	kfree(pcaps);
2126 	return err;
2127 }
2128 
2129 /**
2130  * ice_configure_phy - configure PHY
2131  * @vsi: VSI of PHY
2132  *
2133  * Set the PHY configuration. If the current PHY configuration is the same as
2134  * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
2135  * configure the based get PHY capabilities for topology with media.
2136  */
2137 static int ice_configure_phy(struct ice_vsi *vsi)
2138 {
2139 	struct device *dev = ice_pf_to_dev(vsi->back);
2140 	struct ice_port_info *pi = vsi->port_info;
2141 	struct ice_aqc_get_phy_caps_data *pcaps;
2142 	struct ice_aqc_set_phy_cfg_data *cfg;
2143 	struct ice_phy_info *phy = &pi->phy;
2144 	struct ice_pf *pf = vsi->back;
2145 	int err;
2146 
2147 	/* Ensure we have media as we cannot configure a medialess port */
2148 	if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2149 		return -EPERM;
2150 
2151 	ice_print_topo_conflict(vsi);
2152 
2153 	if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) &&
2154 	    phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
2155 		return -EPERM;
2156 
2157 	if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
2158 		return ice_force_phys_link_state(vsi, true);
2159 
2160 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2161 	if (!pcaps)
2162 		return -ENOMEM;
2163 
2164 	/* Get current PHY config */
2165 	err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
2166 				  NULL);
2167 	if (err) {
2168 		dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n",
2169 			vsi->vsi_num, err);
2170 		goto done;
2171 	}
2172 
2173 	/* If PHY enable link is configured and configuration has not changed,
2174 	 * there's nothing to do
2175 	 */
2176 	if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
2177 	    ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
2178 		goto done;
2179 
2180 	/* Use PHY topology as baseline for configuration */
2181 	memset(pcaps, 0, sizeof(*pcaps));
2182 	if (ice_fw_supports_report_dflt_cfg(pi->hw))
2183 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2184 					  pcaps, NULL);
2185 	else
2186 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2187 					  pcaps, NULL);
2188 	if (err) {
2189 		dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n",
2190 			vsi->vsi_num, err);
2191 		goto done;
2192 	}
2193 
2194 	cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2195 	if (!cfg) {
2196 		err = -ENOMEM;
2197 		goto done;
2198 	}
2199 
2200 	ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2201 
2202 	/* Speed - If default override pending, use curr_user_phy_cfg set in
2203 	 * ice_init_phy_user_cfg_ldo.
2204 	 */
2205 	if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2206 			       vsi->back->state)) {
2207 		cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2208 		cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2209 	} else {
2210 		u64 phy_low = 0, phy_high = 0;
2211 
2212 		ice_update_phy_type(&phy_low, &phy_high,
2213 				    pi->phy.curr_user_speed_req);
2214 		cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2215 		cfg->phy_type_high = pcaps->phy_type_high &
2216 				     cpu_to_le64(phy_high);
2217 	}
2218 
2219 	/* Can't provide what was requested; use PHY capabilities */
2220 	if (!cfg->phy_type_low && !cfg->phy_type_high) {
2221 		cfg->phy_type_low = pcaps->phy_type_low;
2222 		cfg->phy_type_high = pcaps->phy_type_high;
2223 	}
2224 
2225 	/* FEC */
2226 	ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2227 
2228 	/* Can't provide what was requested; use PHY capabilities */
2229 	if (cfg->link_fec_opt !=
2230 	    (cfg->link_fec_opt & pcaps->link_fec_options)) {
2231 		cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2232 		cfg->link_fec_opt = pcaps->link_fec_options;
2233 	}
2234 
2235 	/* Flow Control - always supported; no need to check against
2236 	 * capabilities
2237 	 */
2238 	ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2239 
2240 	/* Enable link and link update */
2241 	cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2242 
2243 	err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2244 	if (err)
2245 		dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
2246 			vsi->vsi_num, err);
2247 
2248 	kfree(cfg);
2249 done:
2250 	kfree(pcaps);
2251 	return err;
2252 }
2253 
2254 /**
2255  * ice_check_media_subtask - Check for media
2256  * @pf: pointer to PF struct
2257  *
2258  * If media is available, then initialize PHY user configuration if it is not
2259  * been, and configure the PHY if the interface is up.
2260  */
2261 static void ice_check_media_subtask(struct ice_pf *pf)
2262 {
2263 	struct ice_port_info *pi;
2264 	struct ice_vsi *vsi;
2265 	int err;
2266 
2267 	/* No need to check for media if it's already present */
2268 	if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2269 		return;
2270 
2271 	vsi = ice_get_main_vsi(pf);
2272 	if (!vsi)
2273 		return;
2274 
2275 	/* Refresh link info and check if media is present */
2276 	pi = vsi->port_info;
2277 	err = ice_update_link_info(pi);
2278 	if (err)
2279 		return;
2280 
2281 	ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
2282 
2283 	if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2284 		if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2285 			ice_init_phy_user_cfg(pi);
2286 
2287 		/* PHY settings are reset on media insertion, reconfigure
2288 		 * PHY to preserve settings.
2289 		 */
2290 		if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2291 		    test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2292 			return;
2293 
2294 		err = ice_configure_phy(vsi);
2295 		if (!err)
2296 			clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2297 
2298 		/* A Link Status Event will be generated; the event handler
2299 		 * will complete bringing the interface up
2300 		 */
2301 	}
2302 }
2303 
2304 /**
2305  * ice_service_task - manage and run subtasks
2306  * @work: pointer to work_struct contained by the PF struct
2307  */
2308 static void ice_service_task(struct work_struct *work)
2309 {
2310 	struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2311 	unsigned long start_time = jiffies;
2312 
2313 	/* subtasks */
2314 
2315 	/* process reset requests first */
2316 	ice_reset_subtask(pf);
2317 
2318 	/* bail if a reset/recovery cycle is pending or rebuild failed */
2319 	if (ice_is_reset_in_progress(pf->state) ||
2320 	    test_bit(ICE_SUSPENDED, pf->state) ||
2321 	    test_bit(ICE_NEEDS_RESTART, pf->state)) {
2322 		ice_service_task_complete(pf);
2323 		return;
2324 	}
2325 
2326 	if (test_and_clear_bit(ICE_AUX_ERR_PENDING, pf->state)) {
2327 		struct iidc_event *event;
2328 
2329 		event = kzalloc(sizeof(*event), GFP_KERNEL);
2330 		if (event) {
2331 			set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2332 			/* report the entire OICR value to AUX driver */
2333 			swap(event->reg, pf->oicr_err_reg);
2334 			ice_send_event_to_aux(pf, event);
2335 			kfree(event);
2336 		}
2337 	}
2338 
2339 	/* unplug aux dev per request, if an unplug request came in
2340 	 * while processing a plug request, this will handle it
2341 	 */
2342 	if (test_and_clear_bit(ICE_FLAG_UNPLUG_AUX_DEV, pf->flags))
2343 		ice_unplug_aux_dev(pf);
2344 
2345 	/* Plug aux device per request */
2346 	if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
2347 		ice_plug_aux_dev(pf);
2348 
2349 	if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
2350 		struct iidc_event *event;
2351 
2352 		event = kzalloc(sizeof(*event), GFP_KERNEL);
2353 		if (event) {
2354 			set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
2355 			ice_send_event_to_aux(pf, event);
2356 			kfree(event);
2357 		}
2358 	}
2359 
2360 	ice_clean_adminq_subtask(pf);
2361 	ice_check_media_subtask(pf);
2362 	ice_check_for_hang_subtask(pf);
2363 	ice_sync_fltr_subtask(pf);
2364 	ice_handle_mdd_event(pf);
2365 	ice_watchdog_subtask(pf);
2366 
2367 	if (ice_is_safe_mode(pf)) {
2368 		ice_service_task_complete(pf);
2369 		return;
2370 	}
2371 
2372 	ice_process_vflr_event(pf);
2373 	ice_clean_mailboxq_subtask(pf);
2374 	ice_clean_sbq_subtask(pf);
2375 	ice_sync_arfs_fltrs(pf);
2376 	ice_flush_fdir_ctx(pf);
2377 
2378 	/* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2379 	ice_service_task_complete(pf);
2380 
2381 	/* If the tasks have taken longer than one service timer period
2382 	 * or there is more work to be done, reset the service timer to
2383 	 * schedule the service task now.
2384 	 */
2385 	if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2386 	    test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2387 	    test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2388 	    test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2389 	    test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2390 	    test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2391 	    test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2392 		mod_timer(&pf->serv_tmr, jiffies);
2393 }
2394 
2395 /**
2396  * ice_set_ctrlq_len - helper function to set controlq length
2397  * @hw: pointer to the HW instance
2398  */
2399 static void ice_set_ctrlq_len(struct ice_hw *hw)
2400 {
2401 	hw->adminq.num_rq_entries = ICE_AQ_LEN;
2402 	hw->adminq.num_sq_entries = ICE_AQ_LEN;
2403 	hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2404 	hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2405 	hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2406 	hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2407 	hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2408 	hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2409 	hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2410 	hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2411 	hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2412 	hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2413 }
2414 
2415 /**
2416  * ice_schedule_reset - schedule a reset
2417  * @pf: board private structure
2418  * @reset: reset being requested
2419  */
2420 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2421 {
2422 	struct device *dev = ice_pf_to_dev(pf);
2423 
2424 	/* bail out if earlier reset has failed */
2425 	if (test_bit(ICE_RESET_FAILED, pf->state)) {
2426 		dev_dbg(dev, "earlier reset has failed\n");
2427 		return -EIO;
2428 	}
2429 	/* bail if reset/recovery already in progress */
2430 	if (ice_is_reset_in_progress(pf->state)) {
2431 		dev_dbg(dev, "Reset already in progress\n");
2432 		return -EBUSY;
2433 	}
2434 
2435 	switch (reset) {
2436 	case ICE_RESET_PFR:
2437 		set_bit(ICE_PFR_REQ, pf->state);
2438 		break;
2439 	case ICE_RESET_CORER:
2440 		set_bit(ICE_CORER_REQ, pf->state);
2441 		break;
2442 	case ICE_RESET_GLOBR:
2443 		set_bit(ICE_GLOBR_REQ, pf->state);
2444 		break;
2445 	default:
2446 		return -EINVAL;
2447 	}
2448 
2449 	ice_service_task_schedule(pf);
2450 	return 0;
2451 }
2452 
2453 /**
2454  * ice_irq_affinity_notify - Callback for affinity changes
2455  * @notify: context as to what irq was changed
2456  * @mask: the new affinity mask
2457  *
2458  * This is a callback function used by the irq_set_affinity_notifier function
2459  * so that we may register to receive changes to the irq affinity masks.
2460  */
2461 static void
2462 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2463 			const cpumask_t *mask)
2464 {
2465 	struct ice_q_vector *q_vector =
2466 		container_of(notify, struct ice_q_vector, affinity_notify);
2467 
2468 	cpumask_copy(&q_vector->affinity_mask, mask);
2469 }
2470 
2471 /**
2472  * ice_irq_affinity_release - Callback for affinity notifier release
2473  * @ref: internal core kernel usage
2474  *
2475  * This is a callback function used by the irq_set_affinity_notifier function
2476  * to inform the current notification subscriber that they will no longer
2477  * receive notifications.
2478  */
2479 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2480 
2481 /**
2482  * ice_vsi_ena_irq - Enable IRQ for the given VSI
2483  * @vsi: the VSI being configured
2484  */
2485 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2486 {
2487 	struct ice_hw *hw = &vsi->back->hw;
2488 	int i;
2489 
2490 	ice_for_each_q_vector(vsi, i)
2491 		ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2492 
2493 	ice_flush(hw);
2494 	return 0;
2495 }
2496 
2497 /**
2498  * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2499  * @vsi: the VSI being configured
2500  * @basename: name for the vector
2501  */
2502 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2503 {
2504 	int q_vectors = vsi->num_q_vectors;
2505 	struct ice_pf *pf = vsi->back;
2506 	struct device *dev;
2507 	int rx_int_idx = 0;
2508 	int tx_int_idx = 0;
2509 	int vector, err;
2510 	int irq_num;
2511 
2512 	dev = ice_pf_to_dev(pf);
2513 	for (vector = 0; vector < q_vectors; vector++) {
2514 		struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2515 
2516 		irq_num = q_vector->irq.virq;
2517 
2518 		if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2519 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2520 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2521 			tx_int_idx++;
2522 		} else if (q_vector->rx.rx_ring) {
2523 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2524 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
2525 		} else if (q_vector->tx.tx_ring) {
2526 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2527 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
2528 		} else {
2529 			/* skip this unused q_vector */
2530 			continue;
2531 		}
2532 		if (vsi->type == ICE_VSI_CTRL && vsi->vf)
2533 			err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2534 					       IRQF_SHARED, q_vector->name,
2535 					       q_vector);
2536 		else
2537 			err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2538 					       0, q_vector->name, q_vector);
2539 		if (err) {
2540 			netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2541 				   err);
2542 			goto free_q_irqs;
2543 		}
2544 
2545 		/* register for affinity change notifications */
2546 		if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2547 			struct irq_affinity_notify *affinity_notify;
2548 
2549 			affinity_notify = &q_vector->affinity_notify;
2550 			affinity_notify->notify = ice_irq_affinity_notify;
2551 			affinity_notify->release = ice_irq_affinity_release;
2552 			irq_set_affinity_notifier(irq_num, affinity_notify);
2553 		}
2554 
2555 		/* assign the mask for this irq */
2556 		irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2557 	}
2558 
2559 	err = ice_set_cpu_rx_rmap(vsi);
2560 	if (err) {
2561 		netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n",
2562 			   vsi->vsi_num, ERR_PTR(err));
2563 		goto free_q_irqs;
2564 	}
2565 
2566 	vsi->irqs_ready = true;
2567 	return 0;
2568 
2569 free_q_irqs:
2570 	while (vector--) {
2571 		irq_num = vsi->q_vectors[vector]->irq.virq;
2572 		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2573 			irq_set_affinity_notifier(irq_num, NULL);
2574 		irq_set_affinity_hint(irq_num, NULL);
2575 		devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2576 	}
2577 	return err;
2578 }
2579 
2580 /**
2581  * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2582  * @vsi: VSI to setup Tx rings used by XDP
2583  *
2584  * Return 0 on success and negative value on error
2585  */
2586 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2587 {
2588 	struct device *dev = ice_pf_to_dev(vsi->back);
2589 	struct ice_tx_desc *tx_desc;
2590 	int i, j;
2591 
2592 	ice_for_each_xdp_txq(vsi, i) {
2593 		u16 xdp_q_idx = vsi->alloc_txq + i;
2594 		struct ice_ring_stats *ring_stats;
2595 		struct ice_tx_ring *xdp_ring;
2596 
2597 		xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2598 		if (!xdp_ring)
2599 			goto free_xdp_rings;
2600 
2601 		ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
2602 		if (!ring_stats) {
2603 			ice_free_tx_ring(xdp_ring);
2604 			goto free_xdp_rings;
2605 		}
2606 
2607 		xdp_ring->ring_stats = ring_stats;
2608 		xdp_ring->q_index = xdp_q_idx;
2609 		xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2610 		xdp_ring->vsi = vsi;
2611 		xdp_ring->netdev = NULL;
2612 		xdp_ring->dev = dev;
2613 		xdp_ring->count = vsi->num_tx_desc;
2614 		WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2615 		if (ice_setup_tx_ring(xdp_ring))
2616 			goto free_xdp_rings;
2617 		ice_set_ring_xdp(xdp_ring);
2618 		spin_lock_init(&xdp_ring->tx_lock);
2619 		for (j = 0; j < xdp_ring->count; j++) {
2620 			tx_desc = ICE_TX_DESC(xdp_ring, j);
2621 			tx_desc->cmd_type_offset_bsz = 0;
2622 		}
2623 	}
2624 
2625 	return 0;
2626 
2627 free_xdp_rings:
2628 	for (; i >= 0; i--) {
2629 		if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc) {
2630 			kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2631 			vsi->xdp_rings[i]->ring_stats = NULL;
2632 			ice_free_tx_ring(vsi->xdp_rings[i]);
2633 		}
2634 	}
2635 	return -ENOMEM;
2636 }
2637 
2638 /**
2639  * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2640  * @vsi: VSI to set the bpf prog on
2641  * @prog: the bpf prog pointer
2642  */
2643 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2644 {
2645 	struct bpf_prog *old_prog;
2646 	int i;
2647 
2648 	old_prog = xchg(&vsi->xdp_prog, prog);
2649 	ice_for_each_rxq(vsi, i)
2650 		WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2651 
2652 	if (old_prog)
2653 		bpf_prog_put(old_prog);
2654 }
2655 
2656 /**
2657  * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2658  * @vsi: VSI to bring up Tx rings used by XDP
2659  * @prog: bpf program that will be assigned to VSI
2660  *
2661  * Return 0 on success and negative value on error
2662  */
2663 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2664 {
2665 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2666 	int xdp_rings_rem = vsi->num_xdp_txq;
2667 	struct ice_pf *pf = vsi->back;
2668 	struct ice_qs_cfg xdp_qs_cfg = {
2669 		.qs_mutex = &pf->avail_q_mutex,
2670 		.pf_map = pf->avail_txqs,
2671 		.pf_map_size = pf->max_pf_txqs,
2672 		.q_count = vsi->num_xdp_txq,
2673 		.scatter_count = ICE_MAX_SCATTER_TXQS,
2674 		.vsi_map = vsi->txq_map,
2675 		.vsi_map_offset = vsi->alloc_txq,
2676 		.mapping_mode = ICE_VSI_MAP_CONTIG
2677 	};
2678 	struct device *dev;
2679 	int i, v_idx;
2680 	int status;
2681 
2682 	dev = ice_pf_to_dev(pf);
2683 	vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2684 				      sizeof(*vsi->xdp_rings), GFP_KERNEL);
2685 	if (!vsi->xdp_rings)
2686 		return -ENOMEM;
2687 
2688 	vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2689 	if (__ice_vsi_get_qs(&xdp_qs_cfg))
2690 		goto err_map_xdp;
2691 
2692 	if (static_key_enabled(&ice_xdp_locking_key))
2693 		netdev_warn(vsi->netdev,
2694 			    "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2695 
2696 	if (ice_xdp_alloc_setup_rings(vsi))
2697 		goto clear_xdp_rings;
2698 
2699 	/* follow the logic from ice_vsi_map_rings_to_vectors */
2700 	ice_for_each_q_vector(vsi, v_idx) {
2701 		struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2702 		int xdp_rings_per_v, q_id, q_base;
2703 
2704 		xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2705 					       vsi->num_q_vectors - v_idx);
2706 		q_base = vsi->num_xdp_txq - xdp_rings_rem;
2707 
2708 		for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2709 			struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2710 
2711 			xdp_ring->q_vector = q_vector;
2712 			xdp_ring->next = q_vector->tx.tx_ring;
2713 			q_vector->tx.tx_ring = xdp_ring;
2714 		}
2715 		xdp_rings_rem -= xdp_rings_per_v;
2716 	}
2717 
2718 	ice_for_each_rxq(vsi, i) {
2719 		if (static_key_enabled(&ice_xdp_locking_key)) {
2720 			vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2721 		} else {
2722 			struct ice_q_vector *q_vector = vsi->rx_rings[i]->q_vector;
2723 			struct ice_tx_ring *ring;
2724 
2725 			ice_for_each_tx_ring(ring, q_vector->tx) {
2726 				if (ice_ring_is_xdp(ring)) {
2727 					vsi->rx_rings[i]->xdp_ring = ring;
2728 					break;
2729 				}
2730 			}
2731 		}
2732 		ice_tx_xsk_pool(vsi, i);
2733 	}
2734 
2735 	/* omit the scheduler update if in reset path; XDP queues will be
2736 	 * taken into account at the end of ice_vsi_rebuild, where
2737 	 * ice_cfg_vsi_lan is being called
2738 	 */
2739 	if (ice_is_reset_in_progress(pf->state))
2740 		return 0;
2741 
2742 	/* tell the Tx scheduler that right now we have
2743 	 * additional queues
2744 	 */
2745 	for (i = 0; i < vsi->tc_cfg.numtc; i++)
2746 		max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2747 
2748 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2749 				 max_txqs);
2750 	if (status) {
2751 		dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2752 			status);
2753 		goto clear_xdp_rings;
2754 	}
2755 
2756 	/* assign the prog only when it's not already present on VSI;
2757 	 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2758 	 * VSI rebuild that happens under ethtool -L can expose us to
2759 	 * the bpf_prog refcount issues as we would be swapping same
2760 	 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2761 	 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2762 	 * this is not harmful as dev_xdp_install bumps the refcount
2763 	 * before calling the op exposed by the driver;
2764 	 */
2765 	if (!ice_is_xdp_ena_vsi(vsi))
2766 		ice_vsi_assign_bpf_prog(vsi, prog);
2767 
2768 	return 0;
2769 clear_xdp_rings:
2770 	ice_for_each_xdp_txq(vsi, i)
2771 		if (vsi->xdp_rings[i]) {
2772 			kfree_rcu(vsi->xdp_rings[i], rcu);
2773 			vsi->xdp_rings[i] = NULL;
2774 		}
2775 
2776 err_map_xdp:
2777 	mutex_lock(&pf->avail_q_mutex);
2778 	ice_for_each_xdp_txq(vsi, i) {
2779 		clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2780 		vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2781 	}
2782 	mutex_unlock(&pf->avail_q_mutex);
2783 
2784 	devm_kfree(dev, vsi->xdp_rings);
2785 	return -ENOMEM;
2786 }
2787 
2788 /**
2789  * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2790  * @vsi: VSI to remove XDP rings
2791  *
2792  * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2793  * resources
2794  */
2795 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2796 {
2797 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2798 	struct ice_pf *pf = vsi->back;
2799 	int i, v_idx;
2800 
2801 	/* q_vectors are freed in reset path so there's no point in detaching
2802 	 * rings; in case of rebuild being triggered not from reset bits
2803 	 * in pf->state won't be set, so additionally check first q_vector
2804 	 * against NULL
2805 	 */
2806 	if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2807 		goto free_qmap;
2808 
2809 	ice_for_each_q_vector(vsi, v_idx) {
2810 		struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2811 		struct ice_tx_ring *ring;
2812 
2813 		ice_for_each_tx_ring(ring, q_vector->tx)
2814 			if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2815 				break;
2816 
2817 		/* restore the value of last node prior to XDP setup */
2818 		q_vector->tx.tx_ring = ring;
2819 	}
2820 
2821 free_qmap:
2822 	mutex_lock(&pf->avail_q_mutex);
2823 	ice_for_each_xdp_txq(vsi, i) {
2824 		clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2825 		vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2826 	}
2827 	mutex_unlock(&pf->avail_q_mutex);
2828 
2829 	ice_for_each_xdp_txq(vsi, i)
2830 		if (vsi->xdp_rings[i]) {
2831 			if (vsi->xdp_rings[i]->desc) {
2832 				synchronize_rcu();
2833 				ice_free_tx_ring(vsi->xdp_rings[i]);
2834 			}
2835 			kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2836 			vsi->xdp_rings[i]->ring_stats = NULL;
2837 			kfree_rcu(vsi->xdp_rings[i], rcu);
2838 			vsi->xdp_rings[i] = NULL;
2839 		}
2840 
2841 	devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2842 	vsi->xdp_rings = NULL;
2843 
2844 	if (static_key_enabled(&ice_xdp_locking_key))
2845 		static_branch_dec(&ice_xdp_locking_key);
2846 
2847 	if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2848 		return 0;
2849 
2850 	ice_vsi_assign_bpf_prog(vsi, NULL);
2851 
2852 	/* notify Tx scheduler that we destroyed XDP queues and bring
2853 	 * back the old number of child nodes
2854 	 */
2855 	for (i = 0; i < vsi->tc_cfg.numtc; i++)
2856 		max_txqs[i] = vsi->num_txq;
2857 
2858 	/* change number of XDP Tx queues to 0 */
2859 	vsi->num_xdp_txq = 0;
2860 
2861 	return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2862 			       max_txqs);
2863 }
2864 
2865 /**
2866  * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2867  * @vsi: VSI to schedule napi on
2868  */
2869 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2870 {
2871 	int i;
2872 
2873 	ice_for_each_rxq(vsi, i) {
2874 		struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2875 
2876 		if (rx_ring->xsk_pool)
2877 			napi_schedule(&rx_ring->q_vector->napi);
2878 	}
2879 }
2880 
2881 /**
2882  * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2883  * @vsi: VSI to determine the count of XDP Tx qs
2884  *
2885  * returns 0 if Tx qs count is higher than at least half of CPU count,
2886  * -ENOMEM otherwise
2887  */
2888 int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2889 {
2890 	u16 avail = ice_get_avail_txq_count(vsi->back);
2891 	u16 cpus = num_possible_cpus();
2892 
2893 	if (avail < cpus / 2)
2894 		return -ENOMEM;
2895 
2896 	vsi->num_xdp_txq = min_t(u16, avail, cpus);
2897 
2898 	if (vsi->num_xdp_txq < cpus)
2899 		static_branch_inc(&ice_xdp_locking_key);
2900 
2901 	return 0;
2902 }
2903 
2904 /**
2905  * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2906  * @vsi: Pointer to VSI structure
2907  */
2908 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
2909 {
2910 	if (test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
2911 		return ICE_RXBUF_1664;
2912 	else
2913 		return ICE_RXBUF_3072;
2914 }
2915 
2916 /**
2917  * ice_xdp_setup_prog - Add or remove XDP eBPF program
2918  * @vsi: VSI to setup XDP for
2919  * @prog: XDP program
2920  * @extack: netlink extended ack
2921  */
2922 static int
2923 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2924 		   struct netlink_ext_ack *extack)
2925 {
2926 	unsigned int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2927 	bool if_running = netif_running(vsi->netdev);
2928 	int ret = 0, xdp_ring_err = 0;
2929 
2930 	if (prog && !prog->aux->xdp_has_frags) {
2931 		if (frame_size > ice_max_xdp_frame_size(vsi)) {
2932 			NL_SET_ERR_MSG_MOD(extack,
2933 					   "MTU is too large for linear frames and XDP prog does not support frags");
2934 			return -EOPNOTSUPP;
2935 		}
2936 	}
2937 
2938 	/* hot swap progs and avoid toggling link */
2939 	if (ice_is_xdp_ena_vsi(vsi) == !!prog) {
2940 		ice_vsi_assign_bpf_prog(vsi, prog);
2941 		return 0;
2942 	}
2943 
2944 	/* need to stop netdev while setting up the program for Rx rings */
2945 	if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2946 		ret = ice_down(vsi);
2947 		if (ret) {
2948 			NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2949 			return ret;
2950 		}
2951 	}
2952 
2953 	if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2954 		xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2955 		if (xdp_ring_err) {
2956 			NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2957 		} else {
2958 			xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2959 			if (xdp_ring_err)
2960 				NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2961 		}
2962 		xdp_features_set_redirect_target(vsi->netdev, true);
2963 		/* reallocate Rx queues that are used for zero-copy */
2964 		xdp_ring_err = ice_realloc_zc_buf(vsi, true);
2965 		if (xdp_ring_err)
2966 			NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Rx resources failed");
2967 	} else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2968 		xdp_features_clear_redirect_target(vsi->netdev);
2969 		xdp_ring_err = ice_destroy_xdp_rings(vsi);
2970 		if (xdp_ring_err)
2971 			NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2972 		/* reallocate Rx queues that were used for zero-copy */
2973 		xdp_ring_err = ice_realloc_zc_buf(vsi, false);
2974 		if (xdp_ring_err)
2975 			NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Rx resources failed");
2976 	}
2977 
2978 	if (if_running)
2979 		ret = ice_up(vsi);
2980 
2981 	if (!ret && prog)
2982 		ice_vsi_rx_napi_schedule(vsi);
2983 
2984 	return (ret || xdp_ring_err) ? -ENOMEM : 0;
2985 }
2986 
2987 /**
2988  * ice_xdp_safe_mode - XDP handler for safe mode
2989  * @dev: netdevice
2990  * @xdp: XDP command
2991  */
2992 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2993 			     struct netdev_bpf *xdp)
2994 {
2995 	NL_SET_ERR_MSG_MOD(xdp->extack,
2996 			   "Please provide working DDP firmware package in order to use XDP\n"
2997 			   "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2998 	return -EOPNOTSUPP;
2999 }
3000 
3001 /**
3002  * ice_xdp - implements XDP handler
3003  * @dev: netdevice
3004  * @xdp: XDP command
3005  */
3006 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
3007 {
3008 	struct ice_netdev_priv *np = netdev_priv(dev);
3009 	struct ice_vsi *vsi = np->vsi;
3010 
3011 	if (vsi->type != ICE_VSI_PF) {
3012 		NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
3013 		return -EINVAL;
3014 	}
3015 
3016 	switch (xdp->command) {
3017 	case XDP_SETUP_PROG:
3018 		return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
3019 	case XDP_SETUP_XSK_POOL:
3020 		return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
3021 					  xdp->xsk.queue_id);
3022 	default:
3023 		return -EINVAL;
3024 	}
3025 }
3026 
3027 /**
3028  * ice_ena_misc_vector - enable the non-queue interrupts
3029  * @pf: board private structure
3030  */
3031 static void ice_ena_misc_vector(struct ice_pf *pf)
3032 {
3033 	struct ice_hw *hw = &pf->hw;
3034 	u32 val;
3035 
3036 	/* Disable anti-spoof detection interrupt to prevent spurious event
3037 	 * interrupts during a function reset. Anti-spoof functionally is
3038 	 * still supported.
