1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, 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 "ice.h"
10 #include "ice_base.h"
11 #include "ice_lib.h"
12 #include "ice_fltr.h"
13 #include "ice_dcb_lib.h"
14 #include "ice_dcb_nl.h"
15 #include "ice_devlink.h"
16 /* Including ice_trace.h with CREATE_TRACE_POINTS defined will generate the
17  * ice tracepoint functions. This must be done exactly once across the
18  * ice driver.
19  */
20 #define CREATE_TRACE_POINTS
21 #include "ice_trace.h"
22 #include "ice_eswitch.h"
23 #include "ice_tc_lib.h"
24 #include "ice_vsi_vlan_ops.h"
25 
26 #define DRV_SUMMARY	"Intel(R) Ethernet Connection E800 Series Linux Driver"
27 static const char ice_driver_string[] = DRV_SUMMARY;
28 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
29 
30 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
31 #define ICE_DDP_PKG_PATH	"intel/ice/ddp/"
32 #define ICE_DDP_PKG_FILE	ICE_DDP_PKG_PATH "ice.pkg"
33 
34 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
35 MODULE_DESCRIPTION(DRV_SUMMARY);
36 MODULE_LICENSE("GPL v2");
37 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
38 
39 static int debug = -1;
40 module_param(debug, int, 0644);
41 #ifndef CONFIG_DYNAMIC_DEBUG
42 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
43 #else
44 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
45 #endif /* !CONFIG_DYNAMIC_DEBUG */
46 
47 static DEFINE_IDA(ice_aux_ida);
48 DEFINE_STATIC_KEY_FALSE(ice_xdp_locking_key);
49 EXPORT_SYMBOL(ice_xdp_locking_key);
50 
51 /**
52  * ice_hw_to_dev - Get device pointer from the hardware structure
53  * @hw: pointer to the device HW structure
54  *
55  * Used to access the device pointer from compilation units which can't easily
56  * include the definition of struct ice_pf without leading to circular header
57  * dependencies.
58  */
59 struct device *ice_hw_to_dev(struct ice_hw *hw)
60 {
61 	struct ice_pf *pf = container_of(hw, struct ice_pf, hw);
62 
63 	return &pf->pdev->dev;
64 }
65 
66 static struct workqueue_struct *ice_wq;
67 static const struct net_device_ops ice_netdev_safe_mode_ops;
68 static const struct net_device_ops ice_netdev_ops;
69 
70 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
71 
72 static void ice_vsi_release_all(struct ice_pf *pf);
73 
74 static int ice_rebuild_channels(struct ice_pf *pf);
75 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_adv_fltr);
76 
77 static int
78 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
79 		     void *cb_priv, enum tc_setup_type type, void *type_data,
80 		     void *data,
81 		     void (*cleanup)(struct flow_block_cb *block_cb));
82 
83 bool netif_is_ice(struct net_device *dev)
84 {
85 	return dev && (dev->netdev_ops == &ice_netdev_ops);
86 }
87 
88 /**
89  * ice_get_tx_pending - returns number of Tx descriptors not processed
90  * @ring: the ring of descriptors
91  */
92 static u16 ice_get_tx_pending(struct ice_tx_ring *ring)
93 {
94 	u16 head, tail;
95 
96 	head = ring->next_to_clean;
97 	tail = ring->next_to_use;
98 
99 	if (head != tail)
100 		return (head < tail) ?
101 			tail - head : (tail + ring->count - head);
102 	return 0;
103 }
104 
105 /**
106  * ice_check_for_hang_subtask - check for and recover hung queues
107  * @pf: pointer to PF struct
108  */
109 static void ice_check_for_hang_subtask(struct ice_pf *pf)
110 {
111 	struct ice_vsi *vsi = NULL;
112 	struct ice_hw *hw;
113 	unsigned int i;
114 	int packets;
115 	u32 v;
116 
117 	ice_for_each_vsi(pf, v)
118 		if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
119 			vsi = pf->vsi[v];
120 			break;
121 		}
122 
123 	if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
124 		return;
125 
126 	if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
127 		return;
128 
129 	hw = &vsi->back->hw;
130 
131 	ice_for_each_txq(vsi, i) {
132 		struct ice_tx_ring *tx_ring = vsi->tx_rings[i];
133 
134 		if (!tx_ring)
135 			continue;
136 		if (ice_ring_ch_enabled(tx_ring))
137 			continue;
138 
139 		if (tx_ring->desc) {
140 			/* If packet counter has not changed the queue is
141 			 * likely stalled, so force an interrupt for this
142 			 * queue.
143 			 *
144 			 * prev_pkt would be negative if there was no
145 			 * pending work.
146 			 */
147 			packets = tx_ring->stats.pkts & INT_MAX;
148 			if (tx_ring->tx_stats.prev_pkt == packets) {
149 				/* Trigger sw interrupt to revive the queue */
150 				ice_trigger_sw_intr(hw, tx_ring->q_vector);
151 				continue;
152 			}
153 
154 			/* Memory barrier between read of packet count and call
155 			 * to ice_get_tx_pending()
156 			 */
157 			smp_rmb();
158 			tx_ring->tx_stats.prev_pkt =
159 			    ice_get_tx_pending(tx_ring) ? packets : -1;
160 		}
161 	}
162 }
163 
164 /**
165  * ice_init_mac_fltr - Set initial MAC filters
166  * @pf: board private structure
167  *
168  * Set initial set of MAC filters for PF VSI; configure filters for permanent
169  * address and broadcast address. If an error is encountered, netdevice will be
170  * unregistered.
171  */
172 static int ice_init_mac_fltr(struct ice_pf *pf)
173 {
174 	struct ice_vsi *vsi;
175 	u8 *perm_addr;
176 
177 	vsi = ice_get_main_vsi(pf);
178 	if (!vsi)
179 		return -EINVAL;
180 
181 	perm_addr = vsi->port_info->mac.perm_addr;
182 	return ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
183 }
184 
185 /**
186  * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
187  * @netdev: the net device on which the sync is happening
188  * @addr: MAC address to sync
189  *
190  * This is a callback function which is called by the in kernel device sync
191  * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
192  * populates the tmp_sync_list, which is later used by ice_add_mac to add the
193  * MAC filters from the hardware.
194  */
195 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
196 {
197 	struct ice_netdev_priv *np = netdev_priv(netdev);
198 	struct ice_vsi *vsi = np->vsi;
199 
200 	if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
201 				     ICE_FWD_TO_VSI))
202 		return -EINVAL;
203 
204 	return 0;
205 }
206 
207 /**
208  * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
209  * @netdev: the net device on which the unsync is happening
210  * @addr: MAC address to unsync
211  *
212  * This is a callback function which is called by the in kernel device unsync
213  * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
214  * populates the tmp_unsync_list, which is later used by ice_remove_mac to
215  * delete the MAC filters from the hardware.
216  */
217 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
218 {
219 	struct ice_netdev_priv *np = netdev_priv(netdev);
220 	struct ice_vsi *vsi = np->vsi;
221 
222 	/* Under some circumstances, we might receive a request to delete our
223 	 * own device address from our uc list. Because we store the device
224 	 * address in the VSI's MAC filter list, we need to ignore such
225 	 * requests and not delete our device address from this list.
226 	 */
227 	if (ether_addr_equal(addr, netdev->dev_addr))
228 		return 0;
229 
230 	if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
231 				     ICE_FWD_TO_VSI))
232 		return -EINVAL;
233 
234 	return 0;
235 }
236 
237 /**
238  * ice_vsi_fltr_changed - check if filter state changed
239  * @vsi: VSI to be checked
240  *
241  * returns true if filter state has changed, false otherwise.
242  */
243 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
244 {
245 	return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
246 	       test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
247 }
248 
249 /**
250  * ice_set_promisc - Enable promiscuous mode for a given PF
251  * @vsi: the VSI being configured
252  * @promisc_m: mask of promiscuous config bits
253  *
254  */
255 static int ice_set_promisc(struct ice_vsi *vsi, u8 promisc_m)
256 {
257 	int status;
258 
259 	if (vsi->type != ICE_VSI_PF)
260 		return 0;
261 
262 	if (ice_vsi_has_non_zero_vlans(vsi)) {
263 		promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
264 		status = ice_fltr_set_vlan_vsi_promisc(&vsi->back->hw, vsi,
265 						       promisc_m);
266 	} else {
267 		status = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
268 						  promisc_m, 0);
269 	}
270 
271 	return status;
272 }
273 
274 /**
275  * ice_clear_promisc - Disable promiscuous mode for a given PF
276  * @vsi: the VSI being configured
277  * @promisc_m: mask of promiscuous config bits
278  *
279  */
280 static int ice_clear_promisc(struct ice_vsi *vsi, u8 promisc_m)
281 {
282 	int status;
283 
284 	if (vsi->type != ICE_VSI_PF)
285 		return 0;
286 
287 	if (ice_vsi_has_non_zero_vlans(vsi)) {
288 		promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
289 		status = ice_fltr_clear_vlan_vsi_promisc(&vsi->back->hw, vsi,
290 							 promisc_m);
291 	} else {
292 		status = ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
293 						    promisc_m, 0);
294 	}
295 
296 	return status;
297 }
298 
299 /**
300  * ice_get_devlink_port - Get devlink port from netdev
301  * @netdev: the netdevice structure
302  */
303 static struct devlink_port *ice_get_devlink_port(struct net_device *netdev)
304 {
305 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
306 
307 	if (!ice_is_switchdev_running(pf))
308 		return NULL;
309 
310 	return &pf->devlink_port;
311 }
312 
313 /**
314  * ice_vsi_sync_fltr - Update the VSI filter list to the HW
315  * @vsi: ptr to the VSI
316  *
317  * Push any outstanding VSI filter changes through the AdminQ.
318  */
319 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
320 {
321 	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
322 	struct device *dev = ice_pf_to_dev(vsi->back);
323 	struct net_device *netdev = vsi->netdev;
324 	bool promisc_forced_on = false;
325 	struct ice_pf *pf = vsi->back;
326 	struct ice_hw *hw = &pf->hw;
327 	u32 changed_flags = 0;
328 	int err;
329 
330 	if (!vsi->netdev)
331 		return -EINVAL;
332 
333 	while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
334 		usleep_range(1000, 2000);
335 
336 	changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
337 	vsi->current_netdev_flags = vsi->netdev->flags;
338 
339 	INIT_LIST_HEAD(&vsi->tmp_sync_list);
340 	INIT_LIST_HEAD(&vsi->tmp_unsync_list);
341 
342 	if (ice_vsi_fltr_changed(vsi)) {
343 		clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
344 		clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
345 
346 		/* grab the netdev's addr_list_lock */
347 		netif_addr_lock_bh(netdev);
348 		__dev_uc_sync(netdev, ice_add_mac_to_sync_list,
349 			      ice_add_mac_to_unsync_list);
350 		__dev_mc_sync(netdev, ice_add_mac_to_sync_list,
351 			      ice_add_mac_to_unsync_list);
352 		/* our temp lists are populated. release lock */
353 		netif_addr_unlock_bh(netdev);
354 	}
355 
356 	/* Remove MAC addresses in the unsync list */
357 	err = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
358 	ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
359 	if (err) {
360 		netdev_err(netdev, "Failed to delete MAC filters\n");
361 		/* if we failed because of alloc failures, just bail */
362 		if (err == -ENOMEM)
363 			goto out;
364 	}
365 
366 	/* Add MAC addresses in the sync list */
367 	err = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
368 	ice_fltr_free_list(dev, &vsi->tmp_sync_list);
369 	/* If filter is added successfully or already exists, do not go into
370 	 * 'if' condition and report it as error. Instead continue processing
371 	 * rest of the function.
372 	 */
373 	if (err && err != -EEXIST) {
374 		netdev_err(netdev, "Failed to add MAC filters\n");
375 		/* If there is no more space for new umac filters, VSI
376 		 * should go into promiscuous mode. There should be some
377 		 * space reserved for promiscuous filters.
378 		 */
379 		if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
380 		    !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
381 				      vsi->state)) {
382 			promisc_forced_on = true;
383 			netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
384 				    vsi->vsi_num);
385 		} else {
386 			goto out;
387 		}
388 	}
389 	err = 0;
390 	/* check for changes in promiscuous modes */
391 	if (changed_flags & IFF_ALLMULTI) {
392 		if (vsi->current_netdev_flags & IFF_ALLMULTI) {
393 			err = ice_set_promisc(vsi, ICE_MCAST_PROMISC_BITS);
394 			if (err) {
395 				vsi->current_netdev_flags &= ~IFF_ALLMULTI;
396 				goto out_promisc;
397 			}
398 		} else {
399 			/* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
400 			err = ice_clear_promisc(vsi, ICE_MCAST_PROMISC_BITS);
401 			if (err) {
402 				vsi->current_netdev_flags |= IFF_ALLMULTI;
403 				goto out_promisc;
404 			}
405 		}
406 	}
407 
408 	if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
409 	    test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
410 		clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
411 		if (vsi->current_netdev_flags & IFF_PROMISC) {
412 			/* Apply Rx filter rule to get traffic from wire */
413 			if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
414 				err = ice_set_dflt_vsi(pf->first_sw, vsi);
415 				if (err && err != -EEXIST) {
416 					netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
417 						   err, vsi->vsi_num);
418 					vsi->current_netdev_flags &=
419 						~IFF_PROMISC;
420 					goto out_promisc;
421 				}
422 				err = 0;
423 				vlan_ops->dis_rx_filtering(vsi);
424 			}
425 		} else {
426 			/* Clear Rx filter to remove traffic from wire */
427 			if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
428 				err = ice_clear_dflt_vsi(pf->first_sw);
429 				if (err) {
430 					netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
431 						   err, vsi->vsi_num);
432 					vsi->current_netdev_flags |=
433 						IFF_PROMISC;
434 					goto out_promisc;
435 				}
436 				if (vsi->current_netdev_flags &
437 				    NETIF_F_HW_VLAN_CTAG_FILTER)
438 					vlan_ops->ena_rx_filtering(vsi);
439 			}
440 		}
441 	}
442 	goto exit;
443 
444 out_promisc:
445 	set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
446 	goto exit;
447 out:
448 	/* if something went wrong then set the changed flag so we try again */
449 	set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
450 	set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
451 exit:
452 	clear_bit(ICE_CFG_BUSY, vsi->state);
453 	return err;
454 }
455 
456 /**
457  * ice_sync_fltr_subtask - Sync the VSI filter list with HW
458  * @pf: board private structure
459  */
460 static void ice_sync_fltr_subtask(struct ice_pf *pf)
461 {
462 	int v;
463 
464 	if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
465 		return;
466 
467 	clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
468 
469 	ice_for_each_vsi(pf, v)
470 		if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
471 		    ice_vsi_sync_fltr(pf->vsi[v])) {
472 			/* come back and try again later */
473 			set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
474 			break;
475 		}
476 }
477 
478 /**
479  * ice_pf_dis_all_vsi - Pause all VSIs on a PF
480  * @pf: the PF
481  * @locked: is the rtnl_lock already held
482  */
483 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
484 {
485 	int node;
486 	int v;
487 
488 	ice_for_each_vsi(pf, v)
489 		if (pf->vsi[v])
490 			ice_dis_vsi(pf->vsi[v], locked);
491 
492 	for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
493 		pf->pf_agg_node[node].num_vsis = 0;
494 
495 	for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
496 		pf->vf_agg_node[node].num_vsis = 0;
497 }
498 
499 /**
500  * ice_clear_sw_switch_recipes - clear switch recipes
501  * @pf: board private structure
502  *
503  * Mark switch recipes as not created in sw structures. There are cases where
504  * rules (especially advanced rules) need to be restored, either re-read from
505  * hardware or added again. For example after the reset. 'recp_created' flag
506  * prevents from doing that and need to be cleared upfront.
507  */
508 static void ice_clear_sw_switch_recipes(struct ice_pf *pf)
509 {
510 	struct ice_sw_recipe *recp;
511 	u8 i;
512 
513 	recp = pf->hw.switch_info->recp_list;
514 	for (i = 0; i < ICE_MAX_NUM_RECIPES; i++)
515 		recp[i].recp_created = false;
516 }
517 
518 /**
519  * ice_prepare_for_reset - prep for reset
520  * @pf: board private structure
521  * @reset_type: reset type requested
522  *
523  * Inform or close all dependent features in prep for reset.
524  */
525 static void
526 ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
527 {
528 	struct ice_hw *hw = &pf->hw;
529 	struct ice_vsi *vsi;
530 	struct ice_vf *vf;
531 	unsigned int bkt;
532 
533 	dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type);
534 
535 	/* already prepared for reset */
536 	if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
537 		return;
538 
539 	ice_unplug_aux_dev(pf);
540 
541 	/* Notify VFs of impending reset */
542 	if (ice_check_sq_alive(hw, &hw->mailboxq))
543 		ice_vc_notify_reset(pf);
544 
545 	/* Disable VFs until reset is completed */
546 	mutex_lock(&pf->vfs.table_lock);
547 	ice_for_each_vf(pf, bkt, vf)
548 		ice_set_vf_state_qs_dis(vf);
549 	mutex_unlock(&pf->vfs.table_lock);
550 
551 	if (ice_is_eswitch_mode_switchdev(pf)) {
552 		if (reset_type != ICE_RESET_PFR)
553 			ice_clear_sw_switch_recipes(pf);
554 	}
555 
556 	/* release ADQ specific HW and SW resources */
557 	vsi = ice_get_main_vsi(pf);
558 	if (!vsi)
559 		goto skip;
560 
561 	/* to be on safe side, reset orig_rss_size so that normal flow
562 	 * of deciding rss_size can take precedence
563 	 */
564 	vsi->orig_rss_size = 0;
565 
566 	if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
567 		if (reset_type == ICE_RESET_PFR) {
568 			vsi->old_ena_tc = vsi->all_enatc;
569 			vsi->old_numtc = vsi->all_numtc;
570 		} else {
571 			ice_remove_q_channels(vsi, true);
572 
573 			/* for other reset type, do not support channel rebuild
574 			 * hence reset needed info
575 			 */
576 			vsi->old_ena_tc = 0;
577 			vsi->all_enatc = 0;
578 			vsi->old_numtc = 0;
579 			vsi->all_numtc = 0;
580 			vsi->req_txq = 0;
581 			vsi->req_rxq = 0;
582 			clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
583 			memset(&vsi->mqprio_qopt, 0, sizeof(vsi->mqprio_qopt));
584 		}
585 	}
586 skip:
587 
588 	/* clear SW filtering DB */
589 	ice_clear_hw_tbls(hw);
590 	/* disable the VSIs and their queues that are not already DOWN */
591 	ice_pf_dis_all_vsi(pf, false);
592 
593 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
594 		ice_ptp_prepare_for_reset(pf);
595 
596 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
597 		ice_gnss_exit(pf);
598 
599 	if (hw->port_info)
600 		ice_sched_clear_port(hw->port_info);
601 
602 	ice_shutdown_all_ctrlq(hw);
603 
604 	set_bit(ICE_PREPARED_FOR_RESET, pf->state);
605 }
606 
607 /**
608  * ice_do_reset - Initiate one of many types of resets
609  * @pf: board private structure
610  * @reset_type: reset type requested before this function was called.
611  */
612 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
613 {
614 	struct device *dev = ice_pf_to_dev(pf);
615 	struct ice_hw *hw = &pf->hw;
616 
617 	dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
618 
619 	ice_prepare_for_reset(pf, reset_type);
620 
621 	/* trigger the reset */
622 	if (ice_reset(hw, reset_type)) {
623 		dev_err(dev, "reset %d failed\n", reset_type);
624 		set_bit(ICE_RESET_FAILED, pf->state);
625 		clear_bit(ICE_RESET_OICR_RECV, pf->state);
626 		clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
627 		clear_bit(ICE_PFR_REQ, pf->state);
628 		clear_bit(ICE_CORER_REQ, pf->state);
629 		clear_bit(ICE_GLOBR_REQ, pf->state);
630 		wake_up(&pf->reset_wait_queue);
631 		return;
632 	}
633 
634 	/* PFR is a bit of a special case because it doesn't result in an OICR
635 	 * interrupt. So for PFR, rebuild after the reset and clear the reset-
636 	 * associated state bits.
637 	 */
638 	if (reset_type == ICE_RESET_PFR) {
639 		pf->pfr_count++;
640 		ice_rebuild(pf, reset_type);
641 		clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
642 		clear_bit(ICE_PFR_REQ, pf->state);
643 		wake_up(&pf->reset_wait_queue);
644 		ice_reset_all_vfs(pf);
645 	}
646 }
647 
648 /**
649  * ice_reset_subtask - Set up for resetting the device and driver
650  * @pf: board private structure
651  */
652 static void ice_reset_subtask(struct ice_pf *pf)
653 {
654 	enum ice_reset_req reset_type = ICE_RESET_INVAL;
655 
656 	/* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
657 	 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
658 	 * of reset is pending and sets bits in pf->state indicating the reset
659 	 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
660 	 * prepare for pending reset if not already (for PF software-initiated
661 	 * global resets the software should already be prepared for it as
662 	 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
663 	 * by firmware or software on other PFs, that bit is not set so prepare
664 	 * for the reset now), poll for reset done, rebuild and return.
665 	 */
666 	if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
667 		/* Perform the largest reset requested */
668 		if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
669 			reset_type = ICE_RESET_CORER;
670 		if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
671 			reset_type = ICE_RESET_GLOBR;
672 		if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
673 			reset_type = ICE_RESET_EMPR;
674 		/* return if no valid reset type requested */
675 		if (reset_type == ICE_RESET_INVAL)
676 			return;
677 		ice_prepare_for_reset(pf, reset_type);
678 
679 		/* make sure we are ready to rebuild */
680 		if (ice_check_reset(&pf->hw)) {
681 			set_bit(ICE_RESET_FAILED, pf->state);
682 		} else {
683 			/* done with reset. start rebuild */
684 			pf->hw.reset_ongoing = false;
685 			ice_rebuild(pf, reset_type);
686 			/* clear bit to resume normal operations, but
687 			 * ICE_NEEDS_RESTART bit is set in case rebuild failed
688 			 */
689 			clear_bit(ICE_RESET_OICR_RECV, pf->state);
690 			clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
691 			clear_bit(ICE_PFR_REQ, pf->state);
692 			clear_bit(ICE_CORER_REQ, pf->state);
693 			clear_bit(ICE_GLOBR_REQ, pf->state);
694 			wake_up(&pf->reset_wait_queue);
695 			ice_reset_all_vfs(pf);
696 		}
697 
698 		return;
699 	}
700 
701 	/* No pending resets to finish processing. Check for new resets */
702 	if (test_bit(ICE_PFR_REQ, pf->state))
703 		reset_type = ICE_RESET_PFR;
704 	if (test_bit(ICE_CORER_REQ, pf->state))
705 		reset_type = ICE_RESET_CORER;
706 	if (test_bit(ICE_GLOBR_REQ, pf->state))
707 		reset_type = ICE_RESET_GLOBR;
708 	/* If no valid reset type requested just return */
709 	if (reset_type == ICE_RESET_INVAL)
710 		return;
711 
712 	/* reset if not already down or busy */
713 	if (!test_bit(ICE_DOWN, pf->state) &&
714 	    !test_bit(ICE_CFG_BUSY, pf->state)) {
715 		ice_do_reset(pf, reset_type);
716 	}
717 }
718 
719 /**
720  * ice_print_topo_conflict - print topology conflict message
721  * @vsi: the VSI whose topology status is being checked
722  */
723 static void ice_print_topo_conflict(struct ice_vsi *vsi)
724 {
725 	switch (vsi->port_info->phy.link_info.topo_media_conflict) {
726 	case ICE_AQ_LINK_TOPO_CONFLICT:
727 	case ICE_AQ_LINK_MEDIA_CONFLICT:
728 	case ICE_AQ_LINK_TOPO_UNREACH_PRT:
729 	case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
730 	case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
731 		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");
732 		break;
733 	case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
734 		if (test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, vsi->back->flags))
735 			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");
736 		else
737 			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");
738 		break;
739 	default:
740 		break;
741 	}
742 }
743 
744 /**
745  * ice_print_link_msg - print link up or down message
746  * @vsi: the VSI whose link status is being queried
747  * @isup: boolean for if the link is now up or down
748  */
749 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
750 {
751 	struct ice_aqc_get_phy_caps_data *caps;
752 	const char *an_advertised;
753 	const char *fec_req;
754 	const char *speed;
755 	const char *fec;
756 	const char *fc;
757 	const char *an;
758 	int status;
759 
760 	if (!vsi)
761 		return;
762 
763 	if (vsi->current_isup == isup)
764 		return;
765 
766 	vsi->current_isup = isup;
767 
768 	if (!isup) {
769 		netdev_info(vsi->netdev, "NIC Link is Down\n");
770 		return;
771 	}
772 
773 	switch (vsi->port_info->phy.link_info.link_speed) {
774 	case ICE_AQ_LINK_SPEED_100GB:
775 		speed = "100 G";
776 		break;
777 	case ICE_AQ_LINK_SPEED_50GB:
778 		speed = "50 G";
779 		break;
780 	case ICE_AQ_LINK_SPEED_40GB:
781 		speed = "40 G";
782 		break;
783 	case ICE_AQ_LINK_SPEED_25GB:
784 		speed = "25 G";
785 		break;
786 	case ICE_AQ_LINK_SPEED_20GB:
787 		speed = "20 G";
788 		break;
789 	case ICE_AQ_LINK_SPEED_10GB:
790 		speed = "10 G";
791 		break;
792 	case ICE_AQ_LINK_SPEED_5GB:
793 		speed = "5 G";
794 		break;
795 	case ICE_AQ_LINK_SPEED_2500MB:
796 		speed = "2.5 G";
797 		break;
798 	case ICE_AQ_LINK_SPEED_1000MB:
799 		speed = "1 G";
800 		break;
801 	case ICE_AQ_LINK_SPEED_100MB:
802 		speed = "100 M";
803 		break;
804 	default:
805 		speed = "Unknown ";
806 		break;
807 	}
808 
809 	switch (vsi->port_info->fc.current_mode) {
810 	case ICE_FC_FULL:
811 		fc = "Rx/Tx";
812 		break;
813 	case ICE_FC_TX_PAUSE:
814 		fc = "Tx";
815 		break;
816 	case ICE_FC_RX_PAUSE:
817 		fc = "Rx";
818 		break;
819 	case ICE_FC_NONE:
820 		fc = "None";
821 		break;
822 	default:
823 		fc = "Unknown";
824 		break;
825 	}
826 
827 	/* Get FEC mode based on negotiated link info */
828 	switch (vsi->port_info->phy.link_info.fec_info) {
829 	case ICE_AQ_LINK_25G_RS_528_FEC_EN:
830 	case ICE_AQ_LINK_25G_RS_544_FEC_EN:
831 		fec = "RS-FEC";
832 		break;
833 	case ICE_AQ_LINK_25G_KR_FEC_EN:
834 		fec = "FC-FEC/BASE-R";
835 		break;
836 	default:
837 		fec = "NONE";
838 		break;
839 	}
840 
841 	/* check if autoneg completed, might be false due to not supported */
842 	if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
843 		an = "True";
844 	else
845 		an = "False";
846 
847 	/* Get FEC mode requested based on PHY caps last SW configuration */
848 	caps = kzalloc(sizeof(*caps), GFP_KERNEL);
849 	if (!caps) {
850 		fec_req = "Unknown";
851 		an_advertised = "Unknown";
852 		goto done;
853 	}
854 
855 	status = ice_aq_get_phy_caps(vsi->port_info, false,
856 				     ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
857 	if (status)
858 		netdev_info(vsi->netdev, "Get phy capability failed.\n");
859 
860 	an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
861 
862 	if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
863 	    caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
864 		fec_req = "RS-FEC";
865 	else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
866 		 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
867 		fec_req = "FC-FEC/BASE-R";
868 	else
869 		fec_req = "NONE";
870 
871 	kfree(caps);
872 
873 done:
874 	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",
875 		    speed, fec_req, fec, an_advertised, an, fc);
876 	ice_print_topo_conflict(vsi);
877 }
878 
879 /**
880  * ice_vsi_link_event - update the VSI's netdev
881  * @vsi: the VSI on which the link event occurred
882  * @link_up: whether or not the VSI needs to be set up or down
883  */
884 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
885 {
886 	if (!vsi)
887 		return;
888 
889 	if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
890 		return;
891 
892 	if (vsi->type == ICE_VSI_PF) {
893 		if (link_up == netif_carrier_ok(vsi->netdev))
894 			return;
895 
896 		if (link_up) {
897 			netif_carrier_on(vsi->netdev);
898 			netif_tx_wake_all_queues(vsi->netdev);
899 		} else {
900 			netif_carrier_off(vsi->netdev);
901 			netif_tx_stop_all_queues(vsi->netdev);
902 		}
903 	}
904 }
905 
906 /**
907  * ice_set_dflt_mib - send a default config MIB to the FW
908  * @pf: private PF struct
909  *
910  * This function sends a default configuration MIB to the FW.
911  *
912  * If this function errors out at any point, the driver is still able to
913  * function.  The main impact is that LFC may not operate as expected.
914  * Therefore an error state in this function should be treated with a DBG
915  * message and continue on with driver rebuild/reenable.
916  */
917 static void ice_set_dflt_mib(struct ice_pf *pf)
918 {
919 	struct device *dev = ice_pf_to_dev(pf);
920 	u8 mib_type, *buf, *lldpmib = NULL;
921 	u16 len, typelen, offset = 0;
922 	struct ice_lldp_org_tlv *tlv;
923 	struct ice_hw *hw = &pf->hw;
924 	u32 ouisubtype;
925 
926 	mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
927 	lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
928 	if (!lldpmib) {
929 		dev_dbg(dev, "%s Failed to allocate MIB memory\n",
930 			__func__);
931 		return;
932 	}
933 
934 	/* Add ETS CFG TLV */
935 	tlv = (struct ice_lldp_org_tlv *)lldpmib;
936 	typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
937 		   ICE_IEEE_ETS_TLV_LEN);
938 	tlv->typelen = htons(typelen);
939 	ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
940 		      ICE_IEEE_SUBTYPE_ETS_CFG);
941 	tlv->ouisubtype = htonl(ouisubtype);
942 
943 	buf = tlv->tlvinfo;
944 	buf[0] = 0;
945 
946 	/* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
947 	 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
948 	 * Octets 13 - 20 are TSA values - leave as zeros
949 	 */
950 	buf[5] = 0x64;
951 	len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
952 	offset += len + 2;
953 	tlv = (struct ice_lldp_org_tlv *)
954 		((char *)tlv + sizeof(tlv->typelen) + len);
955 
956 	/* Add ETS REC TLV */
957 	buf = tlv->tlvinfo;
958 	tlv->typelen = htons(typelen);
959 
960 	ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
961 		      ICE_IEEE_SUBTYPE_ETS_REC);
962 	tlv->ouisubtype = htonl(ouisubtype);
963 
964 	/* First octet of buf is reserved
965 	 * Octets 1 - 4 map UP to TC - all UPs map to zero
966 	 * Octets 5 - 12 are BW values - set TC 0 to 100%.
967 	 * Octets 13 - 20 are TSA value - leave as zeros
968 	 */
969 	buf[5] = 0x64;
970 	offset += len + 2;
971 	tlv = (struct ice_lldp_org_tlv *)
972 		((char *)tlv + sizeof(tlv->typelen) + len);
973 
974 	/* Add PFC CFG TLV */
975 	typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
976 		   ICE_IEEE_PFC_TLV_LEN);
977 	tlv->typelen = htons(typelen);
978 
979 	ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
980 		      ICE_IEEE_SUBTYPE_PFC_CFG);
981 	tlv->ouisubtype = htonl(ouisubtype);
982 
983 	/* Octet 1 left as all zeros - PFC disabled */
984 	buf[0] = 0x08;
985 	len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
986 	offset += len + 2;
987 
988 	if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
989 		dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
990 
991 	kfree(lldpmib);
992 }
993 
994 /**
995  * ice_check_phy_fw_load - check if PHY FW load failed
996  * @pf: pointer to PF struct
997  * @link_cfg_err: bitmap from the link info structure
998  *
999  * check if external PHY FW load failed and print an error message if it did
1000  */
1001 static void ice_check_phy_fw_load(struct ice_pf *pf, u8 link_cfg_err)
1002 {
1003 	if (!(link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE)) {
1004 		clear_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1005 		return;
1006 	}
1007 
1008 	if (test_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags))
1009 		return;
1010 
1011 	if (link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE) {
1012 		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");
1013 		set_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1014 	}
1015 }
1016 
1017 /**
1018  * ice_check_module_power
1019  * @pf: pointer to PF struct
1020  * @link_cfg_err: bitmap from the link info structure
1021  *
1022  * check module power level returned by a previous call to aq_get_link_info
1023  * and print error messages if module power level is not supported
1024  */
1025 static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
1026 {
1027 	/* if module power level is supported, clear the flag */
1028 	if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
1029 			      ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
1030 		clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1031 		return;
1032 	}
1033 
1034 	/* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
1035 	 * above block didn't clear this bit, there's nothing to do
1036 	 */
1037 	if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
1038 		return;
1039 
1040 	if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
1041 		dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
1042 		set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1043 	} else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
1044 		dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
1045 		set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1046 	}
1047 }
1048 
1049 /**
1050  * ice_check_link_cfg_err - check if link configuration failed
1051  * @pf: pointer to the PF struct
1052  * @link_cfg_err: bitmap from the link info structure
1053  *
1054  * print if any link configuration failure happens due to the value in the
1055  * link_cfg_err parameter in the link info structure
1056  */
1057 static void ice_check_link_cfg_err(struct ice_pf *pf, u8 link_cfg_err)
1058 {
1059 	ice_check_module_power(pf, link_cfg_err);
1060 	ice_check_phy_fw_load(pf, link_cfg_err);
1061 }
1062 
1063 /**
1064  * ice_link_event - process the link event
1065  * @pf: PF that the link event is associated with
1066  * @pi: port_info for the port that the link event is associated with
1067  * @link_up: true if the physical link is up and false if it is down
1068  * @link_speed: current link speed received from the link event
1069  *
1070  * Returns 0 on success and negative on failure
1071  */
1072 static int
1073 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
1074 	       u16 link_speed)
1075 {
1076 	struct device *dev = ice_pf_to_dev(pf);
1077 	struct ice_phy_info *phy_info;
1078 	struct ice_vsi *vsi;
1079 	u16 old_link_speed;
1080 	bool old_link;
1081 	int status;
1082 
1083 	phy_info = &pi->phy;
1084 	phy_info->link_info_old = phy_info->link_info;
1085 
1086 	old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
1087 	old_link_speed = phy_info->link_info_old.link_speed;
1088 
1089 	/* update the link info structures and re-enable link events,
1090 	 * don't bail on failure due to other book keeping needed
1091 	 */
1092 	status = ice_update_link_info(pi);
1093 	if (status)
1094 		dev_dbg(dev, "Failed to update link status on port %d, err %d aq_err %s\n",
1095 			pi->lport, status,
1096 			ice_aq_str(pi->hw->adminq.sq_last_status));
1097 
1098 	ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
1099 
1100 	/* Check if the link state is up after updating link info, and treat
1101 	 * this event as an UP event since the link is actually UP now.