3039 	 */
3040 	val = rd32(hw, GL_MDCK_TX_TDPU);
3041 	val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
3042 	wr32(hw, GL_MDCK_TX_TDPU, val);
3043 
3044 	/* clear things first */
3045 	wr32(hw, PFINT_OICR_ENA, 0);	/* disable all */
3046 	rd32(hw, PFINT_OICR);		/* read to clear */
3047 
3048 	val = (PFINT_OICR_ECC_ERR_M |
3049 	       PFINT_OICR_MAL_DETECT_M |
3050 	       PFINT_OICR_GRST_M |
3051 	       PFINT_OICR_PCI_EXCEPTION_M |
3052 	       PFINT_OICR_VFLR_M |
3053 	       PFINT_OICR_HMC_ERR_M |
3054 	       PFINT_OICR_PE_PUSH_M |
3055 	       PFINT_OICR_PE_CRITERR_M);
3056 
3057 	wr32(hw, PFINT_OICR_ENA, val);
3058 
3059 	/* SW_ITR_IDX = 0, but don't change INTENA */
3060 	wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
3061 	     GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
3062 }
3063 
3064 /**
3065  * ice_misc_intr - misc interrupt handler
3066  * @irq: interrupt number
3067  * @data: pointer to a q_vector
3068  */
3069 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
3070 {
3071 	struct ice_pf *pf = (struct ice_pf *)data;
3072 	struct ice_hw *hw = &pf->hw;
3073 	struct device *dev;
3074 	u32 oicr, ena_mask;
3075 
3076 	dev = ice_pf_to_dev(pf);
3077 	set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
3078 	set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
3079 	set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
3080 
3081 	oicr = rd32(hw, PFINT_OICR);
3082 	ena_mask = rd32(hw, PFINT_OICR_ENA);
3083 
3084 	if (oicr & PFINT_OICR_SWINT_M) {
3085 		ena_mask &= ~PFINT_OICR_SWINT_M;
3086 		pf->sw_int_count++;
3087 	}
3088 
3089 	if (oicr & PFINT_OICR_MAL_DETECT_M) {
3090 		ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
3091 		set_bit(ICE_MDD_EVENT_PENDING, pf->state);
3092 	}
3093 	if (oicr & PFINT_OICR_VFLR_M) {
3094 		/* disable any further VFLR event notifications */
3095 		if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
3096 			u32 reg = rd32(hw, PFINT_OICR_ENA);
3097 
3098 			reg &= ~PFINT_OICR_VFLR_M;
3099 			wr32(hw, PFINT_OICR_ENA, reg);
3100 		} else {
3101 			ena_mask &= ~PFINT_OICR_VFLR_M;
3102 			set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
3103 		}
3104 	}
3105 
3106 	if (oicr & PFINT_OICR_GRST_M) {
3107 		u32 reset;
3108 
3109 		/* we have a reset warning */
3110 		ena_mask &= ~PFINT_OICR_GRST_M;
3111 		reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
3112 			GLGEN_RSTAT_RESET_TYPE_S;
3113 
3114 		if (reset == ICE_RESET_CORER)
3115 			pf->corer_count++;
3116 		else if (reset == ICE_RESET_GLOBR)
3117 			pf->globr_count++;
3118 		else if (reset == ICE_RESET_EMPR)
3119 			pf->empr_count++;
3120 		else
3121 			dev_dbg(dev, "Invalid reset type %d\n", reset);
3122 
3123 		/* If a reset cycle isn't already in progress, we set a bit in
3124 		 * pf->state so that the service task can start a reset/rebuild.
3125 		 */
3126 		if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3127 			if (reset == ICE_RESET_CORER)
3128 				set_bit(ICE_CORER_RECV, pf->state);
3129 			else if (reset == ICE_RESET_GLOBR)
3130 				set_bit(ICE_GLOBR_RECV, pf->state);
3131 			else
3132 				set_bit(ICE_EMPR_RECV, pf->state);
3133 
3134 			/* There are couple of different bits at play here.
3135 			 * hw->reset_ongoing indicates whether the hardware is
3136 			 * in reset. This is set to true when a reset interrupt
3137 			 * is received and set back to false after the driver
3138 			 * has determined that the hardware is out of reset.
3139 			 *
3140 			 * ICE_RESET_OICR_RECV in pf->state indicates
3141 			 * that a post reset rebuild is required before the
3142 			 * driver is operational again. This is set above.
3143 			 *
3144 			 * As this is the start of the reset/rebuild cycle, set
3145 			 * both to indicate that.
3146 			 */
3147 			hw->reset_ongoing = true;
3148 		}
3149 	}
3150 
3151 	if (oicr & PFINT_OICR_TSYN_TX_M) {
3152 		ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3153 		if (!hw->reset_ongoing && ice_ptp_pf_handles_tx_interrupt(pf))
3154 			set_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread);
3155 	}
3156 
3157 	if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3158 		u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3159 		u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3160 
3161 		ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3162 
3163 		if (ice_pf_src_tmr_owned(pf)) {
3164 			/* Save EVENTs from GLTSYN register */
3165 			pf->ptp.ext_ts_irq |= gltsyn_stat &
3166 					      (GLTSYN_STAT_EVENT0_M |
3167 					       GLTSYN_STAT_EVENT1_M |
3168 					       GLTSYN_STAT_EVENT2_M);
3169 
3170 			set_bit(ICE_MISC_THREAD_EXTTS_EVENT, pf->misc_thread);
3171 		}
3172 	}
3173 
3174 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3175 	if (oicr & ICE_AUX_CRIT_ERR) {
3176 		pf->oicr_err_reg |= oicr;
3177 		set_bit(ICE_AUX_ERR_PENDING, pf->state);
3178 		ena_mask &= ~ICE_AUX_CRIT_ERR;
3179 	}
3180 
3181 	/* Report any remaining unexpected interrupts */
3182 	oicr &= ena_mask;
3183 	if (oicr) {
3184 		dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3185 		/* If a critical error is pending there is no choice but to
3186 		 * reset the device.
3187 		 */
3188 		if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3189 			    PFINT_OICR_ECC_ERR_M)) {
3190 			set_bit(ICE_PFR_REQ, pf->state);
3191 		}
3192 	}
3193 
3194 	return IRQ_WAKE_THREAD;
3195 }
3196 
3197 /**
3198  * ice_misc_intr_thread_fn - misc interrupt thread function
3199  * @irq: interrupt number
3200  * @data: pointer to a q_vector
3201  */
3202 static irqreturn_t ice_misc_intr_thread_fn(int __always_unused irq, void *data)
3203 {
3204 	struct ice_pf *pf = data;
3205 	struct ice_hw *hw;
3206 
3207 	hw = &pf->hw;
3208 
3209 	if (ice_is_reset_in_progress(pf->state))
3210 		return IRQ_HANDLED;
3211 
3212 	ice_service_task_schedule(pf);
3213 
3214 	if (test_and_clear_bit(ICE_MISC_THREAD_EXTTS_EVENT, pf->misc_thread))
3215 		ice_ptp_extts_event(pf);
3216 
3217 	if (test_and_clear_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread)) {
3218 		/* Process outstanding Tx timestamps. If there is more work,
3219 		 * re-arm the interrupt to trigger again.
3220 		 */
3221 		if (ice_ptp_process_ts(pf) == ICE_TX_TSTAMP_WORK_PENDING) {
3222 			wr32(hw, PFINT_OICR, PFINT_OICR_TSYN_TX_M);
3223 			ice_flush(hw);
3224 		}
3225 	}
3226 
3227 	ice_irq_dynamic_ena(hw, NULL, NULL);
3228 
3229 	return IRQ_HANDLED;
3230 }
3231 
3232 /**
3233  * ice_dis_ctrlq_interrupts - disable control queue interrupts
3234  * @hw: pointer to HW structure
3235  */
3236 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3237 {
3238 	/* disable Admin queue Interrupt causes */
3239 	wr32(hw, PFINT_FW_CTL,
3240 	     rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3241 
3242 	/* disable Mailbox queue Interrupt causes */
3243 	wr32(hw, PFINT_MBX_CTL,
3244 	     rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3245 
3246 	wr32(hw, PFINT_SB_CTL,
3247 	     rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3248 
3249 	/* disable Control queue Interrupt causes */
3250 	wr32(hw, PFINT_OICR_CTL,
3251 	     rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3252 
3253 	ice_flush(hw);
3254 }
3255 
3256 /**
3257  * ice_free_irq_msix_misc - Unroll misc vector setup
3258  * @pf: board private structure
3259  */
3260 static void ice_free_irq_msix_misc(struct ice_pf *pf)
3261 {
3262 	int misc_irq_num = pf->oicr_irq.virq;
3263 	struct ice_hw *hw = &pf->hw;
3264 
3265 	ice_dis_ctrlq_interrupts(hw);
3266 
3267 	/* disable OICR interrupt */
3268 	wr32(hw, PFINT_OICR_ENA, 0);
3269 	ice_flush(hw);
3270 
3271 	synchronize_irq(misc_irq_num);
3272 	devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf);
3273 
3274 	ice_free_irq(pf, pf->oicr_irq);
3275 }
3276 
3277 /**
3278  * ice_ena_ctrlq_interrupts - enable control queue interrupts
3279  * @hw: pointer to HW structure
3280  * @reg_idx: HW vector index to associate the control queue interrupts with
3281  */
3282 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3283 {
3284 	u32 val;
3285 
3286 	val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3287 	       PFINT_OICR_CTL_CAUSE_ENA_M);
3288 	wr32(hw, PFINT_OICR_CTL, val);
3289 
3290 	/* enable Admin queue Interrupt causes */
3291 	val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3292 	       PFINT_FW_CTL_CAUSE_ENA_M);
3293 	wr32(hw, PFINT_FW_CTL, val);
3294 
3295 	/* enable Mailbox queue Interrupt causes */
3296 	val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3297 	       PFINT_MBX_CTL_CAUSE_ENA_M);
3298 	wr32(hw, PFINT_MBX_CTL, val);
3299 
3300 	/* This enables Sideband queue Interrupt causes */
3301 	val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3302 	       PFINT_SB_CTL_CAUSE_ENA_M);
3303 	wr32(hw, PFINT_SB_CTL, val);
3304 
3305 	ice_flush(hw);
3306 }
3307 
3308 /**
3309  * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3310  * @pf: board private structure
3311  *
3312  * This sets up the handler for MSIX 0, which is used to manage the
3313  * non-queue interrupts, e.g. AdminQ and errors. This is not used
3314  * when in MSI or Legacy interrupt mode.
3315  */
3316 static int ice_req_irq_msix_misc(struct ice_pf *pf)
3317 {
3318 	struct device *dev = ice_pf_to_dev(pf);
3319 	struct ice_hw *hw = &pf->hw;
3320 	struct msi_map oicr_irq;
3321 	int err = 0;
3322 
3323 	if (!pf->int_name[0])
3324 		snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3325 			 dev_driver_string(dev), dev_name(dev));
3326 
3327 	/* Do not request IRQ but do enable OICR interrupt since settings are
3328 	 * lost during reset. Note that this function is called only during
3329 	 * rebuild path and not while reset is in progress.
3330 	 */
3331 	if (ice_is_reset_in_progress(pf->state))
3332 		goto skip_req_irq;
3333 
3334 	/* reserve one vector in irq_tracker for misc interrupts */
3335 	oicr_irq = ice_alloc_irq(pf, false);
3336 	if (oicr_irq.index < 0)
3337 		return oicr_irq.index;
3338 
3339 	pf->oicr_irq = oicr_irq;
3340 	err = devm_request_threaded_irq(dev, pf->oicr_irq.virq, ice_misc_intr,
3341 					ice_misc_intr_thread_fn, 0,
3342 					pf->int_name, pf);
3343 	if (err) {
3344 		dev_err(dev, "devm_request_threaded_irq for %s failed: %d\n",
3345 			pf->int_name, err);
3346 		ice_free_irq(pf, pf->oicr_irq);
3347 		return err;
3348 	}
3349 
3350 skip_req_irq:
3351 	ice_ena_misc_vector(pf);
3352 
3353 	ice_ena_ctrlq_interrupts(hw, pf->oicr_irq.index);
3354 	wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_irq.index),
3355 	     ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3356 
3357 	ice_flush(hw);
3358 	ice_irq_dynamic_ena(hw, NULL, NULL);
3359 
3360 	return 0;
3361 }
3362 
3363 /**
3364  * ice_napi_add - register NAPI handler for the VSI
3365  * @vsi: VSI for which NAPI handler is to be registered
3366  *
3367  * This function is only called in the driver's load path. Registering the NAPI
3368  * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3369  * reset/rebuild, etc.)
3370  */
3371 static void ice_napi_add(struct ice_vsi *vsi)
3372 {
3373 	int v_idx;
3374 
3375 	if (!vsi->netdev)
3376 		return;
3377 
3378 	ice_for_each_q_vector(vsi, v_idx)
3379 		netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3380 			       ice_napi_poll);
3381 }
3382 
3383 /**
3384  * ice_set_ops - set netdev and ethtools ops for the given netdev
3385  * @vsi: the VSI associated with the new netdev
3386  */
3387 static void ice_set_ops(struct ice_vsi *vsi)
3388 {
3389 	struct net_device *netdev = vsi->netdev;
3390 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
3391 
3392 	if (ice_is_safe_mode(pf)) {
3393 		netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3394 		ice_set_ethtool_safe_mode_ops(netdev);
3395 		return;
3396 	}
3397 
3398 	netdev->netdev_ops = &ice_netdev_ops;
3399 	netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3400 	ice_set_ethtool_ops(netdev);
3401 
3402 	if (vsi->type != ICE_VSI_PF)
3403 		return;
3404 
3405 	netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
3406 			       NETDEV_XDP_ACT_XSK_ZEROCOPY |
3407 			       NETDEV_XDP_ACT_RX_SG;
3408 	netdev->xdp_zc_max_segs = ICE_MAX_BUF_TXD;
3409 }
3410 
3411 /**
3412  * ice_set_netdev_features - set features for the given netdev
3413  * @netdev: netdev instance
3414  */
3415 static void ice_set_netdev_features(struct net_device *netdev)
3416 {
3417 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
3418 	bool is_dvm_ena = ice_is_dvm_ena(&pf->hw);
3419 	netdev_features_t csumo_features;
3420 	netdev_features_t vlano_features;
3421 	netdev_features_t dflt_features;
3422 	netdev_features_t tso_features;
3423 
3424 	if (ice_is_safe_mode(pf)) {
3425 		/* safe mode */
3426 		netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3427 		netdev->hw_features = netdev->features;
3428 		return;
3429 	}
3430 
3431 	dflt_features = NETIF_F_SG	|
3432 			NETIF_F_HIGHDMA	|
3433 			NETIF_F_NTUPLE	|
3434 			NETIF_F_RXHASH;
3435 
3436 	csumo_features = NETIF_F_RXCSUM	  |
3437 			 NETIF_F_IP_CSUM  |
3438 			 NETIF_F_SCTP_CRC |
3439 			 NETIF_F_IPV6_CSUM;
3440 
3441 	vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3442 			 NETIF_F_HW_VLAN_CTAG_TX     |
3443 			 NETIF_F_HW_VLAN_CTAG_RX;
3444 
3445 	/* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */
3446 	if (is_dvm_ena)
3447 		vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER;
3448 
3449 	tso_features = NETIF_F_TSO			|
3450 		       NETIF_F_TSO_ECN			|
3451 		       NETIF_F_TSO6			|
3452 		       NETIF_F_GSO_GRE			|
3453 		       NETIF_F_GSO_UDP_TUNNEL		|
3454 		       NETIF_F_GSO_GRE_CSUM		|
3455 		       NETIF_F_GSO_UDP_TUNNEL_CSUM	|
3456 		       NETIF_F_GSO_PARTIAL		|
3457 		       NETIF_F_GSO_IPXIP4		|
3458 		       NETIF_F_GSO_IPXIP6		|
3459 		       NETIF_F_GSO_UDP_L4;
3460 
3461 	netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3462 					NETIF_F_GSO_GRE_CSUM;
3463 	/* set features that user can change */
3464 	netdev->hw_features = dflt_features | csumo_features |
3465 			      vlano_features | tso_features;
3466 
3467 	/* add support for HW_CSUM on packets with MPLS header */
3468 	netdev->mpls_features =  NETIF_F_HW_CSUM |
3469 				 NETIF_F_TSO     |
3470 				 NETIF_F_TSO6;
3471 
3472 	/* enable features */
3473 	netdev->features |= netdev->hw_features;
3474 
3475 	netdev->hw_features |= NETIF_F_HW_TC;
3476 	netdev->hw_features |= NETIF_F_LOOPBACK;
3477 
3478 	/* encap and VLAN devices inherit default, csumo and tso features */
3479 	netdev->hw_enc_features |= dflt_features | csumo_features |
3480 				   tso_features;
3481 	netdev->vlan_features |= dflt_features | csumo_features |
3482 				 tso_features;
3483 
3484 	/* advertise support but don't enable by default since only one type of
3485 	 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one
3486 	 * type turns on the other has to be turned off. This is enforced by the
3487 	 * ice_fix_features() ndo callback.
3488 	 */
3489 	if (is_dvm_ena)
3490 		netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX |
3491 			NETIF_F_HW_VLAN_STAG_TX;
3492 
3493 	/* Leave CRC / FCS stripping enabled by default, but allow the value to
3494 	 * be changed at runtime
3495 	 */
3496 	netdev->hw_features |= NETIF_F_RXFCS;
3497 
3498 	netif_set_tso_max_size(netdev, ICE_MAX_TSO_SIZE);
3499 }
3500 
3501 /**
3502  * ice_fill_rss_lut - Fill the RSS lookup table with default values
3503  * @lut: Lookup table
3504  * @rss_table_size: Lookup table size
3505  * @rss_size: Range of queue number for hashing
3506  */
3507 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3508 {
3509 	u16 i;
3510 
3511 	for (i = 0; i < rss_table_size; i++)
3512 		lut[i] = i % rss_size;
3513 }
3514 
3515 /**
3516  * ice_pf_vsi_setup - Set up a PF VSI
3517  * @pf: board private structure
3518  * @pi: pointer to the port_info instance
3519  *
3520  * Returns pointer to the successfully allocated VSI software struct
3521  * on success, otherwise returns NULL on failure.
3522  */
3523 static struct ice_vsi *
3524 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3525 {
3526 	struct ice_vsi_cfg_params params = {};
3527 
3528 	params.type = ICE_VSI_PF;
3529 	params.pi = pi;
3530 	params.flags = ICE_VSI_FLAG_INIT;
3531 
3532 	return ice_vsi_setup(pf, &params);
3533 }
3534 
3535 static struct ice_vsi *
3536 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3537 		   struct ice_channel *ch)
3538 {
3539 	struct ice_vsi_cfg_params params = {};
3540 
3541 	params.type = ICE_VSI_CHNL;
3542 	params.pi = pi;
3543 	params.ch = ch;
3544 	params.flags = ICE_VSI_FLAG_INIT;
3545 
3546 	return ice_vsi_setup(pf, &params);
3547 }
3548 
3549 /**
3550  * ice_ctrl_vsi_setup - Set up a control VSI
3551  * @pf: board private structure
3552  * @pi: pointer to the port_info instance
3553  *
3554  * Returns pointer to the successfully allocated VSI software struct
3555  * on success, otherwise returns NULL on failure.
3556  */
3557 static struct ice_vsi *
3558 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3559 {
3560 	struct ice_vsi_cfg_params params = {};
3561 
3562 	params.type = ICE_VSI_CTRL;
3563 	params.pi = pi;
3564 	params.flags = ICE_VSI_FLAG_INIT;
3565 
3566 	return ice_vsi_setup(pf, &params);
3567 }
3568 
3569 /**
3570  * ice_lb_vsi_setup - Set up a loopback VSI
3571  * @pf: board private structure
3572  * @pi: pointer to the port_info instance
3573  *
3574  * Returns pointer to the successfully allocated VSI software struct
3575  * on success, otherwise returns NULL on failure.
3576  */
3577 struct ice_vsi *
3578 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3579 {
3580 	struct ice_vsi_cfg_params params = {};
3581 
3582 	params.type = ICE_VSI_LB;
3583 	params.pi = pi;
3584 	params.flags = ICE_VSI_FLAG_INIT;
3585 
3586 	return ice_vsi_setup(pf, &params);
3587 }
3588 
3589 /**
3590  * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3591  * @netdev: network interface to be adjusted
3592  * @proto: VLAN TPID
3593  * @vid: VLAN ID to be added
3594  *
3595  * net_device_ops implementation for adding VLAN IDs
3596  */
3597 static int
3598 ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
3599 {
3600 	struct ice_netdev_priv *np = netdev_priv(netdev);
3601 	struct ice_vsi_vlan_ops *vlan_ops;
3602 	struct ice_vsi *vsi = np->vsi;
3603 	struct ice_vlan vlan;
3604 	int ret;
3605 
3606 	/* VLAN 0 is added by default during load/reset */
3607 	if (!vid)
3608 		return 0;
3609 
3610 	while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3611 		usleep_range(1000, 2000);
3612 
3613 	/* Add multicast promisc rule for the VLAN ID to be added if
3614 	 * all-multicast is currently enabled.
3615 	 */
3616 	if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3617 		ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3618 					       ICE_MCAST_VLAN_PROMISC_BITS,
3619 					       vid);
3620 		if (ret)
3621 			goto finish;
3622 	}
3623 
3624 	vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3625 
3626 	/* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3627 	 * packets aren't pruned by the device's internal switch on Rx
3628 	 */
3629 	vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3630 	ret = vlan_ops->add_vlan(vsi, &vlan);
3631 	if (ret)
3632 		goto finish;
3633 
3634 	/* If all-multicast is currently enabled and this VLAN ID is only one
3635 	 * besides VLAN-0 we have to update look-up type of multicast promisc
3636 	 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN.
3637 	 */
3638 	if ((vsi->current_netdev_flags & IFF_ALLMULTI) &&
3639 	    ice_vsi_num_non_zero_vlans(vsi) == 1) {
3640 		ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3641 					   ICE_MCAST_PROMISC_BITS, 0);
3642 		ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3643 					 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3644 	}
3645 
3646 finish:
3647 	clear_bit(ICE_CFG_BUSY, vsi->state);
3648 
3649 	return ret;
3650 }
3651 
3652 /**
3653  * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3654  * @netdev: network interface to be adjusted
3655  * @proto: VLAN TPID
3656  * @vid: VLAN ID to be removed
3657  *
3658  * net_device_ops implementation for removing VLAN IDs
3659  */
3660 static int
3661 ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
3662 {
3663 	struct ice_netdev_priv *np = netdev_priv(netdev);
3664 	struct ice_vsi_vlan_ops *vlan_ops;
3665 	struct ice_vsi *vsi = np->vsi;
3666 	struct ice_vlan vlan;
3667 	int ret;
3668 
3669 	/* don't allow removal of VLAN 0 */
3670 	if (!vid)
3671 		return 0;
3672 
3673 	while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3674 		usleep_range(1000, 2000);
3675 
3676 	ret = ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3677 				    ICE_MCAST_VLAN_PROMISC_BITS, vid);
3678 	if (ret) {
3679 		netdev_err(netdev, "Error clearing multicast promiscuous mode on VSI %i\n",
3680 			   vsi->vsi_num);
3681 		vsi->current_netdev_flags |= IFF_ALLMULTI;
3682 	}
3683 
3684 	vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3685 
3686 	/* Make sure VLAN delete is successful before updating VLAN
3687 	 * information
3688 	 */
3689 	vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3690 	ret = vlan_ops->del_vlan(vsi, &vlan);
3691 	if (ret)
3692 		goto finish;
3693 
3694 	/* Remove multicast promisc rule for the removed VLAN ID if
3695 	 * all-multicast is enabled.
3696 	 */
3697 	if (vsi->current_netdev_flags & IFF_ALLMULTI)
3698 		ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3699 					   ICE_MCAST_VLAN_PROMISC_BITS, vid);
3700 
3701 	if (!ice_vsi_has_non_zero_vlans(vsi)) {
3702 		/* Update look-up type of multicast promisc rule for VLAN 0
3703 		 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when
3704 		 * all-multicast is enabled and VLAN 0 is the only VLAN rule.
3705 		 */
3706 		if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3707 			ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3708 						   ICE_MCAST_VLAN_PROMISC_BITS,
3709 						   0);
3710 			ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3711 						 ICE_MCAST_PROMISC_BITS, 0);
3712 		}
3713 	}
3714 
3715 finish:
3716 	clear_bit(ICE_CFG_BUSY, vsi->state);
3717 
3718 	return ret;
3719 }
3720 
3721 /**
3722  * ice_rep_indr_tc_block_unbind
3723  * @cb_priv: indirection block private data
3724  */
3725 static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3726 {
3727 	struct ice_indr_block_priv *indr_priv = cb_priv;
3728 
3729 	list_del(&indr_priv->list);
3730 	kfree(indr_priv);
3731 }
3732 
3733 /**
3734  * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3735  * @vsi: VSI struct which has the netdev
3736  */
3737 static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3738 {
3739 	struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3740 
3741 	flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3742 				 ice_rep_indr_tc_block_unbind);
3743 }
3744 
3745 /**
3746  * ice_tc_indir_block_register - Register TC indirect block notifications
3747  * @vsi: VSI struct which has the netdev
3748  *
3749  * Returns 0 on success, negative value on failure
3750  */
3751 static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3752 {
3753 	struct ice_netdev_priv *np;
3754 
3755 	if (!vsi || !vsi->netdev)
3756 		return -EINVAL;
3757 
3758 	np = netdev_priv(vsi->netdev);
3759 
3760 	INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3761 	return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3762 }
3763 
3764 /**
3765  * ice_get_avail_q_count - Get count of queues in use
3766  * @pf_qmap: bitmap to get queue use count from
3767  * @lock: pointer to a mutex that protects access to pf_qmap
3768  * @size: size of the bitmap
3769  */
3770 static u16
3771 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3772 {
3773 	unsigned long bit;
3774 	u16 count = 0;
3775 
3776 	mutex_lock(lock);
3777 	for_each_clear_bit(bit, pf_qmap, size)
3778 		count++;
3779 	mutex_unlock(lock);
3780 
3781 	return count;
3782 }
3783 
3784 /**
3785  * ice_get_avail_txq_count - Get count of Tx queues in use
3786  * @pf: pointer to an ice_pf instance
3787  */
3788 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3789 {
3790 	return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3791 				     pf->max_pf_txqs);
3792 }
3793 
3794 /**
3795  * ice_get_avail_rxq_count - Get count of Rx queues in use
3796  * @pf: pointer to an ice_pf instance
3797  */
3798 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3799 {
3800 	return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3801 				     pf->max_pf_rxqs);
3802 }
3803 
3804 /**
3805  * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3806  * @pf: board private structure to initialize
3807  */
3808 static void ice_deinit_pf(struct ice_pf *pf)
3809 {
3810 	ice_service_task_stop(pf);
3811 	mutex_destroy(&pf->lag_mutex);
3812 	mutex_destroy(&pf->adev_mutex);
3813 	mutex_destroy(&pf->sw_mutex);
3814 	mutex_destroy(&pf->tc_mutex);
3815 	mutex_destroy(&pf->avail_q_mutex);
3816 	mutex_destroy(&pf->vfs.table_lock);
3817 
3818 	if (pf->avail_txqs) {
3819 		bitmap_free(pf->avail_txqs);
3820 		pf->avail_txqs = NULL;
3821 	}
3822 
3823 	if (pf->avail_rxqs) {
3824 		bitmap_free(pf->avail_rxqs);
3825 		pf->avail_rxqs = NULL;
3826 	}
3827 
3828 	if (pf->ptp.clock)
3829 		ptp_clock_unregister(pf->ptp.clock);
3830 }
3831 
3832 /**
3833  * ice_set_pf_caps - set PFs capability flags
3834  * @pf: pointer to the PF instance
3835  */
3836 static void ice_set_pf_caps(struct ice_pf *pf)
3837 {
3838 	struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3839 
3840 	clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3841 	if (func_caps->common_cap.rdma)
3842 		set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3843 	clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3844 	if (func_caps->common_cap.dcb)
3845 		set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3846 	clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3847 	if (func_caps->common_cap.sr_iov_1_1) {
3848 		set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3849 		pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
3850 					      ICE_MAX_SRIOV_VFS);
3851 	}
3852 	clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3853 	if (func_caps->common_cap.rss_table_size)
3854 		set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3855 
3856 	clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3857 	if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3858 		u16 unused;
3859 
3860 		/* ctrl_vsi_idx will be set to a valid value when flow director
3861 		 * is setup by ice_init_fdir
3862 		 */
3863 		pf->ctrl_vsi_idx = ICE_NO_VSI;
3864 		set_bit(ICE_FLAG_FD_ENA, pf->flags);
3865 		/* force guaranteed filter pool for PF */
3866 		ice_alloc_fd_guar_item(&pf->hw, &unused,
3867 				       func_caps->fd_fltr_guar);
3868 		/* force shared filter pool for PF */
3869 		ice_alloc_fd_shrd_item(&pf->hw, &unused,
3870 				       func_caps->fd_fltr_best_effort);
3871 	}
3872 
3873 	clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3874 	if (func_caps->common_cap.ieee_1588 &&
3875 	    !(pf->hw.mac_type == ICE_MAC_E830))
3876 		set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3877 
3878 	pf->max_pf_txqs = func_caps->common_cap.num_txq;
3879 	pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3880 }
3881 
3882 /**
3883  * ice_init_pf - Initialize general software structures (struct ice_pf)
3884  * @pf: board private structure to initialize
3885  */
3886 static int ice_init_pf(struct ice_pf *pf)
3887 {
3888 	ice_set_pf_caps(pf);
3889 
3890 	mutex_init(&pf->sw_mutex);
3891 	mutex_init(&pf->tc_mutex);
3892 	mutex_init(&pf->adev_mutex);
3893 	mutex_init(&pf->lag_mutex);
3894 
3895 	INIT_HLIST_HEAD(&pf->aq_wait_list);
3896 	spin_lock_init(&pf->aq_wait_lock);
3897 	init_waitqueue_head(&pf->aq_wait_queue);
3898 
3899 	init_waitqueue_head(&pf->reset_wait_queue);
3900 
3901 	/* setup service timer and periodic service task */
3902 	timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3903 	pf->serv_tmr_period = HZ;
3904 	INIT_WORK(&pf->serv_task, ice_service_task);
3905 	clear_bit(ICE_SERVICE_SCHED, pf->state);
3906 
3907 	mutex_init(&pf->avail_q_mutex);
3908 	pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3909 	if (!pf->avail_txqs)
3910 		return -ENOMEM;
3911 
3912 	pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3913 	if (!pf->avail_rxqs) {
3914 		bitmap_free(pf->avail_txqs);
3915 		pf->avail_txqs = NULL;
3916 		return -ENOMEM;
3917 	}
3918 
3919 	mutex_init(&pf->vfs.table_lock);
3920 	hash_init(pf->vfs.table);
3921 	ice_mbx_init_snapshot(&pf->hw);
3922 
3923 	return 0;
3924 }
3925 
3926 /**
3927  * ice_is_wol_supported - check if WoL is supported
3928  * @hw: pointer to hardware info
3929  *
3930  * Check if WoL is supported based on the HW configuration.