1102 	 */
1103 	if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
1104 		link_up = true;
1105 
1106 	vsi = ice_get_main_vsi(pf);
1107 	if (!vsi || !vsi->port_info)
1108 		return -EINVAL;
1109 
1110 	/* turn off PHY if media was removed */
1111 	if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
1112 	    !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
1113 		set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
1114 		ice_set_link(vsi, false);
1115 	}
1116 
1117 	/* if the old link up/down and speed is the same as the new */
1118 	if (link_up == old_link && link_speed == old_link_speed)
1119 		return 0;
1120 
1121 	if (!ice_is_e810(&pf->hw))
1122 		ice_ptp_link_change(pf, pf->hw.pf_id, link_up);
1123 
1124 	if (ice_is_dcb_active(pf)) {
1125 		if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
1126 			ice_dcb_rebuild(pf);
1127 	} else {
1128 		if (link_up)
1129 			ice_set_dflt_mib(pf);
1130 	}
1131 	ice_vsi_link_event(vsi, link_up);
1132 	ice_print_link_msg(vsi, link_up);
1133 
1134 	ice_vc_notify_link_state(pf);
1135 
1136 	return 0;
1137 }
1138 
1139 /**
1140  * ice_watchdog_subtask - periodic tasks not using event driven scheduling
1141  * @pf: board private structure
1142  */
1143 static void ice_watchdog_subtask(struct ice_pf *pf)
1144 {
1145 	int i;
1146 
1147 	/* if interface is down do nothing */
1148 	if (test_bit(ICE_DOWN, pf->state) ||
1149 	    test_bit(ICE_CFG_BUSY, pf->state))
1150 		return;
1151 
1152 	/* make sure we don't do these things too often */
1153 	if (time_before(jiffies,
1154 			pf->serv_tmr_prev + pf->serv_tmr_period))
1155 		return;
1156 
1157 	pf->serv_tmr_prev = jiffies;
1158 
1159 	/* Update the stats for active netdevs so the network stack
1160 	 * can look at updated numbers whenever it cares to
1161 	 */
1162 	ice_update_pf_stats(pf);
1163 	ice_for_each_vsi(pf, i)
1164 		if (pf->vsi[i] && pf->vsi[i]->netdev)
1165 			ice_update_vsi_stats(pf->vsi[i]);
1166 }
1167 
1168 /**
1169  * ice_init_link_events - enable/initialize link events
1170  * @pi: pointer to the port_info instance
1171  *
1172  * Returns -EIO on failure, 0 on success
1173  */
1174 static int ice_init_link_events(struct ice_port_info *pi)
1175 {
1176 	u16 mask;
1177 
1178 	mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1179 		       ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL |
1180 		       ICE_AQ_LINK_EVENT_PHY_FW_LOAD_FAIL));
1181 
1182 	if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1183 		dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1184 			pi->lport);
1185 		return -EIO;
1186 	}
1187 
1188 	if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1189 		dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1190 			pi->lport);
1191 		return -EIO;
1192 	}
1193 
1194 	return 0;
1195 }
1196 
1197 /**
1198  * ice_handle_link_event - handle link event via ARQ
1199  * @pf: PF that the link event is associated with
1200  * @event: event structure containing link status info
1201  */
1202 static int
1203 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1204 {
1205 	struct ice_aqc_get_link_status_data *link_data;
1206 	struct ice_port_info *port_info;
1207 	int status;
1208 
1209 	link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1210 	port_info = pf->hw.port_info;
1211 	if (!port_info)
1212 		return -EINVAL;
1213 
1214 	status = ice_link_event(pf, port_info,
1215 				!!(link_data->link_info & ICE_AQ_LINK_UP),
1216 				le16_to_cpu(link_data->link_speed));
1217 	if (status)
1218 		dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1219 			status);
1220 
1221 	return status;
1222 }
1223 
1224 enum ice_aq_task_state {
1225 	ICE_AQ_TASK_WAITING = 0,
1226 	ICE_AQ_TASK_COMPLETE,
1227 	ICE_AQ_TASK_CANCELED,
1228 };
1229 
1230 struct ice_aq_task {
1231 	struct hlist_node entry;
1232 
1233 	u16 opcode;
1234 	struct ice_rq_event_info *event;
1235 	enum ice_aq_task_state state;
1236 };
1237 
1238 /**
1239  * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1240  * @pf: pointer to the PF private structure
1241  * @opcode: the opcode to wait for
1242  * @timeout: how long to wait, in jiffies
1243  * @event: storage for the event info
1244  *
1245  * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1246  * current thread will be put to sleep until the specified event occurs or
1247  * until the given timeout is reached.
1248  *
1249  * To obtain only the descriptor contents, pass an event without an allocated
1250  * msg_buf. If the complete data buffer is desired, allocate the
1251  * event->msg_buf with enough space ahead of time.
1252  *
1253  * Returns: zero on success, or a negative error code on failure.
1254  */
1255 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1256 			  struct ice_rq_event_info *event)
1257 {
1258 	struct device *dev = ice_pf_to_dev(pf);
1259 	struct ice_aq_task *task;
1260 	unsigned long start;
1261 	long ret;
1262 	int err;
1263 
1264 	task = kzalloc(sizeof(*task), GFP_KERNEL);
1265 	if (!task)
1266 		return -ENOMEM;
1267 
1268 	INIT_HLIST_NODE(&task->entry);
1269 	task->opcode = opcode;
1270 	task->event = event;
1271 	task->state = ICE_AQ_TASK_WAITING;
1272 
1273 	spin_lock_bh(&pf->aq_wait_lock);
1274 	hlist_add_head(&task->entry, &pf->aq_wait_list);
1275 	spin_unlock_bh(&pf->aq_wait_lock);
1276 
1277 	start = jiffies;
1278 
1279 	ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1280 					       timeout);
1281 	switch (task->state) {
1282 	case ICE_AQ_TASK_WAITING:
1283 		err = ret < 0 ? ret : -ETIMEDOUT;
1284 		break;
1285 	case ICE_AQ_TASK_CANCELED:
1286 		err = ret < 0 ? ret : -ECANCELED;
1287 		break;
1288 	case ICE_AQ_TASK_COMPLETE:
1289 		err = ret < 0 ? ret : 0;
1290 		break;
1291 	default:
1292 		WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1293 		err = -EINVAL;
1294 		break;
1295 	}
1296 
1297 	dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1298 		jiffies_to_msecs(jiffies - start),
1299 		jiffies_to_msecs(timeout),
1300 		opcode);
1301 
1302 	spin_lock_bh(&pf->aq_wait_lock);
1303 	hlist_del(&task->entry);
1304 	spin_unlock_bh(&pf->aq_wait_lock);
1305 	kfree(task);
1306 
1307 	return err;
1308 }
1309 
1310 /**
1311  * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1312  * @pf: pointer to the PF private structure
1313  * @opcode: the opcode of the event
1314  * @event: the event to check
1315  *
1316  * Loops over the current list of pending threads waiting for an AdminQ event.
1317  * For each matching task, copy the contents of the event into the task
1318  * structure and wake up the thread.
1319  *
1320  * If multiple threads wait for the same opcode, they will all be woken up.
1321  *
1322  * Note that event->msg_buf will only be duplicated if the event has a buffer
1323  * with enough space already allocated. Otherwise, only the descriptor and
1324  * message length will be copied.
1325  *
1326  * Returns: true if an event was found, false otherwise
1327  */
1328 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1329 				struct ice_rq_event_info *event)
1330 {
1331 	struct ice_aq_task *task;
1332 	bool found = false;
1333 
1334 	spin_lock_bh(&pf->aq_wait_lock);
1335 	hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1336 		if (task->state || task->opcode != opcode)
1337 			continue;
1338 
1339 		memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1340 		task->event->msg_len = event->msg_len;
1341 
1342 		/* Only copy the data buffer if a destination was set */
1343 		if (task->event->msg_buf &&
1344 		    task->event->buf_len > event->buf_len) {
1345 			memcpy(task->event->msg_buf, event->msg_buf,
1346 			       event->buf_len);
1347 			task->event->buf_len = event->buf_len;
1348 		}
1349 
1350 		task->state = ICE_AQ_TASK_COMPLETE;
1351 		found = true;
1352 	}
1353 	spin_unlock_bh(&pf->aq_wait_lock);
1354 
1355 	if (found)
1356 		wake_up(&pf->aq_wait_queue);
1357 }
1358 
1359 /**
1360  * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1361  * @pf: the PF private structure
1362  *
1363  * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1364  * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1365  */
1366 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1367 {
1368 	struct ice_aq_task *task;
1369 
1370 	spin_lock_bh(&pf->aq_wait_lock);
1371 	hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1372 		task->state = ICE_AQ_TASK_CANCELED;
1373 	spin_unlock_bh(&pf->aq_wait_lock);
1374 
1375 	wake_up(&pf->aq_wait_queue);
1376 }
1377 
1378 /**
1379  * __ice_clean_ctrlq - helper function to clean controlq rings
1380  * @pf: ptr to struct ice_pf
1381  * @q_type: specific Control queue type
1382  */
1383 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1384 {
1385 	struct device *dev = ice_pf_to_dev(pf);
1386 	struct ice_rq_event_info event;
1387 	struct ice_hw *hw = &pf->hw;
1388 	struct ice_ctl_q_info *cq;
1389 	u16 pending, i = 0;
1390 	const char *qtype;
1391 	u32 oldval, val;
1392 
1393 	/* Do not clean control queue if/when PF reset fails */
1394 	if (test_bit(ICE_RESET_FAILED, pf->state))
1395 		return 0;
1396 
1397 	switch (q_type) {
1398 	case ICE_CTL_Q_ADMIN:
1399 		cq = &hw->adminq;
1400 		qtype = "Admin";
1401 		break;
1402 	case ICE_CTL_Q_SB:
1403 		cq = &hw->sbq;
1404 		qtype = "Sideband";
1405 		break;
1406 	case ICE_CTL_Q_MAILBOX:
1407 		cq = &hw->mailboxq;
1408 		qtype = "Mailbox";
1409 		/* we are going to try to detect a malicious VF, so set the
1410 		 * state to begin detection
1411 		 */
1412 		hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1413 		break;
1414 	default:
1415 		dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1416 		return 0;
1417 	}
1418 
1419 	/* check for error indications - PF_xx_AxQLEN register layout for
1420 	 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1421 	 */
1422 	val = rd32(hw, cq->rq.len);
1423 	if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1424 		   PF_FW_ARQLEN_ARQCRIT_M)) {
1425 		oldval = val;
1426 		if (val & PF_FW_ARQLEN_ARQVFE_M)
1427 			dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1428 				qtype);
1429 		if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1430 			dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1431 				qtype);
1432 		}
1433 		if (val & PF_FW_ARQLEN_ARQCRIT_M)
1434 			dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1435 				qtype);
1436 		val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1437 			 PF_FW_ARQLEN_ARQCRIT_M);
1438 		if (oldval != val)
1439 			wr32(hw, cq->rq.len, val);
1440 	}
1441 
1442 	val = rd32(hw, cq->sq.len);
1443 	if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1444 		   PF_FW_ATQLEN_ATQCRIT_M)) {
1445 		oldval = val;
1446 		if (val & PF_FW_ATQLEN_ATQVFE_M)
1447 			dev_dbg(dev, "%s Send Queue VF Error detected\n",
1448 				qtype);
1449 		if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1450 			dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1451 				qtype);
1452 		}
1453 		if (val & PF_FW_ATQLEN_ATQCRIT_M)
1454 			dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1455 				qtype);
1456 		val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1457 			 PF_FW_ATQLEN_ATQCRIT_M);
1458 		if (oldval != val)
1459 			wr32(hw, cq->sq.len, val);
1460 	}
1461 
1462 	event.buf_len = cq->rq_buf_size;
1463 	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1464 	if (!event.msg_buf)
1465 		return 0;
1466 
1467 	do {
1468 		u16 opcode;
1469 		int ret;
1470 
1471 		ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1472 		if (ret == -EALREADY)
1473 			break;
1474 		if (ret) {
1475 			dev_err(dev, "%s Receive Queue event error %d\n", qtype,
1476 				ret);
1477 			break;
1478 		}
1479 
1480 		opcode = le16_to_cpu(event.desc.opcode);
1481 
1482 		/* Notify any thread that might be waiting for this event */
1483 		ice_aq_check_events(pf, opcode, &event);
1484 
1485 		switch (opcode) {
1486 		case ice_aqc_opc_get_link_status:
1487 			if (ice_handle_link_event(pf, &event))
1488 				dev_err(dev, "Could not handle link event\n");
1489 			break;
1490 		case ice_aqc_opc_event_lan_overflow:
1491 			ice_vf_lan_overflow_event(pf, &event);
1492 			break;
1493 		case ice_mbx_opc_send_msg_to_pf:
1494 			if (!ice_is_malicious_vf(pf, &event, i, pending))
1495 				ice_vc_process_vf_msg(pf, &event);
1496 			break;
1497 		case ice_aqc_opc_fw_logging:
1498 			ice_output_fw_log(hw, &event.desc, event.msg_buf);
1499 			break;
1500 		case ice_aqc_opc_lldp_set_mib_change:
1501 			ice_dcb_process_lldp_set_mib_change(pf, &event);
1502 			break;
1503 		default:
1504 			dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1505 				qtype, opcode);
1506 			break;
1507 		}
1508 	} while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1509 
1510 	kfree(event.msg_buf);
1511 
1512 	return pending && (i == ICE_DFLT_IRQ_WORK);
1513 }
1514 
1515 /**
1516  * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1517  * @hw: pointer to hardware info
1518  * @cq: control queue information
1519  *
1520  * returns true if there are pending messages in a queue, false if there aren't
1521  */
1522 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1523 {
1524 	u16 ntu;
1525 
1526 	ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1527 	return cq->rq.next_to_clean != ntu;
1528 }
1529 
1530 /**
1531  * ice_clean_adminq_subtask - clean the AdminQ rings
1532  * @pf: board private structure
1533  */
1534 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1535 {
1536 	struct ice_hw *hw = &pf->hw;
1537 
1538 	if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1539 		return;
1540 
1541 	if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1542 		return;
1543 
1544 	clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1545 
1546 	/* There might be a situation where new messages arrive to a control
1547 	 * queue between processing the last message and clearing the
1548 	 * EVENT_PENDING bit. So before exiting, check queue head again (using
1549 	 * ice_ctrlq_pending) and process new messages if any.
1550 	 */
1551 	if (ice_ctrlq_pending(hw, &hw->adminq))
1552 		__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1553 
1554 	ice_flush(hw);
1555 }
1556 
1557 /**
1558  * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1559  * @pf: board private structure
1560  */
1561 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1562 {
1563 	struct ice_hw *hw = &pf->hw;
1564 
1565 	if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1566 		return;
1567 
1568 	if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1569 		return;
1570 
1571 	clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1572 
1573 	if (ice_ctrlq_pending(hw, &hw->mailboxq))
1574 		__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1575 
1576 	ice_flush(hw);
1577 }
1578 
1579 /**
1580  * ice_clean_sbq_subtask - clean the Sideband Queue rings
1581  * @pf: board private structure
1582  */
1583 static void ice_clean_sbq_subtask(struct ice_pf *pf)
1584 {
1585 	struct ice_hw *hw = &pf->hw;
1586 
1587 	/* Nothing to do here if sideband queue is not supported */
1588 	if (!ice_is_sbq_supported(hw)) {
1589 		clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1590 		return;
1591 	}
1592 
1593 	if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1594 		return;
1595 
1596 	if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1597 		return;
1598 
1599 	clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1600 
1601 	if (ice_ctrlq_pending(hw, &hw->sbq))
1602 		__ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1603 
1604 	ice_flush(hw);
1605 }
1606 
1607 /**
1608  * ice_service_task_schedule - schedule the service task to wake up
1609  * @pf: board private structure
1610  *
1611  * If not already scheduled, this puts the task into the work queue.
1612  */
1613 void ice_service_task_schedule(struct ice_pf *pf)
1614 {
1615 	if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1616 	    !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1617 	    !test_bit(ICE_NEEDS_RESTART, pf->state))
1618 		queue_work(ice_wq, &pf->serv_task);
1619 }
1620 
1621 /**
1622  * ice_service_task_complete - finish up the service task
1623  * @pf: board private structure
1624  */
1625 static void ice_service_task_complete(struct ice_pf *pf)
1626 {
1627 	WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1628 
1629 	/* force memory (pf->state) to sync before next service task */
1630 	smp_mb__before_atomic();
1631 	clear_bit(ICE_SERVICE_SCHED, pf->state);
1632 }
1633 
1634 /**
1635  * ice_service_task_stop - stop service task and cancel works
1636  * @pf: board private structure
1637  *
1638  * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1639  * 1 otherwise.
1640  */
1641 static int ice_service_task_stop(struct ice_pf *pf)
1642 {
1643 	int ret;
1644 
1645 	ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1646 
1647 	if (pf->serv_tmr.function)
1648 		del_timer_sync(&pf->serv_tmr);
1649 	if (pf->serv_task.func)
1650 		cancel_work_sync(&pf->serv_task);
1651 
1652 	clear_bit(ICE_SERVICE_SCHED, pf->state);
1653 	return ret;
1654 }
1655 
1656 /**
1657  * ice_service_task_restart - restart service task and schedule works
1658  * @pf: board private structure
1659  *
1660  * This function is needed for suspend and resume works (e.g WoL scenario)
1661  */
1662 static void ice_service_task_restart(struct ice_pf *pf)
1663 {
1664 	clear_bit(ICE_SERVICE_DIS, pf->state);
1665 	ice_service_task_schedule(pf);
1666 }
1667 
1668 /**
1669  * ice_service_timer - timer callback to schedule service task
1670  * @t: pointer to timer_list
1671  */
1672 static void ice_service_timer(struct timer_list *t)
1673 {
1674 	struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1675 
1676 	mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1677 	ice_service_task_schedule(pf);
1678 }
1679 
1680 /**
1681  * ice_handle_mdd_event - handle malicious driver detect event
1682  * @pf: pointer to the PF structure
1683  *
1684  * Called from service task. OICR interrupt handler indicates MDD event.
1685  * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1686  * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1687  * disable the queue, the PF can be configured to reset the VF using ethtool
1688  * private flag mdd-auto-reset-vf.
1689  */
1690 static void ice_handle_mdd_event(struct ice_pf *pf)
1691 {
1692 	struct device *dev = ice_pf_to_dev(pf);
1693 	struct ice_hw *hw = &pf->hw;
1694 	struct ice_vf *vf;
1695 	unsigned int bkt;
1696 	u32 reg;
1697 
1698 	if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1699 		/* Since the VF MDD event logging is rate limited, check if
1700 		 * there are pending MDD events.
1701 		 */
1702 		ice_print_vfs_mdd_events(pf);
1703 		return;
1704 	}
1705 
1706 	/* find what triggered an MDD event */
1707 	reg = rd32(hw, GL_MDET_TX_PQM);
1708 	if (reg & GL_MDET_TX_PQM_VALID_M) {
1709 		u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1710 				GL_MDET_TX_PQM_PF_NUM_S;
1711 		u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1712 				GL_MDET_TX_PQM_VF_NUM_S;
1713 		u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1714 				GL_MDET_TX_PQM_MAL_TYPE_S;
1715 		u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1716 				GL_MDET_TX_PQM_QNUM_S);
1717 
1718 		if (netif_msg_tx_err(pf))
1719 			dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1720 				 event, queue, pf_num, vf_num);
1721 		wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1722 	}
1723 
1724 	reg = rd32(hw, GL_MDET_TX_TCLAN);
1725 	if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1726 		u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1727 				GL_MDET_TX_TCLAN_PF_NUM_S;
1728 		u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1729 				GL_MDET_TX_TCLAN_VF_NUM_S;
1730 		u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1731 				GL_MDET_TX_TCLAN_MAL_TYPE_S;
1732 		u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1733 				GL_MDET_TX_TCLAN_QNUM_S);
1734 
1735 		if (netif_msg_tx_err(pf))
1736 			dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1737 				 event, queue, pf_num, vf_num);
1738 		wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1739 	}
1740 
1741 	reg = rd32(hw, GL_MDET_RX);
1742 	if (reg & GL_MDET_RX_VALID_M) {
1743 		u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1744 				GL_MDET_RX_PF_NUM_S;
1745 		u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1746 				GL_MDET_RX_VF_NUM_S;
1747 		u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1748 				GL_MDET_RX_MAL_TYPE_S;
1749 		u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1750 				GL_MDET_RX_QNUM_S);
1751 
1752 		if (netif_msg_rx_err(pf))
1753 			dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1754 				 event, queue, pf_num, vf_num);
1755 		wr32(hw, GL_MDET_RX, 0xffffffff);
1756 	}
1757 
1758 	/* check to see if this PF caused an MDD event */
1759 	reg = rd32(hw, PF_MDET_TX_PQM);
1760 	if (reg & PF_MDET_TX_PQM_VALID_M) {
1761 		wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1762 		if (netif_msg_tx_err(pf))
1763 			dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1764 	}
1765 
1766 	reg = rd32(hw, PF_MDET_TX_TCLAN);
1767 	if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1768 		wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1769 		if (netif_msg_tx_err(pf))
1770 			dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1771 	}
1772 
1773 	reg = rd32(hw, PF_MDET_RX);
1774 	if (reg & PF_MDET_RX_VALID_M) {
1775 		wr32(hw, PF_MDET_RX, 0xFFFF);
1776 		if (netif_msg_rx_err(pf))
1777 			dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1778 	}
1779 
1780 	/* Check to see if one of the VFs caused an MDD event, and then
1781 	 * increment counters and set print pending
1782 	 */
1783 	mutex_lock(&pf->vfs.table_lock);
1784 	ice_for_each_vf(pf, bkt, vf) {
1785 		reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id));
1786 		if (reg & VP_MDET_TX_PQM_VALID_M) {
1787 			wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF);
1788 			vf->mdd_tx_events.count++;
1789 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1790 			if (netif_msg_tx_err(pf))
1791 				dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1792 					 vf->vf_id);
1793 		}
1794 
1795 		reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id));
1796 		if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1797 			wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF);
1798 			vf->mdd_tx_events.count++;
1799 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1800 			if (netif_msg_tx_err(pf))
1801 				dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1802 					 vf->vf_id);
1803 		}
1804 
1805 		reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id));
1806 		if (reg & VP_MDET_TX_TDPU_VALID_M) {
1807 			wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF);
1808 			vf->mdd_tx_events.count++;
1809 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1810 			if (netif_msg_tx_err(pf))
1811 				dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1812 					 vf->vf_id);
1813 		}
1814 
1815 		reg = rd32(hw, VP_MDET_RX(vf->vf_id));
1816 		if (reg & VP_MDET_RX_VALID_M) {
1817 			wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF);
1818 			vf->mdd_rx_events.count++;
1819 			set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1820 			if (netif_msg_rx_err(pf))
1821 				dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1822 					 vf->vf_id);
1823 
1824 			/* Since the queue is disabled on VF Rx MDD events, the
1825 			 * PF can be configured to reset the VF through ethtool
1826 			 * private flag mdd-auto-reset-vf.
1827 			 */
1828 			if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1829 				/* VF MDD event counters will be cleared by
1830 				 * reset, so print the event prior to reset.
1831 				 */
1832 				ice_print_vf_rx_mdd_event(vf);
1833 				ice_reset_vf(vf, ICE_VF_RESET_LOCK);
1834 			}
1835 		}
1836 	}
1837 	mutex_unlock(&pf->vfs.table_lock);
1838 
1839 	ice_print_vfs_mdd_events(pf);
1840 }
1841 
1842 /**
1843  * ice_force_phys_link_state - Force the physical link state
1844  * @vsi: VSI to force the physical link state to up/down
1845  * @link_up: true/false indicates to set the physical link to up/down
1846  *
1847  * Force the physical link state by getting the current PHY capabilities from
1848  * hardware and setting the PHY config based on the determined capabilities. If
1849  * link changes a link event will be triggered because both the Enable Automatic
1850  * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1851  *
1852  * Returns 0 on success, negative on failure
1853  */
1854 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1855 {
1856 	struct ice_aqc_get_phy_caps_data *pcaps;
1857 	struct ice_aqc_set_phy_cfg_data *cfg;
1858 	struct ice_port_info *pi;
1859 	struct device *dev;
1860 	int retcode;
1861 
1862 	if (!vsi || !vsi->port_info || !vsi->back)
1863 		return -EINVAL;
1864 	if (vsi->type != ICE_VSI_PF)
1865 		return 0;
1866 
1867 	dev = ice_pf_to_dev(vsi->back);
1868 
1869 	pi = vsi->port_info;
1870 
1871 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1872 	if (!pcaps)
1873 		return -ENOMEM;
1874 
1875 	retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1876 				      NULL);
1877 	if (retcode) {
1878 		dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1879 			vsi->vsi_num, retcode);
1880 		retcode = -EIO;
1881 		goto out;
1882 	}
1883 
1884 	/* No change in link */
1885 	if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1886 	    link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1887 		goto out;
1888 
1889 	/* Use the current user PHY configuration. The current user PHY
1890 	 * configuration is initialized during probe from PHY capabilities
1891 	 * software mode, and updated on set PHY configuration.
1892 	 */
1893 	cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1894 	if (!cfg) {
1895 		retcode = -ENOMEM;
1896 		goto out;
1897 	}
1898 
1899 	cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1900 	if (link_up)
1901 		cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1902 	else
1903 		cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1904 
1905 	retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1906 	if (retcode) {
1907 		dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1908 			vsi->vsi_num, retcode);
1909 		retcode = -EIO;
1910 	}
1911 
1912 	kfree(cfg);
1913 out:
1914 	kfree(pcaps);
1915 	return retcode;
1916 }
1917 
1918 /**
1919  * ice_init_nvm_phy_type - Initialize the NVM PHY type
1920  * @pi: port info structure
1921  *
1922  * Initialize nvm_phy_type_[low|high] for link lenient mode support
1923  */
1924 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1925 {
1926 	struct ice_aqc_get_phy_caps_data *pcaps;
1927 	struct ice_pf *pf = pi->hw->back;
1928 	int err;
1929 
1930 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1931 	if (!pcaps)
1932 		return -ENOMEM;
1933 
1934 	err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA,
1935 				  pcaps, NULL);
1936 
1937 	if (err) {
1938 		dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1939 		goto out;
1940 	}
1941 
1942 	pf->nvm_phy_type_hi = pcaps->phy_type_high;
1943 	pf->nvm_phy_type_lo = pcaps->phy_type_low;
1944 
1945 out:
1946 	kfree(pcaps);
1947 	return err;
1948 }
1949 
1950 /**
1951  * ice_init_link_dflt_override - Initialize link default override
1952  * @pi: port info structure
1953  *
1954  * Initialize link default override and PHY total port shutdown during probe
1955  */
1956 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1957 {
1958 	struct ice_link_default_override_tlv *ldo;
1959 	struct ice_pf *pf = pi->hw->back;
1960 
1961 	ldo = &pf->link_dflt_override;
1962 	if (ice_get_link_default_override(ldo, pi))
1963 		return;
1964 
1965 	if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1966 		return;
1967 
1968 	/* Enable Total Port Shutdown (override/replace link-down-on-close
1969 	 * ethtool private flag) for ports with Port Disable bit set.
1970 	 */
1971 	set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1972 	set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1973 }
1974 
1975 /**
1976  * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1977  * @pi: port info structure
1978  *
1979  * If default override is enabled, initialize the user PHY cfg speed and FEC
1980  * settings using the default override mask from the NVM.
1981  *
1982  * The PHY should only be configured with the default override settings the
1983  * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1984  * is used to indicate that the user PHY cfg default override is initialized
1985  * and the PHY has not been configured with the default override settings. The
1986  * state is set here, and cleared in ice_configure_phy the first time the PHY is
1987  * configured.
1988  *
1989  * This function should be called only if the FW doesn't support default
1990  * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1991  */
1992 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1993 {
1994 	struct ice_link_default_override_tlv *ldo;
1995 	struct ice_aqc_set_phy_cfg_data *cfg;
1996 	struct ice_phy_info *phy = &pi->phy;
1997 	struct ice_pf *pf = pi->hw->back;
1998 
1999 	ldo = &pf->link_dflt_override;
2000 
2001 	/* If link default override is enabled, use to mask NVM PHY capabilities
2002 	 * for speed and FEC default configuration.
2003 	 */
2004 	cfg = &phy->curr_user_phy_cfg;
2005 
2006 	if (ldo->phy_type_low || ldo->phy_type_high) {
2007 		cfg->phy_type_low = pf->nvm_phy_type_lo &
2008 				    cpu_to_le64(ldo->phy_type_low);
2009 		cfg->phy_type_high = pf->nvm_phy_type_hi &
2010 				     cpu_to_le64(ldo->phy_type_high);
2011 	}
2012 	cfg->link_fec_opt = ldo->fec_options;
2013 	phy->curr_user_fec_req = ICE_FEC_AUTO;
2014 
2015 	set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
2016 }
2017 
2018 /**
2019  * ice_init_phy_user_cfg - Initialize the PHY user configuration
2020  * @pi: port info structure
2021  *
2022  * Initialize the current user PHY configuration, speed, FEC, and FC requested
2023  * mode to default. The PHY defaults are from get PHY capabilities topology
2024  * with media so call when media is first available. An error is returned if
2025  * called when media is not available. The PHY initialization completed state is
2026  * set here.
2027  *
2028  * These configurations are used when setting PHY
2029  * configuration. The user PHY configuration is updated on set PHY
2030  * configuration. Returns 0 on success, negative on failure
2031  */
2032 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
2033 {
2034 	struct ice_aqc_get_phy_caps_data *pcaps;
2035 	struct ice_phy_info *phy = &pi->phy;
2036 	struct ice_pf *pf = pi->hw->back;
2037 	int err;
2038 
2039 	if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2040 		return -EIO;
2041 
2042 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2043 	if (!pcaps)
2044 		return -ENOMEM;
2045 
2046 	if (ice_fw_supports_report_dflt_cfg(pi->hw))
2047 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2048 					  pcaps, NULL);
2049 	else
2050 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2051 					  pcaps, NULL);
2052 	if (err) {
2053 		dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
2054 		goto err_out;
2055 	}
2056 
2057 	ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
2058 
2059 	/* check if lenient mode is supported and enabled */
2060 	if (ice_fw_supports_link_override(pi->hw) &&
2061 	    !(pcaps->module_compliance_enforcement &
2062 	      ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
2063 		set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
2064 
2065 		/* if the FW supports default PHY configuration mode, then the driver
2066 		 * does not have to apply link override settings. If not,
2067 		 * initialize user PHY configuration with link override values
2068 		 */
2069 		if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
2070 		    (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
2071 			ice_init_phy_cfg_dflt_override(pi);
2072 			goto out;
2073 		}
2074 	}
2075 
2076 	/* if link default override is not enabled, set user flow control and
2077 	 * FEC settings based on what get_phy_caps returned
2078 	 */
2079 	phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
2080 						      pcaps->link_fec_options);
2081 	phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
2082 
2083 out:
2084 	phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
2085 	set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
2086 err_out:
2087 	kfree(pcaps);
2088 	return err;
2089 }
2090 
2091 /**
2092  * ice_configure_phy - configure PHY
2093  * @vsi: VSI of PHY
2094  *
2095  * Set the PHY configuration. If the current PHY configuration is the same as
2096  * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
2097  * configure the based get PHY capabilities for topology with media.
2098  */
2099 static int ice_configure_phy(struct ice_vsi *vsi)
2100 {
2101 	struct device *dev = ice_pf_to_dev(vsi->back);
2102 	struct ice_port_info *pi = vsi->port_info;
2103 	struct ice_aqc_get_phy_caps_data *pcaps;
2104 	struct ice_aqc_set_phy_cfg_data *cfg;
2105 	struct ice_phy_info *phy = &pi->phy;
2106 	struct ice_pf *pf = vsi->back;
2107 	int err;
2108 
2109 	/* Ensure we have media as we cannot configure a medialess port */
2110 	if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2111 		return -EPERM;
2112 
2113 	ice_print_topo_conflict(vsi);
2114 
2115 	if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) &&
2116 	    phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
2117 		return -EPERM;
2118 
2119 	if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
2120 		return ice_force_phys_link_state(vsi, true);
2121 
2122 	pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2123 	if (!pcaps)
2124 		return -ENOMEM;
2125 
2126 	/* Get current PHY config */
2127 	err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
2128 				  NULL);
2129 	if (err) {
2130 		dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n",
2131 			vsi->vsi_num, err);
2132 		goto done;
2133 	}
2134 
2135 	/* If PHY enable link is configured and configuration has not changed,
2136 	 * there's nothing to do
2137 	 */
2138 	if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
2139 	    ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
2140 		goto done;
2141 
2142 	/* Use PHY topology as baseline for configuration */
2143 	memset(pcaps, 0, sizeof(*pcaps));
2144 	if (ice_fw_supports_report_dflt_cfg(pi->hw))
2145 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2146 					  pcaps, NULL);
2147 	else
2148 		err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2149 					  pcaps, NULL);
2150 	if (err) {
2151 		dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n",
2152 			vsi->vsi_num, err);
2153 		goto done;
2154 	}
2155 
2156 	cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2157 	if (!cfg) {
2158 		err = -ENOMEM;
2159 		goto done;
2160 	}
2161 
2162 	ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2163 
2164 	/* Speed - If default override pending, use curr_user_phy_cfg set in
2165 	 * ice_init_phy_user_cfg_ldo.
2166 	 */
2167 	if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2168 			       vsi->back->state)) {
2169 		cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2170 		cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2171 	} else {
2172 		u64 phy_low = 0, phy_high = 0;
2173 
2174 		ice_update_phy_type(&phy_low, &phy_high,
2175 				    pi->phy.curr_user_speed_req);
2176 		cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2177 		cfg->phy_type_high = pcaps->phy_type_high &
2178 				     cpu_to_le64(phy_high);
2179 	}
2180 
2181 	/* Can't provide what was requested; use PHY capabilities */
2182 	if (!cfg->phy_type_low && !cfg->phy_type_high) {
2183 		cfg->phy_type_low = pcaps->phy_type_low;
2184 		cfg->phy_type_high = pcaps->phy_type_high;
2185 	}
2186 
2187 	/* FEC */
2188 	ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2189 
2190 	/* Can't provide what was requested; use PHY capabilities */
2191 	if (cfg->link_fec_opt !=
2192 	    (cfg->link_fec_opt & pcaps->link_fec_options)) {
2193 		cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2194 		cfg->link_fec_opt = pcaps->link_fec_options;
2195 	}
2196 
2197 	/* Flow Control - always supported; no need to check against
2198 	 * capabilities
2199 	 */
2200 	ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2201 
2202 	/* Enable link and link update */
2203 	cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2204 
2205 	err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2206 	if (err)
2207 		dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
2208 			vsi->vsi_num, err);
2209 
2210 	kfree(cfg);
2211 done:
2212 	kfree(pcaps);
2213 	return err;
2214 }
2215 
2216 /**
2217  * ice_check_media_subtask - Check for media
2218  * @pf: pointer to PF struct
2219  *
2220  * If media is available, then initialize PHY user configuration if it is not
2221  * been, and configure the PHY if the interface is up.
2222  */
2223 static void ice_check_media_subtask(struct ice_pf *pf)
2224 {
2225 	struct ice_port_info *pi;
2226 	struct ice_vsi *vsi;
2227 	int err;
2228 
2229 	/* No need to check for media if it's already present */
2230 	if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2231 		return;
2232 
2233 	vsi = ice_get_main_vsi(pf);
2234 	if (!vsi)
2235 		return;
2236 
2237 	/* Refresh link info and check if media is present */
2238 	pi = vsi->port_info;
2239 	err = ice_update_link_info(pi);
2240 	if (err)
2241 		return;
2242 
2243 	ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
2244 
2245 	if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2246 		if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2247 			ice_init_phy_user_cfg(pi);
2248 
2249 		/* PHY settings are reset on media insertion, reconfigure
2250 		 * PHY to preserve settings.
2251 		 */
2252 		if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2253 		    test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2254 			return;
2255 
2256 		err = ice_configure_phy(vsi);
2257 		if (!err)
2258 			clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2259 
2260 		/* A Link Status Event will be generated; the event handler
2261 		 * will complete bringing the interface up
2262 		 */
2263 	}
2264 }
2265 
2266 /**
2267  * ice_service_task - manage and run subtasks
2268  * @work: pointer to work_struct contained by the PF struct
2269  */
2270 static void ice_service_task(struct work_struct *work)
2271 {
2272 	struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2273 	unsigned long start_time = jiffies;
2274 
2275 	/* subtasks */
2276 
2277 	/* process reset requests first */
2278 	ice_reset_subtask(pf);
2279 
2280 	/* bail if a reset/recovery cycle is pending or rebuild failed */
2281 	if (ice_is_reset_in_progress(pf->state) ||
2282 	    test_bit(ICE_SUSPENDED, pf->state) ||
2283 	    test_bit(ICE_NEEDS_RESTART, pf->state)) {
2284 		ice_service_task_complete(pf);
2285 		return;
2286 	}
2287 
2288 	if (test_and_clear_bit(ICE_AUX_ERR_PENDING, pf->state)) {
2289 		struct iidc_event *event;
2290 
2291 		event = kzalloc(sizeof(*event), GFP_KERNEL);
2292 		if (event) {
2293 			set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2294 			/* report the entire OICR value to AUX driver */
2295 			swap(event->reg, pf->oicr_err_reg);
2296 			ice_send_event_to_aux(pf, event);
2297 			kfree(event);
2298 		}
2299 	}
2300 
2301 	if (test_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) {
2302 		/* Plug aux device per request */
2303 		ice_plug_aux_dev(pf);
2304 
2305 		/* Mark plugging as done but check whether unplug was
2306 		 * requested during ice_plug_aux_dev() call
2307 		 * (e.g. from ice_clear_rdma_cap()) and if so then
2308 		 * plug aux device.
2309 		 */
2310 		if (!test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
2311 			ice_unplug_aux_dev(pf);
2312 	}
2313 
2314 	if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
2315 		struct iidc_event *event;
2316 
2317 		event = kzalloc(sizeof(*event), GFP_KERNEL);
2318 		if (event) {
2319 			set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
2320 			ice_send_event_to_aux(pf, event);
2321 			kfree(event);
2322 		}
2323 	}
2324 
2325 	ice_clean_adminq_subtask(pf);
2326 	ice_check_media_subtask(pf);
2327 	ice_check_for_hang_subtask(pf);
2328 	ice_sync_fltr_subtask(pf);
2329 	ice_handle_mdd_event(pf);
2330 	ice_watchdog_subtask(pf);
2331 
2332 	if (ice_is_safe_mode(pf)) {
2333 		ice_service_task_complete(pf);
2334 		return;
2335 	}
2336 
2337 	ice_process_vflr_event(pf);
2338 	ice_clean_mailboxq_subtask(pf);
2339 	ice_clean_sbq_subtask(pf);
2340 	ice_sync_arfs_fltrs(pf);
2341 	ice_flush_fdir_ctx(pf);
2342 
2343 	/* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2344 	ice_service_task_complete(pf);
2345 
2346 	/* If the tasks have taken longer than one service timer period
2347 	 * or there is more work to be done, reset the service timer to
2348 	 * schedule the service task now.
2349 	 */
2350 	if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2351 	    test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2352 	    test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2353 	    test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2354 	    test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2355 	    test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2356 	    test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2357 		mod_timer(&pf->serv_tmr, jiffies);
2358 }
2359 
2360 /**
2361  * ice_set_ctrlq_len - helper function to set controlq length
2362  * @hw: pointer to the HW instance
2363  */
2364 static void ice_set_ctrlq_len(struct ice_hw *hw)
2365 {
2366 	hw->adminq.num_rq_entries = ICE_AQ_LEN;
2367 	hw->adminq.num_sq_entries = ICE_AQ_LEN;
2368 	hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2369 	hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2370 	hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2371 	hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2372 	hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2373 	hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2374 	hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2375 	hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2376 	hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2377 	hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2378 }
2379 
2380 /**
2381  * ice_schedule_reset - schedule a reset
2382  * @pf: board private structure
2383  * @reset: reset being requested
2384  */
2385 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2386 {
2387 	struct device *dev = ice_pf_to_dev(pf);
2388 
2389 	/* bail out if earlier reset has failed */
2390 	if (test_bit(ICE_RESET_FAILED, pf->state)) {
2391 		dev_dbg(dev, "earlier reset has failed\n");
2392 		return -EIO;
2393 	}
2394 	/* bail if reset/recovery already in progress */
2395 	if (ice_is_reset_in_progress(pf->state)) {
2396 		dev_dbg(dev, "Reset already in progress\n");
2397 		return -EBUSY;
2398 	}
2399 
2400 	ice_unplug_aux_dev(pf);
2401 
2402 	switch (reset) {
2403 	case ICE_RESET_PFR:
2404 		set_bit(ICE_PFR_REQ, pf->state);
2405 		break;
2406 	case ICE_RESET_CORER:
2407 		set_bit(ICE_CORER_REQ, pf->state);
2408 		break;
2409 	case ICE_RESET_GLOBR:
2410 		set_bit(ICE_GLOBR_REQ, pf->state);
2411 		break;
2412 	default:
2413 		return -EINVAL;
2414 	}
2415 
2416 	ice_service_task_schedule(pf);
2417 	return 0;
2418 }
2419 
2420 /**
2421  * ice_irq_affinity_notify - Callback for affinity changes
2422  * @notify: context as to what irq was changed
2423  * @mask: the new affinity mask
2424  *
2425  * This is a callback function used by the irq_set_affinity_notifier function
2426  * so that we may register to receive changes to the irq affinity masks.
2427  */
2428 static void
2429 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2430 			const cpumask_t *mask)
2431 {
2432 	struct ice_q_vector *q_vector =
2433 		container_of(notify, struct ice_q_vector, affinity_notify);
2434 
2435 	cpumask_copy(&q_vector->affinity_mask, mask);
2436 }
2437 
2438 /**
2439  * ice_irq_affinity_release - Callback for affinity notifier release
2440  * @ref: internal core kernel usage
2441  *
2442  * This is a callback function used by the irq_set_affinity_notifier function
2443  * to inform the current notification subscriber that they will no longer
2444  * receive notifications.
2445  */
2446 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2447 
2448 /**
2449  * ice_vsi_ena_irq - Enable IRQ for the given VSI
2450  * @vsi: the VSI being configured
2451  */
2452 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2453 {
2454 	struct ice_hw *hw = &vsi->back->hw;
2455 	int i;
2456 
2457 	ice_for_each_q_vector(vsi, i)
2458 		ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2459 
2460 	ice_flush(hw);
2461 	return 0;
2462 }
2463 
2464 /**
2465  * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2466  * @vsi: the VSI being configured
2467  * @basename: name for the vector
2468  */
2469 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2470 {
2471 	int q_vectors = vsi->num_q_vectors;
2472 	struct ice_pf *pf = vsi->back;
2473 	int base = vsi->base_vector;
2474 	struct device *dev;
2475 	int rx_int_idx = 0;
2476 	int tx_int_idx = 0;
2477 	int vector, err;
2478 	int irq_num;
2479 
2480 	dev = ice_pf_to_dev(pf);
2481 	for (vector = 0; vector < q_vectors; vector++) {
2482 		struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2483 
2484 		irq_num = pf->msix_entries[base + vector].vector;
2485 
2486 		if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2487 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2488 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2489 			tx_int_idx++;
2490 		} else if (q_vector->rx.rx_ring) {
2491 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2492 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
2493 		} else if (q_vector->tx.tx_ring) {
2494 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2495 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
2496 		} else {
2497 			/* skip this unused q_vector */
2498 			continue;
2499 		}
2500 		if (vsi->type == ICE_VSI_CTRL && vsi->vf)
2501 			err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2502 					       IRQF_SHARED, q_vector->name,
2503 					       q_vector);
2504 		else
2505 			err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2506 					       0, q_vector->name, q_vector);
2507 		if (err) {
2508 			netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2509 				   err);
2510 			goto free_q_irqs;
2511 		}
2512 
2513 		/* register for affinity change notifications */
2514 		if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2515 			struct irq_affinity_notify *affinity_notify;
2516 
2517 			affinity_notify = &q_vector->affinity_notify;
2518 			affinity_notify->notify = ice_irq_affinity_notify;
2519 			affinity_notify->release = ice_irq_affinity_release;
2520 			irq_set_affinity_notifier(irq_num, affinity_notify);
2521 		}
2522 
2523 		/* assign the mask for this irq */
2524 		irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2525 	}
2526 
2527 	err = ice_set_cpu_rx_rmap(vsi);
2528 	if (err) {
2529 		netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n",
2530 			   vsi->vsi_num, ERR_PTR(err));
2531 		goto free_q_irqs;
2532 	}
2533 
2534 	vsi->irqs_ready = true;
2535 	return 0;
2536 
2537 free_q_irqs:
2538 	while (vector) {
2539 		vector--;
2540 		irq_num = pf->msix_entries[base + vector].vector;
2541 		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2542 			irq_set_affinity_notifier(irq_num, NULL);
2543 		irq_set_affinity_hint(irq_num, NULL);
2544 		devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2545 	}
2546 	return err;
2547 }
2548 
2549 /**
2550  * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2551  * @vsi: VSI to setup Tx rings used by XDP
2552  *
2553  * Return 0 on success and negative value on error
2554  */
2555 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2556 {
2557 	struct device *dev = ice_pf_to_dev(vsi->back);
2558 	struct ice_tx_desc *tx_desc;
2559 	int i, j;
2560 
2561 	ice_for_each_xdp_txq(vsi, i) {
2562 		u16 xdp_q_idx = vsi->alloc_txq + i;
2563 		struct ice_tx_ring *xdp_ring;
2564 
2565 		xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2566 
2567 		if (!xdp_ring)
2568 			goto free_xdp_rings;
2569 
2570 		xdp_ring->q_index = xdp_q_idx;
2571 		xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2572 		xdp_ring->vsi = vsi;
2573 		xdp_ring->netdev = NULL;
2574 		xdp_ring->dev = dev;
2575 		xdp_ring->count = vsi->num_tx_desc;
2576 		xdp_ring->next_dd = ICE_RING_QUARTER(xdp_ring) - 1;
2577 		xdp_ring->next_rs = ICE_RING_QUARTER(xdp_ring) - 1;
2578 		WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2579 		if (ice_setup_tx_ring(xdp_ring))
2580 			goto free_xdp_rings;
2581 		ice_set_ring_xdp(xdp_ring);
2582 		xdp_ring->xsk_pool = ice_tx_xsk_pool(xdp_ring);
2583 		spin_lock_init(&xdp_ring->tx_lock);
2584 		for (j = 0; j < xdp_ring->count; j++) {
2585 			tx_desc = ICE_TX_DESC(xdp_ring, j);
2586 			tx_desc->cmd_type_offset_bsz = 0;
2587 		}
2588 	}
2589 
2590 	ice_for_each_rxq(vsi, i) {
2591 		if (static_key_enabled(&ice_xdp_locking_key))
2592 			vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2593 		else
2594 			vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i];
2595 	}
2596 
2597 	return 0;
2598 
2599 free_xdp_rings:
2600 	for (; i >= 0; i--)
2601 		if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2602 			ice_free_tx_ring(vsi->xdp_rings[i]);
2603 	return -ENOMEM;
2604 }
2605 
2606 /**
2607  * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2608  * @vsi: VSI to set the bpf prog on
2609  * @prog: the bpf prog pointer
2610  */
2611 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2612 {
2613 	struct bpf_prog *old_prog;
2614 	int i;
2615 
2616 	old_prog = xchg(&vsi->xdp_prog, prog);
2617 	if (old_prog)
2618 		bpf_prog_put(old_prog);
2619 
2620 	ice_for_each_rxq(vsi, i)
2621 		WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2622 }
2623 
2624 /**
2625  * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2626  * @vsi: VSI to bring up Tx rings used by XDP
2627  * @prog: bpf program that will be assigned to VSI
2628  *
2629  * Return 0 on success and negative value on error
2630  */
2631 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2632 {
2633 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2634 	int xdp_rings_rem = vsi->num_xdp_txq;
2635 	struct ice_pf *pf = vsi->back;
2636 	struct ice_qs_cfg xdp_qs_cfg = {
2637 		.qs_mutex = &pf->avail_q_mutex,
2638 		.pf_map = pf->avail_txqs,
2639 		.pf_map_size = pf->max_pf_txqs,
2640 		.q_count = vsi->num_xdp_txq,
2641 		.scatter_count = ICE_MAX_SCATTER_TXQS,
2642 		.vsi_map = vsi->txq_map,
2643 		.vsi_map_offset = vsi->alloc_txq,
2644 		.mapping_mode = ICE_VSI_MAP_CONTIG
2645 	};
2646 	struct device *dev;
2647 	int i, v_idx;
2648 	int status;
2649 
2650 	dev = ice_pf_to_dev(pf);
2651 	vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2652 				      sizeof(*vsi->xdp_rings), GFP_KERNEL);
2653 	if (!vsi->xdp_rings)
2654 		return -ENOMEM;
2655 
2656 	vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2657 	if (__ice_vsi_get_qs(&xdp_qs_cfg))
2658 		goto err_map_xdp;
2659 
2660 	if (static_key_enabled(&ice_xdp_locking_key))
2661 		netdev_warn(vsi->netdev,
2662 			    "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2663 
2664 	if (ice_xdp_alloc_setup_rings(vsi))
2665 		goto clear_xdp_rings;
2666 
2667 	/* follow the logic from ice_vsi_map_rings_to_vectors */
2668 	ice_for_each_q_vector(vsi, v_idx) {
2669 		struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2670 		int xdp_rings_per_v, q_id, q_base;
2671 
2672 		xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2673 					       vsi->num_q_vectors - v_idx);
2674 		q_base = vsi->num_xdp_txq - xdp_rings_rem;
2675 
2676 		for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2677 			struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2678 
2679 			xdp_ring->q_vector = q_vector;
2680 			xdp_ring->next = q_vector->tx.tx_ring;
2681 			q_vector->tx.tx_ring = xdp_ring;
2682 		}
2683 		xdp_rings_rem -= xdp_rings_per_v;
2684 	}
2685 
2686 	/* omit the scheduler update if in reset path; XDP queues will be
2687 	 * taken into account at the end of ice_vsi_rebuild, where
2688 	 * ice_cfg_vsi_lan is being called
2689 	 */
2690 	if (ice_is_reset_in_progress(pf->state))
2691 		return 0;
2692 
2693 	/* tell the Tx scheduler that right now we have
2694 	 * additional queues
2695 	 */
2696 	for (i = 0; i < vsi->tc_cfg.numtc; i++)
2697 		max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2698 
2699 	status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2700 				 max_txqs);
2701 	if (status) {
2702 		dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2703 			status);
2704 		goto clear_xdp_rings;
2705 	}
2706 
2707 	/* assign the prog only when it's not already present on VSI;
2708 	 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2709 	 * VSI rebuild that happens under ethtool -L can expose us to
2710 	 * the bpf_prog refcount issues as we would be swapping same
2711 	 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2712 	 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2713 	 * this is not harmful as dev_xdp_install bumps the refcount
2714 	 * before calling the op exposed by the driver;
2715 	 */
2716 	if (!ice_is_xdp_ena_vsi(vsi))
2717 		ice_vsi_assign_bpf_prog(vsi, prog);
2718 
2719 	return 0;
2720 clear_xdp_rings:
2721 	ice_for_each_xdp_txq(vsi, i)
2722 		if (vsi->xdp_rings[i]) {
2723 			kfree_rcu(vsi->xdp_rings[i], rcu);
2724 			vsi->xdp_rings[i] = NULL;
2725 		}
2726 
2727 err_map_xdp:
2728 	mutex_lock(&pf->avail_q_mutex);
2729 	ice_for_each_xdp_txq(vsi, i) {
2730 		clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2731 		vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2732 	}
2733 	mutex_unlock(&pf->avail_q_mutex);
2734 
2735 	devm_kfree(dev, vsi->xdp_rings);
2736 	return -ENOMEM;
2737 }
2738 
2739 /**
2740  * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2741  * @vsi: VSI to remove XDP rings
2742  *
2743  * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2744  * resources
2745  */
2746 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2747 {
2748 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2749 	struct ice_pf *pf = vsi->back;
2750 	int i, v_idx;
2751 
2752 	/* q_vectors are freed in reset path so there's no point in detaching
2753 	 * rings; in case of rebuild being triggered not from reset bits
2754 	 * in pf->state won't be set, so additionally check first q_vector
2755 	 * against NULL
2756 	 */
2757 	if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2758 		goto free_qmap;
2759 
2760 	ice_for_each_q_vector(vsi, v_idx) {
2761 		struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2762 		struct ice_tx_ring *ring;
2763 
2764 		ice_for_each_tx_ring(ring, q_vector->tx)
2765 			if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2766 				break;
2767 
2768 		/* restore the value of last node prior to XDP setup */
2769 		q_vector->tx.tx_ring = ring;
2770 	}
2771 
2772 free_qmap:
2773 	mutex_lock(&pf->avail_q_mutex);
2774 	ice_for_each_xdp_txq(vsi, i) {
2775 		clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2776 		vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2777 	}
2778 	mutex_unlock(&pf->avail_q_mutex);
2779 
2780 	ice_for_each_xdp_txq(vsi, i)
2781 		if (vsi->xdp_rings[i]) {
2782 			if (vsi->xdp_rings[i]->desc) {
2783 				synchronize_rcu();
2784 				ice_free_tx_ring(vsi->xdp_rings[i]);
2785 			}
2786 			kfree_rcu(vsi->xdp_rings[i], rcu);
2787 			vsi->xdp_rings[i] = NULL;
2788 		}
2789 
2790 	devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2791 	vsi->xdp_rings = NULL;
2792 
2793 	if (static_key_enabled(&ice_xdp_locking_key))
2794 		static_branch_dec(&ice_xdp_locking_key);
2795 
2796 	if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2797 		return 0;
2798 
2799 	ice_vsi_assign_bpf_prog(vsi, NULL);
2800 
2801 	/* notify Tx scheduler that we destroyed XDP queues and bring
2802 	 * back the old number of child nodes
2803 	 */
2804 	for (i = 0; i < vsi->tc_cfg.numtc; i++)
2805 		max_txqs[i] = vsi->num_txq;
2806 
2807 	/* change number of XDP Tx queues to 0 */
2808 	vsi->num_xdp_txq = 0;
2809 
2810 	return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2811 			       max_txqs);
2812 }
2813 
2814 /**
2815  * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2816  * @vsi: VSI to schedule napi on
2817  */
2818 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2819 {
2820 	int i;
2821 
2822 	ice_for_each_rxq(vsi, i) {
2823 		struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2824 
2825 		if (rx_ring->xsk_pool)
2826 			napi_schedule(&rx_ring->q_vector->napi);
2827 	}
2828 }
2829 
2830 /**
2831  * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2832  * @vsi: VSI to determine the count of XDP Tx qs
2833  *
2834  * returns 0 if Tx qs count is higher than at least half of CPU count,
2835  * -ENOMEM otherwise
2836  */
2837 int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2838 {
2839 	u16 avail = ice_get_avail_txq_count(vsi->back);
2840 	u16 cpus = num_possible_cpus();
2841 
2842 	if (avail < cpus / 2)
2843 		return -ENOMEM;
2844 
2845 	vsi->num_xdp_txq = min_t(u16, avail, cpus);
2846 
2847 	if (vsi->num_xdp_txq < cpus)
2848 		static_branch_inc(&ice_xdp_locking_key);
2849 
2850 	return 0;
2851 }
2852 
2853 /**
2854  * ice_xdp_setup_prog - Add or remove XDP eBPF program
2855  * @vsi: VSI to setup XDP for
2856  * @prog: XDP program
2857  * @extack: netlink extended ack
2858  */
2859 static int
2860 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2861 		   struct netlink_ext_ack *extack)
2862 {
2863 	int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2864 	bool if_running = netif_running(vsi->netdev);
2865 	int ret = 0, xdp_ring_err = 0;
2866 
2867 	if (frame_size > vsi->rx_buf_len) {
2868 		NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2869 		return -EOPNOTSUPP;
2870 	}
2871 
2872 	/* need to stop netdev while setting up the program for Rx rings */
2873 	if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2874 		ret = ice_down(vsi);
2875 		if (ret) {
2876 			NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2877 			return ret;
2878 		}
2879 	}
2880 
2881 	if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2882 		xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2883 		if (xdp_ring_err) {
2884 			NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2885 		} else {
2886 			xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2887 			if (xdp_ring_err)
2888 				NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2889 		}
2890 	} else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2891 		xdp_ring_err = ice_destroy_xdp_rings(vsi);
2892 		if (xdp_ring_err)
2893 			NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2894 	} else {
2895 		/* safe to call even when prog == vsi->xdp_prog as
2896 		 * dev_xdp_install in net/core/dev.c incremented prog's
2897 		 * refcount so corresponding bpf_prog_put won't cause
2898 		 * underflow
2899 		 */
2900 		ice_vsi_assign_bpf_prog(vsi, prog);
2901 	}
2902 
2903 	if (if_running)
2904 		ret = ice_up(vsi);
2905 
2906 	if (!ret && prog)
2907 		ice_vsi_rx_napi_schedule(vsi);
2908 
2909 	return (ret || xdp_ring_err) ? -ENOMEM : 0;
2910 }
2911 
2912 /**
2913  * ice_xdp_safe_mode - XDP handler for safe mode
2914  * @dev: netdevice
2915  * @xdp: XDP command
2916  */
2917 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2918 			     struct netdev_bpf *xdp)
2919 {
2920 	NL_SET_ERR_MSG_MOD(xdp->extack,
2921 			   "Please provide working DDP firmware package in order to use XDP\n"
2922 			   "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2923 	return -EOPNOTSUPP;
2924 }
2925 
2926 /**
2927  * ice_xdp - implements XDP handler
2928  * @dev: netdevice
2929  * @xdp: XDP command
2930  */
2931 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2932 {
2933 	struct ice_netdev_priv *np = netdev_priv(dev);
2934 	struct ice_vsi *vsi = np->vsi;
2935 
2936 	if (vsi->type != ICE_VSI_PF) {
2937 		NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2938 		return -EINVAL;
2939 	}
2940 
2941 	switch (xdp->command) {
2942 	case XDP_SETUP_PROG:
2943 		return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2944 	case XDP_SETUP_XSK_POOL:
2945 		return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2946 					  xdp->xsk.queue_id);
2947 	default:
2948 		return -EINVAL;
2949 	}
2950 }
2951 
2952 /**
2953  * ice_ena_misc_vector - enable the non-queue interrupts
2954  * @pf: board private structure
2955  */
2956 static void ice_ena_misc_vector(struct ice_pf *pf)
2957 {
2958 	struct ice_hw *hw = &pf->hw;
2959 	u32 val;
2960 
2961 	/* Disable anti-spoof detection interrupt to prevent spurious event
2962 	 * interrupts during a function reset. Anti-spoof functionally is
2963 	 * still supported.
2964 	 */
2965 	val = rd32(hw, GL_MDCK_TX_TDPU);
2966 	val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2967 	wr32(hw, GL_MDCK_TX_TDPU, val);
2968 
2969 	/* clear things first */
2970 	wr32(hw, PFINT_OICR_ENA, 0);	/* disable all */
2971 	rd32(hw, PFINT_OICR);		/* read to clear */
2972 
2973 	val = (PFINT_OICR_ECC_ERR_M |
2974 	       PFINT_OICR_MAL_DETECT_M |
2975 	       PFINT_OICR_GRST_M |
2976 	       PFINT_OICR_PCI_EXCEPTION_M |
2977 	       PFINT_OICR_VFLR_M |
2978 	       PFINT_OICR_HMC_ERR_M |
2979 	       PFINT_OICR_PE_PUSH_M |
2980 	       PFINT_OICR_PE_CRITERR_M);
2981 
2982 	wr32(hw, PFINT_OICR_ENA, val);
2983 
2984 	/* SW_ITR_IDX = 0, but don't change INTENA */
2985 	wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2986 	     GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2987 }
2988 
2989 /**
2990  * ice_misc_intr - misc interrupt handler
2991  * @irq: interrupt number
2992  * @data: pointer to a q_vector
2993  */
2994 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2995 {
2996 	struct ice_pf *pf = (struct ice_pf *)data;
2997 	struct ice_hw *hw = &pf->hw;
2998 	irqreturn_t ret = IRQ_NONE;
2999 	struct device *dev;
3000 	u32 oicr, ena_mask;
3001 
3002 	dev = ice_pf_to_dev(pf);
3003 	set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
3004 	set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
3005 	set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
3006 
3007 	oicr = rd32(hw, PFINT_OICR);
3008 	ena_mask = rd32(hw, PFINT_OICR_ENA);
3009 
3010 	if (oicr & PFINT_OICR_SWINT_M) {
3011 		ena_mask &= ~PFINT_OICR_SWINT_M;
3012 		pf->sw_int_count++;
3013 	}
3014 
3015 	if (oicr & PFINT_OICR_MAL_DETECT_M) {
3016 		ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
3017 		set_bit(ICE_MDD_EVENT_PENDING, pf->state);
3018 	}
3019 	if (oicr & PFINT_OICR_VFLR_M) {
3020 		/* disable any further VFLR event notifications */
3021 		if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
3022 			u32 reg = rd32(hw, PFINT_OICR_ENA);
3023 
3024 			reg &= ~PFINT_OICR_VFLR_M;
3025 			wr32(hw, PFINT_OICR_ENA, reg);
3026 		} else {
3027 			ena_mask &= ~PFINT_OICR_VFLR_M;
3028 			set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
3029 		}
3030 	}
3031 
3032 	if (oicr & PFINT_OICR_GRST_M) {
3033 		u32 reset;
3034 
3035 		/* we have a reset warning */
3036 		ena_mask &= ~PFINT_OICR_GRST_M;
3037 		reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
3038 			GLGEN_RSTAT_RESET_TYPE_S;
3039 
3040 		if (reset == ICE_RESET_CORER)
3041 			pf->corer_count++;
3042 		else if (reset == ICE_RESET_GLOBR)
3043 			pf->globr_count++;
3044 		else if (reset == ICE_RESET_EMPR)
3045 			pf->empr_count++;
3046 		else
3047 			dev_dbg(dev, "Invalid reset type %d\n", reset);
3048 
3049 		/* If a reset cycle isn't already in progress, we set a bit in
3050 		 * pf->state so that the service task can start a reset/rebuild.
3051 		 */
3052 		if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3053 			if (reset == ICE_RESET_CORER)
3054 				set_bit(ICE_CORER_RECV, pf->state);
3055 			else if (reset == ICE_RESET_GLOBR)
3056 				set_bit(ICE_GLOBR_RECV, pf->state);
3057 			else
3058 				set_bit(ICE_EMPR_RECV, pf->state);
3059 
3060 			/* There are couple of different bits at play here.
3061 			 * hw->reset_ongoing indicates whether the hardware is
3062 			 * in reset. This is set to true when a reset interrupt
3063 			 * is received and set back to false after the driver
3064 			 * has determined that the hardware is out of reset.
3065 			 *
3066 			 * ICE_RESET_OICR_RECV in pf->state indicates
3067 			 * that a post reset rebuild is required before the
3068 			 * driver is operational again. This is set above.
3069 			 *
3070 			 * As this is the start of the reset/rebuild cycle, set
3071 			 * both to indicate that.
3072 			 */
3073 			hw->reset_ongoing = true;
3074 		}
3075 	}
3076 
3077 	if (oicr & PFINT_OICR_TSYN_TX_M) {
3078 		ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3079 		ice_ptp_process_ts(pf);
3080 	}
3081 
3082 	if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3083 		u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3084 		u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3085 
3086 		/* Save EVENTs from GTSYN register */
3087 		pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
3088 						     GLTSYN_STAT_EVENT1_M |
3089 						     GLTSYN_STAT_EVENT2_M);
3090 		ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3091 		kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
3092 	}
3093 
3094 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3095 	if (oicr & ICE_AUX_CRIT_ERR) {
3096 		pf->oicr_err_reg |= oicr;
3097 		set_bit(ICE_AUX_ERR_PENDING, pf->state);
3098 		ena_mask &= ~ICE_AUX_CRIT_ERR;
3099 	}
3100 
3101 	/* Report any remaining unexpected interrupts */
3102 	oicr &= ena_mask;
3103 	if (oicr) {
3104 		dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3105 		/* If a critical error is pending there is no choice but to
3106 		 * reset the device.
3107 		 */
3108 		if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3109 			    PFINT_OICR_ECC_ERR_M)) {
3110 			set_bit(ICE_PFR_REQ, pf->state);
3111 			ice_service_task_schedule(pf);
3112 		}
3113 	}
3114 	ret = IRQ_HANDLED;
3115 
3116 	ice_service_task_schedule(pf);
3117 	ice_irq_dynamic_ena(hw, NULL, NULL);
3118 
3119 	return ret;
3120 }
3121 
3122 /**
3123  * ice_dis_ctrlq_interrupts - disable control queue interrupts
3124  * @hw: pointer to HW structure
3125  */
3126 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3127 {
3128 	/* disable Admin queue Interrupt causes */
3129 	wr32(hw, PFINT_FW_CTL,
3130 	     rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3131 
3132 	/* disable Mailbox queue Interrupt causes */
3133 	wr32(hw, PFINT_MBX_CTL,
3134 	     rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3135 
3136 	wr32(hw, PFINT_SB_CTL,
3137 	     rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3138 
3139 	/* disable Control queue Interrupt causes */
3140 	wr32(hw, PFINT_OICR_CTL,
3141 	     rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3142 
3143 	ice_flush(hw);
3144 }
3145 
3146 /**
3147  * ice_free_irq_msix_misc - Unroll misc vector setup
3148  * @pf: board private structure
3149  */
3150 static void ice_free_irq_msix_misc(struct ice_pf *pf)
3151 {
3152 	struct ice_hw *hw = &pf->hw;
3153 
3154 	ice_dis_ctrlq_interrupts(hw);
3155 
3156 	/* disable OICR interrupt */
3157 	wr32(hw, PFINT_OICR_ENA, 0);
3158 	ice_flush(hw);
3159 
3160 	if (pf->msix_entries) {
3161 		synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
3162 		devm_free_irq(ice_pf_to_dev(pf),
3163 			      pf->msix_entries[pf->oicr_idx].vector, pf);
3164 	}
3165 
3166 	pf->num_avail_sw_msix += 1;
3167 	ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
3168 }
3169 
3170 /**
3171  * ice_ena_ctrlq_interrupts - enable control queue interrupts
3172  * @hw: pointer to HW structure
3173  * @reg_idx: HW vector index to associate the control queue interrupts with
3174  */
3175 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3176 {
3177 	u32 val;
3178 
3179 	val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3180 	       PFINT_OICR_CTL_CAUSE_ENA_M);
3181 	wr32(hw, PFINT_OICR_CTL, val);
3182 
3183 	/* enable Admin queue Interrupt causes */
3184 	val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3185 	       PFINT_FW_CTL_CAUSE_ENA_M);
3186 	wr32(hw, PFINT_FW_CTL, val);
3187 
3188 	/* enable Mailbox queue Interrupt causes */
3189 	val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3190 	       PFINT_MBX_CTL_CAUSE_ENA_M);
3191 	wr32(hw, PFINT_MBX_CTL, val);
3192 
3193 	/* This enables Sideband queue Interrupt causes */
3194 	val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3195 	       PFINT_SB_CTL_CAUSE_ENA_M);
3196 	wr32(hw, PFINT_SB_CTL, val);
3197 
3198 	ice_flush(hw);
3199 }
3200 
3201 /**
3202  * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3203  * @pf: board private structure
3204  *
3205  * This sets up the handler for MSIX 0, which is used to manage the
3206  * non-queue interrupts, e.g. AdminQ and errors. This is not used
3207  * when in MSI or Legacy interrupt mode.
3208  */
3209 static int ice_req_irq_msix_misc(struct ice_pf *pf)
3210 {
3211 	struct device *dev = ice_pf_to_dev(pf);
3212 	struct ice_hw *hw = &pf->hw;
3213 	int oicr_idx, err = 0;
3214 
3215 	if (!pf->int_name[0])
3216 		snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3217 			 dev_driver_string(dev), dev_name(dev));
3218 
3219 	/* Do not request IRQ but do enable OICR interrupt since settings are
3220 	 * lost during reset. Note that this function is called only during
3221 	 * rebuild path and not while reset is in progress.
3222 	 */
3223 	if (ice_is_reset_in_progress(pf->state))
3224 		goto skip_req_irq;
3225 
3226 	/* reserve one vector in irq_tracker for misc interrupts */
3227 	oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3228 	if (oicr_idx < 0)
3229 		return oicr_idx;
3230 
3231 	pf->num_avail_sw_msix -= 1;
3232 	pf->oicr_idx = (u16)oicr_idx;
3233 
3234 	err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
3235 			       ice_misc_intr, 0, pf->int_name, pf);
3236 	if (err) {
3237 		dev_err(dev, "devm_request_irq for %s failed: %d\n",
3238 			pf->int_name, err);
3239 		ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3240 		pf->num_avail_sw_msix += 1;
3241 		return err;
3242 	}
3243 
3244 skip_req_irq:
3245 	ice_ena_misc_vector(pf);
3246 
3247 	ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3248 	wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3249 	     ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3250 
3251 	ice_flush(hw);
3252 	ice_irq_dynamic_ena(hw, NULL, NULL);
3253 
3254 	return 0;
3255 }
3256 
3257 /**
3258  * ice_napi_add - register NAPI handler for the VSI
3259  * @vsi: VSI for which NAPI handler is to be registered
3260  *
3261  * This function is only called in the driver's load path. Registering the NAPI
3262  * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3263  * reset/rebuild, etc.)
3264  */
3265 static void ice_napi_add(struct ice_vsi *vsi)
3266 {
3267 	int v_idx;
3268 
3269 	if (!vsi->netdev)
3270 		return;
3271 
3272 	ice_for_each_q_vector(vsi, v_idx)
3273 		netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3274 			       ice_napi_poll, NAPI_POLL_WEIGHT);
3275 }
3276 
3277 /**
3278  * ice_set_ops - set netdev and ethtools ops for the given netdev
3279  * @netdev: netdev instance
3280  */
3281 static void ice_set_ops(struct net_device *netdev)
3282 {
3283 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
3284 
3285 	if (ice_is_safe_mode(pf)) {
3286 		netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3287 		ice_set_ethtool_safe_mode_ops(netdev);
3288 		return;
3289 	}
3290 
3291 	netdev->netdev_ops = &ice_netdev_ops;
3292 	netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3293 	ice_set_ethtool_ops(netdev);
3294 }
3295 
3296 /**
3297  * ice_set_netdev_features - set features for the given netdev
3298  * @netdev: netdev instance
3299  */
3300 static void ice_set_netdev_features(struct net_device *netdev)
3301 {
3302 	struct ice_pf *pf = ice_netdev_to_pf(netdev);
3303 	bool is_dvm_ena = ice_is_dvm_ena(&pf->hw);
3304 	netdev_features_t csumo_features;
3305 	netdev_features_t vlano_features;
3306 	netdev_features_t dflt_features;
3307 	netdev_features_t tso_features;
3308 
3309 	if (ice_is_safe_mode(pf)) {
3310 		/* safe mode */
3311 		netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3312 		netdev->hw_features = netdev->features;
3313 		return;
3314 	}
3315 
3316 	dflt_features = NETIF_F_SG	|
3317 			NETIF_F_HIGHDMA	|
3318 			NETIF_F_NTUPLE	|
3319 			NETIF_F_RXHASH;
3320 
3321 	csumo_features = NETIF_F_RXCSUM	  |
3322 			 NETIF_F_IP_CSUM  |
3323 			 NETIF_F_SCTP_CRC |
3324 			 NETIF_F_IPV6_CSUM;
3325 
3326 	vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3327 			 NETIF_F_HW_VLAN_CTAG_TX     |
3328 			 NETIF_F_HW_VLAN_CTAG_RX;
3329 
3330 	/* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */
3331 	if (is_dvm_ena)
3332 		vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER;
3333 
3334 	tso_features = NETIF_F_TSO			|
3335 		       NETIF_F_TSO_ECN			|
3336 		       NETIF_F_TSO6			|
3337 		       NETIF_F_GSO_GRE			|
3338 		       NETIF_F_GSO_UDP_TUNNEL		|
3339 		       NETIF_F_GSO_GRE_CSUM		|
3340 		       NETIF_F_GSO_UDP_TUNNEL_CSUM	|
3341 		       NETIF_F_GSO_PARTIAL		|
3342 		       NETIF_F_GSO_IPXIP4		|
3343 		       NETIF_F_GSO_IPXIP6		|
3344 		       NETIF_F_GSO_UDP_L4;
3345 
3346 	netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3347 					NETIF_F_GSO_GRE_CSUM;
3348 	/* set features that user can change */
3349 	netdev->hw_features = dflt_features | csumo_features |
3350 			      vlano_features | tso_features;
3351 
3352 	/* add support for HW_CSUM on packets with MPLS header */
3353 	netdev->mpls_features =  NETIF_F_HW_CSUM |
3354 				 NETIF_F_TSO     |
3355 				 NETIF_F_TSO6;
3356 
3357 	/* enable features */
3358 	netdev->features |= netdev->hw_features;
3359 
3360 	netdev->hw_features |= NETIF_F_HW_TC;
3361 
3362 	/* encap and VLAN devices inherit default, csumo and tso features */
3363 	netdev->hw_enc_features |= dflt_features | csumo_features |
3364 				   tso_features;
3365 	netdev->vlan_features |= dflt_features | csumo_features |
3366 				 tso_features;
3367 
3368 	/* advertise support but don't enable by default since only one type of
3369 	 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one
3370 	 * type turns on the other has to be turned off. This is enforced by the
3371 	 * ice_fix_features() ndo callback.