3931  * Returns true if NVM supports and enables WoL for this port, false otherwise
3932  */
3933 bool ice_is_wol_supported(struct ice_hw *hw)
3934 {
3935 	u16 wol_ctrl;
3936 
3937 	/* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3938 	 * word) indicates WoL is not supported on the corresponding PF ID.
3939 	 */
3940 	if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3941 		return false;
3942 
3943 	return !(BIT(hw->port_info->lport) & wol_ctrl);
3944 }
3945 
3946 /**
3947  * ice_vsi_recfg_qs - Change the number of queues on a VSI
3948  * @vsi: VSI being changed
3949  * @new_rx: new number of Rx queues
3950  * @new_tx: new number of Tx queues
3951  * @locked: is adev device_lock held
3952  *
3953  * Only change the number of queues if new_tx, or new_rx is non-0.
3954  *
3955  * Returns 0 on success.
3956  */
3957 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx, bool locked)
3958 {
3959 	struct ice_pf *pf = vsi->back;
3960 	int err = 0, timeout = 50;
3961 
3962 	if (!new_rx && !new_tx)
3963 		return -EINVAL;
3964 
3965 	while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
3966 		timeout--;
3967 		if (!timeout)
3968 			return -EBUSY;
3969 		usleep_range(1000, 2000);
3970 	}
3971 
3972 	if (new_tx)
3973 		vsi->req_txq = (u16)new_tx;
3974 	if (new_rx)
3975 		vsi->req_rxq = (u16)new_rx;
3976 
3977 	/* set for the next time the netdev is started */
3978 	if (!netif_running(vsi->netdev)) {
3979 		ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
3980 		dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3981 		goto done;
3982 	}
3983 
3984 	ice_vsi_close(vsi);
3985 	ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
3986 	ice_pf_dcb_recfg(pf, locked);
3987 	ice_vsi_open(vsi);
3988 done:
3989 	clear_bit(ICE_CFG_BUSY, pf->state);
3990 	return err;
3991 }
3992 
3993 /**
3994  * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3995  * @pf: PF to configure
3996  *
3997  * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3998  * VSI can still Tx/Rx VLAN tagged packets.
3999  */
4000 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4001 {
4002 	struct ice_vsi *vsi = ice_get_main_vsi(pf);
4003 	struct ice_vsi_ctx *ctxt;
4004 	struct ice_hw *hw;
4005 	int status;
4006 
4007 	if (!vsi)
4008 		return;
4009 
4010 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4011 	if (!ctxt)
4012 		return;
4013 
4014 	hw = &pf->hw;
4015 	ctxt->info = vsi->info;
4016 
4017 	ctxt->info.valid_sections =
4018 		cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4019 			    ICE_AQ_VSI_PROP_SECURITY_VALID |
4020 			    ICE_AQ_VSI_PROP_SW_VALID);
4021 
4022 	/* disable VLAN anti-spoof */
4023 	ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4024 				  ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4025 
4026 	/* disable VLAN pruning and keep all other settings */
4027 	ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4028 
4029 	/* allow all VLANs on Tx and don't strip on Rx */
4030 	ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL |
4031 		ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
4032 
4033 	status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4034 	if (status) {
4035 		dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4036 			status, ice_aq_str(hw->adminq.sq_last_status));
4037 	} else {
4038 		vsi->info.sec_flags = ctxt->info.sec_flags;
4039 		vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4040 		vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags;
4041 	}
4042 
4043 	kfree(ctxt);
4044 }
4045 
4046 /**
4047  * ice_log_pkg_init - log result of DDP package load
4048  * @hw: pointer to hardware info
4049  * @state: state of package load
4050  */
4051 static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4052 {
4053 	struct ice_pf *pf = hw->back;
4054 	struct device *dev;
4055 
4056 	dev = ice_pf_to_dev(pf);
4057 
4058 	switch (state) {
4059 	case ICE_DDP_PKG_SUCCESS:
4060 		dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4061 			 hw->active_pkg_name,
4062 			 hw->active_pkg_ver.major,
4063 			 hw->active_pkg_ver.minor,
4064 			 hw->active_pkg_ver.update,
4065 			 hw->active_pkg_ver.draft);
4066 		break;
4067 	case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4068 		dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4069 			 hw->active_pkg_name,
4070 			 hw->active_pkg_ver.major,
4071 			 hw->active_pkg_ver.minor,
4072 			 hw->active_pkg_ver.update,
4073 			 hw->active_pkg_ver.draft);
4074 		break;
4075 	case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4076 		dev_err(dev, "The device has a DDP package that is not supported by the driver.  The device has package '%s' version %d.%d.x.x.  The driver requires version %d.%d.x.x.  Entering Safe Mode.\n",
4077 			hw->active_pkg_name,
4078 			hw->active_pkg_ver.major,
4079 			hw->active_pkg_ver.minor,
4080 			ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4081 		break;
4082 	case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4083 		dev_info(dev, "The driver could not load the DDP package file because a compatible DDP package is already present on the device.  The device has package '%s' version %d.%d.%d.%d.  The package file found by the driver: '%s' version %d.%d.%d.%d.\n",
4084 			 hw->active_pkg_name,
4085 			 hw->active_pkg_ver.major,
4086 			 hw->active_pkg_ver.minor,
4087 			 hw->active_pkg_ver.update,
4088 			 hw->active_pkg_ver.draft,
4089 			 hw->pkg_name,
4090 			 hw->pkg_ver.major,
4091 			 hw->pkg_ver.minor,
4092 			 hw->pkg_ver.update,
4093 			 hw->pkg_ver.draft);
4094 		break;
4095 	case ICE_DDP_PKG_FW_MISMATCH:
4096 		dev_err(dev, "The firmware loaded on the device is not compatible with the DDP package.  Please update the device's NVM.  Entering safe mode.\n");
4097 		break;
4098 	case ICE_DDP_PKG_INVALID_FILE:
4099 		dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4100 		break;
4101 	case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4102 		dev_err(dev, "The DDP package file version is higher than the driver supports.  Please use an updated driver.  Entering Safe Mode.\n");
4103 		break;
4104 	case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4105 		dev_err(dev, "The DDP package file version is lower than the driver supports.  The driver requires version %d.%d.x.x.  Please use an updated DDP Package file.  Entering Safe Mode.\n",
4106 			ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4107 		break;
4108 	case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4109 		dev_err(dev, "The DDP package could not be loaded because its signature is not valid.  Please use a valid DDP Package.  Entering Safe Mode.\n");
4110 		break;
4111 	case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4112 		dev_err(dev, "The DDP Package could not be loaded because its security revision is too low.  Please use an updated DDP Package.  Entering Safe Mode.\n");
4113 		break;
4114 	case ICE_DDP_PKG_LOAD_ERROR:
4115 		dev_err(dev, "An error occurred on the device while loading the DDP package.  The device will be reset.\n");
4116 		/* poll for reset to complete */
4117 		if (ice_check_reset(hw))
4118 			dev_err(dev, "Error resetting device. Please reload the driver\n");
4119 		break;
4120 	case ICE_DDP_PKG_ERR:
4121 	default:
4122 		dev_err(dev, "An unknown error occurred when loading the DDP package.  Entering Safe Mode.\n");
4123 		break;
4124 	}
4125 }
4126 
4127 /**
4128  * ice_load_pkg - load/reload the DDP Package file
4129  * @firmware: firmware structure when firmware requested or NULL for reload
4130  * @pf: pointer to the PF instance
4131  *
4132  * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4133  * initialize HW tables.
4134  */
4135 static void
4136 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4137 {
4138 	enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4139 	struct device *dev = ice_pf_to_dev(pf);
4140 	struct ice_hw *hw = &pf->hw;
4141 
4142 	/* Load DDP Package */
4143 	if (firmware && !hw->pkg_copy) {
4144 		state = ice_copy_and_init_pkg(hw, firmware->data,
4145 					      firmware->size);
4146 		ice_log_pkg_init(hw, state);
4147 	} else if (!firmware && hw->pkg_copy) {
4148 		/* Reload package during rebuild after CORER/GLOBR reset */
4149 		state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4150 		ice_log_pkg_init(hw, state);
4151 	} else {
4152 		dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4153 	}
4154 
4155 	if (!ice_is_init_pkg_successful(state)) {
4156 		/* Safe Mode */
4157 		clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4158 		return;
4159 	}
4160 
4161 	/* Successful download package is the precondition for advanced
4162 	 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4163 	 */
4164 	set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4165 }
4166 
4167 /**
4168  * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4169  * @pf: pointer to the PF structure
4170  *
4171  * There is no error returned here because the driver should be able to handle
4172  * 128 Byte cache lines, so we only print a warning in case issues are seen,
4173  * specifically with Tx.
4174  */
4175 static void ice_verify_cacheline_size(struct ice_pf *pf)
4176 {
4177 	if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4178 		dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4179 			 ICE_CACHE_LINE_BYTES);
4180 }
4181 
4182 /**
4183  * ice_send_version - update firmware with driver version
4184  * @pf: PF struct
4185  *
4186  * Returns 0 on success, else error code
4187  */
4188 static int ice_send_version(struct ice_pf *pf)
4189 {
4190 	struct ice_driver_ver dv;
4191 
4192 	dv.major_ver = 0xff;
4193 	dv.minor_ver = 0xff;
4194 	dv.build_ver = 0xff;
4195 	dv.subbuild_ver = 0;
4196 	strscpy((char *)dv.driver_string, UTS_RELEASE,
4197 		sizeof(dv.driver_string));
4198 	return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4199 }
4200 
4201 /**
4202  * ice_init_fdir - Initialize flow director VSI and configuration
4203  * @pf: pointer to the PF instance
4204  *
4205  * returns 0 on success, negative on error
4206  */
4207 static int ice_init_fdir(struct ice_pf *pf)
4208 {
4209 	struct device *dev = ice_pf_to_dev(pf);
4210 	struct ice_vsi *ctrl_vsi;
4211 	int err;
4212 
4213 	/* Side Band Flow Director needs to have a control VSI.
4214 	 * Allocate it and store it in the PF.
4215 	 */
4216 	ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4217 	if (!ctrl_vsi) {
4218 		dev_dbg(dev, "could not create control VSI\n");
4219 		return -ENOMEM;
4220 	}
4221 
4222 	err = ice_vsi_open_ctrl(ctrl_vsi);
4223 	if (err) {
4224 		dev_dbg(dev, "could not open control VSI\n");
4225 		goto err_vsi_open;
4226 	}
4227 
4228 	mutex_init(&pf->hw.fdir_fltr_lock);
4229 
4230 	err = ice_fdir_create_dflt_rules(pf);
4231 	if (err)
4232 		goto err_fdir_rule;
4233 
4234 	return 0;
4235 
4236 err_fdir_rule:
4237 	ice_fdir_release_flows(&pf->hw);
4238 	ice_vsi_close(ctrl_vsi);
4239 err_vsi_open:
4240 	ice_vsi_release(ctrl_vsi);
4241 	if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4242 		pf->vsi[pf->ctrl_vsi_idx] = NULL;
4243 		pf->ctrl_vsi_idx = ICE_NO_VSI;
4244 	}
4245 	return err;
4246 }
4247 
4248 static void ice_deinit_fdir(struct ice_pf *pf)
4249 {
4250 	struct ice_vsi *vsi = ice_get_ctrl_vsi(pf);
4251 
4252 	if (!vsi)
4253 		return;
4254 
4255 	ice_vsi_manage_fdir(vsi, false);
4256 	ice_vsi_release(vsi);
4257 	if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4258 		pf->vsi[pf->ctrl_vsi_idx] = NULL;
4259 		pf->ctrl_vsi_idx = ICE_NO_VSI;
4260 	}
4261 
4262 	mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4263 }
4264 
4265 /**
4266  * ice_get_opt_fw_name - return optional firmware file name or NULL
4267  * @pf: pointer to the PF instance
4268  */
4269 static char *ice_get_opt_fw_name(struct ice_pf *pf)
4270 {
4271 	/* Optional firmware name same as default with additional dash
4272 	 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4273 	 */
4274 	struct pci_dev *pdev = pf->pdev;
4275 	char *opt_fw_filename;
4276 	u64 dsn;
4277 
4278 	/* Determine the name of the optional file using the DSN (two
4279 	 * dwords following the start of the DSN Capability).
4280 	 */
4281 	dsn = pci_get_dsn(pdev);
4282 	if (!dsn)
4283 		return NULL;
4284 
4285 	opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4286 	if (!opt_fw_filename)
4287 		return NULL;
4288 
4289 	snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4290 		 ICE_DDP_PKG_PATH, dsn);
4291 
4292 	return opt_fw_filename;
4293 }
4294 
4295 /**
4296  * ice_request_fw - Device initialization routine
4297  * @pf: pointer to the PF instance
4298  */
4299 static void ice_request_fw(struct ice_pf *pf)
4300 {
4301 	char *opt_fw_filename = ice_get_opt_fw_name(pf);
4302 	const struct firmware *firmware = NULL;
4303 	struct device *dev = ice_pf_to_dev(pf);
4304 	int err = 0;
4305 
4306 	/* optional device-specific DDP (if present) overrides the default DDP
4307 	 * package file. kernel logs a debug message if the file doesn't exist,
4308 	 * and warning messages for other errors.
4309 	 */
4310 	if (opt_fw_filename) {
4311 		err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4312 		if (err) {
4313 			kfree(opt_fw_filename);
4314 			goto dflt_pkg_load;
4315 		}
4316 
4317 		/* request for firmware was successful. Download to device */
4318 		ice_load_pkg(firmware, pf);
4319 		kfree(opt_fw_filename);
4320 		release_firmware(firmware);
4321 		return;
4322 	}
4323 
4324 dflt_pkg_load:
4325 	err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4326 	if (err) {
4327 		dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4328 		return;
4329 	}
4330 
4331 	/* request for firmware was successful. Download to device */
4332 	ice_load_pkg(firmware, pf);
4333 	release_firmware(firmware);
4334 }
4335 
4336 /**
4337  * ice_print_wake_reason - show the wake up cause in the log
4338  * @pf: pointer to the PF struct
4339  */
4340 static void ice_print_wake_reason(struct ice_pf *pf)
4341 {
4342 	u32 wus = pf->wakeup_reason;
4343 	const char *wake_str;
4344 
4345 	/* if no wake event, nothing to print */
4346 	if (!wus)
4347 		return;
4348 
4349 	if (wus & PFPM_WUS_LNKC_M)
4350 		wake_str = "Link\n";
4351 	else if (wus & PFPM_WUS_MAG_M)
4352 		wake_str = "Magic Packet\n";
4353 	else if (wus & PFPM_WUS_MNG_M)
4354 		wake_str = "Management\n";
4355 	else if (wus & PFPM_WUS_FW_RST_WK_M)
4356 		wake_str = "Firmware Reset\n";
4357 	else
4358 		wake_str = "Unknown\n";
4359 
4360 	dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4361 }
4362 
4363 /**
4364  * ice_register_netdev - register netdev
4365  * @vsi: pointer to the VSI struct
4366  */
4367 static int ice_register_netdev(struct ice_vsi *vsi)
4368 {
4369 	int err;
4370 
4371 	if (!vsi || !vsi->netdev)
4372 		return -EIO;
4373 
4374 	err = register_netdev(vsi->netdev);
4375 	if (err)
4376 		return err;
4377 
4378 	set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4379 	netif_carrier_off(vsi->netdev);
4380 	netif_tx_stop_all_queues(vsi->netdev);
4381 
4382 	return 0;
4383 }
4384 
4385 static void ice_unregister_netdev(struct ice_vsi *vsi)
4386 {
4387 	if (!vsi || !vsi->netdev)
4388 		return;
4389 
4390 	unregister_netdev(vsi->netdev);
4391 	clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4392 }
4393 
4394 /**
4395  * ice_cfg_netdev - Allocate, configure and register a netdev
4396  * @vsi: the VSI associated with the new netdev
4397  *
4398  * Returns 0 on success, negative value on failure
4399  */
4400 static int ice_cfg_netdev(struct ice_vsi *vsi)
4401 {
4402 	struct ice_netdev_priv *np;
4403 	struct net_device *netdev;
4404 	u8 mac_addr[ETH_ALEN];
4405 
4406 	netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
4407 				    vsi->alloc_rxq);
4408 	if (!netdev)
4409 		return -ENOMEM;
4410 
4411 	set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4412 	vsi->netdev = netdev;
4413 	np = netdev_priv(netdev);
4414 	np->vsi = vsi;
4415 
4416 	ice_set_netdev_features(netdev);
4417 	ice_set_ops(vsi);
4418 
4419 	if (vsi->type == ICE_VSI_PF) {
4420 		SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
4421 		ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4422 		eth_hw_addr_set(netdev, mac_addr);
4423 	}
4424 
4425 	netdev->priv_flags |= IFF_UNICAST_FLT;
4426 
4427 	/* Setup netdev TC information */
4428 	ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
4429 
4430 	netdev->max_mtu = ICE_MAX_MTU;
4431 
4432 	return 0;
4433 }
4434 
4435 static void ice_decfg_netdev(struct ice_vsi *vsi)
4436 {
4437 	clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4438 	free_netdev(vsi->netdev);
4439 	vsi->netdev = NULL;
4440 }
4441 
4442 static int ice_start_eth(struct ice_vsi *vsi)
4443 {
4444 	int err;
4445 
4446 	err = ice_init_mac_fltr(vsi->back);
4447 	if (err)
4448 		return err;
4449 
4450 	err = ice_vsi_open(vsi);
4451 	if (err)
4452 		ice_fltr_remove_all(vsi);
4453 
4454 	return err;
4455 }
4456 
4457 static void ice_stop_eth(struct ice_vsi *vsi)
4458 {
4459 	ice_fltr_remove_all(vsi);
4460 	ice_vsi_close(vsi);
4461 }
4462 
4463 static int ice_init_eth(struct ice_pf *pf)
4464 {
4465 	struct ice_vsi *vsi = ice_get_main_vsi(pf);
4466 	int err;
4467 
4468 	if (!vsi)
4469 		return -EINVAL;
4470 
4471 	/* init channel list */
4472 	INIT_LIST_HEAD(&vsi->ch_list);
4473 
4474 	err = ice_cfg_netdev(vsi);
4475 	if (err)
4476 		return err;
4477 	/* Setup DCB netlink interface */
4478 	ice_dcbnl_setup(vsi);
4479 
4480 	err = ice_init_mac_fltr(pf);
4481 	if (err)
4482 		goto err_init_mac_fltr;
4483 
4484 	err = ice_devlink_create_pf_port(pf);
4485 	if (err)
4486 		goto err_devlink_create_pf_port;
4487 
4488 	SET_NETDEV_DEVLINK_PORT(vsi->netdev, &pf->devlink_port);
4489 
4490 	err = ice_register_netdev(vsi);
4491 	if (err)
4492 		goto err_register_netdev;
4493 
4494 	err = ice_tc_indir_block_register(vsi);
4495 	if (err)
4496 		goto err_tc_indir_block_register;
4497 
4498 	ice_napi_add(vsi);
4499 
4500 	return 0;
4501 
4502 err_tc_indir_block_register:
4503 	ice_unregister_netdev(vsi);
4504 err_register_netdev:
4505 	ice_devlink_destroy_pf_port(pf);
4506 err_devlink_create_pf_port:
4507 err_init_mac_fltr:
4508 	ice_decfg_netdev(vsi);
4509 	return err;
4510 }
4511 
4512 static void ice_deinit_eth(struct ice_pf *pf)
4513 {
4514 	struct ice_vsi *vsi = ice_get_main_vsi(pf);
4515 
4516 	if (!vsi)
4517 		return;
4518 
4519 	ice_vsi_close(vsi);
4520 	ice_unregister_netdev(vsi);
4521 	ice_devlink_destroy_pf_port(pf);
4522 	ice_tc_indir_block_unregister(vsi);
4523 	ice_decfg_netdev(vsi);
4524 }
4525 
4526 /**
4527  * ice_wait_for_fw - wait for full FW readiness
4528  * @hw: pointer to the hardware structure
4529  * @timeout: milliseconds that can elapse before timing out
4530  */
4531 static int ice_wait_for_fw(struct ice_hw *hw, u32 timeout)
4532 {
4533 	int fw_loading;
4534 	u32 elapsed = 0;
4535 
4536 	while (elapsed <= timeout) {
4537 		fw_loading = rd32(hw, GL_MNG_FWSM) & GL_MNG_FWSM_FW_LOADING_M;
4538 
4539 		/* firmware was not yet loaded, we have to wait more */
4540 		if (fw_loading) {
4541 			elapsed += 100;
4542 			msleep(100);
4543 			continue;
4544 		}
4545 		return 0;
4546 	}
4547 
4548 	return -ETIMEDOUT;
4549 }
4550 
4551 static int ice_init_dev(struct ice_pf *pf)
4552 {
4553 	struct device *dev = ice_pf_to_dev(pf);
4554 	struct ice_hw *hw = &pf->hw;
4555 	int err;
4556 
4557 	err = ice_init_hw(hw);
4558 	if (err) {
4559 		dev_err(dev, "ice_init_hw failed: %d\n", err);
4560 		return err;
4561 	}
4562 
4563 	/* Some cards require longer initialization times
4564 	 * due to necessity of loading FW from an external source.
4565 	 * This can take even half a minute.
4566 	 */
4567 	if (ice_is_pf_c827(hw)) {
4568 		err = ice_wait_for_fw(hw, 30000);
4569 		if (err) {
4570 			dev_err(dev, "ice_wait_for_fw timed out");
4571 			return err;
4572 		}
4573 	}
4574 
4575 	ice_init_feature_support(pf);
4576 
4577 	ice_request_fw(pf);
4578 
4579 	/* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4580 	 * set in pf->state, which will cause ice_is_safe_mode to return
4581 	 * true
4582 	 */
4583 	if (ice_is_safe_mode(pf)) {
4584 		/* we already got function/device capabilities but these don't
4585 		 * reflect what the driver needs to do in safe mode. Instead of
4586 		 * adding conditional logic everywhere to ignore these
4587 		 * device/function capabilities, override them.
4588 		 */
4589 		ice_set_safe_mode_caps(hw);
4590 	}
4591 
4592 	err = ice_init_pf(pf);
4593 	if (err) {
4594 		dev_err(dev, "ice_init_pf failed: %d\n", err);
4595 		goto err_init_pf;
4596 	}
4597 
4598 	pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4599 	pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4600 	pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4601 	pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4602 	if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4603 		pf->hw.udp_tunnel_nic.tables[0].n_entries =
4604 			pf->hw.tnl.valid_count[TNL_VXLAN];
4605 		pf->hw.udp_tunnel_nic.tables[0].tunnel_types =
4606 			UDP_TUNNEL_TYPE_VXLAN;
4607 	}
4608 	if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4609 		pf->hw.udp_tunnel_nic.tables[1].n_entries =
4610 			pf->hw.tnl.valid_count[TNL_GENEVE];
4611 		pf->hw.udp_tunnel_nic.tables[1].tunnel_types =
4612 			UDP_TUNNEL_TYPE_GENEVE;
4613 	}
4614 
4615 	err = ice_init_interrupt_scheme(pf);
4616 	if (err) {
4617 		dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4618 		err = -EIO;
4619 		goto err_init_interrupt_scheme;
4620 	}
4621 
4622 	/* In case of MSIX we are going to setup the misc vector right here
4623 	 * to handle admin queue events etc. In case of legacy and MSI
4624 	 * the misc functionality and queue processing is combined in
4625 	 * the same vector and that gets setup at open.
4626 	 */
4627 	err = ice_req_irq_msix_misc(pf);
4628 	if (err) {
4629 		dev_err(dev, "setup of misc vector failed: %d\n", err);
4630 		goto err_req_irq_msix_misc;
4631 	}
4632 
4633 	return 0;
4634 
4635 err_req_irq_msix_misc:
4636 	ice_clear_interrupt_scheme(pf);
4637 err_init_interrupt_scheme:
4638 	ice_deinit_pf(pf);
4639 err_init_pf:
4640 	ice_deinit_hw(hw);
4641 	return err;
4642 }
4643 
4644 static void ice_deinit_dev(struct ice_pf *pf)
4645 {
4646 	ice_free_irq_msix_misc(pf);
4647 	ice_deinit_pf(pf);
4648 	ice_deinit_hw(&pf->hw);
4649 
4650 	/* Service task is already stopped, so call reset directly. */
4651 	ice_reset(&pf->hw, ICE_RESET_PFR);
4652 	pci_wait_for_pending_transaction(pf->pdev);
4653 	ice_clear_interrupt_scheme(pf);
4654 }
4655 
4656 static void ice_init_features(struct ice_pf *pf)
4657 {
4658 	struct device *dev = ice_pf_to_dev(pf);
4659 
4660 	if (ice_is_safe_mode(pf))
4661 		return;
4662 
4663 	/* initialize DDP driven features */
4664 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4665 		ice_ptp_init(pf);
4666 
4667 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
4668 		ice_gnss_init(pf);
4669 
4670 	if (ice_is_feature_supported(pf, ICE_F_CGU) ||
4671 	    ice_is_feature_supported(pf, ICE_F_PHY_RCLK))
4672 		ice_dpll_init(pf);
4673 
4674 	/* Note: Flow director init failure is non-fatal to load */
4675 	if (ice_init_fdir(pf))
4676 		dev_err(dev, "could not initialize flow director\n");
4677 
4678 	/* Note: DCB init failure is non-fatal to load */
4679 	if (ice_init_pf_dcb(pf, false)) {
4680 		clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4681 		clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4682 	} else {
4683 		ice_cfg_lldp_mib_change(&pf->hw, true);
4684 	}
4685 
4686 	if (ice_init_lag(pf))
4687 		dev_warn(dev, "Failed to init link aggregation support\n");
4688 }
4689 
4690 static void ice_deinit_features(struct ice_pf *pf)
4691 {
4692 	if (ice_is_safe_mode(pf))
4693 		return;
4694 
4695 	ice_deinit_lag(pf);
4696 	if (test_bit(ICE_FLAG_DCB_CAPABLE, pf->flags))
4697 		ice_cfg_lldp_mib_change(&pf->hw, false);
4698 	ice_deinit_fdir(pf);
4699 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
4700 		ice_gnss_exit(pf);
4701 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4702 		ice_ptp_release(pf);
4703 	if (test_bit(ICE_FLAG_DPLL, pf->flags))
4704 		ice_dpll_deinit(pf);
4705 }
4706 
4707 static void ice_init_wakeup(struct ice_pf *pf)
4708 {
4709 	/* Save wakeup reason register for later use */
4710 	pf->wakeup_reason = rd32(&pf->hw, PFPM_WUS);
4711 
4712 	/* check for a power management event */
4713 	ice_print_wake_reason(pf);
4714 
4715 	/* clear wake status, all bits */
4716 	wr32(&pf->hw, PFPM_WUS, U32_MAX);
4717 
4718 	/* Disable WoL at init, wait for user to enable */
4719 	device_set_wakeup_enable(ice_pf_to_dev(pf), false);
4720 }
4721 
4722 static int ice_init_link(struct ice_pf *pf)
4723 {
4724 	struct device *dev = ice_pf_to_dev(pf);
4725 	int err;
4726 
4727 	err = ice_init_link_events(pf->hw.port_info);
4728 	if (err) {
4729 		dev_err(dev, "ice_init_link_events failed: %d\n", err);
4730 		return err;
4731 	}
4732 
4733 	/* not a fatal error if this fails */
4734 	err = ice_init_nvm_phy_type(pf->hw.port_info);
4735 	if (err)
4736 		dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4737 
4738 	/* not a fatal error if this fails */
4739 	err = ice_update_link_info(pf->hw.port_info);
4740 	if (err)
4741 		dev_err(dev, "ice_update_link_info failed: %d\n", err);
4742 
4743 	ice_init_link_dflt_override(pf->hw.port_info);
4744 
4745 	ice_check_link_cfg_err(pf,
4746 			       pf->hw.port_info->phy.link_info.link_cfg_err);
4747 
4748 	/* if media available, initialize PHY settings */
4749 	if (pf->hw.port_info->phy.link_info.link_info &
4750 	    ICE_AQ_MEDIA_AVAILABLE) {
4751 		/* not a fatal error if this fails */
4752 		err = ice_init_phy_user_cfg(pf->hw.port_info);
4753 		if (err)
4754 			dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4755 
4756 		if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4757 			struct ice_vsi *vsi = ice_get_main_vsi(pf);
4758 
4759 			if (vsi)
4760 				ice_configure_phy(vsi);
4761 		}
4762 	} else {
4763 		set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4764 	}
4765 
4766 	return err;
4767 }
4768 
4769 static int ice_init_pf_sw(struct ice_pf *pf)
4770 {
4771 	bool dvm = ice_is_dvm_ena(&pf->hw);
4772 	struct ice_vsi *vsi;
4773 	int err;
4774 
4775 	/* create switch struct for the switch element created by FW on boot */
4776 	pf->first_sw = kzalloc(sizeof(*pf->first_sw), GFP_KERNEL);
4777 	if (!pf->first_sw)
4778 		return -ENOMEM;
4779 
4780 	if (pf->hw.evb_veb)
4781 		pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4782 	else
4783 		pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4784 
4785 	pf->first_sw->pf = pf;
4786 
4787 	/* record the sw_id available for later use */
4788 	pf->first_sw->sw_id = pf->hw.port_info->sw_id;
4789 
4790 	err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
4791 	if (err)
4792 		goto err_aq_set_port_params;
4793 
4794 	vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
4795 	if (!vsi) {
4796 		err = -ENOMEM;
4797 		goto err_pf_vsi_setup;
4798 	}
4799 
4800 	return 0;
4801 
4802 err_pf_vsi_setup:
4803 err_aq_set_port_params:
4804 	kfree(pf->first_sw);
4805 	return err;
4806 }
4807 
4808 static void ice_deinit_pf_sw(struct ice_pf *pf)
4809 {
4810 	struct ice_vsi *vsi = ice_get_main_vsi(pf);
4811 
4812 	if (!vsi)
4813 		return;
4814 
4815 	ice_vsi_release(vsi);
4816 	kfree(pf->first_sw);
4817 }
4818 
4819 static int ice_alloc_vsis(struct ice_pf *pf)
4820 {
4821 	struct device *dev = ice_pf_to_dev(pf);
4822 
4823 	pf->num_alloc_vsi = pf->hw.func_caps.guar_num_vsi;
4824 	if (!pf->num_alloc_vsi)
4825 		return -EIO;
4826 
4827 	if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4828 		dev_warn(dev,
4829 			 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4830 			 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4831 		pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4832 	}
4833 
4834 	pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4835 			       GFP_KERNEL);
4836 	if (!pf->vsi)
4837 		return -ENOMEM;
4838 
4839 	pf->vsi_stats = devm_kcalloc(dev, pf->num_alloc_vsi,
4840 				     sizeof(*pf->vsi_stats), GFP_KERNEL);
4841 	if (!pf->vsi_stats) {
4842 		devm_kfree(dev, pf->vsi);
4843 		return -ENOMEM;
4844 	}
4845 
4846 	return 0;
4847 }
4848 
4849 static void ice_dealloc_vsis(struct ice_pf *pf)
4850 {
4851 	devm_kfree(ice_pf_to_dev(pf), pf->vsi_stats);
4852 	pf->vsi_stats = NULL;
4853 
4854 	pf->num_alloc_vsi = 0;
4855 	devm_kfree(ice_pf_to_dev(pf), pf->vsi);
4856 	pf->vsi = NULL;
4857 }
4858 
4859 static int ice_init_devlink(struct ice_pf *pf)
4860 {
4861 	int err;
4862 
4863 	err = ice_devlink_register_params(pf);
4864 	if (err)
4865 		return err;
4866 
4867 	ice_devlink_init_regions(pf);
4868 	ice_devlink_register(pf);
4869 
4870 	return 0;
4871 }
4872 
4873 static void ice_deinit_devlink(struct ice_pf *pf)
4874 {
4875 	ice_devlink_unregister(pf);
4876 	ice_devlink_destroy_regions(pf);
4877 	ice_devlink_unregister_params(pf);
4878 }
4879 
4880 static int ice_init(struct ice_pf *pf)
4881 {
4882 	int err;
4883 
4884 	err = ice_init_dev(pf);
4885 	if (err)
4886 		return err;
4887 
4888 	err = ice_alloc_vsis(pf);
4889 	if (err)
4890 		goto err_alloc_vsis;
4891 
4892 	err = ice_init_pf_sw(pf);
4893 	if (err)
4894 		goto err_init_pf_sw;
4895 
4896 	ice_init_wakeup(pf);
4897 
4898 	err = ice_init_link(pf);
4899 	if (err)
4900 		goto err_init_link;
4901 
4902 	err = ice_send_version(pf);
4903 	if (err)
4904 		goto err_init_link;
4905 
4906 	ice_verify_cacheline_size(pf);
4907 
4908 	if (ice_is_safe_mode(pf))
4909 		ice_set_safe_mode_vlan_cfg(pf);
4910 	else
4911 		/* print PCI link speed and width */
4912 		pcie_print_link_status(pf->pdev);
4913 
4914 	/* ready to go, so clear down state bit */
4915 	clear_bit(ICE_DOWN, pf->state);
4916 	clear_bit(ICE_SERVICE_DIS, pf->state);
4917 
4918 	/* since everything is good, start the service timer */
4919 	mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4920 
4921 	return 0;
4922 
4923 err_init_link:
4924 	ice_deinit_pf_sw(pf);
4925 err_init_pf_sw:
4926 	ice_dealloc_vsis(pf);
4927 err_alloc_vsis:
4928 	ice_deinit_dev(pf);
4929 	return err;
4930 }
4931 
4932 static void ice_deinit(struct ice_pf *pf)
4933 {
4934 	set_bit(ICE_SERVICE_DIS, pf->state);
4935 	set_bit(ICE_DOWN, pf->state);
4936 
4937 	ice_deinit_pf_sw(pf);
4938 	ice_dealloc_vsis(pf);
4939 	ice_deinit_dev(pf);
4940 }
4941 
4942 /**
4943  * ice_load - load pf by init hw and starting VSI
4944  * @pf: pointer to the pf instance
4945  */
4946 int ice_load(struct ice_pf *pf)
4947 {
4948 	struct ice_vsi_cfg_params params = {};
4949 	struct ice_vsi *vsi;
4950 	int err;
4951 
4952 	err = ice_init_dev(pf);
4953 	if (err)
4954 		return err;
4955 
4956 	vsi = ice_get_main_vsi(pf);
4957 
4958 	params = ice_vsi_to_params(vsi);
4959 	params.flags = ICE_VSI_FLAG_INIT;
4960 
4961 	rtnl_lock();
4962 	err = ice_vsi_cfg(vsi, &params);
4963 	if (err)
4964 		goto err_vsi_cfg;
4965 
4966 	err = ice_start_eth(ice_get_main_vsi(pf));
4967 	if (err)
4968 		goto err_start_eth;
4969 	rtnl_unlock();
4970 
4971 	err = ice_init_rdma(pf);
4972 	if (err)
4973 		goto err_init_rdma;
4974 
4975 	ice_init_features(pf);
4976 	ice_service_task_restart(pf);
4977 
4978 	clear_bit(ICE_DOWN, pf->state);
4979 
4980 	return 0;
4981 
4982 err_init_rdma:
4983 	ice_vsi_close(ice_get_main_vsi(pf));
4984 	rtnl_lock();
4985 err_start_eth:
4986 	ice_vsi_decfg(ice_get_main_vsi(pf));
4987 err_vsi_cfg:
4988 	rtnl_unlock();
4989 	ice_deinit_dev(pf);
4990 	return err;
4991 }
4992 
4993 /**
4994  * ice_unload - unload pf by stopping VSI and deinit hw
4995  * @pf: pointer to the pf instance
4996  */
4997 void ice_unload(struct ice_pf *pf)
4998 {
4999 	ice_deinit_features(pf);
5000 	ice_deinit_rdma(pf);
5001 	rtnl_lock();
5002 	ice_stop_eth(ice_get_main_vsi(pf));
5003 	ice_vsi_decfg(ice_get_main_vsi(pf));
5004 	rtnl_unlock();
5005 	ice_deinit_dev(pf);
5006 }
5007 
5008 /**
5009  * ice_probe - Device initialization routine
5010  * @pdev: PCI device information struct
5011  * @ent: entry in ice_pci_tbl
5012  *
5013  * Returns 0 on success, negative on failure
5014  */
5015 static int
5016 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
5017 {
5018 	struct device *dev = &pdev->dev;
5019 	struct ice_pf *pf;
5020 	struct ice_hw *hw;
5021 	int err;
5022 
5023 	if (pdev->is_virtfn) {
5024 		dev_err(dev, "can't probe a virtual function\n");
5025 		return -EINVAL;
5026 	}
5027 
5028 	/* when under a kdump kernel initiate a reset before enabling the
5029 	 * device in order to clear out any pending DMA transactions. These
5030 	 * transactions can cause some systems to machine check when doing
5031 	 * the pcim_enable_device() below.