3372 	 */
3373 	if (is_dvm_ena)
3374 		netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX |
3375 			NETIF_F_HW_VLAN_STAG_TX;
3376 }
3377 
3378 /**
3379  * ice_cfg_netdev - Allocate, configure and register a netdev
3380  * @vsi: the VSI associated with the new netdev
3381  *
3382  * Returns 0 on success, negative value on failure
3383  */
3384 static int ice_cfg_netdev(struct ice_vsi *vsi)
3385 {
3386 	struct ice_netdev_priv *np;
3387 	struct net_device *netdev;
3388 	u8 mac_addr[ETH_ALEN];
3389 
3390 	netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3391 				    vsi->alloc_rxq);
3392 	if (!netdev)
3393 		return -ENOMEM;
3394 
3395 	set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3396 	vsi->netdev = netdev;
3397 	np = netdev_priv(netdev);
3398 	np->vsi = vsi;
3399 
3400 	ice_set_netdev_features(netdev);
3401 
3402 	ice_set_ops(netdev);
3403 
3404 	if (vsi->type == ICE_VSI_PF) {
3405 		SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
3406 		ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3407 		eth_hw_addr_set(netdev, mac_addr);
3408 		ether_addr_copy(netdev->perm_addr, mac_addr);
3409 	}
3410 
3411 	netdev->priv_flags |= IFF_UNICAST_FLT;
3412 
3413 	/* Setup netdev TC information */
3414 	ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3415 
3416 	/* setup watchdog timeout value to be 5 second */
3417 	netdev->watchdog_timeo = 5 * HZ;
3418 
3419 	netdev->min_mtu = ETH_MIN_MTU;
3420 	netdev->max_mtu = ICE_MAX_MTU;
3421 
3422 	return 0;
3423 }
3424 
3425 /**
3426  * ice_fill_rss_lut - Fill the RSS lookup table with default values
3427  * @lut: Lookup table
3428  * @rss_table_size: Lookup table size
3429  * @rss_size: Range of queue number for hashing
3430  */
3431 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3432 {
3433 	u16 i;
3434 
3435 	for (i = 0; i < rss_table_size; i++)
3436 		lut[i] = i % rss_size;
3437 }
3438 
3439 /**
3440  * ice_pf_vsi_setup - Set up a PF VSI
3441  * @pf: board private structure
3442  * @pi: pointer to the port_info instance
3443  *
3444  * Returns pointer to the successfully allocated VSI software struct
3445  * on success, otherwise returns NULL on failure.
3446  */
3447 static struct ice_vsi *
3448 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3449 {
3450 	return ice_vsi_setup(pf, pi, ICE_VSI_PF, NULL, NULL);
3451 }
3452 
3453 static struct ice_vsi *
3454 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3455 		   struct ice_channel *ch)
3456 {
3457 	return ice_vsi_setup(pf, pi, ICE_VSI_CHNL, NULL, ch);
3458 }
3459 
3460 /**
3461  * ice_ctrl_vsi_setup - Set up a control VSI
3462  * @pf: board private structure
3463  * @pi: pointer to the port_info instance
3464  *
3465  * Returns pointer to the successfully allocated VSI software struct
3466  * on success, otherwise returns NULL on failure.
3467  */
3468 static struct ice_vsi *
3469 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3470 {
3471 	return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, NULL, NULL);
3472 }
3473 
3474 /**
3475  * ice_lb_vsi_setup - Set up a loopback VSI
3476  * @pf: board private structure
3477  * @pi: pointer to the port_info instance
3478  *
3479  * Returns pointer to the successfully allocated VSI software struct
3480  * on success, otherwise returns NULL on failure.
3481  */
3482 struct ice_vsi *
3483 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3484 {
3485 	return ice_vsi_setup(pf, pi, ICE_VSI_LB, NULL, NULL);
3486 }
3487 
3488 /**
3489  * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3490  * @netdev: network interface to be adjusted
3491  * @proto: VLAN TPID
3492  * @vid: VLAN ID to be added
3493  *
3494  * net_device_ops implementation for adding VLAN IDs
3495  */
3496 static int
3497 ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
3498 {
3499 	struct ice_netdev_priv *np = netdev_priv(netdev);
3500 	struct ice_vsi_vlan_ops *vlan_ops;
3501 	struct ice_vsi *vsi = np->vsi;
3502 	struct ice_vlan vlan;
3503 	int ret;
3504 
3505 	/* VLAN 0 is added by default during load/reset */
3506 	if (!vid)
3507 		return 0;
3508 
3509 	while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3510 		usleep_range(1000, 2000);
3511 
3512 	/* Add multicast promisc rule for the VLAN ID to be added if
3513 	 * all-multicast is currently enabled.
3514 	 */
3515 	if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3516 		ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3517 					       ICE_MCAST_VLAN_PROMISC_BITS,
3518 					       vid);
3519 		if (ret)
3520 			goto finish;
3521 	}
3522 
3523 	vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3524 
3525 	/* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3526 	 * packets aren't pruned by the device's internal switch on Rx
3527 	 */
3528 	vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3529 	ret = vlan_ops->add_vlan(vsi, &vlan);
3530 	if (ret)
3531 		goto finish;
3532 
3533 	/* If all-multicast is currently enabled and this VLAN ID is only one
3534 	 * besides VLAN-0 we have to update look-up type of multicast promisc
3535 	 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN.
3536 	 */
3537 	if ((vsi->current_netdev_flags & IFF_ALLMULTI) &&
3538 	    ice_vsi_num_non_zero_vlans(vsi) == 1) {
3539 		ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3540 					   ICE_MCAST_PROMISC_BITS, 0);
3541 		ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3542 					 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3543 	}
3544 
3545 finish:
3546 	clear_bit(ICE_CFG_BUSY, vsi->state);
3547 
3548 	return ret;
3549 }
3550 
3551 /**
3552  * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3553  * @netdev: network interface to be adjusted
3554  * @proto: VLAN TPID
3555  * @vid: VLAN ID to be removed
3556  *
3557  * net_device_ops implementation for removing VLAN IDs
3558  */
3559 static int
3560 ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
3561 {
3562 	struct ice_netdev_priv *np = netdev_priv(netdev);
3563 	struct ice_vsi_vlan_ops *vlan_ops;
3564 	struct ice_vsi *vsi = np->vsi;
3565 	struct ice_vlan vlan;
3566 	int ret;
3567 
3568 	/* don't allow removal of VLAN 0 */
3569 	if (!vid)
3570 		return 0;
3571 
3572 	while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3573 		usleep_range(1000, 2000);
3574 
3575 	vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3576 
3577 	/* Make sure VLAN delete is successful before updating VLAN
3578 	 * information
3579 	 */
3580 	vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3581 	ret = vlan_ops->del_vlan(vsi, &vlan);
3582 	if (ret)
3583 		goto finish;
3584 
3585 	/* Remove multicast promisc rule for the removed VLAN ID if
3586 	 * all-multicast is enabled.
3587 	 */
3588 	if (vsi->current_netdev_flags & IFF_ALLMULTI)
3589 		ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3590 					   ICE_MCAST_VLAN_PROMISC_BITS, vid);
3591 
3592 	if (!ice_vsi_has_non_zero_vlans(vsi)) {
3593 		/* Update look-up type of multicast promisc rule for VLAN 0
3594 		 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when
3595 		 * all-multicast is enabled and VLAN 0 is the only VLAN rule.
3596 		 */
3597 		if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3598 			ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3599 						   ICE_MCAST_VLAN_PROMISC_BITS,
3600 						   0);
3601 			ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3602 						 ICE_MCAST_PROMISC_BITS, 0);
3603 		}
3604 	}
3605 
3606 finish:
3607 	clear_bit(ICE_CFG_BUSY, vsi->state);
3608 
3609 	return ret;
3610 }
3611 
3612 /**
3613  * ice_rep_indr_tc_block_unbind
3614  * @cb_priv: indirection block private data
3615  */
3616 static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3617 {
3618 	struct ice_indr_block_priv *indr_priv = cb_priv;
3619 
3620 	list_del(&indr_priv->list);
3621 	kfree(indr_priv);
3622 }
3623 
3624 /**
3625  * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3626  * @vsi: VSI struct which has the netdev
3627  */
3628 static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3629 {
3630 	struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3631 
3632 	flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3633 				 ice_rep_indr_tc_block_unbind);
3634 }
3635 
3636 /**
3637  * ice_tc_indir_block_remove - clean indirect TC block notifications
3638  * @pf: PF structure
3639  */
3640 static void ice_tc_indir_block_remove(struct ice_pf *pf)
3641 {
3642 	struct ice_vsi *pf_vsi = ice_get_main_vsi(pf);
3643 
3644 	if (!pf_vsi)
3645 		return;
3646 
3647 	ice_tc_indir_block_unregister(pf_vsi);
3648 }
3649 
3650 /**
3651  * ice_tc_indir_block_register - Register TC indirect block notifications
3652  * @vsi: VSI struct which has the netdev
3653  *
3654  * Returns 0 on success, negative value on failure
3655  */
3656 static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3657 {
3658 	struct ice_netdev_priv *np;
3659 
3660 	if (!vsi || !vsi->netdev)
3661 		return -EINVAL;
3662 
3663 	np = netdev_priv(vsi->netdev);
3664 
3665 	INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3666 	return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3667 }
3668 
3669 /**
3670  * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3671  * @pf: board private structure
3672  *
3673  * Returns 0 on success, negative value on failure
3674  */
3675 static int ice_setup_pf_sw(struct ice_pf *pf)
3676 {
3677 	struct device *dev = ice_pf_to_dev(pf);
3678 	bool dvm = ice_is_dvm_ena(&pf->hw);
3679 	struct ice_vsi *vsi;
3680 	int status;
3681 
3682 	if (ice_is_reset_in_progress(pf->state))
3683 		return -EBUSY;
3684 
3685 	status = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
3686 	if (status)
3687 		return -EIO;
3688 
3689 	vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3690 	if (!vsi)
3691 		return -ENOMEM;
3692 
3693 	/* init channel list */
3694 	INIT_LIST_HEAD(&vsi->ch_list);
3695 
3696 	status = ice_cfg_netdev(vsi);
3697 	if (status)
3698 		goto unroll_vsi_setup;
3699 	/* netdev has to be configured before setting frame size */
3700 	ice_vsi_cfg_frame_size(vsi);
3701 
3702 	/* init indirect block notifications */
3703 	status = ice_tc_indir_block_register(vsi);
3704 	if (status) {
3705 		dev_err(dev, "Failed to register netdev notifier\n");
3706 		goto unroll_cfg_netdev;
3707 	}
3708 
3709 	/* Setup DCB netlink interface */
3710 	ice_dcbnl_setup(vsi);
3711 
3712 	/* registering the NAPI handler requires both the queues and
3713 	 * netdev to be created, which are done in ice_pf_vsi_setup()
3714 	 * and ice_cfg_netdev() respectively
3715 	 */
3716 	ice_napi_add(vsi);
3717 
3718 	status = ice_init_mac_fltr(pf);
3719 	if (status)
3720 		goto unroll_napi_add;
3721 
3722 	return 0;
3723 
3724 unroll_napi_add:
3725 	ice_tc_indir_block_unregister(vsi);
3726 unroll_cfg_netdev:
3727 	if (vsi) {
3728 		ice_napi_del(vsi);
3729 		if (vsi->netdev) {
3730 			clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3731 			free_netdev(vsi->netdev);
3732 			vsi->netdev = NULL;
3733 		}
3734 	}
3735 
3736 unroll_vsi_setup:
3737 	ice_vsi_release(vsi);
3738 	return status;
3739 }
3740 
3741 /**
3742  * ice_get_avail_q_count - Get count of queues in use
3743  * @pf_qmap: bitmap to get queue use count from
3744  * @lock: pointer to a mutex that protects access to pf_qmap
3745  * @size: size of the bitmap
3746  */
3747 static u16
3748 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3749 {
3750 	unsigned long bit;
3751 	u16 count = 0;
3752 
3753 	mutex_lock(lock);
3754 	for_each_clear_bit(bit, pf_qmap, size)
3755 		count++;
3756 	mutex_unlock(lock);
3757 
3758 	return count;
3759 }
3760 
3761 /**
3762  * ice_get_avail_txq_count - Get count of Tx queues in use
3763  * @pf: pointer to an ice_pf instance
3764  */
3765 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3766 {
3767 	return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3768 				     pf->max_pf_txqs);
3769 }
3770 
3771 /**
3772  * ice_get_avail_rxq_count - Get count of Rx queues in use
3773  * @pf: pointer to an ice_pf instance
3774  */
3775 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3776 {
3777 	return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3778 				     pf->max_pf_rxqs);
3779 }
3780 
3781 /**
3782  * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3783  * @pf: board private structure to initialize
3784  */
3785 static void ice_deinit_pf(struct ice_pf *pf)
3786 {
3787 	ice_service_task_stop(pf);
3788 	mutex_destroy(&pf->adev_mutex);
3789 	mutex_destroy(&pf->sw_mutex);
3790 	mutex_destroy(&pf->tc_mutex);
3791 	mutex_destroy(&pf->avail_q_mutex);
3792 	mutex_destroy(&pf->vfs.table_lock);
3793 
3794 	if (pf->avail_txqs) {
3795 		bitmap_free(pf->avail_txqs);
3796 		pf->avail_txqs = NULL;
3797 	}
3798 
3799 	if (pf->avail_rxqs) {
3800 		bitmap_free(pf->avail_rxqs);
3801 		pf->avail_rxqs = NULL;
3802 	}
3803 
3804 	if (pf->ptp.clock)
3805 		ptp_clock_unregister(pf->ptp.clock);
3806 }
3807 
3808 /**
3809  * ice_set_pf_caps - set PFs capability flags
3810  * @pf: pointer to the PF instance
3811  */
3812 static void ice_set_pf_caps(struct ice_pf *pf)
3813 {
3814 	struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3815 
3816 	clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3817 	if (func_caps->common_cap.rdma)
3818 		set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3819 	clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3820 	if (func_caps->common_cap.dcb)
3821 		set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3822 	clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3823 	if (func_caps->common_cap.sr_iov_1_1) {
3824 		set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3825 		pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
3826 					      ICE_MAX_SRIOV_VFS);
3827 	}
3828 	clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3829 	if (func_caps->common_cap.rss_table_size)
3830 		set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3831 
3832 	clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3833 	if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3834 		u16 unused;
3835 
3836 		/* ctrl_vsi_idx will be set to a valid value when flow director
3837 		 * is setup by ice_init_fdir
3838 		 */
3839 		pf->ctrl_vsi_idx = ICE_NO_VSI;
3840 		set_bit(ICE_FLAG_FD_ENA, pf->flags);
3841 		/* force guaranteed filter pool for PF */
3842 		ice_alloc_fd_guar_item(&pf->hw, &unused,
3843 				       func_caps->fd_fltr_guar);
3844 		/* force shared filter pool for PF */
3845 		ice_alloc_fd_shrd_item(&pf->hw, &unused,
3846 				       func_caps->fd_fltr_best_effort);
3847 	}
3848 
3849 	clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3850 	if (func_caps->common_cap.ieee_1588)
3851 		set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3852 
3853 	pf->max_pf_txqs = func_caps->common_cap.num_txq;
3854 	pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3855 }
3856 
3857 /**
3858  * ice_init_pf - Initialize general software structures (struct ice_pf)
3859  * @pf: board private structure to initialize
3860  */
3861 static int ice_init_pf(struct ice_pf *pf)
3862 {
3863 	ice_set_pf_caps(pf);
3864 
3865 	mutex_init(&pf->sw_mutex);
3866 	mutex_init(&pf->tc_mutex);
3867 	mutex_init(&pf->adev_mutex);
3868 
3869 	INIT_HLIST_HEAD(&pf->aq_wait_list);
3870 	spin_lock_init(&pf->aq_wait_lock);
3871 	init_waitqueue_head(&pf->aq_wait_queue);
3872 
3873 	init_waitqueue_head(&pf->reset_wait_queue);
3874 
3875 	/* setup service timer and periodic service task */
3876 	timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3877 	pf->serv_tmr_period = HZ;
3878 	INIT_WORK(&pf->serv_task, ice_service_task);
3879 	clear_bit(ICE_SERVICE_SCHED, pf->state);
3880 
3881 	mutex_init(&pf->avail_q_mutex);
3882 	pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3883 	if (!pf->avail_txqs)
3884 		return -ENOMEM;
3885 
3886 	pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3887 	if (!pf->avail_rxqs) {
3888 		devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3889 		pf->avail_txqs = NULL;
3890 		return -ENOMEM;
3891 	}
3892 
3893 	mutex_init(&pf->vfs.table_lock);
3894 	hash_init(pf->vfs.table);
3895 
3896 	return 0;
3897 }
3898 
3899 /**
3900  * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3901  * @pf: board private structure
3902  *
3903  * compute the number of MSIX vectors required (v_budget) and request from
3904  * the OS. Return the number of vectors reserved or negative on failure
3905  */
3906 static int ice_ena_msix_range(struct ice_pf *pf)
3907 {
3908 	int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3909 	struct device *dev = ice_pf_to_dev(pf);
3910 	int needed, err, i;
3911 
3912 	v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3913 	num_cpus = num_online_cpus();
3914 
3915 	/* reserve for LAN miscellaneous handler */
3916 	needed = ICE_MIN_LAN_OICR_MSIX;
3917 	if (v_left < needed)
3918 		goto no_hw_vecs_left_err;
3919 	v_budget += needed;
3920 	v_left -= needed;
3921 
3922 	/* reserve for flow director */
3923 	if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3924 		needed = ICE_FDIR_MSIX;
3925 		if (v_left < needed)
3926 			goto no_hw_vecs_left_err;
3927 		v_budget += needed;
3928 		v_left -= needed;
3929 	}
3930 
3931 	/* reserve for switchdev */
3932 	needed = ICE_ESWITCH_MSIX;
3933 	if (v_left < needed)
3934 		goto no_hw_vecs_left_err;
3935 	v_budget += needed;
3936 	v_left -= needed;
3937 
3938 	/* total used for non-traffic vectors */
3939 	v_other = v_budget;
3940 
3941 	/* reserve vectors for LAN traffic */
3942 	needed = num_cpus;
3943 	if (v_left < needed)
3944 		goto no_hw_vecs_left_err;
3945 	pf->num_lan_msix = needed;
3946 	v_budget += needed;
3947 	v_left -= needed;
3948 
3949 	/* reserve vectors for RDMA auxiliary driver */
3950 	if (ice_is_rdma_ena(pf)) {
3951 		needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3952 		if (v_left < needed)
3953 			goto no_hw_vecs_left_err;
3954 		pf->num_rdma_msix = needed;
3955 		v_budget += needed;
3956 		v_left -= needed;
3957 	}
3958 
3959 	pf->msix_entries = devm_kcalloc(dev, v_budget,
3960 					sizeof(*pf->msix_entries), GFP_KERNEL);
3961 	if (!pf->msix_entries) {
3962 		err = -ENOMEM;
3963 		goto exit_err;
3964 	}
3965 
3966 	for (i = 0; i < v_budget; i++)
3967 		pf->msix_entries[i].entry = i;
3968 
3969 	/* actually reserve the vectors */
3970 	v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3971 					 ICE_MIN_MSIX, v_budget);
3972 	if (v_actual < 0) {
3973 		dev_err(dev, "unable to reserve MSI-X vectors\n");
3974 		err = v_actual;
3975 		goto msix_err;
3976 	}
3977 
3978 	if (v_actual < v_budget) {
3979 		dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3980 			 v_budget, v_actual);
3981 
3982 		if (v_actual < ICE_MIN_MSIX) {
3983 			/* error if we can't get minimum vectors */
3984 			pci_disable_msix(pf->pdev);
3985 			err = -ERANGE;
3986 			goto msix_err;
3987 		} else {
3988 			int v_remain = v_actual - v_other;
3989 			int v_rdma = 0, v_min_rdma = 0;
3990 
3991 			if (ice_is_rdma_ena(pf)) {
3992 				/* Need at least 1 interrupt in addition to
3993 				 * AEQ MSIX
3994 				 */
3995 				v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
3996 				v_min_rdma = ICE_MIN_RDMA_MSIX;
3997 			}
3998 
3999 			if (v_actual == ICE_MIN_MSIX ||
4000 			    v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
4001 				dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
4002 				clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
4003 
4004 				pf->num_rdma_msix = 0;
4005 				pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
4006 			} else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
4007 				   (v_remain - v_rdma < v_rdma)) {
4008 				/* Support minimum RDMA and give remaining
4009 				 * vectors to LAN MSIX
4010 				 */
4011 				pf->num_rdma_msix = v_min_rdma;
4012 				pf->num_lan_msix = v_remain - v_min_rdma;
4013 			} else {
4014 				/* Split remaining MSIX with RDMA after
4015 				 * accounting for AEQ MSIX
4016 				 */
4017 				pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
4018 						    ICE_RDMA_NUM_AEQ_MSIX;
4019 				pf->num_lan_msix = v_remain - pf->num_rdma_msix;
4020 			}
4021 
4022 			dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
4023 				   pf->num_lan_msix);
4024 
4025 			if (ice_is_rdma_ena(pf))
4026 				dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
4027 					   pf->num_rdma_msix);
4028 		}
4029 	}
4030 
4031 	return v_actual;
4032 
4033 msix_err:
4034 	devm_kfree(dev, pf->msix_entries);
4035 	goto exit_err;
4036 
4037 no_hw_vecs_left_err:
4038 	dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
4039 		needed, v_left);
4040 	err = -ERANGE;
4041 exit_err:
4042 	pf->num_rdma_msix = 0;
4043 	pf->num_lan_msix = 0;
4044 	return err;
4045 }
4046 
4047 /**
4048  * ice_dis_msix - Disable MSI-X interrupt setup in OS
4049  * @pf: board private structure
4050  */
4051 static void ice_dis_msix(struct ice_pf *pf)
4052 {
4053 	pci_disable_msix(pf->pdev);
4054 	devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
4055 	pf->msix_entries = NULL;
4056 }
4057 
4058 /**
4059  * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
4060  * @pf: board private structure
4061  */
4062 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
4063 {
4064 	ice_dis_msix(pf);
4065 
4066 	if (pf->irq_tracker) {
4067 		devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
4068 		pf->irq_tracker = NULL;
4069 	}
4070 }
4071 
4072 /**
4073  * ice_init_interrupt_scheme - Determine proper interrupt scheme
4074  * @pf: board private structure to initialize
4075  */
4076 static int ice_init_interrupt_scheme(struct ice_pf *pf)
4077 {
4078 	int vectors;
4079 
4080 	vectors = ice_ena_msix_range(pf);
4081 
4082 	if (vectors < 0)
4083 		return vectors;
4084 
4085 	/* set up vector assignment tracking */
4086 	pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
4087 				       struct_size(pf->irq_tracker, list, vectors),
4088 				       GFP_KERNEL);
4089 	if (!pf->irq_tracker) {
4090 		ice_dis_msix(pf);
4091 		return -ENOMEM;
4092 	}
4093 
4094 	/* populate SW interrupts pool with number of OS granted IRQs. */
4095 	pf->num_avail_sw_msix = (u16)vectors;
4096 	pf->irq_tracker->num_entries = (u16)vectors;
4097 	pf->irq_tracker->end = pf->irq_tracker->num_entries;
4098 
4099 	return 0;
4100 }
4101 
4102 /**
4103  * ice_is_wol_supported - check if WoL is supported
4104  * @hw: pointer to hardware info
4105  *
4106  * Check if WoL is supported based on the HW configuration.
4107  * Returns true if NVM supports and enables WoL for this port, false otherwise
4108  */
4109 bool ice_is_wol_supported(struct ice_hw *hw)
4110 {
4111 	u16 wol_ctrl;
4112 
4113 	/* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
4114 	 * word) indicates WoL is not supported on the corresponding PF ID.
4115 	 */
4116 	if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
4117 		return false;
4118 
4119 	return !(BIT(hw->port_info->lport) & wol_ctrl);
4120 }
4121 
4122 /**
4123  * ice_vsi_recfg_qs - Change the number of queues on a VSI
4124  * @vsi: VSI being changed
4125  * @new_rx: new number of Rx queues
4126  * @new_tx: new number of Tx queues
4127  *
4128  * Only change the number of queues if new_tx, or new_rx is non-0.
4129  *
4130  * Returns 0 on success.
4131  */
4132 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
4133 {
4134 	struct ice_pf *pf = vsi->back;
4135 	int err = 0, timeout = 50;
4136 
4137 	if (!new_rx && !new_tx)
4138 		return -EINVAL;
4139 
4140 	while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
4141 		timeout--;
4142 		if (!timeout)
4143 			return -EBUSY;
4144 		usleep_range(1000, 2000);
4145 	}
4146 
4147 	if (new_tx)
4148 		vsi->req_txq = (u16)new_tx;
4149 	if (new_rx)
4150 		vsi->req_rxq = (u16)new_rx;
4151 
4152 	/* set for the next time the netdev is started */
4153 	if (!netif_running(vsi->netdev)) {
4154 		ice_vsi_rebuild(vsi, false);
4155 		dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
4156 		goto done;
4157 	}
4158 
4159 	ice_vsi_close(vsi);
4160 	ice_vsi_rebuild(vsi, false);
4161 	ice_pf_dcb_recfg(pf);
4162 	ice_vsi_open(vsi);
4163 done:
4164 	clear_bit(ICE_CFG_BUSY, pf->state);
4165 	return err;
4166 }
4167 
4168 /**
4169  * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
4170  * @pf: PF to configure
4171  *
4172  * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
4173  * VSI can still Tx/Rx VLAN tagged packets.
4174  */
4175 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4176 {
4177 	struct ice_vsi *vsi = ice_get_main_vsi(pf);
4178 	struct ice_vsi_ctx *ctxt;
4179 	struct ice_hw *hw;
4180 	int status;
4181 
4182 	if (!vsi)
4183 		return;
4184 
4185 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4186 	if (!ctxt)
4187 		return;
4188 
4189 	hw = &pf->hw;
4190 	ctxt->info = vsi->info;
4191 
4192 	ctxt->info.valid_sections =
4193 		cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4194 			    ICE_AQ_VSI_PROP_SECURITY_VALID |
4195 			    ICE_AQ_VSI_PROP_SW_VALID);
4196 
4197 	/* disable VLAN anti-spoof */
4198 	ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4199 				  ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4200 
4201 	/* disable VLAN pruning and keep all other settings */
4202 	ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4203 
4204 	/* allow all VLANs on Tx and don't strip on Rx */
4205 	ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL |
4206 		ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
4207 
4208 	status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4209 	if (status) {
4210 		dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4211 			status, ice_aq_str(hw->adminq.sq_last_status));
4212 	} else {
4213 		vsi->info.sec_flags = ctxt->info.sec_flags;
4214 		vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4215 		vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags;
4216 	}
4217 
4218 	kfree(ctxt);
4219 }
4220 
4221 /**
4222  * ice_log_pkg_init - log result of DDP package load
4223  * @hw: pointer to hardware info
4224  * @state: state of package load
4225  */
4226 static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4227 {
4228 	struct ice_pf *pf = hw->back;
4229 	struct device *dev;
4230 
4231 	dev = ice_pf_to_dev(pf);
4232 
4233 	switch (state) {
4234 	case ICE_DDP_PKG_SUCCESS:
4235 		dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4236 			 hw->active_pkg_name,
4237 			 hw->active_pkg_ver.major,
4238 			 hw->active_pkg_ver.minor,
4239 			 hw->active_pkg_ver.update,
4240 			 hw->active_pkg_ver.draft);
4241 		break;
4242 	case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4243 		dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4244 			 hw->active_pkg_name,
4245 			 hw->active_pkg_ver.major,
4246 			 hw->active_pkg_ver.minor,
4247 			 hw->active_pkg_ver.update,
4248 			 hw->active_pkg_ver.draft);
4249 		break;
4250 	case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4251 		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",
4252 			hw->active_pkg_name,
4253 			hw->active_pkg_ver.major,
4254 			hw->active_pkg_ver.minor,
4255 			ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4256 		break;
4257 	case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4258 		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",
4259 			 hw->active_pkg_name,
4260 			 hw->active_pkg_ver.major,
4261 			 hw->active_pkg_ver.minor,
4262 			 hw->active_pkg_ver.update,
4263 			 hw->active_pkg_ver.draft,
4264 			 hw->pkg_name,
4265 			 hw->pkg_ver.major,
4266 			 hw->pkg_ver.minor,
4267 			 hw->pkg_ver.update,
4268 			 hw->pkg_ver.draft);
4269 		break;
4270 	case ICE_DDP_PKG_FW_MISMATCH:
4271 		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");
4272 		break;
4273 	case ICE_DDP_PKG_INVALID_FILE:
4274 		dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4275 		break;
4276 	case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4277 		dev_err(dev, "The DDP package file version is higher than the driver supports.  Please use an updated driver.  Entering Safe Mode.\n");
4278 		break;
4279 	case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4280 		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",
4281 			ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4282 		break;
4283 	case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4284 		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");
4285 		break;
4286 	case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4287 		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");
4288 		break;
4289 	case ICE_DDP_PKG_LOAD_ERROR:
4290 		dev_err(dev, "An error occurred on the device while loading the DDP package.  The device will be reset.\n");
4291 		/* poll for reset to complete */
4292 		if (ice_check_reset(hw))
4293 			dev_err(dev, "Error resetting device. Please reload the driver\n");
4294 		break;
4295 	case ICE_DDP_PKG_ERR:
4296 	default:
4297 		dev_err(dev, "An unknown error occurred when loading the DDP package.  Entering Safe Mode.\n");
4298 		break;
4299 	}
4300 }
4301 
4302 /**
4303  * ice_load_pkg - load/reload the DDP Package file
4304  * @firmware: firmware structure when firmware requested or NULL for reload
4305  * @pf: pointer to the PF instance
4306  *
4307  * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4308  * initialize HW tables.
4309  */
4310 static void
4311 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4312 {
4313 	enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4314 	struct device *dev = ice_pf_to_dev(pf);
4315 	struct ice_hw *hw = &pf->hw;
4316 
4317 	/* Load DDP Package */
4318 	if (firmware && !hw->pkg_copy) {
4319 		state = ice_copy_and_init_pkg(hw, firmware->data,
4320 					      firmware->size);
4321 		ice_log_pkg_init(hw, state);
4322 	} else if (!firmware && hw->pkg_copy) {
4323 		/* Reload package during rebuild after CORER/GLOBR reset */
4324 		state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4325 		ice_log_pkg_init(hw, state);
4326 	} else {
4327 		dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4328 	}
4329 
4330 	if (!ice_is_init_pkg_successful(state)) {
4331 		/* Safe Mode */
4332 		clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4333 		return;
4334 	}
4335 
4336 	/* Successful download package is the precondition for advanced
4337 	 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4338 	 */
4339 	set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4340 }
4341 
4342 /**
4343  * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4344  * @pf: pointer to the PF structure
4345  *
4346  * There is no error returned here because the driver should be able to handle
4347  * 128 Byte cache lines, so we only print a warning in case issues are seen,
4348  * specifically with Tx.
4349  */
4350 static void ice_verify_cacheline_size(struct ice_pf *pf)
4351 {
4352 	if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4353 		dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4354 			 ICE_CACHE_LINE_BYTES);
4355 }
4356 
4357 /**
4358  * ice_send_version - update firmware with driver version
4359  * @pf: PF struct
4360  *
4361  * Returns 0 on success, else error code
4362  */
4363 static int ice_send_version(struct ice_pf *pf)
4364 {
4365 	struct ice_driver_ver dv;
4366 
4367 	dv.major_ver = 0xff;
4368 	dv.minor_ver = 0xff;
4369 	dv.build_ver = 0xff;
4370 	dv.subbuild_ver = 0;
4371 	strscpy((char *)dv.driver_string, UTS_RELEASE,
4372 		sizeof(dv.driver_string));
4373 	return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4374 }
4375 
4376 /**
4377  * ice_init_fdir - Initialize flow director VSI and configuration
4378  * @pf: pointer to the PF instance
4379  *
4380  * returns 0 on success, negative on error
4381  */
4382 static int ice_init_fdir(struct ice_pf *pf)
4383 {
4384 	struct device *dev = ice_pf_to_dev(pf);
4385 	struct ice_vsi *ctrl_vsi;
4386 	int err;
4387 
4388 	/* Side Band Flow Director needs to have a control VSI.
4389 	 * Allocate it and store it in the PF.
4390 	 */
4391 	ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4392 	if (!ctrl_vsi) {
4393 		dev_dbg(dev, "could not create control VSI\n");
4394 		return -ENOMEM;
4395 	}
4396 
4397 	err = ice_vsi_open_ctrl(ctrl_vsi);
4398 	if (err) {
4399 		dev_dbg(dev, "could not open control VSI\n");
4400 		goto err_vsi_open;
4401 	}
4402 
4403 	mutex_init(&pf->hw.fdir_fltr_lock);
4404 
4405 	err = ice_fdir_create_dflt_rules(pf);
4406 	if (err)
4407 		goto err_fdir_rule;
4408 
4409 	return 0;
4410 
4411 err_fdir_rule:
4412 	ice_fdir_release_flows(&pf->hw);
4413 	ice_vsi_close(ctrl_vsi);
4414 err_vsi_open:
4415 	ice_vsi_release(ctrl_vsi);
4416 	if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4417 		pf->vsi[pf->ctrl_vsi_idx] = NULL;
4418 		pf->ctrl_vsi_idx = ICE_NO_VSI;
4419 	}
4420 	return err;
4421 }
4422 
4423 /**
4424  * ice_get_opt_fw_name - return optional firmware file name or NULL
4425  * @pf: pointer to the PF instance
4426  */
4427 static char *ice_get_opt_fw_name(struct ice_pf *pf)
4428 {
4429 	/* Optional firmware name same as default with additional dash
4430 	 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4431 	 */
4432 	struct pci_dev *pdev = pf->pdev;
4433 	char *opt_fw_filename;
4434 	u64 dsn;
4435 
4436 	/* Determine the name of the optional file using the DSN (two
4437 	 * dwords following the start of the DSN Capability).
4438 	 */
4439 	dsn = pci_get_dsn(pdev);
4440 	if (!dsn)
4441 		return NULL;
4442 
4443 	opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4444 	if (!opt_fw_filename)
4445 		return NULL;
4446 
4447 	snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4448 		 ICE_DDP_PKG_PATH, dsn);
4449 
4450 	return opt_fw_filename;
4451 }
4452 
4453 /**
4454  * ice_request_fw - Device initialization routine
4455  * @pf: pointer to the PF instance
4456  */
4457 static void ice_request_fw(struct ice_pf *pf)
4458 {
4459 	char *opt_fw_filename = ice_get_opt_fw_name(pf);
4460 	const struct firmware *firmware = NULL;
4461 	struct device *dev = ice_pf_to_dev(pf);
4462 	int err = 0;
4463 
4464 	/* optional device-specific DDP (if present) overrides the default DDP
4465 	 * package file. kernel logs a debug message if the file doesn't exist,
4466 	 * and warning messages for other errors.