5032 	 */
5033 	if (is_kdump_kernel()) {
5034 		pci_save_state(pdev);
5035 		pci_clear_master(pdev);
5036 		err = pcie_flr(pdev);
5037 		if (err)
5038 			return err;
5039 		pci_restore_state(pdev);
5040 	}
5041 
5042 	/* this driver uses devres, see
5043 	 * Documentation/driver-api/driver-model/devres.rst
5044 	 */
5045 	err = pcim_enable_device(pdev);
5046 	if (err)
5047 		return err;
5048 
5049 	err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
5050 	if (err) {
5051 		dev_err(dev, "BAR0 I/O map error %d\n", err);
5052 		return err;
5053 	}
5054 
5055 	pf = ice_allocate_pf(dev);
5056 	if (!pf)
5057 		return -ENOMEM;
5058 
5059 	/* initialize Auxiliary index to invalid value */
5060 	pf->aux_idx = -1;
5061 
5062 	/* set up for high or low DMA */
5063 	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
5064 	if (err) {
5065 		dev_err(dev, "DMA configuration failed: 0x%x\n", err);
5066 		return err;
5067 	}
5068 
5069 	pci_set_master(pdev);
5070 
5071 	pf->pdev = pdev;
5072 	pci_set_drvdata(pdev, pf);
5073 	set_bit(ICE_DOWN, pf->state);
5074 	/* Disable service task until DOWN bit is cleared */
5075 	set_bit(ICE_SERVICE_DIS, pf->state);
5076 
5077 	hw = &pf->hw;
5078 	hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
5079 	pci_save_state(pdev);
5080 
5081 	hw->back = pf;
5082 	hw->port_info = NULL;
5083 	hw->vendor_id = pdev->vendor;
5084 	hw->device_id = pdev->device;
5085 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
5086 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
5087 	hw->subsystem_device_id = pdev->subsystem_device;
5088 	hw->bus.device = PCI_SLOT(pdev->devfn);
5089 	hw->bus.func = PCI_FUNC(pdev->devfn);
5090 	ice_set_ctrlq_len(hw);
5091 
5092 	pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
5093 
5094 #ifndef CONFIG_DYNAMIC_DEBUG
5095 	if (debug < -1)
5096 		hw->debug_mask = debug;
5097 #endif
5098 
5099 	err = ice_init(pf);
5100 	if (err)
5101 		goto err_init;
5102 
5103 	err = ice_init_eth(pf);
5104 	if (err)
5105 		goto err_init_eth;
5106 
5107 	err = ice_init_rdma(pf);
5108 	if (err)
5109 		goto err_init_rdma;
5110 
5111 	err = ice_init_devlink(pf);
5112 	if (err)
5113 		goto err_init_devlink;
5114 
5115 	ice_init_features(pf);
5116 
5117 	return 0;
5118 
5119 err_init_devlink:
5120 	ice_deinit_rdma(pf);
5121 err_init_rdma:
5122 	ice_deinit_eth(pf);
5123 err_init_eth:
5124 	ice_deinit(pf);
5125 err_init:
5126 	pci_disable_device(pdev);
5127 	return err;
5128 }
5129 
5130 /**
5131  * ice_set_wake - enable or disable Wake on LAN
5132  * @pf: pointer to the PF struct
5133  *
5134  * Simple helper for WoL control
5135  */
5136 static void ice_set_wake(struct ice_pf *pf)
5137 {
5138 	struct ice_hw *hw = &pf->hw;
5139 	bool wol = pf->wol_ena;
5140 
5141 	/* clear wake state, otherwise new wake events won't fire */
5142 	wr32(hw, PFPM_WUS, U32_MAX);
5143 
5144 	/* enable / disable APM wake up, no RMW needed */
5145 	wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
5146 
5147 	/* set magic packet filter enabled */
5148 	wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
5149 }
5150 
5151 /**
5152  * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
5153  * @pf: pointer to the PF struct
5154  *
5155  * Issue firmware command to enable multicast magic wake, making
5156  * sure that any locally administered address (LAA) is used for
5157  * wake, and that PF reset doesn't undo the LAA.
5158  */
5159 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
5160 {
5161 	struct device *dev = ice_pf_to_dev(pf);
5162 	struct ice_hw *hw = &pf->hw;
5163 	u8 mac_addr[ETH_ALEN];
5164 	struct ice_vsi *vsi;
5165 	int status;
5166 	u8 flags;
5167 
5168 	if (!pf->wol_ena)
5169 		return;
5170 
5171 	vsi = ice_get_main_vsi(pf);
5172 	if (!vsi)
5173 		return;
5174 
5175 	/* Get current MAC address in case it's an LAA */
5176 	if (vsi->netdev)
5177 		ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
5178 	else
5179 		ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
5180 
5181 	flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
5182 		ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
5183 		ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
5184 
5185 	status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
5186 	if (status)
5187 		dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
5188 			status, ice_aq_str(hw->adminq.sq_last_status));
5189 }
5190 
5191 /**
5192  * ice_remove - Device removal routine
5193  * @pdev: PCI device information struct
5194  */
5195 static void ice_remove(struct pci_dev *pdev)
5196 {
5197 	struct ice_pf *pf = pci_get_drvdata(pdev);
5198 	int i;
5199 
5200 	for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
5201 		if (!ice_is_reset_in_progress(pf->state))
5202 			break;
5203 		msleep(100);
5204 	}
5205 
5206 	if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
5207 		set_bit(ICE_VF_RESETS_DISABLED, pf->state);
5208 		ice_free_vfs(pf);
5209 	}
5210 
5211 	ice_service_task_stop(pf);
5212 	ice_aq_cancel_waiting_tasks(pf);
5213 	set_bit(ICE_DOWN, pf->state);
5214 
5215 	if (!ice_is_safe_mode(pf))
5216 		ice_remove_arfs(pf);
5217 	ice_deinit_features(pf);
5218 	ice_deinit_devlink(pf);
5219 	ice_deinit_rdma(pf);
5220 	ice_deinit_eth(pf);
5221 	ice_deinit(pf);
5222 
5223 	ice_vsi_release_all(pf);
5224 
5225 	ice_setup_mc_magic_wake(pf);
5226 	ice_set_wake(pf);
5227 
5228 	pci_disable_device(pdev);
5229 }
5230 
5231 /**
5232  * ice_shutdown - PCI callback for shutting down device
5233  * @pdev: PCI device information struct
5234  */
5235 static void ice_shutdown(struct pci_dev *pdev)
5236 {
5237 	struct ice_pf *pf = pci_get_drvdata(pdev);
5238 
5239 	ice_remove(pdev);
5240 
5241 	if (system_state == SYSTEM_POWER_OFF) {
5242 		pci_wake_from_d3(pdev, pf->wol_ena);
5243 		pci_set_power_state(pdev, PCI_D3hot);
5244 	}
5245 }
5246 
5247 #ifdef CONFIG_PM
5248 /**
5249  * ice_prepare_for_shutdown - prep for PCI shutdown
5250  * @pf: board private structure
5251  *
5252  * Inform or close all dependent features in prep for PCI device shutdown
5253  */
5254 static void ice_prepare_for_shutdown(struct ice_pf *pf)
5255 {
5256 	struct ice_hw *hw = &pf->hw;
5257 	u32 v;
5258 
5259 	/* Notify VFs of impending reset */
5260 	if (ice_check_sq_alive(hw, &hw->mailboxq))
5261 		ice_vc_notify_reset(pf);
5262 
5263 	dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
5264 
5265 	/* disable the VSIs and their queues that are not already DOWN */
5266 	ice_pf_dis_all_vsi(pf, false);
5267 
5268 	ice_for_each_vsi(pf, v)
5269 		if (pf->vsi[v])
5270 			pf->vsi[v]->vsi_num = 0;
5271 
5272 	ice_shutdown_all_ctrlq(hw);
5273 }
5274 
5275 /**
5276  * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
5277  * @pf: board private structure to reinitialize
5278  *
5279  * This routine reinitialize interrupt scheme that was cleared during
5280  * power management suspend callback.
5281  *
5282  * This should be called during resume routine to re-allocate the q_vectors
5283  * and reacquire interrupts.
5284  */
5285 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
5286 {
5287 	struct device *dev = ice_pf_to_dev(pf);
5288 	int ret, v;
5289 
5290 	/* Since we clear MSIX flag during suspend, we need to
5291 	 * set it back during resume...
5292 	 */
5293 
5294 	ret = ice_init_interrupt_scheme(pf);
5295 	if (ret) {
5296 		dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
5297 		return ret;
5298 	}
5299 
5300 	/* Remap vectors and rings, after successful re-init interrupts */
5301 	ice_for_each_vsi(pf, v) {
5302 		if (!pf->vsi[v])
5303 			continue;
5304 
5305 		ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
5306 		if (ret)
5307 			goto err_reinit;
5308 		ice_vsi_map_rings_to_vectors(pf->vsi[v]);
5309 	}
5310 
5311 	ret = ice_req_irq_msix_misc(pf);
5312 	if (ret) {
5313 		dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5314 			ret);
5315 		goto err_reinit;
5316 	}
5317 
5318 	return 0;
5319 
5320 err_reinit:
5321 	while (v--)
5322 		if (pf->vsi[v])
5323 			ice_vsi_free_q_vectors(pf->vsi[v]);
5324 
5325 	return ret;
5326 }
5327 
5328 /**
5329  * ice_suspend
5330  * @dev: generic device information structure
5331  *
5332  * Power Management callback to quiesce the device and prepare
5333  * for D3 transition.
5334  */
5335 static int __maybe_unused ice_suspend(struct device *dev)
5336 {
5337 	struct pci_dev *pdev = to_pci_dev(dev);
5338 	struct ice_pf *pf;
5339 	int disabled, v;
5340 
5341 	pf = pci_get_drvdata(pdev);
5342 
5343 	if (!ice_pf_state_is_nominal(pf)) {
5344 		dev_err(dev, "Device is not ready, no need to suspend it\n");
5345 		return -EBUSY;
5346 	}
5347 
5348 	/* Stop watchdog tasks until resume completion.
5349 	 * Even though it is most likely that the service task is
5350 	 * disabled if the device is suspended or down, the service task's
5351 	 * state is controlled by a different state bit, and we should
5352 	 * store and honor whatever state that bit is in at this point.
5353 	 */
5354 	disabled = ice_service_task_stop(pf);
5355 
5356 	ice_unplug_aux_dev(pf);
5357 
5358 	/* Already suspended?, then there is nothing to do */
5359 	if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5360 		if (!disabled)
5361 			ice_service_task_restart(pf);
5362 		return 0;
5363 	}
5364 
5365 	if (test_bit(ICE_DOWN, pf->state) ||
5366 	    ice_is_reset_in_progress(pf->state)) {
5367 		dev_err(dev, "can't suspend device in reset or already down\n");
5368 		if (!disabled)
5369 			ice_service_task_restart(pf);
5370 		return 0;
5371 	}
5372 
5373 	ice_setup_mc_magic_wake(pf);
5374 
5375 	ice_prepare_for_shutdown(pf);
5376 
5377 	ice_set_wake(pf);
5378 
5379 	/* Free vectors, clear the interrupt scheme and release IRQs
5380 	 * for proper hibernation, especially with large number of CPUs.
5381 	 * Otherwise hibernation might fail when mapping all the vectors back
5382 	 * to CPU0.
5383 	 */
5384 	ice_free_irq_msix_misc(pf);
5385 	ice_for_each_vsi(pf, v) {
5386 		if (!pf->vsi[v])
5387 			continue;
5388 		ice_vsi_free_q_vectors(pf->vsi[v]);
5389 	}
5390 	ice_clear_interrupt_scheme(pf);
5391 
5392 	pci_save_state(pdev);
5393 	pci_wake_from_d3(pdev, pf->wol_ena);
5394 	pci_set_power_state(pdev, PCI_D3hot);
5395 	return 0;
5396 }
5397 
5398 /**
5399  * ice_resume - PM callback for waking up from D3
5400  * @dev: generic device information structure
5401  */
5402 static int __maybe_unused ice_resume(struct device *dev)
5403 {
5404 	struct pci_dev *pdev = to_pci_dev(dev);
5405 	enum ice_reset_req reset_type;
5406 	struct ice_pf *pf;
5407 	struct ice_hw *hw;
5408 	int ret;
5409 
5410 	pci_set_power_state(pdev, PCI_D0);
5411 	pci_restore_state(pdev);
5412 	pci_save_state(pdev);
5413 
5414 	if (!pci_device_is_present(pdev))
5415 		return -ENODEV;
5416 
5417 	ret = pci_enable_device_mem(pdev);
5418 	if (ret) {
5419 		dev_err(dev, "Cannot enable device after suspend\n");
5420 		return ret;
5421 	}
5422 
5423 	pf = pci_get_drvdata(pdev);
5424 	hw = &pf->hw;
5425 
5426 	pf->wakeup_reason = rd32(hw, PFPM_WUS);
5427 	ice_print_wake_reason(pf);
5428 
5429 	/* We cleared the interrupt scheme when we suspended, so we need to
5430 	 * restore it now to resume device functionality.
5431 	 */
5432 	ret = ice_reinit_interrupt_scheme(pf);
5433 	if (ret)
5434 		dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5435 
5436 	clear_bit(ICE_DOWN, pf->state);
5437 	/* Now perform PF reset and rebuild */
5438 	reset_type = ICE_RESET_PFR;
5439 	/* re-enable service task for reset, but allow reset to schedule it */
5440 	clear_bit(ICE_SERVICE_DIS, pf->state);
5441 
5442 	if (ice_schedule_reset(pf, reset_type))
5443 		dev_err(dev, "Reset during resume failed.\n");
5444 
5445 	clear_bit(ICE_SUSPENDED, pf->state);
5446 	ice_service_task_restart(pf);
5447 
5448 	/* Restart the service task */
5449 	mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5450 
5451 	return 0;
5452 }
5453 #endif /* CONFIG_PM */
5454 
5455 /**
5456  * ice_pci_err_detected - warning that PCI error has been detected
5457  * @pdev: PCI device information struct
5458  * @err: the type of PCI error
5459  *
5460  * Called to warn that something happened on the PCI bus and the error handling
5461  * is in progress.  Allows the driver to gracefully prepare/handle PCI errors.
5462  */
5463 static pci_ers_result_t
5464 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5465 {
5466 	struct ice_pf *pf = pci_get_drvdata(pdev);
5467 
5468 	if (!pf) {
5469 		dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5470 			__func__, err);
5471 		return PCI_ERS_RESULT_DISCONNECT;
5472 	}
5473 
5474 	if (!test_bit(ICE_SUSPENDED, pf->state)) {
5475 		ice_service_task_stop(pf);
5476 
5477 		if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5478 			set_bit(ICE_PFR_REQ, pf->state);
5479 			ice_prepare_for_reset(pf, ICE_RESET_PFR);
5480 		}
5481 	}
5482 
5483 	return PCI_ERS_RESULT_NEED_RESET;
5484 }
5485 
5486 /**
5487  * ice_pci_err_slot_reset - a PCI slot reset has just happened
5488  * @pdev: PCI device information struct
5489  *
5490  * Called to determine if the driver can recover from the PCI slot reset by
5491  * using a register read to determine if the device is recoverable.
5492  */
5493 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5494 {
5495 	struct ice_pf *pf = pci_get_drvdata(pdev);
5496 	pci_ers_result_t result;
5497 	int err;
5498 	u32 reg;
5499 
5500 	err = pci_enable_device_mem(pdev);
5501 	if (err) {
5502 		dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5503 			err);
5504 		result = PCI_ERS_RESULT_DISCONNECT;
5505 	} else {
5506 		pci_set_master(pdev);
5507 		pci_restore_state(pdev);
5508 		pci_save_state(pdev);
5509 		pci_wake_from_d3(pdev, false);
5510 
5511 		/* Check for life */
5512 		reg = rd32(&pf->hw, GLGEN_RTRIG);
5513 		if (!reg)
5514 			result = PCI_ERS_RESULT_RECOVERED;
5515 		else
5516 			result = PCI_ERS_RESULT_DISCONNECT;
5517 	}
5518 
5519 	return result;
5520 }
5521 
5522 /**
5523  * ice_pci_err_resume - restart operations after PCI error recovery
5524  * @pdev: PCI device information struct
5525  *
5526  * Called to allow the driver to bring things back up after PCI error and/or
5527  * reset recovery have finished
5528  */
5529 static void ice_pci_err_resume(struct pci_dev *pdev)
5530 {
5531 	struct ice_pf *pf = pci_get_drvdata(pdev);
5532 
5533 	if (!pf) {
5534 		dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5535 			__func__);
5536 		return;
5537 	}
5538 
5539 	if (test_bit(ICE_SUSPENDED, pf->state)) {
5540 		dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5541 			__func__);
5542 		return;
5543 	}
5544 
5545 	ice_restore_all_vfs_msi_state(pf);
5546 
5547 	ice_do_reset(pf, ICE_RESET_PFR);
5548 	ice_service_task_restart(pf);
5549 	mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5550 }
5551 
5552 /**
5553  * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5554  * @pdev: PCI device information struct
5555  */
5556 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5557 {
5558 	struct ice_pf *pf = pci_get_drvdata(pdev);
5559 
5560 	if (!test_bit(ICE_SUSPENDED, pf->state)) {
5561 		ice_service_task_stop(pf);
5562 
5563 		if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5564 			set_bit(ICE_PFR_REQ, pf->state);
5565 			ice_prepare_for_reset(pf, ICE_RESET_PFR);
5566 		}
5567 	}
5568 }
5569 
5570 /**
5571  * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5572  * @pdev: PCI device information struct
5573  */
5574 static void ice_pci_err_reset_done(struct pci_dev *pdev)
5575 {
5576 	ice_pci_err_resume(pdev);
5577 }
5578 
5579 /* ice_pci_tbl - PCI Device ID Table
5580  *
5581  * Wildcard entries (PCI_ANY_ID) should come last
5582  * Last entry must be all 0s
5583  *
5584  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5585  *   Class, Class Mask, private data (not used) }
5586  */
5587 static const struct pci_device_id ice_pci_tbl[] = {
5588 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE) },
5589 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP) },
5590 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP) },
5591 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE) },
5592 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP) },
5593 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP) },
5594 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE) },
5595 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP) },
5596 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP) },
5597 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T) },
5598 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII) },
5599 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE) },
5600 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP) },
5601 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP) },
5602 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T) },
5603 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII) },
5604 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE) },
5605 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP) },
5606 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T) },
5607 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII) },
5608 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE) },
5609 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP) },
5610 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T) },
5611 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE) },
5612 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP) },
5613 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822_SI_DFLT) },
5614 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_BACKPLANE) },
5615 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_QSFP56) },
5616 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_SFP) },
5617 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_SFP_DD) },
5618 	/* required last entry */
5619 	{}
5620 };
5621 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5622 
5623 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5624 
5625 static const struct pci_error_handlers ice_pci_err_handler = {
5626 	.error_detected = ice_pci_err_detected,
5627 	.slot_reset = ice_pci_err_slot_reset,
5628 	.reset_prepare = ice_pci_err_reset_prepare,
5629 	.reset_done = ice_pci_err_reset_done,
5630 	.resume = ice_pci_err_resume
5631 };
5632 
5633 static struct pci_driver ice_driver = {
5634 	.name = KBUILD_MODNAME,
5635 	.id_table = ice_pci_tbl,
5636 	.probe = ice_probe,
5637 	.remove = ice_remove,
5638 #ifdef CONFIG_PM
5639 	.driver.pm = &ice_pm_ops,
5640 #endif /* CONFIG_PM */
5641 	.shutdown = ice_shutdown,
5642 	.sriov_configure = ice_sriov_configure,
5643 	.sriov_get_vf_total_msix = ice_sriov_get_vf_total_msix,
5644 	.sriov_set_msix_vec_count = ice_sriov_set_msix_vec_count,
5645 	.err_handler = &ice_pci_err_handler
5646 };
5647 
5648 /**
5649  * ice_module_init - Driver registration routine
5650  *
5651  * ice_module_init is the first routine called when the driver is
5652  * loaded. All it does is register with the PCI subsystem.
5653  */
5654 static int __init ice_module_init(void)
5655 {
5656 	int status = -ENOMEM;
5657 
5658 	pr_info("%s\n", ice_driver_string);
5659 	pr_info("%s\n", ice_copyright);
5660 
5661 	ice_adv_lnk_speed_maps_init();
5662 
5663 	ice_wq = alloc_workqueue("%s", 0, 0, KBUILD_MODNAME);
5664 	if (!ice_wq) {
5665 		pr_err("Failed to create workqueue\n");
5666 		return status;
5667 	}
5668 
5669 	ice_lag_wq = alloc_ordered_workqueue("ice_lag_wq", 0);
5670 	if (!ice_lag_wq) {
5671 		pr_err("Failed to create LAG workqueue\n");
5672 		goto err_dest_wq;
5673 	}
5674 
5675 	status = pci_register_driver(&ice_driver);
5676 	if (status) {
5677 		pr_err("failed to register PCI driver, err %d\n", status);
5678 		goto err_dest_lag_wq;
5679 	}
5680 
5681 	return 0;
5682 
5683 err_dest_lag_wq:
5684 	destroy_workqueue(ice_lag_wq);
5685 err_dest_wq:
5686 	destroy_workqueue(ice_wq);
5687 	return status;
5688 }
5689 module_init(ice_module_init);
5690 
5691 /**
5692  * ice_module_exit - Driver exit cleanup routine
5693  *
5694  * ice_module_exit is called just before the driver is removed
5695  * from memory.
5696  */
5697 static void __exit ice_module_exit(void)
5698 {
5699 	pci_unregister_driver(&ice_driver);
5700 	destroy_workqueue(ice_wq);
5701 	destroy_workqueue(ice_lag_wq);
5702 	pr_info("module unloaded\n");
5703 }
5704 module_exit(ice_module_exit);
5705 
5706 /**
5707  * ice_set_mac_address - NDO callback to set MAC address
5708  * @netdev: network interface device structure
5709  * @pi: pointer to an address structure
5710  *
5711  * Returns 0 on success, negative on failure
5712  */
5713 static int ice_set_mac_address(struct net_device *netdev, void *pi)
5714 {
5715 	struct ice_netdev_priv *np = netdev_priv(netdev);
5716 	struct ice_vsi *vsi = np->vsi;
5717 	struct ice_pf *pf = vsi->back;
5718 	struct ice_hw *hw = &pf->hw;
5719 	struct sockaddr *addr = pi;
5720 	u8 old_mac[ETH_ALEN];
5721 	u8 flags = 0;
5722 	u8 *mac;
5723 	int err;
5724 
5725 	mac = (u8 *)addr->sa_data;
5726 
5727 	if (!is_valid_ether_addr(mac))
5728 		return -EADDRNOTAVAIL;
5729 
5730 	if (test_bit(ICE_DOWN, pf->state) ||
5731 	    ice_is_reset_in_progress(pf->state)) {
5732 		netdev_err(netdev, "can't set mac %pM. device not ready\n",
5733 			   mac);
5734 		return -EBUSY;
5735 	}
5736 
5737 	if (ice_chnl_dmac_fltr_cnt(pf)) {
5738 		netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5739 			   mac);
5740 		return -EAGAIN;
5741 	}
5742 
5743 	netif_addr_lock_bh(netdev);
5744 	ether_addr_copy(old_mac, netdev->dev_addr);
5745 	/* change the netdev's MAC address */
5746 	eth_hw_addr_set(netdev, mac);
5747 	netif_addr_unlock_bh(netdev);
5748 
5749 	/* Clean up old MAC filter. Not an error if old filter doesn't exist */
5750 	err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5751 	if (err && err != -ENOENT) {
5752 		err = -EADDRNOTAVAIL;
5753 		goto err_update_filters;
5754 	}
5755 
5756 	/* Add filter for new MAC. If filter exists, return success */
5757 	err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5758 	if (err == -EEXIST) {
5759 		/* Although this MAC filter is already present in hardware it's
5760 		 * possible in some cases (e.g. bonding) that dev_addr was
5761 		 * modified outside of the driver and needs to be restored back
5762 		 * to this value.
5763 		 */
5764 		netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5765 
5766 		return 0;
5767 	} else if (err) {
5768 		/* error if the new filter addition failed */
5769 		err = -EADDRNOTAVAIL;
5770 	}
5771 
5772 err_update_filters:
5773 	if (err) {
5774 		netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5775 			   mac);
5776 		netif_addr_lock_bh(netdev);
5777 		eth_hw_addr_set(netdev, old_mac);
5778 		netif_addr_unlock_bh(netdev);
5779 		return err;
5780 	}
5781 
5782 	netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5783 		   netdev->dev_addr);
5784 
5785 	/* write new MAC address to the firmware */
5786 	flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5787 	err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5788 	if (err) {
5789 		netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5790 			   mac, err);
5791 	}
5792 	return 0;
5793 }
5794 
5795 /**
5796  * ice_set_rx_mode - NDO callback to set the netdev filters
5797  * @netdev: network interface device structure
5798  */
5799 static void ice_set_rx_mode(struct net_device *netdev)
5800 {
5801 	struct ice_netdev_priv *np = netdev_priv(netdev);
5802 	struct ice_vsi *vsi = np->vsi;
5803 
5804 	if (!vsi || ice_is_switchdev_running(vsi->back))
5805 		return;
5806 
5807 	/* Set the flags to synchronize filters
5808 	 * ndo_set_rx_mode may be triggered even without a change in netdev
5809 	 * flags
5810 	 */
5811 	set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5812 	set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5813 	set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5814 
5815 	/* schedule our worker thread which will take care of
5816 	 * applying the new filter changes
5817 	 */
5818 	ice_service_task_schedule(vsi->back);
5819 }
5820 
5821 /**
5822  * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5823  * @netdev: network interface device structure
5824  * @queue_index: Queue ID
5825  * @maxrate: maximum bandwidth in Mbps
5826  */
5827 static int
5828 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5829 {
5830 	struct ice_netdev_priv *np = netdev_priv(netdev);
5831 	struct ice_vsi *vsi = np->vsi;
5832 	u16 q_handle;
5833 	int status;
5834 	u8 tc;
5835 
5836 	/* Validate maxrate requested is within permitted range */
5837 	if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5838 		netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5839 			   maxrate, queue_index);
5840 		return -EINVAL;
5841 	}
5842 
5843 	q_handle = vsi->tx_rings[queue_index]->q_handle;
5844 	tc = ice_dcb_get_tc(vsi, queue_index);
5845 
5846 	vsi = ice_locate_vsi_using_queue(vsi, queue_index);
5847 	if (!vsi) {
5848 		netdev_err(netdev, "Invalid VSI for given queue %d\n",
5849 			   queue_index);
5850 		return -EINVAL;
5851 	}
5852 
5853 	/* Set BW back to default, when user set maxrate to 0 */
5854 	if (!maxrate)
5855 		status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5856 					       q_handle, ICE_MAX_BW);
5857 	else
5858 		status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5859 					  q_handle, ICE_MAX_BW, maxrate * 1000);
5860 	if (status)
5861 		netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
5862 			   status);
5863 
5864 	return status;
5865 }
5866 
5867 /**
5868  * ice_fdb_add - add an entry to the hardware database
5869  * @ndm: the input from the stack
5870  * @tb: pointer to array of nladdr (unused)
5871  * @dev: the net device pointer
5872  * @addr: the MAC address entry being added
5873  * @vid: VLAN ID
5874  * @flags: instructions from stack about fdb operation
5875  * @extack: netlink extended ack
5876  */
5877 static int
5878 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5879 	    struct net_device *dev, const unsigned char *addr, u16 vid,
5880 	    u16 flags, struct netlink_ext_ack __always_unused *extack)
5881 {
5882 	int err;
5883 
5884 	if (vid) {
5885 		netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5886 		return -EINVAL;
5887 	}
5888 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5889 		netdev_err(dev, "FDB only supports static addresses\n");
5890 		return -EINVAL;
5891 	}
5892 
5893 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5894 		err = dev_uc_add_excl(dev, addr);
5895 	else if (is_multicast_ether_addr(addr))
5896 		err = dev_mc_add_excl(dev, addr);
5897 	else
5898 		err = -EINVAL;
5899 
5900 	/* Only return duplicate errors if NLM_F_EXCL is set */
5901 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
5902 		err = 0;
5903 
5904 	return err;
5905 }
5906 
5907 /**
5908  * ice_fdb_del - delete an entry from the hardware database
5909  * @ndm: the input from the stack
5910  * @tb: pointer to array of nladdr (unused)
5911  * @dev: the net device pointer
5912  * @addr: the MAC address entry being added
5913  * @vid: VLAN ID
5914  * @extack: netlink extended ack
5915  */
5916 static int
5917 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5918 	    struct net_device *dev, const unsigned char *addr,
5919 	    __always_unused u16 vid, struct netlink_ext_ack *extack)
5920 {
5921 	int err;
5922 
5923 	if (ndm->ndm_state & NUD_PERMANENT) {
5924 		netdev_err(dev, "FDB only supports static addresses\n");
5925 		return -EINVAL;
5926 	}
5927 
5928 	if (is_unicast_ether_addr(addr))
5929 		err = dev_uc_del(dev, addr);
5930 	else if (is_multicast_ether_addr(addr))
5931 		err = dev_mc_del(dev, addr);
5932 	else
5933 		err = -EINVAL;
5934 
5935 	return err;
5936 }
5937 
5938 #define NETIF_VLAN_OFFLOAD_FEATURES	(NETIF_F_HW_VLAN_CTAG_RX | \
5939 					 NETIF_F_HW_VLAN_CTAG_TX | \
5940 					 NETIF_F_HW_VLAN_STAG_RX | \
5941 					 NETIF_F_HW_VLAN_STAG_TX)
5942 
5943 #define NETIF_VLAN_STRIPPING_FEATURES	(NETIF_F_HW_VLAN_CTAG_RX | \
5944 					 NETIF_F_HW_VLAN_STAG_RX)
5945 
5946 #define NETIF_VLAN_FILTERING_FEATURES	(NETIF_F_HW_VLAN_CTAG_FILTER | \
5947 					 NETIF_F_HW_VLAN_STAG_FILTER)
5948 
5949 /**
5950  * ice_fix_features - fix the netdev features flags based on device limitations
5951  * @netdev: ptr to the netdev that flags are being fixed on
5952  * @features: features that need to be checked and possibly fixed
5953  *
5954  * Make sure any fixups are made to features in this callback. This enables the
5955  * driver to not have to check unsupported configurations throughout the driver
5956  * because that's the responsiblity of this callback.
5957  *
5958  * Single VLAN Mode (SVM) Supported Features:
5959  *	NETIF_F_HW_VLAN_CTAG_FILTER
5960  *	NETIF_F_HW_VLAN_CTAG_RX
5961  *	NETIF_F_HW_VLAN_CTAG_TX
5962  *
5963  * Double VLAN Mode (DVM) Supported Features:
5964  *	NETIF_F_HW_VLAN_CTAG_FILTER
5965  *	NETIF_F_HW_VLAN_CTAG_RX
5966  *	NETIF_F_HW_VLAN_CTAG_TX
5967  *
5968  *	NETIF_F_HW_VLAN_STAG_FILTER
5969  *	NETIF_HW_VLAN_STAG_RX
5970  *	NETIF_HW_VLAN_STAG_TX
5971  *
5972  * Features that need fixing:
5973  *	Cannot simultaneously enable CTAG and STAG stripping and/or insertion.
5974  *	These are mutually exlusive as the VSI context cannot support multiple
5975  *	VLAN ethertypes simultaneously for stripping and/or insertion. If this
5976  *	is not done, then default to clearing the requested STAG offload
5977  *	settings.
5978  *
5979  *	All supported filtering has to be enabled or disabled together. For
5980  *	example, in DVM, CTAG and STAG filtering have to be enabled and disabled
5981  *	together. If this is not done, then default to VLAN filtering disabled.
5982  *	These are mutually exclusive as there is currently no way to
5983  *	enable/disable VLAN filtering based on VLAN ethertype when using VLAN
5984  *	prune rules.
5985  */
5986 static netdev_features_t
5987 ice_fix_features(struct net_device *netdev, netdev_features_t features)
5988 {
5989 	struct ice_netdev_priv *np = netdev_priv(netdev);
5990 	netdev_features_t req_vlan_fltr, cur_vlan_fltr;
5991 	bool cur_ctag, cur_stag, req_ctag, req_stag;
5992 
5993 	cur_vlan_fltr = netdev->features & NETIF_VLAN_FILTERING_FEATURES;
5994 	cur_ctag = cur_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
5995 	cur_stag = cur_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
5996 
5997 	req_vlan_fltr = features & NETIF_VLAN_FILTERING_FEATURES;
5998 	req_ctag = req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
5999 	req_stag = req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
6000 
6001 	if (req_vlan_fltr != cur_vlan_fltr) {
6002 		if (ice_is_dvm_ena(&np->vsi->back->hw)) {
6003 			if (req_ctag && req_stag) {
6004 				features |= NETIF_VLAN_FILTERING_FEATURES;
6005 			} else if (!req_ctag && !req_stag) {
6006 				features &= ~NETIF_VLAN_FILTERING_FEATURES;
6007 			} else if ((!cur_ctag && req_ctag && !cur_stag) ||
6008 				   (!cur_stag && req_stag && !cur_ctag)) {
6009 				features |= NETIF_VLAN_FILTERING_FEATURES;
6010 				netdev_warn(netdev,  "802.1Q and 802.1ad VLAN filtering must be either both on or both off. VLAN filtering has been enabled for both types.\n");
6011 			} else if ((cur_ctag && !req_ctag && cur_stag) ||
6012 				   (cur_stag && !req_stag && cur_ctag)) {
6013 				features &= ~NETIF_VLAN_FILTERING_FEATURES;
6014 				netdev_warn(netdev,  "802.1Q and 802.1ad VLAN filtering must be either both on or both off. VLAN filtering has been disabled for both types.\n");
6015 			}
6016 		} else {
6017 			if (req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER)
6018 				netdev_warn(netdev, "cannot support requested 802.1ad filtering setting in SVM mode\n");
6019 
6020 			if (req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER)
6021 				features |= NETIF_F_HW_VLAN_CTAG_FILTER;
6022 		}
6023 	}
6024 
6025 	if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) &&
6026 	    (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) {
6027 		netdev_warn(netdev, "cannot support CTAG and STAG VLAN stripping and/or insertion simultaneously since CTAG and STAG offloads are mutually exclusive, clearing STAG offload settings\n");
6028 		features &= ~(NETIF_F_HW_VLAN_STAG_RX |
6029 			      NETIF_F_HW_VLAN_STAG_TX);
6030 	}
6031 
6032 	if (!(netdev->features & NETIF_F_RXFCS) &&
6033 	    (features & NETIF_F_RXFCS) &&
6034 	    (features & NETIF_VLAN_STRIPPING_FEATURES) &&
6035 	    !ice_vsi_has_non_zero_vlans(np->vsi)) {
6036 		netdev_warn(netdev, "Disabling VLAN stripping as FCS/CRC stripping is also disabled and there is no VLAN configured\n");
6037 		features &= ~NETIF_VLAN_STRIPPING_FEATURES;
6038 	}
6039 
6040 	return features;
6041 }
6042 
6043 /**
6044  * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI
6045  * @vsi: PF's VSI
6046  * @features: features used to determine VLAN offload settings
6047  *
6048  * First, determine the vlan_ethertype based on the VLAN offload bits in
6049  * features. Then determine if stripping and insertion should be enabled or
6050  * disabled. Finally enable or disable VLAN stripping and insertion.
6051  */
6052 static int
6053 ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features)
6054 {
6055 	bool enable_stripping = true, enable_insertion = true;
6056 	struct ice_vsi_vlan_ops *vlan_ops;
6057 	int strip_err = 0, insert_err = 0;
6058 	u16 vlan_ethertype = 0;
6059 
6060 	vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6061 
6062 	if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))
6063 		vlan_ethertype = ETH_P_8021AD;
6064 	else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX))
6065 		vlan_ethertype = ETH_P_8021Q;
6066 
6067 	if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX)))
6068 		enable_stripping = false;
6069 	if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX)))
6070 		enable_insertion = false;
6071 
6072 	if (enable_stripping)
6073 		strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype);
6074 	else
6075 		strip_err = vlan_ops->dis_stripping(vsi);
6076 
6077 	if (enable_insertion)
6078 		insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype);
6079 	else
6080 		insert_err = vlan_ops->dis_insertion(vsi);
6081 
6082 	if (strip_err || insert_err)
6083 		return -EIO;
6084 
6085 	return 0;
6086 }
6087 
6088 /**
6089  * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI
6090  * @vsi: PF's VSI
6091  * @features: features used to determine VLAN filtering settings
6092  *
6093  * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the
6094  * features.
6095  */
6096 static int
6097 ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features)
6098 {
6099 	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6100 	int err = 0;
6101 
6102 	/* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking
6103 	 * if either bit is set
6104 	 */
6105 	if (features &
6106 	    (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER))
6107 		err = vlan_ops->ena_rx_filtering(vsi);
6108 	else
6109 		err = vlan_ops->dis_rx_filtering(vsi);
6110 
6111 	return err;
6112 }
6113 
6114 /**
6115  * ice_set_vlan_features - set VLAN settings based on suggested feature set
6116  * @netdev: ptr to the netdev being adjusted
6117  * @features: the feature set that the stack is suggesting
6118  *
6119  * Only update VLAN settings if the requested_vlan_features are different than
6120  * the current_vlan_features.
6121  */
6122 static int
6123 ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)
6124 {
6125 	netdev_features_t current_vlan_features, requested_vlan_features;
6126 	struct ice_netdev_priv *np = netdev_priv(netdev);
6127 	struct ice_vsi *vsi = np->vsi;
6128 	int err;
6129 
6130 	current_vlan_features = netdev->features & NETIF_VLAN_OFFLOAD_FEATURES;
6131 	requested_vlan_features = features & NETIF_VLAN_OFFLOAD_FEATURES;
6132 	if (current_vlan_features ^ requested_vlan_features) {
6133 		if ((features & NETIF_F_RXFCS) &&
6134 		    (features & NETIF_VLAN_STRIPPING_FEATURES)) {
6135 			dev_err(ice_pf_to_dev(vsi->back),
6136 				"To enable VLAN stripping, you must first enable FCS/CRC stripping\n");
6137 			return -EIO;
6138 		}
6139 
6140 		err = ice_set_vlan_offload_features(vsi, features);
6141 		if (err)
6142 			return err;
6143 	}
6144 
6145 	current_vlan_features = netdev->features &
6146 		NETIF_VLAN_FILTERING_FEATURES;
6147 	requested_vlan_features = features & NETIF_VLAN_FILTERING_FEATURES;
6148 	if (current_vlan_features ^ requested_vlan_features) {
6149 		err = ice_set_vlan_filtering_features(vsi, features);
6150 		if (err)
6151 			return err;
6152 	}
6153 
6154 	return 0;
6155 }
6156 
6157 /**
6158  * ice_set_loopback - turn on/off loopback mode on underlying PF
6159  * @vsi: ptr to VSI
6160  * @ena: flag to indicate the on/off setting
6161  */
6162 static int ice_set_loopback(struct ice_vsi *vsi, bool ena)
6163 {
6164 	bool if_running = netif_running(vsi->netdev);
6165 	int ret;
6166 
6167 	if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6168 		ret = ice_down(vsi);
6169 		if (ret) {
6170 			netdev_err(vsi->netdev, "Preparing device to toggle loopback failed\n");
6171 			return ret;
6172 		}
6173 	}
6174 	ret = ice_aq_set_mac_loopback(&vsi->back->hw, ena, NULL);
6175 	if (ret)
6176 		netdev_err(vsi->netdev, "Failed to toggle loopback state\n");
6177 	if (if_running)
6178 		ret = ice_up(vsi);
6179 
6180 	return ret;
6181 }
6182 
6183 /**
6184  * ice_set_features - set the netdev feature flags
6185  * @netdev: ptr to the netdev being adjusted
6186  * @features: the feature set that the stack is suggesting
6187  */
6188 static int
6189 ice_set_features(struct net_device *netdev, netdev_features_t features)
6190 {
6191 	netdev_features_t changed = netdev->features ^ features;
6192 	struct ice_netdev_priv *np = netdev_priv(netdev);
6193 	struct ice_vsi *vsi = np->vsi;
6194 	struct ice_pf *pf = vsi->back;
6195 	int ret = 0;
6196 
6197 	/* Don't set any netdev advanced features with device in Safe Mode */
6198 	if (ice_is_safe_mode(pf)) {
6199 		dev_err(ice_pf_to_dev(pf),
6200 			"Device is in Safe Mode - not enabling advanced netdev features\n");
6201 		return ret;
6202 	}
6203 
6204 	/* Do not change setting during reset */
6205 	if (ice_is_reset_in_progress(pf->state)) {
6206 		dev_err(ice_pf_to_dev(pf),
6207 			"Device is resetting, changing advanced netdev features temporarily unavailable.\n");
6208 		return -EBUSY;
6209 	}
6210 
6211 	/* Multiple features can be changed in one call so keep features in
6212 	 * separate if/else statements to guarantee each feature is checked
6213 	 */
6214 	if (changed & NETIF_F_RXHASH)
6215 		ice_vsi_manage_rss_lut(vsi, !!(features & NETIF_F_RXHASH));
6216 
6217 	ret = ice_set_vlan_features(netdev, features);
6218 	if (ret)
6219 		return ret;
6220 
6221 	/* Turn on receive of FCS aka CRC, and after setting this
6222 	 * flag the packet data will have the 4 byte CRC appended
6223 	 */
6224 	if (changed & NETIF_F_RXFCS) {
6225 		if ((features & NETIF_F_RXFCS) &&
6226 		    (features & NETIF_VLAN_STRIPPING_FEATURES)) {
6227 			dev_err(ice_pf_to_dev(vsi->back),
6228 				"To disable FCS/CRC stripping, you must first disable VLAN stripping\n");
6229 			return -EIO;
6230 		}
6231 
6232 		ice_vsi_cfg_crc_strip(vsi, !!(features & NETIF_F_RXFCS));
6233 		ret = ice_down_up(vsi);
6234 		if (ret)
6235 			return ret;
6236 	}
6237 
6238 	if (changed & NETIF_F_NTUPLE) {
6239 		bool ena = !!(features & NETIF_F_NTUPLE);
6240 
6241 		ice_vsi_manage_fdir(vsi, ena);
6242 		ena ? ice_init_arfs(vsi) : ice_clear_arfs(vsi);
6243 	}
6244 
6245 	/* don't turn off hw_tc_offload when ADQ is already enabled */
6246 	if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) {
6247 		dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n");
6248 		return -EACCES;
6249 	}
6250 
6251 	if (changed & NETIF_F_HW_TC) {
6252 		bool ena = !!(features & NETIF_F_HW_TC);
6253 
6254 		ena ? set_bit(ICE_FLAG_CLS_FLOWER, pf->flags) :
6255 		      clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
6256 	}
6257 
6258 	if (changed & NETIF_F_LOOPBACK)
6259 		ret = ice_set_loopback(vsi, !!(features & NETIF_F_LOOPBACK));
6260 
6261 	return ret;
6262 }
6263 
6264 /**
6265  * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI
6266  * @vsi: VSI to setup VLAN properties for
6267  */
6268 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
6269 {
6270 	int err;
6271 
6272 	err = ice_set_vlan_offload_features(vsi, vsi->netdev->features);
6273 	if (err)
6274 		return err;
6275 
6276 	err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features);
6277 	if (err)
6278 		return err;
6279 
6280 	return ice_vsi_add_vlan_zero(vsi);
6281 }
6282 
6283 /**
6284  * ice_vsi_cfg_lan - Setup the VSI lan related config
6285  * @vsi: the VSI being configured
6286  *
6287  * Return 0 on success and negative value on error
6288  */
6289 int ice_vsi_cfg_lan(struct ice_vsi *vsi)
6290 {
6291 	int err;
6292 
6293 	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6294 		ice_set_rx_mode(vsi->netdev);
6295 
6296 		err = ice_vsi_vlan_setup(vsi);
6297 		if (err)
6298 			return err;
6299 	}
6300 	ice_vsi_cfg_dcb_rings(vsi);
6301 
6302 	err = ice_vsi_cfg_lan_txqs(vsi);
6303 	if (!err && ice_is_xdp_ena_vsi(vsi))
6304 		err = ice_vsi_cfg_xdp_txqs(vsi);
6305 	if (!err)
6306 		err = ice_vsi_cfg_rxqs(vsi);
6307 
6308 	return err;
6309 }
6310 
6311 /* THEORY OF MODERATION:
6312  * The ice driver hardware works differently than the hardware that DIMLIB was
6313  * originally made for. ice hardware doesn't have packet count limits that
6314  * can trigger an interrupt, but it *does* have interrupt rate limit support,
6315  * which is hard-coded to a limit of 250,000 ints/second.
6316  * If not using dynamic moderation, the INTRL value can be modified
6317  * by ethtool rx-usecs-high.
6318  */
6319 struct ice_dim {
6320 	/* the throttle rate for interrupts, basically worst case delay before
6321 	 * an initial interrupt fires, value is stored in microseconds.
6322 	 */
6323 	u16 itr;
6324 };
6325 
6326 /* Make a different profile for Rx that doesn't allow quite so aggressive
6327  * moderation at the high end (it maxes out at 126us or about 8k interrupts a
6328  * second.
6329  */
6330 static const struct ice_dim rx_profile[] = {
6331 	{2},    /* 500,000 ints/s, capped at 250K by INTRL */
6332 	{8},    /* 125,000 ints/s */
6333 	{16},   /*  62,500 ints/s */
6334 	{62},   /*  16,129 ints/s */
6335 	{126}   /*   7,936 ints/s */
6336 };
6337 
6338 /* The transmit profile, which has the same sorts of values
6339  * as the previous struct
6340  */
6341 static const struct ice_dim tx_profile[] = {
6342 	{2},    /* 500,000 ints/s, capped at 250K by INTRL */
6343 	{8},    /* 125,000 ints/s */
6344 	{40},   /*  16,125 ints/s */
6345 	{128},  /*   7,812 ints/s */
6346 	{256}   /*   3,906 ints/s */
6347 };
6348 
6349 static void ice_tx_dim_work(struct work_struct *work)
6350 {
6351 	struct ice_ring_container *rc;
6352 	struct dim *dim;
6353 	u16 itr;
6354 
6355 	dim = container_of(work, struct dim, work);
6356 	rc = dim->priv;
6357 
6358 	WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
6359 
6360 	/* look up the values in our local table */
6361 	itr = tx_profile[dim->profile_ix].itr;
6362 
6363 	ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
6364 	ice_write_itr(rc, itr);
6365 
6366 	dim->state = DIM_START_MEASURE;
6367 }
6368 
6369 static void ice_rx_dim_work(struct work_struct *work)
6370 {
6371 	struct ice_ring_container *rc;
6372 	struct dim *dim;
6373 	u16 itr;
6374 
6375 	dim = container_of(work, struct dim, work);
6376 	rc = dim->priv;
6377 
6378 	WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
6379 
6380 	/* look up the values in our local table */
6381 	itr = rx_profile[dim->profile_ix].itr;
6382 
6383 	ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
6384 	ice_write_itr(rc, itr);
6385 
6386 	dim->state = DIM_START_MEASURE;
6387 }
6388 
6389 #define ICE_DIM_DEFAULT_PROFILE_IX 1
6390 
6391 /**
6392  * ice_init_moderation - set up interrupt moderation
6393  * @q_vector: the vector containing rings to be configured
6394  *
6395  * Set up interrupt moderation registers, with the intent to do the right thing
6396  * when called from reset or from probe, and whether or not dynamic moderation
6397  * is enabled or not. Take special care to write all the registers in both
6398  * dynamic moderation mode or not in order to make sure hardware is in a known
6399  * state.
6400  */
6401 static void ice_init_moderation(struct ice_q_vector *q_vector)
6402 {
6403 	struct ice_ring_container *rc;
6404 	bool tx_dynamic, rx_dynamic;
6405 
6406 	rc = &q_vector->tx;
6407 	INIT_WORK(&rc->dim.work, ice_tx_dim_work);
6408 	rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6409 	rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6410 	rc->dim.priv = rc;
6411 	tx_dynamic = ITR_IS_DYNAMIC(rc);
6412 
6413 	/* set the initial TX ITR to match the above */
6414 	ice_write_itr(rc, tx_dynamic ?
6415 		      tx_profile[rc->dim.profile_ix].itr : rc->itr_setting);
6416 
6417 	rc = &q_vector->rx;
6418 	INIT_WORK(&rc->dim.work, ice_rx_dim_work);
6419 	rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6420 	rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6421 	rc->dim.priv = rc;
6422 	rx_dynamic = ITR_IS_DYNAMIC(rc);
6423 
6424 	/* set the initial RX ITR to match the above */
6425 	ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
6426 				       rc->itr_setting);
6427 
6428 	ice_set_q_vector_intrl(q_vector);
6429 }
6430 
6431 /**
6432  * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
6433  * @vsi: the VSI being configured
6434  */
6435 static void ice_napi_enable_all(struct ice_vsi *vsi)
6436 {
6437 	int q_idx;
6438 
6439 	if (!vsi->netdev)
6440 		return;
6441 
6442 	ice_for_each_q_vector(vsi, q_idx) {
6443 		struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6444 
6445 		ice_init_moderation(q_vector);
6446 
6447 		if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6448 			napi_enable(&q_vector->napi);
6449 	}
6450 }
6451 
6452 /**
6453  * ice_up_complete - Finish the last steps of bringing up a connection
6454  * @vsi: The VSI being configured
6455  *
6456  * Return 0 on success and negative value on error
6457  */
6458 static int ice_up_complete(struct ice_vsi *vsi)
6459 {
6460 	struct ice_pf *pf = vsi->back;
6461 	int err;
6462 
6463 	ice_vsi_cfg_msix(vsi);
6464 
6465 	/* Enable only Rx rings, Tx rings were enabled by the FW when the
6466 	 * Tx queue group list was configured and the context bits were
6467 	 * programmed using ice_vsi_cfg_txqs
6468 	 */
6469 	err = ice_vsi_start_all_rx_rings(vsi);
6470 	if (err)
6471 		return err;
6472 
6473 	clear_bit(ICE_VSI_DOWN, vsi->state);
6474 	ice_napi_enable_all(vsi);
6475 	ice_vsi_ena_irq(vsi);
6476 
6477 	if (vsi->port_info &&
6478 	    (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
6479 	    vsi->netdev && vsi->type == ICE_VSI_PF) {
6480 		ice_print_link_msg(vsi, true);
6481 		netif_tx_start_all_queues(vsi->netdev);
6482 		netif_carrier_on(vsi->netdev);
6483 		ice_ptp_link_change(pf, pf->hw.pf_id, true);
6484 	}
6485 
6486 	/* Perform an initial read of the statistics registers now to
6487 	 * set the baseline so counters are ready when interface is up
6488 	 */
6489 	ice_update_eth_stats(vsi);
6490 
6491 	if (vsi->type == ICE_VSI_PF)
6492 		ice_service_task_schedule(pf);
6493 
6494 	return 0;
6495 }
6496 
6497 /**
6498  * ice_up - Bring the connection back up after being down
6499  * @vsi: VSI being configured
6500  */
6501 int ice_up(struct ice_vsi *vsi)
6502 {
6503 	int err;
6504 
6505 	err = ice_vsi_cfg_lan(vsi);
6506 	if (!err)
6507 		err = ice_up_complete(vsi);
6508 
6509 	return err;
6510 }
6511 
6512 /**
6513  * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
6514  * @syncp: pointer to u64_stats_sync
6515  * @stats: stats that pkts and bytes count will be taken from
6516  * @pkts: packets stats counter
6517  * @bytes: bytes stats counter
6518  *
6519  * This function fetches stats from the ring considering the atomic operations
6520  * that needs to be performed to read u64 values in 32 bit machine.