4467 	 */
4468 	if (opt_fw_filename) {
4469 		err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4470 		if (err) {
4471 			kfree(opt_fw_filename);
4472 			goto dflt_pkg_load;
4473 		}
4474 
4475 		/* request for firmware was successful. Download to device */
4476 		ice_load_pkg(firmware, pf);
4477 		kfree(opt_fw_filename);
4478 		release_firmware(firmware);
4479 		return;
4480 	}
4481 
4482 dflt_pkg_load:
4483 	err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4484 	if (err) {
4485 		dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4486 		return;
4487 	}
4488 
4489 	/* request for firmware was successful. Download to device */
4490 	ice_load_pkg(firmware, pf);
4491 	release_firmware(firmware);
4492 }
4493 
4494 /**
4495  * ice_print_wake_reason - show the wake up cause in the log
4496  * @pf: pointer to the PF struct
4497  */
4498 static void ice_print_wake_reason(struct ice_pf *pf)
4499 {
4500 	u32 wus = pf->wakeup_reason;
4501 	const char *wake_str;
4502 
4503 	/* if no wake event, nothing to print */
4504 	if (!wus)
4505 		return;
4506 
4507 	if (wus & PFPM_WUS_LNKC_M)
4508 		wake_str = "Link\n";
4509 	else if (wus & PFPM_WUS_MAG_M)
4510 		wake_str = "Magic Packet\n";
4511 	else if (wus & PFPM_WUS_MNG_M)
4512 		wake_str = "Management\n";
4513 	else if (wus & PFPM_WUS_FW_RST_WK_M)
4514 		wake_str = "Firmware Reset\n";
4515 	else
4516 		wake_str = "Unknown\n";
4517 
4518 	dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4519 }
4520 
4521 /**
4522  * ice_register_netdev - register netdev and devlink port
4523  * @pf: pointer to the PF struct
4524  */
4525 static int ice_register_netdev(struct ice_pf *pf)
4526 {
4527 	struct ice_vsi *vsi;
4528 	int err = 0;
4529 
4530 	vsi = ice_get_main_vsi(pf);
4531 	if (!vsi || !vsi->netdev)
4532 		return -EIO;
4533 
4534 	err = register_netdev(vsi->netdev);
4535 	if (err)
4536 		goto err_register_netdev;
4537 
4538 	set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4539 	netif_carrier_off(vsi->netdev);
4540 	netif_tx_stop_all_queues(vsi->netdev);
4541 	err = ice_devlink_create_pf_port(pf);
4542 	if (err)
4543 		goto err_devlink_create;
4544 
4545 	devlink_port_type_eth_set(&pf->devlink_port, vsi->netdev);
4546 
4547 	return 0;
4548 err_devlink_create:
4549 	unregister_netdev(vsi->netdev);
4550 	clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4551 err_register_netdev:
4552 	free_netdev(vsi->netdev);
4553 	vsi->netdev = NULL;
4554 	clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4555 	return err;
4556 }
4557 
4558 /**
4559  * ice_probe - Device initialization routine
4560  * @pdev: PCI device information struct
4561  * @ent: entry in ice_pci_tbl
4562  *
4563  * Returns 0 on success, negative on failure
4564  */
4565 static int
4566 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
4567 {
4568 	struct device *dev = &pdev->dev;
4569 	struct ice_pf *pf;
4570 	struct ice_hw *hw;
4571 	int i, err;
4572 
4573 	if (pdev->is_virtfn) {
4574 		dev_err(dev, "can't probe a virtual function\n");
4575 		return -EINVAL;
4576 	}
4577 
4578 	/* this driver uses devres, see
4579 	 * Documentation/driver-api/driver-model/devres.rst
4580 	 */
4581 	err = pcim_enable_device(pdev);
4582 	if (err)
4583 		return err;
4584 
4585 	err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4586 	if (err) {
4587 		dev_err(dev, "BAR0 I/O map error %d\n", err);
4588 		return err;
4589 	}
4590 
4591 	pf = ice_allocate_pf(dev);
4592 	if (!pf)
4593 		return -ENOMEM;
4594 
4595 	/* initialize Auxiliary index to invalid value */
4596 	pf->aux_idx = -1;
4597 
4598 	/* set up for high or low DMA */
4599 	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4600 	if (err) {
4601 		dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4602 		return err;
4603 	}
4604 
4605 	pci_enable_pcie_error_reporting(pdev);
4606 	pci_set_master(pdev);
4607 
4608 	pf->pdev = pdev;
4609 	pci_set_drvdata(pdev, pf);
4610 	set_bit(ICE_DOWN, pf->state);
4611 	/* Disable service task until DOWN bit is cleared */
4612 	set_bit(ICE_SERVICE_DIS, pf->state);
4613 
4614 	hw = &pf->hw;
4615 	hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4616 	pci_save_state(pdev);
4617 
4618 	hw->back = pf;
4619 	hw->vendor_id = pdev->vendor;
4620 	hw->device_id = pdev->device;
4621 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4622 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
4623 	hw->subsystem_device_id = pdev->subsystem_device;
4624 	hw->bus.device = PCI_SLOT(pdev->devfn);
4625 	hw->bus.func = PCI_FUNC(pdev->devfn);
4626 	ice_set_ctrlq_len(hw);
4627 
4628 	pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4629 
4630 #ifndef CONFIG_DYNAMIC_DEBUG
4631 	if (debug < -1)
4632 		hw->debug_mask = debug;
4633 #endif
4634 
4635 	err = ice_init_hw(hw);
4636 	if (err) {
4637 		dev_err(dev, "ice_init_hw failed: %d\n", err);
4638 		err = -EIO;
4639 		goto err_exit_unroll;
4640 	}
4641 
4642 	ice_init_feature_support(pf);
4643 
4644 	ice_request_fw(pf);
4645 
4646 	/* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4647 	 * set in pf->state, which will cause ice_is_safe_mode to return
4648 	 * true
4649 	 */
4650 	if (ice_is_safe_mode(pf)) {
4651 		/* we already got function/device capabilities but these don't
4652 		 * reflect what the driver needs to do in safe mode. Instead of
4653 		 * adding conditional logic everywhere to ignore these
4654 		 * device/function capabilities, override them.
4655 		 */
4656 		ice_set_safe_mode_caps(hw);
4657 	}
4658 
4659 	err = ice_init_pf(pf);
4660 	if (err) {
4661 		dev_err(dev, "ice_init_pf failed: %d\n", err);
4662 		goto err_init_pf_unroll;
4663 	}
4664 
4665 	ice_devlink_init_regions(pf);
4666 
4667 	pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4668 	pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4669 	pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4670 	pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4671 	i = 0;
4672 	if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4673 		pf->hw.udp_tunnel_nic.tables[i].n_entries =
4674 			pf->hw.tnl.valid_count[TNL_VXLAN];
4675 		pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4676 			UDP_TUNNEL_TYPE_VXLAN;
4677 		i++;
4678 	}
4679 	if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4680 		pf->hw.udp_tunnel_nic.tables[i].n_entries =
4681 			pf->hw.tnl.valid_count[TNL_GENEVE];
4682 		pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4683 			UDP_TUNNEL_TYPE_GENEVE;
4684 		i++;
4685 	}
4686 
4687 	pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4688 	if (!pf->num_alloc_vsi) {
4689 		err = -EIO;
4690 		goto err_init_pf_unroll;
4691 	}
4692 	if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4693 		dev_warn(&pf->pdev->dev,
4694 			 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4695 			 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4696 		pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4697 	}
4698 
4699 	pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4700 			       GFP_KERNEL);
4701 	if (!pf->vsi) {
4702 		err = -ENOMEM;
4703 		goto err_init_pf_unroll;
4704 	}
4705 
4706 	err = ice_init_interrupt_scheme(pf);
4707 	if (err) {
4708 		dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4709 		err = -EIO;
4710 		goto err_init_vsi_unroll;
4711 	}
4712 
4713 	/* In case of MSIX we are going to setup the misc vector right here
4714 	 * to handle admin queue events etc. In case of legacy and MSI
4715 	 * the misc functionality and queue processing is combined in
4716 	 * the same vector and that gets setup at open.
4717 	 */
4718 	err = ice_req_irq_msix_misc(pf);
4719 	if (err) {
4720 		dev_err(dev, "setup of misc vector failed: %d\n", err);
4721 		goto err_init_interrupt_unroll;
4722 	}
4723 
4724 	/* create switch struct for the switch element created by FW on boot */
4725 	pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4726 	if (!pf->first_sw) {
4727 		err = -ENOMEM;
4728 		goto err_msix_misc_unroll;
4729 	}
4730 
4731 	if (hw->evb_veb)
4732 		pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4733 	else
4734 		pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4735 
4736 	pf->first_sw->pf = pf;
4737 
4738 	/* record the sw_id available for later use */
4739 	pf->first_sw->sw_id = hw->port_info->sw_id;
4740 
4741 	err = ice_setup_pf_sw(pf);
4742 	if (err) {
4743 		dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4744 		goto err_alloc_sw_unroll;
4745 	}
4746 
4747 	clear_bit(ICE_SERVICE_DIS, pf->state);
4748 
4749 	/* tell the firmware we are up */
4750 	err = ice_send_version(pf);
4751 	if (err) {
4752 		dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4753 			UTS_RELEASE, err);
4754 		goto err_send_version_unroll;
4755 	}
4756 
4757 	/* since everything is good, start the service timer */
4758 	mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4759 
4760 	err = ice_init_link_events(pf->hw.port_info);
4761 	if (err) {
4762 		dev_err(dev, "ice_init_link_events failed: %d\n", err);
4763 		goto err_send_version_unroll;
4764 	}
4765 
4766 	/* not a fatal error if this fails */
4767 	err = ice_init_nvm_phy_type(pf->hw.port_info);
4768 	if (err)
4769 		dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4770 
4771 	/* not a fatal error if this fails */
4772 	err = ice_update_link_info(pf->hw.port_info);
4773 	if (err)
4774 		dev_err(dev, "ice_update_link_info failed: %d\n", err);
4775 
4776 	ice_init_link_dflt_override(pf->hw.port_info);
4777 
4778 	ice_check_link_cfg_err(pf,
4779 			       pf->hw.port_info->phy.link_info.link_cfg_err);
4780 
4781 	/* if media available, initialize PHY settings */
4782 	if (pf->hw.port_info->phy.link_info.link_info &
4783 	    ICE_AQ_MEDIA_AVAILABLE) {
4784 		/* not a fatal error if this fails */
4785 		err = ice_init_phy_user_cfg(pf->hw.port_info);
4786 		if (err)
4787 			dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4788 
4789 		if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4790 			struct ice_vsi *vsi = ice_get_main_vsi(pf);
4791 
4792 			if (vsi)
4793 				ice_configure_phy(vsi);
4794 		}
4795 	} else {
4796 		set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4797 	}
4798 
4799 	ice_verify_cacheline_size(pf);
4800 
4801 	/* Save wakeup reason register for later use */
4802 	pf->wakeup_reason = rd32(hw, PFPM_WUS);
4803 
4804 	/* check for a power management event */
4805 	ice_print_wake_reason(pf);
4806 
4807 	/* clear wake status, all bits */
4808 	wr32(hw, PFPM_WUS, U32_MAX);
4809 
4810 	/* Disable WoL at init, wait for user to enable */
4811 	device_set_wakeup_enable(dev, false);
4812 
4813 	if (ice_is_safe_mode(pf)) {
4814 		ice_set_safe_mode_vlan_cfg(pf);
4815 		goto probe_done;
4816 	}
4817 
4818 	/* initialize DDP driven features */
4819 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4820 		ice_ptp_init(pf);
4821 
4822 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
4823 		ice_gnss_init(pf);
4824 
4825 	/* Note: Flow director init failure is non-fatal to load */
4826 	if (ice_init_fdir(pf))
4827 		dev_err(dev, "could not initialize flow director\n");
4828 
4829 	/* Note: DCB init failure is non-fatal to load */
4830 	if (ice_init_pf_dcb(pf, false)) {
4831 		clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4832 		clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4833 	} else {
4834 		ice_cfg_lldp_mib_change(&pf->hw, true);
4835 	}
4836 
4837 	if (ice_init_lag(pf))
4838 		dev_warn(dev, "Failed to init link aggregation support\n");
4839 
4840 	/* print PCI link speed and width */
4841 	pcie_print_link_status(pf->pdev);
4842 
4843 probe_done:
4844 	err = ice_register_netdev(pf);
4845 	if (err)
4846 		goto err_netdev_reg;
4847 
4848 	err = ice_devlink_register_params(pf);
4849 	if (err)
4850 		goto err_netdev_reg;
4851 
4852 	/* ready to go, so clear down state bit */
4853 	clear_bit(ICE_DOWN, pf->state);
4854 	if (ice_is_rdma_ena(pf)) {
4855 		pf->aux_idx = ida_alloc(&ice_aux_ida, GFP_KERNEL);
4856 		if (pf->aux_idx < 0) {
4857 			dev_err(dev, "Failed to allocate device ID for AUX driver\n");
4858 			err = -ENOMEM;
4859 			goto err_devlink_reg_param;
4860 		}
4861 
4862 		err = ice_init_rdma(pf);
4863 		if (err) {
4864 			dev_err(dev, "Failed to initialize RDMA: %d\n", err);
4865 			err = -EIO;
4866 			goto err_init_aux_unroll;
4867 		}
4868 	} else {
4869 		dev_warn(dev, "RDMA is not supported on this device\n");
4870 	}
4871 
4872 	ice_devlink_register(pf);
4873 	return 0;
4874 
4875 err_init_aux_unroll:
4876 	pf->adev = NULL;
4877 	ida_free(&ice_aux_ida, pf->aux_idx);
4878 err_devlink_reg_param:
4879 	ice_devlink_unregister_params(pf);
4880 err_netdev_reg:
4881 err_send_version_unroll:
4882 	ice_vsi_release_all(pf);
4883 err_alloc_sw_unroll:
4884 	set_bit(ICE_SERVICE_DIS, pf->state);
4885 	set_bit(ICE_DOWN, pf->state);
4886 	devm_kfree(dev, pf->first_sw);
4887 err_msix_misc_unroll:
4888 	ice_free_irq_msix_misc(pf);
4889 err_init_interrupt_unroll:
4890 	ice_clear_interrupt_scheme(pf);
4891 err_init_vsi_unroll:
4892 	devm_kfree(dev, pf->vsi);
4893 err_init_pf_unroll:
4894 	ice_deinit_pf(pf);
4895 	ice_devlink_destroy_regions(pf);
4896 	ice_deinit_hw(hw);
4897 err_exit_unroll:
4898 	pci_disable_pcie_error_reporting(pdev);
4899 	pci_disable_device(pdev);
4900 	return err;
4901 }
4902 
4903 /**
4904  * ice_set_wake - enable or disable Wake on LAN
4905  * @pf: pointer to the PF struct
4906  *
4907  * Simple helper for WoL control
4908  */
4909 static void ice_set_wake(struct ice_pf *pf)
4910 {
4911 	struct ice_hw *hw = &pf->hw;
4912 	bool wol = pf->wol_ena;
4913 
4914 	/* clear wake state, otherwise new wake events won't fire */
4915 	wr32(hw, PFPM_WUS, U32_MAX);
4916 
4917 	/* enable / disable APM wake up, no RMW needed */
4918 	wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4919 
4920 	/* set magic packet filter enabled */
4921 	wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4922 }
4923 
4924 /**
4925  * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4926  * @pf: pointer to the PF struct
4927  *
4928  * Issue firmware command to enable multicast magic wake, making
4929  * sure that any locally administered address (LAA) is used for
4930  * wake, and that PF reset doesn't undo the LAA.
4931  */
4932 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4933 {
4934 	struct device *dev = ice_pf_to_dev(pf);
4935 	struct ice_hw *hw = &pf->hw;
4936 	u8 mac_addr[ETH_ALEN];
4937 	struct ice_vsi *vsi;
4938 	int status;
4939 	u8 flags;
4940 
4941 	if (!pf->wol_ena)
4942 		return;
4943 
4944 	vsi = ice_get_main_vsi(pf);
4945 	if (!vsi)
4946 		return;
4947 
4948 	/* Get current MAC address in case it's an LAA */
4949 	if (vsi->netdev)
4950 		ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4951 	else
4952 		ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4953 
4954 	flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4955 		ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4956 		ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4957 
4958 	status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4959 	if (status)
4960 		dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
4961 			status, ice_aq_str(hw->adminq.sq_last_status));
4962 }
4963 
4964 /**
4965  * ice_remove - Device removal routine
4966  * @pdev: PCI device information struct
4967  */
4968 static void ice_remove(struct pci_dev *pdev)
4969 {
4970 	struct ice_pf *pf = pci_get_drvdata(pdev);
4971 	int i;
4972 
4973 	ice_devlink_unregister(pf);
4974 	for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4975 		if (!ice_is_reset_in_progress(pf->state))
4976 			break;
4977 		msleep(100);
4978 	}
4979 
4980 	ice_tc_indir_block_remove(pf);
4981 
4982 	if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4983 		set_bit(ICE_VF_RESETS_DISABLED, pf->state);
4984 		ice_free_vfs(pf);
4985 	}
4986 
4987 	ice_service_task_stop(pf);
4988 
4989 	ice_aq_cancel_waiting_tasks(pf);
4990 	ice_unplug_aux_dev(pf);
4991 	if (pf->aux_idx >= 0)
4992 		ida_free(&ice_aux_ida, pf->aux_idx);
4993 	ice_devlink_unregister_params(pf);
4994 	set_bit(ICE_DOWN, pf->state);
4995 
4996 	ice_deinit_lag(pf);
4997 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4998 		ice_ptp_release(pf);
4999 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
5000 		ice_gnss_exit(pf);
5001 	if (!ice_is_safe_mode(pf))
5002 		ice_remove_arfs(pf);
5003 	ice_setup_mc_magic_wake(pf);
5004 	ice_vsi_release_all(pf);
5005 	mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
5006 	ice_set_wake(pf);
5007 	ice_free_irq_msix_misc(pf);
5008 	ice_for_each_vsi(pf, i) {
5009 		if (!pf->vsi[i])
5010 			continue;
5011 		ice_vsi_free_q_vectors(pf->vsi[i]);
5012 	}
5013 	ice_deinit_pf(pf);
5014 	ice_devlink_destroy_regions(pf);
5015 	ice_deinit_hw(&pf->hw);
5016 
5017 	/* Issue a PFR as part of the prescribed driver unload flow.  Do not
5018 	 * do it via ice_schedule_reset() since there is no need to rebuild
5019 	 * and the service task is already stopped.
5020 	 */
5021 	ice_reset(&pf->hw, ICE_RESET_PFR);
5022 	pci_wait_for_pending_transaction(pdev);
5023 	ice_clear_interrupt_scheme(pf);
5024 	pci_disable_pcie_error_reporting(pdev);
5025 	pci_disable_device(pdev);
5026 }
5027 
5028 /**
5029  * ice_shutdown - PCI callback for shutting down device
5030  * @pdev: PCI device information struct
5031  */
5032 static void ice_shutdown(struct pci_dev *pdev)
5033 {
5034 	struct ice_pf *pf = pci_get_drvdata(pdev);
5035 
5036 	ice_remove(pdev);
5037 
5038 	if (system_state == SYSTEM_POWER_OFF) {
5039 		pci_wake_from_d3(pdev, pf->wol_ena);
5040 		pci_set_power_state(pdev, PCI_D3hot);
5041 	}
5042 }
5043 
5044 #ifdef CONFIG_PM
5045 /**
5046  * ice_prepare_for_shutdown - prep for PCI shutdown
5047  * @pf: board private structure
5048  *
5049  * Inform or close all dependent features in prep for PCI device shutdown
5050  */
5051 static void ice_prepare_for_shutdown(struct ice_pf *pf)
5052 {
5053 	struct ice_hw *hw = &pf->hw;
5054 	u32 v;
5055 
5056 	/* Notify VFs of impending reset */
5057 	if (ice_check_sq_alive(hw, &hw->mailboxq))
5058 		ice_vc_notify_reset(pf);
5059 
5060 	dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
5061 
5062 	/* disable the VSIs and their queues that are not already DOWN */
5063 	ice_pf_dis_all_vsi(pf, false);
5064 
5065 	ice_for_each_vsi(pf, v)
5066 		if (pf->vsi[v])
5067 			pf->vsi[v]->vsi_num = 0;
5068 
5069 	ice_shutdown_all_ctrlq(hw);
5070 }
5071 
5072 /**
5073  * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
5074  * @pf: board private structure to reinitialize
5075  *
5076  * This routine reinitialize interrupt scheme that was cleared during
5077  * power management suspend callback.
5078  *
5079  * This should be called during resume routine to re-allocate the q_vectors
5080  * and reacquire interrupts.
5081  */
5082 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
5083 {
5084 	struct device *dev = ice_pf_to_dev(pf);
5085 	int ret, v;
5086 
5087 	/* Since we clear MSIX flag during suspend, we need to
5088 	 * set it back during resume...
5089 	 */
5090 
5091 	ret = ice_init_interrupt_scheme(pf);
5092 	if (ret) {
5093 		dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
5094 		return ret;
5095 	}
5096 
5097 	/* Remap vectors and rings, after successful re-init interrupts */
5098 	ice_for_each_vsi(pf, v) {
5099 		if (!pf->vsi[v])
5100 			continue;
5101 
5102 		ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
5103 		if (ret)
5104 			goto err_reinit;
5105 		ice_vsi_map_rings_to_vectors(pf->vsi[v]);
5106 	}
5107 
5108 	ret = ice_req_irq_msix_misc(pf);
5109 	if (ret) {
5110 		dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5111 			ret);
5112 		goto err_reinit;
5113 	}
5114 
5115 	return 0;
5116 
5117 err_reinit:
5118 	while (v--)
5119 		if (pf->vsi[v])
5120 			ice_vsi_free_q_vectors(pf->vsi[v]);
5121 
5122 	return ret;
5123 }
5124 
5125 /**
5126  * ice_suspend
5127  * @dev: generic device information structure
5128  *
5129  * Power Management callback to quiesce the device and prepare
5130  * for D3 transition.
5131  */
5132 static int __maybe_unused ice_suspend(struct device *dev)
5133 {
5134 	struct pci_dev *pdev = to_pci_dev(dev);
5135 	struct ice_pf *pf;
5136 	int disabled, v;
5137 
5138 	pf = pci_get_drvdata(pdev);
5139 
5140 	if (!ice_pf_state_is_nominal(pf)) {
5141 		dev_err(dev, "Device is not ready, no need to suspend it\n");
5142 		return -EBUSY;
5143 	}
5144 
5145 	/* Stop watchdog tasks until resume completion.
5146 	 * Even though it is most likely that the service task is
5147 	 * disabled if the device is suspended or down, the service task's
5148 	 * state is controlled by a different state bit, and we should
5149 	 * store and honor whatever state that bit is in at this point.
5150 	 */
5151 	disabled = ice_service_task_stop(pf);
5152 
5153 	ice_unplug_aux_dev(pf);
5154 
5155 	/* Already suspended?, then there is nothing to do */
5156 	if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5157 		if (!disabled)
5158 			ice_service_task_restart(pf);
5159 		return 0;
5160 	}
5161 
5162 	if (test_bit(ICE_DOWN, pf->state) ||
5163 	    ice_is_reset_in_progress(pf->state)) {
5164 		dev_err(dev, "can't suspend device in reset or already down\n");
5165 		if (!disabled)
5166 			ice_service_task_restart(pf);
5167 		return 0;
5168 	}
5169 
5170 	ice_setup_mc_magic_wake(pf);
5171 
5172 	ice_prepare_for_shutdown(pf);
5173 
5174 	ice_set_wake(pf);
5175 
5176 	/* Free vectors, clear the interrupt scheme and release IRQs
5177 	 * for proper hibernation, especially with large number of CPUs.
5178 	 * Otherwise hibernation might fail when mapping all the vectors back
5179 	 * to CPU0.
5180 	 */
5181 	ice_free_irq_msix_misc(pf);
5182 	ice_for_each_vsi(pf, v) {
5183 		if (!pf->vsi[v])
5184 			continue;
5185 		ice_vsi_free_q_vectors(pf->vsi[v]);
5186 	}
5187 	ice_clear_interrupt_scheme(pf);
5188 
5189 	pci_save_state(pdev);
5190 	pci_wake_from_d3(pdev, pf->wol_ena);
5191 	pci_set_power_state(pdev, PCI_D3hot);
5192 	return 0;
5193 }
5194 
5195 /**
5196  * ice_resume - PM callback for waking up from D3
5197  * @dev: generic device information structure
5198  */
5199 static int __maybe_unused ice_resume(struct device *dev)
5200 {
5201 	struct pci_dev *pdev = to_pci_dev(dev);
5202 	enum ice_reset_req reset_type;
5203 	struct ice_pf *pf;
5204 	struct ice_hw *hw;
5205 	int ret;
5206 
5207 	pci_set_power_state(pdev, PCI_D0);
5208 	pci_restore_state(pdev);
5209 	pci_save_state(pdev);
5210 
5211 	if (!pci_device_is_present(pdev))
5212 		return -ENODEV;
5213 
5214 	ret = pci_enable_device_mem(pdev);
5215 	if (ret) {
5216 		dev_err(dev, "Cannot enable device after suspend\n");
5217 		return ret;
5218 	}
5219 
5220 	pf = pci_get_drvdata(pdev);
5221 	hw = &pf->hw;
5222 
5223 	pf->wakeup_reason = rd32(hw, PFPM_WUS);
5224 	ice_print_wake_reason(pf);
5225 
5226 	/* We cleared the interrupt scheme when we suspended, so we need to
5227 	 * restore it now to resume device functionality.
5228 	 */
5229 	ret = ice_reinit_interrupt_scheme(pf);
5230 	if (ret)
5231 		dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5232 
5233 	clear_bit(ICE_DOWN, pf->state);
5234 	/* Now perform PF reset and rebuild */
5235 	reset_type = ICE_RESET_PFR;
5236 	/* re-enable service task for reset, but allow reset to schedule it */
5237 	clear_bit(ICE_SERVICE_DIS, pf->state);
5238 
5239 	if (ice_schedule_reset(pf, reset_type))
5240 		dev_err(dev, "Reset during resume failed.\n");
5241 
5242 	clear_bit(ICE_SUSPENDED, pf->state);
5243 	ice_service_task_restart(pf);
5244 
5245 	/* Restart the service task */
5246 	mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5247 
5248 	return 0;
5249 }
5250 #endif /* CONFIG_PM */
5251 
5252 /**
5253  * ice_pci_err_detected - warning that PCI error has been detected
5254  * @pdev: PCI device information struct
5255  * @err: the type of PCI error
5256  *
5257  * Called to warn that something happened on the PCI bus and the error handling
5258  * is in progress.  Allows the driver to gracefully prepare/handle PCI errors.
5259  */
5260 static pci_ers_result_t
5261 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5262 {
5263 	struct ice_pf *pf = pci_get_drvdata(pdev);
5264 
5265 	if (!pf) {
5266 		dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5267 			__func__, err);
5268 		return PCI_ERS_RESULT_DISCONNECT;
5269 	}
5270 
5271 	if (!test_bit(ICE_SUSPENDED, pf->state)) {
5272 		ice_service_task_stop(pf);
5273 
5274 		if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5275 			set_bit(ICE_PFR_REQ, pf->state);
5276 			ice_prepare_for_reset(pf, ICE_RESET_PFR);
5277 		}
5278 	}
5279 
5280 	return PCI_ERS_RESULT_NEED_RESET;
5281 }
5282 
5283 /**
5284  * ice_pci_err_slot_reset - a PCI slot reset has just happened
5285  * @pdev: PCI device information struct
5286  *
5287  * Called to determine if the driver can recover from the PCI slot reset by
5288  * using a register read to determine if the device is recoverable.
5289  */
5290 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5291 {
5292 	struct ice_pf *pf = pci_get_drvdata(pdev);
5293 	pci_ers_result_t result;
5294 	int err;
5295 	u32 reg;
5296 
5297 	err = pci_enable_device_mem(pdev);
5298 	if (err) {
5299 		dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5300 			err);
5301 		result = PCI_ERS_RESULT_DISCONNECT;
5302 	} else {
5303 		pci_set_master(pdev);
5304 		pci_restore_state(pdev);
5305 		pci_save_state(pdev);
5306 		pci_wake_from_d3(pdev, false);
5307 
5308 		/* Check for life */
5309 		reg = rd32(&pf->hw, GLGEN_RTRIG);
5310 		if (!reg)
5311 			result = PCI_ERS_RESULT_RECOVERED;
5312 		else
5313 			result = PCI_ERS_RESULT_DISCONNECT;
5314 	}
5315 
5316 	err = pci_aer_clear_nonfatal_status(pdev);
5317 	if (err)
5318 		dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
5319 			err);
5320 		/* non-fatal, continue */
5321 
5322 	return result;
5323 }
5324 
5325 /**
5326  * ice_pci_err_resume - restart operations after PCI error recovery
5327  * @pdev: PCI device information struct
5328  *
5329  * Called to allow the driver to bring things back up after PCI error and/or
5330  * reset recovery have finished
5331  */
5332 static void ice_pci_err_resume(struct pci_dev *pdev)
5333 {
5334 	struct ice_pf *pf = pci_get_drvdata(pdev);
5335 
5336 	if (!pf) {
5337 		dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5338 			__func__);
5339 		return;
5340 	}
5341 
5342 	if (test_bit(ICE_SUSPENDED, pf->state)) {
5343 		dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5344 			__func__);
5345 		return;
5346 	}
5347 
5348 	ice_restore_all_vfs_msi_state(pdev);
5349 
5350 	ice_do_reset(pf, ICE_RESET_PFR);
5351 	ice_service_task_restart(pf);
5352 	mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5353 }
5354 
5355 /**
5356  * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5357  * @pdev: PCI device information struct
5358  */
5359 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5360 {
5361 	struct ice_pf *pf = pci_get_drvdata(pdev);
5362 
5363 	if (!test_bit(ICE_SUSPENDED, pf->state)) {
5364 		ice_service_task_stop(pf);
5365 
5366 		if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5367 			set_bit(ICE_PFR_REQ, pf->state);
5368 			ice_prepare_for_reset(pf, ICE_RESET_PFR);
5369 		}
5370 	}
5371 }
5372 
5373 /**
5374  * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5375  * @pdev: PCI device information struct
5376  */
5377 static void ice_pci_err_reset_done(struct pci_dev *pdev)
5378 {
5379 	ice_pci_err_resume(pdev);
5380 }
5381 
5382 /* ice_pci_tbl - PCI Device ID Table
5383  *
5384  * Wildcard entries (PCI_ANY_ID) should come last
5385  * Last entry must be all 0s
5386  *
5387  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5388  *   Class, Class Mask, private data (not used) }
5389  */
5390 static const struct pci_device_id ice_pci_tbl[] = {
5391 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
5392 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
5393 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
5394 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE), 0 },
5395 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP), 0 },
5396 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
5397 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
5398 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
5399 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
5400 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
5401 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
5402 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
5403 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
5404 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
5405 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
5406 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
5407 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
5408 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
5409 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
5410 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
5411 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
5412 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
5413 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
5414 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
5415 	{ PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
5416 	/* required last entry */
5417 	{ 0, }
5418 };
5419 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5420 
5421 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5422 
5423 static const struct pci_error_handlers ice_pci_err_handler = {
5424 	.error_detected = ice_pci_err_detected,
5425 	.slot_reset = ice_pci_err_slot_reset,
5426 	.reset_prepare = ice_pci_err_reset_prepare,
5427 	.reset_done = ice_pci_err_reset_done,
5428 	.resume = ice_pci_err_resume
5429 };
5430 
5431 static struct pci_driver ice_driver = {
5432 	.name = KBUILD_MODNAME,
5433 	.id_table = ice_pci_tbl,
5434 	.probe = ice_probe,
5435 	.remove = ice_remove,
5436 #ifdef CONFIG_PM
5437 	.driver.pm = &ice_pm_ops,
5438 #endif /* CONFIG_PM */
5439 	.shutdown = ice_shutdown,
5440 	.sriov_configure = ice_sriov_configure,
5441 	.err_handler = &ice_pci_err_handler
5442 };
5443 
5444 /**
5445  * ice_module_init - Driver registration routine
5446  *
5447  * ice_module_init is the first routine called when the driver is
5448  * loaded. All it does is register with the PCI subsystem.
5449  */
5450 static int __init ice_module_init(void)
5451 {
5452 	int status;
5453 
5454 	pr_info("%s\n", ice_driver_string);
5455 	pr_info("%s\n", ice_copyright);
5456 
5457 	ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
5458 	if (!ice_wq) {
5459 		pr_err("Failed to create workqueue\n");
5460 		return -ENOMEM;
5461 	}
5462 
5463 	status = pci_register_driver(&ice_driver);
5464 	if (status) {
5465 		pr_err("failed to register PCI driver, err %d\n", status);
5466 		destroy_workqueue(ice_wq);
5467 	}
5468 
5469 	return status;
5470 }
5471 module_init(ice_module_init);
5472 
5473 /**
5474  * ice_module_exit - Driver exit cleanup routine
5475  *
5476  * ice_module_exit is called just before the driver is removed
5477  * from memory.
5478  */
5479 static void __exit ice_module_exit(void)
5480 {
5481 	pci_unregister_driver(&ice_driver);
5482 	destroy_workqueue(ice_wq);
5483 	pr_info("module unloaded\n");
5484 }
5485 module_exit(ice_module_exit);
5486 
5487 /**
5488  * ice_set_mac_address - NDO callback to set MAC address
5489  * @netdev: network interface device structure
5490  * @pi: pointer to an address structure
5491  *
5492  * Returns 0 on success, negative on failure
5493  */
5494 static int ice_set_mac_address(struct net_device *netdev, void *pi)
5495 {
5496 	struct ice_netdev_priv *np = netdev_priv(netdev);
5497 	struct ice_vsi *vsi = np->vsi;
5498 	struct ice_pf *pf = vsi->back;
5499 	struct ice_hw *hw = &pf->hw;
5500 	struct sockaddr *addr = pi;
5501 	u8 old_mac[ETH_ALEN];
5502 	u8 flags = 0;
5503 	u8 *mac;
5504 	int err;
5505 
5506 	mac = (u8 *)addr->sa_data;
5507 
5508 	if (!is_valid_ether_addr(mac))
5509 		return -EADDRNOTAVAIL;
5510 
5511 	if (ether_addr_equal(netdev->dev_addr, mac)) {
5512 		netdev_dbg(netdev, "already using mac %pM\n", mac);
5513 		return 0;
5514 	}
5515 
5516 	if (test_bit(ICE_DOWN, pf->state) ||
5517 	    ice_is_reset_in_progress(pf->state)) {
5518 		netdev_err(netdev, "can't set mac %pM. device not ready\n",
5519 			   mac);
5520 		return -EBUSY;
5521 	}
5522 
5523 	if (ice_chnl_dmac_fltr_cnt(pf)) {
5524 		netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5525 			   mac);
5526 		return -EAGAIN;
5527 	}
5528 
5529 	netif_addr_lock_bh(netdev);
5530 	ether_addr_copy(old_mac, netdev->dev_addr);
5531 	/* change the netdev's MAC address */
5532 	eth_hw_addr_set(netdev, mac);
5533 	netif_addr_unlock_bh(netdev);
5534 
5535 	/* Clean up old MAC filter. Not an error if old filter doesn't exist */
5536 	err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5537 	if (err && err != -ENOENT) {
5538 		err = -EADDRNOTAVAIL;
5539 		goto err_update_filters;
5540 	}
5541 
5542 	/* Add filter for new MAC. If filter exists, return success */
5543 	err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5544 	if (err == -EEXIST) {
5545 		/* Although this MAC filter is already present in hardware it's
5546 		 * possible in some cases (e.g. bonding) that dev_addr was
5547 		 * modified outside of the driver and needs to be restored back
5548 		 * to this value.
5549 		 */
5550 		netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5551 
5552 		return 0;
5553 	} else if (err) {
5554 		/* error if the new filter addition failed */
5555 		err = -EADDRNOTAVAIL;
5556 	}
5557 
5558 err_update_filters:
5559 	if (err) {
5560 		netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5561 			   mac);
5562 		netif_addr_lock_bh(netdev);
5563 		eth_hw_addr_set(netdev, old_mac);
5564 		netif_addr_unlock_bh(netdev);
5565 		return err;
5566 	}
5567 
5568 	netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5569 		   netdev->dev_addr);
5570 
5571 	/* write new MAC address to the firmware */
5572 	flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5573 	err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5574 	if (err) {
5575 		netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5576 			   mac, err);
5577 	}
5578 	return 0;
5579 }
5580 
5581 /**
5582  * ice_set_rx_mode - NDO callback to set the netdev filters
5583  * @netdev: network interface device structure
5584  */
5585 static void ice_set_rx_mode(struct net_device *netdev)
5586 {
5587 	struct ice_netdev_priv *np = netdev_priv(netdev);
5588 	struct ice_vsi *vsi = np->vsi;
5589 
5590 	if (!vsi)
5591 		return;
5592 
5593 	/* Set the flags to synchronize filters
5594 	 * ndo_set_rx_mode may be triggered even without a change in netdev
5595 	 * flags
5596 	 */
5597 	set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5598 	set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5599 	set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5600 
5601 	/* schedule our worker thread which will take care of
5602 	 * applying the new filter changes
5603 	 */
5604 	ice_service_task_schedule(vsi->back);
5605 }
5606 
5607 /**
5608  * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5609  * @netdev: network interface device structure
5610  * @queue_index: Queue ID
5611  * @maxrate: maximum bandwidth in Mbps
5612  */
5613 static int
5614 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5615 {
5616 	struct ice_netdev_priv *np = netdev_priv(netdev);
5617 	struct ice_vsi *vsi = np->vsi;
5618 	u16 q_handle;
5619 	int status;
5620 	u8 tc;
5621 
5622 	/* Validate maxrate requested is within permitted range */
5623 	if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5624 		netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5625 			   maxrate, queue_index);
5626 		return -EINVAL;
5627 	}
5628 
5629 	q_handle = vsi->tx_rings[queue_index]->q_handle;
5630 	tc = ice_dcb_get_tc(vsi, queue_index);
5631 
5632 	/* Set BW back to default, when user set maxrate to 0 */
5633 	if (!maxrate)
5634 		status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5635 					       q_handle, ICE_MAX_BW);
5636 	else
5637 		status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5638 					  q_handle, ICE_MAX_BW, maxrate * 1000);
5639 	if (status)
5640 		netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
5641 			   status);
5642 
5643 	return status;
5644 }
5645 
5646 /**
5647  * ice_fdb_add - add an entry to the hardware database
5648  * @ndm: the input from the stack
5649  * @tb: pointer to array of nladdr (unused)
5650  * @dev: the net device pointer
5651  * @addr: the MAC address entry being added
5652  * @vid: VLAN ID
5653  * @flags: instructions from stack about fdb operation
5654  * @extack: netlink extended ack
5655  */
5656 static int
5657 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5658 	    struct net_device *dev, const unsigned char *addr, u16 vid,
5659 	    u16 flags, struct netlink_ext_ack __always_unused *extack)
5660 {
5661 	int err;
5662 
5663 	if (vid) {
5664 		netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5665 		return -EINVAL;
5666 	}
5667 	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5668 		netdev_err(dev, "FDB only supports static addresses\n");
5669 		return -EINVAL;
5670 	}
5671 
5672 	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5673 		err = dev_uc_add_excl(dev, addr);
5674 	else if (is_multicast_ether_addr(addr))
5675 		err = dev_mc_add_excl(dev, addr);
5676 	else
5677 		err = -EINVAL;
5678 
5679 	/* Only return duplicate errors if NLM_F_EXCL is set */
5680 	if (err == -EEXIST && !(flags & NLM_F_EXCL))
5681 		err = 0;
5682 
5683 	return err;
5684 }
5685 
5686 /**
5687  * ice_fdb_del - delete an entry from the hardware database
5688  * @ndm: the input from the stack
5689  * @tb: pointer to array of nladdr (unused)
5690  * @dev: the net device pointer
5691  * @addr: the MAC address entry being added
5692  * @vid: VLAN ID
5693  * @extack: netlink extended ack
5694  */
5695 static int
5696 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5697 	    struct net_device *dev, const unsigned char *addr,
5698 	    __always_unused u16 vid, struct netlink_ext_ack *extack)
5699 {
5700 	int err;
5701 
5702 	if (ndm->ndm_state & NUD_PERMANENT) {
5703 		netdev_err(dev, "FDB only supports static addresses\n");
5704 		return -EINVAL;
5705 	}
5706 
5707 	if (is_unicast_ether_addr(addr))
5708 		err = dev_uc_del(dev, addr);
5709 	else if (is_multicast_ether_addr(addr))
5710 		err = dev_mc_del(dev, addr);
5711 	else
5712 		err = -EINVAL;
5713 
5714 	return err;
5715 }
5716 
5717 #define NETIF_VLAN_OFFLOAD_FEATURES	(NETIF_F_HW_VLAN_CTAG_RX | \
5718 					 NETIF_F_HW_VLAN_CTAG_TX | \
5719 					 NETIF_F_HW_VLAN_STAG_RX | \
5720 					 NETIF_F_HW_VLAN_STAG_TX)
5721 
5722 #define NETIF_VLAN_FILTERING_FEATURES	(NETIF_F_HW_VLAN_CTAG_FILTER | \
5723 					 NETIF_F_HW_VLAN_STAG_FILTER)
5724 
5725 /**
5726  * ice_fix_features - fix the netdev features flags based on device limitations
5727  * @netdev: ptr to the netdev that flags are being fixed on
5728  * @features: features that need to be checked and possibly fixed
5729  *
5730  * Make sure any fixups are made to features in this callback. This enables the
5731  * driver to not have to check unsupported configurations throughout the driver
5732  * because that's the responsiblity of this callback.