6521  */
6522 void
6523 ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp,
6524 			     struct ice_q_stats stats, u64 *pkts, u64 *bytes)
6525 {
6526 	unsigned int start;
6527 
6528 	do {
6529 		start = u64_stats_fetch_begin(syncp);
6530 		*pkts = stats.pkts;
6531 		*bytes = stats.bytes;
6532 	} while (u64_stats_fetch_retry(syncp, start));
6533 }
6534 
6535 /**
6536  * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
6537  * @vsi: the VSI to be updated
6538  * @vsi_stats: the stats struct to be updated
6539  * @rings: rings to work on
6540  * @count: number of rings
6541  */
6542 static void
6543 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
6544 			     struct rtnl_link_stats64 *vsi_stats,
6545 			     struct ice_tx_ring **rings, u16 count)
6546 {
6547 	u16 i;
6548 
6549 	for (i = 0; i < count; i++) {
6550 		struct ice_tx_ring *ring;
6551 		u64 pkts = 0, bytes = 0;
6552 
6553 		ring = READ_ONCE(rings[i]);
6554 		if (!ring || !ring->ring_stats)
6555 			continue;
6556 		ice_fetch_u64_stats_per_ring(&ring->ring_stats->syncp,
6557 					     ring->ring_stats->stats, &pkts,
6558 					     &bytes);
6559 		vsi_stats->tx_packets += pkts;
6560 		vsi_stats->tx_bytes += bytes;
6561 		vsi->tx_restart += ring->ring_stats->tx_stats.restart_q;
6562 		vsi->tx_busy += ring->ring_stats->tx_stats.tx_busy;
6563 		vsi->tx_linearize += ring->ring_stats->tx_stats.tx_linearize;
6564 	}
6565 }
6566 
6567 /**
6568  * ice_update_vsi_ring_stats - Update VSI stats counters
6569  * @vsi: the VSI to be updated
6570  */
6571 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
6572 {
6573 	struct rtnl_link_stats64 *net_stats, *stats_prev;
6574 	struct rtnl_link_stats64 *vsi_stats;
6575 	u64 pkts, bytes;
6576 	int i;
6577 
6578 	vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
6579 	if (!vsi_stats)
6580 		return;
6581 
6582 	/* reset non-netdev (extended) stats */
6583 	vsi->tx_restart = 0;
6584 	vsi->tx_busy = 0;
6585 	vsi->tx_linearize = 0;
6586 	vsi->rx_buf_failed = 0;
6587 	vsi->rx_page_failed = 0;
6588 
6589 	rcu_read_lock();
6590 
6591 	/* update Tx rings counters */
6592 	ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6593 				     vsi->num_txq);
6594 
6595 	/* update Rx rings counters */
6596 	ice_for_each_rxq(vsi, i) {
6597 		struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6598 		struct ice_ring_stats *ring_stats;
6599 
6600 		ring_stats = ring->ring_stats;
6601 		ice_fetch_u64_stats_per_ring(&ring_stats->syncp,
6602 					     ring_stats->stats, &pkts,
6603 					     &bytes);
6604 		vsi_stats->rx_packets += pkts;
6605 		vsi_stats->rx_bytes += bytes;
6606 		vsi->rx_buf_failed += ring_stats->rx_stats.alloc_buf_failed;
6607 		vsi->rx_page_failed += ring_stats->rx_stats.alloc_page_failed;
6608 	}
6609 
6610 	/* update XDP Tx rings counters */
6611 	if (ice_is_xdp_ena_vsi(vsi))
6612 		ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6613 					     vsi->num_xdp_txq);
6614 
6615 	rcu_read_unlock();
6616 
6617 	net_stats = &vsi->net_stats;
6618 	stats_prev = &vsi->net_stats_prev;
6619 
6620 	/* clear prev counters after reset */
6621 	if (vsi_stats->tx_packets < stats_prev->tx_packets ||
6622 	    vsi_stats->rx_packets < stats_prev->rx_packets) {
6623 		stats_prev->tx_packets = 0;
6624 		stats_prev->tx_bytes = 0;
6625 		stats_prev->rx_packets = 0;
6626 		stats_prev->rx_bytes = 0;
6627 	}
6628 
6629 	/* update netdev counters */
6630 	net_stats->tx_packets += vsi_stats->tx_packets - stats_prev->tx_packets;
6631 	net_stats->tx_bytes += vsi_stats->tx_bytes - stats_prev->tx_bytes;
6632 	net_stats->rx_packets += vsi_stats->rx_packets - stats_prev->rx_packets;
6633 	net_stats->rx_bytes += vsi_stats->rx_bytes - stats_prev->rx_bytes;
6634 
6635 	stats_prev->tx_packets = vsi_stats->tx_packets;
6636 	stats_prev->tx_bytes = vsi_stats->tx_bytes;
6637 	stats_prev->rx_packets = vsi_stats->rx_packets;
6638 	stats_prev->rx_bytes = vsi_stats->rx_bytes;
6639 
6640 	kfree(vsi_stats);
6641 }
6642 
6643 /**
6644  * ice_update_vsi_stats - Update VSI stats counters
6645  * @vsi: the VSI to be updated
6646  */
6647 void ice_update_vsi_stats(struct ice_vsi *vsi)
6648 {
6649 	struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6650 	struct ice_eth_stats *cur_es = &vsi->eth_stats;
6651 	struct ice_pf *pf = vsi->back;
6652 
6653 	if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6654 	    test_bit(ICE_CFG_BUSY, pf->state))
6655 		return;
6656 
6657 	/* get stats as recorded by Tx/Rx rings */
6658 	ice_update_vsi_ring_stats(vsi);
6659 
6660 	/* get VSI stats as recorded by the hardware */
6661 	ice_update_eth_stats(vsi);
6662 
6663 	cur_ns->tx_errors = cur_es->tx_errors;
6664 	cur_ns->rx_dropped = cur_es->rx_discards;
6665 	cur_ns->tx_dropped = cur_es->tx_discards;
6666 	cur_ns->multicast = cur_es->rx_multicast;
6667 
6668 	/* update some more netdev stats if this is main VSI */
6669 	if (vsi->type == ICE_VSI_PF) {
6670 		cur_ns->rx_crc_errors = pf->stats.crc_errors;
6671 		cur_ns->rx_errors = pf->stats.crc_errors +
6672 				    pf->stats.illegal_bytes +
6673 				    pf->stats.rx_len_errors +
6674 				    pf->stats.rx_undersize +
6675 				    pf->hw_csum_rx_error +
6676 				    pf->stats.rx_jabber +
6677 				    pf->stats.rx_fragments +
6678 				    pf->stats.rx_oversize;
6679 		cur_ns->rx_length_errors = pf->stats.rx_len_errors;
6680 		/* record drops from the port level */
6681 		cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6682 	}
6683 }
6684 
6685 /**
6686  * ice_update_pf_stats - Update PF port stats counters
6687  * @pf: PF whose stats needs to be updated
6688  */
6689 void ice_update_pf_stats(struct ice_pf *pf)
6690 {
6691 	struct ice_hw_port_stats *prev_ps, *cur_ps;
6692 	struct ice_hw *hw = &pf->hw;
6693 	u16 fd_ctr_base;
6694 	u8 port;
6695 
6696 	port = hw->port_info->lport;
6697 	prev_ps = &pf->stats_prev;
6698 	cur_ps = &pf->stats;
6699 
6700 	if (ice_is_reset_in_progress(pf->state))
6701 		pf->stat_prev_loaded = false;
6702 
6703 	ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6704 			  &prev_ps->eth.rx_bytes,
6705 			  &cur_ps->eth.rx_bytes);
6706 
6707 	ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6708 			  &prev_ps->eth.rx_unicast,
6709 			  &cur_ps->eth.rx_unicast);
6710 
6711 	ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6712 			  &prev_ps->eth.rx_multicast,
6713 			  &cur_ps->eth.rx_multicast);
6714 
6715 	ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6716 			  &prev_ps->eth.rx_broadcast,
6717 			  &cur_ps->eth.rx_broadcast);
6718 
6719 	ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6720 			  &prev_ps->eth.rx_discards,
6721 			  &cur_ps->eth.rx_discards);
6722 
6723 	ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6724 			  &prev_ps->eth.tx_bytes,
6725 			  &cur_ps->eth.tx_bytes);
6726 
6727 	ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6728 			  &prev_ps->eth.tx_unicast,
6729 			  &cur_ps->eth.tx_unicast);
6730 
6731 	ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6732 			  &prev_ps->eth.tx_multicast,
6733 			  &cur_ps->eth.tx_multicast);
6734 
6735 	ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6736 			  &prev_ps->eth.tx_broadcast,
6737 			  &cur_ps->eth.tx_broadcast);
6738 
6739 	ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6740 			  &prev_ps->tx_dropped_link_down,
6741 			  &cur_ps->tx_dropped_link_down);
6742 
6743 	ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6744 			  &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6745 
6746 	ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6747 			  &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6748 
6749 	ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6750 			  &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6751 
6752 	ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6753 			  &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6754 
6755 	ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6756 			  &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6757 
6758 	ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6759 			  &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6760 
6761 	ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6762 			  &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6763 
6764 	ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6765 			  &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6766 
6767 	ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6768 			  &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6769 
6770 	ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6771 			  &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6772 
6773 	ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6774 			  &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6775 
6776 	ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6777 			  &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6778 
6779 	ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6780 			  &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6781 
6782 	ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6783 			  &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6784 
6785 	fd_ctr_base = hw->fd_ctr_base;
6786 
6787 	ice_stat_update40(hw,
6788 			  GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6789 			  pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6790 			  &cur_ps->fd_sb_match);
6791 	ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6792 			  &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6793 
6794 	ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6795 			  &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6796 
6797 	ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6798 			  &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6799 
6800 	ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6801 			  &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6802 
6803 	ice_update_dcb_stats(pf);
6804 
6805 	ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6806 			  &prev_ps->crc_errors, &cur_ps->crc_errors);
6807 
6808 	ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6809 			  &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6810 
6811 	ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6812 			  &prev_ps->mac_local_faults,
6813 			  &cur_ps->mac_local_faults);
6814 
6815 	ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6816 			  &prev_ps->mac_remote_faults,
6817 			  &cur_ps->mac_remote_faults);
6818 
6819 	ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
6820 			  &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
6821 
6822 	ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6823 			  &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6824 
6825 	ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6826 			  &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6827 
6828 	ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6829 			  &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6830 
6831 	ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6832 			  &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6833 
6834 	cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6835 
6836 	pf->stat_prev_loaded = true;
6837 }
6838 
6839 /**
6840  * ice_get_stats64 - get statistics for network device structure
6841  * @netdev: network interface device structure
6842  * @stats: main device statistics structure
6843  */
6844 static
6845 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
6846 {
6847 	struct ice_netdev_priv *np = netdev_priv(netdev);
6848 	struct rtnl_link_stats64 *vsi_stats;
6849 	struct ice_vsi *vsi = np->vsi;
6850 
6851 	vsi_stats = &vsi->net_stats;
6852 
6853 	if (!vsi->num_txq || !vsi->num_rxq)
6854 		return;
6855 
6856 	/* netdev packet/byte stats come from ring counter. These are obtained
6857 	 * by summing up ring counters (done by ice_update_vsi_ring_stats).
6858 	 * But, only call the update routine and read the registers if VSI is
6859 	 * not down.
6860 	 */
6861 	if (!test_bit(ICE_VSI_DOWN, vsi->state))
6862 		ice_update_vsi_ring_stats(vsi);
6863 	stats->tx_packets = vsi_stats->tx_packets;
6864 	stats->tx_bytes = vsi_stats->tx_bytes;
6865 	stats->rx_packets = vsi_stats->rx_packets;
6866 	stats->rx_bytes = vsi_stats->rx_bytes;
6867 
6868 	/* The rest of the stats can be read from the hardware but instead we
6869 	 * just return values that the watchdog task has already obtained from
6870 	 * the hardware.
6871 	 */
6872 	stats->multicast = vsi_stats->multicast;
6873 	stats->tx_errors = vsi_stats->tx_errors;
6874 	stats->tx_dropped = vsi_stats->tx_dropped;
6875 	stats->rx_errors = vsi_stats->rx_errors;
6876 	stats->rx_dropped = vsi_stats->rx_dropped;
6877 	stats->rx_crc_errors = vsi_stats->rx_crc_errors;
6878 	stats->rx_length_errors = vsi_stats->rx_length_errors;
6879 }
6880 
6881 /**
6882  * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
6883  * @vsi: VSI having NAPI disabled
6884  */
6885 static void ice_napi_disable_all(struct ice_vsi *vsi)
6886 {
6887 	int q_idx;
6888 
6889 	if (!vsi->netdev)
6890 		return;
6891 
6892 	ice_for_each_q_vector(vsi, q_idx) {
6893 		struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6894 
6895 		if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6896 			napi_disable(&q_vector->napi);
6897 
6898 		cancel_work_sync(&q_vector->tx.dim.work);
6899 		cancel_work_sync(&q_vector->rx.dim.work);
6900 	}
6901 }
6902 
6903 /**
6904  * ice_down - Shutdown the connection
6905  * @vsi: The VSI being stopped
6906  *
6907  * Caller of this function is expected to set the vsi->state ICE_DOWN bit
6908  */
6909 int ice_down(struct ice_vsi *vsi)
6910 {
6911 	int i, tx_err, rx_err, vlan_err = 0;
6912 
6913 	WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
6914 
6915 	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6916 		vlan_err = ice_vsi_del_vlan_zero(vsi);
6917 		ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
6918 		netif_carrier_off(vsi->netdev);
6919 		netif_tx_disable(vsi->netdev);
6920 	} else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
6921 		ice_eswitch_stop_all_tx_queues(vsi->back);
6922 	}
6923 
6924 	ice_vsi_dis_irq(vsi);
6925 
6926 	tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
6927 	if (tx_err)
6928 		netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
6929 			   vsi->vsi_num, tx_err);
6930 	if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
6931 		tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
6932 		if (tx_err)
6933 			netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
6934 				   vsi->vsi_num, tx_err);
6935 	}
6936 
6937 	rx_err = ice_vsi_stop_all_rx_rings(vsi);
6938 	if (rx_err)
6939 		netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
6940 			   vsi->vsi_num, rx_err);
6941 
6942 	ice_napi_disable_all(vsi);
6943 
6944 	ice_for_each_txq(vsi, i)
6945 		ice_clean_tx_ring(vsi->tx_rings[i]);
6946 
6947 	if (ice_is_xdp_ena_vsi(vsi))
6948 		ice_for_each_xdp_txq(vsi, i)
6949 			ice_clean_tx_ring(vsi->xdp_rings[i]);
6950 
6951 	ice_for_each_rxq(vsi, i)
6952 		ice_clean_rx_ring(vsi->rx_rings[i]);
6953 
6954 	if (tx_err || rx_err || vlan_err) {
6955 		netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
6956 			   vsi->vsi_num, vsi->vsw->sw_id);
6957 		return -EIO;
6958 	}
6959 
6960 	return 0;
6961 }
6962 
6963 /**
6964  * ice_down_up - shutdown the VSI connection and bring it up
6965  * @vsi: the VSI to be reconnected
6966  */
6967 int ice_down_up(struct ice_vsi *vsi)
6968 {
6969 	int ret;
6970 
6971 	/* if DOWN already set, nothing to do */
6972 	if (test_and_set_bit(ICE_VSI_DOWN, vsi->state))
6973 		return 0;
6974 
6975 	ret = ice_down(vsi);
6976 	if (ret)
6977 		return ret;
6978 
6979 	ret = ice_up(vsi);
6980 	if (ret) {
6981 		netdev_err(vsi->netdev, "reallocating resources failed during netdev features change, may need to reload driver\n");
6982 		return ret;
6983 	}
6984 
6985 	return 0;
6986 }
6987 
6988 /**
6989  * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
6990  * @vsi: VSI having resources allocated
6991  *
6992  * Return 0 on success, negative on failure
6993  */
6994 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
6995 {
6996 	int i, err = 0;
6997 
6998 	if (!vsi->num_txq) {
6999 		dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
7000 			vsi->vsi_num);
7001 		return -EINVAL;
7002 	}
7003 
7004 	ice_for_each_txq(vsi, i) {
7005 		struct ice_tx_ring *ring = vsi->tx_rings[i];
7006 
7007 		if (!ring)
7008 			return -EINVAL;
7009 
7010 		if (vsi->netdev)
7011 			ring->netdev = vsi->netdev;
7012 		err = ice_setup_tx_ring(ring);
7013 		if (err)
7014 			break;
7015 	}
7016 
7017 	return err;
7018 }
7019 
7020 /**
7021  * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
7022  * @vsi: VSI having resources allocated
7023  *
7024  * Return 0 on success, negative on failure
7025  */
7026 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
7027 {
7028 	int i, err = 0;
7029 
7030 	if (!vsi->num_rxq) {
7031 		dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
7032 			vsi->vsi_num);
7033 		return -EINVAL;
7034 	}
7035 
7036 	ice_for_each_rxq(vsi, i) {
7037 		struct ice_rx_ring *ring = vsi->rx_rings[i];
7038 
7039 		if (!ring)
7040 			return -EINVAL;
7041 
7042 		if (vsi->netdev)
7043 			ring->netdev = vsi->netdev;
7044 		err = ice_setup_rx_ring(ring);
7045 		if (err)
7046 			break;
7047 	}
7048 
7049 	return err;
7050 }
7051 
7052 /**
7053  * ice_vsi_open_ctrl - open control VSI for use
7054  * @vsi: the VSI to open
7055  *
7056  * Initialization of the Control VSI
7057  *
7058  * Returns 0 on success, negative value on error
7059  */
7060 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
7061 {
7062 	char int_name[ICE_INT_NAME_STR_LEN];
7063 	struct ice_pf *pf = vsi->back;
7064 	struct device *dev;
7065 	int err;
7066 
7067 	dev = ice_pf_to_dev(pf);
7068 	/* allocate descriptors */
7069 	err = ice_vsi_setup_tx_rings(vsi);
7070 	if (err)
7071 		goto err_setup_tx;
7072 
7073 	err = ice_vsi_setup_rx_rings(vsi);
7074 	if (err)
7075 		goto err_setup_rx;
7076 
7077 	err = ice_vsi_cfg_lan(vsi);
7078 	if (err)
7079 		goto err_setup_rx;
7080 
7081 	snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
7082 		 dev_driver_string(dev), dev_name(dev));
7083 	err = ice_vsi_req_irq_msix(vsi, int_name);
7084 	if (err)
7085 		goto err_setup_rx;
7086 
7087 	ice_vsi_cfg_msix(vsi);
7088 
7089 	err = ice_vsi_start_all_rx_rings(vsi);
7090 	if (err)
7091 		goto err_up_complete;
7092 
7093 	clear_bit(ICE_VSI_DOWN, vsi->state);
7094 	ice_vsi_ena_irq(vsi);
7095 
7096 	return 0;
7097 
7098 err_up_complete:
7099 	ice_down(vsi);
7100 err_setup_rx:
7101 	ice_vsi_free_rx_rings(vsi);
7102 err_setup_tx:
7103 	ice_vsi_free_tx_rings(vsi);
7104 
7105 	return err;
7106 }
7107 
7108 /**
7109  * ice_vsi_open - Called when a network interface is made active
7110  * @vsi: the VSI to open
7111  *
7112  * Initialization of the VSI
7113  *
7114  * Returns 0 on success, negative value on error
7115  */
7116 int ice_vsi_open(struct ice_vsi *vsi)
7117 {
7118 	char int_name[ICE_INT_NAME_STR_LEN];
7119 	struct ice_pf *pf = vsi->back;
7120 	int err;
7121 
7122 	/* allocate descriptors */
7123 	err = ice_vsi_setup_tx_rings(vsi);
7124 	if (err)
7125 		goto err_setup_tx;
7126 
7127 	err = ice_vsi_setup_rx_rings(vsi);
7128 	if (err)
7129 		goto err_setup_rx;
7130 
7131 	err = ice_vsi_cfg_lan(vsi);
7132 	if (err)
7133 		goto err_setup_rx;
7134 
7135 	snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
7136 		 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
7137 	err = ice_vsi_req_irq_msix(vsi, int_name);
7138 	if (err)
7139 		goto err_setup_rx;
7140 
7141 	ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
7142 
7143 	if (vsi->type == ICE_VSI_PF) {
7144 		/* Notify the stack of the actual queue counts. */
7145 		err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
7146 		if (err)
7147 			goto err_set_qs;
7148 
7149 		err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
7150 		if (err)
7151 			goto err_set_qs;
7152 	}
7153 
7154 	err = ice_up_complete(vsi);
7155 	if (err)
7156 		goto err_up_complete;
7157 
7158 	return 0;
7159 
7160 err_up_complete:
7161 	ice_down(vsi);
7162 err_set_qs:
7163 	ice_vsi_free_irq(vsi);
7164 err_setup_rx:
7165 	ice_vsi_free_rx_rings(vsi);
7166 err_setup_tx:
7167 	ice_vsi_free_tx_rings(vsi);
7168 
7169 	return err;
7170 }
7171 
7172 /**
7173  * ice_vsi_release_all - Delete all VSIs
7174  * @pf: PF from which all VSIs are being removed
7175  */
7176 static void ice_vsi_release_all(struct ice_pf *pf)
7177 {
7178 	int err, i;
7179 
7180 	if (!pf->vsi)
7181 		return;
7182 
7183 	ice_for_each_vsi(pf, i) {
7184 		if (!pf->vsi[i])
7185 			continue;
7186 
7187 		if (pf->vsi[i]->type == ICE_VSI_CHNL)
7188 			continue;
7189 
7190 		err = ice_vsi_release(pf->vsi[i]);
7191 		if (err)
7192 			dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
7193 				i, err, pf->vsi[i]->vsi_num);
7194 	}
7195 }
7196 
7197 /**
7198  * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
7199  * @pf: pointer to the PF instance
7200  * @type: VSI type to rebuild
7201  *
7202  * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
7203  */
7204 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
7205 {
7206 	struct device *dev = ice_pf_to_dev(pf);
7207 	int i, err;
7208 
7209 	ice_for_each_vsi(pf, i) {
7210 		struct ice_vsi *vsi = pf->vsi[i];
7211 
7212 		if (!vsi || vsi->type != type)
7213 			continue;
7214 
7215 		/* rebuild the VSI */
7216 		err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
7217 		if (err) {
7218 			dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
7219 				err, vsi->idx, ice_vsi_type_str(type));
7220 			return err;
7221 		}
7222 
7223 		/* replay filters for the VSI */
7224 		err = ice_replay_vsi(&pf->hw, vsi->idx);
7225 		if (err) {
7226 			dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
7227 				err, vsi->idx, ice_vsi_type_str(type));
7228 			return err;
7229 		}
7230 
7231 		/* Re-map HW VSI number, using VSI handle that has been
7232 		 * previously validated in ice_replay_vsi() call above
7233 		 */
7234 		vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
7235 
7236 		/* enable the VSI */
7237 		err = ice_ena_vsi(vsi, false);
7238 		if (err) {
7239 			dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
7240 				err, vsi->idx, ice_vsi_type_str(type));
7241 			return err;
7242 		}
7243 
7244 		dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
7245 			 ice_vsi_type_str(type));
7246 	}
7247 
7248 	return 0;
7249 }
7250 
7251 /**
7252  * ice_update_pf_netdev_link - Update PF netdev link status
7253  * @pf: pointer to the PF instance
7254  */
7255 static void ice_update_pf_netdev_link(struct ice_pf *pf)
7256 {
7257 	bool link_up;
7258 	int i;
7259 
7260 	ice_for_each_vsi(pf, i) {
7261 		struct ice_vsi *vsi = pf->vsi[i];
7262 
7263 		if (!vsi || vsi->type != ICE_VSI_PF)
7264 			return;
7265 
7266 		ice_get_link_status(pf->vsi[i]->port_info, &link_up);
7267 		if (link_up) {
7268 			netif_carrier_on(pf->vsi[i]->netdev);
7269 			netif_tx_wake_all_queues(pf->vsi[i]->netdev);
7270 		} else {
7271 			netif_carrier_off(pf->vsi[i]->netdev);
7272 			netif_tx_stop_all_queues(pf->vsi[i]->netdev);
7273 		}
7274 	}
7275 }
7276 
7277 /**
7278  * ice_rebuild - rebuild after reset
7279  * @pf: PF to rebuild
7280  * @reset_type: type of reset
7281  *
7282  * Do not rebuild VF VSI in this flow because that is already handled via
7283  * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
7284  * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
7285  * to reset/rebuild all the VF VSI twice.
7286  */
7287 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
7288 {
7289 	struct device *dev = ice_pf_to_dev(pf);
7290 	struct ice_hw *hw = &pf->hw;
7291 	bool dvm;
7292 	int err;
7293 
7294 	if (test_bit(ICE_DOWN, pf->state))
7295 		goto clear_recovery;
7296 
7297 	dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
7298 
7299 #define ICE_EMP_RESET_SLEEP_MS 5000
7300 	if (reset_type == ICE_RESET_EMPR) {
7301 		/* If an EMP reset has occurred, any previously pending flash
7302 		 * update will have completed. We no longer know whether or
7303 		 * not the NVM update EMP reset is restricted.
7304 		 */
7305 		pf->fw_emp_reset_disabled = false;
7306 
7307 		msleep(ICE_EMP_RESET_SLEEP_MS);
7308 	}
7309 
7310 	err = ice_init_all_ctrlq(hw);
7311 	if (err) {
7312 		dev_err(dev, "control queues init failed %d\n", err);
7313 		goto err_init_ctrlq;
7314 	}
7315 
7316 	/* if DDP was previously loaded successfully */
7317 	if (!ice_is_safe_mode(pf)) {
7318 		/* reload the SW DB of filter tables */
7319 		if (reset_type == ICE_RESET_PFR)
7320 			ice_fill_blk_tbls(hw);
7321 		else
7322 			/* Reload DDP Package after CORER/GLOBR reset */
7323 			ice_load_pkg(NULL, pf);
7324 	}
7325 
7326 	err = ice_clear_pf_cfg(hw);
7327 	if (err) {
7328 		dev_err(dev, "clear PF configuration failed %d\n", err);
7329 		goto err_init_ctrlq;
7330 	}
7331 
7332 	ice_clear_pxe_mode(hw);
7333 
7334 	err = ice_init_nvm(hw);
7335 	if (err) {
7336 		dev_err(dev, "ice_init_nvm failed %d\n", err);
7337 		goto err_init_ctrlq;
7338 	}
7339 
7340 	err = ice_get_caps(hw);
7341 	if (err) {
7342 		dev_err(dev, "ice_get_caps failed %d\n", err);
7343 		goto err_init_ctrlq;
7344 	}
7345 
7346 	err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
7347 	if (err) {
7348 		dev_err(dev, "set_mac_cfg failed %d\n", err);
7349 		goto err_init_ctrlq;
7350 	}
7351 
7352 	dvm = ice_is_dvm_ena(hw);
7353 
7354 	err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
7355 	if (err)
7356 		goto err_init_ctrlq;
7357 
7358 	err = ice_sched_init_port(hw->port_info);
7359 	if (err)
7360 		goto err_sched_init_port;
7361 
7362 	/* start misc vector */
7363 	err = ice_req_irq_msix_misc(pf);
7364 	if (err) {
7365 		dev_err(dev, "misc vector setup failed: %d\n", err);
7366 		goto err_sched_init_port;
7367 	}
7368 
7369 	if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7370 		wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
7371 		if (!rd32(hw, PFQF_FD_SIZE)) {
7372 			u16 unused, guar, b_effort;
7373 
7374 			guar = hw->func_caps.fd_fltr_guar;
7375 			b_effort = hw->func_caps.fd_fltr_best_effort;
7376 
7377 			/* force guaranteed filter pool for PF */
7378 			ice_alloc_fd_guar_item(hw, &unused, guar);
7379 			/* force shared filter pool for PF */
7380 			ice_alloc_fd_shrd_item(hw, &unused, b_effort);
7381 		}
7382 	}
7383 
7384 	if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
7385 		ice_dcb_rebuild(pf);
7386 
7387 	/* If the PF previously had enabled PTP, PTP init needs to happen before
7388 	 * the VSI rebuild. If not, this causes the PTP link status events to
7389 	 * fail.