5733  *
5734  * Single VLAN Mode (SVM) Supported Features:
5735  *	NETIF_F_HW_VLAN_CTAG_FILTER
5736  *	NETIF_F_HW_VLAN_CTAG_RX
5737  *	NETIF_F_HW_VLAN_CTAG_TX
5738  *
5739  * Double VLAN Mode (DVM) Supported Features:
5740  *	NETIF_F_HW_VLAN_CTAG_FILTER
5741  *	NETIF_F_HW_VLAN_CTAG_RX
5742  *	NETIF_F_HW_VLAN_CTAG_TX
5743  *
5744  *	NETIF_F_HW_VLAN_STAG_FILTER
5745  *	NETIF_HW_VLAN_STAG_RX
5746  *	NETIF_HW_VLAN_STAG_TX
5747  *
5748  * Features that need fixing:
5749  *	Cannot simultaneously enable CTAG and STAG stripping and/or insertion.
5750  *	These are mutually exlusive as the VSI context cannot support multiple
5751  *	VLAN ethertypes simultaneously for stripping and/or insertion. If this
5752  *	is not done, then default to clearing the requested STAG offload
5753  *	settings.
5754  *
5755  *	All supported filtering has to be enabled or disabled together. For
5756  *	example, in DVM, CTAG and STAG filtering have to be enabled and disabled
5757  *	together. If this is not done, then default to VLAN filtering disabled.
5758  *	These are mutually exclusive as there is currently no way to
5759  *	enable/disable VLAN filtering based on VLAN ethertype when using VLAN
5760  *	prune rules.
5761  */
5762 static netdev_features_t
5763 ice_fix_features(struct net_device *netdev, netdev_features_t features)
5764 {
5765 	struct ice_netdev_priv *np = netdev_priv(netdev);
5766 	netdev_features_t supported_vlan_filtering;
5767 	netdev_features_t requested_vlan_filtering;
5768 	struct ice_vsi *vsi = np->vsi;
5769 
5770 	requested_vlan_filtering = features & NETIF_VLAN_FILTERING_FEATURES;
5771 
5772 	/* make sure supported_vlan_filtering works for both SVM and DVM */
5773 	supported_vlan_filtering = NETIF_F_HW_VLAN_CTAG_FILTER;
5774 	if (ice_is_dvm_ena(&vsi->back->hw))
5775 		supported_vlan_filtering |= NETIF_F_HW_VLAN_STAG_FILTER;
5776 
5777 	if (requested_vlan_filtering &&
5778 	    requested_vlan_filtering != supported_vlan_filtering) {
5779 		if (requested_vlan_filtering & NETIF_F_HW_VLAN_CTAG_FILTER) {
5780 			netdev_warn(netdev, "cannot support requested VLAN filtering settings, enabling all supported VLAN filtering settings\n");
5781 			features |= supported_vlan_filtering;
5782 		} else {
5783 			netdev_warn(netdev, "cannot support requested VLAN filtering settings, clearing all supported VLAN filtering settings\n");
5784 			features &= ~supported_vlan_filtering;
5785 		}
5786 	}
5787 
5788 	if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) &&
5789 	    (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) {
5790 		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");
5791 		features &= ~(NETIF_F_HW_VLAN_STAG_RX |
5792 			      NETIF_F_HW_VLAN_STAG_TX);
5793 	}
5794 
5795 	return features;
5796 }
5797 
5798 /**
5799  * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI
5800  * @vsi: PF's VSI
5801  * @features: features used to determine VLAN offload settings
5802  *
5803  * First, determine the vlan_ethertype based on the VLAN offload bits in
5804  * features. Then determine if stripping and insertion should be enabled or
5805  * disabled. Finally enable or disable VLAN stripping and insertion.
5806  */
5807 static int
5808 ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features)
5809 {
5810 	bool enable_stripping = true, enable_insertion = true;
5811 	struct ice_vsi_vlan_ops *vlan_ops;
5812 	int strip_err = 0, insert_err = 0;
5813 	u16 vlan_ethertype = 0;
5814 
5815 	vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
5816 
5817 	if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))
5818 		vlan_ethertype = ETH_P_8021AD;
5819 	else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX))
5820 		vlan_ethertype = ETH_P_8021Q;
5821 
5822 	if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX)))
5823 		enable_stripping = false;
5824 	if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX)))
5825 		enable_insertion = false;
5826 
5827 	if (enable_stripping)
5828 		strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype);
5829 	else
5830 		strip_err = vlan_ops->dis_stripping(vsi);
5831 
5832 	if (enable_insertion)
5833 		insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype);
5834 	else
5835 		insert_err = vlan_ops->dis_insertion(vsi);
5836 
5837 	if (strip_err || insert_err)
5838 		return -EIO;
5839 
5840 	return 0;
5841 }
5842 
5843 /**
5844  * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI
5845  * @vsi: PF's VSI
5846  * @features: features used to determine VLAN filtering settings
5847  *
5848  * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the
5849  * features.
5850  */
5851 static int
5852 ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features)
5853 {
5854 	struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
5855 	int err = 0;
5856 
5857 	/* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking
5858 	 * if either bit is set
5859 	 */
5860 	if (features &
5861 	    (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER))
5862 		err = vlan_ops->ena_rx_filtering(vsi);
5863 	else
5864 		err = vlan_ops->dis_rx_filtering(vsi);
5865 
5866 	return err;
5867 }
5868 
5869 /**
5870  * ice_set_vlan_features - set VLAN settings based on suggested feature set
5871  * @netdev: ptr to the netdev being adjusted
5872  * @features: the feature set that the stack is suggesting
5873  *
5874  * Only update VLAN settings if the requested_vlan_features are different than
5875  * the current_vlan_features.
5876  */
5877 static int
5878 ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)
5879 {
5880 	netdev_features_t current_vlan_features, requested_vlan_features;
5881 	struct ice_netdev_priv *np = netdev_priv(netdev);
5882 	struct ice_vsi *vsi = np->vsi;
5883 	int err;
5884 
5885 	current_vlan_features = netdev->features & NETIF_VLAN_OFFLOAD_FEATURES;
5886 	requested_vlan_features = features & NETIF_VLAN_OFFLOAD_FEATURES;
5887 	if (current_vlan_features ^ requested_vlan_features) {
5888 		err = ice_set_vlan_offload_features(vsi, features);
5889 		if (err)
5890 			return err;
5891 	}
5892 
5893 	current_vlan_features = netdev->features &
5894 		NETIF_VLAN_FILTERING_FEATURES;
5895 	requested_vlan_features = features & NETIF_VLAN_FILTERING_FEATURES;
5896 	if (current_vlan_features ^ requested_vlan_features) {
5897 		err = ice_set_vlan_filtering_features(vsi, features);
5898 		if (err)
5899 			return err;
5900 	}
5901 
5902 	return 0;
5903 }
5904 
5905 /**
5906  * ice_set_features - set the netdev feature flags
5907  * @netdev: ptr to the netdev being adjusted
5908  * @features: the feature set that the stack is suggesting
5909  */
5910 static int
5911 ice_set_features(struct net_device *netdev, netdev_features_t features)
5912 {
5913 	struct ice_netdev_priv *np = netdev_priv(netdev);
5914 	struct ice_vsi *vsi = np->vsi;
5915 	struct ice_pf *pf = vsi->back;
5916 	int ret = 0;
5917 
5918 	/* Don't set any netdev advanced features with device in Safe Mode */
5919 	if (ice_is_safe_mode(vsi->back)) {
5920 		dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5921 		return ret;
5922 	}
5923 
5924 	/* Do not change setting during reset */
5925 	if (ice_is_reset_in_progress(pf->state)) {
5926 		dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5927 		return -EBUSY;
5928 	}
5929 
5930 	/* Multiple features can be changed in one call so keep features in
5931 	 * separate if/else statements to guarantee each feature is checked
5932 	 */
5933 	if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5934 		ice_vsi_manage_rss_lut(vsi, true);
5935 	else if (!(features & NETIF_F_RXHASH) &&
5936 		 netdev->features & NETIF_F_RXHASH)
5937 		ice_vsi_manage_rss_lut(vsi, false);
5938 
5939 	ret = ice_set_vlan_features(netdev, features);
5940 	if (ret)
5941 		return ret;
5942 
5943 	if ((features & NETIF_F_NTUPLE) &&
5944 	    !(netdev->features & NETIF_F_NTUPLE)) {
5945 		ice_vsi_manage_fdir(vsi, true);
5946 		ice_init_arfs(vsi);
5947 	} else if (!(features & NETIF_F_NTUPLE) &&
5948 		 (netdev->features & NETIF_F_NTUPLE)) {
5949 		ice_vsi_manage_fdir(vsi, false);
5950 		ice_clear_arfs(vsi);
5951 	}
5952 
5953 	/* don't turn off hw_tc_offload when ADQ is already enabled */
5954 	if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) {
5955 		dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n");
5956 		return -EACCES;
5957 	}
5958 
5959 	if ((features & NETIF_F_HW_TC) &&
5960 	    !(netdev->features & NETIF_F_HW_TC))
5961 		set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
5962 	else
5963 		clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
5964 
5965 	return 0;
5966 }
5967 
5968 /**
5969  * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI
5970  * @vsi: VSI to setup VLAN properties for
5971  */
5972 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5973 {
5974 	int err;
5975 
5976 	err = ice_set_vlan_offload_features(vsi, vsi->netdev->features);
5977 	if (err)
5978 		return err;
5979 
5980 	err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features);
5981 	if (err)
5982 		return err;
5983 
5984 	return ice_vsi_add_vlan_zero(vsi);
5985 }
5986 
5987 /**
5988  * ice_vsi_cfg - Setup the VSI
5989  * @vsi: the VSI being configured
5990  *
5991  * Return 0 on success and negative value on error
5992  */
5993 int ice_vsi_cfg(struct ice_vsi *vsi)
5994 {
5995 	int err;
5996 
5997 	if (vsi->netdev) {
5998 		ice_set_rx_mode(vsi->netdev);
5999 
6000 		err = ice_vsi_vlan_setup(vsi);
6001 
6002 		if (err)
6003 			return err;
6004 	}
6005 	ice_vsi_cfg_dcb_rings(vsi);
6006 
6007 	err = ice_vsi_cfg_lan_txqs(vsi);
6008 	if (!err && ice_is_xdp_ena_vsi(vsi))
6009 		err = ice_vsi_cfg_xdp_txqs(vsi);
6010 	if (!err)
6011 		err = ice_vsi_cfg_rxqs(vsi);
6012 
6013 	return err;
6014 }
6015 
6016 /* THEORY OF MODERATION:
6017  * The ice driver hardware works differently than the hardware that DIMLIB was
6018  * originally made for. ice hardware doesn't have packet count limits that
6019  * can trigger an interrupt, but it *does* have interrupt rate limit support,
6020  * which is hard-coded to a limit of 250,000 ints/second.
6021  * If not using dynamic moderation, the INTRL value can be modified
6022  * by ethtool rx-usecs-high.
6023  */
6024 struct ice_dim {
6025 	/* the throttle rate for interrupts, basically worst case delay before
6026 	 * an initial interrupt fires, value is stored in microseconds.
6027 	 */
6028 	u16 itr;
6029 };
6030 
6031 /* Make a different profile for Rx that doesn't allow quite so aggressive
6032  * moderation at the high end (it maxes out at 126us or about 8k interrupts a
6033  * second.
6034  */
6035 static const struct ice_dim rx_profile[] = {
6036 	{2},    /* 500,000 ints/s, capped at 250K by INTRL */
6037 	{8},    /* 125,000 ints/s */
6038 	{16},   /*  62,500 ints/s */
6039 	{62},   /*  16,129 ints/s */
6040 	{126}   /*   7,936 ints/s */
6041 };
6042 
6043 /* The transmit profile, which has the same sorts of values
6044  * as the previous struct
6045  */
6046 static const struct ice_dim tx_profile[] = {
6047 	{2},    /* 500,000 ints/s, capped at 250K by INTRL */
6048 	{8},    /* 125,000 ints/s */
6049 	{40},   /*  16,125 ints/s */
6050 	{128},  /*   7,812 ints/s */
6051 	{256}   /*   3,906 ints/s */
6052 };
6053 
6054 static void ice_tx_dim_work(struct work_struct *work)
6055 {
6056 	struct ice_ring_container *rc;
6057 	struct dim *dim;
6058 	u16 itr;
6059 
6060 	dim = container_of(work, struct dim, work);
6061 	rc = (struct ice_ring_container *)dim->priv;
6062 
6063 	WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
6064 
6065 	/* look up the values in our local table */
6066 	itr = tx_profile[dim->profile_ix].itr;
6067 
6068 	ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
6069 	ice_write_itr(rc, itr);
6070 
6071 	dim->state = DIM_START_MEASURE;
6072 }
6073 
6074 static void ice_rx_dim_work(struct work_struct *work)
6075 {
6076 	struct ice_ring_container *rc;
6077 	struct dim *dim;
6078 	u16 itr;
6079 
6080 	dim = container_of(work, struct dim, work);
6081 	rc = (struct ice_ring_container *)dim->priv;
6082 
6083 	WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
6084 
6085 	/* look up the values in our local table */
6086 	itr = rx_profile[dim->profile_ix].itr;
6087 
6088 	ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
6089 	ice_write_itr(rc, itr);
6090 
6091 	dim->state = DIM_START_MEASURE;
6092 }
6093 
6094 #define ICE_DIM_DEFAULT_PROFILE_IX 1
6095 
6096 /**
6097  * ice_init_moderation - set up interrupt moderation
6098  * @q_vector: the vector containing rings to be configured
6099  *
6100  * Set up interrupt moderation registers, with the intent to do the right thing
6101  * when called from reset or from probe, and whether or not dynamic moderation
6102  * is enabled or not. Take special care to write all the registers in both
6103  * dynamic moderation mode or not in order to make sure hardware is in a known
6104  * state.
6105  */
6106 static void ice_init_moderation(struct ice_q_vector *q_vector)
6107 {
6108 	struct ice_ring_container *rc;
6109 	bool tx_dynamic, rx_dynamic;
6110 
6111 	rc = &q_vector->tx;
6112 	INIT_WORK(&rc->dim.work, ice_tx_dim_work);
6113 	rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6114 	rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6115 	rc->dim.priv = rc;
6116 	tx_dynamic = ITR_IS_DYNAMIC(rc);
6117 
6118 	/* set the initial TX ITR to match the above */
6119 	ice_write_itr(rc, tx_dynamic ?
6120 		      tx_profile[rc->dim.profile_ix].itr : rc->itr_setting);
6121 
6122 	rc = &q_vector->rx;
6123 	INIT_WORK(&rc->dim.work, ice_rx_dim_work);
6124 	rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6125 	rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6126 	rc->dim.priv = rc;
6127 	rx_dynamic = ITR_IS_DYNAMIC(rc);
6128 
6129 	/* set the initial RX ITR to match the above */
6130 	ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
6131 				       rc->itr_setting);
6132 
6133 	ice_set_q_vector_intrl(q_vector);
6134 }
6135 
6136 /**
6137  * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
6138  * @vsi: the VSI being configured
6139  */
6140 static void ice_napi_enable_all(struct ice_vsi *vsi)
6141 {
6142 	int q_idx;
6143 
6144 	if (!vsi->netdev)
6145 		return;
6146 
6147 	ice_for_each_q_vector(vsi, q_idx) {
6148 		struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6149 
6150 		ice_init_moderation(q_vector);
6151 
6152 		if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6153 			napi_enable(&q_vector->napi);
6154 	}
6155 }
6156 
6157 /**
6158  * ice_up_complete - Finish the last steps of bringing up a connection
6159  * @vsi: The VSI being configured
6160  *
6161  * Return 0 on success and negative value on error
6162  */
6163 static int ice_up_complete(struct ice_vsi *vsi)
6164 {
6165 	struct ice_pf *pf = vsi->back;
6166 	int err;
6167 
6168 	ice_vsi_cfg_msix(vsi);
6169 
6170 	/* Enable only Rx rings, Tx rings were enabled by the FW when the
6171 	 * Tx queue group list was configured and the context bits were
6172 	 * programmed using ice_vsi_cfg_txqs
6173 	 */
6174 	err = ice_vsi_start_all_rx_rings(vsi);
6175 	if (err)
6176 		return err;
6177 
6178 	clear_bit(ICE_VSI_DOWN, vsi->state);
6179 	ice_napi_enable_all(vsi);
6180 	ice_vsi_ena_irq(vsi);
6181 
6182 	if (vsi->port_info &&
6183 	    (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
6184 	    vsi->netdev) {
6185 		ice_print_link_msg(vsi, true);
6186 		netif_tx_start_all_queues(vsi->netdev);
6187 		netif_carrier_on(vsi->netdev);
6188 		if (!ice_is_e810(&pf->hw))
6189 			ice_ptp_link_change(pf, pf->hw.pf_id, true);
6190 	}
6191 
6192 	/* Perform an initial read of the statistics registers now to
6193 	 * set the baseline so counters are ready when interface is up
6194 	 */
6195 	ice_update_eth_stats(vsi);
6196 	ice_service_task_schedule(pf);
6197 
6198 	return 0;
6199 }
6200 
6201 /**
6202  * ice_up - Bring the connection back up after being down
6203  * @vsi: VSI being configured
6204  */
6205 int ice_up(struct ice_vsi *vsi)
6206 {
6207 	int err;
6208 
6209 	err = ice_vsi_cfg(vsi);
6210 	if (!err)
6211 		err = ice_up_complete(vsi);
6212 
6213 	return err;
6214 }
6215 
6216 /**
6217  * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
6218  * @syncp: pointer to u64_stats_sync
6219  * @stats: stats that pkts and bytes count will be taken from
6220  * @pkts: packets stats counter
6221  * @bytes: bytes stats counter
6222  *
6223  * This function fetches stats from the ring considering the atomic operations
6224  * that needs to be performed to read u64 values in 32 bit machine.
6225  */
6226 void
6227 ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp,
6228 			     struct ice_q_stats stats, u64 *pkts, u64 *bytes)
6229 {
6230 	unsigned int start;
6231 
6232 	do {
6233 		start = u64_stats_fetch_begin_irq(syncp);
6234 		*pkts = stats.pkts;
6235 		*bytes = stats.bytes;
6236 	} while (u64_stats_fetch_retry_irq(syncp, start));
6237 }
6238 
6239 /**
6240  * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
6241  * @vsi: the VSI to be updated
6242  * @vsi_stats: the stats struct to be updated
6243  * @rings: rings to work on
6244  * @count: number of rings
6245  */
6246 static void
6247 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
6248 			     struct rtnl_link_stats64 *vsi_stats,
6249 			     struct ice_tx_ring **rings, u16 count)
6250 {
6251 	u16 i;
6252 
6253 	for (i = 0; i < count; i++) {
6254 		struct ice_tx_ring *ring;
6255 		u64 pkts = 0, bytes = 0;
6256 
6257 		ring = READ_ONCE(rings[i]);
6258 		if (!ring)
6259 			continue;
6260 		ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
6261 		vsi_stats->tx_packets += pkts;
6262 		vsi_stats->tx_bytes += bytes;
6263 		vsi->tx_restart += ring->tx_stats.restart_q;
6264 		vsi->tx_busy += ring->tx_stats.tx_busy;
6265 		vsi->tx_linearize += ring->tx_stats.tx_linearize;
6266 	}
6267 }
6268 
6269 /**
6270  * ice_update_vsi_ring_stats - Update VSI stats counters
6271  * @vsi: the VSI to be updated
6272  */
6273 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
6274 {
6275 	struct rtnl_link_stats64 *vsi_stats;
6276 	u64 pkts, bytes;
6277 	int i;
6278 
6279 	vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
6280 	if (!vsi_stats)
6281 		return;
6282 
6283 	/* reset non-netdev (extended) stats */
6284 	vsi->tx_restart = 0;
6285 	vsi->tx_busy = 0;
6286 	vsi->tx_linearize = 0;
6287 	vsi->rx_buf_failed = 0;
6288 	vsi->rx_page_failed = 0;
6289 
6290 	rcu_read_lock();
6291 
6292 	/* update Tx rings counters */
6293 	ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6294 				     vsi->num_txq);
6295 
6296 	/* update Rx rings counters */
6297 	ice_for_each_rxq(vsi, i) {
6298 		struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6299 
6300 		ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
6301 		vsi_stats->rx_packets += pkts;
6302 		vsi_stats->rx_bytes += bytes;
6303 		vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
6304 		vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
6305 	}
6306 
6307 	/* update XDP Tx rings counters */
6308 	if (ice_is_xdp_ena_vsi(vsi))
6309 		ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6310 					     vsi->num_xdp_txq);
6311 
6312 	rcu_read_unlock();
6313 
6314 	vsi->net_stats.tx_packets = vsi_stats->tx_packets;
6315 	vsi->net_stats.tx_bytes = vsi_stats->tx_bytes;
6316 	vsi->net_stats.rx_packets = vsi_stats->rx_packets;
6317 	vsi->net_stats.rx_bytes = vsi_stats->rx_bytes;
6318 
6319 	kfree(vsi_stats);
6320 }
6321 
6322 /**
6323  * ice_update_vsi_stats - Update VSI stats counters
6324  * @vsi: the VSI to be updated
6325  */
6326 void ice_update_vsi_stats(struct ice_vsi *vsi)
6327 {
6328 	struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6329 	struct ice_eth_stats *cur_es = &vsi->eth_stats;
6330 	struct ice_pf *pf = vsi->back;
6331 
6332 	if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6333 	    test_bit(ICE_CFG_BUSY, pf->state))
6334 		return;
6335 
6336 	/* get stats as recorded by Tx/Rx rings */
6337 	ice_update_vsi_ring_stats(vsi);
6338 
6339 	/* get VSI stats as recorded by the hardware */
6340 	ice_update_eth_stats(vsi);
6341 
6342 	cur_ns->tx_errors = cur_es->tx_errors;
6343 	cur_ns->rx_dropped = cur_es->rx_discards;
6344 	cur_ns->tx_dropped = cur_es->tx_discards;
6345 	cur_ns->multicast = cur_es->rx_multicast;
6346 
6347 	/* update some more netdev stats if this is main VSI */
6348 	if (vsi->type == ICE_VSI_PF) {
6349 		cur_ns->rx_crc_errors = pf->stats.crc_errors;
6350 		cur_ns->rx_errors = pf->stats.crc_errors +
6351 				    pf->stats.illegal_bytes +
6352 				    pf->stats.rx_len_errors +
6353 				    pf->stats.rx_undersize +
6354 				    pf->hw_csum_rx_error +
6355 				    pf->stats.rx_jabber +
6356 				    pf->stats.rx_fragments +
6357 				    pf->stats.rx_oversize;
6358 		cur_ns->rx_length_errors = pf->stats.rx_len_errors;
6359 		/* record drops from the port level */
6360 		cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6361 	}
6362 }
6363 
6364 /**
6365  * ice_update_pf_stats - Update PF port stats counters
6366  * @pf: PF whose stats needs to be updated
6367  */
6368 void ice_update_pf_stats(struct ice_pf *pf)
6369 {
6370 	struct ice_hw_port_stats *prev_ps, *cur_ps;
6371 	struct ice_hw *hw = &pf->hw;
6372 	u16 fd_ctr_base;
6373 	u8 port;
6374 
6375 	port = hw->port_info->lport;
6376 	prev_ps = &pf->stats_prev;
6377 	cur_ps = &pf->stats;
6378 
6379 	ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6380 			  &prev_ps->eth.rx_bytes,
6381 			  &cur_ps->eth.rx_bytes);
6382 
6383 	ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6384 			  &prev_ps->eth.rx_unicast,
6385 			  &cur_ps->eth.rx_unicast);
6386 
6387 	ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6388 			  &prev_ps->eth.rx_multicast,
6389 			  &cur_ps->eth.rx_multicast);
6390 
6391 	ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6392 			  &prev_ps->eth.rx_broadcast,
6393 			  &cur_ps->eth.rx_broadcast);
6394 
6395 	ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6396 			  &prev_ps->eth.rx_discards,
6397 			  &cur_ps->eth.rx_discards);
6398 
6399 	ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6400 			  &prev_ps->eth.tx_bytes,
6401 			  &cur_ps->eth.tx_bytes);
6402 
6403 	ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6404 			  &prev_ps->eth.tx_unicast,
6405 			  &cur_ps->eth.tx_unicast);
6406 
6407 	ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6408 			  &prev_ps->eth.tx_multicast,
6409 			  &cur_ps->eth.tx_multicast);
6410 
6411 	ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6412 			  &prev_ps->eth.tx_broadcast,
6413 			  &cur_ps->eth.tx_broadcast);
6414 
6415 	ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6416 			  &prev_ps->tx_dropped_link_down,
6417 			  &cur_ps->tx_dropped_link_down);
6418 
6419 	ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6420 			  &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6421 
6422 	ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6423 			  &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6424 
6425 	ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6426 			  &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6427 
6428 	ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6429 			  &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6430 
6431 	ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6432 			  &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6433 
6434 	ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6435 			  &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6436 
6437 	ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6438 			  &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6439 
6440 	ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6441 			  &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6442 
6443 	ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6444 			  &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6445 
6446 	ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6447 			  &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6448 
6449 	ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6450 			  &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6451 
6452 	ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6453 			  &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6454 
6455 	ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6456 			  &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6457 
6458 	ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6459 			  &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6460 
6461 	fd_ctr_base = hw->fd_ctr_base;
6462 
6463 	ice_stat_update40(hw,
6464 			  GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6465 			  pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6466 			  &cur_ps->fd_sb_match);
6467 	ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6468 			  &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6469 
6470 	ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6471 			  &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6472 
6473 	ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6474 			  &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6475 
6476 	ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6477 			  &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6478 
6479 	ice_update_dcb_stats(pf);
6480 
6481 	ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6482 			  &prev_ps->crc_errors, &cur_ps->crc_errors);
6483 
6484 	ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6485 			  &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6486 
6487 	ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6488 			  &prev_ps->mac_local_faults,
6489 			  &cur_ps->mac_local_faults);
6490 
6491 	ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6492 			  &prev_ps->mac_remote_faults,
6493 			  &cur_ps->mac_remote_faults);
6494 
6495 	ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
6496 			  &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
6497 
6498 	ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6499 			  &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6500 
6501 	ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6502 			  &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6503 
6504 	ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6505 			  &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6506 
6507 	ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6508 			  &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6509 
6510 	cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6511 
6512 	pf->stat_prev_loaded = true;
6513 }
6514 
6515 /**
6516  * ice_get_stats64 - get statistics for network device structure
6517  * @netdev: network interface device structure
6518  * @stats: main device statistics structure
6519  */
6520 static
6521 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
6522 {
6523 	struct ice_netdev_priv *np = netdev_priv(netdev);
6524 	struct rtnl_link_stats64 *vsi_stats;
6525 	struct ice_vsi *vsi = np->vsi;
6526 
6527 	vsi_stats = &vsi->net_stats;
6528 
6529 	if (!vsi->num_txq || !vsi->num_rxq)
6530 		return;
6531 
6532 	/* netdev packet/byte stats come from ring counter. These are obtained
6533 	 * by summing up ring counters (done by ice_update_vsi_ring_stats).
6534 	 * But, only call the update routine and read the registers if VSI is
6535 	 * not down.
6536 	 */
6537 	if (!test_bit(ICE_VSI_DOWN, vsi->state))
6538 		ice_update_vsi_ring_stats(vsi);
6539 	stats->tx_packets = vsi_stats->tx_packets;
6540 	stats->tx_bytes = vsi_stats->tx_bytes;
6541 	stats->rx_packets = vsi_stats->rx_packets;
6542 	stats->rx_bytes = vsi_stats->rx_bytes;
6543 
6544 	/* The rest of the stats can be read from the hardware but instead we
6545 	 * just return values that the watchdog task has already obtained from
6546 	 * the hardware.
6547 	 */
6548 	stats->multicast = vsi_stats->multicast;
6549 	stats->tx_errors = vsi_stats->tx_errors;
6550 	stats->tx_dropped = vsi_stats->tx_dropped;
6551 	stats->rx_errors = vsi_stats->rx_errors;
6552 	stats->rx_dropped = vsi_stats->rx_dropped;
6553 	stats->rx_crc_errors = vsi_stats->rx_crc_errors;
6554 	stats->rx_length_errors = vsi_stats->rx_length_errors;
6555 }
6556 
6557 /**
6558  * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
6559  * @vsi: VSI having NAPI disabled
6560  */
6561 static void ice_napi_disable_all(struct ice_vsi *vsi)
6562 {
6563 	int q_idx;
6564 
6565 	if (!vsi->netdev)
6566 		return;
6567 
6568 	ice_for_each_q_vector(vsi, q_idx) {
6569 		struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6570 
6571 		if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6572 			napi_disable(&q_vector->napi);
6573 
6574 		cancel_work_sync(&q_vector->tx.dim.work);
6575 		cancel_work_sync(&q_vector->rx.dim.work);
6576 	}
6577 }
6578 
6579 /**
6580  * ice_down - Shutdown the connection
6581  * @vsi: The VSI being stopped
6582  *
6583  * Caller of this function is expected to set the vsi->state ICE_DOWN bit
6584  */
6585 int ice_down(struct ice_vsi *vsi)
6586 {
6587 	int i, tx_err, rx_err, link_err = 0, vlan_err = 0;
6588 
6589 	WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
6590 
6591 	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6592 		vlan_err = ice_vsi_del_vlan_zero(vsi);
6593 		if (!ice_is_e810(&vsi->back->hw))
6594 			ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
6595 		netif_carrier_off(vsi->netdev);
6596 		netif_tx_disable(vsi->netdev);
6597 	} else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
6598 		ice_eswitch_stop_all_tx_queues(vsi->back);
6599 	}
6600 
6601 	ice_vsi_dis_irq(vsi);
6602 
6603 	tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
6604 	if (tx_err)
6605 		netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
6606 			   vsi->vsi_num, tx_err);
6607 	if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
6608 		tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
6609 		if (tx_err)
6610 			netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
6611 				   vsi->vsi_num, tx_err);
6612 	}
6613 
6614 	rx_err = ice_vsi_stop_all_rx_rings(vsi);
6615 	if (rx_err)
6616 		netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
6617 			   vsi->vsi_num, rx_err);
6618 
6619 	ice_napi_disable_all(vsi);
6620 
6621 	if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
6622 		link_err = ice_force_phys_link_state(vsi, false);
6623 		if (link_err)
6624 			netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
6625 				   vsi->vsi_num, link_err);
6626 	}
6627 
6628 	ice_for_each_txq(vsi, i)
6629 		ice_clean_tx_ring(vsi->tx_rings[i]);
6630 
6631 	ice_for_each_rxq(vsi, i)
6632 		ice_clean_rx_ring(vsi->rx_rings[i]);
6633 
6634 	if (tx_err || rx_err || link_err || vlan_err) {
6635 		netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
6636 			   vsi->vsi_num, vsi->vsw->sw_id);
6637 		return -EIO;
6638 	}
6639 
6640 	return 0;
6641 }
6642 
6643 /**
6644  * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
6645  * @vsi: VSI having resources allocated
6646  *
6647  * Return 0 on success, negative on failure
6648  */
6649 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
6650 {
6651 	int i, err = 0;
6652 
6653 	if (!vsi->num_txq) {
6654 		dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
6655 			vsi->vsi_num);
6656 		return -EINVAL;
6657 	}
6658 
6659 	ice_for_each_txq(vsi, i) {
6660 		struct ice_tx_ring *ring = vsi->tx_rings[i];
6661 
6662 		if (!ring)
6663 			return -EINVAL;
6664 
6665 		if (vsi->netdev)
6666 			ring->netdev = vsi->netdev;
6667 		err = ice_setup_tx_ring(ring);
6668 		if (err)
6669 			break;
6670 	}
6671 
6672 	return err;
6673 }
6674 
6675 /**
6676  * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
6677  * @vsi: VSI having resources allocated
6678  *
6679  * Return 0 on success, negative on failure
6680  */
6681 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
6682 {
6683 	int i, err = 0;
6684 
6685 	if (!vsi->num_rxq) {
6686 		dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
6687 			vsi->vsi_num);
6688 		return -EINVAL;
6689 	}
6690 
6691 	ice_for_each_rxq(vsi, i) {
6692 		struct ice_rx_ring *ring = vsi->rx_rings[i];
6693 
6694 		if (!ring)
6695 			return -EINVAL;
6696 
6697 		if (vsi->netdev)
6698 			ring->netdev = vsi->netdev;
6699 		err = ice_setup_rx_ring(ring);
6700 		if (err)
6701 			break;
6702 	}
6703 
6704 	return err;
6705 }
6706 
6707 /**
6708  * ice_vsi_open_ctrl - open control VSI for use
6709  * @vsi: the VSI to open
6710  *
6711  * Initialization of the Control VSI
6712  *
6713  * Returns 0 on success, negative value on error
6714  */
6715 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
6716 {
6717 	char int_name[ICE_INT_NAME_STR_LEN];
6718 	struct ice_pf *pf = vsi->back;
6719 	struct device *dev;
6720 	int err;
6721 
6722 	dev = ice_pf_to_dev(pf);
6723 	/* allocate descriptors */
6724 	err = ice_vsi_setup_tx_rings(vsi);
6725 	if (err)
6726 		goto err_setup_tx;
6727 
6728 	err = ice_vsi_setup_rx_rings(vsi);
6729 	if (err)
6730 		goto err_setup_rx;
6731 
6732 	err = ice_vsi_cfg(vsi);
6733 	if (err)
6734 		goto err_setup_rx;
6735 
6736 	snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
6737 		 dev_driver_string(dev), dev_name(dev));
6738 	err = ice_vsi_req_irq_msix(vsi, int_name);
6739 	if (err)
6740 		goto err_setup_rx;
6741 
6742 	ice_vsi_cfg_msix(vsi);
6743 
6744 	err = ice_vsi_start_all_rx_rings(vsi);
6745 	if (err)
6746 		goto err_up_complete;
6747 
6748 	clear_bit(ICE_VSI_DOWN, vsi->state);
6749 	ice_vsi_ena_irq(vsi);
6750 
6751 	return 0;
6752 
6753 err_up_complete:
6754 	ice_down(vsi);
6755 err_setup_rx:
6756 	ice_vsi_free_rx_rings(vsi);
6757 err_setup_tx:
6758 	ice_vsi_free_tx_rings(vsi);
6759 
6760 	return err;
6761 }
6762 
6763 /**
6764  * ice_vsi_open - Called when a network interface is made active
6765  * @vsi: the VSI to open
6766  *
6767  * Initialization of the VSI
6768  *
6769  * Returns 0 on success, negative value on error
6770  */
6771 int ice_vsi_open(struct ice_vsi *vsi)
6772 {
6773 	char int_name[ICE_INT_NAME_STR_LEN];
6774 	struct ice_pf *pf = vsi->back;
6775 	int err;
6776 
6777 	/* allocate descriptors */
6778 	err = ice_vsi_setup_tx_rings(vsi);
6779 	if (err)
6780 		goto err_setup_tx;
6781 
6782 	err = ice_vsi_setup_rx_rings(vsi);
6783 	if (err)
6784 		goto err_setup_rx;
6785 
6786 	err = ice_vsi_cfg(vsi);
6787 	if (err)
6788 		goto err_setup_rx;
6789 
6790 	snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6791 		 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
6792 	err = ice_vsi_req_irq_msix(vsi, int_name);
6793 	if (err)
6794 		goto err_setup_rx;
6795 
6796 	if (vsi->type == ICE_VSI_PF) {
6797 		/* Notify the stack of the actual queue counts. */
6798 		err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
6799 		if (err)
6800 			goto err_set_qs;
6801 
6802 		err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
6803 		if (err)
6804 			goto err_set_qs;
6805 	}
6806 
6807 	err = ice_up_complete(vsi);
6808 	if (err)
6809 		goto err_up_complete;
6810 
6811 	return 0;
6812 
6813 err_up_complete:
6814 	ice_down(vsi);
6815 err_set_qs:
6816 	ice_vsi_free_irq(vsi);
6817 err_setup_rx:
6818 	ice_vsi_free_rx_rings(vsi);
6819 err_setup_tx:
6820 	ice_vsi_free_tx_rings(vsi);
6821 
6822 	return err;
6823 }
6824 
6825 /**
6826  * ice_vsi_release_all - Delete all VSIs
6827  * @pf: PF from which all VSIs are being removed
6828  */
6829 static void ice_vsi_release_all(struct ice_pf *pf)
6830 {
6831 	int err, i;
6832 
6833 	if (!pf->vsi)
6834 		return;
6835 
6836 	ice_for_each_vsi(pf, i) {
6837 		if (!pf->vsi[i])
6838 			continue;
6839 
6840 		if (pf->vsi[i]->type == ICE_VSI_CHNL)
6841 			continue;
6842 
6843 		err = ice_vsi_release(pf->vsi[i]);
6844 		if (err)
6845 			dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
6846 				i, err, pf->vsi[i]->vsi_num);
6847 	}
6848 }
6849 
6850 /**
6851  * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
6852  * @pf: pointer to the PF instance
6853  * @type: VSI type to rebuild
6854  *
6855  * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6856  */
6857 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6858 {
6859 	struct device *dev = ice_pf_to_dev(pf);
6860 	int i, err;
6861 
6862 	ice_for_each_vsi(pf, i) {
6863 		struct ice_vsi *vsi = pf->vsi[i];
6864 
6865 		if (!vsi || vsi->type != type)
6866 			continue;
6867 
6868 		/* rebuild the VSI */
6869 		err = ice_vsi_rebuild(vsi, true);
6870 		if (err) {
6871 			dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6872 				err, vsi->idx, ice_vsi_type_str(type));
6873 			return err;
6874 		}
6875 
6876 		/* replay filters for the VSI */
6877 		err = ice_replay_vsi(&pf->hw, vsi->idx);
6878 		if (err) {
6879 			dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
6880 				err, vsi->idx, ice_vsi_type_str(type));
6881 			return err;
6882 		}
6883 
6884 		/* Re-map HW VSI number, using VSI handle that has been
6885 		 * previously validated in ice_replay_vsi() call above
6886 		 */
6887 		vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6888 
6889 		/* enable the VSI */
6890 		err = ice_ena_vsi(vsi, false);
6891 		if (err) {
6892 			dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6893 				err, vsi->idx, ice_vsi_type_str(type));
6894 			return err;
6895 		}
6896 
6897 		dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6898 			 ice_vsi_type_str(type));
6899 	}
6900 
6901 	return 0;
6902 }
6903 
6904 /**
6905  * ice_update_pf_netdev_link - Update PF netdev link status
6906  * @pf: pointer to the PF instance
6907  */
6908 static void ice_update_pf_netdev_link(struct ice_pf *pf)
6909 {
6910 	bool link_up;
6911 	int i;
6912 
6913 	ice_for_each_vsi(pf, i) {
6914 		struct ice_vsi *vsi = pf->vsi[i];
6915 
6916 		if (!vsi || vsi->type != ICE_VSI_PF)
6917 			return;
6918 
6919 		ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6920 		if (link_up) {
6921 			netif_carrier_on(pf->vsi[i]->netdev);
6922 			netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6923 		} else {
6924 			netif_carrier_off(pf->vsi[i]->netdev);
6925 			netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6926 		}
6927 	}
6928 }
6929 
6930 /**
6931  * ice_rebuild - rebuild after reset
6932  * @pf: PF to rebuild
6933  * @reset_type: type of reset
6934  *
6935  * Do not rebuild VF VSI in this flow because that is already handled via
6936  * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6937  * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6938  * to reset/rebuild all the VF VSI twice.
6939  */
6940 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6941 {
6942 	struct device *dev = ice_pf_to_dev(pf);
6943 	struct ice_hw *hw = &pf->hw;
6944 	bool dvm;
6945 	int err;
6946 
6947 	if (test_bit(ICE_DOWN, pf->state))
6948 		goto clear_recovery;
6949 
6950 	dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
6951 
6952 #define ICE_EMP_RESET_SLEEP_MS 5000
6953 	if (reset_type == ICE_RESET_EMPR) {
6954 		/* If an EMP reset has occurred, any previously pending flash
6955 		 * update will have completed. We no longer know whether or
6956 		 * not the NVM update EMP reset is restricted.
6957 		 */
6958 		pf->fw_emp_reset_disabled = false;
6959 
6960 		msleep(ICE_EMP_RESET_SLEEP_MS);
6961 	}
6962 
6963 	err = ice_init_all_ctrlq(hw);
6964 	if (err) {
6965 		dev_err(dev, "control queues init failed %d\n", err);
6966 		goto err_init_ctrlq;
6967 	}
6968 
6969 	/* if DDP was previously loaded successfully */
6970 	if (!ice_is_safe_mode(pf)) {
6971 		/* reload the SW DB of filter tables */
6972 		if (reset_type == ICE_RESET_PFR)
6973 			ice_fill_blk_tbls(hw);
6974 		else
6975 			/* Reload DDP Package after CORER/GLOBR reset */
6976 			ice_load_pkg(NULL, pf);
6977 	}
6978 
6979 	err = ice_clear_pf_cfg(hw);
6980 	if (err) {
6981 		dev_err(dev, "clear PF configuration failed %d\n", err);
6982 		goto err_init_ctrlq;
6983 	}
6984 
6985 	if (pf->first_sw->dflt_vsi_ena)
6986 		dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
6987 	/* clear the default VSI configuration if it exists */
6988 	pf->first_sw->dflt_vsi = NULL;
6989 	pf->first_sw->dflt_vsi_ena = false;
6990 
6991 	ice_clear_pxe_mode(hw);
6992 
6993 	err = ice_init_nvm(hw);
6994 	if (err) {
6995 		dev_err(dev, "ice_init_nvm failed %d\n", err);
6996 		goto err_init_ctrlq;
6997 	}
6998 
6999 	err = ice_get_caps(hw);
7000 	if (err) {
7001 		dev_err(dev, "ice_get_caps failed %d\n", err);
7002 		goto err_init_ctrlq;
7003 	}
7004 
7005 	err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
7006 	if (err) {
7007 		dev_err(dev, "set_mac_cfg failed %d\n", err);
7008 		goto err_init_ctrlq;
7009 	}
7010 
7011 	dvm = ice_is_dvm_ena(hw);
7012 
7013 	err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
7014 	if (err)
7015 		goto err_init_ctrlq;
7016 
7017 	err = ice_sched_init_port(hw->port_info);
7018 	if (err)
7019 		goto err_sched_init_port;
7020 
7021 	/* start misc vector */
7022 	err = ice_req_irq_msix_misc(pf);
7023 	if (err) {
7024 		dev_err(dev, "misc vector setup failed: %d\n", err);
7025 		goto err_sched_init_port;
7026 	}
7027 
7028 	if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7029 		wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
7030 		if (!rd32(hw, PFQF_FD_SIZE)) {
7031 			u16 unused, guar, b_effort;
7032 
7033 			guar = hw->func_caps.fd_fltr_guar;
7034 			b_effort = hw->func_caps.fd_fltr_best_effort;
7035 
7036 			/* force guaranteed filter pool for PF */
7037 			ice_alloc_fd_guar_item(hw, &unused, guar);
7038 			/* force shared filter pool for PF */
7039 			ice_alloc_fd_shrd_item(hw, &unused, b_effort);
7040 		}
7041 	}
7042 
7043 	if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
7044 		ice_dcb_rebuild(pf);
7045 
7046 	/* If the PF previously had enabled PTP, PTP init needs to happen before
7047 	 * the VSI rebuild. If not, this causes the PTP link status events to
7048 	 * fail.
7049 	 */
7050 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7051 		ice_ptp_reset(pf);
7052 
7053 	if (ice_is_feature_supported(pf, ICE_F_GNSS))
7054 		ice_gnss_init(pf);
7055 
7056 	/* rebuild PF VSI */
7057 	err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
7058 	if (err) {
7059 		dev_err(dev, "PF VSI rebuild failed: %d\n", err);
7060 		goto err_vsi_rebuild;
7061 	}
7062 
7063 	/* configure PTP timestamping after VSI rebuild */
7064 	if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7065 		ice_ptp_cfg_timestamp(pf, false);
7066 
7067 	err = ice_vsi_rebuild_by_type(pf, ICE_VSI_SWITCHDEV_CTRL);
7068 	if (err) {
7069 		dev_err(dev, "Switchdev CTRL VSI rebuild failed: %d\n", err);
7070 		goto err_vsi_rebuild;
7071 	}
7072 
7073 	if (reset_type == ICE_RESET_PFR) {
7074 		err = ice_rebuild_channels(pf);
7075 		if (err) {
7076 			dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
7077 				err);
7078 			goto err_vsi_rebuild;
7079 		}
7080 	}
7081 
7082 	/* If Flow Director is active */
7083 	if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7084 		err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
7085 		if (err) {
7086 			dev_err(dev, "control VSI rebuild failed: %d\n", err);
7087 			goto err_vsi_rebuild;
7088 		}
7089 
7090 		/* replay HW Flow Director recipes */
7091 		if (hw->fdir_prof)
7092 			ice_fdir_replay_flows(hw);
7093 
7094 		/* replay Flow Director filters */
7095 		ice_fdir_replay_fltrs(pf);
7096 
7097 		ice_rebuild_arfs(pf);
7098 	}
7099 
7100 	ice_update_pf_netdev_link(pf);
7101 
7102 	/* tell the firmware we are up */
7103 	err = ice_send_version(pf);
7104 	if (err) {
7105 		dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
7106 			err);
7107 		goto err_vsi_rebuild;
7108 	}
7109 
7110 	ice_replay_post(hw);
7111 
7112 	/* if we get here, reset flow is successful */
7113 	clear_bit(ICE_RESET_FAILED, pf->state);
7114 
7115 	ice_plug_aux_dev(pf);
7116 	return;
7117 
7118 err_vsi_rebuild:
7119 err_sched_init_port:
7120 	ice_sched_cleanup_all(hw);
7121 err_init_ctrlq:
7122 	ice_shutdown_all_ctrlq(hw);
7123 	set_bit(ICE_RESET_FAILED, pf->state);
7124 clear_recovery:
7125 	/* set this bit in PF state to control service task scheduling */
7126 	set_bit(ICE_NEEDS_RESTART, pf->state);
7127 	dev_err(dev, "Rebuild failed, unload and reload driver\n");
7128 }
7129 
7130 /**
7131  * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
7132  * @vsi: Pointer to VSI structure
7133  */
7134 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
7135 {
7136 	if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
7137 		return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
7138 	else
7139 		return ICE_RXBUF_3072;
7140 }
7141 
7142 /**
7143  * ice_change_mtu - NDO callback to change the MTU
7144  * @netdev: network interface device structure
7145  * @new_mtu: new value for maximum frame size
7146  *
7147  * Returns 0 on success, negative on failure
7148  */
7149 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
7150 {
7151 	struct ice_netdev_priv *np = netdev_priv(netdev);
7152 	struct ice_vsi *vsi = np->vsi;
7153 	struct ice_pf *pf = vsi->back;
7154 	u8 count = 0;
7155 	int err = 0;
7156 
7157 	if (new_mtu == (int)netdev->mtu) {
7158 		netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
7159 		return 0;
7160 	}
7161 
7162 	if (ice_is_xdp_ena_vsi(vsi)) {
7163 		int frame_size = ice_max_xdp_frame_size(vsi);
7164 
7165 		if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
7166 			netdev_err(netdev, "max MTU for XDP usage is %d\n",
7167 				   frame_size - ICE_ETH_PKT_HDR_PAD);
7168 			return -EINVAL;
7169 		}
7170 	}
7171 
7172 	/* if a reset is in progress, wait for some time for it to complete */
7173 	do {
7174 		if (ice_is_reset_in_progress(pf->state)) {
7175 			count++;
7176 			usleep_range(1000, 2000);
7177 		} else {
7178 			break;
7179 		}
7180 
7181 	} while (count < 100);
7182 
7183 	if (count == 100) {
7184 		netdev_err(netdev, "can't change MTU. Device is busy\n");
7185 		return -EBUSY;
7186 	}
7187 
7188 	netdev->mtu = (unsigned int)new_mtu;
7189 
7190 	/* if VSI is up, bring it down and then back up */
7191 	if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
7192 		err = ice_down(vsi);
7193 		if (err) {
7194 			netdev_err(netdev, "change MTU if_down err %d\n", err);
7195 			return err;
7196 		}
7197 
7198 		err = ice_up(vsi);
7199 		if (err) {
7200 			netdev_err(netdev, "change MTU if_up err %d\n", err);
7201 			return err;
7202 		}
7203 	}
7204 
7205 	netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
7206 	set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
7207 
7208 	return err;
7209 }
7210 
7211 /**
7212  * ice_eth_ioctl - Access the hwtstamp interface
7213  * @netdev: network interface device structure
7214  * @ifr: interface request data
7215  * @cmd: ioctl command
7216  */
7217 static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7218 {
7219 	struct ice_netdev_priv *np = netdev_priv(netdev);
7220 	struct ice_pf *pf = np->vsi->back;
7221 
7222 	switch (cmd) {
7223 	case SIOCGHWTSTAMP:
7224 		return ice_ptp_get_ts_config(pf, ifr);
7225 	case SIOCSHWTSTAMP:
7226 		return ice_ptp_set_ts_config(pf, ifr);
7227 	default:
7228 		return -EOPNOTSUPP;
7229 	}
7230 }
7231 
7232 /**
7233  * ice_aq_str - convert AQ err code to a string
7234  * @aq_err: the AQ error code to convert
7235  */
7236 const char *ice_aq_str(enum ice_aq_err aq_err)
7237 {
7238 	switch (aq_err) {
7239 	case ICE_AQ_RC_OK:
7240 		return "OK";
7241 	case ICE_AQ_RC_EPERM:
7242 		return "ICE_AQ_RC_EPERM";
7243 	case ICE_AQ_RC_ENOENT:
7244 		return "ICE_AQ_RC_ENOENT";
7245 	case ICE_AQ_RC_ENOMEM:
7246 		return "ICE_AQ_RC_ENOMEM";
7247 	case ICE_AQ_RC_EBUSY:
7248 		return "ICE_AQ_RC_EBUSY";
7249 	case ICE_AQ_RC_EEXIST:
7250 		return "ICE_AQ_RC_EEXIST";
7251 	case ICE_AQ_RC_EINVAL:
7252 		return "ICE_AQ_RC_EINVAL";
7253 	case ICE_AQ_RC_ENOSPC:
7254 		return "ICE_AQ_RC_ENOSPC";
7255 	case ICE_AQ_RC_ENOSYS:
7256 		return "ICE_AQ_RC_ENOSYS";
7257 	case ICE_AQ_RC_EMODE:
7258 		return "ICE_AQ_RC_EMODE";
7259 	case ICE_AQ_RC_ENOSEC:
7260 		return "ICE_AQ_RC_ENOSEC";
7261 	case ICE_AQ_RC_EBADSIG:
7262 		return "ICE_AQ_RC_EBADSIG";
7263 	case ICE_AQ_RC_ESVN:
7264 		return "ICE_AQ_RC_ESVN";
7265 	case ICE_AQ_RC_EBADMAN:
7266 		return "ICE_AQ_RC_EBADMAN";
7267 	case ICE_AQ_RC_EBADBUF:
7268 		return "ICE_AQ_RC_EBADBUF";
7269 	}
7270 
7271 	return "ICE_AQ_RC_UNKNOWN";
7272 }
7273 
7274 /**
7275  * ice_set_rss_lut - Set RSS LUT
7276  * @vsi: Pointer to VSI structure
7277  * @lut: Lookup table
7278  * @lut_size: Lookup table size
7279  *
7280  * Returns 0 on success, negative on failure
7281  */
7282 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7283 {
7284 	struct ice_aq_get_set_rss_lut_params params = {};
7285 	struct ice_hw *hw = &vsi->back->hw;
7286 	int status;
7287 
7288 	if (!lut)
7289 		return -EINVAL;
7290 
7291 	params.vsi_handle = vsi->idx;
7292 	params.lut_size = lut_size;
7293 	params.lut_type = vsi->rss_lut_type;
7294 	params.lut = lut;
7295 
7296 	status = ice_aq_set_rss_lut(hw, &params);
7297 	if (status)
7298 		dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
7299 			status, ice_aq_str(hw->adminq.sq_last_status));
7300 
7301 	return status;
7302 }
7303 
7304 /**
7305  * ice_set_rss_key - Set RSS key
7306  * @vsi: Pointer to the VSI structure
7307  * @seed: RSS hash seed
7308  *
7309  * Returns 0 on success, negative on failure
7310  */
7311 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7312 {
7313 	struct ice_hw *hw = &vsi->back->hw;
7314 	int status;
7315 
7316 	if (!seed)
7317 		return -EINVAL;
7318 
7319 	status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7320 	if (status)
7321 		dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7322 			status, ice_aq_str(hw->adminq.sq_last_status));
7323 
7324 	return status;
7325 }
7326 
7327 /**
7328  * ice_get_rss_lut - Get RSS LUT
7329  * @vsi: Pointer to VSI structure
7330  * @lut: Buffer to store the lookup table entries
7331  * @lut_size: Size of buffer to store the lookup table entries
7332  *
7333  * Returns 0 on success, negative on failure
7334  */
7335 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7336 {
7337 	struct ice_aq_get_set_rss_lut_params params = {};
7338 	struct ice_hw *hw = &vsi->back->hw;
7339 	int status;
7340 
7341 	if (!lut)
7342 		return -EINVAL;
7343 
7344 	params.vsi_handle = vsi->idx;
7345 	params.lut_size = lut_size;
7346 	params.lut_type = vsi->rss_lut_type;
7347 	params.lut = lut;
7348 
7349 	status = ice_aq_get_rss_lut(hw, &params);
7350 	if (status)
7351 		dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7352 			status, ice_aq_str(hw->adminq.sq_last_status));
7353 
7354 	return status;
7355 }
7356 
7357 /**
7358  * ice_get_rss_key - Get RSS key
7359  * @vsi: Pointer to VSI structure
7360  * @seed: Buffer to store the key in
7361  *
7362  * Returns 0 on success, negative on failure
7363  */
7364 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7365 {
7366 	struct ice_hw *hw = &vsi->back->hw;
7367 	int status;
7368 
7369 	if (!seed)
7370 		return -EINVAL;
7371 
7372 	status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7373 	if (status)
7374 		dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7375 			status, ice_aq_str(hw->adminq.sq_last_status));
7376 
7377 	return status;
7378 }
7379 
7380 /**
7381  * ice_bridge_getlink - Get the hardware bridge mode
7382  * @skb: skb buff
7383  * @pid: process ID
7384  * @seq: RTNL message seq
7385  * @dev: the netdev being configured
7386  * @filter_mask: filter mask passed in
7387  * @nlflags: netlink flags passed in
7388  *
7389  * Return the bridge mode (VEB/VEPA)
7390  */
7391 static int
7392 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7393 		   struct net_device *dev, u32 filter_mask, int nlflags)
7394 {
7395 	struct ice_netdev_priv *np = netdev_priv(dev);
7396 	struct ice_vsi *vsi = np->vsi;
7397 	struct ice_pf *pf = vsi->back;
7398 	u16 bmode;
7399 
7400 	bmode = pf->first_sw->bridge_mode;
7401 
7402 	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7403 				       filter_mask, NULL);
7404 }
7405 
7406 /**
7407  * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7408  * @vsi: Pointer to VSI structure
7409  * @bmode: Hardware bridge mode (VEB/VEPA)
7410  *
7411  * Returns 0 on success, negative on failure
7412  */
7413 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7414 {
7415 	struct ice_aqc_vsi_props *vsi_props;
7416 	struct ice_hw *hw = &vsi->back->hw;
7417 	struct ice_vsi_ctx *ctxt;
7418 	int ret;
7419 
7420 	vsi_props = &vsi->info;
7421 
7422 	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7423 	if (!ctxt)
7424 		return -ENOMEM;
7425 
7426 	ctxt->info = vsi->info;
7427 
7428 	if (bmode == BRIDGE_MODE_VEB)
7429 		/* change from VEPA to VEB mode */
7430 		ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7431 	else
7432 		/* change from VEB to VEPA mode */
7433 		ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7434 	ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7435 
7436 	ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7437 	if (ret) {
7438 		dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7439 			bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7440 		goto out;
7441 	}
7442 	/* Update sw flags for book keeping */
7443 	vsi_props->sw_flags = ctxt->info.sw_flags;
7444 
7445 out:
7446 	kfree(ctxt);
7447 	return ret;
7448 }
7449 
7450 /**
7451  * ice_bridge_setlink - Set the hardware bridge mode
7452  * @dev: the netdev being configured
7453  * @nlh: RTNL message
7454  * @flags: bridge setlink flags
7455  * @extack: netlink extended ack
7456  *
7457  * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7458  * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7459  * not already set for all VSIs connected to this switch. And also update the
7460  * unicast switch filter rules for the corresponding switch of the netdev.
7461  */
7462 static int
7463 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
7464 		   u16 __always_unused flags,
7465 		   struct netlink_ext_ack __always_unused *extack)
7466 {
7467 	struct ice_netdev_priv *np = netdev_priv(dev);
7468 	struct ice_pf *pf = np->vsi->back;
7469 	struct nlattr *attr, *br_spec;
7470 	struct ice_hw *hw = &pf->hw;
7471 	struct ice_sw *pf_sw;
7472 	int rem, v, err = 0;
7473 
7474 	pf_sw = pf->first_sw;
7475 	/* find the attribute in the netlink message */
7476 	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7477 
7478 	nla_for_each_nested(attr, br_spec, rem) {
7479 		__u16 mode;
7480 
7481 		if (nla_type(attr) != IFLA_BRIDGE_MODE)
7482 			continue;
7483 		mode = nla_get_u16(attr);
7484 		if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
7485 			return -EINVAL;
7486 		/* Continue  if bridge mode is not being flipped */
7487 		if (mode == pf_sw->bridge_mode)
7488 			continue;
7489 		/* Iterates through the PF VSI list and update the loopback
7490 		 * mode of the VSI
7491 		 */
7492 		ice_for_each_vsi(pf, v) {
7493 			if (!pf->vsi[v])
7494 				continue;
7495 			err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
7496 			if (err)
7497 				return err;
7498 		}
7499 
7500 		hw->evb_veb = (mode == BRIDGE_MODE_VEB);
7501 		/* Update the unicast switch filter rules for the corresponding
7502 		 * switch of the netdev
7503 		 */
7504 		err = ice_update_sw_rule_bridge_mode(hw);
7505 		if (err) {
7506 			netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
7507 				   mode, err,
7508 				   ice_aq_str(hw->adminq.sq_last_status));
7509 			/* revert hw->evb_veb */
7510 			hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
7511 			return err;
7512 		}
7513 
7514 		pf_sw->bridge_mode = mode;
7515 	}
7516 
7517 	return 0;
7518 }
7519 
7520 /**
7521  * ice_tx_timeout - Respond to a Tx Hang
7522  * @netdev: network interface device structure
7523  * @txqueue: Tx queue
7524  */
7525 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
7526 {
7527 	struct ice_netdev_priv *np = netdev_priv(netdev);
7528 	struct ice_tx_ring *tx_ring = NULL;
7529 	struct ice_vsi *vsi = np->vsi;
7530 	struct ice_pf *pf = vsi->back;
7531 	u32 i;
7532 
7533 	pf->tx_timeout_count++;
7534 
7535 	/* Check if PFC is enabled for the TC to which the queue belongs
7536 	 * to. If yes then Tx timeout is not caused by a hung queue, no
7537 	 * need to reset and rebuild
7538 	 */
7539 	if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
7540 		dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
7541 			 txqueue);
7542 		return;
7543 	}
7544 
7545 	/* now that we have an index, find the tx_ring struct */
7546 	ice_for_each_txq(vsi, i)
7547 		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
7548 			if (txqueue == vsi->tx_rings[i]->q_index) {
7549 				tx_ring = vsi->tx_rings[i];
7550 				break;
7551 			}
7552 
7553 	/* Reset recovery level if enough time has elapsed after last timeout.
7554 	 * Also ensure no new reset action happens before next timeout period.
7555 	 */
7556 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
7557 		pf->tx_timeout_recovery_level = 1;
7558 	else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
7559 				       netdev->watchdog_timeo)))
7560 		return;
7561 
7562 	if (tx_ring) {
7563 		struct ice_hw *hw = &pf->hw;
7564 		u32 head, val = 0;
7565 
7566 		head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
7567 			QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
7568 		/* Read interrupt register */
7569 		val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
7570 
7571 		netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
7572 			    vsi->vsi_num, txqueue, tx_ring->next_to_clean,
7573 			    head, tx_ring->next_to_use, val);
7574 	}
7575 
7576 	pf->tx_timeout_last_recovery = jiffies;
7577 	netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
7578 		    pf->tx_timeout_recovery_level, txqueue);
7579 
7580 	switch (pf->tx_timeout_recovery_level) {
7581 	case 1:
7582 		set_bit(ICE_PFR_REQ, pf->state);
7583 		break;
7584 	case 2:
7585 		set_bit(ICE_CORER_REQ, pf->state);
7586 		break;
7587 	case 3:
7588 		set_bit(ICE_GLOBR_REQ, pf->state);
7589 		break;
7590 	default:
7591 		netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
7592 		set_bit(ICE_DOWN, pf->state);
7593 		set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
7594 		set_bit(ICE_SERVICE_DIS, pf->state);
7595 		break;
7596 	}
7597 
7598 	ice_service_task_schedule(pf);
7599 	pf->tx_timeout_recovery_level++;
7600 }
7601 
7602 /**
7603  * ice_setup_tc_cls_flower - flower classifier offloads
7604  * @np: net device to configure
7605  * @filter_dev: device on which filter is added
7606  * @cls_flower: offload data
7607  */
7608 static int
7609 ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
7610 			struct net_device *filter_dev,
7611 			struct flow_cls_offload *cls_flower)
7612 {
7613 	struct ice_vsi *vsi = np->vsi;
7614 
7615 	if (cls_flower->common.chain_index)
7616 		return -EOPNOTSUPP;
7617 
7618 	switch (cls_flower->command) {
7619 	case FLOW_CLS_REPLACE:
7620 		return ice_add_cls_flower(filter_dev, vsi, cls_flower);
7621 	case FLOW_CLS_DESTROY:
7622 		return ice_del_cls_flower(vsi, cls_flower);
7623 	default:
7624 		return -EINVAL;
7625 	}
7626 }
7627 
7628 /**
7629  * ice_setup_tc_block_cb - callback handler registered for TC block
7630  * @type: TC SETUP type
7631  * @type_data: TC flower offload data that contains user input
7632  * @cb_priv: netdev private data
7633  */
7634 static int
7635 ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
7636 {
7637 	struct ice_netdev_priv *np = cb_priv;
7638 
7639 	switch (type) {
7640 	case TC_SETUP_CLSFLOWER:
7641 		return ice_setup_tc_cls_flower(np, np->vsi->netdev,
7642 					       type_data);
7643 	default:
7644 		return -EOPNOTSUPP;
7645 	}
7646 }
7647 
7648 /**
7649  * ice_validate_mqprio_qopt - Validate TCF input parameters
7650  * @vsi: Pointer to VSI
7651  * @mqprio_qopt: input parameters for mqprio queue configuration
7652  *
7653  * This function validates MQPRIO params, such as qcount (power of 2 wherever
7654  * needed), and make sure user doesn't specify qcount and BW rate limit
7655  * for TCs, which are more than "num_tc"
7656  */
7657 static int
7658 ice_validate_mqprio_qopt(struct ice_vsi *vsi,
7659 			 struct tc_mqprio_qopt_offload *mqprio_qopt)
7660 {
7661 	u64 sum_max_rate = 0, sum_min_rate = 0;
7662 	int non_power_of_2_qcount = 0;
7663 	struct ice_pf *pf = vsi->back;
7664 	int max_rss_q_cnt = 0;
7665 	struct device *dev;
7666 	int i, speed;
7667 	u8 num_tc;
7668 
7669 	if (vsi->type != ICE_VSI_PF)
7670 		return -EINVAL;
7671 
7672 	if (mqprio_qopt->qopt.offset[0] != 0 ||
7673 	    mqprio_qopt->qopt.num_tc < 1 ||
7674 	    mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
7675 		return -EINVAL;
7676 
7677 	dev = ice_pf_to_dev(pf);
7678 	vsi->ch_rss_size = 0;
7679 	num_tc = mqprio_qopt->qopt.num_tc;
7680 
7681 	for (i = 0; num_tc; i++) {
7682 		int qcount = mqprio_qopt->qopt.count[i];
7683 		u64 max_rate, min_rate, rem;
7684 
7685 		if (!qcount)
7686 			return -EINVAL;
7687 
7688 		if (is_power_of_2(qcount)) {
7689 			if (non_power_of_2_qcount &&
7690 			    qcount > non_power_of_2_qcount) {
7691 				dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
7692 					qcount, non_power_of_2_qcount);
7693 				return -EINVAL;
7694 			}
7695 			if (qcount > max_rss_q_cnt)
7696 				max_rss_q_cnt = qcount;
7697 		} else {
7698 			if (non_power_of_2_qcount &&
7699 			    qcount != non_power_of_2_qcount) {
7700 				dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
7701 					qcount, non_power_of_2_qcount);
7702 				return -EINVAL;
7703 			}
7704 			if (qcount < max_rss_q_cnt) {
7705 				dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
7706 					qcount, max_rss_q_cnt);
7707 				return -EINVAL;
7708 			}
7709 			max_rss_q_cnt = qcount;
7710 			non_power_of_2_qcount = qcount;
7711 		}
7712 
7713 		/* TC command takes input in K/N/Gbps or K/M/Gbit etc but
7714 		 * converts the bandwidth rate limit into Bytes/s when
7715 		 * passing it down to the driver. So convert input bandwidth
7716 		 * from Bytes/s to Kbps
7717 		 */
7718 		max_rate = mqprio_qopt->max_rate[i];
7719 		max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
7720 		sum_max_rate += max_rate;
7721 
7722 		/* min_rate is minimum guaranteed rate and it can't be zero */
7723 		min_rate = mqprio_qopt->min_rate[i];
7724 		min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
7725 		sum_min_rate += min_rate;
7726 
7727 		if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
7728 			dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
7729 				min_rate, ICE_MIN_BW_LIMIT);
7730 			return -EINVAL;
7731 		}
7732 
7733 		iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
7734 		if (rem) {
7735 			dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
7736 				i, ICE_MIN_BW_LIMIT);
7737 			return -EINVAL;
7738 		}
7739 
7740 		iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
7741 		if (rem) {
7742 			dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
7743 				i, ICE_MIN_BW_LIMIT);
7744 			return -EINVAL;
7745 		}
7746 
7747 		/* min_rate can't be more than max_rate, except when max_rate
7748 		 * is zero (implies max_rate sought is max line rate). In such
7749 		 * a case min_rate can be more than max.
7750 		 */
7751 		if (max_rate && min_rate > max_rate) {
7752 			dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
7753 				min_rate, max_rate);
7754 			return -EINVAL;
7755 		}
7756 
7757 		if (i >= mqprio_qopt->qopt.num_tc - 1)
7758 			break;
7759 		if (mqprio_qopt->qopt.offset[i + 1] !=
7760 		    (mqprio_qopt->qopt.offset[i] + qcount))
7761 			return -EINVAL;
7762 	}
7763 	if (vsi->num_rxq <
7764 	    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7765 		return -EINVAL;
7766 	if (vsi->num_txq <
7767 	    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7768 		return -EINVAL;
7769 
7770 	speed = ice_get_link_speed_kbps(vsi);
7771 	if (sum_max_rate && sum_max_rate > (u64)speed) {
7772 		dev_err(dev, "Invalid max Tx rate(%llu) Kbps > speed(%u) Kbps specified\n",
7773 			sum_max_rate, speed);
7774 		return -EINVAL;
7775 	}
7776 	if (sum_min_rate && sum_min_rate > (u64)speed) {
7777 		dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
7778 			sum_min_rate, speed);
7779 		return -EINVAL;
7780 	}
7781 
7782 	/* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
7783 	vsi->ch_rss_size = max_rss_q_cnt;
7784 
7785 	return 0;
7786 }
7787 
7788 /**
7789  * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
7790  * @pf: ptr to PF device
7791  * @vsi: ptr to VSI
7792  */
7793 static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
7794 {
7795 	struct device *dev = ice_pf_to_dev(pf);
7796 	bool added = false;
7797 	struct ice_hw *hw;
7798 	int flow;
7799 
7800 	if (!(vsi->num_gfltr || vsi->num_bfltr))
7801 		return -EINVAL;
7802 
7803 	hw = &pf->hw;
7804 	for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
7805 		struct ice_fd_hw_prof *prof;
7806 		int tun, status;
7807 		u64 entry_h;
7808 
7809 		if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
7810 		      hw->fdir_prof[flow]->cnt))
7811 			continue;
7812 
7813 		for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
7814 			enum ice_flow_priority prio;
7815 			u64 prof_id;
7816 
7817 			/* add this VSI to FDir profile for this flow */
7818 			prio = ICE_FLOW_PRIO_NORMAL;
7819 			prof = hw->fdir_prof[flow];
7820 			prof_id = flow + tun * ICE_FLTR_PTYPE_MAX;
7821 			status = ice_flow_add_entry(hw, ICE_BLK_FD, prof_id,
7822 						    prof->vsi_h[0], vsi->idx,
7823 						    prio, prof->fdir_seg[tun],
7824 						    &entry_h);
7825 			if (status) {
7826 				dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
7827 					vsi->idx, flow);
7828 				continue;
7829 			}
7830 
7831 			prof->entry_h[prof->cnt][tun] = entry_h;
7832 		}
7833 
7834 		/* store VSI for filter replay and delete */
7835 		prof->vsi_h[prof->cnt] = vsi->idx;
7836 		prof->cnt++;
7837 
7838 		added = true;
7839 		dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
7840 			flow);
7841 	}
7842 
7843 	if (!added)
7844 		dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
7845 
7846 	return 0;
7847 }
7848 
7849 /**
7850  * ice_add_channel - add a channel by adding VSI
7851  * @pf: ptr to PF device
7852  * @sw_id: underlying HW switching element ID
7853  * @ch: ptr to channel structure
7854  *
7855  * Add a channel (VSI) using add_vsi and queue_map
7856  */
7857 static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
7858 {
7859 	struct device *dev = ice_pf_to_dev(pf);
7860 	struct ice_vsi *vsi;
7861 
7862 	if (ch->type != ICE_VSI_CHNL) {
7863 		dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
7864 		return -EINVAL;
7865 	}
7866 
7867 	vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
7868 	if (!vsi || vsi->type != ICE_VSI_CHNL) {
7869 		dev_err(dev, "create chnl VSI failure\n");
7870 		return -EINVAL;
7871 	}
7872 
7873 	ice_add_vsi_to_fdir(pf, vsi);
7874 
7875 	ch->sw_id = sw_id;
7876 	ch->vsi_num = vsi->vsi_num;
7877 	ch->info.mapping_flags = vsi->info.mapping_flags;
7878 	ch->ch_vsi = vsi;
7879 	/* set the back pointer of channel for newly created VSI */
7880 	vsi->ch = ch;
7881 
7882 	memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
7883 	       sizeof(vsi->info.q_mapping));
7884 	memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
7885 	       sizeof(vsi->info.tc_mapping));
7886 
7887 	return 0;
7888 }
7889 
7890 /**
7891  * ice_chnl_cfg_res
7892  * @vsi: the VSI being setup
7893  * @ch: ptr to channel structure
7894  *
7895  * Configure channel specific resources such as rings, vector.