7390 	 */
7391 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7392 		ice_ptp_reset(pf);
7393 
7394 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
7395 		ice_gnss_init(pf);
7396 
7397 	/* rebuild PF VSI */
7398 	err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
7399 	if (err) {
7400 		dev_err(dev, "PF VSI rebuild failed: %d\n", err);
7401 		goto err_vsi_rebuild;
7402 	}
7403 
7404 	err = ice_vsi_rebuild_by_type(pf, ICE_VSI_SWITCHDEV_CTRL);
7405 	if (err) {
7406 		dev_err(dev, "Switchdev CTRL VSI rebuild failed: %d\n", err);
7407 		goto err_vsi_rebuild;
7408 	}
7409 
7410 	if (reset_type == ICE_RESET_PFR) {
7411 		err = ice_rebuild_channels(pf);
7412 		if (err) {
7413 			dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
7414 				err);
7415 			goto err_vsi_rebuild;
7416 		}
7417 	}
7418 
7419 	/* If Flow Director is active */
7420 	if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7421 		err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
7422 		if (err) {
7423 			dev_err(dev, "control VSI rebuild failed: %d\n", err);
7424 			goto err_vsi_rebuild;
7425 		}
7426 
7427 		/* replay HW Flow Director recipes */
7428 		if (hw->fdir_prof)
7429 			ice_fdir_replay_flows(hw);
7430 
7431 		/* replay Flow Director filters */
7432 		ice_fdir_replay_fltrs(pf);
7433 
7434 		ice_rebuild_arfs(pf);
7435 	}
7436 
7437 	ice_update_pf_netdev_link(pf);
7438 
7439 	/* tell the firmware we are up */
7440 	err = ice_send_version(pf);
7441 	if (err) {
7442 		dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
7443 			err);
7444 		goto err_vsi_rebuild;
7445 	}
7446 
7447 	ice_replay_post(hw);
7448 
7449 	/* if we get here, reset flow is successful */
7450 	clear_bit(ICE_RESET_FAILED, pf->state);
7451 
7452 	ice_plug_aux_dev(pf);
7453 	if (ice_is_feature_supported(pf, ICE_F_SRIOV_LAG))
7454 		ice_lag_rebuild(pf);
7455 
7456 	/* Restore timestamp mode settings after VSI rebuild */
7457 	ice_ptp_restore_timestamp_mode(pf);
7458 	return;
7459 
7460 err_vsi_rebuild:
7461 err_sched_init_port:
7462 	ice_sched_cleanup_all(hw);
7463 err_init_ctrlq:
7464 	ice_shutdown_all_ctrlq(hw);
7465 	set_bit(ICE_RESET_FAILED, pf->state);
7466 clear_recovery:
7467 	/* set this bit in PF state to control service task scheduling */
7468 	set_bit(ICE_NEEDS_RESTART, pf->state);
7469 	dev_err(dev, "Rebuild failed, unload and reload driver\n");
7470 }
7471 
7472 /**
7473  * ice_change_mtu - NDO callback to change the MTU
7474  * @netdev: network interface device structure
7475  * @new_mtu: new value for maximum frame size
7476  *
7477  * Returns 0 on success, negative on failure
7478  */
7479 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
7480 {
7481 	struct ice_netdev_priv *np = netdev_priv(netdev);
7482 	struct ice_vsi *vsi = np->vsi;
7483 	struct ice_pf *pf = vsi->back;
7484 	struct bpf_prog *prog;
7485 	u8 count = 0;
7486 	int err = 0;
7487 
7488 	if (new_mtu == (int)netdev->mtu) {
7489 		netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
7490 		return 0;
7491 	}
7492 
7493 	prog = vsi->xdp_prog;
7494 	if (prog && !prog->aux->xdp_has_frags) {
7495 		int frame_size = ice_max_xdp_frame_size(vsi);
7496 
7497 		if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
7498 			netdev_err(netdev, "max MTU for XDP usage is %d\n",
7499 				   frame_size - ICE_ETH_PKT_HDR_PAD);
7500 			return -EINVAL;
7501 		}
7502 	} else if (test_bit(ICE_FLAG_LEGACY_RX, pf->flags)) {
7503 		if (new_mtu + ICE_ETH_PKT_HDR_PAD > ICE_MAX_FRAME_LEGACY_RX) {
7504 			netdev_err(netdev, "Too big MTU for legacy-rx; Max is %d\n",
7505 				   ICE_MAX_FRAME_LEGACY_RX - ICE_ETH_PKT_HDR_PAD);
7506 			return -EINVAL;
7507 		}
7508 	}
7509 
7510 	/* if a reset is in progress, wait for some time for it to complete */
7511 	do {
7512 		if (ice_is_reset_in_progress(pf->state)) {
7513 			count++;
7514 			usleep_range(1000, 2000);
7515 		} else {
7516 			break;
7517 		}
7518 
7519 	} while (count < 100);
7520 
7521 	if (count == 100) {
7522 		netdev_err(netdev, "can't change MTU. Device is busy\n");
7523 		return -EBUSY;
7524 	}
7525 
7526 	netdev->mtu = (unsigned int)new_mtu;
7527 	err = ice_down_up(vsi);
7528 	if (err)
7529 		return err;
7530 
7531 	netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
7532 	set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
7533 
7534 	return err;
7535 }
7536 
7537 /**
7538  * ice_eth_ioctl - Access the hwtstamp interface
7539  * @netdev: network interface device structure
7540  * @ifr: interface request data
7541  * @cmd: ioctl command
7542  */
7543 static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7544 {
7545 	struct ice_netdev_priv *np = netdev_priv(netdev);
7546 	struct ice_pf *pf = np->vsi->back;
7547 
7548 	switch (cmd) {
7549 	case SIOCGHWTSTAMP:
7550 		return ice_ptp_get_ts_config(pf, ifr);
7551 	case SIOCSHWTSTAMP:
7552 		return ice_ptp_set_ts_config(pf, ifr);
7553 	default:
7554 		return -EOPNOTSUPP;
7555 	}
7556 }
7557 
7558 /**
7559  * ice_aq_str - convert AQ err code to a string
7560  * @aq_err: the AQ error code to convert
7561  */
7562 const char *ice_aq_str(enum ice_aq_err aq_err)
7563 {
7564 	switch (aq_err) {
7565 	case ICE_AQ_RC_OK:
7566 		return "OK";
7567 	case ICE_AQ_RC_EPERM:
7568 		return "ICE_AQ_RC_EPERM";
7569 	case ICE_AQ_RC_ENOENT:
7570 		return "ICE_AQ_RC_ENOENT";
7571 	case ICE_AQ_RC_ENOMEM:
7572 		return "ICE_AQ_RC_ENOMEM";
7573 	case ICE_AQ_RC_EBUSY:
7574 		return "ICE_AQ_RC_EBUSY";
7575 	case ICE_AQ_RC_EEXIST:
7576 		return "ICE_AQ_RC_EEXIST";
7577 	case ICE_AQ_RC_EINVAL:
7578 		return "ICE_AQ_RC_EINVAL";
7579 	case ICE_AQ_RC_ENOSPC:
7580 		return "ICE_AQ_RC_ENOSPC";
7581 	case ICE_AQ_RC_ENOSYS:
7582 		return "ICE_AQ_RC_ENOSYS";
7583 	case ICE_AQ_RC_EMODE:
7584 		return "ICE_AQ_RC_EMODE";
7585 	case ICE_AQ_RC_ENOSEC:
7586 		return "ICE_AQ_RC_ENOSEC";
7587 	case ICE_AQ_RC_EBADSIG:
7588 		return "ICE_AQ_RC_EBADSIG";
7589 	case ICE_AQ_RC_ESVN:
7590 		return "ICE_AQ_RC_ESVN";
7591 	case ICE_AQ_RC_EBADMAN:
7592 		return "ICE_AQ_RC_EBADMAN";
7593 	case ICE_AQ_RC_EBADBUF:
7594 		return "ICE_AQ_RC_EBADBUF";
7595 	}
7596 
7597 	return "ICE_AQ_RC_UNKNOWN";
7598 }
7599 
7600 /**
7601  * ice_set_rss_lut - Set RSS LUT
7602  * @vsi: Pointer to VSI structure
7603  * @lut: Lookup table
7604  * @lut_size: Lookup table size
7605  *
7606  * Returns 0 on success, negative on failure
7607  */
7608 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7609 {
7610 	struct ice_aq_get_set_rss_lut_params params = {};
7611 	struct ice_hw *hw = &vsi->back->hw;
7612 	int status;
7613 
7614 	if (!lut)
7615 		return -EINVAL;
7616 
7617 	params.vsi_handle = vsi->idx;
7618 	params.lut_size = lut_size;
7619 	params.lut_type = vsi->rss_lut_type;
7620 	params.lut = lut;
7621 
7622 	status = ice_aq_set_rss_lut(hw, &params);
7623 	if (status)
7624 		dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
7625 			status, ice_aq_str(hw->adminq.sq_last_status));
7626 
7627 	return status;
7628 }
7629 
7630 /**
7631  * ice_set_rss_key - Set RSS key
7632  * @vsi: Pointer to the VSI structure
7633  * @seed: RSS hash seed
7634  *
7635  * Returns 0 on success, negative on failure
7636  */
7637 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7638 {
7639 	struct ice_hw *hw = &vsi->back->hw;
7640 	int status;
7641 
7642 	if (!seed)
7643 		return -EINVAL;
7644 
7645 	status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7646 	if (status)
7647 		dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7648 			status, ice_aq_str(hw->adminq.sq_last_status));
7649 
7650 	return status;
7651 }
7652 
7653 /**
7654  * ice_get_rss_lut - Get RSS LUT
7655  * @vsi: Pointer to VSI structure
7656  * @lut: Buffer to store the lookup table entries
7657  * @lut_size: Size of buffer to store the lookup table entries
7658  *
7659  * Returns 0 on success, negative on failure
7660  */
7661 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7662 {
7663 	struct ice_aq_get_set_rss_lut_params params = {};
7664 	struct ice_hw *hw = &vsi->back->hw;
7665 	int status;
7666 
7667 	if (!lut)
7668 		return -EINVAL;
7669 
7670 	params.vsi_handle = vsi->idx;
7671 	params.lut_size = lut_size;
7672 	params.lut_type = vsi->rss_lut_type;
7673 	params.lut = lut;
7674 
7675 	status = ice_aq_get_rss_lut(hw, &params);
7676 	if (status)
7677 		dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7678 			status, ice_aq_str(hw->adminq.sq_last_status));
7679 
7680 	return status;
7681 }
7682 
7683 /**
7684  * ice_get_rss_key - Get RSS key
7685  * @vsi: Pointer to VSI structure
7686  * @seed: Buffer to store the key in
7687  *
7688  * Returns 0 on success, negative on failure
7689  */
7690 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7691 {
7692 	struct ice_hw *hw = &vsi->back->hw;
7693 	int status;
7694 
7695 	if (!seed)
7696 		return -EINVAL;
7697 
7698 	status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7699 	if (status)
7700 		dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7701 			status, ice_aq_str(hw->adminq.sq_last_status));
7702 
7703 	return status;
7704 }
7705 
7706 /**
7707  * ice_bridge_getlink - Get the hardware bridge mode
7708  * @skb: skb buff
7709  * @pid: process ID
7710  * @seq: RTNL message seq
7711  * @dev: the netdev being configured
7712  * @filter_mask: filter mask passed in
7713  * @nlflags: netlink flags passed in
7714  *
7715  * Return the bridge mode (VEB/VEPA)
7716  */
7717 static int
7718 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7719 		   struct net_device *dev, u32 filter_mask, int nlflags)
7720 {
7721 	struct ice_netdev_priv *np = netdev_priv(dev);
7722 	struct ice_vsi *vsi = np->vsi;
7723 	struct ice_pf *pf = vsi->back;
7724 	u16 bmode;
7725 
7726 	bmode = pf->first_sw->bridge_mode;
7727 
7728 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7729 				       filter_mask, NULL);
7730 }
7731 
7732 /**
7733  * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7734  * @vsi: Pointer to VSI structure
7735  * @bmode: Hardware bridge mode (VEB/VEPA)
7736  *
7737  * Returns 0 on success, negative on failure
7738  */
7739 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7740 {
7741 	struct ice_aqc_vsi_props *vsi_props;
7742 	struct ice_hw *hw = &vsi->back->hw;
7743 	struct ice_vsi_ctx *ctxt;
7744 	int ret;
7745 
7746 	vsi_props = &vsi->info;
7747 
7748 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7749 	if (!ctxt)
7750 		return -ENOMEM;
7751 
7752 	ctxt->info = vsi->info;
7753 
7754 	if (bmode == BRIDGE_MODE_VEB)
7755 		/* change from VEPA to VEB mode */
7756 		ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7757 	else
7758 		/* change from VEB to VEPA mode */
7759 		ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7760 	ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7761 
7762 	ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7763 	if (ret) {
7764 		dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7765 			bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7766 		goto out;
7767 	}
7768 	/* Update sw flags for book keeping */
7769 	vsi_props->sw_flags = ctxt->info.sw_flags;
7770 
7771 out:
7772 	kfree(ctxt);
7773 	return ret;
7774 }
7775 
7776 /**
7777  * ice_bridge_setlink - Set the hardware bridge mode
7778  * @dev: the netdev being configured
7779  * @nlh: RTNL message
7780  * @flags: bridge setlink flags
7781  * @extack: netlink extended ack
7782  *
7783  * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7784  * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7785  * not already set for all VSIs connected to this switch. And also update the
7786  * unicast switch filter rules for the corresponding switch of the netdev.
7787  */
7788 static int
7789 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
7790 		   u16 __always_unused flags,
7791 		   struct netlink_ext_ack __always_unused *extack)
7792 {
7793 	struct ice_netdev_priv *np = netdev_priv(dev);
7794 	struct ice_pf *pf = np->vsi->back;
7795 	struct nlattr *attr, *br_spec;
7796 	struct ice_hw *hw = &pf->hw;
7797 	struct ice_sw *pf_sw;
7798 	int rem, v, err = 0;
7799 
7800 	pf_sw = pf->first_sw;
7801 	/* find the attribute in the netlink message */
7802 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7803 
7804 	nla_for_each_nested(attr, br_spec, rem) {
7805 		__u16 mode;
7806 
7807 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7808 			continue;
7809 		mode = nla_get_u16(attr);
7810 		if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
7811 			return -EINVAL;
7812 		/* Continue  if bridge mode is not being flipped */
7813 		if (mode == pf_sw->bridge_mode)
7814 			continue;
7815 		/* Iterates through the PF VSI list and update the loopback
7816 		 * mode of the VSI
7817 		 */
7818 		ice_for_each_vsi(pf, v) {
7819 			if (!pf->vsi[v])
7820 				continue;
7821 			err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
7822 			if (err)
7823 				return err;
7824 		}
7825 
7826 		hw->evb_veb = (mode == BRIDGE_MODE_VEB);
7827 		/* Update the unicast switch filter rules for the corresponding
7828 		 * switch of the netdev
7829 		 */
7830 		err = ice_update_sw_rule_bridge_mode(hw);
7831 		if (err) {
7832 			netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
7833 				   mode, err,
7834 				   ice_aq_str(hw->adminq.sq_last_status));
7835 			/* revert hw->evb_veb */
7836 			hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
7837 			return err;
7838 		}
7839 
7840 		pf_sw->bridge_mode = mode;
7841 	}
7842 
7843 	return 0;
7844 }
7845 
7846 /**
7847  * ice_tx_timeout - Respond to a Tx Hang
7848  * @netdev: network interface device structure
7849  * @txqueue: Tx queue
7850  */
7851 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
7852 {
7853 	struct ice_netdev_priv *np = netdev_priv(netdev);
7854 	struct ice_tx_ring *tx_ring = NULL;
7855 	struct ice_vsi *vsi = np->vsi;
7856 	struct ice_pf *pf = vsi->back;
7857 	u32 i;
7858 
7859 	pf->tx_timeout_count++;
7860 
7861 	/* Check if PFC is enabled for the TC to which the queue belongs
7862 	 * to. If yes then Tx timeout is not caused by a hung queue, no
7863 	 * need to reset and rebuild
7864 	 */
7865 	if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
7866 		dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
7867 			 txqueue);
7868 		return;
7869 	}
7870 
7871 	/* now that we have an index, find the tx_ring struct */
7872 	ice_for_each_txq(vsi, i)
7873 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
7874 			if (txqueue == vsi->tx_rings[i]->q_index) {
7875 				tx_ring = vsi->tx_rings[i];
7876 				break;
7877 			}
7878 
7879 	/* Reset recovery level if enough time has elapsed after last timeout.
7880 	 * Also ensure no new reset action happens before next timeout period.
7881 	 */
7882 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
7883 		pf->tx_timeout_recovery_level = 1;
7884 	else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
7885 				       netdev->watchdog_timeo)))
7886 		return;
7887 
7888 	if (tx_ring) {
7889 		struct ice_hw *hw = &pf->hw;
7890 		u32 head, val = 0;
7891 
7892 		head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
7893 			QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
7894 		/* Read interrupt register */
7895 		val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
7896 
7897 		netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
7898 			    vsi->vsi_num, txqueue, tx_ring->next_to_clean,
7899 			    head, tx_ring->next_to_use, val);
7900 	}
7901 
7902 	pf->tx_timeout_last_recovery = jiffies;
7903 	netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
7904 		    pf->tx_timeout_recovery_level, txqueue);
7905 
7906 	switch (pf->tx_timeout_recovery_level) {
7907 	case 1:
7908 		set_bit(ICE_PFR_REQ, pf->state);
7909 		break;
7910 	case 2:
7911 		set_bit(ICE_CORER_REQ, pf->state);
7912 		break;
7913 	case 3:
7914 		set_bit(ICE_GLOBR_REQ, pf->state);
7915 		break;
7916 	default:
7917 		netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
7918 		set_bit(ICE_DOWN, pf->state);
7919 		set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
7920 		set_bit(ICE_SERVICE_DIS, pf->state);
7921 		break;
7922 	}
7923 
7924 	ice_service_task_schedule(pf);
7925 	pf->tx_timeout_recovery_level++;
7926 }
7927 
7928 /**
7929  * ice_setup_tc_cls_flower - flower classifier offloads
7930  * @np: net device to configure
7931  * @filter_dev: device on which filter is added
7932  * @cls_flower: offload data
7933  */
7934 static int
7935 ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
7936 			struct net_device *filter_dev,
7937 			struct flow_cls_offload *cls_flower)
7938 {
7939 	struct ice_vsi *vsi = np->vsi;
7940 
7941 	if (cls_flower->common.chain_index)
7942 		return -EOPNOTSUPP;
7943 
7944 	switch (cls_flower->command) {
7945 	case FLOW_CLS_REPLACE:
7946 		return ice_add_cls_flower(filter_dev, vsi, cls_flower);
7947 	case FLOW_CLS_DESTROY:
7948 		return ice_del_cls_flower(vsi, cls_flower);
7949 	default:
7950 		return -EINVAL;
7951 	}
7952 }
7953 
7954 /**
7955  * ice_setup_tc_block_cb - callback handler registered for TC block
7956  * @type: TC SETUP type
7957  * @type_data: TC flower offload data that contains user input
7958  * @cb_priv: netdev private data
7959  */
7960 static int
7961 ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
7962 {
7963 	struct ice_netdev_priv *np = cb_priv;
7964 
7965 	switch (type) {
7966 	case TC_SETUP_CLSFLOWER:
7967 		return ice_setup_tc_cls_flower(np, np->vsi->netdev,
7968 					       type_data);
7969 	default:
7970 		return -EOPNOTSUPP;
7971 	}
7972 }
7973 
7974 /**
7975  * ice_validate_mqprio_qopt - Validate TCF input parameters
7976  * @vsi: Pointer to VSI
7977  * @mqprio_qopt: input parameters for mqprio queue configuration
7978  *
7979  * This function validates MQPRIO params, such as qcount (power of 2 wherever
7980  * needed), and make sure user doesn't specify qcount and BW rate limit
7981  * for TCs, which are more than "num_tc"
7982  */
7983 static int
7984 ice_validate_mqprio_qopt(struct ice_vsi *vsi,
7985 			 struct tc_mqprio_qopt_offload *mqprio_qopt)
7986 {
7987 	int non_power_of_2_qcount = 0;
7988 	struct ice_pf *pf = vsi->back;
7989 	int max_rss_q_cnt = 0;
7990 	u64 sum_min_rate = 0;
7991 	struct device *dev;
7992 	int i, speed;
7993 	u8 num_tc;
7994 
7995 	if (vsi->type != ICE_VSI_PF)
7996 		return -EINVAL;
7997 
7998 	if (mqprio_qopt->qopt.offset[0] != 0 ||
7999 	    mqprio_qopt->qopt.num_tc < 1 ||
8000 	    mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
8001 		return -EINVAL;
8002 
8003 	dev = ice_pf_to_dev(pf);
8004 	vsi->ch_rss_size = 0;
8005 	num_tc = mqprio_qopt->qopt.num_tc;
8006 	speed = ice_get_link_speed_kbps(vsi);
8007 
8008 	for (i = 0; num_tc; i++) {
8009 		int qcount = mqprio_qopt->qopt.count[i];
8010 		u64 max_rate, min_rate, rem;
8011 
8012 		if (!qcount)
8013 			return -EINVAL;
8014 
8015 		if (is_power_of_2(qcount)) {
8016 			if (non_power_of_2_qcount &&
8017 			    qcount > non_power_of_2_qcount) {
8018 				dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
8019 					qcount, non_power_of_2_qcount);
8020 				return -EINVAL;
8021 			}
8022 			if (qcount > max_rss_q_cnt)
8023 				max_rss_q_cnt = qcount;
8024 		} else {
8025 			if (non_power_of_2_qcount &&
8026 			    qcount != non_power_of_2_qcount) {
8027 				dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
8028 					qcount, non_power_of_2_qcount);
8029 				return -EINVAL;
8030 			}
8031 			if (qcount < max_rss_q_cnt) {
8032 				dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
8033 					qcount, max_rss_q_cnt);
8034 				return -EINVAL;
8035 			}
8036 			max_rss_q_cnt = qcount;
8037 			non_power_of_2_qcount = qcount;
8038 		}
8039 
8040 		/* TC command takes input in K/N/Gbps or K/M/Gbit etc but
8041 		 * converts the bandwidth rate limit into Bytes/s when
8042 		 * passing it down to the driver. So convert input bandwidth
8043 		 * from Bytes/s to Kbps
8044 		 */
8045 		max_rate = mqprio_qopt->max_rate[i];
8046 		max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
8047 
8048 		/* min_rate is minimum guaranteed rate and it can't be zero */
8049 		min_rate = mqprio_qopt->min_rate[i];
8050 		min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
8051 		sum_min_rate += min_rate;
8052 
8053 		if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
8054 			dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
8055 				min_rate, ICE_MIN_BW_LIMIT);
8056 			return -EINVAL;
8057 		}
8058 
8059 		if (max_rate && max_rate > speed) {
8060 			dev_err(dev, "TC%d: max_rate(%llu Kbps) > link speed of %u Kbps\n",
8061 				i, max_rate, speed);
8062 			return -EINVAL;
8063 		}
8064 
8065 		iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
8066 		if (rem) {
8067 			dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
8068 				i, ICE_MIN_BW_LIMIT);
8069 			return -EINVAL;
8070 		}
8071 
8072 		iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
8073 		if (rem) {
8074 			dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
8075 				i, ICE_MIN_BW_LIMIT);
8076 			return -EINVAL;
8077 		}
8078 
8079 		/* min_rate can't be more than max_rate, except when max_rate
8080 		 * is zero (implies max_rate sought is max line rate). In such
8081 		 * a case min_rate can be more than max.
8082 		 */
8083 		if (max_rate && min_rate > max_rate) {
8084 			dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
8085 				min_rate, max_rate);
8086 			return -EINVAL;
8087 		}
8088 
8089 		if (i >= mqprio_qopt->qopt.num_tc - 1)
8090 			break;
8091 		if (mqprio_qopt->qopt.offset[i + 1] !=
8092 		    (mqprio_qopt->qopt.offset[i] + qcount))
8093 			return -EINVAL;
8094 	}
8095 	if (vsi->num_rxq <
8096 	    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8097 		return -EINVAL;
8098 	if (vsi->num_txq <
8099 	    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8100 		return -EINVAL;
8101 
8102 	if (sum_min_rate && sum_min_rate > (u64)speed) {
8103 		dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
8104 			sum_min_rate, speed);
8105 		return -EINVAL;
8106 	}
8107 
8108 	/* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
8109 	vsi->ch_rss_size = max_rss_q_cnt;
8110 
8111 	return 0;
8112 }
8113 
8114 /**
8115  * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
8116  * @pf: ptr to PF device
8117  * @vsi: ptr to VSI
8118  */
8119 static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
8120 {
8121 	struct device *dev = ice_pf_to_dev(pf);
8122 	bool added = false;
8123 	struct ice_hw *hw;
8124 	int flow;
8125 
8126 	if (!(vsi->num_gfltr || vsi->num_bfltr))
8127 		return -EINVAL;
8128 
8129 	hw = &pf->hw;
8130 	for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
8131 		struct ice_fd_hw_prof *prof;
8132 		int tun, status;
8133 		u64 entry_h;
8134 
8135 		if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
8136 		      hw->fdir_prof[flow]->cnt))
8137 			continue;
8138 
8139 		for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
8140 			enum ice_flow_priority prio;
8141 			u64 prof_id;
8142 
8143 			/* add this VSI to FDir profile for this flow */
8144 			prio = ICE_FLOW_PRIO_NORMAL;
8145 			prof = hw->fdir_prof[flow];
8146 			prof_id = flow + tun * ICE_FLTR_PTYPE_MAX;
8147 			status = ice_flow_add_entry(hw, ICE_BLK_FD, prof_id,
8148 						    prof->vsi_h[0], vsi->idx,
8149 						    prio, prof->fdir_seg[tun],
8150 						    &entry_h);
8151 			if (status) {
8152 				dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
8153 					vsi->idx, flow);
8154 				continue;
8155 			}
8156 
8157 			prof->entry_h[prof->cnt][tun] = entry_h;
8158 		}
8159 
8160 		/* store VSI for filter replay and delete */
8161 		prof->vsi_h[prof->cnt] = vsi->idx;
8162 		prof->cnt++;
8163 
8164 		added = true;
8165 		dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
8166 			flow);
8167 	}
8168 
8169 	if (!added)
8170 		dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
8171 
8172 	return 0;
8173 }
8174 
8175 /**
8176  * ice_add_channel - add a channel by adding VSI
8177  * @pf: ptr to PF device
8178  * @sw_id: underlying HW switching element ID
8179  * @ch: ptr to channel structure
8180  *
8181  * Add a channel (VSI) using add_vsi and queue_map
8182  */
8183 static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
8184 {
8185 	struct device *dev = ice_pf_to_dev(pf);
8186 	struct ice_vsi *vsi;
8187 
8188 	if (ch->type != ICE_VSI_CHNL) {
8189 		dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
8190 		return -EINVAL;
8191 	}
8192 
8193 	vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
8194 	if (!vsi || vsi->type != ICE_VSI_CHNL) {
8195 		dev_err(dev, "create chnl VSI failure\n");
8196 		return -EINVAL;
8197 	}
8198 
8199 	ice_add_vsi_to_fdir(pf, vsi);
8200 
8201 	ch->sw_id = sw_id;
8202 	ch->vsi_num = vsi->vsi_num;
8203 	ch->info.mapping_flags = vsi->info.mapping_flags;
8204 	ch->ch_vsi = vsi;
8205 	/* set the back pointer of channel for newly created VSI */
8206 	vsi->ch = ch;
8207 
8208 	memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
8209 	       sizeof(vsi->info.q_mapping));
8210 	memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
8211 	       sizeof(vsi->info.tc_mapping));
8212 
8213 	return 0;
8214 }
8215 
8216 /**
8217  * ice_chnl_cfg_res
8218  * @vsi: the VSI being setup
8219  * @ch: ptr to channel structure
8220  *
8221  * Configure channel specific resources such as rings, vector.
8222  */
8223 static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
8224 {
8225 	int i;
8226 
8227 	for (i = 0; i < ch->num_txq; i++) {
8228 		struct ice_q_vector *tx_q_vector, *rx_q_vector;
8229 		struct ice_ring_container *rc;
8230 		struct ice_tx_ring *tx_ring;
8231 		struct ice_rx_ring *rx_ring;
8232 
8233 		tx_ring = vsi->tx_rings[ch->base_q + i];
8234 		rx_ring = vsi->rx_rings[ch->base_q + i];
8235 		if (!tx_ring || !rx_ring)
8236 			continue;
8237 
8238 		/* setup ring being channel enabled */
8239 		tx_ring->ch = ch;
8240 		rx_ring->ch = ch;
8241 
8242 		/* following code block sets up vector specific attributes */
8243 		tx_q_vector = tx_ring->q_vector;
8244 		rx_q_vector = rx_ring->q_vector;
8245 		if (!tx_q_vector && !rx_q_vector)
8246 			continue;
8247 
8248 		if (tx_q_vector) {
8249 			tx_q_vector->ch = ch;
8250 			/* setup Tx and Rx ITR setting if DIM is off */
8251 			rc = &tx_q_vector->tx;
8252 			if (!ITR_IS_DYNAMIC(rc))
8253 				ice_write_itr(rc, rc->itr_setting);
8254 		}
8255 		if (rx_q_vector) {
8256 			rx_q_vector->ch = ch;
8257 			/* setup Tx and Rx ITR setting if DIM is off */
8258 			rc = &rx_q_vector->rx;
8259 			if (!ITR_IS_DYNAMIC(rc))
8260 				ice_write_itr(rc, rc->itr_setting);
8261 		}
8262 	}
8263 
8264 	/* it is safe to assume that, if channel has non-zero num_t[r]xq, then
8265 	 * GLINT_ITR register would have written to perform in-context
8266 	 * update, hence perform flush
8267 	 */
8268 	if (ch->num_txq || ch->num_rxq)
8269 		ice_flush(&vsi->back->hw);
8270 }
8271 
8272 /**
8273  * ice_cfg_chnl_all_res - configure channel resources
8274  * @vsi: pte to main_vsi
8275  * @ch: ptr to channel structure
8276  *
8277  * This function configures channel specific resources such as flow-director
8278  * counter index, and other resources such as queues, vectors, ITR settings
8279  */
8280 static void
8281 ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
8282 {
8283 	/* configure channel (aka ADQ) resources such as queues, vectors,
8284 	 * ITR settings for channel specific vectors and anything else
8285 	 */
8286 	ice_chnl_cfg_res(vsi, ch);
8287 }
8288 
8289 /**
8290  * ice_setup_hw_channel - setup new channel
8291  * @pf: ptr to PF device
8292  * @vsi: the VSI being setup
8293  * @ch: ptr to channel structure
8294  * @sw_id: underlying HW switching element ID
8295  * @type: type of channel to be created (VMDq2/VF)
8296  *
8297  * Setup new channel (VSI) based on specified type (VMDq2/VF)
8298  * and configures Tx rings accordingly
8299  */
8300 static int
8301 ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8302 		     struct ice_channel *ch, u16 sw_id, u8 type)
8303 {
8304 	struct device *dev = ice_pf_to_dev(pf);
8305 	int ret;
8306 
8307 	ch->base_q = vsi->next_base_q;
8308 	ch->type = type;
8309 
8310 	ret = ice_add_channel(pf, sw_id, ch);
8311 	if (ret) {
8312 		dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
8313 		return ret;
8314 	}
8315 
8316 	/* configure/setup ADQ specific resources */
8317 	ice_cfg_chnl_all_res(vsi, ch);
8318 
8319 	/* make sure to update the next_base_q so that subsequent channel's
8320 	 * (aka ADQ) VSI queue map is correct
8321 	 */
8322 	vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
8323 	dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
8324 		ch->num_rxq);
8325 
8326 	return 0;
8327 }
8328 
8329 /**
8330  * ice_setup_channel - setup new channel using uplink element
8331  * @pf: ptr to PF device
8332  * @vsi: the VSI being setup
8333  * @ch: ptr to channel structure
8334  *
8335  * Setup new channel (VSI) based on specified type (VMDq2/VF)
8336  * and uplink switching element
8337  */
8338 static bool
8339 ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8340 		  struct ice_channel *ch)
8341 {
8342 	struct device *dev = ice_pf_to_dev(pf);
8343 	u16 sw_id;
8344 	int ret;
8345 
8346 	if (vsi->type != ICE_VSI_PF) {
8347 		dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
8348 		return false;
8349 	}
8350 
8351 	sw_id = pf->first_sw->sw_id;
8352 
8353 	/* create channel (VSI) */
8354 	ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
8355 	if (ret) {
8356 		dev_err(dev, "failed to setup hw_channel\n");
8357 		return false;
8358 	}
8359 	dev_dbg(dev, "successfully created channel()\n");
8360 
8361 	return ch->ch_vsi ? true : false;
8362 }
8363 
8364 /**
8365  * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
8366  * @vsi: VSI to be configured
8367  * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
8368  * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
8369  */
8370 static int
8371 ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
8372 {
8373 	int err;
8374 
8375 	err = ice_set_min_bw_limit(vsi, min_tx_rate);
8376 	if (err)
8377 		return err;
8378 
8379 	return ice_set_max_bw_limit(vsi, max_tx_rate);
8380 }
8381 
8382 /**
8383  * ice_create_q_channel - function to create channel
8384  * @vsi: VSI to be configured
8385  * @ch: ptr to channel (it contains channel specific params)
8386  *
8387  * This function creates channel (VSI) using num_queues specified by user,
8388  * reconfigs RSS if needed.
8389  */
8390 static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
8391 {
8392 	struct ice_pf *pf = vsi->back;
8393 	struct device *dev;
8394 
8395 	if (!ch)
8396 		return -EINVAL;
8397 
8398 	dev = ice_pf_to_dev(pf);
8399 	if (!ch->num_txq || !ch->num_rxq) {
8400 		dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
8401 		return -EINVAL;
8402 	}
8403 
8404 	if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
8405 		dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
8406 			vsi->cnt_q_avail, ch->num_txq);
8407 		return -EINVAL;
8408 	}
8409 
8410 	if (!ice_setup_channel(pf, vsi, ch)) {
8411 		dev_info(dev, "Failed to setup channel\n");
8412 		return -EINVAL;
8413 	}
8414 	/* configure BW rate limit */
8415 	if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
8416 		int ret;
8417 
8418 		ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
8419 				       ch->min_tx_rate);
8420 		if (ret)
8421 			dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
8422 				ch->max_tx_rate, ch->ch_vsi->vsi_num);
8423 		else
8424 			dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
8425 				ch->max_tx_rate, ch->ch_vsi->vsi_num);
8426 	}
8427 
8428 	vsi->cnt_q_avail -= ch->num_txq;
8429 
8430 	return 0;
8431 }
8432 
8433 /**
8434  * ice_rem_all_chnl_fltrs - removes all channel filters
8435  * @pf: ptr to PF, TC-flower based filter are tracked at PF level
8436  *
8437  * Remove all advanced switch filters only if they are channel specific
8438  * tc-flower based filter
8439  */
8440 static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
8441 {
8442 	struct ice_tc_flower_fltr *fltr;
8443 	struct hlist_node *node;
8444 
8445 	/* to remove all channel filters, iterate an ordered list of filters */
8446 	hlist_for_each_entry_safe(fltr, node,
8447 				  &pf->tc_flower_fltr_list,
8448 				  tc_flower_node) {
8449 		struct ice_rule_query_data rule;
8450 		int status;
8451 
8452 		/* for now process only channel specific filters */
8453 		if (!ice_is_chnl_fltr(fltr))
8454 			continue;
8455 
8456 		rule.rid = fltr->rid;
8457 		rule.rule_id = fltr->rule_id;
8458 		rule.vsi_handle = fltr->dest_vsi_handle;
8459 		status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
8460 		if (status) {
8461 			if (status == -ENOENT)
8462 				dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
8463 					rule.rule_id);
8464 			else
8465 				dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
8466 					status);
8467 		} else if (fltr->dest_vsi) {
8468 			/* update advanced switch filter count */
8469 			if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
8470 				u32 flags = fltr->flags;
8471 
8472 				fltr->dest_vsi->num_chnl_fltr--;
8473 				if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
8474 					     ICE_TC_FLWR_FIELD_ENC_DST_MAC))
8475 					pf->num_dmac_chnl_fltrs--;
8476 			}
8477 		}
8478 
8479 		hlist_del(&fltr->tc_flower_node);
8480 		kfree(fltr);
8481 	}
8482 }
8483 
8484 /**
8485  * ice_remove_q_channels - Remove queue channels for the TCs
8486  * @vsi: VSI to be configured
8487  * @rem_fltr: delete advanced switch filter or not
8488  *
8489  * Remove queue channels for the TCs
8490  */
8491 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
8492 {
8493 	struct ice_channel *ch, *ch_tmp;
8494 	struct ice_pf *pf = vsi->back;
8495 	int i;
8496 
8497 	/* remove all tc-flower based filter if they are channel filters only */
8498 	if (rem_fltr)
8499 		ice_rem_all_chnl_fltrs(pf);
8500 
8501 	/* remove ntuple filters since queue configuration is being changed */
8502 	if  (vsi->netdev->features & NETIF_F_NTUPLE) {
8503 		struct ice_hw *hw = &pf->hw;
8504 
8505 		mutex_lock(&hw->fdir_fltr_lock);
8506 		ice_fdir_del_all_fltrs(vsi);
8507 		mutex_unlock(&hw->fdir_fltr_lock);
8508 	}
8509 
8510 	/* perform cleanup for channels if they exist */
8511 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8512 		struct ice_vsi *ch_vsi;
8513 
8514 		list_del(&ch->list);
8515 		ch_vsi = ch->ch_vsi;
8516 		if (!ch_vsi) {
8517 			kfree(ch);
8518 			continue;
8519 		}
8520 
8521 		/* Reset queue contexts */
8522 		for (i = 0; i < ch->num_rxq; i++) {
8523 			struct ice_tx_ring *tx_ring;
8524 			struct ice_rx_ring *rx_ring;
8525 
8526 			tx_ring = vsi->tx_rings[ch->base_q + i];
8527 			rx_ring = vsi->rx_rings[ch->base_q + i];
8528 			if (tx_ring) {
8529 				tx_ring->ch = NULL;
8530 				if (tx_ring->q_vector)
8531 					tx_ring->q_vector->ch = NULL;
8532 			}
8533 			if (rx_ring) {
8534 				rx_ring->ch = NULL;
8535 				if (rx_ring->q_vector)
8536 					rx_ring->q_vector->ch = NULL;
8537 			}
8538 		}
8539 
8540 		/* Release FD resources for the channel VSI */
8541 		ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
8542 
8543 		/* clear the VSI from scheduler tree */
8544 		ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
8545 
8546 		/* Delete VSI from FW, PF and HW VSI arrays */
8547 		ice_vsi_delete(ch->ch_vsi);
8548 
8549 		/* free the channel */
8550 		kfree(ch);
8551 	}
8552 
8553 	/* clear the channel VSI map which is stored in main VSI */
8554 	ice_for_each_chnl_tc(i)
8555 		vsi->tc_map_vsi[i] = NULL;
8556 
8557 	/* reset main VSI's all TC information */
8558 	vsi->all_enatc = 0;
8559 	vsi->all_numtc = 0;
8560 }
8561 
8562 /**
8563  * ice_rebuild_channels - rebuild channel
8564  * @pf: ptr to PF
8565  *
8566  * Recreate channel VSIs and replay filters
8567  */
8568 static int ice_rebuild_channels(struct ice_pf *pf)
8569 {
8570 	struct device *dev = ice_pf_to_dev(pf);
8571 	struct ice_vsi *main_vsi;
8572 	bool rem_adv_fltr = true;
8573 	struct ice_channel *ch;
8574 	struct ice_vsi *vsi;
8575 	int tc_idx = 1;
8576 	int i, err;
8577 
8578 	main_vsi = ice_get_main_vsi(pf);
8579 	if (!main_vsi)
8580 		return 0;
8581 
8582 	if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
8583 	    main_vsi->old_numtc == 1)
8584 		return 0; /* nothing to be done */
8585 
8586 	/* reconfigure main VSI based on old value of TC and cached values
8587 	 * for MQPRIO opts
8588 	 */
8589 	err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
8590 	if (err) {
8591 		dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
8592 			main_vsi->old_ena_tc, main_vsi->vsi_num);
8593 		return err;
8594 	}
8595 
8596 	/* rebuild ADQ VSIs */
8597 	ice_for_each_vsi(pf, i) {
8598 		enum ice_vsi_type type;
8599 
8600 		vsi = pf->vsi[i];
8601 		if (!vsi || vsi->type != ICE_VSI_CHNL)
8602 			continue;
8603 
8604 		type = vsi->type;
8605 
8606 		/* rebuild ADQ VSI */
8607 		err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
8608 		if (err) {
8609 			dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
8610 				ice_vsi_type_str(type), vsi->idx, err);
8611 			goto cleanup;
8612 		}
8613 
8614 		/* Re-map HW VSI number, using VSI handle that has been
8615 		 * previously validated in ice_replay_vsi() call above
8616 		 */
8617 		vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
8618 
8619 		/* replay filters for the VSI */
8620 		err = ice_replay_vsi(&pf->hw, vsi->idx);
8621 		if (err) {
8622 			dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
8623 				ice_vsi_type_str(type), err, vsi->idx);
8624 			rem_adv_fltr = false;
8625 			goto cleanup;
8626 		}
8627 		dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
8628 			 ice_vsi_type_str(type), vsi->idx);
8629 
8630 		/* store ADQ VSI at correct TC index in main VSI's
8631 		 * map of TC to VSI
8632 		 */
8633 		main_vsi->tc_map_vsi[tc_idx++] = vsi;
8634 	}
8635 
8636 	/* ADQ VSI(s) has been rebuilt successfully, so setup
8637 	 * channel for main VSI's Tx and Rx rings
8638 	 */
8639 	list_for_each_entry(ch, &main_vsi->ch_list, list) {
8640 		struct ice_vsi *ch_vsi;
8641 
8642 		ch_vsi = ch->ch_vsi;
8643 		if (!ch_vsi)
8644 			continue;
8645 
8646 		/* reconfig channel resources */
8647 		ice_cfg_chnl_all_res(main_vsi, ch);
8648 
8649 		/* replay BW rate limit if it is non-zero */
8650 		if (!ch->max_tx_rate && !ch->min_tx_rate)
8651 			continue;
8652 
8653 		err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8654 				       ch->min_tx_rate);
8655 		if (err)
8656 			dev_err(dev, "failed (err:%d) to rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8657 				err, ch->max_tx_rate, ch->min_tx_rate,
8658 				ch_vsi->vsi_num);
8659 		else
8660 			dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8661 				ch->max_tx_rate, ch->min_tx_rate,
8662 				ch_vsi->vsi_num);
8663 	}
8664 
8665 	/* reconfig RSS for main VSI */
8666 	if (main_vsi->ch_rss_size)
8667 		ice_vsi_cfg_rss_lut_key(main_vsi);
8668 
8669 	return 0;
8670 
8671 cleanup:
8672 	ice_remove_q_channels(main_vsi, rem_adv_fltr);
8673 	return err;
8674 }
8675 
8676 /**
8677  * ice_create_q_channels - Add queue channel for the given TCs
8678  * @vsi: VSI to be configured
8679  *
8680  * Configures queue channel mapping to the given TCs
8681  */
8682 static int ice_create_q_channels(struct ice_vsi *vsi)
8683 {
8684 	struct ice_pf *pf = vsi->back;
8685 	struct ice_channel *ch;
8686 	int ret = 0, i;
8687 
8688 	ice_for_each_chnl_tc(i) {
8689 		if (!(vsi->all_enatc & BIT(i)))
8690 			continue;
8691 
8692 		ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8693 		if (!ch) {
8694 			ret = -ENOMEM;
8695 			goto err_free;
8696 		}
8697 		INIT_LIST_HEAD(&ch->list);
8698 		ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8699 		ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8700 		ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8701 		ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8702 		ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8703 
8704 		/* convert to Kbits/s */
8705 		if (ch->max_tx_rate)
8706 			ch->max_tx_rate = div_u64(ch->max_tx_rate,
8707 						  ICE_BW_KBPS_DIVISOR);
8708 		if (ch->min_tx_rate)
8709 			ch->min_tx_rate = div_u64(ch->min_tx_rate,
8710 						  ICE_BW_KBPS_DIVISOR);
8711 
8712 		ret = ice_create_q_channel(vsi, ch);
8713 		if (ret) {
8714 			dev_err(ice_pf_to_dev(pf),
8715 				"failed creating channel TC:%d\n", i);
8716 			kfree(ch);
8717 			goto err_free;
8718 		}
8719 		list_add_tail(&ch->list, &vsi->ch_list);
8720 		vsi->tc_map_vsi[i] = ch->ch_vsi;
8721 		dev_dbg(ice_pf_to_dev(pf),
8722 			"successfully created channel: VSI %pK\n", ch->ch_vsi);
8723 	}
8724 	return 0;
8725 
8726 err_free:
8727 	ice_remove_q_channels(vsi, false);
8728 
8729 	return ret;
8730 }
8731 
8732 /**
8733  * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8734  * @netdev: net device to configure
8735  * @type_data: TC offload data
8736  */
8737 static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8738 {
8739 	struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8740 	struct ice_netdev_priv *np = netdev_priv(netdev);
8741 	struct ice_vsi *vsi = np->vsi;
8742 	struct ice_pf *pf = vsi->back;
8743 	u16 mode, ena_tc_qdisc = 0;
8744 	int cur_txq, cur_rxq;
8745 	u8 hw = 0, num_tcf;
8746 	struct device *dev;
8747 	int ret, i;
8748 
8749 	dev = ice_pf_to_dev(pf);
8750 	num_tcf = mqprio_qopt->qopt.num_tc;
8751 	hw = mqprio_qopt->qopt.hw;
8752 	mode = mqprio_qopt->mode;
8753 	if (!hw) {
8754 		clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8755 		vsi->ch_rss_size = 0;
8756 		memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8757 		goto config_tcf;
8758 	}
8759 
8760 	/* Generate queue region map for number of TCF requested */
8761 	for (i = 0; i < num_tcf; i++)
8762 		ena_tc_qdisc |= BIT(i);
8763 
8764 	switch (mode) {
8765 	case TC_MQPRIO_MODE_CHANNEL:
8766 
8767 		if (pf->hw.port_info->is_custom_tx_enabled) {
8768 			dev_err(dev, "Custom Tx scheduler feature enabled, can't configure ADQ\n");
8769 			return -EBUSY;
8770 		}
8771 		ice_tear_down_devlink_rate_tree(pf);
8772 
8773 		ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8774 		if (ret) {
8775 			netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8776 				   ret);
8777 			return ret;
8778 		}
8779 		memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8780 		set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8781 		/* don't assume state of hw_tc_offload during driver load
8782 		 * and set the flag for TC flower filter if hw_tc_offload
8783 		 * already ON
8784 		 */
8785 		if (vsi->netdev->features & NETIF_F_HW_TC)
8786 			set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
8787 		break;
8788 	default:
8789 		return -EINVAL;
8790 	}
8791 
8792 config_tcf:
8793 
8794 	/* Requesting same TCF configuration as already enabled */
8795 	if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
8796 	    mode != TC_MQPRIO_MODE_CHANNEL)
8797 		return 0;
8798 
8799 	/* Pause VSI queues */
8800 	ice_dis_vsi(vsi, true);
8801 
8802 	if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
8803 		ice_remove_q_channels(vsi, true);
8804 
8805 	if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8806 		vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
8807 				     num_online_cpus());
8808 		vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
8809 				     num_online_cpus());
8810 	} else {
8811 		/* logic to rebuild VSI, same like ethtool -L */
8812 		u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
8813 
8814 		for (i = 0; i < num_tcf; i++) {
8815 			if (!(ena_tc_qdisc & BIT(i)))
8816 				continue;
8817 
8818 			offset = vsi->mqprio_qopt.qopt.offset[i];
8819 			qcount_rx = vsi->mqprio_qopt.qopt.count[i];
8820 			qcount_tx = vsi->mqprio_qopt.qopt.count[i];
8821 		}
8822 		vsi->req_txq = offset + qcount_tx;
8823 		vsi->req_rxq = offset + qcount_rx;
8824 
8825 		/* store away original rss_size info, so that it gets reused
8826 		 * form ice_vsi_rebuild during tc-qdisc delete stage - to
8827 		 * determine, what should be the rss_sizefor main VSI
8828 		 */
8829 		vsi->orig_rss_size = vsi->rss_size;
8830 	}
8831 
8832 	/* save current values of Tx and Rx queues before calling VSI rebuild
8833 	 * for fallback option
8834 	 */
8835 	cur_txq = vsi->num_txq;
8836 	cur_rxq = vsi->num_rxq;
8837 
8838 	/* proceed with rebuild main VSI using correct number of queues */
8839 	ret = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
8840 	if (ret) {
8841 		/* fallback to current number of queues */
8842 		dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
8843 		vsi->req_txq = cur_txq;
8844 		vsi->req_rxq = cur_rxq;
8845 		clear_bit(ICE_RESET_FAILED, pf->state);
8846 		if (ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT)) {
8847 			dev_err(dev, "Rebuild of main VSI failed again\n");
8848 			return ret;
8849 		}
8850 	}
8851 
8852 	vsi->all_numtc = num_tcf;
8853 	vsi->all_enatc = ena_tc_qdisc;
8854 	ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
8855 	if (ret) {
8856 		netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
8857 			   vsi->vsi_num);
8858 		goto exit;
8859 	}
8860 
8861 	if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8862 		u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
8863 		u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
8864 
8865 		/* set TC0 rate limit if specified */
8866 		if (max_tx_rate || min_tx_rate) {
8867 			/* convert to Kbits/s */
8868 			if (max_tx_rate)
8869 				max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
8870 			if (min_tx_rate)
8871 				min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
8872 
8873 			ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
8874 			if (!ret) {
8875 				dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
8876 					max_tx_rate, min_tx_rate, vsi->vsi_num);
8877 			} else {
8878 				dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
8879 					max_tx_rate, min_tx_rate, vsi->vsi_num);
8880 				goto exit;
8881 			}
8882 		}
8883 		ret = ice_create_q_channels(vsi);
8884 		if (ret) {
8885 			netdev_err(netdev, "failed configuring queue channels\n");
8886 			goto exit;
8887 		} else {
8888 			netdev_dbg(netdev, "successfully configured channels\n");
8889 		}
8890 	}
8891 
8892 	if (vsi->ch_rss_size)
8893 		ice_vsi_cfg_rss_lut_key(vsi);
8894 
8895 exit:
8896 	/* if error, reset the all_numtc and all_enatc */
8897 	if (ret) {
8898 		vsi->all_numtc = 0;
8899 		vsi->all_enatc = 0;
8900 	}
8901 	/* resume VSI */
8902 	ice_ena_vsi(vsi, true);
8903 
8904 	return ret;
8905 }
8906 
8907 static LIST_HEAD(ice_block_cb_list);
8908 
8909 static int
8910 ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8911 	     void *type_data)
8912 {
8913 	struct ice_netdev_priv *np = netdev_priv(netdev);
8914 	struct ice_pf *pf = np->vsi->back;
8915 	bool locked = false;
8916 	int err;
8917 
8918 	switch (type) {
8919 	case TC_SETUP_BLOCK:
8920 		return flow_block_cb_setup_simple(type_data,
8921 						  &ice_block_cb_list,
8922 						  ice_setup_tc_block_cb,
8923 						  np, np, true);
8924 	case TC_SETUP_QDISC_MQPRIO:
8925 		if (ice_is_eswitch_mode_switchdev(pf)) {
8926 			netdev_err(netdev, "TC MQPRIO offload not supported, switchdev is enabled\n");
8927 			return -EOPNOTSUPP;
8928 		}
8929 
8930 		if (pf->adev) {
8931 			mutex_lock(&pf->adev_mutex);
8932 			device_lock(&pf->adev->dev);
8933 			locked = true;
8934 			if (pf->adev->dev.driver) {
8935 				netdev_err(netdev, "Cannot change qdisc when RDMA is active\n");
8936 				err = -EBUSY;
8937 				goto adev_unlock;
8938 			}
8939 		}
8940 
8941 		/* setup traffic classifier for receive side */
8942 		mutex_lock(&pf->tc_mutex);
8943 		err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
8944 		mutex_unlock(&pf->tc_mutex);
8945 
8946 adev_unlock:
8947 		if (locked) {
8948 			device_unlock(&pf->adev->dev);
8949 			mutex_unlock(&pf->adev_mutex);
8950 		}
8951 		return err;
8952 	default:
8953 		return -EOPNOTSUPP;
8954 	}
8955 	return -EOPNOTSUPP;
8956 }
8957 
8958 static struct ice_indr_block_priv *
8959 ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
8960 			   struct net_device *netdev)
8961 {
8962 	struct ice_indr_block_priv *cb_priv;
8963 
8964 	list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
8965 		if (!cb_priv->netdev)
8966 			return NULL;
8967 		if (cb_priv->netdev == netdev)
8968 			return cb_priv;
8969 	}
8970 	return NULL;
8971 }
8972 
8973 static int
8974 ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
8975 			void *indr_priv)
8976 {
8977 	struct ice_indr_block_priv *priv = indr_priv;
8978 	struct ice_netdev_priv *np = priv->np;
8979 
8980 	switch (type) {
8981 	case TC_SETUP_CLSFLOWER:
8982 		return ice_setup_tc_cls_flower(np, priv->netdev,
8983 					       (struct flow_cls_offload *)
8984 					       type_data);
8985 	default:
8986 		return -EOPNOTSUPP;
8987 	}
8988 }
8989 
8990 static int
8991 ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
8992 			struct ice_netdev_priv *np,
8993 			struct flow_block_offload *f, void *data,
8994 			void (*cleanup)(struct flow_block_cb *block_cb))
8995 {
8996 	struct ice_indr_block_priv *indr_priv;
8997 	struct flow_block_cb *block_cb;
8998 
8999 	if (!ice_is_tunnel_supported(netdev) &&
9000 	    !(is_vlan_dev(netdev) &&
9001 	      vlan_dev_real_dev(netdev) == np->vsi->netdev))
9002 		return -EOPNOTSUPP;
9003 
9004 	if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9005 		return -EOPNOTSUPP;
9006 
9007 	switch (f->command) {
9008 	case FLOW_BLOCK_BIND:
9009 		indr_priv = ice_indr_block_priv_lookup(np, netdev);
9010 		if (indr_priv)
9011 			return -EEXIST;
9012 
9013 		indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
9014 		if (!indr_priv)
9015 			return -ENOMEM;
9016 
9017 		indr_priv->netdev = netdev;
9018 		indr_priv->np = np;
9019 		list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
9020 
9021 		block_cb =
9022 			flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
9023 						 indr_priv, indr_priv,
9024 						 ice_rep_indr_tc_block_unbind,
9025 						 f, netdev, sch, data, np,
9026 						 cleanup);
9027 
9028 		if (IS_ERR(block_cb)) {
9029 			list_del(&indr_priv->list);
9030 			kfree(indr_priv);
9031 			return PTR_ERR(block_cb);
9032 		}
9033 		flow_block_cb_add(block_cb, f);
9034 		list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
9035 		break;
9036 	case FLOW_BLOCK_UNBIND:
9037 		indr_priv = ice_indr_block_priv_lookup(np, netdev);
9038 		if (!indr_priv)
9039 			return -ENOENT;
9040 
9041 		block_cb = flow_block_cb_lookup(f->block,
9042 						ice_indr_setup_block_cb,
9043 						indr_priv);
9044 		if (!block_cb)
9045 			return -ENOENT;
9046 
9047 		flow_indr_block_cb_remove(block_cb, f);
9048 
9049 		list_del(&block_cb->driver_list);
9050 		break;
9051 	default:
9052 		return -EOPNOTSUPP;
9053 	}
9054 	return 0;
9055 }
9056 
9057 static int
9058 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
9059 		     void *cb_priv, enum tc_setup_type type, void *type_data,
9060 		     void *data,
9061 		     void (*cleanup)(struct flow_block_cb *block_cb))
9062 {
9063 	switch (type) {
9064 	case TC_SETUP_BLOCK:
9065 		return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
9066 					       data, cleanup);
9067 
9068 	default:
9069 		return -EOPNOTSUPP;
9070 	}
9071 }
9072 
9073 /**
9074  * ice_open - Called when a network interface becomes active
9075  * @netdev: network interface device structure
9076  *
9077  * The open entry point is called when a network interface is made
9078  * active by the system (IFF_UP). At this point all resources needed
9079  * for transmit and receive operations are allocated, the interrupt
9080  * handler is registered with the OS, the netdev watchdog is enabled,
9081  * and the stack is notified that the interface is ready.
9082  *
9083  * Returns 0 on success, negative value on failure
9084  */
9085 int ice_open(struct net_device *netdev)
9086 {
9087 	struct ice_netdev_priv *np = netdev_priv(netdev);
9088 	struct ice_pf *pf = np->vsi->back;
9089 
9090 	if (ice_is_reset_in_progress(pf->state)) {
9091 		netdev_err(netdev, "can't open net device while reset is in progress");
9092 		return -EBUSY;
9093 	}
9094 
9095 	return ice_open_internal(netdev);
9096 }
9097 
9098 /**
9099  * ice_open_internal - Called when a network interface becomes active
9100  * @netdev: network interface device structure
9101  *
9102  * Internal ice_open implementation. Should not be used directly except for ice_open and reset
9103  * handling routine
9104  *
9105  * Returns 0 on success, negative value on failure
9106  */
9107 int ice_open_internal(struct net_device *netdev)
9108 {
9109 	struct ice_netdev_priv *np = netdev_priv(netdev);
9110 	struct ice_vsi *vsi = np->vsi;
9111 	struct ice_pf *pf = vsi->back;
9112 	struct ice_port_info *pi;
9113 	int err;
9114 
9115 	if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
9116 		netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
9117 		return -EIO;
9118 	}
9119 
9120 	netif_carrier_off(netdev);
9121 
9122 	pi = vsi->port_info;
9123 	err = ice_update_link_info(pi);
9124 	if (err) {
9125 		netdev_err(netdev, "Failed to get link info, error %d\n", err);
9126 		return err;
9127 	}
9128 
9129 	ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
9130 
9131 	/* Set PHY if there is media, otherwise, turn off PHY */
9132 	if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
9133 		clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9134 		if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
9135 			err = ice_init_phy_user_cfg(pi);
9136 			if (err) {
9137 				netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
9138 					   err);
9139 				return err;
9140 			}
9141 		}
9142 
9143 		err = ice_configure_phy(vsi);
9144 		if (err) {
9145 			netdev_err(netdev, "Failed to set physical link up, error %d\n",
9146 				   err);
9147 			return err;
9148 		}
9149 	} else {
9150 		set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9151 		ice_set_link(vsi, false);
9152 	}
9153 
9154 	err = ice_vsi_open(vsi);
9155 	if (err)
9156 		netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
9157 			   vsi->vsi_num, vsi->vsw->sw_id);
9158 
9159 	/* Update existing tunnels information */
9160 	udp_tunnel_get_rx_info(netdev);
9161 
9162 	return err;
9163 }
9164 
9165 /**
9166  * ice_stop - Disables a network interface
9167  * @netdev: network interface device structure
9168  *
9169  * The stop entry point is called when an interface is de-activated by the OS,
9170  * and the netdevice enters the DOWN state. The hardware is still under the
9171  * driver's control, but the netdev interface is disabled.
9172  *
9173  * Returns success only - not allowed to fail
9174  */
9175 int ice_stop(struct net_device *netdev)
9176 {
9177 	struct ice_netdev_priv *np = netdev_priv(netdev);
9178 	struct ice_vsi *vsi = np->vsi;
9179 	struct ice_pf *pf = vsi->back;
9180 
9181 	if (ice_is_reset_in_progress(pf->state)) {
9182 		netdev_err(netdev, "can't stop net device while reset is in progress");
9183 		return -EBUSY;
9184 	}
9185 
9186 	if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
9187 		int link_err = ice_force_phys_link_state(vsi, false);
9188 
9189 		if (link_err) {
9190 			netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
9191 				   vsi->vsi_num, link_err);
9192 			return -EIO;
9193 		}
9194 	}
9195 
9196 	ice_vsi_close(vsi);
9197 
9198 	return 0;
9199 }
9200 
9201 /**
9202  * ice_features_check - Validate encapsulated packet conforms to limits
9203  * @skb: skb buffer
9204  * @netdev: This port's netdev
9205  * @features: Offload features that the stack believes apply
9206  */
9207 static netdev_features_t
9208 ice_features_check(struct sk_buff *skb,
9209 		   struct net_device __always_unused *netdev,
9210 		   netdev_features_t features)
9211 {
9212 	bool gso = skb_is_gso(skb);
9213 	size_t len;
9214 
9215 	/* No point in doing any of this if neither checksum nor GSO are
9216 	 * being requested for this frame. We can rule out both by just
9217 	 * checking for CHECKSUM_PARTIAL
9218 	 */
9219 	if (skb->ip_summed != CHECKSUM_PARTIAL)
9220 		return features;
9221 
9222 	/* We cannot support GSO if the MSS is going to be less than
9223 	 * 64 bytes. If it is then we need to drop support for GSO.
9224 	 */
9225 	if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
9226 		features &= ~NETIF_F_GSO_MASK;
9227 
9228 	len = skb_network_offset(skb);
9229 	if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
9230 		goto out_rm_features;
9231 
9232 	len = skb_network_header_len(skb);
9233 	if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9234 		goto out_rm_features;
9235 
9236 	if (skb->encapsulation) {
9237 		/* this must work for VXLAN frames AND IPIP/SIT frames, and in
9238 		 * the case of IPIP frames, the transport header pointer is
9239 		 * after the inner header! So check to make sure that this
9240 		 * is a GRE or UDP_TUNNEL frame before doing that math.
9241 		 */
9242 		if (gso && (skb_shinfo(skb)->gso_type &
9243 			    (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
9244 			len = skb_inner_network_header(skb) -
9245 			      skb_transport_header(skb);
9246 			if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
9247 				goto out_rm_features;
9248 		}
9249 
9250 		len = skb_inner_network_header_len(skb);
9251 		if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9252 			goto out_rm_features;
9253 	}
9254 
9255 	return features;
9256 out_rm_features:
9257 	return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9258 }
9259 
9260 static const struct net_device_ops ice_netdev_safe_mode_ops = {
9261 	.ndo_open = ice_open,
9262 	.ndo_stop = ice_stop,
9263 	.ndo_start_xmit = ice_start_xmit,
9264 	.ndo_set_mac_address = ice_set_mac_address,
9265 	.ndo_validate_addr = eth_validate_addr,
9266 	.ndo_change_mtu = ice_change_mtu,
9267 	.ndo_get_stats64 = ice_get_stats64,
9268 	.ndo_tx_timeout = ice_tx_timeout,
9269 	.ndo_bpf = ice_xdp_safe_mode,
9270 };
9271 
9272 static const struct net_device_ops ice_netdev_ops = {
9273 	.ndo_open = ice_open,
9274 	.ndo_stop = ice_stop,
9275 	.ndo_start_xmit = ice_start_xmit,
9276 	.ndo_select_queue = ice_select_queue,
9277 	.ndo_features_check = ice_features_check,
9278 	.ndo_fix_features = ice_fix_features,
9279 	.ndo_set_rx_mode = ice_set_rx_mode,
9280 	.ndo_set_mac_address = ice_set_mac_address,
9281 	.ndo_validate_addr = eth_validate_addr,
9282 	.ndo_change_mtu = ice_change_mtu,
9283 	.ndo_get_stats64 = ice_get_stats64,
9284 	.ndo_set_tx_maxrate = ice_set_tx_maxrate,
9285 	.ndo_eth_ioctl = ice_eth_ioctl,
9286 	.ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
9287 	.ndo_set_vf_mac = ice_set_vf_mac,
9288 	.ndo_get_vf_config = ice_get_vf_cfg,
9289 	.ndo_set_vf_trust = ice_set_vf_trust,
9290 	.ndo_set_vf_vlan = ice_set_vf_port_vlan,
9291 	.ndo_set_vf_link_state = ice_set_vf_link_state,
9292 	.ndo_get_vf_stats = ice_get_vf_stats,
9293 	.ndo_set_vf_rate = ice_set_vf_bw,
9294 	.ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
9295 	.ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
9296 	.ndo_setup_tc = ice_setup_tc,
9297 	.ndo_set_features = ice_set_features,
9298 	.ndo_bridge_getlink = ice_bridge_getlink,
9299 	.ndo_bridge_setlink = ice_bridge_setlink,
9300 	.ndo_fdb_add = ice_fdb_add,
9301 	.ndo_fdb_del = ice_fdb_del,
9302 #ifdef CONFIG_RFS_ACCEL
9303 	.ndo_rx_flow_steer = ice_rx_flow_steer,
9304 #endif
9305 	.ndo_tx_timeout = ice_tx_timeout,
9306 	.ndo_bpf = ice_xdp,
9307 	.ndo_xdp_xmit = ice_xdp_xmit,
9308 	.ndo_xsk_wakeup = ice_xsk_wakeup,
9309 };
9310