7896  */
7897 static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
7898 {
7899 	int i;
7900 
7901 	for (i = 0; i < ch->num_txq; i++) {
7902 		struct ice_q_vector *tx_q_vector, *rx_q_vector;
7903 		struct ice_ring_container *rc;
7904 		struct ice_tx_ring *tx_ring;
7905 		struct ice_rx_ring *rx_ring;
7906 
7907 		tx_ring = vsi->tx_rings[ch->base_q + i];
7908 		rx_ring = vsi->rx_rings[ch->base_q + i];
7909 		if (!tx_ring || !rx_ring)
7910 			continue;
7911 
7912 		/* setup ring being channel enabled */
7913 		tx_ring->ch = ch;
7914 		rx_ring->ch = ch;
7915 
7916 		/* following code block sets up vector specific attributes */
7917 		tx_q_vector = tx_ring->q_vector;
7918 		rx_q_vector = rx_ring->q_vector;
7919 		if (!tx_q_vector && !rx_q_vector)
7920 			continue;
7921 
7922 		if (tx_q_vector) {
7923 			tx_q_vector->ch = ch;
7924 			/* setup Tx and Rx ITR setting if DIM is off */
7925 			rc = &tx_q_vector->tx;
7926 			if (!ITR_IS_DYNAMIC(rc))
7927 				ice_write_itr(rc, rc->itr_setting);
7928 		}
7929 		if (rx_q_vector) {
7930 			rx_q_vector->ch = ch;
7931 			/* setup Tx and Rx ITR setting if DIM is off */
7932 			rc = &rx_q_vector->rx;
7933 			if (!ITR_IS_DYNAMIC(rc))
7934 				ice_write_itr(rc, rc->itr_setting);
7935 		}
7936 	}
7937 
7938 	/* it is safe to assume that, if channel has non-zero num_t[r]xq, then
7939 	 * GLINT_ITR register would have written to perform in-context
7940 	 * update, hence perform flush
7941 	 */
7942 	if (ch->num_txq || ch->num_rxq)
7943 		ice_flush(&vsi->back->hw);
7944 }
7945 
7946 /**
7947  * ice_cfg_chnl_all_res - configure channel resources
7948  * @vsi: pte to main_vsi
7949  * @ch: ptr to channel structure
7950  *
7951  * This function configures channel specific resources such as flow-director
7952  * counter index, and other resources such as queues, vectors, ITR settings
7953  */
7954 static void
7955 ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
7956 {
7957 	/* configure channel (aka ADQ) resources such as queues, vectors,
7958 	 * ITR settings for channel specific vectors and anything else
7959 	 */
7960 	ice_chnl_cfg_res(vsi, ch);
7961 }
7962 
7963 /**
7964  * ice_setup_hw_channel - setup new channel
7965  * @pf: ptr to PF device
7966  * @vsi: the VSI being setup
7967  * @ch: ptr to channel structure
7968  * @sw_id: underlying HW switching element ID
7969  * @type: type of channel to be created (VMDq2/VF)
7970  *
7971  * Setup new channel (VSI) based on specified type (VMDq2/VF)
7972  * and configures Tx rings accordingly
7973  */
7974 static int
7975 ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
7976 		     struct ice_channel *ch, u16 sw_id, u8 type)
7977 {
7978 	struct device *dev = ice_pf_to_dev(pf);
7979 	int ret;
7980 
7981 	ch->base_q = vsi->next_base_q;
7982 	ch->type = type;
7983 
7984 	ret = ice_add_channel(pf, sw_id, ch);
7985 	if (ret) {
7986 		dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
7987 		return ret;
7988 	}
7989 
7990 	/* configure/setup ADQ specific resources */
7991 	ice_cfg_chnl_all_res(vsi, ch);
7992 
7993 	/* make sure to update the next_base_q so that subsequent channel's
7994 	 * (aka ADQ) VSI queue map is correct
7995 	 */
7996 	vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
7997 	dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
7998 		ch->num_rxq);
7999 
8000 	return 0;
8001 }
8002 
8003 /**
8004  * ice_setup_channel - setup new channel using uplink element
8005  * @pf: ptr to PF device
8006  * @vsi: the VSI being setup
8007  * @ch: ptr to channel structure
8008  *
8009  * Setup new channel (VSI) based on specified type (VMDq2/VF)
8010  * and uplink switching element
8011  */
8012 static bool
8013 ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8014 		  struct ice_channel *ch)
8015 {
8016 	struct device *dev = ice_pf_to_dev(pf);
8017 	u16 sw_id;
8018 	int ret;
8019 
8020 	if (vsi->type != ICE_VSI_PF) {
8021 		dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
8022 		return false;
8023 	}
8024 
8025 	sw_id = pf->first_sw->sw_id;
8026 
8027 	/* create channel (VSI) */
8028 	ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
8029 	if (ret) {
8030 		dev_err(dev, "failed to setup hw_channel\n");
8031 		return false;
8032 	}
8033 	dev_dbg(dev, "successfully created channel()\n");
8034 
8035 	return ch->ch_vsi ? true : false;
8036 }
8037 
8038 /**
8039  * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
8040  * @vsi: VSI to be configured
8041  * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
8042  * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
8043  */
8044 static int
8045 ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
8046 {
8047 	int err;
8048 
8049 	err = ice_set_min_bw_limit(vsi, min_tx_rate);
8050 	if (err)
8051 		return err;
8052 
8053 	return ice_set_max_bw_limit(vsi, max_tx_rate);
8054 }
8055 
8056 /**
8057  * ice_create_q_channel - function to create channel
8058  * @vsi: VSI to be configured
8059  * @ch: ptr to channel (it contains channel specific params)
8060  *
8061  * This function creates channel (VSI) using num_queues specified by user,
8062  * reconfigs RSS if needed.
8063  */
8064 static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
8065 {
8066 	struct ice_pf *pf = vsi->back;
8067 	struct device *dev;
8068 
8069 	if (!ch)
8070 		return -EINVAL;
8071 
8072 	dev = ice_pf_to_dev(pf);
8073 	if (!ch->num_txq || !ch->num_rxq) {
8074 		dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
8075 		return -EINVAL;
8076 	}
8077 
8078 	if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
8079 		dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
8080 			vsi->cnt_q_avail, ch->num_txq);
8081 		return -EINVAL;
8082 	}
8083 
8084 	if (!ice_setup_channel(pf, vsi, ch)) {
8085 		dev_info(dev, "Failed to setup channel\n");
8086 		return -EINVAL;
8087 	}
8088 	/* configure BW rate limit */
8089 	if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
8090 		int ret;
8091 
8092 		ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
8093 				       ch->min_tx_rate);
8094 		if (ret)
8095 			dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
8096 				ch->max_tx_rate, ch->ch_vsi->vsi_num);
8097 		else
8098 			dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
8099 				ch->max_tx_rate, ch->ch_vsi->vsi_num);
8100 	}
8101 
8102 	vsi->cnt_q_avail -= ch->num_txq;
8103 
8104 	return 0;
8105 }
8106 
8107 /**
8108  * ice_rem_all_chnl_fltrs - removes all channel filters
8109  * @pf: ptr to PF, TC-flower based filter are tracked at PF level
8110  *
8111  * Remove all advanced switch filters only if they are channel specific
8112  * tc-flower based filter
8113  */
8114 static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
8115 {
8116 	struct ice_tc_flower_fltr *fltr;
8117 	struct hlist_node *node;
8118 
8119 	/* to remove all channel filters, iterate an ordered list of filters */
8120 	hlist_for_each_entry_safe(fltr, node,
8121 				  &pf->tc_flower_fltr_list,
8122 				  tc_flower_node) {
8123 		struct ice_rule_query_data rule;
8124 		int status;
8125 
8126 		/* for now process only channel specific filters */
8127 		if (!ice_is_chnl_fltr(fltr))
8128 			continue;
8129 
8130 		rule.rid = fltr->rid;
8131 		rule.rule_id = fltr->rule_id;
8132 		rule.vsi_handle = fltr->dest_id;
8133 		status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
8134 		if (status) {
8135 			if (status == -ENOENT)
8136 				dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
8137 					rule.rule_id);
8138 			else
8139 				dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
8140 					status);
8141 		} else if (fltr->dest_vsi) {
8142 			/* update advanced switch filter count */
8143 			if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
8144 				u32 flags = fltr->flags;
8145 
8146 				fltr->dest_vsi->num_chnl_fltr--;
8147 				if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
8148 					     ICE_TC_FLWR_FIELD_ENC_DST_MAC))
8149 					pf->num_dmac_chnl_fltrs--;
8150 			}
8151 		}
8152 
8153 		hlist_del(&fltr->tc_flower_node);
8154 		kfree(fltr);
8155 	}
8156 }
8157 
8158 /**
8159  * ice_remove_q_channels - Remove queue channels for the TCs
8160  * @vsi: VSI to be configured
8161  * @rem_fltr: delete advanced switch filter or not
8162  *
8163  * Remove queue channels for the TCs
8164  */
8165 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
8166 {
8167 	struct ice_channel *ch, *ch_tmp;
8168 	struct ice_pf *pf = vsi->back;
8169 	int i;
8170 
8171 	/* remove all tc-flower based filter if they are channel filters only */
8172 	if (rem_fltr)
8173 		ice_rem_all_chnl_fltrs(pf);
8174 
8175 	/* remove ntuple filters since queue configuration is being changed */
8176 	if  (vsi->netdev->features & NETIF_F_NTUPLE) {
8177 		struct ice_hw *hw = &pf->hw;
8178 
8179 		mutex_lock(&hw->fdir_fltr_lock);
8180 		ice_fdir_del_all_fltrs(vsi);
8181 		mutex_unlock(&hw->fdir_fltr_lock);
8182 	}
8183 
8184 	/* perform cleanup for channels if they exist */
8185 	list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8186 		struct ice_vsi *ch_vsi;
8187 
8188 		list_del(&ch->list);
8189 		ch_vsi = ch->ch_vsi;
8190 		if (!ch_vsi) {
8191 			kfree(ch);
8192 			continue;
8193 		}
8194 
8195 		/* Reset queue contexts */
8196 		for (i = 0; i < ch->num_rxq; i++) {
8197 			struct ice_tx_ring *tx_ring;
8198 			struct ice_rx_ring *rx_ring;
8199 
8200 			tx_ring = vsi->tx_rings[ch->base_q + i];
8201 			rx_ring = vsi->rx_rings[ch->base_q + i];
8202 			if (tx_ring) {
8203 				tx_ring->ch = NULL;
8204 				if (tx_ring->q_vector)
8205 					tx_ring->q_vector->ch = NULL;
8206 			}
8207 			if (rx_ring) {
8208 				rx_ring->ch = NULL;
8209 				if (rx_ring->q_vector)
8210 					rx_ring->q_vector->ch = NULL;
8211 			}
8212 		}
8213 
8214 		/* Release FD resources for the channel VSI */
8215 		ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
8216 
8217 		/* clear the VSI from scheduler tree */
8218 		ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
8219 
8220 		/* Delete VSI from FW */
8221 		ice_vsi_delete(ch->ch_vsi);
8222 
8223 		/* Delete VSI from PF and HW VSI arrays */
8224 		ice_vsi_clear(ch->ch_vsi);
8225 
8226 		/* free the channel */
8227 		kfree(ch);
8228 	}
8229 
8230 	/* clear the channel VSI map which is stored in main VSI */
8231 	ice_for_each_chnl_tc(i)
8232 		vsi->tc_map_vsi[i] = NULL;
8233 
8234 	/* reset main VSI's all TC information */
8235 	vsi->all_enatc = 0;
8236 	vsi->all_numtc = 0;
8237 }
8238 
8239 /**
8240  * ice_rebuild_channels - rebuild channel
8241  * @pf: ptr to PF
8242  *
8243  * Recreate channel VSIs and replay filters
8244  */
8245 static int ice_rebuild_channels(struct ice_pf *pf)
8246 {
8247 	struct device *dev = ice_pf_to_dev(pf);
8248 	struct ice_vsi *main_vsi;
8249 	bool rem_adv_fltr = true;
8250 	struct ice_channel *ch;
8251 	struct ice_vsi *vsi;
8252 	int tc_idx = 1;
8253 	int i, err;
8254 
8255 	main_vsi = ice_get_main_vsi(pf);
8256 	if (!main_vsi)
8257 		return 0;
8258 
8259 	if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
8260 	    main_vsi->old_numtc == 1)
8261 		return 0; /* nothing to be done */
8262 
8263 	/* reconfigure main VSI based on old value of TC and cached values
8264 	 * for MQPRIO opts
8265 	 */
8266 	err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
8267 	if (err) {
8268 		dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
8269 			main_vsi->old_ena_tc, main_vsi->vsi_num);
8270 		return err;
8271 	}
8272 
8273 	/* rebuild ADQ VSIs */
8274 	ice_for_each_vsi(pf, i) {
8275 		enum ice_vsi_type type;
8276 
8277 		vsi = pf->vsi[i];
8278 		if (!vsi || vsi->type != ICE_VSI_CHNL)
8279 			continue;
8280 
8281 		type = vsi->type;
8282 
8283 		/* rebuild ADQ VSI */
8284 		err = ice_vsi_rebuild(vsi, true);
8285 		if (err) {
8286 			dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
8287 				ice_vsi_type_str(type), vsi->idx, err);
8288 			goto cleanup;
8289 		}
8290 
8291 		/* Re-map HW VSI number, using VSI handle that has been
8292 		 * previously validated in ice_replay_vsi() call above
8293 		 */
8294 		vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
8295 
8296 		/* replay filters for the VSI */
8297 		err = ice_replay_vsi(&pf->hw, vsi->idx);
8298 		if (err) {
8299 			dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
8300 				ice_vsi_type_str(type), err, vsi->idx);
8301 			rem_adv_fltr = false;
8302 			goto cleanup;
8303 		}
8304 		dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
8305 			 ice_vsi_type_str(type), vsi->idx);
8306 
8307 		/* store ADQ VSI at correct TC index in main VSI's
8308 		 * map of TC to VSI
8309 		 */
8310 		main_vsi->tc_map_vsi[tc_idx++] = vsi;
8311 	}
8312 
8313 	/* ADQ VSI(s) has been rebuilt successfully, so setup
8314 	 * channel for main VSI's Tx and Rx rings
8315 	 */
8316 	list_for_each_entry(ch, &main_vsi->ch_list, list) {
8317 		struct ice_vsi *ch_vsi;
8318 
8319 		ch_vsi = ch->ch_vsi;
8320 		if (!ch_vsi)
8321 			continue;
8322 
8323 		/* reconfig channel resources */
8324 		ice_cfg_chnl_all_res(main_vsi, ch);
8325 
8326 		/* replay BW rate limit if it is non-zero */
8327 		if (!ch->max_tx_rate && !ch->min_tx_rate)
8328 			continue;
8329 
8330 		err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8331 				       ch->min_tx_rate);
8332 		if (err)
8333 			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",
8334 				err, ch->max_tx_rate, ch->min_tx_rate,
8335 				ch_vsi->vsi_num);
8336 		else
8337 			dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8338 				ch->max_tx_rate, ch->min_tx_rate,
8339 				ch_vsi->vsi_num);
8340 	}
8341 
8342 	/* reconfig RSS for main VSI */
8343 	if (main_vsi->ch_rss_size)
8344 		ice_vsi_cfg_rss_lut_key(main_vsi);
8345 
8346 	return 0;
8347 
8348 cleanup:
8349 	ice_remove_q_channels(main_vsi, rem_adv_fltr);
8350 	return err;
8351 }
8352 
8353 /**
8354  * ice_create_q_channels - Add queue channel for the given TCs
8355  * @vsi: VSI to be configured
8356  *
8357  * Configures queue channel mapping to the given TCs
8358  */
8359 static int ice_create_q_channels(struct ice_vsi *vsi)
8360 {
8361 	struct ice_pf *pf = vsi->back;
8362 	struct ice_channel *ch;
8363 	int ret = 0, i;
8364 
8365 	ice_for_each_chnl_tc(i) {
8366 		if (!(vsi->all_enatc & BIT(i)))
8367 			continue;
8368 
8369 		ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8370 		if (!ch) {
8371 			ret = -ENOMEM;
8372 			goto err_free;
8373 		}
8374 		INIT_LIST_HEAD(&ch->list);
8375 		ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8376 		ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8377 		ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8378 		ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8379 		ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8380 
8381 		/* convert to Kbits/s */
8382 		if (ch->max_tx_rate)
8383 			ch->max_tx_rate = div_u64(ch->max_tx_rate,
8384 						  ICE_BW_KBPS_DIVISOR);
8385 		if (ch->min_tx_rate)
8386 			ch->min_tx_rate = div_u64(ch->min_tx_rate,
8387 						  ICE_BW_KBPS_DIVISOR);
8388 
8389 		ret = ice_create_q_channel(vsi, ch);
8390 		if (ret) {
8391 			dev_err(ice_pf_to_dev(pf),
8392 				"failed creating channel TC:%d\n", i);
8393 			kfree(ch);
8394 			goto err_free;
8395 		}
8396 		list_add_tail(&ch->list, &vsi->ch_list);
8397 		vsi->tc_map_vsi[i] = ch->ch_vsi;
8398 		dev_dbg(ice_pf_to_dev(pf),
8399 			"successfully created channel: VSI %pK\n", ch->ch_vsi);
8400 	}
8401 	return 0;
8402 
8403 err_free:
8404 	ice_remove_q_channels(vsi, false);
8405 
8406 	return ret;
8407 }
8408 
8409 /**
8410  * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8411  * @netdev: net device to configure
8412  * @type_data: TC offload data
8413  */
8414 static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8415 {
8416 	struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8417 	struct ice_netdev_priv *np = netdev_priv(netdev);
8418 	struct ice_vsi *vsi = np->vsi;
8419 	struct ice_pf *pf = vsi->back;
8420 	u16 mode, ena_tc_qdisc = 0;
8421 	int cur_txq, cur_rxq;
8422 	u8 hw = 0, num_tcf;
8423 	struct device *dev;
8424 	int ret, i;
8425 
8426 	dev = ice_pf_to_dev(pf);
8427 	num_tcf = mqprio_qopt->qopt.num_tc;
8428 	hw = mqprio_qopt->qopt.hw;
8429 	mode = mqprio_qopt->mode;
8430 	if (!hw) {
8431 		clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8432 		vsi->ch_rss_size = 0;
8433 		memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8434 		goto config_tcf;
8435 	}
8436 
8437 	/* Generate queue region map for number of TCF requested */
8438 	for (i = 0; i < num_tcf; i++)
8439 		ena_tc_qdisc |= BIT(i);
8440 
8441 	switch (mode) {
8442 	case TC_MQPRIO_MODE_CHANNEL:
8443 
8444 		ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8445 		if (ret) {
8446 			netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8447 				   ret);
8448 			return ret;
8449 		}
8450 		memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8451 		set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8452 		/* don't assume state of hw_tc_offload during driver load
8453 		 * and set the flag for TC flower filter if hw_tc_offload
8454 		 * already ON
8455 		 */
8456 		if (vsi->netdev->features & NETIF_F_HW_TC)
8457 			set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
8458 		break;
8459 	default:
8460 		return -EINVAL;
8461 	}
8462 
8463 config_tcf:
8464 
8465 	/* Requesting same TCF configuration as already enabled */
8466 	if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
8467 	    mode != TC_MQPRIO_MODE_CHANNEL)
8468 		return 0;
8469 
8470 	/* Pause VSI queues */
8471 	ice_dis_vsi(vsi, true);
8472 
8473 	if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
8474 		ice_remove_q_channels(vsi, true);
8475 
8476 	if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8477 		vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
8478 				     num_online_cpus());
8479 		vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
8480 				     num_online_cpus());
8481 	} else {
8482 		/* logic to rebuild VSI, same like ethtool -L */
8483 		u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
8484 
8485 		for (i = 0; i < num_tcf; i++) {
8486 			if (!(ena_tc_qdisc & BIT(i)))
8487 				continue;
8488 
8489 			offset = vsi->mqprio_qopt.qopt.offset[i];
8490 			qcount_rx = vsi->mqprio_qopt.qopt.count[i];
8491 			qcount_tx = vsi->mqprio_qopt.qopt.count[i];
8492 		}
8493 		vsi->req_txq = offset + qcount_tx;
8494 		vsi->req_rxq = offset + qcount_rx;
8495 
8496 		/* store away original rss_size info, so that it gets reused
8497 		 * form ice_vsi_rebuild during tc-qdisc delete stage - to
8498 		 * determine, what should be the rss_sizefor main VSI
8499 		 */
8500 		vsi->orig_rss_size = vsi->rss_size;
8501 	}
8502 
8503 	/* save current values of Tx and Rx queues before calling VSI rebuild
8504 	 * for fallback option
8505 	 */
8506 	cur_txq = vsi->num_txq;
8507 	cur_rxq = vsi->num_rxq;
8508 
8509 	/* proceed with rebuild main VSI using correct number of queues */
8510 	ret = ice_vsi_rebuild(vsi, false);
8511 	if (ret) {
8512 		/* fallback to current number of queues */
8513 		dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
8514 		vsi->req_txq = cur_txq;
8515 		vsi->req_rxq = cur_rxq;
8516 		clear_bit(ICE_RESET_FAILED, pf->state);
8517 		if (ice_vsi_rebuild(vsi, false)) {
8518 			dev_err(dev, "Rebuild of main VSI failed again\n");
8519 			return ret;
8520 		}
8521 	}
8522 
8523 	vsi->all_numtc = num_tcf;
8524 	vsi->all_enatc = ena_tc_qdisc;
8525 	ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
8526 	if (ret) {
8527 		netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
8528 			   vsi->vsi_num);
8529 		goto exit;
8530 	}
8531 
8532 	if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8533 		u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
8534 		u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
8535 
8536 		/* set TC0 rate limit if specified */
8537 		if (max_tx_rate || min_tx_rate) {
8538 			/* convert to Kbits/s */
8539 			if (max_tx_rate)
8540 				max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
8541 			if (min_tx_rate)
8542 				min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
8543 
8544 			ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
8545 			if (!ret) {
8546 				dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
8547 					max_tx_rate, min_tx_rate, vsi->vsi_num);
8548 			} else {
8549 				dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
8550 					max_tx_rate, min_tx_rate, vsi->vsi_num);
8551 				goto exit;
8552 			}
8553 		}
8554 		ret = ice_create_q_channels(vsi);
8555 		if (ret) {
8556 			netdev_err(netdev, "failed configuring queue channels\n");
8557 			goto exit;
8558 		} else {
8559 			netdev_dbg(netdev, "successfully configured channels\n");
8560 		}
8561 	}
8562 
8563 	if (vsi->ch_rss_size)
8564 		ice_vsi_cfg_rss_lut_key(vsi);
8565 
8566 exit:
8567 	/* if error, reset the all_numtc and all_enatc */
8568 	if (ret) {
8569 		vsi->all_numtc = 0;
8570 		vsi->all_enatc = 0;
8571 	}
8572 	/* resume VSI */
8573 	ice_ena_vsi(vsi, true);
8574 
8575 	return ret;
8576 }
8577 
8578 static LIST_HEAD(ice_block_cb_list);
8579 
8580 static int
8581 ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8582 	     void *type_data)
8583 {
8584 	struct ice_netdev_priv *np = netdev_priv(netdev);
8585 	struct ice_pf *pf = np->vsi->back;
8586 	int err;
8587 
8588 	switch (type) {
8589 	case TC_SETUP_BLOCK:
8590 		return flow_block_cb_setup_simple(type_data,
8591 						  &ice_block_cb_list,
8592 						  ice_setup_tc_block_cb,
8593 						  np, np, true);
8594 	case TC_SETUP_QDISC_MQPRIO:
8595 		/* setup traffic classifier for receive side */
8596 		mutex_lock(&pf->tc_mutex);
8597 		err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
8598 		mutex_unlock(&pf->tc_mutex);
8599 		return err;
8600 	default:
8601 		return -EOPNOTSUPP;
8602 	}
8603 	return -EOPNOTSUPP;
8604 }
8605 
8606 static struct ice_indr_block_priv *
8607 ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
8608 			   struct net_device *netdev)
8609 {
8610 	struct ice_indr_block_priv *cb_priv;
8611 
8612 	list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
8613 		if (!cb_priv->netdev)
8614 			return NULL;
8615 		if (cb_priv->netdev == netdev)
8616 			return cb_priv;
8617 	}
8618 	return NULL;
8619 }
8620 
8621 static int
8622 ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
8623 			void *indr_priv)
8624 {
8625 	struct ice_indr_block_priv *priv = indr_priv;
8626 	struct ice_netdev_priv *np = priv->np;
8627 
8628 	switch (type) {
8629 	case TC_SETUP_CLSFLOWER:
8630 		return ice_setup_tc_cls_flower(np, priv->netdev,
8631 					       (struct flow_cls_offload *)
8632 					       type_data);
8633 	default:
8634 		return -EOPNOTSUPP;
8635 	}
8636 }
8637 
8638 static int
8639 ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
8640 			struct ice_netdev_priv *np,
8641 			struct flow_block_offload *f, void *data,
8642 			void (*cleanup)(struct flow_block_cb *block_cb))
8643 {
8644 	struct ice_indr_block_priv *indr_priv;
8645 	struct flow_block_cb *block_cb;
8646 
8647 	if (!ice_is_tunnel_supported(netdev) &&
8648 	    !(is_vlan_dev(netdev) &&
8649 	      vlan_dev_real_dev(netdev) == np->vsi->netdev))
8650 		return -EOPNOTSUPP;
8651 
8652 	if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
8653 		return -EOPNOTSUPP;
8654 
8655 	switch (f->command) {
8656 	case FLOW_BLOCK_BIND:
8657 		indr_priv = ice_indr_block_priv_lookup(np, netdev);
8658 		if (indr_priv)
8659 			return -EEXIST;
8660 
8661 		indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
8662 		if (!indr_priv)
8663 			return -ENOMEM;
8664 
8665 		indr_priv->netdev = netdev;
8666 		indr_priv->np = np;
8667 		list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
8668 
8669 		block_cb =
8670 			flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
8671 						 indr_priv, indr_priv,
8672 						 ice_rep_indr_tc_block_unbind,
8673 						 f, netdev, sch, data, np,
8674 						 cleanup);
8675 
8676 		if (IS_ERR(block_cb)) {
8677 			list_del(&indr_priv->list);
8678 			kfree(indr_priv);
8679 			return PTR_ERR(block_cb);
8680 		}
8681 		flow_block_cb_add(block_cb, f);
8682 		list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
8683 		break;
8684 	case FLOW_BLOCK_UNBIND:
8685 		indr_priv = ice_indr_block_priv_lookup(np, netdev);
8686 		if (!indr_priv)
8687 			return -ENOENT;
8688 
8689 		block_cb = flow_block_cb_lookup(f->block,
8690 						ice_indr_setup_block_cb,
8691 						indr_priv);
8692 		if (!block_cb)
8693 			return -ENOENT;
8694 
8695 		flow_indr_block_cb_remove(block_cb, f);
8696 
8697 		list_del(&block_cb->driver_list);
8698 		break;
8699 	default:
8700 		return -EOPNOTSUPP;
8701 	}
8702 	return 0;
8703 }
8704 
8705 static int
8706 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
8707 		     void *cb_priv, enum tc_setup_type type, void *type_data,
8708 		     void *data,
8709 		     void (*cleanup)(struct flow_block_cb *block_cb))
8710 {
8711 	switch (type) {
8712 	case TC_SETUP_BLOCK:
8713 		return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
8714 					       data, cleanup);
8715 
8716 	default:
8717 		return -EOPNOTSUPP;
8718 	}
8719 }
8720 
8721 /**
8722  * ice_open - Called when a network interface becomes active
8723  * @netdev: network interface device structure
8724  *
8725  * The open entry point is called when a network interface is made
8726  * active by the system (IFF_UP). At this point all resources needed
8727  * for transmit and receive operations are allocated, the interrupt
8728  * handler is registered with the OS, the netdev watchdog is enabled,
8729  * and the stack is notified that the interface is ready.
8730  *
8731  * Returns 0 on success, negative value on failure
8732  */
8733 int ice_open(struct net_device *netdev)
8734 {
8735 	struct ice_netdev_priv *np = netdev_priv(netdev);
8736 	struct ice_pf *pf = np->vsi->back;
8737 
8738 	if (ice_is_reset_in_progress(pf->state)) {
8739 		netdev_err(netdev, "can't open net device while reset is in progress");
8740 		return -EBUSY;
8741 	}
8742 
8743 	return ice_open_internal(netdev);
8744 }
8745 
8746 /**
8747  * ice_open_internal - Called when a network interface becomes active
8748  * @netdev: network interface device structure
8749  *
8750  * Internal ice_open implementation. Should not be used directly except for ice_open and reset
8751  * handling routine
8752  *
8753  * Returns 0 on success, negative value on failure
8754  */
8755 int ice_open_internal(struct net_device *netdev)
8756 {
8757 	struct ice_netdev_priv *np = netdev_priv(netdev);
8758 	struct ice_vsi *vsi = np->vsi;
8759 	struct ice_pf *pf = vsi->back;
8760 	struct ice_port_info *pi;
8761 	int err;
8762 
8763 	if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
8764 		netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
8765 		return -EIO;
8766 	}
8767 
8768 	netif_carrier_off(netdev);
8769 
8770 	pi = vsi->port_info;
8771 	err = ice_update_link_info(pi);
8772 	if (err) {
8773 		netdev_err(netdev, "Failed to get link info, error %d\n", err);
8774 		return err;
8775 	}
8776 
8777 	ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
8778 
8779 	/* Set PHY if there is media, otherwise, turn off PHY */
8780 	if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
8781 		clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8782 		if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
8783 			err = ice_init_phy_user_cfg(pi);
8784 			if (err) {
8785 				netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
8786 					   err);
8787 				return err;
8788 			}
8789 		}
8790 
8791 		err = ice_configure_phy(vsi);
8792 		if (err) {
8793 			netdev_err(netdev, "Failed to set physical link up, error %d\n",
8794 				   err);
8795 			return err;
8796 		}
8797 	} else {
8798 		set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8799 		ice_set_link(vsi, false);
8800 	}
8801 
8802 	err = ice_vsi_open(vsi);
8803 	if (err)
8804 		netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
8805 			   vsi->vsi_num, vsi->vsw->sw_id);
8806 
8807 	/* Update existing tunnels information */
8808 	udp_tunnel_get_rx_info(netdev);
8809 
8810 	return err;
8811 }
8812 
8813 /**
8814  * ice_stop - Disables a network interface
8815  * @netdev: network interface device structure
8816  *
8817  * The stop entry point is called when an interface is de-activated by the OS,
8818  * and the netdevice enters the DOWN state. The hardware is still under the
8819  * driver's control, but the netdev interface is disabled.
8820  *
8821  * Returns success only - not allowed to fail
8822  */
8823 int ice_stop(struct net_device *netdev)
8824 {
8825 	struct ice_netdev_priv *np = netdev_priv(netdev);
8826 	struct ice_vsi *vsi = np->vsi;
8827 	struct ice_pf *pf = vsi->back;
8828 
8829 	if (ice_is_reset_in_progress(pf->state)) {
8830 		netdev_err(netdev, "can't stop net device while reset is in progress");
8831 		return -EBUSY;
8832 	}
8833 
8834 	ice_vsi_close(vsi);
8835 
8836 	return 0;
8837 }
8838 
8839 /**
8840  * ice_features_check - Validate encapsulated packet conforms to limits
8841  * @skb: skb buffer
8842  * @netdev: This port's netdev
8843  * @features: Offload features that the stack believes apply
8844  */
8845 static netdev_features_t
8846 ice_features_check(struct sk_buff *skb,
8847 		   struct net_device __always_unused *netdev,
8848 		   netdev_features_t features)
8849 {
8850 	bool gso = skb_is_gso(skb);
8851 	size_t len;
8852 
8853 	/* No point in doing any of this if neither checksum nor GSO are
8854 	 * being requested for this frame. We can rule out both by just
8855 	 * checking for CHECKSUM_PARTIAL
8856 	 */
8857 	if (skb->ip_summed != CHECKSUM_PARTIAL)
8858 		return features;
8859 
8860 	/* We cannot support GSO if the MSS is going to be less than
8861 	 * 64 bytes. If it is then we need to drop support for GSO.
8862 	 */
8863 	if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
8864 		features &= ~NETIF_F_GSO_MASK;
8865 
8866 	len = skb_network_offset(skb);
8867 	if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
8868 		goto out_rm_features;
8869 
8870 	len = skb_network_header_len(skb);
8871 	if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8872 		goto out_rm_features;
8873 
8874 	if (skb->encapsulation) {
8875 		/* this must work for VXLAN frames AND IPIP/SIT frames, and in
8876 		 * the case of IPIP frames, the transport header pointer is
8877 		 * after the inner header! So check to make sure that this
8878 		 * is a GRE or UDP_TUNNEL frame before doing that math.
8879 		 */
8880 		if (gso && (skb_shinfo(skb)->gso_type &
8881 			    (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
8882 			len = skb_inner_network_header(skb) -
8883 			      skb_transport_header(skb);
8884 			if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
8885 				goto out_rm_features;
8886 		}
8887 
8888 		len = skb_inner_network_header_len(skb);
8889 		if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8890 			goto out_rm_features;
8891 	}
8892 
8893 	return features;
8894 out_rm_features:
8895 	return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
8896 }
8897 
8898 static const struct net_device_ops ice_netdev_safe_mode_ops = {
8899 	.ndo_open = ice_open,
8900 	.ndo_stop = ice_stop,
8901 	.ndo_start_xmit = ice_start_xmit,
8902 	.ndo_set_mac_address = ice_set_mac_address,
8903 	.ndo_validate_addr = eth_validate_addr,
8904 	.ndo_change_mtu = ice_change_mtu,
8905 	.ndo_get_stats64 = ice_get_stats64,
8906 	.ndo_tx_timeout = ice_tx_timeout,
8907 	.ndo_bpf = ice_xdp_safe_mode,
8908 };
8909 
8910 static const struct net_device_ops ice_netdev_ops = {
8911 	.ndo_open = ice_open,
8912 	.ndo_stop = ice_stop,
8913 	.ndo_start_xmit = ice_start_xmit,
8914 	.ndo_select_queue = ice_select_queue,
8915 	.ndo_features_check = ice_features_check,
8916 	.ndo_fix_features = ice_fix_features,
8917 	.ndo_set_rx_mode = ice_set_rx_mode,
8918 	.ndo_set_mac_address = ice_set_mac_address,
8919 	.ndo_validate_addr = eth_validate_addr,
8920 	.ndo_change_mtu = ice_change_mtu,
8921 	.ndo_get_stats64 = ice_get_stats64,
8922 	.ndo_set_tx_maxrate = ice_set_tx_maxrate,
8923 	.ndo_eth_ioctl = ice_eth_ioctl,
8924 	.ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
8925 	.ndo_set_vf_mac = ice_set_vf_mac,
8926 	.ndo_get_vf_config = ice_get_vf_cfg,
8927 	.ndo_set_vf_trust = ice_set_vf_trust,
8928 	.ndo_set_vf_vlan = ice_set_vf_port_vlan,
8929 	.ndo_set_vf_link_state = ice_set_vf_link_state,
8930 	.ndo_get_vf_stats = ice_get_vf_stats,
8931 	.ndo_set_vf_rate = ice_set_vf_bw,
8932 	.ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
8933 	.ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
8934 	.ndo_setup_tc = ice_setup_tc,
8935 	.ndo_set_features = ice_set_features,
8936 	.ndo_bridge_getlink = ice_bridge_getlink,
8937 	.ndo_bridge_setlink = ice_bridge_setlink,
8938 	.ndo_fdb_add = ice_fdb_add,
8939 	.ndo_fdb_del = ice_fdb_del,
8940 #ifdef CONFIG_RFS_ACCEL
8941 	.ndo_rx_flow_steer = ice_rx_flow_steer,
8942 #endif
8943 	.ndo_tx_timeout = ice_tx_timeout,
8944 	.ndo_bpf = ice_xdp,
8945 	.ndo_xdp_xmit = ice_xdp_xmit,
8946 	.ndo_xsk_wakeup = ice_xsk_wakeup,
8947 	.ndo_get_devlink_port = ice_get_devlink_port,
8948 };
8949